]> git.hungrycats.org Git - linux/log
linux
2 months agodrm/amdkfd: Check bounds on CRIU restore event id
David Francis [Tue, 12 May 2026 18:49:41 +0000 (14:49 -0400)]
drm/amdkfd: Check bounds on CRIU restore event id

The valid amdkfd event ids go from 0 to KFD_SIGNAL_EVENT_LIMIT - 1.

During CRIU restore, ensure that the provided event ids are
in that range.

v2: No need for lower bound check since idr_alloc rejects negative
inputs

v3: Also change error message to reflect new error condition

Reviewed-by: David Yat Sin <david.yatsin@amd.com>
Signed-off-by: David Francis <David.Francis@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdkfd: expose pasid of secondary contexts by debugfs
Zhu Lingshan [Wed, 1 Jul 2026 03:03:09 +0000 (11:03 +0800)]
drm/amdkfd: expose pasid of secondary contexts by debugfs

Current kfd debugfs interfaces only expose pasid
of the primary process, this commit exposes
pasid of secondary contexts by debugfs

Just like entries under sysfs,
the secondary contexts are named as
context_<id> under its primary kfd process.

The layout:
/sys/kernel/debug/kfd/proc# tree
.
└── 5802
    ├── context_0
    │   ├── pasid_1025
    │   └── pasid_63266
    ├── context_1
    │   ├── pasid_1025
    │   └── pasid_63266
    ├── pasid_1025
    └── pasid_63266

Another fix is, kfd_debugfs_add_process may fail,
this commit change it to return a meaningful
value other than void

Signed-off-by: Zhu Lingshan <lingshan.zhu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amd/pm: fix smu14 power limit range calculation
Yang Wang [Mon, 6 Jul 2026 01:29:24 +0000 (09:29 +0800)]
drm/amd/pm: fix smu14 power limit range calculation

SMU14 derives the default PPT limit from SocketPowerLimitAc/Dc, but
MsgLimits.Power may expose a different firmware limit for the same PPT0
throttler. Using those values independently as fixed min/max bases can
report an incorrect configurable power range.

Keep the socket power limit as the default value and as the fallback for
current-limit queries. Calculate the reported range from both firmware
values instead, using the lower value as the minimum base and the higher
value as the maximum base before applying OD percentages.

Signed-off-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/gmc6: replace BUG() with an error
Alex Deucher [Mon, 15 Jun 2026 23:14:01 +0000 (19:14 -0400)]
drm/amdgpu/gmc6: replace BUG() with an error

There's no need to crash the kernel for this case.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/si: drop BUG()s
Alex Deucher [Mon, 15 Jun 2026 23:09:39 +0000 (19:09 -0400)]
drm/amdgpu/si: drop BUG()s

There's no need to crash the kernel for these cases.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/sdma3.0: replace BUG() with an error
Alex Deucher [Mon, 15 Jun 2026 23:06:42 +0000 (19:06 -0400)]
drm/amdgpu/sdma3.0: replace BUG() with an error

There's no need to crash the kernel for this case.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/sdma2.4: replace BUG() with an error
Alex Deucher [Mon, 15 Jun 2026 23:06:11 +0000 (19:06 -0400)]
drm/amdgpu/sdma2.4: replace BUG() with an error

There's no need to crash the kernel for this case.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/cik_sdma: replace BUG() with an error
Alex Deucher [Mon, 15 Jun 2026 23:03:27 +0000 (19:03 -0400)]
drm/amdgpu/cik_sdma: replace BUG() with an error

There's no need to crash the kernel for this case.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/ucode: WARN() rather than BUG()
Alex Deucher [Mon, 15 Jun 2026 23:02:37 +0000 (19:02 -0400)]
drm/amdgpu/ucode: WARN() rather than BUG()

There's no need to crash the kernel for this case.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/cgs: WARN() rather than BUG()
Alex Deucher [Mon, 15 Jun 2026 22:58:33 +0000 (18:58 -0400)]
drm/amdgpu/cgs: WARN() rather than BUG()

There's no need to crash the kernel for these cases.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/atomfirmware: WARN() rather than BUG()
Alex Deucher [Mon, 15 Jun 2026 22:57:29 +0000 (18:57 -0400)]
drm/amdgpu/atomfirmware: WARN() rather than BUG()

There's no need to crash the kernel for this case.  Just
return an error.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/gfx12.1: WARN() rather than BUG() for invalid SDMA engine
Alex Deucher [Mon, 15 Jun 2026 22:56:39 +0000 (18:56 -0400)]
drm/amdgpu/gfx12.1: WARN() rather than BUG() for invalid SDMA engine

There's no need to crash the kernel for these cases.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/gfx12: WARN() rather than BUG() for invalid SDMA engine
Alex Deucher [Mon, 15 Jun 2026 22:56:14 +0000 (18:56 -0400)]
drm/amdgpu/gfx12: WARN() rather than BUG() for invalid SDMA engine

There's no need to crash the kernel for these cases.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/gfx11: WARN() rather than BUG() for invalid SDMA engine
Alex Deucher [Mon, 15 Jun 2026 22:55:21 +0000 (18:55 -0400)]
drm/amdgpu/gfx11: WARN() rather than BUG() for invalid SDMA engine

There's no need to crash the kernel for these cases.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: return an error instead of BUG() for CSA bo_va
Alex Deucher [Mon, 15 Jun 2026 22:40:49 +0000 (18:40 -0400)]
drm/amdgpu: return an error instead of BUG() for CSA bo_va

If the bo_va is not present, return an error rather than
crashing the kernel.

Reviewed-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/gfx11: enable gfx pipe1 hardware support
Alex Deucher [Tue, 3 Feb 2026 16:31:48 +0000 (11:31 -0500)]
drm/amdgpu/gfx11: enable gfx pipe1 hardware support

Enable gfx pipe1 hardware support.  This is only available
on gfx11 chips using the F32 microcontroller.  Chips using
the RS64 microcontroller are not able to use the second gfx
pipe.  In practice this means the second pipe is only
available on APUs. This explains the stability issues
Pierre-Eric saw previously with this on Navi33.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/gfx12: Program DB_RING_CONTROL
Alex Deucher [Fri, 26 Jun 2026 20:33:56 +0000 (16:33 -0400)]
drm/gfx12: Program DB_RING_CONTROL

This is needed to allocate occlusion counters across
both gfx pipes. GFX 12 only has one gfx pipe, so no functional
change since the default value is 1.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/gfx11: Program DB_RING_CONTROL
Alex Deucher [Fri, 26 Jun 2026 20:33:40 +0000 (16:33 -0400)]
drm/gfx11: Program DB_RING_CONTROL

This is needed to allocate occlusion counters across
both gfx pipes. No functional change since we only
use one gfx pipe at the moment (default value is 1).

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/gfx10: Program DB_RING_CONTROL
Alex Deucher [Fri, 26 Jun 2026 20:29:13 +0000 (16:29 -0400)]
drm/gfx10: Program DB_RING_CONTROL

This is needed to allocate occlusion counters across
both gfx pipes.

Fixes: b7a1a0ef12b8 ("drm/amd/amdgpu: add pipe1 hardware support")
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdkfd: Check bounds in allocate_event_notification_slot
David Francis [Thu, 21 May 2026 13:18:59 +0000 (09:18 -0400)]
drm/amdkfd: Check bounds in allocate_event_notification_slot

The valid event ids go from 0 to KFD_SIGNAL_EVENT_LIMIT

allocate_event_notification_slot has an option to specify
an event id to allocate at, used by CRIU. We weren't checking
the bounds on that value.

Check them.

v2: Lower bounds check is unecessary because of idr_alloc
already rejecting negative numbers. Upper bounds check should
be KFD_SIGNAL_EVENT_LIMIT since the signal mode mappings might
not yet exist

Signed-off-by: David Francis <David.Francis@amd.com>
Reviewed-by: David Yat Sin <david.yatsin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agoamdkfd: properly free secondary context id
Zhu Lingshan [Wed, 1 Jul 2026 08:08:12 +0000 (16:08 +0800)]
amdkfd: properly free secondary context id

Function kfd_process_free_id() should skip over
the primary kfd process because its context id
is fixed assigned, not allocated through the ida table.
This function should only work on secondary contexts.

Fixes: fac682a1d1af ("amdkfd: identify a secondary kfd process by its id")
Signed-off-by: Zhu Lingshan <lingshan.zhu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdkfd: Don't acquire buffers during CRIU queue restore.
David Francis [Tue, 30 Jun 2026 13:58:18 +0000 (09:58 -0400)]
drm/amdkfd: Don't acquire buffers during CRIU queue restore.

kfd_criu_restore_queue's call of kfd_queue_acquire_buffers was
failing for multiple reasons
- The ctl_stack_size set by the CRIU plugin doesn't match
what is expected by acquire_buffers
- The svm buffer cannot be acquired at this point because
CRIU may not have restored it, or may have restored it
to a different address.

The only reason acquire_buffers was necessary here was to
avoid a null ptr dereference in init_user_queue.

Just put in a check for that dereference; it doesn't appear to
come up in real use cases right now. That is, there is no
usage of CRIU with shared MES.

This is a partial revert of
commit 20a5e7ffdfec ("drm/amdkfd: Properly acquire queue buffers in CRIU restore")

Fixes: 20a5e7ffdfec ("drm/amdkfd: Properly acquire queue buffers in CRIU restore")
Reviewed-by: David Yat Sin <david.yatsin@amd.com>
Signed-off-by: David Francis <David.Francis@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdkfd: Check bounds on CRIU restore queue type and mqd size
David Francis [Mon, 6 Jul 2026 14:19:04 +0000 (10:19 -0400)]
drm/amdkfd: Check bounds on CRIU restore queue type and mqd size

We weren't checking whether the values provided in the private
data in kfd CRIU restore were within bounds.

For queue type, add a KFD_QUEUE_TYPE_MAX and ensure the provided
type is less than it.

For mqd_size, add new function mqd_size_from_queue_type and confirm
that the provided mqd_size matches expectations.

Reviewed-by: David Yat Sin <david.yatsin@amd.com>
Signed-off-by: David Francis <David.Francis@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: fix lifetime issue of amdgpu_vm_get_task_info_pasid()
Shahyan Soltani [Mon, 6 Jul 2026 12:15:21 +0000 (08:15 -0400)]
drm/amdgpu: fix lifetime issue of amdgpu_vm_get_task_info_pasid()

The vm pointer returned from amdgpu_vm_get_vm_from_pasid() is only
valid while the lock is still being held. Once xa_unlock_irqrestore is
called and returned, the pointer is no longer under lock and is subject
to modification. Since, the caller still dereferences vm->task_info in
amdgpu_vm_get_task_info_vm() after the lock is removed, this causes a
use after unlock problem.

Remove the lifetime issue present in amdgpu_vm_get_task_info_pasid()
through removing the amdgpu_vm_get_vm_from_pasid() function from
amdgpu_vm.c and making the relevant code inline to hold the lock while
it is still in use.

Signed-off-by: Shahyan Soltani <shahyan.soltani@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: include amdgpu_video_codecs.h only where needed
Shahyan Soltani [Mon, 8 Jun 2026 17:42:21 +0000 (13:42 -0400)]
drm/amdgpu: include amdgpu_video_codecs.h only where needed

Remove #include "amdgpu_video_codecs.h" from amdgpu.h and add forward declaration
of struct amdgpu_video_codecs.

Add #include "amdgpu_video_codecs.h" into files amdgpu_kms.c, amdgpu_virt.c,
cik.c, nv.c, si.c, soc15.c, soc21.c, soc24.c, soc_v1_0.c, and vi.c.

Signed-off-by: Shahyan Soltani <shahyan.soltani@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: move amdgpu_allowed_register_entry into amdgpu_reg_access.h
Shahyan Soltani [Wed, 3 Jun 2026 15:28:54 +0000 (11:28 -0400)]
drm/amdgpu: move amdgpu_allowed_register_entry into amdgpu_reg_access.h

Move struct amdgpu_allowed_register_entry from monolithic amdgpu.h file into
existing amdgpu_reg_access.h file.

This is part of the ongoing effort to reduce the size of amdgpu.h into their own respective
separate headers.

Signed-off-by: Shahyan Soltani <shahyan.soltani@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: move amdgpu_acpi helpers into new header
Shahyan Soltani [Wed, 3 Jun 2026 15:18:59 +0000 (11:18 -0400)]
drm/amdgpu: move amdgpu_acpi helpers into new header

Move struct amdgpu_uma_carveout_option, struct amdgpu_uma_carveout_info,
struct amdgpu_numa_info, and relevant acpi helpers from the monolithic
amdgpu.h header file into a new amdgpu_acpi.h file.

This is part of the ongoing effort to reduce the size of amdgpu.h into their own respective
separate headers.

Signed-off-by: Shahyan Soltani <shahyan.soltani@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amd/display: Fix sign mismatch warning
George Zhang [Fri, 3 Jul 2026 13:44:45 +0000 (09:44 -0400)]
drm/amd/display: Fix sign mismatch warning

Using mismatched signedness (int and uint32_t) causes a -Wsign-compare
warning. Fix it by changing the min macro to min_t to explicitly cast.

Fixes: 8cbe3648aa86 ("drm/amd/display: clamp DMUB AUX reply length to payload buffer")
Signed-off-by: George Zhang <george.zhang@amd.com>
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: move struct amdgpu_mqd and helpers into header file
Shahyan Soltani [Wed, 3 Jun 2026 13:57:18 +0000 (09:57 -0400)]
drm/amdgpu: move struct amdgpu_mqd and helpers into header file

Move struct amdgpu_mqd_prop, struct amdgpu_mqd, and helpers from the monolithic amdgpu.h
into existing amdgpu_mes.h file.

This is part of the ongoing effort to reduce the size of amdgpu.h into their own respective
separate headers.

Signed-off-by: Shahyan Soltani <shahyan.soltani@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: move struct amdgpu_video_codecs and helpers into header file
Shahyan Soltani [Wed, 3 Jun 2026 13:43:12 +0000 (09:43 -0400)]
drm/amdgpu: move struct amdgpu_video_codecs and helpers into header file

Move struct amdgpu_video_codec_info, struct amdgpu_video_codecs, and helpers into a new
amdgpu_video_codecs.h file.

This is part of the ongoing effort to reduce the size of amdgpu.h into their own respective
separate headers.

Signed-off-by: Shahyan Soltani <shahyan.soltani@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: move struct amdgpu_wb and helpers into separate files
Shahyan Soltani [Wed, 3 Jun 2026 13:23:04 +0000 (09:23 -0400)]
drm/amdgpu: move struct amdgpu_wb and helpers into separate files

Move struct amdgpu_wb and helpers out of the monolithic header amdgpu.h
into its own dedicated header amdgpu_wb.h.

Add amdgpu_wb_init() and amdgpu_wb_fini() into amdgpu_wb.h.

Move functions amdgpu_device_wb_get(), amdgpu_device_wb_free(),
amdgpu_device_wb_init(), and amdgpu_device_wb_fini() out of
amdgpu_device.c into new dedicated amdgpu_wb.c file.

Removed static from functions amdgpu_device_wb_init() and amdgpu_device_wb_fini().

Rename functions amdgpu_device_wb_get(), amdgpu_device_wb_free(),
amdgpu_device_wb_init(), and amdgpu_device_wb_fini() into
amdgpu_wb_get(), amdgpu_wb_free(), amdgpu_wb_init(), and amdgpu_wb_fini().

Update amdgpu/Makefile to build amdgpu_wb.o.

This is part of the ongoing effort to reduce the size of amdgpu.h into
their own respective separate headers.

Signed-off-by: Shahyan Soltani <shahyan.soltani@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/mes12: Remove MES self test
Srinivasan Shanmugam [Fri, 3 Jul 2026 12:48:26 +0000 (18:18 +0530)]
drm/amdgpu/mes12: Remove MES self test

The MES self test is no longer needed. Other MES versions already
dropped their self tests since IGT now covers this functionality.

Remove the MES v12 self test as well.

Cc: Alex Deucher <alexander.deucher@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: trigger GPU recovery when userq destroy fails to unmap a hung queue
Jesse Zhang [Wed, 1 Jul 2026 10:17:03 +0000 (18:17 +0800)]
drm/amdgpu: trigger GPU recovery when userq destroy fails to unmap a hung queue

Destroying a hung user queue issues a MES REMOVE_QUEUE that times out,
The destroy path only logged the error and freed the queue, so the
next userq submission failed and forced a GPU reset attributed to an innocent workload.

Kick the userq reset work when unmap fails so the GPU is recovered at
destroy time.

Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amd/amdgpu: disable ASPM on VI if pcie dpm is disabled
Kenneth Feng [Thu, 25 Jun 2026 09:48:22 +0000 (17:48 +0800)]
drm/amd/amdgpu: disable ASPM on VI if pcie dpm is disabled

Disable ASPM on VI if PCIE dpm is disabled.

Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/5370
Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: retire legacy deferred error separate logging
Ce Sun [Wed, 25 Mar 2026 04:06:48 +0000 (12:06 +0800)]
drm/amdgpu: retire legacy deferred error separate logging

Remove the legacy logic that logs deferred errors separately

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: retire legacy get_retire_flip_bits interface for UMC
Ce Sun [Wed, 25 Mar 2026 03:11:03 +0000 (11:11 +0800)]
drm/amdgpu: retire legacy get_retire_flip_bits interface for UMC

Remove the legacy general get_retire_flip_bits interface for UMC

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: retire legacy get_retire_flip_bits for UMC
Ce Sun [Fri, 3 Apr 2026 03:08:29 +0000 (11:08 +0800)]
drm/amdgpu: retire legacy get_retire_flip_bits for UMC

Remove the legacy get_retire_flip_bits implementation for UMC v12

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: retire legacy mca umc status check interface
Ce Sun [Wed, 25 Mar 2026 02:47:38 +0000 (10:47 +0800)]
drm/amdgpu: retire legacy mca umc status check interface

Remove the legacy interface to check mca umc status

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amd/pm: retire legacy ras_smu_drv interface for smu v13.0.12
Ce Sun [Wed, 25 Mar 2026 02:37:04 +0000 (10:37 +0800)]
drm/amd/pm: retire legacy ras_smu_drv interface for smu v13.0.12

Remove the legacy ras_smu_drv interface implementation for SMU v13.0.12

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/mes: Fix hung_queue_db_array loop limit for multi-XCC
Geoffrey McRae [Sun, 28 Jun 2026 05:37:02 +0000 (15:37 +1000)]
drm/amdgpu/mes: Fix hung_queue_db_array loop limit for multi-XCC

The loop iterated only AMDGPU_MAX_MES_PIPES times, leaving entries
uninitialized for multi-XCC GPUs. This causes null pointer dereferences
when accessing arrays indexed by XCC ID >= 2. Extend the loop to cover
all XCCs (AMDGPU_MAX_MES_PIPES * num_xcc), matching other per-XCC
arrays.

Fixes: a132fc9bc2f8 ("drm/amdgpu: Fixup boost mes detect hang array size")
Signed-off-by: Geoffrey McRae <geoffrey.mcrae@amd.com>
Reviewed-by: Amber Lin <amber.lin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdkfd: use iosys_map for CWSR buffer access
James Zhu [Sun, 28 Jun 2026 13:02:07 +0000 (09:02 -0400)]
drm/amdkfd: use iosys_map for CWSR buffer access

After moving TBA/TMA from GTT to VRAM for GFX9.4.2+ in commit
5088a1ba6d6d ("drm/amdkfd: move TBA/TMA from system to device memory"),
direct pointer dereferences to CWSR buffers became unsafe because VRAM
is accessed via MMIO (PCI BAR mappings).

Direct writes like 'tma[2] = enabled' and memcpy() can fail or
produce incorrect results on non-x86 architectures because:
- MMIO requires specific accessor functions (writeq/readq)
- Compiler optimizations may generate invalid instruction sequences
- No guarantee of proper memory barriers or atomic access

This patch converts CWSR buffer access to use struct iosys_map,
which automatically handles both system memory (GTT) and MMIO
(VRAM) correctly by:
- Using writeq/writel/memcpy_toio for MMIO regions
- Using WRITE_ONCE/memcpy for system memory
- Providing proper memory barriers and access guarantees

Changes:
- Replace void *cwsr_kaddr with struct iosys_map cwsr_map
- Detect MMIO vs system memory using TTM_BO_MAP_IOMEM_MASK
- Use iosys_map_wr() for writing trap handler addresses and flags
- Use iosys_map_memcpy_to() for copying CWSR ISA code

This ensures correct operation on all architectures while maintaining
backward compatibility with older GPUs and APUs that use GTT.

Assisted-by: Claude:Claude-Opus-4.6
Signed-off-by: James Zhu <James.Zhu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/mes: Add NULL check for mes_hung_db_array allocation
Geoffrey McRae [Sun, 28 Jun 2026 06:17:01 +0000 (16:17 +1000)]
drm/amdgpu/mes: Add NULL check for mes_hung_db_array allocation

kcalloc but does not check for failure. If the allocation fails, the
pointer remains NULL but the function returns success. Subsequent code
using this buffer will dereference a NULL pointer, causing a kernel
oops. Add a check to return -ENOMEM if the allocation fails.

Signed-off-by: Geoffrey McRae <geoffrey.mcrae@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/jpeg: deduplicate jpeg_v5_3_0 process_interrupt
Tiago Dourado [Fri, 26 Jun 2026 02:59:13 +0000 (23:59 -0300)]
drm/amdgpu/jpeg: deduplicate jpeg_v5_3_0 process_interrupt

The jpeg_v5_3_0_process_interrupt function is identical to
jpeg_v5_0_0_process_interrupt. Remove the duplicate implementation
in jpeg_v5_3_0 and assign the jpeg_v5_0_0 version directly to the irq_funcs struct.
Export jpeg_v5_0_0_process_interrupt through jpeg_v5_0_0.h to allow
cross-version reuse.

Signed-off-by: Tiago Dourado <tiagodourado@usp.br>
Co-developed-by: Luiz Fernandes <luiz.f.f.fernandes@usp.br>
Signed-off-by: Luiz Fernandes <luiz.f.f.fernandes@usp.br>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu/jpeg: deduplicate jpeg_v3_0 process_interrupt
Tiago Dourado [Fri, 26 Jun 2026 02:59:12 +0000 (23:59 -0300)]
drm/amdgpu/jpeg: deduplicate jpeg_v3_0 process_interrupt

The jpeg_v3_0_process_interrupt function is identical to
jpeg_v2_0_process_interrupt. Remove the duplicate implementation
in jpeg_v3_0 and assign the jpeg_v2_0 version directly to the irq_funcs struct.
Export jpeg_v2_0_process_interrupt through jpeg_v2_0.h to allow
cross-version reuse.

Signed-off-by: Tiago Dourado <tiagodourado@usp.br>
Co-developed-by: Luiz Fernandes <luiz.f.f.fernandes@usp.br>
Signed-off-by: Luiz Fernandes <luiz.f.f.fernandes@usp.br>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amd/pm: retire legacy smu ras driver framework
Ce Sun [Fri, 3 Apr 2026 03:06:05 +0000 (11:06 +0800)]
drm/amd/pm: retire legacy smu ras driver framework

Remove the legacy smu ras driver framework

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: retire legacy pmfw eeprom support check
Ce Sun [Fri, 3 Apr 2026 02:58:22 +0000 (10:58 +0800)]
drm/amdgpu: retire legacy pmfw eeprom support check

Remove the legacy function to check pmfw eeprom support

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: Disable JDPG on VCN5_3
Suresh Guttula [Fri, 26 Jun 2026 06:09:51 +0000 (11:39 +0530)]
drm/amdgpu: Disable JDPG on VCN5_3

JDPG does not support on VCN5

This patch will disable JDPG, because DPG is not correctly
copying the JRBC Read/Write Pointers (R/WPTR) from the PG
(Power Gating) block to JRBC.

Signed-off-by: Suresh Guttula <suresh.guttula@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: add support for SMU version 15.0.9
Kanala Ramalingeswara Reddy [Wed, 17 Jun 2026 12:34:53 +0000 (18:04 +0530)]
drm/amdgpu: add support for SMU version 15.0.9

Initialize SMU Version 15_0_9

Signed-off-by: Kanala Ramalingeswara Reddy <Kanala.RamalingeswaraReddy@amd.com>
Signed-off-by: Granthali Vinodkumar Dhandar <granthali.vinodkumardhandar@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: add support for PSP version 15.0.9
Kanala Ramalingeswara Reddy [Wed, 17 Jun 2026 12:34:36 +0000 (18:04 +0530)]
drm/amdgpu: add support for PSP version 15.0.9

Initialize PSP Version 15_0_9

Signed-off-by: Kanala Ramalingeswara Reddy <Kanala.RamalingeswaraReddy@amd.com>
Signed-off-by: Granthali Vinodkumar Dhandar <granthali.vinodkumardhandar@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdgpu: retire legacy pmfw eeprom interface wrapper
Ce Sun [Fri, 3 Apr 2026 02:55:22 +0000 (10:55 +0800)]
drm/amdgpu: retire legacy pmfw eeprom interface wrapper

retire legacy pmfw eeprom interface wrapper functions

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Ce Sun <cesun102@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agodrm/amdkfd: drop duplicate kfd_queue_acquire_buffers()
Alex Deucher [Wed, 8 Jul 2026 18:18:36 +0000 (14:18 -0400)]
drm/amdkfd: drop duplicate kfd_queue_acquire_buffers()

Duplicated code from the merge.  Drop it.

Fixes: 0461ba9a7994 ("Merge tag 'amd-drm-next-7.3-2026-07-02' of https://gitlab.freedesktop.org/agd5f/linux into drm-next")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 months agoMerge tag 'amd-drm-next-7.3-2026-07-02' of https://gitlab.freedesktop.org/agd5f/linux...
Dave Airlie [Tue, 7 Jul 2026 06:47:08 +0000 (16:47 +1000)]
Merge tag 'amd-drm-next-7.3-2026-07-02' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

[airlied: had to reapply
drm/amdgpu: Implement "color format" DRM property by hand to amdgpu_dm_connector.c]

amd-drm-next-7.3-2026-07-02:

amdgpu:
- Queue reset updates
- Initial compute pipe reset support
- Improved boundary checking for bios parsing
- Cleaned up sysfs input parsing
- devcoredump fixes
- RAS updates and rework
- VCN secure submission fixes
- 8K panel fix
- Add display KUnit tests
- Display CRC fixes
- UserQ updates
- Backlight fixes
- Parse panel type info from DisplayID
- Align IP discovery to pci device lifetime
- IOCTL boundary check fixes
- Convert amdgpu_vm_lock_by_pasid() to drm_exec
- Ctx fixes and cleanup
- SOC15 register macro cleanups
- Memory placement fixes for UVD
- Disable KQ support for MI3xx
- GFX9 mode2 reset fix
- BO list cleanup
- Soc24 aborted suspend fix
- Gfx8 soft reset rework
- Enable soft reset on gfx8
- Drop unnecessary BUG() and BUG_ON() in error paths
- Fix power reporting unit conversion
- Improve vbios command table bounds checking
- UVD bounds checking improvements
- VCN bounds checking improvements
- PSR and replay fixes
- DCN 4.2 updates
- Colorop updates
- DC GPIO rework
- ACP fixes
- Fix aperture mapping leak
- Ignore_damage_clips fix
- Fixes for non-4K pages
- JPEG idle check fixes
- Userptr fixes
- GPUVM fixes
- GC 11.7 updates
- SMU 13 fixes

amdkfd:
- Initial compute pipe reset support
- Allow applications to opt out of sigbus on fatal errors
- Fix doorbell/mmio BO cleanup
- Improved CRIU boundary checking
- MQD handling rework
- SMI fixes
- Reset event fixes
- CRIU fixes
- Sysfs teardown fixes
- IOCTL boundary check fixes
- SVM fixes
- Soft IH ring fixes
- Move TBA/TMA from system to device memory

radeon:
- Blit fix for large BOs
- r600 dpm cleanup fix

drm:
- Extract EDID base section header processing into helper
- Parse panel type from DisplayID 2.x Display Parameters

UAPI:
- KFD interface for applications to select sigbus behavior on fatal errors
  Proposed userspace: https://github.com/ROCm/rocm-systems/pull/6190

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patch.msgid.link/20260702141515.67919-1-alexander.deucher@amd.com
Signed-off-by: Dave Airlie <airlied@redhat.com>
2 months agoMerge tag 'drm-xe-next-2026-07-03' of https://gitlab.freedesktop.org/drm/xe/kernel...
Dave Airlie [Tue, 7 Jul 2026 05:19:26 +0000 (15:19 +1000)]
Merge tag 'drm-xe-next-2026-07-03' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-next

UAPI Changes:

- Add additional error components to xe drm_ras (Riana)
- Drop 'force_execlist' module parameter (Roper)

Cross-subsystem Changes:

- Perf events: Export perf_allow_{cpu,tracepoint} to be used by Xe (John)

Display Changes:

- Skip FORCE_WC and vm_bound check for external dma-bufs (Auld)

Driver Changes:

- Gate observation streams with perf_allow_cpu (John Hubbard)
- Documentation updates and fixes (Michal, Zhan)
- Remove unreleased NVL-S GuC (Daniele)
- Fix pcode init path (Michal)
- RTP fixes and improvements (Gustavo, Violet, Thomas, Roper)
- TLB invalidation fixes and improvements (Tilak)
- PXP detachment from HuC for newer platforms (Daniele)
- Multi-queue fix (Niranjana)
- Improve Kconfig.profile help (Rodrigo)
- Xe_drm_ras and hw_error updates and fixes (Raag, Riana)
- NVL-S updated PCI-IDs and W/a (Gustavo, Nitin)
- Fix dma_fence refcound (Wentao)
- Madvise: optimize invalidation path (Arvind)
- Fix a infinite gt-reset loop in the timeout recovery (Rodrigo)
- Fix wa_oob codegen recipe for external module builds (Thomas)
- Avoid global forcewake in cycle query path (Xin)
- Update TTM device benefical_order (Brost)
- Fix buffer overflow in guc capture (Tejas)
- Page-table fixes and improvements (Brian, Francois, Auld, Brost)
- Removing redundant check (Lu)
- General MCR and MMIO clean-up and improvements (Michal)
- Don't whitelist OA registers unconditionally (Ashutosh)
- SVM error return fix (Brost)
- Userptr fix and small related clean-ups (Shuicheng)
- Don't attempt to process FAST_REQ or EVENT relays on PF (Michal)
- Couple fdinfo improvements (Auld)
- Drop manual VF check on i2c (Raag)
- Probe info outside of xe_info_init functions (Gustavo)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patch.msgid.link/akf7xr96MI4Rd6Qj@intel.com
2 months agoMerge tag 'drm-intel-next-2026-07-02' of https://gitlab.freedesktop.org/drm/i915...
Dave Airlie [Tue, 7 Jul 2026 01:15:46 +0000 (11:15 +1000)]
Merge tag 'drm-intel-next-2026-07-02' of https://gitlab.freedesktop.org/drm/i915/kernel into drm-next

drm/i915 feature pull for v7.3:

Features and functionality:
- Novalake (NVL, display versio 35) Common Mode Timing Generator (CMTG) enabling (Animesh)
- NVL DC3CO enabling (Dibin)

Refactoring and cleanups:
- Refactor and document DP link recovery and capability tracking (Imre)
- i915 and xe driver display probe/remove/suspend/resume/shutdown cleanup and unification (Jani)
- Clean up bandwidth/SAGV code (Ville)
- Streamline Pre/Post-CSC LUT loops (Jani)
- Remove unused TGL DC3DO support (Dibin)
- Make read-only array rates static const (Colin Ian King)
- CDCLK sanitization improvements and cleanups (Ville)
- Break i915 and xe panic dependency on struct intel_framebuffer (Jani)
- GPIO pin renames and cleanups (Ville)
- Unify generic irq handler error messages (Jonathan)

Fixes:
- Various color pipeline fixes (Chaitanya)
- Fix step discontinuity in Pre/Post-CSC Gamma LUT (Sean Paul)
- Improve refresh rate changes without full modeset on VRR capable eDP (Ville)
- Display suspend/resume fixes (Imre)
- Fix LPE audio irq for PREEMPT-RT (Maarten)
- Fix HDMI scrambling enable (Jerome Tollet)
- LNL bandwidth buddy programming update (Vinod)
- Streamline display register wait timeouts (Ville)
- Fix DP MSA VTotal (Mitul)
- Fix LTPHY SSC enabling (Suraj)
- Fix LOBF requirements with optimized guardband (Ankit)
- Avoid full modeset for LRR vsync changes (Ankit)
- Fix vtotal-vsync distance when adjusting vtotal for lower refresh rate (Ankit)
- Extend VRR safe window wait for default VRR timing generator usage (Ankit)

Merges:
- Backmerge for v7.2-rc1 (Jani)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Jani Nikula <jani.nikula@intel.com>
Link: https://patch.msgid.link/934cacdbf2a37856eb5cd6337d8774b449405bd8@intel.com
2 months agoMerge tag 'drm-misc-next-2026-07-02' of https://gitlab.freedesktop.org/drm/misc/kerne...
Dave Airlie [Mon, 6 Jul 2026 23:33:58 +0000 (09:33 +1000)]
Merge tag 'drm-misc-next-2026-07-02' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-next

drm-misc-next for 7.3:

UAPI Changes:

Cross-subsystem Changes:

Core Changes:
- bridge: Add atomic_create_state callback and helpers, drop atomic_reset
- dp: Add support for DSC max delta BPP
- edid: Parse panel type from DisplayID 2.x Display Parameters
- sysfb: Improve panel, stride and framebuffer size validation

Driver Changes:
- hibmc: Improvements to the plane formats handling, switch to gem-shmem
- nouveau: race fixes, misc improvements

- bridges:
  - Convert all bridges to atomic_create_state

- panels:
  - panel-edp: New quirks for BOE NE160QDM-NY1, MB116AS01

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@redhat.com>
Link: https://patch.msgid.link/20260702-powerful-successful-raptor-eded34@houat
2 months agoMerge tag 'drm-misc-next-2026-06-19' of https://gitlab.freedesktop.org/drm/misc/kerne...
Dave Airlie [Mon, 6 Jul 2026 07:38:33 +0000 (17:38 +1000)]
Merge tag 'drm-misc-next-2026-06-19' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-next

drm-misc-next for 7.3:

UAPI Changes:
 - connector: Add color format property

Cross-subsystem Changes:

 - dmem: introduce a peak file

Core Changes:
 - atomic: Add create_state callback and helpers to all objects, add
   documentation on atomic commit lifetime
 - buddy: Fix use-after-free, add per-order free and used block
   scoreboards
 - gem: Remove DRIVER_GEM_GPUVA feature flag
 - hdmi: Hook the color format property in the helpers
 - mipi-dsi: Add MIPI_DSI_MODE_DSC_ALL_SLICES_IN_PKT flag
 - sched: Add test suite for concurrent job submissions
 - virtio: Add support for saving and restoring virtio_gpu_objects,
   abort virtqueue wait on device removal to avoid hung task

Driver Changes:
 - amdgpu: Implement "color format" DRM property
 - amdxdna: Disable device buffer exporting
 - ethosu: Add performance counter support
 - msm: Support DSC configurations with slice_per_pkt > 1
 - mxsfb: Fix disable sequence
 - panthor: Support sparse mappings
 - rockchip: Support YUV background color, Fix layer config timeout, add
   edp support for rk3576, cleanups and formats improvements
 - solomon: Add a batch command submission function
 - tegra: Add DSI support for Tegra 20 and 30, fix dsi driver when the
   firmware hasn't enabled the controller,
 - v3d: Reduce PM runtime autosuspend delay, Scheduler and submission
   fixes and refactoring, Deprecate V3D 3.3 and 4.1 support

 - bridge:
   - display-connector: don't autoenable HPD IRQ, trigger initial HPD
     event for DP
   - dw-dp: Null pointer dereference and use-after-free fixes
   - ti-sn65dsi83: Remove NO_HFP and NO_HBP mode flags
 - panel:
   - himax-hx83121a: add backlight regulator support
   - novatek-nt36672a: Inline panel init sequences
   - panel-edp: Add quirks for AUO B116XAT04.3, CMN N116BCP-EA2, CSW
     MNB601LS1-8, BOE NV116WH2-M30, BOE NT116WHM-N21, BOE NV116FH1-M31,
     BOE NV116FH1-M30, NV140FHM-N5B, TM156VDXP25
   - New panels: Samsung ATNA40HQ08-0, Anbernic TD4310, Chipone
     ICNA35XX, Ilitek ILI9488

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@redhat.com>
Link: https://patch.msgid.link/20260619-burgundy-termite-of-whirlwind-f7a4dd@houat
2 months agoLinux 7.2-rc2
Linus Torvalds [Mon, 6 Jul 2026 00:44:06 +0000 (14:44 -1000)]
Linux 7.2-rc2

2 months agoMerge tag 'x86-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 5 Jul 2026 15:37:46 +0000 (05:37 -1000)]
Merge tag 'x86-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull x86 fix from Ingo Molnar:

 - Prevent OOB access in the resctrl code while offlining
   CPUs when Intel SNC (Sub-NUMA Clustering) is enabled
   (Reinette Chatre)

* tag 'x86-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86,fs/resctrl: Prevent out-of-bounds access while offlining CPU when SNC enabled

2 months agoMerge tag 'perf-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 5 Jul 2026 15:34:43 +0000 (05:34 -1000)]
Merge tag 'perf-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull perf events fixes from Ingo Molnar:

 - Fix a perf_event_attr::remove_on_exec bug for group events
   (Taeyang Lee)

 - Fix uprobes CALL emulation interaction with shadow stacks, and
   add a testcase for this (David Windsor)

 - Fix uprobes unregister bug (Jiri Olsa)

* tag 'perf-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  uprobes/x86: Use proper mm_struct in __in_uprobe_trampoline
  selftests/x86: Add shadow stack uprobe CALL test
  x86/uprobes: Keep shadow stack in sync for emulated CALLs
  perf/core: Detach event groups during remove_on_exec

2 months agoMerge tag 'locking-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 5 Jul 2026 15:31:41 +0000 (05:31 -1000)]
Merge tag 'locking-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull futex fix from Ingo Molnar:

 - Fix a futex-requeue deadlock detection regression (Thomas Gleixner)

* tag 'locking-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  futex/requeue: Revert "Prevent NULL pointer dereference in remove_waiter() on self-deadlock""

2 months agoMerge tag 'irq-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 5 Jul 2026 15:29:41 +0000 (05:29 -1000)]
Merge tag 'irq-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull irq fixes from Ingo Molnar:
 "Misc irqchip driver fixes:

   - Fix a resource leak in the RISC-V imsic-early driver (Haoxiang Li)

   - Fix an OF node reference leak in the ARM gic-v3-its driver (Yuho
     Choi)

   - Fix a dangling handler function on module removal bug in the
     TS-4800 ARM board irqchip driver (Qingshuang Fu)"

* tag 'irq-urgent-2026-07-05' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/ts4800: Fix missing chained handler cleanup on remove
  irqchip/gic-v3-its: Fix OF node reference leak
  irqchip/irq-riscv-imsic-early: Fix fwnode leak on state setup failure

2 months agoMerge tag 'sound-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Sun, 5 Jul 2026 15:26:45 +0000 (05:26 -1000)]
Merge tag 'sound-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A standard set of driver-specific fixes and quirks accumulated since
  the merge window:

  ASoC:
   - SOF: Sanity check to prevent OOB reads
   - rsnd: Fix clock leak and double-disable issues with PM
   - tas675x: Misc fixes for register fields, etc
   - lpass-va-macro: Correct codec version for Qualcomm SC7280
   - amd-yc: DMIC quirk for Alienware m15 R7 AMD

  Others:
   - us144mkii: Fix a UAF on disconnect and anchor list corruption
   - HD-audio: Realtek quirks for HP models"

* tag 'sound-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ASoC: rsnd: src: Add missing scu_supply clock to suspend/resume
  Documentation: sound: tas675x: Fix temperature range and impedance documentation
  ASoC: codecs: tas675x: Fix CHx temperature range register bit fields
  ASoC: codecs: tas675x: use READ_ONCE for params to be used concurrently
  ASoC: rsnd: adg: make rsnd_adg_clk_control() idempotent
  ASoC: SOF: validate probe info element counts
  ALSA: usx2y: us144mkii: fix work UAF on disconnect
  ASoC: amd: yc: Add Alienware m15 R7 AMD to DMIC quirk table
  ALSA: hda/realtek: Add quirk for HP Victus 16-e0xxx (88EE) to enable mute LED
  MAINTAINERS: ASoC: SOF: add AMD reviewer for Sound Open Firmware
  ASoC: codecs: lpass-va-macro: Fix LPASS Codec Version for SC7280
  ALSA: us144mkii: capture_urb_complete: redundant usb_anchor_urb corrupts anchor list on each resubmission

2 months agoMerge tag 'spi-fix-v7.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi
Linus Torvalds [Sun, 5 Jul 2026 15:24:06 +0000 (05:24 -1000)]
Merge tag 'spi-fix-v7.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi

Pull spi fixes from Mark Brown:
 "A small set of fixes that came in since -rc1, we have one core fix for
  shutting down target mode properly if the system suspends while it's
  running plus a small set of fairly unremarkable device specific fixes.
  There's also a couple of pure DT binding changes for Renesas SoCs, the
  power domains one allows some SoCs to be correctly described with
  existing code"

* tag 'spi-fix-v7.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: rzv2h-rspi: Fix DMA transfer error handling for signal interruption
  spi: dt-bindings: snps,dw-apb-ssi: add 'power-domains' property
  spi: dt-bindings: snps,dw-apb-ssi: drop superfluous RZ/N1 entry
  spi: dw: use the correct error msg if request_irq() fails
  spi: dw: fix first spi transfer with dma always fallback to PIO
  spi: core: Abort active target transfer on controller suspend
  spi: sh-msiof: abort transfers when reset times out

2 months agoMerge tag 's390-7.2-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Sat, 4 Jul 2026 16:28:45 +0000 (06:28 -1000)]
Merge tag 's390-7.2-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux

Pull s390 fixes from Vasily Gorbik:

 - Fix PKEY_VERIFYPROTK ioctl key type handling by removing the generic
   key-length based type check with its wrong bit-size calculation, and
   leaving protected key verification to the pkey handler

 - Fix monwriter buffer reuse by rejecting records that change the data
   length, preventing out of bounds user copy into the kernel buffer

* tag 's390-7.2-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/monwriter: Reject buffer reuse with different data length
  pkey: Move keytype check from pkey api to handler

2 months agoMerge tag 'mips-fixes_7.2_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips...
Linus Torvalds [Sat, 4 Jul 2026 16:05:28 +0000 (06:05 -1000)]
Merge tag 'mips-fixes_7.2_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux

Pull MIPS fixes from Thomas Bogendoerfer.

* tag 'mips-fixes_7.2_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
  MIPS: configs: Enable the current Ingenic USB PHY symbol
  MIPS: loongson64: add IRQ work based on self-IPI
  MIPS: mm: Add check for highmem before removing memory block
  mips: Add build salt to the vDSO
  MIPS: DEC: Ensure RTC platform device deregistration upon failure

2 months agoMerge tag 'v7.2-rc1-smb3-server-fixes' of git://git.samba.org/ksmbd
Linus Torvalds [Sat, 4 Jul 2026 04:55:34 +0000 (18:55 -1000)]
Merge tag 'v7.2-rc1-smb3-server-fixes' of git://git.samba.org/ksmbd

Pull smb server fixes from Steve French:

 - Fix several use-after-free races in durable handle reconnect,
   supersede, and oplock handling

 - Avoid holding the inode oplock lock while waiting for a lease break
   acknowledgement. This removes delays of up to 35 seconds when cifs.ko
   closes a deferred handle in response to a lease break

 - Fix malformed security descriptor handling, including an undersized
   DACL allocation issue and an out-of-bounds ACE SID read

 - Fix memory leaks in security descriptor and DOS attribute xattr
   encoding/decoding error paths

 - Fix outstanding SMB2 credit leaks on aborted requests and correct the
   QUERY_INFO credit charge calculation

 - Fix hard-link creation without replacement being incorrectly rejected
   when the handle lacks DELETE access

 - Avoid unnecessary zeroing of large SMB2 read buffers

 - Add an oplock list lockdep annotation and update the documented
   support status for durable handles and SMB3.1.1 compression

 - Durable handle fixes to address ownership and lifetime races during
   reconnect, session teardown, oplock handling, and superseding opens,
   preventing stale session and file references from being used by
   concurrent operations

* tag 'v7.2-rc1-smb3-server-fixes' of git://git.samba.org/ksmbd:
  ksmbd: fix app-instance durable supersede session UAF
  ksmbd: snapshot previous oplock state before durable checks
  ksmbd: close superseded durable handles through refcount handoff
  ksmbd: fix use-after-free of fp->owner.name in durable handle owner check
  smb/server: do not require delete access for non-replacing links
  ksmbd: don't hold ci->m_lock while waiting for a lease break ack
  ksmbd: doc: update feature support status for durable handles and compression
  ksmbd: annotate oplock list traversals under m_lock
  ksmbd: fix outstanding credit leak on abort and error paths
  ksmbd: fix credit charge calculation for SMB2 QUERY_INFO
  ksmbd: avoid zeroing the read buffer in smb2_read()
  ksmbd: validate num_subauth when copying ACE in set_ntacl_dacl
  ksmbd: reject undersized DACLs before parsing ACEs
  ksmbd: fix n.data memory leak in ksmbd_vfs_set_dos_attrib_xattr
  ksmbd: Fix acl.sd_buf memory leak and invalid sd_size error handling
  ksmbd: fix sd_ndr.data memory leak in ksmbd_vfs_set_sd_xattr

2 months agoMerge tag 'drm-fixes-2026-07-04' of https://gitlab.freedesktop.org/drm/kernel
Linus Torvalds [Sat, 4 Jul 2026 01:42:20 +0000 (15:42 -1000)]
Merge tag 'drm-fixes-2026-07-04' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Dave Airlie:
 "Weekly fixes for drm. This is large for rc2 but it's just a lot of
  small fixes across a bunch of drivers, xe, amdgpu as usual, plus some
  sashiko-inspired fixes for panthor, and some dma-fence updates.

  core:
   - kernel doc fix
   - include types.h in drm_ras.h

  dma-fence:
   - fix NULL ptr dereference
   - use correct callback
   - make dma_fence_dedup_array more robust

  dp:
   - handle torn down topology gracefully
   - fix kernel doc

  i915:
   - Input validation fixes for BIOS and EDID
   - Fix HDCP code buffer overflow and seq_num_v monotonic increase check
   - Fix near-NULL deref in i915_active during GFP_ATOMIC exhaustion

  xe:
   - Wedge from the timeout handler only after releasing the queue
   - Fix a NULL pointer dereference
   - Remove redundant exec_queue_suspended
   - RTP / OA whitelist fixes
   - Return error on non-migratable faults requiring devmem
   - Skip FORCE_WC and vm_bound check for external dma-bufs
   - Hold notifier lock for write on inject test path
   - Drop bogus static from finish in force_invalidate
   - Fix double-free of managed BO in error path
   - Don't attempt to process FAST_REQ or EVENT relays
   - Fix NPD in bo_meminfo
   - Prevent invalid cursor access for purged BOs
   - Fix offset alignment for MERT WHITELST_OA_MERT_MMIO_TRG

  amdgpu:
   - Soc24 aborted suspend fix
   - Drop unecessary BUG() and BUG_ON() from error paths
   - SCPM fix
   - Power reporting fix
   - DCE HDR fix
   - UVD boundary checks
   - VCN boundary checks
   - VCE boundary checks
   - DCN 4.2 fixes
   - Large stack allocation fixes
   - Fix aperture mapping leak
   - UserQ fixes
   - Ignore_damage_clips fix
   - ACP fixes
   - DC boundary checks
   - GPUVM fixes
   - JPEG idle check fixes
   - Userptr fix
   - GC 11.7 updates
   - Non-4K page fix
   - SMU 13 fixes
   - DP alt mode fix

  amdkfd:
   - Boundary checks
   - CRIU fixes

  amdxdna:
   - fix device removal issues
   - fix use after free in debug BO

  imagination:
   - fix double call to scheduler fini
   - fix ioctl return values
   - fix user array stride

  virtio:
   - handle EDIDs better

  panthor:
   - irq safe fence lock fix
   - reset work fix
   - fix invalid pointer
   - fix iomem access in suspended state
   - sched resume fix
   - unplug suspend fix
   - drop needless check
   - eviction leak fix
   - bail on group start/resume fix
   - keep irqs masked

  malidp:
   - use clock bulk API

  komeda:
   - clock prepare fixes"

* tag 'drm-fixes-2026-07-04' of https://gitlab.freedesktop.org/drm/kernel: (105 commits)
  drm/xe/oa: Fix offset alignment for MERT WHITELIST_OA_MERT_MMIO_TRG
  drm/xe/pt: prevent invalid cursor access for purged BOs
  drm/xe: fix NPD in bo_meminfo()
  drm/xe/pf: Don't attempt to process FAST_REQ or EVENT relays
  drm/xe/hw_engine: Fix double-free of managed BO in error path
  drm/xe/userptr: Drop bogus static from finish in force_invalidate
  drm/xe/userptr: Hold notifier_lock for write on inject test path
  drm/xe/display: skip FORCE_WC and vm_bound check for external dma-bufs
  drm/xe: Return error on non-migratable faults requiring devmem
  drm/xe/rtp: Ensure locking/ref counting for OA whitelists
  drm/xe/oa: (De-)whitelist OA registers on OA stream open/release
  drm/xe/rtp: (De-)whitelist OA registers for all hwe's for a gt
  drm/xe/rtp: Toggle 'deny' bit to (de-)whitelist OA regs
  drm/xe/rtp: Save OA nonpriv registers to register save/restore lists
  drm/xe/rtp: Generalize whitelist_apply_to_hwe
  drm/xe/rtp: Keep track of non-OA nonpriv slots
  drm/xe/rtp: Maintain OA whitelists separately
  drm/xe/rtp: Fix build error with clang < 21 and non-const initializers
  drm/imagination: Fix user array stride in pvr_set_uobj_array()
  drm/imagination: Fix returned size for DRM_IOCTL_PVR_DEV_QUERY
  ...

2 months agoMerge tag 'acpi-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Sat, 4 Jul 2026 01:13:50 +0000 (15:13 -1000)]
Merge tag 'acpi-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm

Pull ACPI support fixes from Rafael Wysocki:
 "These fix a coding mistake in the ACPI TAD (Time and Alarm
  Device) driver introduced by one of its previous updates and
  get rid of the ugly #ifdef __KERNEL__ conditional compilation
  in acpi_ut_safe_strncpy() by redefining that function as an
  alias for strscpy_pad():

   - Add a missing ACPI_TAD_AC_WAKE capability check omitted by mistake
     to the ACPI TAD driver (Xu Rao)

   - Define acpi_ut_safe_strncpy() as an alias for strscpy_pad()
     which is viable because that function is only called from kernel
     code (Rafael Wysocki)"

* tag 'acpi-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPICA: Define acpi_ut_safe_strncpy() as strscpy_pad() alias
  ACPI: TAD: Check AC wake capability before enabling wakeup

2 months agoMerge tag 'riscv-for-linus-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 4 Jul 2026 01:07:24 +0000 (15:07 -1000)]
Merge tag 'riscv-for-linus-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux

Pull RISC-V fixes from Paul Walmsley:

 - Fix a crash when a kretprobe reads from the stack

 - Fix an issue with the build-time mcount sorter that broke ftrace

 - Fix the rv32 IRQ stack frame padding to match the ABI

 - Only defer IOMMU configuration during initialization. This avoids an
   issue where IOMMU configuration could be indefinitely deferred

 - Add the missing build salt to the vDSO

 - Now that RISC-V systems with higher numbers of cores are starting to
   become available, raise NR_CPUS for RISC-V to 256

 - Clean up some warnings from sparse caused by the RISC-V-optimized
   RAID6 code

 - Clean up our __cpu_up() code with a few minor fixes

* tag 'riscv-for-linus-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: probes: save original sp in rethook trampoline
  riscv: Fix 32-bit call_on_irq_stack() frame pointer ABI
  scripts/sorttable: Handle RISC-V patchable ftrace entries
  riscv: smp: use secs_to_jiffies in __cpu_up
  ACPI: RIMT: Only defer the IOMMU configuration in init stage
  riscv: Add build salt to the vDSO
  raid6: fix raid6_recov_rvv symbol undeclared warning
  raid6: fix riscv symbol undeclared warnigns
  riscv: Raise default NR_CPUS for 64BIT to 256

2 months agoMerge tag 'v7.2-rc1-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 4 Jul 2026 01:01:33 +0000 (15:01 -1000)]
Merge tag 'v7.2-rc1-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull smb client fixes from Steve French:

 - Credit fix

 - Fix alignment issue in parse_posix_ctxt

 - SID parsing fix

* tag 'v7.2-rc1-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: Fix missing credit release on failure in cifs_issue_read()
  cifs: update internal module version number
  smb: client: use unaligned reads in parse_posix_ctxt()
  smb: client: harden POSIX SID length parsing

2 months agoMerge branch 'acpi-tad'
Rafael J. Wysocki [Fri, 3 Jul 2026 18:28:08 +0000 (20:28 +0200)]
Merge branch 'acpi-tad'

Merge an ACPI TAD (Time and Alarm Device) driver fix for 7.2-rc2.

* acpi-tad:
  ACPI: TAD: Check AC wake capability before enabling wakeup

2 months agoMerge tag 'vfs-7.2-rc2.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Linus Torvalds [Fri, 3 Jul 2026 15:48:05 +0000 (05:48 -1000)]
Merge tag 'vfs-7.2-rc2.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs

Pull vfs fixes from Christian Brauner:

 - netfs:

    - fix the decision when to disallow write-streaming with fscache in
      use, handling of asynchronous cache object creation, a double fput
      in cachefiles, clearing S_KERNEL_FILE without the inode lock held,
      page extraction bugs in the iov_iter helpers (a potential
      underflow, a missing allocation failure check, a memory leak, and
      a folio offset miscalculation), writeback error and ENOMEM
      handling, DIO write retry for filesystems without a
      ->prepare_write() method, and the replacement of the wb_lock mutex
      with a bit lock plus writethrough collection offload so that
      multiple asynchronous writebacks don't interfere with each other.

    - Fix the barriering when walking the netfs subrequest list during
      retries as it was possible to see a subrequest that was just added
      by the application thread.

 - iomap:

    - Change iomap to submit read bios after each extent instead of
      building them up across extents. The old behavior was considered
      problematic for a while and now caused an actual erofs bug.

    - Guard the ioend io_size EOF trim in iomap against underflow when a
      concurrent truncate moves EOF below the start of the ioend,
      wrapping io_size to a huge value.

 - overlayfs

    - Fix a stale overlayfs comment about the locking order.

    - Store the linked-in upper dentry instead of the disconnected
      O_TMPFILE dentry during overlayfs tmpfile copy-up. With a FUSE or
      virtiofs upper layer ->d_revalidate() would try to look up "/" in
      the workdir and fail, causing persistent ESTALE errors that broke
      dpkg and apt.

 - vfs-bpf:

   Have the bpf_real_data_inode() kfunc take a struct file instead of a
   dentry so it is usable from the bprm_check_security, mmap_file, and
   file_mprotect hooks, and rename it from bpf_real_inode() to make the
   data-inode semantics explicit. The kfunc landed this cycle so the
   change is safe.

 - afs:

   NULL pointer dereferences in the callback service and in
   afs_get_tree(), several memory and refcount leaks, missing locking
   around the dynamic root inode numbers and premature cell exposure
   through /afs, a netns destruction hang caused by a misplaced
   increment of net->cells_outstanding, a bulk lookup malfunction caused
   by the dir_emit() API change, inode (re)initialisation issues, and
   assorted smaller fixes to error codes, seqlock handling, and debug
   output.

 - vfs:

   Refuse O_TMPFILE creation with an unmapped fsuid or fsgid and add a
   selftest for it.

 - vboxsf:

   Add Jori Koolstra as vboxsf maintainer, taking over from Hans de
   Goede.

 - dio:

   Release the pages attached to a short atomic dio bio; the REQ_ATOMIC
   size check error path leaked them.

 - procfs:

   Only bump the parent directory link count when registering
   directories in procfs. Registering regular files inflated the count
   and leaked a link on every create and remove cycle.

 - minix:

   Avoid an unsigned overflow in the minix bitmap block count
   calculation that let crafted images with huge inode or zone counts
   pass superblock validation and crash the kernel during mount.

 - cachefiles:

   Fix a double unlock in the cachefiles nomem_d_alloc error path left
   over from the start_creating() conversion.

 - fat:

   Stop fat from reading directory entries past the 0x00
   end-of-directory marker. If the trailing on-disk slots aren't
   zero-filled the driver surfaced arbitrary garbage as directory
   entries.

 - freexvfs:

   Don't BUG() on unknown typed-extent types in freevxfs, reachable via
   ioctl(FIBMAP) on a crafted image; fail with an I/O error instead.

 - orangefs:

   Keep the readdir entry size 64-bit in orangefs fill_from_part().
   Truncating it to __u32 bypassed the bounds check and led to
   out-of-bounds reads triggerable by the userspace client.

 - xfs:

   Fix the error unwind in xfs_open_devices() which released the rt
   device file twice and left dangling buftarg pointers behind that were
   freed again when the failed mount was torn down.

 - exec:

   Fix an off-by-one in the comment documenting the maximum binfmt
   rewrite depth in exec_binprm(). The code allows five rewrites, not
   four; restricting the code would break userspace so the comment is
   fixed instead.

 - file handles:

   Reject detached mounts in capable_wrt_mount(). A detached mount can
   be dissolved concurrently, leaving a NULL mount namespace that
   open_by_handle_at() would dereference.

* tag 'vfs-7.2-rc2.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (57 commits)
  netfs: Fix barriering when walking subrequest list
  iomap: submit read bio after each extent
  fuse: call fuse_send_readpages explicitly from fuse_readahead
  iomap: consolidate bio submission
  fhandle: reject detached mounts in capable_wrt_mount()
  netfs: Fix DIO write retry for filesystems without a ->prepare_write()
  netfs: Fix folio state after ENOMEM whilst under writeback iteration
  netfs: Fix writeback error handling
  netfs: Fix writethrough to use collection offload
  netfs: Replace wb_lock with a bit lock for asynchronicity
  netfs: Fix kdoc warning
  scatterlist: Fix offset in folio calc in extract_xarray_to_sg()
  iov_iter: Remove unused variable in kunit_iov_iter.c
  iov_iter: Fix a memory leak in iov_iter_extract_user_pages()
  iov_iter: Fix missing alloc fail check in iov_iter_extract_bvec_pages()
  iov_iter: Fix potential underflow in iov_iter_extract_xarray_pages()
  cachefiles: Fix file burial to take lock when unsetting S_KERNEL_FILE
  cachefiles: Fix double fput
  netfs: Fix netfs_create_write_req() to handle async cache object creation
  netfs: Fix decision whether to disallow write-streaming due to fscache use
  ...

2 months agoMerge tag 'xfs-fixes-7.2-rc2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Fri, 3 Jul 2026 15:44:56 +0000 (05:44 -1000)]
Merge tag 'xfs-fixes-7.2-rc2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

Pull xfs fixes from Carlos Maiolino:
 "A collection of bugfixes and some small code refactoring"

* tag 'xfs-fixes-7.2-rc2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: simplify __xfs_buf_ioend
  xfs: fix handling of synchronous errors in xfs_buf_submit
  xfs: remove xfs_buf_ioend
  xfs: improve the xfs_buf_ioend_fail calling convention
  xfs: use null daddr for unset first bad log block
  xfs: fix memory leak in xfs_dqinode_metadir_create()
  xfs: release dquot buffer after dqflush failure
  xfs: also mark the buffer stale on verifier failure in xfs_buf_submit
  xfs: open code xfs_buf_ioend_fail in xfs_buf_submit
  xfs: fix AGFL extent count calculation in xrep_agfl_fill
  xfs: simplify the failure path in xfs_buf_alloc_vmalloc
  xfs: fix incorrect use of gfp flags in xfs_buf_alloc_backing_mem
  xfs: lift setting __GFP_NOFAIL from xfs_buf_alloc_kmem to the caller
  xfs: split up xfs_buf_alloc_backing_mem

2 months agoMerge tag 'for-linus-7.2a-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 3 Jul 2026 15:40:58 +0000 (05:40 -1000)]
Merge tag 'for-linus-7.2a-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:

 - rename function parameters and a comment related to
   xen_exchange_memory() (Jan Beulich)

 - replace __ASSEMBLY__ with __ASSEMBLER__ (Thomas Huth)

 - add some sanity checking to the Xen pvcalls frontend driver (Michael
   Bommarito)

 - fix error handling in the Xen gntdev driver (Wentao Liang)

 - fix several minor bugs in Xen related drivers (Yousef Alhouseen)

* tag 'for-linus-7.2a-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  x86/Xen: correct commentary and parameter naming of xen_exchange_memory()
  xenbus: reject unterminated directory replies
  xen/gntalloc: validate grant count before allocation
  xen/gntalloc: make grant counters unsigned
  xen/front-pgdir-shbuf: free grant reference head on errors
  xen/gntdev: fix error handling in ioctl
  xen: Replace __ASSEMBLY__ with __ASSEMBLER__ in header files
  xen/pvcalls: bound backend response req_id before indexing rsp[]

2 months agoMerge tag 'gpio-fixes-for-v7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 3 Jul 2026 15:38:12 +0000 (05:38 -1000)]
Merge tag 'gpio-fixes-for-v7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux

Pull gpio fixes from Bartosz Golaszewski:

 - check the return value of gpiochip_add_data() in gpio-mvebu and
   gpio-htc-egpio

 - avoid locking context issues with GPIO drivers using the shared GPIO
   proxy by only allowing sleeping operations (atomic GPIO ops don't
   really make sense in shared context anyway)

 - with the above: restore non-sleeping GPIO access in pinctrl-meson

 - fix return value on OOM in gpio-timberdale

 - fix interrupt handling in gpio-mt7621

 - support both A and B variants of NCT6126D in gpio-f7188x

* tag 'gpio-fixes-for-v7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux:
  pinctrl: meson: restore non-sleeping GPIO access
  gpio: timberdale: Return -ENOMEM on dynamic memory allocation in probe
  gpio: mt7621: be sure IRQ domain is created before exposing GPIO chips
  gpio: mt7621: more robust management of IRQ domain teardown
  gpio: mt7621: avoid corruption of shared interrupt trigger state
  gpio: shared-proxy: always serialize with a sleeping mutex
  gpio-f7188x: Add support for NCT6126D version B
  gpio: htc-egpio: use managed gpiochip registration
  gpio: mvebu: fail probe if gpiochip registration fails

2 months agonetfs: Fix barriering when walking subrequest list
David Howells [Thu, 2 Jul 2026 08:23:02 +0000 (09:23 +0100)]
netfs: Fix barriering when walking subrequest list

Fix the barriering used when walking the subrequest list in retry as
there's a possibility of seeing a subreq that's just been added by the
application thread.

Fixes: ee4cdf7ba857 ("netfs: Speed up buffered reading")
Fixes: 288ace2f57c9 ("netfs: New writeback implementation")
Link: https://sashiko.dev/#/patchset/20260608145432.681865-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
Link: https://patch.msgid.link/138807.1782980582@warthog.procyon.org.uk
Reviewed-by: Paulo Alcantara (Red Hat) <pc@manguebit.org>
cc: Paulo Alcantara <pc@manguebit.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
2 months agoMerge tag 'asoc-fix-v7.2-rc1' of https://git.kernel.org/pub/scm/linux/kernel/git...
Takashi Iwai [Fri, 3 Jul 2026 07:30:30 +0000 (09:30 +0200)]
Merge tag 'asoc-fix-v7.2-rc1' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound into for-linus

ASoC: Fixes for v7.2

A fairly standard set of driver specific fixes and quirks that have come
in since the merge window, plus a MAINTAINERS update.  The tas675x
READ_ONCE change is probably not actually fixing issues properly but we
need a whole new approach to concurrency there and it came along with
some good fixes.

2 months agoMerge tag 'device-id-rework' of git://git.kernel.org/pub/scm/linux/kernel/git/ukleine...
Linus Torvalds [Fri, 3 Jul 2026 06:54:26 +0000 (20:54 -1000)]
Merge tag 'device-id-rework' of git://git.kernel.org/pub/scm/linux/kernel/git/ukleinek/linux

Pull mod_devicetable.h header split from Uwe Kleine-König:
 "Split <linux/mod_devicetable.h> in per subsystem headers

  <linux/mod_devicetable.h> is included transitively in nearly every
  driver in an x86_64 allmodconfig build of v7.1:

      $ find drivers -name \*.o -not -name \*.mod.o | wc -l
      21330
      $ find drivers -name \*.o.cmd -not -name \*.mod.o.cmd | xargs grep -l mod_devicetable.h | wc -l
      17038

  The result of this mixture of different and unrelated subsystem
  details is that even when touching an obscure device id struct most of
  the kernel needs to be recompiled. Given that each driver typically
  only needs one or two of these structures, splitting into per
  subsystem headers and only including what is really needed reduces the
  amount of needed recompilation.

  This split is implemented in the first commit and then after some
  preparatory work in the following commits, the last two replace
  includes of <linux/mod_devicetable.h> by the actually needed more
  specific headers.

  There are still a few instances left, but the ones with high impact
  (that is in headers that are used a lot) and the easy ones (.c files)
  are handled. These remaining includes will be addressed during the
  next merge window"

* tag 'device-id-rework' of git://git.kernel.org/pub/scm/linux/kernel/git/ukleinek/linux:
  Replace <linux/mod_devicetable.h> by more specific <linux/device-id/*.h> (c files)
  Replace <linux/mod_devicetable.h> by more specific <linux/device-id/*.h> (headers)
  parisc: #include <linux/compiler.h> for unlikely() in <asm/ptrace.h>
  media: em28xx: Add include for struct usb_device_id
  LoongArch: KVM: Add include defining struct cpu_feature
  ALSA: hda/core: Add include defining struct hda_device_id
  usb: dwc2: Add include defining struct pci_device_id
  platform/x86: int3472: Add include defining struct dmi_system_id
  platform/x86: x86-android-tablets: Add include defining struct dmi_system_id
  i2c: Let i2c-core.h include <linux/i2c.h>
  of: Explicitly include <linux/types.h> and <linux/err.h>
  platform/x86: msi-ec: Ensure dmi_system_id is defined
  usb: serial: Include <linux/usb.h> in <linux/usb/serial.h>
  driver core: platform: Include header for struct platform_device_id
  driver: core: Include headers for acpi_device_id and of_device_id for struct device_driver
  media: ti: vpe: #include <linux/platform_device.h> explicitly
  mod_devicetable.h: Split into per subsystem headers

2 months agoMerge tag 'ata-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/libata/linux
Linus Torvalds [Fri, 3 Jul 2026 06:05:43 +0000 (20:05 -1000)]
Merge tag 'ata-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/libata/linux

Pull ata fixes from Damien Le Moal:

 - Quirk the Phison PS3111-S11 SSD with NOLPM due to its defective
   link power management (Bryam)

 - Strengthen checks on a device concurrent positioning range
   information to make sure to reject any invalid report (Bryam)

 - Fix probe error handling in the pata_pxa and sata_gemini
   drivers (Myeonghun, Wentao)

 - Limit buffer size of replies from translated commands to what
   libata actually generated (Karuna)

* tag 'ata-7.2-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/libata/linux:
  ata: libata-scsi: limit simulated SCSI command copy to response length
  ata: pata_pxa: Fix DMA channel leak on probe error
  ata: sata_gemini: unwind clocks on IDE pinctrl errors
  ata: libata-core: Reject an invalid concurrent positioning ranges count
  ata: libata-core: Add NOLPM quirk for PNY CS900 1TB SSD

2 months agoReplace <linux/mod_devicetable.h> by more specific <linux/device-id/*.h> (c files)
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:36 +0000 (11:24 +0200)]
Replace <linux/mod_devicetable.h> by more specific <linux/device-id/*.h> (c files)

Replace the #include of <linux/mod_devicetable.h> by the more specific
<linux/device-id/*.h> where applicable. For most cases the include
can be dropped completely, only a few drivers need one or two headers
added.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://patch.msgid.link/1a3f2007c5c5dcf555c09a4035ce3ae8ef1b6c49.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoReplace <linux/mod_devicetable.h> by more specific <linux/device-id/*.h> (headers)
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:35 +0000 (11:24 +0200)]
Replace <linux/mod_devicetable.h> by more specific <linux/device-id/*.h> (headers)

<linux/mod_devicetable.h> is included in a many files:

$ git grep '<linux/mod_devicetable.h>' ef0c9f75a195 | wc -l
1598

; some of them are widely used headers. To stop mixing up different and
unrelated driver( type)s let the subsystem headers only use the subset
of the recently split <linux/mod_devicetable.h> that are relevant for
them.

The fallout (I hope) is addressed in the previous commits that handle
sources relying on e.g. <linux/i2c.h> pulling in the full legacy header
and thus providing pci_device_id.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/199fe46b624ba07fb9bd3e0cd6ff13757932cb5f.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoparisc: #include <linux/compiler.h> for unlikely() in <asm/ptrace.h>
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:34 +0000 (11:24 +0200)]
parisc: #include <linux/compiler.h> for unlikely() in <asm/ptrace.h>

Currently <linux/compiler.h> isn't included at all (not even
transitively) in <asm/ptrace.h>.
arch/parisc/kernel/asm-offsets.c just happens to include the following
chain of includes before <asm/ptrace.h>:

<linux/sched.h>
-> <asm/processor.h>
-> <asm/hardware.h>
-> <linux/mod_devicetable.h>
-> <linux/uuid.h>
-> <linux/string.h>
-> <linux/compiler.h>

. That chain will be broken, because in one of the next commits
<asm/hardware.h> is changed to only include <linux/device-id/parisc.h>
instead of <linux/mod_devicetable.h>. So to ensure
arch/parisc/kernel/asm-offsets.c knows about unlikely() even after that
change, #include <linux/compiler.h> explicitly.

Link: https://patch.msgid.link/0574a2b73363c3cbf21c55c27455c3cecfb33583.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agomedia: em28xx: Add include for struct usb_device_id
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:33 +0000 (11:24 +0200)]
media: em28xx: Add include for struct usb_device_id

Traditionally <linux/mod_devicetable.h> was a header defining a plethora
of structs, among them struct usb_device_id. This was split now with the
objective that only the relevant bits are included.

Currently <linux/mod_devicetable.h> is transitively included in
drivers/media/usb/em28xx/em28xx.h via:

drivers/media/usb/em28xx/em28xx.h ->
<linux/i2c.h> ->
<linux/acpi.h> ->
<linux/device.h> ->
<linux/device/driver.h> ->
<linux/mod_devicetable.h

To keep struct usb_device_id available once <linux/device/driver.h>
stops including <linux/mod_devicetable.h>, include it the header
providing that struct explictly.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/e72de5b4b9f1aa77a3c19a5e698a195dfd81ae0b.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoLoongArch: KVM: Add include defining struct cpu_feature
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:32 +0000 (11:24 +0200)]
LoongArch: KVM: Add include defining struct cpu_feature

Traditionally <linux/mod_devicetable.h> was a header defining a plethora
of structs, among them struct cpu_features. This was split now with the
objective that only the relevant bits are included.

Currently <linux/mod_devicetable.h> is transitively included in
arch/loongarch/kvm/main.c via:

arch/loongarch/kvm/main.c ->
<linux/kvm_host.h> ->
<linux/entry-virt.h> ->
<linux/resume_user_mode.h> ->
<linux/memcontrol.h> ->
<linux/cgroup.h> ->
<linux/kernel_stat.h> ->
<linux/interrupt.h> ->
<linux/hardirq> ->
<asm/hardirq.h> ->
<linux/irq.h> ->
<asm/irq.h> ->
<linux/irqdomain.h> ->
<linux/of.h> ->
<linux/mod_devicetable.h>

To keep struct cpu_features available once <linux/of.h> stops including
<linux/mod_devicetable.h>, include it here explicitly.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Reviewed-by: Bibo Mao <maobibo@loongson.cn>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/052feec0e04ea8f5b2706a19a5b236679eed0aba.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoALSA: hda/core: Add include defining struct hda_device_id
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:31 +0000 (11:24 +0200)]
ALSA: hda/core: Add include defining struct hda_device_id

Traditionally all *_device_id were defined in a single header
<linux/mod_devicetable.h>. This was split now with the objective that
only the relevant bits are included. So including <linux/pci.h> won't be
enough to get a definition of (the unrelated to pci) struct
hda_device_id.

Add an explicit include for the header defining struct hda_device_id to
keep working when <linux/pci.h> stops providing this defintion.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Reviewed-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/376883bc5889d5cca01efb6f8d4e07a20158f2b8.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agousb: dwc2: Add include defining struct pci_device_id
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:30 +0000 (11:24 +0200)]
usb: dwc2: Add include defining struct pci_device_id

Up to now <linux/acpi.h> includes <linux/mod_devicetable.h> that
provides struct pci_device_id. However <linux/mod_devicetable.h> was
split into per bus headers and <linux/acpi.h> will only include the acpi
related one (and similar for other bus headers).

As struct pci_device_id is used in drivers/usb/dwc2/core.h, add an
include to ensure it's defined also after the includes in <linux/acpi.h>
are tightened.

Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/bddfcdfaf36d735c244e03efada6083ef98ebd51.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoplatform/x86: int3472: Add include defining struct dmi_system_id
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:29 +0000 (11:24 +0200)]
platform/x86: int3472: Add include defining struct dmi_system_id

Currently <linux/mod_devicetable.h> is included transitively in
int3472.h via

<linux/clk-provider.h> ->
<linux/of.h> ->
<linux/mod_devicetable.h>

However these includes will be tightend such that only the bits relevant
for of will be provided by <linux/of.h>. To ensure that dmi_system_id
stays around, include the respective header explicitly.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Link: https://patch.msgid.link/0ba52730f67dc995d9d896b81fa6a7320bf8cb4b.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoplatform/x86: x86-android-tablets: Add include defining struct dmi_system_id
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:28 +0000 (11:24 +0200)]
platform/x86: x86-android-tablets: Add include defining struct dmi_system_id

Currently <linux/i2c.h> includes <linux/mod_devicetable.h> transitively
which ensures that struct dmi_system_id is defined in
drivers/platform/x86/x86-android-tablets/x86-android-tablets.h. However
this include in <linux/i2c.h> will be replaced by one for i2c_device_id
only. To ensure that dmi_system_id is available add the include for that
explicitly.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/32928d9ee47cefc7dfc4c385c06bd5e598b0fca1.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoi2c: Let i2c-core.h include <linux/i2c.h>
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:27 +0000 (11:24 +0200)]
i2c: Let i2c-core.h include <linux/i2c.h>

The subsystem private header i2c-core.h uses several symbols defined in
<linux/i2c.h>, e.g. struct i2c_board_info and i2c_lock_bus()). This
doesn't pose a problem in practise because all files including
"i2c-core.h" also include <linux/i2c.h>.

To make this more robust add an include statement for <linux/i2c.h>
making the header self-contained.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/46aa85ab3dc4e63bfb5bd8ff1fd212a3d0e31f58.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoof: Explicitly include <linux/types.h> and <linux/err.h>
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:26 +0000 (11:24 +0200)]
of: Explicitly include <linux/types.h> and <linux/err.h>

<linux/of_platform.h> uses resource_size_t and relies on the transitive
include <linux/mod_devicetable.h> -> <linux/types.h>. It also uses error
constants and thus relying on the include chain
<linux/mod_devicetable.h> -> <linux/uuid.h> -> <linux/string.h> ->
<linux/err.h>.

With the plan to split <linux/mod_devicetable.h> per subsystem and then
only letting of_platform.h include the of-specific bits (which don't
require these two headers), add the needed includes explicitly to keep
the header self-contained.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/a730991bc8813cf70c2445064ea425291538f709.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoplatform/x86: msi-ec: Ensure dmi_system_id is defined
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:25 +0000 (11:24 +0200)]
platform/x86: msi-ec: Ensure dmi_system_id is defined

Currently <linux/acpi.h> includes <linux/mod_devicetable.h> and thus
dmi_system_id is available for the driver. To disentangle includes
<linux/acpi.h> will be changed to only include the header for
acpi_device_id instead of the full <linux/mod_devicetable.h>. To prepare
for that include the dedicated header for struct dmi_device_id.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/600c7ab3263dcb8cee39b43dbd313eba8abef376.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agousb: serial: Include <linux/usb.h> in <linux/usb/serial.h>
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:24 +0000 (11:24 +0200)]
usb: serial: Include <linux/usb.h> in <linux/usb/serial.h>

All consumers of the latter also include the former, but without that
struct usb_driver and struct usb_device_id (and maybe more) are not
defined. Add an include for <linux/usb.h> to make the header
self-contained.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/82219ab65d16ee5bfe5a35d11bc938baac3fd3bc.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agodriver core: platform: Include header for struct platform_device_id
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:23 +0000 (11:24 +0200)]
driver core: platform: Include header for struct platform_device_id

Platform drivers can define an array containing the supported device
variants to be assigned to the struct platform_driver's .id_table.

While a forward declaration of struct platform_device_id is technically
enough to make the driver self-contained, it's reasonable to provide the
(very lightweight) data type definition for that array in
<linux/platform_device.h> to not add that burden to all platform drivers
with an id-table.

Note that currently <linux/device.h> transitively includes
<linux/mod_devicetable.h> that provides struct platform_device_id. But
that include is planned to be replaced by a tighter set of includes that
only define the structures relevant for the stuff in <linux/device.h>.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/4ca29592c9d1c6d528a65e05b80af7355f3c79c5.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agodriver: core: Include headers for acpi_device_id and of_device_id for struct device_d...
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:22 +0000 (11:24 +0200)]
driver: core: Include headers for acpi_device_id and of_device_id for struct device_driver

struct device_driver contains pointers of type struct of_device_id* and
struct acpi_device_id* but doesn't ensure these are defined. To make the
header self-contained add the (very lightweight) includes that contain
the respective definitions.

Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Rafael J. Wysocki (Intel) <rafael@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/199ba71b4ac73f4b4d9f5d2be635c96eec73c70e.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agomedia: ti: vpe: #include <linux/platform_device.h> explicitly
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:21 +0000 (11:24 +0200)]
media: ti: vpe: #include <linux/platform_device.h> explicitly

The driver uses several symbols and structs defined in that header. The
header is currently included transitively via

"vip.h" ->
<media/v4l2-ctrls.h> ->
<media/media-request.h> ->
<media/media-device.h> ->
<linux/platform_device.h>

which seems to be on the lower end of the scale between random and
reliable.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Reviewed-by: Yemike Abhilash Chandra <y-abhilashchandra@ti.com>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://patch.msgid.link/9f2e0e001eec087f00ac2c5af2de2e8f6d0978c1.1782808461.git.u.kleine-koenig@baylibre.com
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agomod_devicetable.h: Split into per subsystem headers
Uwe Kleine-König (The Capable Hub) [Tue, 30 Jun 2026 09:24:20 +0000 (11:24 +0200)]
mod_devicetable.h: Split into per subsystem headers

<linux/mod_devicetable.h> is included transitively in nearly every
driver in an x86_64 allmodconfig build of v7.1:

$ find drivers -name \*.o -not -name \*.mod.o | wc -l
21330
$ find drivers -name \*.o.cmd -not -name \*.mod.o.cmd | xargs grep -l mod_devicetable.h | wc -l
17038

The result is that even when touching an obscure device id struct most
of the kernel needs to be recompiled. Given that each driver typically
only needs one or two of these structures, splitting into per subsystem
headers and only including what is really needed reduces the amount of
needed recompilation.

Implement the first step and define each device id struct in a separate
header (together with its associated #defines).

<linux/mod_devicetable.h> is modified to include all the new headers to
continue to provide the same symbols.

Several headers currently include <linux/mod_devicetable.h>, those that
are most lukrative to include only their subsystem headers only are:

$ git -C source grep -l mod_devicetable.h include/linux | while read h; do echo -n "$h:"; find drivers -name \*.o.cmd -not -name \*.mod.o.cmd | xargs grep -l $h | wc -l; done | sort -t: -k2 -n -r | head
include/linux/of.h:10897
include/linux/pci.h:7920
include/linux/acpi.h:7097
include/linux/i2c.h:5402
include/linux/spi/spi.h:1897
include/linux/dmi.h:1643
include/linux/usb.h:1222
include/linux/input.h:1205
include/linux/mdio.h:835
include/linux/phy.h:733

struct cpu_feature isn't really a device_id struct. That is kept in
<linux/mod_devicetable.h> for now.

Acked-by: Danilo Krummrich <dakr@kernel.org>
Acked-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org> # zorro
Link: https://patch.msgid.link/41400e323be8640702b906d04327e833c5bdaf4a.1782808461.git.u.kleine-koenig@baylibre.com
[Drop "MOD" from the header guards]
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
2 months agoata: libata-scsi: limit simulated SCSI command copy to response length
Karuna Ramkumar [Thu, 2 Jul 2026 02:01:42 +0000 (02:01 +0000)]
ata: libata-scsi: limit simulated SCSI command copy to response length

The function ata_scsi_rbuf_fill() is used to copy the response of
emulated SCSI commands from ata_scsi_rbuf to the SCSI command's
scatterlist.

Currently, sg_copy_from_buffer() is called with the size argument
set to ATA_SCSI_RBUF_SIZE (2048 bytes). Since ata_scsi_rbuf is
zeroed out before the simulation actor is invoked, copying the
full buffer size causes the remainder of the SCSI command's
transfer buffer (beyond the actual response length 'len') to be
overwritten with zeroes. This clobbers any pre-existing sentinel
values or data in the caller's buffer tail, even though the
correct residual count is reported via scsi_set_resid().

Fix this by passing the actual response length 'len' as the copy
size to sg_copy_from_buffer(), ensuring that the tail of the
caller's buffer remains untouched. Also, add a defensive check
to ensure that the actor does not return a length exceeding the
static buffer capacity. If this occurs, trigger a WARN_ON(),
fail the command with an aborted command error, and return
immediately without copying any data.

The fix was tested by invoking an SCSI SG_IO INQUIRY on
an ATA disk on vanilla build, and build with the fix. Confirmed
that the input buffer's tail end remains unmodified with the fix.

Fixes: 5251ae224d8d ("ata: libata-scsi: Return residual for emulated SCSI commands")
Assisted-by: Antigravity:gemini-3.5-flash
Signed-off-by: Karuna Ramkumar <rkaruna@google.com>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
2 months agoata: pata_pxa: Fix DMA channel leak on probe error
Wentao Liang [Thu, 25 Jun 2026 14:18:37 +0000 (22:18 +0800)]
ata: pata_pxa: Fix DMA channel leak on probe error

When dmaengine_slave_config() fails, the DMA channel acquired by
dma_request_chan() is not released before returning the error,
leaking the channel reference.

Fix by adding dma_release_channel() in the error path.

The ata_host_activate() error path already correctly releases the
DMA channel.

Cc: stable@vger.kernel.org
Fixes: 88622d80af82 ("ata: pata_pxa: dmaengine conversion")
Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
2 months agoata: sata_gemini: unwind clocks on IDE pinctrl errors
Myeonghun Pak [Fri, 26 Jun 2026 08:58:37 +0000 (17:58 +0900)]
ata: sata_gemini: unwind clocks on IDE pinctrl errors

gemini_sata_bridge_init() prepares and enables both SATA PCLKs, then
disables them again while keeping the clocks prepared for later bridge
start and stop operations. If gemini_setup_ide_pins() fails after that,
gemini_sata_probe() returns directly and skips the existing
out_unprep_clk unwind path.

Route the IDE pinctrl failure through out_unprep_clk so the clocks
prepared by gemini_sata_bridge_init() are unprepared before probe
fails.

Fixes: d872ced29d5f ("ata: sata_gemini: Introduce explicit IDE pin control")
Co-developed-by: Ijae Kim <ae878000@gmail.com>
Signed-off-by: Ijae Kim <ae878000@gmail.com>
Signed-off-by: Myeonghun Pak <mhun512@gmail.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Linus Walleij <linusw@kernel.org>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
2 months agoata: libata-core: Reject an invalid concurrent positioning ranges count
Bryam Vargas [Tue, 23 Jun 2026 03:23:45 +0000 (22:23 -0500)]
ata: libata-core: Reject an invalid concurrent positioning ranges count

ata_dev_config_cpr() takes the number of range descriptors from buf[0]
of the concurrent positioning ranges log (up to 255), which the device
reports independently of the log size in the GPL directory. The count is
then walked at a fixed 32-byte stride in two places with no bound: the
log read here, and the INQUIRY VPD page B9h emitter, which writes one
descriptor per range into the fixed 2048-byte ata_scsi_rbuf. A device
reporting a count larger than its own log overflows the read buffer (up
to 7704 bytes past a 512-byte slab), and a count above 62 overflows the
response buffer on the emit side.

Bound the count once, on probe, against both the log the device returned
and the number of descriptors the VPD B9h response buffer can hold
(ATA_DEV_MAX_CPR, derived from the rbuf size). Reject an out-of-range
count with a warning; this keeps the emitter in bounds with no separate
change there.

Suggested-by: Damien Le Moal <dlemoal@kernel.org>
Fixes: fe22e1c2f705 ("libata: support concurrent positioning ranges log")
Fixes: c745dfc541e7 ("libata: fix reading concurrent positioning ranges log")
Cc: stable@vger.kernel.org
Signed-off-by: Bryam Vargas <hexlabsecurity@proton.me>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>