aboutsummaryrefslogtreecommitdiffstats
path: root/fs/btrfs
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2009-10-15 18:06:37 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2009-10-15 18:06:37 -0400
commitdcbeb0bec5f2695c3ff53f174efb8e03c209f3f3 (patch)
tree30d223a3a3c7470c657284ef030657bd1753d4d3 /fs/btrfs
parent2b650df2cea96e487f2fd9ecaa68e533ea9b5ed7 (diff)
parent444528b3e614f7f2391488d9bca8e0b872db909b (diff)
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable
* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable: Btrfs: always pin metadata in discard mode Btrfs: enable discard support Btrfs: add -o discard option Btrfs: properly wait log writers during log sync Btrfs: fix possible ENOSPC problems with truncate Btrfs: fix btrfs acl #ifdef checks Btrfs: streamline tree-log btree block writeout Btrfs: avoid tree log commit when there are no changes Btrfs: only write one super copy during fsync
Diffstat (limited to 'fs/btrfs')
-rw-r--r--fs/btrfs/acl.c6
-rw-r--r--fs/btrfs/btrfs_inode.h6
-rw-r--r--fs/btrfs/ctree.h4
-rw-r--r--fs/btrfs/disk-io.c2
-rw-r--r--fs/btrfs/extent-tree.c17
-rw-r--r--fs/btrfs/file.c41
-rw-r--r--fs/btrfs/inode.c33
-rw-r--r--fs/btrfs/super.c9
-rw-r--r--fs/btrfs/transaction.c45
-rw-r--r--fs/btrfs/transaction.h5
-rw-r--r--fs/btrfs/tree-log.c48
-rw-r--r--fs/btrfs/tree-log.h3
-rw-r--r--fs/btrfs/xattr.c2
13 files changed, 179 insertions, 42 deletions
diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c
index 69b355ae7f49..361604244271 100644
--- a/fs/btrfs/acl.c
+++ b/fs/btrfs/acl.c
@@ -27,7 +27,7 @@
27#include "btrfs_inode.h" 27#include "btrfs_inode.h"
28#include "xattr.h" 28#include "xattr.h"
29 29
30#ifdef CONFIG_BTRFS_POSIX_ACL 30#ifdef CONFIG_BTRFS_FS_POSIX_ACL
31 31
32static struct posix_acl *btrfs_get_acl(struct inode *inode, int type) 32static struct posix_acl *btrfs_get_acl(struct inode *inode, int type)
33{ 33{
@@ -313,7 +313,7 @@ struct xattr_handler btrfs_xattr_acl_access_handler = {
313 .set = btrfs_xattr_acl_access_set, 313 .set = btrfs_xattr_acl_access_set,
314}; 314};
315 315
316#else /* CONFIG_BTRFS_POSIX_ACL */ 316#else /* CONFIG_BTRFS_FS_POSIX_ACL */
317 317
318int btrfs_acl_chmod(struct inode *inode) 318int btrfs_acl_chmod(struct inode *inode)
319{ 319{
@@ -325,4 +325,4 @@ int btrfs_init_acl(struct inode *inode, struct inode *dir)
325 return 0; 325 return 0;
326} 326}
327 327
328#endif /* CONFIG_BTRFS_POSIX_ACL */ 328#endif /* CONFIG_BTRFS_FS_POSIX_ACL */
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index c71abec0ab90..f6783a42f010 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -86,6 +86,12 @@ struct btrfs_inode {
86 * transid of the trans_handle that last modified this inode 86 * transid of the trans_handle that last modified this inode
87 */ 87 */
88 u64 last_trans; 88 u64 last_trans;
89
90 /*
91 * log transid when this inode was last modified
92 */
93 u64 last_sub_trans;
94
89 /* 95 /*
90 * transid that last logged this inode 96 * transid that last logged this inode
91 */ 97 */
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 1bb897ecdeeb..444b3e9b92a4 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1009,6 +1009,7 @@ struct btrfs_root {
1009 atomic_t log_writers; 1009 atomic_t log_writers;
1010 atomic_t log_commit[2]; 1010 atomic_t log_commit[2];
1011 unsigned long log_transid; 1011 unsigned long log_transid;
1012 unsigned long last_log_commit;
1012 unsigned long log_batch; 1013 unsigned long log_batch;
1013 pid_t log_start_pid; 1014 pid_t log_start_pid;
1014 bool log_multiple_pids; 1015 bool log_multiple_pids;
@@ -1152,6 +1153,7 @@ struct btrfs_root {
1152#define BTRFS_MOUNT_FLUSHONCOMMIT (1 << 7) 1153#define BTRFS_MOUNT_FLUSHONCOMMIT (1 << 7)
1153#define BTRFS_MOUNT_SSD_SPREAD (1 << 8) 1154#define BTRFS_MOUNT_SSD_SPREAD (1 << 8)
1154#define BTRFS_MOUNT_NOSSD (1 << 9) 1155#define BTRFS_MOUNT_NOSSD (1 << 9)
1156#define BTRFS_MOUNT_DISCARD (1 << 10)
1155 1157
1156#define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) 1158#define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt)
1157#define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) 1159#define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt)
@@ -2373,7 +2375,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options);
2373int btrfs_sync_fs(struct super_block *sb, int wait); 2375int btrfs_sync_fs(struct super_block *sb, int wait);
2374 2376
2375/* acl.c */ 2377/* acl.c */
2376#ifdef CONFIG_BTRFS_POSIX_ACL 2378#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2377int btrfs_check_acl(struct inode *inode, int mask); 2379int btrfs_check_acl(struct inode *inode, int mask);
2378#else 2380#else
2379#define btrfs_check_acl NULL 2381#define btrfs_check_acl NULL
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 100551a66c46..02b6afbd7450 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -917,6 +917,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
917 atomic_set(&root->log_writers, 0); 917 atomic_set(&root->log_writers, 0);
918 root->log_batch = 0; 918 root->log_batch = 0;
919 root->log_transid = 0; 919 root->log_transid = 0;
920 root->last_log_commit = 0;
920 extent_io_tree_init(&root->dirty_log_pages, 921 extent_io_tree_init(&root->dirty_log_pages,
921 fs_info->btree_inode->i_mapping, GFP_NOFS); 922 fs_info->btree_inode->i_mapping, GFP_NOFS);
922 923
@@ -1087,6 +1088,7 @@ int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1087 WARN_ON(root->log_root); 1088 WARN_ON(root->log_root);
1088 root->log_root = log_root; 1089 root->log_root = log_root;
1089 root->log_transid = 0; 1090 root->log_transid = 0;
1091 root->last_log_commit = 0;
1090 return 0; 1092 return 0;
1091} 1093}
1092 1094
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index d0c4d584efad..e238a0cdac67 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1568,23 +1568,23 @@ static int remove_extent_backref(struct btrfs_trans_handle *trans,
1568 return ret; 1568 return ret;
1569} 1569}
1570 1570
1571#ifdef BIO_RW_DISCARD
1572static void btrfs_issue_discard(struct block_device *bdev, 1571static void btrfs_issue_discard(struct block_device *bdev,
1573 u64 start, u64 len) 1572 u64 start, u64 len)
1574{ 1573{
1575 blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL, 1574 blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL,
1576 DISCARD_FL_BARRIER); 1575 DISCARD_FL_BARRIER);
1577} 1576}
1578#endif
1579 1577
1580static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr, 1578static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1581 u64 num_bytes) 1579 u64 num_bytes)
1582{ 1580{
1583#ifdef BIO_RW_DISCARD
1584 int ret; 1581 int ret;
1585 u64 map_length = num_bytes; 1582 u64 map_length = num_bytes;
1586 struct btrfs_multi_bio *multi = NULL; 1583 struct btrfs_multi_bio *multi = NULL;
1587 1584
1585 if (!btrfs_test_opt(root, DISCARD))
1586 return 0;
1587
1588 /* Tell the block device(s) that the sectors can be discarded */ 1588 /* Tell the block device(s) that the sectors can be discarded */
1589 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ, 1589 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
1590 bytenr, &map_length, &multi, 0); 1590 bytenr, &map_length, &multi, 0);
@@ -1604,9 +1604,6 @@ static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1604 } 1604 }
1605 1605
1606 return ret; 1606 return ret;
1607#else
1608 return 0;
1609#endif
1610} 1607}
1611 1608
1612int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, 1609int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
@@ -3690,6 +3687,14 @@ static int pin_down_bytes(struct btrfs_trans_handle *trans,
3690 if (is_data) 3687 if (is_data)
3691 goto pinit; 3688 goto pinit;
3692 3689
3690 /*
3691 * discard is sloooow, and so triggering discards on
3692 * individual btree blocks isn't a good plan. Just
3693 * pin everything in discard mode.
3694 */
3695 if (btrfs_test_opt(root, DISCARD))
3696 goto pinit;
3697
3693 buf = btrfs_find_tree_block(root, bytenr, num_bytes); 3698 buf = btrfs_find_tree_block(root, bytenr, num_bytes);
3694 if (!buf) 3699 if (!buf)
3695 goto pinit; 3700 goto pinit;
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 2d623aa0625f..06550affbd27 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1086,8 +1086,10 @@ out_nolock:
1086 btrfs_end_transaction(trans, root); 1086 btrfs_end_transaction(trans, root);
1087 else 1087 else
1088 btrfs_commit_transaction(trans, root); 1088 btrfs_commit_transaction(trans, root);
1089 } else { 1089 } else if (ret != BTRFS_NO_LOG_SYNC) {
1090 btrfs_commit_transaction(trans, root); 1090 btrfs_commit_transaction(trans, root);
1091 } else {
1092 btrfs_end_transaction(trans, root);
1091 } 1093 }
1092 } 1094 }
1093 if (file->f_flags & O_DIRECT) { 1095 if (file->f_flags & O_DIRECT) {
@@ -1137,6 +1139,13 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
1137 int ret = 0; 1139 int ret = 0;
1138 struct btrfs_trans_handle *trans; 1140 struct btrfs_trans_handle *trans;
1139 1141
1142
1143 /* we wait first, since the writeback may change the inode */
1144 root->log_batch++;
1145 /* the VFS called filemap_fdatawrite for us */
1146 btrfs_wait_ordered_range(inode, 0, (u64)-1);
1147 root->log_batch++;
1148
1140 /* 1149 /*
1141 * check the transaction that last modified this inode 1150 * check the transaction that last modified this inode
1142 * and see if its already been committed 1151 * and see if its already been committed
@@ -1144,6 +1153,11 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
1144 if (!BTRFS_I(inode)->last_trans) 1153 if (!BTRFS_I(inode)->last_trans)
1145 goto out; 1154 goto out;
1146 1155
1156 /*
1157 * if the last transaction that changed this file was before
1158 * the current transaction, we can bail out now without any
1159 * syncing
1160 */
1147 mutex_lock(&root->fs_info->trans_mutex); 1161 mutex_lock(&root->fs_info->trans_mutex);
1148 if (BTRFS_I(inode)->last_trans <= 1162 if (BTRFS_I(inode)->last_trans <=
1149 root->fs_info->last_trans_committed) { 1163 root->fs_info->last_trans_committed) {
@@ -1153,13 +1167,6 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
1153 } 1167 }
1154 mutex_unlock(&root->fs_info->trans_mutex); 1168 mutex_unlock(&root->fs_info->trans_mutex);
1155 1169
1156 root->log_batch++;
1157 filemap_fdatawrite(inode->i_mapping);
1158 btrfs_wait_ordered_range(inode, 0, (u64)-1);
1159 root->log_batch++;
1160
1161 if (datasync && !(inode->i_state & I_DIRTY_PAGES))
1162 goto out;
1163 /* 1170 /*
1164 * ok we haven't committed the transaction yet, lets do a commit 1171 * ok we haven't committed the transaction yet, lets do a commit
1165 */ 1172 */
@@ -1188,14 +1195,18 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
1188 */ 1195 */
1189 mutex_unlock(&dentry->d_inode->i_mutex); 1196 mutex_unlock(&dentry->d_inode->i_mutex);
1190 1197
1191 if (ret > 0) { 1198 if (ret != BTRFS_NO_LOG_SYNC) {
1192 ret = btrfs_commit_transaction(trans, root); 1199 if (ret > 0) {
1193 } else {
1194 ret = btrfs_sync_log(trans, root);
1195 if (ret == 0)
1196 ret = btrfs_end_transaction(trans, root);
1197 else
1198 ret = btrfs_commit_transaction(trans, root); 1200 ret = btrfs_commit_transaction(trans, root);
1201 } else {
1202 ret = btrfs_sync_log(trans, root);
1203 if (ret == 0)
1204 ret = btrfs_end_transaction(trans, root);
1205 else
1206 ret = btrfs_commit_transaction(trans, root);
1207 }
1208 } else {
1209 ret = btrfs_end_transaction(trans, root);
1199 } 1210 }
1200 mutex_lock(&dentry->d_inode->i_mutex); 1211 mutex_lock(&dentry->d_inode->i_mutex);
1201out: 1212out:
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 9e138b793dc7..dae12dc7e159 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3032,12 +3032,22 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
3032 3032
3033 if ((offset & (blocksize - 1)) == 0) 3033 if ((offset & (blocksize - 1)) == 0)
3034 goto out; 3034 goto out;
3035 ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE);
3036 if (ret)
3037 goto out;
3038
3039 ret = btrfs_reserve_metadata_for_delalloc(root, inode, 1);
3040 if (ret)
3041 goto out;
3035 3042
3036 ret = -ENOMEM; 3043 ret = -ENOMEM;
3037again: 3044again:
3038 page = grab_cache_page(mapping, index); 3045 page = grab_cache_page(mapping, index);
3039 if (!page) 3046 if (!page) {
3047 btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
3048 btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
3040 goto out; 3049 goto out;
3050 }
3041 3051
3042 page_start = page_offset(page); 3052 page_start = page_offset(page);
3043 page_end = page_start + PAGE_CACHE_SIZE - 1; 3053 page_end = page_start + PAGE_CACHE_SIZE - 1;
@@ -3070,6 +3080,10 @@ again:
3070 goto again; 3080 goto again;
3071 } 3081 }
3072 3082
3083 clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end,
3084 EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
3085 GFP_NOFS);
3086
3073 ret = btrfs_set_extent_delalloc(inode, page_start, page_end); 3087 ret = btrfs_set_extent_delalloc(inode, page_start, page_end);
3074 if (ret) { 3088 if (ret) {
3075 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 3089 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
@@ -3088,6 +3102,9 @@ again:
3088 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 3102 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3089 3103
3090out_unlock: 3104out_unlock:
3105 if (ret)
3106 btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
3107 btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
3091 unlock_page(page); 3108 unlock_page(page);
3092 page_cache_release(page); 3109 page_cache_release(page);
3093out: 3110out:
@@ -3111,7 +3128,9 @@ int btrfs_cont_expand(struct inode *inode, loff_t size)
3111 if (size <= hole_start) 3128 if (size <= hole_start)
3112 return 0; 3129 return 0;
3113 3130
3114 btrfs_truncate_page(inode->i_mapping, inode->i_size); 3131 err = btrfs_truncate_page(inode->i_mapping, inode->i_size);
3132 if (err)
3133 return err;
3115 3134
3116 while (1) { 3135 while (1) {
3117 struct btrfs_ordered_extent *ordered; 3136 struct btrfs_ordered_extent *ordered;
@@ -3480,6 +3499,7 @@ static noinline void init_btrfs_i(struct inode *inode)
3480 bi->generation = 0; 3499 bi->generation = 0;
3481 bi->sequence = 0; 3500 bi->sequence = 0;
3482 bi->last_trans = 0; 3501 bi->last_trans = 0;
3502 bi->last_sub_trans = 0;
3483 bi->logged_trans = 0; 3503 bi->logged_trans = 0;
3484 bi->delalloc_bytes = 0; 3504 bi->delalloc_bytes = 0;
3485 bi->reserved_bytes = 0; 3505 bi->reserved_bytes = 0;
@@ -4980,7 +5000,9 @@ again:
4980 set_page_dirty(page); 5000 set_page_dirty(page);
4981 SetPageUptodate(page); 5001 SetPageUptodate(page);
4982 5002
4983 BTRFS_I(inode)->last_trans = root->fs_info->generation + 1; 5003 BTRFS_I(inode)->last_trans = root->fs_info->generation;
5004 BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;
5005
4984 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 5006 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
4985 5007
4986out_unlock: 5008out_unlock:
@@ -5005,7 +5027,9 @@ static void btrfs_truncate(struct inode *inode)
5005 if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) 5027 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
5006 return; 5028 return;
5007 5029
5008 btrfs_truncate_page(inode->i_mapping, inode->i_size); 5030 ret = btrfs_truncate_page(inode->i_mapping, inode->i_size);
5031 if (ret)
5032 return;
5009 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1); 5033 btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
5010 5034
5011 trans = btrfs_start_transaction(root, 1); 5035 trans = btrfs_start_transaction(root, 1);
@@ -5100,6 +5124,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
5100 if (!ei) 5124 if (!ei)
5101 return NULL; 5125 return NULL;
5102 ei->last_trans = 0; 5126 ei->last_trans = 0;
5127 ei->last_sub_trans = 0;
5103 ei->logged_trans = 0; 5128 ei->logged_trans = 0;
5104 ei->outstanding_extents = 0; 5129 ei->outstanding_extents = 0;
5105 ei->reserved_extents = 0; 5130 ei->reserved_extents = 0;
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 9de9b2236419..752a5463bf53 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -66,7 +66,8 @@ enum {
66 Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow, 66 Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
67 Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, 67 Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
68 Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, 68 Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl,
69 Opt_compress, Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_err, 69 Opt_compress, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
70 Opt_discard, Opt_err,
70}; 71};
71 72
72static match_table_t tokens = { 73static match_table_t tokens = {
@@ -88,6 +89,7 @@ static match_table_t tokens = {
88 {Opt_notreelog, "notreelog"}, 89 {Opt_notreelog, "notreelog"},
89 {Opt_flushoncommit, "flushoncommit"}, 90 {Opt_flushoncommit, "flushoncommit"},
90 {Opt_ratio, "metadata_ratio=%d"}, 91 {Opt_ratio, "metadata_ratio=%d"},
92 {Opt_discard, "discard"},
91 {Opt_err, NULL}, 93 {Opt_err, NULL},
92}; 94};
93 95
@@ -257,6 +259,9 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
257 info->metadata_ratio); 259 info->metadata_ratio);
258 } 260 }
259 break; 261 break;
262 case Opt_discard:
263 btrfs_set_opt(info->mount_opt, DISCARD);
264 break;
260 default: 265 default:
261 break; 266 break;
262 } 267 }
@@ -344,7 +349,7 @@ static int btrfs_fill_super(struct super_block *sb,
344 sb->s_export_op = &btrfs_export_ops; 349 sb->s_export_op = &btrfs_export_ops;
345 sb->s_xattr = btrfs_xattr_handlers; 350 sb->s_xattr = btrfs_xattr_handlers;
346 sb->s_time_gran = 1; 351 sb->s_time_gran = 1;
347#ifdef CONFIG_BTRFS_POSIX_ACL 352#ifdef CONFIG_BTRFS_FS_POSIX_ACL
348 sb->s_flags |= MS_POSIXACL; 353 sb->s_flags |= MS_POSIXACL;
349#endif 354#endif
350 355
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 0b8f36d4400a..bca82a4ca8e6 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -344,10 +344,10 @@ int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
344/* 344/*
345 * when btree blocks are allocated, they have some corresponding bits set for 345 * when btree blocks are allocated, they have some corresponding bits set for
346 * them in one of two extent_io trees. This is used to make sure all of 346 * them in one of two extent_io trees. This is used to make sure all of
347 * those extents are on disk for transaction or log commit 347 * those extents are sent to disk but does not wait on them
348 */ 348 */
349int btrfs_write_and_wait_marked_extents(struct btrfs_root *root, 349int btrfs_write_marked_extents(struct btrfs_root *root,
350 struct extent_io_tree *dirty_pages) 350 struct extent_io_tree *dirty_pages)
351{ 351{
352 int ret; 352 int ret;
353 int err = 0; 353 int err = 0;
@@ -394,6 +394,29 @@ int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
394 page_cache_release(page); 394 page_cache_release(page);
395 } 395 }
396 } 396 }
397 if (err)
398 werr = err;
399 return werr;
400}
401
402/*
403 * when btree blocks are allocated, they have some corresponding bits set for
404 * them in one of two extent_io trees. This is used to make sure all of
405 * those extents are on disk for transaction or log commit. We wait
406 * on all the pages and clear them from the dirty pages state tree
407 */
408int btrfs_wait_marked_extents(struct btrfs_root *root,
409 struct extent_io_tree *dirty_pages)
410{
411 int ret;
412 int err = 0;
413 int werr = 0;
414 struct page *page;
415 struct inode *btree_inode = root->fs_info->btree_inode;
416 u64 start = 0;
417 u64 end;
418 unsigned long index;
419
397 while (1) { 420 while (1) {
398 ret = find_first_extent_bit(dirty_pages, 0, &start, &end, 421 ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
399 EXTENT_DIRTY); 422 EXTENT_DIRTY);
@@ -424,6 +447,22 @@ int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
424 return werr; 447 return werr;
425} 448}
426 449
450/*
451 * when btree blocks are allocated, they have some corresponding bits set for
452 * them in one of two extent_io trees. This is used to make sure all of
453 * those extents are on disk for transaction or log commit
454 */
455int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
456 struct extent_io_tree *dirty_pages)
457{
458 int ret;
459 int ret2;
460
461 ret = btrfs_write_marked_extents(root, dirty_pages);
462 ret2 = btrfs_wait_marked_extents(root, dirty_pages);
463 return ret || ret2;
464}
465
427int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, 466int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
428 struct btrfs_root *root) 467 struct btrfs_root *root)
429{ 468{
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 663c67404918..d4e3e7a6938c 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -79,6 +79,7 @@ static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans,
79 struct inode *inode) 79 struct inode *inode)
80{ 80{
81 BTRFS_I(inode)->last_trans = trans->transaction->transid; 81 BTRFS_I(inode)->last_trans = trans->transaction->transid;
82 BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;
82} 83}
83 84
84int btrfs_end_transaction(struct btrfs_trans_handle *trans, 85int btrfs_end_transaction(struct btrfs_trans_handle *trans,
@@ -107,5 +108,9 @@ int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
107 struct btrfs_root *root); 108 struct btrfs_root *root);
108int btrfs_write_and_wait_marked_extents(struct btrfs_root *root, 109int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
109 struct extent_io_tree *dirty_pages); 110 struct extent_io_tree *dirty_pages);
111int btrfs_write_marked_extents(struct btrfs_root *root,
112 struct extent_io_tree *dirty_pages);
113int btrfs_wait_marked_extents(struct btrfs_root *root,
114 struct extent_io_tree *dirty_pages);
110int btrfs_transaction_in_commit(struct btrfs_fs_info *info); 115int btrfs_transaction_in_commit(struct btrfs_fs_info *info);
111#endif 116#endif
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 4edfdc2acc5f..741666a7676a 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -1980,6 +1980,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
1980 int ret; 1980 int ret;
1981 struct btrfs_root *log = root->log_root; 1981 struct btrfs_root *log = root->log_root;
1982 struct btrfs_root *log_root_tree = root->fs_info->log_root_tree; 1982 struct btrfs_root *log_root_tree = root->fs_info->log_root_tree;
1983 u64 log_transid = 0;
1983 1984
1984 mutex_lock(&root->log_mutex); 1985 mutex_lock(&root->log_mutex);
1985 index1 = root->log_transid % 2; 1986 index1 = root->log_transid % 2;
@@ -1994,12 +1995,13 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
1994 if (atomic_read(&root->log_commit[(index1 + 1) % 2])) 1995 if (atomic_read(&root->log_commit[(index1 + 1) % 2]))
1995 wait_log_commit(trans, root, root->log_transid - 1); 1996 wait_log_commit(trans, root, root->log_transid - 1);
1996 1997
1997 while (root->log_multiple_pids) { 1998 while (1) {
1998 unsigned long batch = root->log_batch; 1999 unsigned long batch = root->log_batch;
1999 mutex_unlock(&root->log_mutex); 2000 if (root->log_multiple_pids) {
2000 schedule_timeout_uninterruptible(1); 2001 mutex_unlock(&root->log_mutex);
2001 mutex_lock(&root->log_mutex); 2002 schedule_timeout_uninterruptible(1);
2002 2003 mutex_lock(&root->log_mutex);
2004 }
2003 wait_for_writer(trans, root); 2005 wait_for_writer(trans, root);
2004 if (batch == root->log_batch) 2006 if (batch == root->log_batch)
2005 break; 2007 break;
@@ -2012,12 +2014,16 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
2012 goto out; 2014 goto out;
2013 } 2015 }
2014 2016
2015 ret = btrfs_write_and_wait_marked_extents(log, &log->dirty_log_pages); 2017 /* we start IO on all the marked extents here, but we don't actually
2018 * wait for them until later.
2019 */
2020 ret = btrfs_write_marked_extents(log, &log->dirty_log_pages);
2016 BUG_ON(ret); 2021 BUG_ON(ret);
2017 2022
2018 btrfs_set_root_node(&log->root_item, log->node); 2023 btrfs_set_root_node(&log->root_item, log->node);
2019 2024
2020 root->log_batch = 0; 2025 root->log_batch = 0;
2026 log_transid = root->log_transid;
2021 root->log_transid++; 2027 root->log_transid++;
2022 log->log_transid = root->log_transid; 2028 log->log_transid = root->log_transid;
2023 root->log_start_pid = 0; 2029 root->log_start_pid = 0;
@@ -2046,6 +2052,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
2046 2052
2047 index2 = log_root_tree->log_transid % 2; 2053 index2 = log_root_tree->log_transid % 2;
2048 if (atomic_read(&log_root_tree->log_commit[index2])) { 2054 if (atomic_read(&log_root_tree->log_commit[index2])) {
2055 btrfs_wait_marked_extents(log, &log->dirty_log_pages);
2049 wait_log_commit(trans, log_root_tree, 2056 wait_log_commit(trans, log_root_tree,
2050 log_root_tree->log_transid); 2057 log_root_tree->log_transid);
2051 mutex_unlock(&log_root_tree->log_mutex); 2058 mutex_unlock(&log_root_tree->log_mutex);
@@ -2065,6 +2072,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
2065 * check the full commit flag again 2072 * check the full commit flag again
2066 */ 2073 */
2067 if (root->fs_info->last_trans_log_full_commit == trans->transid) { 2074 if (root->fs_info->last_trans_log_full_commit == trans->transid) {
2075 btrfs_wait_marked_extents(log, &log->dirty_log_pages);
2068 mutex_unlock(&log_root_tree->log_mutex); 2076 mutex_unlock(&log_root_tree->log_mutex);
2069 ret = -EAGAIN; 2077 ret = -EAGAIN;
2070 goto out_wake_log_root; 2078 goto out_wake_log_root;
@@ -2073,6 +2081,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
2073 ret = btrfs_write_and_wait_marked_extents(log_root_tree, 2081 ret = btrfs_write_and_wait_marked_extents(log_root_tree,
2074 &log_root_tree->dirty_log_pages); 2082 &log_root_tree->dirty_log_pages);
2075 BUG_ON(ret); 2083 BUG_ON(ret);
2084 btrfs_wait_marked_extents(log, &log->dirty_log_pages);
2076 2085
2077 btrfs_set_super_log_root(&root->fs_info->super_for_commit, 2086 btrfs_set_super_log_root(&root->fs_info->super_for_commit,
2078 log_root_tree->node->start); 2087 log_root_tree->node->start);
@@ -2092,9 +2101,14 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
2092 * the running transaction open, so a full commit can't hop 2101 * the running transaction open, so a full commit can't hop
2093 * in and cause problems either. 2102 * in and cause problems either.
2094 */ 2103 */
2095 write_ctree_super(trans, root->fs_info->tree_root, 2); 2104 write_ctree_super(trans, root->fs_info->tree_root, 1);
2096 ret = 0; 2105 ret = 0;
2097 2106
2107 mutex_lock(&root->log_mutex);
2108 if (root->last_log_commit < log_transid)
2109 root->last_log_commit = log_transid;
2110 mutex_unlock(&root->log_mutex);
2111
2098out_wake_log_root: 2112out_wake_log_root:
2099 atomic_set(&log_root_tree->log_commit[index2], 0); 2113 atomic_set(&log_root_tree->log_commit[index2], 0);
2100 smp_mb(); 2114 smp_mb();
@@ -2862,6 +2876,21 @@ out:
2862 return ret; 2876 return ret;
2863} 2877}
2864 2878
2879static int inode_in_log(struct btrfs_trans_handle *trans,
2880 struct inode *inode)
2881{
2882 struct btrfs_root *root = BTRFS_I(inode)->root;
2883 int ret = 0;
2884
2885 mutex_lock(&root->log_mutex);
2886 if (BTRFS_I(inode)->logged_trans == trans->transid &&
2887 BTRFS_I(inode)->last_sub_trans <= root->last_log_commit)
2888 ret = 1;
2889 mutex_unlock(&root->log_mutex);
2890 return ret;
2891}
2892
2893
2865/* 2894/*
2866 * helper function around btrfs_log_inode to make sure newly created 2895 * helper function around btrfs_log_inode to make sure newly created
2867 * parent directories also end up in the log. A minimal inode and backref 2896 * parent directories also end up in the log. A minimal inode and backref
@@ -2901,6 +2930,11 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
2901 if (ret) 2930 if (ret)
2902 goto end_no_trans; 2931 goto end_no_trans;
2903 2932
2933 if (inode_in_log(trans, inode)) {
2934 ret = BTRFS_NO_LOG_SYNC;
2935 goto end_no_trans;
2936 }
2937
2904 start_log_trans(trans, root); 2938 start_log_trans(trans, root);
2905 2939
2906 ret = btrfs_log_inode(trans, root, inode, inode_only); 2940 ret = btrfs_log_inode(trans, root, inode, inode_only);
diff --git a/fs/btrfs/tree-log.h b/fs/btrfs/tree-log.h
index d09c7609e16b..0776eacb5083 100644
--- a/fs/btrfs/tree-log.h
+++ b/fs/btrfs/tree-log.h
@@ -19,6 +19,9 @@
19#ifndef __TREE_LOG_ 19#ifndef __TREE_LOG_
20#define __TREE_LOG_ 20#define __TREE_LOG_
21 21
22/* return value for btrfs_log_dentry_safe that means we don't need to log it at all */
23#define BTRFS_NO_LOG_SYNC 256
24
22int btrfs_sync_log(struct btrfs_trans_handle *trans, 25int btrfs_sync_log(struct btrfs_trans_handle *trans,
23 struct btrfs_root *root); 26 struct btrfs_root *root);
24int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root); 27int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root);
diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c
index b0fc93f95fd0..b6dd5967c48a 100644
--- a/fs/btrfs/xattr.c
+++ b/fs/btrfs/xattr.c
@@ -260,7 +260,7 @@ err:
260 * attributes are handled directly. 260 * attributes are handled directly.
261 */ 261 */
262struct xattr_handler *btrfs_xattr_handlers[] = { 262struct xattr_handler *btrfs_xattr_handlers[] = {
263#ifdef CONFIG_BTRFS_POSIX_ACL 263#ifdef CONFIG_BTRFS_FS_POSIX_ACL
264 &btrfs_xattr_acl_access_handler, 264 &btrfs_xattr_acl_access_handler,
265 &btrfs_xattr_acl_default_handler, 265 &btrfs_xattr_acl_default_handler,
266#endif 266#endif