amdkfd: Only ignore -ENOENT for KFD init failuires
When compiled without CONFIG_HSA_AMD KFD will return -ENOENT.
As other errors will cause KFD functionality issues this is the
only error code that should be ignored at init.
Reviewed-by: Kent Russell <kent.russell@amd.com> Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 4259a25341abf77939767215706f4e3cfd4b73b8)
drm/amdgpu: avoid double drm_exec_fini() in userq validate
When new_addition is true, amdgpu_userq_vm_validate() calls
drm_exec_fini(&exec) before iterating over the collected HMM ranges and
calling amdgpu_ttm_tt_get_user_pages().
If amdgpu_ttm_tt_get_user_pages() fails in that path, the code jumps to
unlock_all and calls drm_exec_fini(&exec) a second time on the same
exec object. drm_exec_fini() is not idempotent: it frees exec->objects
and may also drop exec->contended and finalize the ww acquire context.
Route that error path directly to the range cleanup once exec has
already been finalized.
Fixes: 42f148788469 ("drm/amdgpu/userqueue: validate userptrs for userqueues")
Issue found using a prototype static analysis tool
and confirmed by code review.
Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Hongyan Xu <getshell@seu.edu.cn> Signed-off-by: Slavin Liu <220245772@seu.edu.cn> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 2802952e4a07306da6ebe813ff1acacc5691851a)
Fixes: e56e3cff2a1b ("drm/amd/display: Sync dcn42 with DC 3.2.373") Cc: Timur Kristóf <timur.kristof@gmail.com> Signed-off-by: Roman Li <Roman.Li@amd.com> Reviewed-by: Alex Hung <alex.hung@amd.com> Reviewed-by: Timur Kristóf <timur.kristof@gmail.com> Tested-by: Timur Kristóf <timur.kristof@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit cea8349e4494d2892ea57eef3fe4a8987464a876)
Arjan van de Ven [Mon, 20 Apr 2026 21:57:15 +0000 (14:57 -0700)]
drm/amdgpu: fix zero-size GDS range init on RDNA4
RDNA4 (GFX 12) hardware removes the GDS, GWS, and OA on-chip memory
resources. The gfx_v12_0 initialisation code correctly leaves
adev->gds.gds_size, adev->gds.gws_size, and adev->gds.oa_size at
zero to reflect this.
amdgpu_ttm_init() unconditionally calls amdgpu_ttm_init_on_chip() for
each of these resources regardless of size. When the size is zero,
amdgpu_ttm_init_on_chip() forwards the call to ttm_range_man_init(),
which calls drm_mm_init(mm, 0, 0). drm_mm_init() immediately fires
DRM_MM_BUG_ON(start + size <= start) -- trivially true when size is
zero -- crashing the kernel during modprobe of amdgpu on an RX 9070 XT.
Guard against this by returning 0 early from
amdgpu_ttm_init_on_chip() when size_in_page is zero. This skips TTM
resource manager registration for hardware resources that are absent,
without affecting any other GPU type.
DRM_MM_BUG_ON() only asserts if CONFIG_DRM_DEBUG_MM is enabled in
the kernel config. This is apparently rarely enabled as these chips
have been in the market for over a year and this issue was only reported
now.
Link: https://lore.kernel.org/all/bug-221376-2300@https.bugzilla.kernel.org%2F/ Link: https://bugzilla.kernel.org/show_bug.cgi?id=221376
Oops-Analysis: http://oops.fenrus.org/reports/bugzilla.korg/221376/report.html Assisted-by: GitHub Copilot:Claude Sonnet 4.6 linux-kernel-oops-x86. Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: "Christian König" <christian.koenig@amd.com> Cc: amd-gfx@lists.freedesktop.org Cc: dri-devel@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 5719ce5865279cad4fd5f01011fe037168503f2d) Cc: stable@vger.kernel.org
YuanShang [Thu, 26 Mar 2026 10:27:30 +0000 (18:27 +0800)]
drm/amdkfd: check if vm ready in svm map and unmap to gpu
Don't map or unmap svm range to gpu if vm is not ready for updates.
Why: DRM entity may already be killed when the svm worker try to
update gpu vm.
Signed-off-by: YuanShang <YuanShang.Mao@amd.com> Reviewed-by: Philip Yang <philip.yang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdkfd: Limit per-process xnack support for GFX12.1
GFX12.1 doesn't support different processes with different xnack
setting. Hence, xnack changes from its default setting needs driver
reload with noretry module parameter.
However, if driver boots-up with xnack on and process sets xnack off,
this condition is supported. In this case, only driver takes xnack off
path whereas hardware / firmware considers xnack is on.
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com> Reviewed-by: Philip.Yang <Philip.Yang@amd.com> Reviewed-by: Felix Kuehling <felix.kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Hawking Zhang [Fri, 29 Aug 2025 04:07:54 +0000 (12:07 +0800)]
drm/amdgpu: Add nbio v6_3_2 ip headers v4
Add header files for nbio v6_3_2 register offsets
and shift masks
v2: Update nbio v6_3_2 ip headers up to CL7337280
v3: Update nbio v6_3_2 ip headers up to CL7749557
v4: Clean up registers (Alex)
v5: squash in updates (Alex)
drm/amdgpu: Avoid reset in AMDGPU unload path for APUs with GFX V11 and higher.
GFX V11 has GC block as default off IP.
Every time AMDGPU driver sends a request to PMFW
to unload MP1, PMFW will put GC in reset and
power down the voltage.Hence, skipping reset
for APUs with GFX V11 or later to avoid reset
related failures.
Fixes: 34355e61835e ("drm/amdgpu: Fix GFX hang on SteamDeck when amdgpu is reloaded") Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Shubhankar Milind Sardeshpande <Shubhankar.MilindSardeshpande@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Kent Russell [Wed, 22 Apr 2026 13:34:04 +0000 (09:34 -0400)]
drm/amdgpu: Only send RMA CPER when threshold is exceeded
According to our documentation, the RMA should only occur when the
threshold has been exceeded, not met.
Fixes: 5028a24aa89a ("drm/amdgpu: Send applicable RMA CPERs at end of RAS init") Signed-off-by: Kent Russell <kent.russell@amd.com> Reviewed-by: Tao Zhou <tao.zhou1@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amd/display: Fix unsigned underflow in SubVP vblank schedulability check v2
subvp_vblank_schedulable() checks whether the SubVP active region can
fit the prefetch time, vblank frame time, and the larger of the vblank
blanking time and MALL region.
Commit 11236ac905e7 ("drm/amd/display: Fix dml2_0 narrowing boundaries")
changed the intermediate timing variables from u16 to u32. After that
change, the schedulability test:
is evaluated in unsigned arithmetic. If the required time exceeds the
available active time, the subtraction can underflow and wrap, producing
a large positive value instead of a negative result.
Fix this by comparing the available time against the required time
directly, using u64 for the accumulated required duration.
v2:
- Use uint64_t instead of u64 for consistency with DC style (Dillon)
- Make all terms explicitly uint64_t in the accumulated sum & use
single-definition form for required_us (Gaghik)
Fixes: 11236ac905e7 ("drm/amd/display: Fix dml2_0 narrowing boundaries") Reported-by: Dan Carpenter <error27@gmail.com> Cc: Roman Li <roman.li@amd.com> Cc: Alex Hung <alex.hung@amd.com> Cc: Tom Chung <chiahsuan.chung@amd.com> Cc: Dillon Varone <dillon.varone@amd.com> Cc: Gaghik Khachatrian <gaghik.khachatrian@amd.com> Cc: Aurabindo Pillai <aurabindo.pillai@amd.com> Cc: Chenyu Chen <chen-yu.chen@amd.com> Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com> Reviewed-by: Dillon Varone <dillon.varone@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This can trigger a divide-by-zero bug if an invalid refresh rate is
passed in.
Rewrite the calculation to directly compute the number of frames needed
for at least 30 ms using DIV_ROUND_UP(). This avoids the invalid divisor
and removes the temporary frame_time_microsec calculation.
Keep the existing fallback behavior of 2 static frames when
vsync_rate_hz is zero.
Fixes: 4cef2ac4c795 ("drm/amd/display: Introduce power module on Linux") Reported-by: Dan Carpenter <error27@gmail.com> Cc: Leo Li <sunpeng.li@amd.com> Cc: Ray Wu <ray.wu@amd.com> Cc: Chenyu Chen <chen-yu.chen@amd.com> Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com> Reviewed-by: Leo Li <sunpeng.li@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: fix root reservation in amdgpu_vm_handle_fault
svm_range_restore_pages might reserve the root bo so it must
be called after unreserving it.
Fixes: 1b135c6da061 ("drm/amdgpu: extract amdgpu_vm_lock_by_pasid from amdgpu_vm_handle_fault") Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Timur Kristóf [Sat, 18 Apr 2026 21:49:33 +0000 (23:49 +0200)]
drm/amdgpu/gfx6: Support harvested SI chips with disabled TCCs (v2)
This commit fixes amdgpu to work on the Radeon HD 7870 XT
which has never worked with the Linux open source drivers before.
Some boards have "harvested" chips, meaning that some parts of
the chip are disabled and fused, and it's sold for cheaper and
under a different marketing name.
On a harvested chip, any of the following can be disabled:
- CUs (Compute Units)
- RBs (Render Backend, aka. ROP)
- Memory channels (ie. the chip has a lower bandwidth)
- TCCs (ie. less L2 cache)
Handle chips with harvested TCCs by patching the registers
that configure how TCCs are mapped.
If some TCCs are disabled, we need to make sure that
the disabled TCCs are not used, and the remaining TCCs
are used optimally.
TCP_CHAN_STEER_LO/HI control which TCC is used by TCP channels.
TCP_ADDR_CONFIG.NUM_TCC_BANKS controls how many channels are used.
Note that the TCC configuration is highly relevant to performance.
Suboptimal configuration (eg. CHAN_STEER=0) can significantly
reduce gaming performance.
For optimal performance:
- Rely on the CHAN_STEER from the golden registers table,
only skip disabled TCCs but keep the mapping order.
- Limit NUM_TCC_BANKS to number of active TCCs to avoid thrashing,
which performs better than using the same TCC twice.
v2:
- Also consider CGTS_USER_TCC_DISABLE for disabled TCCs.
Link: https://bugs.freedesktop.org/show_bug.cgi?id=60879 Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/2664 Fixes: 2cd46ad22383 ("drm/amdgpu: add graphic pipeline implementation for si v8") Signed-off-by: Timur Kristóf <timur.kristof@gmail.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Timur Kristóf [Sat, 18 Apr 2026 21:49:32 +0000 (23:49 +0200)]
Documentation/gpu: Add TCC, update TCP in amdgpu glossary
These are the L2 and L1 cache on some AMD GPU architectures.
Add them to the glossary, keeping existing alphabetical order.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Timur Kristóf [Sat, 18 Apr 2026 21:49:31 +0000 (23:49 +0200)]
drm/amdgpu/uvd3.1: Don't validate the firmware when already validated
UVD 3.1 firmware validation seems to always fail after
attempting it when it had already been validated.
(This works similarly with the VCE 1.0 as well.)
Don't attempt repeating the validation when it's already done.
This caused issues in situations when the system isn't able
to suspend the GPU properly and so the GPU isn't actually
powered down. Then amdgpu would fail when calling the IP
block resume function.
Closes: https://gitlab.freedesktop.org/drm/amd/-/work_items/2887 Fixes: bb7978111dd3 ("drm/amdgpu: fix SI UVD firmware validate resume fail") Signed-off-by: Timur Kristóf <timur.kristof@gmail.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Christian König [Fri, 17 Apr 2026 13:52:45 +0000 (15:52 +0200)]
drm/amdgpu: fix AMDGPU_INFO_READ_MMR_REG
There were multiple issues in that code.
First of all the order between the reset semaphore and the mm_lock was
wrong (e.g. copy_to_user) was called while holding the lock.
Then we allocated memory while holding the reset semaphore which is also
a pretty big bug and can deadlock.
Then we used down_read_trylock() instead of waiting for the reset to
finish.
Signed-off-by: Christian König <christian.koenig@amd.com> Fixes: 9e823f307074 ("drm/amdgpu: Block MMR_READ IOCTL in reset") Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Yang Wang [Wed, 22 Apr 2026 10:41:42 +0000 (18:41 +0800)]
drm/amd/pm: fix missing fine-grained dpm table flag on aldebaran
Add the missing SMU_DPM_TABLE_FINE_GRAINED flag to aldebaran DPM table.
This fixes the pp_dpm_sclk node issue caused by missing flag configuration.
Fixes: 7ea1c722fe1d ("drm/amd/pm: Use common helper for aldebaran dpm table") Signed-off-by: Yang Wang <kevinyang.wang@amd.com> Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Yang Wang [Wed, 18 Mar 2026 00:58:07 +0000 (20:58 -0400)]
drm/amd/pm: remove trailing semicolon from AMDGPU_PM_POLICY_ATTR macro
macros should not include a trailing semicolon as per kernel coding
style (checkpatch.pl warning).
move the semicolon from the macro definition to the invocation sites instead.
checkpatch.pl logs:
WARNING: macros should not use a trailing semicolon
+#define AMDGPU_PM_POLICY_ATTR(_name, _id) \
+ static struct amdgpu_pm_policy_attr pm_policy_attr_##_name = { \
+ .dev_attr = __ATTR(_name, 0644, amdgpu_get_pm_policy_attr, \
+ amdgpu_set_pm_policy_attr), \
+ .id = PP_PM_POLICY_##_id, \
+ };
Fixes: 4d154b1ca580 ("drm/amd/pm: Add support for DPM policies") Signed-off-by: Yang Wang <kevinyang.wang@amd.com> Acked-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Timur Kristóf [Sat, 18 Apr 2026 21:49:30 +0000 (23:49 +0200)]
drm/amdgpu/gmc: Fix AMDGPU_GART_PLACEMENT_LOW to not overlap with VRAM
When the GART placement is set to AMDGPU_GART_PLACEMENT_LOW:
Make sure that GART does not overlap with VRAM when
VRAM is configured to be in the low address space.
Solve this according to the following logic:
- When GART fits before VRAM, use zero address for GART
- Otherwise, put GART after the end of VRAM, aligned to 4 GiB
Previously, I had assumed this was not possible
so it was OK to not handle it, but now we got a report
from a user who has a board that is configured this way.
Fixes: 917f91d8d8e8 ("drm/amdgpu/gmc: add a way to force a particular placement for GART") Signed-off-by: Timur Kristóf <timur.kristof@gmail.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
amdkfd: Only ignore -ENOENT for KFD init failuires
When compiled without CONFIG_HSA_AMD KFD will return -ENOENT.
As other errors will cause KFD functionality issues this is the
only error code that should be ignored at init.
Reviewed-by: Kent Russell <kent.russell@amd.com> Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: avoid double drm_exec_fini() in userq validate
When new_addition is true, amdgpu_userq_vm_validate() calls
drm_exec_fini(&exec) before iterating over the collected HMM ranges and
calling amdgpu_ttm_tt_get_user_pages().
If amdgpu_ttm_tt_get_user_pages() fails in that path, the code jumps to
unlock_all and calls drm_exec_fini(&exec) a second time on the same
exec object. drm_exec_fini() is not idempotent: it frees exec->objects
and may also drop exec->contended and finalize the ww acquire context.
Route that error path directly to the range cleanup once exec has
already been finalized.
Fixes: 42f148788469 ("drm/amdgpu/userqueue: validate userptrs for userqueues")
Issue found using a prototype static analysis tool
and confirmed by code review.
Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Hongyan Xu <getshell@seu.edu.cn> Signed-off-by: Slavin Liu <220245772@seu.edu.cn> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fixes: e56e3cff2a1b ("drm/amd/display: Sync dcn42 with DC 3.2.373") Cc: Timur Kristóf <timur.kristof@gmail.com> Signed-off-by: Roman Li <Roman.Li@amd.com> Reviewed-by: Alex Hung <alex.hung@amd.com> Reviewed-by: Timur Kristóf <timur.kristof@gmail.com> Tested-by: Timur Kristóf <timur.kristof@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Arjan van de Ven [Mon, 20 Apr 2026 21:57:15 +0000 (14:57 -0700)]
drm/amdgpu: fix zero-size GDS range init on RDNA4
RDNA4 (GFX 12) hardware removes the GDS, GWS, and OA on-chip memory
resources. The gfx_v12_0 initialisation code correctly leaves
adev->gds.gds_size, adev->gds.gws_size, and adev->gds.oa_size at
zero to reflect this.
amdgpu_ttm_init() unconditionally calls amdgpu_ttm_init_on_chip() for
each of these resources regardless of size. When the size is zero,
amdgpu_ttm_init_on_chip() forwards the call to ttm_range_man_init(),
which calls drm_mm_init(mm, 0, 0). drm_mm_init() immediately fires
DRM_MM_BUG_ON(start + size <= start) -- trivially true when size is
zero -- crashing the kernel during modprobe of amdgpu on an RX 9070 XT.
Guard against this by returning 0 early from
amdgpu_ttm_init_on_chip() when size_in_page is zero. This skips TTM
resource manager registration for hardware resources that are absent,
without affecting any other GPU type.
DRM_MM_BUG_ON() only asserts if CONFIG_DRM_DEBUG_MM is enabled in
the kernel config. This is apparently rarely enabled as these chips
have been in the market for over a year and this issue was only reported
now.
Link: https://lore.kernel.org/all/bug-221376-2300@https.bugzilla.kernel.org%2F/ Link: https://bugzilla.kernel.org/show_bug.cgi?id=221376
Oops-Analysis: http://oops.fenrus.org/reports/bugzilla.korg/221376/report.html Assisted-by: GitHub Copilot:Claude Sonnet 4.6 linux-kernel-oops-x86. Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: "Christian König" <christian.koenig@amd.com> Cc: amd-gfx@lists.freedesktop.org Cc: dri-devel@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm/amdgpu: Replace VKMS vblank timer with common implementation
Replace amdgpu's custom vblank timers with the shared implementation
in DRM's vblank code. Both are built upon hrtimers. The vblank logic
is identical.
The shared helpers contain all initialization internally. They also
handle a number of deadlocks and race conditions that are present in
amdgpu.
Also remove the set-but-unused field vsync_timer_enabled from struct
amdgpu_crtc.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
All known issues have been adressed.
Allow to select RT.
Acked-by: Linus Walleij <linus.walleij@linaro.org> Reviewed-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Discussions in the mailing list mentioned that arch/arm is also affected
but the fix was never applied to it [1][2]. Apply the change now, since
the race condition can cause sporadic SIGILL's and SEGV's especially
while under high memory pressure.
Dave Hansen [Tue, 21 Apr 2026 15:19:09 +0000 (08:19 -0700)]
x86/mm: Revert INVLPGB optimization for set_memory code
tl;dr: Revert an INVLPGB optimization that did not properly handle
discontiguous virtual addresses.
Full story:
I got a report from some graphics (i915) folks that bisected a
regression in their test suite to 86e6815b316e ("x86/mm: Change
cpa_flush() to call flush_kernel_range() directly"). There was a bit
of flip-flopping on the exact bisect, but the code here does seem
wrong to me. The i915 folks were calling set_pages_array_wc(), so
using the CPA_PAGES_ARRAY mode.
Basically, the 'struct cpa_data' can wrap up all kinds of page table
changes. Some of these are virtually contiguous, but some are very
much not which is one reason why there are ->vaddr and ->pages arrays.
86e6815b316e made the mistake of assuming that the virtual addresses
in the cpa_data are always contiguous. It got things right when neither
CPA_ARRAY/CPA_PAGES_ARRAY is used, but theoretically wrong when either
of those is used.
In the i915 case, it probably failed to flush some WB TLB entries and
install WC ones, leaving some data in the caches and not flushing it
out to where the device could see it. That eventually caused graphics
problems.
Revert the INVLPGB optimization. It can be reintroduced later, but it
will need to be a bit careful about the array modes.
Fixes: 86e6815b316ec ("x86/mm: Change cpa_flush() to call flush_kernel_range()") Reported-by: Cui, Ling <ling.cui@intel.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Signed-off-by: Ingo Molnar <mingo@kernel.org> Reviewed-by: Rick Edgecombe <rick.p.edgecombe@intel.com> Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> Link: https://patch.msgid.link/20260421151909.6B3281C6@davehans-spike.ostc.intel.com
Thomas Weißschuh [Wed, 22 Apr 2026 15:10:27 +0000 (17:10 +0200)]
kbuild: Never respect CONFIG_WERROR / W=e to fixdep
The fixdep hostprog may be built multiple times during a single build.
Once during the configuration phase and later during the regular phase.
As only the regular build phase respects CONFIG_WERROR / W=e, the
compiler flags might change between the phases, leading to rebuilds.
Example, the rebuilds will happen twice on each invocation of the build:
$ make allyesconfig prepare
make[1]: Entering directory '/tmp/deleteme'
HOSTCC scripts/basic/fixdep
#
# No change to .config
#
HOSTCC scripts/basic/fixdep
DESCEND objtool
INSTALL libsubcmd_headers
make[1]: Leaving directory '/tmp/deleteme'
Fix the compilation flags used for scripts/basic/ before
scripts/Makefile.warn is evaluated to stop CONFIG_WERROR / W=e
influencing the fixdep build to avoid the spurious rebuilds.
Marc Zyngier [Thu, 23 Apr 2026 16:36:07 +0000 (17:36 +0100)]
KVM: arm64: Wake-up from WFI when iqrchip is in userspace
It appears that there is nothing in the wake-up path that
evaluates whether the in-kernel interrupts are pending unless
we have a vgic.
This means that the userspace irqchip support has been broken for
about four years, and nobody noticed. It was also broken before
as we wouldn't wake-up on a PMU interrupt, but hey, who cares...
It is probably time to remove the feature altogether, because it
was a terrible idea 10 years ago, and it still is.
KVM: arm64: Fix initialisation order in __pkvm_init_finalise()
fix_host_ownership() walks the hypervisor's stage-1 page-table to
adjust the host's stage-2 accordingly. Any such adjustment that
requires cache maintenance operations depends on the per-CPU hyp
fixmap being present. However, fix_host_ownership() is currently
called before fix_hyp_pgtable_refcnt() and hyp_create_fixmap(), so
the fixmap does not yet exist when it runs.
This is benign today because the host stage-2 starts empty and no
CMOs are needed, but it becomes a latent crash as soon as
fix_host_ownership() is extended to operate on a non-empty
page-table.
Reorder the calls so that fix_hyp_pgtable_refcnt() and
hyp_create_fixmap() complete before fix_host_ownership() is invoked.
Fixes: 0d16d12eb26e ("KVM: arm64: Fix-up hyp stage-1 refcounts for all pages mapped at EL2") Signed-off-by: Quentin Perret <qperret@google.com> Signed-off-by: Fuad Tabba <tabba@google.com> Link: https://patch.msgid.link/20260424084908.370776-7-tabba@google.com Signed-off-by: Marc Zyngier <maz@kernel.org> Cc: stable@vger.kernel.org
KVM: arm64: Fix pin leak and publication ordering in __pkvm_init_vcpu()
Two bugs exist in the vCPU initialisation path:
1. If a check fails after hyp_pin_shared_mem() succeeds, the cleanup
path jumps to 'unlock' without calling unpin_host_vcpu() or
unpin_host_sve_state(), permanently leaking pin references on the
host vCPU and SVE state pages.
Extract a register_hyp_vcpu() helper that performs the checks and
the store. When register_hyp_vcpu() returns an error, call
unpin_host_vcpu() and unpin_host_sve_state() inline before falling
through to the existing 'unlock' label.
2. register_hyp_vcpu() publishes the new vCPU pointer into
'hyp_vm->vcpus[]' with a bare store, allowing a concurrent caller
of pkvm_load_hyp_vcpu() to observe a partially initialised vCPU
object.
Ensure the store uses smp_store_release() and the load uses
smp_load_acquire(). While 'vm_table_lock' currently serialises the
store and the load, these barriers ensure the reader sees the fully
initialised 'hyp_vcpu' object even if there were a lockless path or
if the lock's own ordering guarantees were insufficient for nested
object initialization.
Fixes: 49af6ddb8e5c ("KVM: arm64: Add infrastructure to create and track pKVM instances at EL2") Reported-by: Ben Simner <ben.simner@cl.cam.ac.uk> Co-developed-by: Will Deacon <willdeacon@google.com> Signed-off-by: Will Deacon <willdeacon@google.com> Signed-off-by: Fuad Tabba <tabba@google.com> Link: https://patch.msgid.link/20260424084908.370776-6-tabba@google.com Signed-off-by: Marc Zyngier <maz@kernel.org> Cc: stable@vger.kernel.org
The macro is defined with parameter 'v' but the body references the
literal token 'vcpu' instead, causing it to silently operate on whatever
'vcpu' resolves to in the caller's scope rather than the value passed by
the caller. All current call sites happen to use a variable named 'vcpu',
so the bug is latent.
Fixes: e016333745c7 ("KVM: arm64: Only reset vCPU-scoped feature ID regs once") Signed-off-by: Fuad Tabba <tabba@google.com> Link: https://patch.msgid.link/20260424084908.370776-5-tabba@google.com Signed-off-by: Marc Zyngier <maz@kernel.org> Cc: stable@vger.kernel.org
KVM: arm64: Fix FEAT_SPE_FnE to use PMSIDR_EL1.FnE, not PMSVer
FEAT_SPE_FnE is architecturally detected via PMSIDR_EL1.FnE [6], not
ID_AA64DFR0_EL1.PMSVer. The FEAT_X macro form (register, field, value)
cannot encode a PMSIDR_EL1-based feature, so FEAT_SPE_FnE was defined
identically to FEAT_SPEv1p2 (ID_AA64DFR0_EL1, PMSVer, V1P2), producing
a duplicate that used PMSVer >= V1P2 as a proxy.
Replace the macro with feat_spe_fne(), following the same pattern as
the sibling feat_spe_fds(): guard on FEAT_SPEv1p2 and read
PMSIDR_EL1.FnE [6] directly. Wire the two NEEDS_FEAT consumers to use
the new function.
KVM: arm64: Fix FEAT_Debugv8p9 to check DebugVer, not PMUVer
FEAT_Debugv8p9 is incorrectly defined against ID_AA64DFR0_EL1.PMUVer
instead of ID_AA64DFR0_EL1.DebugVer. All three consumers of the macro
gate features that are architecturally tied to FEAT_Debugv8p9
(DebugVer = 0b1011, DDI0487 M.b A2.2.10):
- HDFGRTR2_EL2.nMDSELR_EL1, HDFGWTR2_EL2.nMDSELR_EL1: MDSELR_EL1
is present only when FEAT_Debugv8p9 is implemented (D24.3.21).
- MDCR_EL2.EBWE: the Extended Breakpoint and Watchpoint Enable bit
is RES0 unless FEAT_Debugv8p9 is implemented (D24.3.17).
Neither register has any dependency on PMUVer.
FEAT_Debugv8p9 and FEAT_PMUv3p9 are independent. Per DDI0487 M.b
A2.2.10, FEAT_Debugv8p9 is unconditionally mandatory from Armv8.9,
whereas FEAT_PMUv3p9 is mandatory only when FEAT_PMUv3 is implemented.
An Armv8.9 CPU without a PMU has DebugVer = 0b1011 but PMUVer = 0b0000,
so the wrong field check would cause KVM to incorrectly treat EBWE and
MDSELR_EL1 as RES0 on such hardware.
Sebastian Ene [Wed, 8 Apr 2026 11:41:18 +0000 (11:41 +0000)]
KVM: arm64: Reject non compliant SMCCC function calls in pKVM
Prevent the propagation of a function-id that has the top bits set since
this is not compliant with the SMCCC spec and can overlap with the
already known function-id decoders. (eg. if we invoke an smc with
0xffffffffc4000012 it will be decoded as a PSCI reset call). Instead,
make it clear that we don't support it and return an error.
David Woodhouse [Tue, 7 Apr 2026 20:27:02 +0000 (21:27 +0100)]
KVM: arm64: vgic: Fix IIDR revision field extracted from wrong value
The uaccess write handlers for GICD_IIDR in both GICv2 and GICv3
extract the revision field from 'reg' (the current IIDR value read back
from the emulated distributor) instead of 'val' (the value userspace is
trying to write). This means userspace can never actually change the
implementation revision — the extracted value is always the current one.
Fix the FIELD_GET to use 'val' so that userspace can select a different
revision for migration compatibility.
Fixes: 49a1a2c70a7f ("KVM: arm64: vgic-v3: Advertise GICR_CTLR.{IR, CES} as a new GICD_IIDR revision") Signed-off-by: David Woodhouse <dwmw@amazon.co.uk> Link: https://patch.msgid.link/20260407210949.2076251-2-dwmw2@infradead.org Signed-off-by: Marc Zyngier <maz@kernel.org> Cc: stable@vger.kernel.org
Jani Nikula [Wed, 15 Apr 2026 11:51:03 +0000 (14:51 +0300)]
drm/i915/ddi: add wrapper for calling ->get_buf_trans() hook
Add intel_ddi_buf_trans_get() wrapper for calling the
encoder->get_buf_trans() hooks, allowing for a centralized place to
e.g. check for VBT overrides for the buf translations.
Various names for Qualcomm as a company are used in user-visible config
options: QCOM, Qualcomm and Qualcomm Technologies. Switch to unified
"Qualcomm" so it will be easier for users to identify the options when
for example running menuconfig.
Dave Airlie [Fri, 24 Apr 2026 03:17:50 +0000 (13:17 +1000)]
Merge tag 'drm-misc-fixes-2026-04-23' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-fixes
This week in drm-misc-fixes, we have:
- A patch to raise the vblank timeout to avoid it on virtual drivers
- a state check fix for stm_lvds
- a use-after-free fix for dma-buf
- a mapping fix for panthor
- a device_node reference leak fix for arcgpu
- a bridge reference leak fix for dw-mipi-dsi
- a sparse warning fix for dma-fence
- a kconfig fix for hv
- a memory leak fix for nouveau
- a fix to duplicate colorop when duplicating states
- a panel initialisation order fix for visionox-rm69299
- a fix to prevent an infinite loop for v3d
- an overflow fix for nouveau
Weiming Shi [Thu, 23 Apr 2026 18:38:32 +0000 (11:38 -0700)]
bpf: Fix NULL pointer dereference in bpf_skb_fib_lookup()
When tot_len is not provided by the user, bpf_skb_fib_lookup()
resolves the FIB result's output device via dev_get_by_index_rcu()
to check skb forwardability and fill in mtu_result. The returned
pointer is dereferenced without a NULL check. If the device is
concurrently unregistered, dev_get_by_index_rcu() returns NULL and
is_skb_forwardable() crashes at dev->flags:
KASAN: null-ptr-deref in range
[0x00000000000000b0-0x00000000000000b7]
Call Trace:
is_skb_forwardable (include/linux/netdevice.h:4365)
bpf_skb_fib_lookup (net/core/filter.c:6446)
bpf_prog_test_run_skb (net/bpf/test_run.c)
__sys_bpf (kernel/bpf/syscall.c)
Add the missing NULL check, returning -ENODEV to be consistent
with how bpf_ipv4_fib_lookup() and bpf_ipv6_fib_lookup() handle
the same condition.
Fixes: 4f74fede40df ("bpf: Add mtu checking to FIB forwarding helper") Reported-by: Xiang Mei <xmei5@asu.edu> Signed-off-by: Weiming Shi <bestswngs@gmail.com> Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org> Acked-by: Paul Chaignon <paul.chaignon@gmail.com> Link: https://patch.msgid.link/20260423183831.1325480-2-bestswngs@gmail.com
sockmap: Fix sk_psock_drop() race vs sock_map_{unhash,close,destroy}().
syzbot reported a splat in sock_map_destroy() [0], where psock was
NULL even though sk->sk_prot still pointed to tcp_bpf_prots[][].
The stack trace shows how badly the path was excercised, see
inet_release() calls tcp_close(), not sock_map_close() yet, but
finally reaching sock_map_destroy().
The root cause is a lack of synchronisation.
Even if sk_psock_get() fails to bump psock->refcnt, it does not
guarantee that sk_psock_drop() has finished, and thus sk->sk_prot
might not have been restored to the original one.
Commit 4b4647add7d3 ("sock_map: avoid race between sock_map_close
and sk_psock_put") attempted to address this, but it was insufficient
for two reasons.
It did not cover sock_map_unhash() and sock_map_destroy(), and
it missed the corner case where sk_psock() is NULL.
On non-x86 platforms, sk_psock_restore_proto(sk, psock) and
rcu_assign_sk_user_data(sk, NULL) can be reordered because there
is no address dependency between sk->sk_prot and sk->sk_user_data.
sk_psock_get() returning NULL implies nothing about sk->sk_prot.
Let's simply retry sk_psock_get() in the unlikely case.
Note that we cannot avoid loop even if we added memory barrier
in sk_psock_drop() and sock_map_psock_get_checked().
Also note that sock_map_destroy() cannot be called from softirq
while sock_map_close() has also been running.
It is because sock_map_destroy() requires SOCK_DEAD, so sock_map_destroy()
cannot happen until sock_map_close() has finished the saved_close()
(which is tcp_close()).
Weiming Shi [Wed, 22 Apr 2026 06:54:12 +0000 (23:54 -0700)]
bpf: Fix NULL pointer dereference in bpf_sk_storage_clone and diag paths
bpf_selem_unlink_nofail() sets SDATA(selem)->smap to NULL before
removing the selem from the storage hlist. A concurrent RCU reader in
bpf_sk_storage_clone() can observe the selem still on the list with
smap already NULL, causing a NULL pointer dereference.
general protection fault, probably for non-canonical address 0xdffffc000000000a:
KASAN: null-ptr-deref in range [0x0000000000000050-0x0000000000000057]
RIP: 0010:bpf_sk_storage_clone+0x1cd/0xaa0 net/core/bpf_sk_storage.c:174
Call Trace:
<IRQ>
sk_clone+0xfed/0x1980 net/core/sock.c:2591
inet_csk_clone_lock+0x30/0x760 net/ipv4/inet_connection_sock.c:1222
tcp_create_openreq_child+0x35/0x2680 net/ipv4/tcp_minisocks.c:571
tcp_v4_syn_recv_sock+0x123/0xf90 net/ipv4/tcp_ipv4.c:1729
tcp_check_req+0x8e1/0x2580 include/net/tcp.h:855
tcp_v4_rcv+0x1845/0x3b80 net/ipv4/tcp_ipv4.c:2347
Add a NULL check for smap in bpf_sk_storage_clone().
bpf_sk_storage_diag_put_all() has the same issue. Add a NULL check
and pass the validated smap directly to diag_get(), which is refactored
to take smap as a parameter instead of reading it internally.
bpf_sk_storage_diag_put() uses diag->maps[i] which is always valid
under its refcount, so diag->maps[i] is passed directly to diag_get().
Fixes: 5d800f87d0a5 ("bpf: Support lockless unlink when freeing map or local storage") Reported-by: Xiang Mei <xmei5@asu.edu> Acked-by: Amery Hung <ameryhung@gmail.com> Signed-off-by: Weiming Shi <bestswngs@gmail.com> Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org> Link: https://patch.msgid.link/20260422065411.1007737-2-bestswngs@gmail.com
Merge tag 'vfs-7.1-rc1.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Pull vfs fixes from Christian Brauner:
- eventpoll: fix ep_remove() UAF and follow-up cleanup
- fs: aio: set VMA_DONTCOPY_BIT in mmap to fix NULL-pointer-dereference
error
- writeback: Fix use after free in inode_switch_wbs_work_fn()
- fuse: reject oversized dirents in page cache
- fs: aio: reject partial mremap to avoid Null-pointer-dereference
error
- nstree: fix func. parameter kernel-doc warnings
- fs: Handle multiply claimed blocks more gracefully with mmb
* tag 'vfs-7.1-rc1.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
eventpoll: drop vestigial epi->dying flag
eventpoll: drop dead bool return from ep_remove_epi()
eventpoll: refresh eventpoll_release() fast-path comment
eventpoll: move f_lock acquisition into ep_remove_file()
eventpoll: fix ep_remove struct eventpoll / struct file UAF
eventpoll: move epi_fget() up
eventpoll: rename ep_remove_safe() back to ep_remove()
eventpoll: drop vestigial __ prefix from ep_remove_{file,epi}()
eventpoll: kill __ep_remove()
eventpoll: split __ep_remove()
eventpoll: use hlist_is_singular_node() in __ep_remove()
fs: Handle multiply claimed blocks more gracefully with mmb
nstree: fix func. parameter kernel-doc warnings
fs: aio: reject partial mremap to avoid Null-pointer-dereference error
fuse: reject oversized dirents in page cache
writeback: Fix use after free in inode_switch_wbs_work_fn()
fs: aio: set VMA_DONTCOPY_BIT in mmap to fix NULL-pointer-dereference error
Merge tag 'v7.1-rc-part2-ksmbd-fixes' of git://git.samba.org/ksmbd
Pull more smb server updates from Steve French:
- move fs/smb/common/smbdirect to fs/smb/smbdirect
- change signature calc to use AES-CMAC library, simpler and faster
- invalid signature fix
- multichannel fix
- open create options fix
- fix durable handle leak
- cap maximum lock count to avoid potential denial of service
- four connection fixes: connection free and session destroy IDA fixes,
refcount fix, connection leak fix, max_connections off by one fix
- IPC validation fix
- fix out of bounds write in getting xattrs
- fix use after free in durable handle reconnect
- three ACL fixes: fix potential ACL overflow, harden num_aces check,
and fix minimum ACE size check
* tag 'v7.1-rc-part2-ksmbd-fixes' of git://git.samba.org/ksmbd:
smb: smbdirect: move fs/smb/common/smbdirect/ to fs/smb/smbdirect/
smb: server: stop sending fake security descriptors
ksmbd: scope conn->binding slowpath to bound sessions only
ksmbd: fix CreateOptions sanitization clobbering the whole field
ksmbd: fix durable fd leak on ClientGUID mismatch in durable v2 open
ksmbd: fix O(N^2) DoS in smb2_lock via unbounded LockCount
ksmbd: destroy async_ida in ksmbd_conn_free()
ksmbd: destroy tree_conn_ida in ksmbd_session_destroy()
ksmbd: Use AES-CMAC library for SMB3 signature calculation
ksmbd: reset rcount per connection in ksmbd_conn_wait_idle_sess_id()
ksmbd: fix out-of-bounds write in smb2_get_ea() EA alignment
ksmbd: use check_add_overflow() to prevent u16 DACL size overflow
ksmbd: fix use-after-free in smb2_open during durable reconnect
ksmbd: validate num_aces and harden ACE walk in smb_inherit_dacl()
smb: server: fix max_connections off-by-one in tcp accept path
ksmbd: require minimum ACE size in smb_check_perm_dacl()
ksmbd: validate response sizes in ipc_validate_msg()
smb: server: fix active_num_conn leak on transport allocation failure
Merge tag 'v7.1-rc1-part3-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6
Pull smb client fixes from Steve French:
- Four bug fixes: OOB read in ioctl query info, 3 ACL fixes
- SMB1 Unix extensions mount fix
- Four crypto improvements: move to AES-CMAC library, simpler and faster
- Remove drop_dir_cache to avoid potential crash, and move to /procfs
- Seven SMB3.1.1 compression fixes
* tag 'v7.1-rc1-part3-smb3-client-fixes' of git://git.samba.org/sfrench/cifs-2.6:
smb: client: Drop 'allocate_crypto' arg from smb*_calc_signature()
smb: client: Make generate_key() return void
smb: client: Remove obsolete cmac(aes) allocation
smb: client: Use AES-CMAC library for SMB3 signature calculation
smb: common: add SMB3_COMPRESS_MAX_ALGS
smb: client: compress: add code docs to lz77.c
smb: client: compress: LZ77 optimizations
smb: client: compress: increase LZ77_MATCH_MAX_DIST
smb: client: compress: fix counting in LZ77 match finding
smb: client: compress: fix buffer overrun in lz77_compress()
smb: client: scope end_of_dacl to CIFS_DEBUG2 use in parse_dacl
smb: client: fix (remove) drop_dir_cache module parameter
smb: client: require a full NFS mode SID before reading mode bits
smb: client: validate the whole DACL before rewriting it in cifsacl
smb: client: fix OOB read in smb2_ioctl_query_info QUERY_INFO path
cifs: update internal module version number
smb: client: compress: fix bad encoding on last LZ77 flag
smb: client: fix dir separator in SMB1 UNIX mounts
Merge tag 'net-7.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Pull networking fixes from Jakub Kicinski:
"Including fixes from Netfilter.
Steady stream of fixes. Last two weeks feel comparable to the two
weeks before the merge window. Lots of AI-aided bug discovery. A newer
big source is Sashiko/Gemini (Roman Gushchin's system), which points
out issues in existing code during patch review (maybe 25% of fixes
here likely originating from Sashiko). Nice thing is these are often
fixed by the respective maintainers, not drive-bys.
Current release - new code bugs:
- kconfig: MDIO_PIC64HPSC should depend on ARCH_MICROCHIP
Previous releases - regressions:
- add async ndo_set_rx_mode and switch drivers which we promised to
be called under the per-netdev mutex to it
- dsa: remove duplicate netdev_lock_ops() for conduit ethtool ops
- hv_sock: report EOF instead of -EIO for FIN
- vsock/virtio: fix MSG_PEEK calculation on bytes to copy
Previous releases - always broken:
- ipv6: fix possible UAF in icmpv6_rcv()
- icmp: validate reply type before using icmp_pointers
- af_unix: drop all SCM attributes for SOCKMAP
- netfilter: fix a number of bugs in the osf (OS fingerprinting)
- eth: intel: fix timestamp interrupt configuration for E825C
Misc:
- bunch of data-race annotations"
* tag 'net-7.1-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (148 commits)
rxrpc: Fix error handling in rxgk_extract_token()
rxrpc: Fix re-decryption of RESPONSE packets
rxrpc: Fix rxrpc_input_call_event() to only unshare DATA packets
rxrpc: Fix missing validation of ticket length in non-XDR key preparsing
rxgk: Fix potential integer overflow in length check
rxrpc: Fix conn-level packet handling to unshare RESPONSE packets
rxrpc: Fix potential UAF after skb_unshare() failure
rxrpc: Fix rxkad crypto unalignment handling
rxrpc: Fix memory leaks in rxkad_verify_response()
net: rds: fix MR cleanup on copy error
m68k: mvme147: Make me the maintainer
net: txgbe: fix firmware version check
selftests/bpf: check epoll readiness during reuseport migration
tcp: call sk_data_ready() after listener migration
vhost_net: fix sleeping with preempt-disabled in vhost_net_busy_poll()
ipv6: Cap TLV scan in ip6_tnl_parse_tlv_enc_lim
tipc: fix double-free in tipc_buf_append()
llc: Return -EINPROGRESS from llc_ui_connect()
ipv4: icmp: validate reply type before using icmp_pointers
selftests/net: packetdrill: cover RFC 5961 5.2 challenge ACK on both edges
...
Merge tag 'i2c-for-7.1-rc1-part2' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux
Pull more i2c updates from Wolfram Sang:
- cx92755: convert I2C bindings to DT schema
- mediatek: add optional bus power management during transfers
- pxa: handle early bus busy condition
- MAINTAINERS: update I2C RUST entry
* tag 'i2c-for-7.1-rc1-part2' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
MAINTAINERS: add Rust I2C tree and update Igor Korotin's email
i2c: mediatek: add bus regulator control for power saving
dt-bindings: i2c: cnxt,cx92755-i2c: Convert to DT schema
i2c: pxa: handle 'Early Bus Busy' condition on Armada 3700
Merge tag 'xtensa-20260422' of https://github.com/jcmvbkbc/linux-xtensa
Pull Xtensa updates from Max Filippov:
- use register_sys_off_handler(SYS_OFF_MODE_RESTART) instead of
the deprecated register_restart_handler()
- drop custom ucontext.h and reuse asm-generic ucontext.h
* tag 'xtensa-20260422' of https://github.com/jcmvbkbc/linux-xtensa:
xtensa: uapi: Reuse asm-generic ucontext.h
xtensa: xtfpga: Use register_sys_off_handler(SYS_OFF_MODE_RESTART)
xtensa: xt2000: Use register_sys_off_handler(SYS_OFF_MODE_RESTART)
xtensa: ISS: Use register_sys_off_handler(SYS_OFF_MODE_RESTART)
====================
Remove a number of ISA and PCMCIA Ethernet drivers
These old drivers have not been much of a Maintenance burden until
recently. Now there are more newbies using AI and fuzzers finding
issues, resulting in more work for Maintainers. Fixing these old
drivers make little sense, if it is not clear they have users.
These mostly ISA and PCMCIA Ethernet devices, mostly from the last
century, a couple from 2001 or 2002. It seems unlikely they are still
used. However, remove them one patch at a time so they can be brought
back if somebody still has the hardware, runs modern kernels and wants
to take up the roll of driver Maintainer.
====================
Merge patch series "eventpoll: fix ep_remove() UAF and follow-up cleanup"
Christian Brauner <brauner@kernel.org> says:
ep_remove() (via __ep_remove_file()) cleared file->f_ep under
file->f_lock but then kept using @file in the same critical section:
is_file_epoll(), hlist_del_rcu() through the head, spin_unlock. A
concurrent __fput() on the watched eventpoll caught the transient
NULL in eventpoll_release()'s lockless fast path, skipped
eventpoll_release_file() entirely, and ran to ep_eventpoll_release()
-> ep_clear_and_put() -> ep_free(). That kfree()s the struct
eventpoll whose embedded ->refs hlist_head is exactly where
epi->fllink.pprev points and the subsequent hlist_del_rcu()'s
"*pprev = next" scribbles into freed kmalloc-192 memory, which is
the slab-use-after-free KASAN caught.
struct file is SLAB_TYPESAFE_BY_RCU on top of that so the same window
also lets the slot recycle while ep_remove() is still nominally
inside file->f_lock. The upshot is an attacker-influencable
kmem_cache_free() against the wrong slab cache.
The comment on eventpoll_release()'s fast path - "False positives
simply cannot happen because the file in on the way to be removed
and nobody ( but eventpoll ) has still a reference to this file" -
was itself the wrong invariant this race exploits.
The fix pins @file via epi_fget() at the top of ep_remove() and
gates the f_ep clear / hlist_del_rcu() on the pin succeeding. With
the pin held __fput() cannot start which transitively keeps the
watched struct eventpoll alive across the critical section and also
prevents the struct file slot from recycling. Both UAFs are closed.
If the pin fails __fput() is already in flight on @file. Because we
bail before clearing f_ep that path takes eventpoll_release()'s slow
path into eventpoll_release_file() which blocks on ep->mtx until
ep_clear_and_put() drops it and then cleans up the orphaned epi. The
bailed epi's share of ep->refcount stays intact so
ep_clear_and_put()'s trailing ep_refcount_dec_and_test() cannot free
the eventpoll out from under eventpoll_release_file().
With epi_fget() now gating every ep_remove() call the epi->dying
flag becomes vestigial. epi->dying == true always coincides with
file_ref_get() == false because __fput() is reachable only once the
refcount hits zero and the refcount is monotone there. The last
patch drops the flag and leaves a single coordination mechanism
instead of two.
* patches from https://patch.msgid.link/20260423-work-epoll-uaf-v1-0-2470f9eec0f5@kernel.org:
eventpoll: drop vestigial epi->dying flag
eventpoll: drop dead bool return from __ep_remove_epi()
eventpoll: refresh eventpoll_release() fast-path comment
eventpoll: move f_lock acquisition into __ep_remove_file()
eventpoll: fix ep_remove struct eventpoll / struct file UAF
eventpoll: move epi_fget() up
eventpoll: rename ep_remove_safe() back to ep_remove()
eventpoll: kill __ep_remove()
eventpoll: split __ep_remove()
eventpoll: use hlist_is_singular_node() in __ep_remove()
With ep_remove() now pinning @file via epi_fget() across the
f_ep clear and hlist_del_rcu(), the dying flag no longer
orchestrates anything: it was set in eventpoll_release_file()
(which only runs from __fput(), i.e. after @file's refcount has
reached zero) and read in __ep_remove() / ep_remove() as a cheap
bail before attempting the same synchronization epi_fget() now
provides unconditionally.
The implication is simple: epi->dying == true always coincides
with file_ref_get(&file->f_ref) == false, because __fput() is
reachable only once the refcount hits zero and the refcount is
monotone in that state. The READ_ONCE(epi->dying) in ep_remove()
therefore selects exactly the same callers that epi_fget() would
reject, just one atomic cheaper. That's not worth a struct
field, a second coordination mechanism, and the comments on
both.
Refresh the eventpoll_release_file() comment to describe what
actually makes the path race-free now (the pin in ep_remove()).
No functional change: the correctness argument is unchanged,
only the mechanism is now a single one instead of two.
eventpoll: drop dead bool return from ep_remove_epi()
ep_remove_epi() always returns true -- the "can be disposed"
answer was meaningful back when the dying-check lived inside the
pre-split __ep_remove(), but after that check moved to ep_remove()
the return value is just noise. Both callers gate on it
unconditionally:
if (ep_remove_epi(ep, epi))
WARN_ON_ONCE(ep_refcount_dec_and_test(ep));
dispose = ep_remove_epi(ep, epi);
...
if (dispose && ep_refcount_dec_and_test(ep))
ep_free(ep);
Make ep_remove_epi() return void, drop the dispose local in
eventpoll_release_file(), and the useless conditionals at both
callers. No functional change.
The old comment justified the lockless READ_ONCE(file->f_ep) check
with "False positives simply cannot happen because the file is on
the way to be removed and nobody ( but eventpoll ) has still a
reference to this file." That reasoning was the root of the UAF
fixed in "eventpoll: fix ep_remove struct eventpoll / struct file
UAF": __ep_remove() could clear f_ep while another close raced
past the fast path and freed the watched eventpoll / recycled the
struct file slot.
With ep_remove() now pinning @file via epi_fget() across the f_ep
clear and hlist_del_rcu(), the invariant is re-established for the
right reason: anyone who might clear f_ep holds @file alive for
the duration, so a NULL observation really does mean no
concurrent eventpoll path has work left on this file. Refresh the
comment accordingly so the next reader doesn't inherit the broken
model.
eventpoll: move f_lock acquisition into ep_remove_file()
Let the helper own its critical section end-to-end: take &file->f_lock
at the top, read file->f_ep inside the lock, release on exit. Callers
(ep_remove() and eventpoll_release_file()) no longer need to wrap the
call, and the function-comment lock-handoff contract is gone.
ep_remove() (via ep_remove_file()) cleared file->f_ep under
file->f_lock but then kept using @file inside the critical section
(is_file_epoll(), hlist_del_rcu() through the head, spin_unlock).
A concurrent __fput() taking the eventpoll_release() fastpath in
that window observed the transient NULL, skipped
eventpoll_release_file() and ran to f_op->release / file_free().
For the epoll-watches-epoll case, f_op->release is
ep_eventpoll_release() -> ep_clear_and_put() -> ep_free(), which
kfree()s the watched struct eventpoll. Its embedded ->refs
hlist_head is exactly where epi->fllink.pprev points, so the
subsequent hlist_del_rcu()'s "*pprev = next" scribbles into freed
kmalloc-192 memory.
In addition, struct file is SLAB_TYPESAFE_BY_RCU, so the slot
backing @file could be recycled by alloc_empty_file() --
reinitializing f_lock and f_ep -- while ep_remove() is still
nominally inside that lock. The upshot is an attacker-controllable
kmem_cache_free() against the wrong slab cache.
Pin @file via epi_fget() at the top of ep_remove() and gate the
critical section on the pin succeeding. With the pin held @file
cannot reach refcount zero, which holds __fput() off and
transitively keeps the watched struct eventpoll alive across the
hlist_del_rcu() and the f_lock use, closing both UAFs.
If the pin fails @file has already reached refcount zero and its
__fput() is in flight. Because we bailed before clearing f_ep,
that path takes the eventpoll_release() slow path into
eventpoll_release_file() and blocks on ep->mtx until the waiter
side's ep_clear_and_put() drops it. The bailed epi's share of
ep->refcount stays intact, so the trailing ep_refcount_dec_and_test()
in ep_clear_and_put() cannot free the eventpoll out from under
eventpoll_release_file(); the orphaned epi is then cleaned up
there.
A successful pin also proves we are not racing
eventpoll_release_file() on this epi, so drop the now-redundant
re-check of epi->dying under f_lock. The cheap lockless
READ_ONCE(epi->dying) fast-path bailout stays.
eventpoll: drop vestigial __ prefix from ep_remove_{file,epi}()
With __ep_remove() gone, the double-underscore on __ep_remove_file()
and __ep_remove_epi() no longer contrasts with a __-less parent and
just reads as noise. Rename both to ep_remove_file() and
ep_remove_epi(). No functional change.
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
Remove the boolean conditional in __ep_remove() and restructure the code
so the check for racing with eventpoll_release_file() are only done in
the ep_remove_safe() path where they belong.
eventpoll: use hlist_is_singular_node() in __ep_remove()
Replace the open-coded "epi is the only entry in file->f_ep" check
with hlist_is_singular_node(). Same semantics, and the helper avoids
the head-cacheline access in the common false case.
Jan Kara [Thu, 23 Apr 2026 09:03:12 +0000 (11:03 +0200)]
fs: Handle multiply claimed blocks more gracefully with mmb
When a metadata block is referenced by multiple inodes and tracked by
metadata bh infrastructure (which is forbidden and generally indicates
filesystem corruption), it can happen that mmb_mark_buffer_dirty() is
called for two different mmb structures in parallel. This can lead to a
corruption of mmb linked list. Handle that situation gracefully (at
least from mmb POV) by serializing on setting bh->b_mmb.
Randy Dunlap [Thu, 16 Apr 2026 21:54:29 +0000 (14:54 -0700)]
nstree: fix func. parameter kernel-doc warnings
Use the correct parameter name ("__ns") for function parameter kernel-doc
to avoid 3 warnings:
Warning: include/linux/nstree.h:68 function parameter '__ns' not described in 'ns_tree_add_raw'
Warning: include/linux/nstree.h:77 function parameter '__ns' not described in 'ns_tree_add'
Warning: include/linux/nstree.h:88 function parameter '__ns' not described in 'ns_tree_remove'
Zizhi Wo [Sat, 18 Apr 2026 06:06:34 +0000 (14:06 +0800)]
fs: aio: reject partial mremap to avoid Null-pointer-dereference error
[BUG]
Recently, our internal syzkaller testing uncovered a null pointer
dereference issue:
BUG: kernel NULL pointer dereference, address: 0000000000000000
...
[ 51.111664] filemap_read_folio+0x25/0xe0
[ 51.112410] filemap_fault+0xad7/0x1250
[ 51.113112] __do_fault+0x4b/0x460
[ 51.113699] do_pte_missing+0x5bc/0x1db0
[ 51.114250] ? __pte_offset_map+0x23/0x170
[ 51.114822] __handle_mm_fault+0x9f8/0x1680
...
Crash analysis showed the file involved was an AIO ring file. The
phenomenon triggered is the same as the issue described in [1].
[CAUSE]
Consider the following scenario: userspace sets up an AIO context via
io_setup(), which creates a VMA covering the entire ring buffer. Then
userspace calls mremap() with the AIO ring address as the source, a smaller
old_len (less than the full ring size), MREMAP_MAYMOVE set, and without
MREMAP_DONTUNMAP. The kernel will relocate the requested portion to a new
destination address.
During this move, __split_vma() splits the original AIO ring VMA. The
requested portion is unmapped from the source and re-established at the
destination, while the remainder stays at the original source address as
an orphan VMA. The aio_ring_mremap() callback fires on the new destination
VMA, updating ctx->mmap_base to the destination address. But the callback
is unaware that only a partial region was moved and that an orphan VMA
still exists at the source:
source(AIO):
+-------------------+---------------------+
| moved to dest | orphan VMA (AIO) |
+-------------------+---------------------+
A A+partial_len A+ctx->mmap_size
dest:
+-------------------+
| moved VMA (AIO) |
+-------------------+
B B+partial_len
Later, io_destroy() calls vm_munmap(ctx->mmap_base, ctx->mmap_size), which
unmaps the destination. This not only fails to unmap the orphan VMA at the
source, but also overshoots the destination VMA and may unmap unrelated
mappings adjacent to it! After put_aio_ring_file() calls truncate_setsize()
to remove all pages from the pagecache, any subsequent access to the orphan
VMA triggers filemap_fault(), which calls a_ops->read_folio(). Since aio
does not implement read_folio, this results in a NULL pointer dereference.
[FIX]
Note that expanding mremap (new_len > old_len) is already rejected because
AIO ring VMAs are created with VM_DONTEXPAND. The only problematic case is
a partial move where "old_len == new_len" but both are smaller than the
full ring size.
Fix this by checking in aio_ring_mremap() that the new VMA covers the
entire ring. This ensures the AIO ring is always moved as a whole,
preventing orphan VMAs and the subsequent crash.
Samuel Page [Mon, 20 Apr 2026 09:01:37 +0000 (11:01 +0200)]
fuse: reject oversized dirents in page cache
fuse_add_dirent_to_cache() computes a serialized dirent size from the
server-controlled namelen field and copies the dirent into a single
page-cache page. The existing logic only checks whether the dirent fits
in the remaining space of the current page and advances to a fresh page
if not. It never checks whether the dirent itself exceeds PAGE_SIZE.
As a result, a malicious FUSE server can return a dirent with
namelen=4095, producing a serialized record size of 4120 bytes. On 4 KiB
page systems this causes memcpy() to overflow the cache page by 24 bytes
into the following kernel page.
Reject dirents that cannot fit in a single page before copying them into
the readdir cache.
Fixes: 69e34551152a ("fuse: allow caching readdir") Cc: stable@vger.kernel.org # v6.16+ Assisted-by: Bynario AI Signed-off-by: Samuel Page <sam@bynar.io> Reported-by: Qi Tang <tpluszz77@gmail.com> Reported-by: Zijun Hu <nightu@northwestern.edu> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com> Link: https://patch.msgid.link/20260420090139.662772-1-mszeredi@redhat.com Signed-off-by: Christian Brauner <brauner@kernel.org>
Jan Kara [Mon, 13 Apr 2026 09:36:19 +0000 (11:36 +0200)]
writeback: Fix use after free in inode_switch_wbs_work_fn()
inode_switch_wbs_work_fn() has a loop like:
wb_get(new_wb);
while (1) {
list = llist_del_all(&new_wb->switch_wbs_ctxs);
/* Nothing to do? */
if (!list)
break;
... process the items ...
}
Now adding of items to the list looks like:
wb_queue_isw()
if (llist_add(&isw->list, &wb->switch_wbs_ctxs))
queue_work(isw_wq, &wb->switch_work);
Because inode_switch_wbs_work_fn() loops when processing isw items, it
can happen that wb->switch_work is pending while wb->switch_wbs_ctxs is
empty. This is a problem because in that case wb can get freed (no isw
items -> no wb reference) while the work is still pending causing
use-after-free issues.
We cannot just fix this by cancelling work when freeing wb because that
could still trigger problematic 0 -> 1 transitions on wb refcount due to
wb_get() in inode_switch_wbs_work_fn(). It could be all handled with
more careful code but that seems unnecessarily complex so let's avoid
that until it is proven that the looping actually brings practical
benefit. Just remove the loop from inode_switch_wbs_work_fn() instead.
That way when wb_queue_isw() queues work, we are guaranteed we have
added the first item to wb->switch_wbs_ctxs and nobody is going to
remove it (and drop the wb reference it holds) until the queued work
runs.
Fixes: e1b849cfa6b6 ("writeback: Avoid contention on wb->list_lock when switching inodes") CC: stable@vger.kernel.org Signed-off-by: Jan Kara <jack@suse.cz> Link: https://patch.msgid.link/20260413093618.17244-2-jack@suse.cz Acked-by: Tejun Heo <tj@kernel.org> Signed-off-by: Christian Brauner <brauner@kernel.org>
At t=0, the parent process calls io_setup and then fork. The child process
gets its own VMA but without any PTEs. The parent then calls io_destroy.
Before i_size is truncated to 0, at t=1 the child process accesses this AIO
ctx address and triggers a pagefault. After the max_idx check passes, at
t=2 the parent calls truncate_setsize and truncate_pagecache. At t=3 the
child fails to obtain the folio, falls into the "page_not_uptodate" path,
and hits this problem because AIO does not implement "read_folio".
[Fix]
Fix this by marking the AIO ring buffer VMA with VM_DONTCOPY so
that fork()'s dup_mmap() skips it entirely. This is the correct
semantic because:
1) The child's ioctx_table is already reset to NULL by mm_init_aio() during
fork(), so the child has no AIO context and no way to perform any AIO
operations on this mapping.
2) The AIO ring VMA is only meaningful in conjunction with its associated
kioctx, which is never inherited across fork(). So child process with no
AIO context has no legitimate reason to access the ring buffer. Delivering
SIGSEGV on such an erroneous access is preferable to a kernel crash.
As discussed with Andrew Lunn, this patch series removes the obsolete
hamachi and yellowfin PCI drivers. Both drivers support hardware that
is over two decades old and no longer in active use.
Removing them eliminates dead code and reduces the overall maintenance
burden on the netdev subsystem.
====================
Mingyu Wang [Wed, 22 Apr 2026 04:48:20 +0000 (12:48 +0800)]
net: packetengines: remove obsolete yellowfin driver and vendor dir
Similar to the hamachi driver, the yellowfin driver supports hardware
that is over two decades old and no longer in active use.
Since yellowfin was the last remaining driver in the packetengines
vendor directory, we can now safely remove the entire directory and
drop its associated references from the parent Kconfig and Makefile.
This eliminates dead code and reduces the overall maintenance burden
on the netdev subsystem.
projects / thirdparty / linux.git / log
