linux/drivers/gpu/drm/nouveau/nvc0_grgpc.fuc
Ben Skeggs 0411de8548 drm/nvc0/gr: import and use our own fuc by default
The ability to use NVIDIA's fuc has been retained *temporarily* in order
to better debug any issues that may be lingering in our initial attempt
at writing this ucode.  Once I'm fairly confident we're okay, it'll be
removed.

There's a number of things not implemented by this fuc currently, but
most of it is sets of state that our context setup would not have used
anyway.  No doubt we'll find out what they're for at some point, and
implement it if required.

This has been tested on 0xc0/0xc4 thus far, and from what I could tell
it worked as well as NVIDIA's.  It's also been tested on 0xc1, but even
with NVIDIA's fuc that chipset doesn't work correctly with nouveau yet.

0xc3/0xc8/0xce should in theory be supported too, but I don't have the
hardware to check that.

There's no doubt numerous bugs to squash yet, please report any!

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
2011-06-23 15:57:38 +10:00

475 lines
11 KiB
Plaintext

/* fuc microcode for nvc0 PGRAPH/GPC
*
* Copyright 2011 Red Hat Inc.
*
* 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 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 COPYRIGHT HOLDER(S) OR AUTHOR(S) 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: Ben Skeggs
*/
/* To build:
* m4 nvc0_grgpc.fuc | envyas -a -w -m fuc -V nva3 -o nvc0_grgpc.fuc.h
*/
/* TODO
* - bracket certain functions with scratch writes, useful for debugging
* - watchdog timer around ctx operations
*/
.section nvc0_grgpc_data
include(`nvc0_graph.fuc')
gpc_id: .b32 0
gpc_mmio_list_head: .b32 0
gpc_mmio_list_tail: .b32 0
tpc_count: .b32 0
tpc_mask: .b32 0
tpc_mmio_list_head: .b32 0
tpc_mmio_list_tail: .b32 0
cmd_queue: queue_init
// chipset descriptions
chipsets:
.b8 0xc0 0 0 0
.b16 nvc0_gpc_mmio_head
.b16 nvc0_gpc_mmio_tail
.b16 nvc0_tpc_mmio_head
.b16 nvc0_tpc_mmio_tail
.b8 0xc1 0 0 0
.b16 nvc0_gpc_mmio_head
.b16 nvc1_gpc_mmio_tail
.b16 nvc0_tpc_mmio_head
.b16 nvc1_tpc_mmio_tail
.b8 0xc3 0 0 0
.b16 nvc0_gpc_mmio_head
.b16 nvc0_gpc_mmio_tail
.b16 nvc0_tpc_mmio_head
.b16 nvc3_tpc_mmio_tail
.b8 0xc4 0 0 0
.b16 nvc0_gpc_mmio_head
.b16 nvc0_gpc_mmio_tail
.b16 nvc0_tpc_mmio_head
.b16 nvc3_tpc_mmio_tail
.b8 0xc8 0 0 0
.b16 nvc0_gpc_mmio_head
.b16 nvc0_gpc_mmio_tail
.b16 nvc0_tpc_mmio_head
.b16 nvc0_tpc_mmio_tail
.b8 0xce 0 0 0
.b16 nvc0_gpc_mmio_head
.b16 nvc0_gpc_mmio_tail
.b16 nvc0_tpc_mmio_head
.b16 nvc3_tpc_mmio_tail
.b8 0 0 0 0
// GPC mmio lists
nvc0_gpc_mmio_head:
mmctx_data(0x000380, 1)
mmctx_data(0x000400, 6)
mmctx_data(0x000450, 9)
mmctx_data(0x000600, 1)
mmctx_data(0x000684, 1)
mmctx_data(0x000700, 5)
mmctx_data(0x000800, 1)
mmctx_data(0x000808, 3)
mmctx_data(0x000828, 1)
mmctx_data(0x000830, 1)
mmctx_data(0x0008d8, 1)
mmctx_data(0x0008e0, 1)
mmctx_data(0x0008e8, 6)
mmctx_data(0x00091c, 1)
mmctx_data(0x000924, 3)
mmctx_data(0x000b00, 1)
mmctx_data(0x000b08, 6)
mmctx_data(0x000bb8, 1)
mmctx_data(0x000c08, 1)
mmctx_data(0x000c10, 8)
mmctx_data(0x000c80, 1)
mmctx_data(0x000c8c, 1)
mmctx_data(0x001000, 3)
mmctx_data(0x001014, 1)
nvc0_gpc_mmio_tail:
mmctx_data(0x000c6c, 1);
nvc1_gpc_mmio_tail:
// TPC mmio lists
nvc0_tpc_mmio_head:
mmctx_data(0x000018, 1)
mmctx_data(0x00003c, 1)
mmctx_data(0x000048, 1)
mmctx_data(0x000064, 1)
mmctx_data(0x000088, 1)
mmctx_data(0x000200, 6)
mmctx_data(0x00021c, 2)
mmctx_data(0x000300, 6)
mmctx_data(0x0003d0, 1)
mmctx_data(0x0003e0, 2)
mmctx_data(0x000400, 3)
mmctx_data(0x000420, 1)
mmctx_data(0x0004b0, 1)
mmctx_data(0x0004e8, 1)
mmctx_data(0x0004f4, 1)
mmctx_data(0x000520, 2)
mmctx_data(0x000604, 4)
mmctx_data(0x000644, 20)
mmctx_data(0x000698, 1)
mmctx_data(0x000750, 2)
nvc0_tpc_mmio_tail:
mmctx_data(0x000758, 1)
mmctx_data(0x0002c4, 1)
mmctx_data(0x0004bc, 1)
mmctx_data(0x0006e0, 1)
nvc3_tpc_mmio_tail:
mmctx_data(0x000544, 1)
nvc1_tpc_mmio_tail:
.section nvc0_grgpc_code
bra init
define(`include_code')
include(`nvc0_graph.fuc')
// reports an exception to the host
//
// In: $r15 error code (see nvc0_graph.fuc)
//
error:
push $r14
mov $r14 -0x67ec // 0x9814
sethi $r14 0x400000
call nv_wr32 // HUB_CTXCTL_CC_SCRATCH[5] = error code
add b32 $r14 0x41c
mov $r15 1
call nv_wr32 // HUB_CTXCTL_INTR_UP_SET
pop $r14
ret
// GPC fuc initialisation, executed by triggering ucode start, will
// fall through to main loop after completion.
//
// Input:
// CC_SCRATCH[0]: chipset (PMC_BOOT_0 read returns 0x0bad0bad... sigh)
// CC_SCRATCH[1]: context base
//
// Output:
// CC_SCRATCH[0]:
// 31:31: set to signal completion
// CC_SCRATCH[1]:
// 31:0: GPC context size
//
init:
clear b32 $r0
mov $sp $r0
// enable fifo access
mov $r1 0x1200
mov $r2 2
iowr I[$r1 + 0x000] $r2 // FIFO_ENABLE
// setup i0 handler, and route all interrupts to it
mov $r1 ih
mov $iv0 $r1
mov $r1 0x400
iowr I[$r1 + 0x300] $r0 // INTR_DISPATCH
// enable fifo interrupt
mov $r2 4
iowr I[$r1 + 0x000] $r2 // INTR_EN_SET
// enable interrupts
bset $flags ie0
// figure out which GPC we are, and how many TPCs we have
mov $r1 0x608
shl b32 $r1 6
iord $r2 I[$r1 + 0x000] // UNITS
mov $r3 1
and $r2 0x1f
shl b32 $r3 $r2
sub b32 $r3 1
st b32 D[$r0 + tpc_count] $r2
st b32 D[$r0 + tpc_mask] $r3
add b32 $r1 0x400
iord $r2 I[$r1 + 0x000] // MYINDEX
st b32 D[$r0 + gpc_id] $r2
// find context data for this chipset
mov $r2 0x800
shl b32 $r2 6
iord $r2 I[$r2 + 0x000] // CC_SCRATCH[0]
mov $r1 chipsets - 12
init_find_chipset:
add b32 $r1 12
ld b32 $r3 D[$r1 + 0x00]
cmpu b32 $r3 $r2
bra e init_context
cmpu b32 $r3 0
bra ne init_find_chipset
// unknown chipset
ret
// initialise context base, and size tracking
init_context:
mov $r2 0x800
shl b32 $r2 6
iord $r2 I[$r2 + 0x100] // CC_SCRATCH[1], initial base
clear b32 $r3 // track GPC context size here
// set mmctx base addresses now so we don't have to do it later,
// they don't currently ever change
mov $r4 0x700
shl b32 $r4 6
shr b32 $r5 $r2 8
iowr I[$r4 + 0x000] $r5 // MMCTX_SAVE_SWBASE
iowr I[$r4 + 0x100] $r5 // MMCTX_LOAD_SWBASE
// calculate GPC mmio context size, store the chipset-specific
// mmio list pointers somewhere we can get at them later without
// re-parsing the chipset list
clear b32 $r14
clear b32 $r15
ld b16 $r14 D[$r1 + 4]
ld b16 $r15 D[$r1 + 6]
st b16 D[$r0 + gpc_mmio_list_head] $r14
st b16 D[$r0 + gpc_mmio_list_tail] $r15
call mmctx_size
add b32 $r2 $r15
add b32 $r3 $r15
// calculate per-TPC mmio context size, store the list pointers
ld b16 $r14 D[$r1 + 8]
ld b16 $r15 D[$r1 + 10]
st b16 D[$r0 + tpc_mmio_list_head] $r14
st b16 D[$r0 + tpc_mmio_list_tail] $r15
call mmctx_size
ld b32 $r14 D[$r0 + tpc_count]
mulu $r14 $r15
add b32 $r2 $r14
add b32 $r3 $r14
// round up base/size to 256 byte boundary (for strand SWBASE)
add b32 $r4 0x1300
shr b32 $r3 2
iowr I[$r4 + 0x000] $r3 // MMCTX_LOAD_COUNT, wtf for?!?
shr b32 $r2 8
shr b32 $r3 6
add b32 $r2 1
add b32 $r3 1
shl b32 $r2 8
shl b32 $r3 8
// calculate size of strand context data
mov b32 $r15 $r2
call strand_ctx_init
add b32 $r3 $r15
// save context size, and tell HUB we're done
mov $r1 0x800
shl b32 $r1 6
iowr I[$r1 + 0x100] $r3 // CC_SCRATCH[1] = context size
add b32 $r1 0x800
clear b32 $r2
bset $r2 31
iowr I[$r1 + 0x000] $r2 // CC_SCRATCH[0] |= 0x80000000
// Main program loop, very simple, sleeps until woken up by the interrupt
// handler, pulls a command from the queue and executes its handler
//
main:
bset $flags $p0
sleep $p0
mov $r13 cmd_queue
call queue_get
bra $p1 main
// 0x0000-0x0003 are all context transfers
cmpu b32 $r14 0x04
bra nc main_not_ctx_xfer
// fetch $flags and mask off $p1/$p2
mov $r1 $flags
mov $r2 0x0006
not b32 $r2
and $r1 $r2
// set $p1/$p2 according to transfer type
shl b32 $r14 1
or $r1 $r14
mov $flags $r1
// transfer context data
call ctx_xfer
bra main
main_not_ctx_xfer:
shl b32 $r15 $r14 16
or $r15 E_BAD_COMMAND
call error
bra main
// interrupt handler
ih:
push $r8
mov $r8 $flags
push $r8
push $r9
push $r10
push $r11
push $r13
push $r14
push $r15
// incoming fifo command?
iord $r10 I[$r0 + 0x200] // INTR
and $r11 $r10 0x00000004
bra e ih_no_fifo
// queue incoming fifo command for later processing
mov $r11 0x1900
mov $r13 cmd_queue
iord $r14 I[$r11 + 0x100] // FIFO_CMD
iord $r15 I[$r11 + 0x000] // FIFO_DATA
call queue_put
add b32 $r11 0x400
mov $r14 1
iowr I[$r11 + 0x000] $r14 // FIFO_ACK
// ack, and wake up main()
ih_no_fifo:
iowr I[$r0 + 0x100] $r10 // INTR_ACK
pop $r15
pop $r14
pop $r13
pop $r11
pop $r10
pop $r9
pop $r8
mov $flags $r8
pop $r8
bclr $flags $p0
iret
// Set this GPC's bit in HUB_BAR, used to signal completion of various
// activities to the HUB fuc
//
hub_barrier_done:
mov $r15 1
ld b32 $r14 D[$r0 + gpc_id]
shl b32 $r15 $r14
mov $r14 -0x6be8 // 0x409418 - HUB_BAR_SET
sethi $r14 0x400000
call nv_wr32
ret
// Disables various things, waits a bit, and re-enables them..
//
// Not sure how exactly this helps, perhaps "ENABLE" is not such a
// good description for the bits we turn off? Anyways, without this,
// funny things happen.
//
ctx_redswitch:
mov $r14 0x614
shl b32 $r14 6
mov $r15 0x020
iowr I[$r14] $r15 // GPC_RED_SWITCH = POWER
mov $r15 8
ctx_redswitch_delay:
sub b32 $r15 1
bra ne ctx_redswitch_delay
mov $r15 0xa20
iowr I[$r14] $r15 // GPC_RED_SWITCH = UNK11, ENABLE, POWER
ret
// Transfer GPC context data between GPU and storage area
//
// In: $r15 context base address
// $p1 clear on save, set on load
// $p2 set if opposite direction done/will be done, so:
// on save it means: "a load will follow this save"
// on load it means: "a save preceeded this load"
//
ctx_xfer:
// set context base address
mov $r1 0xa04
shl b32 $r1 6
iowr I[$r1 + 0x000] $r15// MEM_BASE
bra not $p1 ctx_xfer_not_load
call ctx_redswitch
ctx_xfer_not_load:
// strands
mov $r1 0x4afc
sethi $r1 0x20000
mov $r2 0xc
iowr I[$r1] $r2 // STRAND_CMD(0x3f) = 0x0c
call strand_wait
mov $r2 0x47fc
sethi $r2 0x20000
iowr I[$r2] $r0 // STRAND_FIRST_GENE(0x3f) = 0x00
xbit $r2 $flags $p1
add b32 $r2 3
iowr I[$r1] $r2 // STRAND_CMD(0x3f) = 0x03/0x04 (SAVE/LOAD)
// mmio context
xbit $r10 $flags $p1 // direction
or $r10 2 // first
mov $r11 0x0000
sethi $r11 0x500000
ld b32 $r12 D[$r0 + gpc_id]
shl b32 $r12 15
add b32 $r11 $r12 // base = NV_PGRAPH_GPCn
ld b32 $r12 D[$r0 + gpc_mmio_list_head]
ld b32 $r13 D[$r0 + gpc_mmio_list_tail]
mov $r14 0 // not multi
call mmctx_xfer
// per-TPC mmio context
xbit $r10 $flags $p1 // direction
or $r10 4 // last
mov $r11 0x4000
sethi $r11 0x500000 // base = NV_PGRAPH_GPC0_TPC0
ld b32 $r12 D[$r0 + gpc_id]
shl b32 $r12 15
add b32 $r11 $r12 // base = NV_PGRAPH_GPCn_TPC0
ld b32 $r12 D[$r0 + tpc_mmio_list_head]
ld b32 $r13 D[$r0 + tpc_mmio_list_tail]
ld b32 $r15 D[$r0 + tpc_mask]
mov $r14 0x800 // stride = 0x800
call mmctx_xfer
// wait for strands to finish
call strand_wait
// if load, or a save without a load following, do some
// unknown stuff that's done after finishing a block of
// strand commands
bra $p1 ctx_xfer_post
bra not $p2 ctx_xfer_done
ctx_xfer_post:
mov $r1 0x4afc
sethi $r1 0x20000
mov $r2 0xd
iowr I[$r1] $r2 // STRAND_CMD(0x3f) = 0x0d
call strand_wait
// mark completion in HUB's barrier
ctx_xfer_done:
call hub_barrier_done
ret
.align 256