Skip to contentSkip to conversation

drm/msm: Separate locking of buffer resources from struct_mutex

PATCH In review Sushmita Susheelendra 5 participants 13 messages Jun 13, 2017 – Jun 16, 2017 7 messages not in the archive

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 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.

Change-Id I4fba9f8c38a6cd13f80f660639d1e74d4336e3fb
---
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c    |   2 -
 drivers/gpu/drm/msm/adreno/a5xx_power.c  |   2 -
 drivers/gpu/drm/msm/adreno/adreno_gpu.c  |   2 -
 drivers/gpu/drm/msm/dsi/dsi_host.c       |   5 +-
 drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c |   2 +-
 drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c  |   2 -
 drivers/gpu/drm/msm/msm_drv.c            |   1 +
 drivers/gpu/drm/msm/msm_drv.h            |   7 +-
 drivers/gpu/drm/msm/msm_fbdev.c          |   6 +-
 drivers/gpu/drm/msm/msm_gem.c            | 190 +++++++++++++++----------------
 drivers/gpu/drm/msm/msm_gem.h            |   2 +
 drivers/gpu/drm/msm/msm_gem_submit.c     |   6 +-
 drivers/gpu/drm/msm/msm_gem_vma.c        |  10 +-
 drivers/gpu/drm/msm/msm_gpu.c            |   4 +-
 drivers/gpu/drm/msm/msm_rd.c             |   4 +-
 drivers/gpu/drm/msm/msm_ringbuffer.c     |   2 +-
 16 files changed, 120 insertions(+), 127 deletions(-)
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +0 -2
index 31a9bce..7893de1 100644
@@ -235,9 +235,7 @@ static struct drm_gem_object *a5xx_ucode_load_bo(struct msm_gpu *gpu,
 	struct drm_gem_object *bo;
 	void *ptr;
 
-	mutex_lock(&drm->struct_mutex);
 	bo = msm_gem_new(drm, fw->size - 4, MSM_BO_UNCACHED);
-	mutex_unlock(&drm->struct_mutex);
 
 	if (IS_ERR(bo))
 		return bo;
drivers/gpu/drm/msm/adreno/a5xx_power.c +0 -2
index 72d52c7..eb88f44 100644
@@ -294,9 +294,7 @@ void a5xx_gpmu_ucode_init(struct msm_gpu *gpu)
 	 */
 	bosize = (cmds_size + (cmds_size / TYPE4_MAX_PAYLOAD) + 1) << 2;
 
-	mutex_lock(&drm->struct_mutex);
 	a5xx_gpu->gpmu_bo = msm_gem_new(drm, bosize, MSM_BO_UNCACHED);
-	mutex_unlock(&drm->struct_mutex);
 
 	if (IS_ERR(a5xx_gpu->gpmu_bo))
 		goto err;
drivers/gpu/drm/msm/adreno/adreno_gpu.c +0 -2
index 5b63fc6..1162c15 100644
@@ -392,10 +392,8 @@ 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 +1 -4
index 4f79b10..6a1b0da 100644
@@ -980,19 +980,16 @@ 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, 0, &iova);
-		mutex_unlock(&dev->struct_mutex);
+		ret = msm_gem_get_iova(msm_host->tx_gem_obj, 0, &iova);
 		if (ret) {
 			pr_err("%s: failed to get iova, %d\n", __func__, ret);
 			return ret;
drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +1 -1
index f29194a..15478f8 100644
@@ -372,7 +372,7 @@ 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, mdp4_kms->id, &iova);
+			msm_gem_get_iova(next_bo, mdp4_kms->id, &iova);
 
 			/* enable cursor: */
 			mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_SIZE(dma),
drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c +0 -2
index 6295204..3a464b3 100644
@@ -561,9 +561,7 @@ 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 +1 -0
index 87b5695..bb1f3ee 100644
@@ -349,6 +349,7 @@ 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 +3 -4
index 28b6f9b..567737d 100644
@@ -157,6 +157,7 @@ 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;
@@ -209,8 +210,6 @@ 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, int id,
-		uint64_t *iova);
 int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova);
 uint64_t msm_gem_iova(struct drm_gem_object *obj, int id);
 struct page **msm_gem_get_pages(struct drm_gem_object *obj);
@@ -228,9 +227,7 @@ 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);
@@ -247,6 +244,8 @@ 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 +2 -4
index 951e40f..785925e9 100644
@@ -95,10 +95,8 @@ 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;
@@ -124,7 +122,7 @@ 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, 0, &paddr);
+	ret = msm_gem_get_iova(fbdev->bo, 0, &paddr);
 	if (ret) {
 		dev_err(dev->dev, "failed to get buffer obj iova: %d\n", ret);
 		goto fail_unlock;
@@ -153,7 +151,7 @@ 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;
drivers/gpu/drm/msm/msm_gem.c +94 -96

402 lines not on this page · raw message

drivers/gpu/drm/msm/msm_gem.h +2 -0
index 1b4cf20..b1bfc10 100644
@@ -31,6 +31,7 @@ struct msm_gem_address_space {
 	 * and position mm_node->start is in # of pages:
 	 */
 	struct drm_mm mm;
+	spinlock_t lock; /* Protects drm_mm node allocation/removal */
 	struct msm_mmu *mmu;
 	struct kref kref;
 };
@@ -87,6 +88,7 @@ struct msm_gem_object {
 	 * an IOMMU.  Also used for stolen/splashscreen buffer.
 	 */
 	struct drm_mm_node *vram_node;
+	struct mutex lock; /* Protects resources associated with bo */
 };
 #define to_msm_bo(x) container_of(x, struct msm_gem_object, base)
 
drivers/gpu/drm/msm/msm_gem_submit.c +3 -3
index 1c545eb..8420551 100644
@@ -245,7 +245,7 @@ static int submit_pin_objects(struct msm_gem_submit *submit)
 		uint64_t iova;
 
 		/* if locking succeeded, pin bo: */
-		ret = msm_gem_get_iova_locked(&msm_obj->base,
+		ret = msm_gem_get_iova(&msm_obj->base,
 				submit->gpu->id, &iova);
 
 		if (ret)
@@ -301,7 +301,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
 	/* For now, just map the entire thing.  Eventually we probably
 	 * to do it page-by-page, w/ kmap() if not vmap()d..
 	 */
-	ptr = msm_gem_get_vaddr_locked(&obj->base);
+	ptr = msm_gem_get_vaddr(&obj->base);
 
 	if (IS_ERR(ptr)) {
 		ret = PTR_ERR(ptr);
@@ -359,7 +359,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
 	}
 
 out:
-	msm_gem_put_vaddr_locked(&obj->base);
+	msm_gem_put_vaddr(&obj->base);
 
 	return ret;
 }
drivers/gpu/drm/msm/msm_gem_vma.c +9 -1
index f285d7e..c36321bc 100644
@@ -50,7 +50,9 @@ void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
 		aspace->mmu->funcs->unmap(aspace->mmu, vma->iova, sgt, size);
 	}
 
+	spin_lock(&aspace->lock);
 	drm_mm_remove_node(&vma->node);
+	spin_unlock(&aspace->lock);
 
 	vma->iova = 0;
 
@@ -63,10 +65,15 @@ void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
 {
 	int ret;
 
-	if (WARN_ON(drm_mm_node_allocated(&vma->node)))
+	spin_lock(&aspace->lock);
+	if (WARN_ON(drm_mm_node_allocated(&vma->node))) {
+		spin_unlock(&aspace->lock);
 		return 0;
+	}
 
 	ret = drm_mm_insert_node(&aspace->mm, &vma->node, npages);
+	spin_unlock(&aspace->lock);
+
 	if (ret)
 		return ret;
 
@@ -94,6 +101,7 @@ struct msm_gem_address_space *
 	if (!aspace)
 		return ERR_PTR(-ENOMEM);
 
+	spin_lock_init(&aspace->lock);
 	aspace->name = name;
 	aspace->mmu = msm_iommu_new(dev, domain);
 
drivers/gpu/drm/msm/msm_gpu.c +1 -3
index 97b9c38..d6e5cb2 100644
@@ -495,7 +495,7 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
 		/* submit takes a reference to the bo and iova until retired: */
 		drm_gem_object_reference(&msm_obj->base);
-		msm_gem_get_iova_locked(&msm_obj->base,
+		msm_gem_get_iova(&msm_obj->base,
 				submit->gpu->id, &iova);
 
 		if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
@@ -662,9 +662,7 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 
 
 	/* Create ringbuffer: */
-	mutex_lock(&drm->struct_mutex);
 	gpu->rb = msm_ringbuffer_new(gpu, ringsz);
-	mutex_unlock(&drm->struct_mutex);
 	if (IS_ERR(gpu->rb)) {
 		ret = PTR_ERR(gpu->rb);
 		gpu->rb = NULL;
drivers/gpu/drm/msm/msm_rd.c +2 -2
index 0e81faa..0366b80 100644
@@ -268,7 +268,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
 	struct msm_gem_object *obj = submit->bos[idx].obj;
 	const char *buf;
 
-	buf = msm_gem_get_vaddr_locked(&obj->base);
+	buf = msm_gem_get_vaddr(&obj->base);
 	if (IS_ERR(buf))
 		return;
 
@@ -283,7 +283,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
 			(uint32_t[3]){ iova, size, iova >> 32 }, 12);
 	rd_write_section(rd, RD_BUFFER_CONTENTS, buf, size);
 
-	msm_gem_put_vaddr_locked(&obj->base);
+	msm_gem_put_vaddr(&obj->base);
 }
 
 /* called under struct_mutex */
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -1
index 67b34e0..791bca3 100644
@@ -40,7 +40,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int size)
 		goto fail;
 	}
 
-	ring->start = msm_gem_get_vaddr_locked(ring->bo);
+	ring->start = msm_gem_get_vaddr(ring->bo);
 	if (IS_ERR(ring->start)) {
 		ret = PTR_ERR(ring->start);
 		goto fail;
Signature · 1 line
1.9.1

A message that is not in the archive (sent off-list or not yet received): 1497394374-19982-1-git-send-email-ssusheel-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org

#2 lore raw
Quoted 7 lines · Sushmita Susheelendra

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.

oh, cool.. this might be a bit conflicty w/ "drm/msm: fix locking inconsistency for gpu->hw_init()" and possibly the patchset to support arbitrary # of vma's per gem bo. But it is definitely the direction we want to go. It should also help us move away from the mutex_trylock_recursive() hack in the shrinker code.

I'll have a go at rebasing it, and in the process taking a closer look, tomorrow since this is something that has been on the TODO list for quite a while now and is pretty badly needed.

BR,
-R

Quoted 781 lines
Change-Id I4fba9f8c38a6cd13f80f660639d1e74d4336e3fb
---
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c    |   2 -
 drivers/gpu/drm/msm/adreno/a5xx_power.c  |   2 -
 drivers/gpu/drm/msm/adreno/adreno_gpu.c  |   2 -
 drivers/gpu/drm/msm/dsi/dsi_host.c       |   5 +-
 drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c |   2 +-
 drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c  |   2 -
 drivers/gpu/drm/msm/msm_drv.c            |   1 +
 drivers/gpu/drm/msm/msm_drv.h            |   7 +-
 drivers/gpu/drm/msm/msm_fbdev.c          |   6 +-
 drivers/gpu/drm/msm/msm_gem.c            | 190 +++++++++++++++----------------
 drivers/gpu/drm/msm/msm_gem.h            |   2 +
 drivers/gpu/drm/msm/msm_gem_submit.c     |   6 +-
 drivers/gpu/drm/msm/msm_gem_vma.c        |  10 +-
 drivers/gpu/drm/msm/msm_gpu.c            |   4 +-
 drivers/gpu/drm/msm/msm_rd.c             |   4 +-
 drivers/gpu/drm/msm/msm_ringbuffer.c     |   2 +-
 16 files changed, 120 insertions(+), 127 deletions(-)
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +0 -2
index 31a9bce..7893de1 100644
@@ -235,9 +235,7 @@ static struct drm_gem_object *a5xx_ucode_load_bo(struct msm_gpu *gpu,
        struct drm_gem_object *bo;
        void *ptr;
 
-       mutex_lock(&drm->struct_mutex);
        bo = msm_gem_new(drm, fw->size - 4, MSM_BO_UNCACHED);
-       mutex_unlock(&drm->struct_mutex);
 
        if (IS_ERR(bo))
                return bo;
drivers/gpu/drm/msm/adreno/a5xx_power.c +0 -2
index 72d52c7..eb88f44 100644
@@ -294,9 +294,7 @@ void a5xx_gpmu_ucode_init(struct msm_gpu *gpu)
         */
        bosize = (cmds_size + (cmds_size / TYPE4_MAX_PAYLOAD) + 1) << 2;
 
-       mutex_lock(&drm->struct_mutex);
        a5xx_gpu->gpmu_bo = msm_gem_new(drm, bosize, MSM_BO_UNCACHED);
-       mutex_unlock(&drm->struct_mutex);
 
        if (IS_ERR(a5xx_gpu->gpmu_bo))
                goto err;
drivers/gpu/drm/msm/adreno/adreno_gpu.c +0 -2
index 5b63fc6..1162c15 100644
@@ -392,10 +392,8 @@ 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 +1 -4
index 4f79b10..6a1b0da 100644
@@ -980,19 +980,16 @@ 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, 0, &iova);
-               mutex_unlock(&dev->struct_mutex);
+               ret = msm_gem_get_iova(msm_host->tx_gem_obj, 0, &iova);
                if (ret) {
                        pr_err("%s: failed to get iova, %d\n", __func__, ret);
                        return ret;
drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +1 -1
index f29194a..15478f8 100644
@@ -372,7 +372,7 @@ 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, mdp4_kms->id, &iova);
+                       msm_gem_get_iova(next_bo, mdp4_kms->id, &iova);
 
                        /* enable cursor: */
                        mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_SIZE(dma),
drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c +0 -2
index 6295204..3a464b3 100644
@@ -561,9 +561,7 @@ 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 +1 -0
index 87b5695..bb1f3ee 100644
@@ -349,6 +349,7 @@ 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 +3 -4
index 28b6f9b..567737d 100644
@@ -157,6 +157,7 @@ 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;
@@ -209,8 +210,6 @@ 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, int id,
-               uint64_t *iova);
 int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova);
 uint64_t msm_gem_iova(struct drm_gem_object *obj, int id);
 struct page **msm_gem_get_pages(struct drm_gem_object *obj);
@@ -228,9 +227,7 @@ 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);
@@ -247,6 +244,8 @@ 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 +2 -4
index 951e40f..785925e9 100644
@@ -95,10 +95,8 @@ 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;
@@ -124,7 +122,7 @@ 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, 0, &paddr);
+       ret = msm_gem_get_iova(fbdev->bo, 0, &paddr);
        if (ret) {
                dev_err(dev->dev, "failed to get buffer obj iova: %d\n", ret);
                goto fail_unlock;
@@ -153,7 +151,7 @@ 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;
drivers/gpu/drm/msm/msm_gem.c +94 -96
index 68e509b..1e803fb 100644
@@ -41,8 +41,7 @@ static bool use_pages(struct drm_gem_object *obj)
 }
 
 /* allocate pages from VRAM carveout, used when no IOMMU: */
-static struct page **get_pages_vram(struct drm_gem_object *obj,
-               int npages)
+static struct page **get_pages_vram(struct drm_gem_object *obj, int npages)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
        struct msm_drm_private *priv = obj->dev->dev_private;
@@ -54,7 +53,9 @@ static struct page **get_pages_vram(struct drm_gem_object *obj,
        if (!p)
                return ERR_PTR(-ENOMEM);
 
+       spin_lock(&priv->vram.lock);
        ret = drm_mm_insert_node(&priv->vram.mm, msm_obj->vram_node, npages);
+       spin_unlock(&priv->vram.lock);
        if (ret) {
                drm_free_large(p);
                return ERR_PTR(ret);
@@ -69,7 +70,6 @@ static struct page **get_pages_vram(struct drm_gem_object *obj,
        return p;
 }
 
-/* called with dev->struct_mutex held */
 static struct page **get_pages(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -109,6 +109,18 @@ static struct page **get_pages(struct drm_gem_object *obj)
        return msm_obj->pages;
 }
 
+static void put_pages_vram(struct drm_gem_object *obj)
+{
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+       struct msm_drm_private *priv = obj->dev->dev_private;
+
+       spin_lock(&priv->vram.lock);
+       drm_mm_remove_node(msm_obj->vram_node);
+       spin_unlock(&priv->vram.lock);
+
+       drm_free_large(msm_obj->pages);
+}
+
 static void put_pages(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -125,10 +137,8 @@ static void put_pages(struct drm_gem_object *obj)
 
                if (use_pages(obj))
                        drm_gem_put_pages(obj, msm_obj->pages, true, false);
-               else {
-                       drm_mm_remove_node(msm_obj->vram_node);
-                       drm_free_large(msm_obj->pages);
-               }
+               else
+                       put_pages_vram(obj);
 
                msm_obj->pages = NULL;
        }
@@ -136,11 +146,12 @@ static void put_pages(struct drm_gem_object *obj)
 
 struct page **msm_gem_get_pages(struct drm_gem_object *obj)
 {
-       struct drm_device *dev = obj->dev;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
        struct page **p;
-       mutex_lock(&dev->struct_mutex);
+
+       mutex_lock(&msm_obj->lock);
        p = get_pages(obj);
-       mutex_unlock(&dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
        return p;
 }
 
@@ -195,25 +206,17 @@ int msm_gem_fault(struct vm_fault *vmf)
 {
        struct vm_area_struct *vma = vmf->vma;
        struct drm_gem_object *obj = vma->vm_private_data;
-       struct drm_device *dev = obj->dev;
-       struct msm_drm_private *priv = dev->dev_private;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
        struct page **pages;
        unsigned long pfn;
        pgoff_t pgoff;
        int ret;
 
-       /* This should only happen if userspace tries to pass a mmap'd
-        * but unfaulted gem bo vaddr into submit ioctl, triggering
-        * a page fault while struct_mutex is already held.  This is
-        * not a valid use-case so just bail.
+       /*
+        * vm_ops.open/drm_gem_mmap_obj and close get and put
+        * a reference on obj. So, we dont need to hold one here.
         */
-       if (priv->struct_mutex_task == current)
-               return VM_FAULT_SIGBUS;
-
-       /* Make sure we don't parallel update on a fault, nor move or remove
-        * something from beneath our feet
-        */
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
+       ret = mutex_lock_interruptible(&msm_obj->lock);
        if (ret)
                goto out;
 
@@ -235,7 +238,7 @@ int msm_gem_fault(struct vm_fault *vmf)
        ret = vm_insert_mixed(vma, vmf->address, __pfn_to_pfn_t(pfn, PFN_DEV));
 
 out_unlock:
-       mutex_unlock(&dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
 out:
        switch (ret) {
        case -EAGAIN:
@@ -259,9 +262,10 @@ int msm_gem_fault(struct vm_fault *vmf)
 static uint64_t mmap_offset(struct drm_gem_object *obj)
 {
        struct drm_device *dev = obj->dev;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
        int ret;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
 
        /* Make it mmapable */
        ret = drm_gem_create_mmap_offset(obj);
@@ -277,21 +281,23 @@ static uint64_t mmap_offset(struct drm_gem_object *obj)
 uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj)
 {
        uint64_t offset;
-       mutex_lock(&obj->dev->struct_mutex);
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+
+       mutex_lock(&msm_obj->lock);
        offset = mmap_offset(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
        return offset;
 }
 
+/* Called with msm_obj->lock locked */
 static void
 put_iova(struct drm_gem_object *obj)
 {
-       struct drm_device *dev = obj->dev;
        struct msm_drm_private *priv = obj->dev->dev_private;
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
        int id;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
 
        for (id = 0; id < ARRAY_SIZE(msm_obj->domain); id++) {
                if (!priv->aspace[id])
@@ -301,25 +307,23 @@ uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj)
        }
 }
 
-/* should be called under struct_mutex.. although it can be called
- * from atomic context without struct_mutex to acquire an extra
- * iova ref if you know one is already held.
- *
- * That means when I do eventually need to add support for unpinning
- * the refcnt counter needs to be atomic_t.
- */
-int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
+/* A reference to obj must be held before calling this function. */
+int msm_gem_get_iova(struct drm_gem_object *obj, int id,
                uint64_t *iova)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
        int ret = 0;
 
+       mutex_lock(&msm_obj->lock);
+
        if (!msm_obj->domain[id].iova) {
                struct msm_drm_private *priv = obj->dev->dev_private;
                struct page **pages = get_pages(obj);
 
-               if (IS_ERR(pages))
+               if (IS_ERR(pages)) {
+                       mutex_unlock(&msm_obj->lock);
                        return PTR_ERR(pages);
+               }
 
                if (iommu_present(&platform_bus_type)) {
                        ret = msm_gem_map_vma(priv->aspace[id], &msm_obj->domain[id],
@@ -332,26 +336,7 @@ int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
        if (!ret)
                *iova = msm_obj->domain[id].iova;
 
-       return ret;
-}
-
-/* get iova, taking a reference.  Should have a matching put */
-int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova)
-{
-       struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       int ret;
-
-       /* this is safe right now because we don't unmap until the
-        * bo is deleted:
-        */
-       if (msm_obj->domain[id].iova) {
-               *iova = msm_obj->domain[id].iova;
-               return 0;
-       }
-
-       mutex_lock(&obj->dev->struct_mutex);
-       ret = msm_gem_get_iova_locked(obj, id, iova);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
        return ret;
 }
 
@@ -405,45 +390,37 @@ int msm_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev,
        return ret;
 }
 
-void *msm_gem_get_vaddr_locked(struct drm_gem_object *obj)
+void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+
+       mutex_lock(&msm_obj->lock);
        if (!msm_obj->vaddr) {
                struct page **pages = get_pages(obj);
-               if (IS_ERR(pages))
+               if (IS_ERR(pages)) {
+                       mutex_unlock(&msm_obj->lock);
                        return ERR_CAST(pages);
+               }
                msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
                                VM_MAP, pgprot_writecombine(PAGE_KERNEL));
-               if (msm_obj->vaddr == NULL)
+               if (msm_obj->vaddr == NULL) {
+                       mutex_unlock(&msm_obj->lock);
                        return ERR_PTR(-ENOMEM);
+               }
        }
        msm_obj->vmap_count++;
+       mutex_unlock(&msm_obj->lock);
        return msm_obj->vaddr;
 }
 
-void *msm_gem_get_vaddr(struct drm_gem_object *obj)
-{
-       void *ret;
-       mutex_lock(&obj->dev->struct_mutex);
-       ret = msm_gem_get_vaddr_locked(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
-       return ret;
-}
-
-void msm_gem_put_vaddr_locked(struct drm_gem_object *obj)
+void msm_gem_put_vaddr(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+
+       mutex_lock(&msm_obj->lock);
        WARN_ON(msm_obj->vmap_count < 1);
        msm_obj->vmap_count--;
-}
-
-void msm_gem_put_vaddr(struct drm_gem_object *obj)
-{
-       mutex_lock(&obj->dev->struct_mutex);
-       msm_gem_put_vaddr_locked(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
 }
 
 /* Update madvise status, returns true if not purged, else
@@ -697,6 +674,8 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
        list_del(&msm_obj->mm_list);
 
+       mutex_lock(&msm_obj->lock);
+
        put_iova(obj);
 
        if (obj->import_attach) {
@@ -720,6 +699,7 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
        drm_gem_object_release(obj);
 
+       mutex_unlock(&msm_obj->lock);
        kfree(msm_obj);
 }
 
@@ -730,14 +710,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file,
        struct drm_gem_object *obj;
        int ret;
 
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
-       if (ret)
-               return ret;
-
        obj = msm_gem_new(dev, size, flags);
 
-       mutex_unlock(&dev->struct_mutex);
-
        if (IS_ERR(obj))
                return PTR_ERR(obj);
 
@@ -752,7 +726,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file,
 static int msm_gem_new_impl(struct drm_device *dev,
                uint32_t size, uint32_t flags,
                struct reservation_object *resv,
-               struct drm_gem_object **obj)
+               struct drm_gem_object **obj,
+               bool struct_mutex_locked)
 {
        struct msm_drm_private *priv = dev->dev_private;
        struct msm_gem_object *msm_obj;
@@ -781,6 +756,8 @@ static int msm_gem_new_impl(struct drm_device *dev,
        if (!msm_obj)
                return -ENOMEM;
 
+       mutex_init(&msm_obj->lock);
+
        if (use_vram)
                msm_obj->vram_node = &msm_obj->domain[0].node;
 
@@ -795,21 +772,25 @@ static int msm_gem_new_impl(struct drm_device *dev,
        }
 
        INIT_LIST_HEAD(&msm_obj->submit_entry);
-       list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+       if (struct_mutex_locked) {
+               list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+       } else {
+               mutex_lock(&dev->struct_mutex);
+               list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+               mutex_unlock(&dev->struct_mutex);
+       }
 
        *obj = &msm_obj->base;
 
        return 0;
 }
 
-struct drm_gem_object *msm_gem_new(struct drm_device *dev,
-               uint32_t size, uint32_t flags)
+static struct drm_gem_object *_msm_gem_new(struct drm_device *dev,
+               uint32_t size, uint32_t flags, bool struct_mutex_locked)
 {
        struct drm_gem_object *obj = NULL;
        int ret;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
-
        size = PAGE_ALIGN(size);
 
        /* Disallow zero sized objects as they make the underlying
@@ -818,7 +799,7 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev,
        if (size == 0)
                return ERR_PTR(-EINVAL);
 
-       ret = msm_gem_new_impl(dev, size, flags, NULL, &obj);
+       ret = msm_gem_new_impl(dev, size, flags, NULL, &obj, struct_mutex_locked);
        if (ret)
                goto fail;
 
@@ -833,10 +814,22 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev,
        return obj;
 
 fail:
-       drm_gem_object_unreference(obj);
+       drm_gem_object_unreference_unlocked(obj);
        return ERR_PTR(ret);
 }
 
+struct drm_gem_object *msm_gem_new_locked(struct drm_device *dev,
+               uint32_t size, uint32_t flags)
+{
+       return _msm_gem_new(dev, size, flags, true);
+}
+
+struct drm_gem_object *msm_gem_new(struct drm_device *dev,
+               uint32_t size, uint32_t flags)
+{
+       return _msm_gem_new(dev, size, flags, false);
+}
+
 struct drm_gem_object *msm_gem_import(struct drm_device *dev,
                struct dma_buf *dmabuf, struct sg_table *sgt)
 {
@@ -853,7 +846,7 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev,
 
        size = PAGE_ALIGN(dmabuf->size);
 
-       ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj);
+       ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj, false);
        if (ret)
                goto fail;
 
@@ -862,17 +855,22 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev,
        npages = size / PAGE_SIZE;
 
        msm_obj = to_msm_bo(obj);
+       mutex_lock(&msm_obj->lock);
        msm_obj->sgt = sgt;
        msm_obj->pages = drm_malloc_ab(npages, sizeof(struct page *));
        if (!msm_obj->pages) {
+               mutex_unlock(&msm_obj->lock);
                ret = -ENOMEM;
                goto fail;
        }
 
        ret = drm_prime_sg_to_page_addr_arrays(sgt, msm_obj->pages, NULL, npages);
-       if (ret)
+       if (ret) {
+               mutex_unlock(&msm_obj->lock);
                goto fail;
+       }
 
+       mutex_unlock(&msm_obj->lock);
        return obj;
 
 fail:
drivers/gpu/drm/msm/msm_gem.h +2 -0
index 1b4cf20..b1bfc10 100644
@@ -31,6 +31,7 @@ struct msm_gem_address_space {
         * and position mm_node->start is in # of pages:
         */
        struct drm_mm mm;
+       spinlock_t lock; /* Protects drm_mm node allocation/removal */
        struct msm_mmu *mmu;
        struct kref kref;
 };
@@ -87,6 +88,7 @@ struct msm_gem_object {
         * an IOMMU.  Also used for stolen/splashscreen buffer.
         */
        struct drm_mm_node *vram_node;
+       struct mutex lock; /* Protects resources associated with bo */
 };
 #define to_msm_bo(x) container_of(x, struct msm_gem_object, base)
 
drivers/gpu/drm/msm/msm_gem_submit.c +3 -3
index 1c545eb..8420551 100644
@@ -245,7 +245,7 @@ static int submit_pin_objects(struct msm_gem_submit *submit)
                uint64_t iova;
 
                /* if locking succeeded, pin bo: */
-               ret = msm_gem_get_iova_locked(&msm_obj->base,
+               ret = msm_gem_get_iova(&msm_obj->base,
                                submit->gpu->id, &iova);
 
                if (ret)
@@ -301,7 +301,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
        /* For now, just map the entire thing.  Eventually we probably
         * to do it page-by-page, w/ kmap() if not vmap()d..
         */
-       ptr = msm_gem_get_vaddr_locked(&obj->base);
+       ptr = msm_gem_get_vaddr(&obj->base);
 
        if (IS_ERR(ptr)) {
                ret = PTR_ERR(ptr);
@@ -359,7 +359,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
        }
 
 out:
-       msm_gem_put_vaddr_locked(&obj->base);
+       msm_gem_put_vaddr(&obj->base);
 
        return ret;
 }
drivers/gpu/drm/msm/msm_gem_vma.c +9 -1
index f285d7e..c36321bc 100644
@@ -50,7 +50,9 @@ void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
                aspace->mmu->funcs->unmap(aspace->mmu, vma->iova, sgt, size);
        }
 
+       spin_lock(&aspace->lock);
        drm_mm_remove_node(&vma->node);
+       spin_unlock(&aspace->lock);
 
        vma->iova = 0;
 
@@ -63,10 +65,15 @@ void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
 {
        int ret;
 
-       if (WARN_ON(drm_mm_node_allocated(&vma->node)))
+       spin_lock(&aspace->lock);
+       if (WARN_ON(drm_mm_node_allocated(&vma->node))) {
+               spin_unlock(&aspace->lock);
                return 0;
+       }
 
        ret = drm_mm_insert_node(&aspace->mm, &vma->node, npages);
+       spin_unlock(&aspace->lock);
+
        if (ret)
                return ret;
 
@@ -94,6 +101,7 @@ struct msm_gem_address_space *
        if (!aspace)
                return ERR_PTR(-ENOMEM);
 
+       spin_lock_init(&aspace->lock);
        aspace->name = name;
        aspace->mmu = msm_iommu_new(dev, domain);
 
drivers/gpu/drm/msm/msm_gpu.c +1 -3
index 97b9c38..d6e5cb2 100644
@@ -495,7 +495,7 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
                /* submit takes a reference to the bo and iova until retired: */
                drm_gem_object_reference(&msm_obj->base);
-               msm_gem_get_iova_locked(&msm_obj->base,
+               msm_gem_get_iova(&msm_obj->base,
                                submit->gpu->id, &iova);
 
                if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
@@ -662,9 +662,7 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 
 
        /* Create ringbuffer: */
-       mutex_lock(&drm->struct_mutex);
        gpu->rb = msm_ringbuffer_new(gpu, ringsz);
-       mutex_unlock(&drm->struct_mutex);
        if (IS_ERR(gpu->rb)) {
                ret = PTR_ERR(gpu->rb);
                gpu->rb = NULL;
drivers/gpu/drm/msm/msm_rd.c +2 -2
index 0e81faa..0366b80 100644
@@ -268,7 +268,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
        struct msm_gem_object *obj = submit->bos[idx].obj;
        const char *buf;
 
-       buf = msm_gem_get_vaddr_locked(&obj->base);
+       buf = msm_gem_get_vaddr(&obj->base);
        if (IS_ERR(buf))
                return;
 
@@ -283,7 +283,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
                        (uint32_t[3]){ iova, size, iova >> 32 }, 12);
        rd_write_section(rd, RD_BUFFER_CONTENTS, buf, size);
 
-       msm_gem_put_vaddr_locked(&obj->base);
+       msm_gem_put_vaddr(&obj->base);
 }
 
 /* called under struct_mutex */
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -2
index 67b34e0..791bca3 100644
@@ -40,7 +40,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int size)
                goto fail;
        }
 
-       ring->start = msm_gem_get_vaddr_locked(ring->bo);
+       ring->start = msm_gem_get_vaddr(ring->bo);
        if (IS_ERR(ring->start)) {
                ret = PTR_ERR(ring->start);
                goto fail;
--

1.9.1

Quoted 4 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-arm-msm" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno

#3 lore raw
In reply to Sushmita Susheelendra
Quoted 7 lines · Sushmita Susheelendra

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.

I've rebased this on top of the other changes that are in the queue for 4.13:

https://github.com/freedreno/kernel-msm/commit/4e0c4e139a647914cfcf7c413da5c19f9f124885

I've done some light testing on a5xx.. although between this and moving gpu->hw_init() under struct_mutex, I need to double check on a3xx/a4xx.

I do think we probably need a bit more work for shrinker. In particular msm_gem_purge() still assumes everything is protected by a single struct_mutex, which is no longer true. The tricky thing here is that shrinker can be triggered by code-paths where we already hold msm_obj->lock.

I think we probably want anything that ends up in get_pages() or vmap() to have something along the lines of

  if (WARN_ON(msm_obj->madv == DONTNEED)) {
    mutex_unlock(&msm_obj->lock);
    return -EINVAL;   // or -EBUSY, or ??
  }

it's really only something that a badly behaved userspace could triger.. but otoh we probably don't want to let a badly behaved userspace deadlock things in the kernel. I guess I probably should have written some test cases for this by now.

Quoted 1 line

Change-Id: I4fba9f8c38a6cd13f80f660639d1e74d4336e3fb

jfyi, no need for the Change-Id.. I usually strip them out when I apply patches, but I appreciate if you can strip them out before sending (because sometimes I forget)

BR,
-R

Quoted 784 lines
---
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c    |   2 -
 drivers/gpu/drm/msm/adreno/a5xx_power.c  |   2 -
 drivers/gpu/drm/msm/adreno/adreno_gpu.c  |   2 -
 drivers/gpu/drm/msm/dsi/dsi_host.c       |   5 +-
 drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c |   2 +-
 drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c  |   2 -
 drivers/gpu/drm/msm/msm_drv.c            |   1 +
 drivers/gpu/drm/msm/msm_drv.h            |   7 +-
 drivers/gpu/drm/msm/msm_fbdev.c          |   6 +-
 drivers/gpu/drm/msm/msm_gem.c            | 190 +++++++++++++++----------------
 drivers/gpu/drm/msm/msm_gem.h            |   2 +
 drivers/gpu/drm/msm/msm_gem_submit.c     |   6 +-
 drivers/gpu/drm/msm/msm_gem_vma.c        |  10 +-
 drivers/gpu/drm/msm/msm_gpu.c            |   4 +-
 drivers/gpu/drm/msm/msm_rd.c             |   4 +-
 drivers/gpu/drm/msm/msm_ringbuffer.c     |   2 +-
 16 files changed, 120 insertions(+), 127 deletions(-)
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +0 -2
index 31a9bce..7893de1 100644
@@ -235,9 +235,7 @@ static struct drm_gem_object *a5xx_ucode_load_bo(struct msm_gpu *gpu,
        struct drm_gem_object *bo;
        void *ptr;
 
-       mutex_lock(&drm->struct_mutex);
        bo = msm_gem_new(drm, fw->size - 4, MSM_BO_UNCACHED);
-       mutex_unlock(&drm->struct_mutex);
 
        if (IS_ERR(bo))
                return bo;
drivers/gpu/drm/msm/adreno/a5xx_power.c +0 -2
index 72d52c7..eb88f44 100644
@@ -294,9 +294,7 @@ void a5xx_gpmu_ucode_init(struct msm_gpu *gpu)
         */
        bosize = (cmds_size + (cmds_size / TYPE4_MAX_PAYLOAD) + 1) << 2;
 
-       mutex_lock(&drm->struct_mutex);
        a5xx_gpu->gpmu_bo = msm_gem_new(drm, bosize, MSM_BO_UNCACHED);
-       mutex_unlock(&drm->struct_mutex);
 
        if (IS_ERR(a5xx_gpu->gpmu_bo))
                goto err;
drivers/gpu/drm/msm/adreno/adreno_gpu.c +0 -2
index 5b63fc6..1162c15 100644
@@ -392,10 +392,8 @@ 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 +1 -4
index 4f79b10..6a1b0da 100644
@@ -980,19 +980,16 @@ 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, 0, &iova);
-               mutex_unlock(&dev->struct_mutex);
+               ret = msm_gem_get_iova(msm_host->tx_gem_obj, 0, &iova);
                if (ret) {
                        pr_err("%s: failed to get iova, %d\n", __func__, ret);
                        return ret;
drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +1 -1
index f29194a..15478f8 100644
@@ -372,7 +372,7 @@ 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, mdp4_kms->id, &iova);
+                       msm_gem_get_iova(next_bo, mdp4_kms->id, &iova);
 
                        /* enable cursor: */
                        mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_SIZE(dma),
drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c +0 -2
index 6295204..3a464b3 100644
@@ -561,9 +561,7 @@ 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 +1 -0
index 87b5695..bb1f3ee 100644
@@ -349,6 +349,7 @@ 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 +3 -4
index 28b6f9b..567737d 100644
@@ -157,6 +157,7 @@ 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;
@@ -209,8 +210,6 @@ 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, int id,
-               uint64_t *iova);
 int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova);
 uint64_t msm_gem_iova(struct drm_gem_object *obj, int id);
 struct page **msm_gem_get_pages(struct drm_gem_object *obj);
@@ -228,9 +227,7 @@ 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);
@@ -247,6 +244,8 @@ 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 +2 -4
index 951e40f..785925e9 100644
@@ -95,10 +95,8 @@ 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;
@@ -124,7 +122,7 @@ 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, 0, &paddr);
+       ret = msm_gem_get_iova(fbdev->bo, 0, &paddr);
        if (ret) {
                dev_err(dev->dev, "failed to get buffer obj iova: %d\n", ret);
                goto fail_unlock;
@@ -153,7 +151,7 @@ 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;
drivers/gpu/drm/msm/msm_gem.c +94 -96
index 68e509b..1e803fb 100644
@@ -41,8 +41,7 @@ static bool use_pages(struct drm_gem_object *obj)
 }
 
 /* allocate pages from VRAM carveout, used when no IOMMU: */
-static struct page **get_pages_vram(struct drm_gem_object *obj,
-               int npages)
+static struct page **get_pages_vram(struct drm_gem_object *obj, int npages)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
        struct msm_drm_private *priv = obj->dev->dev_private;
@@ -54,7 +53,9 @@ static struct page **get_pages_vram(struct drm_gem_object *obj,
        if (!p)
                return ERR_PTR(-ENOMEM);
 
+       spin_lock(&priv->vram.lock);
        ret = drm_mm_insert_node(&priv->vram.mm, msm_obj->vram_node, npages);
+       spin_unlock(&priv->vram.lock);
        if (ret) {
                drm_free_large(p);
                return ERR_PTR(ret);
@@ -69,7 +70,6 @@ static struct page **get_pages_vram(struct drm_gem_object *obj,
        return p;
 }
 
-/* called with dev->struct_mutex held */
 static struct page **get_pages(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -109,6 +109,18 @@ static struct page **get_pages(struct drm_gem_object *obj)
        return msm_obj->pages;
 }
 
+static void put_pages_vram(struct drm_gem_object *obj)
+{
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+       struct msm_drm_private *priv = obj->dev->dev_private;
+
+       spin_lock(&priv->vram.lock);
+       drm_mm_remove_node(msm_obj->vram_node);
+       spin_unlock(&priv->vram.lock);
+
+       drm_free_large(msm_obj->pages);
+}
+
 static void put_pages(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -125,10 +137,8 @@ static void put_pages(struct drm_gem_object *obj)
 
                if (use_pages(obj))
                        drm_gem_put_pages(obj, msm_obj->pages, true, false);
-               else {
-                       drm_mm_remove_node(msm_obj->vram_node);
-                       drm_free_large(msm_obj->pages);
-               }
+               else
+                       put_pages_vram(obj);
 
                msm_obj->pages = NULL;
        }
@@ -136,11 +146,12 @@ static void put_pages(struct drm_gem_object *obj)
 
 struct page **msm_gem_get_pages(struct drm_gem_object *obj)
 {
-       struct drm_device *dev = obj->dev;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
        struct page **p;
-       mutex_lock(&dev->struct_mutex);
+
+       mutex_lock(&msm_obj->lock);
        p = get_pages(obj);
-       mutex_unlock(&dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
        return p;
 }
 
@@ -195,25 +206,17 @@ int msm_gem_fault(struct vm_fault *vmf)
 {
        struct vm_area_struct *vma = vmf->vma;
        struct drm_gem_object *obj = vma->vm_private_data;
-       struct drm_device *dev = obj->dev;
-       struct msm_drm_private *priv = dev->dev_private;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
        struct page **pages;
        unsigned long pfn;
        pgoff_t pgoff;
        int ret;
 
-       /* This should only happen if userspace tries to pass a mmap'd
-        * but unfaulted gem bo vaddr into submit ioctl, triggering
-        * a page fault while struct_mutex is already held.  This is
-        * not a valid use-case so just bail.
+       /*
+        * vm_ops.open/drm_gem_mmap_obj and close get and put
+        * a reference on obj. So, we dont need to hold one here.
         */
-       if (priv->struct_mutex_task == current)
-               return VM_FAULT_SIGBUS;
-
-       /* Make sure we don't parallel update on a fault, nor move or remove
-        * something from beneath our feet
-        */
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
+       ret = mutex_lock_interruptible(&msm_obj->lock);
        if (ret)
                goto out;
 
@@ -235,7 +238,7 @@ int msm_gem_fault(struct vm_fault *vmf)
        ret = vm_insert_mixed(vma, vmf->address, __pfn_to_pfn_t(pfn, PFN_DEV));
 
 out_unlock:
-       mutex_unlock(&dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
 out:
        switch (ret) {
        case -EAGAIN:
@@ -259,9 +262,10 @@ int msm_gem_fault(struct vm_fault *vmf)
 static uint64_t mmap_offset(struct drm_gem_object *obj)
 {
        struct drm_device *dev = obj->dev;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
        int ret;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
 
        /* Make it mmapable */
        ret = drm_gem_create_mmap_offset(obj);
@@ -277,21 +281,23 @@ static uint64_t mmap_offset(struct drm_gem_object *obj)
 uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj)
 {
        uint64_t offset;
-       mutex_lock(&obj->dev->struct_mutex);
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+
+       mutex_lock(&msm_obj->lock);
        offset = mmap_offset(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
        return offset;
 }
 
+/* Called with msm_obj->lock locked */
 static void
 put_iova(struct drm_gem_object *obj)
 {
-       struct drm_device *dev = obj->dev;
        struct msm_drm_private *priv = obj->dev->dev_private;
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
        int id;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
 
        for (id = 0; id < ARRAY_SIZE(msm_obj->domain); id++) {
                if (!priv->aspace[id])
@@ -301,25 +307,23 @@ uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj)
        }
 }
 
-/* should be called under struct_mutex.. although it can be called
- * from atomic context without struct_mutex to acquire an extra
- * iova ref if you know one is already held.
- *
- * That means when I do eventually need to add support for unpinning
- * the refcnt counter needs to be atomic_t.
- */
-int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
+/* A reference to obj must be held before calling this function. */
+int msm_gem_get_iova(struct drm_gem_object *obj, int id,
                uint64_t *iova)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
        int ret = 0;
 
+       mutex_lock(&msm_obj->lock);
+
        if (!msm_obj->domain[id].iova) {
                struct msm_drm_private *priv = obj->dev->dev_private;
                struct page **pages = get_pages(obj);
 
-               if (IS_ERR(pages))
+               if (IS_ERR(pages)) {
+                       mutex_unlock(&msm_obj->lock);
                        return PTR_ERR(pages);
+               }
 
                if (iommu_present(&platform_bus_type)) {
                        ret = msm_gem_map_vma(priv->aspace[id], &msm_obj->domain[id],
@@ -332,26 +336,7 @@ int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
        if (!ret)
                *iova = msm_obj->domain[id].iova;
 
-       return ret;
-}
-
-/* get iova, taking a reference.  Should have a matching put */
-int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova)
-{
-       struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       int ret;
-
-       /* this is safe right now because we don't unmap until the
-        * bo is deleted:
-        */
-       if (msm_obj->domain[id].iova) {
-               *iova = msm_obj->domain[id].iova;
-               return 0;
-       }
-
-       mutex_lock(&obj->dev->struct_mutex);
-       ret = msm_gem_get_iova_locked(obj, id, iova);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
        return ret;
 }
 
@@ -405,45 +390,37 @@ int msm_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev,
        return ret;
 }
 
-void *msm_gem_get_vaddr_locked(struct drm_gem_object *obj)
+void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+
+       mutex_lock(&msm_obj->lock);
        if (!msm_obj->vaddr) {
                struct page **pages = get_pages(obj);
-               if (IS_ERR(pages))
+               if (IS_ERR(pages)) {
+                       mutex_unlock(&msm_obj->lock);
                        return ERR_CAST(pages);
+               }
                msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
                                VM_MAP, pgprot_writecombine(PAGE_KERNEL));
-               if (msm_obj->vaddr == NULL)
+               if (msm_obj->vaddr == NULL) {
+                       mutex_unlock(&msm_obj->lock);
                        return ERR_PTR(-ENOMEM);
+               }
        }
        msm_obj->vmap_count++;
+       mutex_unlock(&msm_obj->lock);
        return msm_obj->vaddr;
 }
 
-void *msm_gem_get_vaddr(struct drm_gem_object *obj)
-{
-       void *ret;
-       mutex_lock(&obj->dev->struct_mutex);
-       ret = msm_gem_get_vaddr_locked(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
-       return ret;
-}
-
-void msm_gem_put_vaddr_locked(struct drm_gem_object *obj)
+void msm_gem_put_vaddr(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+
+       mutex_lock(&msm_obj->lock);
        WARN_ON(msm_obj->vmap_count < 1);
        msm_obj->vmap_count--;
-}
-
-void msm_gem_put_vaddr(struct drm_gem_object *obj)
-{
-       mutex_lock(&obj->dev->struct_mutex);
-       msm_gem_put_vaddr_locked(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
 }
 
 /* Update madvise status, returns true if not purged, else
@@ -697,6 +674,8 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
        list_del(&msm_obj->mm_list);
 
+       mutex_lock(&msm_obj->lock);
+
        put_iova(obj);
 
        if (obj->import_attach) {
@@ -720,6 +699,7 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
        drm_gem_object_release(obj);
 
+       mutex_unlock(&msm_obj->lock);
        kfree(msm_obj);
 }
 
@@ -730,14 +710,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file,
        struct drm_gem_object *obj;
        int ret;
 
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
-       if (ret)
-               return ret;
-
        obj = msm_gem_new(dev, size, flags);
 
-       mutex_unlock(&dev->struct_mutex);
-
        if (IS_ERR(obj))
                return PTR_ERR(obj);
 
@@ -752,7 +726,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file,
 static int msm_gem_new_impl(struct drm_device *dev,
                uint32_t size, uint32_t flags,
                struct reservation_object *resv,
-               struct drm_gem_object **obj)
+               struct drm_gem_object **obj,
+               bool struct_mutex_locked)
 {
        struct msm_drm_private *priv = dev->dev_private;
        struct msm_gem_object *msm_obj;
@@ -781,6 +756,8 @@ static int msm_gem_new_impl(struct drm_device *dev,
        if (!msm_obj)
                return -ENOMEM;
 
+       mutex_init(&msm_obj->lock);
+
        if (use_vram)
                msm_obj->vram_node = &msm_obj->domain[0].node;
 
@@ -795,21 +772,25 @@ static int msm_gem_new_impl(struct drm_device *dev,
        }
 
        INIT_LIST_HEAD(&msm_obj->submit_entry);
-       list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+       if (struct_mutex_locked) {
+               list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+       } else {
+               mutex_lock(&dev->struct_mutex);
+               list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+               mutex_unlock(&dev->struct_mutex);
+       }
 
        *obj = &msm_obj->base;
 
        return 0;
 }
 
-struct drm_gem_object *msm_gem_new(struct drm_device *dev,
-               uint32_t size, uint32_t flags)
+static struct drm_gem_object *_msm_gem_new(struct drm_device *dev,
+               uint32_t size, uint32_t flags, bool struct_mutex_locked)
 {
        struct drm_gem_object *obj = NULL;
        int ret;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
-
        size = PAGE_ALIGN(size);
 
        /* Disallow zero sized objects as they make the underlying
@@ -818,7 +799,7 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev,
        if (size == 0)
                return ERR_PTR(-EINVAL);
 
-       ret = msm_gem_new_impl(dev, size, flags, NULL, &obj);
+       ret = msm_gem_new_impl(dev, size, flags, NULL, &obj, struct_mutex_locked);
        if (ret)
                goto fail;
 
@@ -833,10 +814,22 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev,
        return obj;
 
 fail:
-       drm_gem_object_unreference(obj);
+       drm_gem_object_unreference_unlocked(obj);
        return ERR_PTR(ret);
 }
 
+struct drm_gem_object *msm_gem_new_locked(struct drm_device *dev,
+               uint32_t size, uint32_t flags)
+{
+       return _msm_gem_new(dev, size, flags, true);
+}
+
+struct drm_gem_object *msm_gem_new(struct drm_device *dev,
+               uint32_t size, uint32_t flags)
+{
+       return _msm_gem_new(dev, size, flags, false);
+}
+
 struct drm_gem_object *msm_gem_import(struct drm_device *dev,
                struct dma_buf *dmabuf, struct sg_table *sgt)
 {
@@ -853,7 +846,7 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev,
 
        size = PAGE_ALIGN(dmabuf->size);
 
-       ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj);
+       ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj, false);
        if (ret)
                goto fail;
 
@@ -862,17 +855,22 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev,
        npages = size / PAGE_SIZE;
 
        msm_obj = to_msm_bo(obj);
+       mutex_lock(&msm_obj->lock);
        msm_obj->sgt = sgt;
        msm_obj->pages = drm_malloc_ab(npages, sizeof(struct page *));
        if (!msm_obj->pages) {
+               mutex_unlock(&msm_obj->lock);
                ret = -ENOMEM;
                goto fail;
        }
 
        ret = drm_prime_sg_to_page_addr_arrays(sgt, msm_obj->pages, NULL, npages);
-       if (ret)
+       if (ret) {
+               mutex_unlock(&msm_obj->lock);
                goto fail;
+       }
 
+       mutex_unlock(&msm_obj->lock);
        return obj;
 
 fail:
drivers/gpu/drm/msm/msm_gem.h +2 -0
index 1b4cf20..b1bfc10 100644
@@ -31,6 +31,7 @@ struct msm_gem_address_space {
         * and position mm_node->start is in # of pages:
         */
        struct drm_mm mm;
+       spinlock_t lock; /* Protects drm_mm node allocation/removal */
        struct msm_mmu *mmu;
        struct kref kref;
 };
@@ -87,6 +88,7 @@ struct msm_gem_object {
         * an IOMMU.  Also used for stolen/splashscreen buffer.
         */
        struct drm_mm_node *vram_node;
+       struct mutex lock; /* Protects resources associated with bo */
 };
 #define to_msm_bo(x) container_of(x, struct msm_gem_object, base)
 
drivers/gpu/drm/msm/msm_gem_submit.c +3 -3
index 1c545eb..8420551 100644
@@ -245,7 +245,7 @@ static int submit_pin_objects(struct msm_gem_submit *submit)
                uint64_t iova;
 
                /* if locking succeeded, pin bo: */
-               ret = msm_gem_get_iova_locked(&msm_obj->base,
+               ret = msm_gem_get_iova(&msm_obj->base,
                                submit->gpu->id, &iova);
 
                if (ret)
@@ -301,7 +301,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
        /* For now, just map the entire thing.  Eventually we probably
         * to do it page-by-page, w/ kmap() if not vmap()d..
         */
-       ptr = msm_gem_get_vaddr_locked(&obj->base);
+       ptr = msm_gem_get_vaddr(&obj->base);
 
        if (IS_ERR(ptr)) {
                ret = PTR_ERR(ptr);
@@ -359,7 +359,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
        }
 
 out:
-       msm_gem_put_vaddr_locked(&obj->base);
+       msm_gem_put_vaddr(&obj->base);
 
        return ret;
 }
drivers/gpu/drm/msm/msm_gem_vma.c +9 -1
index f285d7e..c36321bc 100644
@@ -50,7 +50,9 @@ void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
                aspace->mmu->funcs->unmap(aspace->mmu, vma->iova, sgt, size);
        }
 
+       spin_lock(&aspace->lock);
        drm_mm_remove_node(&vma->node);
+       spin_unlock(&aspace->lock);
 
        vma->iova = 0;
 
@@ -63,10 +65,15 @@ void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
 {
        int ret;
 
-       if (WARN_ON(drm_mm_node_allocated(&vma->node)))
+       spin_lock(&aspace->lock);
+       if (WARN_ON(drm_mm_node_allocated(&vma->node))) {
+               spin_unlock(&aspace->lock);
                return 0;
+       }
 
        ret = drm_mm_insert_node(&aspace->mm, &vma->node, npages);
+       spin_unlock(&aspace->lock);
+
        if (ret)
                return ret;
 
@@ -94,6 +101,7 @@ struct msm_gem_address_space *
        if (!aspace)
                return ERR_PTR(-ENOMEM);
 
+       spin_lock_init(&aspace->lock);
        aspace->name = name;
        aspace->mmu = msm_iommu_new(dev, domain);
 
drivers/gpu/drm/msm/msm_gpu.c +1 -3
index 97b9c38..d6e5cb2 100644
@@ -495,7 +495,7 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
                /* submit takes a reference to the bo and iova until retired: */
                drm_gem_object_reference(&msm_obj->base);
-               msm_gem_get_iova_locked(&msm_obj->base,
+               msm_gem_get_iova(&msm_obj->base,
                                submit->gpu->id, &iova);
 
                if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
@@ -662,9 +662,7 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 
 
        /* Create ringbuffer: */
-       mutex_lock(&drm->struct_mutex);
        gpu->rb = msm_ringbuffer_new(gpu, ringsz);
-       mutex_unlock(&drm->struct_mutex);
        if (IS_ERR(gpu->rb)) {
                ret = PTR_ERR(gpu->rb);
                gpu->rb = NULL;
drivers/gpu/drm/msm/msm_rd.c +2 -2
index 0e81faa..0366b80 100644
@@ -268,7 +268,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
        struct msm_gem_object *obj = submit->bos[idx].obj;
        const char *buf;
 
-       buf = msm_gem_get_vaddr_locked(&obj->base);
+       buf = msm_gem_get_vaddr(&obj->base);
        if (IS_ERR(buf))
                return;
 
@@ -283,7 +283,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
                        (uint32_t[3]){ iova, size, iova >> 32 }, 12);
        rd_write_section(rd, RD_BUFFER_CONTENTS, buf, size);
 
-       msm_gem_put_vaddr_locked(&obj->base);
+       msm_gem_put_vaddr(&obj->base);
 }
 
 /* called under struct_mutex */
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -2
index 67b34e0..791bca3 100644
@@ -40,7 +40,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int size)
                goto fail;
        }
 
-       ring->start = msm_gem_get_vaddr_locked(ring->bo);
+       ring->start = msm_gem_get_vaddr(ring->bo);
        if (IS_ERR(ring->start)) {
                ret = PTR_ERR(ring->start);
                goto fail;
--

1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-arm-msm" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

A message that is not in the archive (sent off-list or not yet received): CAF6AEGuUNmpbssk0nsf6Nrb3Z-H5ug-gthEkZTqg_9GQAA3iUA-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#5 lore raw

Hi Rob,

Yes, I’d completely missed the shrinker path in this cleanup. But, yeah, I see how get_pages (which is called with msm_obj->lock held) -> drm_gem_get_pages could trigger shrinker_scan which calls msm_gem_purge. It makes sense to prevent any get_pages/vmap on objects that’ve been marked as DONTNEED. I’ll send you a patch soon for that.

Thanks,
Sushmita

Quoted 828 lines · Rob Clark
Quoted 7 lines · On Tue, Jun 13, 2017 at 6:52 PM, Sushmita Susheelendra…

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.

I've rebased this on top of the other changes that are in the queue for 4.13:

https://github.com/freedreno/kernel-msm/commit/4e0c4e139a647914cfcf7c413da5c19f9f124885 <https://github.com/freedreno/kernel-msm/commit/4e0c4e139a647914cfcf7c413da5c19f9f124885>

I've done some light testing on a5xx.. although between this and moving gpu->hw_init() under struct_mutex, I need to double check on a3xx/a4xx.

I do think we probably need a bit more work for shrinker. In particular msm_gem_purge() still assumes everything is protected by a single struct_mutex, which is no longer true. The tricky thing here is that shrinker can be triggered by code-paths where we already hold msm_obj->lock.

I think we probably want anything that ends up in get_pages() or vmap() to have something along the lines of

if (WARN_ON(msm_obj->madv == DONTNEED)) {
mutex_unlock(&msm_obj->lock);
return -EINVAL; // or -EBUSY, or ??
}

it's really only something that a badly behaved userspace could triger.. but otoh we probably don't want to let a badly behaved userspace deadlock things in the kernel. I guess I probably should have written some test cases for this by now.

Quoted 1 line

Change-Id: I4fba9f8c38a6cd13f80f660639d1e74d4336e3fb

jfyi, no need for the Change-Id.. I usually strip them out when I apply patches, but I appreciate if you can strip them out before sending (because sometimes I forget)

BR,
-R

Quoted 780 lines

---

drivers/gpu/drm/msm/adreno/a5xx_gpu.c    |   2 -
drivers/gpu/drm/msm/adreno/a5xx_power.c  |   2 -
drivers/gpu/drm/msm/adreno/adreno_gpu.c  |   2 -
drivers/gpu/drm/msm/dsi/dsi_host.c       |   5 +-
drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c |   2 +-
drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c  |   2 -
drivers/gpu/drm/msm/msm_drv.c            |   1 +
drivers/gpu/drm/msm/msm_drv.h            |   7 +-
drivers/gpu/drm/msm/msm_fbdev.c          |   6 +-
drivers/gpu/drm/msm/msm_gem.c            | 190 +++++++++++++++----------------
drivers/gpu/drm/msm/msm_gem.h            |   2 +
drivers/gpu/drm/msm/msm_gem_submit.c     |   6 +-
drivers/gpu/drm/msm/msm_gem_vma.c        |  10 +-
drivers/gpu/drm/msm/msm_gpu.c            |   4 +-
drivers/gpu/drm/msm/msm_rd.c             |   4 +-
drivers/gpu/drm/msm/msm_ringbuffer.c     |   2 +-
16 files changed, 120 insertions(+), 127 deletions(-)
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +0 -2
index 31a9bce..7893de1 100644
@@ -235,9 +235,7 @@ static struct drm_gem_object *a5xx_ucode_load_bo(struct msm_gpu *gpu,
       struct drm_gem_object *bo;
       void *ptr;
 
-       mutex_lock(&drm->struct_mutex);
       bo = msm_gem_new(drm, fw->size - 4, MSM_BO_UNCACHED);
-       mutex_unlock(&drm->struct_mutex);
 
       if (IS_ERR(bo))
               return bo;
drivers/gpu/drm/msm/adreno/a5xx_power.c +0 -2
index 72d52c7..eb88f44 100644
@@ -294,9 +294,7 @@ void a5xx_gpmu_ucode_init(struct msm_gpu *gpu)
        */
       bosize = (cmds_size + (cmds_size / TYPE4_MAX_PAYLOAD) + 1) << 2;
 
-       mutex_lock(&drm->struct_mutex);
       a5xx_gpu->gpmu_bo = msm_gem_new(drm, bosize, MSM_BO_UNCACHED);
-       mutex_unlock(&drm->struct_mutex);
 
       if (IS_ERR(a5xx_gpu->gpmu_bo))
               goto err;
drivers/gpu/drm/msm/adreno/adreno_gpu.c +0 -2
index 5b63fc6..1162c15 100644
@@ -392,10 +392,8 @@ 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 +1 -4
index 4f79b10..6a1b0da 100644
@@ -980,19 +980,16 @@ 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, 0, &iova);
-               mutex_unlock(&dev->struct_mutex);
+               ret = msm_gem_get_iova(msm_host->tx_gem_obj, 0, &iova);
               if (ret) {
                       pr_err("%s: failed to get iova, %d\n", __func__, ret);
                       return ret;
drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +1 -1
index f29194a..15478f8 100644
@@ -372,7 +372,7 @@ 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, mdp4_kms->id, &iova);
+                       msm_gem_get_iova(next_bo, mdp4_kms->id, &iova);
 
                       /* enable cursor: */
                       mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_SIZE(dma),
drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c +0 -2
index 6295204..3a464b3 100644
@@ -561,9 +561,7 @@ 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 +1 -0
index 87b5695..bb1f3ee 100644
@@ -349,6 +349,7 @@ 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 +3 -4
index 28b6f9b..567737d 100644
@@ -157,6 +157,7 @@ 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;
@@ -209,8 +210,6 @@ 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, int id,
-               uint64_t *iova);
 int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova);
 uint64_t msm_gem_iova(struct drm_gem_object *obj, int id);
 struct page **msm_gem_get_pages(struct drm_gem_object *obj);
@@ -228,9 +227,7 @@ 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);
@@ -247,6 +244,8 @@ 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 +2 -4
index 951e40f..785925e9 100644
@@ -95,10 +95,8 @@ 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;
@@ -124,7 +122,7 @@ 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, 0, &paddr);
+       ret = msm_gem_get_iova(fbdev->bo, 0, &paddr);
       if (ret) {
               dev_err(dev->dev, "failed to get buffer obj iova: %d\n", ret);
               goto fail_unlock;
@@ -153,7 +151,7 @@ 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;
drivers/gpu/drm/msm/msm_gem.c +26 -23
index 68e509b..1e803fb 100644
@@ -41,8 +41,7 @@ static bool use_pages(struct drm_gem_object *obj)
 }
 
 /* allocate pages from VRAM carveout, used when no IOMMU: */
-static struct page **get_pages_vram(struct drm_gem_object *obj,
-               int npages)
+static struct page **get_pages_vram(struct drm_gem_object *obj, int npages)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       struct msm_drm_private *priv = obj->dev->dev_private;
@@ -54,7 +53,9 @@ static struct page **get_pages_vram(struct drm_gem_object *obj,
       if (!p)
               return ERR_PTR(-ENOMEM);
 
+       spin_lock(&priv->vram.lock);
       ret = drm_mm_insert_node(&priv->vram.mm, msm_obj->vram_node, npages);
+       spin_unlock(&priv->vram.lock);
       if (ret) {
               drm_free_large(p);
               return ERR_PTR(ret);
@@ -69,7 +70,6 @@ static struct page **get_pages_vram(struct drm_gem_object *obj,
       return p;
 }
 
-/* called with dev->struct_mutex held */
 static struct page **get_pages(struct drm_gem_object *obj)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -109,6 +109,18 @@ static struct page **get_pages(struct drm_gem_object *obj)
       return msm_obj->pages;
 }
 
+static void put_pages_vram(struct drm_gem_object *obj)
+{
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+       struct msm_drm_private *priv = obj->dev->dev_private;
+
+       spin_lock(&priv->vram.lock);
+       drm_mm_remove_node(msm_obj->vram_node);
+       spin_unlock(&priv->vram.lock);
+
+       drm_free_large(msm_obj->pages);
+}
+
 static void put_pages(struct drm_gem_object *obj)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -125,10 +137,8 @@ static void put_pages(struct drm_gem_object *obj)
 
               if (use_pages(obj))
                       drm_gem_put_pages(obj, msm_obj->pages, true, false);
-               else {
-                       drm_mm_remove_node(msm_obj->vram_node);
-                       drm_free_large(msm_obj->pages);
-               }
+               else
+                       put_pages_vram(obj);
 
               msm_obj->pages = NULL;
       }
@@ -136,11 +146,12 @@ static void put_pages(struct drm_gem_object *obj)
 
 struct page **msm_gem_get_pages(struct drm_gem_object *obj)
 {
-       struct drm_device *dev = obj->dev;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       struct page **p;
-       mutex_lock(&dev->struct_mutex);
+
+       mutex_lock(&msm_obj->lock);
       p = get_pages(obj);
-       mutex_unlock(&dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
       return p;
 }
 
@@ -195,25 +206,17 @@ int msm_gem_fault(struct vm_fault *vmf)
 {
       struct vm_area_struct *vma = vmf->vma;
       struct drm_gem_object *obj = vma->vm_private_data;
-       struct drm_device *dev = obj->dev;
-       struct msm_drm_private *priv = dev->dev_private;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       struct page **pages;
       unsigned long pfn;
       pgoff_t pgoff;
       int ret;
 
-       /* This should only happen if userspace tries to pass a mmap'd
-        * but unfaulted gem bo vaddr into submit ioctl, triggering
-        * a page fault while struct_mutex is already held.  This is
-        * not a valid use-case so just bail.
+       /*
+        * vm_ops.open/drm_gem_mmap_obj and close get and put
+        * a reference on obj. So, we dont need to hold one here.
        */
-       if (priv->struct_mutex_task == current)
-               return VM_FAULT_SIGBUS;
-
-       /* Make sure we don't parallel update on a fault, nor move or remove
-        * something from beneath our feet
-        */
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
+       ret = mutex_lock_interruptible(&msm_obj->lock);
       if (ret)
               goto out;
 
@@ -235,7 +238,7 @@ int msm_gem_fault(struct vm_fault *vmf)
       ret = vm_insert_mixed(vma, vmf->address, __pfn_to_pfn_t(pfn, PFN_DEV));
 

out_unlock:

drivers/gpu/drm/msm/msm_gem.c +1 -1
-       mutex_unlock(&dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);

out:

       switch (ret) {
       case -EAGAIN:
drivers/gpu/drm/msm/msm_gem.c +7 -3
@@ -259,9 +262,10 @@ int msm_gem_fault(struct vm_fault *vmf)
 static uint64_t mmap_offset(struct drm_gem_object *obj)
 {
       struct drm_device *dev = obj->dev;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       int ret;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
 
       /* Make it mmapable */
       ret = drm_gem_create_mmap_offset(obj);
@@ -277,21 +281,23 @@ static uint64_t mmap_offset(struct drm_gem_object *obj)
 uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj)
 {
       uint64_t offset;
-       mutex_lock(&obj->dev->struct_mutex);
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+
+       mutex_lock(&msm_obj->lock);
       offset = mmap_offset(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
       return offset;
 }
 
+/* Called with msm_obj->lock locked */

static void
put_iova(struct drm_gem_object *obj)
{

drivers/gpu/drm/msm/msm_gem.c +40 -66
-       struct drm_device *dev = obj->dev;
       struct msm_drm_private *priv = obj->dev->dev_private;
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       int id;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
 
       for (id = 0; id < ARRAY_SIZE(msm_obj->domain); id++) {
               if (!priv->aspace[id])
@@ -301,25 +307,23 @@ uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj)
       }
 }
 
-/* should be called under struct_mutex.. although it can be called
- * from atomic context without struct_mutex to acquire an extra
- * iova ref if you know one is already held.
- *
- * That means when I do eventually need to add support for unpinning
- * the refcnt counter needs to be atomic_t.
- */
-int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
+/* A reference to obj must be held before calling this function. */
+int msm_gem_get_iova(struct drm_gem_object *obj, int id,
               uint64_t *iova)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       int ret = 0;
 
+       mutex_lock(&msm_obj->lock);
+
       if (!msm_obj->domain[id].iova) {
               struct msm_drm_private *priv = obj->dev->dev_private;
               struct page **pages = get_pages(obj);
 
-               if (IS_ERR(pages))
+               if (IS_ERR(pages)) {
+                       mutex_unlock(&msm_obj->lock);
                       return PTR_ERR(pages);
+               }
 
               if (iommu_present(&platform_bus_type)) {
                       ret = msm_gem_map_vma(priv->aspace[id], &msm_obj->domain[id],
@@ -332,26 +336,7 @@ int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
       if (!ret)
               *iova = msm_obj->domain[id].iova;
 
-       return ret;
-}
-
-/* get iova, taking a reference.  Should have a matching put */
-int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova)
-{
-       struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       int ret;
-
-       /* this is safe right now because we don't unmap until the
-        * bo is deleted:
-        */
-       if (msm_obj->domain[id].iova) {
-               *iova = msm_obj->domain[id].iova;
-               return 0;
-       }
-
-       mutex_lock(&obj->dev->struct_mutex);
-       ret = msm_gem_get_iova_locked(obj, id, iova);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
       return ret;
 }
 
@@ -405,45 +390,37 @@ int msm_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev,
       return ret;
 }
 
-void *msm_gem_get_vaddr_locked(struct drm_gem_object *obj)
+void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+
+       mutex_lock(&msm_obj->lock);
       if (!msm_obj->vaddr) {
               struct page **pages = get_pages(obj);
-               if (IS_ERR(pages))
+               if (IS_ERR(pages)) {
+                       mutex_unlock(&msm_obj->lock);
                       return ERR_CAST(pages);
+               }
               msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
                               VM_MAP, pgprot_writecombine(PAGE_KERNEL));
-               if (msm_obj->vaddr == NULL)
+               if (msm_obj->vaddr == NULL) {
+                       mutex_unlock(&msm_obj->lock);
                       return ERR_PTR(-ENOMEM);
+               }
       }
       msm_obj->vmap_count++;
+       mutex_unlock(&msm_obj->lock);
       return msm_obj->vaddr;
 }
 
-void *msm_gem_get_vaddr(struct drm_gem_object *obj)
-{
-       void *ret;
-       mutex_lock(&obj->dev->struct_mutex);
-       ret = msm_gem_get_vaddr_locked(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
-       return ret;
-}
-
-void msm_gem_put_vaddr_locked(struct drm_gem_object *obj)
+void msm_gem_put_vaddr(struct drm_gem_object *obj)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+
+       mutex_lock(&msm_obj->lock);
       WARN_ON(msm_obj->vmap_count < 1);
       msm_obj->vmap_count--;
-}
-
-void msm_gem_put_vaddr(struct drm_gem_object *obj)
-{
-       mutex_lock(&obj->dev->struct_mutex);
-       msm_gem_put_vaddr_locked(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
 }
 
 /* Update madvise status, returns true if not purged, else
@@ -697,6 +674,8 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
       list_del(&msm_obj->mm_list);
 
+       mutex_lock(&msm_obj->lock);
+
       put_iova(obj);
 
       if (obj->import_attach) {
@@ -720,6 +699,7 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
       drm_gem_object_release(obj);
 
+       mutex_unlock(&msm_obj->lock);
       kfree(msm_obj);
 }
 
@@ -730,14 +710,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file,
       struct drm_gem_object *obj;
       int ret;
 
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
-       if (ret)
-               return ret;
-
       obj = msm_gem_new(dev, size, flags);
 
-       mutex_unlock(&dev->struct_mutex);
-
       if (IS_ERR(obj))
               return PTR_ERR(obj);
 
@@ -752,7 +726,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file,
 static int msm_gem_new_impl(struct drm_device *dev,
               uint32_t size, uint32_t flags,
               struct reservation_object *resv,
-               struct drm_gem_object **obj)
+               struct drm_gem_object **obj,
+               bool struct_mutex_locked)
 {
       struct msm_drm_private *priv = dev->dev_private;
       struct msm_gem_object *msm_obj;
@@ -781,6 +756,8 @@ static int msm_gem_new_impl(struct drm_device *dev,
       if (!msm_obj)
               return -ENOMEM;
 
+       mutex_init(&msm_obj->lock);
+
       if (use_vram)
               msm_obj->vram_node = &msm_obj->domain[0].node;
 
@@ -795,21 +772,25 @@ static int msm_gem_new_impl(struct drm_device *dev,
       }
 
       INIT_LIST_HEAD(&msm_obj->submit_entry);
-       list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+       if (struct_mutex_locked) {
+               list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+       } else {
+               mutex_lock(&dev->struct_mutex);
+               list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+               mutex_unlock(&dev->struct_mutex);
+       }
 
       *obj = &msm_obj->base;
 
       return 0;
 }
 
-struct drm_gem_object *msm_gem_new(struct drm_device *dev,
-               uint32_t size, uint32_t flags)
+static struct drm_gem_object *_msm_gem_new(struct drm_device *dev,
+               uint32_t size, uint32_t flags, bool struct_mutex_locked)
 {
       struct drm_gem_object *obj = NULL;
       int ret;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
-
       size = PAGE_ALIGN(size);
 
       /* Disallow zero sized objects as they make the underlying
@@ -818,7 +799,7 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev,
       if (size == 0)
               return ERR_PTR(-EINVAL);
 
-       ret = msm_gem_new_impl(dev, size, flags, NULL, &obj);
+       ret = msm_gem_new_impl(dev, size, flags, NULL, &obj, struct_mutex_locked);
       if (ret)
               goto fail;
 
@@ -833,10 +814,22 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev,
       return obj;
 

fail:

drivers/gpu/drm/msm/msm_gem.c +1 -1
-       drm_gem_object_unreference(obj);
+       drm_gem_object_unreference_unlocked(obj);
       return ERR_PTR(ret);

}

drivers/gpu/drm/msm/msm_gem.c +12 -0
+struct drm_gem_object *msm_gem_new_locked(struct drm_device *dev,
+               uint32_t size, uint32_t flags)
+{
+       return _msm_gem_new(dev, size, flags, true);
+}
+
+struct drm_gem_object *msm_gem_new(struct drm_device *dev,
+               uint32_t size, uint32_t flags)
+{
+       return _msm_gem_new(dev, size, flags, false);
+}
+

struct drm_gem_object *msm_gem_import(struct drm_device *dev,
struct dma_buf *dmabuf, struct sg_table *sgt)
{

drivers/gpu/drm/msm/msm_gem.c +7 -2
@@ -853,7 +846,7 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev,
 
       size = PAGE_ALIGN(dmabuf->size);
 
-       ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj);
+       ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj, false);
       if (ret)
               goto fail;
 
@@ -862,17 +855,22 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev,
       npages = size / PAGE_SIZE;
 
       msm_obj = to_msm_bo(obj);
+       mutex_lock(&msm_obj->lock);
       msm_obj->sgt = sgt;
       msm_obj->pages = drm_malloc_ab(npages, sizeof(struct page *));
       if (!msm_obj->pages) {
+               mutex_unlock(&msm_obj->lock);
               ret = -ENOMEM;
               goto fail;
       }
 
       ret = drm_prime_sg_to_page_addr_arrays(sgt, msm_obj->pages, NULL, npages);
-       if (ret)
+       if (ret) {
+               mutex_unlock(&msm_obj->lock);
               goto fail;
+       }
 
+       mutex_unlock(&msm_obj->lock);
       return obj;
 

fail:

drivers/gpu/drm/msm/msm_gem.h +2 -0
index 1b4cf20..b1bfc10 100644
@@ -31,6 +31,7 @@ struct msm_gem_address_space {
        * and position mm_node->start is in # of pages:
        */
       struct drm_mm mm;
+       spinlock_t lock; /* Protects drm_mm node allocation/removal */
       struct msm_mmu *mmu;
       struct kref kref;
 };
@@ -87,6 +88,7 @@ struct msm_gem_object {
        * an IOMMU.  Also used for stolen/splashscreen buffer.
        */
       struct drm_mm_node *vram_node;
+       struct mutex lock; /* Protects resources associated with bo */
 };
 #define to_msm_bo(x) container_of(x, struct msm_gem_object, base)
 
drivers/gpu/drm/msm/msm_gem_submit.c +2 -2
index 1c545eb..8420551 100644
@@ -245,7 +245,7 @@ static int submit_pin_objects(struct msm_gem_submit *submit)
               uint64_t iova;
 
               /* if locking succeeded, pin bo: */
-               ret = msm_gem_get_iova_locked(&msm_obj->base,
+               ret = msm_gem_get_iova(&msm_obj->base,
                               submit->gpu->id, &iova);
 
               if (ret)
@@ -301,7 +301,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
       /* For now, just map the entire thing.  Eventually we probably
        * to do it page-by-page, w/ kmap() if not vmap()d..
        */
-       ptr = msm_gem_get_vaddr_locked(&obj->base);
+       ptr = msm_gem_get_vaddr(&obj->base);
 
       if (IS_ERR(ptr)) {
               ret = PTR_ERR(ptr);
@@ -359,7 +359,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
       }
 

out:

drivers/gpu/drm/msm/msm_gem_submit.c +1 -1
-       msm_gem_put_vaddr_locked(&obj->base);
+       msm_gem_put_vaddr(&obj->base);
 
       return ret;

}

drivers/gpu/drm/msm/msm_gem_vma.c +9 -1
index f285d7e..c36321bc 100644
@@ -50,7 +50,9 @@ void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
               aspace->mmu->funcs->unmap(aspace->mmu, vma->iova, sgt, size);
       }
 
+       spin_lock(&aspace->lock);
       drm_mm_remove_node(&vma->node);
+       spin_unlock(&aspace->lock);
 
       vma->iova = 0;
 
@@ -63,10 +65,15 @@ void msm_gem_address_space_put(struct msm_gem_address_space *aspace)
 {
       int ret;
 
-       if (WARN_ON(drm_mm_node_allocated(&vma->node)))
+       spin_lock(&aspace->lock);
+       if (WARN_ON(drm_mm_node_allocated(&vma->node))) {
+               spin_unlock(&aspace->lock);
               return 0;
+       }
 
       ret = drm_mm_insert_node(&aspace->mm, &vma->node, npages);
+       spin_unlock(&aspace->lock);
+
       if (ret)
               return ret;
 
@@ -94,6 +101,7 @@ struct msm_gem_address_space *
       if (!aspace)
               return ERR_PTR(-ENOMEM);
 
+       spin_lock_init(&aspace->lock);
       aspace->name = name;
       aspace->mmu = msm_iommu_new(dev, domain);
 
drivers/gpu/drm/msm/msm_gpu.c +1 -3
index 97b9c38..d6e5cb2 100644
@@ -495,7 +495,7 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
               /* submit takes a reference to the bo and iova until retired: */
               drm_gem_object_reference(&msm_obj->base);
-               msm_gem_get_iova_locked(&msm_obj->base,
+               msm_gem_get_iova(&msm_obj->base,
                               submit->gpu->id, &iova);
 
               if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
@@ -662,9 +662,7 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 
 
       /* Create ringbuffer: */
-       mutex_lock(&drm->struct_mutex);
       gpu->rb = msm_ringbuffer_new(gpu, ringsz);
-       mutex_unlock(&drm->struct_mutex);
       if (IS_ERR(gpu->rb)) {
               ret = PTR_ERR(gpu->rb);
               gpu->rb = NULL;
drivers/gpu/drm/msm/msm_rd.c +2 -2
index 0e81faa..0366b80 100644
@@ -268,7 +268,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
       struct msm_gem_object *obj = submit->bos[idx].obj;
       const char *buf;
 
-       buf = msm_gem_get_vaddr_locked(&obj->base);
+       buf = msm_gem_get_vaddr(&obj->base);
       if (IS_ERR(buf))
               return;
 
@@ -283,7 +283,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
                       (uint32_t[3]){ iova, size, iova >> 32 }, 12);
       rd_write_section(rd, RD_BUFFER_CONTENTS, buf, size);
 
-       msm_gem_put_vaddr_locked(&obj->base);
+       msm_gem_put_vaddr(&obj->base);
 }
 
 /* called under struct_mutex */
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -2
index 67b34e0..791bca3 100644
@@ -40,7 +40,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int size)
               goto fail;
       }
 
-       ring->start = msm_gem_get_vaddr_locked(ring->bo);
+       ring->start = msm_gem_get_vaddr(ring->bo);
       if (IS_ERR(ring->start)) {
               ret = PTR_ERR(ring->start);
               goto fail;
--

1.9.1

Quoted 4 lines
Quoted 4 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-arm-msm" in
the body of a message to [email protected] <mailto:[email protected]> More majordomo info at http://vger.kernel.org/majordomo-info.html <http://vger.kernel.org/majordomo-info.html>

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno

#6 lore raw

I think the trick is what is the best way to synchronize access to msm_obj->madv in the shrinker path, since we can't reliably just take msm_obj->lock in shrinker path since we might already hold it:
https://hastebin.com/ubafepahov.xml

fwiw, that was with my work-in-progress attempt to deal with shrinker:
https://hastebin.com/hucijugufo.cs

Anyways, got distracted on other things and didn't get much further, but hopefully should have some time for another look tomorrow. Suggestions welcome ;-)

BR,
-R

Quoted 895 lines · Sushmita

Hi Rob,

Yes, I’d completely missed the shrinker path in this cleanup. But, yeah, I see how get_pages (which is called with msm_obj->lock held) -> drm_gem_get_pages could trigger shrinker_scan which calls msm_gem_purge. It makes sense to prevent any get_pages/vmap on objects that’ve been marked as DONTNEED. I’ll send you a patch soon for that.

Thanks,
Sushmita

On Jun 14, 2017, at 10:49 AM, Rob Clark <[email protected]> wrote:

On Tue, Jun 13, 2017 at 6:52 PM, Sushmita Susheelendra <[email protected]> wrote:

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.

I've rebased this on top of the other changes that are in the queue for 4.13:

https://github.com/freedreno/kernel-msm/commit/4e0c4e139a647914cfcf7c413da5c19f9f124885

I've done some light testing on a5xx.. although between this and moving gpu->hw_init() under struct_mutex, I need to double check on a3xx/a4xx.

I do think we probably need a bit more work for shrinker. In particular msm_gem_purge() still assumes everything is protected by a single struct_mutex, which is no longer true. The tricky thing here is that shrinker can be triggered by code-paths where we already hold msm_obj->lock.

I think we probably want anything that ends up in get_pages() or vmap() to have something along the lines of

if (WARN_ON(msm_obj->madv == DONTNEED)) {
mutex_unlock(&msm_obj->lock);
return -EINVAL; // or -EBUSY, or ??
}

it's really only something that a badly behaved userspace could triger.. but otoh we probably don't want to let a badly behaved userspace deadlock things in the kernel. I guess I probably should have written some test cases for this by now.

Change-Id: I4fba9f8c38a6cd13f80f660639d1e74d4336e3fb

jfyi, no need for the Change-Id.. I usually strip them out when I apply patches, but I appreciate if you can strip them out before sending (because sometimes I forget)

BR,
-R

---

drivers/gpu/drm/msm/adreno/a5xx_gpu.c    |   2 -
drivers/gpu/drm/msm/adreno/a5xx_power.c  |   2 -
drivers/gpu/drm/msm/adreno/adreno_gpu.c  |   2 -
drivers/gpu/drm/msm/dsi/dsi_host.c       |   5 +-
drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c |   2 +-
drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c  |   2 -
drivers/gpu/drm/msm/msm_drv.c            |   1 +
drivers/gpu/drm/msm/msm_drv.h            |   7 +-
drivers/gpu/drm/msm/msm_fbdev.c          |   6 +-
drivers/gpu/drm/msm/msm_gem.c            | 190
+++++++++++++++----------------
drivers/gpu/drm/msm/msm_gem.h            |   2 +
drivers/gpu/drm/msm/msm_gem_submit.c     |   6 +-
drivers/gpu/drm/msm/msm_gem_vma.c        |  10 +-
drivers/gpu/drm/msm/msm_gpu.c            |   4 +-
drivers/gpu/drm/msm/msm_rd.c             |   4 +-
drivers/gpu/drm/msm/msm_ringbuffer.c     |   2 +-
16 files changed, 120 insertions(+), 127 deletions(-)
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +0 -2
index 31a9bce..7893de1 100644
@@ -235,9 +235,7 @@ static struct drm_gem_object *a5xx_ucode_load_bo(struct
 msm_gpu *gpu,
       struct drm_gem_object *bo;
       void *ptr;
 
-       mutex_lock(&drm->struct_mutex);
       bo = msm_gem_new(drm, fw->size - 4, MSM_BO_UNCACHED);
-       mutex_unlock(&drm->struct_mutex);
 
       if (IS_ERR(bo))
               return bo;
drivers/gpu/drm/msm/adreno/a5xx_power.c +0 -2
index 72d52c7..eb88f44 100644
@@ -294,9 +294,7 @@ void a5xx_gpmu_ucode_init(struct msm_gpu *gpu)
        */
       bosize = (cmds_size + (cmds_size / TYPE4_MAX_PAYLOAD) + 1) << 2;
 
-       mutex_lock(&drm->struct_mutex);
       a5xx_gpu->gpmu_bo = msm_gem_new(drm, bosize, MSM_BO_UNCACHED);
-       mutex_unlock(&drm->struct_mutex);
 
       if (IS_ERR(a5xx_gpu->gpmu_bo))
               goto err;
drivers/gpu/drm/msm/adreno/adreno_gpu.c +0 -2
index 5b63fc6..1162c15 100644
@@ -392,10 +392,8 @@ 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 +1 -4
index 4f79b10..6a1b0da 100644
@@ -980,19 +980,16 @@ 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, 0,
 &iova);
-               mutex_unlock(&dev->struct_mutex);
+               ret = msm_gem_get_iova(msm_host->tx_gem_obj, 0, &iova);
               if (ret) {
                       pr_err("%s: failed to get iova, %d\n", __func__,
ret);
                       return ret;
drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +0 -0
index f29194a..15478f8 100644
@@ -372,7 +372,7 @@ 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, mdp4_kms->id,
&iova);

drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c +1 -0
+                       msm_gem_get_iova(next_bo, mdp4_kms->id, &iova);
 
                       /* enable cursor: */
                       mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_SIZE(dma),
drivers/gpu/drm/msm/mdp/mdp4/mdp4_kms.c +0 -2
index 6295204..3a464b3 100644
@@ -561,9 +561,7 @@ 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 +1 -0
index 87b5695..bb1f3ee 100644
@@ -349,6 +349,7 @@ 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 +1 -0
index 28b6f9b..567737d 100644
@@ -157,6 +157,7 @@ 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;
drivers/gpu/drm/msm/msm_drv.h +0 -4
@@ -209,8 +210,6 @@ 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, int id,
-               uint64_t *iova);
 int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova);
 uint64_t msm_gem_iova(struct drm_gem_object *obj, int id);
 struct page **msm_gem_get_pages(struct drm_gem_object *obj);
@@ -228,9 +227,7 @@ 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);

drivers/gpu/drm/msm/msm_drv.h +2 -0
@@ -247,6 +244,8 @@ 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 +1 -3
index 951e40f..785925e9 100644
@@ -95,10 +95,8 @@ 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;
@@ -124,7 +122,7 @@ 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, 0, &paddr);
+       ret = msm_gem_get_iova(fbdev->bo, 0, &paddr);
       if (ret) {
               dev_err(dev->dev, "failed to get buffer obj iova: %d\n",
ret);
               goto fail_unlock;
drivers/gpu/drm/msm/msm_fbdev.c +1 -1
@@ -153,7 +151,7 @@ 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;
drivers/gpu/drm/msm/msm_gem.c +3 -2
index 68e509b..1e803fb 100644
@@ -41,8 +41,7 @@ static bool use_pages(struct drm_gem_object *obj)
 }
 
 /* allocate pages from VRAM carveout, used when no IOMMU: */
-static struct page **get_pages_vram(struct drm_gem_object *obj,
-               int npages)
+static struct page **get_pages_vram(struct drm_gem_object *obj, int npages)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       struct msm_drm_private *priv = obj->dev->dev_private;
@@ -54,7 +53,9 @@ static struct page **get_pages_vram(struct drm_gem_object
 *obj,
       if (!p)
               return ERR_PTR(-ENOMEM);
 
+       spin_lock(&priv->vram.lock);
       ret = drm_mm_insert_node(&priv->vram.mm, msm_obj->vram_node, npages);
+       spin_unlock(&priv->vram.lock);
       if (ret) {
               drm_free_large(p);
               return ERR_PTR(ret);
drivers/gpu/drm/msm/msm_gem.c +0 -1
@@ -69,7 +70,6 @@ static struct page **get_pages_vram(struct drm_gem_object
 *obj,
       return p;
 }
 
-/* called with dev->struct_mutex held */
 static struct page **get_pages(struct drm_gem_object *obj)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
drivers/gpu/drm/msm/msm_gem.c +12 -0
@@ -109,6 +109,18 @@ static struct page **get_pages(struct drm_gem_object
 *obj)
       return msm_obj->pages;
 }
 
+static void put_pages_vram(struct drm_gem_object *obj)
+{
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+       struct msm_drm_private *priv = obj->dev->dev_private;
+
+       spin_lock(&priv->vram.lock);
+       drm_mm_remove_node(msm_obj->vram_node);
+       spin_unlock(&priv->vram.lock);
+
+       drm_free_large(msm_obj->pages);
+}
+
 static void put_pages(struct drm_gem_object *obj)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
drivers/gpu/drm/msm/msm_gem.c +11 -20
@@ -125,10 +137,8 @@ static void put_pages(struct drm_gem_object *obj)
 
               if (use_pages(obj))
                       drm_gem_put_pages(obj, msm_obj->pages, true, false);
-               else {
-                       drm_mm_remove_node(msm_obj->vram_node);
-                       drm_free_large(msm_obj->pages);
-               }
+               else
+                       put_pages_vram(obj);
 
               msm_obj->pages = NULL;
       }
@@ -136,11 +146,12 @@ static void put_pages(struct drm_gem_object *obj)
 
 struct page **msm_gem_get_pages(struct drm_gem_object *obj)
 {
-       struct drm_device *dev = obj->dev;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       struct page **p;
-       mutex_lock(&dev->struct_mutex);
+
+       mutex_lock(&msm_obj->lock);
       p = get_pages(obj);
-       mutex_unlock(&dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
       return p;
 }
 
@@ -195,25 +206,17 @@ int msm_gem_fault(struct vm_fault *vmf)
 {
       struct vm_area_struct *vma = vmf->vma;
       struct drm_gem_object *obj = vma->vm_private_data;
-       struct drm_device *dev = obj->dev;
-       struct msm_drm_private *priv = dev->dev_private;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       struct page **pages;
       unsigned long pfn;
       pgoff_t pgoff;
       int ret;
 
-       /* This should only happen if userspace tries to pass a mmap'd
-        * but unfaulted gem bo vaddr into submit ioctl, triggering
-        * a page fault while struct_mutex is already held.  This is
-        * not a valid use-case so just bail.
+       /*
+        * vm_ops.open/drm_gem_mmap_obj and close get and put
+        * a reference on obj. So, we dont need to hold one here.
        */
-       if (priv->struct_mutex_task == current)
-               return VM_FAULT_SIGBUS;
-
-       /* Make sure we don't parallel update on a fault, nor move or remove
-        * something from beneath our feet
-        */
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
+       ret = mutex_lock_interruptible(&msm_obj->lock);
       if (ret)
               goto out;
 
@@ -235,7 +238,7 @@ int msm_gem_fault(struct vm_fault *vmf)
       ret = vm_insert_mixed(vma, vmf->address, __pfn_to_pfn_t(pfn,
 PFN_DEV));
 

out_unlock:

drivers/gpu/drm/msm/msm_gem.c +1 -1
-       mutex_unlock(&dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);

out:

       switch (ret) {
       case -EAGAIN:
drivers/gpu/drm/msm/msm_gem.c +7 -3
@@ -259,9 +262,10 @@ int msm_gem_fault(struct vm_fault *vmf)
 static uint64_t mmap_offset(struct drm_gem_object *obj)
 {
       struct drm_device *dev = obj->dev;
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       int ret;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
 
       /* Make it mmapable */
       ret = drm_gem_create_mmap_offset(obj);
@@ -277,21 +281,23 @@ static uint64_t mmap_offset(struct drm_gem_object
 *obj)
 uint64_t msm_gem_mmap_offset(struct drm_gem_object *obj)
 {
       uint64_t offset;
-       mutex_lock(&obj->dev->struct_mutex);
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+
+       mutex_lock(&msm_obj->lock);
       offset = mmap_offset(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
       return offset;
 }
 
+/* Called with msm_obj->lock locked */

static void
put_iova(struct drm_gem_object *obj)
{

drivers/gpu/drm/msm/msm_gem.c +8 -11
-       struct drm_device *dev = obj->dev;
       struct msm_drm_private *priv = obj->dev->dev_private;
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       int id;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
 
       for (id = 0; id < ARRAY_SIZE(msm_obj->domain); id++) {
               if (!priv->aspace[id])
@@ -301,25 +307,23 @@ uint64_t msm_gem_mmap_offset(struct drm_gem_object
 *obj)
       }
 }
 
-/* should be called under struct_mutex.. although it can be called
- * from atomic context without struct_mutex to acquire an extra
- * iova ref if you know one is already held.
- *
- * That means when I do eventually need to add support for unpinning
- * the refcnt counter needs to be atomic_t.
- */
-int msm_gem_get_iova_locked(struct drm_gem_object *obj, int id,
+/* A reference to obj must be held before calling this function. */
+int msm_gem_get_iova(struct drm_gem_object *obj, int id,
               uint64_t *iova)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
       int ret = 0;
 
+       mutex_lock(&msm_obj->lock);
+
       if (!msm_obj->domain[id].iova) {
               struct msm_drm_private *priv = obj->dev->dev_private;
               struct page **pages = get_pages(obj);
 
-               if (IS_ERR(pages))
+               if (IS_ERR(pages)) {
+                       mutex_unlock(&msm_obj->lock);
                       return PTR_ERR(pages);
+               }
 
               if (iommu_present(&platform_bus_type)) {
                       ret = msm_gem_map_vma(priv->aspace[id],

&msm_obj->domain[id],

drivers/gpu/drm/msm/msm_gem.c +15 -42
@@ -332,26 +336,7 @@ int msm_gem_get_iova_locked(struct drm_gem_object *obj,
 int id,
       if (!ret)
               *iova = msm_obj->domain[id].iova;
 
-       return ret;
-}
-
-/* get iova, taking a reference.  Should have a matching put */
-int msm_gem_get_iova(struct drm_gem_object *obj, int id, uint64_t *iova)
-{
-       struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       int ret;
-
-       /* this is safe right now because we don't unmap until the
-        * bo is deleted:
-        */
-       if (msm_obj->domain[id].iova) {
-               *iova = msm_obj->domain[id].iova;
-               return 0;
-       }
-
-       mutex_lock(&obj->dev->struct_mutex);
-       ret = msm_gem_get_iova_locked(obj, id, iova);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
       return ret;
 }
@@ -405,45 +390,37 @@ int msm_gem_dumb_map_offset(struct drm_file *file,
 struct drm_device *dev,
       return ret;
 }
 
-void *msm_gem_get_vaddr_locked(struct drm_gem_object *obj)
+void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+
+       mutex_lock(&msm_obj->lock);
       if (!msm_obj->vaddr) {
               struct page **pages = get_pages(obj);
-               if (IS_ERR(pages))
+               if (IS_ERR(pages)) {
+                       mutex_unlock(&msm_obj->lock);
                       return ERR_CAST(pages);
+               }
               msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
                               VM_MAP, pgprot_writecombine(PAGE_KERNEL));
-               if (msm_obj->vaddr == NULL)
+               if (msm_obj->vaddr == NULL) {
+                       mutex_unlock(&msm_obj->lock);
                       return ERR_PTR(-ENOMEM);
+               }
       }
       msm_obj->vmap_count++;
+       mutex_unlock(&msm_obj->lock);
       return msm_obj->vaddr;
 }
 
-void *msm_gem_get_vaddr(struct drm_gem_object *obj)
-{
-       void *ret;
-       mutex_lock(&obj->dev->struct_mutex);
-       ret = msm_gem_get_vaddr_locked(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
-       return ret;
-}
-
-void msm_gem_put_vaddr_locked(struct drm_gem_object *obj)
+void msm_gem_put_vaddr(struct drm_gem_object *obj)
 {
       struct msm_gem_object *msm_obj = to_msm_bo(obj);
-       WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+
+       mutex_lock(&msm_obj->lock);
       WARN_ON(msm_obj->vmap_count < 1);
       msm_obj->vmap_count--;
-}
-
-void msm_gem_put_vaddr(struct drm_gem_object *obj)
-{
-       mutex_lock(&obj->dev->struct_mutex);
-       msm_gem_put_vaddr_locked(obj);
-       mutex_unlock(&obj->dev->struct_mutex);
+       mutex_unlock(&msm_obj->lock);
 }
 

/* Update madvise status, returns true if not purged, else

drivers/gpu/drm/msm/msm_gem.c +5 -7
@@ -697,6 +674,8 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
       list_del(&msm_obj->mm_list);
 
+       mutex_lock(&msm_obj->lock);
+
       put_iova(obj);
 
       if (obj->import_attach) {
@@ -720,6 +699,7 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
       drm_gem_object_release(obj);
 
+       mutex_unlock(&msm_obj->lock);
       kfree(msm_obj);
 }
 
@@ -730,14 +710,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct
 drm_file *file,
       struct drm_gem_object *obj;
       int ret;
 
-       ret = mutex_lock_interruptible(&dev->struct_mutex);
-       if (ret)
-               return ret;
-
       obj = msm_gem_new(dev, size, flags);
 
-       mutex_unlock(&dev->struct_mutex);
-
       if (IS_ERR(obj))
               return PTR_ERR(obj);
@@ -752,7 +726,8 @@ int msm_gem_new_handle(struct drm_device *dev, struct
 drm_file *file,
 static int msm_gem_new_impl(struct drm_device *dev,
               uint32_t size, uint32_t flags,
               struct reservation_object *resv,
-               struct drm_gem_object **obj)
+               struct drm_gem_object **obj,
+               bool struct_mutex_locked)
 {
       struct msm_drm_private *priv = dev->dev_private;
       struct msm_gem_object *msm_obj;
drivers/gpu/drm/msm/msm_gem.c +12 -6
@@ -781,6 +756,8 @@ static int msm_gem_new_impl(struct drm_device *dev,
       if (!msm_obj)
               return -ENOMEM;
 
+       mutex_init(&msm_obj->lock);
+
       if (use_vram)
               msm_obj->vram_node = &msm_obj->domain[0].node;
 
@@ -795,21 +772,25 @@ static int msm_gem_new_impl(struct drm_device *dev,
       }
 
       INIT_LIST_HEAD(&msm_obj->submit_entry);
-       list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+       if (struct_mutex_locked) {
+               list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+       } else {
+               mutex_lock(&dev->struct_mutex);
+               list_add_tail(&msm_obj->mm_list, &priv->inactive_list);
+               mutex_unlock(&dev->struct_mutex);
+       }
 
       *obj = &msm_obj->base;
 
       return 0;
 }
 
-struct drm_gem_object *msm_gem_new(struct drm_device *dev,
-               uint32_t size, uint32_t flags)
+static struct drm_gem_object *_msm_gem_new(struct drm_device *dev,
+               uint32_t size, uint32_t flags, bool struct_mutex_locked)
 {
       struct drm_gem_object *obj = NULL;
       int ret;
 
-       WARN_ON(!mutex_is_locked(&dev->struct_mutex));
-
       size = PAGE_ALIGN(size);
 
       /* Disallow zero sized objects as they make the underlying
@@ -818,7 +799,7 @@ struct drm_gem_object *msm_gem_new(struct drm_device
 *dev,
       if (size == 0)
               return ERR_PTR(-EINVAL);
 
-       ret = msm_gem_new_impl(dev, size, flags, NULL, &obj);
+       ret = msm_gem_new_impl(dev, size, flags, NULL, &obj,
 struct_mutex_locked);
       if (ret)
               goto fail;
drivers/gpu/drm/msm/msm_gem.c +0 -0
@@ -833,10 +814,22 @@ struct drm_gem_object *msm_gem_new(struct drm_device
 *dev,
       return obj;
 

fail:

drivers/gpu/drm/msm/msm_gem.c +1 -1
-       drm_gem_object_unreference(obj);
+       drm_gem_object_unreference_unlocked(obj);
       return ERR_PTR(ret);

}

drivers/gpu/drm/msm/msm_gem.c +12 -0
+struct drm_gem_object *msm_gem_new_locked(struct drm_device *dev,
+               uint32_t size, uint32_t flags)
+{
+       return _msm_gem_new(dev, size, flags, true);
+}
+
+struct drm_gem_object *msm_gem_new(struct drm_device *dev,
+               uint32_t size, uint32_t flags)
+{
+       return _msm_gem_new(dev, size, flags, false);
+}
+

struct drm_gem_object *msm_gem_import(struct drm_device *dev,
struct dma_buf *dmabuf, struct sg_table *sgt)
{

drivers/gpu/drm/msm/msm_gem.c +1 -1
@@ -853,7 +846,7 @@ struct drm_gem_object *msm_gem_import(struct drm_device
 *dev,
 
       size = PAGE_ALIGN(dmabuf->size);
 
-       ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj);
+       ret = msm_gem_new_impl(dev, size, MSM_BO_WC, dmabuf->resv, &obj,
 false);
       if (ret)
               goto fail;
drivers/gpu/drm/msm/msm_gem.c +6 -1
@@ -862,17 +855,22 @@ struct drm_gem_object *msm_gem_import(struct
 drm_device *dev,
       npages = size / PAGE_SIZE;
 
       msm_obj = to_msm_bo(obj);
+       mutex_lock(&msm_obj->lock);
       msm_obj->sgt = sgt;
       msm_obj->pages = drm_malloc_ab(npages, sizeof(struct page *));
       if (!msm_obj->pages) {
+               mutex_unlock(&msm_obj->lock);
               ret = -ENOMEM;
               goto fail;
       }
 
       ret = drm_prime_sg_to_page_addr_arrays(sgt, msm_obj->pages, NULL,
 npages);
-       if (ret)
+       if (ret) {
+               mutex_unlock(&msm_obj->lock);
               goto fail;
+       }
 
+       mutex_unlock(&msm_obj->lock);
       return obj;

fail:

drivers/gpu/drm/msm/msm_gem.h +2 -0
index 1b4cf20..b1bfc10 100644
@@ -31,6 +31,7 @@ struct msm_gem_address_space {
        * and position mm_node->start is in # of pages:
        */
       struct drm_mm mm;
+       spinlock_t lock; /* Protects drm_mm node allocation/removal */
       struct msm_mmu *mmu;
       struct kref kref;
 };
@@ -87,6 +88,7 @@ struct msm_gem_object {
        * an IOMMU.  Also used for stolen/splashscreen buffer.
        */
       struct drm_mm_node *vram_node;
+       struct mutex lock; /* Protects resources associated with bo */
 };
 #define to_msm_bo(x) container_of(x, struct msm_gem_object, base)
 
drivers/gpu/drm/msm/msm_gem_submit.c +1 -1
index 1c545eb..8420551 100644
@@ -245,7 +245,7 @@ static int submit_pin_objects(struct msm_gem_submit
 *submit)
               uint64_t iova;
 
               /* if locking succeeded, pin bo: */
-               ret = msm_gem_get_iova_locked(&msm_obj->base,
+               ret = msm_gem_get_iova(&msm_obj->base,
                               submit->gpu->id, &iova);
 

if (ret)

drivers/gpu/drm/msm/msm_gem_submit.c +0 -0
@@ -301,7 +301,7 @@ static int submit_reloc(struct msm_gem_submit *submit,

struct msm_gem_object *ob

       /* For now, just map the entire thing.  Eventually we probably
        * to do it page-by-page, w/ kmap() if not vmap()d..
        */
drivers/gpu/drm/msm/msm_gem_submit.c +1 -1
-       ptr = msm_gem_get_vaddr_locked(&obj->base);
+       ptr = msm_gem_get_vaddr(&obj->base);
 
       if (IS_ERR(ptr)) {
               ret = PTR_ERR(ptr);
@@ -359,7 +359,7 @@ static int submit_reloc(struct msm_gem_submit *submit,

struct msm_gem_object *ob

       }

out:

drivers/gpu/drm/msm/msm_gem_submit.c +1 -1
-       msm_gem_put_vaddr_locked(&obj->base);
+       msm_gem_put_vaddr(&obj->base);
 
       return ret;

}

drivers/gpu/drm/msm/msm_gem_vma.c +9 -1
index f285d7e..c36321bc 100644
@@ -50,7 +50,9 @@ void msm_gem_address_space_put(struct
 msm_gem_address_space *aspace)
               aspace->mmu->funcs->unmap(aspace->mmu, vma->iova, sgt, size);
       }
 
+       spin_lock(&aspace->lock);
       drm_mm_remove_node(&vma->node);
+       spin_unlock(&aspace->lock);
 
       vma->iova = 0;
@@ -63,10 +65,15 @@ void msm_gem_address_space_put(struct
 msm_gem_address_space *aspace)
 {
       int ret;
 
-       if (WARN_ON(drm_mm_node_allocated(&vma->node)))
+       spin_lock(&aspace->lock);
+       if (WARN_ON(drm_mm_node_allocated(&vma->node))) {
+               spin_unlock(&aspace->lock);
               return 0;
+       }
 
       ret = drm_mm_insert_node(&aspace->mm, &vma->node, npages);
+       spin_unlock(&aspace->lock);
+
       if (ret)
               return ret;
@@ -94,6 +101,7 @@ struct msm_gem_address_space *
       if (!aspace)
               return ERR_PTR(-ENOMEM);
 
+       spin_lock_init(&aspace->lock);
       aspace->name = name;
       aspace->mmu = msm_iommu_new(dev, domain);
 
drivers/gpu/drm/msm/msm_gpu.c +1 -1
index 97b9c38..d6e5cb2 100644
@@ -495,7 +495,7 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct
 msm_gem_submit *submit,
 
               /* submit takes a reference to the bo and iova until retired:
 */
               drm_gem_object_reference(&msm_obj->base);
-               msm_gem_get_iova_locked(&msm_obj->base,
+               msm_gem_get_iova(&msm_obj->base,
                               submit->gpu->id, &iova);
               if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
drivers/gpu/drm/msm/msm_gpu.c +0 -2
@@ -662,9 +662,7 @@ int msm_gpu_init(struct drm_device *drm, struct
 platform_device *pdev,
 
 
       /* Create ringbuffer: */
-       mutex_lock(&drm->struct_mutex);
       gpu->rb = msm_ringbuffer_new(gpu, ringsz);
-       mutex_unlock(&drm->struct_mutex);
       if (IS_ERR(gpu->rb)) {
               ret = PTR_ERR(gpu->rb);
               gpu->rb = NULL;
drivers/gpu/drm/msm/msm_rd.c +2 -2
index 0e81faa..0366b80 100644
@@ -268,7 +268,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
       struct msm_gem_object *obj = submit->bos[idx].obj;
       const char *buf;
 
-       buf = msm_gem_get_vaddr_locked(&obj->base);
+       buf = msm_gem_get_vaddr(&obj->base);
       if (IS_ERR(buf))
               return;
 
@@ -283,7 +283,7 @@ static void snapshot_buf(struct msm_rd_state *rd,
                       (uint32_t[3]){ iova, size, iova >> 32 }, 12);
       rd_write_section(rd, RD_BUFFER_CONTENTS, buf, size);
 
-       msm_gem_put_vaddr_locked(&obj->base);
+       msm_gem_put_vaddr(&obj->base);
 }
 
 /* called under struct_mutex */
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -1
index 67b34e0..791bca3 100644
@@ -40,7 +40,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu
 *gpu, int size)
               goto fail;
       }
 
-       ring->start = msm_gem_get_vaddr_locked(ring->bo);
+       ring->start = msm_gem_get_vaddr(ring->bo);
       if (IS_ERR(ring->start)) {
               ret = PTR_ERR(ring->start);
               goto fail;
Quoted 9 lines

--
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-arm-msm" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#7 lore raw
In reply to a message not in the archive

Quoting Rob Clark (2017-06-14 17:49:02)

Quoted 18 lines
Quoted 7 lines · Sushmita Susheelendra

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.

I've rebased this on top of the other changes that are in the queue for 4.13:

https://github.com/freedreno/kernel-msm/commit/4e0c4e139a647914cfcf7c413da5c19f9f124885

I've done some light testing on a5xx.. although between this and moving gpu->hw_init() under struct_mutex, I need to double check on a3xx/a4xx.

Quoted 5 lines

I do think we probably need a bit more work for shrinker. In particular msm_gem_purge() still assumes everything is protected by a single struct_mutex, which is no longer true. The tricky thing here is that shrinker can be triggered by code-paths where we already hold msm_obj->lock.

Make sure that msm_obj->lock *only* covers the backing store (and its operations). (e.g. I called it obj->mm.lock and moved everything that it covers into obj->mm). Then you can ensure that you don't attempt to shrink msm_obj itself whilst holding msm_obj->mm.lock (e.g. by tracking a obj->mm.pages_pin_count and not attempting to shrink any object that has its pages still pinned, and you then only allocate underneath the msm_obj->mm.lock whilst holding a page reference). After all that you can use mutex_lock_nested(.subclass = MSM_SHRINKER) for the special case where the shrinker may be acquiring the lock on other objects whilst holding msm_obj->mm.lock.
-Chris
_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno

#8 lore raw

PATCH 1/1 fixup! drm/msm: Separate locking of buffer resources from struct_mutex

In reply to Sushmita Susheelendra

---
This is roughly based on Chris's suggestion, in particular the part about using mutex_lock_nested(). It's not *exactly* the same, in particular msm_obj->lock protects a bit more than just backing store and we don't currently track a pin_count. (Instead we currently keep pages pinned until the object is purged or freed.)

Instead of making msm_obj->lock only cover backing store, it is easier to split out madv, which is still protected by struct_mutex, which is still held by the shrinker, so the shrinker does not need to grab msm_obj->lock until it purges an object. We avoid going down any path that could trigger shrinker by ensuring that msm_obj->madv == WILLNEED. To synchronize access to msm_obj->madv it is protected by msm_obj->lock inside struct_mutex.

This seems to keep lockdep happy in my testing so far.

 drivers/gpu/drm/msm/msm_gem.c          | 54 ++++++++++++++++++++++++++++++++--
 drivers/gpu/drm/msm/msm_gem.h          |  1 +
 drivers/gpu/drm/msm/msm_gem_shrinker.c | 12 ++++++++
 3 files changed, 65 insertions(+), 2 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +52 -2
index e132548..f5d1f84 100644
@@ -26,6 +26,22 @@
 #include "msm_gpu.h"
 #include "msm_mmu.h"
 
+/* The shrinker can be triggered while we hold objA->lock, and need
+ * to grab objB->lock to purge it.  Lockdep just sees these as a single
+ * class of lock, so we use subclasses to teach it the difference.
+ *
+ * OBJ_LOCK_NORMAL is implicit (ie. normal mutex_lock() call), and
+ * OBJ_LOCK_SHRINKER is used in msm_gem_purge().
+ *
+ * It is *essential* that we never go down paths that could trigger the
+ * shrinker for a purgable object.  This is ensured by checking that
+ * msm_obj->madv == MSM_MADV_WILLNEED.
+ */
+enum {
+	OBJ_LOCK_NORMAL,
+	OBJ_LOCK_SHRINKER,
+};
+
 static dma_addr_t physaddr(struct drm_gem_object *obj)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -150,6 +166,12 @@ struct page **msm_gem_get_pages(struct drm_gem_object *obj)
 	struct page **p;
 
 	mutex_lock(&msm_obj->lock);
+
+	if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+		mutex_unlock(&msm_obj->lock);
+		return ERR_PTR(-EBUSY);
+	}
+
 	p = get_pages(obj);
 	mutex_unlock(&msm_obj->lock);
 	return p;
@@ -220,6 +242,11 @@ int msm_gem_fault(struct vm_fault *vmf)
 	if (ret)
 		goto out;
 
+	if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+		mutex_unlock(&msm_obj->lock);
+		return VM_FAULT_SIGBUS;
+	}
+
 	/* make sure we have pages attached now */
 	pages = get_pages(obj);
 	if (IS_ERR(pages)) {
@@ -358,6 +385,11 @@ int msm_gem_get_iova(struct drm_gem_object *obj,
 
 	mutex_lock(&msm_obj->lock);
 
+	if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+		mutex_unlock(&msm_obj->lock);
+		return -EBUSY;
+	}
+
 	vma = lookup_vma(obj, aspace);
 
 	if (!vma) {
@@ -454,6 +486,12 @@ void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
 	mutex_lock(&msm_obj->lock);
+
+	if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+		mutex_unlock(&msm_obj->lock);
+		return ERR_PTR(-EBUSY);
+	}
+
 	if (!msm_obj->vaddr) {
 		struct page **pages = get_pages(obj);
 		if (IS_ERR(pages)) {
@@ -489,12 +527,18 @@ int msm_gem_madvise(struct drm_gem_object *obj, unsigned madv)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
+	mutex_lock(&msm_obj->lock);
+
 	WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
 
 	if (msm_obj->madv != __MSM_MADV_PURGED)
 		msm_obj->madv = madv;
 
-	return (msm_obj->madv != __MSM_MADV_PURGED);
+	madv = msm_obj->madv;
+
+	mutex_unlock(&msm_obj->lock);
+
+	return (madv != __MSM_MADV_PURGED);
 }
 
 void msm_gem_purge(struct drm_gem_object *obj)
@@ -506,6 +550,8 @@ void msm_gem_purge(struct drm_gem_object *obj)
 	WARN_ON(!is_purgeable(msm_obj));
 	WARN_ON(obj->import_attach);
 
+	mutex_lock_nested(&msm_obj->lock, OBJ_LOCK_SHRINKER);
+
 	put_iova(obj);
 
 	msm_gem_vunmap(obj);
@@ -526,6 +572,8 @@ void msm_gem_purge(struct drm_gem_object *obj)
 
 	invalidate_mapping_pages(file_inode(obj->filp)->i_mapping,
 			0, (loff_t)-1);
+
+	mutex_unlock(&msm_obj->lock);
 }
 
 void msm_gem_vunmap(struct drm_gem_object *obj)
@@ -660,7 +708,7 @@ void msm_gem_describe(struct drm_gem_object *obj, struct seq_file *m)
 	uint64_t off = drm_vma_node_start(&obj->vma_node);
 	const char *madv;
 
-	WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+	mutex_lock(&msm_obj->lock);
 
 	switch (msm_obj->madv) {
 	case __MSM_MADV_PURGED:
@@ -701,6 +749,8 @@ void msm_gem_describe(struct drm_gem_object *obj, struct seq_file *m)
 	if (fence)
 		describe_fence(fence, "Exclusive", m);
 	rcu_read_unlock();
+
+	mutex_unlock(&msm_obj->lock);
 }
 
 void msm_gem_describe_objects(struct list_head *list, struct seq_file *m)
drivers/gpu/drm/msm/msm_gem.h +1 -0
index 9ad5ba4c..2b9b8e9 100644
@@ -101,6 +101,7 @@ static inline bool is_active(struct msm_gem_object *msm_obj)
 
 static inline bool is_purgeable(struct msm_gem_object *msm_obj)
 {
+	WARN_ON(!mutex_is_locked(&msm_obj->base.dev->struct_mutex));
 	return (msm_obj->madv == MSM_MADV_DONTNEED) && msm_obj->sgt &&
 			!msm_obj->base.dma_buf && !msm_obj->base.import_attach;
 }
drivers/gpu/drm/msm/msm_gem_shrinker.c +12 -0
index ab1dd02..e1db4ad 100644
@@ -20,6 +20,18 @@
 
 static bool msm_gem_shrinker_lock(struct drm_device *dev, bool *unlock)
 {
+	/* NOTE: we are *closer* to being able to get rid of
+	 * mutex_trylock_recursive().. the msm_gem code itself does
+	 * not need struct_mutex, although codepaths that can trigger
+	 * shrinker are still called in code-paths that hold the
+	 * struct_mutex.
+	 *
+	 * Also, msm_obj->madv is protected by struct_mutex.
+	 *
+	 * The next step is probably split out a seperate lock for
+	 * protecting inactive_list, so that shrinker does not need
+	 * struct_mutex.
+	 */
 	switch (mutex_trylock_recursive(&dev->struct_mutex)) {
 	case MUTEX_TRYLOCK_FAILED:
 		return false;
Signature · 1 line
2.9.4

A message that is not in the archive (sent off-list or not yet received): 20170615132050.1196-1-robdclark-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org

#10 lore raw

Re: [PATCH] fixup! drm/msm: Separate locking of buffer resources from struct_mutex

Hi Rob,

I can see how we can trigger the shrinker on objB while holding objA->lock. So, the nested lock with class SHRINKER makes sense.
However, I’m trying to figure how the get_pages/vmap/fault path on an objA can end up triggering the shrinker on objA itself. As objA itself would not be purgeable (msm_obj->sgt would not be set)/vunmappable (msm_obj->vaddr would not be set) yet at that point, we would never end up calling msm_gem_purge or msm_gem_vunmap on objA itself right? If that is the case, we may not need the (msm_obj->madv == MSM_MADV_WILLNEED) check? Or am I missing something here?

I think shrinker_vmap would also need the nested SHRINKER lock before it calls msm_gem_vunmap because a vmap on objA could trigger msm_gem_vunmap on objB.

I really like the idea of protecting priv->inactive_list with a separate lock as that is pretty much why the shrinker needs to hold struct_mutex.

Thanks,
Sushmita

Quoted 195 lines · Rob Clark

---
This is roughly based on Chris's suggestion, in particular the part about using mutex_lock_nested(). It's not *exactly* the same, in particular msm_obj->lock protects a bit more than just backing store and we don't currently track a pin_count. (Instead we currently keep pages pinned until the object is purged or freed.)

Instead of making msm_obj->lock only cover backing store, it is easier to split out madv, which is still protected by struct_mutex, which is still held by the shrinker, so the shrinker does not need to grab msm_obj->lock until it purges an object. We avoid going down any path that could trigger shrinker by ensuring that msm_obj->madv == WILLNEED. To synchronize access to msm_obj->madv it is protected by msm_obj->lock inside struct_mutex.

This seems to keep lockdep happy in my testing so far.

drivers/gpu/drm/msm/msm_gem.c          | 54 ++++++++++++++++++++++++++++++++--
drivers/gpu/drm/msm/msm_gem.h          |  1 +
drivers/gpu/drm/msm/msm_gem_shrinker.c | 12 ++++++++
3 files changed, 65 insertions(+), 2 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +52 -2
index e132548..f5d1f84 100644
@@ -26,6 +26,22 @@
 #include "msm_gpu.h"
 #include "msm_mmu.h"
 
+/* The shrinker can be triggered while we hold objA->lock, and need
+ * to grab objB->lock to purge it.  Lockdep just sees these as a single
+ * class of lock, so we use subclasses to teach it the difference.
+ *
+ * OBJ_LOCK_NORMAL is implicit (ie. normal mutex_lock() call), and
+ * OBJ_LOCK_SHRINKER is used in msm_gem_purge().
+ *
+ * It is *essential* that we never go down paths that could trigger the
+ * shrinker for a purgable object.  This is ensured by checking that
+ * msm_obj->madv == MSM_MADV_WILLNEED.
+ */
+enum {
+	OBJ_LOCK_NORMAL,
+	OBJ_LOCK_SHRINKER,
+};
+
 static dma_addr_t physaddr(struct drm_gem_object *obj)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -150,6 +166,12 @@ struct page **msm_gem_get_pages(struct drm_gem_object *obj)
 	struct page **p;
 
 	mutex_lock(&msm_obj->lock);
+
+	if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+		mutex_unlock(&msm_obj->lock);
+		return ERR_PTR(-EBUSY);
+	}
+
 	p = get_pages(obj);
 	mutex_unlock(&msm_obj->lock);
 	return p;
@@ -220,6 +242,11 @@ int msm_gem_fault(struct vm_fault *vmf)
 	if (ret)
 		goto out;
 
+	if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+		mutex_unlock(&msm_obj->lock);
+		return VM_FAULT_SIGBUS;
+	}
+
 	/* make sure we have pages attached now */
 	pages = get_pages(obj);
 	if (IS_ERR(pages)) {
@@ -358,6 +385,11 @@ int msm_gem_get_iova(struct drm_gem_object *obj,
 
 	mutex_lock(&msm_obj->lock);
 
+	if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+		mutex_unlock(&msm_obj->lock);
+		return -EBUSY;
+	}
+
 	vma = lookup_vma(obj, aspace);
 
 	if (!vma) {
@@ -454,6 +486,12 @@ void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
 	mutex_lock(&msm_obj->lock);
+
+	if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+		mutex_unlock(&msm_obj->lock);
+		return ERR_PTR(-EBUSY);
+	}
+
 	if (!msm_obj->vaddr) {
 		struct page **pages = get_pages(obj);
 		if (IS_ERR(pages)) {
@@ -489,12 +527,18 @@ int msm_gem_madvise(struct drm_gem_object *obj, unsigned madv)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
+	mutex_lock(&msm_obj->lock);
+
 	WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
 
 	if (msm_obj->madv != __MSM_MADV_PURGED)
 		msm_obj->madv = madv;
 
-	return (msm_obj->madv != __MSM_MADV_PURGED);
+	madv = msm_obj->madv;
+
+	mutex_unlock(&msm_obj->lock);
+
+	return (madv != __MSM_MADV_PURGED);
 }
 
 void msm_gem_purge(struct drm_gem_object *obj)
@@ -506,6 +550,8 @@ void msm_gem_purge(struct drm_gem_object *obj)
 	WARN_ON(!is_purgeable(msm_obj));
 	WARN_ON(obj->import_attach);
 
+	mutex_lock_nested(&msm_obj->lock, OBJ_LOCK_SHRINKER);
+
 	put_iova(obj);
 
 	msm_gem_vunmap(obj);
@@ -526,6 +572,8 @@ void msm_gem_purge(struct drm_gem_object *obj)
 
 	invalidate_mapping_pages(file_inode(obj->filp)->i_mapping,
 			0, (loff_t)-1);
+
+	mutex_unlock(&msm_obj->lock);
 }
 
 void msm_gem_vunmap(struct drm_gem_object *obj)
@@ -660,7 +708,7 @@ void msm_gem_describe(struct drm_gem_object *obj, struct seq_file *m)
 	uint64_t off = drm_vma_node_start(&obj->vma_node);
 	const char *madv;
 
-	WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+	mutex_lock(&msm_obj->lock);
 
 	switch (msm_obj->madv) {
 	case __MSM_MADV_PURGED:
@@ -701,6 +749,8 @@ void msm_gem_describe(struct drm_gem_object *obj, struct seq_file *m)
 	if (fence)
 		describe_fence(fence, "Exclusive", m);
 	rcu_read_unlock();
+
+	mutex_unlock(&msm_obj->lock);
 }
 
 void msm_gem_describe_objects(struct list_head *list, struct seq_file *m)
drivers/gpu/drm/msm/msm_gem.h +1 -0
index 9ad5ba4c..2b9b8e9 100644
@@ -101,6 +101,7 @@ static inline bool is_active(struct msm_gem_object *msm_obj)
 
 static inline bool is_purgeable(struct msm_gem_object *msm_obj)
 {
+	WARN_ON(!mutex_is_locked(&msm_obj->base.dev->struct_mutex));
 	return (msm_obj->madv == MSM_MADV_DONTNEED) && msm_obj->sgt &&
 			!msm_obj->base.dma_buf && !msm_obj->base.import_attach;
 }
drivers/gpu/drm/msm/msm_gem_shrinker.c +12 -0
index ab1dd02..e1db4ad 100644
@@ -20,6 +20,18 @@
 
 static bool msm_gem_shrinker_lock(struct drm_device *dev, bool *unlock)
 {
+	/* NOTE: we are *closer* to being able to get rid of
+	 * mutex_trylock_recursive().. the msm_gem code itself does
+	 * not need struct_mutex, although codepaths that can trigger
+	 * shrinker are still called in code-paths that hold the
+	 * struct_mutex.
+	 *
+	 * Also, msm_obj->madv is protected by struct_mutex.
+	 *
+	 * The next step is probably split out a seperate lock for
+	 * protecting inactive_list, so that shrinker does not need
+	 * struct_mutex.
+	 */
 	switch (mutex_trylock_recursive(&dev->struct_mutex)) {
 	case MUTEX_TRYLOCK_FAILED:
 		return false;
Signature · 1 line
2.9.4

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno

A message that is not in the archive (sent off-list or not yet received): D38DFC6B-2CE4-459F-9D35-C21CEE5B362D-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org

#12 lore raw

Re: [PATCH] fixup! drm/msm: Separate locking of buffer resources from struct_mutex

Quoted 11 lines · Sushmita

Hi Rob,

I can see how we can trigger the shrinker on objB while holding objA->lock. So, the nested lock with class SHRINKER makes sense.
However, I’m trying to figure how the get_pages/vmap/fault path on an objA can end up triggering the shrinker on objA itself. As objA itself would not be purgeable (msm_obj->sgt would not be set)/vunmappable (msm_obj->vaddr would not be set) yet at that point, we would never end up calling msm_gem_purge or msm_gem_vunmap on objA itself right? If that is the case, we may not need the (msm_obj->madv == MSM_MADV_WILLNEED) check? Or am I missing something here?

get_pages() would set msm_obj->sgt.. I guess that is protected by msm_obj->lock, so maybe it would be safe to drop the WILLNEED check. Otoh, I think that does make things more clear to include the check and a bit more future-proof.

We do check is_purgable() outside of msm_obj->lock.. I'd have to think about it for more than a few seconds, but it seems like keeping the WILLNEED check is a good idea.

Quoted 3 lines

I think shrinker_vmap would also need the nested SHRINKER lock before it calls msm_gem_vunmap because a vmap on objA could trigger msm_gem_vunmap on objB.

hmm, right.. I guess I still need to test this w/ 32b build (where I guess vmap shrinker is more likely), but I think you are right about needing the nested lock in _vunmap() as well.

Quoted 2 lines

I really like the idea of protecting priv->inactive_list with a separate lock as that is pretty much why the shrinker needs to hold struct_mutex.

yeah, rough idea is split out (probably?) spin-locks to protect the list and madv. Then I think we could drop struct_mutex lock for shrinker.

I think this first patch is pretty close to being ready in time to queue up for 4.13 (which I probably need to do this weekend). We should try and tackle the list+madv locks for 4.14, I think, since this is already a pretty big change.

BR,
-R

Quoted 200 lines

Thanks,
Sushmita

On Jun 15, 2017, at 7:20 AM, Rob Clark <[email protected]> wrote:

---
This is roughly based on Chris's suggestion, in particular the part about using mutex_lock_nested(). It's not *exactly* the same, in particular msm_obj->lock protects a bit more than just backing store and we don't currently track a pin_count. (Instead we currently keep pages pinned until the object is purged or freed.)

Instead of making msm_obj->lock only cover backing store, it is easier to split out madv, which is still protected by struct_mutex, which is still held by the shrinker, so the shrinker does not need to grab msm_obj->lock until it purges an object. We avoid going down any path that could trigger shrinker by ensuring that msm_obj->madv == WILLNEED. To synchronize access to msm_obj->madv it is protected by msm_obj->lock inside struct_mutex.

This seems to keep lockdep happy in my testing so far.

drivers/gpu/drm/msm/msm_gem.c | 54
++++++++++++++++++++++++++++++++--
drivers/gpu/drm/msm/msm_gem.h | 1 +
drivers/gpu/drm/msm/msm_gem_shrinker.c | 12 ++++++++

3 files changed, 65 insertions(+), 2 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +22 -0
index e132548..f5d1f84 100644
@@ -26,6 +26,22 @@
 #include "msm_gpu.h"
 #include "msm_mmu.h"
 
+/* The shrinker can be triggered while we hold objA->lock, and need
+ * to grab objB->lock to purge it.  Lockdep just sees these as a single
+ * class of lock, so we use subclasses to teach it the difference.
+ *
+ * OBJ_LOCK_NORMAL is implicit (ie. normal mutex_lock() call), and
+ * OBJ_LOCK_SHRINKER is used in msm_gem_purge().
+ *
+ * It is *essential* that we never go down paths that could trigger the
+ * shrinker for a purgable object.  This is ensured by checking that
+ * msm_obj->madv == MSM_MADV_WILLNEED.
+ */
+enum {
+ OBJ_LOCK_NORMAL,
+ OBJ_LOCK_SHRINKER,
+};
+
 static dma_addr_t physaddr(struct drm_gem_object *obj)
 {
 struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -150,6 +166,12 @@ struct page **msm_gem_get_pages(struct drm_gem_object
 *obj)
 struct page **p;
 
 mutex_lock(&msm_obj->lock);
+
+ if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+ mutex_unlock(&msm_obj->lock);
+ return ERR_PTR(-EBUSY);
+ }
+
 p = get_pages(obj);
 mutex_unlock(&msm_obj->lock);

return p;

drivers/gpu/drm/msm/msm_gem.c +23 -1
@@ -220,6 +242,11 @@ int msm_gem_fault(struct vm_fault *vmf)
 if (ret)
 goto out;
 
+ if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+ mutex_unlock(&msm_obj->lock);
+ return VM_FAULT_SIGBUS;
+ }
+
 /* make sure we have pages attached now */
 pages = get_pages(obj);
 if (IS_ERR(pages)) {
@@ -358,6 +385,11 @@ int msm_gem_get_iova(struct drm_gem_object *obj,
 
 mutex_lock(&msm_obj->lock);
 
+ if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+ mutex_unlock(&msm_obj->lock);
+ return -EBUSY;
+ }
+
 vma = lookup_vma(obj, aspace);
 
 if (!vma) {
@@ -454,6 +486,12 @@ void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
 mutex_lock(&msm_obj->lock);
+
+ if (WARN_ON(msm_obj->madv != MSM_MADV_WILLNEED)) {
+ mutex_unlock(&msm_obj->lock);
+ return ERR_PTR(-EBUSY);
+ }
+
 if (!msm_obj->vaddr) {
 struct page **pages = get_pages(obj);
 if (IS_ERR(pages)) {
@@ -489,12 +527,18 @@ int msm_gem_madvise(struct drm_gem_object *obj,
 unsigned madv)
 {
 struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
+ mutex_lock(&msm_obj->lock);
+
 WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
 
 if (msm_obj->madv != __MSM_MADV_PURGED)
 msm_obj->madv = madv;
 
- return (msm_obj->madv != __MSM_MADV_PURGED);
+ madv = msm_obj->madv;
+
+ mutex_unlock(&msm_obj->lock);
+
+ return (madv != __MSM_MADV_PURGED);
 }
 

void msm_gem_purge(struct drm_gem_object *obj)

drivers/gpu/drm/msm/msm_gem.c +5 -1
@@ -506,6 +550,8 @@ void msm_gem_purge(struct drm_gem_object *obj)
 WARN_ON(!is_purgeable(msm_obj));
 WARN_ON(obj->import_attach);
 
+ mutex_lock_nested(&msm_obj->lock, OBJ_LOCK_SHRINKER);
+
 put_iova(obj);
 
 msm_gem_vunmap(obj);
@@ -526,6 +572,8 @@ void msm_gem_purge(struct drm_gem_object *obj)
 
 invalidate_mapping_pages(file_inode(obj->filp)->i_mapping,
 0, (loff_t)-1);
+
+ mutex_unlock(&msm_obj->lock);
 }
 
 void msm_gem_vunmap(struct drm_gem_object *obj)
@@ -660,7 +708,7 @@ void msm_gem_describe(struct drm_gem_object *obj, struct
 seq_file *m)
 uint64_t off = drm_vma_node_start(&obj->vma_node);
 const char *madv;
 
- WARN_ON(!mutex_is_locked(&obj->dev->struct_mutex));
+ mutex_lock(&msm_obj->lock);
 
 switch (msm_obj->madv) {

case __MSM_MADV_PURGED:

drivers/gpu/drm/msm/msm_gem.c +2 -0
@@ -701,6 +749,8 @@ void msm_gem_describe(struct drm_gem_object *obj, struct
 seq_file *m)
 if (fence)
 describe_fence(fence, "Exclusive", m);
 rcu_read_unlock();
+
+ mutex_unlock(&msm_obj->lock);
 }
 

void msm_gem_describe_objects(struct list_head *list, struct seq_file *m)

drivers/gpu/drm/msm/msm_gem.h +1 -0
index 9ad5ba4c..2b9b8e9 100644
@@ -101,6 +101,7 @@ static inline bool is_active(struct msm_gem_object
 *msm_obj)
 
 static inline bool is_purgeable(struct msm_gem_object *msm_obj)
 {
+ WARN_ON(!mutex_is_locked(&msm_obj->base.dev->struct_mutex));

return (msm_obj->madv == MSM_MADV_DONTNEED) && msm_obj->sgt && !msm_obj->base.dma_buf && !msm_obj->base.import_attach; }

drivers/gpu/drm/msm/msm_gem_shrinker.c +12 -0
index ab1dd02..e1db4ad 100644
@@ -20,6 +20,18 @@
 
 static bool msm_gem_shrinker_lock(struct drm_device *dev, bool *unlock)
 {
+ /* NOTE: we are *closer* to being able to get rid of
+ * mutex_trylock_recursive().. the msm_gem code itself does
+ * not need struct_mutex, although codepaths that can trigger
+ * shrinker are still called in code-paths that hold the
+ * struct_mutex.
+ *
+ * Also, msm_obj->madv is protected by struct_mutex.
+ *
+ * The next step is probably split out a seperate lock for
+ * protecting inactive_list, so that shrinker does not need
+ * struct_mutex.
+ */
 switch (mutex_trylock_recursive(&dev->struct_mutex)) {
Quoted 6 lines

case MUTEX_TRYLOCK_FAILED:
return false;
--
2.9.4

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno

A message that is not in the archive (sent off-list or not yet received): CAF6AEGst=tPn30N-0u5J8-FkBb=ZKKNkjEBGE-M84TxnC054Vw-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#14 lore raw

PATCH 1/1 fixup! drm/msm: Separate locking of buffer resources from struct_mutex

---
Ok, 2nd fixup to handle vmap shrinking.

 drivers/gpu/drm/msm/msm_gem.c | 44 +++++++++++++++++++++++++++++++++++--------
 1 file changed, 36 insertions(+), 8 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +36 -8
index f5d1f84..3190e05 100644
@@ -42,6 +42,9 @@ enum {
 	OBJ_LOCK_SHRINKER,
 };
 
+static void msm_gem_vunmap_locked(struct drm_gem_object *obj);
+
+
 static dma_addr_t physaddr(struct drm_gem_object *obj)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -484,6 +487,7 @@ int msm_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev,
 void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
+	int ret = 0;
 
 	mutex_lock(&msm_obj->lock);
 
@@ -492,22 +496,35 @@ void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 		return ERR_PTR(-EBUSY);
 	}
 
+	/* increment vmap_count *before* vmap() call, so shrinker can
+	 * check vmap_count (is_vunmapable()) outside of msm_obj->lock.
+	 * This guarantees that we won't try to msm_gem_vunmap() this
+	 * same object from within the vmap() call (while we already
+	 * hold msm_obj->lock)
+	 */
+	msm_obj->vmap_count++;
+
 	if (!msm_obj->vaddr) {
 		struct page **pages = get_pages(obj);
 		if (IS_ERR(pages)) {
-			mutex_unlock(&msm_obj->lock);
-			return ERR_CAST(pages);
+			ret = PTR_ERR(pages);
+			goto fail;
 		}
 		msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
 				VM_MAP, pgprot_writecombine(PAGE_KERNEL));
 		if (msm_obj->vaddr == NULL) {
-			mutex_unlock(&msm_obj->lock);
-			return ERR_PTR(-ENOMEM);
+			ret = -ENOMEM;
+			goto fail;
 		}
 	}
-	msm_obj->vmap_count++;
+
 	mutex_unlock(&msm_obj->lock);
 	return msm_obj->vaddr;
+
+fail:
+	msm_obj->vmap_count--;
+	mutex_unlock(&msm_obj->lock);
+	return ERR_PTR(ret);
 }
 
 void msm_gem_put_vaddr(struct drm_gem_object *obj)
@@ -554,7 +571,7 @@ void msm_gem_purge(struct drm_gem_object *obj)
 
 	put_iova(obj);
 
-	msm_gem_vunmap(obj);
+	msm_gem_vunmap_locked(obj);
 
 	put_pages(obj);
 
@@ -576,10 +593,12 @@ void msm_gem_purge(struct drm_gem_object *obj)
 	mutex_unlock(&msm_obj->lock);
 }
 
-void msm_gem_vunmap(struct drm_gem_object *obj)
+static void msm_gem_vunmap_locked(struct drm_gem_object *obj)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
+	WARN_ON(!mutex_is_locked(&msm_obj->lock));
+
 	if (!msm_obj->vaddr || WARN_ON(!is_vunmapable(msm_obj)))
 		return;
 
@@ -587,6 +606,15 @@ void msm_gem_vunmap(struct drm_gem_object *obj)
 	msm_obj->vaddr = NULL;
 }
 
+void msm_gem_vunmap(struct drm_gem_object *obj)
+{
+	struct msm_gem_object *msm_obj = to_msm_bo(obj);
+
+	mutex_lock_nested(&msm_obj->lock, OBJ_LOCK_SHRINKER);
+	msm_gem_vunmap_locked(obj);
+	mutex_unlock(&msm_obj->lock);
+}
+
 /* must be called before _move_to_active().. */
 int msm_gem_sync_object(struct drm_gem_object *obj,
 		struct msm_fence_context *fctx, bool exclusive)
@@ -799,7 +827,7 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
 		drm_prime_gem_destroy(obj, msm_obj->sgt);
 	} else {
-		msm_gem_vunmap(obj);
+		msm_gem_vunmap_locked(obj);
 		put_pages(obj);
 	}
 
Signature · 6 lines
2.9.4

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno

A message that is not in the archive (sent off-list or not yet received): 20170616142207.20821-1-robdclark-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org

#16 lore raw

Re: [PATCH] fixup! drm/msm: Separate locking of buffer resources from struct_mutex

Hi Rob,

This looks good to me!

Just one nit: msm_gem_vunmap becomes very shrinker specific as it holds the msm_obj->lock with the shrinker class. Should we have the caller i.e. msm_gem_shrinker_vmap hold it instead? Or change it's name to msm_gem_vunmap_shrinker or something like that...?

It does make sense to make the madv/priv->inactive_list locking cleanup separately.

Thanks,
Sushmita

Quoted 112 lines · Rob Clark

---
Ok, 2nd fixup to handle vmap shrinking.

 drivers/gpu/drm/msm/msm_gem.c | 44 
+++++++++++++++++++++++++++++++++++--------
 1 file changed, 36 insertions(+), 8 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +22 -5
index f5d1f84..3190e05 100644
@@ -42,6 +42,9 @@ enum {
 	OBJ_LOCK_SHRINKER,
 };
 
+static void msm_gem_vunmap_locked(struct drm_gem_object *obj);
+
+
 static dma_addr_t physaddr(struct drm_gem_object *obj)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -484,6 +487,7 @@ int msm_gem_dumb_map_offset(struct drm_file *file,
 struct drm_device *dev,
 void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
+	int ret = 0;
 
 	mutex_lock(&msm_obj->lock);
@@ -492,22 +496,35 @@ void *msm_gem_get_vaddr(struct drm_gem_object 
 *obj)
 		return ERR_PTR(-EBUSY);
 	}
 
+	/* increment vmap_count *before* vmap() call, so shrinker can
+	 * check vmap_count (is_vunmapable()) outside of msm_obj->lock.
+	 * This guarantees that we won't try to msm_gem_vunmap() this
+	 * same object from within the vmap() call (while we already
+	 * hold msm_obj->lock)
+	 */
+	msm_obj->vmap_count++;
+
 	if (!msm_obj->vaddr) {
 		struct page **pages = get_pages(obj);
 		if (IS_ERR(pages)) {
-			mutex_unlock(&msm_obj->lock);
-			return ERR_CAST(pages);
+			ret = PTR_ERR(pages);
+			goto fail;
 		}
 		msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
 				VM_MAP, pgprot_writecombine(PAGE_KERNEL));
 		if (msm_obj->vaddr == NULL) {
-			mutex_unlock(&msm_obj->lock);
-			return ERR_PTR(-ENOMEM);
+			ret = -ENOMEM;
+			goto fail;
 		}
 	}
-	msm_obj->vmap_count++;
+
 	mutex_unlock(&msm_obj->lock);
 	return msm_obj->vaddr;
+
+fail:
+	msm_obj->vmap_count--;
+	mutex_unlock(&msm_obj->lock);
+	return ERR_PTR(ret);
 }
 

void msm_gem_put_vaddr(struct drm_gem_object *obj)

drivers/gpu/drm/msm/msm_gem.c +13 -2
@@ -554,7 +571,7 @@ void msm_gem_purge(struct drm_gem_object *obj)
 
 	put_iova(obj);
 
-	msm_gem_vunmap(obj);
+	msm_gem_vunmap_locked(obj);
 
 	put_pages(obj);
 
@@ -576,10 +593,12 @@ void msm_gem_purge(struct drm_gem_object *obj)
 	mutex_unlock(&msm_obj->lock);
 }
 
-void msm_gem_vunmap(struct drm_gem_object *obj)
+static void msm_gem_vunmap_locked(struct drm_gem_object *obj)
 {
 	struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
+	WARN_ON(!mutex_is_locked(&msm_obj->lock));
+
 	if (!msm_obj->vaddr || WARN_ON(!is_vunmapable(msm_obj)))
 		return;
 
@@ -587,6 +606,15 @@ void msm_gem_vunmap(struct drm_gem_object *obj)
 	msm_obj->vaddr = NULL;
 }
 
+void msm_gem_vunmap(struct drm_gem_object *obj)
+{
+	struct msm_gem_object *msm_obj = to_msm_bo(obj);
+
+	mutex_lock_nested(&msm_obj->lock, OBJ_LOCK_SHRINKER);
+	msm_gem_vunmap_locked(obj);
+	mutex_unlock(&msm_obj->lock);
+}
+
 /* must be called before _move_to_active().. */
 int msm_gem_sync_object(struct drm_gem_object *obj,
 		struct msm_fence_context *fctx, bool exclusive)

drivers/gpu/drm/msm/msm_gem.c void msm_gem_free_object(struct drm_gem_object · line 833

Quoted 8 lines
drivers/gpu/drm/msm/msm_gem.c +1 -1
@@ -799,7 +827,7 @@ void msm_gem_free_object(struct drm_gem_object 
 *obj)
 
 		drm_prime_gem_destroy(obj, msm_obj->sgt);
 	} else {
-		msm_gem_vunmap(obj);
+		msm_gem_vunmap_locked(obj);
 		put_pages(obj);
 	}

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno

#17 lore raw

Re: [Freedreno] [PATCH] fixup! drm/msm: Separate locking of buffer resources from struct_mutex

Quoted 8 lines · On Fri, Jun 16, 2017 at 5:32 PM, <[email protected]> wrote:

Hi Rob,

This looks good to me!

Just one nit: msm_gem_vunmap becomes very shrinker specific as it holds the msm_obj->lock with the shrinker class. Should we have the caller i.e. msm_gem_shrinker_vmap hold it instead? Or change it's name to msm_gem_vunmap_shrinker or something like that...?

Hmm, I suppose I could pass the lock class in to msm_gem_vunmap() in case we ever want to call it elsewhere (iirc, i915 did something like that)

BR,
-R

Quoted 128 lines

It does make sense to make the madv/priv->inactive_list locking cleanup separately.

Thanks,
Sushmita

Quoted 119 lines · Rob Clark

---
Ok, 2nd fixup to handle vmap shrinking.

 drivers/gpu/drm/msm/msm_gem.c | 44
+++++++++++++++++++++++++++++++++++--------
 1 file changed, 36 insertions(+), 8 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +36 -8
index f5d1f84..3190e05 100644
@@ -42,6 +42,9 @@ enum {
        OBJ_LOCK_SHRINKER,
 };
 
+static void msm_gem_vunmap_locked(struct drm_gem_object *obj);
+
+
 static dma_addr_t physaddr(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -484,6 +487,7 @@ int msm_gem_dumb_map_offset(struct drm_file *file,
 struct drm_device *dev,
 void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
+       int ret = 0;
 
        mutex_lock(&msm_obj->lock);
@@ -492,22 +496,35 @@ void *msm_gem_get_vaddr(struct drm_gem_object *obj)
                return ERR_PTR(-EBUSY);
        }
 
+       /* increment vmap_count *before* vmap() call, so shrinker can
+        * check vmap_count (is_vunmapable()) outside of msm_obj->lock.
+        * This guarantees that we won't try to msm_gem_vunmap() this
+        * same object from within the vmap() call (while we already
+        * hold msm_obj->lock)
+        */
+       msm_obj->vmap_count++;
+
        if (!msm_obj->vaddr) {
                struct page **pages = get_pages(obj);
                if (IS_ERR(pages)) {
-                       mutex_unlock(&msm_obj->lock);
-                       return ERR_CAST(pages);
+                       ret = PTR_ERR(pages);
+                       goto fail;
                }
                msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
                                VM_MAP, pgprot_writecombine(PAGE_KERNEL));
                if (msm_obj->vaddr == NULL) {
-                       mutex_unlock(&msm_obj->lock);
-                       return ERR_PTR(-ENOMEM);
+                       ret = -ENOMEM;
+                       goto fail;
                }
        }
-       msm_obj->vmap_count++;
+
        mutex_unlock(&msm_obj->lock);
        return msm_obj->vaddr;
+
+fail:
+       msm_obj->vmap_count--;
+       mutex_unlock(&msm_obj->lock);
+       return ERR_PTR(ret);
 }
 
 void msm_gem_put_vaddr(struct drm_gem_object *obj)
@@ -554,7 +571,7 @@ void msm_gem_purge(struct drm_gem_object *obj)
 
        put_iova(obj);
 
-       msm_gem_vunmap(obj);
+       msm_gem_vunmap_locked(obj);
 
        put_pages(obj);
 
@@ -576,10 +593,12 @@ void msm_gem_purge(struct drm_gem_object *obj)
        mutex_unlock(&msm_obj->lock);
 }
 
-void msm_gem_vunmap(struct drm_gem_object *obj)
+static void msm_gem_vunmap_locked(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
+
        if (!msm_obj->vaddr || WARN_ON(!is_vunmapable(msm_obj)))
                return;
 
@@ -587,6 +606,15 @@ void msm_gem_vunmap(struct drm_gem_object *obj)
        msm_obj->vaddr = NULL;
 }
 
+void msm_gem_vunmap(struct drm_gem_object *obj)
+{
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+
+       mutex_lock_nested(&msm_obj->lock, OBJ_LOCK_SHRINKER);
+       msm_gem_vunmap_locked(obj);
+       mutex_unlock(&msm_obj->lock);
+}
+
 /* must be called before _move_to_active().. */
 int msm_gem_sync_object(struct drm_gem_object *obj,
                struct msm_fence_context *fctx, bool exclusive)
@@ -799,7 +827,7 @@ void msm_gem_free_object(struct drm_gem_object *obj)
 
                drm_prime_gem_destroy(obj, msm_obj->sgt);
        } else {
-               msm_gem_vunmap(obj);
+               msm_gem_vunmap_locked(obj);
                put_pages(obj);
        }

A message that is not in the archive (sent off-list or not yet received): CAF6AEGvbf_bwDz9CHPd-BhNHFaBngspfyV1OE=m8YAtm+W1T2Q-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

#19 lore raw

Re: [PATCH] fixup! drm/msm: Separate locking of buffer resources from struct_mutex

Sounds good.

Quoted 144 lines · Rob Clark
Quoted 8 lines · On Fri, Jun 16, 2017 at 5:32 PM, <[email protected]> wrote:

Hi Rob,

This looks good to me!

Just one nit: msm_gem_vunmap becomes very shrinker specific as it holds the msm_obj->lock with the shrinker class. Should we have the caller i.e. msm_gem_shrinker_vmap hold it instead? Or change it's name to msm_gem_vunmap_shrinker or something like that...?

Hmm, I suppose I could pass the lock class in to msm_gem_vunmap() in case we ever want to call it elsewhere (iirc, i915 did something like that)

BR,
-R

Quoted 127 lines

It does make sense to make the madv/priv->inactive_list locking cleanup separately.

Thanks,
Sushmita

Quoted 118 lines · Rob Clark

---
Ok, 2nd fixup to handle vmap shrinking.

 drivers/gpu/drm/msm/msm_gem.c | 44
+++++++++++++++++++++++++++++++++++--------
 1 file changed, 36 insertions(+), 8 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +4 -0
index f5d1f84..3190e05 100644
@@ -42,6 +42,9 @@ enum {
        OBJ_LOCK_SHRINKER,
 };
 
+static void msm_gem_vunmap_locked(struct drm_gem_object *obj);
+
+
 static dma_addr_t physaddr(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
@@ -484,6 +487,7 @@ int msm_gem_dumb_map_offset(struct drm_file 
 *file,
 struct drm_device *dev,
 void *msm_gem_get_vaddr(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
+       int ret = 0;
 
        mutex_lock(&msm_obj->lock);
drivers/gpu/drm/msm/msm_gem.c +2 -0
@@ -492,22 +496,35 @@ void *msm_gem_get_vaddr(struct drm_gem_object 
 *obj)
                return ERR_PTR(-EBUSY);
        }
 
+       /* increment vmap_count *before* vmap() call, so shrinker can
+        * check vmap_count (is_vunmapable()) outside of 

msm_obj->lock.

drivers/gpu/drm/msm/msm_gem.c +8 -2
+        * This guarantees that we won't try to msm_gem_vunmap() this
+        * same object from within the vmap() call (while we already
+        * hold msm_obj->lock)
+        */
+       msm_obj->vmap_count++;
+
        if (!msm_obj->vaddr) {
                struct page **pages = get_pages(obj);
                if (IS_ERR(pages)) {
-                       mutex_unlock(&msm_obj->lock);
-                       return ERR_CAST(pages);
+                       ret = PTR_ERR(pages);
+                       goto fail;
                }
                msm_obj->vaddr = vmap(pages, obj->size >> PAGE_SHIFT,
                                VM_MAP, 
pgprot_writecombine(PAGE_KERNEL));
                if (msm_obj->vaddr == NULL) {
drivers/gpu/drm/msm/msm_gem.c +21 -5
-                       mutex_unlock(&msm_obj->lock);
-                       return ERR_PTR(-ENOMEM);
+                       ret = -ENOMEM;
+                       goto fail;
                }
        }
-       msm_obj->vmap_count++;
+
        mutex_unlock(&msm_obj->lock);
        return msm_obj->vaddr;
+
+fail:
+       msm_obj->vmap_count--;
+       mutex_unlock(&msm_obj->lock);
+       return ERR_PTR(ret);
 }
 
 void msm_gem_put_vaddr(struct drm_gem_object *obj)
@@ -554,7 +571,7 @@ void msm_gem_purge(struct drm_gem_object *obj)
 
        put_iova(obj);
 
-       msm_gem_vunmap(obj);
+       msm_gem_vunmap_locked(obj);
 
        put_pages(obj);
 
@@ -576,10 +593,12 @@ void msm_gem_purge(struct drm_gem_object *obj)
        mutex_unlock(&msm_obj->lock);
 }
 
-void msm_gem_vunmap(struct drm_gem_object *obj)
+static void msm_gem_vunmap_locked(struct drm_gem_object *obj)
 {
        struct msm_gem_object *msm_obj = to_msm_bo(obj);
 
+       WARN_ON(!mutex_is_locked(&msm_obj->lock));
+
        if (!msm_obj->vaddr || WARN_ON(!is_vunmapable(msm_obj)))
                return;
 
@@ -587,6 +606,15 @@ void msm_gem_vunmap(struct drm_gem_object *obj)
        msm_obj->vaddr = NULL;
 }
 
+void msm_gem_vunmap(struct drm_gem_object *obj)
+{
+       struct msm_gem_object *msm_obj = to_msm_bo(obj);
+
+       mutex_lock_nested(&msm_obj->lock, OBJ_LOCK_SHRINKER);
+       msm_gem_vunmap_locked(obj);
+       mutex_unlock(&msm_obj->lock);
+}
+
 /* must be called before _move_to_active().. */
 int msm_gem_sync_object(struct drm_gem_object *obj,
                struct msm_fence_context *fctx, bool exclusive)
@@ -799,7 +827,7 @@ void msm_gem_free_object(struct drm_gem_object 
 *obj)
 
Quoted 10 lines
Quoted 6 lines

drivers/gpu/drm/msm/msm_gem.c void msm_gem_free_object(struct drm_gem_object *obj) · line 832

Quoted 6 lines
drivers/gpu/drm/msm/msm_gem.c +1 -1
                drm_prime_gem_destroy(obj, msm_obj->sgt);
        } else {
-               msm_gem_vunmap(obj);
+               msm_gem_vunmap_locked(obj);
                put_pages(obj);
        }

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno

_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno