Remove dummy znode from zvol_state

struct zvol_state contains a dummy znode, which is around 1KB on x64,
only for zfs_range_lock. But in reality, other than z_range_lock and
z_range_avl, zfs_range_lock only need znode on regular file, which
means we add 1KB on a structure and gain nothing.

In this patch, we remove the dummy znode for zvol_state. In order to
do that, we also need to refactor zfs_range_lock a bit. We move
z_range_lock and z_range_avl pair out of znode_t to form zfs_rlock_t.
This new struct replaces znode_t as the main handle inside the range
lock functions.

We also add pointers to z_size, z_blksz, and z_max_blksz so range lock
code doesn't depend on znode_t.  This allows non-ZPL consumers like
Lustre to use the range locks with their equivalent znode_t structure.

Signed-off-by: Chunwei Chen <[email protected]>
Signed-off-by: Boris Protopopov <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Closes #4510

5 files changed, 68 insertions, 65 deletions

diff --git a/module/zfs/zfs_ctldir.c b/module/zfs/zfs_ctldir.c
index fc46a7307..05d841d4b 100644
--- a/module/zfs/zfs_ctldir.c
+++ b/module/zfs/zfs_ctldir.c
@@ -486,7 +486,6 @@ zfsctl_inode_alloc(zfs_sb_t *zsb, uint64_t id,
 	zp->z_gid = 0;
 	zp->z_mode = 0;
 	zp->z_sync_cnt = 0;
-	zp->z_is_zvol = B_FALSE;
 	zp->z_is_mapped = B_FALSE;
 	zp->z_is_ctldir = B_TRUE;
 	zp->z_is_sa = B_FALSE;
diff --git a/module/zfs/zfs_rlock.c b/module/zfs/zfs_rlock.c
index 5064eb796..fd3e8a68d 100644
--- a/module/zfs/zfs_rlock.c
+++ b/module/zfs/zfs_rlock.c
@@ -101,9 +101,9 @@
  * Check if a write lock can be grabbed, or wait and recheck until available.
  */
 static void
-zfs_range_lock_writer(znode_t *zp, rl_t *new)
+zfs_range_lock_writer(zfs_rlock_t *zrl, rl_t *new)
 {
-	avl_tree_t *tree = &zp->z_range_avl;
+	avl_tree_t *tree = &zrl->zr_avl;
 	rl_t *rl;
 	avl_index_t where;
 	uint64_t end_size;
@@ -112,32 +112,32 @@ zfs_range_lock_writer(znode_t *zp, rl_t *new)
 
 	for (;;) {
 		/*
-		 * Range locking is also used by zvol and uses a
-		 * dummied up znode. However, for zvol, we don't need to
-		 * append or grow blocksize, and besides we don't have
-		 * a "sa" data or zfs_sb_t - so skip that processing.
+		 * Range locking is also used by zvol. However, for zvol, we
+		 * don't need to append or grow blocksize, so skip that
+		 * processing.
 		 *
 		 * Yes, this is ugly, and would be solved by not handling
 		 * grow or append in range lock code. If that was done then
 		 * we could make the range locking code generically available
 		 * to other non-zfs consumers.
 		 */
-		if (!zp->z_is_zvol) { /* caller is ZPL */
+		if (zrl->zr_size) { /* caller is ZPL */
 			/*
 			 * If in append mode pick up the current end of file.
 			 * This is done under z_range_lock to avoid races.
 			 */
 			if (new->r_type == RL_APPEND)
-				new->r_off = zp->z_size;
+				new->r_off = *zrl->zr_size;
 
 			/*
 			 * If we need to grow the block size then grab the whole
 			 * file range. This is also done under z_range_lock to
 			 * avoid races.
 			 */
-			end_size = MAX(zp->z_size, new->r_off + len);
-			if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) ||
-			    zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) {
+			end_size = MAX(*zrl->zr_size, new->r_off + len);
+			if (end_size > *zrl->zr_blksz &&
+			    (!ISP2(*zrl->zr_blksz) ||
+			    *zrl->zr_blksz < *zrl->zr_max_blksz)) {
 				new->r_off = 0;
 				new->r_len = UINT64_MAX;
 			}
@@ -175,7 +175,7 @@ wait:
 			cv_init(&rl->r_wr_cv, NULL, CV_DEFAULT, NULL);
 			rl->r_write_wanted = B_TRUE;
 		}
-		cv_wait(&rl->r_wr_cv, &zp->z_range_lock);
+		cv_wait(&rl->r_wr_cv, &zrl->zr_mutex);
 
 		/* reset to original */
 		new->r_off = off;
@@ -353,9 +353,9 @@ zfs_range_add_reader(avl_tree_t *tree, rl_t *new, rl_t *prev, avl_index_t where)
  * Check if a reader lock can be grabbed, or wait and recheck until available.
  */
 static void
-zfs_range_lock_reader(znode_t *zp, rl_t *new)
+zfs_range_lock_reader(zfs_rlock_t *zrl, rl_t *new)
 {
-	avl_tree_t *tree = &zp->z_range_avl;
+	avl_tree_t *tree = &zrl->zr_avl;
 	rl_t *prev, *next;
 	avl_index_t where;
 	uint64_t off = new->r_off;
@@ -378,7 +378,7 @@ retry:
 				cv_init(&prev->r_rd_cv, NULL, CV_DEFAULT, NULL);
 				prev->r_read_wanted = B_TRUE;
 			}
-			cv_wait(&prev->r_rd_cv, &zp->z_range_lock);
+			cv_wait(&prev->r_rd_cv, &zrl->zr_mutex);
 			goto retry;
 		}
 		if (off + len < prev->r_off + prev->r_len)
@@ -401,7 +401,7 @@ retry:
 				cv_init(&next->r_rd_cv, NULL, CV_DEFAULT, NULL);
 				next->r_read_wanted = B_TRUE;
 			}
-			cv_wait(&next->r_rd_cv, &zp->z_range_lock);
+			cv_wait(&next->r_rd_cv, &zrl->zr_mutex);
 			goto retry;
 		}
 		if (off + len <= next->r_off + next->r_len)
@@ -423,14 +423,14 @@ got_lock:
  * previously locked as RL_WRITER).
  */
 rl_t *
-zfs_range_lock(znode_t *zp, uint64_t off, uint64_t len, rl_type_t type)
+zfs_range_lock(zfs_rlock_t *zrl, uint64_t off, uint64_t len, rl_type_t type)
 {
 	rl_t *new;
 
 	ASSERT(type == RL_READER || type == RL_WRITER || type == RL_APPEND);
 
 	new = kmem_alloc(sizeof (rl_t), KM_SLEEP);
-	new->r_zp = zp;
+	new->r_zrl = zrl;
 	new->r_off = off;
 	if (len + off < off)	/* overflow */
 		len = UINT64_MAX - off;
@@ -441,18 +441,18 @@ zfs_range_lock(znode_t *zp, uint64_t off, uint64_t len, rl_type_t type)
 	new->r_write_wanted = B_FALSE;
 	new->r_read_wanted = B_FALSE;
 
-	mutex_enter(&zp->z_range_lock);
+	mutex_enter(&zrl->zr_mutex);
 	if (type == RL_READER) {
 		/*
 		 * First check for the usual case of no locks
 		 */
-		if (avl_numnodes(&zp->z_range_avl) == 0)
-			avl_add(&zp->z_range_avl, new);
+		if (avl_numnodes(&zrl->zr_avl) == 0)
+			avl_add(&zrl->zr_avl, new);
 		else
-			zfs_range_lock_reader(zp, new);
-	} else
-		zfs_range_lock_writer(zp, new); /* RL_WRITER or RL_APPEND */
-	mutex_exit(&zp->z_range_lock);
+			zfs_range_lock_reader(zrl, new);
+	} else /* RL_WRITER or RL_APPEND */
+		zfs_range_lock_writer(zrl, new);
+	mutex_exit(&zrl->zr_mutex);
 	return (new);
 }
 
@@ -474,9 +474,9 @@ zfs_range_free(void *arg)
  * Unlock a reader lock
  */
 static void
-zfs_range_unlock_reader(znode_t *zp, rl_t *remove, list_t *free_list)
+zfs_range_unlock_reader(zfs_rlock_t *zrl, rl_t *remove, list_t *free_list)
 {
-	avl_tree_t *tree = &zp->z_range_avl;
+	avl_tree_t *tree = &zrl->zr_avl;
 	rl_t *rl, *next = NULL;
 	uint64_t len;
 
@@ -543,7 +543,7 @@ zfs_range_unlock_reader(znode_t *zp, rl_t *remove, list_t *free_list)
 void
 zfs_range_unlock(rl_t *rl)
 {
-	znode_t *zp = rl->r_zp;
+	zfs_rlock_t *zrl = rl->r_zrl;
 	list_t free_list;
 	rl_t *free_rl;
 
@@ -552,10 +552,10 @@ zfs_range_unlock(rl_t *rl)
 	ASSERT(!rl->r_proxy);
 	list_create(&free_list, sizeof (rl_t), offsetof(rl_t, rl_node));
 
-	mutex_enter(&zp->z_range_lock);
+	mutex_enter(&zrl->zr_mutex);
 	if (rl->r_type == RL_WRITER) {
 		/* writer locks can't be shared or split */
-		avl_remove(&zp->z_range_avl, rl);
+		avl_remove(&zrl->zr_avl, rl);
 		if (rl->r_write_wanted)
 			cv_broadcast(&rl->r_wr_cv);
 
@@ -568,9 +568,9 @@ zfs_range_unlock(rl_t *rl)
 		 * lock may be shared, let zfs_range_unlock_reader()
 		 * release the zp->z_range_lock lock and free the rl_t
 		 */
-		zfs_range_unlock_reader(zp, rl, &free_list);
+		zfs_range_unlock_reader(zrl, rl, &free_list);
 	}
-	mutex_exit(&zp->z_range_lock);
+	mutex_exit(&zrl->zr_mutex);
 
 	while ((free_rl = list_head(&free_list)) != NULL) {
 		list_remove(&free_list, free_rl);
@@ -588,17 +588,17 @@ zfs_range_unlock(rl_t *rl)
 void
 zfs_range_reduce(rl_t *rl, uint64_t off, uint64_t len)
 {
-	znode_t *zp = rl->r_zp;
+	zfs_rlock_t *zrl = rl->r_zrl;
 
 	/* Ensure there are no other locks */
-	ASSERT(avl_numnodes(&zp->z_range_avl) == 1);
+	ASSERT(avl_numnodes(&zrl->zr_avl) == 1);
 	ASSERT(rl->r_off == 0);
 	ASSERT(rl->r_type == RL_WRITER);
 	ASSERT(!rl->r_proxy);
 	ASSERT3U(rl->r_len, ==, UINT64_MAX);
 	ASSERT3U(rl->r_cnt, ==, 1);
 
-	mutex_enter(&zp->z_range_lock);
+	mutex_enter(&zrl->zr_mutex);
 	rl->r_off = off;
 	rl->r_len = len;
 
@@ -607,7 +607,7 @@ zfs_range_reduce(rl_t *rl, uint64_t off, uint64_t len)
 	if (rl->r_read_wanted)
 		cv_broadcast(&rl->r_rd_cv);
 
-	mutex_exit(&zp->z_range_lock);
+	mutex_exit(&zrl->zr_mutex);
 }
 
 /*
@@ -626,3 +626,10 @@ zfs_range_compare(const void *arg1, const void *arg2)
 		return (-1);
 	return (0);
 }
+
+#ifdef _KERNEL
+EXPORT_SYMBOL(zfs_range_lock);
+EXPORT_SYMBOL(zfs_range_unlock);
+EXPORT_SYMBOL(zfs_range_reduce);
+EXPORT_SYMBOL(zfs_range_compare);
+#endif
diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c
index 43c3a3c20..9b79e977e 100644
--- a/module/zfs/zfs_vnops.c
+++ b/module/zfs/zfs_vnops.c
@@ -481,7 +481,8 @@ zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 	/*
 	 * Lock the range against changes.
 	 */
-	rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER);
+	rl = zfs_range_lock(&zp->z_range_lock, uio->uio_loffset, uio->uio_resid,
+	    RL_READER);
 
 	/*
 	 * If we are reading past end-of-file we can skip
@@ -663,7 +664,7 @@ zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 		 * Obtain an appending range lock to guarantee file append
 		 * semantics.  We reset the write offset once we have the lock.
 		 */
-		rl = zfs_range_lock(zp, 0, n, RL_APPEND);
+		rl = zfs_range_lock(&zp->z_range_lock, 0, n, RL_APPEND);
 		woff = rl->r_off;
 		if (rl->r_len == UINT64_MAX) {
 			/*
@@ -680,7 +681,7 @@ zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr)
 		 * this write, then this range lock will lock the entire file
 		 * so that we can re-write the block safely.
 		 */
-		rl = zfs_range_lock(zp, woff, n, RL_WRITER);
+		rl = zfs_range_lock(&zp->z_range_lock, woff, n, RL_WRITER);
 	}
 
 	if (woff >= limit) {
@@ -1007,7 +1008,8 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 	 * we don't have to write the data twice.
 	 */
 	if (buf != NULL) { /* immediate write */
-		zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER);
+		zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, size,
+		    RL_READER);
 		/* test for truncation needs to be done while range locked */
 		if (offset >= zp->z_size) {
 			error = SET_ERROR(ENOENT);
@@ -1028,8 +1030,8 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 			size = zp->z_blksz;
 			blkoff = ISP2(size) ? P2PHASE(offset, size) : offset;
 			offset -= blkoff;
-			zgd->zgd_rl = zfs_range_lock(zp, offset, size,
-			    RL_READER);
+			zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset,
+			    size, RL_READER);
 			if (zp->z_blksz == size)
 				break;
 			offset += blkoff;
@@ -3947,7 +3949,7 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
 	redirty_page_for_writepage(wbc, pp);
 	unlock_page(pp);
 
-	rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER);
+	rl = zfs_range_lock(&zp->z_range_lock, pgoff, pglen, RL_WRITER);
 	lock_page(pp);
 
 	/* Page mapping changed or it was no longer dirty, we're done */
diff --git a/module/zfs/zfs_znode.c b/module/zfs/zfs_znode.c
index 1742a51ee..abfb50e63 100644
--- a/module/zfs/zfs_znode.c
+++ b/module/zfs/zfs_znode.c
@@ -113,9 +113,7 @@ zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
 	mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
 	rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL);
 
-	mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
-	avl_create(&zp->z_range_avl, zfs_range_compare,
-	    sizeof (rl_t), offsetof(rl_t, r_node));
+	zfs_rlock_init(&zp->z_range_lock);
 
 	zp->z_dirlocks = NULL;
 	zp->z_acl_cached = NULL;
@@ -137,8 +135,7 @@ zfs_znode_cache_destructor(void *buf, void *arg)
 	rw_destroy(&zp->z_name_lock);
 	mutex_destroy(&zp->z_acl_lock);
 	rw_destroy(&zp->z_xattr_lock);
-	avl_destroy(&zp->z_range_avl);
-	mutex_destroy(&zp->z_range_lock);
+	zfs_rlock_destroy(&zp->z_range_lock);
 
 	ASSERT(zp->z_dirlocks == NULL);
 	ASSERT(zp->z_acl_cached == NULL);
@@ -531,10 +528,12 @@ zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
 	zp->z_blksz = blksz;
 	zp->z_seq = 0x7A4653;
 	zp->z_sync_cnt = 0;
-	zp->z_is_zvol = B_FALSE;
 	zp->z_is_mapped = B_FALSE;
 	zp->z_is_ctldir = B_FALSE;
 	zp->z_is_stale = B_FALSE;
+	zp->z_range_lock.zr_size = &zp->z_size;
+	zp->z_range_lock.zr_blksz = &zp->z_blksz;
+	zp->z_range_lock.zr_max_blksz = &ZTOZSB(zp)->z_max_blksz;
 
 	zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
 
@@ -1437,7 +1436,7 @@ zfs_extend(znode_t *zp, uint64_t end)
 	/*
 	 * We will change zp_size, lock the whole file.
 	 */
-	rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
+	rl = zfs_range_lock(&zp->z_range_lock, 0, UINT64_MAX, RL_WRITER);
 
 	/*
 	 * Nothing to do if file already at desired length.
@@ -1554,7 +1553,7 @@ zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
 	/*
 	 * Lock the range being freed.
 	 */
-	rl = zfs_range_lock(zp, off, len, RL_WRITER);
+	rl = zfs_range_lock(&zp->z_range_lock, off, len, RL_WRITER);
 
 	/*
 	 * Nothing to do if file already at desired length.
@@ -1636,7 +1635,7 @@ zfs_trunc(znode_t *zp, uint64_t end)
 	/*
 	 * We will change zp_size, lock the whole file.
 	 */
-	rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
+	rl = zfs_range_lock(&zp->z_range_lock, 0, UINT64_MAX, RL_WRITER);
 
 	/*
 	 * Nothing to do if file already at desired length.
diff --git a/module/zfs/zvol.c b/module/zfs/zvol.c
index a7a79e6dd..be084b1ba 100644
--- a/module/zfs/zvol.c
+++ b/module/zfs/zvol.c
@@ -74,7 +74,7 @@ typedef struct zvol_state {
 	uint32_t		zv_open_count;	/* open counts */
 	uint32_t		zv_changed;	/* disk changed */
 	zilog_t			*zv_zilog;	/* ZIL handle */
-	znode_t			zv_znode;	/* for range locking */
+	zfs_rlock_t		zv_range_lock;	/* range lock */
 	dmu_buf_t		*zv_dbuf;	/* bonus handle */
 	dev_t			zv_dev;		/* device id */
 	struct gendisk		*zv_disk;	/* generic disk */
@@ -632,7 +632,7 @@ zvol_write(struct bio *bio)
 	if (size == 0)
 		goto out;
 
-	rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
+	rl = zfs_range_lock(&zv->zv_range_lock, offset, size, RL_WRITER);
 
 	tx = dmu_tx_create(zv->zv_objset);
 	dmu_tx_hold_write(tx, ZVOL_OBJ, offset, size);
@@ -695,14 +695,13 @@ zvol_discard(struct bio *bio)
 	if (start >= end)
 		return (0);
 
-	rl = zfs_range_lock(&zv->zv_znode, start, size, RL_WRITER);
+	rl = zfs_range_lock(&zv->zv_range_lock, start, size, RL_WRITER);
 
 	error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, start, size);
 
 	/*
 	 * TODO: maybe we should add the operation to the log.
 	 */
-
 	zfs_range_unlock(rl);
 
 	return (error);
@@ -722,7 +721,7 @@ zvol_read(struct bio *bio)
 	if (len == 0)
 		return (0);
 
-	rl = zfs_range_lock(&zv->zv_znode, offset, len, RL_READER);
+	rl = zfs_range_lock(&zv->zv_range_lock, offset, len, RL_READER);
 
 	error = dmu_read_bio(zv->zv_objset, ZVOL_OBJ, bio);
 
@@ -823,7 +822,8 @@ zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
 
 	zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
 	zgd->zgd_zilog = zv->zv_zilog;
-	zgd->zgd_rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER);
+	zgd->zgd_rl = zfs_range_lock(&zv->zv_range_lock, offset, size,
+	    RL_READER);
 
 	/*
 	 * Write records come in two flavors: immediate and indirect.
@@ -1250,10 +1250,7 @@ zvol_alloc(dev_t dev, const char *name)
 	zv->zv_open_count = 0;
 	strlcpy(zv->zv_name, name, MAXNAMELEN);
 
-	mutex_init(&zv->zv_znode.z_range_lock, NULL, MUTEX_DEFAULT, NULL);
-	avl_create(&zv->zv_znode.z_range_avl, zfs_range_compare,
-	    sizeof (rl_t), offsetof(rl_t, r_node));
-	zv->zv_znode.z_is_zvol = TRUE;
+	zfs_rlock_init(&zv->zv_range_lock);
 
 	zv->zv_disk->major = zvol_major;
 	zv->zv_disk->first_minor = (dev & MINORMASK);
@@ -1282,8 +1279,7 @@ zvol_free(zvol_state_t *zv)
 	ASSERT(MUTEX_HELD(&zvol_state_lock));
 	ASSERT(zv->zv_open_count == 0);
 
-	avl_destroy(&zv->zv_znode.z_range_avl);
-	mutex_destroy(&zv->zv_znode.z_range_lock);
+	zfs_rlock_destroy(&zv->zv_range_lock);
 
 	zv->zv_disk->private_data = NULL;