| Age | Commit message (Collapse) | Author |
|---|
|
# Conflicts:
# drivers/firmware/qcom/Kconfig
# drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
# drivers/gpu/drm/vc4/vc4_v3d.c
# drivers/soc/qcom/qcom-geni-se.c
# rust/kernel/mem.rs
|
|
The amdxdna has a soft dependency on the amd_pmf module. It will use an
exported function if it is available and fall back to another function
if it is not enabled.
amdxdna and amd_pmf are tridstate modules supporting both built [y] and
modularized [m] in the .config file. Since the amdxdna decides to use
the amd_pmf if it is either a module or built in (see
drivers/accel/amdxdna/aie2_pci.h), then there is an issue if they are
not compiled either both as modules or both as built-in.
Adding the optional dependency to Kconfig does not allow the
configuration from being selected.
Fixes: f66d6cc6891e4 ("accel/amdxdna: Support sensors for column utilization")
Signed-off-by: Liam R. Howlett <liam@infradead.org>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20261009011010.3521631-1-liam@infradead.org
|
|
aie2_get_dev_revision() fails to recognize AIE2 revision 9 hardware.
Add AIE2_DEV_REVISION_GRGH to the list of supported AIE2 revisions.
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260929181459.928468-1-lizhi.hou@amd.com
|
|
alloc_ordered_workqueue
create_singlethread_workqueue() is deprecated. Replace it with
alloc_ordered_workqueue() to use the shared system workqueue
infrastructure more efficiently.
WQ_MEM_RECLAIM is added to ensure the mailbox can process
interrupts under memory pressure.
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: George Karagounis <mail@taterr.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260925124804.58808-2-mail@taterr.org
|
|
The hardware context scheduler fence is exposed through
drm_syncobj_add_point(). After a hardware context is destroyed,
userspace can still call drm_sched_fence_get_timeline_name() on the
fence through SYNC_IOC_FILE_INFO. This results in a use-after-free
when accessing fence->sched->name.
Preallocate hardware context schedulers and free them during device
teardown instead. A destroyed hardware context no longer frees its
scheduler, ensuring that the scheduler remains valid while its fences
are still accessible.
Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Reported-by: Jonghyuk Kim(MalHyuk) <malhyuk97@gmail.com>
Link: https://lore.kernel.org/dri-devel/20260902012712.880520-1-malhyuk97@gmail.com/
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260923200501.281208-1-lizhi.hou@amd.com
|
|
amdxdna_drm_sync_bo_ioctl() calls amdxdna_hwctx_sync_debug_bo() for every
FROM_DEVICE sync, which answers -EINVAL when the BO's assigned_hwctx names
no context. Only a BO attached with ATTACH_DEBUG_BO is ever given one, so
an ordinary read-back sync reports failure after its flush has already run.
Ask for the debug sync only when the BO has a context. An unattached BO
carries AMDXDNA_INVALID_CTX_HANDLE and hwctx ids are allocated above it, so
the test is exact, -EINVAL keeps meaning that the named context is gone,
and the handle is not resolved twice. The field is written under dev_lock
and read here without it; the context is still resolved under that lock, so
a racing attach only decides whether this sync sees the buffer.
Fixes: 7ea046838021 ("accel/amdxdna: Support firmware debug buffer")
Suggested-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Taimuraz Kaitmazov <taimuraz@kaitmazov.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260819224458.257346-6-taimuraz@kaitmazov.com
|
|
amdxdna_drm_sync_bo_ioctl() answers a failed amdxdna_flush_bo() with
drm_WARN(). Both of that function's error returns are decided by the
ioctl's arguments, so SYNC_BO with an offset past the end of the BO
splats and taints the kernel from an unprivileged caller.
Log it at debug level, since the same caller can repeat it.
Fixes: dbc8fd7a03cb ("accel/amdxdna: Add expandable device heap support")
Signed-off-by: Taimuraz Kaitmazov <taimuraz@kaitmazov.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260819224458.257346-4-taimuraz@kaitmazov.com
|
|
aie2_create_context() allocates a mailbox channel, starts it, and on a
failed start frees it again:
hwctx->priv->mbox_chann = xdna_mailbox_alloc_channel(ndev->mbox);
...
ret = xdna_mailbox_start_channel(hwctx->priv->mbox_chann, ...);
if (ret)
goto free_channel;
...
free_channel:
xdna_mailbox_free_channel(hwctx->priv->mbox_chann);
hwctx->priv->mbox_chann keeps pointing at the freed channel. That
pointer is the driver's own test for whether a hardware context has a
usable channel: aie2_destroy_context() returns early on NULL,
aie2_sched_job_run() refuses to run a job on NULL, and the message
helpers return -ENODEV on NULL. A stale pointer passes all of them.
It matters on the restart path. aie2_hwctx_restart() calls
aie2_create_context() again on a hardware context that is already live,
after aie2_hwctx_stop() has torn its channel down, and the context
survives a failure there: aie2_sched_job_timedout() discards the return
value entirely, and aie2_hwctx_resume() only propagates it. What is
left is a live hardware context holding a freed channel, and every
later user of it takes the non-NULL branch:
- the next job submitted reaches xdna_mailbox_send_msg() on the freed
channel, via the !mbox_chann guard in aie2_sched_job_run(),
- a second command timeout reaches aie2_hwctx_stop() ->
aie2_destroy_context(),
- closing the device reaches aie2_hwctx_fini() ->
aie2_release_resource() -> xrs_release_resource() ->
aie2_xrs_unload() -> aie2_destroy_context(),
and aie2_destroy_context() then calls xdna_mailbox_stop_channel()
followed by xdna_mailbox_free_channel() on memory that was already
freed: a use-after-free, and a second free of the same channel.
The AIE4 management channel already does this correctly -
aie4_mailbox_start() clears ndev->aie.mgmt_chann right after freeing
it. Do the same here.
The create path itself is not affected: when aie2_xrs_load() fails,
xrs_allocate_resource() removes the solver node without calling
->unload, and aie2_hwctx_init() frees hwctx->priv, so the stale pointer
never outlives the structure holding it.
Starting a channel can fail today without this patch: the ring buffer
sizes firmware reports are rejected unless both are powers of two, and
request_irq() can fail.
Fixes: d5b8b0347fa8 ("accel/amdxdna: Split mailbox channel create function")
Cc: stable@vger.kernel.org
Signed-off-by: Eva Crystal <0xiviel@gmail.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/2fcf0eca215090dbbd95cb86e1bf5076056d3818.1789334558.git.0xiviel@gmail.com
|
|
When aie2_populate_range() faults in pages, it only sets
HMM_PFN_REQ_FAULT. Without HMM_PFN_REQ_WRITE, handle_mm_fault() does not
request write permissions and may populate the CPU page tables with
read-only PTEs.
Check the VMA flags and set HMM_PFN_REQ_WRITE when the VMA has VM_WRITE.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260917171952.1883974-1-lizhi.hou@amd.com
|
|
A userspace process can concurrently trigger an unmap, which queues
hmm_unreg_work, and free BO. If amdxdna_hmm_unreg_work() executes
list_del(&mapp->node) and is preempted before calling amdxdna_umap_put(),
amdxdna_hmm_unregister_all() can fail to find the mapping in the list
and return without calling cancel_work_sync(). This allows object
destruction to proceed while the worker still holds a pointer to the BO
in the umap structure.
Moving the work item from struct amdxdna_umap to struct amdxdna_gem_obj.
The worker no longer holds a pointer to an individual mapping, so there
is no longer a window where the umap has been unlinked from umap_list but
work is still pending against it. The work item now lives in the BO
itself, so use a single cancel_work_sync() in the free path to drain the
work.
Fixes: 445d20910429 ("accel/amdxdna: Fix unexpected wait when flushing notifier_wq")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260917154824.1872220-1-lizhi.hou@amd.com
|
|
The map structure is currently only partially initialized when the
notifier is registered. Although holding mmap_write_lock() prevents
the notifier callback from running before the remaining initialization
is complete, the ordering can be confusing during code review.
Fully initialize the map structure before registering the notifier so
that the structure is completely constructed before it is published.
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260911163907.1193852-1-lizhi.hou@amd.com
|
|
handle_mm_fault() must not be called from the mmap callback because
the VMA has not yet been linked. The handle_mm_fault() API contract
assumes that the VMA is already linked.
Remove the handle_mm_fault() call from the mmap callback. For imported
BOs, mark the mapping as invalid and rely on the first command submission
to fault in the pages.
For shmem BOs, set the VM_MIXEDMAP flag and use vm_insert_pages().
Implement amdxdna_gem_mixed_vm_ops to handle the page faults.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260910211338.1102315-1-lizhi.hou@amd.com
|
|
In amdxdna_gem_obj_open(), abo->lock is held when calling
amdxdna_gem_add_bo_usage(), which then acquires client->mm_lock.
However, the heap update path may acquire these locks in the reverse
order, creating a potential deadlock.
Fix this by saving the client pointer locally before acquiring abo->lock,
and releasing abo->lock before calling amdxdna_gem_add_bo_usage().
Apply the same change to amdxdna_gem_obj_close().
Fixes: 1f513a3ec3a9 ("accel/amdxdna: Add per-process BO memory usage query support")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260902180955.3999242-2-lizhi.hou@amd.com
|
|
Accessing abo->client in amdxdna_gem_del_bo_usage() may result in a
use-after-free when the BO is imported via flink.
Disable flink import by verifying that filp->driver_priv matches the client
stored in abo->client before accessing the client.
Fixes: 3cc5d7a59519 ("accel/amdxdna: Add carveout memory support for non-IOMMU systems")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260902180955.3999242-1-lizhi.hou@amd.com
|
|
Backmerging to get drm-misc-next up to v7.3-rc2. Requested for
commit 3a2c4d55e32a ("treewide: refresh kmalloc_obj() conversions").
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
|
|
SYNC_BO clflushes an imported BO's scatterlist. An importer may not do
that: the memory belongs to the exporter, and dma-buf gives the importer
no interface to ask for maintenance on it. Refuse the request instead.
is_import_bo() is (obj)->attach, which covers more than foreign buffers.
A userptr BO arrives through a ubuf, and on a carveout device every share
BO and the device heap arrive through a cbuf, so SYNC_BO answers
-EOPNOTSUPP for those too, including the AMDXDNA_BO_DEV path that flushes
through its heap.
Only the ubuf case gives up maintenance it was getting: on a 64 MiB
userptr BO a 4 KiB sync and a full sync both cost 659 us, this arm having
ignored the range. amdxdna_cbuf_map() fills in only the DMA address and
length, so drm_clflush_sg() already walks zero pages on carveout memory.
Userspace maintains these through the mapping it already holds, as XRT's
buffer::sync() does unless it is told to sync through the driver.
Fixes: dbc8fd7a03cb ("accel/amdxdna: Add expandable device heap support")
Reported-by: Christian König <christian.koenig@amd.com>
Link: https://lore.kernel.org/dri-devel/a505f9e5-b416-43e9-934d-c5c29b8a70e9@amd.com/
Suggested-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Taimuraz Kaitmazov <taimuraz@kaitmazov.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260819224458.257346-5-taimuraz@kaitmazov.com
|
|
When the DRM file descriptor is closed, amdxdna_client_cleanup() frees
the client structure. If device BOs are still alive, freeing them may
access abo->client and result in a use-after-free.
Move amdxdna_gem_heap_free(), which accesses abo->client, before clearing
abo->client in the BO close callback. Also track the heap BOs referenced
by device BOs so they can be accessed without dereferencing abo->client.
Fixes: dbc8fd7a03cb ("accel/amdxdna: Add expandable device heap support")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260901160129.3812405-1-lizhi.hou@amd.com
|
|
In amdxdna_gem_obj_free(), flush_workqueue(xdna->notifier_wq) waits for
all pending work items on the device-global notifier workqueue, rather
than only the work items associated with the BO being freed.
If another BO has a pending hmm_unreg_work, freeing an unrelated BO can
be unnecessarily blocked until that work completes.
mmu_interval_notifier_remove() is deferred to a workqueue because it
cannot be called from the MMU notifier callback itself. The BO free path
is not a notifier callback, so call mmu_interval_notifier_remove()
directly there and avoid the workqueue.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260813205016.941743-1-lizhi.hou@amd.com
|
|
Let's start the 7.3 drm-misc-fixes cycle.
Signed-off-by: Maxime Ripard <mripard@kernel.org>
|
|
amdxdna_cmd_set_error() looks up the first BO of a command chain, which
takes a reference, and drops it at the end of the function. The mapping of
that BO is established in between, and the failure path returns without the
put, so the reference is leaked.
Ordinary use does not reach it. The chain has been submitted before any of
this runs, so aie2_cmdlist_fill_slot() has already called
amdxdna_cmd_get_op() on that BO and amdxdna_gem_vmap() has cached its
address. What makes it reachable is that the BO is resolved again by
handle here, and the handle is userspace's to recycle: closing it after
submission and importing a dma-buf whose exporter implements no vmap onto
the same id leaves amdxdna_gem_get_obj() returning an object this cannot
map, since prime_import() types every import AMDXDNA_BO_SHARE.
Fixes: d76856beb4a4 ("accel/amdxdna: Refactor GEM BO handling and add helper APIs for address retrieval")
Signed-off-by: Taimuraz Kaitmazov <taimuraz@kaitmazov.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260819230852.287751-1-taimuraz@kaitmazov.com
|
|
A chain whose command_count is zero passes the payload length check,
because struct_size(payload, data, 0) is just the header. The fill loop
then does not run, so offset stays zero and the request is submitted with
a zero-length buffer.
On firmware without AIE2_NPU_COMMAND that ends at the opcode check, since
op is still ERT_INVALID_CMD and aie2_get_chain_msg_op() answers
MSG_OP_MAX_OPCODE. aie2_get_npu_chain_msg_op() answers
MSG_OP_CHAIN_EXEC_NPU whatever it is given, so there the submission
continues to drm_clflush_virt_range(cmd_buf, 0), which reads the byte
before the buffer and faults on the vmap guard page. EXEC_CMD is
reachable by any process that can open the render node.
Reject the request instead.
Fixes: 8ed8b0239617 ("accel/amdxdna: Add debug prints for command submission")
Signed-off-by: Taimuraz Kaitmazov <taimuraz@kaitmazov.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260818000019.369366-1-taimuraz@kaitmazov.com
|
|
struct amdxdna_cmd_chain contains a flexible array annotated with
__counted_by(command_count). Since the structure is stored in shared
AMDXDNA_BO_SHARE memory, userspace can modify command_count concurrently.
If command_count is changed to zero, the bounds check generated from
__counted_by may fail and trigger a kernel panic.
Remove __counted_by to avoid relying on the userspace-controlled
command_count for the flexible array bounds check.
Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260821033543.1839719-1-lizhi.hou@amd.com
|
|
Pull drm updates from Dave Airlie:
"Highlights:
- dmemcg eviction support is good for low VRAM things like Steam
Machine
- AMD adds gfx6-8 modifier support for older GPUs that enables a
bunch of wayland stuff
- i915/xe has some new hw support but also a lot of display
refactoring
Everything:
perf:
- export perf_allow_ APIs for xe
udmabuf:
- remove default size limit of 64MB
rust:
- i/o rework (signed tag from driver-core tree)
- add registration guard and registration data
- fix unbounded lifetimes in ioctl handler args
- fix a drm_dev_register race
- gem_shmem: add DmaResvGuard helper
- gpuvm: require send/sync for driver data
- implement send/sync for GpuVaAlloc and GpuVmBo
- add SmContext lifetime
- rename dma_handle to dma_address
- change pci_sriov_get_totalvfs return to unsigned int
core:
- create drm_of_get_panel_orientation
- send per-connector hotplug events
- add thunderbolt UBHR tunneling support
connector:
- add color format property
dmem:
- introduce a peak file
- accept one region per limit
- add dmemcg support for eviction
gpusvm:
- reorg code to give drivers more flexibility
atomic:
- add create_state callback and helper
- add documentation on atomic commit lifetime
buddy:
- add per-order free
- add used block scoreboard
- fix UAF
- test buffer clearance on resume
- add phys_addr->block helper
gem:
- drop DRIVER_GEM_GPUVA flag
ttm:
- be more aggressive allocating below protection limit
sched:
- add test suite for concurrent job submissions
hdmi:
- hook the color format property in helpers
mipi-dsi:
- add MIPI_DSI_MODE_DSC_ALL_SLICES_IN_PKT
bridge:
- add atomic create callbacks
- drop atomic reset
- display-connector: don't autoenable HPD IRQ
- trigger initial HPD for DP
- ti-sn65dsi83: remove NO_HFP and NO_HBP mode flags
- analogix_dp: switch to DP link training helpers
dp:
- add support for DSC max delta BPP
edid:
- parse panel type from DisplayID 2.x Display Parameters
sysfb:
- improve panel, stride, framebuffer size validation
panel:
- implement ref counting for struct drm_panel
- himax-hx83121a: add backlight regulator support
- novatek-nt36672a: Inline panel init sequences
- visionox-vtdr6130: enable DSC
- novatek-nt37801: Use mipi_dsi_*_multi() functions
- samsung-s6d16d0: Fix prepare error handling
- support Novatek NT36536 plus DT bindings
- sofef00: fix backlight updates
- osd101t2587: use mipi_dsi_*_multi interface
- panel-edp: adjust timing for AUO displays
- panel-lvds: support Opto Logic SCX1001511GGC49
- panel-simple: support Kyocera tcg070wvlq
- panel-edp: quirks
- AUO B116XAT04.3, CMN N116BCP-EA2, CSW MNB601LS1-8
- BOE NV116WH2-M30, BOE NT116WHM-N21, BOE NV116FH1-M31
- BOE NV116FH1-M30, NV140FHM-N5B, TM156VDXP25
- BOE NE160QDM-NY1, MB116AS01
- new:
- Samsung ATNA40HQ08-0, Anbernic TD4310
- Chipone ICNA35XX, Ilitek ILI9488
- Ilitek ILI7807S, Renesas R63419
- MNE001BS6-2, MNF601BS4-1, Sharp LQ120P1JX51
virtio:
- add support for save/restore virtio_gpu_objects
- abort vq wait on device removal
amdgpu:
- add color format DRM property
- initial compute pipe reset support
- add GFX 6-8 modifier support
- initial DCN 6.0.0 support
- dmemcg eviction support
- improved boundary checking for bios parsing
- RAS updates and rework
- VCN secure submission fixes
- 8K panel fix
- Display KUNIT tests
- parse panel type from DisplayID
- Align IP discovery to pci device lifetime
- SOC15 register macro cleanups
- UVD memory placement fixes
- GFX9 mode2 reset fixes
- drop unnecessary BUG/BUG_ON
- GFX8 soft reset rework
- enable soft reset on GFX8
- PSP/SMU 15.0.9 update
- VI ASPM fix
- userq fixes
- amdgpu_vm_get_task_info_pasid lifetime fix
- DC CACP support
- change system_unbound_wq with system_dfl_wq
- Loosen VFCT bios parsing to deal with pci=realloc
- SI/SMU7 AC/DC switch fix
- VM fence handling fix
- GEM close optimisation
- Apple Studio Display fixes
- DC FRL fixes
amdkfd:
- initial compute pipe reset support
- allow applications to opt out of sigbus on fatal errors
- improve CRIU boundary checks
- MQD handling rework
- move TBA/TMA from system to device memory
- avoid topology-lock in kfd_mmap
- SVM eviction fixes
radeon:
- fix unset CONFIG_ACPI build
i915:
- Novalake (NVL display version 35) timing generator enabling
- NVL DC3CO enabling
- enable UBHR link rates on thunderbolt tunnels
- Reduce Xe3+ PM demand peak bandwidth
- enable pipe DMC error interrupts for display 30+
- add kunit tests for DP link config selection
- refactor and document DP link recovery
- i915/xe driver display probe/remove/suspend/resume/shutdown cleanup
and unification
- i915/xe display runtime PM unified
- Break i915 and xe panic dependency on struct intel_framebuffer
- Streamline Pre/Post-CSC LUT loops
- drop TGL DC3DO support
- CDCLK santization
- fix HDMI scrambling enable
- fix phys bo pread/pwrite with offset
- add missing nospec on parallel submit slot
- fix some NULL derefs
xe:
- drop force_execlist module param
- gate observation streams with perf_allow_cpu
- skip FORCE_WC and vm_bound check for external dma-bufs
- dmemcg eviction support
- remove unused NVL-S GuC
- TLB invalidation improvements
- NVL-S updated PCI-IDs and w/a
- madvise: optimise invalidation path
- fix infinite gt-reset loop in timeout recovery
- update TTM device benefical_order
- wait on external BO kernel fences in exec ioctl
- add/use more KLV helpers
- sriov: disable display in admin only PF mode
- add RAS GPU health indicator
- optimise TTM populate for DONTNEED BO
- drop force_probe for NVL-s
- add debugfs for pcode info
amdxdna:
- disable device buffer export
nova:
- build nova-core/nova-drm from drivers/gpu
- export nova-core rust symbols (workaround)
- GSP boot process consolidation
- Boot GSP with vGPU enabled
- TLV firmware image format support
- Hopper/Blackwell fixes and cleanups
- I/O projection adoption
tyr:
- firmware loading and MCU boot
- add generic slot manager + MMU
- GPU VM support ARM64 LPAE page tables
- add kernel buffer object for internal allocations
- add parser for Mali CSF
- add MCU booting
nouveau:
- race fixes
- check instmem iomapping at first use
- add dmemcg support
- expose NVDEC channels
- add scanline position/head state support for GSP
qxl:
- convert simple encoder to regular
ethosu:
- add perf counter support
etnaviv:
- force flush on power register ops
msm:
- support DSC configuration with slice_per_pkt > 1
mxsfb:
- fix disable sequence
panthor:
- support sparse mappings
rockchip:
- switch away from simple helpers
- support YUV background color
- fix layer config timeout
- add edp support for rk3576
- add batch command submission function
rocket:
- error handling and NULL ptr deref fixes
sun4i:
- switch away from simple helpers
imagination:
- mark BXM-4-64 MC1 as support
host1x:
- support tegra264
tegra:
- add DSI for tegra 20/30
v3d:
- reduce PM runtime autosuspend delay
- scheduler fixes and refactoring
- deprecate v3d 3.3 and 4.1
- validate CPU job query boundaries
hibmc:
- improve plane format handling
- switch to gem shmem
mediatek:
- cec: correct compat for mt7623-8167?
exynos:
- remove simple dependency
- add error handling to encoder paths
- take i2c adapter module reference"
* tag 'drm-next-2026-08-20' of https://gitlab.freedesktop.org/drm/kernel: (2074 commits)
drm/xe/mcr: Take vcs1/vecs1 into account for first media slice
drm/xe: Fix a bug in pc_adjust_freq_bounds()
drm/xe: Fix xe_device_probe() failure
drm/xe/drm_ras: Move has_drm_ras check to drm_ras layer
drm/xe/ras: Fix boot-time ras error processing
drm/amd/display: make DC_RUN_WITH_PREEMPTION_ENABLED misuse a build error
drm/amd/pm: silence uninitialized variable warnings
drm/amdgpu: skip BOs being torn down during GTT recovery
drm/amdgpu: Reject UVD message with invalid number of h265 refs
drm/amdgpu: keep PRT mappings off the vm_bo state lists
drm/amdgpu: fix nbif 6.3.1 l1 low power not functional
drm/amd/display: fix BT.2020 YCbCr output CSC matrices for DCE
drm/amd/display: fix BT.2020 YCbCr limited output CSC matrix
drm/amdgpu: Implement insert_end for VCE 3
drm/amdgpu: Fix UVD min buffer sizes
drm/amdgpu: Fix UVD decode image min size calculation
drm/amdgpu: Fix UVD dpb min size calculation for H264
drm/amdgpu: Reject UVD message with dimensions above 4096
drm/amdgpu: check ASPM on the dGPU host link
drm/radeon: fix autosuspend cleanup during teardown
...
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull MM updates from Andrew Morton:
- "mm: drop "sub" prefix from various places" (Dev Jain)
page->folio conversion and a naming cleanup
- "mm/kasan: remove redundant initialization for kasan_flag_write_only"
(Igor Putko)
KASAN cleanup work
- "mm/filemap: reduce unnecessary xarray lookups" (Chi Zhiling)
Small speedup in the pagecaache read code
- "mm/percpu: Fix possible NOFS/NOIO reclaim recursion" (Kaitao Cheng)
Improve the vmalloc code - mainly the avoidance of GFP_KERNEL
allocations when the caller asked for GFP_NOFS or GFP_NOIO
- "mm/kmemleak: avoid soft lockup when scanning task stacks" (Breno
Leitao)
Avoid a soft lockup watchdog trigger from the kmemleak scanning code
in extreme situations
- "mm/page_owner: misc cleanups" (Ye Liu)
Cleanups to the page_owner code. For some reason lots of people have
been working on the page_owner code this cycle.
- "mm: convert to walk_page_range_vma() to eliminate find_vma()"
(Kefeng Wang)
Simplify and accelerate the page walking library function
- "mm/migrate: preparatory cleanups for batch copy and offload"
(Shivank Garg)
Cleanups in the migration code
- "mm/page_owner: add per-fd filter infrastructure for print_mode and
NUMA filtering" (Zhen Ni)
Per-fd filtering to page_owner in order to reduce the sometimes vast
amount of output it can produce
- "mm: Refactor bootmem gigantic hugepage allocation" (Muchun Song)
Fixes and preparatory cleanups around bootmem HugeTLB handling,
sparse initialization ordering, and related vmemmap setup
- "mm/zsmalloc: reduce lock contention in zs_free()" (Wenchao Hao)
Reduce lock contention in zs_free(), which dominates the unmap path
under memory pressure on Android (LMK kills) and on x86 servers
running zswap-heavy workloads.
Up to 1.83x improvement in microbenchmarking.
- "move alloc_tag.c file under mm/" (Suren Baghdasaryan)
- "samples/damon: handle damon_{start,stop}() failures" (SJ Park)
Fix improper handling of damon_start(), damon_stop(), and
damon_call() failures across DAMON sample modules to prevent
potential memory leaks, operation disruptions and use-after-free
bugs
- "mm/damon/sysfs: kobject_del() directories that users can
create/remove" (SJ Park)
Fix delayed sysfs directory removal under DEBUG_KOBJECT_RELEASE
causeing creation failures due to duplicate directory names by adding
missing kobject_del() calls before creating new directories
- "mm: cleanup clear_not_present_full_ptes()" (David Hildenbrand)
Clean up the core pte handling code
- "selftests/damon: misc fixes for test bugs" (Kunwu Chan)
Fix several bugs in the DAMON selftests
- "selftests/damon: fix memcg_path staging handling" (Cheng Nie)
Fix a bug in _damon_sysfs.py for damos_filter memcg_path setup, and
add a test case for it in sysfs.py.
- "selftests/damon: test kdamond refresh_ms" (Ruslan Valiyev)
Selftest coverage for DAMON's refresh_ms sysfs feature by updating
the test control module and verifying that scheme stats update
automatically without manual intervention
- "mm/damon: five misc fixups" (Akinobu Mita)
Miscellaneous DAMON fixups.
- "mm/damon/core: detect internal variation above max_nr_regions/2"
(Jiayuan Chen)
Fix DAMON's region splitting behavior when region counts exceed half
the maximum budget by dynamically scaling down the split fraction as
the limit approaches, preventing large regions from staying un-split,
and add corresponding KUnit test coverage
- "mm: preparatory patches for PMD level swap entries" (Usama Arif)
Refactor and clean up PMD softleaf helpers, call sites, and
architecture flags to lay the groundwork for a follow-up series that
introduces PMD page table swap entries
- "mm/damon: update, optimize, and clean up doc, tests, and code" (SJ
Park)
Update DAMON design and ABI documentation, expands unit and selftest
coverage, optimize damon_commit_target_regions(), and clean up
recently added sysfs interface code for better readability
- "mm/vmpressure: reduce CPU, memory and code overhead on cgroup v2"
(Usama Arif)
Optimize vmpressure() by skipping unnecessary work on cgroup v2 for
userspace event notifications and refactor v1-only eventfd handling
into mm/memcontrol-v1.c to reduce memory overhead and code complexity
- "selftests/mm: refactor pkey helpers and fix mmap error handling"
(Hongfu Li)
Refactor pkeys shared tracing and assertion helpers into a common
file, unify protection key selftests to use consistent diagnostic
logging and assertions, and enforce standardized MAP_FAILED return
checks for mmap() calls across the tests
- "mm/damon: optimize out nr_accesses_bp" (SJ Park)
Replace the error-prone, continuously updated nr_accesses_bp field in
damon_region with an on-demand moving sum function, reducing
structure memory overhead and avoiding state corruption bugs
- "Open HugeTLB allocation routine for more generic use" (Ackerley Tng)
Decouple HugeTLB folio allocation from VMA dependencies by
introducing hugetlb_alloc_folio(), enabling subsystems like
guest_memfd to allocate HugeTLB folios without standard VMA
reservations or pseudo-VMAs
- "mm/damon: provide pseudo moving sum probe_hits" (SJ Park)
Integrate DAMON's probe_hits attribute counter into the pseudo moving
sum infrastructure, enabling real-time, online monitoring without
waiting for full aggregation intervals
- "mm: Some cleanups for page allocator APIs" (Brendan Jackman)
Simplify and refactor the page allocator entry points and flags by
unifying allocation paths, adding internal alloc_flags arguments, and
eliminating redundant __ prefixed alloc_pages variants.
- "Fix incorrect access of hugetlb pte entries" (Dev Jain)
Enforce the consistent use of huge_ptep_get() instead of ptep_get()
for HugeTLB entries and fixes an unaligned address issue in arm64's
huge_ptep_get() implementation
- "mm/damon: validate all parameters in the core" (SJ Park)
Consolidate parameter validation into the DAMON core specifically
within damon_start() and damon_commit_ctx() to centralize error
checking, eliminate caller-side redundant checks and to improve
maintenance efficiency
- "tools/mm/page_owner_sort: fix filtering and cleanup issues" (Yichong
Chen)
Rename is_need() to filter_record() for clearer return semantics, fix
per-record allocation memory leaks and bound output copies in
search_pattern() to address an existing buffer issue
- "memcg: bail out reclaim when memcg is dying" (Jiayuan Chen)
Mitigate a system-wide stall which occurs when a cgroup is removed
while one of its memory control files is doing synchronous reclaim
- "mm/memory-failure: add panic option for unrecoverable pages" (Breno
Leitao)
Introduce an opt-in vm.panic_on_unrecoverable_memory_failure sysctl
that immediately panics the kernel on unrecoverable memory errors in
kernel-owned pages to preserve error context and prevent delayed,
silent data corruption
- "mm/damon: refactor damon_{start,stop,commit}() for simple error
handling" (SJ Park)
Refactor the DAMON core API functions to guarantee that all contexts
are fully stopped when damon_start(), damon_stop(), or damon_commit()
fail, eliminating the need for complex and error-prone caller-side
cleanup code
- "Keep tail page private zero at free and folio split" (Zi Yan)
Add checks to ensure tail_page->private is zero when freeing compound
or high-order pages and when promoting tail pages during large folio
splits. By validating these fields at free and split time, it allows
the removal of redundant private field clearing inside
prep_compound_tail()
- "mm: drop redundant lru_add_drain in anon folio reuse paths" (Barry
Song)
Eliminate redundant lru_add_drain() calls in
wp_can_reuse_anon_folio() and do_swap_page() to reduce LRU lock
contention and system overhead
By validating folio refcounts against the LRU cache before draining
and removing unnecessary drains in the swap path, it achieves up to a
30.5% reduction in drain calls during heavy swap workloads
- "mm: clean up folio LRU and swap declarations" (Jianyue Wu)
Reorganize folio LRU and swap code by relocating page-cluster state
to mm/swap_state.c, renaming mm/swap.c to mm/folio.c, and moving
MM-internal reclaim declarations into mm/internal.h.
- "userfaultfd: working set tracking for VM guest memory" (Kiryl
Shutsemau)
Add userfaultfd support for tracking the working set of VM guest
memory, so a VMM can identify hot pages and reclaim cold ones to
tiered or remote storage
- "mm: remove CONFIG_HAVE_BOOTMEM_INFO_NODE (Part 2)" (David
Hildenbrand)
Remove the remaining pieces of CONFIG_HAVE_BOOTMEM_INFO_NODE,
performing some smaller cleanups around freeing of reserved vmemmap
pages on the way.
- "mm/damon: update probe hits for runtime parameter commits" (SJ Park)
Ensure that DAMON's probe_hits attribute counter is properly updated
when monitoring intervals are changed at runtime, matching the
behavior of nr_accesses. To achieve this, it refactors and renames
existing helper functions for shared use, applies the updates to
probe_hits, and handles edge cases in damon_probe_hits_mvsum() to
maintain measurement accuracy.
- "KSM: performance optimizations for rmap_walk_ksm" (xu xin)
Resolve a severe KSM reverse-mapping performance bottleneck where
thousands of split VMAs sharing a single anon_vma cause extended lock
contention.
By adding an interval-filtering check during the rmap walk, it
reduces worst-case anon_vma lock hold times from over 500ms down to
under 2ms, preventing application freezes and latency spikes under
memory pressure.
- "mm: split a couple of headers from internal.h" (Mike Rapoport)
Split declarations related to mm_init, memblock, vmalloc and sparse
into new headers
- "KSM: use linear_page_index in collect_procs_ksm()" (xu xin)
Apply the interval tree optimization from rmap_walk_ksm() to
collect_procs_ksm() to avoid iterating over non-matching VMAs during
KSM memory error handling.
It hoists loop-invariant address initialization and restricts the
anon_vma_interval_tree_foreach walk to a targeted page offset range,
reducing redundant checks and improving lookup efficiency.
- "selftests/mm: avoid false failures in hugetlb and KSM tests" (Sayali
Patil)
Fix issues in the hugetlb and KSM MM selftest categories that can
report failures when the prerequisites for the tests are not
satisfied
- "mm/damon: introduce data attributes only monitoring" (SJ Park)
Introduce attribute-weighted region management in DAMON, allowing
users to prioritize specific data attributes (such as page sizes or
cgroups) over or instead of access monitoring.
By assigning weights to attribute probes, DAMON can completely
disable access tracking and adjust monitoring regions based on
weighted probe-hit counters to optimize monitoring quality for
attribute-focused workloads.
- "mm/hmm: Add mmap lock-drop support for userfaultfd-backed mappings"
(Stanislav Kinsburskii)
Extend hmm_range_fault() to support userfaultfd-backed regions by
allowing the mmap lock to be dropped during fault handling via a new
hmm_range_fault_locked() helper.
By accepting a locked pointer and signaling retry status when lock
release occurs, it enables page fault resolution in userfaultfd
regions while preserving backward compatibility for existing callers.
- "mm: make VMA page offset handling more consistent" (Lorenzo Stoakes)
Clean up and standardize how vma->vm_pgoff is accessed and
manipulated across file-backed and anonymous mappings in the kernel
It introduces dedicated helper functions such as vma_start_pgoff(),
vma_end_pgoff(), vma_set_pgoff() and linear_page_delta() while
renaming rmap interval tree helpers to better reflect their
functionality.
These changes establish a cleaner foundation for future work that
will unify virtual page offset indexing for all anonymous and CoW'd
folios.
- "mm: handle device-private PMDs in walk callbacks" (Usama Arif)
Address kernel panics and state corruption caused by MM walk
callbacks reaching non-present device-private PMD swap entries
created during HMM migrations
It ensures that functions which acquire pmd_trans_huge_lock()
properly recognize device-private PMDs instead of assuming a present
THP or a standard migration entry.
- "mm/rmap: Refactor try_to_unmap_one" (Dev Jain)
Refactor try_to_unmap_one by modularizing Hugetlb,
anonymous-lazyfree, and anonymous-swapbacked logic into dedicated
functions, laying the structural groundwork for batched anonymous
large folio unmapping.
- "Docs/ABI/damon: sysfs ABI document fixes and additions" (Song Hu)
Fix typos and fills in missing entries in the DAMON sysfs ABI
document
- "dax/kmem: atomic whole-device hotplug via sysfs" (Gregory Price)
Introduce an atomic sysfs state attribute and supporting DAX/MM
infrastructure to prevent userland races when offlining and removing
entire memory regions
By adding an unplugged state alongside standard online modes, it
enables whole-device atomic hotplug control while preserving backward
compatibility.
- "mm: convert more vm_flags_t users to vma_flags_t" (Lorenzo Stoakes)
Continue transitioning the kernel from the deprecated vm_flags_t type
to vma_flags_t across core memory management infrastructure.
It replaces legacy type usage in core functions such as do_mmap(),
unmapped area allocation, mm->def_vma_flags, and VMA operations like
mlock, mprotect, and mremap.
- "Two small patches to clean up mm/mm_slot.h" (xu xin)
Refactor mm_slot.h by introducing mm_slot_remove() to unify duplicate
slot deletion sequences in khugepaged and KSM. It also adds code
documentation explaining why mm_slot_lookup and mm_slot_insert must
remain as preprocessor macros rather than static inline functions.
- "mm/damon/core: hide core-private struct fields" (SJ Park)
Clean up DAMON core structures by consistently marking internal-only
fields with private: comment tags to prevent improper direct access
from outer layers.
It enforces encapsulation across core structures including
damon_region, damon_target, and damon_ctx and updates DAMON_SYSFS to
interact through approved access APIs instead of exposing raw struct
members.
- "mm/damon: unurgent fixes for infinite loop, NULL de-ref and races"
(SJ Park)
Address potential infinite loops, NULL dereferences, and race
conditions identified in DAMON
It fixes an infinite loop triggered by extreme user configurations, a
NULL pointer dereference within unit tests and minor monitoring
accuracy degradation caused by subtle runtime races.
- "mm/page_alloc: fixes for free_pages_nolock() on RT/UP" (Brendan
Jackman)
Fix an NMI safety flaw in __free_frozen_pages() where freeing pages
on non-SMP or PREEMPT_RT kernels can bypass can_spin_trylock() checks
via non-PCP or isolated migration paths.
It also resolves potential kernel crashes and privilege escalation
risks triggered when BPF tracing runs in NMI context alongside memory
hotplug or large allocation frees.
- "mm/page_alloc: couple of followups for recent cleanups" (Brendan
Jackman)
Clean up and update page allocator nomenclature, documentation, and
debug assertions.
It aligns internal FPI_ flags with the public "nolock" naming
convention, removes outdated internal implementation details from
high-level page allocator comments, and eliminates obsolete
VM_BUG_ON() assertions in allocation paths.
- "mm/mseal: further cleanups" (Lorenzo Stoakes)
Refactor and simplify the mseal implementation by clarifying API
boundaries and removing unnecessary code complexity.
It replaces generic do_mseal() usage outside the syscall with a
dedicated mseal_mmap_page_zero() helper for MMAP_PAGE_ZERO,
eliminates mm_struct parameters to enforce that sealing applies only
to current->mm, and streamlines overall logic and comments with no
functional changes intended.
- "mm/vmscan: fix swappiness=max and clean up per-node proactive
reclaim" (Ridong Chen)
Resolve reclaim behavior bugs and clean up function parameters across
memory reclaim paths
It fixes swappiness=max in both standard reclaim and MGLRU so
unswappable anonymous memory no longer falls back to evicting page
cache, ensures reclaim_store() returns accurate error codes instead
of collapsing all failures into -EAGAIN, and removes the obsolete
gfp_mask parameter from __node_reclaim().
- "mm: mincore: misc cleanups" (Kefeng Wang)
Clean up and simplifies the mincore code. Most importantly, it
removes the historical special behavior that always reports VM_PFNMAP
pages as non-resident.
- "mm/huge_memory: drop dead split helper variants" (Kiryl Shutsemau)
Two trivial cleanups in the folio split API
- "mm/damon: fix uninitialized DAMOS field and kunit exec expectation
bugs" (SJ Park)
Resolve minor operational and testing bugs in DAMON identified by
Sashiko. It initializes the damos->last_applied field to prevent
occasional efficiency degradation and fixes invalid memory accesses
in DAMON KUnit tests during test failure handling.
- "cleanup for stable_page_flags()" (Jinjiang Tu)
Clean up and refactor stable_page_flags() used by /proc/kpageflags
without altering functionality.
It uses BIT_ULL() to prevent shift-overflow warnings on 64-bit flag
bits, converts folio-specific flag checks to standard folio_test_*()
helpers, and removes redundant CONFIG_PAGE_IDLE_FLAG handling.
- "Batch unmap of uffd-wp file folios" (Dev Jain)
Extend batched folio unmapping support to file folios within
userfaultfd write-protect (uffd-wp) VMAs by adding batching
capabilities to pte_install_uffd_wp_if_needed().
This removes special-case restrictions on uffd-wp VMAs in
try_to_unmap_one(), significantly simplifying the function's control
flow and complexity.
- "mm/early_ioremap: clarify and clean up early_ioremap_reset()"
(Sang-Heon Jeon)
Clarify and clean up the architecture-specific usage of
__late_set_fixmap() and __late_clear_fixmap() after
early_ioremap_reset()
It adds explicit documentation regarding when early_ioremap_reset()
must be called and removes redundant macro definitions and reset
calls in the RISC-V and ARM64 architectures.
- "mm: fix reclaim storms in defrag_mode" (Johannes Weiner)
Address severe performance regressions, swap storms, and spurious
OOMs caused by vm.defrag_mode=1 under high memory pressure in Meta
production
It updates the page allocator slowpath so non-movable allocation
requests actively trigger direct reclaim and direct compaction at
pageblock_order scale, allowing them to claim whole pageblocks rather
than spinning unproductively.
- "zram: lockmap tweaks" (Sebastian Siewior)
Optimize and fix lockdep tracking for zram devices by consolidating
per-entry lockmaps and isolate lock classes across multiple instances
This reduces memory overhead by replacing per-entry lockdep_map
instances with a single map per struct zram, and assigns a dynamic
lock_class_key to each instance to prevent false deadlock reports
when different zram devices are backed by distinct filesystems.
* tag 'mm-stable-2026-08-18-18-39' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (501 commits)
selftests/mm: thuge-gen: fix test_shmget() for PAGE_SIZE check
selftests/mm: unpoison pages in memory-failure teardown
mm/shmem: downgrade final i_blocks check in shmem_evict_inode() to pr_warn()
mm/khugepaged: replace mutex_lock/mutex_unlock usage with guard macro
mm/zsmalloc: fix release order of locks in zs_page_migrate()
Documentation: zram: remove sections numbering
ksm: stop iterating VMAs when ksm_test_exit returns true
mm: fold userfaultfd_rwp() to false without CONFIG_ARCH_HAS_PTE_PROTNONE
mm/migrate: report RCU-tasks quiescent states in migrate_pages_batch()
zram: use a custom key for each zram object
zram: move lockmap to be per-zram instead per table
selftests/mm: fix gup_longterm EINVAL error message
mm: page_alloc: fix non-movable reclaim storm in defrag_mode
mm: page_alloc: move capture_control to the page allocator
mm: compaction: support non-movable compaction for pageblock requests
mm: page_alloc: __GFP_FS lockdep annotation for direct compaction
hugetlb: evaluate subpool free state while locked
mm/damon: remove trailing semicolons after function definitions
mm/damon/ops-common: prevent migration fallback to non-target nodes
mm/damon: update outdated comment about DAMOS filter handling
...
|
|
Linux 7.2
There was a lot of conflicts this round between fixes and next,
and I'd like to get the merge resolutions that we have in drm-tip.
Signed-off-by: Dave Airlie <airlied@redhat.com>
|
|
SYNC_BO does not constrain its size, so a request for zero bytes reaches
drm_clflush_virt_range(), which ends with an unconditional
clflushopt(end - 1). For an empty range that is the byte before the
mapping, and abo->mem.kva comes from vmap(), so the access lands in the
guard page below the vmalloc area and faults:
BUG: unable to handle page fault for address: ffffd16fbbc70fff
#PF: supervisor read access in kernel mode
Oops: Oops: 0000 [#1] SMP NOPTI
CPU: 7 UID: 1000 Comm: sync_bo_probe
RIP: 0010:drm_clflush_virt_range+0x3c/0x70
Call Trace:
amdxdna_drm_sync_bo_ioctl+0x124/0x430 [amdxdna]
drm_ioctl+0x301/0x4c0
__x64_sys_ioctl+0x115/0x2f0
do_syscall_64+0xa6/0x3d0
Any process that can open the render node can do this. Reproduced 3 of 3
times on a Strix Point NPU (1022:17f0), by calling SYNC_BO with size 0 on
an AMDXDNA_BO_SHARE object. The import arm takes the same request but
flushes the whole scatterlist, so it survives it.
Nothing needs flushing for an empty range, so answer before choosing a
path.
Fixes: e252e3f3488a ("accel/amdxdna: Revise device bo creation and free")
Cc: stable@vger.kernel.org
Signed-off-by: Taimuraz Kaitmazov <taimuraz@kaitmazov.com>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260817230655.356785-1-taimuraz@kaitmazov.com
|
|
aie2_populate_range() incorrectly failed jobs for BOs with multiple
mmaps: if the unmapped entry appeared first in umap_list, the loop would
pick it up, call hmm_range_fault() on a gone VMA, and return -EFAULT
without ever trying the remaining valid mapps.
Fix it by skipping unmapped entries. After the loop, if the map list is
empty or all maps are valid, map_invalid can be cleared normally.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260812205628.810816-1-lizhi.hou@amd.com
|
|
This reverts commit ac58121339db0178186d256a956bb65feb8b6e45.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Cc: Luke.Wildhardt@proton.me
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Danilo Krummrich <dakr@kernel.org>
Cc: Philipp Stanner <phasta@kernel.org>
Cc: Christian König <ckoenig.leichtzumerken@gmail.com>
Signed-off-by: Tvrtko Ursulin <tursulin@ursulin.net>
Link: https://lore.kernel.org/r/20260811163139.99746-16-tvrtko.ursulin@igalia.com
|
|
aie2_populate_range() takes mmap_read_lock() only around
hmm_range_fault(). It also open-codes the mmu interval sequence setup
before each HMM walk and retries -EBUSY until HMM_RANGE_DEFAULT_TIMEOUT
expires.
Use hmm_range_fault_unlocked_timeout() instead. The HMM helper now owns
the mmap lock and refreshes mapp->range.notifier_seq for its internal
retries, so the driver only needs to call the helper and then validate the
sequence before marking the mapping populated.
Pass HMM_RANGE_DEFAULT_TIMEOUT as the helper retry budget for each HMM
population attempt. This scopes the timeout to repeated HMM notifier
retries while preserving the existing outer loop that moves between
invalid mappings and restarts when the interval is invalidated before the
driver updates its mapping state.
Keep returning -ETIME when the HMM retry budget expires, matching the
driver's existing timeout error convention.
Link: https://lore.kernel.org/20260723-hmm-v10-v11-7-c55b003a4b61@gmail.com
Signed-off-by: Stanislav Kinsburskii <skinsburskii@gmail.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Cc: Danilo Krummrich <dakr@kernel.org>
Cc: Dave Airlie <airlied@gmail.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Dexuan Cui <decui@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: Leon Romanovsky <leon@kernel.org>
Cc: Liam R. Howlett <liam@infradead.org>
Cc: Lizhi Hou <lizhi.hou@amd.com>
Cc: Long Li <longli@microsoft.com>
Cc: Lorenzo Stoakes <ljs@kernel.org>
Cc: Lyude <lyude@redhat.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Maxime Ripard <mripard@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Oded Gabbay <ogabbay@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Thomas Zimemrmann <tzimmermann@suse.de>
Cc: Vlastimil Babka <vbabka@kernel.org>
Cc: Wei Liu <wei.liu@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
|
|
In amdxdna_insert_pages(), vm_flags_mod() sets VM_MIXEDMAP and clears
VM_PFNMAP. If an unprivileged userspace process mmaps a non-imported GEM
object and then calls madvise(MADV_DONTNEED), the PTEs will be
successfully cleared because VM_MIXEDMAP allows this (unlike VM_PFNMAP).
When userspace subsequently accesses the memory, drm_gem_shmem_fault()
handles the page fault and attempts to map the backing shmem page via
vmf_insert_pfn() which calls vmf_insert_pfn_prot(). Because the backing
shmem page is normal system memory (pfn_valid(pfn) is true) and the VMA
now has VM_MIXEDMAP set, won't this predictably trigger the explicit
assertion BUG_ON((vma->vm_flags & VM_MIXEDMAP) && pfn_valid(pfn))
Fix by removing the vm_flags_mod() call and replacing the vm_insert_pages()
pre-population with the handle_mm_fault() loop that was already used for
the import (dma-buf) path.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260731185955.3449311-1-lizhi.hou@amd.com
|
|
Two error paths in amdxdna_insert_pages() called vma->vm_ops->close(vma)
before returning an error code to the caller. This is incorrect:
amdxdna_gem_obj_mmap() registers an HMM interval notifier before calling
amdxdna_insert_pages(), and on a hard error it jumps to hmm_unreg to undo
that registration. Calling vm_ops->close() manually — which drops the
shmem pages_pin_count and the GEM object reference that backs the VMA —
before the mmap syscall has even returned causes those resources to be
released while the VMA is still alive. The kernel VMA teardown will call
vm_ops->close() a second time when the process later unmaps the range,
producing a reference count underflow.
Replace both hard-error returns with a deferred-fault approach that keeps
the VMA alive and retries page insertion through the HMM range-fault path.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260723074256.2435143-1-lizhi.hou@amd.com
|
|
When two commands enter aie2_sched_job_timedout() concurrently, both
check the timeout detection state. The first scheduler thread observes
tdr_status as SIGNALED and updates it to WAIT. The second thread then
observes the updated state instead of the original SIGNALED state, which
may cause the command timeout to be handled incorrectly.
Replace tdr_status with last_signal_ts, which records the timestamp of
the last driver signal. Timeout detection now only reads
last_signal_ts and never modifies it, allowing multiple serialized
detect() calls under dev_lock to evaluate the same signal timestamp
independently. If there is not any new job scheduled or completed
within tdr_timeout_ms, the command will timeout.
Fixes: 9022f010977f ("accel/amdxdna: Check for device hang on job timeout")
Signed-off-by: Wendy Liang <wendy.liang@amd.com>
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260718083409.1825940-1-lizhi.hou@amd.com
|
|
amdxdna_cmd_submit() stores current->mm in job->mm without holding any
reference. aie2_sched_job_run() later access job->mm from the DRM
scheduler worker thread. With only a raw pointer and no structural
reference, the mm_struct can be freed before the scheduler runs the job.
Fix this by calling mmgrab() to hold a structural mm_count reference for
the lifetime of the job, paired with mmdrop() in every cleanup path.
Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260716151305.1595780-1-lizhi.hou@amd.com
|
|
amdxdna_cmd_submit() calls xdna->dev_info->ops->cmd_submit()
unconditionally, but only aie2_dev_ops defines that callback.
aie4_vf_ops (the AIE4 SR-IOV virtual function) does not, so a user
AMDXDNA_EXEC_CMD ioctl on an AIE4 device reaches a NULL function-pointer
call and oopses the kernel. AIE4 submits work through a mapped user queue
and doorbell, not this ioctl path.
Reject the submission early with -EOPNOTSUPP when the device provides no
cmd_submit op, so the shared EXEC ioctl is a clean no-op on such devices.
Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Cc: stable@vger.kernel.org
Found by 0sec automated security-research tooling (https://0sec.ai).
Assisted-by: 0sec:claude-opus-4-8
Signed-off-by: Doruk Tan Ozturk <doruk@0sec.ai>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260713173030.87541-3-doruk@0sec.ai
|
|
amdxdna_drm_submit_execbuf() passes the user-supplied command BO handle
straight into amdxdna_cmd_submit() with drv_cmd == NULL. When the handle
is AMDXDNA_INVALID_BO_HANDLE (0), the block that fetches job->cmd_bo is
skipped, leaving it NULL, and no check rejects it on the user path (the
!job->cmd_bo guard lives inside the != INVALID branch).
The job is then armed and pushed to the DRM scheduler.
aie2_sched_job_run() takes the drv_cmd == NULL path and calls
amdxdna_cmd_set_state(job->cmd_bo) -> amdxdna_gem_vmap(NULL) ->
to_gobj(NULL)->dev, a NULL pointer dereference in the drm_sched worker.
A process with access to the accel node on a system with a probed AMD NPU
can trigger a kernel oops with a single AMDXDNA_EXEC_CMD ioctl
(cmd_handles = 0).
Only internal driver commands (SYNC_DEBUG_BO / ATTACH_DEBUG_BO)
legitimately pass AMDXDNA_INVALID_BO_HANDLE, and they always set drv_cmd.
Reject the invalid handle for user submissions (drv_cmd == NULL) at the
submit choke point so every user path is covered.
Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Cc: stable@vger.kernel.org
Found by 0sec automated security-research tooling (https://0sec.ai).
Assisted-by: 0sec:claude-opus-4-8
Signed-off-by: Doruk Tan Ozturk <doruk@0sec.ai>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260713173030.87541-2-doruk@0sec.ai
|
|
Closing a BO handle clears abo->client, while the underlying GEM object
may remain alive due to internal kernel references. As a result, code
executed after the BO handle is closed may dereference a NULL abo->client
pointer.
Remove accesses to abo->client from code paths that may execute after the
BO handle has been closed.
Fixes: d76856beb4a4 ("accel/amdxdna: Refactor GEM BO handling and add helper APIs for address retrieval")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260707201556.562191-1-lizhi.hou@amd.com
|
|
The return value of init_srcu_struct() is currently ignored. If
initialization fails, subsequent use of hwctx_srcu may result in invalid
memory accesses.
Check the return value of init_srcu_struct() and propagate the error to
the caller.
Fixes: aac243092b70 ("accel/amdxdna: Add command execution")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260707172323.539721-1-lizhi.hou@amd.com
|
|
drmm_mutex_init() may fail and return an error. Check the return value
and abort initialization if mutex creation fails.
Fixes: 8c9ff1b181ba ("accel/amdxdna: Add a new driver for AMD AI Engine")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260707172307.539676-1-lizhi.hou@amd.com
|
|
amdxdna_update_heap() iterates over hardware contexts while holding
xdna->dev_lock. During the iteration, amdxdna_pm_resume_get_locked() may
temporarily release and reacquire the lock, allowing hardware contexts to
be modified concurrently.
Fix the race by calling amdxdna_pm_resume_get_locked() before
iterating over hardware contexts.
Fixes: dbc8fd7a03cb ("accel/amdxdna: Add expandable device heap support")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260707055818.479165-1-lizhi.hou@amd.com
|
|
Both amdxdna_hwctx_sync_debug_bo() and amdxdna_drm_config_hwctx_ioctl()
hold xdna->dev_lock while invoking backend operations. If the hardware
hangs, aie2_cmd_wait() blocks waiting for a firmware response. When the
DRM scheduler timeout expires, aie2_sched_job_timedout() is invoked to
reset the hardware. However, the timeout handler also attempts to acquire
dev_lock, resulting in a deadlock.
Avoid this by releasing dev_lock before waiting for the firmware
response and reacquiring it after the wait completes. This allows the
timeout handler to proceed with device recovery when a debug BO command
times out.
Fixes: 7ea046838021 ("accel/amdxdna: Support firmware debug buffer")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260707055732.479103-1-lizhi.hou@amd.com
|
|
amdxdna_gem_obj_open() increments open_ref before attempting to set up
the DMA address mapping. When amdxdna_dma_map_bo() fails, the function
returned immediately without rolling back either change made on the first
open (open_ref == 1 path).
Fix it by decrementing open_ref and clearing abo->client on the error path.
Fixes: ece3e8980907 ("accel/amdxdna: Allow forcing IOVA-based DMA via module parameter")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260707055658.479049-1-lizhi.hou@amd.com
|
|
amdxdna_umap_release() calls the blocking mmu_interval_notifier_remove()
before removing the object from abo->mem.umap_list. If
aie2_populate_range() runs concurrently, it may obtain a reference to an
amdxdna_umap that is being released, leading to a potential use-after-free.
Use kref_get_unless_zero() in aie2_populate_range() when acquiring a
reference. If the reference count has already dropped to zero, release
is in progress and the entry is skipped.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260706221228.434099-1-lizhi.hou@amd.com
|
|
aie2_populate_range() and amdxdna_umap_release() access a saved VMA
pointer that may have already been freed, leading to a potential
use-after-free.
Remove the VMA accesses from these functions to avoid the race.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260609011242.2833740-1-lizhi.hou@amd.com
|
|
nr_pages is declared as u32 in amdxdna_hmm_register(), which may not be
large enough to represent the number of pages for large mappings.
Use unsigned long for nr_pages to avoid potential overflow.
Fixes: ac49797c1815 ("accel/amdxdna: Add GEM buffer object management")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260616061532.3533469-1-lizhi.hou@amd.com
|
|
amdxdna_hwctx_sync_debug_bo() invokes the hardware hwctx_sync_debug_bo()
callback while holding xdna->dev_lock.
The callback may call amdxdna_cmd_submit(), which in turn calls
amdxdna_pm_resume_get(). If the device is suspended,
amdxdna_pm_resume_get() may synchronously execute amdxdna_pm_resume(),
which also acquires xdna->dev_lock, resulting in a deadlock.
Avoid the deadlock by calling amdxdna_pm_resume_get() before holding
xdna->dev_lock in both amdxdna_hwctx_sync_debug_bo() and
amdxdna_drm_config_hwctx_ioctl()
Fixes: 7ea046838021 ("accel/amdxdna: Support firmware debug buffer")
Reviewed-by: Max Zhen <max.zhen@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260616212429.3620645-1-lizhi.hou@amd.com
|
|
https://gitlab.freedesktop.org/drm/misc/kernel into drm-next
drm-misc-next for 7.3:
UAPI Changes:
- connector: Add color format property
Cross-subsystem Changes:
- dmem: introduce a peak file
Core Changes:
- atomic: Add create_state callback and helpers to all objects, add
documentation on atomic commit lifetime
- buddy: Fix use-after-free, add per-order free and used block
scoreboards
- gem: Remove DRIVER_GEM_GPUVA feature flag
- hdmi: Hook the color format property in the helpers
- mipi-dsi: Add MIPI_DSI_MODE_DSC_ALL_SLICES_IN_PKT flag
- sched: Add test suite for concurrent job submissions
- virtio: Add support for saving and restoring virtio_gpu_objects,
abort virtqueue wait on device removal to avoid hung task
Driver Changes:
- amdgpu: Implement "color format" DRM property
- amdxdna: Disable device buffer exporting
- ethosu: Add performance counter support
- msm: Support DSC configurations with slice_per_pkt > 1
- mxsfb: Fix disable sequence
- panthor: Support sparse mappings
- rockchip: Support YUV background color, Fix layer config timeout, add
edp support for rk3576, cleanups and formats improvements
- solomon: Add a batch command submission function
- tegra: Add DSI support for Tegra 20 and 30, fix dsi driver when the
firmware hasn't enabled the controller,
- v3d: Reduce PM runtime autosuspend delay, Scheduler and submission
fixes and refactoring, Deprecate V3D 3.3 and 4.1 support
- bridge:
- display-connector: don't autoenable HPD IRQ, trigger initial HPD
event for DP
- dw-dp: Null pointer dereference and use-after-free fixes
- ti-sn65dsi83: Remove NO_HFP and NO_HBP mode flags
- panel:
- himax-hx83121a: add backlight regulator support
- novatek-nt36672a: Inline panel init sequences
- panel-edp: Add quirks for AUO B116XAT04.3, CMN N116BCP-EA2, CSW
MNB601LS1-8, BOE NV116WH2-M30, BOE NT116WHM-N21, BOE NV116FH1-M31,
BOE NV116FH1-M30, NV140FHM-N5B, TM156VDXP25
- New panels: Samsung ATNA40HQ08-0, Anbernic TD4310, Chipone
ICNA35XX, Ilitek ILI9488
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@redhat.com>
Link: https://patch.msgid.link/20260619-burgundy-termite-of-whirlwind-f7a4dd@houat
|
|
When vm_insert_pages() fails, the error path calls vma->vm_ops->close(vma)
which internally calls drm_gem_vm_close() → drm_gem_object_put(),
releasing the GEM object reference acquired at the start of the function.
However, the close_vma label then falls through to put_obj, which calls
drm_gem_object_put() a second time on the same object.
If the first put releases the last reference, the object is freed and the
second put accesses freed memory, causing a use-after-free.
Fix by returning directly from close_vma instead of falling through to
put_obj, since the close handler already performs all necessary cleanup
including the object put.
Cc: stable@vger.kernel.org
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
Reviewed-by: Lizhi Hou <lizhi.hou@amd.com>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260625113239.49764-1-vulab@iscas.ac.cn
|
|
When a debug BO command completes, job->drv_cmd may already have been
freed. Accessing it from aie2_sched_drvcmd_resp_handler() can result in
a use-after-free and memory corruption.
Fix this by introducing reference counting for drv_cmd objects and
transferring ownership to the job while it is in flight. This ensures
that the command remains valid until the completion handler finishes
processing it.
Fixes: 7ea046838021 ("accel/amdxdna: Support firmware debug buffer")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260701155556.663541-1-lizhi.hou@amd.com
|
|
When force_iova mode is enabled, amdxdna_remove() frees xdna->domain. If
amdxdna_gem_obj_free() is called after device removal, it may attempt to
access xdna->domain, resulting in a use-after-free.
Fix the race by adding freeing xdna->domain as a managed release action,
so its lifetime is managed by DRM and remains valid until all managed
resources are released.
Fixes: ece3e8980907 ("accel/amdxdna: Allow forcing IOVA-based DMA via module parameter")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260611055150.3070216-3-lizhi.hou@amd.com
|
|
amdxdna_remove() destroys notifier_wq. If amdxdna_gem_obj_free() is
called after device removal, it may attempt to flush notifier_wq,
resulting in a use-after-free.
Fix the race by allocating notifier_wq with
drmm_alloc_ordered_workqueue(), so its lifetime is managed by DRM and
remains valid until all managed resources are released.
Fixes: e486147c912f ("accel/amdxdna: Add BO import and export")
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20260611055150.3070216-2-lizhi.hou@amd.com
|
