When Wa_22010954014 and Wa_14022085890 were first implemented, we didn't
have a device workaround infrastructure so we hacked them into the GT
workaround list. Now that we have proper device workaround support,
move them to the proper place. Note that Wa_14022085890 specifically
applies to BMG-G21 platforms, so this requires defining a BMG
subplatform to capture the correct subset of device IDs.
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-40-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
The display part of Wa_22019338487 (i.e., avoiding use of stolen memory)
is using a platform test rather than an graphics/media IP test. Since
this workaround is focused on non-GT uses of stolen memory, it makes
sense that we'd want to still apply the workaround on affected platforms
even if the GTs themselves are disabled via configfs.
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-38-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
If the primary GT is disabled via configfs, we need to make sure that we
don't search for this workaround on a NULL xe_gt pointer. Since we can
disable the primary GT only on igpu platforms, the media GT is the one
we'd want to check anyway for this workaround.
The ternary operators in ggtt_update_access_counter() were getting a bit
long/complicated, so rewrite them with regular if/else statements.
While we're at it, throw in a couple extra assertions to make sure that
we're truly picking the expected GT according to igpu/dgpu type.
v2:
- Adjust indentation/wrapping; it's easier to read this with longer,
unwrapped lines. (Lucas)
- Tweak wording of commit message to remove ambiguity. (Gustavo)
Cc: Gustavo Sousa <gustavo.sousa@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-37-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
During the early days of the Xe driver, there were cases where we
accessed some fields in the primary GT's xe_gt structure before the GT
itself was formally initialized; this required that the structure itself
be allocated during xe_tile_init_early(). A lot of refactoring of the
device probe has happened since that time and there's no longer a need
to allocate the primary GT early. Move the allocation into
xe_info_init() where GT initialization happens and where we're doing the
allocation of the media GT.
v2:
- Only make this change after a separate patch to perform VF GMD_ID
lookup with a dummy GT instead of xe_root_mmio_gt().
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-33-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
SRIOV VF initialization has a bit of a chicken and egg design problem.
Determining the IP version of the graphics and media IPs can't be done
via direct register reads as it is on PF or native and instead requires
querying the GuC. However initialization of the GT, including its GuC,
needs to wait until after we know the IP versions so that the proper
initialization steps for the platform/IP are followed.
Currently the (somewhat hacky) solution is to manually fill out just
enough fields in tile 0's primary GT structure to make it look as if the
GT has been initialized so that the GuC can be partially initialized and
queried to obtain the GMD_ID values. When the GT gets properly
initialized during the regular flows, the hacked-up values will get
overwritten as part of the general initialization flows.
Rather than using tile 0's primary GT structure to hold the hacked up
values for querying every GT on every tile, instead allocate a dedicated
dummy structure. This will allow us to move the tile->primary_gt's
allocation to a more consistent place later in the initialization flow
in future patches (i.e., we shouldn't even allocate this GT structure if
the GT is disabled/unavailable). It also helps ensure there can't be
any accidental leakage of initialization or state between the dummy
initialization for GMD_ID and the real driver initialization of the GT.
v2:
- Initialize gt->tile for temporary GT. (CI, Michal)
- Use scope-based cleanup handler to free temp GT. (Michal)
- Propagate actual error code from xe_gt_sriov_vf_bootstrap() rather
than just setting IP version to 0.0 now that read_gmdid() can return
an error. (Michal)
v3:
- Explicitly initialize gt to NULL, just in case something else gets
inserted before the kzalloc() in the future. (Lucas)
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-32-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
FlatCCS presence/absence is a flag that should be tracked at the
platform level rather than the IP level. FlatCCS affects the
device-wide memory initialization and reservations so its effects are
not confined to a single IP block or GT. This is also a trait that
should be tied to the platform even if the graphics IP itself is not
present (e.g., if we disable the primary GT via configfs).
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-31-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Restrictions and requirements on VRAM alignment are something that
should be tracked at the platform level rather than the IP level. Even
when mixing and matching various graphics, media, and display IP blocks,
the platform as a whole has to have consistent memory allocation
handling. This is also a trait that should be tied to the platform even
if the graphics IP itself is not present (e.g., if we disable the
primary GT via configfs).
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-30-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
The number of page table levels for PPGTT virtual addresses is something
that should be tracked at the platform level rather than the IP level.
Even when mixing and matching various graphics, media, and display IP
blocks, the platform as a whole has to have consistent page table
handling. This is also a trait that should be tied to the platform even
if the graphics IP itself is not present (e.g., if we disable the
primary GT via configfs).
v2:
- Drop default value of 4 and explicitly set the value in each platform
descriptor. (Lucas)
v3:
- Drop outdated code comment and commit message paragraph about default
value. (Gustavo)
v4:
- Add missing setting for tgl_desc. (Gustavo)
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Gustavo Sousa <gustavo.sousa@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-29-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
The number of virtual address bits is something that should be tracked
at the platform level rather than the IP level. Even when mixing and
matching various graphics, media, and display IP blocks, the platform as
a whole has to have consistent page table handling. This is also a
trait that should be tied to the platform even if the graphics IP itself
is not present (e.g., if we disable the primary GT via configfs).
v2:
- Drop the default value of 48 and explicitly set it in each relevant
descriptor. (Lucas, Michal)
v3:
- Drop an outdated comment about default value. (Michal)
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-28-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Display interrupt handling has no relation to GT(s) on the platforms
supported by the Xe driver. We only call xe_display_irq_postinstall
with the first tile's primary GT, so the single condition that uses the
GT pointer within the function always evaluates to true. Drop the
unnecessary parameter and the condition.
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-27-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
The HuC initialization code determines whether a platform can have a HuC
on the primary GT by checking whether tile->media_gt is NULL; old Xe1
platforms that combined render+media into a single GT will always have a
NULL media_gt pointer. However once we allow media to be disabled via
configfs, there will also be cases where tile->media_gt is NULL on more
modern platforms, causing this condition to behave incorrectly.
To handle cases where media gets disabled via configfs (or theoretical
cases where media is truly fused off in hardware), change the condition
to consider the graphics version of the primary GT; only the old Xe1
platforms with graphics versions 12.55 or earlier should try to
initialize a HuC on the primary GT.
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
Link: https://lore.kernel.org/r/20251013200944.2499947-26-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Kunit's generate_params() was recently updated to take an additional
test context parameter. Xe's IP and platform parameter generators were
updated accordingly at the same time, but the new parameter was not
added to the functions' kerneldoc, resulting in the following warnings:
Warning: drivers/gpu/drm/xe/tests/xe_pci.c:78 function parameter 'test' not described in 'xe_pci_fake_data_gen_params'
Warning: drivers/gpu/drm/xe/tests/xe_pci.c:254 function parameter 'test' not described in 'xe_pci_graphics_ip_gen_param'
Warning: drivers/gpu/drm/xe/tests/xe_pci.c:278 function parameter 'test' not described in 'xe_pci_media_ip_gen_param'
Warning: drivers/gpu/drm/xe/tests/xe_pci.c:302 function parameter 'test' not described in 'xe_pci_id_gen_param'
Warning: drivers/gpu/drm/xe/tests/xe_pci.c:390 function parameter 'test' not described in 'xe_pci_live_device_gen_param'
5 warnings as errors
Document the new parameter to eliminate the warnings and make CI happy.
Fixes: b9a214b5f6 ("kunit: Pass parameterized test context to generate_params()")
Reviewed-by: Shuicheng Lin <shuicheng.lin@intel.com>
Link: https://lore.kernel.org/r/20251013153014.2362879-2-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
If the location madvise() is set to
DRM_XE_PREFERRED_LOC_DEFAULT_SYSTEM, the drm_pagemap in the
SVM gpu fault handler will be set to NULL. However there is nothing
that explicitly migrates the data to system if it is already present
in device memory.
In that case, set the device memory owner to NULL to ensure
data gets properly migrated to system on page-fault.
v2:
- Remove redundant dpagemap assignment (Himal Prasad Ghimiray)
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com> #v1
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Link: https://lore.kernel.org/r/20251010104149.72783-2-thomas.hellstrom@linux.intel.com
Fixes: 10aa5c8060 ("drm/gpusvm, drm/xe: Fix userptr to not allow device private pages")
The function `guc_ct_send_locked()` is really quite simple, but still
looks complex due to exposed internals. It is sending a message,
and in case of lack of space, waiting for a proper moment to send
a retry.
Clear separation of send function and wait function will help with
readability. This is a cosmetic change only, no functional difference
is expected.
This patch introduces `guc_ct_send_wait_for_retry()`, and uses it to
greatly simplify `guc_ct_send_locked()`.
Signed-off-by: Tomasz Lis <tomasz.lis@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://lore.kernel.org/r/20251009170844.178199-1-tomasz.lis@intel.com
Using 2M pages in xe_migrate_vram has two benefits: we issue fewer
instructions per 2M copy (1 vs. 512), and the cache hit rate should be
higher. This results in increased copy engine bandwidth, as shown by
benchmark IGTs.
Enable 2M pages by reserving PDEs in the migrate VM and using 2M pages
in xe_migrate_vram if the DMA address order matches 2M.
v2:
- Reuse build_pt_update_batch_sram (Stuart)
- Fix build_pt_update_batch_sram for PAGE_SIZE > 4K
v3:
- More fixes for PAGE_SIZE > 4K, align chunk, decrement chunk as needed
- Use stack incr var in xe_migrate_vram_use_pde (Stuart)
v4:
- Split PAGE_SIZE > 4K fix out in different patch (Stuart)
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Stuart Summers <stuart.summers@intel.com>
Link: https://lore.kernel.org/r/20251013034555.4121168-3-matthew.brost@intel.com
Correct several spelling and grammar issues in xe_gt struct
documentation to improve readability:
- Fix "to not" -> "do not".
- Fix "mmigrations" -> "migrations".
- Fix "Multiple queues exists" -> "Multiple queues exist".
- Fix "are be processed" -> "to be processed".
- Fix "have being processed" -> "have been processed".
These changes are purely cosmetic and do not affect functionality.
v2: drop kernel-doc formatting change. (Jani)
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20251006151317.2553182-2-shuicheng.lin@intel.com
In normal operation, a registered exec queue is disabled and
deregistered through the GuC, and freed only after the GuC confirms
completion. However, if the driver is forced to unbind while the exec
queue is still running, the user may call exec_destroy() after the GuC
has already been stopped and CT communication disabled.
In this case, the driver cannot receive a response from the GuC,
preventing proper cleanup of exec queue resources. Fix this by directly
releasing the resources when GuC is not running.
Here is the failure dmesg log:
"
[ 468.089581] ---[ end trace 0000000000000000 ]---
[ 468.089608] pci 0000:03:00.0: [drm] *ERROR* GT0: GUC ID manager unclean (1/65535)
[ 468.090558] pci 0000:03:00.0: [drm] GT0: total 65535
[ 468.090562] pci 0000:03:00.0: [drm] GT0: used 1
[ 468.090564] pci 0000:03:00.0: [drm] GT0: range 1..1 (1)
[ 468.092716] ------------[ cut here ]------------
[ 468.092719] WARNING: CPU: 14 PID: 4775 at drivers/gpu/drm/xe/xe_ttm_vram_mgr.c:298 ttm_vram_mgr_fini+0xf8/0x130 [xe]
"
v2: use xe_uc_fw_is_running() instead of xe_guc_ct_enabled().
As CT may go down and come back during VF migration.
Fixes: dd08ebf6c3 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Cc: stable@vger.kernel.org
Cc: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20251010172529.2967639-2-shuicheng.lin@intel.com
Pull more drm fixes from Dave Airlie:
"Just the follow up fixes for rc1 from the next branch, amdgpu and xe
mostly with a single v3d fix in there.
amdgpu:
- DC DCE6 fixes
- GPU reset fixes
- Secure diplay messaging cleanup
- MES fix
- GPUVM locking fixes
- PMFW messaging cleanup
- PCI US/DS switch handling fix
- VCN queue reset fix
- DC FPU handling fix
- DCN 3.5 fix
- DC mirroring fix
amdkfd:
- Fix kfd process ref leak
- mmap write lock handling fix
- Fix comments in IOCTL
xe:
- Fix build with clang 16
- Fix handling of invalid configfs syntax usage and spell out the
expected syntax in the documentation
- Do not try late bind firmware when running as VF since it shouldn't
handle firmware loading
- Fix idle assertion for local BOs
- Fix uninitialized variable for late binding
- Do not require perfmon_capable to expose free memory at page
granularity. Handle it like other drm drivers do
- Fix lock handling on suspend error path
- Fix I2C controller resume after S3
v3d:
- fix fence locking"
* tag 'drm-next-2025-10-11-1' of https://gitlab.freedesktop.org/drm/kernel: (34 commits)
drm/amd/display: Incorrect Mirror Cositing
drm/amd/display: Enable Dynamic DTBCLK Switch
drm/amdgpu: Report individual reset error
drm/amdgpu: partially revert "revert to old status lock handling v3"
drm/amd/display: Fix unsafe uses of kernel mode FPU
drm/amd/pm: Disable VCN queue reset on SMU v13.0.6 due to regression
drm/amdgpu: Fix general protection fault in amdgpu_vm_bo_reset_state_machine
drm/amdgpu: Check swus/ds for switch state save
drm/amdkfd: Fix two comments in kfd_ioctl.h
drm/amd/pm: Avoid interface mismatch messaging
drm/amdgpu: Merge amdgpu_vm_set_pasid into amdgpu_vm_init
drm/amd/amdgpu: Fix the mes version that support inv_tlbs
drm/amd: Check whether secure display TA loaded successfully
drm/amdkfd: Fix mmap write lock not release
drm/amdkfd: Fix kfd process ref leaking when userptr unmapping
drm/amdgpu: Fix for GPU reset being blocked by KIQ I/O.
drm/amd/display: Disable scaling on DCE6 for now
drm/amd/display: Properly disable scaling on DCE6
drm/amd/display: Properly clear SCL_*_FILTER_CONTROL on DCE6
drm/amd/display: Add missing DCE6 SCL_HORZ_FILTER_INIT* SRIs
...
Pull drm fixes from Dave Airlie:
"Some fixes leftover from our fixes branch, just nouveau and vmwgfx:
nouveau:
- Return errno code from TTM move helper
vmwgfx:
- Fix null-ptr access in cursor code
- Fix UAF in validation
- Use correct iterator in validation"
* tag 'drm-fixes-2025-10-11' of https://gitlab.freedesktop.org/drm/kernel:
drm/nouveau: fix bad ret code in nouveau_bo_move_prep
drm/vmwgfx: Fix copy-paste typo in validation
drm/vmwgfx: Fix Use-after-free in validation
drm/vmwgfx: Fix a null-ptr access in the cursor snooper
Moving to VRAM will fail if mixed mappings are present or if the page is
already located in VRAM. Atomic faults that require a move to VRAM
currently retry without attempting to evict mixed mappings or locate
existing VRAM mappings.
This patch fixes the issue by attempting to evict mixed mappings or find
existing VRAM pages when a move to VRAM fails during atomic fault
handling.
Fixes: a9ac0fa455 ("drm/xe: Strict migration policy for atomic SVM faults")
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Link: https://lore.kernel.org/r/20251009130629.3531962-1-matthew.brost@intel.com
There may be cases in which the BAR0 also needs to move to accommodate
the bigger BAR2. However if it's not released, the BAR2 resize fails.
During the vram probe it can't be released as it's already in use by
xe_mmio for early register access.
Add a new function in xe_vram and let xe_pci call it directly before
even early device probe. This allows the BAR2 to resize in cases BAR0
also needs to move, assuming there aren't other reasons to hold that
move:
[] xe 0000:03:00.0: vgaarb: deactivate vga console
[] xe 0000:03:00.0: [drm] Attempting to resize bar from 8192MiB -> 16384MiB
[] xe 0000:03:00.0: BAR 0 [mem 0x83000000-0x83ffffff 64bit]: releasing
[] xe 0000:03:00.0: BAR 2 [mem 0x4000000000-0x41ffffffff 64bit pref]: releasing
[] pcieport 0000:02:01.0: bridge window [mem 0x4000000000-0x41ffffffff 64bit pref]: releasing
[] pcieport 0000:01:00.0: bridge window [mem 0x4000000000-0x41ffffffff 64bit pref]: releasing
[] pcieport 0000:01:00.0: bridge window [mem 0x4000000000-0x43ffffffff 64bit pref]: assigned
[] pcieport 0000:02:01.0: bridge window [mem 0x4000000000-0x43ffffffff 64bit pref]: assigned
[] xe 0000:03:00.0: BAR 2 [mem 0x4000000000-0x43ffffffff 64bit pref]: assigned
[] xe 0000:03:00.0: BAR 0 [mem 0x83000000-0x83ffffff 64bit]: assigned
[] pcieport 0000:00:01.0: PCI bridge to [bus 01-04]
[] pcieport 0000:00:01.0: bridge window [mem 0x83000000-0x840fffff]
[] pcieport 0000:00:01.0: bridge window [mem 0x4000000000-0x44007fffff 64bit pref]
[] pcieport 0000:01:00.0: PCI bridge to [bus 02-04]
[] pcieport 0000:01:00.0: bridge window [mem 0x83000000-0x840fffff]
[] pcieport 0000:01:00.0: bridge window [mem 0x4000000000-0x43ffffffff 64bit pref]
[] pcieport 0000:02:01.0: PCI bridge to [bus 03]
[] pcieport 0000:02:01.0: bridge window [mem 0x83000000-0x83ffffff]
[] pcieport 0000:02:01.0: bridge window [mem 0x4000000000-0x43ffffffff 64bit pref]
[] xe 0000:03:00.0: [drm] BAR2 resized to 16384M
[] xe 0000:03:00.0: [drm:xe_pci_probe [xe]] BATTLEMAGE e221:0000 dgfx:1 gfx:Xe2_HPG (20.02) ...
For BMG there are additional fix needed in the PCI side, but this
helps getting it to a working resize.
All the rebar logic is more pci-specific than xe-specific and can be
done very early in the probe sequence. In future it would be good to
move it out of xe_vram.c, but this refactor is left for later.
Cc: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Cc: stable@vger.kernel.org # 6.12+
Link: https://lore.kernel.org/intel-xe/fafda2a3-fc63-ce97-d22b-803f771a4d19@linux.intel.com
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://lore.kernel.org/r/20250918-xe-pci-rebar-2-v1-2-6c094702a074@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
It is possible that the media GT's VF post-migration recovery work item
gets scheduled before the primary GT's work item. Since the media GT
depends on the primary GT's work item to complete CCS restore, if the
media GT's work item is scheduled first, detect this condition and
re-queue the media GT's work item for a later time.
v5:
- Adjust debug message (Tomasz)
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Tomasz Lis <tomasz.lis@intel.com>
Link: https://lore.kernel.org/r/20251008214532.3442967-33-matthew.brost@intel.com
VF CCS restore is a primary GT operation on which the media GT depends.
Therefore, it doesn't make much sense to run these operations in
parallel. To address this, point the media GT's ordered work queue to
the primary GT's ordered work queue on platforms that require (PTL VFs)
CCS restore as part of VF post-migration recovery.
v7:
- Remove bool from xe_gt_alloc (Lucas)
v9:
- Fix typo (Lucas)
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20251008214532.3442967-32-matthew.brost@intel.com
A race condition exists where a paused VF's H2G request can be processed
and subsequently rejected. This rejection results in a FAST_REQ failure
being delivered to the KMD, which then terminates the CT via a dead
worker and triggers a GT reset—an undesirable outcome.
This workaround mitigates the issue by checking if a VF post-migration
recovery is in progress and aborting these adverse actions accordingly.
The GuC firmware will address this bug in an upcoming release. Once that
version is available and VF migration depends on it, this workaround can
be safely removed.
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Tomasz Lis <tomasz.lis@intel.com>
Link: https://lore.kernel.org/r/20251008214532.3442967-30-matthew.brost@intel.com
A queue must be in the submission backend's tracking state before the
LRC is created to avoid a race condition where the LRC's GGTT addresses
are not properly fixed up during VF post-migration recovery.
Move the queue initialization—which adds the queue to the submission
backend's tracking state—before LRC creation.
Also wait on pending GGTT fixups before allocating LRCs to avoid racing
with fixups.
v2:
- Wait on VF GGTT fixes before creating LRC (testing)
v5:
- Adjust comment in code (Tomasz)
- Reduce race window
v7:
- Only wakeup waiters in recovery path (CI)
- Wakeup waiters on abort
- Use GT warn on (Michal)
- Fix kernel doc for LRC ring size function (Tomasz)
v8:
- Guard against migration not supported or no memirq (CI)
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Tomasz Lis <tomasz.lis@intel.com>
Link: https://lore.kernel.org/r/20251008214532.3442967-28-matthew.brost@intel.com
Fixup GuC submission pause / unpause functions to properly replay any
possible state lost during VF post migration recovery.
v3:
- Add helpers for revert / replay (Tomasz)
- Add comment around WQ NOPs (Tomasz)
v7:
- Only fixup / replay parallel queues once (Testing)
- Skip unpause step on queues created after resfix done (Testing)
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Tomasz Lis <tomasz.lis@intel.com>
Link: https://lore.kernel.org/r/20251008214532.3442967-27-matthew.brost@intel.com
