aboutsummaryrefslogtreecommitdiff
path: root/fs/btrfs
diff options
context:
space:
mode:
Diffstat (limited to 'fs/btrfs')
-rw-r--r--fs/btrfs/Makefile19
-rw-r--r--fs/btrfs/acl.c20
-rw-r--r--fs/btrfs/async-thread.c67
-rw-r--r--fs/btrfs/async-thread.h2
-rw-r--r--fs/btrfs/ctree.c327
-rw-r--r--fs/btrfs/ctree.h90
-rw-r--r--fs/btrfs/delayed-ref.c1
-rw-r--r--fs/btrfs/disk-io.c110
-rw-r--r--fs/btrfs/extent-tree.c447
-rw-r--r--fs/btrfs/extent_io.c183
-rw-r--r--fs/btrfs/extent_map.c18
-rw-r--r--fs/btrfs/file.c95
-rw-r--r--fs/btrfs/free-space-cache.c535
-rw-r--r--fs/btrfs/free-space-cache.h44
-rw-r--r--fs/btrfs/inode-map.c2
-rw-r--r--fs/btrfs/inode.c188
-rw-r--r--fs/btrfs/ioctl.c60
-rw-r--r--fs/btrfs/locking.c4
-rw-r--r--fs/btrfs/ordered-data.c2
-rw-r--r--fs/btrfs/super.c86
-rw-r--r--fs/btrfs/transaction.c13
-rw-r--r--fs/btrfs/tree-log.c14
-rw-r--r--fs/btrfs/volumes.c200
-rw-r--r--fs/btrfs/volumes.h18
24 files changed, 1543 insertions, 1002 deletions
diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 9adf5e4f7e96..94212844a9bc 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -1,25 +1,10 @@
-ifneq ($(KERNELRELEASE),)
-# kbuild part of makefile
obj-$(CONFIG_BTRFS_FS) := btrfs.o
-btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
+
+btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
file-item.o inode-item.o inode-map.o disk-io.o \
transaction.o inode.o file.o tree-defrag.o \
extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \
extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
ref-cache.o export.o tree-log.o acl.o free-space-cache.o zlib.o \
compression.o delayed-ref.o
-else
-
-# Normal Makefile
-
-KERNELDIR := /lib/modules/`uname -r`/build
-all:
- $(MAKE) -C $(KERNELDIR) M=`pwd` CONFIG_BTRFS_FS=m modules
-
-modules_install:
- $(MAKE) -C $(KERNELDIR) M=`pwd` modules_install
-clean:
- $(MAKE) -C $(KERNELDIR) M=`pwd` clean
-
-endif
diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c
index 1d53b62dbba5..cbba000dccbe 100644
--- a/fs/btrfs/acl.c
+++ b/fs/btrfs/acl.c
@@ -60,15 +60,20 @@ static struct posix_acl *btrfs_get_acl(struct inode *inode, int type)
return ERR_PTR(-EINVAL);
}
+ /* Handle the cached NULL acl case without locking */
+ acl = ACCESS_ONCE(*p_acl);
+ if (!acl)
+ return acl;
+
spin_lock(&inode->i_lock);
- if (*p_acl != BTRFS_ACL_NOT_CACHED)
- acl = posix_acl_dup(*p_acl);
+ acl = *p_acl;
+ if (acl != BTRFS_ACL_NOT_CACHED)
+ acl = posix_acl_dup(acl);
spin_unlock(&inode->i_lock);
- if (acl)
+ if (acl != BTRFS_ACL_NOT_CACHED)
return acl;
-
size = __btrfs_getxattr(inode, name, "", 0);
if (size > 0) {
value = kzalloc(size, GFP_NOFS);
@@ -80,9 +85,12 @@ static struct posix_acl *btrfs_get_acl(struct inode *inode, int type)
btrfs_update_cached_acl(inode, p_acl, acl);
}
kfree(value);
- } else if (size == -ENOENT) {
+ } else if (size == -ENOENT || size == -ENODATA || size == 0) {
+ /* FIXME, who returns -ENOENT? I think nobody */
acl = NULL;
btrfs_update_cached_acl(inode, p_acl, acl);
+ } else {
+ acl = ERR_PTR(-EIO);
}
return acl;
@@ -256,7 +264,7 @@ int btrfs_init_acl(struct inode *inode, struct inode *dir)
}
if (!acl)
- inode->i_mode &= ~current->fs->umask;
+ inode->i_mode &= ~current_umask();
}
if (IS_POSIXACL(dir) && acl) {
diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c
index c84ca1f5259a..502c3d61de62 100644
--- a/fs/btrfs/async-thread.c
+++ b/fs/btrfs/async-thread.c
@@ -20,12 +20,12 @@
#include <linux/list.h>
#include <linux/spinlock.h>
#include <linux/freezer.h>
-#include <linux/ftrace.h>
#include "async-thread.h"
#define WORK_QUEUED_BIT 0
#define WORK_DONE_BIT 1
#define WORK_ORDER_DONE_BIT 2
+#define WORK_HIGH_PRIO_BIT 3
/*
* container for the kthread task pointer and the list of pending work
@@ -37,6 +37,7 @@ struct btrfs_worker_thread {
/* list of struct btrfs_work that are waiting for service */
struct list_head pending;
+ struct list_head prio_pending;
/* list of worker threads from struct btrfs_workers */
struct list_head worker_list;
@@ -104,10 +105,16 @@ static noinline int run_ordered_completions(struct btrfs_workers *workers,
spin_lock_irqsave(&workers->lock, flags);
- while (!list_empty(&workers->order_list)) {
- work = list_entry(workers->order_list.next,
- struct btrfs_work, order_list);
-
+ while (1) {
+ if (!list_empty(&workers->prio_order_list)) {
+ work = list_entry(workers->prio_order_list.next,
+ struct btrfs_work, order_list);
+ } else if (!list_empty(&workers->order_list)) {
+ work = list_entry(workers->order_list.next,
+ struct btrfs_work, order_list);
+ } else {
+ break;
+ }
if (!test_bit(WORK_DONE_BIT, &work->flags))
break;
@@ -144,8 +151,14 @@ static int worker_loop(void *arg)
do {
spin_lock_irq(&worker->lock);
again_locked:
- while (!list_empty(&worker->pending)) {
- cur = worker->pending.next;
+ while (1) {
+ if (!list_empty(&worker->prio_pending))
+ cur = worker->prio_pending.next;
+ else if (!list_empty(&worker->pending))
+ cur = worker->pending.next;
+ else
+ break;
+
work = list_entry(cur, struct btrfs_work, list);
list_del(&work->list);
clear_bit(WORK_QUEUED_BIT, &work->flags);
@@ -164,7 +177,6 @@ again_locked:
spin_lock_irq(&worker->lock);
check_idle_worker(worker);
-
}
if (freezing(current)) {
worker->working = 0;
@@ -179,7 +191,8 @@ again_locked:
* jump_in?
*/
smp_mb();
- if (!list_empty(&worker->pending))
+ if (!list_empty(&worker->pending) ||
+ !list_empty(&worker->prio_pending))
continue;
/*
@@ -192,13 +205,18 @@ again_locked:
*/
schedule_timeout(1);
smp_mb();
- if (!list_empty(&worker->pending))
+ if (!list_empty(&worker->pending) ||
+ !list_empty(&worker->prio_pending))
continue;
+ if (kthread_should_stop())
+ break;
+
/* still no more work?, sleep for real */
spin_lock_irq(&worker->lock);
set_current_state(TASK_INTERRUPTIBLE);
- if (!list_empty(&worker->pending))
+ if (!list_empty(&worker->pending) ||
+ !list_empty(&worker->prio_pending))
goto again_locked;
/*
@@ -208,7 +226,8 @@ again_locked:
worker->working = 0;
spin_unlock_irq(&worker->lock);
- schedule();
+ if (!kthread_should_stop())
+ schedule();
}
__set_current_state(TASK_RUNNING);
}
@@ -245,6 +264,7 @@ void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max)
INIT_LIST_HEAD(&workers->worker_list);
INIT_LIST_HEAD(&workers->idle_list);
INIT_LIST_HEAD(&workers->order_list);
+ INIT_LIST_HEAD(&workers->prio_order_list);
spin_lock_init(&workers->lock);
workers->max_workers = max;
workers->idle_thresh = 32;
@@ -270,6 +290,7 @@ int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
}
INIT_LIST_HEAD(&worker->pending);
+ INIT_LIST_HEAD(&worker->prio_pending);
INIT_LIST_HEAD(&worker->worker_list);
spin_lock_init(&worker->lock);
atomic_set(&worker->num_pending, 0);
@@ -393,7 +414,10 @@ int btrfs_requeue_work(struct btrfs_work *work)
goto out;
spin_lock_irqsave(&worker->lock, flags);
- list_add_tail(&work->list, &worker->pending);
+ if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags))
+ list_add_tail(&work->list, &worker->prio_pending);
+ else
+ list_add_tail(&work->list, &worker->pending);
atomic_inc(&worker->num_pending);
/* by definition we're busy, take ourselves off the idle
@@ -419,6 +443,11 @@ out:
return 0;
}
+void btrfs_set_work_high_prio(struct btrfs_work *work)
+{
+ set_bit(WORK_HIGH_PRIO_BIT, &work->flags);
+}
+
/*
* places a struct btrfs_work into the pending queue of one of the kthreads
*/
@@ -435,7 +464,12 @@ int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work)
worker = find_worker(workers);
if (workers->ordered) {
spin_lock_irqsave(&workers->lock, flags);
- list_add_tail(&work->order_list, &workers->order_list);
+ if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags)) {
+ list_add_tail(&work->order_list,
+ &workers->prio_order_list);
+ } else {
+ list_add_tail(&work->order_list, &workers->order_list);
+ }
spin_unlock_irqrestore(&workers->lock, flags);
} else {
INIT_LIST_HEAD(&work->order_list);
@@ -443,7 +477,10 @@ int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work)
spin_lock_irqsave(&worker->lock, flags);
- list_add_tail(&work->list, &worker->pending);
+ if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags))
+ list_add_tail(&work->list, &worker->prio_pending);
+ else
+ list_add_tail(&work->list, &worker->pending);
atomic_inc(&worker->num_pending);
check_busy_worker(worker);
diff --git a/fs/btrfs/async-thread.h b/fs/btrfs/async-thread.h
index 31be4ed8b63e..1b511c109db6 100644
--- a/fs/btrfs/async-thread.h
+++ b/fs/btrfs/async-thread.h
@@ -85,6 +85,7 @@ struct btrfs_workers {
* of work items waiting for completion
*/
struct list_head order_list;
+ struct list_head prio_order_list;
/* lock for finding the next worker thread to queue on */
spinlock_t lock;
@@ -98,4 +99,5 @@ int btrfs_start_workers(struct btrfs_workers *workers, int num_workers);
int btrfs_stop_workers(struct btrfs_workers *workers);
void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max);
int btrfs_requeue_work(struct btrfs_work *work);
+void btrfs_set_work_high_prio(struct btrfs_work *work);
#endif
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index dbb724124633..a99f1c2a710d 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -1244,9 +1244,9 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
* readahead one full node of leaves, finding things that are close
* to the block in 'slot', and triggering ra on them.
*/
-static noinline void reada_for_search(struct btrfs_root *root,
- struct btrfs_path *path,
- int level, int slot, u64 objectid)
+static void reada_for_search(struct btrfs_root *root,
+ struct btrfs_path *path,
+ int level, int slot, u64 objectid)
{
struct extent_buffer *node;
struct btrfs_disk_key disk_key;
@@ -1325,12 +1325,12 @@ static noinline int reada_for_balance(struct btrfs_root *root,
int ret = 0;
int blocksize;
- parent = path->nodes[level - 1];
+ parent = path->nodes[level + 1];
if (!parent)
return 0;
nritems = btrfs_header_nritems(parent);
- slot = path->slots[level];
+ slot = path->slots[level + 1];
blocksize = btrfs_level_size(root, level);
if (slot > 0) {
@@ -1341,7 +1341,7 @@ static noinline int reada_for_balance(struct btrfs_root *root,
block1 = 0;
free_extent_buffer(eb);
}
- if (slot < nritems) {
+ if (slot + 1 < nritems) {
block2 = btrfs_node_blockptr(parent, slot + 1);
gen = btrfs_node_ptr_generation(parent, slot + 1);
eb = btrfs_find_tree_block(root, block2, blocksize);
@@ -1351,7 +1351,11 @@ static noinline int reada_for_balance(struct btrfs_root *root,
}
if (block1 || block2) {
ret = -EAGAIN;
+
+ /* release the whole path */
btrfs_release_path(root, path);
+
+ /* read the blocks */
if (block1)
readahead_tree_block(root, block1, blocksize, 0);
if (block2)
@@ -1361,7 +1365,7 @@ static noinline int reada_for_balance(struct btrfs_root *root,
eb = read_tree_block(root, block1, blocksize, 0);
free_extent_buffer(eb);
}
- if (block1) {
+ if (block2) {
eb = read_tree_block(root, block2, blocksize, 0);
free_extent_buffer(eb);
}
@@ -1447,6 +1451,120 @@ noinline void btrfs_unlock_up_safe(struct btrfs_path *path, int level)
}
/*
+ * helper function for btrfs_search_slot. The goal is to find a block
+ * in cache without setting the path to blocking. If we find the block
+ * we return zero and the path is unchanged.
+ *
+ * If we can't find the block, we set the path blocking and do some
+ * reada. -EAGAIN is returned and the search must be repeated.
+ */
+static int
+read_block_for_search(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, struct btrfs_path *p,
+ struct extent_buffer **eb_ret, int level, int slot,
+ struct btrfs_key *key)
+{
+ u64 blocknr;
+ u64 gen;
+ u32 blocksize;
+ struct extent_buffer *b = *eb_ret;
+ struct extent_buffer *tmp;
+
+ blocknr = btrfs_node_blockptr(b, slot);
+ gen = btrfs_node_ptr_generation(b, slot);
+ blocksize = btrfs_level_size(root, level - 1);
+
+ tmp = btrfs_find_tree_block(root, blocknr, blocksize);
+ if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
+ *eb_ret = tmp;
+ return 0;
+ }
+
+ /*
+ * reduce lock contention at high levels
+ * of the btree by dropping locks before
+ * we read.
+ */
+ btrfs_unlock_up_safe(p, level + 1);
+ btrfs_set_path_blocking(p);
+
+ if (tmp)
+ free_extent_buffer(tmp);
+ if (p->reada)
+ reada_for_search(root, p, level, slot, key->objectid);
+
+ btrfs_release_path(NULL, p);
+ tmp = read_tree_block(root, blocknr, blocksize, gen);
+ if (tmp)
+ free_extent_buffer(tmp);
+ return -EAGAIN;
+}
+
+/*
+ * helper function for btrfs_search_slot. This does all of the checks
+ * for node-level blocks and does any balancing required based on
+ * the ins_len.
+ *
+ * If no extra work was required, zero is returned. If we had to
+ * drop the path, -EAGAIN is returned and btrfs_search_slot must
+ * start over
+ */
+static int
+setup_nodes_for_search(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, struct btrfs_path *p,
+ struct extent_buffer *b, int level, int ins_len)
+{
+ int ret;
+ if ((p->search_for_split || ins_len > 0) && btrfs_header_nritems(b) >=
+ BTRFS_NODEPTRS_PER_BLOCK(root) - 3) {
+ int sret;
+
+ sret = reada_for_balance(root, p, level);
+ if (sret)
+ goto again;
+
+ btrfs_set_path_blocking(p);
+ sret = split_node(trans, root, p, level);
+ btrfs_clear_path_blocking(p, NULL);
+
+ BUG_ON(sret > 0);
+ if (sret) {
+ ret = sret;
+ goto done;
+ }
+ b = p->nodes[level];
+ } else if (ins_len < 0 && btrfs_header_nritems(b) <
+ BTRFS_NODEPTRS_PER_BLOCK(root) / 4) {
+ int sret;
+
+ sret = reada_for_balance(root, p, level);
+ if (sret)
+ goto again;
+
+ btrfs_set_path_blocking(p);
+ sret = balance_level(trans, root, p, level);
+ btrfs_clear_path_blocking(p, NULL);
+
+ if (sret) {
+ ret = sret;
+ goto done;
+ }
+ b = p->nodes[level];
+ if (!b) {
+ btrfs_release_path(NULL, p);
+ goto again;
+ }
+ BUG_ON(btrfs_header_nritems(b) == 1);
+ }
+ return 0;
+
+again:
+ ret = -EAGAIN;
+done:
+ return ret;
+}
+
+/*
* look for key in the tree. path is filled in with nodes along the way
* if key is found, we return zero and you can find the item in the leaf
* level of the path (level 0)
@@ -1464,16 +1582,11 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
ins_len, int cow)
{
struct extent_buffer *b;
- struct extent_buffer *tmp;
int slot;
int ret;
int level;
- int should_reada = p->reada;
int lowest_unlock = 1;
- int blocksize;
u8 lowest_level = 0;
- u64 blocknr;
- u64 gen;
lowest_level = p->lowest_level;
WARN_ON(lowest_level && ins_len > 0);
@@ -1502,7 +1615,11 @@ again:
if (cow) {
int wret;
- /* is a cow on this block not required */
+ /*
+ * if we don't really need to cow this block
+ * then we don't want to set the path blocking,
+ * so we test it here
+ */
if (btrfs_header_generation(b) == trans->transid &&
btrfs_header_owner(b) == root->root_key.objectid &&
!btrfs_header_flag(b, BTRFS_HEADER_FLAG_WRITTEN)) {
@@ -1557,51 +1674,15 @@ cow_done:
if (ret && slot > 0)
slot -= 1;
p->slots[level] = slot;
- if ((p->search_for_split || ins_len > 0) &&
- btrfs_header_nritems(b) >=
- BTRFS_NODEPTRS_PER_BLOCK(root) - 3) {
- int sret;
-
- sret = reada_for_balance(root, p, level);
- if (sret)
- goto again;
-
- btrfs_set_path_blocking(p);
- sret = split_node(trans, root, p, level);
- btrfs_clear_path_blocking(p, NULL);
-
- BUG_ON(sret > 0);
- if (sret) {
- ret = sret;
- goto done;
- }
- b = p->nodes[level];
- slot = p->slots[level];
- } else if (ins_len < 0 &&
- btrfs_header_nritems(b) <
- BTRFS_NODEPTRS_PER_BLOCK(root) / 4) {
- int sret;
-
- sret = reada_for_balance(root, p, level);
- if (sret)
- goto again;
-
- btrfs_set_path_blocking(p);
- sret = balance_level(trans, root, p, level);
- btrfs_clear_path_blocking(p, NULL);
+ ret = setup_nodes_for_search(trans, root, p, b, level,
+ ins_len);
+ if (ret == -EAGAIN)
+ goto again;
+ else if (ret)
+ goto done;
+ b = p->nodes[level];
+ slot = p->slots[level];
- if (sret) {
- ret = sret;
- goto done;
- }
- b = p->nodes[level];
- if (!b) {
- btrfs_release_path(NULL, p);
- goto again;
- }
- slot = p->slots[level];
- BUG_ON(btrfs_header_nritems(b) == 1);
- }
unlock_up(p, level, lowest_unlock);
/* this is only true while dropping a snapshot */
@@ -1610,44 +1691,11 @@ cow_done:
goto done;
}
- blocknr = btrfs_node_blockptr(b, slot);
- gen = btrfs_node_ptr_generation(b, slot);
- blocksize = btrfs_level_size(root, level - 1);
+ ret = read_block_for_search(trans, root, p,
+ &b, level, slot, key);
+ if (ret == -EAGAIN)
+ goto again;
- tmp = btrfs_find_tree_block(root, blocknr, blocksize);
- if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
- b = tmp;
- } else {
- /*
- * reduce lock contention at high levels
- * of the btree by dropping locks before
- * we read.
- */
- if (level > 0) {
- btrfs_release_path(NULL, p);
- if (tmp)
- free_extent_buffer(tmp);
- if (should_reada)
- reada_for_search(root, p,
- level, slot,
- key->objectid);
-
- tmp = read_tree_block(root, blocknr,
- blocksize, gen);
- if (tmp)
- free_extent_buffer(tmp);
- goto again;
- } else {
- btrfs_set_path_blocking(p);
- if (tmp)
- free_extent_buffer(tmp);
- if (should_reada)
- reada_for_search(root, p,
- level, slot,
- key->objectid);
- b = read_node_slot(root, b, slot);
- }
- }
if (!p->skip_locking) {
int lret;
@@ -2116,8 +2164,7 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
BUG_ON(!path->nodes[level]);
lower = path->nodes[level];
nritems = btrfs_header_nritems(lower);
- if (slot > nritems)
- BUG();
+ BUG_ON(slot > nritems);
if (nritems == BTRFS_NODEPTRS_PER_BLOCK(root))
BUG();
if (slot != nritems) {
@@ -4086,28 +4133,44 @@ next:
int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
{
int slot;
- int level = 1;
+ int level;
struct extent_buffer *c;
- struct extent_buffer *next = NULL;
+ struct extent_buffer *next;
struct btrfs_key key;
u32 nritems;
int ret;
+ int old_spinning = path->leave_spinning;
+ int force_blocking = 0;
nritems = btrfs_header_nritems(path->nodes[0]);
if (nritems == 0)
return 1;
- btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
+ /*
+ * we take the blocks in an order that upsets lockdep. Using
+ * blocking mode is the only way around it.
+ */
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+ force_blocking = 1;
+#endif
+ btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
+again:
+ level = 1;
+ next = NULL;
btrfs_release_path(root, path);
+
path->keep_locks = 1;
+
+ if (!force_blocking)
+ path->leave_spinning = 1;
+
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
path->keep_locks = 0;
if (ret < 0)
return ret;
- btrfs_set_path_blocking(path);
nritems = btrfs_header_nritems(path->nodes[0]);
/*
* by releasing the path above we dropped all our locks. A balance
@@ -4117,19 +4180,24 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
*/
if (nritems > 0 && path->slots[0] < nritems - 1) {
path->slots[0]++;
+ ret = 0;
goto done;
}
while (level < BTRFS_MAX_LEVEL) {
- if (!path->nodes[level])
- return 1;
+ if (!path->nodes[level]) {
+ ret = 1;
+ goto done;
+ }
slot = path->slots[level] + 1;
c = path->nodes[level];
if (slot >= btrfs_header_nritems(c)) {
level++;
- if (level == BTRFS_MAX_LEVEL)
- return 1;
+ if (level == BTRFS_MAX_LEVEL) {
+ ret = 1;
+ goto done;
+ }
continue;
}
@@ -4138,16 +4206,22 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
free_extent_buffer(next);
}
- /* the path was set to blocking above */
- if (level == 1 && (path->locks[1] || path->skip_locking) &&
- path->reada)
- reada_for_search(root, path, level, slot, 0);
+ next = c;
+ ret = read_block_for_search(NULL, root, path, &next, level,
+ slot, &key);
+ if (ret == -EAGAIN)
+ goto again;
- next = read_node_slot(root, c, slot);
if (!path->skip_locking) {
- btrfs_assert_tree_locked(c);
- btrfs_tree_lock(next);
- btrfs_set_lock_blocking(next);
+ ret = btrfs_try_spin_lock(next);
+ if (!ret) {
+ btrfs_set_path_blocking(path);
+ btrfs_tree_lock(next);
+ if (!force_blocking)
+ btrfs_clear_path_blocking(path, next);
+ }
+ if (force_blocking)
+ btrfs_set_lock_blocking(next);
}
break;
}
@@ -4157,27 +4231,42 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
c = path->nodes[level];
if (path->locks[level])
btrfs_tree_unlock(c);
+
free_extent_buffer(c);
path->nodes[level] = next;
path->slots[level] = 0;
if (!path->skip_locking)
path->locks[level] = 1;
+
if (!level)
break;
- btrfs_set_path_blocking(path);
- if (level == 1 && path->locks[1] && path->reada)
- reada_for_search(root, path, level, slot, 0);
- next = read_node_slot(root, next, 0);
+ ret = read_block_for_search(NULL, root, path, &next, level,
+ 0, &key);
+ if (ret == -EAGAIN)
+ goto again;
+
if (!path->skip_locking) {
btrfs_assert_tree_locked(path->nodes[level]);
- btrfs_tree_lock(next);
- btrfs_set_lock_blocking(next);
+ ret = btrfs_try_spin_lock(next);
+ if (!ret) {
+ btrfs_set_path_blocking(path);
+ btrfs_tree_lock(next);
+ if (!force_blocking)
+ btrfs_clear_path_blocking(path, next);
+ }
+ if (force_blocking)
+ btrfs_set_lock_blocking(next);
}
}
+ ret = 0;
done:
unlock_up(path, 0, 1);
- return 0;
+ path->leave_spinning = old_spinning;
+ if (!old_spinning)
+ btrfs_set_path_blocking(path);
+
+ return ret;
}
/*
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 9417713542a2..4414a5d9983a 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -143,12 +143,15 @@ static int btrfs_csum_sizes[] = { 4, 0 };
#define BTRFS_FT_MAX 9
/*
- * the key defines the order in the tree, and so it also defines (optimal)
- * block layout. objectid corresonds to the inode number. The flags
- * tells us things about the object, and is a kind of stream selector.
- * so for a given inode, keys with flags of 1 might refer to the inode
- * data, flags of 2 may point to file data in the btree and flags == 3
- * may point to extents.
+ * The key defines the order in the tree, and so it also defines (optimal)
+ * block layout.
+ *
+ * objectid corresponds to the inode number.
+ *
+ * type tells us things about the object, and is a kind of stream selector.
+ * so for a given inode, keys with type of 1 might refer to the inode data,
+ * type of 2 may point to file data in the btree and type == 3 may point to
+ * extents.
*
* offset is the starting byte offset for this key in the stream.
*
@@ -200,7 +203,7 @@ struct btrfs_dev_item {
/*
* starting byte of this partition on the device,
- * to allowr for stripe alignment in the future
+ * to allow for stripe alignment in the future
*/
__le64 start_offset;
@@ -633,18 +636,35 @@ struct btrfs_space_info {
struct rw_semaphore groups_sem;
};
-struct btrfs_free_space {
- struct rb_node bytes_index;
- struct rb_node offset_index;
- u64 offset;
- u64 bytes;
+/*
+ * free clusters are used to claim free space in relatively large chunks,
+ * allowing us to do less seeky writes. They are used for all metadata
+ * allocations and data allocations in ssd mode.
+ */
+struct btrfs_free_cluster {
+ spinlock_t lock;
+ spinlock_t refill_lock;
+ struct rb_root root;
+
+ /* largest extent in this cluster */
+ u64 max_size;
+
+ /* first extent starting offset */
+ u64 window_start;
+
+ struct btrfs_block_group_cache *block_group;
+ /*
+ * when a cluster is allocated from a block group, we put the
+ * cluster onto a list in the block group so that it can
+ * be freed before the block group is freed.
+ */
+ struct list_head block_group_list;
};
struct btrfs_block_group_cache {
struct btrfs_key key;
struct btrfs_block_group_item item;
spinlock_t lock;
- struct mutex alloc_mutex;
struct mutex cache_mutex;
u64 pinned;
u64 reserved;
@@ -656,6 +676,7 @@ struct btrfs_block_group_cache {
struct btrfs_space_info *space_info;
/* free space cache stuff */
+ spinlock_t tree_lock;
struct rb_root free_space_bytes;
struct rb_root free_space_offset;
@@ -667,6 +688,11 @@ struct btrfs_block_group_cache {
/* usage count */
atomic_t count;
+
+ /* List of struct btrfs_free_clusters for this block group.
+ * Today it will only have one thing on it, but that may change
+ */
+ struct list_head cluster_list;
};
struct btrfs_leaf_ref_tree {
@@ -728,7 +754,6 @@ struct btrfs_fs_info {
struct mutex tree_log_mutex;
struct mutex transaction_kthread_mutex;
struct mutex cleaner_mutex;
- struct mutex pinned_mutex;
struct mutex chunk_mutex;
struct mutex drop_mutex;
struct mutex volume_mutex;
@@ -839,8 +864,12 @@ struct btrfs_fs_info {
spinlock_t delalloc_lock;
spinlock_t new_trans_lock;
u64 delalloc_bytes;
- u64 last_alloc;
- u64 last_data_alloc;
+
+ /* data_alloc_cluster is only used in ssd mode */
+ struct btrfs_free_cluster data_alloc_cluster;
+
+ /* all metadata allocations go through this cluster */
+ struct btrfs_free_cluster meta_alloc_cluster;
spinlock_t ref_cache_lock;
u64 total_ref_cache_size;
@@ -852,6 +881,9 @@ struct btrfs_fs_info {
u64 metadata_alloc_profile;
u64 system_alloc_profile;
+ unsigned data_chunk_allocations;
+ unsigned metadata_ratio;
+
void *bdev_holder;
};
@@ -932,7 +964,6 @@ struct btrfs_root {
};
/*
-
* inode items have the data typically returned from stat and store other
* info about object characteristics. There is one for every file and dir in
* the FS
@@ -963,7 +994,7 @@ struct btrfs_root {
#define BTRFS_EXTENT_CSUM_KEY 128
/*
- * root items point to tree roots. There are typically in the root
+ * root items point to tree roots. They are typically in the root
* tree used by the super block to find all the other trees
*/
#define BTRFS_ROOT_ITEM_KEY 132
@@ -1010,6 +1041,8 @@ struct btrfs_root {
#define BTRFS_MOUNT_SSD (1 << 3)
#define BTRFS_MOUNT_DEGRADED (1 << 4)
#define BTRFS_MOUNT_COMPRESS (1 << 5)
+#define BTRFS_MOUNT_NOTREELOG (1 << 6)
+#define BTRFS_MOUNT_FLUSHONCOMMIT (1 << 7)
#define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt)
#define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt)
@@ -1748,6 +1781,7 @@ static inline struct dentry *fdentry(struct file *file)
}
/* extent-tree.c */
+void btrfs_put_block_group(struct btrfs_block_group_cache *cache);
int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
struct btrfs_root *root, unsigned long count);
int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len);
@@ -2143,7 +2177,8 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode);
extern struct file_operations btrfs_file_operations;
int btrfs_drop_extents(struct btrfs_trans_handle *trans,
struct btrfs_root *root, struct inode *inode,
- u64 start, u64 end, u64 inline_limit, u64 *hint_block);
+ u64 start, u64 end, u64 locked_end,
+ u64 inline_limit, u64 *hint_block);
int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct inode *inode, u64 start, u64 end);
@@ -2174,21 +2209,4 @@ int btrfs_check_acl(struct inode *inode, int mask);
int btrfs_init_acl(struct inode *inode, struct inode *dir);
int btrfs_acl_chmod(struct inode *inode);
-/* free-space-cache.c */
-int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
- u64 bytenr, u64 size);
-int btrfs_add_free_space_lock(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes);
-int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
- u64 bytenr, u64 size);
-int btrfs_remove_free_space_lock(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes);
-void btrfs_remove_free_space_cache(struct btrfs_block_group_cache
- *block_group);
-struct btrfs_free_space *btrfs_find_free_space(struct btrfs_block_group_cache
- *block_group, u64 offset,
- u64 bytes);
-void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
- u64 bytes);
-u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group);
#endif
diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index cbf7dc8ae3ec..d6c01c096a40 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -18,7 +18,6 @@
#include <linux/sched.h>
#include <linux/sort.h>
-#include <linux/ftrace.h>
#include "ctree.h"
#include "delayed-ref.h"
#include "transaction.h"
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 92d73929d381..0ff16d3331da 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -38,6 +38,7 @@
#include "locking.h"
#include "ref-cache.h"
#include "tree-log.h"
+#include "free-space-cache.h"
static struct extent_io_ops btree_extent_io_ops;
static void end_workqueue_fn(struct btrfs_work *work);
@@ -231,10 +232,14 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
memcpy(&found, result, csum_size);
read_extent_buffer(buf, &val, 0, csum_size);
- printk(KERN_INFO "btrfs: %s checksum verify failed "
- "on %llu wanted %X found %X level %d\n",
- root->fs_info->sb->s_id,
- buf->start, val, found, btrfs_header_level(buf));
+ if (printk_ratelimit()) {
+ printk(KERN_INFO "btrfs: %s checksum verify "
+ "failed on %llu wanted %X found %X "
+ "level %d\n",
+ root->fs_info->sb->s_id,
+ (unsigned long long)buf->start, val, found,
+ btrfs_header_level(buf));
+ }
if (result != (char *)&inline_result)
kfree(result);
return 1;
@@ -267,10 +272,13 @@ static int verify_parent_transid(struct extent_io_tree *io_tree,
ret = 0;
goto out;
}
- printk("parent transid verify failed on %llu wanted %llu found %llu\n",
- (unsigned long long)eb->start,
- (unsigned long long)parent_transid,
- (unsigned long long)btrfs_header_generation(eb));
+ if (printk_ratelimit()) {
+ printk("parent transid verify failed on %llu wanted %llu "
+ "found %llu\n",
+ (unsigned long long)eb->start,
+ (unsigned long long)parent_transid,
+ (unsigned long long)btrfs_header_generation(eb));
+ }
ret = 1;
clear_extent_buffer_uptodate(io_tree, eb);
out:
@@ -414,9 +422,12 @@ static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
found_start = btrfs_header_bytenr(eb);
if (found_start != start) {
- printk(KERN_INFO "btrfs bad tree block start %llu %llu\n",
- (unsigned long long)found_start,
- (unsigned long long)eb->start);
+ if (printk_ratelimit()) {
+ printk(KERN_INFO "btrfs bad tree block start "
+ "%llu %llu\n",
+ (unsigned long long)found_start,
+ (unsigned long long)eb->start);
+ }
ret = -EIO;
goto err;
}
@@ -428,8 +439,10 @@ static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
goto err;
}
if (check_tree_block_fsid(root, eb)) {
- printk(KERN_INFO "btrfs bad fsid on block %llu\n",
- (unsigned long long)eb->start);
+ if (printk_ratelimit()) {
+ printk(KERN_INFO "btrfs bad fsid on block %llu\n",
+ (unsigned long long)eb->start);
+ }
ret = -EIO;
goto err;
}
@@ -578,19 +591,12 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
async->bio_flags = bio_flags;
atomic_inc(&fs_info->nr_async_submits);
+
+ if (rw & (1 << BIO_RW_SYNCIO))
+ btrfs_set_work_high_prio(&async->work);
+
btrfs_queue_worker(&fs_info->workers, &async->work);
-#if 0
- int limit = btrfs_async_submit_limit(fs_info);
- if (atomic_read(&fs_info->nr_async_submits) > limit) {
- wait_event_timeout(fs_info->async_submit_wait,
- (atomic_read(&fs_info->nr_async_submits) < limit),
- HZ/10);
- wait_event_timeout(fs_info->async_submit_wait,
- (atomic_read(&fs_info->nr_async_bios) < limit),
- HZ/10);
- }
-#endif
while (atomic_read(&fs_info->async_submit_draining) &&
atomic_read(&fs_info->nr_async_submits)) {
wait_event(fs_info->async_submit_wait,
@@ -655,6 +661,7 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
mirror_num, 0);
}
+
/*
* kthread helpers are used to submit writes so that checksumming
* can happen in parallel across all CPUs
@@ -764,27 +771,6 @@ static void btree_invalidatepage(struct page *page, unsigned long offset)
}
}
-#if 0
-static int btree_writepage(struct page *page, struct writeback_control *wbc)
-{
- struct buffer_head *bh;
- struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
- struct buffer_head *head;
- if (!page_has_buffers(page)) {
- create_empty_buffers(page, root->fs_info->sb->s_blocksize,
- (1 << BH_Dirty)|(1 << BH_Uptodate));
- }
- head = page_buffers(page);
- bh = head;
- do {
- if (buffer_dirty(bh))
- csum_tree_block(root, bh, 0);
- bh = bh->b_this_page;
- } while (bh != head);
- return block_write_full_page(page, btree_get_block, wbc);
-}
-#endif
-
static struct address_space_operations btree_aops = {
.readpage = btree_readpage,
.writepage = btree_writepage,
@@ -1272,11 +1258,7 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits)
int ret = 0;
struct btrfs_device *device;
struct backing_dev_info *bdi;
-#if 0
- if ((bdi_bits & (1 << BDI_write_congested)) &&
- btrfs_congested_async(info, 0))
- return 1;
-#endif
+
list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
if (!device->bdev)
continue;
@@ -1412,8 +1394,6 @@ static int bio_ready_for_csum(struct bio *bio)
ret = extent_range_uptodate(io_tree, start + length,
start + buf_len - 1);
- if (ret == 1)
- return ret;
return ret;
}
@@ -1600,6 +1580,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
fs_info->btree_inode = new_inode(sb);
fs_info->btree_inode->i_ino = 1;
fs_info->btree_inode->i_nlink = 1;
+ fs_info->metadata_ratio = 8;
fs_info->thread_pool_size = min_t(unsigned long,
num_online_cpus() + 2, 8);
@@ -1647,12 +1628,15 @@ struct btrfs_root *open_ctree(struct super_block *sb,
mutex_init(&fs_info->ordered_operations_mutex);
mutex_init(&fs_info->tree_log_mutex);
mutex_init(&fs_info->drop_mutex);
- mutex_init(&fs_info->pinned_mutex);
mutex_init(&fs_info->chunk_mutex);
mutex_init(&fs_info->transaction_kthread_mutex);
mutex_init(&fs_info->cleaner_mutex);
mutex_init(&fs_info->volume_mutex);
mutex_init(&fs_info->tree_reloc_mutex);
+
+ btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
+ btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
+
init_waitqueue_head(&fs_info->transaction_throttle);
init_waitqueue_head(&fs_info->transaction_wait);
init_waitqueue_head(&fs_info->async_submit_wait);
@@ -1687,7 +1671,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
if (features) {
printk(KERN_ERR "BTRFS: couldn't mount because of "
"unsupported optional features (%Lx).\n",
- features);
+ (unsigned long long)features);
err = -EINVAL;
goto fail_iput;
}
@@ -1697,7 +1681,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
if (!(sb->s_flags & MS_RDONLY) && features) {
printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
"unsupported option features (%Lx).\n",
- features);
+ (unsigned long long)features);
err = -EINVAL;
goto fail_iput;
}
@@ -2093,10 +2077,10 @@ static int write_dev_supers(struct btrfs_device *device,
device->barriers = 0;
get_bh(bh);
lock_buffer(bh);
- ret = submit_bh(WRITE, bh);
+ ret = submit_bh(WRITE_SYNC, bh);
}
} else {
- ret = submit_bh(WRITE, bh);
+ ret = submit_bh(WRITE_SYNC, bh);
}
if (!ret && wait) {
@@ -2289,7 +2273,7 @@ int close_ctree(struct btrfs_root *root)
if (fs_info->delalloc_bytes) {
printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
- fs_info->delalloc_bytes);
+ (unsigned long long)fs_info->delalloc_bytes);
}
if (fs_info->total_ref_cache_size) {
printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
@@ -2326,16 +2310,6 @@ int close_ctree(struct btrfs_root *root)
btrfs_stop_workers(&fs_info->endio_write_workers);
btrfs_stop_workers(&fs_info->submit_workers);
-#if 0
- while (!list_empty(&fs_info->hashers)) {
- struct btrfs_hasher *hasher;
- hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher,
- hashers);
- list_del(&hasher->hashers);
- crypto_free_hash(&fs_info->hash_tfm);
- kfree(hasher);
- }
-#endif
btrfs_close_devices(fs_info->fs_devices);
btrfs_mapping_tree_free(&fs_info->mapping_tree);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index f5e7cae63d80..e4966444811b 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -31,6 +31,7 @@
#include "volumes.h"
#include "locking.h"
#include "ref-cache.h"
+#include "free-space-cache.h"
#define PENDING_EXTENT_INSERT 0
#define PENDING_EXTENT_DELETE 1
@@ -166,7 +167,6 @@ static int add_new_free_space(struct btrfs_block_group_cache *block_group,
u64 extent_start, extent_end, size;
int ret;
- mutex_lock(&info->pinned_mutex);
while (start < end) {
ret = find_first_extent_bit(&info->pinned_extents, start,
&extent_start, &extent_end,
@@ -192,7 +192,6 @@ static int add_new_free_space(struct btrfs_block_group_cache *block_group,
ret = btrfs_add_free_space(block_group, start, size);
BUG_ON(ret);
}
- mutex_unlock(&info->pinned_mutex);
return 0;
}
@@ -291,8 +290,8 @@ next:
block_group->key.objectid +
block_group->key.offset);
- remove_sb_from_cache(root, block_group);
block_group->cached = 1;
+ remove_sb_from_cache(root, block_group);
ret = 0;
err:
btrfs_free_path(path);
@@ -326,7 +325,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(
return cache;
}
-static inline void put_block_group(struct btrfs_block_group_cache *cache)
+void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
{
if (atomic_dec_and_test(&cache->count))
kfree(cache);
@@ -399,12 +398,12 @@ again:
div_factor(cache->key.offset, factor)) {
group_start = cache->key.objectid;
spin_unlock(&cache->lock);
- put_block_group(cache);
+ btrfs_put_block_group(cache);
goto found;
}
}
spin_unlock(&cache->lock);
- put_block_group(cache);
+ btrfs_put_block_group(cache);
cond_resched();
}
if (!wrapped) {
@@ -1594,7 +1593,7 @@ int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
if (!block_group || block_group->ro)
readonly = 1;
if (block_group)
- put_block_group(block_group);
+ btrfs_put_block_group(block_group);
return readonly;
}
@@ -1845,10 +1844,14 @@ again:
printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
", %llu bytes_used, %llu bytes_reserved, "
"%llu bytes_pinned, %llu bytes_readonly, %llu may use"
- "%llu total\n", bytes, data_sinfo->bytes_delalloc,
- data_sinfo->bytes_used, data_sinfo->bytes_reserved,
- data_sinfo->bytes_pinned, data_sinfo->bytes_readonly,
- data_sinfo->bytes_may_use, data_sinfo->total_bytes);
+ "%llu total\n", (unsigned long long)bytes,
+ (unsigned long long)data_sinfo->bytes_delalloc,
+ (unsigned long long)data_sinfo->bytes_used,
+ (unsigned long long)data_sinfo->bytes_reserved,
+ (unsigned long long)data_sinfo->bytes_pinned,
+ (unsigned long long)data_sinfo->bytes_readonly,
+ (unsigned long long)data_sinfo->bytes_may_use,
+ (unsigned long long)data_sinfo->total_bytes);
return -ENOSPC;
}
data_sinfo->bytes_may_use += bytes;
@@ -1919,15 +1922,29 @@ void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
spin_unlock(&info->lock);
}
+static void force_metadata_allocation(struct btrfs_fs_info *info)
+{
+ struct list_head *head = &info->space_info;
+ struct btrfs_space_info *found;
+
+ rcu_read_lock();
+ list_for_each_entry_rcu(found, head, list) {
+ if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
+ found->force_alloc = 1;
+ }
+ rcu_read_unlock();
+}
+
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root, u64 alloc_bytes,
u64 flags, int force)
{
struct btrfs_space_info *space_info;
+ struct btrfs_fs_info *fs_info = extent_root->fs_info;
u64 thresh;
int ret = 0;
- mutex_lock(&extent_root->fs_info->chunk_mutex);
+ mutex_lock(&fs_info->chunk_mutex);
flags = btrfs_reduce_alloc_profile(extent_root, flags);
@@ -1959,6 +1976,18 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans,
}
spin_unlock(&space_info->lock);
+ /*
+ * if we're doing a data chunk, go ahead and make sure that
+ * we keep a reasonable number of metadata chunks allocated in the
+ * FS as well.
+ */
+ if (flags & BTRFS_BLOCK_GROUP_DATA) {
+ fs_info->data_chunk_allocations++;
+ if (!(fs_info->data_chunk_allocations %
+ fs_info->metadata_ratio))
+ force_metadata_allocation(fs_info);
+ }
+
ret = btrfs_alloc_chunk(trans, extent_root, flags);
if (ret)
space_info->full = 1;
@@ -2018,7 +2047,7 @@ static int update_block_group(struct btrfs_trans_handle *trans,
WARN_ON(ret);
}
}
- put_block_group(cache);
+ btrfs_put_block_group(cache);
total -= num_bytes;
bytenr += num_bytes;
}
@@ -2035,7 +2064,7 @@ static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
return 0;
bytenr = cache->key.objectid;
- put_block_group(cache);
+ btrfs_put_block_group(cache);
return bytenr;
}
@@ -2047,7 +2076,6 @@ int btrfs_update_pinned_extents(struct btrfs_root *root,
struct btrfs_block_group_cache *cache;
struct btrfs_fs_info *fs_info = root->fs_info;
- WARN_ON(!mutex_is_locked(&root->fs_info->pinned_mutex));
if (pin) {
set_extent_dirty(&fs_info->pinned_extents,
bytenr, bytenr + num - 1, GFP_NOFS);
@@ -2055,7 +2083,6 @@ int btrfs_update_pinned_extents(struct btrfs_root *root,
clear_extent_dirty(&fs_info->pinned_extents,
bytenr, bytenr + num - 1, GFP_NOFS);
}
- mutex_unlock(&root->fs_info->pinned_mutex);
while (num > 0) {
cache = btrfs_lookup_block_group(fs_info, bytenr);
@@ -2081,7 +2108,7 @@ int btrfs_update_pinned_extents(struct btrfs_root *root,
if (cache->cached)
btrfs_add_free_space(cache, bytenr, len);
}
- put_block_group(cache);
+ btrfs_put_block_group(cache);
bytenr += len;
num -= len;
}
@@ -2112,7 +2139,7 @@ static int update_reserved_extents(struct btrfs_root *root,
}
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
- put_block_group(cache);
+ btrfs_put_block_group(cache);
bytenr += len;
num -= len;
}
@@ -2127,7 +2154,6 @@ int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
int ret;
- mutex_lock(&root->fs_info->pinned_mutex);
while (1) {
ret = find_first_extent_bit(pinned_extents, last,
&start, &end, EXTENT_DIRTY);
@@ -2136,7 +2162,6 @@ int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
set_extent_dirty(copy, start, end, GFP_NOFS);
last = end + 1;
}
- mutex_unlock(&root->fs_info->pinned_mutex);
return 0;
}
@@ -2149,7 +2174,6 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
int ret;
while (1) {
- mutex_lock(&root->fs_info->pinned_mutex);
ret = find_first_extent_bit(unpin, 0, &start, &end,
EXTENT_DIRTY);
if (ret)
@@ -2163,7 +2187,6 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
cond_resched();
}
- mutex_unlock(&root->fs_info->pinned_mutex);
return ret;
}
@@ -2205,7 +2228,6 @@ static int pin_down_bytes(struct btrfs_trans_handle *trans,
free_extent_buffer(buf);
pinit:
btrfs_set_path_blocking(path);
- mutex_lock(&root->fs_info->pinned_mutex);
/* unlocks the pinned mutex */
btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
@@ -2511,8 +2533,6 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans,
*/
if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID &&
owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
- mutex_lock(&root->fs_info->pinned_mutex);
-
/* unlocks the pinned mutex */
btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
update_reserved_extents(root, bytenr, num_bytes, 0);
@@ -2554,228 +2574,237 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans,
{
int ret = 0;
struct btrfs_root *root = orig_root->fs_info->extent_root;
- u64 total_needed = num_bytes;
- u64 *last_ptr = NULL;
- u64 last_wanted = 0;
+ struct btrfs_free_cluster *last_ptr = NULL;
struct btrfs_block_group_cache *block_group = NULL;
- int chunk_alloc_done = 0;
int empty_cluster = 2 * 1024 * 1024;
int allowed_chunk_alloc = 0;
- struct list_head *head = NULL, *cur = NULL;
- int loop = 0;
- int extra_loop = 0;
struct btrfs_space_info *space_info;
+ int last_ptr_loop = 0;
+ int loop = 0;
WARN_ON(num_bytes < root->sectorsize);
btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
ins->objectid = 0;
ins->offset = 0;
+ space_info = __find_space_info(root->fs_info, data);
+
if (orig_root->ref_cows || empty_size)
allowed_chunk_alloc = 1;
if (data & BTRFS_BLOCK_GROUP_METADATA) {
- last_ptr = &root->fs_info->last_alloc;
+ last_ptr = &root->fs_info->meta_alloc_cluster;
if (!btrfs_test_opt(root, SSD))
empty_cluster = 64 * 1024;
}
- if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD))
- last_ptr = &root->fs_info->last_data_alloc;
+ if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
+ last_ptr = &root->fs_info->data_alloc_cluster;
+ }
if (last_ptr) {
- if (*last_ptr) {
- hint_byte = *last_ptr;
- last_wanted = *last_ptr;
- } else
- empty_size += empty_cluster;
- } else {
- empty_cluster = 0;
+ spin_lock(&last_ptr->lock);
+ if (last_ptr->block_group)
+ hint_byte = last_ptr->window_start;
+ spin_unlock(&last_ptr->lock);
}
+
search_start = max(search_start, first_logical_byte(root, 0));
search_start = max(search_start, hint_byte);
- if (last_wanted && search_start != last_wanted) {
- last_wanted = 0;
- empty_size += empty_cluster;
+ if (!last_ptr) {
+ empty_cluster = 0;
+ loop = 1;
}
- total_needed += empty_size;
- block_group = btrfs_lookup_block_group(root->fs_info, search_start);
- if (!block_group)
- block_group = btrfs_lookup_first_block_group(root->fs_info,
- search_start);
- space_info = __find_space_info(root->fs_info, data);
+ if (search_start == hint_byte) {
+ block_group = btrfs_lookup_block_group(root->fs_info,
+ search_start);
+ if (block_group && block_group_bits(block_group, data)) {
+ down_read(&space_info->groups_sem);
+ goto have_block_group;
+ } else if (block_group) {
+ btrfs_put_block_group(block_group);
+ }
+ }
+search:
down_read(&space_info->groups_sem);
- while (1) {
- struct btrfs_free_space *free_space;
- /*
- * the only way this happens if our hint points to a block
- * group thats not of the proper type, while looping this
- * should never happen
- */
- if (empty_size)
- extra_loop = 1;
+ list_for_each_entry(block_group, &space_info->block_groups, list) {
+ u64 offset;
- if (!block_group)
- goto new_group_no_lock;
+ atomic_inc(&block_group->count);
+ search_start = block_group->key.objectid;
+have_block_group:
if (unlikely(!block_group->cached)) {
mutex_lock(&block_group->cache_mutex);
ret = cache_block_group(root, block_group);
mutex_unlock(&block_group->cache_mutex);
- if (ret)
+ if (ret) {
+ btrfs_put_block_group(block_group);
break;
+ }
}
- mutex_lock(&block_group->alloc_mutex);
- if (unlikely(!block_group_bits(block_group, data)))
- goto new_group;
-
if (unlikely(block_group->ro))
- goto new_group;
+ goto loop;
- free_space = btrfs_find_free_space(block_group, search_start,
- total_needed);
- if (free_space) {
- u64 start = block_group->key.objectid;
- u64 end = block_group->key.objectid +
- block_group->key.offset;
+ if (last_ptr) {
+ /*
+ * the refill lock keeps out other
+ * people trying to start a new cluster
+ */
+ spin_lock(&last_ptr->refill_lock);
+ offset = btrfs_alloc_from_cluster(block_group, last_ptr,
+ num_bytes, search_start);
+ if (offset) {
+ /* we have a block, we're done */
+ spin_unlock(&last_ptr->refill_lock);
+ goto checks;
+ }
- search_start = stripe_align(root, free_space->offset);
+ spin_lock(&last_ptr->lock);
+ /*
+ * whoops, this cluster doesn't actually point to
+ * this block group. Get a ref on the block
+ * group is does point to and try again
+ */
+ if (!last_ptr_loop && last_ptr->block_group &&
+ last_ptr->block_group != block_group) {
+
+ btrfs_put_block_group(block_group);
+ block_group = last_ptr->block_group;
+ atomic_inc(&block_group->count);
+ spin_unlock(&last_ptr->lock);
+ spin_unlock(&last_ptr->refill_lock);
+
+ last_ptr_loop = 1;
+ search_start = block_group->key.objectid;
+ goto have_block_group;
+ }
+ spin_unlock(&last_ptr->lock);
- /* move on to the next group */
- if (search_start + num_bytes >= search_end)
- goto new_group;
+ /*
+ * this cluster didn't work out, free it and
+ * start over
+ */
+ btrfs_return_cluster_to_free_space(NULL, last_ptr);
- /* move on to the next group */
- if (search_start + num_bytes > end)
- goto new_group;
+ last_ptr_loop = 0;
- if (last_wanted && search_start != last_wanted) {
- total_needed += empty_cluster;
- empty_size += empty_cluster;
- last_wanted = 0;
+ /* allocate a cluster in this block group */
+ ret = btrfs_find_space_cluster(trans,
+ block_group, last_ptr,
+ offset, num_bytes,
+ empty_cluster + empty_size);
+ if (ret == 0) {
/*
- * if search_start is still in this block group
- * then we just re-search this block group
+ * now pull our allocation out of this
+ * cluster
*/
- if (search_start >= start &&
- search_start < end) {
- mutex_unlock(&block_group->alloc_mutex);
- continue;
+ offset = btrfs_alloc_from_cluster(block_group,
+ last_ptr, num_bytes,
+ search_start);
+ if (offset) {
+ /* we found one, proceed */
+ spin_unlock(&last_ptr->refill_lock);
+ goto checks;
}
-
- /* else we go to the next block group */
- goto new_group;
}
-
- if (exclude_nr > 0 &&
- (search_start + num_bytes > exclude_start &&
- search_start < exclude_start + exclude_nr)) {
- search_start = exclude_start + exclude_nr;
- /*
- * if search_start is still in this block group
- * then we just re-search this block group
- */
- if (search_start >= start &&
- search_start < end) {
- mutex_unlock(&block_group->alloc_mutex);
- last_wanted = 0;
- continue;
- }
-
- /* else we go to the next block group */
- goto new_group;
+ /*
+ * at this point we either didn't find a cluster
+ * or we weren't able to allocate a block from our
+ * cluster. Free the cluster we've been trying
+ * to use, and go to the next block group
+ */
+ if (loop < 2) {
+ btrfs_return_cluster_to_free_space(NULL,
+ last_ptr);
+ spin_unlock(&last_ptr->refill_lock);
+ goto loop;
}
+ spin_unlock(&last_ptr->refill_lock);
+ }
- ins->objectid = search_start;
- ins->offset = num_bytes;
+ offset = btrfs_find_space_for_alloc(block_group, search_start,
+ num_bytes, empty_size);
+ if (!offset)
+ goto loop;
+checks:
+ search_start = stripe_align(root, offset);
+
+ /* move on to the next group */
+ if (search_start + num_bytes >= search_end) {
+ btrfs_add_free_space(block_group, offset, num_bytes);
+ goto loop;
+ }
- btrfs_remove_free_space_lock(block_group, search_start,
- num_bytes);
- /* we are all good, lets return */
- mutex_unlock(&block_group->alloc_mutex);
- break;
+ /* move on to the next group */
+ if (search_start + num_bytes >
+ block_group->key.objectid + block_group->key.offset) {
+ btrfs_add_free_space(block_group, offset, num_bytes);
+ goto loop;
}
-new_group:
- mutex_unlock(&block_group->alloc_mutex);
- put_block_group(block_group);
- block_group = NULL;
-new_group_no_lock:
- /* don't try to compare new allocations against the
- * last allocation any more
- */
- last_wanted = 0;
- /*
- * Here's how this works.
- * loop == 0: we were searching a block group via a hint
- * and didn't find anything, so we start at
- * the head of the block groups and keep searching
- * loop == 1: we're searching through all of the block groups
- * if we hit the head again we have searched
- * all of the block groups for this space and we
- * need to try and allocate, if we cant error out.
- * loop == 2: we allocated more space and are looping through
- * all of the block groups again.
- */
- if (loop == 0) {
- head = &space_info->block_groups;
- cur = head->next;
- loop++;
- } else if (loop == 1 && cur == head) {
- int keep_going;
-
- /* at this point we give up on the empty_size
- * allocations and just try to allocate the min
- * space.
- *
- * The extra_loop field was set if an empty_size
- * allocation was attempted above, and if this
- * is try we need to try the loop again without
- * the additional empty_size.
+ if (exclude_nr > 0 &&
+ (search_start + num_bytes > exclude_start &&
+ search_start < exclude_start + exclude_nr)) {
+ search_start = exclude_start + exclude_nr;
+
+ btrfs_add_free_space(block_group, offset, num_bytes);
+ /*
+ * if search_start is still in this block group
+ * then we just re-search this block group
*/
- total_needed -= empty_size;
- empty_size = 0;
- keep_going = extra_loop;
- loop++;
+ if (search_start >= block_group->key.objectid &&
+ search_start < (block_group->key.objectid +
+ block_group->key.offset))
+ goto have_block_group;
+ goto loop;
+ }
- if (allowed_chunk_alloc && !chunk_alloc_done) {
- up_read(&space_info->groups_sem);
- ret = do_chunk_alloc(trans, root, num_bytes +
- 2 * 1024 * 1024, data, 1);
- down_read(&space_info->groups_sem);
- if (ret < 0)
- goto loop_check;
- head = &space_info->block_groups;
- /*
- * we've allocated a new chunk, keep
- * trying
- */
- keep_going = 1;
- chunk_alloc_done = 1;
- } else if (!allowed_chunk_alloc) {
- space_info->force_alloc = 1;
- }
-loop_check:
- if (keep_going) {
- cur = head->next;
- extra_loop = 0;
- } else {
- break;
- }
- } else if (cur == head) {
- break;
+ ins->objectid = search_start;
+ ins->offset = num_bytes;
+
+ if (offset < search_start)
+ btrfs_add_free_space(block_group, offset,
+ search_start - offset);
+ BUG_ON(offset > search_start);
+
+ /* we are all good, lets return */
+ break;
+loop:
+ btrfs_put_block_group(block_group);
+ }
+ up_read(&space_info->groups_sem);
+
+ /* loop == 0, try to find a clustered alloc in every block group
+ * loop == 1, try again after forcing a chunk allocation
+ * loop == 2, set empty_size and empty_cluster to 0 and try again
+ */
+ if (!ins->objectid && loop < 3 &&
+ (empty_size || empty_cluster || allowed_chunk_alloc)) {
+ if (loop >= 2) {
+ empty_size = 0;
+ empty_cluster = 0;
}
- block_group = list_entry(cur, struct btrfs_block_group_cache,
- list);
- atomic_inc(&block_group->count);
+ if (allowed_chunk_alloc) {
+ ret = do_chunk_alloc(trans, root, num_bytes +
+ 2 * 1024 * 1024, data, 1);
+ allowed_chunk_alloc = 0;
+ } else {
+ space_info->force_alloc = 1;
+ }
- search_start = block_group->key.objectid;
- cur = cur->next;
+ if (loop < 3) {
+ loop++;
+ goto search;
+ }
+ ret = -ENOSPC;
+ } else if (!ins->objectid) {
+ ret = -ENOSPC;
}
/* we found what we needed */
@@ -2783,21 +2812,10 @@ loop_check:
if (!(data & BTRFS_BLOCK_GROUP_DATA))
trans->block_group = block_group->key.objectid;
- if (last_ptr)
- *last_ptr = ins->objectid + ins->offset;
+ btrfs_put_block_group(block_group);
ret = 0;
- } else if (!ret) {
- printk(KERN_ERR "btrfs searching for %llu bytes, "
- "num_bytes %llu, loop %d, allowed_alloc %d\n",
- (unsigned long long)total_needed,
- (unsigned long long)num_bytes,
- loop, allowed_chunk_alloc);
- ret = -ENOSPC;
}
- if (block_group)
- put_block_group(block_group);
- up_read(&space_info->groups_sem);
return ret;
}
@@ -2810,9 +2828,12 @@ static void dump_space_info(struct btrfs_space_info *info, u64 bytes)
info->bytes_pinned - info->bytes_reserved),
(info->full) ? "" : "not ");
printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
- " may_use=%llu, used=%llu\n", info->total_bytes,
- info->bytes_pinned, info->bytes_delalloc, info->bytes_may_use,
- info->bytes_used);
+ " may_use=%llu, used=%llu\n",
+ (unsigned long long)info->total_bytes,
+ (unsigned long long)info->bytes_pinned,
+ (unsigned long long)info->bytes_delalloc,
+ (unsigned long long)info->bytes_may_use,
+ (unsigned long long)info->bytes_used);
down_read(&info->groups_sem);
list_for_each_entry(cache, &info->block_groups, list) {
@@ -2902,7 +2923,7 @@ int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
ret = btrfs_discard_extent(root, start, len);
btrfs_add_free_space(cache, start, len);
- put_block_group(cache);
+ btrfs_put_block_group(cache);
update_reserved_extents(root, start, len, 0);
return ret;
@@ -3040,7 +3061,7 @@ int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
ret = btrfs_remove_free_space(block_group, ins->objectid,
ins->offset);
BUG_ON(ret);
- put_block_group(block_group);
+ btrfs_put_block_group(block_group);
ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid,
ref_generation, owner, ins, 1);
return ret;
@@ -5729,7 +5750,7 @@ next:
WARN_ON(block_group->reserved > 0);
WARN_ON(btrfs_block_group_used(&block_group->item) > 0);
spin_unlock(&block_group->lock);
- put_block_group(block_group);
+ btrfs_put_block_group(block_group);
ret = 0;
out:
btrfs_free_path(path);
@@ -5856,9 +5877,10 @@ int btrfs_read_block_groups(struct btrfs_root *root)
atomic_set(&cache->count, 1);
spin_lock_init(&cache->lock);
- mutex_init(&cache->alloc_mutex);
+ spin_lock_init(&cache->tree_lock);
mutex_init(&cache->cache_mutex);
INIT_LIST_HEAD(&cache->list);
+ INIT_LIST_HEAD(&cache->cluster_list);
read_extent_buffer(leaf, &cache->item,
btrfs_item_ptr_offset(leaf, path->slots[0]),
sizeof(cache->item));
@@ -5912,9 +5934,10 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
atomic_set(&cache->count, 1);
spin_lock_init(&cache->lock);
- mutex_init(&cache->alloc_mutex);
+ spin_lock_init(&cache->tree_lock);
mutex_init(&cache->cache_mutex);
INIT_LIST_HEAD(&cache->list);
+ INIT_LIST_HEAD(&cache->cluster_list);
btrfs_set_block_group_used(&cache->item, bytes_used);
btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
@@ -5974,8 +5997,8 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
spin_unlock(&block_group->space_info->lock);
block_group->space_info->full = 0;
- put_block_group(block_group);
- put_block_group(block_group);
+ btrfs_put_block_group(block_group);
+ btrfs_put_block_group(block_group);
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret > 0)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 08085af089e2..fe9eb990e443 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -17,12 +17,6 @@
#include "ctree.h"
#include "btrfs_inode.h"
-/* temporary define until extent_map moves out of btrfs */
-struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
- unsigned long extra_flags,
- void (*ctor)(void *, struct kmem_cache *,
- unsigned long));
-
static struct kmem_cache *extent_state_cache;
static struct kmem_cache *extent_buffer_cache;
@@ -50,20 +44,23 @@ struct extent_page_data {
/* tells writepage not to lock the state bits for this range
* it still does the unlocking
*/
- int extent_locked;
+ unsigned int extent_locked:1;
+
+ /* tells the submit_bio code to use a WRITE_SYNC */
+ unsigned int sync_io:1;
};
int __init extent_io_init(void)
{
- extent_state_cache = btrfs_cache_create("extent_state",
- sizeof(struct extent_state), 0,
- NULL);
+ extent_state_cache = kmem_cache_create("extent_state",
+ sizeof(struct extent_state), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
if (!extent_state_cache)
return -ENOMEM;
- extent_buffer_cache = btrfs_cache_create("extent_buffers",
- sizeof(struct extent_buffer), 0,
- NULL);
+ extent_buffer_cache = kmem_cache_create("extent_buffers",
+ sizeof(struct extent_buffer), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
if (!extent_buffer_cache)
goto free_state_cache;
return 0;
@@ -1404,69 +1401,6 @@ out:
return total_bytes;
}
-#if 0
-/*
- * helper function to lock both pages and extents in the tree.
- * pages must be locked first.
- */
-static int lock_range(struct extent_io_tree *tree, u64 start, u64 end)
-{
- unsigned long index = start >> PAGE_CACHE_SHIFT;
- unsigned long end_index = end >> PAGE_CACHE_SHIFT;
- struct page *page;
- int err;
-
- while (index <= end_index) {
- page = grab_cache_page(tree->mapping, index);
- if (!page) {
- err = -ENOMEM;
- goto failed;
- }
- if (IS_ERR(page)) {
- err = PTR_ERR(page);
- goto failed;
- }
- index++;
- }
- lock_extent(tree, start, end, GFP_NOFS);
- return 0;
-
-failed:
- /*
- * we failed above in getting the page at 'index', so we undo here
- * up to but not including the page at 'index'
- */
- end_index = index;
- index = start >> PAGE_CACHE_SHIFT;
- while (index < end_index) {
- page = find_get_page(tree->mapping, index);
- unlock_page(page);
- page_cache_release(page);
- index++;
- }
- return err;
-}
-
-/*
- * helper function to unlock both pages and extents in the tree.
- */
-static int unlock_range(struct extent_io_tree *tree, u64 start, u64 end)
-{
- unsigned long index = start >> PAGE_CACHE_SHIFT;
- unsigned long end_index = end >> PAGE_CACHE_SHIFT;
- struct page *page;
-
- while (index <= end_index) {
- page = find_get_page(tree->mapping, index);
- unlock_page(page);
- page_cache_release(page);
- index++;
- }
- unlock_extent(tree, start, end, GFP_NOFS);
- return 0;
-}
-#endif
-
/*
* set the private field for a given byte offset in the tree. If there isn't
* an extent_state there already, this does nothing.
@@ -2101,6 +2035,16 @@ int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
return ret;
}
+static noinline void update_nr_written(struct page *page,
+ struct writeback_control *wbc,
+ unsigned long nr_written)
+{
+ wbc->nr_to_write -= nr_written;
+ if (wbc->range_cyclic || (wbc->nr_to_write > 0 &&
+ wbc->range_start == 0 && wbc->range_end == LLONG_MAX))
+ page->mapping->writeback_index = page->index + nr_written;
+}
+
/*
* the writepage semantics are similar to regular writepage. extent
* records are inserted to lock ranges in the tree, and as dirty areas
@@ -2136,8 +2080,14 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
u64 delalloc_end;
int page_started;
int compressed;
+ int write_flags;
unsigned long nr_written = 0;
+ if (wbc->sync_mode == WB_SYNC_ALL)
+ write_flags = WRITE_SYNC_PLUG;
+ else
+ write_flags = WRITE;
+
WARN_ON(!PageLocked(page));
pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
if (page->index > end_index ||
@@ -2164,6 +2114,12 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
delalloc_end = 0;
page_started = 0;
if (!epd->extent_locked) {
+ /*
+ * make sure the wbc mapping index is at least updated
+ * to this page.
+ */
+ update_nr_written(page, wbc, 0);
+
while (delalloc_end < page_end) {
nr_delalloc = find_lock_delalloc_range(inode, tree,
page,
@@ -2185,7 +2141,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
*/
if (page_started) {
ret = 0;
- goto update_nr_written;
+ /*
+ * we've unlocked the page, so we can't update
+ * the mapping's writeback index, just update
+ * nr_to_write.
+ */
+ wbc->nr_to_write -= nr_written;
+ goto done_unlocked;
}
}
lock_extent(tree, start, page_end, GFP_NOFS);
@@ -2198,13 +2160,18 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
if (ret == -EAGAIN) {
unlock_extent(tree, start, page_end, GFP_NOFS);
redirty_page_for_writepage(wbc, page);
+ update_nr_written(page, wbc, nr_written);
unlock_page(page);
ret = 0;
- goto update_nr_written;
+ goto done_unlocked;
}
}
- nr_written++;
+ /*
+ * we don't want to touch the inode after unlocking the page,
+ * so we update the mapping writeback index now
+ */
+ update_nr_written(page, wbc, nr_written + 1);
end = page_end;
if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0))
@@ -2314,9 +2281,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
(unsigned long long)end);
}
- ret = submit_extent_page(WRITE, tree, page, sector,
- iosize, pg_offset, bdev,
- &epd->bio, max_nr,
+ ret = submit_extent_page(write_flags, tree, page,
+ sector, iosize, pg_offset,
+ bdev, &epd->bio, max_nr,
end_bio_extent_writepage,
0, 0, 0);
if (ret)
@@ -2336,11 +2303,8 @@ done:
unlock_extent(tree, unlock_start, page_end, GFP_NOFS);
unlock_page(page);
-update_nr_written:
- wbc->nr_to_write -= nr_written;
- if (wbc->range_cyclic || (wbc->nr_to_write > 0 &&
- wbc->range_start == 0 && wbc->range_end == LLONG_MAX))
- page->mapping->writeback_index = page->index + nr_written;
+done_unlocked:
+
return 0;
}
@@ -2460,15 +2424,23 @@ retry:
return ret;
}
-static noinline void flush_write_bio(void *data)
+static void flush_epd_write_bio(struct extent_page_data *epd)
{
- struct extent_page_data *epd = data;
if (epd->bio) {
- submit_one_bio(WRITE, epd->bio, 0, 0);
+ if (epd->sync_io)
+ submit_one_bio(WRITE_SYNC, epd->bio, 0, 0);
+ else
+ submit_one_bio(WRITE, epd->bio, 0, 0);
epd->bio = NULL;
}
}
+static noinline void flush_write_bio(void *data)
+{
+ struct extent_page_data *epd = data;
+ flush_epd_write_bio(epd);
+}
+
int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
get_extent_t *get_extent,
struct writeback_control *wbc)
@@ -2480,23 +2452,22 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
.tree = tree,
.get_extent = get_extent,
.extent_locked = 0,
+ .sync_io = wbc->sync_mode == WB_SYNC_ALL,
};
struct writeback_control wbc_writepages = {
.bdi = wbc->bdi,
- .sync_mode = WB_SYNC_NONE,
+ .sync_mode = wbc->sync_mode,
.older_than_this = NULL,
.nr_to_write = 64,
.range_start = page_offset(page) + PAGE_CACHE_SIZE,
.range_end = (loff_t)-1,
};
-
ret = __extent_writepage(page, wbc, &epd);
extent_write_cache_pages(tree, mapping, &wbc_writepages,
__extent_writepage, &epd, flush_write_bio);
- if (epd.bio)
- submit_one_bio(WRITE, epd.bio, 0, 0);
+ flush_epd_write_bio(&epd);
return ret;
}
@@ -2515,6 +2486,7 @@ int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
.tree = tree,
.get_extent = get_extent,
.extent_locked = 1,
+ .sync_io = mode == WB_SYNC_ALL,
};
struct writeback_control wbc_writepages = {
.bdi = inode->i_mapping->backing_dev_info,
@@ -2540,8 +2512,7 @@ int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
start += PAGE_CACHE_SIZE;
}
- if (epd.bio)
- submit_one_bio(WRITE, epd.bio, 0, 0);
+ flush_epd_write_bio(&epd);
return ret;
}
@@ -2556,13 +2527,13 @@ int extent_writepages(struct extent_io_tree *tree,
.tree = tree,
.get_extent = get_extent,
.extent_locked = 0,
+ .sync_io = wbc->sync_mode == WB_SYNC_ALL,
};
ret = extent_write_cache_pages(tree, mapping, wbc,
__extent_writepage, &epd,
flush_write_bio);
- if (epd.bio)
- submit_one_bio(WRITE, epd.bio, 0, 0);
+ flush_epd_write_bio(&epd);
return ret;
}
@@ -2884,25 +2855,19 @@ int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
disko = 0;
flags = 0;
- switch (em->block_start) {
- case EXTENT_MAP_LAST_BYTE:
+ if (em->block_start == EXTENT_MAP_LAST_BYTE) {
end = 1;
flags |= FIEMAP_EXTENT_LAST;
- break;
- case EXTENT_MAP_HOLE:
+ } else if (em->block_start == EXTENT_MAP_HOLE) {
flags |= FIEMAP_EXTENT_UNWRITTEN;
- break;
- case EXTENT_MAP_INLINE:
+ } else if (em->block_start == EXTENT_MAP_INLINE) {
flags |= (FIEMAP_EXTENT_DATA_INLINE |
FIEMAP_EXTENT_NOT_ALIGNED);
- break;
- case EXTENT_MAP_DELALLOC:
+ } else if (em->block_start == EXTENT_MAP_DELALLOC) {
flags |= (FIEMAP_EXTENT_DELALLOC |
FIEMAP_EXTENT_UNKNOWN);
- break;
- default:
+ } else {
disko = em->block_start;
- break;
}
if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
flags |= FIEMAP_EXTENT_ENCODED;
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 50da69da20ce..30c9365861e6 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -6,19 +6,14 @@
#include <linux/hardirq.h>
#include "extent_map.h"
-/* temporary define until extent_map moves out of btrfs */
-struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
- unsigned long extra_flags,
- void (*ctor)(void *, struct kmem_cache *,
- unsigned long));
static struct kmem_cache *extent_map_cache;
int __init extent_map_init(void)
{
- extent_map_cache = btrfs_cache_create("extent_map",
- sizeof(struct extent_map), 0,
- NULL);
+ extent_map_cache = kmem_cache_create("extent_map",
+ sizeof(struct extent_map), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
if (!extent_map_cache)
return -ENOMEM;
return 0;
@@ -43,7 +38,6 @@ void extent_map_tree_init(struct extent_map_tree *tree, gfp_t mask)
tree->map.rb_node = NULL;
spin_lock_init(&tree->lock);
}
-EXPORT_SYMBOL(extent_map_tree_init);
/**
* alloc_extent_map - allocate new extent map structure
@@ -64,7 +58,6 @@ struct extent_map *alloc_extent_map(gfp_t mask)
atomic_set(&em->refs, 1);
return em;
}
-EXPORT_SYMBOL(alloc_extent_map);
/**
* free_extent_map - drop reference count of an extent_map
@@ -83,7 +76,6 @@ void free_extent_map(struct extent_map *em)
kmem_cache_free(extent_map_cache, em);
}
}
-EXPORT_SYMBOL(free_extent_map);
static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
struct rb_node *node)
@@ -234,7 +226,6 @@ int add_extent_mapping(struct extent_map_tree *tree,
rb = tree_insert(&tree->map, em->start, &em->rb_node);
if (rb) {
ret = -EEXIST;
- free_extent_map(merge);
goto out;
}
atomic_inc(&em->refs);
@@ -265,7 +256,6 @@ int add_extent_mapping(struct extent_map_tree *tree,
out:
return ret;
}
-EXPORT_SYMBOL(add_extent_mapping);
/* simple helper to do math around the end of an extent, handling wrap */
static u64 range_end(u64 start, u64 len)
@@ -327,7 +317,6 @@ found:
out:
return em;
}
-EXPORT_SYMBOL(lookup_extent_mapping);
/**
* remove_extent_mapping - removes an extent_map from the extent tree
@@ -347,4 +336,3 @@ int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em)
em->in_tree = 0;
return ret;
}
-EXPORT_SYMBOL(remove_extent_mapping);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 9c9fb46ccd08..1d51dc38bb49 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -272,83 +272,6 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
return 0;
}
-int btrfs_check_file(struct btrfs_root *root, struct inode *inode)
-{
- return 0;
-#if 0
- struct btrfs_path *path;
- struct btrfs_key found_key;
- struct extent_buffer *leaf;
- struct btrfs_file_extent_item *extent;
- u64 last_offset = 0;
- int nritems;
- int slot;
- int found_type;
- int ret;
- int err = 0;
- u64 extent_end = 0;
-
- path = btrfs_alloc_path();
- ret = btrfs_lookup_file_extent(NULL, root, path, inode->i_ino,
- last_offset, 0);
- while (1) {
- nritems = btrfs_header_nritems(path->nodes[0]);
- if (path->slots[0] >= nritems) {
- ret = btrfs_next_leaf(root, path);
- if (ret)
- goto out;
- nritems = btrfs_header_nritems(path->nodes[0]);
- }
- slot = path->slots[0];
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &found_key, slot);
- if (found_key.objectid != inode->i_ino)
- break;
- if (found_key.type != BTRFS_EXTENT_DATA_KEY)
- goto out;
-
- if (found_key.offset < last_offset) {
- WARN_ON(1);
- btrfs_print_leaf(root, leaf);
- printk(KERN_ERR "inode %lu found offset %llu "
- "expected %llu\n", inode->i_ino,
- (unsigned long long)found_key.offset,
- (unsigned long long)last_offset);
- err = 1;
- goto out;
- }
- extent = btrfs_item_ptr(leaf, slot,
- struct btrfs_file_extent_item);
- found_type = btrfs_file_extent_type(leaf, extent);
- if (found_type == BTRFS_FILE_EXTENT_REG) {
- extent_end = found_key.offset +
- btrfs_file_extent_num_bytes(leaf, extent);
- } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
- struct btrfs_item *item;
- item = btrfs_item_nr(leaf, slot);
- extent_end = found_key.offset +
- btrfs_file_extent_inline_len(leaf, extent);
- extent_end = (extent_end + root->sectorsize - 1) &
- ~((u64)root->sectorsize - 1);
- }
- last_offset = extent_end;
- path->slots[0]++;
- }
- if (0 && last_offset < inode->i_size) {
- WARN_ON(1);
- btrfs_print_leaf(root, leaf);
- printk(KERN_ERR "inode %lu found offset %llu size %llu\n",
- inode->i_ino, (unsigned long long)last_offset,
- (unsigned long long)inode->i_size);
- err = 1;
-
- }
-out:
- btrfs_free_path(path);
- return err;
-#endif
-}
-
/*
* this is very complex, but the basic idea is to drop all extents
* in the range start - end. hint_block is filled in with a block number
@@ -363,15 +286,16 @@ out:
*/
noinline int btrfs_drop_extents(struct btrfs_trans_handle *trans,
struct btrfs_root *root, struct inode *inode,
- u64 start, u64 end, u64 inline_limit, u64 *hint_byte)
+ u64 start, u64 end, u64 locked_end,
+ u64 inline_limit, u64 *hint_byte)
{
u64 extent_end = 0;
- u64 locked_end = end;
u64 search_start = start;
u64 leaf_start;
u64 ram_bytes = 0;
u64 orig_parent = 0;
u64 disk_bytenr = 0;
+ u64 orig_locked_end = locked_end;
u8 compression;
u8 encryption;
u16 other_encoding = 0;
@@ -684,11 +608,10 @@ next_slot:
}
out:
btrfs_free_path(path);
- if (locked_end > end) {
- unlock_extent(&BTRFS_I(inode)->io_tree, end, locked_end - 1,
- GFP_NOFS);
+ if (locked_end > orig_locked_end) {
+ unlock_extent(&BTRFS_I(inode)->io_tree, orig_locked_end,
+ locked_end - 1, GFP_NOFS);
}
- btrfs_check_file(root, inode);
return ret;
}
@@ -830,7 +753,7 @@ again:
ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
BUG_ON(ret);
- goto done;
+ goto release;
} else if (split == start) {
if (locked_end < extent_end) {
ret = try_lock_extent(&BTRFS_I(inode)->io_tree,
@@ -926,6 +849,8 @@ again:
}
done:
btrfs_mark_buffer_dirty(leaf);
+
+release:
btrfs_release_path(root, path);
if (split_end && split == start) {
split = end;
@@ -1131,7 +1056,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
if (will_write) {
btrfs_fdatawrite_range(inode->i_mapping, pos,
pos + write_bytes - 1,
- WB_SYNC_NONE);
+ WB_SYNC_ALL);
} else {
balance_dirty_pages_ratelimited_nr(inode->i_mapping,
num_pages);
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index d1e5f0e84c58..0bc93657b460 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -18,6 +18,15 @@
#include <linux/sched.h>
#include "ctree.h"
+#include "free-space-cache.h"
+#include "transaction.h"
+
+struct btrfs_free_space {
+ struct rb_node bytes_index;
+ struct rb_node offset_index;
+ u64 offset;
+ u64 bytes;
+};
static int tree_insert_offset(struct rb_root *root, u64 offset,
struct rb_node *node)
@@ -68,14 +77,24 @@ static int tree_insert_bytes(struct rb_root *root, u64 bytes,
}
/*
- * searches the tree for the given offset. If contains is set we will return
- * the free space that contains the given offset. If contains is not set we
- * will return the free space that starts at or after the given offset and is
- * at least bytes long.
+ * searches the tree for the given offset.
+ *
+ * fuzzy == 1: this is used for allocations where we are given a hint of where
+ * to look for free space. Because the hint may not be completely on an offset
+ * mark, or the hint may no longer point to free space we need to fudge our
+ * results a bit. So we look for free space starting at or after offset with at
+ * least bytes size. We prefer to find as close to the given offset as we can.
+ * Also if the offset is within a free space range, then we will return the free
+ * space that contains the given offset, which means we can return a free space
+ * chunk with an offset before the provided offset.
+ *
+ * fuzzy == 0: this is just a normal tree search. Give us the free space that
+ * starts at the given offset which is at least bytes size, and if its not there
+ * return NULL.
*/
static struct btrfs_free_space *tree_search_offset(struct rb_root *root,
u64 offset, u64 bytes,
- int contains)
+ int fuzzy)
{
struct rb_node *n = root->rb_node;
struct btrfs_free_space *entry, *ret = NULL;
@@ -84,13 +103,14 @@ static struct btrfs_free_space *tree_search_offset(struct rb_root *root,
entry = rb_entry(n, struct btrfs_free_space, offset_index);
if (offset < entry->offset) {
- if (!contains &&
+ if (fuzzy &&
(!ret || entry->offset < ret->offset) &&
(bytes <= entry->bytes))
ret = entry;
n = n->rb_left;
} else if (offset > entry->offset) {
- if ((entry->offset + entry->bytes - 1) >= offset &&
+ if (fuzzy &&
+ (entry->offset + entry->bytes - 1) >= offset &&
bytes <= entry->bytes) {
ret = entry;
break;
@@ -171,6 +191,7 @@ static int link_free_space(struct btrfs_block_group_cache *block_group,
int ret = 0;
+ BUG_ON(!info->bytes);
ret = tree_insert_offset(&block_group->free_space_offset, info->offset,
&info->offset_index);
if (ret)
@@ -184,108 +205,70 @@ static int link_free_space(struct btrfs_block_group_cache *block_group,
return ret;
}
-static int __btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes)
+int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
+ u64 offset, u64 bytes)
{
struct btrfs_free_space *right_info;
struct btrfs_free_space *left_info;
struct btrfs_free_space *info = NULL;
- struct btrfs_free_space *alloc_info;
int ret = 0;
- alloc_info = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
- if (!alloc_info)
+ info = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
+ if (!info)
return -ENOMEM;
+ info->offset = offset;
+ info->bytes = bytes;
+
+ spin_lock(&block_group->tree_lock);
+
/*
* first we want to see if there is free space adjacent to the range we
* are adding, if there is remove that struct and add a new one to
* cover the entire range
*/
right_info = tree_search_offset(&block_group->free_space_offset,
- offset+bytes, 0, 1);
+ offset+bytes, 0, 0);
left_info = tree_search_offset(&block_group->free_space_offset,
offset-1, 0, 1);
- if (right_info && right_info->offset == offset+bytes) {
+ if (right_info) {
unlink_free_space(block_group, right_info);
- info = right_info;
- info->offset = offset;
- info->bytes += bytes;
- } else if (right_info && right_info->offset != offset+bytes) {
- printk(KERN_ERR "btrfs adding space in the middle of an "
- "existing free space area. existing: "
- "offset=%llu, bytes=%llu. new: offset=%llu, "
- "bytes=%llu\n", (unsigned long long)right_info->offset,
- (unsigned long long)right_info->bytes,
- (unsigned long long)offset,
- (unsigned long long)bytes);
- BUG();
+ info->bytes += right_info->bytes;
+ kfree(right_info);
}
- if (left_info) {
+ if (left_info && left_info->offset + left_info->bytes == offset) {
unlink_free_space(block_group, left_info);
-
- if (unlikely((left_info->offset + left_info->bytes) !=
- offset)) {
- printk(KERN_ERR "btrfs free space to the left "
- "of new free space isn't "
- "quite right. existing: offset=%llu, "
- "bytes=%llu. new: offset=%llu, bytes=%llu\n",
- (unsigned long long)left_info->offset,
- (unsigned long long)left_info->bytes,
- (unsigned long long)offset,
- (unsigned long long)bytes);
- BUG();
- }
-
- if (info) {
- info->offset = left_info->offset;
- info->bytes += left_info->bytes;
- kfree(left_info);
- } else {
- info = left_info;
- info->bytes += bytes;
- }
+ info->offset = left_info->offset;
+ info->bytes += left_info->bytes;
+ kfree(left_info);
}
- if (info) {
- ret = link_free_space(block_group, info);
- if (!ret)
- info = NULL;
- goto out;
- }
-
- info = alloc_info;
- alloc_info = NULL;
- info->offset = offset;
- info->bytes = bytes;
-
ret = link_free_space(block_group, info);
if (ret)
kfree(info);
-out:
+
+ spin_unlock(&block_group->tree_lock);
+
if (ret) {
printk(KERN_ERR "btrfs: unable to add free space :%d\n", ret);
- if (ret == -EEXIST)
- BUG();
+ BUG_ON(ret == -EEXIST);
}
- kfree(alloc_info);
-
return ret;
}
-static int
-__btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes)
+int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
+ u64 offset, u64 bytes)
{
struct btrfs_free_space *info;
int ret = 0;
+ spin_lock(&block_group->tree_lock);
+
info = tree_search_offset(&block_group->free_space_offset, offset, 0,
1);
-
if (info && info->offset == offset) {
if (info->bytes < bytes) {
printk(KERN_ERR "Found free space at %llu, size %llu,"
@@ -295,12 +278,14 @@ __btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
(unsigned long long)bytes);
WARN_ON(1);
ret = -EINVAL;
+ spin_unlock(&block_group->tree_lock);
goto out;
}
unlink_free_space(block_group, info);
if (info->bytes == bytes) {
kfree(info);
+ spin_unlock(&block_group->tree_lock);
goto out;
}
@@ -308,6 +293,7 @@ __btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
info->bytes -= bytes;
ret = link_free_space(block_group, info);
+ spin_unlock(&block_group->tree_lock);
BUG_ON(ret);
} else if (info && info->offset < offset &&
info->offset + info->bytes >= offset + bytes) {
@@ -333,70 +319,37 @@ __btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
*/
kfree(info);
}
-
+ spin_unlock(&block_group->tree_lock);
/* step two, insert a new info struct to cover anything
* before the hole
*/
- ret = __btrfs_add_free_space(block_group, old_start,
- offset - old_start);
+ ret = btrfs_add_free_space(block_group, old_start,
+ offset - old_start);
BUG_ON(ret);
} else {
+ spin_unlock(&block_group->tree_lock);
+ if (!info) {
+ printk(KERN_ERR "couldn't find space %llu to free\n",
+ (unsigned long long)offset);
+ printk(KERN_ERR "cached is %d, offset %llu bytes %llu\n",
+ block_group->cached,
+ (unsigned long long)block_group->key.objectid,
+ (unsigned long long)block_group->key.offset);
+ btrfs_dump_free_space(block_group, bytes);
+ } else if (info) {
+ printk(KERN_ERR "hmm, found offset=%llu bytes=%llu, "
+ "but wanted offset=%llu bytes=%llu\n",
+ (unsigned long long)info->offset,
+ (unsigned long long)info->bytes,
+ (unsigned long long)offset,
+ (unsigned long long)bytes);
+ }
WARN_ON(1);
}
out:
return ret;
}
-int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes)
-{
- int ret;
- struct btrfs_free_space *sp;
-
- mutex_lock(&block_group->alloc_mutex);
- ret = __btrfs_add_free_space(block_group, offset, bytes);
- sp = tree_search_offset(&block_group->free_space_offset, offset, 0, 1);
- BUG_ON(!sp);
- mutex_unlock(&block_group->alloc_mutex);
-
- return ret;
-}
-
-int btrfs_add_free_space_lock(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes)
-{
- int ret;
- struct btrfs_free_space *sp;
-
- ret = __btrfs_add_free_space(block_group, offset, bytes);
- sp = tree_search_offset(&block_group->free_space_offset, offset, 0, 1);
- BUG_ON(!sp);
-
- return ret;
-}
-
-int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes)
-{
- int ret = 0;
-
- mutex_lock(&block_group->alloc_mutex);
- ret = __btrfs_remove_free_space(block_group, offset, bytes);
- mutex_unlock(&block_group->alloc_mutex);
-
- return ret;
-}
-
-int btrfs_remove_free_space_lock(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes)
-{
- int ret;
-
- ret = __btrfs_remove_free_space(block_group, offset, bytes);
-
- return ret;
-}
-
void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
u64 bytes)
{
@@ -408,6 +361,9 @@ void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
info = rb_entry(n, struct btrfs_free_space, offset_index);
if (info->bytes >= bytes)
count++;
+ printk(KERN_ERR "entry offset %llu, bytes %llu\n",
+ (unsigned long long)info->offset,
+ (unsigned long long)info->bytes);
}
printk(KERN_INFO "%d blocks of free space at or bigger than bytes is"
"\n", count);
@@ -428,68 +384,337 @@ u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group)
return ret;
}
+/*
+ * for a given cluster, put all of its extents back into the free
+ * space cache. If the block group passed doesn't match the block group
+ * pointed to by the cluster, someone else raced in and freed the
+ * cluster already. In that case, we just return without changing anything
+ */
+static int
+__btrfs_return_cluster_to_free_space(
+ struct btrfs_block_group_cache *block_group,
+ struct btrfs_free_cluster *cluster)
+{
+ struct btrfs_free_space *entry;
+ struct rb_node *node;
+
+ spin_lock(&cluster->lock);
+ if (cluster->block_group != block_group)
+ goto out;
+
+ cluster->window_start = 0;
+ node = rb_first(&cluster->root);
+ while(node) {
+ entry = rb_entry(node, struct btrfs_free_space, offset_index);
+ node = rb_next(&entry->offset_index);
+ rb_erase(&entry->offset_index, &cluster->root);
+ link_free_space(block_group, entry);
+ }
+ list_del_init(&cluster->block_group_list);
+
+ btrfs_put_block_group(cluster->block_group);
+ cluster->block_group = NULL;
+ cluster->root.rb_node = NULL;
+out:
+ spin_unlock(&cluster->lock);
+ return 0;
+}
+
void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
{
struct btrfs_free_space *info;
struct rb_node *node;
+ struct btrfs_free_cluster *cluster;
+ struct btrfs_free_cluster *safe;
+
+ spin_lock(&block_group->tree_lock);
+
+ list_for_each_entry_safe(cluster, safe, &block_group->cluster_list,
+ block_group_list) {
+
+ WARN_ON(cluster->block_group != block_group);
+ __btrfs_return_cluster_to_free_space(block_group, cluster);
+ }
- mutex_lock(&block_group->alloc_mutex);
while ((node = rb_last(&block_group->free_space_bytes)) != NULL) {
info = rb_entry(node, struct btrfs_free_space, bytes_index);
unlink_free_space(block_group, info);
kfree(info);
if (need_resched()) {
- mutex_unlock(&block_group->alloc_mutex);
+ spin_unlock(&block_group->tree_lock);
cond_resched();
- mutex_lock(&block_group->alloc_mutex);
+ spin_lock(&block_group->tree_lock);
}
}
- mutex_unlock(&block_group->alloc_mutex);
+ spin_unlock(&block_group->tree_lock);
}
-#if 0
-static struct btrfs_free_space *btrfs_find_free_space_offset(struct
- btrfs_block_group_cache
- *block_group, u64 offset,
- u64 bytes)
+u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
+ u64 offset, u64 bytes, u64 empty_size)
{
- struct btrfs_free_space *ret;
+ struct btrfs_free_space *entry = NULL;
+ u64 ret = 0;
- mutex_lock(&block_group->alloc_mutex);
- ret = tree_search_offset(&block_group->free_space_offset, offset,
- bytes, 0);
- mutex_unlock(&block_group->alloc_mutex);
+ spin_lock(&block_group->tree_lock);
+ entry = tree_search_offset(&block_group->free_space_offset, offset,
+ bytes + empty_size, 1);
+ if (!entry)
+ entry = tree_search_bytes(&block_group->free_space_bytes,
+ offset, bytes + empty_size);
+ if (entry) {
+ unlink_free_space(block_group, entry);
+ ret = entry->offset;
+ entry->offset += bytes;
+ entry->bytes -= bytes;
+
+ if (!entry->bytes)
+ kfree(entry);
+ else
+ link_free_space(block_group, entry);
+ }
+ spin_unlock(&block_group->tree_lock);
return ret;
}
-static struct btrfs_free_space *btrfs_find_free_space_bytes(struct
- btrfs_block_group_cache
- *block_group, u64 offset,
- u64 bytes)
+/*
+ * given a cluster, put all of its extents back into the free space
+ * cache. If a block group is passed, this function will only free
+ * a cluster that belongs to the passed block group.
+ *
+ * Otherwise, it'll get a reference on the block group pointed to by the
+ * cluster and remove the cluster from it.
+ */
+int btrfs_return_cluster_to_free_space(
+ struct btrfs_block_group_cache *block_group,
+ struct btrfs_free_cluster *cluster)
{
- struct btrfs_free_space *ret;
+ int ret;
- mutex_lock(&block_group->alloc_mutex);
+ /* first, get a safe pointer to the block group */
+ spin_lock(&cluster->lock);
+ if (!block_group) {
+ block_group = cluster->block_group;
+ if (!block_group) {
+ spin_unlock(&cluster->lock);
+ return 0;
+ }
+ } else if (cluster->block_group != block_group) {
+ /* someone else has already freed it don't redo their work */
+ spin_unlock(&cluster->lock);
+ return 0;
+ }
+ atomic_inc(&block_group->count);
+ spin_unlock(&cluster->lock);
- ret = tree_search_bytes(&block_group->free_space_bytes, offset, bytes);
- mutex_unlock(&block_group->alloc_mutex);
+ /* now return any extents the cluster had on it */
+ spin_lock(&block_group->tree_lock);
+ ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
+ spin_unlock(&block_group->tree_lock);
+ /* finally drop our ref */
+ btrfs_put_block_group(block_group);
return ret;
}
-#endif
-struct btrfs_free_space *btrfs_find_free_space(struct btrfs_block_group_cache
- *block_group, u64 offset,
- u64 bytes)
+/*
+ * given a cluster, try to allocate 'bytes' from it, returns 0
+ * if it couldn't find anything suitably large, or a logical disk offset
+ * if things worked out
+ */
+u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
+ struct btrfs_free_cluster *cluster, u64 bytes,
+ u64 min_start)
{
- struct btrfs_free_space *ret = NULL;
+ struct btrfs_free_space *entry = NULL;
+ struct rb_node *node;
+ u64 ret = 0;
- ret = tree_search_offset(&block_group->free_space_offset, offset,
- bytes, 0);
- if (!ret)
- ret = tree_search_bytes(&block_group->free_space_bytes,
- offset, bytes);
+ spin_lock(&cluster->lock);
+ if (bytes > cluster->max_size)
+ goto out;
+ if (cluster->block_group != block_group)
+ goto out;
+
+ node = rb_first(&cluster->root);
+ if (!node)
+ goto out;
+
+ entry = rb_entry(node, struct btrfs_free_space, offset_index);
+
+ while(1) {
+ if (entry->bytes < bytes || entry->offset < min_start) {
+ struct rb_node *node;
+
+ node = rb_next(&entry->offset_index);
+ if (!node)
+ break;
+ entry = rb_entry(node, struct btrfs_free_space,
+ offset_index);
+ continue;
+ }
+ ret = entry->offset;
+
+ entry->offset += bytes;
+ entry->bytes -= bytes;
+
+ if (entry->bytes == 0) {
+ rb_erase(&entry->offset_index, &cluster->root);
+ kfree(entry);
+ }
+ break;
+ }
+out:
+ spin_unlock(&cluster->lock);
return ret;
}
+
+/*
+ * here we try to find a cluster of blocks in a block group. The goal
+ * is to find at least bytes free and up to empty_size + bytes free.
+ * We might not find them all in one contiguous area.
+ *
+ * returns zero and sets up cluster if things worked out, otherwise
+ * it returns -enospc
+ */
+int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
+ struct btrfs_block_group_cache *block_group,
+ struct btrfs_free_cluster *cluster,
+ u64 offset, u64 bytes, u64 empty_size)
+{
+ struct btrfs_free_space *entry = NULL;
+ struct rb_node *node;
+ struct btrfs_free_space *next;
+ struct btrfs_free_space *last;
+ u64 min_bytes;
+ u64 window_start;
+ u64 window_free;
+ u64 max_extent = 0;
+ int total_retries = 0;
+ int ret;
+
+ /* for metadata, allow allocates with more holes */
+ if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
+ /*
+ * we want to do larger allocations when we are
+ * flushing out the delayed refs, it helps prevent
+ * making more work as we go along.
+ */
+ if (trans->transaction->delayed_refs.flushing)
+ min_bytes = max(bytes, (bytes + empty_size) >> 1);
+ else
+ min_bytes = max(bytes, (bytes + empty_size) >> 4);
+ } else
+ min_bytes = max(bytes, (bytes + empty_size) >> 2);
+
+ spin_lock(&block_group->tree_lock);
+ spin_lock(&cluster->lock);
+
+ /* someone already found a cluster, hooray */
+ if (cluster->block_group) {
+ ret = 0;
+ goto out;
+ }
+again:
+ min_bytes = min(min_bytes, bytes + empty_size);
+ entry = tree_search_bytes(&block_group->free_space_bytes,
+ offset, min_bytes);
+ if (!entry) {
+ ret = -ENOSPC;
+ goto out;
+ }
+ window_start = entry->offset;
+ window_free = entry->bytes;
+ last = entry;
+ max_extent = entry->bytes;
+
+ while(1) {
+ /* out window is just right, lets fill it */
+ if (window_free >= bytes + empty_size)
+ break;
+
+ node = rb_next(&last->offset_index);
+ if (!node) {
+ ret = -ENOSPC;
+ goto out;
+ }
+ next = rb_entry(node, struct btrfs_free_space, offset_index);
+
+ /*
+ * we haven't filled the empty size and the window is
+ * very large. reset and try again
+ */
+ if (next->offset - window_start > (bytes + empty_size) * 2) {
+ entry = next;
+ window_start = entry->offset;
+ window_free = entry->bytes;
+ last = entry;
+ max_extent = 0;
+ total_retries++;
+ if (total_retries % 256 == 0) {
+ if (min_bytes >= (bytes + empty_size)) {
+ ret = -ENOSPC;
+ goto out;
+ }
+ /*
+ * grow our allocation a bit, we're not having
+ * much luck
+ */
+ min_bytes *= 2;
+ goto again;
+ }
+ } else {
+ last = next;
+ window_free += next->bytes;
+ if (entry->bytes > max_extent)
+ max_extent = entry->bytes;
+ }
+ }
+
+ cluster->window_start = entry->offset;
+
+ /*
+ * now we've found our entries, pull them out of the free space
+ * cache and put them into the cluster rbtree
+ *
+ * The cluster includes an rbtree, but only uses the offset index
+ * of each free space cache entry.
+ */
+ while(1) {
+ node = rb_next(&entry->offset_index);
+ unlink_free_space(block_group, entry);
+ ret = tree_insert_offset(&cluster->root, entry->offset,
+ &entry->offset_index);
+ BUG_ON(ret);
+
+ if (!node || entry == last)
+ break;
+
+ entry = rb_entry(node, struct btrfs_free_space, offset_index);
+ }
+ ret = 0;
+ cluster->max_size = max_extent;
+ atomic_inc(&block_group->count);
+ list_add_tail(&cluster->block_group_list, &block_group->cluster_list);
+ cluster->block_group = block_group;
+out:
+ spin_unlock(&cluster->lock);
+ spin_unlock(&block_group->tree_lock);
+
+ return ret;
+}
+
+/*
+ * simple code to zero out a cluster
+ */
+void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
+{
+ spin_lock_init(&cluster->lock);
+ spin_lock_init(&cluster->refill_lock);
+ cluster->root.rb_node = NULL;
+ cluster->max_size = 0;
+ INIT_LIST_HEAD(&cluster->block_group_list);
+ cluster->block_group = NULL;
+}
+
diff --git a/fs/btrfs/free-space-cache.h b/fs/btrfs/free-space-cache.h
new file mode 100644
index 000000000000..ab0bdc0a63ce
--- /dev/null
+++ b/fs/btrfs/free-space-cache.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (C) 2009 Oracle. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_FREE_SPACE_CACHE
+#define __BTRFS_FREE_SPACE_CACHE
+
+int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
+ u64 bytenr, u64 size);
+int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
+ u64 bytenr, u64 size);
+void btrfs_remove_free_space_cache(struct btrfs_block_group_cache
+ *block_group);
+u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
+ u64 offset, u64 bytes, u64 empty_size);
+void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
+ u64 bytes);
+u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group);
+int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
+ struct btrfs_block_group_cache *block_group,
+ struct btrfs_free_cluster *cluster,
+ u64 offset, u64 bytes, u64 empty_size);
+void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster);
+u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
+ struct btrfs_free_cluster *cluster, u64 bytes,
+ u64 min_start);
+int btrfs_return_cluster_to_free_space(
+ struct btrfs_block_group_cache *block_group,
+ struct btrfs_free_cluster *cluster);
+#endif
diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c
index cc7334d833c9..9abbced1123d 100644
--- a/fs/btrfs/inode-map.c
+++ b/fs/btrfs/inode-map.c
@@ -79,7 +79,7 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
}
path = btrfs_alloc_path();
BUG_ON(!path);
- search_start = max(search_start, BTRFS_FIRST_FREE_OBJECTID);
+ search_start = max(search_start, (u64)BTRFS_FIRST_FREE_OBJECTID);
search_key.objectid = search_start;
search_key.type = 0;
search_key.offset = 0;
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 06d8db5afb08..90c23eb28829 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -70,7 +70,6 @@ static struct extent_io_ops btrfs_extent_io_ops;
static struct kmem_cache *btrfs_inode_cachep;
struct kmem_cache *btrfs_trans_handle_cachep;
struct kmem_cache *btrfs_transaction_cachep;
-struct kmem_cache *btrfs_bit_radix_cachep;
struct kmem_cache *btrfs_path_cachep;
#define S_SHIFT 12
@@ -234,7 +233,7 @@ static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
}
ret = btrfs_drop_extents(trans, root, inode, start,
- aligned_end, start, &hint_byte);
+ aligned_end, aligned_end, start, &hint_byte);
BUG_ON(ret);
if (isize > actual_end)
@@ -1439,6 +1438,7 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
struct inode *inode, u64 file_pos,
u64 disk_bytenr, u64 disk_num_bytes,
u64 num_bytes, u64 ram_bytes,
+ u64 locked_end,
u8 compression, u8 encryption,
u16 other_encoding, int extent_type)
{
@@ -1455,7 +1455,8 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
path->leave_spinning = 1;
ret = btrfs_drop_extents(trans, root, inode, file_pos,
- file_pos + num_bytes, file_pos, &hint);
+ file_pos + num_bytes, locked_end,
+ file_pos, &hint);
BUG_ON(ret);
ins.objectid = inode->i_ino;
@@ -1590,6 +1591,8 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
ordered_extent->disk_len,
ordered_extent->len,
ordered_extent->len,
+ ordered_extent->file_offset +
+ ordered_extent->len,
compressed, 0, 0,
BTRFS_FILE_EXTENT_REG);
BUG_ON(ret);
@@ -1819,10 +1822,12 @@ good:
return 0;
zeroit:
- printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
- "private %llu\n", page->mapping->host->i_ino,
- (unsigned long long)start, csum,
- (unsigned long long)private);
+ if (printk_ratelimit()) {
+ printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
+ "private %llu\n", page->mapping->host->i_ino,
+ (unsigned long long)start, csum,
+ (unsigned long long)private);
+ }
memset(kaddr + offset, 1, end - start + 1);
flush_dcache_page(page);
kunmap_atomic(kaddr, KM_USER0);
@@ -2011,6 +2016,57 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
}
/*
+ * very simple check to peek ahead in the leaf looking for xattrs. If we
+ * don't find any xattrs, we know there can't be any acls.
+ *
+ * slot is the slot the inode is in, objectid is the objectid of the inode
+ */
+static noinline int acls_after_inode_item(struct extent_buffer *leaf,
+ int slot, u64 objectid)
+{
+ u32 nritems = btrfs_header_nritems(leaf);
+ struct btrfs_key found_key;
+ int scanned = 0;
+
+ slot++;
+ while (slot < nritems) {
+ btrfs_item_key_to_cpu(leaf, &found_key, slot);
+
+ /* we found a different objectid, there must not be acls */
+ if (found_key.objectid != objectid)
+ return 0;
+
+ /* we found an xattr, assume we've got an acl */
+ if (found_key.type == BTRFS_XATTR_ITEM_KEY)
+ return 1;
+
+ /*
+ * we found a key greater than an xattr key, there can't
+ * be any acls later on
+ */
+ if (found_key.type > BTRFS_XATTR_ITEM_KEY)
+ return 0;
+
+ slot++;
+ scanned++;
+
+ /*
+ * it goes inode, inode backrefs, xattrs, extents,
+ * so if there are a ton of hard links to an inode there can
+ * be a lot of backrefs. Don't waste time searching too hard,
+ * this is just an optimization
+ */
+ if (scanned >= 8)
+ break;
+ }
+ /* we hit the end of the leaf before we found an xattr or
+ * something larger than an xattr. We have to assume the inode
+ * has acls
+ */
+ return 1;
+}
+
+/*
* read an inode from the btree into the in-memory inode
*/
void btrfs_read_locked_inode(struct inode *inode)
@@ -2021,6 +2077,7 @@ void btrfs_read_locked_inode(struct inode *inode)
struct btrfs_timespec *tspec;
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_key location;
+ int maybe_acls;
u64 alloc_group_block;
u32 rdev;
int ret;
@@ -2067,6 +2124,16 @@ void btrfs_read_locked_inode(struct inode *inode)
alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
+ /*
+ * try to precache a NULL acl entry for files that don't have
+ * any xattrs or acls
+ */
+ maybe_acls = acls_after_inode_item(leaf, path->slots[0], inode->i_ino);
+ if (!maybe_acls) {
+ BTRFS_I(inode)->i_acl = NULL;
+ BTRFS_I(inode)->i_default_acl = NULL;
+ }
+
BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0,
alloc_group_block, 0);
btrfs_free_path(path);
@@ -2877,6 +2944,7 @@ int btrfs_cont_expand(struct inode *inode, loff_t size)
err = btrfs_drop_extents(trans, root, inode,
cur_offset,
cur_offset + hole_size,
+ block_end,
cur_offset, &hint_byte);
if (err)
break;
@@ -3041,8 +3109,8 @@ static noinline void init_btrfs_i(struct inode *inode)
{
struct btrfs_inode *bi = BTRFS_I(inode);
- bi->i_acl = NULL;
- bi->i_default_acl = NULL;
+ bi->i_acl = BTRFS_ACL_NOT_CACHED;
+ bi->i_default_acl = BTRFS_ACL_NOT_CACHED;
bi->generation = 0;
bi->sequence = 0;
@@ -3481,8 +3549,10 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
if (dir) {
ret = btrfs_set_inode_index(dir, index);
- if (ret)
+ if (ret) {
+ iput(inode);
return ERR_PTR(ret);
+ }
}
/*
* index_cnt is ignored for everything but a dir,
@@ -3565,6 +3635,7 @@ fail:
if (dir)
BTRFS_I(dir)->index_cnt--;
btrfs_free_path(path);
+ iput(inode);
return ERR_PTR(ret);
}
@@ -4631,47 +4702,36 @@ void btrfs_destroy_cachep(void)
kmem_cache_destroy(btrfs_trans_handle_cachep);
if (btrfs_transaction_cachep)
kmem_cache_destroy(btrfs_transaction_cachep);
- if (btrfs_bit_radix_cachep)
- kmem_cache_destroy(btrfs_bit_radix_cachep);
if (btrfs_path_cachep)
kmem_cache_destroy(btrfs_path_cachep);
}
-struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
- unsigned long extra_flags,
- void (*ctor)(void *))
-{
- return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
- SLAB_MEM_SPREAD | extra_flags), ctor);
-}
-
int btrfs_init_cachep(void)
{
- btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
- sizeof(struct btrfs_inode),
- 0, init_once);
+ btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
+ sizeof(struct btrfs_inode), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, init_once);
if (!btrfs_inode_cachep)
goto fail;
- btrfs_trans_handle_cachep =
- btrfs_cache_create("btrfs_trans_handle_cache",
- sizeof(struct btrfs_trans_handle),
- 0, NULL);
+
+ btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
+ sizeof(struct btrfs_trans_handle), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
if (!btrfs_trans_handle_cachep)
goto fail;
- btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
- sizeof(struct btrfs_transaction),
- 0, NULL);
+
+ btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
+ sizeof(struct btrfs_transaction), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
if (!btrfs_transaction_cachep)
goto fail;
- btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
- sizeof(struct btrfs_path),
- 0, NULL);
+
+ btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
+ sizeof(struct btrfs_path), 0,
+ SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
if (!btrfs_path_cachep)
goto fail;
- btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
- SLAB_DESTROY_BY_RCU, NULL);
- if (!btrfs_bit_radix_cachep)
- goto fail;
+
return 0;
fail:
btrfs_destroy_cachep();
@@ -4967,10 +5027,10 @@ out_fail:
return err;
}
-static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
- u64 alloc_hint, int mode)
+static int prealloc_file_range(struct btrfs_trans_handle *trans,
+ struct inode *inode, u64 start, u64 end,
+ u64 locked_end, u64 alloc_hint, int mode)
{
- struct btrfs_trans_handle *trans;
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_key ins;
u64 alloc_size;
@@ -4978,10 +5038,6 @@ static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
u64 num_bytes = end - start;
int ret = 0;
- trans = btrfs_join_transaction(root, 1);
- BUG_ON(!trans);
- btrfs_set_trans_block_group(trans, inode);
-
while (num_bytes > 0) {
alloc_size = min(num_bytes, root->fs_info->max_extent);
ret = btrfs_reserve_extent(trans, root, alloc_size,
@@ -4994,7 +5050,8 @@ static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
ret = insert_reserved_file_extent(trans, inode,
cur_offset, ins.objectid,
ins.offset, ins.offset,
- ins.offset, 0, 0, 0,
+ ins.offset, locked_end,
+ 0, 0, 0,
BTRFS_FILE_EXTENT_PREALLOC);
BUG_ON(ret);
num_bytes -= ins.offset;
@@ -5012,7 +5069,6 @@ out:
BUG_ON(ret);
}
- btrfs_end_transaction(trans, root);
return ret;
}
@@ -5024,13 +5080,21 @@ static long btrfs_fallocate(struct inode *inode, int mode,
u64 alloc_start;
u64 alloc_end;
u64 alloc_hint = 0;
+ u64 locked_end;
u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
struct extent_map *em;
+ struct btrfs_trans_handle *trans;
int ret;
alloc_start = offset & ~mask;
alloc_end = (offset + len + mask) & ~mask;
+ /*
+ * wait for ordered IO before we have any locks. We'll loop again
+ * below with the locks held.
+ */
+ btrfs_wait_ordered_range(inode, alloc_start, alloc_end - alloc_start);
+
mutex_lock(&inode->i_mutex);
if (alloc_start > inode->i_size) {
ret = btrfs_cont_expand(inode, alloc_start);
@@ -5038,10 +5102,21 @@ static long btrfs_fallocate(struct inode *inode, int mode,
goto out;
}
+ locked_end = alloc_end - 1;
while (1) {
struct btrfs_ordered_extent *ordered;
- lock_extent(&BTRFS_I(inode)->io_tree, alloc_start,
- alloc_end - 1, GFP_NOFS);
+
+ trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1);
+ if (!trans) {
+ ret = -EIO;
+ goto out;
+ }
+
+ /* the extent lock is ordered inside the running
+ * transaction
+ */
+ lock_extent(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
+ GFP_NOFS);
ordered = btrfs_lookup_first_ordered_extent(inode,
alloc_end - 1);
if (ordered &&
@@ -5049,7 +5124,13 @@ static long btrfs_fallocate(struct inode *inode, int mode,
ordered->file_offset < alloc_end) {
btrfs_put_ordered_extent(ordered);
unlock_extent(&BTRFS_I(inode)->io_tree,
- alloc_start, alloc_end - 1, GFP_NOFS);
+ alloc_start, locked_end, GFP_NOFS);
+ btrfs_end_transaction(trans, BTRFS_I(inode)->root);
+
+ /*
+ * we can't wait on the range with the transaction
+ * running or with the extent lock held
+ */
btrfs_wait_ordered_range(inode, alloc_start,
alloc_end - alloc_start);
} else {
@@ -5067,8 +5148,9 @@ static long btrfs_fallocate(struct inode *inode, int mode,
last_byte = min(extent_map_end(em), alloc_end);
last_byte = (last_byte + mask) & ~mask;
if (em->block_start == EXTENT_MAP_HOLE) {
- ret = prealloc_file_range(inode, cur_offset,
- last_byte, alloc_hint, mode);
+ ret = prealloc_file_range(trans, inode, cur_offset,
+ last_byte, locked_end + 1,
+ alloc_hint, mode);
if (ret < 0) {
free_extent_map(em);
break;
@@ -5084,8 +5166,10 @@ static long btrfs_fallocate(struct inode *inode, int mode,
break;
}
}
- unlock_extent(&BTRFS_I(inode)->io_tree, alloc_start, alloc_end - 1,
+ unlock_extent(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
GFP_NOFS);
+
+ btrfs_end_transaction(trans, BTRFS_I(inode)->root);
out:
mutex_unlock(&inode->i_mutex);
return ret;
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index bca729fc80c8..5e94ea6e1cbe 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -267,7 +267,7 @@ static noinline int btrfs_mksubvol(struct path *parent, char *name,
goto out_dput;
if (!IS_POSIXACL(parent->dentry->d_inode))
- mode &= ~current->fs->umask;
+ mode &= ~current_umask();
error = mnt_want_write(parent->mnt);
if (error)
@@ -461,15 +461,9 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
- vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
-
- if (!vol_args)
- return -ENOMEM;
-
- if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
- ret = -EFAULT;
- goto out;
- }
+ vol_args = memdup_user(arg, sizeof(*vol_args));
+ if (IS_ERR(vol_args))
+ return PTR_ERR(vol_args);
vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
namelen = strlen(vol_args->name);
@@ -483,11 +477,13 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
*devstr = '\0';
devstr = vol_args->name;
devid = simple_strtoull(devstr, &end, 10);
- printk(KERN_INFO "resizing devid %llu\n", devid);
+ printk(KERN_INFO "resizing devid %llu\n",
+ (unsigned long long)devid);
}
device = btrfs_find_device(root, devid, NULL, NULL);
if (!device) {
- printk(KERN_INFO "resizer unable to find device %llu\n", devid);
+ printk(KERN_INFO "resizer unable to find device %llu\n",
+ (unsigned long long)devid);
ret = -EINVAL;
goto out_unlock;
}
@@ -545,7 +541,6 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
out_unlock:
mutex_unlock(&root->fs_info->volume_mutex);
-out:
kfree(vol_args);
return ret;
}
@@ -565,15 +560,9 @@ static noinline int btrfs_ioctl_snap_create(struct file *file,
if (root->fs_info->sb->s_flags & MS_RDONLY)
return -EROFS;
- vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
-
- if (!vol_args)
- return -ENOMEM;
-
- if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
- ret = -EFAULT;
- goto out;
- }
+ vol_args = memdup_user(arg, sizeof(*vol_args));
+ if (IS_ERR(vol_args))
+ return PTR_ERR(vol_args);
vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
namelen = strlen(vol_args->name);
@@ -675,19 +664,13 @@ static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
- vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
+ vol_args = memdup_user(arg, sizeof(*vol_args));
+ if (IS_ERR(vol_args))
+ return PTR_ERR(vol_args);
- if (!vol_args)
- return -ENOMEM;
-
- if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
- ret = -EFAULT;
- goto out;
- }
vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
ret = btrfs_init_new_device(root, vol_args->name);
-out:
kfree(vol_args);
return ret;
}
@@ -703,19 +686,13 @@ static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
if (root->fs_info->sb->s_flags & MS_RDONLY)
return -EROFS;
- vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
+ vol_args = memdup_user(arg, sizeof(*vol_args));
+ if (IS_ERR(vol_args))
+ return PTR_ERR(vol_args);
- if (!vol_args)
- return -ENOMEM;
-
- if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
- ret = -EFAULT;
- goto out;
- }
vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
ret = btrfs_rm_device(root, vol_args->name);
-out:
kfree(vol_args);
return ret;
}
@@ -830,7 +807,8 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
BUG_ON(!trans);
/* punch hole in destination first */
- btrfs_drop_extents(trans, root, inode, off, off+len, 0, &hint_byte);
+ btrfs_drop_extents(trans, root, inode, off, off + len,
+ off + len, 0, &hint_byte);
/* clone data */
key.objectid = src->i_ino;
diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c
index a5310c0f41e2..1c36e5cd8f55 100644
--- a/fs/btrfs/locking.c
+++ b/fs/btrfs/locking.c
@@ -60,8 +60,8 @@ void btrfs_clear_lock_blocking(struct extent_buffer *eb)
/*
* unfortunately, many of the places that currently set a lock to blocking
- * don't end up blocking for every long, and often they don't block
- * at all. For a dbench 50 run, if we don't spin one the blocking bit
+ * don't end up blocking for very long, and often they don't block
+ * at all. For a dbench 50 run, if we don't spin on the blocking bit
* at all, the context switch rate can jump up to 400,000/sec or more.
*
* So, we're still stuck with this crummy spin on the blocking bit,
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index 53c87b197d70..d6f0806c682f 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -489,7 +489,7 @@ again:
/* start IO across the range first to instantiate any delalloc
* extents
*/
- btrfs_fdatawrite_range(inode->i_mapping, start, orig_end, WB_SYNC_NONE);
+ btrfs_fdatawrite_range(inode->i_mapping, start, orig_end, WB_SYNC_ALL);
/* The compression code will leave pages locked but return from
* writepage without setting the page writeback. Starting again
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 19a4daf03ccb..3536bdb2d7cb 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -24,6 +24,7 @@
#include <linux/highmem.h>
#include <linux/time.h>
#include <linux/init.h>
+#include <linux/seq_file.h>
#include <linux/string.h>
#include <linux/smp_lock.h>
#include <linux/backing-dev.h>
@@ -66,7 +67,8 @@ static void btrfs_put_super(struct super_block *sb)
enum {
Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
- Opt_ssd, Opt_thread_pool, Opt_noacl, Opt_compress, Opt_err,
+ Opt_ssd, Opt_thread_pool, Opt_noacl, Opt_compress, Opt_notreelog,
+ Opt_ratio, Opt_flushoncommit, Opt_err,
};
static match_table_t tokens = {
@@ -83,6 +85,9 @@ static match_table_t tokens = {
{Opt_compress, "compress"},
{Opt_ssd, "ssd"},
{Opt_noacl, "noacl"},
+ {Opt_notreelog, "notreelog"},
+ {Opt_flushoncommit, "flushoncommit"},
+ {Opt_ratio, "metadata_ratio=%d"},
{Opt_err, NULL},
};
@@ -191,7 +196,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
info->max_extent = max_t(u64,
info->max_extent, root->sectorsize);
printk(KERN_INFO "btrfs: max_extent at %llu\n",
- info->max_extent);
+ (unsigned long long)info->max_extent);
}
break;
case Opt_max_inline:
@@ -206,7 +211,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
root->sectorsize);
}
printk(KERN_INFO "btrfs: max_inline at %llu\n",
- info->max_inline);
+ (unsigned long long)info->max_inline);
}
break;
case Opt_alloc_start:
@@ -216,12 +221,29 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
kfree(num);
printk(KERN_INFO
"btrfs: allocations start at %llu\n",
- info->alloc_start);
+ (unsigned long long)info->alloc_start);
}
break;
case Opt_noacl:
root->fs_info->sb->s_flags &= ~MS_POSIXACL;
break;
+ case Opt_notreelog:
+ printk(KERN_INFO "btrfs: disabling tree log\n");
+ btrfs_set_opt(info->mount_opt, NOTREELOG);
+ break;
+ case Opt_flushoncommit:
+ printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
+ btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
+ break;
+ case Opt_ratio:
+ intarg = 0;
+ match_int(&args[0], &intarg);
+ if (intarg) {
+ info->metadata_ratio = intarg;
+ printk(KERN_INFO "btrfs: metadata ratio %d\n",
+ info->metadata_ratio);
+ }
+ break;
default:
break;
}
@@ -363,9 +385,8 @@ fail_close:
int btrfs_sync_fs(struct super_block *sb, int wait)
{
struct btrfs_trans_handle *trans;
- struct btrfs_root *root;
+ struct btrfs_root *root = btrfs_sb(sb);
int ret;
- root = btrfs_sb(sb);
if (sb->s_flags & MS_RDONLY)
return 0;
@@ -385,6 +406,44 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
return ret;
}
+static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
+{
+ struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
+ struct btrfs_fs_info *info = root->fs_info;
+
+ if (btrfs_test_opt(root, DEGRADED))
+ seq_puts(seq, ",degraded");
+ if (btrfs_test_opt(root, NODATASUM))
+ seq_puts(seq, ",nodatasum");
+ if (btrfs_test_opt(root, NODATACOW))
+ seq_puts(seq, ",nodatacow");
+ if (btrfs_test_opt(root, NOBARRIER))
+ seq_puts(seq, ",nobarrier");
+ if (info->max_extent != (u64)-1)
+ seq_printf(seq, ",max_extent=%llu",
+ (unsigned long long)info->max_extent);
+ if (info->max_inline != 8192 * 1024)
+ seq_printf(seq, ",max_inline=%llu",
+ (unsigned long long)info->max_inline);
+ if (info->alloc_start != 0)
+ seq_printf(seq, ",alloc_start=%llu",
+ (unsigned long long)info->alloc_start);
+ if (info->thread_pool_size != min_t(unsigned long,
+ num_online_cpus() + 2, 8))
+ seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
+ if (btrfs_test_opt(root, COMPRESS))
+ seq_puts(seq, ",compress");
+ if (btrfs_test_opt(root, SSD))
+ seq_puts(seq, ",ssd");
+ if (btrfs_test_opt(root, NOTREELOG))
+ seq_puts(seq, ",no-treelog");
+ if (btrfs_test_opt(root, FLUSHONCOMMIT))
+ seq_puts(seq, ",flush-on-commit");
+ if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
+ seq_puts(seq, ",noacl");
+ return 0;
+}
+
static void btrfs_write_super(struct super_block *sb)
{
sb->s_dirt = 0;
@@ -589,14 +648,9 @@ static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
- vol = kmalloc(sizeof(*vol), GFP_KERNEL);
- if (!vol)
- return -ENOMEM;
-
- if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
- ret = -EFAULT;
- goto out;
- }
+ vol = memdup_user((void __user *)arg, sizeof(*vol));
+ if (IS_ERR(vol))
+ return PTR_ERR(vol);
switch (cmd) {
case BTRFS_IOC_SCAN_DEV:
@@ -604,7 +658,7 @@ static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
&btrfs_fs_type, &fs_devices);
break;
}
-out:
+
kfree(vol);
return ret;
}
@@ -630,7 +684,7 @@ static struct super_operations btrfs_super_ops = {
.put_super = btrfs_put_super,
.write_super = btrfs_write_super,
.sync_fs = btrfs_sync_fs,
- .show_options = generic_show_options,
+ .show_options = btrfs_show_options,
.write_inode = btrfs_write_inode,
.dirty_inode = btrfs_dirty_inode,
.alloc_inode = btrfs_alloc_inode,
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 664782c6a2df..01b143605ec1 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -53,8 +53,6 @@ static noinline int join_transaction(struct btrfs_root *root)
GFP_NOFS);
BUG_ON(!cur_trans);
root->fs_info->generation++;
- root->fs_info->last_alloc = 0;
- root->fs_info->last_data_alloc = 0;
cur_trans->num_writers = 1;
cur_trans->num_joined = 0;
cur_trans->transid = root->fs_info->generation;
@@ -689,7 +687,13 @@ static noinline int wait_transaction_pre_flush(struct btrfs_fs_info *info)
prepare_to_wait(&info->transaction_wait, &wait,
TASK_UNINTERRUPTIBLE);
mutex_unlock(&info->trans_mutex);
+
+ atomic_dec(&info->throttles);
+ wake_up(&info->transaction_throttle);
+
schedule();
+
+ atomic_inc(&info->throttles);
mutex_lock(&info->trans_mutex);
finish_wait(&info->transaction_wait, &wait);
}
@@ -974,6 +978,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
int ret;
int should_grow = 0;
unsigned long now = get_seconds();
+ int flush_on_commit = btrfs_test_opt(root, FLUSHONCOMMIT);
btrfs_run_ordered_operations(root, 0);
@@ -1053,7 +1058,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
mutex_unlock(&root->fs_info->trans_mutex);
- if (snap_pending) {
+ if (flush_on_commit || snap_pending) {
+ if (flush_on_commit)
+ btrfs_start_delalloc_inodes(root);
ret = btrfs_wait_ordered_extents(root, 1);
BUG_ON(ret);
}
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index fc9b87a7975b..db5e212e8445 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -262,11 +262,9 @@ static int process_one_buffer(struct btrfs_root *log,
struct extent_buffer *eb,
struct walk_control *wc, u64 gen)
{
- if (wc->pin) {
- mutex_lock(&log->fs_info->pinned_mutex);
+ if (wc->pin)
btrfs_update_pinned_extents(log->fs_info->extent_root,
eb->start, eb->len, 1);
- }
if (btrfs_buffer_uptodate(eb, gen)) {
if (wc->write)
@@ -538,7 +536,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
saved_nbytes = inode_get_bytes(inode);
/* drop any overlapping extents */
ret = btrfs_drop_extents(trans, root, inode,
- start, extent_end, start, &alloc_hint);
+ start, extent_end, extent_end, start, &alloc_hint);
BUG_ON(ret);
if (found_type == BTRFS_FILE_EXTENT_REG ||
@@ -1224,8 +1222,7 @@ insert:
ret = insert_one_name(trans, root, path, key->objectid, key->offset,
name, name_len, log_type, &log_key);
- if (ret && ret != -ENOENT)
- BUG();
+ BUG_ON(ret && ret != -ENOENT);
goto out;
}
@@ -2900,6 +2897,11 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
sb = inode->i_sb;
+ if (btrfs_test_opt(root, NOTREELOG)) {
+ ret = 1;
+ goto end_no_trans;
+ }
+
if (root->fs_info->last_trans_log_full_commit >
root->fs_info->last_trans_committed) {
ret = 1;
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index dd06e18e5aac..5f01dad4b696 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -20,6 +20,7 @@
#include <linux/buffer_head.h>
#include <linux/blkdev.h>
#include <linux/random.h>
+#include <linux/iocontext.h>
#include <asm/div64.h>
#include "compat.h"
#include "ctree.h"
@@ -124,6 +125,20 @@ static noinline struct btrfs_fs_devices *find_fsid(u8 *fsid)
return NULL;
}
+static void requeue_list(struct btrfs_pending_bios *pending_bios,
+ struct bio *head, struct bio *tail)
+{
+
+ struct bio *old_head;
+
+ old_head = pending_bios->head;
+ pending_bios->head = head;
+ if (pending_bios->tail)
+ tail->bi_next = old_head;
+ else
+ pending_bios->tail = tail;
+}
+
/*
* we try to collect pending bios for a device so we don't get a large
* number of procs sending bios down to the same device. This greatly
@@ -140,31 +155,44 @@ static noinline int run_scheduled_bios(struct btrfs_device *device)
struct bio *pending;
struct backing_dev_info *bdi;
struct btrfs_fs_info *fs_info;
+ struct btrfs_pending_bios *pending_bios;
struct bio *tail;
struct bio *cur;
int again = 0;
- unsigned long num_run = 0;
+ unsigned long num_run;
+ unsigned long num_sync_run;
unsigned long limit;
+ unsigned long last_waited = 0;
- bdi = device->bdev->bd_inode->i_mapping->backing_dev_info;
+ bdi = blk_get_backing_dev_info(device->bdev);
fs_info = device->dev_root->fs_info;
limit = btrfs_async_submit_limit(fs_info);
limit = limit * 2 / 3;
+ /* we want to make sure that every time we switch from the sync
+ * list to the normal list, we unplug
+ */
+ num_sync_run = 0;
+
loop:
spin_lock(&device->io_lock);
+ num_run = 0;
loop_lock:
+
/* take all the bios off the list at once and process them
* later on (without the lock held). But, remember the
* tail and other pointers so the bios can be properly reinserted
* into the list if we hit congestion
*/
- pending = device->pending_bios;
- tail = device->pending_bio_tail;
+ if (device->pending_sync_bios.head)
+ pending_bios = &device->pending_sync_bios;
+ else
+ pending_bios = &device->pending_bios;
+
+ pending = pending_bios->head;
+ tail = pending_bios->tail;
WARN_ON(pending && !tail);
- device->pending_bios = NULL;
- device->pending_bio_tail = NULL;
/*
* if pending was null this time around, no bios need processing
@@ -174,16 +202,41 @@ loop_lock:
* device->running_pending is used to synchronize with the
* schedule_bio code.
*/
- if (pending) {
- again = 1;
- device->running_pending = 1;
- } else {
+ if (device->pending_sync_bios.head == NULL &&
+ device->pending_bios.head == NULL) {
again = 0;
device->running_pending = 0;
+ } else {
+ again = 1;
+ device->running_pending = 1;
}
+
+ pending_bios->head = NULL;
+ pending_bios->tail = NULL;
+
spin_unlock(&device->io_lock);
+ /*
+ * if we're doing the regular priority list, make sure we unplug
+ * for any high prio bios we've sent down
+ */
+ if (pending_bios == &device->pending_bios && num_sync_run > 0) {
+ num_sync_run = 0;
+ blk_run_backing_dev(bdi, NULL);
+ }
+
while (pending) {
+
+ rmb();
+ if (pending_bios != &device->pending_sync_bios &&
+ device->pending_sync_bios.head &&
+ num_run > 16) {
+ cond_resched();
+ spin_lock(&device->io_lock);
+ requeue_list(pending_bios, pending, tail);
+ goto loop_lock;
+ }
+
cur = pending;
pending = pending->bi_next;
cur->bi_next = NULL;
@@ -194,10 +247,18 @@ loop_lock:
wake_up(&fs_info->async_submit_wait);
BUG_ON(atomic_read(&cur->bi_cnt) == 0);
- bio_get(cur);
submit_bio(cur->bi_rw, cur);
- bio_put(cur);
num_run++;
+ if (bio_sync(cur))
+ num_sync_run++;
+
+ if (need_resched()) {
+ if (num_sync_run) {
+ blk_run_backing_dev(bdi, NULL);
+ num_sync_run = 0;
+ }
+ cond_resched();
+ }
/*
* we made progress, there is more work to do and the bdi
@@ -206,17 +267,41 @@ loop_lock:
*/
if (pending && bdi_write_congested(bdi) && num_run > 16 &&
fs_info->fs_devices->open_devices > 1) {
- struct bio *old_head;
+ struct io_context *ioc;
- spin_lock(&device->io_lock);
-
- old_head = device->pending_bios;
- device->pending_bios = pending;
- if (device->pending_bio_tail)
- tail->bi_next = old_head;
- else
- device->pending_bio_tail = tail;
+ ioc = current->io_context;
+ /*
+ * the main goal here is that we don't want to
+ * block if we're going to be able to submit
+ * more requests without blocking.
+ *
+ * This code does two great things, it pokes into
+ * the elevator code from a filesystem _and_
+ * it makes assumptions about how batching works.
+ */
+ if (ioc && ioc->nr_batch_requests > 0 &&
+ time_before(jiffies, ioc->last_waited + HZ/50UL) &&
+ (last_waited == 0 ||
+ ioc->last_waited == last_waited)) {
+ /*
+ * we want to go through our batch of
+ * requests and stop. So, we copy out
+ * the ioc->last_waited time and test
+ * against it before looping
+ */
+ last_waited = ioc->last_waited;
+ if (need_resched()) {
+ if (num_sync_run) {
+ blk_run_backing_dev(bdi, NULL);
+ num_sync_run = 0;
+ }
+ cond_resched();
+ }
+ continue;
+ }
+ spin_lock(&device->io_lock);
+ requeue_list(pending_bios, pending, tail);
device->running_pending = 1;
spin_unlock(&device->io_lock);
@@ -224,13 +309,32 @@ loop_lock:
goto done;
}
}
+
+ if (num_sync_run) {
+ num_sync_run = 0;
+ blk_run_backing_dev(bdi, NULL);
+ }
+
+ cond_resched();
if (again)
goto loop;
spin_lock(&device->io_lock);
- if (device->pending_bios)
+ if (device->pending_bios.head || device->pending_sync_bios.head)
goto loop_lock;
spin_unlock(&device->io_lock);
+
+ /*
+ * IO has already been through a long path to get here. Checksumming,
+ * async helper threads, perhaps compression. We've done a pretty
+ * good job of collecting a batch of IO and should just unplug
+ * the device right away.
+ *
+ * This will help anyone who is waiting on the IO, they might have
+ * already unplugged, but managed to do so before the bio they
+ * cared about found its way down here.
+ */
+ blk_run_backing_dev(bdi, NULL);
done:
return 0;
}
@@ -1439,7 +1543,7 @@ static noinline int btrfs_update_device(struct btrfs_trans_handle *trans,
btrfs_set_device_io_align(leaf, dev_item, device->io_align);
btrfs_set_device_io_width(leaf, dev_item, device->io_width);
btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
- btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
+ btrfs_set_device_total_bytes(leaf, dev_item, device->disk_total_bytes);
btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
btrfs_mark_buffer_dirty(leaf);
@@ -1836,14 +1940,6 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
device->total_bytes = new_size;
if (device->writeable)
device->fs_devices->total_rw_bytes -= diff;
- ret = btrfs_update_device(trans, device);
- if (ret) {
- unlock_chunks(root);
- btrfs_end_transaction(trans, root);
- goto done;
- }
- WARN_ON(diff > old_total);
- btrfs_set_super_total_bytes(super_copy, old_total - diff);
unlock_chunks(root);
btrfs_end_transaction(trans, root);
@@ -1875,7 +1971,7 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
length = btrfs_dev_extent_length(l, dev_extent);
if (key.offset + length <= new_size)
- goto done;
+ break;
chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent);
chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent);
@@ -1888,6 +1984,26 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
goto done;
}
+ /* Shrinking succeeded, else we would be at "done". */
+ trans = btrfs_start_transaction(root, 1);
+ if (!trans) {
+ ret = -ENOMEM;
+ goto done;
+ }
+ lock_chunks(root);
+
+ device->disk_total_bytes = new_size;
+ /* Now btrfs_update_device() will change the on-disk size. */
+ ret = btrfs_update_device(trans, device);
+ if (ret) {
+ unlock_chunks(root);
+ btrfs_end_transaction(trans, root);
+ goto done;
+ }
+ WARN_ON(diff > old_total);
+ btrfs_set_super_total_bytes(super_copy, old_total - diff);
+ unlock_chunks(root);
+ btrfs_end_transaction(trans, root);
done:
btrfs_free_path(path);
return ret;
@@ -2458,7 +2574,7 @@ again:
max_errors = 1;
}
}
- if (multi_ret && rw == WRITE &&
+ if (multi_ret && (rw & (1 << BIO_RW)) &&
stripes_allocated < stripes_required) {
stripes_allocated = map->num_stripes;
free_extent_map(em);
@@ -2723,6 +2839,7 @@ static noinline int schedule_bio(struct btrfs_root *root,
int rw, struct bio *bio)
{
int should_queue = 1;
+ struct btrfs_pending_bios *pending_bios;
/* don't bother with additional async steps for reads, right now */
if (!(rw & (1 << BIO_RW))) {
@@ -2744,13 +2861,17 @@ static noinline int schedule_bio(struct btrfs_root *root,
bio->bi_rw |= rw;
spin_lock(&device->io_lock);
+ if (bio_sync(bio))
+ pending_bios = &device->pending_sync_bios;
+ else
+ pending_bios = &device->pending_bios;
- if (device->pending_bio_tail)
- device->pending_bio_tail->bi_next = bio;
+ if (pending_bios->tail)
+ pending_bios->tail->bi_next = bio;
- device->pending_bio_tail = bio;
- if (!device->pending_bios)
- device->pending_bios = bio;
+ pending_bios->tail = bio;
+ if (!pending_bios->head)
+ pending_bios->head = bio;
if (device->running_pending)
should_queue = 0;
@@ -2967,7 +3088,8 @@ static int fill_device_from_item(struct extent_buffer *leaf,
unsigned long ptr;
device->devid = btrfs_device_id(leaf, dev_item);
- device->total_bytes = btrfs_device_total_bytes(leaf, dev_item);
+ device->disk_total_bytes = btrfs_device_total_bytes(leaf, dev_item);
+ device->total_bytes = device->disk_total_bytes;
device->bytes_used = btrfs_device_bytes_used(leaf, dev_item);
device->type = btrfs_device_type(leaf, dev_item);
device->io_align = btrfs_device_io_align(leaf, dev_item);
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 86c44e9ae110..5c3ff6d02fd7 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -23,13 +23,22 @@
#include "async-thread.h"
struct buffer_head;
+struct btrfs_pending_bios {
+ struct bio *head;
+ struct bio *tail;
+};
+
struct btrfs_device {
struct list_head dev_list;
struct list_head dev_alloc_list;
struct btrfs_fs_devices *fs_devices;
struct btrfs_root *dev_root;
- struct bio *pending_bios;
- struct bio *pending_bio_tail;
+
+ /* regular prio bios */
+ struct btrfs_pending_bios pending_bios;
+ /* WRITE_SYNC bios */
+ struct btrfs_pending_bios pending_sync_bios;
+
int running_pending;
u64 generation;
@@ -52,6 +61,9 @@ struct btrfs_device {
/* size of the device */
u64 total_bytes;
+ /* size of the disk */
+ u64 disk_total_bytes;
+
/* bytes used */
u64 bytes_used;
@@ -76,7 +88,7 @@ struct btrfs_device {
struct btrfs_fs_devices {
u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
- /* the device with this id has the most recent coyp of the super */
+ /* the device with this id has the most recent copy of the super */
u64 latest_devid;
u64 latest_trans;
u64 num_devices;
Powered by cgit, Gruvboxed by Blaster4385.