Merge drm-upstream/drm-next into drm-intel-next-queued

Needed for timer_setup() and drm_dev_{get,put}() conversions in i915.
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drivers/dma-buf/dma-buf.c

@@ -625,7 +625,7 @@ EXPORT_SYMBOL_GPL(dma_buf_detach);
 struct sg_table *dma_buf_map_attachment(struct dma_buf_attachment *attach,
 					enum dma_data_direction direction)
 {
-	struct sg_table *sg_table = ERR_PTR(-EINVAL);
+	struct sg_table *sg_table;
 
 	might_sleep();
 

drivers/dma-buf/reservation.c

@@ -266,8 +266,7 @@ EXPORT_SYMBOL(reservation_object_add_excl_fence);
 * @dst: the destination reservation object
 * @src: the source reservation object
 *
-* Copy all fences from src to dst. Both src->lock as well as dst-lock must be
-* held.
+* Copy all fences from src to dst. dst-lock must be held.
 */
 int reservation_object_copy_fences(struct reservation_object *dst,
 				   struct reservation_object *src)
@@ -277,33 +276,62 @@ int reservation_object_copy_fences(struct reservation_object *dst,
 	size_t size;
 	unsigned i;
 
-	src_list = reservation_object_get_list(src);
+	rcu_read_lock();
+	src_list = rcu_dereference(src->fence);
 
+retry:
 	if (src_list) {
-		size = offsetof(typeof(*src_list),
-				shared[src_list->shared_count]);
+		unsigned shared_count = src_list->shared_count;
+
+		size = offsetof(typeof(*src_list), shared[shared_count]);
+		rcu_read_unlock();
+
 		dst_list = kmalloc(size, GFP_KERNEL);
 		if (!dst_list)
 			return -ENOMEM;
 
-		dst_list->shared_count = src_list->shared_count;
-		dst_list->shared_max = src_list->shared_count;
-		for (i = 0; i < src_list->shared_count; ++i)
-			dst_list->shared[i] =
-				dma_fence_get(src_list->shared[i]);
+		rcu_read_lock();
+		src_list = rcu_dereference(src->fence);
+		if (!src_list || src_list->shared_count > shared_count) {
+			kfree(dst_list);
+			goto retry;
+		}
+
+		dst_list->shared_count = 0;
+		dst_list->shared_max = shared_count;
+		for (i = 0; i < src_list->shared_count; ++i) {
+			struct dma_fence *fence;
+
+			fence = rcu_dereference(src_list->shared[i]);
+			if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT,
+				     &fence->flags))
+				continue;
+
+			if (!dma_fence_get_rcu(fence)) {
+				kfree(dst_list);
+				src_list = rcu_dereference(src->fence);
+				goto retry;
+			}
+
+			if (dma_fence_is_signaled(fence)) {
+				dma_fence_put(fence);
+				continue;
+			}
+
+			dst_list->shared[dst_list->shared_count++] = fence;
+		}
 	} else {
 		dst_list = NULL;
 	}
 
+	new = dma_fence_get_rcu_safe(&src->fence_excl);
+	rcu_read_unlock();
+
 	kfree(dst->staged);
 	dst->staged = NULL;
 
 	src_list = reservation_object_get_list(dst);
-
 	old = reservation_object_get_excl(dst);
-	new = reservation_object_get_excl(src);
-
-	dma_fence_get(new);
 
 	preempt_disable();
 	write_seqcount_begin(&dst->seq);

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

@@ -121,6 +121,7 @@ extern int amdgpu_cntl_sb_buf_per_se;
 extern int amdgpu_param_buf_per_se;
 extern int amdgpu_job_hang_limit;
 extern int amdgpu_lbpw;
+extern int amdgpu_compute_multipipe;
 
 #ifdef CONFIG_DRM_AMDGPU_SI
 extern int amdgpu_si_support;
@@ -1310,6 +1311,8 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp);
 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
+int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data,
+				    struct drm_file *filp);
 int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
 int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp);
@@ -1524,7 +1527,6 @@ struct amdgpu_device {
 
 	/* powerplay */
 	struct amd_powerplay		powerplay;
-	bool				pp_enabled;
 	bool				pp_force_state_enabled;
 
 	/* dpm */

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c

@@ -338,6 +338,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	struct cik_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, wptr_val, data;
+	bool valid_wptr = false;
 
 	m = get_mqd(mqd);
 
@@ -356,7 +357,14 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
 	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, data);
 
-	if (read_user_wptr(mm, wptr, wptr_val))
+	/* read_user_ptr may take the mm->mmap_sem.
+	 * release srbm_mutex to avoid circular dependency between
+	 * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex.
+	 */
+	release_queue(kgd);
+	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
+	acquire_queue(kgd, pipe_id, queue_id);
+	if (valid_wptr)
 		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask);
 
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c

@@ -292,6 +292,7 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 	struct vi_mqd *m;
 	uint32_t *mqd_hqd;
 	uint32_t reg, wptr_val, data;
+	bool valid_wptr = false;
 
 	m = get_mqd(mqd);
 
@@ -339,7 +340,14 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
 	WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, data);
 
-	if (read_user_wptr(mm, wptr, wptr_val))
+	/* read_user_ptr may take the mm->mmap_sem.
+	 * release srbm_mutex to avoid circular dependency between
+	 * srbm_mutex->mm_sem->reservation_ww_class_mutex->srbm_mutex.
+	 */
+	release_queue(kgd);
+	valid_wptr = read_user_wptr(mm, wptr, wptr_val);
+	acquire_queue(kgd, pipe_id, queue_id);
+	if (valid_wptr)
 		WREG32(mmCP_HQD_PQ_WPTR, (wptr_val << wptr_shift) & wptr_mask);
 
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);

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

@@ -42,6 +42,28 @@ struct amdgpu_cgs_device {
 	struct amdgpu_device *adev =					\
 		((struct amdgpu_cgs_device *)cgs_device)->adev
 
+static void *amdgpu_cgs_register_pp_handle(struct cgs_device *cgs_device,
+			int (*call_back_func)(struct amd_pp_init *, void **))
+{
+	CGS_FUNC_ADEV;
+	struct amd_pp_init pp_init;
+	struct amd_powerplay *amd_pp;
+
+	if (call_back_func == NULL)
+		return NULL;
+
+	amd_pp = &(adev->powerplay);
+	pp_init.chip_family = adev->family;
+	pp_init.chip_id = adev->asic_type;
+	pp_init.pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false;
+	pp_init.feature_mask = amdgpu_pp_feature_mask;
+	pp_init.device = cgs_device;
+	if (call_back_func(&pp_init, &(amd_pp->pp_handle)))
+		return NULL;
+
+	return adev->powerplay.pp_handle;
+}
+
 static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device,
 				    enum cgs_gpu_mem_type type,
 				    uint64_t size, uint64_t align,
@@ -1179,6 +1201,7 @@ static const struct cgs_ops amdgpu_cgs_ops = {
 	.is_virtualization_enabled = amdgpu_cgs_is_virtualization_enabled,
 	.enter_safe_mode = amdgpu_cgs_enter_safe_mode,
 	.lock_grbm_idx = amdgpu_cgs_lock_grbm_idx,
+	.register_pp_handle = amdgpu_cgs_register_pp_handle,
 };
 
 static const struct cgs_os_ops amdgpu_cgs_os_ops = {

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

@@ -231,7 +231,7 @@ amdgpu_connector_update_scratch_regs(struct drm_connector *connector,
 		if (connector->encoder_ids[i] == 0)
 			break;
 
-		encoder = drm_encoder_find(connector->dev,
+		encoder = drm_encoder_find(connector->dev, NULL,
 					connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
@@ -256,7 +256,7 @@ amdgpu_connector_find_encoder(struct drm_connector *connector,
 	for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) {
 		if (connector->encoder_ids[i] == 0)
 			break;
-		encoder = drm_encoder_find(connector->dev,
+		encoder = drm_encoder_find(connector->dev, NULL,
 					connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
@@ -372,7 +372,7 @@ amdgpu_connector_best_single_encoder(struct drm_connector *connector)
 
 	/* pick the encoder ids */
 	if (enc_id)
-		return drm_encoder_find(connector->dev, enc_id);
+		return drm_encoder_find(connector->dev, NULL, enc_id);
 	return NULL;
 }
 
@@ -1077,7 +1077,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force)
 			if (connector->encoder_ids[i] == 0)
 				break;
 
-			encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]);
+			encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]);
 			if (!encoder)
 				continue;
 
@@ -1134,7 +1134,7 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector)
 		if (connector->encoder_ids[i] == 0)
 			break;
 
-		encoder = drm_encoder_find(connector->dev, connector->encoder_ids[i]);
+		encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
 
@@ -1153,7 +1153,7 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector)
 	/* then check use digitial */
 	/* pick the first one */
 	if (enc_id)
-		return drm_encoder_find(connector->dev, enc_id);
+		return drm_encoder_find(connector->dev, NULL, enc_id);
 	return NULL;
 }
 
@@ -1294,7 +1294,7 @@ u16 amdgpu_connector_encoder_get_dp_bridge_encoder_id(struct drm_connector *conn
 		if (connector->encoder_ids[i] == 0)
 			break;
 
-		encoder = drm_encoder_find(connector->dev,
+		encoder = drm_encoder_find(connector->dev, NULL,
 					connector->encoder_ids[i]);
 		if (!encoder)
 			continue;
@@ -1323,7 +1323,7 @@ static bool amdgpu_connector_encoder_is_hbr2(struct drm_connector *connector)
 	for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) {
 		if (connector->encoder_ids[i] == 0)
 			break;
-		encoder = drm_encoder_find(connector->dev,
+		encoder = drm_encoder_find(connector->dev, NULL,
 					connector->encoder_ids[i]);
 		if (!encoder)
 			continue;

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

@@ -25,6 +25,7 @@
  *    Jerome Glisse <glisse@freedesktop.org>
  */
 #include <linux/pagemap.h>
+#include <linux/sync_file.h>
 #include <drm/drmP.h>
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_syncobj.h>
@@ -1330,6 +1331,66 @@ static struct dma_fence *amdgpu_cs_get_fence(struct amdgpu_device *adev,
 	return fence;
 }
 
+int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data,
+				    struct drm_file *filp)
+{
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	union drm_amdgpu_fence_to_handle *info = data;
+	struct dma_fence *fence;
+	struct drm_syncobj *syncobj;
+	struct sync_file *sync_file;
+	int fd, r;
+
+	if (amdgpu_kms_vram_lost(adev, fpriv))
+		return -ENODEV;
+
+	fence = amdgpu_cs_get_fence(adev, filp, &info->in.fence);
+	if (IS_ERR(fence))
+		return PTR_ERR(fence);
+
+	switch (info->in.what) {
+	case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ:
+		r = drm_syncobj_create(&syncobj, 0, fence);
+		dma_fence_put(fence);
+		if (r)
+			return r;
+		r = drm_syncobj_get_handle(filp, syncobj, &info->out.handle);
+		drm_syncobj_put(syncobj);
+		return r;
+
+	case AMDGPU_FENCE_TO_HANDLE_GET_SYNCOBJ_FD:
+		r = drm_syncobj_create(&syncobj, 0, fence);
+		dma_fence_put(fence);
+		if (r)
+			return r;
+		r = drm_syncobj_get_fd(syncobj, (int*)&info->out.handle);
+		drm_syncobj_put(syncobj);
+		return r;
+
+	case AMDGPU_FENCE_TO_HANDLE_GET_SYNC_FILE_FD:
+		fd = get_unused_fd_flags(O_CLOEXEC);
+		if (fd < 0) {
+			dma_fence_put(fence);
+			return fd;
+		}
+
+		sync_file = sync_file_create(fence);
+		dma_fence_put(fence);
+		if (!sync_file) {
+			put_unused_fd(fd);
+			return -ENOMEM;
+		}
+
+		fd_install(fd, sync_file->file);
+		info->out.handle = fd;
+		return 0;
+
+	default:
+		return -EINVAL;
+	}
+}
+
 /**
  * amdgpu_cs_wait_all_fence - wait on all fences to signal
  *

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

@@ -56,6 +56,7 @@
 #include "amdgpu_vf_error.h"
 
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_pm.h"
 
 MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin");
@@ -1603,6 +1604,7 @@ static int amdgpu_init(struct amdgpu_device *adev)
 			return r;
 		}
 		adev->ip_blocks[i].status.sw = true;
+
 		/* need to do gmc hw init early so we can allocate gpu mem */
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
 			r = amdgpu_vram_scratch_init(adev);
@@ -1633,6 +1635,11 @@ static int amdgpu_init(struct amdgpu_device *adev)
 		}
 	}
 
+	mutex_lock(&adev->firmware.mutex);
+	if (amdgpu_ucode_init_bo(adev))
+		adev->firmware.load_type = AMDGPU_FW_LOAD_DIRECT;
+	mutex_unlock(&adev->firmware.mutex);
+
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (!adev->ip_blocks[i].status.sw)
 			continue;
@@ -1768,6 +1775,8 @@ static int amdgpu_fini(struct amdgpu_device *adev)
 
 		adev->ip_blocks[i].status.hw = false;
 	}
+	if (adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT)
+		amdgpu_ucode_fini_bo(adev);
 
 	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
 		if (!adev->ip_blocks[i].status.sw)
@@ -2040,6 +2049,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	mutex_init(&adev->srbm_mutex);
 	mutex_init(&adev->grbm_idx_mutex);
 	mutex_init(&adev->mn_lock);
+	mutex_init(&adev->virt.vf_errors.lock);
 	hash_init(adev->mn_hash);
 
 	amdgpu_check_arguments(adev);
@@ -2125,7 +2135,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_atombios_init(adev);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_atombios_init failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
 		goto failed;
 	}
 
@@ -2136,7 +2146,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	if (amdgpu_vpost_needed(adev)) {
 		if (!adev->bios) {
 			dev_err(adev->dev, "no vBIOS found\n");
-			amdgpu_vf_error_put(AMDGIM_ERROR_VF_NO_VBIOS, 0, 0);
+			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_NO_VBIOS, 0, 0);
 			r = -EINVAL;
 			goto failed;
 		}
@@ -2144,7 +2154,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		r = amdgpu_atom_asic_init(adev->mode_info.atom_context);
 		if (r) {
 			dev_err(adev->dev, "gpu post error!\n");
-			amdgpu_vf_error_put(AMDGIM_ERROR_VF_GPU_POST_ERROR, 0, 0);
+			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_POST_ERROR, 0, 0);
 			goto failed;
 		}
 	} else {
@@ -2156,7 +2166,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		r = amdgpu_atomfirmware_get_clock_info(adev);
 		if (r) {
 			dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n");
-			amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
 			goto failed;
 		}
 	} else {
@@ -2164,7 +2174,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		r = amdgpu_atombios_get_clock_info(adev);
 		if (r) {
 			dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
-			amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
 			goto failed;
 		}
 		/* init i2c buses */
@@ -2175,7 +2185,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_fence_driver_init(adev);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_fence_driver_init failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0);
 		goto failed;
 	}
 
@@ -2185,7 +2195,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_init(adev);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_init failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
 		amdgpu_fini(adev);
 		goto failed;
 	}
@@ -2205,7 +2215,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_ib_pool_init(adev);
 	if (r) {
 		dev_err(adev->dev, "IB initialization failed (%d).\n", r);
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
 		goto failed;
 	}
 
@@ -2215,6 +2225,10 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 
 	amdgpu_fbdev_init(adev);
 
+	r = amdgpu_pm_sysfs_init(adev);
+	if (r)
+		DRM_ERROR("registering pm debugfs failed (%d).\n", r);
+
 	r = amdgpu_gem_debugfs_init(adev);
 	if (r)
 		DRM_ERROR("registering gem debugfs failed (%d).\n", r);
@@ -2254,7 +2268,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	r = amdgpu_late_init(adev);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_late_init failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r);
 		goto failed;
 	}
 
@@ -2311,6 +2325,7 @@ void amdgpu_device_fini(struct amdgpu_device *adev)
 	iounmap(adev->rmmio);
 	adev->rmmio = NULL;
 	amdgpu_doorbell_fini(adev);
+	amdgpu_pm_sysfs_fini(adev);
 	amdgpu_debugfs_regs_cleanup(adev);
 }
 
@@ -2936,7 +2951,7 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
 		}
 	} else {
 		dev_err(adev->dev, "asic resume failed (%d).\n", r);
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_ASIC_RESUME_FAIL, 0, r);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ASIC_RESUME_FAIL, 0, r);
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			if (adev->rings[i] && adev->rings[i]->sched.thread) {
 				kthread_unpark(adev->rings[i]->sched.thread);
@@ -2950,7 +2965,7 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
 	if (r) {
 		/* bad news, how to tell it to userspace ? */
 		dev_info(adev->dev, "GPU reset failed\n");
-		amdgpu_vf_error_put(AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r);
+		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r);
 	}
 	else {
 		dev_info(adev->dev, "GPU reset successed!\n");

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

@@ -356,6 +356,10 @@ enum amdgpu_pcie_gen {
 		((adev)->powerplay.pp_funcs->switch_power_profile(\
 			(adev)->powerplay.pp_handle, type))
 
+#define amdgpu_dpm_set_clockgating_by_smu(adev, msg_id) \
+		((adev)->powerplay.pp_funcs->set_clockgating_by_smu(\
+			(adev)->powerplay.pp_handle, msg_id))
+
 struct amdgpu_dpm {
 	struct amdgpu_ps        *ps;
 	/* number of valid power states */

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

@@ -70,9 +70,10 @@
  * - 3.18.0 - Export gpu always on cu bitmap
  * - 3.19.0 - Add support for UVD MJPEG decode
  * - 3.20.0 - Add support for local BOs
+ * - 3.21.0 - Add DRM_AMDGPU_FENCE_TO_HANDLE ioctl
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	20
+#define KMS_DRIVER_MINOR	21
 #define KMS_DRIVER_PATCHLEVEL	0
 
 int amdgpu_vram_limit = 0;
@@ -122,6 +123,7 @@ int amdgpu_cntl_sb_buf_per_se = 0;
 int amdgpu_param_buf_per_se = 0;
 int amdgpu_job_hang_limit = 0;
 int amdgpu_lbpw = -1;
+int amdgpu_compute_multipipe = -1;
 
 MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes");
 module_param_named(vramlimit, amdgpu_vram_limit, int, 0600);
@@ -265,6 +267,9 @@ module_param_named(job_hang_limit, amdgpu_job_hang_limit, int ,0444);
 MODULE_PARM_DESC(lbpw, "Load Balancing Per Watt (LBPW) support (1 = enable, 0 = disable, -1 = auto)");
 module_param_named(lbpw, amdgpu_lbpw, int, 0444);
 
+MODULE_PARM_DESC(compute_multipipe, "Force compute queues to be spread across pipes (1 = enable, 0 = disable, -1 = auto)");
+module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444);
+
 #ifdef CONFIG_DRM_AMDGPU_SI
 
 #if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE)

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

@@ -109,9 +109,26 @@ void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_s
 	}
 }
 
+static bool amdgpu_gfx_is_multipipe_capable(struct amdgpu_device *adev)
+{
+	if (amdgpu_compute_multipipe != -1) {
+		DRM_INFO("amdgpu: forcing compute pipe policy %d\n",
+			 amdgpu_compute_multipipe);
+		return amdgpu_compute_multipipe == 1;
+	}
+
+	/* FIXME: spreading the queues across pipes causes perf regressions
+	 * on POLARIS11 compute workloads */
+	if (adev->asic_type == CHIP_POLARIS11)
+		return false;
+
+	return adev->gfx.mec.num_mec > 1;
+}
+
 void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
 {
 	int i, queue, pipe, mec;
+	bool multipipe_policy = amdgpu_gfx_is_multipipe_capable(adev);
 
 	/* policy for amdgpu compute queue ownership */
 	for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) {
@@ -125,8 +142,7 @@ void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
 		if (mec >= adev->gfx.mec.num_mec)
 			break;
 
-		/* FIXME: spreading the queues across pipes causes perf regressions */
-		if (0) {
+		if (multipipe_policy) {
 			/* policy: amdgpu owns the first two queues of the first MEC */
 			if (mec == 0 && queue < 2)
 				set_bit(i, adev->gfx.mec.queue_bitmap);

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

@@ -1024,6 +1024,7 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {
 	DRM_IOCTL_DEF_DRV(AMDGPU_CTX, amdgpu_ctx_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_VM, amdgpu_vm_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_BO_LIST, amdgpu_bo_list_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_FENCE_TO_HANDLE, amdgpu_cs_fence_to_handle_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	/* KMS */
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_MMAP, amdgpu_gem_mmap_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_WAIT_IDLE, amdgpu_gem_wait_idle_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),

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

@@ -64,10 +64,6 @@ static const struct cg_flag_name clocks[] = {
 
 void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev)
 {
-	if (adev->pp_enabled)
-		/* TODO */
-		return;
-
 	if (adev->pm.dpm_enabled) {
 		mutex_lock(&adev->pm.mutex);
 		if (power_supply_is_system_supplied() > 0)
@@ -118,7 +114,7 @@ static ssize_t amdgpu_set_dpm_state(struct device *dev,
 		goto fail;
 	}
 
-	if (adev->pp_enabled) {
+	if (adev->powerplay.pp_funcs->dispatch_tasks) {
 		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL);
 	} else {
 		mutex_lock(&adev->pm.mutex);
@@ -303,7 +299,8 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev,
 
 	if (strlen(buf) == 1)
 		adev->pp_force_state_enabled = false;
-	else if (adev->pp_enabled) {
+	else if (adev->powerplay.pp_funcs->dispatch_tasks &&
+			adev->powerplay.pp_funcs->get_pp_num_states) {
 		struct pp_states_info data;
 
 		ret = kstrtoul(buf, 0, &idx);
@@ -531,7 +528,7 @@ static ssize_t amdgpu_set_pp_sclk_od(struct device *dev,
 	if (adev->powerplay.pp_funcs->set_sclk_od)
 		amdgpu_dpm_set_sclk_od(adev, (uint32_t)value);
 
-	if (adev->pp_enabled) {
+	if (adev->powerplay.pp_funcs->dispatch_tasks) {
 		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL);
 	} else {
 		adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
@@ -575,7 +572,7 @@ static ssize_t amdgpu_set_pp_mclk_od(struct device *dev,
 	if (adev->powerplay.pp_funcs->set_mclk_od)
 		amdgpu_dpm_set_mclk_od(adev, (uint32_t)value);
 
-	if (adev->pp_enabled) {
+	if (adev->powerplay.pp_funcs->dispatch_tasks) {
 		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL);
 	} else {
 		adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps;
@@ -959,9 +956,6 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj,
 	     attr == &sensor_dev_attr_pwm1_min.dev_attr.attr))
 		return 0;
 
-	if (adev->pp_enabled)
-		return effective_mode;
-
 	/* Skip fan attributes if fan is not present */
 	if (adev->pm.no_fan &&
 	    (attr == &sensor_dev_attr_pwm1.dev_attr.attr ||
@@ -1317,6 +1311,9 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
 	if (adev->pm.sysfs_initialized)
 		return 0;
 
+	if (adev->pm.dpm_enabled == 0)
+		return 0;
+
 	if (adev->powerplay.pp_funcs->get_temperature == NULL)
 		return 0;
 
@@ -1341,27 +1338,26 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
 		return ret;
 	}
 
-	if (adev->pp_enabled) {
-		ret = device_create_file(adev->dev, &dev_attr_pp_num_states);
-		if (ret) {
-			DRM_ERROR("failed to create device file pp_num_states\n");
-			return ret;
-		}
-		ret = device_create_file(adev->dev, &dev_attr_pp_cur_state);
-		if (ret) {
-			DRM_ERROR("failed to create device file pp_cur_state\n");
-			return ret;
-		}
-		ret = device_create_file(adev->dev, &dev_attr_pp_force_state);
-		if (ret) {
-			DRM_ERROR("failed to create device file pp_force_state\n");
-			return ret;
-		}
-		ret = device_create_file(adev->dev, &dev_attr_pp_table);
-		if (ret) {
-			DRM_ERROR("failed to create device file pp_table\n");
-			return ret;
-		}
+
+	ret = device_create_file(adev->dev, &dev_attr_pp_num_states);
+	if (ret) {
+		DRM_ERROR("failed to create device file pp_num_states\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_pp_cur_state);
+	if (ret) {
+		DRM_ERROR("failed to create device file pp_cur_state\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_pp_force_state);
+	if (ret) {
+		DRM_ERROR("failed to create device file pp_force_state\n");
+		return ret;
+	}
+	ret = device_create_file(adev->dev, &dev_attr_pp_table);
+	if (ret) {
+		DRM_ERROR("failed to create device file pp_table\n");
+		return ret;
 	}
 
 	ret = device_create_file(adev->dev, &dev_attr_pp_dpm_sclk);
@@ -1417,16 +1413,19 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
 
 void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev)
 {
+	if (adev->pm.dpm_enabled == 0)
+		return;
+
 	if (adev->pm.int_hwmon_dev)
 		hwmon_device_unregister(adev->pm.int_hwmon_dev);
 	device_remove_file(adev->dev, &dev_attr_power_dpm_state);
 	device_remove_file(adev->dev, &dev_attr_power_dpm_force_performance_level);
-	if (adev->pp_enabled) {
-		device_remove_file(adev->dev, &dev_attr_pp_num_states);
-		device_remove_file(adev->dev, &dev_attr_pp_cur_state);
-		device_remove_file(adev->dev, &dev_attr_pp_force_state);
-		device_remove_file(adev->dev, &dev_attr_pp_table);
-	}
+
+	device_remove_file(adev->dev, &dev_attr_pp_num_states);
+	device_remove_file(adev->dev, &dev_attr_pp_cur_state);
+	device_remove_file(adev->dev, &dev_attr_pp_force_state);
+	device_remove_file(adev->dev, &dev_attr_pp_table);
+
 	device_remove_file(adev->dev, &dev_attr_pp_dpm_sclk);
 	device_remove_file(adev->dev, &dev_attr_pp_dpm_mclk);
 	device_remove_file(adev->dev, &dev_attr_pp_dpm_pcie);
@@ -1457,7 +1456,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev)
 			amdgpu_fence_wait_empty(ring);
 	}
 
-	if (adev->pp_enabled) {
+	if (adev->powerplay.pp_funcs->dispatch_tasks) {
 		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL, NULL);
 	} else {
 		mutex_lock(&adev->pm.mutex);
@@ -1592,15 +1591,15 @@ static int amdgpu_debugfs_pm_info(struct seq_file *m, void *data)
 	if  ((adev->flags & AMD_IS_PX) &&
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) {
 		seq_printf(m, "PX asic powered off\n");
-	} else if (adev->pp_enabled) {
-		return amdgpu_debugfs_pm_info_pp(m, adev);
-	} else {
+	} else if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level) {
 		mutex_lock(&adev->pm.mutex);
 		if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level)
 			adev->powerplay.pp_funcs->debugfs_print_current_performance_level(adev, m);
 		else
 			seq_printf(m, "Debugfs support not implemented for this asic\n");
 		mutex_unlock(&adev->pm.mutex);
+	} else {
+		return amdgpu_debugfs_pm_info_pp(m, adev);
 	}
 
 	return 0;

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

@@ -34,24 +34,6 @@
 #include "cik_dpm.h"
 #include "vi_dpm.h"
 
-static int amdgpu_create_pp_handle(struct amdgpu_device *adev)
-{
-	struct amd_pp_init pp_init;
-	struct amd_powerplay *amd_pp;
-	int ret;
-
-	amd_pp = &(adev->powerplay);
-	pp_init.chip_family = adev->family;
-	pp_init.chip_id = adev->asic_type;
-	pp_init.pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false;
-	pp_init.feature_mask = amdgpu_pp_feature_mask;
-	pp_init.device = amdgpu_cgs_create_device(adev);
-	ret = amd_powerplay_create(&pp_init, &(amd_pp->pp_handle));
-	if (ret)
-		return -EINVAL;
-	return 0;
-}
-
 static int amdgpu_pp_early_init(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -59,7 +41,6 @@ static int amdgpu_pp_early_init(void *handle)
 	int ret = 0;
 
 	amd_pp = &(adev->powerplay);
-	adev->pp_enabled = false;
 	amd_pp->pp_handle = (void *)adev;
 
 	switch (adev->asic_type) {
@@ -73,9 +54,7 @@ static int amdgpu_pp_early_init(void *handle)
 	case CHIP_STONEY:
 	case CHIP_VEGA10:
 	case CHIP_RAVEN:
-		adev->pp_enabled = true;
-		if (amdgpu_create_pp_handle(adev))
-			return -EINVAL;
+		amd_pp->cgs_device = amdgpu_cgs_create_device(adev);
 		amd_pp->ip_funcs = &pp_ip_funcs;
 		amd_pp->pp_funcs = &pp_dpm_funcs;
 		break;
@@ -97,9 +76,7 @@ static int amdgpu_pp_early_init(void *handle)
 			amd_pp->ip_funcs = &ci_dpm_ip_funcs;
 			amd_pp->pp_funcs = &ci_dpm_funcs;
 		} else {
-			adev->pp_enabled = true;
-			if (amdgpu_create_pp_handle(adev))
-				return -EINVAL;
+			amd_pp->cgs_device = amdgpu_cgs_create_device(adev);
 			amd_pp->ip_funcs = &pp_ip_funcs;
 			amd_pp->pp_funcs = &pp_dpm_funcs;
 		}
@@ -118,12 +95,9 @@ static int amdgpu_pp_early_init(void *handle)
 
 	if (adev->powerplay.ip_funcs->early_init)
 		ret = adev->powerplay.ip_funcs->early_init(
-					adev->powerplay.pp_handle);
+					amd_pp->cgs_device ? amd_pp->cgs_device :
+					amd_pp->pp_handle);
 
-	if (ret == PP_DPM_DISABLED) {
-		adev->pm.dpm_enabled = false;
-		return 0;
-	}
 	return ret;
 }
 
@@ -137,11 +111,6 @@ static int amdgpu_pp_late_init(void *handle)
 		ret = adev->powerplay.ip_funcs->late_init(
 					adev->powerplay.pp_handle);
 
-	if (adev->pp_enabled && adev->pm.dpm_enabled) {
-		amdgpu_pm_sysfs_init(adev);
-		amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_COMPLETE_INIT, NULL, NULL);
-	}
-
 	return ret;
 }
 
@@ -176,21 +145,11 @@ static int amdgpu_pp_hw_init(void *handle)
 	int ret = 0;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	if (adev->pp_enabled && adev->firmware.load_type == AMDGPU_FW_LOAD_SMU)
-		amdgpu_ucode_init_bo(adev);
 
 	if (adev->powerplay.ip_funcs->hw_init)
 		ret = adev->powerplay.ip_funcs->hw_init(
 					adev->powerplay.pp_handle);
 
-	if (ret == PP_DPM_DISABLED) {
-		adev->pm.dpm_enabled = false;
-		return 0;
-	}
-
-	if ((amdgpu_dpm != 0) && !amdgpu_sriov_vf(adev))
-		adev->pm.dpm_enabled = true;
-
 	return ret;
 }
 
@@ -199,16 +158,10 @@ static int amdgpu_pp_hw_fini(void *handle)
 	int ret = 0;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	if (adev->pp_enabled && adev->pm.dpm_enabled)
-		amdgpu_pm_sysfs_fini(adev);
-
 	if (adev->powerplay.ip_funcs->hw_fini)
 		ret = adev->powerplay.ip_funcs->hw_fini(
 					adev->powerplay.pp_handle);
 
-	if (adev->pp_enabled && adev->firmware.load_type == AMDGPU_FW_LOAD_SMU)
-		amdgpu_ucode_fini_bo(adev);
-
 	return ret;
 }
 
@@ -220,9 +173,8 @@ static void amdgpu_pp_late_fini(void *handle)
 		adev->powerplay.ip_funcs->late_fini(
 			  adev->powerplay.pp_handle);
 
-
-	if (adev->pp_enabled)
-		amd_powerplay_destroy(adev->powerplay.pp_handle);
+	if (adev->powerplay.cgs_device)
+		amdgpu_cgs_destroy_device(adev->powerplay.cgs_device);
 }
 
 static int amdgpu_pp_suspend(void *handle)

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

@@ -411,13 +411,6 @@ static int psp_hw_init(void *handle)
 		return 0;
 
 	mutex_lock(&adev->firmware.mutex);
-	/*
-	 * This sequence is just used on hw_init only once, no need on
-	 * resume.
-	 */
-	ret = amdgpu_ucode_init_bo(adev);
-	if (ret)
-		goto failed;
 
 	ret = psp_load_fw(adev);
 	if (ret) {
@@ -442,8 +435,6 @@ static int psp_hw_fini(void *handle)
 	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
 		return 0;
 
-	amdgpu_ucode_fini_bo(adev);
-
 	psp_ring_destroy(psp, PSP_RING_TYPE__KM);
 
 	amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf);

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

@@ -121,7 +121,7 @@ static enum amdgpu_ring_type amdgpu_hw_ip_to_ring_type(int hw_ip)
 
 static int amdgpu_lru_map(struct amdgpu_device *adev,
 			  struct amdgpu_queue_mapper *mapper,
-			  int user_ring,
+			  int user_ring, bool lru_pipe_order,
 			  struct amdgpu_ring **out_ring)
 {
 	int r, i, j;
@@ -139,7 +139,7 @@ static int amdgpu_lru_map(struct amdgpu_device *adev,
 	}
 
 	r = amdgpu_ring_lru_get(adev, ring_type, ring_blacklist,
-				j, out_ring);
+				j, lru_pipe_order, out_ring);
 	if (r)
 		return r;
 
@@ -284,8 +284,10 @@ int amdgpu_queue_mgr_map(struct amdgpu_device *adev,
 		r = amdgpu_identity_map(adev, mapper, ring, out_ring);
 		break;
 	case AMDGPU_HW_IP_DMA:
+		r = amdgpu_lru_map(adev, mapper, ring, false, out_ring);
+		break;
 	case AMDGPU_HW_IP_COMPUTE:
-		r = amdgpu_lru_map(adev, mapper, ring, out_ring);
+		r = amdgpu_lru_map(adev, mapper, ring, true, out_ring);
 		break;
 	default:
 		*out_ring = NULL;

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

@@ -315,14 +315,16 @@ static bool amdgpu_ring_is_blacklisted(struct amdgpu_ring *ring,
  * @type: amdgpu_ring_type enum
  * @blacklist: blacklisted ring ids array
  * @num_blacklist: number of entries in @blacklist
+ * @lru_pipe_order: find a ring from the least recently used pipe
  * @ring: output ring
  *
  * Retrieve the amdgpu_ring structure for the least recently used ring of
  * a specific IP block (all asics).
  * Returns 0 on success, error on failure.
  */
-int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist,
-			int num_blacklist, struct amdgpu_ring **ring)
+int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type,
+			int *blacklist,	int num_blacklist,
+			bool lru_pipe_order, struct amdgpu_ring **ring)
 {
 	struct amdgpu_ring *entry;
 
@@ -337,10 +339,23 @@ int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist,
 		if (amdgpu_ring_is_blacklisted(entry, blacklist, num_blacklist))
 			continue;
 
-		*ring = entry;
-		amdgpu_ring_lru_touch_locked(adev, *ring);
-		break;
+		if (!*ring) {
+			*ring = entry;
+
+			/* We are done for ring LRU */
+			if (!lru_pipe_order)
+				break;
+		}
+
+		/* Move all rings on the same pipe to the end of the list */
+		if (entry->pipe == (*ring)->pipe)
+			amdgpu_ring_lru_touch_locked(adev, entry);
 	}
+
+	/* Move the ring we found to the end of the list */
+	if (*ring)
+		amdgpu_ring_lru_touch_locked(adev, *ring);
+
 	spin_unlock(&adev->ring_lru_list_lock);
 
 	if (!*ring) {

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

@@ -201,8 +201,9 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 		     unsigned ring_size, struct amdgpu_irq_src *irq_src,
 		     unsigned irq_type);
 void amdgpu_ring_fini(struct amdgpu_ring *ring);
-int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist,
-			int num_blacklist, struct amdgpu_ring **ring);
+int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type,
+			int *blacklist, int num_blacklist,
+			bool lru_pipe_order, struct amdgpu_ring **ring);
 void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring);
 static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring)
 {

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

@@ -25,30 +25,21 @@
 #include "amdgpu_vf_error.h"
 #include "mxgpu_ai.h"
 
-#define AMDGPU_VF_ERROR_ENTRY_SIZE    16 
-
-/* struct error_entry - amdgpu VF error information. */
-struct amdgpu_vf_error_buffer {
-	int read_count;
-	int write_count;
-	uint16_t code[AMDGPU_VF_ERROR_ENTRY_SIZE];
-	uint16_t flags[AMDGPU_VF_ERROR_ENTRY_SIZE];
-	uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE];
-};
-
-struct amdgpu_vf_error_buffer admgpu_vf_errors;
-
-
-void amdgpu_vf_error_put(uint16_t sub_error_code, uint16_t error_flags, uint64_t error_data)
+void amdgpu_vf_error_put(struct amdgpu_device *adev,
+			 uint16_t sub_error_code,
+			 uint16_t error_flags,
+			 uint64_t error_data)
 {
 	int index;
 	uint16_t error_code = AMDGIM_ERROR_CODE(AMDGIM_ERROR_CATEGORY_VF, sub_error_code);
 
-	index = admgpu_vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
-	admgpu_vf_errors.code [index] = error_code;
-	admgpu_vf_errors.flags [index] = error_flags;
-	admgpu_vf_errors.data [index] = error_data;
-	admgpu_vf_errors.write_count ++;
+	mutex_lock(&adev->virt.vf_errors.lock);
+	index = adev->virt.vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
+	adev->virt.vf_errors.code [index] = error_code;
+	adev->virt.vf_errors.flags [index] = error_flags;
+	adev->virt.vf_errors.data [index] = error_data;
+	adev->virt.vf_errors.write_count ++;
+	mutex_unlock(&adev->virt.vf_errors.lock);
 }
 
 
@@ -58,7 +49,8 @@ void amdgpu_vf_error_trans_all(struct amdgpu_device *adev)
 	u32 data1, data2, data3;
 	int index;
 
-	if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) || (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) {
+	if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) ||
+	    (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) {
 		return;
 	}
 /*
@@ -68,18 +60,22 @@ void amdgpu_vf_error_trans_all(struct amdgpu_device *adev)
 		return;
 	}
 */
+
+	mutex_lock(&adev->virt.vf_errors.lock);
 	/* The errors are overlay of array, correct read_count as full. */
-	if (admgpu_vf_errors.write_count - admgpu_vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) {
-		admgpu_vf_errors.read_count = admgpu_vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE;
+	if (adev->virt.vf_errors.write_count - adev->virt.vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) {
+		adev->virt.vf_errors.read_count = adev->virt.vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE;
 	}
 
-	while (admgpu_vf_errors.read_count < admgpu_vf_errors.write_count) {
-		index =admgpu_vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
-		data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX (admgpu_vf_errors.code[index], admgpu_vf_errors.flags[index]);
-		data2 = admgpu_vf_errors.data[index] & 0xFFFFFFFF;
-		data3 = (admgpu_vf_errors.data[index] >> 32) & 0xFFFFFFFF;
+	while (adev->virt.vf_errors.read_count < adev->virt.vf_errors.write_count) {
+		index =adev->virt.vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
+		data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX(adev->virt.vf_errors.code[index],
+							   adev->virt.vf_errors.flags[index]);
+		data2 = adev->virt.vf_errors.data[index] & 0xFFFFFFFF;
+		data3 = (adev->virt.vf_errors.data[index] >> 32) & 0xFFFFFFFF;
 
 		adev->virt.ops->trans_msg(adev, IDH_LOG_VF_ERROR, data1, data2, data3);
-		admgpu_vf_errors.read_count ++;
+		adev->virt.vf_errors.read_count ++;
 	}
+	mutex_unlock(&adev->virt.vf_errors.lock);
 }