drm/msm/mdp5: handle SMP block allocations "atomically"

Previously, SMP block allocation was not checked in the plane's
atomic_check() fxn, so we could fail allocation SMP block allocation at
atomic_update() time.  Re-work the block allocation to request blocks
during atomic_check(), but not update the hw until committing the atomic
update.

Since SMP blocks allocated at atomic_check() time, we need to manage the
SMP state as part of mdp5_state (global atomic state).  This actually
ends up significantly simplifying the SMP management, as the SMP module
does not need to manage the intermediate state between assigning new
blocks before setting flush bits and releasing old blocks after vblank.
(The SMP registers and SMP allocation is not double-buffered, so newly
allocated blocks need to be updated in kms->prepare_commit() released
blocks in kms->complete_commit().)

Signed-off-by: Rob Clark <robdclark@gmail.com>

1 files changed, 87 insertions, 183 deletions

diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c
index 27d7b55b52c9..ef1120a3c0b4 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_smp.c
@@ -21,72 +21,6 @@
 #include "mdp5_smp.h"
 
 
-/* SMP - Shared Memory Pool
- *
- * These are shared between all the clients, where each plane in a
- * scanout buffer is a SMP client.  Ie. scanout of 3 plane I420 on
- * pipe VIG0 => 3 clients: VIG0_Y, VIG0_CB, VIG0_CR.
- *
- * Based on the size of the attached scanout buffer, a certain # of
- * blocks must be allocated to that client out of the shared pool.
- *
- * In some hw, some blocks are statically allocated for certain pipes
- * and CANNOT be re-allocated (eg: MMB0 and MMB1 both tied to RGB0).
- *
- * For each block that can be dynamically allocated, it can be either
- *     free:
- *     The block is free.
- *
- *     pending:
- *     The block is allocated to some client and not free.
- *
- *     configured:
- *     The block is allocated to some client, and assigned to that
- *     client in MDP5_SMP_ALLOC registers.
- *
- *     inuse:
- *     The block is being actively used by a client.
- *
- * The updates happen in the following steps:
- *
- *  1) mdp5_smp_request():
- *     When plane scanout is setup, calculate required number of
- *     blocks needed per client, and request. Blocks neither inuse nor
- *     configured nor pending by any other client are added to client's
- *     pending set.
- *     For shrinking, blocks in pending but not in configured can be freed
- *     directly, but those already in configured will be freed later by
- *     mdp5_smp_commit.
- *
- *  2) mdp5_smp_configure():
- *     As hw is programmed, before FLUSH, MDP5_SMP_ALLOC registers
- *     are configured for the union(pending, inuse)
- *     Current pending is copied to configured.
- *     It is assumed that mdp5_smp_request and mdp5_smp_configure not run
- *     concurrently for the same pipe.
- *
- *  3) mdp5_smp_commit():
- *     After next vblank, copy configured -> inuse.  Optionally update
- *     MDP5_SMP_ALLOC registers if there are newly unused blocks
- *
- *  4) mdp5_smp_release():
- *     Must be called after the pipe is disabled and no longer uses any SMB
- *
- * On the next vblank after changes have been committed to hw, the
- * client's pending blocks become it's in-use blocks (and no-longer
- * in-use blocks become available to other clients).
- *
- * btw, hurray for confusing overloaded acronyms!  :-/
- *
- * NOTE: for atomic modeset/pageflip NONBLOCK operations, step #1
- * should happen at (or before)? atomic->check().  And we'd need
- * an API to discard previous requests if update is aborted or
- * (test-only).
- *
- * TODO would perhaps be nice to have debugfs to dump out kernel
- * inuse and pending state of all clients..
- */
-
 struct mdp5_smp {
 	struct drm_device *dev;
 
@@ -94,16 +28,8 @@ struct mdp5_smp {
 
 	int blk_cnt;
 	int blk_size;
-
-	spinlock_t state_lock;
-	mdp5_smp_state_t state; /* to track smp allocation amongst pipes: */
-
-	struct mdp5_client_smp_state client_state[MAX_CLIENTS];
 };
 
-static void update_smp_state(struct mdp5_smp *smp,
-		u32 cid, mdp5_smp_state_t *assigned);
-
 static inline
 struct mdp5_kms *get_kms(struct mdp5_smp *smp)
 {
@@ -134,57 +60,38 @@ static inline u32 pipe2client(enum mdp5_pipe pipe, int plane)
 	return mdp5_cfg->smp.clients[pipe] + plane;
 }
 
-/* step #1: update # of blocks pending for the client: */
+/* allocate blocks for the specified request: */
 static int smp_request_block(struct mdp5_smp *smp,
+		struct mdp5_smp_state *state,
 		u32 cid, int nblks)
 {
-	struct mdp5_kms *mdp5_kms = get_kms(smp);
-	struct mdp5_client_smp_state *ps = &smp->client_state[cid];
-	int i, ret, avail, cur_nblks, cnt = smp->blk_cnt;
+	void *cs = state->client_state[cid];
+	int i, avail, cnt = smp->blk_cnt;
 	uint8_t reserved;
-	unsigned long flags;
 
-	reserved = smp->reserved[cid];
+	/* we shouldn't be requesting blocks for an in-use client: */
+	WARN_ON(bitmap_weight(cs, cnt) > 0);
 
-	spin_lock_irqsave(&smp->state_lock, flags);
+	reserved = smp->reserved[cid];
 
 	if (reserved) {
 		nblks = max(0, nblks - reserved);
 		DBG("%d MMBs allocated (%d reserved)", nblks, reserved);
 	}
 
-	avail = cnt - bitmap_weight(smp->state, cnt);
+	avail = cnt - bitmap_weight(state->state, cnt);
 	if (nblks > avail) {
-		dev_err(mdp5_kms->dev->dev, "out of blks (req=%d > avail=%d)\n",
+		dev_err(smp->dev->dev, "out of blks (req=%d > avail=%d)\n",
 				nblks, avail);
-		ret = -ENOSPC;
-		goto fail;
+		return -ENOSPC;
 	}
 
-	cur_nblks = bitmap_weight(ps->pending, cnt);
-	if (nblks > cur_nblks) {
-		/* grow the existing pending reservation: */
-		for (i = cur_nblks; i < nblks; i++) {
-			int blk = find_first_zero_bit(smp->state, cnt);
-			set_bit(blk, ps->pending);
-			set_bit(blk, smp->state);
-		}
-	} else {
-		/* shrink the existing pending reservation: */
-		for (i = cur_nblks; i > nblks; i--) {
-			int blk = find_first_bit(ps->pending, cnt);
-			clear_bit(blk, ps->pending);
-
-			/* clear in global smp_state if not in configured
-			 * otherwise until _commit()
-			 */
-			if (!test_bit(blk, ps->configured))
-				clear_bit(blk, smp->state);
-		}
+	for (i = 0; i < nblks; i++) {
+		int blk = find_first_zero_bit(state->state, cnt);
+		set_bit(blk, cs);
+		set_bit(blk, state->state);
 	}
 
-fail:
-	spin_unlock_irqrestore(&smp->state_lock, flags);
 	return 0;
 }
 
@@ -209,14 +116,15 @@ static void set_fifo_thresholds(struct mdp5_smp *smp,
  * decimated width.  Ie. SMP buffering sits downstream of decimation (which
  * presumably happens during the dma from scanout buffer).
  */
-int mdp5_smp_request(struct mdp5_smp *smp, enum mdp5_pipe pipe,
-		const struct mdp_format *format, u32 width, bool hdecim)
+uint32_t mdp5_smp_calculate(struct mdp5_smp *smp,
+		const struct mdp_format *format,
+		u32 width, bool hdecim)
 {
 	struct mdp5_kms *mdp5_kms = get_kms(smp);
-	struct drm_device *dev = mdp5_kms->dev;
 	int rev = mdp5_cfg_get_hw_rev(mdp5_kms->cfg);
-	int i, hsub, nplanes, nlines, nblks, ret;
+	int i, hsub, nplanes, nlines;
 	u32 fmt = format->base.pixel_format;
+	uint32_t blkcfg = 0;
 
 	nplanes = drm_format_num_planes(fmt);
 	hsub = drm_format_horz_chroma_subsampling(fmt);
@@ -239,7 +147,7 @@ int mdp5_smp_request(struct mdp5_smp *smp, enum mdp5_pipe pipe,
 			hsub = 1;
 	}
 
-	for (i = 0, nblks = 0; i < nplanes; i++) {
+	for (i = 0; i < nplanes; i++) {
 		int n, fetch_stride, cpp;
 
 		cpp = drm_format_plane_cpp(fmt, i);
@@ -251,60 +159,72 @@ int mdp5_smp_request(struct mdp5_smp *smp, enum mdp5_pipe pipe,
 		if (rev == 0)
 			n = roundup_pow_of_two(n);
 
+		blkcfg |= (n << (8 * i));
+	}
+
+	return blkcfg;
+}
+
+int mdp5_smp_assign(struct mdp5_smp *smp, struct mdp5_smp_state *state,
+		enum mdp5_pipe pipe, uint32_t blkcfg)
+{
+	struct mdp5_kms *mdp5_kms = get_kms(smp);
+	struct drm_device *dev = mdp5_kms->dev;
+	int i, ret;
+
+	for (i = 0; i < pipe2nclients(pipe); i++) {
+		u32 cid = pipe2client(pipe, i);
+		int n = blkcfg & 0xff;
+
+		if (!n)
+			continue;
+
 		DBG("%s[%d]: request %d SMP blocks", pipe2name(pipe), i, n);
-		ret = smp_request_block(smp, pipe2client(pipe, i), n);
+		ret = smp_request_block(smp, state, cid, n);
 		if (ret) {
 			dev_err(dev->dev, "Cannot allocate %d SMP blocks: %d\n",
 					n, ret);
 			return ret;
 		}
 
-		nblks += n;
+		blkcfg >>= 8;
 	}
 
-	set_fifo_thresholds(smp, pipe, nblks);
+	state->assigned |= (1 << pipe);
 
 	return 0;
 }
 
 /* Release SMP blocks for all clients of the pipe */
-void mdp5_smp_release(struct mdp5_smp *smp, enum mdp5_pipe pipe)
+void mdp5_smp_release(struct mdp5_smp *smp, struct mdp5_smp_state *state,
+		enum mdp5_pipe pipe)
 {
 	int i;
-	unsigned long flags;
 	int cnt = smp->blk_cnt;
 
 	for (i = 0; i < pipe2nclients(pipe); i++) {
-		mdp5_smp_state_t assigned;
 		u32 cid = pipe2client(pipe, i);
-		struct mdp5_client_smp_state *ps = &smp->client_state[cid];
-
-		spin_lock_irqsave(&smp->state_lock, flags);
-
-		/* clear hw assignment */
-		bitmap_or(assigned, ps->inuse, ps->configured, cnt);
-		update_smp_state(smp, CID_UNUSED, &assigned);
+		void *cs = state->client_state[cid];
 
-		/* free to global pool */
-		bitmap_andnot(smp->state, smp->state, ps->pending, cnt);
-		bitmap_andnot(smp->state, smp->state, assigned, cnt);
+		/* update global state: */
+		bitmap_andnot(state->state, state->state, cs, cnt);
 
-		/* clear client's infor */
-		bitmap_zero(ps->pending, cnt);
-		bitmap_zero(ps->configured, cnt);
-		bitmap_zero(ps->inuse, cnt);
-
-		spin_unlock_irqrestore(&smp->state_lock, flags);
+		/* clear client's state */
+		bitmap_zero(cs, cnt);
 	}
 
-	set_fifo_thresholds(smp, pipe, 0);
+	state->released |= (1 << pipe);
 }
 
-static void update_smp_state(struct mdp5_smp *smp,
+/* NOTE: SMP_ALLOC_* regs are *not* double buffered, so release has to
+ * happen after scanout completes.
+ */
+static unsigned update_smp_state(struct mdp5_smp *smp,
 		u32 cid, mdp5_smp_state_t *assigned)
 {
 	struct mdp5_kms *mdp5_kms = get_kms(smp);
 	int cnt = smp->blk_cnt;
+	unsigned nblks = 0;
 	u32 blk, val;
 
 	for_each_set_bit(blk, *assigned, cnt) {
@@ -330,62 +250,46 @@ static void update_smp_state(struct mdp5_smp *smp,
 
 		mdp5_write(mdp5_kms, REG_MDP5_SMP_ALLOC_W_REG(idx), val);
 		mdp5_write(mdp5_kms, REG_MDP5_SMP_ALLOC_R_REG(idx), val);
+
+		nblks++;
 	}
+
+	return nblks;
 }
 
-/* step #2: configure hw for union(pending, inuse): */
-void mdp5_smp_configure(struct mdp5_smp *smp, enum mdp5_pipe pipe)
+void mdp5_smp_prepare_commit(struct mdp5_smp *smp, struct mdp5_smp_state *state)
 {
-	int cnt = smp->blk_cnt;
-	mdp5_smp_state_t assigned;
-	int i;
-
-	for (i = 0; i < pipe2nclients(pipe); i++) {
-		u32 cid = pipe2client(pipe, i);
-		struct mdp5_client_smp_state *ps = &smp->client_state[cid];
+	enum mdp5_pipe pipe;
 
-		/*
-		 * if vblank has not happened since last smp_configure
-		 * skip the configure for now
-		 */
-		if (!bitmap_equal(ps->inuse, ps->configured, cnt))
-			continue;
+	for_each_set_bit(pipe, &state->assigned, sizeof(state->assigned) * 8) {
+		unsigned i, nblks = 0;
 
-		bitmap_copy(ps->configured, ps->pending, cnt);
-		bitmap_or(assigned, ps->inuse, ps->configured, cnt);
-		update_smp_state(smp, cid, &assigned);
-	}
-}
+		for (i = 0; i < pipe2nclients(pipe); i++) {
+			u32 cid = pipe2client(pipe, i);
+			void *cs = state->client_state[cid];
 
-/* step #3: after vblank, copy configured -> inuse: */
-void mdp5_smp_commit(struct mdp5_smp *smp, enum mdp5_pipe pipe)
-{
-	int cnt = smp->blk_cnt;
-	mdp5_smp_state_t released;
-	int i;
+			nblks += update_smp_state(smp, cid, cs);
 
-	for (i = 0; i < pipe2nclients(pipe); i++) {
-		u32 cid = pipe2client(pipe, i);
-		struct mdp5_client_smp_state *ps = &smp->client_state[cid];
+			DBG("assign %s:%u, %u blks",
+				pipe2name(pipe), i, nblks);
+		}
 
-		/*
-		 * Figure out if there are any blocks we where previously
-		 * using, which can be released and made available to other
-		 * clients:
-		 */
-		if (bitmap_andnot(released, ps->inuse, ps->configured, cnt)) {
-			unsigned long flags;
+		set_fifo_thresholds(smp, pipe, nblks);
+	}
 
-			spin_lock_irqsave(&smp->state_lock, flags);
-			/* clear released blocks: */
-			bitmap_andnot(smp->state, smp->state, released, cnt);
-			spin_unlock_irqrestore(&smp->state_lock, flags);
+	state->assigned = 0;
+}
 
-			update_smp_state(smp, CID_UNUSED, &released);
-		}
+void mdp5_smp_complete_commit(struct mdp5_smp *smp, struct mdp5_smp_state *state)
+{
+	enum mdp5_pipe pipe;
 
-		bitmap_copy(ps->inuse, ps->configured, cnt);
+	for_each_set_bit(pipe, &state->released, sizeof(state->released) * 8) {
+		DBG("release %s", pipe2name(pipe));
+		set_fifo_thresholds(smp, pipe, 0);
 	}
+
+	state->released = 0;
 }
 
 void mdp5_smp_destroy(struct mdp5_smp *smp)
@@ -393,8 +297,9 @@ void mdp5_smp_destroy(struct mdp5_smp *smp)
 	kfree(smp);
 }
 
-struct mdp5_smp *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_block *cfg)
+struct mdp5_smp *mdp5_smp_init(struct mdp5_kms *mdp5_kms, const struct mdp5_smp_block *cfg)
 {
+	struct mdp5_smp_state *state = &mdp5_kms->state->smp;
 	struct mdp5_smp *smp = NULL;
 	int ret;
 
@@ -404,14 +309,13 @@ struct mdp5_smp *mdp5_smp_init(struct drm_device *dev, const struct mdp5_smp_blo
 		goto fail;
 	}
 
-	smp->dev = dev;
+	smp->dev = mdp5_kms->dev;
 	smp->blk_cnt = cfg->mmb_count;
 	smp->blk_size = cfg->mmb_size;
 
 	/* statically tied MMBs cannot be re-allocated: */
-	bitmap_copy(smp->state, cfg->reserved_state, smp->blk_cnt);
+	bitmap_copy(state->state, cfg->reserved_state, smp->blk_cnt);
 	memcpy(smp->reserved, cfg->reserved, sizeof(smp->reserved));
-	spin_lock_init(&smp->state_lock);
 
 	return smp;
 fail: