Commit 5bf06c4c authored by Dave Airlie's avatar Dave Airlie
Browse files

Merge tag 'amd-drm-fixes-6.1-2022-11-09' of... 

Merge tag 'amd-drm-fixes-6.1-2022-11-09' of https://gitlab.freedesktop.org/agd5f/linux

 into drm-fixes

amd-drm-fixes-6.1-2022-11-09:

amdgpu:
- SMU 13.0.4 update
- GPUVM TLB race fix
- DCN 3.1.4 fixes
- DCN 3.2.x fixes
- Vega10 fan fix
- BACO fix for Beige Goby board
- PSR fix
- GPU VM PT locking fixes

amdkfd:
- CRIU fixes

Signed-off-by: default avatarDave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221109234554.6028-1-alexander.deucher@amd.com
parents f0c4d9fc 675d8462

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

+1 −0
Original line number Diff line number Diff line
@@ -992,6 +992,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, 
		kvfree(e->user_pages);

		e->user_pages = NULL;

	}

	mutex_unlock(&p->bo_list->bo_list_mutex);

	return r;

}

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

+0 −26
Original line number Diff line number Diff line
@@ -143,32 +143,6 @@ int amdgpu_vm_set_pasid(struct amdgpu_device *adev, struct amdgpu_vm *vm, 
	return 0;

}



/*

 * vm eviction_lock can be taken in MMU notifiers. Make sure no reclaim-FS

 * happens while holding this lock anywhere to prevent deadlocks when

 * an MMU notifier runs in reclaim-FS context.

 */

static inline void amdgpu_vm_eviction_lock(struct amdgpu_vm *vm)

{

	mutex_lock(&vm->eviction_lock);

	vm->saved_flags = memalloc_noreclaim_save();

}



static inline int amdgpu_vm_eviction_trylock(struct amdgpu_vm *vm)

{

	if (mutex_trylock(&vm->eviction_lock)) {

		vm->saved_flags = memalloc_noreclaim_save();

		return 1;

	}

	return 0;

}



static inline void amdgpu_vm_eviction_unlock(struct amdgpu_vm *vm)

{

	memalloc_noreclaim_restore(vm->saved_flags);

	mutex_unlock(&vm->eviction_lock);

}



/**

 * amdgpu_vm_bo_evicted - vm_bo is evicted

 *

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h

+41 −0
Original line number Diff line number Diff line
@@ -492,7 +492,48 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m); 
 */

static inline uint64_t amdgpu_vm_tlb_seq(struct amdgpu_vm *vm)

{

	unsigned long flags;

	spinlock_t *lock;



	/*

	 * Workaround to stop racing between the fence signaling and handling

	 * the cb. The lock is static after initially setting it up, just make

	 * sure that the dma_fence structure isn't freed up.

	 */

	rcu_read_lock();

	lock = vm->last_tlb_flush->lock;

	rcu_read_unlock();



	spin_lock_irqsave(lock, flags);

	spin_unlock_irqrestore(lock, flags);



	return atomic64_read(&vm->tlb_seq);

}



/*

 * vm eviction_lock can be taken in MMU notifiers. Make sure no reclaim-FS

 * happens while holding this lock anywhere to prevent deadlocks when

 * an MMU notifier runs in reclaim-FS context.

 */

static inline void amdgpu_vm_eviction_lock(struct amdgpu_vm *vm)

{

	mutex_lock(&vm->eviction_lock);

	vm->saved_flags = memalloc_noreclaim_save();

}



static inline bool amdgpu_vm_eviction_trylock(struct amdgpu_vm *vm)

{

	if (mutex_trylock(&vm->eviction_lock)) {

		vm->saved_flags = memalloc_noreclaim_save();

		return true;

	}

	return false;

}



static inline void amdgpu_vm_eviction_unlock(struct amdgpu_vm *vm)

{

	memalloc_noreclaim_restore(vm->saved_flags);

	mutex_unlock(&vm->eviction_lock);

}



#endif

drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c

+2 −0
Original line number Diff line number Diff line
@@ -597,7 +597,9 @@ static int amdgpu_vm_pt_alloc(struct amdgpu_device *adev, 
	if (entry->bo)

		return 0;



	amdgpu_vm_eviction_unlock(vm);

	r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt);

	amdgpu_vm_eviction_lock(vm);

	if (r)

		return r;

drivers/gpu/drm/amd/amdkfd/kfd_chardev.c

+15 −19
Original line number Diff line number Diff line
@@ -1950,7 +1950,7 @@ static int criu_checkpoint(struct file *filep, 
{

	int ret;

	uint32_t num_devices, num_bos, num_objects;

	uint64_t priv_size, priv_offset = 0;

	uint64_t priv_size, priv_offset = 0, bo_priv_offset;



	if (!args->devices || !args->bos || !args->priv_data)

		return -EINVAL;
@@ -1994,38 +1994,34 @@ static int criu_checkpoint(struct file *filep, 
	if (ret)

		goto exit_unlock;



	ret = criu_checkpoint_bos(p, num_bos, (uint8_t __user *)args->bos,

			    (uint8_t __user *)args->priv_data, &priv_offset);

	if (ret)

		goto exit_unlock;

	/* Leave room for BOs in the private data. They need to be restored

	 * before events, but we checkpoint them last to simplify the error

	 * handling.

	 */

	bo_priv_offset = priv_offset;

	priv_offset += num_bos * sizeof(struct kfd_criu_bo_priv_data);



	if (num_objects) {

		ret = kfd_criu_checkpoint_queues(p, (uint8_t __user *)args->priv_data,

						 &priv_offset);

		if (ret)

			goto close_bo_fds;

			goto exit_unlock;



		ret = kfd_criu_checkpoint_events(p, (uint8_t __user *)args->priv_data,

						 &priv_offset);

		if (ret)

			goto close_bo_fds;

			goto exit_unlock;



		ret = kfd_criu_checkpoint_svm(p, (uint8_t __user *)args->priv_data, &priv_offset);

		if (ret)

			goto close_bo_fds;

			goto exit_unlock;

	}



close_bo_fds:

	if (ret) {

		/* If IOCTL returns err, user assumes all FDs opened in criu_dump_bos are closed */

		uint32_t i;

		struct kfd_criu_bo_bucket *bo_buckets = (struct kfd_criu_bo_bucket *) args->bos;



		for (i = 0; i < num_bos; i++) {

			if (bo_buckets[i].alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM)

				close_fd(bo_buckets[i].dmabuf_fd);

		}

	}

	/* This must be the last thing in this function that can fail.

	 * Otherwise we leak dmabuf file descriptors.

	 */

	ret = criu_checkpoint_bos(p, num_bos, (uint8_t __user *)args->bos,

			   (uint8_t __user *)args->priv_data, &bo_priv_offset);



exit_unlock:

	mutex_unlock(&p->mutex);