drm/xe: Prevent VFs from exposing the CCS mode sysfs file

Skip creating CCS sysfs files in VF mode to ensure VFs do not
try to change CCS mode, as it is predefined and immutable in
the SR-IOV mode.

Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Nareshkumar Gollakoti <naresh.kumar.g@intel.com>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patch.msgid.link/20260202170810.1393147-5-naresh.kumar.g@intel.com
(cherry picked from commit 4e8f602ac3574cf1ebc7acfb6624d06e04b30c91)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/hwmon: Prevent unintended VRAM channel creation

Remove the unnecessary VRAM channel entry introduced in xe_hwmon_channel.
Without this, adding any new hwmon channel causes extra VRAM channel
to appear. This remained unnoticed earlier because VRAM was the
final xe hwmon channel.

v2: Use MAX_VRAM_CHANNELS with in_range() instead of
    CHANNEL_VRAM_N_MAX. (Raag)

Fixes: 49a498338417 ("drm/xe/hwmon: Expose individual VRAM channel temperature")
Signed-off-by: Karthik Poosa <karthik.poosa@intel.com>
Reviewed-by: Raag Jadav <raag.jadav@intel.com>
Link: https://patch.msgid.link/20260206081655.2115439-1-karthik.poosa@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
(cherry picked from commit 48eb073c7d95883eca2789447f94e1e8cafbabe5)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/pagemap: pass pagemap_addr by reference

Passing a structure by value into a function is sometimes problematic,
for a number of reasons. Of of these is a warning from the 32-bit arm
compiler:

drivers/gpu/drm/drm_gpusvm.c: In function '__drm_gpusvm_unmap_pages':
drivers/gpu/drm/drm_gpusvm.c:1152:33: note: parameter passing for argument of type 'struct drm_pagemap_addr' changed in GCC 9.1
 1152 |                                 dpagemap->ops->device_unmap(dpagemap,
      |                                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 1153 |                                                             dev, *addr);
      |                                                             ~~~~~~~~~~~

This particular problem is harmless since we are not mixing compiler versions
inside of the compiler. However, passing this by reference avoids the warning
along with providing slightly better calling conventions as it avoids an
extra copy on the stack.

Fixes: 75af93b3f5d0 ("drm/pagemap, drm/xe: Support destination migration over interconnect")
Fixes: 2df55d9e66a2 ("drm/xe: Support pcie p2p dma as a fast interconnect")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patch.msgid.link/20260216134644.1025365-1-arnd@kernel.org
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
(cherry picked from commit 95162db0208aee122d10ac1342fe97a1721cd258)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/bo: Redirect faults to dummy page for wedged device

As per uapi documentation[1], the prerequisite for wedged device is to
redirected page faults to a dummy page. Follow it.

[1] Documentation/gpu/drm-uapi.rst

v2: Add uapi reference and fixes tag (Matthew Brost)

Fixes: 7bc00751f877 ("drm/xe: Use device wedged event")
Signed-off-by: Raag Jadav <raag.jadav@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260212055622.2054991-1-raag.jadav@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit c020fff70d757612933711dd3cc3751d7d782d3c)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Make xe_modparam.force_vram_bar_size signed

vram_bar_size is registered as an int module parameter and is documented
to accept negative values to disable BAR resizing.
Store it as an int in xe_modparam as well, so negative values work as
intended and the module_param type matches.

Fixes: 80742a1aa26e ("drm/xe: Allow to drop vram resizing")
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Link: https://patch.msgid.link/20260202181853.1095736-2-shuicheng.lin@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit 25c9aa4dcb5ef2ad9f354d19f8f1eeb690d1c161)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/vf: Avoid reading media version when media GT is disabled

When the media GT is not allowed, a VF must not attempt to read
the media version from the GuC. The GuC may not be loaded, and
any attempt to communicate with it would result in a timeout
and a VF probe failure:

(...)
[ 1912.406046] xe 0000:01:00.1: [drm] *ERROR* Tile0: GT1: GuC mmio request 0x5507: no reply 0x5507
[ 1912.407277] xe 0000:01:00.1: [drm] *ERROR* Tile0: GT1: [GUC COMMUNICATION] MMIO send failed (-ETIMEDOUT)
[ 1912.408689] xe 0000:01:00.1: [drm] *ERROR* VF: Tile0: GT1: Failed to reset GuC state (-ETIMEDOUT)
[ 1912.413986] xe 0000:01:00.1: probe with driver xe failed with error -110

Let's skip reading the media version for VFs when the media GT is not
allowed.

v2: move the condition directly to the VF path

Fixes: 7abd69278bb5 ("drm/xe/configfs: Add attribute to disable GT types")
Signed-off-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Shuicheng Lin <shuicheng.lin@intel.com>
Link: https://patch.msgid.link/20260202115041.2863357-1-piotr.piorkowski@intel.com
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
(cherry picked from commit 0bcacf56dc0b265f9c47056c6a4f0c1394a8a3f0)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/xe2_hpg: Fix handling of Wa_14019988906 & Wa_14019877138

The PSS_CHICKEN register has been part of the RCS engine's LRC since it
was first introduced in Xe_LP.  That means that any workarounds that
adjust its value (such as Wa_14019988906 and Wa_14019877138) need to be
implemented in the lrc_was[] table so that they become part of the
default LRC from which all subsequent LRCs are copied.  Although these
workarounds were implemented correctly on most platforms, they were
incorrectly placed on the engine_was[] table for Xe2_HPG.

Move the workarounds to the proper lrc_was[] table and switch the
'xe_rtp_match_first_render_or_compute' rule to specifically match the
RCS since that's the engine whose LRC manages the register.

Bspec: 65182
Fixes: 7f3ee7d88058 ("drm/xe/xe2hpg: Add initial GT workarounds")
Reviewed-by: Shekhar Chauhan <shekhar.chauhan@intel.com>
Link: https://patch.msgid.link/20260205220508.51905-2-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit e04c609eedf4d6748ac0bcada4de1275b034fed6)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/mmio: Avoid double-adjust in 64-bit reads

xe_mmio_read64_2x32() was adjusting register addresses and then
calling xe_mmio_read32(), which applies the adjustment again.
This may shift accesses twice if adj_offset < adj_limit. There is
no issue currently, as for media gt, adj_offset > adj_limit, so
the 2nd adjust will be a no-op. But it may not work in future.

To fix it, replace the adjusted-address comparison with a direct
sanity check that ensures the MMIO address adjustment cutoff never
falls within the 8-byte range of a 64-bit register. And let
xe_mmio_read32() handle address translation.

v2: rewrite the sanity check in a more natural way. (Matt)
v3: Add Fixes tag. (Jani)

Fixes: 07431945d8ae ("drm/xe: Avoid 64-bit register reads")
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Link: https://patch.msgid.link/20260130165621.471408-2-shuicheng.lin@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit a30f999681126b128a43137793ac84b6a5b7443f)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Add bounds check on pat_index to prevent OOB kernel read in madvise

When user provides a bogus pat_index value through the madvise IOCTL, the
xe_pat_index_get_coh_mode() function performs an array access without
validating bounds. This allows a malicious user to trigger an out-of-bounds
kernel read from the xe->pat.table array.

The vulnerability exists because the validation in madvise_args_are_sane()
directly calls xe_pat_index_get_coh_mode(xe, args->pat_index.val) without
first checking if pat_index is within [0, xe->pat.n_entries).

Although xe_pat_index_get_coh_mode() has a WARN_ON to catch this in debug
builds, it still performs the unsafe array access in production kernels.

v2(Matthew Auld)
- Using array_index_nospec() to mitigate spectre attacks when the value
is used

v3(Matthew Auld)
- Put the declarations at the start of the block

Fixes: ada7486c5668 ("drm/xe: Implement madvise ioctl for xe")
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Cc: <stable@vger.kernel.org> # v6.18+
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Shuicheng Lin <shuicheng.lin@intel.com>
Cc: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Cc: "Thomas Hellström" <thomas.hellstrom@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Jia Yao <jia.yao@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patch.msgid.link/20260205161529.1819276-1-jia.yao@intel.com
(cherry picked from commit 944a3329b05510d55c69c2ef455136e2fc02de29)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/configfs: Fix 'parameter name omitted' errors

On some configs and old compilers we can get following build errors:

  ../drivers/gpu/drm/xe/xe_configfs.h: In function 'xe_configfs_get_ctx_restore_mid_bb':
  ../drivers/gpu/drm/xe/xe_configfs.h:40:76: error: parameter name omitted
   static inline u32 xe_configfs_get_ctx_restore_mid_bb(struct pci_dev *pdev, enum xe_engine_class,
                                                                            ^~~~~~~~~~~~~~~~~~~~
  ../drivers/gpu/drm/xe/xe_configfs.h: In function 'xe_configfs_get_ctx_restore_post_bb':
  ../drivers/gpu/drm/xe/xe_configfs.h:42:77: error: parameter name omitted
   static inline u32 xe_configfs_get_ctx_restore_post_bb(struct pci_dev *pdev, enum xe_engine_class,
                                                                             ^~~~~~~~~~~~~~~~~~~~
when trying to define our configfs stub functions. Fix that.

Fixes: 7a4756b2fd04 ("drm/xe/lrc: Allow to add user commands mid context switch")
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Shuicheng Lin <shuicheng.lin@intel.com>
Link: https://patch.msgid.link/20260203193745.576-1-michal.wajdeczko@intel.com
(cherry picked from commit f59cde8a2452b392115d2af8f1143a94725f4827)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/pf: Fix sysfs initialization

In case of devm_add_action_or_reset() failure the provided cleanup
action will be run immediately on the not yet initialized kobject.
This may lead to errors like:

 [ ] kobject: '(null)' (ff110001393608e0): is not initialized, yet kobject_put() is being called.
 [ ] WARNING: lib/kobject.c:734 at kobject_put+0xd9/0x250, CPU#0: kworker/0:0/9
 [ ] RIP: 0010:kobject_put+0xdf/0x250
 [ ] Call Trace:
 [ ]  xe_sriov_pf_sysfs_init+0x21/0x100 [xe]
 [ ]  xe_sriov_pf_init_late+0x87/0x2b0 [xe]
 [ ]  xe_sriov_init_late+0x5f/0x2c0 [xe]
 [ ]  xe_device_probe+0x5f2/0xc20 [xe]
 [ ]  xe_pci_probe+0x396/0x610 [xe]
 [ ]  local_pci_probe+0x47/0xb0

 [ ] refcount_t: underflow; use-after-free.
 [ ] WARNING: lib/refcount.c:28 at refcount_warn_saturate+0x68/0xb0, CPU#0: kworker/0:0/9
 [ ] RIP: 0010:refcount_warn_saturate+0x68/0xb0
 [ ] Call Trace:
 [ ]  kobject_put+0x174/0x250
 [ ]  xe_sriov_pf_sysfs_init+0x21/0x100 [xe]
 [ ]  xe_sriov_pf_init_late+0x87/0x2b0 [xe]
 [ ]  xe_sriov_init_late+0x5f/0x2c0 [xe]
 [ ]  xe_device_probe+0x5f2/0xc20 [xe]
 [ ]  xe_pci_probe+0x396/0x610 [xe]
 [ ]  local_pci_probe+0x47/0xb0

Fix that by calling kobject_init() and kobject_add() separately
and register cleanup action after the kobject is initialized.

Also make this cleanup registration a part of the create helper to
fix another mistake, as in the loop we were wrongly passing parent
kobject while registering cleanup action, and this resulted in some
undetected leaks.

Fixes: 5c170a4d9c53 ("drm/xe/pf: Prepare sysfs for SR-IOV admin attributes")
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Shuicheng Lin <shuicheng.lin@intel.com>
Link: https://patch.msgid.link/20260203235332.1350-1-michal.wajdeczko@intel.com
(cherry picked from commit 98b16727f07e26a5d4de84d88805ce7ffcfdd324)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

Merge tag 'drm-xe-next-fixes-2026-02-05' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-next

- Fix CFI violation in debugfs access (Daniele)
- Kernel-doc fixes (Chaitanya, Shuicheng)
- Disable D3Cold for BMG only on specific platforms (Karthik)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patch.msgid.link/aYStaLZVJWwKCDZt@intel.com

Merge tag 'drm-misc-next-fixes-2026-02-05' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-next

Several fixes for amdxdna around PM handling, error reporting and
memory safety, a compilation fix for ilitek-ili9882t, a NULL pointer
dereference fix for imx8qxp-pixel-combiner and several PTE fixes for
nouveau

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@redhat.com>
Link: https://patch.msgid.link/20260205-refreshing-natural-vole-4c73af@houat

Merge tag 'drm-intel-next-fixes-2026-02-05' of https://gitlab.freedesktop.org/drm/i915/kernel into drm-next

- Fix the pixel normalization handling for xe3p_lpd display

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patch.msgid.link/aYROngKfyUIyoQW0@jlahtine-mobl

drm/xe/pm: Disable D3Cold for BMG only on specific platforms

Restrict D3Cold disablement for BMG to unsupported NUC platforms,
instead of disabling it on all platforms.

Signed-off-by: Karthik Poosa <karthik.poosa@intel.com>
Fixes: 3e331a6715ee ("drm/xe/pm: Temporarily disable D3Cold on BMG")
Link: https://patch.msgid.link/20260123173238.1642383-1-karthik.poosa@intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
(cherry picked from commit 39125eaf8863ab09d70c4b493f58639b08d5a897)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Fix kerneldoc for xe_tlb_inval_job_alloc_dep

Correct the function name in the kerneldoc.
It is for below warning:
"Warning: drivers/gpu/drm/xe/xe_tlb_inval_job.c:210 expecting prototype for
xe_tlb_inval_alloc_dep(). Prototype was for xe_tlb_inval_job_alloc_dep()
instead"

Fixes: 15366239e2130 ("drm/xe: Decouple TLB invalidations from GT")
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patch.msgid.link/20260129233834.419977-8-shuicheng.lin@intel.com
(cherry picked from commit 9f9c117ac566cb567dd56cc5b7564c45653f7a2a)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Fix kerneldoc for xe_gt_tlb_inval_init_early

Correct the function name in the kerneldoc.
It is for below warning:
"Warning: drivers/gpu/drm/xe/xe_tlb_inval.c:136 expecting prototype for
xe_gt_tlb_inval_init(). Prototype was for xe_gt_tlb_inval_init_early()
instead"

v2: add () for the function. (Michal)

Fixes: db16f9d90c1d9 ("drm/xe: Split TLB invalidation code in frontend and backend")
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patch.msgid.link/20260129233834.419977-7-shuicheng.lin@intel.com
(cherry picked from commit 0651dbb9d6a72e99569576fbec4681fd8160d161)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Fix kerneldoc for xe_migrate_exec_queue

Correct the function name in the kerneldoc.
It is for below warning:
"Warning: drivers/gpu/drm/xe/xe_migrate.c:1262 expecting prototype for
xe_get_migrate_exec_queue(). Prototype was for xe_migrate_exec_queue()
instead"

Fixes: 916ee4704a865 ("drm/xe/vf: Register CCS read/write contexts with Guc")
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patch.msgid.link/20260129233834.419977-6-shuicheng.lin@intel.com
(cherry picked from commit 9fd8da717934f05125b9ba6782622c459a368dc0)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/query: Fix topology query pointer advance

The topology query helper advanced the user pointer by the size
of the pointer, not the size of the structure. This can misalign
the output blob and corrupt the following mask. Fix the increment
to use sizeof(*topo).
There is no issue currently, as sizeof(*topo) happens to be equal
to sizeof(topo) on 64-bit systems (both evaluate to 8 bytes).

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patch.msgid.link/20260130043907.465128-2-shuicheng.lin@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit c2a6859138e7f73ad904be17dd7d1da6cc7f06b3)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/guc: Fix kernel-doc warning in GuC scheduler ABI header

The GuC scheduler ABI header contains a file-level comment that is not
intended to document a kernel-doc symbol. Using kernel-doc comment
syntax (/** */) triggers kernel-doc warnings.

With "-Werror", this causes the build to fail. Convert the comment to a
regular block comment.

HDRTEST drivers/gpu/drm/xe/abi/guc_scheduler_abi.h
Warning: drivers/gpu/drm/xe/abi/guc_scheduler_abi.h:11 This comment starts with '/**', but isn't a kernel-doc comment. Refer to Documentation/doc-guide/kernel-doc.rst
 * Generic defines required for registration with and submissions to the GuC
1 warnings as errors
make[6]: *** [drivers/gpu/drm/xe/Makefile:377: drivers/gpu/drm/xe/abi/guc_scheduler_abi.hdrtest] Error 3
make[5]: *** [scripts/Makefile.build:544: drivers/gpu/drm/xe] Error 2
make[4]: *** [scripts/Makefile.build:544: drivers/gpu/drm] Error 2
make[3]: *** [scripts/Makefile.build:544: drivers/gpu] Error 2
make[2]: *** [scripts/Makefile.build:544: drivers] Error 2
make[1]: *** [/home/kbuild2/kernel/Makefile:2088: .] Error 2
make: *** [Makefile:248: __sub-make] Error 2

v2:
 - Add Fixes tag (Daniele)

Fixes: b0c5cf4f5917 ("drm/gt/guc: extract scheduler-related defines from guc_fwif.h")
Signed-off-by: Chaitanya Kumar Borah <chaitanya.kumar.borah@intel.com>
Reviewed-by: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Link: https://patch.msgid.link/20260130135210.2659200-1-chaitanya.kumar.borah@intel.com
(cherry picked from commit f89dbe14a0c8854b7aaf960dd842c10698b3ff19)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/guc: Fix CFI violation in debugfs access.

xe_guc_print_info is void-returning, but the function pointer it is
assigned to expects an int-returning function, leading to the following
CFI error:

[  206.873690] CFI failure at guc_debugfs_show+0xa1/0xf0 [xe]
(target: xe_guc_print_info+0x0/0x370 [xe]; expected type: 0xbe3bc66a)

Fix this by updating xe_guc_print_info to return an integer.

Fixes: e15826bb3c2c ("drm/xe/guc: Refactor GuC debugfs initialization")
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: George D Sworo <george.d.sworo@intel.com>
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patch.msgid.link/20260129182547.32899-2-daniele.ceraolospurio@intel.com
(cherry picked from commit dd8ea2f2ab71b98887fdc426b0651dbb1d1ea760)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

accel/amdxdna: Move RPM resume into job run function

Currently, amdxdna_pm_resume_get() is called during job creation, and
amdxdna_pm_suspend_put() is called when the hardware notifies job
completion. If a job is canceled before it is run, no hardware
completion notification is generated, resulting in an unbalanced
runtime PM resume/suspend pair.

Fix this by moving amdxdna_pm_resume_get() to the job run path, ensuring
runtime PM is only resumed for jobs that are actually executed.

Fixes: 063db451832b ("accel/amdxdna: Enhance runtime power management")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260204171118.3165607-1-lizhi.hou@amd.com

accel/amdxdna: Fix incorrect DPM level after suspend/resume

The suspend routine sets the DPM level to 0, which unintentionally
overwrites the previously saved DPM level. As a result, the device always
resumes with DPM level 0 instead of restoring the original value.

Fix this by ensuring the suspend path does not overwrite the saved DPM
level, allowing the correct DPM level to be restored during resume.

Fixes: f4d7b8a6bc8c ("accel/amdxdna: Enhance power management settings")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260204171048.3165580-1-lizhi.hou@amd.com

nouveau/vmm: start tracking if the LPT PTE is valid. (v6)

When NVK enabled large pages userspace tests were seeing fault
reports at a valid address.

There was a case where an address moving from 64k page to 4k pages
could expose a race between unmapping the 4k page, mapping the 64k
page and unref the 4k pages.

Unref 4k pages would cause the dual-page table handling to always
set the LPTE entry to SPARSE or INVALID, but if we'd mapped a valid
LPTE in the meantime, it would get trashed. Keep track of when
a valid LPTE has been referenced, and don't reset in that case.

This adds an lpte valid tracker and lpte reference count.

Whenever an lpte is referenced, it gets made valid and the ref count
increases, whenever it gets unreference the refcount is tracked.

Link: https://gitlab.freedesktop.org/mesa/mesa/-/issues/14610
Reviewed-by: Mary Guillemard <mary@mary.zone>
Tested-by: Mary Guillemard <mary@mary.zone>
Tested-by: Mel Henning <mhenning@darkrefraction.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patch.msgid.link/20260204030208.2313241-4-airlied@gmail.com

nouveau/vmm: increase size of vmm pte tracker struct to u32 (v2)

We need to tracker large counts of spte than previously due to unref
getting delayed sometimes.

This doesn't fix LPT tracking yet, it just creates space for it.

Reviewed-by: Mary Guillemard <mary@mary.zone>
Tested-by: Mary Guillemard <mary@mary.zone>
Tested-by: Mel Henning <mhenning@darkrefraction.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patch.msgid.link/20260204030208.2313241-3-airlied@gmail.com

nouveau/vmm: rewrite pte tracker using a struct and bitfields.

I want to increase the counters here and start tracking LPTs as well
as there are certain situations where userspace with mixed page sizes
can cause ref/unrefs to live longer so need better reference counting.

This should be entirely non-functional.

Reviewed-by: Mary Guillemard <mary@mary.zone>
Tested-by: Mary Guillemard <mary@mary.zone>
Tested-by: Mel Henning <mhenning@darkrefraction.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patch.msgid.link/20260204030208.2313241-2-airlied@gmail.com

accel/amdxdna: Fix incorrect error code returned for failed chain command

The driver currently returns an incorrect error code when a chain command
fails. In this case, ERT_CMD_STATE_ERROR is expected to be reported for
failed chain commands.

Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Maciej Falkowski <maciej.falkowski@linux.intel.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260203184037.2751889-1-lizhi.hou@amd.com

accel/amdxdna: Remove hardware context status

One newly supported command does not require hardware context configuration
to be performed upfront. As a result, checking hardware context status
causes this command to fail incorrectly.

Remove hardware context status handling entirely. For other commands,
if userspace submits a request without configuring the hardware context
first, the firmware will report an error or time out as appropriate.

Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260202212450.2681273-1-lizhi.hou@amd.com

drm/bridge: imx8qxp-pixel-combiner: Fix bailout for imx8qxp_pc_bridge_probe()

In case the channel0 is unavailable and bailing out from free_child is
needed when we fail to add a DRM bridge for the available channel1,
pointer pc->ch[0] in the bailout path would be NULL and it would be
dereferenced as pc->ch[0]->bridge.next_bridge.  Fix this by checking
pc->ch[0] before dereferencing it.

Fixes: ae754f049ce1 ("drm/bridge: imx8qxp-pixel-combiner: get/put the next bridge")
Fixes: 99764593528f ("drm/bridge: imx8qxp-pixel-combiner: convert to devm_drm_bridge_alloc() API")
Signed-off-by: Liu Ying <victor.liu@nxp.com>
Reviewed-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20260123-imx8qxp-drm-bridge-fixes-v1-3-8bb85ada5866@nxp.com
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>

drm/panel: ilitek-ili9882t: Remove duplicate initializers in tianma_il79900a_dsc

Clang warns (or errors with CONFIG_WERROR=y / W=e):

  drivers/gpu/drm/panel/panel-ilitek-ili9882t.c:95:16: error: initializer overrides prior initialization of this subobject [-Werror,-Winitializer-overrides]
     95 |         .vbr_enable = 0,
        |                       ^
  drivers/gpu/drm/panel/panel-ilitek-ili9882t.c:90:16: note: previous initialization is here
     90 |         .vbr_enable = false,
        |                       ^~~~~
  drivers/gpu/drm/panel/panel-ilitek-ili9882t.c:97:19: error: initializer overrides prior initialization of this subobject [-Werror,-Winitializer-overrides]
     97 |         .rc_model_size = DSC_RC_MODEL_SIZE_CONST,
        |                          ^~~~~~~~~~~~~~~~~~~~~~~
  include/drm/display/drm_dsc.h:22:38: note: expanded from macro 'DSC_RC_MODEL_SIZE_CONST'
     22 | #define DSC_RC_MODEL_SIZE_CONST             8192
        |                                             ^~~~
  drivers/gpu/drm/panel/panel-ilitek-ili9882t.c:91:19: note: previous initialization is here
     91 |         .rc_model_size = DSC_RC_MODEL_SIZE_CONST,
        |                          ^~~~~~~~~~~~~~~~~~~~~~~
  include/drm/display/drm_dsc.h:22:38: note: expanded from macro 'DSC_RC_MODEL_SIZE_CONST'
     22 | #define DSC_RC_MODEL_SIZE_CONST             8192
        |                                             ^~~~
  drivers/gpu/drm/panel/panel-ilitek-ili9882t.c:132:25: error: initializer overrides prior initialization of this subobject [-Werror,-Winitializer-overrides]
    132 |         .initial_scale_value = 32,
        |                                ^~
  drivers/gpu/drm/panel/panel-ilitek-ili9882t.c:126:25: note: previous initialization is here
    126 |         .initial_scale_value = 32,
        |                                ^~
  drivers/gpu/drm/panel/panel-ilitek-ili9882t.c:133:20: error: initializer overrides prior initialization of this subobject [-Werror,-Winitializer-overrides]
    133 |         .nfl_bpg_offset = 3511,
        |                           ^~~~
  drivers/gpu/drm/panel/panel-ilitek-ili9882t.c:108:20: note: previous initialization is here
    108 |         .nfl_bpg_offset = 1402,
        |                           ^~~~

GCC would warn about this in the same manner but its version,
-Woverride-init, is disabled for a normal kernel build in
scripts/Makefile.warn. For clang, -Wextra in drivers/gpu/drm/Makefile
turns it back but GCC respects turning it off earlier in the command
line.

Of all the duplicate fields in the initializer, only nfl_bpg_offset is a
different value. Clear up the duplicate initializers, keeping the
'false' value for .vbr_enable, as it is bool, and the second value for
.nfl_bpg_offset, assuming it is the correct one since it was the one
tested in the original change.

Fixes: 65ce1f5834e9 ("drm/panel: ilitek-ili9882t: Switch Tianma TL121BVMS07 to DSC 120Hz mode")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Link: https://patch.msgid.link/20260114-panel-ilitek-ili9882t-fix-override-init-v1-1-1d69a2b096df@kernel.org
Signed-off-by: Maxime Ripard <mripard@kernel.org>

drm/i915/display: fix the pixel normalization handling for xe3p_lpd

Pixel normalizer is enabled with normalization factor as 1.0 for
FP16 formats in order to support FBC for those formats in xe3p_lpd.
Previously pixel normalizer gets disabled during the plane disable
routine. But there could be plane format settings without explicitly
calling the plane disable in-between and we could endup keeping the
pixel normalizer enabled for formats which we don't require that.
This is causing crc mismatches in yuv formats and FIFO underruns in
planar formats like NV12. Fix this by updating the pixel normalizer
configuration based on the pixel formats explicitly during the plane
settings arm calls itself - enable it for FP16 and disable it for
other formats in HDR capable planes.

v2: avoid redundant pixel normalization setting updates

v3: moved the normalization factor definition to intel_fbc.c and some
    updates to comments

v4: simplified the pixel normalizer setting handling

Fixes: 5298eea7ed20 ("drm/i915/xe3p_lpd: use pixel normalizer for fp16 formats for FBC")
Signed-off-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Link: https://patch.msgid.link/20260130095919.107805-1-vinod.govindapillai@intel.com
(cherry picked from commit c0dc68f4e2aa7eddb9ec6d95931f9576d8fe7334)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Merge tag 'exynos-drm-next-for-v6.20' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-next

Fix three regressions
. Fix a regression where vidi_connection_ioctl() used the wrong device
  to look up the vidi context. It stores the vidi device in exynos_drm_private
  and uses it in ioctl(), preventing invalid pointer access and related bugs.
. Fix a security regression where vidi_connection_ioctl() directly dereferenced
  a user pointer for EDID data. It copies EDID from user space
  with copy_from_user() into kernel memory before use, preventing arbitrary
  kernel memory access.
. Fix a concurrency regression where vidi_context members related
  to EDID memory were accessed without locking. It protects alloc/free and
  state updates with ctx->lock, preventing race conditions and use-after-free bugs.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Inki Dae <inki.dae@samsung.com>
Link: https://patch.msgid.link/20260201143939.27074-1-inki.dae@samsung.com

Merge tag 'amd-drm-next-6.20-2026-01-30' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.20-2026-01-30:

amdgpu:
- Misc cleanups
- SMU 13 fixes
- SMU 14 fixes
- GPUVM fault filter fix
- USB4 fixes
- DC FP guard fixes
- Powergating fix
- JPEG ring reset fix
- RAS fixes
- Xclk fix for soc21 APUs
- Fix COND_EXEC handling for GC 11
- UserQ fixes
- MQD size alignment fixes
- SMU feature interface cleanup
- GC 10-12 KGQ init fixes
- GC 11-12 KGQ reset fixes

amdkfd:
- Fix device snapshot reporting
- GC 12.1 trap handler fixes
- MQD size alignment fixes

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patch.msgid.link/20260130183257.28879-1-alexander.deucher@amd.com

drm/exynos: vidi: use ctx->lock to protect struct vidi_context member variables related to memory alloc/free

Exynos Virtual Display driver performs memory alloc/free operations
without lock protection, which easily causes concurrency problem.

For example, use-after-free can occur in race scenario like this:
```
CPU0 CPU1 CPU2
---- ---- ----
  vidi_connection_ioctl()
    if (vidi->connection) // true
      drm_edid = drm_edid_alloc(); // alloc drm_edid
      ...
      ctx->raw_edid = drm_edid;
      ...
drm_mode_getconnector()
  drm_helper_probe_single_connector_modes()
    vidi_get_modes()
      if (ctx->raw_edid) // true
        drm_edid_dup(ctx->raw_edid);
          if (!drm_edid) // false
          ...
vidi_connection_ioctl()
  if (vidi->connection) // false
    drm_edid_free(ctx->raw_edid); // free drm_edid
    ...
          drm_edid_alloc(drm_edid->edid)
            kmemdup(edid); // UAF!!
            ...
```

To prevent these vulns, at least in vidi_context, member variables related
to memory alloc/free should be protected with ctx->lock.

Cc: <stable@vger.kernel.org>
Signed-off-by: Jeongjun Park <aha310510@gmail.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: vidi: fix to avoid directly dereferencing user pointer

In vidi_connection_ioctl(), vidi->edid(user pointer) is directly
dereferenced in the kernel.

This allows arbitrary kernel memory access from the user space, so instead
of directly accessing the user pointer in the kernel, we should modify it
to copy edid to kernel memory using copy_from_user() and use it.

Cc: <stable@vger.kernel.org>
Signed-off-by: Jeongjun Park <aha310510@gmail.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

drm/exynos: vidi: use priv->vidi_dev for ctx lookup in vidi_connection_ioctl()

vidi_connection_ioctl() retrieves the driver_data from drm_dev->dev to
obtain a struct vidi_context pointer. However, drm_dev->dev is the
exynos-drm master device, and the driver_data contained therein is not
the vidi component device, but a completely different device.

This can lead to various bugs, ranging from null pointer dereferences and
garbage value accesses to, in unlucky cases, out-of-bounds errors,
use-after-free errors, and more.

To resolve this issue, we need to store/delete the vidi device pointer in
exynos_drm_private->vidi_dev during bind/unbind, and then read this
exynos_drm_private->vidi_dev within ioctl() to obtain the correct
struct vidi_context pointer.

Cc: <stable@vger.kernel.org>
Signed-off-by: Jeongjun Park <aha310510@gmail.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>

accel/amdxdna: Fix memory leak in amdxdna_ubuf_map

The amdxdna_ubuf_map() function allocates memory for sg and
internal sg table structures, but it fails to free them if subsequent
operations (sg_alloc_table_from_pages or dma_map_sgtable) fail.

Fixes: bd72d4acda10 ("accel/amdxdna: Support user space allocated buffer")
Signed-off-by: Zishun Yi <zishun.yi.dev@gmail.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Reviewed-by: Min Ma <mamin506@gmail.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260129171022.68578-1-zishun.yi.dev@gmail.com

accel/amdxdna: Stop job scheduling across aie2_release_resource()

Running jobs on a hardware context while it is in the process of
releasing resources can lead to use-after-free and crashes.

Fix this by stopping job scheduling before calling
aie2_release_resource() and restarting it after the release completes.
Additionally, aie2_sched_job_run() now checks whether the hardware
context is still active.

Fixes: 4fd6ca90fc7f ("accel/amdxdna: Refactor hardware context destroy routine")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260130003255.2083255-1-lizhi.hou@amd.com

accel/amdxdna: Hold mm structure across iommu_sva_unbind_device()

Some tests trigger a crash in iommu_sva_unbind_device() due to
accessing iommu_mm after the associated mm structure has been
freed.

Fix this by taking an explicit reference to the mm structure
after successfully binding the device, and releasing it only
after the device is unbound. This ensures the mm remains valid
for the entire SVA bind/unbind lifetime.

Fixes: be462c97b7df ("accel/amdxdna: Add hardware context")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260128002356.1858122-1-lizhi.hou@amd.com

Merge tag 'drm-xe-next-fixes-2026-01-29' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-next

- Reduce LRC timestamp stuck message on VFs to notice (Brost)
- Disable GuC Power DCC strategy on PTL (Vinay)
- Unregister drm device on probe error (Lin)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patch.msgid.link/aXuyrtsnlAOmj_OB@intel.com

Merge tag 'drm-misc-next-fixes-2026-01-29' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-next

Two fixes for NULL pointer dereference in imx8 following the bridge
refcounting conversions, and one for the bridge connector following the
HDMI audio reworks.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@redhat.com>
Link: https://patch.msgid.link/20260129-efficient-jerboa-of-ecstasy-822832@houat

Merge tag 'drm-intel-next-fixes-2026-01-29' of https://gitlab.freedesktop.org/drm/i915/kernel into drm-next

- Prevent u64 underflow in intel_fbc_stolen_end

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patch.msgid.link/aXsWGWjacEJ03rTs@jlahtine-mobl

drm/amdgpu/gfx12: adjust KGQ reset sequence

Kernel gfx queues do not need to be reinitialized or
remapped after a reset.  Align with gfx11.

v2: preserve init and remap for MMIO case.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx11: adjust KGQ reset sequence

Kernel gfx queues do not need to be reinitialized or
remapped after a reset.  This fixes queue reset failures
on APUs.

v2: preserve init and remap for MMIO case.

Fixes: b3e9bfd86658 ("drm/amdgpu/gfx11: add ring reset callbacks")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4789
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx12: fix wptr reset in KGQ init

wptr is a 64 bit value and we need to update the
full value, not just 32 bits. Align with what we
already do for KCQs.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx11: fix wptr reset in KGQ init

wptr is a 64 bit value and we need to update the
full value, not just 32 bits. Align with what we
already do for KCQs.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx10: fix wptr reset in KGQ init

wptr is a 64 bit value and we need to update the
full value, not just 32 bits. Align with what we
already do for KCQs.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Use AMDGPU_MQD_SIZE_ALIGN in gfx11+ kfd mqd manager

MES is enabled by default from gfx11+, use AMDGPU_MQD_SIZE_ALIGN
unconditionally for gfx11+.

Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: Adjust parameter of allocate_mqd

Make allocate_mqd consistent with other callbacks.
Prepare for next patch to use mqd_manager->mqd_size.

Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Use AMDGPU_MQD_SIZE_ALIGN in KGD

Use AMDGPU_MQD_SIZE_ALIGN for both kernel and user queue.

Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: David Belanger <david.belanger@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Initialize allowed feature list

Instead of returning feature bit mask of allowed features, initialize
the allowed features in the callback implementation itself.

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>

drm/amd/pm: Remove unused logic in SMUv14.0.2

Remove commented and redundant logic in get_allowed_feature_mask
implementation.

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>

drm/amd/pm: Add smu feature interface functions

Instead of using bitmap operations, add wrapper interface functions to
operate on smu features.

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>

drm/amd/pm: Add smu feature bits data struct

Add a bitmap struct to represent smu feature bits and functions to set/clear features.

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>

drm/amdgpu: Add a helper macro to align mqd size

MES FW uses address(mqd_addr + sizeof(struct mqd) + 3*sizeof(uint32_t))
as fence address and writes a 32 bit fence value to this address. Driver
needs to allocate some extra memory(at least 4 DWs) in addition to
sizeof(struct mqd) as mqd memory(limited to gfx/compute/sdma queue).

For gfx11/12, sizeof(struct mqd) < PAGE_SIZE, KGD allocates mqd memory with
PAGE_SIZE aligned works. For gfx12.1, sizeof(struct mqd) == PAGE_SIZE,
it doesn't work.

KFD mqd manager hardcodes mqd size to PAGE_SIZE/MQD_SIZE across different
IP versions to solve this issue.

To avoid hardcoding in differnet places and across different IP versions.
Let's use AMDGPU_MQD_SIZE_ALIGN instead. It is used in two places.

1. mqd memory alloction
2. mqd stride handling for multi xcc config

v2: Use AMDGPU_GPU_PAGE_ALIGN. (Mukul)

Signed-off-by: Lang Yu <lang.yu@amd.com>
Reviewed-by: David Belanger <david.belanger@amd.com> (v1)
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: validate user queue size constraints

Add validation to ensure user queue sizes meet hardware requirements:
- Size must be a power of two for efficient ring buffer wrapping
- Size must be at least AMDGPU_GPU_PAGE_SIZE to prevent undersized allocations

This prevents invalid configurations that could lead to GPU faults or
unexpected behavior.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix cond_exec handling in amdgpu_ib_schedule()

The EXEC_COUNT field must be > 0.  In the gfx shadow
handling we always emit a cond_exec packet after the gfx_shadow
packet, but the EXEC_COUNT never gets patched.  This leads
to a hang when we try and reset queues on gfx11 APUs.

Fixes: c68cbbfd54c6 ("drm/amdgpu: cleanup conditional execution")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4789
Reviewed-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/soc21: fix xclk for APUs

The reference clock is supposed to be 100Mhz, but it
appears to actually be slightly lower (99.81Mhz).

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/14451
Reviewed-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdkfd: gfx12.1 trap handler instruction fixup for VOP3PX

A trap may occur in the middle of VOP3PX instruction co-issue.
The PC would be restored incorrectly if left unmodified.

Identify this case by examining the instruction opcode and
rewind the PC 8 bytes if it occurs.

Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Reviewed-by: Lancelot Six <lancelot.six@amd.com>
Reviewed-by: Vladimir Indic <vladimir.indic@amd.com>
Cc: Shweta Khatri <shweta.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Fix null pointer dereference issue

If SMU is disabled, during RAS initialization,
there will be null pointer dereference issue here.

Signed-off-by: Jinzhou Su <jinzhou.su@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Merge tag 'drm-rust-next-2026-01-26' of https://gitlab.freedesktop.org/drm/rust/kernel into drm-next

DRM Rust changes for v7.0-rc1

DRM:
  - Fix documentation for Registration constructors.
  - Use pin_init::zeroed() for fops initialization.
  - Annotate DRM helpers with __rust_helper.
  - Improve safety documentation for gem::Object::new().
  - Update AlwaysRefCounted imports.

MM:
  - Prevent integer overflow in page_align().

Nova (Core):
  - Prepare for Turing support. This includes parsing and handling
    Turing-specific firmware headers and sections as well as a Turing
    Falcon HAL implementation.
  - Get rid of the Result<impl PinInit<T, E>> anti-pattern.
  - Relocate initializer-specific code into the appropriate initializer.
  - Use CStr::from_bytes_until_nul() to remove custom helpers.
  - Improve handling of unexpected firmware values.
  - Clean up redundant debug prints.
  - Replace c_str!() with native Rust C-string literals.
  - Update nova-core task list.

Nova (DRM):
  - Align GEM object size to system page size.

Tyr:
  - Use generated uAPI bindings for GpuInfo.
  - Replace manual sleeps with read_poll_timeout().
  - Replace c_str!() with native Rust C-string literals.
  - Suppress warnings for unread fields.
  - Fix incorrect register name in print statement.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: "Danilo Krummrich" <dakr@kernel.org>
Link: https://patch.msgid.link/DFYW1WV6DUCG.3K8V2DAVD1Q4A@kernel.org

Merge tag 'drm-msm-next-2026-01-23' of https://gitlab.freedesktop.org/drm/msm into drm-next

Changes for v6.20

GPU:
- Document a612/RGMU dt bindings
- UBWC 6.0 support (for A840 / Kaanapali)
- a225 support
- Fixes

DPU:
- Switched to use virtual planes by default
- Fixed DSI CMD panels on DPU 3.x
- Rewrote format handling to remove intermediate representation
- Fixed watchdog on DPU 8.x+
- Fixed TE / Vsync source setting on DPU 8.x+
- Added 3D_Mux on SC7280
- Kaanapali platform support
- Fixed UBWC register programming
- Made RM reserve DSPP-enabled mixers for CRTCs with LMs.
- Gamma correction support

DP:
- Enabled support for eDP 1.4+ link rate tables
- Fixed MDSS1 DP indices on SA8775P, making them to work
- Fixed msm_dp_ctrl_config_msa() to work with LLVM 20

DSI:
- Documented QCS8300 as compatible with SA8775P
- Kaanapali platform support

DSI PHY:
- switched to divider_determine_rate()

MDP5:
- Dropped support for MSM8998, SDM660 and SDM630 (switched over
  to DPU)

MDSS:
- Kaanapali platform support
- Fixed UBWC register programming

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rob Clark <rob.clark@oss.qualcomm.com>
Link: https://patch.msgid.link/CACSVV03Sbeca93A+gGh-TKpzFYVabbkWVgPCCicG0_NQG+5Y2A@mail.gmail.com

BackMerge tag 'v6.19-rc7' into drm-next

Linux 6.19-rc7

This is needed for msm and rust trees.

Signed-off-by: Dave Airlie <airlied@redhat.com>

drm/amdkfd: add extended capabilities to device snapshot

Add additional capabilities reporting.

Signed-off-by: Jonathan Kim <jonathan.kim@amd.com>
Reviewed-by: James Zhu <james.zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Send RMA CPER at bad page loading

Some older builds weren't sending RMA CPERs when the bad page threshold
was exceeded. Newer builds have resolved this, but there could be
systems out there with bad page numbers higher than the threshold, that
haven't sent out an RMA CPER. To be thorough and safe, send an RMA CPER
when we load the table, if the threshold is met or exceeded, instead of
waiting for the next UE to trigger the CPER.

Signed-off-by: Kent Russell <kent.russell@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Fix jpeg ring test order in vcn_v4_0_3

Fix the vcn reset sequence in vcn_v4_0_3_ring_reset() to restore
JPEG power state and unlock the JPEG powergating mutex before
running the JPEG ring post-reset helper.

Fixes: d25c67fd9d6f ("drm/amdgpu/vcn4.0.3: rework reset handling")
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix race in power state check before mutex lock

The power state check in amdgpu_dpm_set_powergating_by_smu() is done
before acquiring the pm mutex, leading to a race condition where:
1. Thread A checks state and thinks no change is needed
2. Thread B acquires mutex and modifies the state
3. Thread A returns without updating state, causing inconsistency

Fix this by moving the mutex lock before the power state check,
ensuring atomicity of the state check and modification.

Fixes: 6ee27ee27ba8 ("drm/amd/pm: avoid duplicate powergate/ungate setting")
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Promote DC to 3.2.367

* Fw release 0.1.44.0
* Fixes for corruption on platforms older than DCN4x.
* Bug fixes related to USB4 link training
* Fixes related to FP guard
* Debug helpers and other stability fixes.
* Some refactors to improve code quality

Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: [FW Promotion] Release 0.1.44.0

* Panel Replay related features/bugfixes
* BootCRC feature

Signed-off-by: Taimur Hassan <Syed.Hassan@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Migrate HUBBUB register access from hwseq to hubbub component.

[why]
Direct HUBBUB register access in the hwseq layer was creating register conflicts.

[how]
Migrated HUBBUB registers from hwseq to the hubbub component.

Reviewed-by: Martin Leung <martin.leung@amd.com>
Signed-off-by: Bhuvanachandra Pinninti <bpinnint@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: mouse event trigger to boost RR when idle

[WHY+HOW]
Add trigger event to boost refresh rate on mouse movement.

Reviewed-by: Jun Lei <jun.lei@amd.com>
Signed-off-by: Muaaz Nisar <muanisar@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add debug flag to override min dispclk

[WHY]
Enable dynamic ODM testing without needing a valid dispclk table

[HOW]
Create a debug flag to specify an override value for min dispclk

Reviewed-by: Dmytro Laktyushkin <dmytro.laktyushkin@amd.com>
Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: avoid dig reg access timeout on usb4 link training fail

[Why]
When usb4 link training fails, the dpia sym clock will be disabled and SYMCLK
source should be changed back to phy clock. In enable_streams, it is
assumed that link training succeeded and will switch from refclk to
phy clock. But phy clk here might not be on. Dig reg access timeout
will occur.

[How]
When enable_stream is hit, check if link training failed for usb4.
If it did, fall back to the ref clock to avoid reg access timeout.

Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Signed-off-by: Zhongwei <Zhongwei.Zhang@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Remove unnecessary DC FP guard

[Why & How]
For dcn2x_fast_validate_bw(), not only populate_dml_pipes needs FP guard
but also dml_get_voltage_level().

Remove unnecessary DC_FP_START/DC_FP_END guard in dcn20_fast_validate_bw
and dcn21_fast_validate_bw. FP guard is already there before calling
dcn2x_validate_bandwidth_fp().

Reviewed-by: ChiaHsuan (Tom) Chung <chiahsuan.chung@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: add setup_stereo for dcn4x or later

[why]
stereo_sync pin is removed, but we still support display stereo

Reviewed-by: Ovidiu (Ovi) Bunea <ovidiu.bunea@amd.com>
Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: perform clear update flags for all DCN asics

Existing version check that limits the sequence to clear update flags
should be performed for all asics. Exclude DCE asics for now.

Reviewed-by: Sun peng (Leo) Li <sunpeng.li@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Enable bootcrc on FW side

[Why]
The bootcrc feature is controlled on the FW side.

[How]
Pass the control bits in boot options to FW.

Reviewed-by: ChiaHsuan (Tom) Chung <chiahsuan.chung@amd.com>
Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Add FR skipping CTS functions

1. To check whether Sink reaches maximum skipping number

Reviewed-by: Robin Chen <robin.chen@amd.com>
Signed-off-by: Jack Chang <jack.chang@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Fix GFX12 family constant checks

Using >=, <= for checking the family is not always correct.

Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Matthew Stewart <Matthew.Stewart2@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/display: Enable vstateup hook for DCN401 to be reused

Add the hook to the DCN401 header file so that it can be reused in other
files

Reviewed-by: Leo Chen <leo.chen@amd.com>
Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Signed-off-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix NULL pointer dereference in amdgpu_gmc_filter_faults_remove

On APUs such as Raven and Renoir (GC 9.1.0, 9.2.2, 9.3.0), the ih1 and
ih2 interrupt ring buffers are not initialized. This is by design, as
these secondary IH rings are only available on discrete GPUs. See
vega10_ih_sw_init() which explicitly skips ih1/ih2 initialization when
AMD_IS_APU is set.

However, amdgpu_gmc_filter_faults_remove() unconditionally uses ih1 to
get the timestamp of the last interrupt entry. When retry faults are
enabled on APUs (noretry=0), this function is called from the SVM page
fault recovery path, resulting in a NULL pointer dereference when
amdgpu_ih_decode_iv_ts_helper() attempts to access ih->ring[].

The crash manifests as:

  BUG: kernel NULL pointer dereference, address: 0000000000000004
  RIP: 0010:amdgpu_ih_decode_iv_ts_helper+0x22/0x40 [amdgpu]
  Call Trace:
   amdgpu_gmc_filter_faults_remove+0x60/0x130 [amdgpu]
   svm_range_restore_pages+0xae5/0x11c0 [amdgpu]
   amdgpu_vm_handle_fault+0xc8/0x340 [amdgpu]
   gmc_v9_0_process_interrupt+0x191/0x220 [amdgpu]
   amdgpu_irq_dispatch+0xed/0x2c0 [amdgpu]
   amdgpu_ih_process+0x84/0x100 [amdgpu]

This issue was exposed by commit 1446226d32a4 ("drm/amdgpu: Remove GC HW
IP 9.3.0 from noretry=1") which changed the default for Renoir APU from
noretry=1 to noretry=0, enabling retry fault handling and thus
exercising the buggy code path.

Fix this by adding a check for ih1.ring_size before attempting to use
it. Also restore the soft_ih support from commit dd299441654f ("drm/amdgpu:
Rework retry fault removal").  This is needed if the hardware doesn't
support secondary HW IH rings.

v2: additional updates (Alex)

Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3814
Fixes: dd299441654f ("drm/amdgpu: Rework retry fault removal")
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Jon Doron <jond@wiz.io>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix smu v14 soft clock frequency setting issue

v1:
resolve the issue where some freq frequencies cannot be set correctly
due to insufficient floating-point precision.

v2:
patch this convert on 'max' value only.

Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: fix smu v13 soft clock frequency setting issue

v1:
resolve the issue where some freq frequencies cannot be set correctly
due to insufficient floating-point precision.

v2:
patch this convert on 'max' value only.

Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/include : Update MES v12 API header - SUSPEND

Update SUSPEND API to support sdma queues.
It's been supportted since 0x82 for gfx12

Signed-off-by: Shaoyun Liu <shaoyun.liu@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Simplify sorting of the bo list

Sort function only cares about the sign so we can replace the conditionals
with a single subtraction.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/mes: Remove idr leftovers v2

Commit
cb17fff3a254 ("drm/amdgpu/mes: remove unused functions")
removed most of the code using these IDRs but forgot to remove the struct
members and init/destroy paths.

There is also interrupt handling code in SDMA 5.0 and 5.2 which appears to
be using it, but is is unreachable since nothing ever allocates the
relevant IDR. We replace those with one time warnings just to avoid any
functional difference, but it is also possible they should be removed.

v2: also fix up gfx_v12_1.c and sdma_v7_1.c

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
References: cb17fff3a254 ("drm/amdgpu/mes: remove unused functions")
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/xe: Unregister drm device on probe error

Call drm_dev_unregister() when xe_device_probe() fails after successful
drm_dev_register(). This ensures the DRM device is promptly unregistered
before returning an error, avoiding leaving it registered on the failure
path.
Otherwise, there is warn message if xe_device_probe() is called again:
"
[  207.322365] [drm:drm_minor_register]
[  207.322381] debugfs: '128' already exists in 'dri'
[  207.322432] sysfs: cannot create duplicate filename '/devices/pci0000:00/0000:00:01.0/0000:01:00.0/0000:02:01.0/0000:03:00.0/drm/renderD128'
[  207.322435] CPU: 5 UID: 0 PID: 10261 Comm: modprobe Tainted: G    B   W           6.19.0-rc2-lgci-xe-kernel+ #223 PREEMPT(voluntary)
[  207.322439] Tainted: [B]=BAD_PAGE, [W]=WARN
[  207.322440] Hardware name: ASUS System Product Name/PRIME Z790-P WIFI, BIOS 0812 02/24/2023
[  207.322441] Call Trace:
[  207.322442]  <TASK>
[  207.322443]  dump_stack_lvl+0xa0/0xc0
[  207.322446]  dump_stack+0x10/0x20
[  207.322448]  sysfs_warn_dup+0xd5/0x110
[  207.322451]  sysfs_create_dir_ns+0x1f6/0x280
[  207.322453]  ? __pfx_sysfs_create_dir_ns+0x10/0x10
[  207.322455]  ? lock_acquire+0x1a4/0x2e0
[  207.322458]  ? __kasan_check_read+0x11/0x20
[  207.322461]  kobject_add_internal+0x28d/0x8e0
[  207.322464]  kobject_add+0x11f/0x1f0
[  207.322465]  ? lock_acquire+0x1a4/0x2e0
[  207.322467]  ? __pfx_kobject_add+0x10/0x10
[  207.322469]  ? __kasan_check_write+0x14/0x20
[  207.322471]  ? kobject_put+0x62/0x4a0
[  207.322473]  ? get_device_parent.isra.0+0x1bb/0x4c0
[  207.322475]  ? kobject_put+0x62/0x4a0
[  207.322477]  device_add+0x2d7/0x1500
[  207.322479]  ? __pfx_device_add+0x10/0x10
[  207.322481]  ? drm_debugfs_add_file+0xfa/0x170
[  207.322483]  ? drm_debugfs_add_files+0x82/0xd0
[  207.322485]  ? drm_debugfs_add_files+0x82/0xd0
[  207.322487]  drm_minor_register+0x10a/0x2d0
[  207.322489]  drm_dev_register+0x143/0x860
[  207.322491]  ? xe_configfs_get_psmi_enabled+0x12/0x90 [xe]
[  207.322667]  xe_device_probe+0x185b/0x2c40 [xe]
[  207.322812]  ? __pfx___drm_dev_dbg+0x10/0x10
[  207.322815]  ? add_dr+0x180/0x220
[  207.322818]  ? __pfx___drmm_mutex_release+0x10/0x10
[  207.322821]  ? __pfx_xe_device_probe+0x10/0x10 [xe]
[  207.322966]  ? xe_pm_init_early+0x33a/0x410 [xe]
[  207.323136]  xe_pci_probe+0x936/0x1250 [xe]
[  207.323298]  ? lock_acquire+0x1a4/0x2e0
[  207.323302]  ? __pfx_xe_pci_probe+0x10/0x10 [xe]
[  207.323464]  local_pci_probe+0xe6/0x1a0
[  207.323468]  pci_device_probe+0x523/0x840
[  207.323470]  ? __pfx_pci_device_probe+0x10/0x10
[  207.323473]  ? sysfs_do_create_link_sd.isra.0+0x8c/0x110
[  207.323476]  ? sysfs_create_link+0x48/0xc0
[  207.323479]  really_probe+0x1fd/0x8a0
...
"

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Signed-off-by: Shuicheng Lin <shuicheng.lin@intel.com>
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patch.msgid.link/20260109211041.2446012-2-shuicheng.lin@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit 60bfb8baf8f0d5b0d521744dfd01c880ce1a23f3)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/bridge: imx8qxp-pxl2dpi: Fix NULL pointer dereference in imx8qxp_pxl2dpi_bridge_destroy()

Pointer bridge->driver_private in imx8qxp_pxl2dpi_bridge_destroy()
is NULL when imx8qxp_pxl2dpi_bridge_probe() returns error, because
the pointer is initialized only when imx8qxp_pxl2dpi_bridge_probe()
returns 0.  The NULL pointer would be set to pointer p2d and then
NULL pointer p2d would be dereferenced.  Fix this by returning early
from imx8qxp_pxl2dpi_bridge_destroy() if !p2d is true.

Fixes: 900699ba830f ("drm/bridge: imx8qxp-pxl2dpi: get/put the companion bridge")
Signed-off-by: Liu Ying <victor.liu@nxp.com>
Reviewed-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20260123-imx8qxp-drm-bridge-fixes-v1-2-8bb85ada5866@nxp.com
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>

drm/bridge: imx8qxp-ldb: Fix NULL pointer dereference in imx8qxp_ldb_bridge_destroy()

Pointer bridge->driver_private in imx8qxp_ldb_bridge_destroy() is NULL
when a LDB channel is unavailable or imx8qxp_ldb_probe() returns error,
because ldb_add_bridge_helper() is the last function called from
imx8qxp_ldb_probe() and it doesn't initialize bridge->driver_private if
a LDB channel is unavailable.  The NULL pointer would be set to pointer
ldb_ch and then NULL pointer ldb_ch would be dereferenced.  Fix this
by returning early from imx8qxp_ldb_bridge_destroy() if !ldb_ch is true.

Fixes: 32529d384cea ("drm/bridge: imx8qxp-ldb: convert to of_drm_find_and_get_bridge()")
Signed-off-by: Liu Ying <victor.liu@nxp.com>
Reviewed-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20260123-imx8qxp-drm-bridge-fixes-v1-1-8bb85ada5866@nxp.com
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>

Merge tag 'amd-drm-next-6.20-2026-01-23' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.20-2026-01-23:

amdgpu:
- GC 12 fix
- UserQ fixes
- Misc error path fixes
- IP discovery fixes
- Misc cleanups
- HDMI fixes
- Documentation update
- Panel replay fixes
- Panel type handling fixes
- DCN 3.1.x fixes
- DC analog display fix
- SMU 6 fixes
- VCN 4.0.3 queue reset fix
- VCN 5.0.1 queue reset fix
- GPUVM TLB flush fix
- RAS fixes
- DC idle optimization fix

amdkfd:
- MQD fix for GC 9.4.3 and 9.5
- GC 11 cooperative launch fix
- GC 12.1 trap handler fixes
- Misc cleanups

radeon:
- Use drm_* logging helpers for VCE

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patch.msgid.link/20260123173537.17221-1-alexander.deucher@amd.com

drm/xe/ptl: Disable DCC on PTL

On PTL, the recommendation is to disable DCC(Duty Cycle Control) as
it may cause some regressions due to added latencies. Upcoming GuC
releases will disable DCC on PTL as well, but we need to force it in
KMD so that this behavior is propagated to older kernels.

v2: Update commit message (Rodrigo)
v3: Rebase
v4: Fix typo: s/propagted/propagated

Fixes: 5cdb71d3b0db ("drm/xe/ptl: Add GuC FW definition for PTL")
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Link: https://patch.msgid.link/20260124005917.398522-1-vinay.belgaumkar@intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
(cherry picked from commit 40ee63f5df2d5c6471b583df800aac89dc0502a4)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Reduce LRC timestamp stuck message on VFs to notice

An LRC timestamp getting stuck is a somewhat normal occurrence. If a
single VF submits a job that does not get timesliced, the LRC timestamp
will not increment. Reduce the LRC timestamp stuck message on VFs to
notice (same log level as job timeout) to avoid false CI bugs in tests
where a VF submits a job that does not get timesliced.

Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/7032
Fixes: bb63e7257e63 ("drm/xe: Avoid toggling schedule state to check LRC timestamp in TDR")
Suggested-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Stuart Summers <stuart.summers@intel.com>
Link: https://patch.msgid.link/20260114184905.4189026-1-matthew.brost@intel.com
(cherry picked from commit e89aacd1ecdd3d13e8f347aa082687878621e03c)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/display: bridge_connector: move audio_infoframe checks to OP_HDMI

There are DRM_BRIDGE_OP_HDMI_AUDIO bridges (e.g. Lontium LT9611UXC)
which don't implement DRM_BRIDGE_OP_HDMI and don't implement
hdmi_clear_audio_infoframe / hdmi_write_audio_infoframe callbacks.

Move corresponding checks under the DRM_BRIDGE_OP_HDMI condition, making
sure that we require those callbacks only from the bridges which are
actually going to use them.

Fixes: b626b1a1c9cc ("drm/bridge: refactor HDMI InfoFrame callbacks")
Acked-by: Maxime Ripard <mripard@kernel.org>
Link: https://patch.msgid.link/20260126-drm-fix-lt9611uxc-v1-1-04a635a8f5cf@oss.qualcomm.com
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>

Documentation: nova: update pending tasks

Some tasks, while already having some progress, are still pending, hence
update them.

Link: https://patch.msgid.link/20260119153129.59876-2-dakr@kernel.org
[ Remove PCI config space, as it is completed. - Danilo ]
Signed-off-by: Danilo Krummrich <dakr@kernel.org>

Documentation: nova: remove completed tasks

Remove completed tasks from Nova's task list.

Link: https://patch.msgid.link/20260119153129.59876-1-dakr@kernel.org
Signed-off-by: Danilo Krummrich <dakr@kernel.org>

drm/i915/display: Prevent u64 underflow in intel_fbc_stolen_end

Static analysis reveals a potential integer underflow in
intel_fbc_stolen_end.  This can apparently occur if
intel_parent_stolen_area_size returns zero (or, theoretically, any value
less than 2^23), as 2^23 is subtracted from the return value and stored
in a u64.  While this doesn't appear to cause any issues due to the use
of the min() function to clamp the return values from the
intel_fbc_stolen_end function, it would be best practice to avoid
undeflowing values like this on principle.  So, rework the function to
prevent the underflow from occurring.  Note that the underflow at
present would result in the value of intel_fbc_cfb_base_max being
returned at the end of intel_fbc_stolen_end, so just return that if the
value of intel_parent_stolen_area_size is too small.

While we're here, fix the other comments here and modify the execution
path for readability.

v2: (Jani)
- Fix the comments in intel_fbc_stolen_end
- Use check_sub_overflow
- Remove macro that mirrors SZ_8M, as it is now only referenced once
- Misc. formatting fixes

Fixes: a9da512b3ed7 ("drm/i915: avoid the last 8mb of stolen on BDW/SKL")
Signed-off-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patch.msgid.link/20260107162935.8123-2-jonathan.cavitt@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
(cherry picked from commit 6695dc279820a50cb20ecd8b5250e05234dac780)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

Linux 6.19-rc7

Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Only one core change, the rest are drivers.

  The core change reorders some state operations in the error handler to
  try to prevent missed wake ups of the error handler (which can halt
  error processing and effectively freeze the entire system)"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: qla2xxx: Sanitize payload size to prevent member overflow
  scsi: target: iscsi: Fix use-after-free in iscsit_dec_session_usage_count()
  scsi: target: iscsi: Fix use-after-free in iscsit_dec_conn_usage_count()
  scsi: core: Wake up the error handler when final completions race against each other
  scsi: storvsc: Process unsupported MODE_SENSE_10
  scsi: xen: scsiback: Fix potential memory leak in scsiback_remove()

Merge tag 'keys-trusted-next-6.19-rc7' of git://git./linux/kernel/git/jarkko/linux-tpmdd

Pull keys fix from Jarkko Sakkinen.

* tag 'keys-trusted-next-6.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-tpmdd:
  keys/trusted_keys: fix handle passed to tpm_buf_append_name during unseal

Merge tag 'char-misc-6.19-rc7' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc/iio driver fixes from Greg KH:
 "Here are some small char/misc/iio and some other minor driver
  subsystem fixes for 6.19-rc7. Nothing huge here, just some fixes for
  reported issues including:

   - lots of little iio driver fixes

   - comedi driver fixes

   - mux driver fix

   - w1 driver fixes

   - uio driver fix

   - slimbus driver fixes

   - hwtracing bugfix

   - other tiny bugfixes

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'char-misc-6.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc: (36 commits)
  comedi: dmm32at: serialize use of paged registers
  mei: trace: treat reg parameter as string
  uio: pci_sva: correct '-ENODEV' check logic
  uacce: ensure safe queue release with state management
  uacce: implement mremap in uacce_vm_ops to return -EPERM
  uacce: fix isolate sysfs check condition
  uacce: fix cdev handling in the cleanup path
  slimbus: core: clean up of_slim_get_device()
  slimbus: core: fix of_slim_get_device() kernel doc
  slimbus: core: amend slim_get_device() kernel doc
  slimbus: core: fix device reference leak on report present
  slimbus: core: fix runtime PM imbalance on report present
  slimbus: core: fix OF node leak on registration failure
  intel_th: rename error label
  intel_th: fix device leak on output open()
  comedi: Fix getting range information for subdevices 16 to 255
  mux: mmio: Fix IS_ERR() vs NULL check in probe()
  interconnect: debugfs: initialize src_node and dst_node to empty strings
  iio: dac: ad3552r-hs: fix out-of-bound write in ad3552r_hs_write_data_source
  iio: accel: iis328dq: fix gain values
  ...