With no platform using graphics/media IP_VER(12, 50), replace the
checks throughout the code with IP_VER(12, 55) so the code makes sense
by itself with no additional explanation of previous baggage.
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Tvrtko Ursulin <tursulin@ursulin.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20240320060543.4034215-5-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
PCI IDs for XEHPSDV were never added and platform always marked with
force_probe. Drop what's not used and rename some places to either be
xehp or dg2, depending on the platform/IP checks.
The registers not used anymore are also removed.
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Tvrtko Ursulin <tursulin@ursulin.net>
Link: https://patchwork.freedesktop.org/patch/msgid/20240320060543.4034215-2-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
i915_request contains direct alias to i915, there is no point to go via
rq->engine->i915.
v2: added missing rq.i915 initialization in measure_breadcrumb_dw.
Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Nirmoy Das <nirmoy.das@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230720113002.1541572-1-andrzej.hajda@intel.com
Currently the KMD is using enum i915_cache_level to set caching policy for
buffer objects. This is flaky because the PAT index which really controls
the caching behavior in PTE has far more levels than what's defined in the
enum. In addition, the PAT index is platform dependent, having to translate
between i915_cache_level and PAT index is not reliable, and makes the code
more complicated.
From UMD's perspective there is also a necessity to set caching policy for
performance fine tuning. It's much easier for the UMD to directly use PAT
index because the behavior of each PAT index is clearly defined in Bspec.
Having the abstracted i915_cache_level sitting in between would only cause
more ambiguity. PAT is expected to work much like MOCS already works today,
and by design userspace is expected to select the index that exactly
matches the desired behavior described in the hardware specification.
For these reasons this patch replaces i915_cache_level with PAT index. Also
note, the cache_level is not completely removed yet, because the KMD still
has the need of creating buffer objects with simple cache settings such as
cached, uncached, or writethrough. For kernel objects, cache_level is used
for simplicity and backward compatibility. For Pre-gen12 platforms PAT can
have 1:1 mapping to i915_cache_level, so these two are interchangeable. see
the use of LEGACY_CACHELEVEL.
One consequence of this change is that gen8_pte_encode is no longer working
for gen12 platforms due to the fact that gen12 platforms has different PAT
definitions. In the meantime the mtl_pte_encode introduced specfically for
MTL becomes generic for all gen12 platforms. This patch renames the MTL
PTE encode function into gen12_pte_encode and apply it to all gen12. Even
though this change looks unrelated, but separating them would temporarily
break gen12 PTE encoding, thus squash them in one patch.
Special note: this patch changes the way caching behavior is controlled in
the sense that some objects are left to be managed by userspace. For such
objects we need to be careful not to change the userspace settings.There
are kerneldoc and comments added around obj->cache_coherent, cache_dirty,
and how to bypass the checkings by i915_gem_object_has_cache_level. For
full understanding, these changes need to be looked at together with the
two follow-up patches, one disables the {set|get}_caching ioctl's and the
other adds set_pat extension to the GEM_CREATE uAPI.
Bspec: 63019
Cc: Chris Wilson <chris.p.wilson@linux.intel.com>
Signed-off-by: Fei Yang <fei.yang@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230509165200.1740-3-fei.yang@intel.com
In the case of lmem -> lmem transfers, which is currently only possible
with small-bar systems, we need to ensure we copy the CCS aux state
as-is, rather than nuke it. This should fix some nasty display
corruption sometimes seen on DG2 small-bar systems, when also using
DG2_RC_CCS_CC for the surface.
Fixes: e3afc69018 ("drm/i915/display: consider DG2_RC_CCS_CC when migrating buffers")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Nirmoy Das <nirmoy.das@intel.com>
Cc: Andrzej Hajda <andrzej.hajda@intel.com>
Cc: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221212171958.82593-1-matthew.auld@intel.com
Simple regression test to check that we don't trample the
rq->reserved_space when returning from emit_pte(), if the ring is nearly
full.
v2: Make spinner_kill() static
v3: Reduce the ring size further, which should mean we need to execute less
noops; hopefully this appeases bsw. Also add some debug logging.
v4: Fix the min request construction to account for reserved_space +
I915_EMIT_PTE_NUM_DWORDS
v5: Use a simple on-stack timer to kill the spinner instead of kthread (Chris)
References: https://gitlab.freedesktop.org/drm/intel/-/issues/7535
References: https://gitlab.freedesktop.org/drm/intel/-/issues/6889
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Chris Wilson <chris.p.wilson@intel.com>
Cc: Andi Shyti <andi.shyti@linux.intel.com>
Cc: Andrzej Hajda <andrzej.hajda@intel.com>
Cc: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221202122844.428006-3-matthew.auld@intel.com
If the ring is nearly full when calling into emit_pte(), we might
incorrectly trample the reserved_space when constructing the packet to
emit the PTEs. This then triggers the GEM_BUG_ON(rq->reserved_space >
ring->space) when later submitting the request, since the request itself
doesn't have enough space left in the ring to emit things like
workarounds, breadcrumbs etc.
v2: Fix the whitespace errors
Testcase: igt@i915_selftests@live_emit_pte_full_ring
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/7535
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/6889
Fixes: cf58602164 ("drm/i915/gt: Pipelined page migration")
Signed-off-by: Chris Wilson <chris.p.wilson@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Andrzej Hajda <andrzej.hajda@intel.com>
Cc: Andi Shyti <andi.shyti@linux.intel.com>
Cc: Nirmoy Das <nirmoy.das@intel.com>
Cc: <stable@vger.kernel.org> # v5.15+
Tested-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221202122844.428006-1-matthew.auld@intel.com
The inline function has no place in i915_drv.h. Move it away, un-inline,
and untangle some header dependencies while at it.
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220914163514.1837467-1-jani.nikula@intel.com
Crucible + recent Mesa seems to sometimes hit:
GEM_BUG_ON(num_ccs_blks > NUM_CCS_BLKS_PER_XFER)
And it looks like we can also trigger this with gem_lmem_swapping, if we
modify the test to use slightly larger object sizes.
Looking closer it looks like we have the following issues in
migrate_copy():
- We are using plain integer in various places, which we can easily
overflow with a large object.
- We pass the entire object size (when the src is lmem) into
emit_pte() and then try to copy it, which doesn't work, since we
only have a few fixed sized windows in which to map the pages and
perform the copy. With an object > 8M we therefore aren't properly
copying the pages. And then with an object > 64M we trigger the
GEM_BUG_ON(num_ccs_blks > NUM_CCS_BLKS_PER_XFER).
So it looks like our copy handling for any object > 8M (which is our
CHUNK_SZ) is currently broken on DG2.
Fixes: da0595ae91 ("drm/i915/migrate: Evict and restore the flatccs capable lmem obj")
Testcase: igt@gem_lmem_swapping
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C<ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220805132240.442747-2-matthew.auld@intel.com
The kernel only manages the ccs state with lmem-only objects, however
the kernel should still take care not to leak the CCS state from the
previous user.
Fixes: 48760ffe92 ("drm/i915/gt: Clear compress metadata for Flat-ccs objects")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220727164346.282407-1-matthew.auld@intel.com
Capture the eviction details for Flat-CCS capable, lmem objects.
v2:
Fix the Flat-ccs capbility of lmem obj with smem residency
possibility [Thomas]
v3:
Fixed the suggestions [Matt]
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
cc: Thomas Hellstrom <thomas.hellstrom@linux.intel.com>
cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220502142618.2704-4-ramalingam.c@intel.com
Calculate the ccs_sz that needs to be emitted based on the src
and dst pages emitted per chunk. And handle the return value of emit_pte
for the ccs pages.
v2:
ccs_sz moved to the reduced scope [Matt]
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220502142618.2704-3-ramalingam.c@intel.com
When we are swapping out the local memory obj on flat-ccs capable platform,
we need to capture the ccs data too along with main meory and we need to
restore it when we are swapping in the content.
When lmem object is swapped into a smem obj, smem obj will
have the extra pages required to hold the ccs data corresponding to the
lmem main memory. So main memory of lmem will be copied into the initial
pages of the smem and then ccs data corresponding to the main memory
will be copied to the subsequent pages of smem. ccs data is 1/256 of
lmem size.
Swapin happens exactly in reverse order. First main memory of lmem is
restored from the smem's initial pages and the ccs data will be restored
from the subsequent pages of smem.
Extracting and restoring the CCS data is done through a special cmd called
XY_CTRL_SURF_COPY_BLT
v2: Fixing the ccs handling
v3: Handle the ccs data at same loop as main memory [Thomas]
v4: changes for emit_copy_ccs
v5: handle non-flat-ccs scenario
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Thomas Hellstrom <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220405150840.29351-10-ramalingam.c@intel.com
Xe-HP and latest devices support Flat CCS which reserved a portion of
the device memory to store compression metadata, during the clearing of
device memory buffer object we also need to clear the associated
CCS buffer.
XY_CTRL_SURF_COPY_BLT is a BLT cmd used for reading and writing the
ccs surface of a lmem memory. So on Flat-CCS capable platform we use
XY_CTRL_SURF_COPY_BLT to clear the CCS meta data.
v2: Fixed issues with platform naming [Lucas]
v3: Rebased [Ram]
Used the round_up funcs [Bob]
v4: Fixed ccs blk calculation [Ram]
Added Kdoc on flat-ccs.
v5: GENMASK is used [Matt]
mocs fix [Matt]
Comments Fix [Matt]
Flush address programming [Ram]
v6: FLUSH_DW is fixed
Few coding style fix
v7: Adopting the XY_FAST_COLOR_BLT (Thomas]
v8: XY_CTRL_SURF_COPY_BLT for ccs clearing.
v9: emit_copy_ccs is used.
v10: ctrl_surf cmds are filled in caller itself. [Thomas]
only one ctrl surf cmd is used as size of lmem is <=8M [Thomas]
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Signed-off-by: Ayaz A Siddiqui <ayaz.siddiqui@intel.com>
Reviewed-by: Thomas Hellstrom <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220405150840.29351-6-ramalingam.c@intel.com
Use faster XY_FAST_COLOR_BLT cmd on graphics version of 12 and more,
for clearing (Zero out) the pages of the newly allocated object.
XY_FAST_COLOR_BLT is faster than the older XY_COLOR_BLT.
v2:
Typo fix at title [Thomas]
v3:
XY_FAST_COLOR_BLT is used only for FLAT_CCS capable gen12+
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Thomas Hellstrom <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220405150840.29351-3-ramalingam.c@intel.com
Move the sanity check that both src and dst are never both system
memory, which should never happen on discrete, and likely means we have
a bug. The only exception is on integrated where we trigger this path in
the selftests.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Nirmoy Das <nirmoy.das@linux.intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Nirmoy Das <nirmoy.das@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220324172143.377104-2-matthew.auld@intel.com
This is all kinds of awkward since we now have to contend with using 64K
GTT pages when mapping anything in LMEM(including the page-tables
themselves).
v2(Ram)
- Document the ppGTT layout and add a better description for the
different windows.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220218184752.7524-12-ramalingam.c@intel.com
Since the gt migration code was using only a single fence for
dependencies, these were collected in a dma_fence_array. However, it
turns out that it's illegal to use some dma_fences in a dma_fence_array,
in particular other dma_fence_arrays and dma_fence_chains, and this
causes trouble for us moving forward.
Have the gt migration code instead take a const struct i915_deps for
dependencies. This means we can skip the dma_fence_array creation
and instead pass the struct i915_deps instead to circumvent the
problem.
v2:
- Make the prev_deps() function static. (kernel test robot <lkp@intel.com>)
- Update the struct i915_deps kerneldoc.
v4:
- Rebase.
Fixes: 5652df829b ("drm/i915/ttm: Update i915_gem_obj_copy_ttm() to be asynchronous")
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211221200050.436316-2-thomas.hellstrom@linux.intel.com
No need to insert PTEs for the PTE window itself, also foreach expects a
length not an end offset, which could be gigantic here with a second
engine.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206112539.3149779-3-matthew.auld@intel.com
Ensure we add the engine base only after we calculate the qword offset
into the PTE window.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206112539.3149779-2-matthew.auld@intel.com
The scratch page might not be allocated in LMEM(like on DG2), so instead
of using that as the deciding factor for where the paging structures
live, let's just query the pt before mapping it.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206112539.3149779-1-matthew.auld@intel.com
We really only need memcpy restore for objects that affect the
operability of the migrate context. That is, primarily the page-table
objects of the migrate VM.
Add an object flag, I915_BO_ALLOC_PM_EARLY for objects that need early
restores using memcpy and a way to assign LMEM page-table object flags
to be used by the vms.
Restore objects without this flag with the gpu blitter and only objects
carrying the flag using TTM memcpy.
Initially mark the migrate, gt, gtt and vgpu vms to use this flag, and
defer for a later audit which vms actually need it. Most importantly, user-
allocated vms with pinned page-table objects can be restored using the
blitter.
Performance-wise memcpy restore is probably as fast as gpu restore if not
faster, but using gpu restore will help tackling future restrictions in
mappable LMEM size.
v4:
- Don't mark the aliasing ppgtt page table flags for early resume, but
rather the ggtt page table flags as intended. (Matthew Auld)
- The check for user buffer objects during early resume is pointless, since
they are never marked I915_BO_ALLOC_PM_EARLY. (Matthew Auld)
v5:
- Mark GuC LMEM objects with I915_BO_ALLOC_PM_EARLY to have them restored
before we fire up the migrate context.
Cc: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-8-thomas.hellstrom@linux.intel.com
If the intel_engine_create_pinned_context() function returns an error
pointer, then dereferencing "ce" will Oops. Use "vm" instead of
"ce->vm".
Fixes: cf58602164 ("drm/i915/gt: Pipelined page migration")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210813113600.GC30697@kili
Previously, we were storing the ring size in the ring pointer before it
was actually allocated. We would then guard setting the ring size on
checking for CONTEXT_ALLOC_BIT. This is error-prone at best and really
only saves us a few bytes on something that already burns at least 4K.
Instead, this patch adds a new ring_size field and makes everything use
that.
v2 (Daniel Vetter):
- Replace 512 * SZ_4K with SZ_2M
v2 (Jason Ekstrand):
- Rebase on top of page migration code
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20210708154835.528166-3-jason@jlekstrand.net
Commit c816723b6b ("drm/i915/gt: replace IS_GEN and friends with
GRAPHICS_VER") converted INTEL_GEN and friends to the new version check
macros. Meanwhile, some changes sneaked in to use INTEL_GEN. Remove the
last users so we can remove the macros.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210707181325.2130821-2-lucas.demarchi@intel.com
Set up a default migration context on the GT and use it from the
selftests.
Add a perf selftest and make sure we exercise LMEM if available.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Co-developed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210617063018.92802-10-thomas.hellstrom@linux.intel.com
Update the PTE and emit a clear within a single unpreemptible packet
such that we can schedule and pipeline clears.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Co-developed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210617063018.92802-9-thomas.hellstrom@linux.intel.com
If we pipeline the PTE updates and then do the copy of those pages
within a single unpreemptible command packet, we can submit the copies
and leave them to be scheduled without having to synchronously wait
under a global lock. In order to manage migration, we need to
preallocate the page tables (and keep them pinned and available for use
at any time), causing a bottleneck for migrations as all clients must
contend on the limited resources. By inlining the ppGTT updates and
performing the blit atomically, each client only owns the PTE while in
use, and so we can reschedule individual operations however we see fit.
And most importantly, we do not need to take a global lock on the shared
vm, and wait until the operation is complete before releasing the lock
for others to claim the PTE for themselves.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Co-developed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210617063018.92802-8-thomas.hellstrom@linux.intel.com
