linux-2.6-microblaze.git
3 months agoMerge tag 'drm-misc-fixes-2026-05-21' of https://gitlab.freedesktop.org/drm/misc...
Dave Airlie [Thu, 21 May 2026 21:00:45 +0000 (07:00 +1000)]
Merge tag 'drm-misc-fixes-2026-05-21' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-fixes

Short summary of fixes pull:

amdxdna:
- remove mmap and export for ubuf

bridge:
- chipone-icn6211: managed bridge cleanup
- lt66121: acquire reset GPIO
- megachips: fix clean up on failed IRQ requests

gem:
- clean up LRU locking

v3d:
- fix UAF in error code paths
- release GEM-object ref on free'd jobs

virtio:
- use uninterruptible resv locking in plane updates

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patch.msgid.link/20260521071456.GA14644@localhost.localdomain
3 months agoMerge tag 'amd-drm-fixes-7.1-2026-05-20' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Thu, 21 May 2026 06:29:38 +0000 (16:29 +1000)]
Merge tag 'amd-drm-fixes-7.1-2026-05-20' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-7.1-2026-05-20:

amdgpu:
- Userq fixes
- VPE fix
- SMU 15 fix
- Misc fixes
- VCE fixes
- DC bios parsing fixes
- DC aux fix
- Mode1 reset fix
- RAS fixes

amdkfd:
- Misc fixes

radeon:
- CS parser fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patch.msgid.link/20260520181359.28421-1-alexander.deucher@amd.com
3 months agoMerge tag 'drm-intel-fixes-2026-05-20' of https://gitlab.freedesktop.org/drm/i915...
Dave Airlie [Thu, 21 May 2026 01:50:27 +0000 (11:50 +1000)]
Merge tag 'drm-intel-fixes-2026-05-20' of https://gitlab.freedesktop.org/drm/i915/kernel into drm-fixes

- Fix joiner color pipeline selection [display] (Chaitanya Kumar Borah)
- Fix readback for target_rr in Adaptive Sync SDP [dp] (Ankit Nautiyal)
- Apply Intel DPCD workaround when SDP on prior line used [psr] (Jouni Högander)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Tvrtko Ursulin <tursulin@igalia.com>
Link: https://patch.msgid.link/ag1hKBRKwwv9JOMW@linux
3 months agoMerge tag 'drm-msm-fixes-2026-05-17' of https://gitlab.freedesktop.org/drm/msm into...
Dave Airlie [Thu, 21 May 2026 00:12:21 +0000 (10:12 +1000)]
Merge tag 'drm-msm-fixes-2026-05-17' of https://gitlab.freedesktop.org/drm/msm into drm-fixes

Fixes for v7.1:

Core:
- Fixed bindings for SM8650, SM8750 and Eliza
- Don't use UTS_RELEASE directly
- Fix typo in clock-names property

DPU:
- Fixed CWB description on Kaanapali
- Fixed scanline strides for YUV UBWC formats
- Stopped DSI register dumping to access past the end of region

DSI:
- Fix dumping unaligned regions

GPU:
- Fix GMEM_BASE for a6xx gen3
- Fix userspace reachable crash on a2xx-a4xx
- Fix sysprof_active for counter collection with IFPC enabled GPUs
- Fix shrinker lockdep

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rob Clark <rob.clark@oss.qualcomm.com>
Link: https://patch.msgid.link/CACSVV02cTK7h=d0uqanRE-cj35THDqFjqsTB_2zQV1Mcw77aNw@mail.gmail.com
3 months agodrm/virtio: use uninterruptible resv lock for plane updates
Deepanshu Kartikey [Tue, 19 May 2026 08:22:47 +0000 (13:52 +0530)]
drm/virtio: use uninterruptible resv lock for plane updates

virtio_gpu_cursor_plane_update() and virtio_gpu_resource_flush() lock
the framebuffer BO's dma_resv via virtio_gpu_array_lock_resv() and
ignore its return value. The function can fail with -EINTR from
dma_resv_lock_interruptible() (signal during lock wait) or with
-ENOMEM from dma_resv_reserve_fences() (fence slot allocation),
leaving the resv lock not held. The queue path then walks the object
array and calls dma_resv_add_fence(), which requires the lock held;
with lockdep enabled this trips dma_resv_assert_held():

  WARNING: drivers/dma-buf/dma-resv.c:296 at dma_resv_add_fence+0x71e/0x840
  Call Trace:
   virtio_gpu_array_add_fence
   virtio_gpu_queue_ctrl_sgs
   virtio_gpu_queue_fenced_ctrl_buffer
   virtio_gpu_cursor_plane_update
   drm_atomic_helper_commit_planes
   drm_atomic_helper_commit_tail
   commit_tail
   drm_atomic_helper_commit
   drm_atomic_commit
   drm_atomic_helper_update_plane
   __setplane_atomic
   drm_mode_cursor_universal
   drm_mode_cursor_common
   drm_mode_cursor_ioctl
   drm_ioctl
   __x64_sys_ioctl

Beyond the WARN, mutating the dma_resv fence list without the lock
races with concurrent readers/writers and can corrupt the list.

Both call sites run inside the .atomic_update plane callback, which
DRM atomic helpers do not allow to fail (by the time it runs, the
commit has been signed off to userspace and there is no clean
rollback path). Moving the lock acquisition to .prepare_fb was
rejected because the broader lock scope deadlocks against other BO
locking paths in the same atomic commit.

Introduce virtio_gpu_lock_one_resv_uninterruptible() that uses
dma_resv_lock() instead of dma_resv_lock_interruptible(). This
eliminates the -EINTR failure mode -- the realistic syzbot trigger
-- without extending the lock hold across the commit. The helper
locks a single BO and rejects nents > 1 with -EINVAL; both fix
sites lock exactly one BO.

Use it from virtio_gpu_cursor_plane_update() and
virtio_gpu_resource_flush(); check the return value to handle the
remaining -ENOMEM case from dma_resv_reserve_fences() by freeing
the objs and skipping the plane update for that frame. The
framebuffer BOs touched here are not shared with other contexts
and lock contention is expected to be brief, so the loss of
signal-interruptibility is acceptable.

Other callers of virtio_gpu_array_lock_resv() (the ioctl paths)
continue to use the interruptible variant.

The bug was reported by syzbot, triggered via fault injection
(fail_nth) on the DRM_IOCTL_MODE_CURSOR path, which forces the
-ENOMEM branch in dma_resv_reserve_fences().

Reported-by: syzbot+72bd3dd3a5d5f39a0271@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=72bd3dd3a5d5f39a0271
Fixes: 5cfd31c5b3a3 ("drm/virtio: fix virtio_gpu_cursor_plane_update().")
Cc: stable@vger.kernel.org
Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Link: https://patch.msgid.link/20260519082247.34470-1-kartikey406@gmail.com
3 months agodrm/amdgpu: fix handling in amdgpu_userq_create
Christian König [Mon, 27 Apr 2026 14:31:31 +0000 (16:31 +0200)]
drm/amdgpu: fix handling in amdgpu_userq_create

Well mostly the same issues the other code had as well:

1. Memory allocation while holding the userq_mutex lock is forbidden!
2. Things were created/started/published in the wrong order.
3. The reset lock was taken in the wrong order and seems to be
   unecessary in the first place.
4. Error messages on invalid input parameters can spam the logs.
5. Error messages on memory allocation failures are usually superflous
   as well.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 89e50de5654dbe7a137e03d78629542e17ba7202)

3 months agodrm/radeon/evergreen_cs: Add missing NULL prefix check in surface check
Vitaliy Triang3l Kuzmin [Fri, 15 May 2026 21:48:32 +0000 (00:48 +0300)]
drm/radeon/evergreen_cs: Add missing NULL prefix check in surface check

'evergreen_surface_check' is called with a NULL warning prefix when
handling potentially recoverable issues or just to compute the alignment
requirements, and 'evergreen_surface_check' is called again in case of
failure (with the correct prefix, as opposed to NULL), therefore, the
initial check must not print a warning, because the surface may be
accepted successfully after having been corrected, however if it isn't,
the final check will print the warning anyway. The surface check
functions specific to array modes already implement this behavior, but
the 'evergreen_surface_check' function itself doesn't.

This is also supposed to fix the "'%s' directive argument is null
[-Werror=format-overflow=]" compiler warning.

Fixes: 285484e2d55e ("drm/radeon: add support for evergreen/ni tiling informations v11")
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vitaliy Triang3l Kuzmin <ml@triang3l.ru>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit e20ea411c99f6968af35fd03e9ee21f70d799144)

3 months agodrm/amdgpu: userq_va_mapped should remain true once done
Sunil Khatri [Wed, 13 May 2026 07:59:35 +0000 (13:29 +0530)]
drm/amdgpu: userq_va_mapped should remain true once done

Multiple queues needs these bo_va objects belonging to
the same uq_mgr. So once they are mapped lets not unmap
them as at any point of time any of the queues might be
using it.

Also userq_va_mapped should be a boolean than atomic.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 5c02889ea22575c3bcfdf212e65fac316cbc6c6a)

3 months agodrm/amdgpu: avoid integer overflow in VA range check
Ce Sun [Mon, 11 May 2026 10:04:57 +0000 (18:04 +0800)]
drm/amdgpu: avoid integer overflow in VA range check

The original addition operation in 64-bit unsigned type may encounter
overflow situations. To prevent such issues and safely reject invalid
inputs, the check_add_overflow() function is used.

Signed-off-by: Ce Sun <cesun102@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit cc768f4dd0bb9083c813683eeec44fc23921f771)

3 months agodrm/amd/ras: Fix UMC error address allocation leak
Xiang Liu [Mon, 11 May 2026 13:28:59 +0000 (21:28 +0800)]
drm/amd/ras: Fix UMC error address allocation leak

amdgpu_umc_handle_bad_pages() allocates err_data->err_addr before
querying UMC error information. In the direct and firmware query paths,
the pointer is reassigned to a fresh allocation before the original
buffer is released, so the initial allocation is leaked on each handled
event.

Free the existing buffer before replacing it in those query paths so the
function exit cleanup only owns the active allocation.

Signed-off-by: Xiang Liu <xiang.liu@amd.com>
Reviewed-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 911b1bdd22c3712a22b60fcc58f7b9f2d07b0803)

3 months agodrm/amdgpu: unmap all user mappings of framebuffer and doorbell before mode1 reset
Yifan Zhang [Mon, 11 May 2026 14:14:23 +0000 (22:14 +0800)]
drm/amdgpu: unmap all user mappings of framebuffer and doorbell before mode1 reset

During Mode 1 reset, the ASIC undergoes a reset cycle and becomes temporarily
inaccessible via PCIe. Any attempt to access framebuffer or MMIO registers during
this window can result in uncompleted PCIe transactions, leading to NMI panics or
system hangs.

To prevent this, Unmap all of the applications mappings of the framebuffer
and doorbell BARs before mode1 reset. Also prevent new mappings from coming in
during the reset process.

v2: remove inode in kfd_dev (Christian)
v3: correct unmap offset (Felix), remove prevent new mappings part
to avoid deadlock (Christian)

Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 70cadefcc6160c575b04f763ada34c20e868d577)

3 months agodrm/amd/display: Validate payload length and link_index in dc_process_dmub_aux_transf...
Harry Wentland [Thu, 7 May 2026 20:26:31 +0000 (16:26 -0400)]
drm/amd/display: Validate payload length and link_index in dc_process_dmub_aux_transfer_async

[Why&How]
dc_process_dmub_aux_transfer_async() copies payload->length bytes into a
16-byte stack buffer (dpaux.data[16]) guarded only by an ASSERT(), which
is a no-op in release builds. If a caller ever passes length > 16 this
results in a stack buffer overflow via memcpy.

Additionally, link_index is used to dereference dc->links[] without
bounds checking against dc->link_count, risking an out-of-bounds access.

Replace the ASSERT with a hard runtime check that returns false when
payload->length exceeds the destination buffer size, and add a bounds
check for link_index before it is used.

Assisted-by: GitHub Copilot:Claude claude-4-opus
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit ba4caa9fecdf7a38f98c878ad05a8a64148b6881)
Cc: stable@vger.kernel.org
3 months agodrm/amd/display: Validate GPIO pin LUT table size before iterating
Harry Wentland [Mon, 4 May 2026 20:14:11 +0000 (16:14 -0400)]
drm/amd/display: Validate GPIO pin LUT table size before iterating

[Why&How]
The GPIO pin table parsers in get_gpio_i2c_info() and
bios_parser_get_gpio_pin_info() derive an element count from the VBIOS
table_header.structuresize field, then iterate over gpio_pin[] entries.
However, GET_IMAGE() only validates that the table header itself fits
within the BIOS image. If the VBIOS reports a structuresize larger than
the actual mapped data, the loop reads past the end of the BIOS image,
causing an out-of-bounds read.

Fix this by calling bios_get_image() to validate that the full claimed
structuresize is accessible within the BIOS image before entering the
loop in both functions.

Assisted-by: GitHub Copilot:claude-opus-4-6
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit ba5e95b43b773ae1bf1f66ee6b31eb774e65afe3)
Cc: stable@vger.kernel.org
3 months agodrm/amd/display: Fix integer overflow in bios_get_image()
Harry Wentland [Mon, 4 May 2026 15:14:45 +0000 (11:14 -0400)]
drm/amd/display: Fix integer overflow in bios_get_image()

[Why&How]
The bounds check in bios_get_image() computes 'offset + size' using
unsigned 32-bit arithmetic before comparing against bios_size. If a
VBIOS image contains a near-UINT32_MAX offset the addition wraps to a
small value, the comparison passes, and the function returns a wild
pointer past the VBIOS mapping.

Additionally, the comparison uses '<' (strict), which incorrectly
rejects the valid exact-fit case where offset + size == bios_size.

Fix both issues by restructuring the check to avoid the addition
entirely: first reject if offset alone exceeds bios_size, then check
size against the remaining space (bios_size - offset). This eliminates
the overflow and correctly permits exact-fit accesses.

Assisted-by: GitHub Copilot:claude-opus-4.6
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit d40fb392af659c4a02b560319f226842f6ec1a95)
Cc: stable@vger.kernel.org
3 months agodrm/amdkfd: Check bounds for allocate_sdma_queue restore_sdma_id
David Francis [Tue, 12 May 2026 19:18:18 +0000 (15:18 -0400)]
drm/amdkfd: Check bounds for allocate_sdma_queue restore_sdma_id

allocate_sdma_queue has an option where the sdma queue id can be
specified (used by CRIU). We weren't bounds-checking that
value.

Confirm it's less than the maximum number of queues.

Signed-off-by: David Francis <David.Francis@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit bfe9a7545b2a7be1c543f1741e16f2d5ec4116ae)

3 months agodrm/amdgpu: use atomic operation to achieve lockless serialization
Sunil Khatri [Thu, 14 May 2026 07:01:00 +0000 (12:31 +0530)]
drm/amdgpu: use atomic operation to achieve lockless serialization

In amdgpu_seq64_alloc there is a possibility that two difference cores
from two separate NODES can try to and could get the same free slot.
So this fixes that race here using atomic test_and_set clear operations.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 4d50a14d346141e03a7c3905e496d91e048bc30c)

3 months agodrm/amdkfd: Check bounds on allocate_doorbell
David Francis [Tue, 12 May 2026 19:15:33 +0000 (15:15 -0400)]
drm/amdkfd: Check bounds on allocate_doorbell

allocated_doorbell has an option to set the doorbell id
to a specific value (used by CRIU). This value was not
bounds checked.

Check to confirm it's less than KFD_MAX_NUM_OF_QUEUES_PER_PROCESS.

Signed-off-by: David Francis <David.Francis@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 1f087bb8cf9e8797633da35c85435e557ef74d06)

3 months agodrm/amdgpu/vce3: Fix VCE 3 firmware size and offsets
Timur Kristóf [Wed, 13 May 2026 20:04:16 +0000 (22:04 +0200)]
drm/amdgpu/vce3: Fix VCE 3 firmware size and offsets

The VCPU BO contains the actual FW at an offset, but
it was not calculated into the VCPU BO size.
Subtract this from the FW size to make sure there is
no out of bounds access.

This may fix VM faults when using VCE 3.

Cc: John Olender <john.olender@gmail.com>
Fixes: e98226221467 ("drm/amdgpu: recalculate VCE firmware BO size")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 15c369257bd85f47a514744f960c5a51c867716f)

3 months agodrm/amdgpu/vce2: Fix VCE 2 firmware size and offsets
Timur Kristóf [Wed, 13 May 2026 20:04:15 +0000 (22:04 +0200)]
drm/amdgpu/vce2: Fix VCE 2 firmware size and offsets

The VCPU BO contains the actual FW at an offset, but
it was not calculated into the VCPU BO size.
Subtract this from the FW size to make sure there is
no out of bounds access.

Additionally, increase the VCE_V2_0_DATA_SIZE to
have extra space after the VCE handles.

Also increase the data size used for each VCE handle.
The FW needs 23744 bytes, use 24K to be safe.

This fixes VM faults when using VCE 2.

Cc: John Olender <john.olender@gmail.com>
Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/4802
Fixes: e98226221467 ("drm/amdgpu: recalculate VCE firmware BO size")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit a20d21df625548c1738c0745f753c5d6eb823bc3)

3 months agodrm/amdgpu/vce1: Stop using amdgpu_vce_resume
Timur Kristóf [Wed, 13 May 2026 20:04:14 +0000 (22:04 +0200)]
drm/amdgpu/vce1: Stop using amdgpu_vce_resume

The VCE1 firmware works slightly differently and is already
loaded by vce_v1_0_load_fw(). It doesn't actually need to
call amdgpu_vce_resume().

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 33d8951405e2dd81ac61edebc680e2dfb6b4fc9f)

3 months agodrm/amdgpu/vce1: Fix VCE 1 firmware size and offsets
Timur Kristóf [Wed, 13 May 2026 20:04:13 +0000 (22:04 +0200)]
drm/amdgpu/vce1: Fix VCE 1 firmware size and offsets

The VCPU BO contains the actual FW at an offset, but
it was not calculated into the VCPU BO size.
Subtract this from the FW size to make sure there is
no out of bounds access.

Make sure the stack and data offsets are aligned to
the 32K TLB size.

Check that the FW microcode actually fits in the
space that is reserved for it.

Fixes: d4a640d4b9f3 ("drm/amdgpu/vce1: Implement VCE1 IP block (v2)")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit c16fe59f622a080fc457a57b3e8f14c780699449)

3 months agodrm/amdgpu/vce1: Don't repeat GTT MGR node allocation
Timur Kristóf [Wed, 13 May 2026 20:04:12 +0000 (22:04 +0200)]
drm/amdgpu/vce1: Don't repeat GTT MGR node allocation

Only allocate entries from the GTT manager when the
VCE GTT node is not allocated yet. This prevents the
possibility of allocating them multiple times, which
causes issues during GPU reset and suspend/resume.

Fixes: 71aec08f80e7 ("amdgpu/vce: use amdgpu_gtt_mgr_alloc_entries")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 8d2a20c1721cb17e22821e1b4ecbb02d475d91c5)

3 months agodrm/amdgpu/vce1: Check if VRAM address is lower than GART.
Timur Kristóf [Wed, 13 May 2026 20:04:11 +0000 (22:04 +0200)]
drm/amdgpu/vce1: Check if VRAM address is lower than GART.

Previously, I had assumed this was not possible
so it was OK to not handle it, but now we got a report
from a user who has a board that is configured this way.

When the VCPU BO is already located in a low 32-bit address
in VRAM (eg. when VRAM is mapped to the low address space),
don't do the workaround.

Fixes: 71aec08f80e7 ("amdgpu/vce: use amdgpu_gtt_mgr_alloc_entries")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit f370ec9b164698a9ca1a7b59bfbea07f70df769d)

3 months agodrm/amdgpu/vce1: Remove superfluous address check
Timur Kristóf [Wed, 13 May 2026 20:04:10 +0000 (22:04 +0200)]
drm/amdgpu/vce1: Remove superfluous address check

The same thing is already checked a few lines above.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit c1dc555e760dbfc4a4710f7270f525a03d433af8)

3 months agodrm/amdgpu/vce1: Check that the GPU address is < 128 MiB
Timur Kristóf [Wed, 13 May 2026 20:04:09 +0000 (22:04 +0200)]
drm/amdgpu/vce1: Check that the GPU address is < 128 MiB

When ensuring the low 32-bit address, make sure it is
less than 128 MiB, otherwise the VCE seems to fail to initialize.
This seems to be an undocumented limitation of the firmware
validation mechanism. Note that in case of VCE1 the BAR
address is zero and we can't change it also due to the
firmware validator.

When programming the mmVCE_VCPU_CACHE_OFFSETn registers,
don't AND them with a mask. This is incorrect because
the register mask is actually 0x0fffffff and useless because
we already ensure the addresses are below the limit.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit e729ae5f3ac73c861c062080ac8c3d666c972404)

3 months agodrm/amdgpu: Align amdgpu_gtt_mgr entries to TLB size on Tahiti (v2)
Timur Kristóf [Wed, 13 May 2026 20:04:08 +0000 (22:04 +0200)]
drm/amdgpu: Align amdgpu_gtt_mgr entries to TLB size on Tahiti (v2)

The TLB is organized in groups of 8 entries, each one is 4K.
On Tahiti, the HW requires these GART entries to be 32K-aligned.

This fixes a VCE 1 firmware validation failure that can happen
after suspend/resume since we use amdgpu_gtt_mgr for VCE 1.

v2:
- Change variable declaration order
- Add comment about "V bit HW bug"

Fixes: 698fa62f56aa ("drm/amdgpu: Add helper to alloc GART entries")
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 530411b465ef0b2c0cc18c2e3d7e38422b1117d1)

3 months agodrm/amdkfd: Fix OOB memory exposure in get_wave_state()
Sunday Clement [Wed, 13 May 2026 15:22:19 +0000 (11:22 -0400)]
drm/amdkfd: Fix OOB memory exposure in get_wave_state()

The get_wave_state() function for v9 trusts cp_hqd_cntl_stack_size and
cp_hqd_cntl_stack_offset values read directly from the MQD, which are
written by GPU microcode and fully attacker-controlled on the
CRIU-restore path (via AMDKFD_IOC_RESTORE_PROCESS with H3).

this leads to an unbounded copy_to_user() that can leak adjacent
GTT/kernel memory. If offset > size, integer underflow produces a ~4 GiB
read length, if size is set to 1 MiB against a 4 KiB allocation, we leak
1 MiB of adjacent kernel memory (other queues' MQDs, ring buffers, KASLR
pointers).

Fix by clamping both cp_hqd_cntl_stack_size to the actual allocated
buffer size (q->ctl_stack_size) and cp_hqd_cntl_stack_offset to the
clamped size before performing arithmetic and copy_to_user().

This ensures we never read beyond the allocated kernel BO regardless of
attacker-supplied MQD field values.

Signed-off-by: Sunday Clement <Sunday.Clement@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 7ef144458f48d5589e36f1b3d83e83db2e5c5ba5)

3 months agodrm/amd/pm: fix memleak of dpm_policies on smu v15
Yang Wang [Sat, 9 May 2026 07:20:39 +0000 (15:20 +0800)]
drm/amd/pm: fix memleak of dpm_policies on smu v15

In smu_v15_0_fini_smc_tables, dpm_policies was not freed or NULLed, causing a memory leak.
Add kfree() and NULL assignment to properly release memory and avoid dangling pointers.

Fixes: 2beedc3a92b7 ("drm/amd/pm: Add initial support for smu v15_0_8");
Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Asad Kamal <asad.kamal@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 014f329074f688b9b49383e8b70e79e9ef99359e)

3 months agodrm/amdgpu: Fix discovery offset check under VF
Lijo Lazar [Tue, 12 May 2026 14:59:52 +0000 (20:29 +0530)]
drm/amdgpu: Fix discovery offset check under VF

Discovery table may be kept at offset 0 by host driver. Remove the
validation check.

Fixes: 01bdc7e219c4 ("drm/amdgpu: New interface to get IP discovery binary v3")
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Ellen Pan <yunru.pan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit d3f5bbd007133c64a20e81ef290a93e46c75df40)

3 months agodrm/amdgpu: remove va cursors for all mappings
Sunil Khatri [Tue, 12 May 2026 16:59:48 +0000 (22:29 +0530)]
drm/amdgpu: remove va cursors for all mappings

va_cursor struct needs to be cleaned even if the mapping
has been removed already.

Also simplify it by make it a void function as return value
check isn't needed as its called during tear down.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 4d35a45c9b4c1ac5b6e3219f83c3db706b675fa2)

3 months agodrm/amdgpu: reject non-user addresses early in GEM_USERPTR ioctl
Amir Shetaia [Thu, 7 May 2026 17:24:55 +0000 (13:24 -0400)]
drm/amdgpu: reject non-user addresses early in GEM_USERPTR ioctl

amdgpu_gem_userptr_ioctl() currently accepts any value of args->addr
and only discovers an out-of-range pointer much later, inside
amdgpu_gem_object_create() and the HMM mirror registration path.
Userspace can drive that path with kernel-side virtual addresses;
the get_user_pages() layer rejects them, but only after the driver
has already allocated a GEM object and started wiring up notifier
state that then has to be torn down on failure.

Add an access_ok() guard at the top of the ioctl, right after the
existing page-alignment check and before flag validation, so any
address that does not lie within the calling task's user address
range is rejected with -EFAULT before any allocation occurs. No
legitimate ROCm/HSA userspace passes kernel-mode pointers through
this interface, so this is defense-in-depth rather than a behaviour
change for valid callers; -EFAULT matches the convention already
used by other uaccess-style rejections in the kernel.

Also add an explicit #include <linux/uaccess.h>; access_ok() is
otherwise only available transitively through other headers in
this translation unit.

Signed-off-by: Amir Shetaia <Amir.Shetaia@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 7a076df36397d780d7e4fb595287b4980451a7f5)

3 months agodrm/amdgpu/vpe: Force collaborate sync after TRAP
Alan Liu [Fri, 1 May 2026 04:35:48 +0000 (12:35 +0800)]
drm/amdgpu/vpe: Force collaborate sync after TRAP

VPE1 could possibly hang and fail to power off at the end of commands in
collaboration mode. This workaround adds a COLLAB_SYNC after TRAP to
force instances synchronized to avoid VPE1 fail to power off.

Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Alan liu <haoping.liu@amd.com>
Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/5171
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit a8b749c5c5afb7e5daa2bfb95d958fb3c6b8f055)
Cc: stable@vger.kernel.org
3 months agodrm/amdgpu/userq: update the vm task info during signal ioctl
Sunil Khatri [Tue, 12 May 2026 10:30:18 +0000 (16:00 +0530)]
drm/amdgpu/userq: update the vm task info during signal ioctl

Pagefaults does not have process information correctly populated
as vm->task is not set during vm_init but should be updated while
real submission. So setting that up during signal_ioctl to get
the correct submission process details.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit a9b14d88b4d83e21ab965f23d1fb7b07b87e0517)

3 months agodrm/amdgpu/userq: cancel reset work while tear down in progress
Sunil Khatri [Tue, 12 May 2026 09:22:40 +0000 (14:52 +0530)]
drm/amdgpu/userq: cancel reset work while tear down in progress

While tear down of a userq_mgr is happening when all the queues
are free we should cancel any reset work if pending before exiting.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 160164609f71f774c4f661227a9b7a370a86b112)

3 months agodrm/amdgpu: rework userq reset work handling
Christian König [Tue, 21 Apr 2026 10:39:54 +0000 (12:39 +0200)]
drm/amdgpu: rework userq reset work handling

It is illegal to schedule reset work from another reset work!

Fix this by scheduling the userq reset work directly on the work queue
of the reset domain.

Not fully tested, I leave that to the IGT test cases.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit fd9200ccefab94f27877d1943761d6b0ccbd89c8)

3 months agodrm/amdgpu/userq: pin mqd and fw object bo to avoid eviction
Sunil Khatri [Fri, 8 May 2026 10:28:09 +0000 (15:58 +0530)]
drm/amdgpu/userq: pin mqd and fw object bo to avoid eviction

mqd and fw objects are queue core objects which should remain
valid and never be unmapped and evicted for user queues to work
properly.

During eviction if these buffers are evicted the hw continue to
use the invalid addresses and caused page faults and system hung.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit a3bbf32a336939a1d21b9561f8e53333b684b7ef)

3 months agodrm/amdgpu/userq: use drm_exec in amdgpu_userq_fence_read_wptr
Sunil Khatri [Fri, 8 May 2026 06:51:20 +0000 (12:21 +0530)]
drm/amdgpu/userq: use drm_exec in amdgpu_userq_fence_read_wptr

To access the bo from vm mapping first lock the root bo and
then the object bo of the mapping to make sure both locks
are taken safely.

Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 3aab50410653fe7eb35eb6f9c2b27e3549ab09e6)

3 months agoaccel/amdxdna: Remove mmap and export support for ubuf
Lizhi Hou [Mon, 18 May 2026 15:57:05 +0000 (08:57 -0700)]
accel/amdxdna: Remove mmap and export support for ubuf

Ubuf pages should not be mmaped or exported. Remove the ubuf mmap callback
and return -EOPNOTSUPP when exporting ubuf objects.

ubuf vmap is also removed for there is not a real use case yet.

Fixes: bd72d4acda10 ("accel/amdxdna: Support user space allocated buffer")
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260518155706.937461-1-lizhi.hou@amd.com
3 months agodrm/bridge: megachips: remove bridge when irq request fails
Osama Abdelkader [Thu, 30 Apr 2026 19:56:59 +0000 (21:56 +0200)]
drm/bridge: megachips: remove bridge when irq request fails

If devm_request_threaded_irq() fails after drm_bridge_add(), remove the
bridge before returning.

Keep drm_bridge_add() rather than devm_drm_bridge_add(): registration is
tied to the STDP4028 device while ge_b850v3_register() may complete from
either I2C probe; devm would not unwind the bridge if the other client's
probe fails.

Signed-off-by: Osama Abdelkader <osama.abdelkader@gmail.com>
Fixes: fcfa0ddc18ed ("drm/bridge: Drivers for megachips-stdpxxxx-ge-b850v3-fw (LVDS-DP++)")
Cc: stable@vger.kernel.org
Reviewed-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Tested-by: Ian Ray <ian.ray@gehealthcare.com>
Link: https://patch.msgid.link/20260430195700.80317-1-osama.abdelkader@gmail.com
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
3 months agodrm/bridge: chipone-icn6211: use devm_drm_bridge_add in i2c probe
Osama Abdelkader [Thu, 30 Apr 2026 19:49:42 +0000 (21:49 +0200)]
drm/bridge: chipone-icn6211: use devm_drm_bridge_add in i2c probe

Use devm_drm_bridge_add() so the bridge is released if probe
fails after registration, and drop drm_bridge_remove() in chipone_i2c_probe.

Signed-off-by: Osama Abdelkader <osama.abdelkader@gmail.com>
Fixes: 8dde6f7452a1 ("drm: bridge: icn6211: Add I2C configuration support")
Cc: stable@vger.kernel.org
Reviewed-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Link: https://patch.msgid.link/20260430194944.78119-1-osama.abdelkader@gmail.com
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
3 months agodrm/i915/psr: Apply Intel DPCD workaround when SDP on prior line used
Jouni Högander [Fri, 15 May 2026 09:57:55 +0000 (12:57 +0300)]
drm/i915/psr: Apply Intel DPCD workaround when SDP on prior line used

There is Intel specific workaround DPCD address containing workaround for
case where SDP is on prior line. Apply this workaround according to values
in the offset.

Fixes: 61e887329e33 ("drm/i915/xelpd: Handle PSR2 SDP indication in the prior scanline")
Cc: <stable@vger.kernel.org> # v5.15+
Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Link: https://patch.msgid.link/20260515095756.2799483-4-jouni.hogander@intel.com
(cherry picked from commit c3fe899fbeac86ea4a5ca9dd845b2cbc0da46249)
Signed-off-by: Tvrtko Ursulin <tursulin@ursulin.net>
3 months agodrm/i915/psr: Read Intel DPCD workaround register
Jouni Högander [Fri, 15 May 2026 09:57:54 +0000 (12:57 +0300)]
drm/i915/psr: Read Intel DPCD workaround register

Read Intel DPCD workaround register and store it into
intel_connector->dp.psr_caps. psr_caps was chosen as currently it contains
only PSR workaround for PSR2 SDP on prior scanline implementation.

Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Link: https://patch.msgid.link/20260515095756.2799483-3-jouni.hogander@intel.com
(cherry picked from commit c48ff24d0f4ab7ad696b2d35ad64ce7e049c668c)
Signed-off-by: Tvrtko Ursulin <tursulin@ursulin.net>
3 months agodrm/i915/psr: Add defininitions for INTEL_WA_REGISTER_CAPS DPCD register
Jouni Högander [Fri, 15 May 2026 09:57:53 +0000 (12:57 +0300)]
drm/i915/psr: Add defininitions for INTEL_WA_REGISTER_CAPS DPCD register

EDP specification says:

"If either VSC SDP is unable to be transmitted 100 ns before the SU region,
the Source device may optionally transmit the VSC SDP during the prior
video scan line’s HBlank period There is a Intel specific drm dp register
currently containing bits related how TCON can support PSR2 with SDP on
prior line."

Unfortunately many panels are having problems in implementing this. So
there is a custom Intel specific DPCD register (INTEL_WA_REGISTER_CAPS) to
figure out if this is properly implemented on a panel or if panel doesn't
require that 100 ns delay before the SU region. Here are the definitions in
this custom DPCD address:

0 = Panel doesn't support SDP on prior line
1 = Panel supports SDP on prior line
2 = Panel doesn't have 100ns requirement
3 = Reserved

Add definitions for this new register and it's values into new header
intel_dpcd.h.

v2: add INTEL_DPCD_ prefix to definitions

Bspec: 74741
Signed-off-by: Jouni Högander <jouni.hogander@intel.com>
Reviewed-by: Suraj Kandpal <suraj.kandpal@intel.com>
Link: https://patch.msgid.link/20260515095756.2799483-2-jouni.hogander@intel.com
(cherry picked from commit 1da1c9294825f08f622c473480d185680c2a3b75)
Signed-off-by: Tvrtko Ursulin <tursulin@ursulin.net>
3 months agodrm/i915/dp: Fix readback for target_rr in Adaptive Sync SDP
Ankit Nautiyal [Mon, 11 May 2026 12:32:15 +0000 (18:02 +0530)]
drm/i915/dp: Fix readback for target_rr in Adaptive Sync SDP

Correct the bit-shift logic to properly readback the 10 bit target_rr from
DB3 and DB4.

v2: Align the style with readback for vtotal. (Ville)

Fixes: 12ea89291603 ("drm/i915/dp: Add Read/Write support for Adaptive Sync SDP")
Cc: Mitul Golani <mitulkumar.ajitkumar.golani@intel.com>
Cc: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patch.msgid.link/20260511123218.1589830-2-ankit.k.nautiyal@intel.com
(cherry picked from commit f7abc4af2b19240a145a221461dfe756cc01d74a)
Signed-off-by: Tvrtko Ursulin <tursulin@ursulin.net>
3 months agodrm/i915/display: Copy color pipeline from plane in the primary joiner pipe
Chaitanya Kumar Borah [Mon, 11 May 2026 05:32:10 +0000 (11:02 +0530)]
drm/i915/display: Copy color pipeline from plane in the primary joiner pipe

When copying plane color state in a joiner configuration, use the plane in
the primary joiner pipe since it carries the pipeline number selected by
the user-space.

This assumes that all pipes in the joiner are symmetric in their plane
color capabilities.

Cc: stable@vger.kernel.org # v6.19+
Fixes: a78f1b6baf4d ("drm/i915/color: Add framework to program CSC")
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>
Signed-off-by: Chaitanya Kumar Borah <chaitanya.kumar.borah@intel.com>
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Link: https://patch.msgid.link/20260511053213.3122314-2-chaitanya.kumar.borah@intel.com
(cherry picked from commit e8308fb5e05ca08ddfb8b46f6d947a6e3fd80cd7)
Signed-off-by: Tvrtko Ursulin <tursulin@ursulin.net>
3 months agodrm/v3d: Release indirect CSD GEM reference on CPU job free
Maíra Canal [Fri, 15 May 2026 15:07:15 +0000 (12:07 -0300)]
drm/v3d: Release indirect CSD GEM reference on CPU job free

v3d_get_cpu_indirect_csd_params() takes a reference to the indirect BO via
drm_gem_object_lookup() and stashes it in cpu_job->indirect_csd.indirect,
but nothing on the CPU job teardown path ever drops that reference.

Drop the extra reference in v3d_cpu_job_free(). The NULL check covers ioctl
errors before the lookup ran and CPU job types other than
V3D_CPU_JOB_TYPE_INDIRECT_CSD, which leave the field zero-initialised.

Cc: stable@vger.kernel.org
Fixes: 18b8413b25b7 ("drm/v3d: Create a CPU job extension for a indirect CSD job")
Assisted-by: Claude:claude-opus-4.7
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patch.msgid.link/20260515-v3d-cpu-job-leaks-v1-2-7f147cbbf935@igalia.com
Signed-off-by: Maíra Canal <mcanal@igalia.com>
3 months agodrm/v3d: Fix use-after-free of CPU job query arrays on error path
Maíra Canal [Fri, 15 May 2026 15:07:14 +0000 (12:07 -0300)]
drm/v3d: Fix use-after-free of CPU job query arrays on error path

The CPU job ioctl's fail label calls kvfree() on cpu_job's timestamp and
performance query arrays after v3d_job_cleanup(), which drops the job's
last reference and frees cpu_job. Reading cpu_job at that point is a
use-after-free. Also, on the early v3d_job_init() failure path, it is a
NULL dereference, since v3d_job_deallocate() zeroes the local pointer.

In the success path, the arrays are released from the scheduler's
.free_job callback, but on the error path, they are freed manually, as
the job was never pushed to the scheduler. While the success path deals
with this correctly, the fail path doesn't.

On top of that, the manual kvfree() calls only free the array storage;
they don't drm_syncobj_put() the per-query syncobjs that
v3d_timestamp_query_info_free() and v3d_performance_query_info_free()
release on the success path. So the same fail path that triggers the
use-after-free also leaks one syncobj reference per query.

Unify the CPU job teardown into the CPU job's kref destructor, mirroring
v3d_render_job_free(). The scheduler's .free_job slot reverts to the
generic v3d_sched_job_free() and the fail label drops the manual
kvfree() calls, leaving a single teardown path that is reached from both
the scheduler and the ioctl error path. That removes the use-after-free,
the NULL dereference, and the syncobj leak by construction.

Cc: stable@vger.kernel.org
Fixes: 9ba0ff3e083f ("drm/v3d: Create a CPU job extension for the timestamp query job")
Assisted-by: Claude:claude-opus-4.7
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patch.msgid.link/20260515-v3d-cpu-job-leaks-v1-1-7f147cbbf935@igalia.com
Signed-off-by: Maíra Canal <mcanal@igalia.com>
3 months agodrm/gem: Make the GEM LRU lock part of drm_device
Boris Brezillon [Mon, 18 May 2026 11:41:45 +0000 (13:41 +0200)]
drm/gem: Make the GEM LRU lock part of drm_device

Recently, a few races have been discovered in the GEM LRU logic, all
of them caused by the fact the LRU lock is accessed through
gem->lru->lock, and that very same lock also protects changes to
gem->lru, leading to situations where gem->lru needs to first be
accessed without the lock held, to then get the lru to access the lock
through and finally take the lock and do the expected operation.

Currently, the only driver making use of this API (MSM) declares a
device-wide lock, and the user we're about to add (panthor) will
do the same. There's no evidence that we will ever have a driver
that wants different pools of LRUs protected by different locks under
the same drm_device. So we're better off moving this lock to drm_device
and always locking it through obj->dev->gem_lru_mutex, or directly
through dev->gem_lru_mutex.

If anyone ever needs more fine-grained locking, this can be revisited
to pass some drm_gem_lru_pool object representing the pool of LRUs
under a specific lock, but for now, the per-device lock seems to be
enough.

Fixes: e7c2af13f811 ("drm/gem: Add LRU/shrinker helper")
Reported-by: Chia-I Wu <olvaffe@gmail.com>
Closes: https://gitlab.freedesktop.org/panfrost/linux/-/work_items/86
Reviewed-by: Rob Clark <rob.clark@oss.qualcomm.com>
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Link: https://patch.msgid.link/20260518-panthor-shrinker-fixes-v4-1-1920234470d5@collabora.com
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
3 months agodrm/bridge: it66121: acquire reset GPIO in probe
Julien Chauveau [Tue, 24 Mar 2026 19:30:11 +0000 (20:30 +0100)]
drm/bridge: it66121: acquire reset GPIO in probe

The it66121_ctx structure has a gpio_reset field, and it66121_hw_reset()
calls gpiod_set_value() on it. However, the GPIO descriptor is never
acquired via devm_gpiod_get(), leaving gpio_reset as NULL throughout
the driver lifetime.

gpiod_set_value() silently returns when passed a NULL descriptor, so
the hardware reset sequence in it66121_hw_reset() is a no-op. This
leaves the chip in an undefined state at probe time, which can prevent
it from responding on the I2C bus.

The DT binding marks reset-gpios as a required property, so all
compliant device trees provide this GPIO. Add the missing
devm_gpiod_get() call after enabling power supplies and before the
hardware reset, so the chip is properly reset with power applied.

Fixes: 988156dc2fc9 ("drm: bridge: add it66121 driver")
Cc: stable@vger.kernel.org
Signed-off-by: Julien Chauveau <chauveau.julien@gmail.com>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Tested-by: Javier Martinez Canillas <javierm@redhat.com>
Link: https://patch.msgid.link/20260324193011.16583-1-chauveau.julien@gmail.com
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
3 months agoLinux 7.1-rc4
Linus Torvalds [Sun, 17 May 2026 20:59:58 +0000 (13:59 -0700)]
Linux 7.1-rc4

3 months agoMerge tag 'trace-v7.1-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/trace...
Linus Torvalds [Sun, 17 May 2026 19:02:31 +0000 (12:02 -0700)]
Merge tag 'trace-v7.1-rc3' of git://git./linux/kernel/git/trace/linux-trace

Pull tracing fixes from Steven Rostedt:

 - Add more functions to the remote allowed list

   randconfig found more functions that are allowed for the remote code
   for s390 and arm. Add them to the allowed list.

 - Fix remote_test error path

   If one of the simple ring buffers fails to load, the code is supposed
   to rollback its initialized buffers. Instead of rolling back the
   buffers for the failed load, it uses the global variable and rolls
   back all the successfully loaded buffers.

* tag 'trace-v7.1-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
  tracing: Fix desc in error path for the trace remote test module
  ring-buffer remote: Avoid unexpected symbol warnings (arm, s390)

3 months agovirt: sev-guest: Do not use host-controlled page order in cleanup path
Carlos López [Tue, 12 May 2026 10:00:41 +0000 (12:00 +0200)]
virt: sev-guest: Do not use host-controlled page order in cleanup path

When issuing an extended guest request (SVM_VMGEXIT_EXT_GUEST_REQUEST),
get_ext_report() allocates a buffer to retrieve a certificate blob from the
host, keeping track of its size in report_req->certs_len.

However, the host may return SNP_GUEST_VMM_ERR_INVALID_LEN, indicating
an invalid buffer size, as well as the expected length of such buffer.
get_ext_report() subsequently updates report_req->certs_len with the
host-controlled value, and cleans up the buffer by computing a page order
from such value. This is incorrect, as the host-provided length may not
match the page order of the original allocation, potentially resulting
in corruption in the page allocator.

Fix this by using alloc_pages_exact() instead, and reusing @npages to
compute the size passed to free_pages_exact(). For consistency, also
use @npages to compute the size when allocating the pages, even though
this last change has no functional effect.

Fixes: 3e385c0d6ce8 ("virt: sev-guest: Move SNP Guest Request data pages handling under snp_cmd_mutex")
Signed-off-by: Carlos López <clopez@suse.de>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Tested-by: Michael Roth <michael.roth@amd.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 months agoMerge tag 'x86-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 17 May 2026 18:40:18 +0000 (11:40 -0700)]
Merge tag 'x86-urgent-2026-05-17' of git://git./linux/kernel/git/tip/tip

Pull x86 fix from Ingo Molnar:

 - Fix x86 boot crash for non-kjump kexecs (David Woodhouse)

* tag 'x86-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/kexec: Push kjump return address even for non-kjump kexec

3 months agoMerge tag 'timers-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 17 May 2026 18:07:09 +0000 (11:07 -0700)]
Merge tag 'timers-urgent-2026-05-17' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Ingo Molnar:

 - Fix potential garbage reads in the vDSO gettimeofday code
   (Thomas Weißschuh)

* tag 'timers-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  vdso/gettimeofday: Reload sequence counter after switch to time page in do_aux()

3 months agoMerge tag 'sched-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 17 May 2026 17:59:32 +0000 (10:59 -0700)]
Merge tag 'sched-urgent-2026-05-17' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Ingo Molnar:

 - Fix ARM64-specific rseq regressions (Mark Rutland)

* tag 'sched-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  arm64/entry: Fix arm64-specific rseq brokenness

3 months agoMerge tag 'ras-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 17 May 2026 17:50:13 +0000 (10:50 -0700)]
Merge tag 'ras-urgent-2026-05-17' of git://git./linux/kernel/git/tip/tip

Pull MCE fix from Ingo Molnar:

 - Fix an MCE polling interval adjustment regression (Borislav Petkov)

* tag 'ras-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mce: Restore MCA polling interval halving

3 months agoMerge tag 'irq-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 17 May 2026 17:34:15 +0000 (10:34 -0700)]
Merge tag 'irq-urgent-2026-05-17' of git://git./linux/kernel/git/tip/tip

Pull IRQ fixes from Ingo Molnar:

 - Fix use-after-free in irq_work_single() on PREEMPT_RT (Jiayuan Chen)

 - Don't call add_interrupt_randomness() for NMIs in
   handle_percpu_devid_irq() (Mark Rutland)

 - Remove unused function in the ath79-cpu irqchip driver causing LKP
   CI build warnings (Rosen Penev)

 - Fix IRQ allocation/teardown leakage regressions in the GICv5 irqchip
   driver (Sascha Bischoff)

 - Fix an IRQ trigger type regression in the Meson S4 SoC irqchip driver
   (Xianwei Zhao)

 - Fix CPU offlining regression in the RiscV IMSIC irqchip driver
   (Yong-Xuan Wang)

* tag 'irq-urgent-2026-05-17' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irq_work: Fix use-after-free in irq_work_single() on PREEMPT_RT
  irqchip/riscv-imsic: Clear interrupt move state during CPU offlining
  irqchip/meson-gpio: Use the correct register in meson_s4_gpio_irq_set_type()
  irqchip/ath79-cpu: Remove unused function
  genirq/chip: Don't call add_interrupt_randomness() for NMIs
  irqchip/gic-v5: Allocate ITS parent LPIs as a range
  irqchip/gic-v5: Support range allocation for LPIs
  irqchip/gic-v5: Move LPI allocation into the LPI domain

3 months agoMerge tag 'riscv-for-linus-7.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 17 May 2026 16:33:49 +0000 (09:33 -0700)]
Merge tag 'riscv-for-linus-7.1-rc4' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Paul Walmsley:
 "Relatively low-impact fixes. Probably the most notable one is that we
  no longer ask the monitor-mode firmware to delegate misaligned access
  handling to the kernel by default, since the kernel code needs
  significant improvement to match the functionality of the firmware.
  This change avoids functional problems at some cost in performance,
  but shouldn't affect any system with misaligned access handling in
  hardware.

   - Disable satp register probing when no5lvl is specified on the
     kernel command line

   - Fix a CFI-related issue with the misaligned access speed
     measurement code

   - Reduce the CFI shadow stack size limit from 4GB to 2GB (following
     ARM64 GCS)

   - Prevent the kernel from requesting delegation of misaligned access
     faults unless a new Kconfig option, RISCV_SBI_FWFT_DELEGATE_MISALIGNED,
     is enabled. This will depend on CONFIG_NONPORTABLE until the
     deficiencies of the kernel misaligned access fixup code are fixed

   - Fix some potential uninitialized memory accesses in error paths in
     compat_riscv_gpr_set() and compat_restore_sigcontext()

   - Fix a bug in the RISC-V MIPS vendor errata patching code where a
     logical-and was used in place of a bitwise-and

   - Drop some unnecessary code in riscv_fill_hwcap_from_isa_string()

   - Use macros for isa2hwcap indices in riscv_fill_hwcap(), rather than
     open-coding them

   - Fix some documentation typos (one affecting 'make htmldocs')"

* tag 'riscv-for-linus-7.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: misaligned: Make enabling delegation depend on NONPORTABLE
  riscv: Docs: fix unmatched quote warning
  riscv: cfi: reduce shadow stack size limit from 4GB to 2GB
  riscv: cpufeature: Use pre-defined ISA ext macros to index isa2hwcap
  riscv: mm: Fixup no5lvl failure when vaddr is invalid
  riscv: Fix register corruption from uninitialized cregs on error
  riscv: errata: Fix bitwise vs logical AND in MIPS errata patching
  Documentation: riscv: cmodx: fix typos
  riscv: cpufeature: Drop this_hwcap clear in T-Head vector workaround
  riscv: Define __riscv_copy_{,vec_}{words,bytes}_unaligned() using SYM_TYPED_FUNC_START

3 months agoMerge tag 'hwmon-for-v7.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/groec...
Linus Torvalds [Sun, 17 May 2026 16:23:28 +0000 (09:23 -0700)]
Merge tag 'hwmon-for-v7.1-rc4' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon fixes from Guenter Roeck:

 - asus_atk0110, acpi_power_meter: Add missing NULL pointer checks

 - lm90: Fix locking and UAF issues

 - sy7636a: Fix sysfs attribute name in documentation

* tag 'hwmon-for-v7.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (lm90) Add lock protection to lm90_alert
  hwmon: (lm90) Stop work before releasing hwmon device
  docs: hwmon: sy7636a: fix temperature sysfs attribute name
  hwmon: (asus_atk0110) Check ACPI_COMPANION() against NULL
  hwmon: (acpi_power_meter) Check ACPI_COMPANION() against NULL

3 months agodt-bindings: display/msm: Fix typo in clock-names property
Lad Prabhakar [Tue, 28 Apr 2026 10:03:38 +0000 (11:03 +0100)]
dt-bindings: display/msm: Fix typo in clock-names property

Fix the typo "clocks-names" to "clock-names" in the allOf/if conditional
blocks.

Fixes: 9be5c47908e66 ("dt-bindings: display/msm: expand to support MST")
Fixes: 7403e87c13847 ("dt-bindings: display: msm: Fix reg ranges and clocks on Glymur")
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>
Patchwork: https://patchwork.freedesktop.org/patch/721682/
Message-ID: <20260428100338.3179722-1-prabhakar.mahadev-lad.rj@bp.renesas.com>
Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com>
3 months agotracing: Fix desc in error path for the trace remote test module
Vincent Donnefort [Fri, 15 May 2026 20:16:16 +0000 (21:16 +0100)]
tracing: Fix desc in error path for the trace remote test module

During initialisation in remote_test_load(), if one of the
simple_ring_buffer fails to initialise, the error path attempts to
rollback initialised buffers. However, the rollback incorrectly uses the
global pointer to the trace descriptor, which is only set upon
successful load completion. Fix the error path by using the local
pointer to the descriptor.

Link: https://patch.msgid.link/20260515201616.337469-1-vdonnefort@google.com
Fixes: ea908a2b79c8 ("tracing: Add a trace remote module for testing")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Signed-off-by: Vincent Donnefort <vdonnefort@google.com>
base-commit: 5d6919055dec134de3c40167a490f33c74c12581
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
3 months agoMerge tag 'powerpc-7.1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sat, 16 May 2026 16:53:14 +0000 (09:53 -0700)]
Merge tag 'powerpc-7.1-3' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Madhavan Srinivasan:

 - Fix preempt count leak in sysfs show paths

 - Fix error handling in pika_dtm_thread

 - Remove pmac_low_i2c_{lock,unlock}()

 - Enable all windfarms by default

 - Fix dead default for GUEST_STATE_BUFFER_TEST

 - Remove redundant preempt_disable|enable() calls from
   arch_irq_work_raise()

Thanks to Aboorva Devarajan, Ally Heev, Amit Machhiwal, Bart Van Assche,
Christophe Leroy, Christophe Leroy (CS GROUP), Dan Carpenter, Gautam
Menghani, Harsh Prateek Bora, Julian Braha, Krzysztof Kozlowski, Linus
Walleij, Ma Ke, Ritesh Harjani (IBM), and Sayali Patil

* tag 'powerpc-7.1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/time: Remove redundant preempt_disable|enable() calls from arch_irq_work_raise()
  powerpc/hv-gpci: fix preempt count leak in sysfs show paths
  powerpc: fix dead default for GUEST_STATE_BUFFER_TEST
  powerpc/powermac: Remove pmac_low_i2c_{lock,unlock}()
  powerpc/warp: Fix error handling in pika_dtm_thread
  powerpc: 82xx: fix uninitialized pointers with free attribute
  powerpc/g5: Enable all windfarms by default

3 months agoMerge tag 'sound-7.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Sat, 16 May 2026 16:32:30 +0000 (09:32 -0700)]
Merge tag 'sound-7.1-rc4' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A collection of small fixes.  All device-specific small changes:

  HD-audio:
   - Fix NULL pointer dereference in snd_hda_ctl_add()
   - ACPI and Kconfig fixes for Cirrus drivers
   - A regression fix CA0132 codec
   - Various device-specific quirks for HP, Lenovo, Samsung, Framework etc
   - Documentation path fix

  USB-audio:
   - Boundary checks for MIDI endpoint descriptors
   - Offload mapping error handling for Qualcomm
   - A new device quirk for TTGK Technology USB-C Audio
   - A fix for Focusrite Scarlett2 mixer"

* tag 'sound-7.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda/ca0132: Disable auto-detect on manual output select
  ALSA: hda/realtek: Add mute LED quirk for HP Pavilion Laptop 16-ag0xxx
  ALSA: hda/realtek: ALC269 fixup for Lenovo Yoga Pro 7 15ASH111 audio
  ALSA: hda: Fix NULL pointer dereference in snd_hda_ctl_add()
  ALSA: hda/realtek: Add quirk for Samsung Galaxy Book5 360 headphone
  ALSA: hda/cs35l56: Drop malformed default N from Kconfig
  ALSA: hda/realtek: fix mic boost on Framework PTL
  ALSA: hda/realtek: Limit mic boost on Positivo DN50E
  ALSA: doc: cs35l56: Update path to HDA driver source
  ALSA: usb-audio: qcom: Check offload mapping failures
  ALSA: hda/realtek: Fix Legion 7 16ITHG6 speaker amp binding
  ALSA: usb-audio: Add iface reset and delay quirk for TTGK Technology USB-C Audio
  ALSA: scarlett2: Add missing error check when initialise Autogain Status
  ALSA: hda: cs35l41: Put ACPI device on missing physical node
  ALSA: hda: cs35l56: Put ACPI device after setting companion
  ALSA: usb-audio: Bound MIDI 2.0 endpoint descriptor scans
  ALSA: usb-audio: Bound MIDI endpoint descriptor scans
  ALSA: hda/realtek: Add codec SSID quirk for Lenovo Yoga Pro 9 16IMH9 (17aa:38d5)

3 months agohwmon: (lm90) Add lock protection to lm90_alert
Guenter Roeck [Thu, 14 May 2026 21:41:00 +0000 (14:41 -0700)]
hwmon: (lm90) Add lock protection to lm90_alert

Sashiko reports:

lm90_alert() executes in the smbus alert context and calls
lm90_update_confreg() to disable the hardware alert line, without
acquiring hwmon_lock.

Concurrently, sysfs write operations (such as lm90_write_convrate) hold
the hwmon_lock, temporarily modify data->config, and then restore it.

If an alert interrupt occurs concurrently with a sysfs write, the sysfs
path will overwrite the alert handler's modifications to data->config
and the hardware register.

This unintentionally re-enables the hardware alert line while the alarm is
still active, causing an interrupt storm.

Add the missing lock to lm90_alert() to solve the problem.

Fixes: 7a1d220ccb0cc ("hwmon: (lm90) Introduce function to update configuration register")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
3 months agohwmon: (lm90) Stop work before releasing hwmon device
Guenter Roeck [Thu, 14 May 2026 21:31:49 +0000 (14:31 -0700)]
hwmon: (lm90) Stop work before releasing hwmon device

Sashiko reports:

In lm90_probe(), the devm action to cancel the alert_work and report_work
(lm90_restore_conf) is registered in lm90_init_client() before
devm_hwmon_device_register_with_info() is called.

Because devm executes cleanup actions in reverse order during module
unbind or probe failure, the hwmon device is unregistered and freed first.

If lm90_alert_work() or lm90_report_alarms() runs in the window between
the hwmon device being freed and the delayed works being cancelled,
lm90_update_alarms() will dereference the freed data->hwmon_dev here.

Fix the problem by canceling the workers separately after registering
the hwmon device and before registering the interrupt handler. This ensures
that the workers are canceled after interrupts are disabled and before
the hwmon device is released. Add "shutdown" flag to indicate that device
shutdown is in progress to prevent workers from being re-armed.

Fixes: f6d0775119fb9 ("hwmon: (lm90) Rework alarm/status handling")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
3 months agodrm/msm/snapshot: fix dumping of the unaligned regions
Dmitry Baryshkov [Sat, 16 May 2026 11:53:45 +0000 (14:53 +0300)]
drm/msm/snapshot: fix dumping of the unaligned regions

The snapshotting code internally aligns data segment to 16 bytes. This
works fine for DPU code (where most of the regions are aligned), but
fails for snapshotting of the DSI data (because DSI data region is
shifted by 4 bytes). Fix the code by removing length alignment and by
accurately printing last registers in the region. While reworking the
code also fix the 16x memory overallocation in
msm_disp_state_dump_regs().

Fixes: 98659487b845 ("drm/msm: add support to take dpu snapshot")
Reported-by: Salendarsingh Gaud <sgaud@qti.qualcomm.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Patchwork: https://patchwork.freedesktop.org/patch/725449/
Message-ID: <20260516-msm-fix-dsi-dump-2-v2-1-9e49fb2d240e@oss.qualcomm.com>
Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com>
3 months agoMerge tag 'drm-fixes-2026-05-16' of https://gitlab.freedesktop.org/drm/kernel
Linus Torvalds [Sat, 16 May 2026 00:00:45 +0000 (17:00 -0700)]
Merge tag 'drm-fixes-2026-05-16' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Dave Airlie:
 "Weekly fixes pull, small and all over fixes, mostly xe and amdgpu,
  with some ttm and a core fix for the handle change pain.

  core:
   - fix for the fix for the handle change race

  ttm:
   - avoid infinite loop in swap out
   - avoid infinite loop in BO shrinking
   - convert -EAGAIN from dmem_cgroup_try_charge to -ENOSPC

  bridge:
   - imx8qxp-pxl2dpi: avoid ERR_PTR with device_node cleanup

  i915:
   - Skip __i915_request_skip() for already signaled requests
   - Fix VSC dynamic range signaling for RGB formats [dp]

  xe:
   - Madvise fix around purgeability tracking
   - Restore engine mask for specific blitter style
   - Couple UAF fixes
   - Drop unused ggtt_balloon field

  amdgpu:
   - Userq fixes
   - DCN 3.2 fix
   - RAS fix
   - GC 12 fix

  gma500:
   - oaktrail_lvds: fix i2c handling

  loongson:
   - use managed cleanup for connector polling

  panfrost:
   - handle results from reservation locking correctly

  qaic:
   - check for integer overflows in mmap logic

  rocket:
   - handle results from reservation locking correctly"

* tag 'drm-fixes-2026-05-16' of https://gitlab.freedesktop.org/drm/kernel: (26 commits)
  drm: Replace old pointer to new idr
  drm/loongson: Use managed KMS polling
  drm/ttm: Fix ttm_bo_shrink() infinite LRU walk on backup failure
  drm/ttm: Convert -EAGAIN from dmem_cgroup_try_charge to -ENOSPC
  drm/gma500/oaktrail_lvds: fix i2c adapter leaks on init
  drm/gma500/oaktrail_lvds: fix hang on init failure
  drm/gma500/oaktrail_hdmi: fix i2c adapter leak on setup
  drm/xe: Drop unused ggtt_balloon field
  accel/qaic: Add overflow check to remap_pfn_range during mmap
  drm/i915/dp: Fix VSC dynamic range signaling for RGB formats
  drm/i915: skip __i915_request_skip() for already signaled requests
  drm/bridge: imx8qxp-pxl2dpi: avoid ERR_PTR with device_node cleanup
  drm/amdgpu/gfx_v12_0: set gfx.rs64_enable from PFP header on GFX12
  drm/amd/ras: Fix CPER ring debugfs read overflow
  drm/amd/display: Wrap DCN32 phantom-plane allocation in DC_RUN_WITH_PREEMPTION_ENABLED
  drm/amdgpu: fix userq hang detection and reset
  drm/amdgpu: remove almost all calls to amdgpu_userq_detect_and_reset_queues
  drm/amdgpu: rework amdgpu_userq_signal_ioctl v3
  drm/amdgpu: remove deadlocks from amdgpu_userq_pre_reset
  drm/xe/dma-buf: fix UAF with retry loop
  ...

3 months agodrm: Replace old pointer to new idr
Edward Adam Davis [Wed, 13 May 2026 04:30:50 +0000 (12:30 +0800)]
drm: Replace old pointer to new idr

Commit 5e28b7b94408 introduced a logical error by failing to replace the
newly generated IDR pointer to old id's pointer at the correct location
within the "change handle" logic; this resulted in the issue reported by
syzbot [1].

Specifically, the new IDR object pointer is intended to replace the original
id's pointer during the normal execution flow.

Additionally, an unnecessary conditional check for the ret exit path has
been removed.

[1]
!RB_EMPTY_ROOT(&prime_fpriv->dmabufs)
WARNING: drivers/gpu/drm/drm_prime.c:224 at drm_prime_destroy_file_private+0x48/0x60 drivers/gpu/drm/drm_prime.c:224, CPU#0: syz.0.17/5833
Call Trace:
 drm_file_free.part.0+0x7e6/0xcc0 drivers/gpu/drm/drm_file.c:269
 drm_file_free drivers/gpu/drm/drm_file.c:237 [inline]
 drm_close_helper.isra.0+0x186/0x200 drivers/gpu/drm/drm_file.c:290
 drm_release+0x1ab/0x360 drivers/gpu/drm/drm_file.c:438

Fixes: 5e28b7b94408 ("drm: Set old handle to NULL before prime swap in change_handle")
Reported-by: syzbot+d7c9eed171647e421013@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=d7c9eed171647e421013
Cc: stable@vger.kernel.org
Tested-by: syzbot+d7c9eed171647e421013@syzkaller.appspotmail.com
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patch.msgid.link/tencent_C267296443AAA4567771176886DFF364A305@qq.com
3 months agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 15 May 2026 22:40:25 +0000 (15:40 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 MPAM fixes from Catalin Marinas:

 - Fix NULL dereference and a false-positive warning when the driver
   probes hardware with surprising version numbers

 - Fix writing values to the wrong registers when probing
   cache-utilisation counters. Replace 'NRDY' probing with a version
   that is robust for platforms where the bit is writeable by both
   hardware and software

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm_mpam: Check whether the config array is allocated before destroying it
  arm_mpam: Fix false positive assert failure during mpam_disable()
  arm_mpam: Improve check for whether or not NRDY is hardware managed
  arm_mpam: Pretend that NRDY is always hardware managed
  arm_mpam: Fix monitor instance selection when checking for hardware NRDY

3 months agoMerge tag 'iommu-fixes-v7.1-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 15 May 2026 22:22:26 +0000 (15:22 -0700)]
Merge tag 'iommu-fixes-v7.1-rc3' of git://git./linux/kernel/git/iommu/linux

Pull iommu fixes from Joerg Roedel:
 "This is probably the largest fixes pull-request ever sent for IOMMU. I
  partially blame it on AI code review which found some issues but there
  is also some rework in here to fix issues in the iommu parts of PCI
  device reset.

  AMD-Vi:
   - Add bounds checks to debugfs and table lookups

  Intel VT-d:
   - Apply an existing quirk for Q35 graphic device
   - Skip dev_pasid teardown for the blocked domain to avoid
     out-of-bounds access
   - Return early if dev_pasid is missing to prevent NULL dereference
     or UAF

  Core:
   - Fix bugs and corner cases in pci_dev_reset_iommu_prepare/done()
   - Fix various issues found by AI in iommupt code

  MAINTAINERS email address update for RISCV IOMMU"

* tag 'iommu-fixes-v7.1-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/iommu/linux:
  MAINTAINERS: update Tomasz Jeznach's email address
  iommupt: Fix the end_index calculation in __map_range_leaf()
  iommupt: Check for missing PAGE_SIZE in the pgsize_bitmap
  iommu: Handle unmap error when iommu_debug is enabled
  iommu: Fix up map/unmap debugging for iommupt domains
  iommu: Fix loss of errno on map failure for classic ops
  iommu/vt-d: Avoid NULL pointer dereference or refcount corruption
  iommu/vt-d: Fix oops due to out of scope access
  iommu/vt-d: Disable DMAR for Intel Q35 IGFX
  iommu: Warn on premature unblock during DMA aliased sibling reset
  iommu: Fix WARN_ON in __iommu_group_set_domain_nofail() due to reset
  iommu: Fix ATS invalidation timeouts during __iommu_remove_group_pasid()
  iommu: Fix nested pci_dev_reset_iommu_prepare/done()
  iommu: Fix pasid attach in pci_dev_reset_iommu_prepare/done()
  iommu: Replace per-group resetting_domain with per-gdev blocked flag
  iommu: Fix kdocs of pci_dev_reset_iommu_done()
  iommu: Fix NULL group->domain dereference in pci_dev_reset_iommu_done()
  iommu/amd: Bounds-check devid in __rlookup_amd_iommu()
  iommu/amd: Remove latent out-of-bounds access in IOMMU debugfs

3 months agoMerge tag 'vfio-v7.1-rc4' of https://github.com/awilliam/linux-vfio
Linus Torvalds [Fri, 15 May 2026 22:13:02 +0000 (15:13 -0700)]
Merge tag 'vfio-v7.1-rc4' of https://github.com/awilliam/linux-vfio

Pull VFIO fixes from Alex Williamson:

 - Convert vfio-pci BAR resource requests and iomaps initialization
   from a lazy, on-demand model to an eager pre-allocation model to
   avoid races while preserving legacy error behavior.  Fix unchecked
   barmap access in dma-buf export path (Matt Evans)

 - Introduce an implicit unsigned cast in converting vfio-pci device
   offsets to region indexes, closing a potential out-of-bounds
   access through the vfio_pci_ioeventfd() interface (Matt Evans)

 - Fix a dma-buf kref underflow and stuck wait_for_completion() when
   closing a previously revoked dma-buf (Alex Williamson)

* tag 'vfio-v7.1-rc4' of https://github.com/awilliam/linux-vfio:
  vfio/pci: Check BAR resources before exporting a DMABUF
  vfio/pci: Set up BAR resources and maps in vfio_pci_core_enable()
  vfio/pci: Make VFIO_PCI_OFFSET_TO_INDEX() return unsigned
  vfio/pci: fix dma-buf kref underflow after revoke

3 months agoMerge tag 'drm-misc-fixes-2026-05-15' of https://gitlab.freedesktop.org/drm/misc...
Dave Airlie [Fri, 15 May 2026 21:55:49 +0000 (07:55 +1000)]
Merge tag 'drm-misc-fixes-2026-05-15' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-fixes

Short summary of fixes pull:

bridge:
- imx8qxp-pxl2dpi: avoid ERR_PTR with device_node cleanup

gma500:
- oaktrail_lvds: fix i2c handling

loongson:
- use managed cleanup for connector polling

panfrost:
- handle results from reservation locking correctly

qaic:
- check for integer overflows in mmap logic

rocket:
- handle results from reservation locking correctly

ttm:
- avoid infinite loop in swap out
- avoid infinite loop in BO shrinking
- convert -EAGAIN from dmem_cgroup_try_charge to -ENOSPC

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patch.msgid.link/20260515070816.GA88575@2a02-2455-9062-2500-7dec-552d-233d-9fe0.dyn6.pyur.net
3 months agoMerge tag 'v7.1-rc4-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Fri, 15 May 2026 21:52:17 +0000 (14:52 -0700)]
Merge tag 'v7.1-rc4-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull smb client fixes from Steve French:

 - Fix integer overflow in read

 - Fix smbdirect error cleanup

 - Multichannel reconnect fix

 - Add some missing defines and correct some references to protocol spec

 - Fix oob symlink read

* tag 'v7.1-rc4-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  smbdirect: Fix error cleanup in smbdirect_map_sges_from_iter()
  smb: client: avoid integer overflow in SMB2 READ length check
  cifs: client: stage smb3_reconfigure() updates and restore ctx on failure
  smb/client: fix possible infinite loop and oob read in symlink_data()
  SMB3.1.1: add missing QUERY_DIR info levels

3 months agoMerge tag 'ceph-for-7.1-rc4' of https://github.com/ceph/ceph-client
Linus Torvalds [Fri, 15 May 2026 21:48:09 +0000 (14:48 -0700)]
Merge tag 'ceph-for-7.1-rc4' of https://github.com/ceph/ceph-client

Pull ceph fixes from Ilya Dryomov:
 "An important patch from Hristo that squashes a folio reference leak
  that could lead to OOM kills in CephFS and a number of miscellaneous
  fixes from Raphael and Slava.

  All but two are marked for stable"

* tag 'ceph-for-7.1-rc4' of https://github.com/ceph/ceph-client:
  libceph: Fix potential null-ptr-deref in decode_choose_args()
  libceph: handle rbtree insertion error in decode_choose_args()
  libceph: Fix potential out-of-bounds access in osdmap_decode()
  ceph: put folios not suitable for writeback
  ceph: add ceph_has_realms_with_quotas() check to ceph_quota_update_statfs()
  libceph: Fix potential out-of-bounds access in __ceph_x_decrypt()
  ceph: fix BUG_ON in __ceph_build_xattrs_blob() due to stale blob size
  ceph: fix a buffer leak in __ceph_setxattr()
  libceph: Fix unnecessarily high ceph_decode_need() for uniform bucket
  libceph: Fix potential out-of-bounds access in crush_decode()

3 months agoMerge tag 'drm-xe-fixes-2026-05-14' of https://gitlab.freedesktop.org/drm/xe/kernel...
Dave Airlie [Fri, 15 May 2026 21:47:06 +0000 (07:47 +1000)]
Merge tag 'drm-xe-fixes-2026-05-14' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes

- Madvise fix around purgeability tracking (Arvind)
- Restore engine mask for specific blitter style (Roper)
- Couple UAF fixes (Auld)
- Drop unused ggtt_balloon field (Wajdeczko)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patch.msgid.link/agXWkM3Y98bqt6TG@intel.com
3 months agoMerge tag 'for-7.1-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Fri, 15 May 2026 20:22:07 +0000 (13:22 -0700)]
Merge tag 'for-7.1-rc3-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - fixup warning when allocating memory for readahead, __GFP_NOWARN was
   accidentally dropped when setting mapping constraints

 - in tracepoint of file sync, fix sleeping in atomic context when
   handling dentries

 - harden initial loading of block group on crafted/fuzzed images,
   iterate all chunk mapping entries unconditionally

 - fix freeing pages of submitted io after checking for errors

 - fix incorrect inode size after remount when using fallocate KEEP_SIZE
   mode (also requires disabled 'no-holes' feature)

* tag 'for-7.1-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: fix incorrect i_size after remount caused by KEEP_SIZE prealloc gap
  btrfs: only release the dirty pages io tree after successful writes
  btrfs: tracepoints: fix sleep while in atomic context in btrfs_sync_file()
  btrfs: always pass __GFP_NOWARN from add_ra_bio_pages()
  btrfs: fix check_chunk_block_group_mappings() to iterate all chunk maps

3 months agoMerge tag 'xfs-fixes-7.1-rc4' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Fri, 15 May 2026 20:17:46 +0000 (13:17 -0700)]
Merge tag 'xfs-fixes-7.1-rc4' of git://git./fs/xfs/xfs-linux

Pull xfs fixes from Carlos Maiolino:
 "A few bug fixes, nothing really special stands out"

* tag 'xfs-fixes-7.1-rc4' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: Fix typo in comment
  xfs: fix the "limiting open zones" message
  xfs: flush delalloc blocks on ENOSPC in xfs_trans_alloc_icreate
  xfs: check da node block pad field during scrub
  xfs: fix memory leak for data allocated by xfs_zone_gc_data_alloc()
  xfs: fix memory leak on error in xfs_alloc_zone_info()
  xfs: check directory data block header padding in scrub
  xfs: zero directory data block padding on write verification
  xfs: zero entire directory data block header region at init
  xfs: remove the meaningless XFS_ALLOC_FLAG_FREEING

3 months agoMerge tag 'nfsd-7.1-1' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux
Linus Torvalds [Fri, 15 May 2026 20:11:41 +0000 (13:11 -0700)]
Merge tag 'nfsd-7.1-1' of git://git./linux/kernel/git/cel/linux

Pull nfsd fixes from Chuck Lever:
 "Fixes for this release:
   - Correctness fix for the new sunrpc cache netlink protocol

  Marked for stable:
   - Correctness fixes for delegated attributes
   - Prevent an infinite loop when revoking layouts"

* tag 'nfsd-7.1-1' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux:
  NFSD: Fix infinite loop in layout state revocation
  sunrpc: start cache request seqno at 1 to fix netlink GET_REQS
  nfsd: update mtime/ctime on COPY in presence of delegated attributes
  nfsd: update mtime/ctime on CLONE in presense of delegated attributes
  nfsd: fix file change detection in CB_GETATTR
  nfsd: fix GET_DIR_DELEGATION when VFS leases are disabled

3 months agoMerge tag 'block-7.1-20260515' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe...
Linus Torvalds [Fri, 15 May 2026 19:47:00 +0000 (12:47 -0700)]
Merge tag 'block-7.1-20260515' of git://git./linux/kernel/git/axboe/linux

Pull block fixes from Jens Axboe:

 - NVMe merge request via Keith:
     - Fix memory leak on a passthrough integrity mapping failure (Keith)
     - Hide secrets behind debug option (Hannes)
     - Fix pci use-after-free for host memory buffer (Chia-Lin Kao)
     - Fix tcp taregt use-after-free for data digest (Sagi)
     - Revert a mistaken quirk (Alan Cui)
     - Fix uevent and controller state race condition (Maurizio)
     - Fix apple submission queue re-initialization (Nick Chan)

 - Three fixes for blk-integrity, fixing an issue with the user data
   mapping and two problems with recomputing number of segments

 - Two fixes for the iov_iter bounce buffering

 - Fix for the handling of dead zoned write plugs

 - ublk max_sectors validation fix, with associated selftest addition

* tag 'block-7.1-20260515' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux:
  nvme-apple: Reset q->sq_tail during queue init
  block: align down bounces bios
  block: pass a minsize argument to bio_iov_iter_bounce
  selftests: ublk: cap nthreads to kernel's actual nr_hw_queues
  block: fix handling of dead zone write plugs
  block: bio-integrity: Fix null-ptr-deref in bio_integrity_map_user()
  block: recompute nr_integrity_segments in blk_insert_cloned_request
  block: don't overwrite bip_vcnt in bio_integrity_copy_user()
  nvme: fix race condition between connected uevent and STARTED_ONCE flag
  Revert "nvme: add quirk NVME_QUIRK_IGNORE_DEV_SUBNQN for 144d:a808"
  nvmet-tcp: Fix potential UAF when ddgst mismatch
  nvme-pci: fix use-after-free in nvme_free_host_mem()
  nvmet-auth: Do not print DH-HMAC-CHAP secrets
  nvme: fix bio leak on mapping failure
  nvme: make prp passthrough usage less scary
  ublk: reject max_sectors smaller than PAGE_SECTORS in parameter validation

3 months agoMerge tag 'io_uring-7.1-20260515' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 15 May 2026 19:34:02 +0000 (12:34 -0700)]
Merge tag 'io_uring-7.1-20260515' of git://git./linux/kernel/git/axboe/linux

Pull io_uring fixes from Jens Axboe:

 - Small series sanitizing the locking done for either modifying or
   reading a chain of requests

 - If the application has a pid namespace, ensure that the sqthread pid
   is correctly printed in fdinfo

 - Fix for a hashing issue in the io-wq thread pool, which could lead to
   a use-after-free

 - Kill dead argument from io_prep_rw_pi()

 - Fix for a missed validation of the CQ ring head, affecting CQE refill

* tag 'io_uring-7.1-20260515' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux:
  io_uring: validate user-controlled cq.head in io_cqe_cache_refill()
  io-wq: check that the predecessor is hashed in io_wq_remove_pending()
  io_uring/rw: drop unused attr_type_mask from io_prep_rw_pi()
  io_uring: hold uring_lock across io_kill_timeouts() in cancel path
  io_uring: defer linked-timeout chain splice out of hrtimer context
  io_uring: hold uring_lock when walking link chain in io_wq_free_work()
  io_uring/fdinfo: translate SqThread PID through caller's pid_ns

3 months agoMerge tag 'hardening-v7.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/kees...
Linus Torvalds [Fri, 15 May 2026 19:27:03 +0000 (12:27 -0700)]
Merge tag 'hardening-v7.1-rc4' of git://git./linux/kernel/git/kees/linux

Pull hardening fix from Kees Cook:

 - gcc-plugins: Fix GCC 16 removal of CONST_CAST macros

* tag 'hardening-v7.1-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
  gcc-plugins: Always define CONST_CAST_GIMPLE and CONST_CAST_TREE

3 months agoMerge tag 'docs-7.1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/docs...
Linus Torvalds [Fri, 15 May 2026 19:24:09 +0000 (12:24 -0700)]
Merge tag 'docs-7.1-fixes' of git://git./linux/kernel/git/docs/linux

Pull documentation fixes from Jonathan Corbet:
 "This is Willy Tarreau's new document clarifying the definition and
  handling of security-related bugs, which we're trying to get out there
  quickly on the theory that some of the bug reporters might actually
  read and pay attention to it"

* tag 'docs-7.1-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/docs/linux:
  docs: threat-model: don't limit root capabilities to CAP_SYS_ADMIN
  docs: security-bugs: add a link to the threat-model documentation
  Documentation: security-bugs: clarify requirements for AI-assisted reports
  Documentation: security-bugs: explain what is and is not a security bug
  Documentation: security-bugs: do not systematically Cc the security team

3 months agoring-buffer remote: Avoid unexpected symbol warnings (arm, s390)
Arnd Bergmann [Fri, 15 May 2026 10:57:09 +0000 (12:57 +0200)]
ring-buffer remote: Avoid unexpected symbol warnings (arm, s390)

The now more verbose check found more architecture specific symbol
missing from the whitelist, during randconfig testing on s390
and 32-bit arm:

Unexpected symbols in kernel/trace/simple_ring_buffer.o:
         U __aeabi_unwind_cpp_pr1

Unexpected symbols in kernel/trace/simple_ring_buffer.o:
                 U __s390_indirect_jump_r1
                 U __s390_indirect_jump_r10
                 U __s390_indirect_jump_r14
                 U __s390_indirect_jump_r2
                 U __s390_indirect_jump_r5
                 U __s390_indirect_jump_r7
                 U __s390_indirect_jump_r8
                 U __s390_indirect_jump_r9
make[6]: *** [/home/arnd/arm-soc/kernel/trace/Makefile:160: kernel/trace/simple_ring_buffer.o.checked] Error 1

Add these to the list and keep it roughly sorted into sanitizer
and architecture symbols.

Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Marc Zyngier <maz@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Vincent Donnefort <vdonnefort@google.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Link: https://patch.msgid.link/20260515105717.1023007-1-arnd@kernel.org
Fixes: 1211907ac0b5 ("tracing: Generate undef symbols allowlist for simple_ring_buffer")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
3 months agoMerge tag 'for-linus-7.1b-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 15 May 2026 18:24:51 +0000 (11:24 -0700)]
Merge tag 'for-linus-7.1b-rc4-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:

 - one simple cleanup

 - a fix for a corner case when running as Xen PV dom0

 - a fix of a regression for Xen PV guests, introduced in 7.0

* tag 'for-linus-7.1b-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  x86/xen: Tolerate nested XEN_LAZY_MMU entering/leaving
  x86/xen: Fix xen_e820_swap_entry_with_ram()
  xen/arm: Replace __ASSEMBLY__ with __ASSEMBLER__ in interface.h

3 months agoMerge tag 'platform-drivers-x86-v7.1-3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 15 May 2026 18:12:54 +0000 (11:12 -0700)]
Merge tag 'platform-drivers-x86-v7.1-3' of git://git./linux/kernel/git/pdx86/platform-drivers-x86

Pull x86 platform driver fixes from Ilpo Järvinen:

 - asus-nb-wmi:
    - Use existing keyboard quirk for ASUS Zenbook Duo UX8407AA

 - hp-wmi:
    - Add support for Victus 16-r0xxx (8BC2)

 - intel/vsec_tpmi:
    - Move debugfs register before creating devices
    - Prevent fault during unbind

 - lenovo-wmi-*:
    - Fix memory leak in lwmi_dev_evaluate_int()
    - Balance IDA id allocation and free
    - Balance component bind and unbind
    - Prevent sending uninitialized WMI arguments to the device
    - Decouple lenovo-wmi-gamezone and lenovo-wmi-other to simplify
      module dependency graph
    - Limit adding attributes to supported devices

 - samsung-galaxybook:
    - Handle kbd backlight, mic mute and camera block hotkeys

* tag 'platform-drivers-x86-v7.1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86:
  platform/x86: asus-nb-wmi: add DMI quirk for ASUS Zenbook Duo UX8407AA
  platform/x86: lenovo-wmi-other: Limit adding attributes to supported devices
  platform/x86: lenovo-wmi-other: Add Attribute ID helper functions
  platform/x86: lenovo-wmi-helpers: Move gamezone enums to wmi-helpers
  platform/x86: lenovo: Decouple lenovo-wmi-gamezone and lenovo-wmi-other
  platform/x86: lenovo-wmi-other: Fix tunable_attr_01 struct members
  platform/x86: lenovo-wmi-other: Zero initialize WMI arguments
  platform/x86: lenovo-wmi-other: Balance component bind and unbind
  platform/x86: lenovo-wmi-other: Balance IDA id allocation and free
  platform/x86: lenovo-wmi-helpers: Fix memory leak in lwmi_dev_evaluate_int()
  platform/x86: hp-wmi: Add support for Victus 16-r0xxx (8BC2)
  platform/x86/intel/tpmi/plr: Prevent fault during unbind
  platform/x86: intel: Add notifiers support
  platform/x86: intel: Move debugfs register before creating devices
  platform/x86: samsung-galaxybook: Handle ACPI hotkey notifications
  platform/x86: samsung-galaxybook: Refactor camera lens cover input device

3 months agoMerge tag 'v7.1-p4' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Fri, 15 May 2026 17:38:37 +0000 (10:38 -0700)]
Merge tag 'v7.1-p4' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fixes from Herbert Xu:

 - Fix potential dead-lock in rhashtable when used by xattr

 - Avoid calling kvfree on atomic path in rhashtable

* tag 'v7.1-p4' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  rhashtable: Add bucket_table_free_atomic() helper
  mm/slab: Add kvfree_atomic() helper
  rhashtable: drop ht->mutex in rhashtable_free_and_destroy()

3 months agodrm/msm: Fix shrinker deadlock
Daniel J Blueman [Fri, 8 May 2026 06:57:21 +0000 (14:57 +0800)]
drm/msm: Fix shrinker deadlock

With PROVE_LOCKING on an Snapdragon X1 and VM reclaim pressure, we see:

   ======================================================
   WARNING: possible circular locking dependency detected
   7.0.0-debug+ #43 Tainted: G        W
   ------------------------------------------------------
   kswapd0/82 is trying to acquire lock:
   ffff800080ec3870 (reservation_ww_class_acquire){+.+.}-{0:0}, at: msm_gem_shrinker_scan+0x17c/0x400 [msm]

   but task is already holding lock:
   ffffc31709b263b8 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0x88/0x988

   which lock already depends on the new lock.

   the existing dependency chain (in reverse order) is:

   -> #2 (fs_reclaim){+.+.}-{0:0}:
          __lock_acquire+0x4d0/0xad0
          lock_acquire.part.0+0xc4/0x248
          lock_acquire+0x8c/0x248
          fs_reclaim_acquire+0xd0/0xf0
          dma_resv_lockdep+0x224/0x348
          do_one_initcall+0x84/0x5d0
          do_initcalls+0x194/0x1d8
          kernel_init_freeable+0x128/0x180
          kernel_init+0x2c/0x160
          ret_from_fork+0x10/0x20

   -> #1 (reservation_ww_class_mutex){+.+.}-{4:4}:
          __lock_acquire+0x4d0/0xad0
          lock_acquire.part.0+0xc4/0x248
          lock_acquire+0x8c/0x248
          dma_resv_lockdep+0x1a8/0x348
          do_one_initcall+0x84/0x5d0
          do_initcalls+0x194/0x1d8
          kernel_init_freeable+0x128/0x180
          kernel_init+0x2c/0x160
          ret_from_fork+0x10/0x20

   -> #0 (reservation_ww_class_acquire){+.+.}-{0:0}:
          check_prev_add+0x114/0x790
          validate_chain+0x594/0x6f0
          __lock_acquire+0x4d0/0xad0
          lock_acquire.part.0+0xc4/0x248
          lock_acquire+0x8c/0x248
          drm_gem_lru_scan+0x1ac/0x440
          msm_gem_shrinker_scan+0x17c/0x400 [msm]
          do_shrink_slab+0x150/0x4a0
          shrink_slab+0x144/0x460
          shrink_one+0x9c/0x1b0
          shrink_many+0x27c/0x5c0
          shrink_node+0x344/0x550
          balance_pgdat+0x2c0/0x988
          kswapd+0x11c/0x318
          kthread+0x10c/0x128
          ret_from_fork+0x10/0x20

   other info that might help us debug this:
   Chain exists of:
     reservation_ww_class_acquire --> reservation_ww_class_mutex --> fs_reclaim
    Possible unsafe locking scenario:
          CPU0                    CPU1
          ----                    ----
     lock(fs_reclaim);
                                  lock(reservation_ww_class_mutex);
                                  lock(fs_reclaim);
     lock(reservation_ww_class_acquire);

    *** DEADLOCK ***
   1 lock held by kswapd0/82:
    #0: ffffc31709b263b8 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0x88/0x988

   stack backtrace:
   CPU: 4 UID: 0 PID: 82 Comm: kswapd0 Tainted: G        W           7.0.0-debug+ #43 PREEMPT(full)
   Tainted: [W]=WARN
   Hardware name: LENOVO 21BX0016US/21BX0016US, BIOS N3HET94W (1.66 ) 09/15/2025
   Call trace:
    show_stack+0x20/0x40 (C)
    dump_stack_lvl+0x9c/0xd0
    dump_stack+0x18/0x30
    print_circular_bug+0x114/0x120
    check_noncircular+0x178/0x198
    check_prev_add+0x114/0x790
    validate_chain+0x594/0x6f0
    __lock_acquire+0x4d0/0xad0
    lock_acquire.part.0+0xc4/0x248
    lock_acquire+0x8c/0x248
    drm_gem_lru_scan+0x1ac/0x440
    msm_gem_shrinker_scan+0x17c/0x400 [msm]
    do_shrink_slab+0x150/0x4a0
    shrink_slab+0x144/0x460
    shrink_one+0x9c/0x1b0
    shrink_many+0x27c/0x5c0
    shrink_node+0x344/0x550
    balance_pgdat+0x2c0/0x988
    kswapd+0x11c/0x318
    kthread+0x10c/0x128
    ret_from_fork+0x10/0x20

kswapd0 holding fs_reclaim calls the MSM shrinker, which calls
dma_resv_lock. This in turn acquires fs_reclaim.

Fix this deadlock by using dma_resv_trylock() instead, dropping the
subsequently unused passed wait-wound lock 'ticket'.

Cc: stable@vger.kernel.org
Signed-off-by: Daniel J Blueman <daniel@quora.org>
Fixes: fe4952b5f27c ("drm/msm: Convert vm locking")
Patchwork: https://patchwork.freedesktop.org/patch/723564/
Message-ID: <20260508065722.18785-1-daniel@quora.org>
[rob: fixup compile errors, replace lockdep splat with something legible]
Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com>
3 months agodrm/msm: Don't use UTS_RELEASE directly
Uwe Kleine-König (The Capable Hub) [Tue, 28 Apr 2026 14:45:53 +0000 (16:45 +0200)]
drm/msm: Don't use UTS_RELEASE directly

UTS_RELEASE evaluates to a static string and changes quite easily (e.g.
uncommitted changes in the source tree or new commits). So when checking
if a patch introduces changes to the resulting binary each usage of
UTS_RELEASE is source of annoyance.

Instead of using UTS_RELEASE directly use init_utsname()->release which
evaluates to the same string but with that a change of UTS_RELEASE
doesn't affect msm_disp_snapshot_util.o or msm_gpu.o.

Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
Patchwork: https://patchwork.freedesktop.org/patch/721948/
Message-ID: <20260428144553.1103785-2-u.kleine-koenig@baylibre.com>
Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com>
3 months agoALSA: hda/ca0132: Disable auto-detect on manual output select
Matt DeVillier [Thu, 7 May 2026 14:58:41 +0000 (09:58 -0500)]
ALSA: hda/ca0132: Disable auto-detect on manual output select

Commit 778031e1658d ("ALSA: hda/ca0132: Set HP/Speaker
auto-detect default from headphone pin verb") enables HP/Speaker
auto-detect by default when the headphone pin supports presence detect.

With auto-detect enabled, ca0132_select_out() and ca0132_alt_select_out()
choose the output from jack presence instead of the manual HP/Speaker
selection. This means selecting speaker output while headphones are
plugged in updates the control state, but audio still routes to the
headphones.

Treat an explicit manual output selection as a request to leave
auto-detect mode. Clear the HP/Speaker auto-detect switch before applying
the manual selection, and notify userspace so the auto-detect control
state is updated in mixers. Do this for both the normal HP/Speaker
Playback Switch and the alternate Output Select control used by desktop
cards.

This keeps auto-detect enabled by default for devices with jack presence
detection, while preserving the expected behavior that a manual output
choice takes effect immediately.

Fixes: 778031e1658d ("ALSA: hda/ca0132: Set HP/Speaker auto-detect default from headphone pin verb")
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Link: https://lore.kernel.org/CAFTm+6AfeXKf=b2frG4xC5yC4jjM9TkD6c8+dOWWFw6BDjDESw@mail.gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agoALSA: hda/realtek: Add mute LED quirk for HP Pavilion Laptop 16-ag0xxx
Adrien Burnett [Thu, 14 May 2026 16:59:05 +0000 (18:59 +0200)]
ALSA: hda/realtek: Add mute LED quirk for HP Pavilion Laptop 16-ag0xxx

Add a SND_PCI_QUIRK entry for the HP Pavilion Laptop 16-ag0xxx
(subsystem 0x103c:0x8cbc, Realtek ALC245). The
ALC245_FIXUP_HP_X360_MUTE_LEDS fixup is already used by the
neighbouring HP Pavilion Aero Laptop 13-bg0xxx (0x103c:0x8cbd);
it chains the master-mute COEF handler with the GPIO mic-mute
LED handler, which is what this machine needs.

Tested on the affected hardware: both the mute and mic-mute key
LEDs respond correctly to the keyboard hotkeys after this change.

Cc: <stable@vger.kernel.org>
Signed-off-by: Adrien Burnett <an.arctic.pigeon@gmail.com>
Link: https://patch.msgid.link/20260514165905.21175-1-an.arctic.pigeon@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agoALSA: hda/realtek: ALC269 fixup for Lenovo Yoga Pro 7 15ASH111 audio
Jackie Dong [Thu, 14 May 2026 15:39:40 +0000 (23:39 +0800)]
ALSA: hda/realtek: ALC269 fixup for Lenovo Yoga Pro 7 15ASH111 audio

Volume control for the speakers on the Lenovo Yoga Pro 7 15ASH11 laptop
doesn't work.
The DAC routing is the same as on the ThinkPad X1 Gen7 function, so reuse
the alc285_fixup_thinkpad_x1_gen7 to get it working.

Signed-off-by: Jackie Dong <xy-jackie@139.com>
Link: https://patch.msgid.link/20260514153940.7320-1-xy-jackie@139.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agoALSA: hda: Fix NULL pointer dereference in snd_hda_ctl_add()
Quan Sun [Thu, 14 May 2026 13:22:45 +0000 (21:22 +0800)]
ALSA: hda: Fix NULL pointer dereference in snd_hda_ctl_add()

snd_hda_ctl_add() dereferences kctl->id.subdevice without checking
whether kctl is NULL. Multiple callers in sound/hda/codecs/ca0132.c
pass the return value of snd_ctl_new1() directly to snd_hda_ctl_add()
without a NULL check:

    return snd_hda_ctl_add(codec, nid, snd_ctl_new1(&knew, codec));

snd_ctl_new1() returns NULL when the underlying snd_ctl_new() fails
on memory allocation (kzalloc_flex),which can occur under memory
pressure or via fault injection.

Add a NULL check at the entry of snd_hda_ctl_add(), matching the
pattern already used by snd_ctl_add_replace() at the same call
path (sound/core/control.c:515). Return -EINVAL to let callers
handle the error gracefully.

Fixes: 44f0c9782cc6 ("ALSA: hda/ca0132: Add tuning controls")
Signed-off-by: Quan Sun <2022090917019@std.uestc.edu.cn>
Link: https://patch.msgid.link/20260514132245.3062884-1-2022090917019@std.uestc.edu.cn
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agoALSA: hda/realtek: Add quirk for Samsung Galaxy Book5 360 headphone
Markus Kramer [Wed, 13 May 2026 22:28:18 +0000 (00:28 +0200)]
ALSA: hda/realtek: Add quirk for Samsung Galaxy Book5 360 headphone

The Samsung Galaxy Book5 360 (NP750QHA, PCI subsystem ID 0x144d:0xc902)
has severe audio distortion on the 3.5mm headphone jack. Applying
ALC256_FIXUP_SAMSUNG_HEADPHONE_VERY_QUIET corrects the output path
configuration, consistent with fixes already applied to other Samsung
Galaxy Book models using the same ALC256 codec.

Cc: stable@vger.kernel.org
Link: https://github.com/thesofproject/linux/issues/5648
Signed-off-by: Markus Kramer <linux@markus-kramer.de>
Link: https://patch.msgid.link/20260513222818.14351-1-linux@markus-kramer.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agoALSA: hda/cs35l56: Drop malformed default N from Kconfig
Andy Shevchenko [Wed, 13 May 2026 16:27:58 +0000 (18:27 +0200)]
ALSA: hda/cs35l56: Drop malformed default N from Kconfig

First of all, it has to be 'default n' (small letter n), otherwise
it looks for CONFIG_N which is absent and in case of appearance
will enable something unrelated. Second and most important is that
'n' *is* the default 'default' already. Hence just drop malformed
line.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260513162758.365972-1-andriy.shevchenko@linux.intel.com
3 months agoALSA: hda/realtek: fix mic boost on Framework PTL
Daniel Schaefer [Wed, 13 May 2026 15:55:13 +0000 (23:55 +0800)]
ALSA: hda/realtek: fix mic boost on Framework PTL

In addition to the mic jack fix, also need to avoid boosting the
internal mic too much, otherwise >50% input volume clips a lot.

Also add a second SSID. We have one for the classic chassis/speaker and
one for the new Pro chassis/speaker.

To: Jaroslav Kysela <perex@perex.cz>
To: Takashi Iwai <tiwai@suse.com>
To: linux-sound@vger.kernel.org
Cc: Dustin L. Howett <dustin@howett.net>
Cc: linux@frame.work
Signed-off-by: Daniel Schaefer <dhs@frame.work>
Link: https://patch.msgid.link/20260513155513.11683-1-dhs@frame.work
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agoALSA: hda/realtek: Limit mic boost on Positivo DN50E
Edson Juliano Drosdeck [Mon, 11 May 2026 18:15:58 +0000 (15:15 -0300)]
ALSA: hda/realtek: Limit mic boost on Positivo DN50E

The internal mic boost on the Positivo DN50E is too high.
Fix this by applying the ALC269_FIXUP_LIMIT_INT_MIC_BOOST fixup to the machine
to limit the gain.

Signed-off-by: Edson Juliano Drosdeck <edson.drosdeck@gmail.com>
Link: https://patch.msgid.link/20260511181558.670563-1-edson.drosdeck@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agoALSA: doc: cs35l56: Update path to HDA driver source
Richard Fitzgerald [Mon, 11 May 2026 10:41:48 +0000 (11:41 +0100)]
ALSA: doc: cs35l56: Update path to HDA driver source

The HDA drivers were moved to sound/hda/... so update a Documentation
reference that still pointed to the old location.

Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Link: https://patch.msgid.link/20260511104148.36382-1-rf@opensource.cirrus.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agoALSA: usb-audio: qcom: Check offload mapping failures
Cássio Gabriel [Mon, 11 May 2026 04:36:37 +0000 (01:36 -0300)]
ALSA: usb-audio: qcom: Check offload mapping failures

uaudio_transfer_buffer_setup() calls dma_get_sgtable() and then passes
the sg_table to uaudio_iommu_map_xfer_buf() without checking whether sg
table construction succeeded. If dma_get_sgtable() fails, the sg_table
contents are not valid.

uaudio_iommu_map_pa() also ignores iommu_map() failures for the event and
transfer rings and still returns the allocated IOVA to the QMI response.
That can expose an unmapped IOVA to the audio DSP. For transfer rings,
the failed mapping also leaves the IOVA allocator state marked in use.

Check both operations. Free the coherent transfer buffer when sg table
construction fails, free the sg table when transfer-buffer IOMMU mapping
fails, and release the transfer-ring IOVA if iommu_map() fails. Also
return the existing event-ring IOVA when the event ring is already mapped,
matching the pre-split helper behavior.

Fixes: 326bbc348298 ("ALSA: usb-audio: qcom: Introduce QC USB SND offloading support")
Fixes: 44499ecb4f28 ("ALSA: usb: qcom: Fix false-positive address space check")
Cc: stable@vger.kernel.org
Signed-off-by: Cássio Gabriel <cassiogabrielcontato@gmail.com>
Link: https://patch.msgid.link/20260511-alsa-usb-qcom-offload-map-errors-v1-1-6502695e58bc@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
3 months agodrm/loongson: Use managed KMS polling
Myeonghun Pak [Wed, 13 May 2026 06:57:00 +0000 (15:57 +0900)]
drm/loongson: Use managed KMS polling

lsdc_pci_probe() initializes KMS polling before setting up vblank support,
requesting the IRQ and registering the DRM device. If any of those later
steps fails, probe returns without finalizing polling. The driver also
never finalizes polling on regular removal.

Use drmm_kms_helper_poll_init() so polling is tied to the DRM device
lifetime and automatically finalized on probe failure and device removal.

This issue was identified during our ongoing static-analysis research while
reviewing kernel code.

Fixes: f39db26c5428 ("drm: Add kms driver for loongson display controller")
Cc: stable@vger.kernel.org
Co-developed-by: Ijae Kim <ae878000@gmail.com>
Signed-off-by: Ijae Kim <ae878000@gmail.com>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Jianmin Lv <lvjianmin@loongson.cn>
Reviewed-by: Huacai Chen <chenhuacai@loongson.cn>
Signed-off-by: Myeonghun Pak <mhun512@gmail.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patch.msgid.link/20260513065706.23803-1-mhun512@gmail.com
3 months agoALSA: hda/realtek: Fix Legion 7 16ITHG6 speaker amp binding
Nicholas Bonello [Fri, 8 May 2026 22:55:07 +0000 (18:55 -0400)]
ALSA: hda/realtek: Fix Legion 7 16ITHG6 speaker amp binding

The Lenovo Legion 7 16ITHG6 uses codec SSID 17aa:3855, but its PCI
SSID is 17aa:3811.  The latter is now also used by the Legion S7 15IMH05
quirk, which is matched before codec SSID fallback and incorrectly
routes Legion 7 16ITHG6 machines to ALC287_FIXUP_LEGION_15IMHG05_SPEAKERS.

That fixup does not bind the CLSA0101 CS35L41 companion amplifiers,
making the built-in speakers silent even though playback appears to be
active.

Add a codec SSID quirk for 17aa:3855 before the conflicting PCI SSID
quirk so that the Legion 7 16ITHG6 uses ALC287_FIXUP_LEGION_16ITHG6.
This restores CS35L41 firmware loading and binds both speaker
amplifiers.

Fixes: 67f4c61a73e9 ("ALSA: hda/realtek: Add quirk for Legion S7 15IMH")
Cc: stable@vger.kernel.org
Tested-by: Nicholas Bonello <hadobedo@gmail.com>
Assisted-by: Codex:GPT-5
Signed-off-by: Nicholas Bonello <hadobedo@gmail.com>
Link: https://patch.msgid.link/20260508225507.47667-1-hadobedo@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>