We have to reject unknown flags for uAPI considerations, and also
because the curent implementation limits their i915 storage space
to two bits.
v2: (Chris Wilson)
* Fix fail in ABI check.
* Added unknown flags and BUILD_BUG_ON.
v3:
* Use ARCH_KMALLOC_MINALIGN instead of alignof. (Chris Wilson)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: cf6e7bac63 ("drm/i915: Add support for drm syncobjs")
Cc: Jason Ekstrand <jason@jlekstrand.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: David Airlie <airlied@linux.ie>
Cc: intel-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031102326.9738-1-tvrtko.ursulin@linux.intel.com
In case the object has changed tiling between calls to execbuf, we need
to check if the existing offset inside the GTT matches the new tiling
constraint. We even need to do this for "unfenced" tiled objects, where
the 3D commands use an implied fence and so the object still needs to
match the physical fence restrictions on alignment (only required for
gen2 and early gen3).
In commit 2889caa923 ("drm/i915: Eliminate lots of iterations over
the execobjects array"), the idea was to remove the second guessing and
only set the NEEDS_MAP flag when required. However, the entire check
for an unusable offset for fencing was removed and not just the
secondary check. I.e.
/* avoid costly ping-pong once a batch bo ended up non-mappable */
if (entry->flags & __EXEC_OBJECT_NEEDS_MAP &&
!i915_vma_is_map_and_fenceable(vma))
return !only_mappable_for_reloc(entry->flags);
was entirely removed as the ping-pong between execbuf passes was fixed,
but its primary purpose in forcing unaligned unfenced access to be
rebound was forgotten.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103502
Fixes: 2889caa923 ("drm/i915: Eliminate lots of iterations over the execobjects array")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171031103607.17836-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
As we have a lightweight fallback to insert a single page into the
aperture, try to avoid any heavier evictions when attempting to insert
the entire object.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009084401.29090-5-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Following the pattern now used for obj->mm.pages, use just pin_fence and
unpin_fence to control access to the fence registers. I.e. instead of
calling get_fence(); pin_fence(), we now just need to call pin_fence().
This will make it easier to reduce the locking requirements around
fence registers.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171009084401.29090-2-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Getting started with v4.15 features:
- Cannonlake workarounds (Rodrigo, Oscar)
- Infoframe refactoring and fixes to enable infoframes for DP (Ville)
- VBT definition updates (Jani)
- Sparse warning fixes (Ville, Chris)
- Crtc state usage fixes and cleanups (Ville)
- DP vswing, pre-emph and buffer translation refactoring and fixes (Rodrigo)
- Prevent IPS from interfering with CRC capture (Ville, Marta)
- Enable Mesa to advertise ARB_timer_query (Nanley)
- Refactor GT number into intel_device_info (Lionel)
- Avoid eDP DP AUX CH timeouts harder (Manasi)
- CDCLK check improvements (Ville)
- Restore GPU clock boost on missed pageflip vblanks (Chris)
- Fence register reservation API for vGPU (Changbin)
- First batch of CCS fixes (Ville)
- Finally, numerous GEM fixes, cleanups and improvements (Chris)
* tag 'drm-intel-next-2017-09-07' of git://anongit.freedesktop.org/git/drm-intel: (100 commits)
drm/i915: Update DRIVER_DATE to 20170907
drm/i915/cnl: WaThrottleEUPerfToAvoidTDBackPressure:cnl(pre-prod)
drm/i915: Lift has-pinned-pages assert to caller of ____i915_gem_object_get_pages
drm/i915: Display WA #1133 WaFbcSkipSegments:cnl, glk
drm/i915/cnl: Allow the reg_read ioctl to read the RCS TIMESTAMP register
drm/i915: Move device_info.has_snoop into the static tables
drm/i915: Disable MI_STORE_DATA_IMM for i915g/i915gm
drm/i915: Re-enable GTT following a device reset
drm/i915/cnp: Wa 1181: Fix Backlight issue
drm/i915: Annotate user relocs with __user
drm/i915: Constify load detect mode
drm/i915/perf: Remove __user from u64 in drm_i915_perf_oa_config
drm/i915: Silence sparse by using gfp_t
drm/i915: io unmap functions want __iomem
drm/i915: Add __rcu to radix tree slot pointer
drm/i915: Wake up the device for the fbdev setup
drm/i915: Add interface to reserve fence registers for vGPU
drm/i915: Use correct path to trace include
drm/i915: Fix the missing PPAT cache attributes on CNL
drm/i915: Fix enum pipe vs. enum transcoder for the PCH transcoder
...
Our global struct with params is named exactly the same way
as new preferred name for the drm_i915_private function parameter.
To avoid such name reuse lets use different name for the global.
v5: pure rename
v6: fix
Credits-to: Coccinelle
@@
identifier n;
@@
(
- i915.n
+ i915_modparams.n
)
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Ville Syrjala <ville.syrjala@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170919193846.38060-1-michal.wajdeczko@intel.com
Since we reuse the same field for the user passing in their control
flags, and for the kernel to track a couple of bits of state, document
and check that those do not overlap.
Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170921110135.15990-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
GFP_TEMPORARY was introduced by commit e12ba74d8f ("Group short-lived
and reclaimable kernel allocations") along with __GFP_RECLAIMABLE. It's
primary motivation was to allow users to tell that an allocation is
short lived and so the allocator can try to place such allocations close
together and prevent long term fragmentation. As much as this sounds
like a reasonable semantic it becomes much less clear when to use the
highlevel GFP_TEMPORARY allocation flag. How long is temporary? Can the
context holding that memory sleep? Can it take locks? It seems there is
no good answer for those questions.
The current implementation of GFP_TEMPORARY is basically GFP_KERNEL |
__GFP_RECLAIMABLE which in itself is tricky because basically none of
the existing caller provide a way to reclaim the allocated memory. So
this is rather misleading and hard to evaluate for any benefits.
I have checked some random users and none of them has added the flag
with a specific justification. I suspect most of them just copied from
other existing users and others just thought it might be a good idea to
use without any measuring. This suggests that GFP_TEMPORARY just
motivates for cargo cult usage without any reasoning.
I believe that our gfp flags are quite complex already and especially
those with highlevel semantic should be clearly defined to prevent from
confusion and abuse. Therefore I propose dropping GFP_TEMPORARY and
replace all existing users to simply use GFP_KERNEL. Please note that
SLAB users with shrinkers will still get __GFP_RECLAIMABLE heuristic and
so they will be placed properly for memory fragmentation prevention.
I can see reasons we might want some gfp flag to reflect shorterm
allocations but I propose starting from a clear semantic definition and
only then add users with proper justification.
This was been brought up before LSF this year by Matthew [1] and it
turned out that GFP_TEMPORARY really doesn't have a clear semantic. It
seems to be a heuristic without any measured advantage for most (if not
all) its current users. The follow up discussion has revealed that
opinions on what might be temporary allocation differ a lot between
developers. So rather than trying to tweak existing users into a
semantic which they haven't expected I propose to simply remove the flag
and start from scratch if we really need a semantic for short term
allocations.
[1] http://lkml.kernel.org/r/20170118054945.GD18349@bombadil.infradead.org
[akpm@linux-foundation.org: fix typo]
[akpm@linux-foundation.org: coding-style fixes]
[sfr@canb.auug.org.au: drm/i915: fix up]
Link: http://lkml.kernel.org/r/20170816144703.378d4f4d@canb.auug.org.au
Link: http://lkml.kernel.org/r/20170728091904.14627-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Neil Brown <neilb@suse.de>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Following the simplification to a single lookup loop in commit
170fa29b14 ("drm/i915: Simplify eb_lookup_vmas()") and commit
d1b48c1e71 ("drm/i915: Replace execbuf vma ht with an idr"), we can go
one step further and reorder the error paths so that the state of the
local variable obj is always known to the compiler and doesn't need the
uninitialized_var markup to squelch a compiler warning.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170912150752.20411-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Add the missing __user to the urelocs cast to fix the following sparse
warning:
i915_gem_execbuffer.c:1541:47: warning: cast removes address space of expression
i915_gem_execbuffer.c:1541:62: warning: incorrect type in argument 2 (different address spaces)
i915_gem_execbuffer.c:1541:62: expected void const [noderef] <asn:1>*from
i915_gem_execbuffer.c:1541:62: got char *
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Fixes: 2889caa923 ("drm/i915: Eliminate lots of iterations over the execobjects array")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901165434.24636-1-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> #irc
(cherry picked from commit 908a610557)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
The early gen3 machines (i915g/Grantsdale and i915gm/Alviso) share a lot
of characteristics in their MI/GTT blocks with gen2, and in particular
can only use physical addresses in MI_STORE_DATA_IMM. This makes it
incompatible with our usage, so include those two machines in the
blacklist to prevent usage.
v2: Make it easy for gcc and rewrite it as a switch to save some space.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> #v1
Link: https://patchwork.freedesktop.org/patch/msgid/20170906152859.5304-1-chris@chris-wilson.co.uk
Add the missing __user to the urelocs cast to fix the following sparse
warning:
i915_gem_execbuffer.c:1541:47: warning: cast removes address space of expression
i915_gem_execbuffer.c:1541:62: warning: incorrect type in argument 2 (different address spaces)
i915_gem_execbuffer.c:1541:62: expected void const [noderef] <asn:1>*from
i915_gem_execbuffer.c:1541:62: got char *
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Fixes: 2889caa923 ("drm/i915: Eliminate lots of iterations over the execobjects array")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170901165434.24636-1-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> #irc
By using drm_gem_flink/drm_gem_open on an object using the same fd, it
is possible for a client to create multiple handles pointing to the same
object (tied to the same contexts and VMA), as exemplified by
igt::gem_handle_to_libdrm_bo(). Since this duplication has been possible
since forever, we cannot assume that the handle:(fpriv, object) is
unique and so must handle the multiple users of a single VMA.
v2: Added commentary noise.
Testcase: igt/gem_close
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102355
Fixes: d1b48c1e71 ("drm/i915: Replace execbuf vma ht with an idr")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170822110517.22277-3-chris@chris-wilson.co.uk
Tested-by: Marta Lofstedt <marta.lofstedt@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
(cherry-picked from commit 3ffff01749)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Sometimes we know we are the only user of the bo, but since we take a
protective pin_pages early on, an attempt to change the vmap on the
object is denied because it is busy. i915_gem_object_pin_map() cannot
tell from our single pin_count if the operation is safe. Instead we must
pass that information down from the caller in the manner of
I915_MAP_OVERRIDE.
This issue has existed from the introduction of the mapping, but was
never noticed as the only place where this conflict might happen is for
cached kernel buffers (such as allocated by i915_gem_batch_pool_get()).
Until recently there was only a single user (the cmdparser) so no
conflicts ever occurred. However, we now use it to allocate batches for
different operations (using MAP_WC on !llc for writes) in addition to the
existing shadow batch (using MAP_WB for reads).
We could either keep both mappings cached, or use a different write
mechanism if we detect a MAP_WB already exists (i.e. clflush
afterwards), but as we haven't seen this issue in the wild (it requires
hitting the GPU reloc path in addition to the cmdparser) for simplicity
just allow the mappings to be recreated.
v2: Include the i915_MAP_OVERRIDE bit in the enum so the compiler knows
about all the valid values.
Fixes: 7dd4f6729f ("drm/i915: Async GPU relocation processing")
Testcase: igt/gem_lut_handle # byt, completely by accident
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170828104631.8606-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
(cherry picked from commit a575c67617)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
If we are using the cmdparser, we will have to copy the batch and so
stall for the relocations. Rather than prolong that stall by adding more
relocation requests, just use CPU relocations and do the stall upfront.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170826135620.25949-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Sometimes we know we are the only user of the bo, but since we take a
protective pin_pages early on, an attempt to change the vmap on the
object is denied because it is busy. i915_gem_object_pin_map() cannot
tell from our single pin_count if the operation is safe. Instead we must
pass that information down from the caller in the manner of
I915_MAP_OVERRIDE.
This issue has existed from the introduction of the mapping, but was
never noticed as the only place where this conflict might happen is for
cached kernel buffers (such as allocated by i915_gem_batch_pool_get()).
Until recently there was only a single user (the cmdparser) so no
conflicts ever occurred. However, we now use it to allocate batches for
different operations (using MAP_WC on !llc for writes) in addition to the
existing shadow batch (using MAP_WB for reads).
We could either keep both mappings cached, or use a different write
mechanism if we detect a MAP_WB already exists (i.e. clflush
afterwards), but as we haven't seen this issue in the wild (it requires
hitting the GPU reloc path in addition to the cmdparser) for simplicity
just allow the mappings to be recreated.
v2: Include the i915_MAP_OVERRIDE bit in the enum so the compiler knows
about all the valid values.
Fixes: 7dd4f6729f ("drm/i915: Async GPU relocation processing")
Testcase: igt/gem_lut_handle # byt, completely by accident
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170828104631.8606-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
By using drm_gem_flink/drm_gem_open on an object using the same fd, it
is possible for a client to create multiple handles pointing to the same
object (tied to the same contexts and VMA), as exemplified by
igt::gem_handle_to_libdrm_bo(). Since this duplication has been possible
since forever, we cannot assume that the handle:(fpriv, object) is
unique and so must handle the multiple users of a single VMA.
v2: Added commentary noise.
Testcase: igt/gem_close
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102355
Fixes: d1b48c1e71 ("drm/i915: Replace execbuf vma ht with an idr")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170822110517.22277-3-chris@chris-wilson.co.uk
Tested-by: Marta Lofstedt <marta.lofstedt@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
This was the competing idea long ago, but it was only with the rewrite
of the idr as an radixtree and using the radixtree directly ourselves,
along with the realisation that we can store the vma directly in the
radixtree and only need a list for the reverse mapping, that made the
patch performant enough to displace using a hashtable. Though the vma ht
is fast and doesn't require any extra allocation (as we can embed the node
inside the vma), it does require a thread for resizing and serialization
and will have the occasional slow lookup. That is hairy enough to
investigate alternatives and favour them if equivalent in peak performance.
One advantage of allocating an indirection entry is that we can support a
single shared bo between many clients, something that was done on a
first-come first-serve basis for shared GGTT vma previously. To offset
the extra allocations, we create yet another kmem_cache for them.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170816085210.4199-5-chris@chris-wilson.co.uk
Since the introduction of being able to perform a lockless lookup of an
object (i915_gem_object_get_rcu() in fbbd37b36f ("drm/i915: Move object
release to a freelist + worker") we no longer need to split the
object/vma lookup into 3 phases and so combine them into a much simpler
single loop.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170816085210.4199-4-chris@chris-wilson.co.uk
When userspace is doing most of the work, avoiding relocs (using
NO_RELOC) and opting out of implicit synchronisation (using ASYNC), we
still spend a lot of time processing the arrays in execbuf, even though
we now should have nothing to do most of the time. One issue that
becomes readily apparent in profiling anv is that iterating over the
large execobj[] is unfriendly to the loop prefetchers of the CPU and it
much prefers iterating over a pair of arrays rather than one big array.
v2: Clear vma[] on construction to handle errors during vma lookup
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170816085210.4199-3-chris@chris-wilson.co.uk
Since we keep the context around across the slow lookup where we may
drop the struct_mutex, we should double check that the context is still
valid upon reacquisition.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170816085210.4199-2-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
MI_STORE_DWORD_IMM just doesn't work on the video decode engine under
Sandybridge, so refrain from using it. Then switch the selftests over to
using the now common test prior to using MI_STORE_DWORD_IMM.
Fixes: 7dd4f6729f ("drm/i915: Async GPU relocation processing")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.13-rc1+
Link: https://patchwork.freedesktop.org/patch/msgid/20170816085210.4199-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
This commit adds support for waiting on or signaling DRM syncobjs as
part of execbuf. It does so by hijacking the currently unused cliprects
pointer to instead point to an array of i915_gem_exec_fence structs
which containe a DRM syncobj and a flags parameter which specifies
whether to wait on it or to signal it. This implementation
theoretically allows for both flags to be set in which case it waits on
the dma_fence that was in the syncobj and then immediately replaces it
with the dma_fence from the current execbuf.
v2:
- Rebase on new syncobj API
v3:
- Pull everything out into helpers
- Do all allocation in gem_execbuffer2
- Pack the flags in the bottom 2 bits of the drm_syncobj*
v4:
- Prevent a potential race on syncobj->fence
Testcase: igt/gem_exec_fence/syncobj*
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Link: https://patchwork.freedesktop.org/patch/msgid/1499289202-25441-1-git-send-email-jason.ekstrand@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20170815145733.4562-1-chris@chris-wilson.co.uk
Another month, another story in the cache coherency saga. This time, we
come to the realisation that i915_gem_object_is_coherent() has been
reporting whether we can read from the target without requiring a cache
invalidate; but we were using it in places for testing whether we could
write into the object without requiring a cache flush. So split the
tracking into two, one to decide before reads, one after writes.
See commit e27ab73d17 ("drm/i915: Mark CPU cache as dirty on every
transition for CPU writes") for the previous entry in this saga.
v2: Be verbose
v3: Remove unused function (i915_gem_object_is_coherent)
v4: Fix inverted coherency check prior to execbuf (from v2)
v5: Add comment for nasty code where we are optimising on gcc's behalf.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101109
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101555
Testcase: igt/kms_mmap_write_crc
Testcase: igt/kms_pwrite_crc
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Dongwon Kim <dongwon.kim@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Tested-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170811111116.10373-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The goal here was to minimise doing any thing or any check inside the
kernel that was not strictly required. For a userspace that assumes
complete control over the cache domains, the kernel is usually using
outdated information and may trigger clflushes where none were
required.
However, swapping is a situation where userspace has no knowledge of the
domain transfer, and will leave the object in the CPU cache. The kernel
must flush this out to the backing storage prior to use with the GPU. As
we use an asynchronous task tracked by an implicit fence for this, we
also need to cancel the ASYNC flag on the object so that the object will
wait for the clflush to complete before being executed. This also absolves
userspace of the responsibility imposed by commit 77ae995789 ("drm/i915:
Enable userspace to opt-out of implicit fencing") that its needed to ensure
that the object was out of the CPU cache prior to use on the GPU.
Fixes: 77ae995789 ("drm/i915: Enable userspace to opt-out of implicit fencing")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101571
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20170721145037.25105-5-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I was being overly paranoid in not updating the execobject.offset after
performing the fallback copy where we set reloc.presumed_offset to -1.
The thinking was to ensure that a subsequent NORELOC execbuf would be
forced to process the invalid relocations. However this is overkill so
long as we *only* update the execobject.offset following a successful
update of the relocation value witin the batch. If we have to repeat the
execbuf due to a later interruption, then we may skip the relocations on
the second pass (honouring NORELOC) since the execobject.offset match
the actual offsets (even though reloc.presumed_offset is garbage).
Subsequent calls to execbuf with NORELOC should themselves ensure that
the reloc.presumed_offset have been corrected in case of future
migration.
Reporting back the actual execobject.offset, even when
reloc.presumed_offset is garbage, ensures that reuse of those objects
use the latest information to avoid relocations.
Fixes: 2889caa923 ("drm/i915: Eliminate lots of iterations over the execobjects array")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101635
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170721145037.25105-4-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Fix the sizeof(ptr) vs. sizeof(*ptr) typo.
Fixes: 2889caa923 ("drm/i915: Eliminate lots of iterations over the execobjects array")
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170714151242.517-2-imre.deak@intel.com
When computing a hash for looking up relocation target handles in an
execbuf, we start with a large size for the hashtable and proceed to
halve it until the allocation succeeds. The final attempt is with an
order of 0 (i.e. a single element). This means that we then pass bits=0
to hash_32() which then computes "hash >> (32 - 0)" to lookup the single
element. Right shifting a value by the width of the operand is
undefined, so limit the smallest hash table we use to order 1.
v2: Keep the retry allocation flag for the final pass
Fixes: 4ff4b44cbb ("drm/i915: Store a direct lookup from object handle to vma")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170629150425.27508-1-chris@chris-wilson.co.uk
commit 2889caa923 ("drm/i915: Eliminate lots of iterations over the
execobjects array") jiggled around the error handling and replace a test
that we cleaned up properly after ourselves with an assertion. That
assertion failed because in the release function (moments after the
assertion) we were indeed forgetting to mark the vma as cleared. The
consequence was when testing an invalid relocation address, we would try
to release the vma twice (following the couple of attempts to verify the
address) and on the second release notice that the first release was
incomplete.
Testcase: igt/gem_reloc_overflow/invalid-address
Fixes: 2889caa923 ("drm/i915: Eliminate lots of iterations over the execobjects array")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170622104722.2583-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Whilst the contents of the context is still protected by the big
struct_mutex, this is not much of an improvement. It is just one tiny
step towards reducing our BKL.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170620110547.15947-3-chris@chris-wilson.co.uk
During execbuf, a mandatory step is that we add this request (this
fence) to each object's reservation_object. Inside execbuf, we track the
vma, and to add the fence to the reservation_object then means having to
first chase the obj, incurring another cache miss. We can reduce the
number of cache misses by stashing a pointer to the reservation_object
in the vma itself.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170616140525.6394-1-chris@chris-wilson.co.uk
If the user requires patching of their batch or auxiliary buffers, we
currently make the alterations on the cpu. If they are active on the GPU
at the time, we wait under the struct_mutex for them to finish executing
before we rewrite the contents. This happens if shared relocation trees
are used between different contexts with separate address space (and the
buffers then have different addresses in each), the 3D state will need
to be adjusted between execution on each context. However, we don't need
to use the CPU to do the relocation patching, as we could queue commands
to the GPU to perform it and use fences to serialise the operation with
the current activity and future - so the operation on the GPU appears
just as atomic as performing it immediately. Performing the relocation
rewrites on the GPU is not free, in terms of pure throughput, the number
of relocations/s is about halved - but more importantly so is the time
under the struct_mutex.
v2: Break out the request/batch allocation for clearer error flow.
v3: A few asserts to ensure rq ordering is maintained
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Currently, the last object in the execlist is the always the batch.
However, when building the batch buffer we often know the batch object
first and if we can use the first slot in the execlist we can emit
relocation instructions relative to it immediately and avoid a separate
pass to adjust the relocations to point to the last execlist slot.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
This simply hides the EAGAIN caused by userptr when userspace causes
resource contention. However, it is quite beneficial with highly
contended userptr users as we avoid repeating the setup costs and
kernel-user context switches.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
When choosing a slot for an execbuffer, we ideally want to use the same
address as last time (so that we don't have to rebind it) and the same
address as expected by the user (so that we don't have to fixup any
relocations pointing to it). If we first try to bind the incoming
execbuffer->offset from the user, or the currently bound offset that
should hopefully achieve the goal of avoiding the rebind cost and the
relocation penalty. However, if the object is not currently bound there
we don't want to arbitrarily unbind an object in our chosen position and
so choose to rebind/relocate the incoming object instead. After we
report the new position back to the user, on the next pass the
relocations should have settled down.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtien@linux.intel.com>
If we take a reference to the object/vma when it is first used in an
execbuf, we can keep that reference until the object's file-local handle
is closed. Thereby saving a frequent ref/unref pair.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The major scaling bottleneck in execbuffer is the processing of the
execobjects. Creating an auxiliary list is inefficient when compared to
using the execobject array we already have allocated.
Reservation is then split into phases. As we lookup up the VMA, we
try and bind it back into active location. Only if that fails, do we add
it to the unbound list for phase 2. In phase 2, we try and add all those
objects that could not fit into their previous location, with fallback
to retrying all objects and evicting the VM in case of severe
fragmentation. (This is the same as before, except that phase 1 is now
done inline with looking up the VMA to avoid an iteration over the
execobject array. In the ideal case, we eliminate the separate reservation
phase). During the reservation phase, we only evict from the VM between
passes (rather than currently as we try to fit every new VMA). In
testing with Unreal Engine's Atlantis demo which stresses the eviction
logic on gen7 class hardware, this speed up the framerate by a factor of
2.
The second loop amalgamation is between move_to_gpu and move_to_active.
As we always submit the request, even if incomplete, we can use the
current request to track active VMA as we perform the flushes and
synchronisation required.
The next big advancement is to avoid copying back to the user any
execobjects and relocations that are not changed.
v2: Add a Theory of Operation spiel.
v3: Fall back to slow relocations in preparation for flushing userptrs.
v4: Document struct members, factor out eb_validate_vma(), add a few
more comments to explain some magic and hide other magic behind macros.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
If we write a relocation into the buffer, we require our own implicit
synchronisation added after the start of the execbuf, outside of the
user's control. As we may end up clflushing, or doing the patch itself
on the GPU, asynchronously we need to look at the implicit serialisation
on obj->resv and hence need to disable EXEC_OBJECT_ASYNC for this
object.
If the user does trigger a stall for relocations, we make sure the stall
is complete enough so that the batch is not submitted before we complete
those relocations.
Fixes: 77ae995789 ("drm/i915: Enable userspace to opt-out of implicit fencing")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
We can simplify our tracking of pending writes in an execbuf to the
single bit in the vma->exec_entry->flags, but that requires the
relocation function knowing the object's vma. Pass it along.
Note we have only been using a single bit to track flushing since
commit cc889e0f6c
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Wed Jun 13 20:45:19 2012 +0200
drm/i915: disable flushing_list/gpu_write_list
unconditionally flushed all render caches before the breadcrumb and
commit 6ac42f4148
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Sat Jul 21 12:25:01 2012 +0200
drm/i915: Replace the complex flushing logic with simple invalidate/flush all
did away with the explicit GPU domain tracking. This was then codified
into the ABI with NO_RELOC in
commit ed5982e6ce
Author: Daniel Vetter <daniel.vetter@ffwll.ch> # Oi! Patch stealer!
Date: Thu Jan 17 22:23:36 2013 +0100
drm/i915: Allow userspace to hint that the relocations were known
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The advent of full-ppgtt lead to an extra indirection between the object
and its binding. That extra indirection has a noticeable impact on how
fast we can convert from the user handles to our internal vma for
execbuffer. In order to bypass the extra indirection, we use a
resizable hashtable to jump from the object to the per-ctx vma.
rhashtable was considered but we don't need the online resizing feature
and the extra complexity proved to undermine its usefulness. Instead, we
simply reallocate the hastable on demand in a background task and
serialize it before iterating.
In non-full-ppgtt modes, multiple files and multiple contexts can share
the same vma. This leads to having multiple possible handle->vma links,
so we only use the first to establish the fast path. The majority of
buffers are not shared and so we should still be able to realise
speedups with multiple clients.
v2: Prettier names, more magic.
v3: Many style tweaks, most notably hiding the misuse of execobj[].rsvd2
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
For ease of use (i.e. avoiding a few checks and function calls), store
the object's cache coherency next to the cache is dirty bit.
Specifically this patch aims to reduce the frequency of no-op calls to
i915_gem_object_clflush() to counter-act the increase of such calls for
GPU only objects in the previous patch.
v2: Replace cache_dirty & ~cache_coherent with cache_dirty &&
!cache_coherent as gcc generates much better code for the latter
(Tvrtko)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dongwon Kim <dongwon.kim@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Tested-by: Dongwon Kim <dongwon.kim@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170616105455.16977-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Currently, we only mark the CPU cache as dirty if we skip a clflush.
This leads to some confusion where we have to ask if the object is in
the write domain or missed a clflush. If we always mark the cache as
dirty, this becomes a much simply question to answer.
The goal remains to do as few clflushes as required and to do them as
late as possible, in the hope of deferring the work to a kthread and not
block the caller (e.g. execbuf, flips).
v2: Always call clflush before GPU execution when the cache_dirty flag
is set. This may cause some extra work on llc systems that migrate dirty
buffers back and forth - but we do try to limit that by only setting
cache_dirty at the end of the gpu sequence.
v3: Always mark the cache as dirty upon a level change, as we need to
invalidate any stale cachelines due to external writes.
Reported-by: Dongwon Kim <dongwon.kim@intel.com>
Fixes: a6a7cc4b7d ("drm/i915: Always flush the dirty CPU cache when pinning the scanout")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dongwon Kim <dongwon.kim@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Tested-by: Dongwon Kim <dongwon.kim@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170615123850.26843-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>