A refcounting fix for bridges, revert a previous framebuffer
use-after-free fix that turned out to be causing more problems, a hang
fix for qaic, an initialization fix for ast, a error handling fix for
sysfb, and a speculation fix for drm_compat_ioctl.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@redhat.com>
Link: https://patch.msgid.link/20260402-vivid-perfect-caiman-ca055e@houat
- Fix for #12045: Huawei Matebook E (DRR-WXX): Persistent Black Screen on Boot with i915 and Gen11: Modesetting and Backlight Control Malfunction
- Fix for #15826: i915: Raptor Lake-P [UHD Graphics] display flicker/corruption on eDP panel
- Use crtc_state->enhanced_framing properly on ivb/hsw CPU eDP
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patch.msgid.link/ac5DM1IpBkuaT58e@jlahtine-mobl
Description:
- Commit b82f075934 ("drm/amd/display: Migrate DIO registers access
from hwseq to dio component") moved DIO_MEM_PWR_CTRL register access
behind the new dio abstraction layer but only created the dio object for
DCN 4.01. On all other generations (DCN 10/20/21/201/30/301/302/303/
31/314/315/316/32/321/35/351/36), the dio pointer is NULL, causing the
register write to be silently skipped.
This results in AFMT HDMI memory not being powered on during init_hw,
which can cause HDMI audio failures and display issues on affected
hardware including Renoir/Cezanne (DCN 2.1) APUs that use dcn10_init_hw.
Call dcn10_dio_construct() in each older DCN generation's resource.c
to create the dio object, following the same pattern as DCN 4.01. This
ensures the dio pointer is non-NULL and the mem_pwr_ctrl callback works
through the dio abstraction for all DCN generations.
Fixes: b82f075934 ("drm/amd/display: Migrate DIO registers access from hwseq to dio component.")
Reviewed-by: Ivan Lipski <ivan.lipski@amd.com>
Signed-off-by: Ionut Nechita <ionut_n2001@yahoo.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fix incorrect error checking and memory type confusion in
efidrm_device_create(). devm_memremap() returns error pointers, not
NULL, and returns system memory while devm_ioremap() returns I/O memory.
The code incorrectly passes system memory to iosys_map_set_vaddr_iomem().
Restructure to handle each memory type separately. Use devm_ioremap*()
with ERR_PTR(-ENXIO) for WC/UC, and devm_memremap() with ERR_CAST() for
WT/WB.
Fixes: 32ae90c66f ("drm/sysfb: Add efidrm for EFI displays")
Signed-off-by: Chen Ni <nichen@iscas.ac.cn>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patch.msgid.link/20260311064652.2903449-1-nichen@iscas.ac.cn
Apparently I forgot about the pipe min_voltage_level when I
decoupled the CDCLK calculations from modesets. Even if the
CDCLK frequency doesn't need changing we may still need to
bump the voltage level to accommodate an increase in the
port clock frequency.
Currently, even if there is a full modeset, we won't notice the
need to go through the full CDCLK calculations/programming,
unless the set of enabled/active pipes changes, or the
pipe/dbuf min CDCLK changes.
Duplicate the same logic we use the pipe's min CDCLK frequency
to also deal with its min voltage level.
Note that the 'allow_voltage_level_decrease' stuff isn't
really useful here since the min voltage level can only
change during a full modeset. But I think sticking to the
same approach in the three similar parts (pipe min cdclk,
pipe min voltage level, dbuf min cdclk) is a good idea.
Cc: stable@vger.kernel.org
Tested-by: Mikhail Rudenko <mike.rudenko@gmail.com>
Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/15826
Fixes: ba91b9eecb ("drm/i915/cdclk: Decouple cdclk from state->modeset")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patch.msgid.link/20260325135849.12603-2-ville.syrjala@linux.intel.com
Reviewed-by: Michał Grzelak <michal.grzelak@intel.com>
(cherry picked from commit 0f21a14987ebae3c05ad1184ea872e7b7a7b8695)
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The control stack size is calculated based on the number of CUs and
waves, and is then aligned to PAGE_SIZE. When the resulting control
stack size is aligned to 64 KB, GPU hangs and queue preemption
failures are observed while running RCCL unit tests on systems with
more than two GPUs.
amdgpu 0048:0f:00.0: amdgpu: Queue preemption failed for queue with
doorbell_id: 80030008
amdgpu 0048:0f:00.0: amdgpu: Failed to evict process queues
amdgpu 0048:0f:00.0: amdgpu: GPU reset begin!. Source: 4
amdgpu 0048:0f:00.0: amdgpu: Queue preemption failed for queue with
doorbell_id: 80030008
amdgpu 0048:0f:00.0: amdgpu: Failed to evict process queues
amdgpu 0048:0f:00.0: amdgpu: Failed to restore process queues
This issue is observed on both 4 KB and 64 KB system page-size
configurations.
This patch fixes the issue by aligning the control stack size to
AMDGPU_GPU_PAGE_SIZE instead of PAGE_SIZE, so the control stack size
will not be 64 KB on systems with a 64 KB page size and queue
preemption works correctly.
Additionally, In the current code, wg_data_size is aligned to PAGE_SIZE,
which can waste memory if the system page size is large. In this patch,
wg_data_size is aligned to AMDGPU_GPU_PAGE_SIZE. The cwsr_size, calculated
from wg_data_size and the control stack size, is aligned to PAGE_SIZE.
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Donet Tom <donettom@linux.ibm.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit a3e1443630)
dcn401_init_hw() assumes that update_bw_bounding_box() is valid when
entering the update path. However, the existing condition:
((!fams2_enable && update_bw_bounding_box) || freq_changed)
does not guarantee this, as the freq_changed branch can evaluate to true
independently of the callback pointer.
This can result in calling update_bw_bounding_box() when it is NULL.
Fix this by separating the update condition from the pointer checks and
ensuring the callback, dc->clk_mgr, and bw_params are validated before
use.
Fixes the below:
../dc/hwss/dcn401/dcn401_hwseq.c:367 dcn401_init_hw() error: we previously assumed 'dc->res_pool->funcs->update_bw_bounding_box' could be null (see line 362)
Fixes: ca0fb243c3 ("drm/amd/display: Underflow Seen on DCN401 eGPU")
Cc: Daniel Sa <Daniel.Sa@amd.com>
Cc: Alvin Lee <alvin.lee2@amd.com>
Cc: Roman Li <roman.li@amd.com>
Cc: Alex Hung <alex.hung@amd.com>
Cc: Tom Chung <chiahsuan.chung@amd.com>
Cc: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 86117c5ab4)
Cc: stable@vger.kernel.org
Currently, AMDGPU_VA_RESERVED_TRAP_SIZE is hardcoded to 8KB, while
KFD_CWSR_TBA_TMA_SIZE is defined as 2 * PAGE_SIZE. On systems with
4K pages, both values match (8KB), so allocation and reserved space
are consistent.
However, on 64K page-size systems, KFD_CWSR_TBA_TMA_SIZE becomes 128KB,
while the reserved trap area remains 8KB. This mismatch causes the
kernel to crash when running rocminfo or rccl unit tests.
Kernel attempted to read user page (2) - exploit attempt? (uid: 1001)
BUG: Kernel NULL pointer dereference on read at 0x00000002
Faulting instruction address: 0xc0000000002c8a64
Oops: Kernel access of bad area, sig: 11 [#1]
LE PAGE_SIZE=64K MMU=Radix SMP NR_CPUS=2048 NUMA pSeries
CPU: 34 UID: 1001 PID: 9379 Comm: rocminfo Tainted: G E
6.19.0-rc4-amdgpu-00320-gf23176405700 #56 VOLUNTARY
Tainted: [E]=UNSIGNED_MODULE
Hardware name: IBM,9105-42A POWER10 (architected) 0x800200 0xf000006
of:IBM,FW1060.30 (ML1060_896) hv:phyp pSeries
NIP: c0000000002c8a64 LR: c00000000125dbc8 CTR: c00000000125e730
REGS: c0000001e0957580 TRAP: 0300 Tainted: G E
MSR: 8000000000009033 <SF,EE,ME,IR,DR,RI,LE> CR: 24008268
XER: 00000036
CFAR: c00000000125dbc4 DAR: 0000000000000002 DSISR: 40000000
IRQMASK: 1
GPR00: c00000000125d908 c0000001e0957820 c0000000016e8100
c00000013d814540
GPR04: 0000000000000002 c00000013d814550 0000000000000045
0000000000000000
GPR08: c00000013444d000 c00000013d814538 c00000013d814538
0000000084002268
GPR12: c00000000125e730 c000007e2ffd5f00 ffffffffffffffff
0000000000020000
GPR16: 0000000000000000 0000000000000002 c00000015f653000
0000000000000000
GPR20: c000000138662400 c00000013d814540 0000000000000000
c00000013d814500
GPR24: 0000000000000000 0000000000000002 c0000001e0957888
c0000001e0957878
GPR28: c00000013d814548 0000000000000000 c00000013d814540
c0000001e0957888
NIP [c0000000002c8a64] __mutex_add_waiter+0x24/0xc0
LR [c00000000125dbc8] __mutex_lock.constprop.0+0x318/0xd00
Call Trace:
0xc0000001e0957890 (unreliable)
__mutex_lock.constprop.0+0x58/0xd00
amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu+0x6fc/0xb60 [amdgpu]
kfd_process_alloc_gpuvm+0x54/0x1f0 [amdgpu]
kfd_process_device_init_cwsr_dgpu+0xa4/0x1a0 [amdgpu]
kfd_process_device_init_vm+0xd8/0x2e0 [amdgpu]
kfd_ioctl_acquire_vm+0xd0/0x130 [amdgpu]
kfd_ioctl+0x514/0x670 [amdgpu]
sys_ioctl+0x134/0x180
system_call_exception+0x114/0x300
system_call_vectored_common+0x15c/0x2ec
This patch changes AMDGPU_VA_RESERVED_TRAP_SIZE to 64 KB and
KFD_CWSR_TBA_TMA_SIZE to the AMD GPU page size. This means we reserve
64 KB for the trap in the address space, but only allocate 8 KB within
it. With this approach, the allocation size never exceeds the reserved
area.
Fixes: 34a1de0f79 ("drm/amdkfd: Relocate TBA/TMA to opposite side of VM hole")
Reviewed-by: Christian König <christian.koenig@amd.com>
Suggested-by: Felix Kuehling <felix.kuehling@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Donet Tom <donettom@linux.ibm.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 31b8de5e55)
Cc: stable@vger.kernel.org
In mes_userq_mqd_create(), the memdup_user() allocations for
IP-specific MQD structs are not freed when subsequent VA validation
fails. The goto free_mqd label only cleans up the MQD BO object and
userq_props.
Fix by adding kfree() before each goto free_mqd on VA validation
failure in the COMPUTE, GFX, and SDMA branches.
Fixes: 9e46b8bb05 ("drm/amdgpu: validate userq buffer virtual address and size")
Reported-by: Yuhao Jiang <danisjiang@gmail.com>
Signed-off-by: Junrui Luo <moonafterrain@outlook.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 27f5ff9e4a)
Cc: stable@vger.kernel.org
For gfxV9, due to a hardware bug ("based on the comments in the code
here [1]"), the control stack of a user-mode compute queue must be
allocated immediately after the page boundary of its regular MQD buffer.
To handle this, we allocate an enlarged MQD buffer where the first page
is used as the MQD and the remaining pages store the control stack.
Although these regions share the same BO, they require different memory
types: the MQD must be UC (uncached), while the control stack must be
NC (non-coherent), matching the behavior when the control stack is
allocated in user space.
This logic works correctly on systems where the CPU page size matches
the GPU page size (4K). However, the current implementation aligns both
the MQD and the control stack to the CPU PAGE_SIZE. On systems with a
larger CPU page size, the entire first CPU page is marked UC—even though
that page may contain multiple GPU pages. The GPU treats the second 4K
GPU page inside that CPU page as part of the control stack, but it is
incorrectly mapped as UC.
This patch fixes the issue by aligning both the MQD and control stack
sizes to the GPU page size (4K). The first 4K page is correctly marked
as UC for the MQD, and the remaining GPU pages are marked NC for the
control stack. This ensures proper memory type assignment on systems
with larger CPU page sizes.
[1]: https://elixir.bootlin.com/linux/v6.18/source/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c#L118
Acked-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Donet Tom <donettom@linux.ibm.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 998d678141)
The AQL queue size can be 4K, but the minimum buffer object (BO)
allocation size is PAGE_SIZE. On systems with a page size larger
than 4K, the expected queue size does not match the allocated BO
size, causing queue creation to fail.
Align the expected queue size to PAGE_SIZE so that it matches the
allocated BO size and allows queue creation to succeed.
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Donet Tom <donettom@linux.ibm.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit b01cd158a2)
amdgpu_userq_get_doorbell_index() passes the user-provided
doorbell_offset to amdgpu_doorbell_index_on_bar() without bounds
checking. An arbitrarily large doorbell_offset can cause the
calculated doorbell index to fall outside the allocated doorbell BO,
potentially corrupting kernel doorbell space.
Validate that doorbell_offset falls within the doorbell BO before
computing the BAR index, using u64 arithmetic to prevent overflow.
Fixes: f09c1e6077 ("drm/amdgpu: generate doorbell index for userqueue")
Reported-by: Yuhao Jiang <danisjiang@gmail.com>
Signed-off-by: Junrui Luo <moonafterrain@outlook.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit de1ef4ffd7)
Cc: stable@vger.kernel.org
xe_device_declare_wedged() runs in the DMA-fence signaling path, where
GFP_KERNEL memory allocations are not allowed. However, registering
xe_device_wedged_fini via drmm_add_action_or_reset() triggers a
GFP_KERNEL allocation.
Fix this by deferring the registration of xe_device_wedged_fini until
late in the driver load sequence. Additionally, drop the wedged PM
reference only if the device is actually wedged in
xe_device_wedged_fini.
Fixes: 452bca0edb ("drm/xe: Don't suspend device upon wedge")
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patch.msgid.link/20260326210116.202585-2-matthew.brost@intel.com
(cherry picked from commit b08ceb443866808b881b12d4183008d214d816c1)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
This reverts commit 6bee098b91.
Den 2026-03-25 kl. 22:11, skrev Simona Vetter:
> On Wed, Mar 25, 2026 at 10:26:40AM -0700, Guenter Roeck wrote:
>> Hi,
>>
>> On Fri, Mar 13, 2026 at 04:17:27PM +0100, Maarten Lankhorst wrote:
>>> When trying to do a rather aggressive test of igt's "xe_module_load
>>> --r reload" with a full desktop environment and game running I noticed
>>> a few OOPSes when dereferencing freed pointers, related to
>>> framebuffers and property blobs after the compositor exits.
>>>
>>> Solve this by guarding the freeing in drm_file with drm_dev_enter/exit,
>>> and immediately put the references from struct drm_file objects during
>>> drm_dev_unplug().
>>>
>>
>> With this patch in v6.18.20, I get the warning backtraces below.
>> The backtraces are gone with the patch reverted.
>
> Yeah, this needs to be reverted, reasoning below. Maarten, can you please
> take care of that and feed the revert through the usual channels? I don't
> think it's critical enough that we need to fast-track this into drm.git
> directly.
>
> Quoting the patch here again:
>
>> drivers/gpu/drm/drm_file.c | 5 ++++-
>> drivers/gpu/drm/drm_mode_config.c | 9 ++++++---
>> 2 files changed, 10 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/drm_file.c b/drivers/gpu/drm/drm_file.c
>> index ec820686b3021..f52141f842a1f 100644
>> --- a/drivers/gpu/drm/drm_file.c
>> +++ b/drivers/gpu/drm/drm_file.c
>> @@ -233,6 +233,7 @@ static void drm_events_release(struct drm_file *file_priv)
>> void drm_file_free(struct drm_file *file)
>> {
>> struct drm_device *dev;
>> + int idx;
>>
>> if (!file)
>> return;
>> @@ -249,9 +250,11 @@ void drm_file_free(struct drm_file *file)
>>
>> drm_events_release(file);
>>
>> - if (drm_core_check_feature(dev, DRIVER_MODESET)) {
>> + if (drm_core_check_feature(dev, DRIVER_MODESET) &&
>> + drm_dev_enter(dev, &idx)) {
>
> This is misplaced for two reasons:
>
> - Even if we'd want to guarantee that we hold a drm_dev_enter/exit
> reference during framebuffer teardown, we'd need to do this
> _consistently over all callsites. Not ad-hoc in just one place that a
> testcase hits. This also means kerneldoc updates of the relevant hooks
> and at least a bunch of acks from other driver people to document the
> consensus.
>
> - More importantly, this is driver responsibilities in general unless we
> have extremely good reasons to the contrary. Which means this must be
> placed in xe.
>
>> drm_fb_release(file);
>> drm_property_destroy_user_blobs(dev, file);
>> + drm_dev_exit(idx);
>> }
>>
>> if (drm_core_check_feature(dev, DRIVER_SYNCOBJ))
>> diff --git a/drivers/gpu/drm/drm_mode_config.c b/drivers/gpu/drm/drm_mode_config.c
>> index 84ae8a23a3678..e349418978f79 100644
>> --- a/drivers/gpu/drm/drm_mode_config.c
>> +++ b/drivers/gpu/drm/drm_mode_config.c
>> @@ -583,10 +583,13 @@ void drm_mode_config_cleanup(struct drm_device *dev)
>> */
>> WARN_ON(!list_empty(&dev->mode_config.fb_list));
>> list_for_each_entry_safe(fb, fbt, &dev->mode_config.fb_list, head) {
>> - struct drm_printer p = drm_dbg_printer(dev, DRM_UT_KMS, "[leaked fb]");
>> + if (list_empty(&fb->filp_head) || drm_framebuffer_read_refcount(fb) > 1) {
>> + struct drm_printer p = drm_dbg_printer(dev, DRM_UT_KMS, "[leaked fb]");
>
> This is also wrong:
>
> - Firstly, it's a completely independent bug, we do not smash two bugfixes
> into one patch.
>
> - Secondly, it's again a driver bug: drm_mode_cleanup must be called when
> the last drm_device reference disappears (hence the existence of
> drmm_mode_config_init), not when the driver gets unbound. The fact that
> this shows up in a callchain from a devres cleanup means the intel
> driver gets this wrong (like almost everyone else because historically
> we didn't know better).
>
> If we don't follow this rule, then we get races with this code here
> running concurrently with drm_file fb cleanups, which just does not
> work. Review pointed that out, but then shrugged it off with a confused
> explanation:
>
> https://lore.kernel.org/all/e61e64c796ccfb17ae673331a3df4b877bf42d82.camel@linux.intel.com/
>
> Yes this also means a lot of the other drm_device teardown that drivers
> do happens way too early. There is a massive can of worms here of a
> magnitude that most likely is much, much bigger than what you can
> backport to stable kernels. Hotunplug is _hard_.
Back to the drawing board, and fixing it in the intel display driver
instead.
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Fixes: 6bee098b91 ("drm: Fix use-after-free on framebuffers and property blobs when calling drm_dev_unplug")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Simona Vetter <simona.vetter@ffwll.ch>
Signed-off-by: Maarten Lankhorst <dev@lankhorst.se>
Link: https://patch.msgid.link/20260326082217.39941-2-dev@lankhorst.se
A typical bridge refcount value is 3 after a bridge chain is formed:
- devm_drm_bridge_alloc() initializes the refcount value to be 1.
- drm_bridge_add() gets an additional reference hence 2.
- drm_bridge_attach() gets the third reference hence 3.
This typical refcount value aligns with allbridges_show()'s behaviour.
However, since encoder_bridges_show() uses
drm_for_each_bridge_in_chain_scoped() to automatically get/put the
bridge reference while iterating, a bogus reference is accidentally
got when showing the wrong typical refcount value as 4 to users via
debugfs. Fix this by caching the refcount value returned from
kref_read() while iterating and explicitly decreasing the cached
refcount value by 1 before showing it to users.
Fixes: bd57048e45 ("drm/bridge: use drm_for_each_bridge_in_chain_scoped()")
Signed-off-by: Liu Ying <victor.liu@nxp.com>
Reviewed-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Tested-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Link: https://patch.msgid.link/20260318-drm-misc-next-2026-03-05-fix-encoder-bridges-refcount-v3-1-147fea581279@nxp.com
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
During 3D workload, user is reporting hitting:
[ 413.361679] WARNING: drivers/gpu/drm/xe/xe_vm.c:1217 at vm_bind_ioctl_ops_unwind+0x1e2/0x2e0 [xe], CPU#7: vkd3d_queue/9925
[ 413.361944] CPU: 7 UID: 1000 PID: 9925 Comm: vkd3d_queue Kdump: loaded Not tainted 7.0.0-070000rc3-generic #202603090038 PREEMPT(lazy)
[ 413.361949] RIP: 0010:vm_bind_ioctl_ops_unwind+0x1e2/0x2e0 [xe]
[ 413.362074] RSP: 0018:ffffd4c25c3df930 EFLAGS: 00010282
[ 413.362077] RAX: 0000000000000000 RBX: ffff8f3ee817ed10 RCX: 0000000000000000
[ 413.362078] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
[ 413.362079] RBP: ffffd4c25c3df980 R08: 0000000000000000 R09: 0000000000000000
[ 413.362081] R10: 0000000000000000 R11: 0000000000000000 R12: ffff8f41fbf99380
[ 413.362082] R13: ffff8f3ee817e968 R14: 00000000ffffffef R15: ffff8f43d00bd380
[ 413.362083] FS: 00000001040ff6c0(0000) GS:ffff8f4696d89000(0000) knlGS:00000000330b0000
[ 413.362085] CS: 0010 DS: 002b ES: 002b CR0: 0000000080050033
[ 413.362086] CR2: 00007ddfc4747000 CR3: 00000002e6262005 CR4: 0000000000f72ef0
[ 413.362088] PKRU: 55555554
[ 413.362089] Call Trace:
[ 413.362092] <TASK>
[ 413.362096] xe_vm_bind_ioctl+0xa9a/0xc60 [xe]
Which seems to hint that the vma we are re-inserting for the ops unwind
is either invalid or overlapping with something already inserted in the
vm. It shouldn't be invalid since this is a re-insertion, so must have
worked before. Leaving the likely culprit as something already placed
where we want to insert the vma.
Following from that, for the case where we do something like a rebind in
the middle of a vma, and one or both mapped ends are already compatible,
we skip doing the rebind of those vma and set next/prev to NULL. As well
as then adjust the original unmap va range, to avoid unmapping the ends.
However, if we trigger the unwind path, we end up with three va, with
the two ends never being removed and the original va range in the middle
still being the shrunken size.
If this occurs, one failure mode is when another unwind op needs to
interact with that range, which can happen with a vector of binds. For
example, if we need to re-insert something in place of the original va.
In this case the va is still the shrunken version, so when removing it
and then doing a re-insert it can overlap with the ends, which were
never removed, triggering a warning like above, plus leaving the vm in a
bad state.
With that, we need two things here:
1) Stop nuking the prev/next tracking for the skip cases. Instead
relying on checking for skip prev/next, where needed. That way on the
unwind path, we now correctly remove both ends.
2) Undo the unmap va shrinkage, on the unwind path. With the two ends
now removed the unmap va should expand back to the original size again,
before re-insertion.
v2:
- Update the explanation in the commit message, based on an actual IGT of
triggering this issue, rather than conjecture.
- Also undo the unmap shrinkage, for the skip case. With the two ends
now removed, the original unmap va range should expand back to the
original range.
v3:
- Track the old start/range separately. vma_size/start() uses the va
info directly.
Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/7602
Fixes: 8f33b4f054 ("drm/xe: Avoid doing rebinds")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <stable@vger.kernel.org> # v6.8+
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patch.msgid.link/20260318100208.78097-2-matthew.auld@intel.com
(cherry picked from commit aec6969f75)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
During a GPU page fault, the driver restores the SVM range and then maps it
into the GPU page tables. The current implementation passes a GPU-page-size
(4K-based) PFN to svm_range_restore_pages() to restore the range.
SVM ranges are tracked using system-page-size PFNs. On systems where the
system page size is larger than 4K, using GPU-page-size PFNs to restore the
range causes two problems:
Range lookup fails:
Because the restore function receives PFNs in GPU (4K) units, the SVM
range lookup does not find the existing range. This will result in a
duplicate SVM range being created.
VMA lookup failure:
The restore function also tries to locate the VMA for the faulting address.
It converts the GPU-page-size PFN into an address using the system page
size, which results in an incorrect address on non-4K page-size systems.
As a result, the VMA lookup fails with the message: "address 0xxxx VMA is
removed".
This patch passes the system-page-size PFN to svm_range_restore_pages() so
that the SVM range is restored correctly on non-4K page systems.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Donet Tom <donettom@linux.ibm.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 074fe395fb)
Forcibly disable the OD_FAN_CURVE feature when temperature or PWM range is invalid,
otherwise PMFW will reject this configuration on smu v14.0.2/14.0.3.
example:
$ sudo cat /sys/bus/pci/devices/<BDF>/gpu_od/fan_ctrl/fan_curve
OD_FAN_CURVE:
0: 0C 0%
1: 0C 0%
2: 0C 0%
3: 0C 0%
4: 0C 0%
OD_RANGE:
FAN_CURVE(hotspot temp): 0C 0C
FAN_CURVE(fan speed): 0% 0%
$ echo "0 50 40" | sudo tee fan_curve
kernel log:
[ 969.761627] amdgpu 0000:03:00.0: amdgpu: Fan curve temp setting(50) must be within [0, 0]!
[ 1010.897800] amdgpu 0000:03:00.0: amdgpu: Fan curve temp setting(50) must be within [0, 0]!
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>
(cherry picked from commit ab4905d466)
Cc: stable@vger.kernel.org
In kfd_ioctl_create_process(), the pointer 'p' is used before checking
if it is NULL.
The code accesses p->context_id before validating 'p'. This can lead
to a possible NULL pointer dereference.
Move the NULL check before using 'p' so that the pointer is validated
before access.
Fixes the below:
drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_chardev.c:3177 kfd_ioctl_create_process() warn: variable dereferenced before check 'p' (see line 3174)
Fixes: cc6b66d661 ("amdkfd: introduce new ioctl AMDKFD_IOC_CREATE_PROCESS")
Cc: Zhu Lingshan <lingshan.zhu@amd.com>
Cc: Felix Kuehling <felix.kuehling@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Dan Carpenter <dan.carpenter@linaro.org>
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>
(cherry picked from commit 19d4149b22)
amdgpu_amdkfd_submit_ib() submits a GPU job and gets a fence
from amdgpu_ib_schedule(). This fence is used to wait for job
completion.
Currently, the code drops the fence reference using dma_fence_put()
before calling dma_fence_wait().
If dma_fence_put() releases the last reference, the fence may be
freed before dma_fence_wait() is called. This can lead to a
use-after-free.
Fix this by waiting on the fence first and releasing the reference
only after dma_fence_wait() completes.
Fixes the below:
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c:697 amdgpu_amdkfd_submit_ib() warn: passing freed memory 'f' (line 696)
Fixes: 9ae55f030d ("drm/amdgpu: Follow up change to previous drm scheduler change.")
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Dan Carpenter <dan.carpenter@linaro.org>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@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>
(cherry picked from commit 8b9e5259ad)
Forcibly disable the OD_FAN_CURVE feature when temperature or PWM range is invalid,
otherwise PMFW will reject this configuration on smu v13.0.x
example:
$ sudo cat /sys/bus/pci/devices/<BDF>/gpu_od/fan_ctrl/fan_curve
OD_FAN_CURVE:
0: 0C 0%
1: 0C 0%
2: 0C 0%
3: 0C 0%
4: 0C 0%
OD_RANGE:
FAN_CURVE(hotspot temp): 0C 0C
FAN_CURVE(fan speed): 0% 0%
$ echo "0 50 40" | sudo tee fan_curve
kernel log:
[ 756.442527] amdgpu 0000:03:00.0: amdgpu: Fan curve temp setting(50) must be within [0, 0]!
[ 777.345800] amdgpu 0000:03:00.0: amdgpu: Fan curve temp setting(50) must be within [0, 0]!
Closes: https://github.com/ROCm/amdgpu/issues/208
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>
(cherry picked from commit 470891606c)
Cc: stable@vger.kernel.org
When SET_UCLK_MAX capability is absent, return -EOPNOTSUPP from
smu_v13_0_6_emit_clk_levels() for OD_MCLK instead of 0. This makes
unsupported OD_MCLK reporting consistent with other clock types
and allows callers to skip the entry cleanly.
Signed-off-by: Asad Kamal <asad.kamal@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit d82e0a72d9)
Cc: stable@vger.kernel.org
Only reapply UCLK soft limits during PP_OD_RESTORE_DEFAULT when the
current max differs from the DPM table max. This avoids redundant
SMC updates and prevents -EINVAL on restore when no change is needed.
Fixes: b7a9003445 ("drm/amd/pm: Allow setting max UCLK on SMU v13.0.6")
Signed-off-by: Asad Kamal <asad.kamal@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 17f11bbbc7)
[WHAT]
When a sink is connected, aconnector->drm_edid was overwritten without
freeing the previous allocation, causing a memory leak on resume.
[HOW]
Free the previous drm_edid before updating it.
Reviewed-by: Roman Li <roman.li@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Chuanyu Tseng <chuanyu.tseng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 52024a94e7)
Cc: stable@vger.kernel.org
PASID resue could cause interrupt issue when process
immediately runs into hw state left by previous
process exited with the same PASID, it's possible that
page faults are still pending in the IH ring buffer when
the process exits and frees up its PASID. To prevent the
case, it uses idr cyclic allocator same as kernel pid's.
Signed-off-by: Eric Huang <jinhuieric.huang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 8f1de51f49)
Cc: stable@vger.kernel.org
