linux/drivers/gpu/drm/i915/intel_ringbuffer.c

2202 lines
58 KiB
C
Raw Normal View History

/*
* Copyright © 2008-2010 Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*
* Authors:
* Eric Anholt <eric@anholt.net>
* Zou Nan hai <nanhai.zou@intel.com>
* Xiang Hai hao<haihao.xiang@intel.com>
*
*/
#include <linux/log2.h>
#include <drm/drmP.h>
#include "i915_drv.h"
#include <drm/i915_drm.h>
#include "i915_trace.h"
#include "intel_drv.h"
/* Rough estimate of the typical request size, performing a flush,
* set-context and then emitting the batch.
*/
#define LEGACY_REQUEST_SIZE 200
static unsigned int __intel_ring_space(unsigned int head,
unsigned int tail,
unsigned int size)
{
/*
* "If the Ring Buffer Head Pointer and the Tail Pointer are on the
* same cacheline, the Head Pointer must not be greater than the Tail
* Pointer."
*/
GEM_BUG_ON(!is_power_of_2(size));
return (head - tail - CACHELINE_BYTES) & (size - 1);
}
unsigned int intel_ring_update_space(struct intel_ring *ring)
{
unsigned int space;
space = __intel_ring_space(ring->head, ring->emit, ring->size);
ring->space = space;
return space;
}
static int
gen2_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 cmd, *cs;
cmd = MI_FLUSH;
if (mode & EMIT_INVALIDATE)
cmd |= MI_READ_FLUSH;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = cmd;
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
return 0;
}
static int
gen4_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 cmd, *cs;
/*
* read/write caches:
*
* I915_GEM_DOMAIN_RENDER is always invalidated, but is
* only flushed if MI_NO_WRITE_FLUSH is unset. On 965, it is
* also flushed at 2d versus 3d pipeline switches.
*
* read-only caches:
*
* I915_GEM_DOMAIN_SAMPLER is flushed on pre-965 if
* MI_READ_FLUSH is set, and is always flushed on 965.
*
* I915_GEM_DOMAIN_COMMAND may not exist?
*
* I915_GEM_DOMAIN_INSTRUCTION, which exists on 965, is
* invalidated when MI_EXE_FLUSH is set.
*
* I915_GEM_DOMAIN_VERTEX, which exists on 965, is
* invalidated with every MI_FLUSH.
*
* TLBs:
*
* On 965, TLBs associated with I915_GEM_DOMAIN_COMMAND
* and I915_GEM_DOMAIN_CPU in are invalidated at PTE write and
* I915_GEM_DOMAIN_RENDER and I915_GEM_DOMAIN_SAMPLER
* are flushed at any MI_FLUSH.
*/
cmd = MI_FLUSH;
if (mode & EMIT_INVALIDATE) {
cmd |= MI_EXE_FLUSH;
if (IS_G4X(req->i915) || IS_GEN5(req->i915))
cmd |= MI_INVALIDATE_ISP;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = cmd;
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
return 0;
}
/**
* Emits a PIPE_CONTROL with a non-zero post-sync operation, for
* implementing two workarounds on gen6. From section 1.4.7.1
* "PIPE_CONTROL" of the Sandy Bridge PRM volume 2 part 1:
*
* [DevSNB-C+{W/A}] Before any depth stall flush (including those
* produced by non-pipelined state commands), software needs to first
* send a PIPE_CONTROL with no bits set except Post-Sync Operation !=
* 0.
*
* [Dev-SNB{W/A}]: Before a PIPE_CONTROL with Write Cache Flush Enable
* =1, a PIPE_CONTROL with any non-zero post-sync-op is required.
*
* And the workaround for these two requires this workaround first:
*
* [Dev-SNB{W/A}]: Pipe-control with CS-stall bit set must be sent
* BEFORE the pipe-control with a post-sync op and no write-cache
* flushes.
*
* And this last workaround is tricky because of the requirements on
* that bit. From section 1.4.7.2.3 "Stall" of the Sandy Bridge PRM
* volume 2 part 1:
*
* "1 of the following must also be set:
* - Render Target Cache Flush Enable ([12] of DW1)
* - Depth Cache Flush Enable ([0] of DW1)
* - Stall at Pixel Scoreboard ([1] of DW1)
* - Depth Stall ([13] of DW1)
* - Post-Sync Operation ([13] of DW1)
* - Notify Enable ([8] of DW1)"
*
* The cache flushes require the workaround flush that triggered this
* one, so we can't use it. Depth stall would trigger the same.
* Post-sync nonzero is what triggered this second workaround, so we
* can't use that one either. Notify enable is IRQs, which aren't
* really our business. That leaves only stall at scoreboard.
*/
static int
intel_emit_post_sync_nonzero_flush(struct drm_i915_gem_request *req)
{
u32 scratch_addr =
i915_ggtt_offset(req->engine->scratch) + 2 * CACHELINE_BYTES;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
cs = intel_ring_begin(req, 6);
if (IS_ERR(cs))
return PTR_ERR(cs);
*cs++ = GFX_OP_PIPE_CONTROL(5);
*cs++ = PIPE_CONTROL_CS_STALL | PIPE_CONTROL_STALL_AT_SCOREBOARD;
*cs++ = scratch_addr | PIPE_CONTROL_GLOBAL_GTT;
*cs++ = 0; /* low dword */
*cs++ = 0; /* high dword */
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
cs = intel_ring_begin(req, 6);
if (IS_ERR(cs))
return PTR_ERR(cs);
*cs++ = GFX_OP_PIPE_CONTROL(5);
*cs++ = PIPE_CONTROL_QW_WRITE;
*cs++ = scratch_addr | PIPE_CONTROL_GLOBAL_GTT;
*cs++ = 0;
*cs++ = 0;
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
return 0;
}
static int
gen6_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
u32 scratch_addr =
i915_ggtt_offset(req->engine->scratch) + 2 * CACHELINE_BYTES;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs, flags = 0;
int ret;
/* Force SNB workarounds for PIPE_CONTROL flushes */
ret = intel_emit_post_sync_nonzero_flush(req);
if (ret)
return ret;
/* Just flush everything. Experiments have shown that reducing the
* number of bits based on the write domains has little performance
* impact.
*/
if (mode & EMIT_FLUSH) {
flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH;
flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH;
/*
* Ensure that any following seqno writes only happen
* when the render cache is indeed flushed.
*/
flags |= PIPE_CONTROL_CS_STALL;
}
if (mode & EMIT_INVALIDATE) {
flags |= PIPE_CONTROL_TLB_INVALIDATE;
flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE;
/*
* TLB invalidate requires a post-sync write.
*/
flags |= PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_CS_STALL;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 4);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = GFX_OP_PIPE_CONTROL(4);
*cs++ = flags;
*cs++ = scratch_addr | PIPE_CONTROL_GLOBAL_GTT;
*cs++ = 0;
intel_ring_advance(req, cs);
return 0;
}
drm/i915: add workarounds to gen7_render_ring_flush From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section detailing the various workarounds: "[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit set must be issued before a pipe-control command that has the State Cache Invalidate bit set." Note that public Bspec has different numbering, it's Vol2Part1, Section 1.10.4.1 "PIPE_CONTROL" there. There's also a second workaround for the PIPE_CONTROL command itself: "[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set, must have a CS_STALL bit set" For simplicity we simply set the CS_STALL bit on every pipe_control on gen7+ Note that this massively helps on some hsw machines, together with the following patch to unconditionally set the CS_STALL bit on every pipe_control it prevents a gpu hang every few seconds. This is a regression that has been introduced in the pipe_control cleanup: commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Jul 20 18:02:28 2012 +0100 drm/i915: Only apply the SNB pipe control w/a to gen6 It looks like the massive snb pipe_control workaround also papered over any issues on ivb and hsw. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: squashed both workarounds together, pimped commit message with Bsepc citations, regression commit citation and changed the comment in the code a bit to clarify that we unconditionally set CS_STALL to avoid being hurt by trying to be clever.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-17 21:35:43 +00:00
static int
gen7_render_ring_cs_stall_wa(struct drm_i915_gem_request *req)
drm/i915: add workarounds to gen7_render_ring_flush From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section detailing the various workarounds: "[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit set must be issued before a pipe-control command that has the State Cache Invalidate bit set." Note that public Bspec has different numbering, it's Vol2Part1, Section 1.10.4.1 "PIPE_CONTROL" there. There's also a second workaround for the PIPE_CONTROL command itself: "[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set, must have a CS_STALL bit set" For simplicity we simply set the CS_STALL bit on every pipe_control on gen7+ Note that this massively helps on some hsw machines, together with the following patch to unconditionally set the CS_STALL bit on every pipe_control it prevents a gpu hang every few seconds. This is a regression that has been introduced in the pipe_control cleanup: commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Jul 20 18:02:28 2012 +0100 drm/i915: Only apply the SNB pipe control w/a to gen6 It looks like the massive snb pipe_control workaround also papered over any issues on ivb and hsw. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: squashed both workarounds together, pimped commit message with Bsepc citations, regression commit citation and changed the comment in the code a bit to clarify that we unconditionally set CS_STALL to avoid being hurt by trying to be clever.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-17 21:35:43 +00:00
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: add workarounds to gen7_render_ring_flush From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section detailing the various workarounds: "[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit set must be issued before a pipe-control command that has the State Cache Invalidate bit set." Note that public Bspec has different numbering, it's Vol2Part1, Section 1.10.4.1 "PIPE_CONTROL" there. There's also a second workaround for the PIPE_CONTROL command itself: "[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set, must have a CS_STALL bit set" For simplicity we simply set the CS_STALL bit on every pipe_control on gen7+ Note that this massively helps on some hsw machines, together with the following patch to unconditionally set the CS_STALL bit on every pipe_control it prevents a gpu hang every few seconds. This is a regression that has been introduced in the pipe_control cleanup: commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Jul 20 18:02:28 2012 +0100 drm/i915: Only apply the SNB pipe control w/a to gen6 It looks like the massive snb pipe_control workaround also papered over any issues on ivb and hsw. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: squashed both workarounds together, pimped commit message with Bsepc citations, regression commit citation and changed the comment in the code a bit to clarify that we unconditionally set CS_STALL to avoid being hurt by trying to be clever.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-17 21:35:43 +00:00
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 4);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: add workarounds to gen7_render_ring_flush From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section detailing the various workarounds: "[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit set must be issued before a pipe-control command that has the State Cache Invalidate bit set." Note that public Bspec has different numbering, it's Vol2Part1, Section 1.10.4.1 "PIPE_CONTROL" there. There's also a second workaround for the PIPE_CONTROL command itself: "[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set, must have a CS_STALL bit set" For simplicity we simply set the CS_STALL bit on every pipe_control on gen7+ Note that this massively helps on some hsw machines, together with the following patch to unconditionally set the CS_STALL bit on every pipe_control it prevents a gpu hang every few seconds. This is a regression that has been introduced in the pipe_control cleanup: commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Jul 20 18:02:28 2012 +0100 drm/i915: Only apply the SNB pipe control w/a to gen6 It looks like the massive snb pipe_control workaround also papered over any issues on ivb and hsw. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: squashed both workarounds together, pimped commit message with Bsepc citations, regression commit citation and changed the comment in the code a bit to clarify that we unconditionally set CS_STALL to avoid being hurt by trying to be clever.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-17 21:35:43 +00:00
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = GFX_OP_PIPE_CONTROL(4);
*cs++ = PIPE_CONTROL_CS_STALL | PIPE_CONTROL_STALL_AT_SCOREBOARD;
*cs++ = 0;
*cs++ = 0;
intel_ring_advance(req, cs);
drm/i915: add workarounds to gen7_render_ring_flush From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section detailing the various workarounds: "[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit set must be issued before a pipe-control command that has the State Cache Invalidate bit set." Note that public Bspec has different numbering, it's Vol2Part1, Section 1.10.4.1 "PIPE_CONTROL" there. There's also a second workaround for the PIPE_CONTROL command itself: "[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set, must have a CS_STALL bit set" For simplicity we simply set the CS_STALL bit on every pipe_control on gen7+ Note that this massively helps on some hsw machines, together with the following patch to unconditionally set the CS_STALL bit on every pipe_control it prevents a gpu hang every few seconds. This is a regression that has been introduced in the pipe_control cleanup: commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Jul 20 18:02:28 2012 +0100 drm/i915: Only apply the SNB pipe control w/a to gen6 It looks like the massive snb pipe_control workaround also papered over any issues on ivb and hsw. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: squashed both workarounds together, pimped commit message with Bsepc citations, regression commit citation and changed the comment in the code a bit to clarify that we unconditionally set CS_STALL to avoid being hurt by trying to be clever.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-17 21:35:43 +00:00
return 0;
}
static int
gen7_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
u32 scratch_addr =
i915_ggtt_offset(req->engine->scratch) + 2 * CACHELINE_BYTES;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs, flags = 0;
drm/i915: add workarounds to gen7_render_ring_flush From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section detailing the various workarounds: "[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit set must be issued before a pipe-control command that has the State Cache Invalidate bit set." Note that public Bspec has different numbering, it's Vol2Part1, Section 1.10.4.1 "PIPE_CONTROL" there. There's also a second workaround for the PIPE_CONTROL command itself: "[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set, must have a CS_STALL bit set" For simplicity we simply set the CS_STALL bit on every pipe_control on gen7+ Note that this massively helps on some hsw machines, together with the following patch to unconditionally set the CS_STALL bit on every pipe_control it prevents a gpu hang every few seconds. This is a regression that has been introduced in the pipe_control cleanup: commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Jul 20 18:02:28 2012 +0100 drm/i915: Only apply the SNB pipe control w/a to gen6 It looks like the massive snb pipe_control workaround also papered over any issues on ivb and hsw. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: squashed both workarounds together, pimped commit message with Bsepc citations, regression commit citation and changed the comment in the code a bit to clarify that we unconditionally set CS_STALL to avoid being hurt by trying to be clever.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-17 21:35:43 +00:00
/*
* Ensure that any following seqno writes only happen when the render
* cache is indeed flushed.
*
* Workaround: 4th PIPE_CONTROL command (except the ones with only
* read-cache invalidate bits set) must have the CS_STALL bit set. We
* don't try to be clever and just set it unconditionally.
*/
flags |= PIPE_CONTROL_CS_STALL;
/* Just flush everything. Experiments have shown that reducing the
* number of bits based on the write domains has little performance
* impact.
*/
if (mode & EMIT_FLUSH) {
flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH;
flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH;
flags |= PIPE_CONTROL_DC_FLUSH_ENABLE;
flags |= PIPE_CONTROL_FLUSH_ENABLE;
}
if (mode & EMIT_INVALIDATE) {
flags |= PIPE_CONTROL_TLB_INVALIDATE;
flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_MEDIA_STATE_CLEAR;
/*
* TLB invalidate requires a post-sync write.
*/
flags |= PIPE_CONTROL_QW_WRITE;
flags |= PIPE_CONTROL_GLOBAL_GTT_IVB;
drm/i915: add workarounds to gen7_render_ring_flush From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section detailing the various workarounds: "[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit set must be issued before a pipe-control command that has the State Cache Invalidate bit set." Note that public Bspec has different numbering, it's Vol2Part1, Section 1.10.4.1 "PIPE_CONTROL" there. There's also a second workaround for the PIPE_CONTROL command itself: "[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set, must have a CS_STALL bit set" For simplicity we simply set the CS_STALL bit on every pipe_control on gen7+ Note that this massively helps on some hsw machines, together with the following patch to unconditionally set the CS_STALL bit on every pipe_control it prevents a gpu hang every few seconds. This is a regression that has been introduced in the pipe_control cleanup: commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Jul 20 18:02:28 2012 +0100 drm/i915: Only apply the SNB pipe control w/a to gen6 It looks like the massive snb pipe_control workaround also papered over any issues on ivb and hsw. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: squashed both workarounds together, pimped commit message with Bsepc citations, regression commit citation and changed the comment in the code a bit to clarify that we unconditionally set CS_STALL to avoid being hurt by trying to be clever.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-17 21:35:43 +00:00
flags |= PIPE_CONTROL_STALL_AT_SCOREBOARD;
drm/i915: add workarounds to gen7_render_ring_flush From Bspec, Vol 2a, Section 1.9.3.4 "PIPE_CONTROL", intro section detailing the various workarounds: "[DevIVB {W/A}, DevHSW {W/A}]: Pipe_control with CS-stall bit set must be issued before a pipe-control command that has the State Cache Invalidate bit set." Note that public Bspec has different numbering, it's Vol2Part1, Section 1.10.4.1 "PIPE_CONTROL" there. There's also a second workaround for the PIPE_CONTROL command itself: "[DevIVB, DevVLV, DevHSW] {WA}: Every 4th PIPE_CONTROL command, not counting the PIPE_CONTROL with only read-cache-invalidate bit(s) set, must have a CS_STALL bit set" For simplicity we simply set the CS_STALL bit on every pipe_control on gen7+ Note that this massively helps on some hsw machines, together with the following patch to unconditionally set the CS_STALL bit on every pipe_control it prevents a gpu hang every few seconds. This is a regression that has been introduced in the pipe_control cleanup: commit 6c6cf5aa9c583478b19e23149feaa92d01fb8c2d Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Jul 20 18:02:28 2012 +0100 drm/i915: Only apply the SNB pipe control w/a to gen6 It looks like the massive snb pipe_control workaround also papered over any issues on ivb and hsw. Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> [danvet: squashed both workarounds together, pimped commit message with Bsepc citations, regression commit citation and changed the comment in the code a bit to clarify that we unconditionally set CS_STALL to avoid being hurt by trying to be clever.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-08-17 21:35:43 +00:00
/* Workaround: we must issue a pipe_control with CS-stall bit
* set before a pipe_control command that has the state cache
* invalidate bit set. */
gen7_render_ring_cs_stall_wa(req);
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 4);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = GFX_OP_PIPE_CONTROL(4);
*cs++ = flags;
*cs++ = scratch_addr;
*cs++ = 0;
intel_ring_advance(req, cs);
return 0;
}
static int
gen8_render_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
u32 flags;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
cs = intel_ring_begin(req, mode & EMIT_INVALIDATE ? 12 : 6);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
if (IS_ERR(cs))
return PTR_ERR(cs);
flags = PIPE_CONTROL_CS_STALL;
if (mode & EMIT_FLUSH) {
flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH;
flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH;
flags |= PIPE_CONTROL_DC_FLUSH_ENABLE;
flags |= PIPE_CONTROL_FLUSH_ENABLE;
}
if (mode & EMIT_INVALIDATE) {
flags |= PIPE_CONTROL_TLB_INVALIDATE;
flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE;
flags |= PIPE_CONTROL_QW_WRITE;
flags |= PIPE_CONTROL_GLOBAL_GTT_IVB;
/* WaCsStallBeforeStateCacheInvalidate:bdw,chv */
cs = gen8_emit_pipe_control(cs,
PIPE_CONTROL_CS_STALL |
PIPE_CONTROL_STALL_AT_SCOREBOARD,
0);
}
cs = gen8_emit_pipe_control(cs, flags,
i915_ggtt_offset(req->engine->scratch) +
2 * CACHELINE_BYTES);
intel_ring_advance(req, cs);
return 0;
}
static void ring_setup_phys_status_page(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
u32 addr;
addr = dev_priv->status_page_dmah->busaddr;
if (INTEL_GEN(dev_priv) >= 4)
addr |= (dev_priv->status_page_dmah->busaddr >> 28) & 0xf0;
I915_WRITE(HWS_PGA, addr);
}
static void intel_ring_setup_status_page(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
drm/i915: Type safe register read/write Make I915_READ and I915_WRITE more type safe by wrapping the register offset in a struct. This should eliminate most of the fumbles we've had with misplaced parens. This only takes care of normal mmio registers. We could extend the idea to other register types and define each with its own struct. That way you wouldn't be able to accidentally pass the wrong thing to a specific register access function. The gpio_reg setup is probably the ugliest thing left. But I figure I'd just leave it for now, and wait for some divine inspiration to strike before making it nice. As for the generated code, it's actually a bit better sometimes. Eg. looking at i915_irq_handler(), we can see the following change: lea 0x70024(%rdx,%rax,1),%r9d mov $0x1,%edx - movslq %r9d,%r9 - mov %r9,%rsi - mov %r9,-0x58(%rbp) - callq *0xd8(%rbx) + mov %r9d,%esi + mov %r9d,-0x48(%rbp) callq *0xd8(%rbx) So previously gcc thought the register offset might be signed and decided to sign extend it, just in case. The rest appears to be mostly just minor shuffling of instructions. v2: i915_mmio_reg_{offset,equal,valid}() helpers added s/_REG/_MMIO/ in the register defines mo more switch statements left to worry about ring_emit stuff got sorted in a prep patch cmd parser, lrc context and w/a batch buildup also in prep patch vgpu stuff cleaned up and moved to a prep patch all other unrelated changes split out v3: Rebased due to BXT DSI/BLC, MOCS, etc. v4: Rebased due to churn, s/i915_mmio_reg_t/i915_reg_t/ Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Link: http://patchwork.freedesktop.org/patch/msgid/1447853606-2751-1-git-send-email-ville.syrjala@linux.intel.com
2015-11-18 13:33:26 +00:00
i915_reg_t mmio;
/* The ring status page addresses are no longer next to the rest of
* the ring registers as of gen7.
*/
if (IS_GEN7(dev_priv)) {
switch (engine->id) {
/*
* No more rings exist on Gen7. Default case is only to shut up
* gcc switch check warning.
*/
default:
GEM_BUG_ON(engine->id);
case RCS:
mmio = RENDER_HWS_PGA_GEN7;
break;
case BCS:
mmio = BLT_HWS_PGA_GEN7;
break;
case VCS:
mmio = BSD_HWS_PGA_GEN7;
break;
case VECS:
mmio = VEBOX_HWS_PGA_GEN7;
break;
}
} else if (IS_GEN6(dev_priv)) {
mmio = RING_HWS_PGA_GEN6(engine->mmio_base);
} else {
/* XXX: gen8 returns to sanity */
mmio = RING_HWS_PGA(engine->mmio_base);
}
if (INTEL_GEN(dev_priv) >= 6)
I915_WRITE(RING_HWSTAM(engine->mmio_base), 0xffffffff);
I915_WRITE(mmio, engine->status_page.ggtt_offset);
POSTING_READ(mmio);
/*
* Flush the TLB for this page
*
* FIXME: These two bits have disappeared on gen8, so a question
* arises: do we still need this and if so how should we go about
* invalidating the TLB?
*/
if (IS_GEN(dev_priv, 6, 7)) {
i915_reg_t reg = RING_INSTPM(engine->mmio_base);
/* ring should be idle before issuing a sync flush*/
WARN_ON((I915_READ_MODE(engine) & MODE_IDLE) == 0);
I915_WRITE(reg,
_MASKED_BIT_ENABLE(INSTPM_TLB_INVALIDATE |
INSTPM_SYNC_FLUSH));
if (intel_wait_for_register(dev_priv,
reg, INSTPM_SYNC_FLUSH, 0,
1000))
DRM_ERROR("%s: wait for SyncFlush to complete for TLB invalidation timed out\n",
engine->name);
}
}
static bool stop_ring(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
if (INTEL_GEN(dev_priv) > 2) {
I915_WRITE_MODE(engine, _MASKED_BIT_ENABLE(STOP_RING));
if (intel_wait_for_register(dev_priv,
RING_MI_MODE(engine->mmio_base),
MODE_IDLE,
MODE_IDLE,
1000)) {
DRM_ERROR("%s : timed out trying to stop ring\n",
engine->name);
/* Sometimes we observe that the idle flag is not
* set even though the ring is empty. So double
* check before giving up.
*/
if (I915_READ_HEAD(engine) != I915_READ_TAIL(engine))
return false;
}
}
I915_WRITE_CTL(engine, 0);
I915_WRITE_HEAD(engine, 0);
I915_WRITE_TAIL(engine, 0);
return (I915_READ_HEAD(engine) & HEAD_ADDR) == 0;
}
static int init_ring_common(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
struct intel_ring *ring = engine->buffer;
int ret = 0;
intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
if (!stop_ring(engine)) {
/* G45 ring initialization often fails to reset head to zero */
DRM_DEBUG_KMS("%s head not reset to zero "
"ctl %08x head %08x tail %08x start %08x\n",
engine->name,
I915_READ_CTL(engine),
I915_READ_HEAD(engine),
I915_READ_TAIL(engine),
I915_READ_START(engine));
if (!stop_ring(engine)) {
DRM_ERROR("failed to set %s head to zero "
"ctl %08x head %08x tail %08x start %08x\n",
engine->name,
I915_READ_CTL(engine),
I915_READ_HEAD(engine),
I915_READ_TAIL(engine),
I915_READ_START(engine));
ret = -EIO;
goto out;
}
}
if (HWS_NEEDS_PHYSICAL(dev_priv))
ring_setup_phys_status_page(engine);
else
intel_ring_setup_status_page(engine);
intel_engine_reset_breadcrumbs(engine);
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
/* Enforce ordering by reading HEAD register back */
I915_READ_HEAD(engine);
/* Initialize the ring. This must happen _after_ we've cleared the ring
* registers with the above sequence (the readback of the HEAD registers
* also enforces ordering), otherwise the hw might lose the new ring
* register values. */
I915_WRITE_START(engine, i915_ggtt_offset(ring->vma));
/* WaClearRingBufHeadRegAtInit:ctg,elk */
if (I915_READ_HEAD(engine))
DRM_DEBUG("%s initialization failed [head=%08x], fudging\n",
engine->name, I915_READ_HEAD(engine));
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
intel_ring_update_space(ring);
I915_WRITE_HEAD(engine, ring->head);
I915_WRITE_TAIL(engine, ring->tail);
(void)I915_READ_TAIL(engine);
I915_WRITE_CTL(engine, RING_CTL_SIZE(ring->size) | RING_VALID);
/* If the head is still not zero, the ring is dead */
if (intel_wait_for_register(dev_priv, RING_CTL(engine->mmio_base),
RING_VALID, RING_VALID,
50)) {
DRM_ERROR("%s initialization failed "
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
"ctl %08x (valid? %d) head %08x [%08x] tail %08x [%08x] start %08x [expected %08x]\n",
engine->name,
I915_READ_CTL(engine),
I915_READ_CTL(engine) & RING_VALID,
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
I915_READ_HEAD(engine), ring->head,
I915_READ_TAIL(engine), ring->tail,
I915_READ_START(engine),
i915_ggtt_offset(ring->vma));
ret = -EIO;
goto out;
}
intel_engine_init_hangcheck(engine);
if (INTEL_GEN(dev_priv) > 2)
I915_WRITE_MODE(engine, _MASKED_BIT_DISABLE(STOP_RING));
out:
intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
return ret;
}
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
static void reset_ring_common(struct intel_engine_cs *engine,
struct drm_i915_gem_request *request)
{
/*
* RC6 must be prevented until the reset is complete and the engine
* reinitialised. If it occurs in the middle of this sequence, the
* state written to/loaded from the power context is ill-defined (e.g.
* the PP_BASE_DIR may be lost).
*/
assert_forcewakes_active(engine->i915, FORCEWAKE_ALL);
/*
* Try to restore the logical GPU state to match the continuation
* of the request queue. If we skip the context/PD restore, then
* the next request may try to execute assuming that its context
* is valid and loaded on the GPU and so may try to access invalid
* memory, prompting repeated GPU hangs.
*
* If the request was guilty, we still restore the logical state
* in case the next request requires it (e.g. the aliasing ppgtt),
* but skip over the hung batch.
*
* If the request was innocent, we try to replay the request with
* the restored context.
*/
if (request) {
struct drm_i915_private *dev_priv = request->i915;
struct intel_context *ce = &request->ctx->engine[engine->id];
struct i915_hw_ppgtt *ppgtt;
/* FIXME consider gen8 reset */
if (ce->state) {
I915_WRITE(CCID,
i915_ggtt_offset(ce->state) |
BIT(8) /* must be set! */ |
CCID_EXTENDED_STATE_SAVE |
CCID_EXTENDED_STATE_RESTORE |
CCID_EN);
}
ppgtt = request->ctx->ppgtt ?: engine->i915->mm.aliasing_ppgtt;
if (ppgtt) {
u32 pd_offset = ppgtt->pd.base.ggtt_offset << 10;
I915_WRITE(RING_PP_DIR_DCLV(engine), PP_DIR_DCLV_2G);
I915_WRITE(RING_PP_DIR_BASE(engine), pd_offset);
/* Wait for the PD reload to complete */
if (intel_wait_for_register(dev_priv,
RING_PP_DIR_BASE(engine),
BIT(0), 0,
10))
DRM_ERROR("Wait for reload of ppgtt page-directory timed out\n");
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine);
}
/* If the rq hung, jump to its breadcrumb and skip the batch */
if (request->fence.error == -EIO)
request->ring->head = request->postfix;
} else {
engine->legacy_active_context = NULL;
}
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
}
static int intel_rcs_ctx_init(struct drm_i915_gem_request *req)
{
int ret;
ret = intel_ring_workarounds_emit(req);
if (ret != 0)
return ret;
ret = i915_gem_render_state_emit(req);
if (ret)
return ret;
return 0;
}
static int init_render_ring(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
int ret = init_ring_common(engine);
if (ret)
return ret;
/* WaTimedSingleVertexDispatch:cl,bw,ctg,elk,ilk,snb */
if (IS_GEN(dev_priv, 4, 6))
I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(VS_TIMER_DISPATCH));
/* We need to disable the AsyncFlip performance optimisations in order
* to use MI_WAIT_FOR_EVENT within the CS. It should already be
* programmed to '1' on all products.
*
* WaDisableAsyncFlipPerfMode:snb,ivb,hsw,vlv
*/
if (IS_GEN(dev_priv, 6, 7))
I915_WRITE(MI_MODE, _MASKED_BIT_ENABLE(ASYNC_FLIP_PERF_DISABLE));
/* Required for the hardware to program scanline values for waiting */
/* WaEnableFlushTlbInvalidationMode:snb */
if (IS_GEN6(dev_priv))
I915_WRITE(GFX_MODE,
_MASKED_BIT_ENABLE(GFX_TLB_INVALIDATE_EXPLICIT));
/* WaBCSVCSTlbInvalidationMode:ivb,vlv,hsw */
if (IS_GEN7(dev_priv))
I915_WRITE(GFX_MODE_GEN7,
_MASKED_BIT_ENABLE(GFX_TLB_INVALIDATE_EXPLICIT) |
_MASKED_BIT_ENABLE(GFX_REPLAY_MODE));
if (IS_GEN6(dev_priv)) {
/* From the Sandybridge PRM, volume 1 part 3, page 24:
* "If this bit is set, STCunit will have LRA as replacement
* policy. [...] This bit must be reset. LRA replacement
* policy is not supported."
*/
I915_WRITE(CACHE_MODE_0,
_MASKED_BIT_DISABLE(CM0_STC_EVICT_DISABLE_LRA_SNB));
}
if (IS_GEN(dev_priv, 6, 7))
I915_WRITE(INSTPM, _MASKED_BIT_ENABLE(INSTPM_FORCE_ORDERING));
if (INTEL_INFO(dev_priv)->gen >= 6)
I915_WRITE_IMR(engine, ~engine->irq_keep_mask);
return init_workarounds_ring(engine);
}
static void render_ring_cleanup(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
i915_vma_unpin_and_release(&dev_priv->semaphore);
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
static u32 *gen8_rcs_signal(struct drm_i915_gem_request *req, u32 *cs)
{
struct drm_i915_private *dev_priv = req->i915;
struct intel_engine_cs *waiter;
enum intel_engine_id id;
drm/i915: Allocate intel_engine_cs structure only for the enabled engines With the possibility of addition of many more number of rings in future, the drm_i915_private structure could bloat as an array, of type intel_engine_cs, is embedded inside it. struct intel_engine_cs engine[I915_NUM_ENGINES]; Though this is still fine as generally there is only a single instance of drm_i915_private structure used, but not all of the possible rings would be enabled or active on most of the platforms. Some memory can be saved by allocating intel_engine_cs structure only for the enabled/active engines. Currently the engine/ring ID is kept static and dev_priv->engine[] is simply indexed using the enums defined in intel_engine_id. To save memory and continue using the static engine/ring IDs, 'engine' is defined as an array of pointers. struct intel_engine_cs *engine[I915_NUM_ENGINES]; dev_priv->engine[engine_ID] will be NULL for disabled engine instances. There is a text size reduction of 928 bytes, from 1028200 to 1027272, for i915.o file (but for i915.ko file text size remain same as 1193131 bytes). v2: - Remove the engine iterator field added in drm_i915_private structure, instead pass a local iterator variable to the for_each_engine** macros. (Chris) - Do away with intel_engine_initialized() and instead directly use the NULL pointer check on engine pointer. (Chris) v3: - Remove for_each_engine_id() macro, as the updated macro for_each_engine() can be used in place of it. (Chris) - Protect the access to Render engine Fault register with a NULL check, as engine specific init is done later in Driver load sequence. v4: - Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris) - Kill the superfluous init_engine_lists(). v5: - Cleanup the intel_engines_init() & intel_engines_setup(), with respect to allocation of intel_engine_cs structure. (Chris) v6: - Rebase. v7: - Optimize the for_each_engine_masked() macro. (Chris) - Change the type of 'iter' local variable to enum intel_engine_id. (Chris) - Rebase. v8: Rebase. v9: Rebase. v10: - For index calculation use engine ID instead of pointer based arithmetic in intel_engine_sync_index() as engine pointers are not contiguous now (Chris) - For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas) - Use for_each_engine macro for cleanup in intel_engines_init() and remove check for NULL engine pointer in cleanup() routines. (Joonas) v11: Rebase. Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Akash Goel <akash.goel@intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com
2016-10-13 17:14:48 +00:00
for_each_engine(waiter, dev_priv, id) {
u64 gtt_offset = req->engine->semaphore.signal_ggtt[id];
if (gtt_offset == MI_SEMAPHORE_SYNC_INVALID)
continue;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = GFX_OP_PIPE_CONTROL(6);
*cs++ = PIPE_CONTROL_GLOBAL_GTT_IVB | PIPE_CONTROL_QW_WRITE |
PIPE_CONTROL_CS_STALL;
*cs++ = lower_32_bits(gtt_offset);
*cs++ = upper_32_bits(gtt_offset);
*cs++ = req->global_seqno;
*cs++ = 0;
*cs++ = MI_SEMAPHORE_SIGNAL |
MI_SEMAPHORE_TARGET(waiter->hw_id);
*cs++ = 0;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
return cs;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
static u32 *gen8_xcs_signal(struct drm_i915_gem_request *req, u32 *cs)
{
struct drm_i915_private *dev_priv = req->i915;
struct intel_engine_cs *waiter;
enum intel_engine_id id;
drm/i915: Allocate intel_engine_cs structure only for the enabled engines With the possibility of addition of many more number of rings in future, the drm_i915_private structure could bloat as an array, of type intel_engine_cs, is embedded inside it. struct intel_engine_cs engine[I915_NUM_ENGINES]; Though this is still fine as generally there is only a single instance of drm_i915_private structure used, but not all of the possible rings would be enabled or active on most of the platforms. Some memory can be saved by allocating intel_engine_cs structure only for the enabled/active engines. Currently the engine/ring ID is kept static and dev_priv->engine[] is simply indexed using the enums defined in intel_engine_id. To save memory and continue using the static engine/ring IDs, 'engine' is defined as an array of pointers. struct intel_engine_cs *engine[I915_NUM_ENGINES]; dev_priv->engine[engine_ID] will be NULL for disabled engine instances. There is a text size reduction of 928 bytes, from 1028200 to 1027272, for i915.o file (but for i915.ko file text size remain same as 1193131 bytes). v2: - Remove the engine iterator field added in drm_i915_private structure, instead pass a local iterator variable to the for_each_engine** macros. (Chris) - Do away with intel_engine_initialized() and instead directly use the NULL pointer check on engine pointer. (Chris) v3: - Remove for_each_engine_id() macro, as the updated macro for_each_engine() can be used in place of it. (Chris) - Protect the access to Render engine Fault register with a NULL check, as engine specific init is done later in Driver load sequence. v4: - Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris) - Kill the superfluous init_engine_lists(). v5: - Cleanup the intel_engines_init() & intel_engines_setup(), with respect to allocation of intel_engine_cs structure. (Chris) v6: - Rebase. v7: - Optimize the for_each_engine_masked() macro. (Chris) - Change the type of 'iter' local variable to enum intel_engine_id. (Chris) - Rebase. v8: Rebase. v9: Rebase. v10: - For index calculation use engine ID instead of pointer based arithmetic in intel_engine_sync_index() as engine pointers are not contiguous now (Chris) - For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas) - Use for_each_engine macro for cleanup in intel_engines_init() and remove check for NULL engine pointer in cleanup() routines. (Joonas) v11: Rebase. Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Akash Goel <akash.goel@intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com
2016-10-13 17:14:48 +00:00
for_each_engine(waiter, dev_priv, id) {
u64 gtt_offset = req->engine->semaphore.signal_ggtt[id];
if (gtt_offset == MI_SEMAPHORE_SYNC_INVALID)
continue;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = (MI_FLUSH_DW + 1) | MI_FLUSH_DW_OP_STOREDW;
*cs++ = lower_32_bits(gtt_offset) | MI_FLUSH_DW_USE_GTT;
*cs++ = upper_32_bits(gtt_offset);
*cs++ = req->global_seqno;
*cs++ = MI_SEMAPHORE_SIGNAL |
MI_SEMAPHORE_TARGET(waiter->hw_id);
*cs++ = 0;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
return cs;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
static u32 *gen6_signal(struct drm_i915_gem_request *req, u32 *cs)
{
struct drm_i915_private *dev_priv = req->i915;
struct intel_engine_cs *engine;
drm/i915: Allocate intel_engine_cs structure only for the enabled engines With the possibility of addition of many more number of rings in future, the drm_i915_private structure could bloat as an array, of type intel_engine_cs, is embedded inside it. struct intel_engine_cs engine[I915_NUM_ENGINES]; Though this is still fine as generally there is only a single instance of drm_i915_private structure used, but not all of the possible rings would be enabled or active on most of the platforms. Some memory can be saved by allocating intel_engine_cs structure only for the enabled/active engines. Currently the engine/ring ID is kept static and dev_priv->engine[] is simply indexed using the enums defined in intel_engine_id. To save memory and continue using the static engine/ring IDs, 'engine' is defined as an array of pointers. struct intel_engine_cs *engine[I915_NUM_ENGINES]; dev_priv->engine[engine_ID] will be NULL for disabled engine instances. There is a text size reduction of 928 bytes, from 1028200 to 1027272, for i915.o file (but for i915.ko file text size remain same as 1193131 bytes). v2: - Remove the engine iterator field added in drm_i915_private structure, instead pass a local iterator variable to the for_each_engine** macros. (Chris) - Do away with intel_engine_initialized() and instead directly use the NULL pointer check on engine pointer. (Chris) v3: - Remove for_each_engine_id() macro, as the updated macro for_each_engine() can be used in place of it. (Chris) - Protect the access to Render engine Fault register with a NULL check, as engine specific init is done later in Driver load sequence. v4: - Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris) - Kill the superfluous init_engine_lists(). v5: - Cleanup the intel_engines_init() & intel_engines_setup(), with respect to allocation of intel_engine_cs structure. (Chris) v6: - Rebase. v7: - Optimize the for_each_engine_masked() macro. (Chris) - Change the type of 'iter' local variable to enum intel_engine_id. (Chris) - Rebase. v8: Rebase. v9: Rebase. v10: - For index calculation use engine ID instead of pointer based arithmetic in intel_engine_sync_index() as engine pointers are not contiguous now (Chris) - For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas) - Use for_each_engine macro for cleanup in intel_engines_init() and remove check for NULL engine pointer in cleanup() routines. (Joonas) v11: Rebase. Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Akash Goel <akash.goel@intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com
2016-10-13 17:14:48 +00:00
enum intel_engine_id id;
int num_rings = 0;
drm/i915: Allocate intel_engine_cs structure only for the enabled engines With the possibility of addition of many more number of rings in future, the drm_i915_private structure could bloat as an array, of type intel_engine_cs, is embedded inside it. struct intel_engine_cs engine[I915_NUM_ENGINES]; Though this is still fine as generally there is only a single instance of drm_i915_private structure used, but not all of the possible rings would be enabled or active on most of the platforms. Some memory can be saved by allocating intel_engine_cs structure only for the enabled/active engines. Currently the engine/ring ID is kept static and dev_priv->engine[] is simply indexed using the enums defined in intel_engine_id. To save memory and continue using the static engine/ring IDs, 'engine' is defined as an array of pointers. struct intel_engine_cs *engine[I915_NUM_ENGINES]; dev_priv->engine[engine_ID] will be NULL for disabled engine instances. There is a text size reduction of 928 bytes, from 1028200 to 1027272, for i915.o file (but for i915.ko file text size remain same as 1193131 bytes). v2: - Remove the engine iterator field added in drm_i915_private structure, instead pass a local iterator variable to the for_each_engine** macros. (Chris) - Do away with intel_engine_initialized() and instead directly use the NULL pointer check on engine pointer. (Chris) v3: - Remove for_each_engine_id() macro, as the updated macro for_each_engine() can be used in place of it. (Chris) - Protect the access to Render engine Fault register with a NULL check, as engine specific init is done later in Driver load sequence. v4: - Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris) - Kill the superfluous init_engine_lists(). v5: - Cleanup the intel_engines_init() & intel_engines_setup(), with respect to allocation of intel_engine_cs structure. (Chris) v6: - Rebase. v7: - Optimize the for_each_engine_masked() macro. (Chris) - Change the type of 'iter' local variable to enum intel_engine_id. (Chris) - Rebase. v8: Rebase. v9: Rebase. v10: - For index calculation use engine ID instead of pointer based arithmetic in intel_engine_sync_index() as engine pointers are not contiguous now (Chris) - For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas) - Use for_each_engine macro for cleanup in intel_engines_init() and remove check for NULL engine pointer in cleanup() routines. (Joonas) v11: Rebase. Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Akash Goel <akash.goel@intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com
2016-10-13 17:14:48 +00:00
for_each_engine(engine, dev_priv, id) {
i915_reg_t mbox_reg;
if (!(BIT(engine->hw_id) & GEN6_SEMAPHORES_MASK))
continue;
drm/i915: Type safe register read/write Make I915_READ and I915_WRITE more type safe by wrapping the register offset in a struct. This should eliminate most of the fumbles we've had with misplaced parens. This only takes care of normal mmio registers. We could extend the idea to other register types and define each with its own struct. That way you wouldn't be able to accidentally pass the wrong thing to a specific register access function. The gpio_reg setup is probably the ugliest thing left. But I figure I'd just leave it for now, and wait for some divine inspiration to strike before making it nice. As for the generated code, it's actually a bit better sometimes. Eg. looking at i915_irq_handler(), we can see the following change: lea 0x70024(%rdx,%rax,1),%r9d mov $0x1,%edx - movslq %r9d,%r9 - mov %r9,%rsi - mov %r9,-0x58(%rbp) - callq *0xd8(%rbx) + mov %r9d,%esi + mov %r9d,-0x48(%rbp) callq *0xd8(%rbx) So previously gcc thought the register offset might be signed and decided to sign extend it, just in case. The rest appears to be mostly just minor shuffling of instructions. v2: i915_mmio_reg_{offset,equal,valid}() helpers added s/_REG/_MMIO/ in the register defines mo more switch statements left to worry about ring_emit stuff got sorted in a prep patch cmd parser, lrc context and w/a batch buildup also in prep patch vgpu stuff cleaned up and moved to a prep patch all other unrelated changes split out v3: Rebased due to BXT DSI/BLC, MOCS, etc. v4: Rebased due to churn, s/i915_mmio_reg_t/i915_reg_t/ Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Link: http://patchwork.freedesktop.org/patch/msgid/1447853606-2751-1-git-send-email-ville.syrjala@linux.intel.com
2015-11-18 13:33:26 +00:00
mbox_reg = req->engine->semaphore.mbox.signal[engine->hw_id];
drm/i915: Type safe register read/write Make I915_READ and I915_WRITE more type safe by wrapping the register offset in a struct. This should eliminate most of the fumbles we've had with misplaced parens. This only takes care of normal mmio registers. We could extend the idea to other register types and define each with its own struct. That way you wouldn't be able to accidentally pass the wrong thing to a specific register access function. The gpio_reg setup is probably the ugliest thing left. But I figure I'd just leave it for now, and wait for some divine inspiration to strike before making it nice. As for the generated code, it's actually a bit better sometimes. Eg. looking at i915_irq_handler(), we can see the following change: lea 0x70024(%rdx,%rax,1),%r9d mov $0x1,%edx - movslq %r9d,%r9 - mov %r9,%rsi - mov %r9,-0x58(%rbp) - callq *0xd8(%rbx) + mov %r9d,%esi + mov %r9d,-0x48(%rbp) callq *0xd8(%rbx) So previously gcc thought the register offset might be signed and decided to sign extend it, just in case. The rest appears to be mostly just minor shuffling of instructions. v2: i915_mmio_reg_{offset,equal,valid}() helpers added s/_REG/_MMIO/ in the register defines mo more switch statements left to worry about ring_emit stuff got sorted in a prep patch cmd parser, lrc context and w/a batch buildup also in prep patch vgpu stuff cleaned up and moved to a prep patch all other unrelated changes split out v3: Rebased due to BXT DSI/BLC, MOCS, etc. v4: Rebased due to churn, s/i915_mmio_reg_t/i915_reg_t/ Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Link: http://patchwork.freedesktop.org/patch/msgid/1447853606-2751-1-git-send-email-ville.syrjala@linux.intel.com
2015-11-18 13:33:26 +00:00
if (i915_mmio_reg_valid(mbox_reg)) {
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_LOAD_REGISTER_IMM(1);
*cs++ = i915_mmio_reg_offset(mbox_reg);
*cs++ = req->global_seqno;
num_rings++;
}
}
if (num_rings & 1)
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_NOOP;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
return cs;
}
static void cancel_requests(struct intel_engine_cs *engine)
{
struct drm_i915_gem_request *request;
unsigned long flags;
spin_lock_irqsave(&engine->timeline->lock, flags);
/* Mark all submitted requests as skipped. */
list_for_each_entry(request, &engine->timeline->requests, link) {
GEM_BUG_ON(!request->global_seqno);
if (!i915_gem_request_completed(request))
dma_fence_set_error(&request->fence, -EIO);
}
/* Remaining _unready_ requests will be nop'ed when submitted */
spin_unlock_irqrestore(&engine->timeline->lock, flags);
}
static void i9xx_submit_request(struct drm_i915_gem_request *request)
{
struct drm_i915_private *dev_priv = request->i915;
i915_gem_request_submit(request);
I915_WRITE_TAIL(request->engine,
intel_ring_set_tail(request->ring, request->tail));
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
static void i9xx_emit_breadcrumb(struct drm_i915_gem_request *req, u32 *cs)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_STORE_DWORD_INDEX;
*cs++ = I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT;
*cs++ = req->global_seqno;
*cs++ = MI_USER_INTERRUPT;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
req->tail = intel_ring_offset(req, cs);
assert_ring_tail_valid(req->ring, req->tail);
}
static const int i9xx_emit_breadcrumb_sz = 4;
/**
* gen6_sema_emit_breadcrumb - Update the semaphore mailbox registers
*
* @request - request to write to the ring
*
* Update the mailbox registers in the *other* rings with the current seqno.
* This acts like a signal in the canonical semaphore.
*/
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
static void gen6_sema_emit_breadcrumb(struct drm_i915_gem_request *req, u32 *cs)
{
return i9xx_emit_breadcrumb(req,
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
req->engine->semaphore.signal(req, cs));
}
static void gen8_render_emit_breadcrumb(struct drm_i915_gem_request *req,
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs)
{
struct intel_engine_cs *engine = req->engine;
if (engine->semaphore.signal)
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = engine->semaphore.signal(req, cs);
*cs++ = GFX_OP_PIPE_CONTROL(6);
*cs++ = PIPE_CONTROL_GLOBAL_GTT_IVB | PIPE_CONTROL_CS_STALL |
PIPE_CONTROL_QW_WRITE;
*cs++ = intel_hws_seqno_address(engine);
*cs++ = 0;
*cs++ = req->global_seqno;
/* We're thrashing one dword of HWS. */
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = 0;
*cs++ = MI_USER_INTERRUPT;
*cs++ = MI_NOOP;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
req->tail = intel_ring_offset(req, cs);
assert_ring_tail_valid(req->ring, req->tail);
}
static const int gen8_render_emit_breadcrumb_sz = 8;
/**
* intel_ring_sync - sync the waiter to the signaller on seqno
*
* @waiter - ring that is waiting
* @signaller - ring which has, or will signal
* @seqno - seqno which the waiter will block on
*/
static int
gen8_ring_sync_to(struct drm_i915_gem_request *req,
struct drm_i915_gem_request *signal)
{
struct drm_i915_private *dev_priv = req->i915;
u64 offset = GEN8_WAIT_OFFSET(req->engine, signal->engine->id);
struct i915_hw_ppgtt *ppgtt;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 4);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_SEMAPHORE_WAIT | MI_SEMAPHORE_GLOBAL_GTT |
MI_SEMAPHORE_SAD_GTE_SDD;
*cs++ = signal->global_seqno;
*cs++ = lower_32_bits(offset);
*cs++ = upper_32_bits(offset);
intel_ring_advance(req, cs);
/* When the !RCS engines idle waiting upon a semaphore, they lose their
* pagetables and we must reload them before executing the batch.
* We do this on the i915_switch_context() following the wait and
* before the dispatch.
*/
ppgtt = req->ctx->ppgtt;
if (ppgtt && req->engine->id != RCS)
ppgtt->pd_dirty_rings |= intel_engine_flag(req->engine);
return 0;
}
static int
gen6_ring_sync_to(struct drm_i915_gem_request *req,
struct drm_i915_gem_request *signal)
{
u32 dw1 = MI_SEMAPHORE_MBOX |
MI_SEMAPHORE_COMPARE |
MI_SEMAPHORE_REGISTER;
u32 wait_mbox = signal->engine->semaphore.mbox.wait[req->engine->hw_id];
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
WARN_ON(wait_mbox == MI_SEMAPHORE_SYNC_INVALID);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 4);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = dw1 | wait_mbox;
/* Throughout all of the GEM code, seqno passed implies our current
* seqno is >= the last seqno executed. However for hardware the
* comparison is strictly greater than.
*/
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = signal->global_seqno - 1;
*cs++ = 0;
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
return 0;
}
drm/i915: Add a delay between interrupt and inspecting the final seqno (ilk) On Ironlake, there is no command nor register to ensure that the write from a MI_STORE command is completed (and coherent on the CPU) before the command parser continues. This means that the ordering between the seqno write and the subsequent user interrupt is undefined (like gen6+). So to ensure that the seqno write is completed after the final user interrupt we need to delay the read sufficiently to allow the write to complete. This delay is undefined by the bspec, and empirically requires 75us even though a register read combined with a clflush is less than 500ns. Hence, the delay is due to an on-chip buffer rather than the latency of the write to memory. Note that the render ring controls this by filling the PIPE_CONTROL fifo with stalling commands that force the earliest pipe-control with the seqno to be completed before the command parser continues. Given that we need a barrier operation for BSD, we may as well forgo the extra per-batch latency by using a common per-interrupt barrier. Studying the impact of adding the usleep shows that in both sequences of and individual synchronous no-op batches is negligible for the media engine (where the write now is unordered with the interrupt). Converting the render engine over from the current glutton of pie-controls over to the per-interrupt delays speeds up both the sequential and individual synchronous no-ops by 20% and 60%, respectively. This speed up holds even when looking at the throughput of small copies (4KiB->4MiB), both serial and synchronous, by about 20%. This is because despite adding a significant delay to the interrupt, in all likelihood we will see the seqno write without having to apply the barrier (only in the rare corner cases where the write is delayed on the last required is the delay necessary). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94307 Testcase: igt/gem_sync #ilk Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-12-git-send-email-chris@chris-wilson.co.uk
2016-07-01 16:23:21 +00:00
static void
gen5_seqno_barrier(struct intel_engine_cs *engine)
{
drm/i915: Add a delay between interrupt and inspecting the final seqno (ilk) On Ironlake, there is no command nor register to ensure that the write from a MI_STORE command is completed (and coherent on the CPU) before the command parser continues. This means that the ordering between the seqno write and the subsequent user interrupt is undefined (like gen6+). So to ensure that the seqno write is completed after the final user interrupt we need to delay the read sufficiently to allow the write to complete. This delay is undefined by the bspec, and empirically requires 75us even though a register read combined with a clflush is less than 500ns. Hence, the delay is due to an on-chip buffer rather than the latency of the write to memory. Note that the render ring controls this by filling the PIPE_CONTROL fifo with stalling commands that force the earliest pipe-control with the seqno to be completed before the command parser continues. Given that we need a barrier operation for BSD, we may as well forgo the extra per-batch latency by using a common per-interrupt barrier. Studying the impact of adding the usleep shows that in both sequences of and individual synchronous no-op batches is negligible for the media engine (where the write now is unordered with the interrupt). Converting the render engine over from the current glutton of pie-controls over to the per-interrupt delays speeds up both the sequential and individual synchronous no-ops by 20% and 60%, respectively. This speed up holds even when looking at the throughput of small copies (4KiB->4MiB), both serial and synchronous, by about 20%. This is because despite adding a significant delay to the interrupt, in all likelihood we will see the seqno write without having to apply the barrier (only in the rare corner cases where the write is delayed on the last required is the delay necessary). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94307 Testcase: igt/gem_sync #ilk Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-12-git-send-email-chris@chris-wilson.co.uk
2016-07-01 16:23:21 +00:00
/* MI_STORE are internally buffered by the GPU and not flushed
* either by MI_FLUSH or SyncFlush or any other combination of
* MI commands.
*
drm/i915: Add a delay between interrupt and inspecting the final seqno (ilk) On Ironlake, there is no command nor register to ensure that the write from a MI_STORE command is completed (and coherent on the CPU) before the command parser continues. This means that the ordering between the seqno write and the subsequent user interrupt is undefined (like gen6+). So to ensure that the seqno write is completed after the final user interrupt we need to delay the read sufficiently to allow the write to complete. This delay is undefined by the bspec, and empirically requires 75us even though a register read combined with a clflush is less than 500ns. Hence, the delay is due to an on-chip buffer rather than the latency of the write to memory. Note that the render ring controls this by filling the PIPE_CONTROL fifo with stalling commands that force the earliest pipe-control with the seqno to be completed before the command parser continues. Given that we need a barrier operation for BSD, we may as well forgo the extra per-batch latency by using a common per-interrupt barrier. Studying the impact of adding the usleep shows that in both sequences of and individual synchronous no-op batches is negligible for the media engine (where the write now is unordered with the interrupt). Converting the render engine over from the current glutton of pie-controls over to the per-interrupt delays speeds up both the sequential and individual synchronous no-ops by 20% and 60%, respectively. This speed up holds even when looking at the throughput of small copies (4KiB->4MiB), both serial and synchronous, by about 20%. This is because despite adding a significant delay to the interrupt, in all likelihood we will see the seqno write without having to apply the barrier (only in the rare corner cases where the write is delayed on the last required is the delay necessary). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94307 Testcase: igt/gem_sync #ilk Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-12-git-send-email-chris@chris-wilson.co.uk
2016-07-01 16:23:21 +00:00
* "Only the submission of the store operation is guaranteed.
* The write result will be complete (coherent) some time later
* (this is practically a finite period but there is no guaranteed
* latency)."
*
* Empirically, we observe that we need a delay of at least 75us to
* be sure that the seqno write is visible by the CPU.
*/
drm/i915: Add a delay between interrupt and inspecting the final seqno (ilk) On Ironlake, there is no command nor register to ensure that the write from a MI_STORE command is completed (and coherent on the CPU) before the command parser continues. This means that the ordering between the seqno write and the subsequent user interrupt is undefined (like gen6+). So to ensure that the seqno write is completed after the final user interrupt we need to delay the read sufficiently to allow the write to complete. This delay is undefined by the bspec, and empirically requires 75us even though a register read combined with a clflush is less than 500ns. Hence, the delay is due to an on-chip buffer rather than the latency of the write to memory. Note that the render ring controls this by filling the PIPE_CONTROL fifo with stalling commands that force the earliest pipe-control with the seqno to be completed before the command parser continues. Given that we need a barrier operation for BSD, we may as well forgo the extra per-batch latency by using a common per-interrupt barrier. Studying the impact of adding the usleep shows that in both sequences of and individual synchronous no-op batches is negligible for the media engine (where the write now is unordered with the interrupt). Converting the render engine over from the current glutton of pie-controls over to the per-interrupt delays speeds up both the sequential and individual synchronous no-ops by 20% and 60%, respectively. This speed up holds even when looking at the throughput of small copies (4KiB->4MiB), both serial and synchronous, by about 20%. This is because despite adding a significant delay to the interrupt, in all likelihood we will see the seqno write without having to apply the barrier (only in the rare corner cases where the write is delayed on the last required is the delay necessary). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94307 Testcase: igt/gem_sync #ilk Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-12-git-send-email-chris@chris-wilson.co.uk
2016-07-01 16:23:21 +00:00
usleep_range(125, 250);
}
static void
gen6_seqno_barrier(struct intel_engine_cs *engine)
2012-12-14 15:01:25 +00:00
{
struct drm_i915_private *dev_priv = engine->i915;
2012-12-14 15:01:25 +00:00
/* Workaround to force correct ordering between irq and seqno writes on
* ivb (and maybe also on snb) by reading from a CS register (like
* ACTHD) before reading the status page.
*
* Note that this effectively stalls the read by the time it takes to
* do a memory transaction, which more or less ensures that the write
* from the GPU has sufficient time to invalidate the CPU cacheline.
* Alternatively we could delay the interrupt from the CS ring to give
* the write time to land, but that would incur a delay after every
* batch i.e. much more frequent than a delay when waiting for the
* interrupt (with the same net latency).
*
* Also note that to prevent whole machine hangs on gen7, we have to
* take the spinlock to guard against concurrent cacheline access.
*/
spin_lock_irq(&dev_priv->uncore.lock);
POSTING_READ_FW(RING_ACTHD(engine->mmio_base));
spin_unlock_irq(&dev_priv->uncore.lock);
2012-12-14 15:01:25 +00:00
}
static void
gen5_irq_enable(struct intel_engine_cs *engine)
{
gen5_enable_gt_irq(engine->i915, engine->irq_enable_mask);
}
static void
gen5_irq_disable(struct intel_engine_cs *engine)
{
gen5_disable_gt_irq(engine->i915, engine->irq_enable_mask);
}
static void
i9xx_irq_enable(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
dev_priv->irq_mask &= ~engine->irq_enable_mask;
I915_WRITE(IMR, dev_priv->irq_mask);
POSTING_READ_FW(RING_IMR(engine->mmio_base));
}
static void
i9xx_irq_disable(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
dev_priv->irq_mask |= engine->irq_enable_mask;
I915_WRITE(IMR, dev_priv->irq_mask);
}
static void
i8xx_irq_enable(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
dev_priv->irq_mask &= ~engine->irq_enable_mask;
I915_WRITE16(IMR, dev_priv->irq_mask);
POSTING_READ16(RING_IMR(engine->mmio_base));
}
static void
i8xx_irq_disable(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
dev_priv->irq_mask |= engine->irq_enable_mask;
I915_WRITE16(IMR, dev_priv->irq_mask);
}
static int
bsd_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_FLUSH;
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
return 0;
}
static void
gen6_irq_enable(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
I915_WRITE_IMR(engine,
~(engine->irq_enable_mask |
engine->irq_keep_mask));
gen5_enable_gt_irq(dev_priv, engine->irq_enable_mask);
}
static void
gen6_irq_disable(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
I915_WRITE_IMR(engine, ~engine->irq_keep_mask);
gen5_disable_gt_irq(dev_priv, engine->irq_enable_mask);
}
static void
hsw_vebox_irq_enable(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
I915_WRITE_IMR(engine, ~engine->irq_enable_mask);
gen6_unmask_pm_irq(dev_priv, engine->irq_enable_mask);
}
static void
hsw_vebox_irq_disable(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
I915_WRITE_IMR(engine, ~0);
gen6_mask_pm_irq(dev_priv, engine->irq_enable_mask);
}
static void
gen8_irq_enable(struct intel_engine_cs *engine)
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
{
struct drm_i915_private *dev_priv = engine->i915;
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
I915_WRITE_IMR(engine,
~(engine->irq_enable_mask |
engine->irq_keep_mask));
POSTING_READ_FW(RING_IMR(engine->mmio_base));
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
}
static void
gen8_irq_disable(struct intel_engine_cs *engine)
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
{
struct drm_i915_private *dev_priv = engine->i915;
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
I915_WRITE_IMR(engine, ~engine->irq_keep_mask);
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
}
static int
i965_emit_bb_start(struct drm_i915_gem_request *req,
u64 offset, u32 length,
unsigned int dispatch_flags)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_BATCH_BUFFER_START | MI_BATCH_GTT | (dispatch_flags &
I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE_I965);
*cs++ = offset;
intel_ring_advance(req, cs);
return 0;
}
/* Just userspace ABI convention to limit the wa batch bo to a resonable size */
#define I830_BATCH_LIMIT (256*1024)
#define I830_TLB_ENTRIES (2)
#define I830_WA_SIZE max(I830_TLB_ENTRIES*4096, I830_BATCH_LIMIT)
static int
i830_emit_bb_start(struct drm_i915_gem_request *req,
u64 offset, u32 len,
unsigned int dispatch_flags)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs, cs_offset = i915_ggtt_offset(req->engine->scratch);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 6);
if (IS_ERR(cs))
return PTR_ERR(cs);
/* Evict the invalid PTE TLBs */
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = COLOR_BLT_CMD | BLT_WRITE_RGBA;
*cs++ = BLT_DEPTH_32 | BLT_ROP_COLOR_COPY | 4096;
*cs++ = I830_TLB_ENTRIES << 16 | 4; /* load each page */
*cs++ = cs_offset;
*cs++ = 0xdeadbeef;
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
if ((dispatch_flags & I915_DISPATCH_PINNED) == 0) {
if (len > I830_BATCH_LIMIT)
return -ENOSPC;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 6 + 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
/* Blit the batch (which has now all relocs applied) to the
* stable batch scratch bo area (so that the CS never
* stumbles over its tlb invalidation bug) ...
*/
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = SRC_COPY_BLT_CMD | BLT_WRITE_RGBA;
*cs++ = BLT_DEPTH_32 | BLT_ROP_SRC_COPY | 4096;
*cs++ = DIV_ROUND_UP(len, 4096) << 16 | 4096;
*cs++ = cs_offset;
*cs++ = 4096;
*cs++ = offset;
*cs++ = MI_FLUSH;
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
/* ... and execute it. */
offset = cs_offset;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_BATCH_BUFFER_START | MI_BATCH_GTT;
*cs++ = offset | (dispatch_flags & I915_DISPATCH_SECURE ? 0 :
MI_BATCH_NON_SECURE);
intel_ring_advance(req, cs);
return 0;
}
static int
i915_emit_bb_start(struct drm_i915_gem_request *req,
u64 offset, u32 len,
unsigned int dispatch_flags)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_BATCH_BUFFER_START | MI_BATCH_GTT;
*cs++ = offset | (dispatch_flags & I915_DISPATCH_SECURE ? 0 :
MI_BATCH_NON_SECURE);
intel_ring_advance(req, cs);
return 0;
}
int intel_ring_pin(struct intel_ring *ring,
struct drm_i915_private *i915,
unsigned int offset_bias)
drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo <oscar.mateo@intel.com> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com> Reviewed-by: Akash Goel <akash.goels@gmail.com> Reviewed-by: Deepak S<deepak.s@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-11-13 10:28:56 +00:00
{
enum i915_map_type map = HAS_LLC(i915) ? I915_MAP_WB : I915_MAP_WC;
struct i915_vma *vma = ring->vma;
unsigned int flags;
void *addr;
drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo <oscar.mateo@intel.com> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com> Reviewed-by: Akash Goel <akash.goels@gmail.com> Reviewed-by: Deepak S<deepak.s@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-11-13 10:28:56 +00:00
int ret;
GEM_BUG_ON(ring->vaddr);
drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo <oscar.mateo@intel.com> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com> Reviewed-by: Akash Goel <akash.goels@gmail.com> Reviewed-by: Deepak S<deepak.s@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-11-13 10:28:56 +00:00
flags = PIN_GLOBAL;
if (offset_bias)
flags |= PIN_OFFSET_BIAS | offset_bias;
if (vma->obj->stolen)
flags |= PIN_MAPPABLE;
if (!(vma->flags & I915_VMA_GLOBAL_BIND)) {
if (flags & PIN_MAPPABLE || map == I915_MAP_WC)
ret = i915_gem_object_set_to_gtt_domain(vma->obj, true);
else
ret = i915_gem_object_set_to_cpu_domain(vma->obj, true);
if (unlikely(ret))
return ret;
}
drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo <oscar.mateo@intel.com> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com> Reviewed-by: Akash Goel <akash.goels@gmail.com> Reviewed-by: Deepak S<deepak.s@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-11-13 10:28:56 +00:00
ret = i915_vma_pin(vma, 0, PAGE_SIZE, flags);
if (unlikely(ret))
return ret;
if (i915_vma_is_map_and_fenceable(vma))
addr = (void __force *)i915_vma_pin_iomap(vma);
else
addr = i915_gem_object_pin_map(vma->obj, map);
if (IS_ERR(addr))
goto err;
drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo <oscar.mateo@intel.com> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com> Reviewed-by: Akash Goel <akash.goels@gmail.com> Reviewed-by: Deepak S<deepak.s@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-11-13 10:28:56 +00:00
vma->obj->pin_global++;
ring->vaddr = addr;
drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo <oscar.mateo@intel.com> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com> Reviewed-by: Akash Goel <akash.goels@gmail.com> Reviewed-by: Deepak S<deepak.s@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-11-13 10:28:56 +00:00
return 0;
err:
i915_vma_unpin(vma);
return PTR_ERR(addr);
drm/i915/bdw: Pin the ringbuffer backing object to GGTT on-demand Same as with the context, pinning to GGTT regardless is harmful (it badly fragments the GGTT and can even exhaust it). Unfortunately, this case is also more complex than the previous one because we need to map and access the ringbuffer in several places along the execbuffer path (and we cannot make do by leaving the default ringbuffer pinned, as before). Also, the context object itself contains a pointer to the ringbuffer address that we have to keep updated if we are going to allow the ringbuffer to move around. v2: Same as with the context pinning, we cannot really do it during an interrupt. Also, pin the default ringbuffers objects regardless (makes error capture a lot easier). v3: Rebased. Take a pin reference of the ringbuffer for each item in the execlist request queue because the hardware may still be using the ringbuffer after the MI_USER_INTERRUPT to notify the seqno update is executed. The ringbuffer must remain pinned until the context save is complete. No longer pin and unpin ringbuffer in populate_lr_context() - this transient address is meaningless and the pinning can cause a sleep while atomic. v4: Moved ringbuffer pin and unpin into the lr_context_pin functions. Downgraded pinning check BUG_ONs to WARN_ONs. v5: Reinstated WARN_ONs for unexpected execlist states. Removed unused variable. Issue: VIZ-4277 Signed-off-by: Oscar Mateo <oscar.mateo@intel.com> Signed-off-by: Thomas Daniel <thomas.daniel@intel.com> Reviewed-by: Akash Goel <akash.goels@gmail.com> Reviewed-by: Deepak S<deepak.s@linux.intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2014-11-13 10:28:56 +00:00
}
void intel_ring_reset(struct intel_ring *ring, u32 tail)
{
GEM_BUG_ON(!list_empty(&ring->request_list));
ring->tail = tail;
ring->head = tail;
ring->emit = tail;
intel_ring_update_space(ring);
}
void intel_ring_unpin(struct intel_ring *ring)
{
GEM_BUG_ON(!ring->vma);
GEM_BUG_ON(!ring->vaddr);
/* Discard any unused bytes beyond that submitted to hw. */
intel_ring_reset(ring, ring->tail);
if (i915_vma_is_map_and_fenceable(ring->vma))
i915_vma_unpin_iomap(ring->vma);
else
i915_gem_object_unpin_map(ring->vma->obj);
ring->vaddr = NULL;
ring->vma->obj->pin_global--;
i915_vma_unpin(ring->vma);
}
static struct i915_vma *
intel_ring_create_vma(struct drm_i915_private *dev_priv, int size)
{
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
obj = i915_gem_object_create_stolen(dev_priv, size);
if (!obj)
obj = i915_gem_object_create_internal(dev_priv, size);
if (IS_ERR(obj))
return ERR_CAST(obj);
/* mark ring buffers as read-only from GPU side by default */
obj->gt_ro = 1;
vma = i915_vma_instance(obj, &dev_priv->ggtt.base, NULL);
if (IS_ERR(vma))
goto err;
return vma;
err:
i915_gem_object_put(obj);
return vma;
}
struct intel_ring *
intel_engine_create_ring(struct intel_engine_cs *engine, int size)
{
struct intel_ring *ring;
struct i915_vma *vma;
GEM_BUG_ON(!is_power_of_2(size));
GEM_BUG_ON(RING_CTL_SIZE(size) & ~RING_NR_PAGES);
ring = kzalloc(sizeof(*ring), GFP_KERNEL);
if (!ring)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&ring->request_list);
ring->size = size;
/* Workaround an erratum on the i830 which causes a hang if
* the TAIL pointer points to within the last 2 cachelines
* of the buffer.
*/
ring->effective_size = size;
if (IS_I830(engine->i915) || IS_I845G(engine->i915))
ring->effective_size -= 2 * CACHELINE_BYTES;
intel_ring_update_space(ring);
vma = intel_ring_create_vma(engine->i915, size);
if (IS_ERR(vma)) {
kfree(ring);
return ERR_CAST(vma);
}
ring->vma = vma;
return ring;
}
void
intel_ring_free(struct intel_ring *ring)
{
struct drm_i915_gem_object *obj = ring->vma->obj;
i915_vma_close(ring->vma);
__i915_gem_object_release_unless_active(obj);
kfree(ring);
}
static int context_pin(struct i915_gem_context *ctx)
drm/i915: Unify active context tracking between legacy/execlists/guc The requests conversion introduced a nasty bug where we could generate a new request in the middle of constructing a request if we needed to idle the system in order to evict space for a context. The request to idle would be executed (and waited upon) before the current one, creating a minor havoc in the seqno accounting, as we will consider the current request to already be completed (prior to deferred seqno assignment) but ring->last_retired_head would have been updated and still could allow us to overwrite the current request before execution. We also employed two different mechanisms to track the active context until it was switched out. The legacy method allowed for waiting upon an active context (it could forcibly evict any vma, including context's), but the execlists method took a step backwards by pinning the vma for the entire active lifespan of the context (the only way to evict was to idle the entire GPU, not individual contexts). However, to circumvent the tricky issue of locking (i.e. we cannot take struct_mutex at the time of i915_gem_request_submit(), where we would want to move the previous context onto the active tracker and unpin it), we take the execlists approach and keep the contexts pinned until retirement. The benefit of the execlists approach, more important for execlists than legacy, was the reduction in work in pinning the context for each request - as the context was kept pinned until idle, it could short circuit the pinning for all active contexts. We introduce new engine vfuncs to pin and unpin the context respectively. The context is pinned at the start of the request, and only unpinned when the following request is retired (this ensures that the context is idle and coherent in main memory before we unpin it). We move the engine->last_context tracking into the retirement itself (rather than during request submission) in order to allow the submission to be reordered or unwound without undue difficultly. And finally an ulterior motive for unifying context handling was to prepare for mock requests. v2: Rename to last_retired_context, split out legacy_context tracking for MI_SET_CONTEXT. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161218153724.8439-3-chris@chris-wilson.co.uk
2016-12-18 15:37:20 +00:00
{
struct i915_vma *vma = ctx->engine[RCS].state;
int ret;
/* Clear this page out of any CPU caches for coherent swap-in/out.
* We only want to do this on the first bind so that we do not stall
* on an active context (which by nature is already on the GPU).
*/
if (!(vma->flags & I915_VMA_GLOBAL_BIND)) {
ret = i915_gem_object_set_to_gtt_domain(vma->obj, false);
if (ret)
return ret;
}
return i915_vma_pin(vma, 0, I915_GTT_MIN_ALIGNMENT,
PIN_GLOBAL | PIN_HIGH);
drm/i915: Unify active context tracking between legacy/execlists/guc The requests conversion introduced a nasty bug where we could generate a new request in the middle of constructing a request if we needed to idle the system in order to evict space for a context. The request to idle would be executed (and waited upon) before the current one, creating a minor havoc in the seqno accounting, as we will consider the current request to already be completed (prior to deferred seqno assignment) but ring->last_retired_head would have been updated and still could allow us to overwrite the current request before execution. We also employed two different mechanisms to track the active context until it was switched out. The legacy method allowed for waiting upon an active context (it could forcibly evict any vma, including context's), but the execlists method took a step backwards by pinning the vma for the entire active lifespan of the context (the only way to evict was to idle the entire GPU, not individual contexts). However, to circumvent the tricky issue of locking (i.e. we cannot take struct_mutex at the time of i915_gem_request_submit(), where we would want to move the previous context onto the active tracker and unpin it), we take the execlists approach and keep the contexts pinned until retirement. The benefit of the execlists approach, more important for execlists than legacy, was the reduction in work in pinning the context for each request - as the context was kept pinned until idle, it could short circuit the pinning for all active contexts. We introduce new engine vfuncs to pin and unpin the context respectively. The context is pinned at the start of the request, and only unpinned when the following request is retired (this ensures that the context is idle and coherent in main memory before we unpin it). We move the engine->last_context tracking into the retirement itself (rather than during request submission) in order to allow the submission to be reordered or unwound without undue difficultly. And finally an ulterior motive for unifying context handling was to prepare for mock requests. v2: Rename to last_retired_context, split out legacy_context tracking for MI_SET_CONTEXT. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161218153724.8439-3-chris@chris-wilson.co.uk
2016-12-18 15:37:20 +00:00
}
static struct i915_vma *
alloc_context_vma(struct intel_engine_cs *engine)
{
struct drm_i915_private *i915 = engine->i915;
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
obj = i915_gem_object_create(i915, engine->context_size);
if (IS_ERR(obj))
return ERR_CAST(obj);
/*
* Try to make the context utilize L3 as well as LLC.
*
* On VLV we don't have L3 controls in the PTEs so we
* shouldn't touch the cache level, especially as that
* would make the object snooped which might have a
* negative performance impact.
*
* Snooping is required on non-llc platforms in execlist
* mode, but since all GGTT accesses use PAT entry 0 we
* get snooping anyway regardless of cache_level.
*
* This is only applicable for Ivy Bridge devices since
* later platforms don't have L3 control bits in the PTE.
*/
if (IS_IVYBRIDGE(i915)) {
/* Ignore any error, regard it as a simple optimisation */
i915_gem_object_set_cache_level(obj, I915_CACHE_L3_LLC);
}
vma = i915_vma_instance(obj, &i915->ggtt.base, NULL);
if (IS_ERR(vma))
i915_gem_object_put(obj);
return vma;
}
static struct intel_ring *
intel_ring_context_pin(struct intel_engine_cs *engine,
struct i915_gem_context *ctx)
{
struct intel_context *ce = &ctx->engine[engine->id];
int ret;
lockdep_assert_held(&ctx->i915->drm.struct_mutex);
if (likely(ce->pin_count++))
goto out;
GEM_BUG_ON(!ce->pin_count); /* no overflow please! */
if (!ce->state && engine->context_size) {
struct i915_vma *vma;
vma = alloc_context_vma(engine);
if (IS_ERR(vma)) {
ret = PTR_ERR(vma);
goto err;
}
ce->state = vma;
}
if (ce->state) {
ret = context_pin(ctx);
drm/i915: Unify active context tracking between legacy/execlists/guc The requests conversion introduced a nasty bug where we could generate a new request in the middle of constructing a request if we needed to idle the system in order to evict space for a context. The request to idle would be executed (and waited upon) before the current one, creating a minor havoc in the seqno accounting, as we will consider the current request to already be completed (prior to deferred seqno assignment) but ring->last_retired_head would have been updated and still could allow us to overwrite the current request before execution. We also employed two different mechanisms to track the active context until it was switched out. The legacy method allowed for waiting upon an active context (it could forcibly evict any vma, including context's), but the execlists method took a step backwards by pinning the vma for the entire active lifespan of the context (the only way to evict was to idle the entire GPU, not individual contexts). However, to circumvent the tricky issue of locking (i.e. we cannot take struct_mutex at the time of i915_gem_request_submit(), where we would want to move the previous context onto the active tracker and unpin it), we take the execlists approach and keep the contexts pinned until retirement. The benefit of the execlists approach, more important for execlists than legacy, was the reduction in work in pinning the context for each request - as the context was kept pinned until idle, it could short circuit the pinning for all active contexts. We introduce new engine vfuncs to pin and unpin the context respectively. The context is pinned at the start of the request, and only unpinned when the following request is retired (this ensures that the context is idle and coherent in main memory before we unpin it). We move the engine->last_context tracking into the retirement itself (rather than during request submission) in order to allow the submission to be reordered or unwound without undue difficultly. And finally an ulterior motive for unifying context handling was to prepare for mock requests. v2: Rename to last_retired_context, split out legacy_context tracking for MI_SET_CONTEXT. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161218153724.8439-3-chris@chris-wilson.co.uk
2016-12-18 15:37:20 +00:00
if (ret)
goto err;
ce->state->obj->mm.dirty = true;
ce->state->obj->pin_global++;
}
/* The kernel context is only used as a placeholder for flushing the
* active context. It is never used for submitting user rendering and
* as such never requires the golden render context, and so we can skip
* emitting it when we switch to the kernel context. This is required
* as during eviction we cannot allocate and pin the renderstate in
* order to initialise the context.
*/
if (i915_gem_context_is_kernel(ctx))
ce->initialised = true;
i915_gem_context_get(ctx);
out:
/* One ringbuffer to rule them all */
return engine->buffer;
err:
ce->pin_count = 0;
return ERR_PTR(ret);
}
drm/i915: Unify active context tracking between legacy/execlists/guc The requests conversion introduced a nasty bug where we could generate a new request in the middle of constructing a request if we needed to idle the system in order to evict space for a context. The request to idle would be executed (and waited upon) before the current one, creating a minor havoc in the seqno accounting, as we will consider the current request to already be completed (prior to deferred seqno assignment) but ring->last_retired_head would have been updated and still could allow us to overwrite the current request before execution. We also employed two different mechanisms to track the active context until it was switched out. The legacy method allowed for waiting upon an active context (it could forcibly evict any vma, including context's), but the execlists method took a step backwards by pinning the vma for the entire active lifespan of the context (the only way to evict was to idle the entire GPU, not individual contexts). However, to circumvent the tricky issue of locking (i.e. we cannot take struct_mutex at the time of i915_gem_request_submit(), where we would want to move the previous context onto the active tracker and unpin it), we take the execlists approach and keep the contexts pinned until retirement. The benefit of the execlists approach, more important for execlists than legacy, was the reduction in work in pinning the context for each request - as the context was kept pinned until idle, it could short circuit the pinning for all active contexts. We introduce new engine vfuncs to pin and unpin the context respectively. The context is pinned at the start of the request, and only unpinned when the following request is retired (this ensures that the context is idle and coherent in main memory before we unpin it). We move the engine->last_context tracking into the retirement itself (rather than during request submission) in order to allow the submission to be reordered or unwound without undue difficultly. And finally an ulterior motive for unifying context handling was to prepare for mock requests. v2: Rename to last_retired_context, split out legacy_context tracking for MI_SET_CONTEXT. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161218153724.8439-3-chris@chris-wilson.co.uk
2016-12-18 15:37:20 +00:00
static void intel_ring_context_unpin(struct intel_engine_cs *engine,
struct i915_gem_context *ctx)
{
struct intel_context *ce = &ctx->engine[engine->id];
lockdep_assert_held(&ctx->i915->drm.struct_mutex);
drm/i915: Unify active context tracking between legacy/execlists/guc The requests conversion introduced a nasty bug where we could generate a new request in the middle of constructing a request if we needed to idle the system in order to evict space for a context. The request to idle would be executed (and waited upon) before the current one, creating a minor havoc in the seqno accounting, as we will consider the current request to already be completed (prior to deferred seqno assignment) but ring->last_retired_head would have been updated and still could allow us to overwrite the current request before execution. We also employed two different mechanisms to track the active context until it was switched out. The legacy method allowed for waiting upon an active context (it could forcibly evict any vma, including context's), but the execlists method took a step backwards by pinning the vma for the entire active lifespan of the context (the only way to evict was to idle the entire GPU, not individual contexts). However, to circumvent the tricky issue of locking (i.e. we cannot take struct_mutex at the time of i915_gem_request_submit(), where we would want to move the previous context onto the active tracker and unpin it), we take the execlists approach and keep the contexts pinned until retirement. The benefit of the execlists approach, more important for execlists than legacy, was the reduction in work in pinning the context for each request - as the context was kept pinned until idle, it could short circuit the pinning for all active contexts. We introduce new engine vfuncs to pin and unpin the context respectively. The context is pinned at the start of the request, and only unpinned when the following request is retired (this ensures that the context is idle and coherent in main memory before we unpin it). We move the engine->last_context tracking into the retirement itself (rather than during request submission) in order to allow the submission to be reordered or unwound without undue difficultly. And finally an ulterior motive for unifying context handling was to prepare for mock requests. v2: Rename to last_retired_context, split out legacy_context tracking for MI_SET_CONTEXT. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161218153724.8439-3-chris@chris-wilson.co.uk
2016-12-18 15:37:20 +00:00
GEM_BUG_ON(ce->pin_count == 0);
if (--ce->pin_count)
return;
if (ce->state) {
ce->state->obj->pin_global--;
i915_vma_unpin(ce->state);
}
i915_gem_context_put(ctx);
}
static int intel_init_ring_buffer(struct intel_engine_cs *engine)
{
struct intel_ring *ring;
int err;
intel_engine_setup_common(engine);
err = intel_engine_init_common(engine);
if (err)
goto err;
ring = intel_engine_create_ring(engine, 32 * PAGE_SIZE);
if (IS_ERR(ring)) {
err = PTR_ERR(ring);
goto err;
}
/* Ring wraparound at offset 0 sometimes hangs. No idea why. */
err = intel_ring_pin(ring, engine->i915, I915_GTT_PAGE_SIZE);
if (err)
goto err_ring;
GEM_BUG_ON(engine->buffer);
engine->buffer = ring;
return 0;
err_ring:
intel_ring_free(ring);
err:
intel_engine_cleanup_common(engine);
return err;
}
void intel_engine_cleanup(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
WARN_ON(INTEL_GEN(dev_priv) > 2 &&
(I915_READ_MODE(engine) & MODE_IDLE) == 0);
intel_ring_unpin(engine->buffer);
intel_ring_free(engine->buffer);
if (engine->cleanup)
engine->cleanup(engine);
intel_engine_cleanup_common(engine);
drm/i915: Allocate intel_engine_cs structure only for the enabled engines With the possibility of addition of many more number of rings in future, the drm_i915_private structure could bloat as an array, of type intel_engine_cs, is embedded inside it. struct intel_engine_cs engine[I915_NUM_ENGINES]; Though this is still fine as generally there is only a single instance of drm_i915_private structure used, but not all of the possible rings would be enabled or active on most of the platforms. Some memory can be saved by allocating intel_engine_cs structure only for the enabled/active engines. Currently the engine/ring ID is kept static and dev_priv->engine[] is simply indexed using the enums defined in intel_engine_id. To save memory and continue using the static engine/ring IDs, 'engine' is defined as an array of pointers. struct intel_engine_cs *engine[I915_NUM_ENGINES]; dev_priv->engine[engine_ID] will be NULL for disabled engine instances. There is a text size reduction of 928 bytes, from 1028200 to 1027272, for i915.o file (but for i915.ko file text size remain same as 1193131 bytes). v2: - Remove the engine iterator field added in drm_i915_private structure, instead pass a local iterator variable to the for_each_engine** macros. (Chris) - Do away with intel_engine_initialized() and instead directly use the NULL pointer check on engine pointer. (Chris) v3: - Remove for_each_engine_id() macro, as the updated macro for_each_engine() can be used in place of it. (Chris) - Protect the access to Render engine Fault register with a NULL check, as engine specific init is done later in Driver load sequence. v4: - Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris) - Kill the superfluous init_engine_lists(). v5: - Cleanup the intel_engines_init() & intel_engines_setup(), with respect to allocation of intel_engine_cs structure. (Chris) v6: - Rebase. v7: - Optimize the for_each_engine_masked() macro. (Chris) - Change the type of 'iter' local variable to enum intel_engine_id. (Chris) - Rebase. v8: Rebase. v9: Rebase. v10: - For index calculation use engine ID instead of pointer based arithmetic in intel_engine_sync_index() as engine pointers are not contiguous now (Chris) - For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas) - Use for_each_engine macro for cleanup in intel_engines_init() and remove check for NULL engine pointer in cleanup() routines. (Joonas) v11: Rebase. Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Akash Goel <akash.goel@intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com
2016-10-13 17:14:48 +00:00
dev_priv->engine[engine->id] = NULL;
kfree(engine);
}
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
void intel_legacy_submission_resume(struct drm_i915_private *dev_priv)
{
struct intel_engine_cs *engine;
drm/i915: Allocate intel_engine_cs structure only for the enabled engines With the possibility of addition of many more number of rings in future, the drm_i915_private structure could bloat as an array, of type intel_engine_cs, is embedded inside it. struct intel_engine_cs engine[I915_NUM_ENGINES]; Though this is still fine as generally there is only a single instance of drm_i915_private structure used, but not all of the possible rings would be enabled or active on most of the platforms. Some memory can be saved by allocating intel_engine_cs structure only for the enabled/active engines. Currently the engine/ring ID is kept static and dev_priv->engine[] is simply indexed using the enums defined in intel_engine_id. To save memory and continue using the static engine/ring IDs, 'engine' is defined as an array of pointers. struct intel_engine_cs *engine[I915_NUM_ENGINES]; dev_priv->engine[engine_ID] will be NULL for disabled engine instances. There is a text size reduction of 928 bytes, from 1028200 to 1027272, for i915.o file (but for i915.ko file text size remain same as 1193131 bytes). v2: - Remove the engine iterator field added in drm_i915_private structure, instead pass a local iterator variable to the for_each_engine** macros. (Chris) - Do away with intel_engine_initialized() and instead directly use the NULL pointer check on engine pointer. (Chris) v3: - Remove for_each_engine_id() macro, as the updated macro for_each_engine() can be used in place of it. (Chris) - Protect the access to Render engine Fault register with a NULL check, as engine specific init is done later in Driver load sequence. v4: - Use !!dev_priv->engine[VCS] style for the engine check in getparam. (Chris) - Kill the superfluous init_engine_lists(). v5: - Cleanup the intel_engines_init() & intel_engines_setup(), with respect to allocation of intel_engine_cs structure. (Chris) v6: - Rebase. v7: - Optimize the for_each_engine_masked() macro. (Chris) - Change the type of 'iter' local variable to enum intel_engine_id. (Chris) - Rebase. v8: Rebase. v9: Rebase. v10: - For index calculation use engine ID instead of pointer based arithmetic in intel_engine_sync_index() as engine pointers are not contiguous now (Chris) - For appropriateness, rename local enum variable 'iter' to 'id'. (Joonas) - Use for_each_engine macro for cleanup in intel_engines_init() and remove check for NULL engine pointer in cleanup() routines. (Joonas) v11: Rebase. Cc: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Akash Goel <akash.goel@intel.com> Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1476378888-7372-1-git-send-email-akash.goel@intel.com
2016-10-13 17:14:48 +00:00
enum intel_engine_id id;
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
/* Restart from the beginning of the rings for convenience */
for_each_engine(engine, dev_priv, id)
intel_ring_reset(engine->buffer, 0);
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
}
static int ring_request_alloc(struct drm_i915_gem_request *request)
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Unify active context tracking between legacy/execlists/guc The requests conversion introduced a nasty bug where we could generate a new request in the middle of constructing a request if we needed to idle the system in order to evict space for a context. The request to idle would be executed (and waited upon) before the current one, creating a minor havoc in the seqno accounting, as we will consider the current request to already be completed (prior to deferred seqno assignment) but ring->last_retired_head would have been updated and still could allow us to overwrite the current request before execution. We also employed two different mechanisms to track the active context until it was switched out. The legacy method allowed for waiting upon an active context (it could forcibly evict any vma, including context's), but the execlists method took a step backwards by pinning the vma for the entire active lifespan of the context (the only way to evict was to idle the entire GPU, not individual contexts). However, to circumvent the tricky issue of locking (i.e. we cannot take struct_mutex at the time of i915_gem_request_submit(), where we would want to move the previous context onto the active tracker and unpin it), we take the execlists approach and keep the contexts pinned until retirement. The benefit of the execlists approach, more important for execlists than legacy, was the reduction in work in pinning the context for each request - as the context was kept pinned until idle, it could short circuit the pinning for all active contexts. We introduce new engine vfuncs to pin and unpin the context respectively. The context is pinned at the start of the request, and only unpinned when the following request is retired (this ensures that the context is idle and coherent in main memory before we unpin it). We move the engine->last_context tracking into the retirement itself (rather than during request submission) in order to allow the submission to be reordered or unwound without undue difficultly. And finally an ulterior motive for unifying context handling was to prepare for mock requests. v2: Rename to last_retired_context, split out legacy_context tracking for MI_SET_CONTEXT. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161218153724.8439-3-chris@chris-wilson.co.uk
2016-12-18 15:37:20 +00:00
GEM_BUG_ON(!request->ctx->engine[request->engine->id].pin_count);
/* Flush enough space to reduce the likelihood of waiting after
* we start building the request - in which case we will just
* have to repeat work.
*/
request->reserved_space += LEGACY_REQUEST_SIZE;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(request, 0);
if (IS_ERR(cs))
return PTR_ERR(cs);
request->reserved_space -= LEGACY_REQUEST_SIZE;
return 0;
drm/i915: Preallocate next seqno before touching the ring Based on the work by Mika Kuoppala, we realised that we need to handle seqno wraparound prior to committing our changes to the ring. The most obvious point then is to grab the seqno inside intel_ring_begin(), and then to reuse that seqno for all ring operations until the next request. As intel_ring_begin() can fail, the callers must already be prepared to handle such failure and so we can safely add further checks. This patch looks like it should be split up into the interface changes and the tweaks to move seqno wrapping from the execbuffer into the core seqno increment. However, I found no easy way to break it into incremental steps without introducing further broken behaviour. v2: Mika found a silly mistake and a subtle error in the existing code; inside i915_gem_retire_requests() we were resetting the sync_seqno of the target ring based on the seqno from this ring - which are only related by the order of their allocation, not retirement. Hence we were applying the optimisation that the rings were synchronised too early, fortunately the only real casualty there is the handling of seqno wrapping. v3: Do not forget to reset the sync_seqno upon module reinitialisation, ala resume. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861 Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2012-11-27 16:22:52 +00:00
}
static noinline int wait_for_space(struct drm_i915_gem_request *req,
unsigned int bytes)
{
struct intel_ring *ring = req->ring;
struct drm_i915_gem_request *target;
long timeout;
lockdep_assert_held(&req->i915->drm.struct_mutex);
if (intel_ring_update_space(ring) >= bytes)
return 0;
/*
* Space is reserved in the ringbuffer for finalising the request,
* as that cannot be allowed to fail. During request finalisation,
* reserved_space is set to 0 to stop the overallocation and the
* assumption is that then we never need to wait (which has the
* risk of failing with EINTR).
*
* See also i915_gem_request_alloc() and i915_add_request().
*/
GEM_BUG_ON(!req->reserved_space);
list_for_each_entry(target, &ring->request_list, ring_link) {
/* Would completion of this request free enough space? */
if (bytes <= __intel_ring_space(target->postfix,
ring->emit, ring->size))
break;
}
drm/i915: Reserve ring buffer space for i915_add_request() commands It is a bad idea for i915_add_request() to fail. The work will already have been send to the ring and will be processed, but there will not be any tracking or management of that work. The only way the add request call can fail is if it can't write its epilogue commands to the ring (cache flushing, seqno updates, interrupt signalling). The reasons for that are mostly down to running out of ring buffer space and the problems associated with trying to get some more. This patch prevents that situation from happening in the first place. When a request is created, it marks sufficient space as reserved for the epilogue commands. Thus guaranteeing that by the time the epilogue is written, there will be plenty of space for it. Note that a ring_begin() call is required to actually reserve the space (and do any potential waiting). However, that is not currently done at request creation time. This is because the ring_begin() code can allocate a request. Hence calling begin() from the request allocation code would lead to infinite recursion! Later patches in this series remove the need for begin() to do the allocate. At that point, it becomes safe for the allocate to call begin() and really reserve the space. Until then, there is a potential for insufficient space to be available at the point of calling i915_add_request(). However, that would only be in the case where the request was created and immediately submitted without ever calling ring_begin() and adding any work to that request. Which should never happen. And even if it does, and if that request happens to fall down the tiny window of opportunity for failing due to being out of ring space then does it really matter because the request wasn't doing anything in the first place? v2: Updated the 'reserved space too small' warning to include the offending sizes. Added a 'cancel' operation to clean up when a request is abandoned. Added re-initialisation of tracking state after a buffer wrap to keep the sanity checks accurate. v3: Incremented the reserved size to accommodate Ironlake (after finally managing to run on an ILK system). Also fixed missing wrap code in LRC mode. v4: Added extra comment and removed duplicate WARN (feedback from Tomas). For: VIZ-5115 CC: Tomas Elf <tomas.elf@intel.com> Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-18 12:10:09 +00:00
if (WARN_ON(&target->ring_link == &ring->request_list))
return -ENOSPC;
timeout = i915_wait_request(target,
I915_WAIT_INTERRUPTIBLE | I915_WAIT_LOCKED,
MAX_SCHEDULE_TIMEOUT);
if (timeout < 0)
return timeout;
i915_gem_request_retire_upto(target);
intel_ring_update_space(ring);
GEM_BUG_ON(ring->space < bytes);
return 0;
drm/i915: Reserve ring buffer space for i915_add_request() commands It is a bad idea for i915_add_request() to fail. The work will already have been send to the ring and will be processed, but there will not be any tracking or management of that work. The only way the add request call can fail is if it can't write its epilogue commands to the ring (cache flushing, seqno updates, interrupt signalling). The reasons for that are mostly down to running out of ring buffer space and the problems associated with trying to get some more. This patch prevents that situation from happening in the first place. When a request is created, it marks sufficient space as reserved for the epilogue commands. Thus guaranteeing that by the time the epilogue is written, there will be plenty of space for it. Note that a ring_begin() call is required to actually reserve the space (and do any potential waiting). However, that is not currently done at request creation time. This is because the ring_begin() code can allocate a request. Hence calling begin() from the request allocation code would lead to infinite recursion! Later patches in this series remove the need for begin() to do the allocate. At that point, it becomes safe for the allocate to call begin() and really reserve the space. Until then, there is a potential for insufficient space to be available at the point of calling i915_add_request(). However, that would only be in the case where the request was created and immediately submitted without ever calling ring_begin() and adding any work to that request. Which should never happen. And even if it does, and if that request happens to fall down the tiny window of opportunity for failing due to being out of ring space then does it really matter because the request wasn't doing anything in the first place? v2: Updated the 'reserved space too small' warning to include the offending sizes. Added a 'cancel' operation to clean up when a request is abandoned. Added re-initialisation of tracking state after a buffer wrap to keep the sanity checks accurate. v3: Incremented the reserved size to accommodate Ironlake (after finally managing to run on an ILK system). Also fixed missing wrap code in LRC mode. v4: Added extra comment and removed duplicate WARN (feedback from Tomas). For: VIZ-5115 CC: Tomas Elf <tomas.elf@intel.com> Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-18 12:10:09 +00:00
}
u32 *intel_ring_begin(struct drm_i915_gem_request *req,
unsigned int num_dwords)
{
struct intel_ring *ring = req->ring;
const unsigned int remain_usable = ring->effective_size - ring->emit;
const unsigned int bytes = num_dwords * sizeof(u32);
unsigned int need_wrap = 0;
unsigned int total_bytes;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Reserve ring buffer space for i915_add_request() commands It is a bad idea for i915_add_request() to fail. The work will already have been send to the ring and will be processed, but there will not be any tracking or management of that work. The only way the add request call can fail is if it can't write its epilogue commands to the ring (cache flushing, seqno updates, interrupt signalling). The reasons for that are mostly down to running out of ring buffer space and the problems associated with trying to get some more. This patch prevents that situation from happening in the first place. When a request is created, it marks sufficient space as reserved for the epilogue commands. Thus guaranteeing that by the time the epilogue is written, there will be plenty of space for it. Note that a ring_begin() call is required to actually reserve the space (and do any potential waiting). However, that is not currently done at request creation time. This is because the ring_begin() code can allocate a request. Hence calling begin() from the request allocation code would lead to infinite recursion! Later patches in this series remove the need for begin() to do the allocate. At that point, it becomes safe for the allocate to call begin() and really reserve the space. Until then, there is a potential for insufficient space to be available at the point of calling i915_add_request(). However, that would only be in the case where the request was created and immediately submitted without ever calling ring_begin() and adding any work to that request. Which should never happen. And even if it does, and if that request happens to fall down the tiny window of opportunity for failing due to being out of ring space then does it really matter because the request wasn't doing anything in the first place? v2: Updated the 'reserved space too small' warning to include the offending sizes. Added a 'cancel' operation to clean up when a request is abandoned. Added re-initialisation of tracking state after a buffer wrap to keep the sanity checks accurate. v3: Incremented the reserved size to accommodate Ironlake (after finally managing to run on an ILK system). Also fixed missing wrap code in LRC mode. v4: Added extra comment and removed duplicate WARN (feedback from Tomas). For: VIZ-5115 CC: Tomas Elf <tomas.elf@intel.com> Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-18 12:10:09 +00:00
/* Packets must be qword aligned. */
GEM_BUG_ON(num_dwords & 1);
total_bytes = bytes + req->reserved_space;
GEM_BUG_ON(total_bytes > ring->effective_size);
drm/i915: Reserve ring buffer space for i915_add_request() commands It is a bad idea for i915_add_request() to fail. The work will already have been send to the ring and will be processed, but there will not be any tracking or management of that work. The only way the add request call can fail is if it can't write its epilogue commands to the ring (cache flushing, seqno updates, interrupt signalling). The reasons for that are mostly down to running out of ring buffer space and the problems associated with trying to get some more. This patch prevents that situation from happening in the first place. When a request is created, it marks sufficient space as reserved for the epilogue commands. Thus guaranteeing that by the time the epilogue is written, there will be plenty of space for it. Note that a ring_begin() call is required to actually reserve the space (and do any potential waiting). However, that is not currently done at request creation time. This is because the ring_begin() code can allocate a request. Hence calling begin() from the request allocation code would lead to infinite recursion! Later patches in this series remove the need for begin() to do the allocate. At that point, it becomes safe for the allocate to call begin() and really reserve the space. Until then, there is a potential for insufficient space to be available at the point of calling i915_add_request(). However, that would only be in the case where the request was created and immediately submitted without ever calling ring_begin() and adding any work to that request. Which should never happen. And even if it does, and if that request happens to fall down the tiny window of opportunity for failing due to being out of ring space then does it really matter because the request wasn't doing anything in the first place? v2: Updated the 'reserved space too small' warning to include the offending sizes. Added a 'cancel' operation to clean up when a request is abandoned. Added re-initialisation of tracking state after a buffer wrap to keep the sanity checks accurate. v3: Incremented the reserved size to accommodate Ironlake (after finally managing to run on an ILK system). Also fixed missing wrap code in LRC mode. v4: Added extra comment and removed duplicate WARN (feedback from Tomas). For: VIZ-5115 CC: Tomas Elf <tomas.elf@intel.com> Signed-off-by: John Harrison <John.C.Harrison@Intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-06-18 12:10:09 +00:00
if (unlikely(total_bytes > remain_usable)) {
const int remain_actual = ring->size - ring->emit;
if (bytes > remain_usable) {
/*
* Not enough space for the basic request. So need to
* flush out the remainder and then wait for
* base + reserved.
*/
total_bytes += remain_actual;
need_wrap = remain_actual | 1;
} else {
/*
* The base request will fit but the reserved space
* falls off the end. So we don't need an immediate
* wrap and only need to effectively wait for the
* reserved size from the start of ringbuffer.
*/
total_bytes = req->reserved_space + remain_actual;
}
}
if (unlikely(total_bytes > ring->space)) {
int ret = wait_for_space(req, total_bytes);
if (unlikely(ret))
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
return ERR_PTR(ret);
}
if (unlikely(need_wrap)) {
need_wrap &= ~1;
GEM_BUG_ON(need_wrap > ring->space);
GEM_BUG_ON(ring->emit + need_wrap > ring->size);
/* Fill the tail with MI_NOOP */
memset(ring->vaddr + ring->emit, 0, need_wrap);
ring->emit = 0;
ring->space -= need_wrap;
}
GEM_BUG_ON(ring->emit > ring->size - bytes);
GEM_BUG_ON(ring->space < bytes);
cs = ring->vaddr + ring->emit;
GEM_DEBUG_EXEC(memset(cs, POISON_INUSE, bytes));
ring->emit += bytes;
ring->space -= bytes;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
return cs;
}
/* Align the ring tail to a cacheline boundary */
int intel_ring_cacheline_align(struct drm_i915_gem_request *req)
{
int num_dwords =
(req->ring->emit & (CACHELINE_BYTES - 1)) / sizeof(uint32_t);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
if (num_dwords == 0)
return 0;
num_dwords = CACHELINE_BYTES / sizeof(uint32_t) - num_dwords;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, num_dwords);
if (IS_ERR(cs))
return PTR_ERR(cs);
while (num_dwords--)
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_NOOP;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
intel_ring_advance(req, cs);
return 0;
}
static void gen6_bsd_submit_request(struct drm_i915_gem_request *request)
{
struct drm_i915_private *dev_priv = request->i915;
intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
/* Every tail move must follow the sequence below */
/* Disable notification that the ring is IDLE. The GT
* will then assume that it is busy and bring it out of rc6.
*/
I915_WRITE_FW(GEN6_BSD_SLEEP_PSMI_CONTROL,
_MASKED_BIT_ENABLE(GEN6_BSD_SLEEP_MSG_DISABLE));
/* Clear the context id. Here be magic! */
I915_WRITE64_FW(GEN6_BSD_RNCID, 0x0);
/* Wait for the ring not to be idle, i.e. for it to wake up. */
if (__intel_wait_for_register_fw(dev_priv,
GEN6_BSD_SLEEP_PSMI_CONTROL,
GEN6_BSD_SLEEP_INDICATOR,
0,
1000, 0, NULL))
DRM_ERROR("timed out waiting for the BSD ring to wake up\n");
/* Now that the ring is fully powered up, update the tail */
i9xx_submit_request(request);
/* Let the ring send IDLE messages to the GT again,
* and so let it sleep to conserve power when idle.
*/
I915_WRITE_FW(GEN6_BSD_SLEEP_PSMI_CONTROL,
_MASKED_BIT_DISABLE(GEN6_BSD_SLEEP_MSG_DISABLE));
intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
}
static int gen6_bsd_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 cmd, *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 4);
if (IS_ERR(cs))
return PTR_ERR(cs);
cmd = MI_FLUSH_DW;
if (INTEL_GEN(req->i915) >= 8)
cmd += 1;
/* We always require a command barrier so that subsequent
* commands, such as breadcrumb interrupts, are strictly ordered
* wrt the contents of the write cache being flushed to memory
* (and thus being coherent from the CPU).
*/
cmd |= MI_FLUSH_DW_STORE_INDEX | MI_FLUSH_DW_OP_STOREDW;
/*
* Bspec vol 1c.5 - video engine command streamer:
* "If ENABLED, all TLBs will be invalidated once the flush
* operation is complete. This bit is only valid when the
* Post-Sync Operation field is a value of 1h or 3h."
*/
if (mode & EMIT_INVALIDATE)
cmd |= MI_INVALIDATE_TLB | MI_INVALIDATE_BSD;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = cmd;
*cs++ = I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT;
if (INTEL_GEN(req->i915) >= 8) {
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = 0; /* upper addr */
*cs++ = 0; /* value */
} else {
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = 0;
*cs++ = MI_NOOP;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
intel_ring_advance(req, cs);
return 0;
}
static int
gen8_emit_bb_start(struct drm_i915_gem_request *req,
u64 offset, u32 len,
unsigned int dispatch_flags)
{
bool ppgtt = USES_PPGTT(req->i915) &&
!(dispatch_flags & I915_DISPATCH_SECURE);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 4);
if (IS_ERR(cs))
return PTR_ERR(cs);
/* FIXME(BDW): Address space and security selectors. */
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_BATCH_BUFFER_START_GEN8 | (ppgtt << 8) | (dispatch_flags &
I915_DISPATCH_RS ? MI_BATCH_RESOURCE_STREAMER : 0);
*cs++ = lower_32_bits(offset);
*cs++ = upper_32_bits(offset);
*cs++ = MI_NOOP;
intel_ring_advance(req, cs);
return 0;
}
static int
hsw_emit_bb_start(struct drm_i915_gem_request *req,
u64 offset, u32 len,
unsigned int dispatch_flags)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_BATCH_BUFFER_START | (dispatch_flags & I915_DISPATCH_SECURE ?
0 : MI_BATCH_PPGTT_HSW | MI_BATCH_NON_SECURE_HSW) |
(dispatch_flags & I915_DISPATCH_RS ?
MI_BATCH_RESOURCE_STREAMER : 0);
/* bit0-7 is the length on GEN6+ */
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = offset;
intel_ring_advance(req, cs);
return 0;
}
static int
gen6_emit_bb_start(struct drm_i915_gem_request *req,
u64 offset, u32 len,
unsigned int dispatch_flags)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 2);
if (IS_ERR(cs))
return PTR_ERR(cs);
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = MI_BATCH_BUFFER_START | (dispatch_flags & I915_DISPATCH_SECURE ?
0 : MI_BATCH_NON_SECURE_I965);
/* bit0-7 is the length on GEN6+ */
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = offset;
intel_ring_advance(req, cs);
return 0;
}
/* Blitter support (SandyBridge+) */
static int gen6_ring_flush(struct drm_i915_gem_request *req, u32 mode)
{
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
u32 cmd, *cs;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
cs = intel_ring_begin(req, 4);
if (IS_ERR(cs))
return PTR_ERR(cs);
cmd = MI_FLUSH_DW;
if (INTEL_GEN(req->i915) >= 8)
cmd += 1;
/* We always require a command barrier so that subsequent
* commands, such as breadcrumb interrupts, are strictly ordered
* wrt the contents of the write cache being flushed to memory
* (and thus being coherent from the CPU).
*/
cmd |= MI_FLUSH_DW_STORE_INDEX | MI_FLUSH_DW_OP_STOREDW;
/*
* Bspec vol 1c.3 - blitter engine command streamer:
* "If ENABLED, all TLBs will be invalidated once the flush
* operation is complete. This bit is only valid when the
* Post-Sync Operation field is a value of 1h or 3h."
*/
if (mode & EMIT_INVALIDATE)
cmd |= MI_INVALIDATE_TLB;
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = cmd;
*cs++ = I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT;
if (INTEL_GEN(req->i915) >= 8) {
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = 0; /* upper addr */
*cs++ = 0; /* value */
} else {
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
*cs++ = 0;
*cs++ = MI_NOOP;
}
drm/i915: Emit to ringbuffer directly This removes the usage of intel_ring_emit in favour of directly writing to the ring buffer. intel_ring_emit was preventing the compiler for optimising fetch and increment of the current ring buffer pointer and therefore generating very verbose code for every write. It had no useful purpose since all ringbuffer operations are started and ended with intel_ring_begin and intel_ring_advance respectively, with no bail out in the middle possible, so it is fine to increment the tail in intel_ring_begin and let the code manage the pointer itself. Useless instruction removal amounts to approximately two and half kilobytes of saved text on my build. Not sure if this has any measurable performance implications but executing a ton of useless instructions on fast paths cannot be good. v2: * Change return from intel_ring_begin to error pointer by popular demand. * Move tail increment to intel_ring_advance to enable some error checking. v3: * Move tail advance back into intel_ring_begin. * Rebase and tidy. v4: * Complete rebase after a few months since v3. v5: * Remove unecessary cast and fix !debug compile. (Chris Wilson) v6: * Make intel_ring_offset take request as well. * Fix recording of request postfix plus a sprinkle of asserts. (Chris Wilson) v7: * Use intel_ring_offset to get the postfix. (Chris Wilson) * Convert GVT code as well. v8: * Rename *out++ to *cs++. v9: * Fix GVT out to cs conversion in GVT. v10: * Rebase for new intel_ring_begin in selftests. Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com
2017-02-14 11:32:42 +00:00
intel_ring_advance(req, cs);
return 0;
}
static void intel_ring_init_semaphores(struct drm_i915_private *dev_priv,
struct intel_engine_cs *engine)
{
struct drm_i915_gem_object *obj;
int ret, i;
if (!i915_modparams.semaphores)
return;
if (INTEL_GEN(dev_priv) >= 8 && !dev_priv->semaphore) {
struct i915_vma *vma;
obj = i915_gem_object_create(dev_priv, PAGE_SIZE);
if (IS_ERR(obj))
goto err;
vma = i915_vma_instance(obj, &dev_priv->ggtt.base, NULL);
if (IS_ERR(vma))
goto err_obj;
ret = i915_gem_object_set_to_gtt_domain(obj, false);
if (ret)
goto err_obj;
ret = i915_vma_pin(vma, 0, 0, PIN_GLOBAL | PIN_HIGH);
if (ret)
goto err_obj;
dev_priv->semaphore = vma;
}
if (INTEL_GEN(dev_priv) >= 8) {
u32 offset = i915_ggtt_offset(dev_priv->semaphore);
engine->semaphore.sync_to = gen8_ring_sync_to;
engine->semaphore.signal = gen8_xcs_signal;
for (i = 0; i < I915_NUM_ENGINES; i++) {
u32 ring_offset;
if (i != engine->id)
ring_offset = offset + GEN8_SEMAPHORE_OFFSET(engine->id, i);
else
ring_offset = MI_SEMAPHORE_SYNC_INVALID;
engine->semaphore.signal_ggtt[i] = ring_offset;
}
} else if (INTEL_GEN(dev_priv) >= 6) {
engine->semaphore.sync_to = gen6_ring_sync_to;
engine->semaphore.signal = gen6_signal;
/*
* The current semaphore is only applied on pre-gen8
* platform. And there is no VCS2 ring on the pre-gen8
* platform. So the semaphore between RCS and VCS2 is
* initialized as INVALID. Gen8 will initialize the
* sema between VCS2 and RCS later.
*/
for (i = 0; i < GEN6_NUM_SEMAPHORES; i++) {
static const struct {
u32 wait_mbox;
i915_reg_t mbox_reg;
} sem_data[GEN6_NUM_SEMAPHORES][GEN6_NUM_SEMAPHORES] = {
[RCS_HW] = {
[VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RV, .mbox_reg = GEN6_VRSYNC },
[BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RB, .mbox_reg = GEN6_BRSYNC },
[VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_RVE, .mbox_reg = GEN6_VERSYNC },
},
[VCS_HW] = {
[RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VR, .mbox_reg = GEN6_RVSYNC },
[BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VB, .mbox_reg = GEN6_BVSYNC },
[VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VVE, .mbox_reg = GEN6_VEVSYNC },
},
[BCS_HW] = {
[RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BR, .mbox_reg = GEN6_RBSYNC },
[VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BV, .mbox_reg = GEN6_VBSYNC },
[VECS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_BVE, .mbox_reg = GEN6_VEBSYNC },
},
[VECS_HW] = {
[RCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VER, .mbox_reg = GEN6_RVESYNC },
[VCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VEV, .mbox_reg = GEN6_VVESYNC },
[BCS_HW] = { .wait_mbox = MI_SEMAPHORE_SYNC_VEB, .mbox_reg = GEN6_BVESYNC },
},
};
u32 wait_mbox;
i915_reg_t mbox_reg;
if (i == engine->hw_id) {
wait_mbox = MI_SEMAPHORE_SYNC_INVALID;
mbox_reg = GEN6_NOSYNC;
} else {
wait_mbox = sem_data[engine->hw_id][i].wait_mbox;
mbox_reg = sem_data[engine->hw_id][i].mbox_reg;
}
engine->semaphore.mbox.wait[i] = wait_mbox;
engine->semaphore.mbox.signal[i] = mbox_reg;
}
}
return;
err_obj:
i915_gem_object_put(obj);
err:
DRM_DEBUG_DRIVER("Failed to allocate space for semaphores, disabling\n");
i915_modparams.semaphores = 0;
}
static void intel_ring_init_irq(struct drm_i915_private *dev_priv,
struct intel_engine_cs *engine)
{
engine->irq_enable_mask = GT_RENDER_USER_INTERRUPT << engine->irq_shift;
if (INTEL_GEN(dev_priv) >= 8) {
engine->irq_enable = gen8_irq_enable;
engine->irq_disable = gen8_irq_disable;
engine->irq_seqno_barrier = gen6_seqno_barrier;
} else if (INTEL_GEN(dev_priv) >= 6) {
engine->irq_enable = gen6_irq_enable;
engine->irq_disable = gen6_irq_disable;
engine->irq_seqno_barrier = gen6_seqno_barrier;
} else if (INTEL_GEN(dev_priv) >= 5) {
engine->irq_enable = gen5_irq_enable;
engine->irq_disable = gen5_irq_disable;
drm/i915: Add a delay between interrupt and inspecting the final seqno (ilk) On Ironlake, there is no command nor register to ensure that the write from a MI_STORE command is completed (and coherent on the CPU) before the command parser continues. This means that the ordering between the seqno write and the subsequent user interrupt is undefined (like gen6+). So to ensure that the seqno write is completed after the final user interrupt we need to delay the read sufficiently to allow the write to complete. This delay is undefined by the bspec, and empirically requires 75us even though a register read combined with a clflush is less than 500ns. Hence, the delay is due to an on-chip buffer rather than the latency of the write to memory. Note that the render ring controls this by filling the PIPE_CONTROL fifo with stalling commands that force the earliest pipe-control with the seqno to be completed before the command parser continues. Given that we need a barrier operation for BSD, we may as well forgo the extra per-batch latency by using a common per-interrupt barrier. Studying the impact of adding the usleep shows that in both sequences of and individual synchronous no-op batches is negligible for the media engine (where the write now is unordered with the interrupt). Converting the render engine over from the current glutton of pie-controls over to the per-interrupt delays speeds up both the sequential and individual synchronous no-ops by 20% and 60%, respectively. This speed up holds even when looking at the throughput of small copies (4KiB->4MiB), both serial and synchronous, by about 20%. This is because despite adding a significant delay to the interrupt, in all likelihood we will see the seqno write without having to apply the barrier (only in the rare corner cases where the write is delayed on the last required is the delay necessary). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94307 Testcase: igt/gem_sync #ilk Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-12-git-send-email-chris@chris-wilson.co.uk
2016-07-01 16:23:21 +00:00
engine->irq_seqno_barrier = gen5_seqno_barrier;
} else if (INTEL_GEN(dev_priv) >= 3) {
engine->irq_enable = i9xx_irq_enable;
engine->irq_disable = i9xx_irq_disable;
} else {
engine->irq_enable = i8xx_irq_enable;
engine->irq_disable = i8xx_irq_disable;
}
}
static void i9xx_set_default_submission(struct intel_engine_cs *engine)
{
engine->submit_request = i9xx_submit_request;
engine->cancel_requests = cancel_requests;
engine->park = NULL;
engine->unpark = NULL;
}
static void gen6_bsd_set_default_submission(struct intel_engine_cs *engine)
{
i9xx_set_default_submission(engine);
engine->submit_request = gen6_bsd_submit_request;
}
static void intel_ring_default_vfuncs(struct drm_i915_private *dev_priv,
struct intel_engine_cs *engine)
{
intel_ring_init_irq(dev_priv, engine);
intel_ring_init_semaphores(dev_priv, engine);
engine->init_hw = init_ring_common;
drm/i915: Update reset path to fix incomplete requests Update reset path in preparation for engine reset which requires identification of incomplete requests and associated context and fixing their state so that engine can resume correctly after reset. The request that caused the hang will be skipped and head is reset to the start of breadcrumb. This allows us to resume from where we left-off. Since this request didn't complete normally we also need to cleanup elsp queue manually. This is vital if we employ nonblocking request submission where we may have a web of dependencies upon the hung request and so advancing the seqno manually is no longer trivial. ABI: gem_reset_stats / DRM_IOCTL_I915_GET_RESET_STATS We change the way we count pending batches. Only the active context involved in the reset is marked as either innocent or guilty, and not mark the entire world as pending. By inspection this only affects igt/gem_reset_stats (which assumes implementation details) and not piglit. ARB_robustness gives this guide on how we expect the user of this interface to behave: * Provide a mechanism for an OpenGL application to learn about graphics resets that affect the context. When a graphics reset occurs, the OpenGL context becomes unusable and the application must create a new context to continue operation. Detecting a graphics reset happens through an inexpensive query. And with regards to the actual meaning of the reset values: Certain events can result in a reset of the GL context. Such a reset causes all context state to be lost. Recovery from such events requires recreation of all objects in the affected context. The current status of the graphics reset state is returned by enum GetGraphicsResetStatusARB(); The symbolic constant returned indicates if the GL context has been in a reset state at any point since the last call to GetGraphicsResetStatusARB. NO_ERROR indicates that the GL context has not been in a reset state since the last call. GUILTY_CONTEXT_RESET_ARB indicates that a reset has been detected that is attributable to the current GL context. INNOCENT_CONTEXT_RESET_ARB indicates a reset has been detected that is not attributable to the current GL context. UNKNOWN_CONTEXT_RESET_ARB indicates a detected graphics reset whose cause is unknown. The language here is explicit in that we must mark up the guilty batch, but is loose enough for us to relax the innocent (i.e. pending) accounting as only the active batches are involved with the reset. In the future, we are looking towards single engine resetting (with minimal locking), where it seems inappropriate to mark the entire world as innocent since the reset occurred on a different engine. Reducing the information available means we only have to encounter the pain once, and also reduces the information leaking from one context to another. v2: Legacy ringbuffer submission required a reset following hibernation, or else we restore stale values to the RING_HEAD and walked over stolen garbage. v3: GuC requires replaying the requests after a reset. v4: Restore engine IRQ after reset (so waiters will be woken!) Rearm hangcheck if resetting with a waiter. Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Arun Siluvery <arun.siluvery@linux.intel.com> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-13-chris@chris-wilson.co.uk
2016-09-09 13:11:53 +00:00
engine->reset_hw = reset_ring_common;
drm/i915: Unify active context tracking between legacy/execlists/guc The requests conversion introduced a nasty bug where we could generate a new request in the middle of constructing a request if we needed to idle the system in order to evict space for a context. The request to idle would be executed (and waited upon) before the current one, creating a minor havoc in the seqno accounting, as we will consider the current request to already be completed (prior to deferred seqno assignment) but ring->last_retired_head would have been updated and still could allow us to overwrite the current request before execution. We also employed two different mechanisms to track the active context until it was switched out. The legacy method allowed for waiting upon an active context (it could forcibly evict any vma, including context's), but the execlists method took a step backwards by pinning the vma for the entire active lifespan of the context (the only way to evict was to idle the entire GPU, not individual contexts). However, to circumvent the tricky issue of locking (i.e. we cannot take struct_mutex at the time of i915_gem_request_submit(), where we would want to move the previous context onto the active tracker and unpin it), we take the execlists approach and keep the contexts pinned until retirement. The benefit of the execlists approach, more important for execlists than legacy, was the reduction in work in pinning the context for each request - as the context was kept pinned until idle, it could short circuit the pinning for all active contexts. We introduce new engine vfuncs to pin and unpin the context respectively. The context is pinned at the start of the request, and only unpinned when the following request is retired (this ensures that the context is idle and coherent in main memory before we unpin it). We move the engine->last_context tracking into the retirement itself (rather than during request submission) in order to allow the submission to be reordered or unwound without undue difficultly. And finally an ulterior motive for unifying context handling was to prepare for mock requests. v2: Rename to last_retired_context, split out legacy_context tracking for MI_SET_CONTEXT. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/20161218153724.8439-3-chris@chris-wilson.co.uk
2016-12-18 15:37:20 +00:00
engine->context_pin = intel_ring_context_pin;
engine->context_unpin = intel_ring_context_unpin;
engine->request_alloc = ring_request_alloc;
engine->emit_breadcrumb = i9xx_emit_breadcrumb;
engine->emit_breadcrumb_sz = i9xx_emit_breadcrumb_sz;
if (i915_modparams.semaphores) {
int num_rings;
engine->emit_breadcrumb = gen6_sema_emit_breadcrumb;
num_rings = INTEL_INFO(dev_priv)->num_rings - 1;
if (INTEL_GEN(dev_priv) >= 8) {
engine->emit_breadcrumb_sz += num_rings * 6;
} else {
engine->emit_breadcrumb_sz += num_rings * 3;
if (num_rings & 1)
engine->emit_breadcrumb_sz++;
}
}
engine->set_default_submission = i9xx_set_default_submission;
if (INTEL_GEN(dev_priv) >= 8)
engine->emit_bb_start = gen8_emit_bb_start;
else if (INTEL_GEN(dev_priv) >= 6)
engine->emit_bb_start = gen6_emit_bb_start;
else if (INTEL_GEN(dev_priv) >= 4)
engine->emit_bb_start = i965_emit_bb_start;
else if (IS_I830(dev_priv) || IS_I845G(dev_priv))
engine->emit_bb_start = i830_emit_bb_start;
else
engine->emit_bb_start = i915_emit_bb_start;
}
int intel_init_render_ring_buffer(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
int ret;
intel_ring_default_vfuncs(dev_priv, engine);
if (HAS_L3_DPF(dev_priv))
engine->irq_keep_mask = GT_RENDER_L3_PARITY_ERROR_INTERRUPT;
drm/i915: Add a delay between interrupt and inspecting the final seqno (ilk) On Ironlake, there is no command nor register to ensure that the write from a MI_STORE command is completed (and coherent on the CPU) before the command parser continues. This means that the ordering between the seqno write and the subsequent user interrupt is undefined (like gen6+). So to ensure that the seqno write is completed after the final user interrupt we need to delay the read sufficiently to allow the write to complete. This delay is undefined by the bspec, and empirically requires 75us even though a register read combined with a clflush is less than 500ns. Hence, the delay is due to an on-chip buffer rather than the latency of the write to memory. Note that the render ring controls this by filling the PIPE_CONTROL fifo with stalling commands that force the earliest pipe-control with the seqno to be completed before the command parser continues. Given that we need a barrier operation for BSD, we may as well forgo the extra per-batch latency by using a common per-interrupt barrier. Studying the impact of adding the usleep shows that in both sequences of and individual synchronous no-op batches is negligible for the media engine (where the write now is unordered with the interrupt). Converting the render engine over from the current glutton of pie-controls over to the per-interrupt delays speeds up both the sequential and individual synchronous no-ops by 20% and 60%, respectively. This speed up holds even when looking at the throughput of small copies (4KiB->4MiB), both serial and synchronous, by about 20%. This is because despite adding a significant delay to the interrupt, in all likelihood we will see the seqno write without having to apply the barrier (only in the rare corner cases where the write is delayed on the last required is the delay necessary). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94307 Testcase: igt/gem_sync #ilk Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-12-git-send-email-chris@chris-wilson.co.uk
2016-07-01 16:23:21 +00:00
if (INTEL_GEN(dev_priv) >= 8) {
engine->init_context = intel_rcs_ctx_init;
engine->emit_breadcrumb = gen8_render_emit_breadcrumb;
engine->emit_breadcrumb_sz = gen8_render_emit_breadcrumb_sz;
engine->emit_flush = gen8_render_ring_flush;
if (i915_modparams.semaphores) {
int num_rings;
engine->semaphore.signal = gen8_rcs_signal;
num_rings = INTEL_INFO(dev_priv)->num_rings - 1;
engine->emit_breadcrumb_sz += num_rings * 8;
}
} else if (INTEL_GEN(dev_priv) >= 6) {
engine->init_context = intel_rcs_ctx_init;
engine->emit_flush = gen7_render_ring_flush;
if (IS_GEN6(dev_priv))
engine->emit_flush = gen6_render_ring_flush;
} else if (IS_GEN5(dev_priv)) {
engine->emit_flush = gen4_render_ring_flush;
} else {
if (INTEL_GEN(dev_priv) < 4)
engine->emit_flush = gen2_render_ring_flush;
else
engine->emit_flush = gen4_render_ring_flush;
engine->irq_enable_mask = I915_USER_INTERRUPT;
}
if (IS_HASWELL(dev_priv))
engine->emit_bb_start = hsw_emit_bb_start;
engine->init_hw = init_render_ring;
engine->cleanup = render_ring_cleanup;
ret = intel_init_ring_buffer(engine);
if (ret)
return ret;
drm/i915: Add a delay between interrupt and inspecting the final seqno (ilk) On Ironlake, there is no command nor register to ensure that the write from a MI_STORE command is completed (and coherent on the CPU) before the command parser continues. This means that the ordering between the seqno write and the subsequent user interrupt is undefined (like gen6+). So to ensure that the seqno write is completed after the final user interrupt we need to delay the read sufficiently to allow the write to complete. This delay is undefined by the bspec, and empirically requires 75us even though a register read combined with a clflush is less than 500ns. Hence, the delay is due to an on-chip buffer rather than the latency of the write to memory. Note that the render ring controls this by filling the PIPE_CONTROL fifo with stalling commands that force the earliest pipe-control with the seqno to be completed before the command parser continues. Given that we need a barrier operation for BSD, we may as well forgo the extra per-batch latency by using a common per-interrupt barrier. Studying the impact of adding the usleep shows that in both sequences of and individual synchronous no-op batches is negligible for the media engine (where the write now is unordered with the interrupt). Converting the render engine over from the current glutton of pie-controls over to the per-interrupt delays speeds up both the sequential and individual synchronous no-ops by 20% and 60%, respectively. This speed up holds even when looking at the throughput of small copies (4KiB->4MiB), both serial and synchronous, by about 20%. This is because despite adding a significant delay to the interrupt, in all likelihood we will see the seqno write without having to apply the barrier (only in the rare corner cases where the write is delayed on the last required is the delay necessary). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94307 Testcase: igt/gem_sync #ilk Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-12-git-send-email-chris@chris-wilson.co.uk
2016-07-01 16:23:21 +00:00
if (INTEL_GEN(dev_priv) >= 6) {
ret = intel_engine_create_scratch(engine, PAGE_SIZE);
if (ret)
return ret;
} else if (HAS_BROKEN_CS_TLB(dev_priv)) {
ret = intel_engine_create_scratch(engine, I830_WA_SIZE);
if (ret)
return ret;
}
return 0;
}
int intel_init_bsd_ring_buffer(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
intel_ring_default_vfuncs(dev_priv, engine);
if (INTEL_GEN(dev_priv) >= 6) {
/* gen6 bsd needs a special wa for tail updates */
if (IS_GEN6(dev_priv))
engine->set_default_submission = gen6_bsd_set_default_submission;
engine->emit_flush = gen6_bsd_ring_flush;
if (INTEL_GEN(dev_priv) < 8)
engine->irq_enable_mask = GT_BSD_USER_INTERRUPT;
} else {
engine->mmio_base = BSD_RING_BASE;
engine->emit_flush = bsd_ring_flush;
if (IS_GEN5(dev_priv))
engine->irq_enable_mask = ILK_BSD_USER_INTERRUPT;
else
engine->irq_enable_mask = I915_BSD_USER_INTERRUPT;
}
return intel_init_ring_buffer(engine);
}
int intel_init_blt_ring_buffer(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
intel_ring_default_vfuncs(dev_priv, engine);
engine->emit_flush = gen6_ring_flush;
if (INTEL_GEN(dev_priv) < 8)
engine->irq_enable_mask = GT_BLT_USER_INTERRUPT;
return intel_init_ring_buffer(engine);
}
int intel_init_vebox_ring_buffer(struct intel_engine_cs *engine)
{
struct drm_i915_private *dev_priv = engine->i915;
intel_ring_default_vfuncs(dev_priv, engine);
engine->emit_flush = gen6_ring_flush;
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
if (INTEL_GEN(dev_priv) < 8) {
engine->irq_enable_mask = PM_VEBOX_USER_INTERRUPT;
engine->irq_enable = hsw_vebox_irq_enable;
engine->irq_disable = hsw_vebox_irq_disable;
drm/i915/bdw: Implement interrupt changes The interrupt handling implementation remains the same as previous generations with the 4 types of registers, status, identity, mask, and enable. However the layout of where the bits go have changed entirely. To address these changes, all of the interrupt vfuncs needed special gen8 code. The way it works is there is a top level status register now which informs the interrupt service routine which unit caused the interrupt, and therefore which interrupt registers to read to process the interrupt. For display the division is quite logical, a set of interrupt registers for each pipe, and in addition to those, a set each for "misc" and port. For GT the things get a bit hairy, as seen by the code. Each of the GT units has it's own bits defined. They all look *very similar* and resides in 16 bits of a GT register. As an example, RCS and BCS share register 0. To compact the code a bit, at a slight expense to complexity, this is exactly how the code works as well. 2 structures are added to the ring buffer so that our ring buffer interrupt handling code knows which ring shares the interrupt registers, and a shift value (ie. the top or bottom 16 bits of the register). The above allows us to kept the interrupt register caching scheme, the per interrupt enables, and the code to mask and unmask interrupts relatively clean (again at the cost of some more complexity). Most of the GT units mentioned above are command streamers, and so the symmetry should work quite well for even the yet to be implemented rings which Broadwell adds. v2: Fixes up a couple of bugs, and is more verbose about errors in the Broadwell interrupt handler. v3: fix DE_MISC IER offset v4: Simplify interrupts: I totally misread the docs the first time I implemented interrupts, and so this should greatly simplify the mess. Unlike GEN6, we never touch the regular mask registers in irq_get/put. v5: Rebased on to of recent pch hotplug setup changes. v6: Fixup on top of moving num_pipes to intel_info. v7: Rebased on top of Egbert Eich's hpd irq handling rework. Also wired up ibx_hpd_irq_setup for gen8. v8: Rebase on top of Jani's asle handling rework. v9: Rebase on top of Ben's VECS enabling for Haswell, where he unfortunately went OCD on the gt irq #defines. Not that they're still not yet fully consistent: - Used the GT_RENDER_ #defines + bdw shifts. - Dropped the shift from the L3_PARITY stuff, seemed clearer. - s/irq_refcount/irq_refcount.gt/ v10: Squash in VECS enabling patches and the gen8_gt_irq_handler refactoring from Zhao Yakui <yakui.zhao@intel.com> v11: Rebase on top of the interrupt cleanups in upstream. v12: Rebase on top of Ben's DPF changes in upstream. v13: Drop bdw from the HAS_L3_DPF feature flag for now, it's unclear what exactly needs to be done. Requested by Ben. v14: Fix the patch. - Drop the mask of reserved bits and assorted logic, it doesn't match the spec. - Do the posting read inconditionally instead of commenting it out. - Add a GEN8_MASTER_IRQ_CONTROL definition and use it. - Fix up the GEN8_PIPE interrupt defines and give the GEN8_ prefixes - we actually will need to use them. - Enclose macros in do {} while (0) (checkpatch). - Clear DE_MISC interrupt bits only after having processed them. - Fix whitespace fail (checkpatch). - Fix overtly long lines where appropriate (checkpatch). - Don't use typedef'ed private_t (maintainer-scripts). - Align the function parameter list correctly. Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v4) Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> bikeshed
2013-11-03 04:07:09 +00:00
}
return intel_init_ring_buffer(engine);
}