linux-stable.git/drivers/gpu/drm/xe, branch linux-6.15.y

drm/xe: Make dma-fences compliant with the safe access rules

2025-08-20T16:36:14+00:00

[ Upstream commit 6bd90e700b4285e6a7541e00f969cab0d696adde ]

Xe can free some of the data pointed to by the dma-fences it exports. Most
notably the timeline name can get freed if userspace closes the associated
submit queue. At the same time the fence could have been exported to a
third party (for example a sync_fence fd) which will then cause an use-
after-free on subsequent access.

To make this safe we need to make the driver compliant with the newly
documented dma-fence rules. Driver has to ensure a RCU grace period
between signalling a fence and freeing any data pointed to by said fence.

For the timeline name we simply make the queue be freed via kfree_rcu and
for the shared lock associated with multiple queues we add a RCU grace
period before freeing the per GT structure holding the lock.

Signed-off-by: Tvrtko Ursulin 
Reviewed-by: Matthew Brost 
Acked-by: Lucas De Marchi 
Signed-off-by: Tvrtko Ursulin 
Link: https://lore.kernel.org/r/20250610164226.10817-5-tvrtko.ursulin@igalia.com
Signed-off-by: Sasha Levin

drm/xe/xe_query: Use separate iterator while filling GT list

2025-08-20T16:36:08+00:00

[ Upstream commit d4eb4a010262ea7801e576d1033b355910f2f7d4 ]

The 'id' value updated by for_each_gt() is the uapi GT ID of the GTs
being iterated over, and may skip over values if a GT is not present on
the device.  Use a separate iterator for GT list array assignments to
ensure that the array will be filled properly on future platforms where
index in the GT query list may not match the uapi ID.

v2:
 - Include the missing increment of the iterator.  (Jonathan)

Cc: Jonathan Cavitt 
Reviewed-by: Jonathan Cavitt 
Link: https://lore.kernel.org/r/20250701201320.2514369-16-matthew.d.roper@intel.com
Signed-off-by: Matt Roper 
Signed-off-by: Sasha Levin

drm/xe/pf: Disable PF restart worker on device removal

2025-08-15T10:17:00+00:00

[ Upstream commit c286ce6b01f633806b4db3e4ec8e0162928299cd ]

We can't let restart worker run once device is removed, since other
data that it might want to access could be already released.
Explicitly disable worker as part of device cleanup action.

Fixes: a4d1c5d0b99b ("drm/xe/pf: Move VFs reprovisioning to worker")
Signed-off-by: Michal Wajdeczko 
Reviewed-by: Piotr Piórkowski 
Cc: Jonathan Cavitt 
Link: https://lore.kernel.org/r/20250801142822.180530-2-michal.wajdeczko@intel.com
(cherry picked from commit a424353937c24554bb242a6582ed8f018b4a411c)
Signed-off-by: Rodrigo Vivi 
Signed-off-by: Sasha Levin

drm/xe/vf: Disable CSC support on VF

2025-08-15T10:16:49+00:00

[ Upstream commit f62408efc8669b82541295a4611494c8c8c52684 ]

CSC is not accessible by VF drivers, so disable its support flag on VF
to prevent further initialization attempts.

Fixes: e02cea83d32d ("drm/xe/gsc: add Battlemage support")
Signed-off-by: Lukasz Laguna 
Cc: Alexander Usyskin 
Cc: Michal Wajdeczko 
Reviewed-by: Michal Wajdeczko 
Signed-off-by: Michal Wajdeczko 
Link: https://lore.kernel.org/r/20250729123437.5933-1-lukasz.laguna@intel.com
(cherry picked from commit 552dbba1caaf0cb40ce961806d757615e26ec668)
Signed-off-by: Rodrigo Vivi 
Signed-off-by: Sasha Levin

drm/xe: Correct BMG VSEC header sizing

2025-08-15T10:16:22+00:00

[ Upstream commit 5b27388171a18cf6842c700520086ec50194e858 ]

The intel_vsec_header information for the crashlog feature is
incorrect.

Update the VSEC header with correct sizing and count.

Since the crashlog entries are "merged" (num_entries = 2), the
separate capabilities entries must be merged as well.

Fixes: 0c45e76fcc62 ("drm/xe/vsec: Support BMG devices")
Acked-by: Rodrigo Vivi 
Signed-off-by: Michael J. Ruhl 
Reviewed-by: David E. Box 
Link: https://lore.kernel.org/r/20250713172943.7335-4-michael.j.ruhl@intel.com
Signed-off-by: Ilpo Järvinen 
Signed-off-by: Sasha Levin

drm/xe: Correct the rev value for the DVSEC entries

2025-08-15T10:16:22+00:00

[ Upstream commit 0ba9e9cf76f2487654bc9bca38218780fa53030e ]

By definition, the Designated Vendor Specific Extended Capability
(DVSEC) revision should be 1.

Add the rev value to be correct.

Fixes: 0c45e76fcc62 ("drm/xe/vsec: Support BMG devices")
Signed-off-by: Michael J. Ruhl 
Reviewed-by: David E. Box 
Link: https://lore.kernel.org/r/20250713172943.7335-3-michael.j.ruhl@intel.com
Reviewed-by: Ilpo Järvinen 
Signed-off-by: Ilpo Järvinen 
Signed-off-by: Sasha Levin

drm/xe: Make WA BB part of LRC BO

2025-08-01T08:51:28+00:00

commit afcad92411772a1f361339f22c49f855c6cc7d0f upstream.

No idea why, but without this GuC context switches randomly fail when
running IGTs in a loop. Need to follow up why this fixes the
aforementioned issue but can live with a stable driver for now.

Fixes: 617d824c5323 ("drm/xe: Add WA BB to capture active context utilization")
Cc: stable@vger.kernel.org
Signed-off-by: Matthew Brost 
Reviewed-by: Lucas De Marchi 
Tested-by: Shuicheng Lin 
Link: https://lore.kernel.org/r/20250612031925.4009701-1-matthew.brost@intel.com
(cherry picked from commit 3a1edef8f4b58b0ba826bc68bf4bce4bdf59ecf3)
Signed-off-by: Lucas De Marchi 
[ adapted xe_bo_create_pin_map() call ]
Signed-off-by: Sasha Levin 
Signed-off-by: Greg Kroah-Hartman

drm/xe: Move page fault init after topology init

2025-07-24T06:58:43+00:00

commit 3155ac89251dcb5e35a3ec2f60a74a6ed22c56fd upstream.

We need the topology to determine GT page fault queue size, move page
fault init after topology init.

Cc: stable@vger.kernel.org
Fixes: 3338e4f90c14 ("drm/xe: Use topology to determine page fault queue size")
Signed-off-by: Matthew Brost 
Reviewed-by: Jonathan Cavitt 
Reviewed-by: Stuart Summers 
Link: https://lore.kernel.org/r/20250710191208.1040215-1-matthew.brost@intel.com
(cherry picked from commit beb72acb5b38dbe670d8eb752d1ad7a32f9c4119)
Signed-off-by: Lucas De Marchi 
Signed-off-by: Sasha Levin 
Signed-off-by: Greg Kroah-Hartman

drm/xe/mocs: Initialize MOCS index early

2025-07-24T06:58:43+00:00

commit 2a58b21adee3df10ca6f4491af965c4890d2d8e3 upstream.

MOCS uc_index is used even before it is initialized in the following
callstack
    guc_prepare_xfer()
    __xe_guc_upload()
    xe_guc_min_load_for_hwconfig()
    xe_uc_init_hwconfig()
    xe_gt_init_hwconfig()

Do MOCS index initialization earlier in the device probe.

Signed-off-by: Balasubramani Vivekanandan 
Reviewed-by: Ravi Kumar Vodapalli 
Link: https://lore.kernel.org/r/20250520142445.2792824-1-balasubramani.vivekanandan@intel.com
Signed-off-by: Matt Roper 
(cherry picked from commit 241cc827c0987d7173714fc5a95a7c8fc9bf15c0)
Signed-off-by: Lucas De Marchi 
Stable-dep-of: 3155ac89251d ("drm/xe: Move page fault init after topology init")
Signed-off-by: Sasha Levin 
Signed-off-by: Greg Kroah-Hartman

drm/xe/pf: Resend PF provisioning after GT reset

2025-07-24T06:58:38+00:00

[ Upstream commit 5c244eeca57ff4e47e1f60310d059346d1b86b9b ]

If we reload the GuC due to suspend/resume or GT reset then we
have to resend not only any VFs provisioning data, but also PF
configuration, like scheduling parameters (EQ, PT), as otherwise
GuC will continue to use default values.

Fixes: 411220808cee ("drm/xe/pf: Restart VFs provisioning after GT reset")
Signed-off-by: Michal Wajdeczko 
Reviewed-by: Piotr Piórkowski 
Link: https://lore.kernel.org/r/20250711193316.1920-3-michal.wajdeczko@intel.com
(cherry picked from commit 1c38dd6afa4a8ecce28e94da794fd1d205c30f51)
Signed-off-by: Lucas De Marchi 
Signed-off-by: Sasha Levin