drm/msm: Separate locking of buffer resources from struct_mutex

Buffer object specific resources like pages, domains, sg list
need not be protected with struct_mutex. They can be protected
with a buffer object level lock. This simplifies locking and
makes it easier to avoid potential recursive locking scenarios
for SVM involving mmap_sem and struct_mutex. This also removes
unnecessary serialization when creating buffer objects, and also
between buffer object creation and GPU command submission.

Signed-off-by: Sushmita Susheelendra <ssusheel@codeaurora.org>
[robclark: squash in handling new locking for shrinker]
Signed-off-by: Rob Clark <robdclark@gmail.com>

drivers/gpu/drm/msm/adreno/a5xx_gpu.c

@@ -297,18 +297,18 @@ static struct drm_gem_object *a5xx_ucode_load_bo(struct msm_gpu *gpu,
 	struct drm_gem_object *bo;
 	void *ptr;
 
-	bo = msm_gem_new(drm, fw->size - 4, MSM_BO_UNCACHED);
+	bo = msm_gem_new_locked(drm, fw->size - 4, MSM_BO_UNCACHED);
 	if (IS_ERR(bo))
 		return bo;
 
-	ptr = msm_gem_get_vaddr_locked(bo);
+	ptr = msm_gem_get_vaddr(bo);
 	if (!ptr) {
 		drm_gem_object_unreference(bo);
 		return ERR_PTR(-ENOMEM);
 	}
 
 	if (iova) {
-		int ret = msm_gem_get_iova_locked(bo, gpu->aspace, iova);
+		int ret = msm_gem_get_iova(bo, gpu->aspace, iova);
 
 		if (ret) {
 			drm_gem_object_unreference(bo);
@@ -318,7 +318,7 @@ static struct drm_gem_object *a5xx_ucode_load_bo(struct msm_gpu *gpu,
 
 	memcpy(ptr, &fw->data[4], fw->size - 4);
 
-	msm_gem_put_vaddr_locked(bo);
+	msm_gem_put_vaddr(bo);
 	return bo;
 }
 

drivers/gpu/drm/msm/adreno/a5xx_power.c

@@ -294,15 +294,15 @@ void a5xx_gpmu_ucode_init(struct msm_gpu *gpu)
 	 */
 	bosize = (cmds_size + (cmds_size / TYPE4_MAX_PAYLOAD) + 1) << 2;
 
-	a5xx_gpu->gpmu_bo = msm_gem_new(drm, bosize, MSM_BO_UNCACHED);
+	a5xx_gpu->gpmu_bo = msm_gem_new_locked(drm, bosize, MSM_BO_UNCACHED);
 	if (IS_ERR(a5xx_gpu->gpmu_bo))
 		goto err;
 
-	if (msm_gem_get_iova_locked(a5xx_gpu->gpmu_bo, gpu->aspace,
+	if (msm_gem_get_iova(a5xx_gpu->gpmu_bo, gpu->aspace,
 			&a5xx_gpu->gpmu_iova))
 		goto err;
 
-	ptr = msm_gem_get_vaddr_locked(a5xx_gpu->gpmu_bo);
+	ptr = msm_gem_get_vaddr(a5xx_gpu->gpmu_bo);
 	if (!ptr)
 		goto err;
 
@@ -321,7 +321,7 @@ void a5xx_gpmu_ucode_init(struct msm_gpu *gpu)
 		cmds_size -= _size;
 	}
 
-	msm_gem_put_vaddr_locked(a5xx_gpu->gpmu_bo);
+	msm_gem_put_vaddr(a5xx_gpu->gpmu_bo);
 	a5xx_gpu->gpmu_dwords = dwords;
 
 	goto out;

drivers/gpu/drm/msm/adreno/adreno_gpu.c

@@ -64,7 +64,7 @@ int adreno_hw_init(struct msm_gpu *gpu)
 
 	DBG("%s", gpu->name);
 
-	ret = msm_gem_get_iova_locked(gpu->rb->bo, gpu->aspace, &gpu->rb_iova);
+	ret = msm_gem_get_iova(gpu->rb->bo, gpu->aspace, &gpu->rb_iova);
 	if (ret) {
 		gpu->rb_iova = 0;
 		dev_err(gpu->dev->dev, "could not map ringbuffer: %d\n", ret);
@@ -397,10 +397,8 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 			return ret;
 	}
 
-	mutex_lock(&drm->struct_mutex);
 	adreno_gpu->memptrs_bo = msm_gem_new(drm, sizeof(*adreno_gpu->memptrs),
 			MSM_BO_UNCACHED);
-	mutex_unlock(&drm->struct_mutex);
 	if (IS_ERR(adreno_gpu->memptrs_bo)) {
 		ret = PTR_ERR(adreno_gpu->memptrs_bo);
 		adreno_gpu->memptrs_bo = NULL;

drivers/gpu/drm/msm/dsi/dsi_host.c

@@ -982,18 +982,16 @@ static int dsi_tx_buf_alloc(struct msm_dsi_host *msm_host, int size)
 	uint64_t iova;
 
 	if (cfg_hnd->major == MSM_DSI_VER_MAJOR_6G) {
-		mutex_lock(&dev->struct_mutex);
 		msm_host->tx_gem_obj = msm_gem_new(dev, size, MSM_BO_UNCACHED);
 		if (IS_ERR(msm_host->tx_gem_obj)) {
 			ret = PTR_ERR(msm_host->tx_gem_obj);
 			pr_err("%s: failed to allocate gem, %d\n",
 				__func__, ret);
 			msm_host->tx_gem_obj = NULL;
-			mutex_unlock(&dev->struct_mutex);
 			return ret;
 		}
 
-		ret = msm_gem_get_iova_locked(msm_host->tx_gem_obj,
+		ret = msm_gem_get_iova(msm_host->tx_gem_obj,
 				priv->kms->aspace, &iova);
 		mutex_unlock(&dev->struct_mutex);
 		if (ret) {

drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c

@@ -374,7 +374,7 @@ static void update_cursor(struct drm_crtc *crtc)
 		if (next_bo) {
 			/* take a obj ref + iova ref when we start scanning out: */
 			drm_gem_object_reference(next_bo);
-			msm_gem_get_iova_locked(next_bo, kms->aspace, &iova);
+			msm_gem_get_iova(next_bo, kms->aspace, &iova);
 
 			/* enable cursor: */
 			mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_SIZE(dma),

drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c

@@ -528,9 +528,7 @@ struct msm_kms *mdp4_kms_init(struct drm_device *dev)
 		goto fail;
 	}
 
-	mutex_lock(&dev->struct_mutex);
 	mdp4_kms->blank_cursor_bo = msm_gem_new(dev, SZ_16K, MSM_BO_WC);
-	mutex_unlock(&dev->struct_mutex);
 	if (IS_ERR(mdp4_kms->blank_cursor_bo)) {
 		ret = PTR_ERR(mdp4_kms->blank_cursor_bo);
 		dev_err(dev->dev, "could not allocate blank-cursor bo: %d\n", ret);

drivers/gpu/drm/msm/msm_drv.c

@@ -336,6 +336,7 @@ static int msm_init_vram(struct drm_device *dev)
 		priv->vram.size = size;
 
 		drm_mm_init(&priv->vram.mm, 0, (size >> PAGE_SHIFT) - 1);
+		spin_lock_init(&priv->vram.lock);
 
 		attrs |= DMA_ATTR_NO_KERNEL_MAPPING;
 		attrs |= DMA_ATTR_WRITE_COMBINE;

drivers/gpu/drm/msm/msm_drv.h

@@ -149,6 +149,7 @@ struct msm_drm_private {
 		 * and position mm_node->start is in # of pages:
 		 */
 		struct drm_mm mm;
+		spinlock_t lock; /* Protects drm_mm node allocation/removal */
 	} vram;
 
 	struct notifier_block vmap_notifier;
@@ -198,8 +199,6 @@ int msm_gem_mmap_obj(struct drm_gem_object *obj,
 int msm_gem_mmap(struct file *filp, struct vm_area_struct *vma);
 int msm_gem_fault(struct vm_fault *vmf);
 uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj);
-int msm_gem_get_iova_locked(struct drm_gem_object *obj,
-		struct msm_gem_address_space *aspace, uint64_t *iova);
 int msm_gem_get_iova(struct drm_gem_object *obj,
 		struct msm_gem_address_space *aspace, uint64_t *iova);
 uint64_t msm_gem_iova(struct drm_gem_object *obj,
@@ -221,13 +220,9 @@ struct drm_gem_object *msm_gem_prime_import_sg_table(struct drm_device *dev,
 		struct dma_buf_attachment *attach, struct sg_table *sg);
 int msm_gem_prime_pin(struct drm_gem_object *obj);
 void msm_gem_prime_unpin(struct drm_gem_object *obj);
-void *msm_gem_get_vaddr_locked(struct drm_gem_object *obj);
 void *msm_gem_get_vaddr(struct drm_gem_object *obj);
-void msm_gem_put_vaddr_locked(struct drm_gem_object *obj);
 void msm_gem_put_vaddr(struct drm_gem_object *obj);
 int msm_gem_madvise(struct drm_gem_object *obj, unsigned madv);
-void msm_gem_purge(struct drm_gem_object *obj);
-void msm_gem_vunmap(struct drm_gem_object *obj);
 int msm_gem_sync_object(struct drm_gem_object *obj,
 		struct msm_fence_context *fctx, bool exclusive);
 void msm_gem_move_to_active(struct drm_gem_object *obj,
@@ -240,6 +235,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file,
 		uint32_t size, uint32_t flags, uint32_t *handle);
 struct drm_gem_object *msm_gem_new(struct drm_device *dev,
 		uint32_t size, uint32_t flags);
+struct drm_gem_object *msm_gem_new_locked(struct drm_device *dev,
+		uint32_t size, uint32_t flags);
 struct drm_gem_object *msm_gem_import(struct drm_device *dev,
 		struct dma_buf *dmabuf, struct sg_table *sgt);
 

drivers/gpu/drm/msm/msm_fbdev.c

@@ -97,10 +97,8 @@ static int msm_fbdev_create(struct drm_fb_helper *helper,
 	/* allocate backing bo */
 	size = mode_cmd.pitches[0] * mode_cmd.height;
 	DBG("allocating %d bytes for fb %d", size, dev->primary->index);
-	mutex_lock(&dev->struct_mutex);
 	fbdev->bo = msm_gem_new(dev, size, MSM_BO_SCANOUT |
 			MSM_BO_WC | MSM_BO_STOLEN);
-	mutex_unlock(&dev->struct_mutex);
 	if (IS_ERR(fbdev->bo)) {
 		ret = PTR_ERR(fbdev->bo);
 		fbdev->bo = NULL;
@@ -126,7 +124,7 @@ static int msm_fbdev_create(struct drm_fb_helper *helper,
 	 * in panic (ie. lock-safe, etc) we could avoid pinning the
 	 * buffer now:
 	 */
-	ret = msm_gem_get_iova_locked(fbdev->bo, priv->kms->aspace, &paddr);
+	ret = msm_gem_get_iova(fbdev->bo, priv->kms->aspace, &paddr);
 	if (ret) {
 		dev_err(dev->dev, "failed to get buffer obj iova: %d\n", ret);
 		goto fail_unlock;
@@ -155,7 +153,7 @@ static int msm_fbdev_create(struct drm_fb_helper *helper,
 
 	dev->mode_config.fb_base = paddr;
 
-	fbi->screen_base = msm_gem_get_vaddr_locked(fbdev->bo);
+	fbi->screen_base = msm_gem_get_vaddr(fbdev->bo);
 	if (IS_ERR(fbi->screen_base)) {
 		ret = PTR_ERR(fbi->screen_base);
 		goto fail_unlock;