From 8eb2d829ffea3677c21bd038f19e5d8ca6b43e36 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 9 Nov 2010 14:48:01 +0800 Subject: btrfs: Fix threshold calculation for block groups smaller than 1GB If a block group is smaller than 1GB, the extent entry threadhold calculation will always set the threshold to 0. So as free space gets fragmented, btrfs will switch to use bitmap to manage free space, but then will never switch back to extents due to this bug. Reviewed-by: Josef Bacik Signed-off-by: Li Zefan --- fs/btrfs/free-space-cache.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'fs/btrfs') diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 60d6842..42f4015 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -1016,14 +1016,18 @@ static void recalculate_thresholds(struct btrfs_block_group_cache *block_group) u64 max_bytes; u64 bitmap_bytes; u64 extent_bytes; + u64 size = block_group->key.offset; /* * The goal is to keep the total amount of memory used per 1gb of space * at or below 32k, so we need to adjust how much memory we allow to be * used by extent based free space tracking */ - max_bytes = MAX_CACHE_BYTES_PER_GIG * - (div64_u64(block_group->key.offset, 1024 * 1024 * 1024)); + if (size < 1024 * 1024 * 1024) + max_bytes = MAX_CACHE_BYTES_PER_GIG; + else + max_bytes = MAX_CACHE_BYTES_PER_GIG * + div64_u64(size, 1024 * 1024 * 1024); /* * we want to account for 1 more bitmap than what we have so we can make -- cgit v1.1 From edf6e2d1ddbac7f326b34a27adbca71ece53ccce Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 9 Nov 2010 14:50:07 +0800 Subject: btrfs: Add helper function free_bitmap() Remove some duplicated code. This prepares for the next patch. Reviewed-by: Josef Bacik Signed-off-by: Li Zefan --- fs/btrfs/free-space-cache.c | 37 ++++++++++++++++--------------------- 1 file changed, 16 insertions(+), 21 deletions(-) (limited to 'fs/btrfs') diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 42f4015..850104f 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -1175,6 +1175,16 @@ static void add_new_bitmap(struct btrfs_block_group_cache *block_group, recalculate_thresholds(block_group); } +static void free_bitmap(struct btrfs_block_group_cache *block_group, + struct btrfs_free_space *bitmap_info) +{ + unlink_free_space(block_group, bitmap_info); + kfree(bitmap_info->bitmap); + kfree(bitmap_info); + block_group->total_bitmaps--; + recalculate_thresholds(block_group); +} + static noinline int remove_from_bitmap(struct btrfs_block_group_cache *block_group, struct btrfs_free_space *bitmap_info, u64 *offset, u64 *bytes) @@ -1215,13 +1225,8 @@ again: if (*bytes) { struct rb_node *next = rb_next(&bitmap_info->offset_index); - if (!bitmap_info->bytes) { - unlink_free_space(block_group, bitmap_info); - kfree(bitmap_info->bitmap); - kfree(bitmap_info); - block_group->total_bitmaps--; - recalculate_thresholds(block_group); - } + if (!bitmap_info->bytes) + free_bitmap(block_group, bitmap_info); /* * no entry after this bitmap, but we still have bytes to @@ -1254,13 +1259,8 @@ again: return -EAGAIN; goto again; - } else if (!bitmap_info->bytes) { - unlink_free_space(block_group, bitmap_info); - kfree(bitmap_info->bitmap); - kfree(bitmap_info); - block_group->total_bitmaps--; - recalculate_thresholds(block_group); - } + } else if (!bitmap_info->bytes) + free_bitmap(block_group, bitmap_info); return 0; } @@ -1689,13 +1689,8 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group, ret = offset; if (entry->bitmap) { bitmap_clear_bits(block_group, entry, offset, bytes); - if (!entry->bytes) { - unlink_free_space(block_group, entry); - kfree(entry->bitmap); - kfree(entry); - block_group->total_bitmaps--; - recalculate_thresholds(block_group); - } + if (!entry->bytes) + free_bitmap(block_group, entry); } else { unlink_free_space(block_group, entry); entry->offset += bytes; -- cgit v1.1 From 70b7da304f9f9bbf1566085155895e32e775a745 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 9 Nov 2010 14:51:45 +0800 Subject: btrfs: Free fully occupied bitmap in cluster If there's no more free space in a bitmap, we should free it. Reviewed-by: Josef Bacik Signed-off-by: Li Zefan --- fs/btrfs/free-space-cache.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs') diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 850104f..cb0137e 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -1788,6 +1788,8 @@ static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group, ret = search_start; bitmap_clear_bits(block_group, entry, ret, bytes); + if (entry->bytes == 0) + free_bitmap(block_group, entry); out: spin_unlock(&cluster->lock); spin_unlock(&block_group->tree_lock); -- cgit v1.1 From 5e71b5d5ec07e4b3fb4c78c4e4b108ff667f123f Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 9 Nov 2010 14:55:34 +0800 Subject: btrfs: Update stats when allocating from a cluster When allocating extent entry from a cluster, we should update the free_space and free_extents fields of the block group. Reviewed-by: Josef Bacik Signed-off-by: Li Zefan --- fs/btrfs/free-space-cache.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) (limited to 'fs/btrfs') diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index cb0137e..2974c47 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -1843,15 +1843,26 @@ u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group, entry->offset += bytes; entry->bytes -= bytes; - if (entry->bytes == 0) { + if (entry->bytes == 0) rb_erase(&entry->offset_index, &cluster->root); - kfree(entry); - } break; } out: spin_unlock(&cluster->lock); + if (!ret) + return 0; + + spin_lock(&block_group->tree_lock); + + block_group->free_space -= bytes; + if (entry->bytes == 0) { + block_group->free_extents--; + kfree(entry); + } + + spin_unlock(&block_group->tree_lock); + return ret; } -- cgit v1.1 From 120d66eec0dcb966fbd03f743598b2ff2513436b Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 9 Nov 2010 14:56:50 +0800 Subject: btrfs: Add a helper try_merge_free_space() When adding a new extent, we'll firstly see if we can merge this extent to the left or/and right extent. Extract this as a helper try_merge_free_space(). As a side effect, we fix a small bug that if the new extent has non-bitmap left entry but is unmergeble, we'll directly link the extent without trying to drop it into bitmap. This also prepares for the next patch. Reviewed-by: Josef Bacik Signed-off-by: Li Zefan --- fs/btrfs/free-space-cache.c | 75 ++++++++++++++++++++++++++------------------- 1 file changed, 43 insertions(+), 32 deletions(-) (limited to 'fs/btrfs') diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 2974c47..cf67dc3 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -1363,22 +1363,14 @@ out: return ret; } -int btrfs_add_free_space(struct btrfs_block_group_cache *block_group, - u64 offset, u64 bytes) +bool try_merge_free_space(struct btrfs_block_group_cache *block_group, + struct btrfs_free_space *info) { - struct btrfs_free_space *right_info = NULL; - struct btrfs_free_space *left_info = NULL; - struct btrfs_free_space *info = NULL; - int ret = 0; - - 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); + struct btrfs_free_space *left_info; + struct btrfs_free_space *right_info; + bool merged = false; + u64 offset = info->offset; + u64 bytes = info->bytes; /* * first we want to see if there is free space adjacent to the range we @@ -1392,27 +1384,11 @@ int btrfs_add_free_space(struct btrfs_block_group_cache *block_group, else left_info = tree_search_offset(block_group, offset - 1, 0, 0); - /* - * If there was no extent directly to the left or right of this new - * extent then we know we're going to have to allocate a new extent, so - * before we do that see if we need to drop this into a bitmap - */ - if ((!left_info || left_info->bitmap) && - (!right_info || right_info->bitmap)) { - ret = insert_into_bitmap(block_group, info); - - if (ret < 0) { - goto out; - } else if (ret) { - ret = 0; - goto out; - } - } - if (right_info && !right_info->bitmap) { unlink_free_space(block_group, right_info); info->bytes += right_info->bytes; kfree(right_info); + merged = true; } if (left_info && !left_info->bitmap && @@ -1421,8 +1397,43 @@ int btrfs_add_free_space(struct btrfs_block_group_cache *block_group, info->offset = left_info->offset; info->bytes += left_info->bytes; kfree(left_info); + merged = true; } + return merged; +} + +int btrfs_add_free_space(struct btrfs_block_group_cache *block_group, + u64 offset, u64 bytes) +{ + struct btrfs_free_space *info; + int ret = 0; + + 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); + + if (try_merge_free_space(block_group, info)) + goto link; + + /* + * There was no extent directly to the left or right of this new + * extent then we know we're going to have to allocate a new extent, so + * before we do that see if we need to drop this into a bitmap + */ + ret = insert_into_bitmap(block_group, info); + if (ret < 0) { + goto out; + } else if (ret) { + ret = 0; + goto out; + } +link: ret = link_free_space(block_group, info); if (ret) kfree(info); -- cgit v1.1 From f333adb5d64bc1c4d6099072fc341c3c8f84e0cf Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Tue, 9 Nov 2010 14:57:39 +0800 Subject: btrfs: Check mergeable free space when removing a cluster After returing extents from a cluster to the block group, some extents in the block group may be mergeable. Reviewed-by: Josef Bacik Signed-off-by: Li Zefan --- fs/btrfs/free-space-cache.c | 26 ++++++++++++++++++++------ 1 file changed, 20 insertions(+), 6 deletions(-) (limited to 'fs/btrfs') diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index cf67dc3..a5501ed 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -987,11 +987,18 @@ tree_search_offset(struct btrfs_block_group_cache *block_group, return entry; } -static void unlink_free_space(struct btrfs_block_group_cache *block_group, - struct btrfs_free_space *info) +static inline void +__unlink_free_space(struct btrfs_block_group_cache *block_group, + struct btrfs_free_space *info) { rb_erase(&info->offset_index, &block_group->free_space_offset); block_group->free_extents--; +} + +static void unlink_free_space(struct btrfs_block_group_cache *block_group, + struct btrfs_free_space *info) +{ + __unlink_free_space(block_group, info); block_group->free_space -= info->bytes; } @@ -1364,7 +1371,7 @@ out: } bool try_merge_free_space(struct btrfs_block_group_cache *block_group, - struct btrfs_free_space *info) + struct btrfs_free_space *info, bool update_stat) { struct btrfs_free_space *left_info; struct btrfs_free_space *right_info; @@ -1385,7 +1392,10 @@ bool try_merge_free_space(struct btrfs_block_group_cache *block_group, left_info = tree_search_offset(block_group, offset - 1, 0, 0); if (right_info && !right_info->bitmap) { - unlink_free_space(block_group, right_info); + if (update_stat) + unlink_free_space(block_group, right_info); + else + __unlink_free_space(block_group, right_info); info->bytes += right_info->bytes; kfree(right_info); merged = true; @@ -1393,7 +1403,10 @@ bool try_merge_free_space(struct btrfs_block_group_cache *block_group, if (left_info && !left_info->bitmap && left_info->offset + left_info->bytes == offset) { - unlink_free_space(block_group, left_info); + if (update_stat) + unlink_free_space(block_group, left_info); + else + __unlink_free_space(block_group, left_info); info->offset = left_info->offset; info->bytes += left_info->bytes; kfree(left_info); @@ -1418,7 +1431,7 @@ int btrfs_add_free_space(struct btrfs_block_group_cache *block_group, spin_lock(&block_group->tree_lock); - if (try_merge_free_space(block_group, info)) + if (try_merge_free_space(block_group, info, true)) goto link; /* @@ -1636,6 +1649,7 @@ __btrfs_return_cluster_to_free_space( node = rb_next(&entry->offset_index); rb_erase(&entry->offset_index, &cluster->root); BUG_ON(entry->bitmap); + try_merge_free_space(block_group, entry, false); tree_insert_offset(&block_group->free_space_offset, entry->offset, &entry->offset_index, 0); } -- cgit v1.1 From 83a4d54840c88a4a45c49670f044b8c7ddeaa8c7 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Mon, 27 Dec 2010 16:19:53 +0800 Subject: Btrfs: Fix memory leak at umount fs_info, which is allocated in open_ctree(), should be freed in close_ctree(). Signed-off-by: Li Zefan --- fs/btrfs/disk-io.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs') diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index a5d2249..089871e 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -2513,6 +2513,8 @@ int close_ctree(struct btrfs_root *root) kfree(fs_info->chunk_root); kfree(fs_info->dev_root); kfree(fs_info->csum_root); + kfree(fs_info); + return 0; } -- cgit v1.1 From bdc924bb4cdac92b945945c3149ab8191c92d75d Mon Sep 17 00:00:00 2001 From: Ian Kent Date: Mon, 27 Dec 2010 16:33:15 +0800 Subject: Btrfs: Fix memory leak on finding existing super We missed a memory deallocation in commit 450ba0ea. If an existing super block is found at mount and there is no error condition then the pre-allocated tree_root and fs_info are no not used and are not freeded. Signed-off-by: Ian Kent Signed-off-by: Li Zefan --- fs/btrfs/super.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs/btrfs') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 61bd79a..f50253c 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -654,6 +654,8 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags, } btrfs_close_devices(fs_devices); + kfree(fs_info); + kfree(tree_root); } else { char b[BDEVNAME_SIZE]; -- cgit v1.1 From 3f3d0bc0df041236fad4ffa82188a6e4ef9af75e Mon Sep 17 00:00:00 2001 From: Tero Roponen Date: Mon, 27 Dec 2010 16:43:13 +0800 Subject: Btrfs: Free correct pointer after using strsep We must save and free the original kstrdup()'ed pointer because strsep() modifies its first argument. Signed-off-by: Tero Roponen Signed-off-by: Li Zefan --- fs/btrfs/super.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'fs/btrfs') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index f50253c..78ee681 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -277,7 +277,7 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags, struct btrfs_fs_devices **fs_devices) { substring_t args[MAX_OPT_ARGS]; - char *opts, *p; + char *opts, *orig, *p; int error = 0; int intarg; @@ -291,6 +291,7 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags, opts = kstrdup(options, GFP_KERNEL); if (!opts) return -ENOMEM; + orig = opts; while ((p = strsep(&opts, ",")) != NULL) { int token; @@ -326,7 +327,7 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags, } out_free_opts: - kfree(opts); + kfree(orig); out: /* * If no subvolume name is specified we use the default one. Allocate -- cgit v1.1 From d0f69686c2ae775529aadc7a8acc6f13ad41de66 Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Tue, 25 Jan 2011 15:46:17 +0800 Subject: Btrfs: Don't return acl info when mounting with noacl option Steps to reproduce: # mkfs.btrfs /dev/sda2 # mount /dev/sda2 /mnt # touch /mnt/file0 # setfacl -m 'u:root:x,g::x,o::x' /mnt/file0 # umount /mnt # mount /dev/sda2 -o noacl /mnt # getfacl /mnt/file0 ... user::rw- user:root:--x group::--x mask::--x other::--x The output should be: user::rw- group::--x other::--x Signed-off-by: Miao Xie Signed-off-by: Li Zefan --- fs/btrfs/acl.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'fs/btrfs') diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c index 2222d16..3c52fc8 100644 --- a/fs/btrfs/acl.c +++ b/fs/btrfs/acl.c @@ -37,6 +37,9 @@ static struct posix_acl *btrfs_get_acl(struct inode *inode, int type) char *value = NULL; struct posix_acl *acl; + if (!IS_POSIXACL(inode)) + return NULL; + acl = get_cached_acl(inode, type); if (acl != ACL_NOT_CACHED) return acl; @@ -82,6 +85,9 @@ static int btrfs_xattr_acl_get(struct dentry *dentry, const char *name, struct posix_acl *acl; int ret = 0; + if (!IS_POSIXACL(dentry->d_inode)) + return -EOPNOTSUPP; + acl = btrfs_get_acl(dentry->d_inode, type); if (IS_ERR(acl)) -- cgit v1.1 From b897abec032deb7cc3ce67392a1f544ac965ddea Mon Sep 17 00:00:00 2001 From: Miao Xie Date: Wed, 26 Jan 2011 16:19:22 +0800 Subject: Btrfs: Fix memory leak in writepage fixup work fixup, which is allocated when starting page write to fix up the extent without ORDERED bit set, should be freed after this work is done. Signed-off-by: Miao Xie Signed-off-by: Li Zefan --- fs/btrfs/inode.c | 1 + 1 file changed, 1 insertion(+) (limited to 'fs/btrfs') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5f91944..3a6edc4 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -1544,6 +1544,7 @@ out: out_page: unlock_page(page); page_cache_release(page); + kfree(fixup); } /* -- cgit v1.1 From 4d728ec7aefdca5419d2ebfb28c147e81a4b59f4 Mon Sep 17 00:00:00 2001 From: Li Zefan Date: Wed, 26 Jan 2011 14:10:43 +0800 Subject: Btrfs: Fix file clone when source offset is not 0 Suppose: - the source extent is: [0, 100] - the src offset is 10 - the clone length is 90 - the dest offset is 0 This statement: new_key.offset = key.offset + destoff - off will produce such an extent for the dest file: [ino, BTRFS_EXTENT_DATA_KEY, -10] , which is obviously wrong. Signed-off-by: Li Zefan --- fs/btrfs/ioctl.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'fs/btrfs') diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index f87552a..1b61dab 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -1788,7 +1788,10 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, memcpy(&new_key, &key, sizeof(new_key)); new_key.objectid = inode->i_ino; - new_key.offset = key.offset + destoff - off; + if (off <= key.offset) + new_key.offset = key.offset + destoff - off; + else + new_key.offset = destoff; trans = btrfs_start_transaction(root, 1); if (IS_ERR(trans)) { -- cgit v1.1