user/sven/linux.git/drivers/accel, branch v6.15.7

accel/ivpu: Fix warning in ivpu_gem_bo_free()

2025-06-27T10:13:10Z

commit 91274fd4ed9ba110b02c53d71d2778b7d13b49ac upstream. Don't WARN if imported buffers are in use in ivpu_gem_bo_free() as they can be indeed used in the original context/driver. Fixes: 647371a6609d ("accel/ivpu: Add GEM buffer object management") Cc: stable@vger.kernel.org # v6.3 Reviewed-by: Jeff Hugo Reviewed-by: Lizhi Hou Signed-off-by: Jacek Lawrynowicz Link: https://lore.kernel.org/r/20250528171220.513225-1-jacek.lawrynowicz@linux.intel.com Signed-off-by: Greg Kroah-Hartman

accel/ivpu: Use dma_resv_lock() instead of a custom mutex

2025-06-27T10:13:10Z

commit 98d3f772ca7d6822bdfc8c960f5f909574db97c9 upstream. This fixes a potential race conditions in: - ivpu_bo_unbind_locked() where we modified the shmem->sgt without holding the dma_resv_lock(). - ivpu_bo_print_info() where we read the shmem->pages without holding the dma_resv_lock(). Using dma_resv_lock() also protects against future syncronisation issues that may arise when accessing drm_gem_shmem_object or drm_gem_object members. Fixes: 42328003ecb6 ("accel/ivpu: Refactor BO creation functions") Cc: stable@vger.kernel.org # v6.9+ Reviewed-by: Lizhi Hou Signed-off-by: Jacek Lawrynowicz Link: https://lore.kernel.org/r/20250528154325.500684-1-jacek.lawrynowicz@linux.intel.com Signed-off-by: Greg Kroah-Hartman

accel/ivpu: Trigger device recovery on engine reset/resume failure

2025-06-27T10:13:10Z

commit a47e36dc5d90dc664cac87304c17d50f1595d634 upstream. Trigger full device recovery when the driver fails to restore device state via engine reset and resume operations. This is necessary because, even if submissions from a faulty context are blocked, the NPU may still process previously submitted faulty jobs if the engine reset fails to abort them. Such jobs can continue to generate faults and occupy device resources. When engine reset is ineffective, the only way to recover is to perform a full device recovery. Fixes: dad945c27a42 ("accel/ivpu: Add handling of VPU_JSM_STATUS_MVNCI_CONTEXT_VIOLATION_HW") Cc: stable@vger.kernel.org # v6.15+ Signed-off-by: Karol Wachowski Reviewed-by: Lizhi Hou Signed-off-by: Jacek Lawrynowicz Link: https://lore.kernel.org/r/20250528154253.500556-1-jacek.lawrynowicz@linux.intel.com Signed-off-by: Greg Kroah-Hartman

accel/ivpu: Use firmware names from upstream repo

2025-06-27T10:13:09Z

commit 1c2c0e29f24360b3130c005a3c261cb8c7b363c6 upstream. Use FW names from linux-firmware repo instead of deprecated ones. The vpu_37xx.bin style names were never released and were only used for internal testing, so it is safe to remove them. Fixes: c140244f0cfb ("accel/ivpu: Add initial Panther Lake support") Cc: stable@vger.kernel.org # v6.13+ Reviewed-by: Lizhi Hou Reviewed-by: Jeff Hugo Signed-off-by: Jacek Lawrynowicz Link: https://lore.kernel.org/r/20250506092030.280276-1-jacek.lawrynowicz@linux.intel.com Signed-off-by: Greg Kroah-Hartman

accel/ivpu: Improve buffer object logging

2025-06-27T10:13:09Z

commit a01e93ee44f7ed76f872d0ede82f8d31bf0a048a upstream. - Fix missing alloc log when drm_gem_handle_create() fails in drm_vma_node_allow() and open callback is not called - Add ivpu_bo->ctx_id that enables to log the actual context id instead of using 0 as default - Add couple WARNs and errors so we can catch more memory corruption issues Fixes: 37dee2a2f433 ("accel/ivpu: Improve buffer object debug logs") Cc: stable@vger.kernel.org # v6.8+ Reviewed-by: Jeff Hugo Reviewed-by: Lizhi Hou Signed-off-by: Jacek Lawrynowicz Link: https://lore.kernel.org/r/20250506091303.262034-1-jacek.lawrynowicz@linux.intel.com Signed-off-by: Greg Kroah-Hartman

accel/amdxdna: Fix incorrect PSP firmware size

2025-06-19T13:40:54Z

[ Upstream commit 779a0c9e06a91d0007f2a4f4071e3b9a8102b15e ] The incorrect PSP firmware size is used for initializing. It may cause error for newer version firmware. Fixes: 8c9ff1b181ba ("accel/amdxdna: Add a new driver for AMD AI Engine") Acked-by: Alex Deucher Signed-off-by: Lizhi Hou Link: https://lore.kernel.org/r/20250604143217.1386272-1-lizhi.hou@amd.com Signed-off-by: Sasha Levin

accel/ivpu: Reorder Doorbell Unregister and Command Queue Destruction

2025-06-19T13:40:34Z

[ Upstream commit 4557cc834712eca4eae7adbd9f0a06bdd8f79c99 ] Refactor ivpu_cmdq_unregister() to ensure the doorbell is unregistered before destroying the command queue. The NPU firmware requires doorbells to be unregistered prior to command queue destruction. If doorbell remains registered when command queue destroy command is sent firmware will automatically unregister the doorbell, making subsequent unregister attempts no-operations (NOPs). Ensure compliance with firmware expectations by moving the doorbell unregister call ahead of the command queue destruction logic, thus preventing unnecessary NOP operation. Fixes: 465a3914b254 ("accel/ivpu: Add API for command queue create/destroy/submit") Signed-off-by: Karol Wachowski Reviewed-by: Jeff Hugo Signed-off-by: Jacek Lawrynowicz Link: https://lore.kernel.org/r/20250515094124.255141-1-jacek.lawrynowicz@linux.intel.com Signed-off-by: Sasha Levin

accel/amdxdna: Fix incorrect size of ERT_START_NPU commands

2025-06-19T13:39:30Z

[ Upstream commit 6c161732ea6467c6dea0c35810ca8e8d1ae135f1 ] When multiple ERT_START_NPU commands are combined in one buffer, the buffer size calculation is incorrect. Also, the condition to make sure the buffer size is not beyond 4K is also fixed. Fixes: aac243092b70 ("accel/amdxdna: Add command execution") Reviewed-by: Jacek Lawrynowicz Reviewed-by: Maciej Falkowski Signed-off-by: Lizhi Hou Link: https://lore.kernel.org/r/20250409210013.10854-1-lizhi.hou@amd.com Signed-off-by: Sasha Levin

accel/ivpu: Use effective buffer size for zero terminator

2025-05-12T13:12:44Z

Use the effective written size instead of original size as index for zero termination. If the input from user-space is to larger and the input is truncated, the original size is out-of-bound. Since there is an upfront size check here, the change is for consistency. Signed-off-by: Markus Burri Reviewed-by: Jacek Lawrynowicz Signed-off-by: Jacek Lawrynowicz Link: https://lore.kernel.org/r/20250508130612.82270-3-markus.burri@mt.com

accel/ivpu: Correct mutex unlock order in job submission

2025-04-30T09:27:37Z

The mutex unlock for vdev->submitted_jobs_lock was incorrectly placed before unlocking file_priv->lock. Change order of unlocks to avoid potential race conditions. Fixes: 5bbccadaf33e ("accel/ivpu: Abort all jobs after command queue unregister") Signed-off-by: Karol Wachowski Reviewed-by: Jeff Hugo Signed-off-by: Jacek Lawrynowicz Link: https://lore.kernel.org/r/20250425093656.2228168-1-jacek.lawrynowicz@linux.intel.com