Merge tag 'drm-msm-next-2023-08-20' of https://gitlab.freedesktop.org/drm/msm into drm-next

Updates for v6.6, which includes a backmerge of msm-fixes to avoid conficts.

Core:
- SM6125 MDSS support

DPU:
- SM6125 DPU support
- Added subblocks to display snapshot
- Use UBWC data from MDSS driver rather than duplicating it
- dpu_core_perf cleanup

DSI:
- Enabled burst mode to fix CMD mode panels
- Runtime PM support
- refgen regulator support

DSI PHY:
- SM6125 support in 14nm DSI PHY driver

GPU:
- Rework GPU identification to prepare for a7xx, and other a7xx prep
- Cleanups and fixes
- Disallow legacy relocs on a6xx and newer
- a690: switch to using a660_gmu.bin fw as this is what we have in
  linux-firmware and we see no evidence that it should be different
  from other a660 family (a6xx subgen 4) devices
- Submit overhead opts, 1.6x faster for NO_IMPLICIT_SYNC commits with
  100 BOs to 2.5x faster for 1000 BOs

Signed-off-by: Dave Airlie <[email protected]>
From: Rob Clark <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/CAF6AEGv_01g-edjdfKLWWcb-rO5aSyLsv5FpbKrTkXVL9+ngTQ@mail.gmail.com

1 files changed, 49 insertions, 30 deletions

diff --git a/drivers/gpu/drm/msm/msm_gem_submit.c b/drivers/gpu/drm/msm/msm_gem_submit.c
index 63c96416e183..99744de6c05a 100644
--- a/drivers/gpu/drm/msm/msm_gem_submit.c
+++ b/drivers/gpu/drm/msm/msm_gem_submit.c
@@ -165,7 +165,7 @@ static int submit_lookup_objects(struct msm_gem_submit *submit,
 
 		drm_gem_object_get(obj);
 
-		submit->bos[i].obj = to_msm_bo(obj);
+		submit->bos[i].obj = obj;
 	}
 
 out_unlock:
@@ -251,7 +251,7 @@ out:
 static void submit_cleanup_bo(struct msm_gem_submit *submit, int i,
 		unsigned cleanup_flags)
 {
-	struct drm_gem_object *obj = &submit->bos[i].obj->base;
+	struct drm_gem_object *obj = submit->bos[i].obj;
 	unsigned flags = submit->bos[i].flags & cleanup_flags;
 
 	/*
@@ -261,10 +261,7 @@ static void submit_cleanup_bo(struct msm_gem_submit *submit, int i,
 	 */
 	submit->bos[i].flags &= ~cleanup_flags;
 
-	if (flags & BO_VMA_PINNED)
-		msm_gem_vma_unpin(submit->bos[i].vma);
-
-	if (flags & BO_OBJ_PINNED)
+	if (flags & BO_PINNED)
 		msm_gem_unpin_locked(obj);
 
 	if (flags & BO_LOCKED)
@@ -273,7 +270,7 @@ static void submit_cleanup_bo(struct msm_gem_submit *submit, int i,
 
 static void submit_unlock_unpin_bo(struct msm_gem_submit *submit, int i)
 {
-	unsigned cleanup_flags = BO_VMA_PINNED | BO_OBJ_PINNED | BO_LOCKED;
+	unsigned cleanup_flags = BO_PINNED | BO_LOCKED;
 	submit_cleanup_bo(submit, i, cleanup_flags);
 
 	if (!(submit->bos[i].flags & BO_VALID))
@@ -287,7 +284,7 @@ static int submit_lock_objects(struct msm_gem_submit *submit)
 
 retry:
 	for (i = 0; i < submit->nr_bos; i++) {
-		struct msm_gem_object *msm_obj = submit->bos[i].obj;
+		struct drm_gem_object *obj = submit->bos[i].obj;
 
 		if (slow_locked == i)
 			slow_locked = -1;
@@ -295,7 +292,7 @@ retry:
 		contended = i;
 
 		if (!(submit->bos[i].flags & BO_LOCKED)) {
-			ret = dma_resv_lock_interruptible(msm_obj->base.resv,
+			ret = dma_resv_lock_interruptible(obj->resv,
 							  &submit->ticket);
 			if (ret)
 				goto fail;
@@ -321,9 +318,9 @@ fail:
 		submit_unlock_unpin_bo(submit, slow_locked);
 
 	if (ret == -EDEADLK) {
-		struct msm_gem_object *msm_obj = submit->bos[contended].obj;
+		struct drm_gem_object *obj = submit->bos[contended].obj;
 		/* we lost out in a seqno race, lock and retry.. */
-		ret = dma_resv_lock_slow_interruptible(msm_obj->base.resv,
+		ret = dma_resv_lock_slow_interruptible(obj->resv,
 						       &submit->ticket);
 		if (!ret) {
 			submit->bos[contended].flags |= BO_LOCKED;
@@ -346,7 +343,7 @@ static int submit_fence_sync(struct msm_gem_submit *submit, bool no_implicit)
 	int i, ret = 0;
 
 	for (i = 0; i < submit->nr_bos; i++) {
-		struct drm_gem_object *obj = &submit->bos[i].obj->base;
+		struct drm_gem_object *obj = submit->bos[i].obj;
 		bool write = submit->bos[i].flags & MSM_SUBMIT_BO_WRITE;
 
 		/* NOTE: _reserve_shared() must happen before
@@ -384,12 +381,13 @@ static int submit_fence_sync(struct msm_gem_submit *submit, bool no_implicit)
 
 static int submit_pin_objects(struct msm_gem_submit *submit)
 {
+	struct msm_drm_private *priv = submit->dev->dev_private;
 	int i, ret = 0;
 
 	submit->valid = true;
 
 	for (i = 0; i < submit->nr_bos; i++) {
-		struct drm_gem_object *obj = &submit->bos[i].obj->base;
+		struct drm_gem_object *obj = submit->bos[i].obj;
 		struct msm_gem_vma *vma;
 
 		/* if locking succeeded, pin bo: */
@@ -403,9 +401,6 @@ static int submit_pin_objects(struct msm_gem_submit *submit)
 		if (ret)
 			break;
 
-		submit->bos[i].flags |= BO_OBJ_PINNED | BO_VMA_PINNED;
-		submit->bos[i].vma = vma;
-
 		if (vma->iova == submit->bos[i].iova) {
 			submit->bos[i].flags |= BO_VALID;
 		} else {
@@ -416,6 +411,20 @@ static int submit_pin_objects(struct msm_gem_submit *submit)
 		}
 	}
 
+	/*
+	 * A second loop while holding the LRU lock (a) avoids acquiring/dropping
+	 * the LRU lock for each individual bo, while (b) avoiding holding the
+	 * LRU lock while calling msm_gem_pin_vma_locked() (which could trigger
+	 * get_pages() which could trigger reclaim.. and if we held the LRU lock
+	 * could trigger deadlock with the shrinker).
+	 */
+	mutex_lock(&priv->lru.lock);
+	for (i = 0; i < submit->nr_bos; i++) {
+		msm_gem_pin_obj_locked(submit->bos[i].obj);
+		submit->bos[i].flags |= BO_PINNED;
+	}
+	mutex_unlock(&priv->lru.lock);
+
 	return ret;
 }
 
@@ -424,7 +433,7 @@ static void submit_attach_object_fences(struct msm_gem_submit *submit)
 	int i;
 
 	for (i = 0; i < submit->nr_bos; i++) {
-		struct drm_gem_object *obj = &submit->bos[i].obj->base;
+		struct drm_gem_object *obj = submit->bos[i].obj;
 
 		if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
 			dma_resv_add_fence(obj->resv, submit->user_fence,
@@ -436,7 +445,7 @@ static void submit_attach_object_fences(struct msm_gem_submit *submit)
 }
 
 static int submit_bo(struct msm_gem_submit *submit, uint32_t idx,
-		struct msm_gem_object **obj, uint64_t *iova, bool *valid)
+		struct drm_gem_object **obj, uint64_t *iova, bool *valid)
 {
 	if (idx >= submit->nr_bos) {
 		DRM_ERROR("invalid buffer index: %u (out of %u)\n",
@@ -455,7 +464,7 @@ static int submit_bo(struct msm_gem_submit *submit, uint32_t idx,
 }
 
 /* process the reloc's and patch up the cmdstream as needed: */
-static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *obj,
+static int submit_reloc(struct msm_gem_submit *submit, struct drm_gem_object *obj,
 		uint32_t offset, uint32_t nr_relocs, struct drm_msm_gem_submit_reloc *relocs)
 {
 	uint32_t i, last_offset = 0;
@@ -473,7 +482,7 @@ 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_locked(obj);
 
 	if (IS_ERR(ptr)) {
 		ret = PTR_ERR(ptr);
@@ -497,7 +506,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
 		/* offset in dwords: */
 		off = submit_reloc.submit_offset / 4;
 
-		if ((off >= (obj->base.size / 4)) ||
+		if ((off >= (obj->size / 4)) ||
 				(off < last_offset)) {
 			DRM_ERROR("invalid offset %u at reloc %u\n", off, i);
 			ret = -EINVAL;
@@ -524,7 +533,7 @@ 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_locked(obj);
 
 	return ret;
 }
@@ -539,13 +548,13 @@ static void submit_cleanup(struct msm_gem_submit *submit, bool error)
 	unsigned i;
 
 	if (error)
-		cleanup_flags |= BO_VMA_PINNED | BO_OBJ_PINNED;
+		cleanup_flags |= BO_PINNED;
 
 	for (i = 0; i < submit->nr_bos; i++) {
-		struct msm_gem_object *msm_obj = submit->bos[i].obj;
+		struct drm_gem_object *obj = submit->bos[i].obj;
 		submit_cleanup_bo(submit, i, cleanup_flags);
 		if (error)
-			drm_gem_object_put(&msm_obj->base);
+			drm_gem_object_put(obj);
 	}
 }
 
@@ -554,7 +563,7 @@ void msm_submit_retire(struct msm_gem_submit *submit)
 	int i;
 
 	for (i = 0; i < submit->nr_bos; i++) {
-		struct drm_gem_object *obj = &submit->bos[i].obj->base;
+		struct drm_gem_object *obj = submit->bos[i].obj;
 
 		drm_gem_object_put(obj);
 	}
@@ -861,17 +870,17 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 		goto out;
 
 	for (i = 0; i < args->nr_cmds; i++) {
-		struct msm_gem_object *msm_obj;
+		struct drm_gem_object *obj;
 		uint64_t iova;
 
 		ret = submit_bo(submit, submit->cmd[i].idx,
-				&msm_obj, &iova, NULL);
+				&obj, &iova, NULL);
 		if (ret)
 			goto out;
 
 		if (!submit->cmd[i].size ||
 			((submit->cmd[i].size + submit->cmd[i].offset) >
-				msm_obj->base.size / 4)) {
+				obj->size / 4)) {
 			DRM_ERROR("invalid cmdstream size: %u\n", submit->cmd[i].size * 4);
 			ret = -EINVAL;
 			goto out;
@@ -882,7 +891,17 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 		if (submit->valid)
 			continue;
 
-		ret = submit_reloc(submit, msm_obj, submit->cmd[i].offset * 4,
+		if (!gpu->allow_relocs) {
+			if (submit->cmd[i].nr_relocs) {
+				DRM_ERROR("relocs not allowed\n");
+				ret = -EINVAL;
+				goto out;
+			}
+
+			continue;
+		}
+
+		ret = submit_reloc(submit, obj, submit->cmd[i].offset * 4,
 				submit->cmd[i].nr_relocs, submit->cmd[i].relocs);
 		if (ret)
 			goto out;