drm/msm/gem: Add fenced vma unpin

With userspace allocated iova (next patch), we can have a race condition
where userspace observes the fence completion and deletes the vma before
retire_submit() gets around to unpinning the vma.  To handle this, add a
fenced unpin which drops the refcount but tracks the fence, and update
msm_gem_vma_inuse() to check any previously unsignaled fences.

v2: Fix inuse underflow (duplicate unpin)
v3: Fix msm_job_run() vs submit_cleanup() race condition

Signed-off-by: Rob Clark <robdclark@chromium.org>
Link: https://lore.kernel.org/r/20220411215849.297838-10-robdclark@gmail.com
Signed-off-by: Rob Clark <robdclark@chromium.org>

drivers/gpu/drm/msm/msm_fence.c

@@ -15,6 +15,7 @@ msm_fence_context_alloc(struct drm_device *dev, volatile uint32_t *fenceptr,
 		const char *name)
 {
 	struct msm_fence_context *fctx;
+	static int index = 0;
 
 	fctx = kzalloc(sizeof(*fctx), GFP_KERNEL);
 	if (!fctx)
@@ -23,6 +24,7 @@ msm_fence_context_alloc(struct drm_device *dev, volatile uint32_t *fenceptr,
 	fctx->dev = dev;
 	strncpy(fctx->name, name, sizeof(fctx->name));
 	fctx->context = dma_fence_context_alloc(1);
+	fctx->index = index++;
 	fctx->fenceptr = fenceptr;
 	spin_lock_init(&fctx->spinlock);
 
@@ -34,7 +36,7 @@ void msm_fence_context_free(struct msm_fence_context *fctx)
 	kfree(fctx);
 }
 
-static inline bool fence_completed(struct msm_fence_context *fctx, uint32_t fence)
+bool msm_fence_completed(struct msm_fence_context *fctx, uint32_t fence)
 {
 	/*
 	 * Note: Check completed_fence first, as fenceptr is in a write-combine
@@ -76,7 +78,7 @@ static const char *msm_fence_get_timeline_name(struct dma_fence *fence)
 static bool msm_fence_signaled(struct dma_fence *fence)
 {
 	struct msm_fence *f = to_msm_fence(fence);
-	return fence_completed(f->fctx, f->base.seqno);
+	return msm_fence_completed(f->fctx, f->base.seqno);
 }
 
 static const struct dma_fence_ops msm_fence_ops = {

drivers/gpu/drm/msm/msm_fence.h

@@ -21,6 +21,8 @@ struct msm_fence_context {
 	char name[32];
 	/** context: see dma_fence_context_alloc() */
 	unsigned context;
+	/** index: similar to context, but local to msm_fence_context's */
+	unsigned index;
 
 	/**
 	 * last_fence:
@@ -56,6 +58,7 @@ struct msm_fence_context * msm_fence_context_alloc(struct drm_device *dev,
 		volatile uint32_t *fenceptr, const char *name);
 void msm_fence_context_free(struct msm_fence_context *fctx);
 
+bool msm_fence_completed(struct msm_fence_context *fctx, uint32_t fence);
 void msm_update_fence(struct msm_fence_context *fctx, uint32_t fence);
 
 struct dma_fence * msm_fence_alloc(struct msm_fence_context *fctx);

drivers/gpu/drm/msm/msm_gem.c

@@ -445,7 +445,7 @@ void msm_gem_unpin_vma_locked(struct drm_gem_object *obj, struct msm_gem_vma *vm
 
 	GEM_WARN_ON(!msm_gem_is_locked(obj));
 
-	msm_gem_unmap_vma(vma->aspace, vma);
+	msm_gem_unpin_vma(vma);
 
 	msm_obj->pin_count--;
 	GEM_WARN_ON(msm_obj->pin_count < 0);

drivers/gpu/drm/msm/msm_gem.h

@@ -49,13 +49,18 @@ struct msm_gem_address_space *
 msm_gem_address_space_create(struct msm_mmu *mmu, const char *name,
 		u64 va_start, u64 size);
 
+struct msm_fence_context;
+
 struct msm_gem_vma {
 	struct drm_mm_node node;
 	uint64_t iova;
 	struct msm_gem_address_space *aspace;
 	struct list_head list;    /* node in msm_gem_object::vmas */
 	bool mapped;
 	int inuse;
+	uint32_t fence_mask;
+	uint32_t fence[MSM_GPU_MAX_RINGS];
+	struct msm_fence_context *fctx[MSM_GPU_MAX_RINGS];
 };
 
 int msm_gem_init_vma(struct msm_gem_address_space *aspace,
@@ -64,8 +69,8 @@ int msm_gem_init_vma(struct msm_gem_address_space *aspace,
 bool msm_gem_vma_inuse(struct msm_gem_vma *vma);
 void msm_gem_purge_vma(struct msm_gem_address_space *aspace,
 		struct msm_gem_vma *vma);
-void msm_gem_unmap_vma(struct msm_gem_address_space *aspace,
-		struct msm_gem_vma *vma);
+void msm_gem_unpin_vma(struct msm_gem_vma *vma);
+void msm_gem_unpin_vma_fenced(struct msm_gem_vma *vma, struct msm_fence_context *fctx);
 int msm_gem_map_vma(struct msm_gem_address_space *aspace,
 		struct msm_gem_vma *vma, int prot,
 		struct sg_table *sgt, int size);
@@ -363,6 +368,11 @@ struct msm_gem_submit {
 		struct drm_msm_gem_submit_reloc *relocs;
 	} *cmd;  /* array of size nr_cmds */
 	struct {
+/* make sure these don't conflict w/ MSM_SUBMIT_BO_x */
+#define BO_VALID    0x8000   /* is current addr in cmdstream correct/valid? */
+#define BO_LOCKED   0x4000   /* obj lock is held */
+#define BO_ACTIVE   0x2000   /* active refcnt is held */
+#define BO_PINNED   0x1000   /* obj is pinned and on active list */
 		uint32_t flags;
 		union {
 			struct msm_gem_object *obj;

drivers/gpu/drm/msm/msm_gem_submit.c

@@ -21,12 +21,6 @@
  * Cmdstream submission:
  */
 
-/* make sure these don't conflict w/ MSM_SUBMIT_BO_x */
-#define BO_VALID    0x8000   /* is current addr in cmdstream correct/valid? */
-#define BO_LOCKED   0x4000   /* obj lock is held */
-#define BO_ACTIVE   0x2000   /* active refcnt is held */
-#define BO_PINNED   0x1000   /* obj is pinned and on active list */
-
 static struct msm_gem_submit *submit_create(struct drm_device *dev,
 		struct msm_gpu *gpu,
 		struct msm_gpu_submitqueue *queue, uint32_t nr_bos,
@@ -231,6 +225,13 @@ static void submit_cleanup_bo(struct msm_gem_submit *submit, int i,
 	struct drm_gem_object *obj = &submit->bos[i].obj->base;
 	unsigned flags = submit->bos[i].flags & cleanup_flags;
 
+	/*
+	 * Clear flags bit before dropping lock, so that the msm_job_run()
+	 * path isn't racing with submit_cleanup() (ie. the read/modify/
+	 * write is protected by the obj lock in all paths)
+	 */
+	submit->bos[i].flags &= ~cleanup_flags;
+
 	if (flags & BO_PINNED)
 		msm_gem_unpin_vma_locked(obj, submit->bos[i].vma);
 
@@ -239,8 +240,6 @@ static void submit_cleanup_bo(struct msm_gem_submit *submit, int i,
 
 	if (flags & BO_LOCKED)
 		dma_resv_unlock(obj->resv);
-
-	submit->bos[i].flags &= ~cleanup_flags;
 }
 
 static void submit_unlock_unpin_bo(struct msm_gem_submit *submit, int i)

drivers/gpu/drm/msm/msm_gem_vma.c

@@ -5,6 +5,7 @@
  */
 
 #include "msm_drv.h"
+#include "msm_fence.h"
 #include "msm_gem.h"
 #include "msm_mmu.h"
 
@@ -39,7 +40,19 @@ msm_gem_address_space_get(struct msm_gem_address_space *aspace)
 
 bool msm_gem_vma_inuse(struct msm_gem_vma *vma)
 {
-	return !!vma->inuse;
+	if (vma->inuse > 0)
+		return true;
+
+	while (vma->fence_mask) {
+		unsigned idx = ffs(vma->fence_mask) - 1;
+
+		if (!msm_fence_completed(vma->fctx[idx], vma->fence[idx]))
+			return true;
+
+		vma->fence_mask &= ~BIT(idx);
+	}
+
+	return false;
 }
 
 /* Actually unmap memory for the vma */
@@ -63,13 +76,24 @@ void msm_gem_purge_vma(struct msm_gem_address_space *aspace,
 }
 
 /* Remove reference counts for the mapping */
-void msm_gem_unmap_vma(struct msm_gem_address_space *aspace,
-		struct msm_gem_vma *vma)
+void msm_gem_unpin_vma(struct msm_gem_vma *vma)
 {
+	if (GEM_WARN_ON(!vma->inuse))
+		return;
 	if (!GEM_WARN_ON(!vma->iova))
 		vma->inuse--;
 }
 
+/* Replace pin reference with fence: */
+void msm_gem_unpin_vma_fenced(struct msm_gem_vma *vma, struct msm_fence_context *fctx)
+{
+	vma->fctx[fctx->index] = fctx;
+	vma->fence[fctx->index] = fctx->last_fence;
+	vma->fence_mask |= BIT(fctx->index);
+	msm_gem_unpin_vma(vma);
+}
+
+/* Map and pin vma: */
 int
 msm_gem_map_vma(struct msm_gem_address_space *aspace,
 		struct msm_gem_vma *vma, int prot,

drivers/gpu/drm/msm/msm_ringbuffer.c

@@ -14,9 +14,20 @@ module_param(num_hw_submissions, uint, 0600);
 static struct dma_fence *msm_job_run(struct drm_sched_job *job)
 {
 	struct msm_gem_submit *submit = to_msm_submit(job);
+	struct msm_fence_context *fctx = submit->ring->fctx;
 	struct msm_gpu *gpu = submit->gpu;
+	int i;
 
-	submit->hw_fence = msm_fence_alloc(submit->ring->fctx);
+	submit->hw_fence = msm_fence_alloc(fctx);
+
+	for (i = 0; i < submit->nr_bos; i++) {
+		struct drm_gem_object *obj = &submit->bos[i].obj->base;
+
+		msm_gem_lock(obj);
+		msm_gem_unpin_vma_fenced(submit->bos[i].vma, fctx);
+		submit->bos[i].flags &= ~BO_PINNED;
+		msm_gem_unlock(obj);
+	}
 
 	pm_runtime_get_sync(&gpu->pdev->dev);