linux.git/drivers/accel/rocket, branch v6.18.21

accel/rocket: fix unwinding in error path in rocket_probe

2026-03-12T11:09:26+00:00

[ Upstream commit 34f4495a7f72895776b81969639f527c99eb12b9 ]

When rocket_core_init() fails (as could be the case with EPROBE_DEFER),
we need to properly unwind by decrementing the counter we just
incremented and if this is the first core we failed to probe, remove the
rocket DRM device with rocket_device_fini() as well. This matches the
logic in rocket_remove(). Failing to properly unwind results in
out-of-bounds accesses.

Fixes: 0810d5ad88a1 ("accel/rocket: Add job submission IOCTL")
Cc: stable@vger.kernel.org
Signed-off-by: Quentin Schulz 
Reviewed-by: Tomeu Vizoso 
Signed-off-by: Tomeu Vizoso 
Link: https://patch.msgid.link/20251215-rocket-error-path-v1-2-eec3bf29dc3b@cherry.de
Signed-off-by: Sasha Levin

accel/rocket: fix unwinding in error path in rocket_core_init

2026-03-12T11:09:26+00:00

[ Upstream commit f509a081f6a289f7c66856333b3becce7a33c97e ]

When rocket_job_init() is called, iommu_group_get() has already been
called, therefore we should call iommu_group_put() and make the
iommu_group pointer NULL. This aligns with what's done in
rocket_core_fini().

If pm_runtime_resume_and_get() somehow fails, not only should
rocket_job_fini() be called but we should also unwind everything done
before that, that is, disable PM, put the iommu_group, NULLify it and
then call rocket_job_fini(). This is exactly what's done in
rocket_core_fini() so let's call that function instead of duplicating
the code.

Fixes: 0810d5ad88a1 ("accel/rocket: Add job submission IOCTL")
Cc: stable@vger.kernel.org
Signed-off-by: Quentin Schulz 
Reviewed-by: Tomeu Vizoso 
Signed-off-by: Tomeu Vizoso 
Link: https://patch.msgid.link/20251215-rocket-error-path-v1-1-eec3bf29dc3b@cherry.de
Signed-off-by: Sasha Levin

accel/rocket: Fix some error checking in rocket_core_init()

2025-09-01T10:11:28+00:00

The problem is that pm_runtime_get_sync() can return 1 on success so
checking for zero doesn't work.  Use the pm_runtime_resume_and_get()
function instead.  The pm_runtime_resume_and_get() function does
additional cleanup as well so that's a bonus as well.

Fixes: 0810d5ad88a1 ("accel/rocket: Add job submission IOCTL")
Signed-off-by: Dan Carpenter 
Signed-off-by: Tomeu Vizoso 
Link: https://lore.kernel.org/r/aKcRW6fsRP_o5C_y@stanley.mountain

accel/rocket: Check the correct DMA irq status to warn about

2025-09-01T10:11:28+00:00

Right now, the code checks the DMA_READ_ERROR state 2 times, while
I guess it was supposed to warn about both read and write errors.

Change the 2nd check to look at the write-error flag.

Fixes: 0810d5ad88a1 ("accel/rocket: Add job submission IOCTL")
Signed-off-by: Heiko Stuebner 
Signed-off-by: Tomeu Vizoso 
Link: https://lore.kernel.org/r/20250818185658.2585696-1-heiko@sntech.de

accel/rocket: Fix usages of kfree() and sizeof()

2025-09-01T10:11:28+00:00

Replace usages of kfree() with kvfree() for pointers which were
allocated using kvmalloc(), as required by the kernel memory management
API.

Use sizeof() on the type that a pointer references instead of the
pointer itself. In this case, scheds and *scheds both happen to be
pointers, so sizeof() will expand to the same value in either case, but
using *scheds is more technically correct since scheds is an array of
drm_gpu_scheduler *.

Reported-by: kernel test robot 
Reported-by: Julia Lawall 
Closes: https://lore.kernel.org/r/202508120730.PLbjlKbI-lkp@intel.com/
Signed-off-by: Brigham Campbell 
Signed-off-by: Tomeu Vizoso 
Link: https://lore.kernel.org/r/20250813-rocket-free-fix-v1-1-51f00a7a1271@brighamcampbell.com
Fixes: 0810d5ad88a1 ("accel/rocket: Add job submission IOCTL")

accel/rocket: Depend on DRM_ACCEL not just DRM

2025-09-01T10:11:28+00:00

With the current dependency on only DRM, a config of

CONFIG_DRM_ACCEL_ROCKET=y

is possible, but of course wrong, because without DRM_ACCEL the build-
system will never even enter drivers/accel/* .

So depend on DRM_ACCEL instead of just DRM.

Fixes: ed98261b4168 ("accel/rocket: Add a new driver for Rockchip's NPU")
Signed-off-by: Heiko Stuebner 
Signed-off-by: Tomeu Vizoso 
Link: https://lore.kernel.org/r/20250814113519.1551855-3-heiko@sntech.de

accel/rocket: Fix indentation of Kconfig entry

2025-09-01T10:11:28+00:00

The general indentation for the Kconfig lines is one tab, so adapt the
lines accordingly.

The description is correctly indented (1 tab + 2 spaces) so doesn't need
changes.

Fixes: ed98261b4168 ("accel/rocket: Add a new driver for Rockchip's NPU")
Signed-off-by: Heiko Stuebner 
Signed-off-by: Tomeu Vizoso 
Link: https://lore.kernel.org/r/20250814113519.1551855-2-heiko@sntech.de

accel/rocket: Fix undeclared const rocket_pm_ops

2025-08-03T15:40:23+00:00

Fix sparse warning regarding an undeclared const rocket_pm_ops, which is
used in rocket_drv.c.

Reported-by: kernel test robot 
Closes: https://lore.kernel.org/oe-kbuild-all/202508030021.uwdr4P08-lkp@intel.com/
Signed-off-by: Brigham Campbell 
Signed-off-by: Tomeu Vizoso 
Link: https://lore.kernel.org/r/20250802-fix-rockchip-npu-build-v1-2-fb0f0dacb3fe@brighamcampbell.com

accel/rocket: Fix Rockchip NPU compilation

2025-08-03T15:40:23+00:00

Replace DRM_GPU_SCHED_STAT_NOMINAL with GPU_DRM_SCHED_STAT_RESET, in
accordance with commit 0a5dc1b67ef5 ("drm/sched: Rename
DRM_GPU_SCHED_STAT_NOMINAL to DRM_GPU_SCHED_STAT_RESET")

Pass extra parameter to drm_sched_job_init, as required by commit
2956554823ce ("drm/sched: Store the drm client_id in drm_sched_fence")

Signed-off-by: Brigham Campbell 
Signed-off-by: Tomeu Vizoso 
Link: https://lore.kernel.org/r/20250802-fix-rockchip-npu-build-v1-1-fb0f0dacb3fe@brighamcampbell.com

accel/rocket: Add IOCTLs for synchronizing memory accesses

2025-07-25T16:04:46+00:00

The NPU cores have their own access to the memory bus, and this isn't
cache coherent with the CPUs.

Add IOCTLs so userspace can mark when the caches need to be flushed, and
also when a writer job needs to be waited for before the buffer can be
accessed from the CPU.

Initially based on the same IOCTLs from the Etnaviv driver.

v2:
- Don't break UABI by reordering the IOCTL IDs (Jeff Hugo)

v3:
- Check that padding fields in IOCTLs are zero (Jeff Hugo)

v6:
- Fix conversion logic to make sure we use DMA_BIDIRECTIONAL when needed
  (Lucas Stach)

v8:
- Always sync BOs in both directions (Robin Murphy)

Reviewed-by: Jeff Hugo 
Tested-by: Heiko Stuebner 
Signed-off-by: Tomeu Vizoso 
Signed-off-by: Jeff Hugo 
Link: https://lore.kernel.org/r/20250721-6-10-rocket-v9-5-77ebd484941e@tomeuvizoso.net