ccs_mode_store() calls xe_gt_reset() which internally invokes
xe_pm_runtime_get_noresume(). That function requires the caller
to already hold an outer runtime PM reference and warns if none
is held:
[46.891177] xe 0000:03:00.0: [drm] Missing outer runtime PM protection
[46.891178] WARNING: drivers/gpu/drm/xe/xe_pm.c:885 at
xe_pm_runtime_get_noresume+0x8b/0xc0
Fix this by protecting xe_gt_reset() with the scope-based
guard(xe_pm_runtime)(xe), which is the preferred form when
the reference lifetime matches a single scope.
v2:
- Use scope-based guard(xe_pm_runtime)(xe) (Shuicheng)
- Update commit message accordingly
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/7593
Fixes: 480b358e7d ("drm/xe: Do not wake device during a GT reset")
Cc: <stable@vger.kernel.org> # v6.19+
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Shuicheng Lin <shuicheng.lin@intel.com>
Suggested-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Sanjay Yadav <sanjay.kumar.yadav@intel.com>
Reviewed-by: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patch.msgid.link/20260313071608.3459480-2-sanjay.kumar.yadav@intel.com
Getting engine specific CTX TIMESTAMP register can fail. In that case,
if the context is active, new_ts is uninitialized. Fix that case by
initializing new_ts to the last value that was sampled in SW -
lrc->ctx_timestamp.
Flagged by static analysis.
v2: Fix new_ts initialization (Ashutosh)
Fixes: bb63e7257e ("drm/xe: Avoid toggling schedule state to check LRC timestamp in TDR")
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Link: https://patch.msgid.link/20260312125308.3126607-2-umesh.nerlige.ramappa@intel.com
Some OA data might be present in the OA buffer when OA stream is
disabled. Allow UMD's to retrieve this data, so that all data till the
point when OA stream is disabled can be retrieved.
v2: Update tail pointer after disable (Umesh)
Fixes: efb315d0a0 ("drm/xe/oa/uapi: Read file_operation")
Cc: stable@vger.kernel.org
Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Reviewed-by: Umesh Nerlige Ramappa<umesh.nerlige.ramappa@intel.com>
Link: https://patch.msgid.link/20260313053630.3176100-1-ashutosh.dixit@intel.com
NULL VMAs have no corresponding PTE, so skip adding a PRL entry to avoid
an unnecessary PRL abort during unbind.
Signed-off-by: Brian Nguyen <brian3.nguyen@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260305171546.67691-8-brian3.nguyen@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Originally, page reclamation is handled by the same fence as tlb
invalidation and uses its seqno, so there was no reason to separate out
the handlers. However in hindsight, for readability, and possible
future changes, it seems more beneficial to move this all out to its own
function.
Signed-off-by: Brian Nguyen <brian3.nguyen@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Stuart Summers <stuart.summers@intel.com>
Link: https://patch.msgid.link/20260305171546.67691-7-brian3.nguyen@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
The check using xe_child->base.children was insufficient in determining
if a pte was a leaf node. So explicitly skip over every non-leaf pt and
conditionally abort if there is a scenario where a non-leaf pt is
interleaved between leaf pt, which results in the page walker skipping
over some leaf pt.
Note that the behavior being targeted for abort is
PD[0] = 2M PTE
PD[1] = PT -> 512 4K PTEs
PD[2] = 2M PTE
results in abort, page walker won't descend PD[1].
With new abort, ensuring valid PRL before handling a second abort.
v2:
- Revert to previous assert.
- Revised non-leaf handling for interleaf child pt and leaf pte.
- Update comments to specifications. (Stuart)
- Remove unnecessary XE_PTE_PS64. (Matthew B)
v3:
- Modify secondary abort to only check non-leaf PTEs. (Matthew B)
Fixes: b912138df2 ("drm/xe: Create page reclaim list on unbind")
Signed-off-by: Brian Nguyen <brian3.nguyen@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Cc: Stuart Summers <stuart.summers@intel.com>
Link: https://patch.msgid.link/20260305171546.67691-6-brian3.nguyen@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Printing a running dword offset in the default_lrc_* debugfs entries
makes it easier for developers to find the right offsets to use in
regs/xe_lrc_layout.h and/or compare the default LRC contents against the
bspec-documented LRC layout.
Reviewed-by: Dnyaneshwar Bhadane <dnyaneshwar.bhadane@intel.com>
Link: https://patch.msgid.link/20260311-default_lrc_offsets-v1-1-58d8ed3aa081@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
I2C IRQ is triggered using virtual wire. Assert/Deassert it in IRQ
handler to allow subsequent interrupt generation.
Signed-off-by: Raag Jadav <raag.jadav@intel.com>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://patch.msgid.link/20260313080438.4166251-1-raag.jadav@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Add .framebuffer_init, .framebuffer_fini and .framebuffer_lookup to the
bo parent interface. While they're about framebuffers, they're
specifically about framebuffer objects, so the bo interface is a good
enough fit, and there's no need to add another interface struct.
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Link: https://patch.msgid.link/848d32a44bf844cba3d66e44ba9f20bea4a8352d.1773238670.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
intel_fb_bo.c is i915 core specific code, and should use struct
drm_i915_private instead of struct intel_display.
Switch one DISPLAY_VER() to GRAPHICS_VER(). The check is for < 4, where
they're effectively the same thing.
Reviewed-by: Suraj Kandpal@intel.com>
Link: https://patch.msgid.link/13087bd24bd5af5265ca6af67f086b93e26e311f.1773238670.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Continue i915 and xe separation from display by moving the bo calls to
the display parent interface. Instead of adding all these functions to
intel_parent.[ch], reuse the now vacated intel_bo.[ch], and avoid mass
renames to calls of these functions. This is similar to
intel_display_rpm.[ch].
Make many of the hooks optional to avoid having to implement dummy
functions in xe. Indeed now we can remove many of the existing dummy
functions.
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Link: https://patch.msgid.link/7899eef2ccf0cd603df69099df065226a0df917b.1773238670.git.jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
amdgpu:
- SMU13 fix
- SMU14 fix
- Fixes for bring up hw testing
- Kerneldoc fix
- GC12 idle power fix for compute workloads
- DCCG fixes
- UserQ fixes
- Move test for fbdev object to a generic helper
- GC 12.1 updates
- Use struct drm_edid in non-DC code
- Include IP discovery data in devcoredump
- SMU 13.x updates
- Misc cleanups
- DML 2.1 fixes
- Enable NV12/P010 support on primary planes
- Enable color encoding and color range on overlay planes
- DC underflow fixes
- HWSS fast path fixes
- Replay fixes
- DCN 4.2 updates
- Support newer IP discovery tables
- LSDMA 7.1 support
- IH 7.1 fixes
- SoC v1 updates
- GC12.1 updates
- PSP 15 updates
- XGMI fixes
- GPUVM locking fix
amdkfd:
- Fix missing BO unreserve in an error path
radeon:
- Move test for fbdev object to a generic helper
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQQgO5Idg2tXNTSZAr293/aFa7yZ2AUCabMItwAKCRC93/aFa7yZ
2BiwAQD3yQsA6xuspaMXduQwFE0tu6rL6bB+NSB7bfM2YYdE1AEAqi2GQPOlfohH
DK/zb58ZPoeJfCdShPPQ85cSkPSu2wY=
=OtzL
-----END PGP SIGNATURE-----
Merge tag 'amd-drm-next-7.1-2026-03-12' of https://gitlab.freedesktop.org/agd5f/linux into drm-next
amd-drm-next-7.1-2026-03-12:
amdgpu:
- SMU13 fix
- SMU14 fix
- Fixes for bring up hw testing
- Kerneldoc fix
- GC12 idle power fix for compute workloads
- DCCG fixes
- UserQ fixes
- Move test for fbdev object to a generic helper
- GC 12.1 updates
- Use struct drm_edid in non-DC code
- Include IP discovery data in devcoredump
- SMU 13.x updates
- Misc cleanups
- DML 2.1 fixes
- Enable NV12/P010 support on primary planes
- Enable color encoding and color range on overlay planes
- DC underflow fixes
- HWSS fast path fixes
- Replay fixes
- DCN 4.2 updates
- Support newer IP discovery tables
- LSDMA 7.1 support
- IH 7.1 fixes
- SoC v1 updates
- GC12.1 updates
- PSP 15 updates
- XGMI fixes
- GPUVM locking fix
amdkfd:
- Fix missing BO unreserve in an error path
radeon:
- Move test for fbdev object to a generic helper
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patch.msgid.link/20260312184425.3875669-1-alexander.deucher@amd.com
Signed-off-by: Dave Airlie <airlied@redhat.com>
This enables support for Transparent Huge Pages (THP) for device pages by
using MIGRATE_VMA_SELECT_COMPOUND during migration. It removes the need to
split folios and loop multiple times over all pages to perform required
operations at page level. Instead, we rely on newly introduced support for
higher orders in drm_pagemap and folio-level API.
In Xe, this drastically improves performance when using SVM. The GT stats
below collected after a 2MB page fault show overall servicing is more than
7 times faster, and thanks to reduced CPU overhead the time spent on the
actual copy goes from 23% without THP to 80% with THP:
Without THP:
svm_2M_pagefault_us: 966
svm_2M_migrate_us: 942
svm_2M_device_copy_us: 223
svm_2M_get_pages_us: 9
svm_2M_bind_us: 10
With THP:
svm_2M_pagefault_us: 132
svm_2M_migrate_us: 128
svm_2M_device_copy_us: 106
svm_2M_get_pages_us: 1
svm_2M_bind_us: 2
v2:
- Fix one occurrence of drm_pagemap_get_devmem_page() (Matthew Brost)
v3:
- Remove migrate_device_split_page() and folio_split_lock, instead rely on
free_zone_device_folio() to split folios before freeing (Matthew Brost)
- Assert folio order is HPAGE_PMD_ORDER (Matthew Brost)
- Always use folio_set_zone_device_data() in split (Matthew Brost)
v4:
- Warn on compound device page, s/continue/goto next/ (Matthew Brost)
v5:
- Revert warn on compound device page
- s/zone_device_page_init()/zone_device_folio_init() (Matthew Brost)
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Michal Mrozek <michal.mrozek@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Balbir Singh <balbirs@nvidia.com>
Cc: linux-mm@kvack.org
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Francois Dugast <francois.dugast@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260312192126.2024853-5-francois.dugast@intel.com
cpages returned from migrate_vma_setup represents the total number of
individual pages found, not the number of 4K pages. The math in
drm_pagemap_migrate_to_devmem for npages is based on the number of 4K
pages, so cpages != npages can fail even if the entire memory range is
found in migrate_vma_setup (e.g., when a single 2M page is found).
Add drm_pagemap_cpages, which converts cpages to the number of 4K pages
found.
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Balbir Singh <balbirs@nvidia.com>
Cc: linux-mm@kvack.org
Reviewed-by: Francois Dugast <francois.dugast@intel.com>
Signed-off-by: Francois Dugast <francois.dugast@intel.com>
Reviewed-by: Balbir Singh <balbirs@nvidia.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260312192126.2024853-4-francois.dugast@intel.com
This new helper helps ensure all accesses to zone_device_data use the
correct API whether the page is part of a folio or not.
v2:
- Move to drm_pagemap.h, stick to folio_zone_device_data (Matthew Brost)
- Return struct drm_pagemap_zdd * (Matthew Brost)
v3:
- Add stub for !CONFIG_ZONE_DEVICE (CI)
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Balbir Singh <balbirs@nvidia.com>
Cc: linux-mm@kvack.org
Suggested-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Francois Dugast <francois.dugast@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260312192126.2024853-3-francois.dugast@intel.com
If the page is part of a folio, unlock and put the whole folio at once
instead of individual pages one after the other. This will reduce the
amount of operations once device THP are in use.
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@kernel.org>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Balbir Singh <balbirs@nvidia.com>
Cc: linux-mm@kvack.org
Suggested-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Francois Dugast <francois.dugast@intel.com>
Reviewed-by: Balbir Singh <balbirs@nvidia.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260312192126.2024853-2-francois.dugast@intel.com
GGTT MMIO access is currently protected by hotplug (drm_dev_enter),
which works correctly when the driver loads successfully and is later
unbound or unloaded. However, if driver load fails, this protection is
insufficient because drm_dev_unplug() is never called.
Additionally, devm release functions cannot guarantee that all BOs with
GGTT mappings are destroyed before the GGTT MMIO region is removed, as
some BOs may be freed asynchronously by worker threads.
To address this, introduce an open-coded flag, protected by the GGTT
lock, that guards GGTT MMIO access. The flag is cleared during the
dev_fini_ggtt devm release function to ensure MMIO access is disabled
once teardown begins.
Cc: stable@vger.kernel.org
Fixes: 919bb54e98 ("drm/xe: Fix missing runtime outer protection for ggtt_remove_node")
Reviewed-by: Zhanjun Dong <zhanjun.dong@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260310225039.1320161-8-zhanjun.dong@intel.com
If the firmware fails to load in GT resets the device is wedged also
initiating a GuC state cleanup.
Signed-off-by: Zhanjun Dong <zhanjun.dong@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260310225039.1320161-7-zhanjun.dong@intel.com
The GuC CT state transition requires moving to the STOP state before
entering the DISABLED state. Update the driver teardown sequence to make
the proper state machine transitions.
Fixes: ee4b32220a ("drm/xe/guc: Add devm release action to safely tear down CT")
Cc: stable@vger.kernel.org
Signed-off-by: Zhanjun Dong <zhanjun.dong@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260310225039.1320161-6-zhanjun.dong@intel.com
Replace the magic number 2 with the proper enum value
XE_WEDGED_MODE_UPON_ANY_HANG_NO_RESET for better code readability
and maintainability.
Signed-off-by: Zhanjun Dong <zhanjun.dong@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260310225039.1320161-5-zhanjun.dong@intel.com
The intent of wedging a device is to allow queues to continue running
only in wedged mode 2. In other modes, queues should initiate cleanup
and signal all remaining fences. Fix xe_guc_submit_wedge to correctly
clean up queues when wedge mode != 2.
Fixes: 7dbe8af13c ("drm/xe: Wedge the entire device")
Cc: stable@vger.kernel.org
Reviewed-by: Zhanjun Dong <zhanjun.dong@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260310225039.1320161-4-zhanjun.dong@intel.com
In GuC submit fini, forcefully tear down any exec queues by disabling
CTs, stopping the scheduler (which cleans up lost G2H), killing all
remaining queues, and resuming scheduling to allow any remaining cleanup
actions to complete and signal any remaining fences.
Split guc_submit_fini into device related and software only part. Using
device-managed and drm-managed action guarantees the correct ordering of
cleanup.
Fixes: dd08ebf6c3 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Cc: stable@vger.kernel.org
Reviewed-by: Zhanjun Dong <zhanjun.dong@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260310225039.1320161-3-zhanjun.dong@intel.com
xe_guc_submit_pause_abort is intended to be called after something
disastrous occurs (e.g., VF migration fails, device wedging, or driver
unload) and should immediately trigger the teardown of remaining
submission state. With that, kill any remaining queues in this function.
Fixes: 7c4b7e34c8 ("drm/xe/vf: Abort VF post migration recovery on failure")
Cc: stable@vger.kernel.org
Signed-off-by: Zhanjun Dong <zhanjun.dong@intel.com>
Reviewed-by: Stuart Summers <stuart.summers@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260310225039.1320161-2-zhanjun.dong@intel.com
By using the same variable for both the return of poll_timeout_us and
the return of the polled function guc_wait_ucode, the return value of
the latter is overwritten and lost after exiting the polling loop. Since
guc_wait_ucode returns -1 on GuC load failure, we lose that information
and always continue as if the GuC had been loaded correctly.
This is fixed by simply using 2 separate variables.
Fixes: a4916b4da4 ("drm/xe/guc: Refactor GuC load to use poll_timeout_us()")
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Link: https://patch.msgid.link/20260303001732.2540493-2-daniele.ceraolospurio@intel.com
Since both the DP SST and MST HPD IRQ handlers call
intel_dp_handle_link_service_irq() with LINK_STATUS_CHANGED set in
irq_mask if intel_dp->link.force_retrain is set, checking for the former
flag is sufficient to determine if the link status needs to be checked
(which includes retraining the link if this is forced); remove checking
for the latter flag.
Since LINK_STATUS_CHANGED is currently set unconditionally for DP SST,
extend the related comment to note that it must be set if
intel_dp->link.force_retrain is set (in case setting LINK_STATUS_CHANGED
becomes conditional on DPCD_REV).
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patch.msgid.link/20260311153152.133744-2-imre.deak@intel.com
Handling of a forced link retraining debugfs request via the DP MST HPD
IRQ handler is incorrectly skipped, if the IRQ handler doesn't see any
HPD IRQs raised by the sink. Fix this by ensuring that the request is
always handled (in the Fixes: commit below by directly calling
intel_dp_check_link_state(), later by the same call moved to
intel_dp_handle_link_service_irq()).
Cc: Luca Coelho <luciano.coelho@intel.com>
Fixes: db4855d903 ("drm/i915/dp_mst: Reuse intel_dp_check_link_state() in the HPD IRQ handler")
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patch.msgid.link/20260311153152.133744-1-imre.deak@intel.com
There is kernel-doc warning for DRM_XE_VM_BIND_FLAG_DECOMPRESS:
./include/uapi/drm/xe_drm.h:1060: WARNING: Block quote ends without
a blank line; unexpected unindent.
Fix the warning by adding the missing '%' prefix to
DRM_XE_VM_BIND_FLAG_DECOMPRESS in the kernel-doc list entry for
struct drm_xe_vm_bind_op.
Fixes: 2270bd7124 ("drm/xe: add VM_BIND DECOMPRESS uapi flag")
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202603121515.gEMrFlTL-lkp@intel.com/
Cc: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Nitin Gote <nitin.r.gote@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patch.msgid.link/20260312160244.809849-2-nitin.r.gote@intel.com
Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
During intel_hdcp_check_link phase we need to take into account
if we are currently forcing HDCP 1.4 or not. This is because
we check for HDCP 2.x Link first and only if HDCP 2.x is not being
used check for HDCP 1.4. With force_hdcp14 in picture we should not
be going into intel_hdcp2_check_link because of which we may end
up trying to disable HDCP2.x even if HDCP 1.4 was enabled causing
a lot of issues while IGT tests this.
Signed-off-by: Suraj Kandpal <suraj.kandpal@intel.com>
Reviewed-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Link: https://patch.msgid.link/20260225065045.3040787-1-suraj.kandpal@intel.com
The Xe2_HPM-specific RTP table entries for Wa_16021867713 and
Wa_14019449301 were removed by commit 941f538b0a ("drm/xe: Consolidate
workaround entries for Wa_16021867713") and commit aa0f0a6783
("drm/xe: Consolidate workaround entries for Wa_14019449301") in favor
of alternate entries earlier in the table that cover a wider range of IP
versions. However these Xe2_HPM-specific entries were accidentally
resurrected during a backmerge, which causes the Xe driver to complain
on probe about two entries trying to program the same registers+bits:
<3> [48.491155] xe 0000:03:00.0: [drm] *ERROR* Tile0: GT1: discarding save-restore reg 1c3f1c (clear: 00000008, set: 00000008, masked: no, mcr: no): ret=-22
<3> [48.491211] xe 0000:03:00.0: [drm] *ERROR* Tile0: GT1: discarding save-restore reg 1d3f1c (clear: 00000008, set: 00000008, masked: no, mcr: no): ret=-22
<3> [48.491225] xe 0000:03:00.0: [drm] *ERROR* Tile0: GT1: discarding save-restore reg 1c3f08 (clear: 00000020, set: 00000020, masked: no, mcr: no): ret=-22
<3> [48.491238] xe 0000:03:00.0: [drm] *ERROR* Tile0: GT1: discarding save-restore reg 1d3f08 (clear: 00000020, set: 00000020, masked: no, mcr: no): ret=-22
Re-drop the redundant Xe2_HPM-specific entries to eliminate the dmesg
errors.
Fixes: 58351f46de ("Merge v7.0-rc3 into drm-next")
Cc: Simona Vetter <simona.vetter@ffwll.ch>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/7608
Reviewed-by: Simona Vetter <simona.vetter@ffwll.ch>
Link: https://patch.msgid.link/20260312-wa_merge_fix-v1-1-2ec6607f1e0c@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
snapshot->ctb is u32*, so pointer arithmetic on it scales
the byte offset from xe_bo_size() by 4, overshooting the
intended start of the g2h portion and writing past the
allocated buffer.
Fix this by using void * to get the arithmetic right and
prevent future mishaps.
v2: s/u8/void for memcpy and iosys_map consistency (Matt)
Fixes: af3de6cf06 ("drm/xe: Split H2G and G2H into separate buffer objects")
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: intel-xe@lists.freedesktop.org
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260304211728.249104-1-mika.kuoppala@linux.intel.com
Implement handling of VM_BIND(..., DECOMPRESS) in xe_vm_bind_ioctl.
Key changes:
- Parse and record per-op intent (op->map.request_decompress) when the
DECOMPRESS flag is present.
- Use xe_pat_index_get_comp_en() helper to check if a PAT index
has compression enabled via the XE2_COMP_EN bit.
- Validate DECOMPRESS preconditions in the ioctl path:
- Only valid for MAP ops.
- The provided pat_index must select the device's "no-compression" PAT.
- Only meaningful on devices with flat CCS and the required XE2+
otherwise return -EOPNOTSUPP.
- Use XE_IOCTL_DBG for uAPI sanity checks.
- Implement xe_bo_decompress():
For VRAM BOs run xe_bo_move_notify(), reserve one fence slot,
schedule xe_migrate_resolve(), and attach the returned fence
with DMA_RESV_USAGE_KERNEL. Non-VRAM cases are silent no-ops.
- Wire scheduling into vma_lock_and_validate() so VM_BIND will schedule
decompression when request_decompress is set.
- Handle fault-mode VMs by performing decompression synchronously during
the bind process, ensuring that the resolve is completed before the bind
finishes.
This schedules an in-place GPU resolve (xe_migrate_resolve) for
decompression.
Compute PR: https://github.com/intel/compute-runtime/pull/898
IGT PR: https://patchwork.freedesktop.org/series/157553/
v7: Rebase on latest drm-tip and add compute and igt pr info
v6: (Matt Auld)
- Rebase as xe_pat_index_get_comp_en() is added in separate
patch
- Drop vm param from xe_bo_decompress(), instead of it
extract tile from bo
- Reject decompression on igpu instead of silent skipping
to avoid any failure on Xe2+igpu as xe_device_has_flat_ccs()
can sometimes be false on igpu due some setting in the BIOS
to turn off compression on igpu.
- Nits
v5: (Matt)
- Correct the condition check of xe_pat_index_get_comp_en
v4: (Matt)
- Introduce xe_pat_index_get_comp_en(), which checks
XE2_COMP_EN for the pat_index
- .interruptible should be true, everything else false
v3: (Matt)
- s/xe_bo_schedule_decompress/xe_bo_decompress
- skip the decrompress step if the BO isn't in VRAM
- start/size not required in xe_bo_schedule_decompress
- Use xe_bo_move_notify instead of xe_vm_invalidate_vma
with respect to invalidation.
- Nits
v2:
- Move decompression work out of vm_bind ioctl. (Matt)
- Put that work in a small helper at the BO/migrate layer invoke it
from vma_lock_and_validate which already runs under drm_exec.
- Move lightweight checks to vm_bind_ioctl_check_args (Matthew Auld)
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Acked-by: Michal Mrozek <michal.mrozek@intel.com>
Signed-off-by: Nitin Gote <nitin.r.gote@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patch.msgid.link/20260304123758.3050386-8-nitin.r.gote@intel.com
Introduce an internal __xe_migrate_copy(..., is_vram_resolve) path and
expose a small wrapper xe_migrate_resolve() that calls it with
is_vram_resolve=true.
For resolve/decompression operations we must ensure the copy code uses
the compression PAT index when appropriate; this change centralizes that
behavior and allows callers to schedule a resolve (decompress) operation
via the migrate API.
v3: Fix kernel-doc warnings
v2: (Matt)
- Simplify xe_migrate_resolve(), use single BO/resource;
remove copy_only_ccs argument as it's always false.
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Nitin Gote <nitin.r.gote@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patch.msgid.link/20260304123758.3050386-7-nitin.r.gote@intel.com
Add a new VM_BIND flag, DRM_XE_VM_BIND_FLAG_DECOMPRESS, that lets userspace
express intent for the driver to perform on-device in-place decompression
for the GPU mapping created by a MAP bind operation.
This flag is used by subsequent driver changes to trigger scheduling of
GPU work that resolves compressed VRAM pages into an uncompressed PAT
VM mapping.
Behavior and semantics:
- Valid only for DRM_XE_VM_BIND_OP_MAP. IOCTLs using this flag on other ops
are rejected (-EINVAL).
- The bind's pat_index must select the device "no-compression" PAT entry;
otherwise the ioctl is rejected (-EINVAL).
- Only meaningful for VRAM-backed BOs on devices that support Flat CCS and
the required hardware generation (driver will return -EOPNOTSUPP if not).
- On success the driver schedules a migrate/resolve and installs the
returned dma_fence into the BO's kernel reservation
(DMA_RESV_USAGE_KERNEL).
Compute PR: https://github.com/intel/compute-runtime/pull/898
v3: Rebase on latest drm-tip and add compute pr info
v2: Add kernel doc (Matt)
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Mrozek, Michal <michal.mrozek@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Nitin Gote <nitin.r.gote@intel.com>
Acked-by: Michal Mrozek <michal.mrozek@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patch.msgid.link/20260304123758.3050386-6-nitin.r.gote@intel.com
The AMD PMF driver provides realtime column utilization (npu_busy)
metrics for the NPU. Extend the DRM_IOCTL_AMDXDNA_GET_INFO sensor
query to expose these metrics to userspace.
Add AMDXDNA_SENSOR_TYPE_COLUMN_UTILIZATION to the sensor type enum
and update aie2_get_sensors() to return both the total power and up
to 8 column utilization sensors if the user buffer permits.
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
[lizhi: support legacy tool which uses small buffer. checkpatch cleanup]
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260311171842.473453-1-lizhi.hou@amd.com
The AMD PMF driver provides an interface to obtain realtime power
estimates for the NPU. Expose this information to userspace through a
new DRM_IOCTL_AMDXDNA_GET_INFO parameter, allowing applications to query
the current NPU power level.
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
(Update comment to indicate power and utilization)
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260228061109.361239-2-superm1@kernel.org
The amdxdna driver uses amd_pmf_get_npu_data() which is exported in the
AMD_PMF namespace. Import the AMD_PMF namespace.
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260301005028.367618-1-superm1@kernel.org
Avoid code duplication by extracting the logic for selection of the
correct PAT_PTA entry for a GT into function gt_pta_entry() and using
that function whenever necessary.
Reviewed-by: Dnyaneshwar Bhadane <dnyaneshwar.bhadane@intel.com>
Link: https://patch.msgid.link/20260303-pat-gt_pta_entry-v1-1-0dee8e1e7bd9@intel.com
Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
It turned out that protecting the status of each bo_va with a
spinlock was just hiding problems instead of solving them.
Revert the whole approach, add a separate stats_lock and lockdep
assertions that the correct reservation lock is held all over the place.
This not only allows for better checks if a state transition is properly
protected by a lock, but also switching back to using list macros to
iterate over the state of lists protected by the dma_resv lock of the
root PD.
v2: re-add missing check
v3: split into two patches
v4: re-apply by fixing holding the VM lock at the right places.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If discovery has failed for any reason (such as no support for a block)
then there is no need to unwind all the IP blocks in fini. In this
condition there can actually be failures during the unwind too.
Reset num_ip_blocks to zero during failure path and skip the unnecessary
cleanup path.
Suggested-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fix XGMI max bitrate/width reporting on SMUv13.0.12 SOCs. The data
format got changed when moved to static table from dynamic metrics
table.
Fixes: 1bec2f2707 ("drm/amd/pm: Fetch SMUv13.0.12 xgmi max speed/width")
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Asad Kamal <asad.kamal@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The mes.fw[] is per-pipe resource shared accross xcc inst.
And enlarge hung_queue array to max inst_pipes.
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
