Li Zefan
2011-Jun-21 05:49 UTC
[PATCH 1/2] Btrfs: fix space leak when skipping small extents during trimming
We''re taking a free space extent out of the free space cache, trimming
it and then putting it back into the cache.
However for an extent that is smaller than the specified minimum length,
it''s taken out but won''t be put back, which causes space leak.
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
---
Unfortunately I have no trim-able device to test the patch.
---
fs/btrfs/free-space-cache.c | 34 +++++++++++++++++-----------------
1 files changed, 17 insertions(+), 17 deletions(-)
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 9f985a4..292c0d9 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -2460,6 +2460,7 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache
*block_group,
u64 bytes = 0;
u64 actually_trimmed;
int ret = 0;
+ int update_ret;
*trimmed = 0;
@@ -2483,6 +2484,7 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache
*block_group,
}
if (entry->bitmap) {
+ bytes = 0;
ret = search_bitmap(ctl, entry, &start, &bytes);
if (!ret) {
if (start >= end) {
@@ -2490,6 +2492,8 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache
*block_group,
break;
}
bytes = min(bytes, end - start);
+ if (bytes < minlen)
+ goto next;
bitmap_clear_bits(ctl, entry, start, bytes);
if (entry->bytes == 0)
free_bitmap(ctl, entry);
@@ -2503,33 +2507,29 @@ int btrfs_trim_block_group(struct
btrfs_block_group_cache *block_group,
} else {
start = entry->offset;
bytes = min(entry->bytes, end - start);
+ if (bytes < minlen)
+ goto next;
unlink_free_space(ctl, entry);
kmem_cache_free(btrfs_free_space_cachep, entry);
}
spin_unlock(&ctl->tree_lock);
- if (bytes >= minlen) {
- int update_ret;
- update_ret = btrfs_update_reserved_bytes(block_group,
- bytes, 1, 1);
+ update_ret = btrfs_update_reserved_bytes(block_group,
+ bytes, 1, 1);
- ret = btrfs_error_discard_extent(fs_info->extent_root,
- start,
- bytes,
- &actually_trimmed);
+ ret = btrfs_error_discard_extent(fs_info->extent_root, start,
+ bytes, &actually_trimmed);
- btrfs_add_free_space(block_group, start, bytes);
- if (!update_ret)
- btrfs_update_reserved_bytes(block_group,
- bytes, 0, 1);
+ btrfs_add_free_space(block_group, start, bytes);
+ if (!update_ret)
+ btrfs_update_reserved_bytes(block_group, bytes, 0, 1);
- if (ret)
- break;
- *trimmed += actually_trimmed;
- }
+ if (ret)
+ break;
+ *trimmed += actually_trimmed;
+next:
start += bytes;
- bytes = 0;
if (fatal_signal_pending(current)) {
ret = -ERESTARTSYS;
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Li Zefan
2011-Jun-21 05:50 UTC
[PATCH 2/2] Btrfs: fix space leak when trimming free extents
When the end of an extent exceeds the end of the specified range,
the extent will be accidentally truncated.
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
---
fs/btrfs/free-space-cache.c | 9 ++++++++-
1 files changed, 8 insertions(+), 1 deletions(-)
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 292c0d9..185cf8e 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -2509,8 +2509,15 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache
*block_group,
bytes = min(entry->bytes, end - start);
if (bytes < minlen)
goto next;
+
unlink_free_space(ctl, entry);
- kmem_cache_free(btrfs_free_space_cachep, entry);
+ if (bytes < entry->bytes) {
+ entry->offset = entry->offset + bytes;
+ entry->bytes = entry->bytes - bytes;
+ link_free_space(ctl, entry);
+ } else {
+ kmem_cache_free(btrfs_free_space_cachep, entry);
+ }
}
spin_unlock(&ctl->tree_lock);
-- 1.7.3.1
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Li Dongyang
2011-Jun-21 09:45 UTC
Re: [PATCH 2/2] Btrfs: fix space leak when trimming free extents
On Tuesday, June 21, 2011 01:50:10 PM Li Zefan wrote:> When the end of an extent exceeds the end of the specified range, > the extent will be accidentally truncated. > > Signed-off-by: Li Zefan <lizf@cn.fujitsu.com> > --- > fs/btrfs/free-space-cache.c | 9 ++++++++- > 1 files changed, 8 insertions(+), 1 deletions(-) > > diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c > index 292c0d9..185cf8e 100644 > --- a/fs/btrfs/free-space-cache.c > +++ b/fs/btrfs/free-space-cache.c > @@ -2509,8 +2509,15 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group, > bytes = min(entry->bytes, end - start); > if (bytes < minlen) > goto next; > + > unlink_free_space(ctl, entry); > - kmem_cache_free(btrfs_free_space_cachep, entry); > + if (bytes < entry->bytes) { > + entry->offset = entry->offset + bytes; > + entry->bytes = entry->bytes - bytes; > + link_free_space(ctl, entry);yes, I forgot to link the rest extent to the free space cache. Thanks for the fix!> + } else { > + kmem_cache_free(btrfs_free_space_cachep, entry); > + } > } > > spin_unlock(&ctl->tree_lock); > -- 1.7.3.1 >-- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Li Dongyang
2011-Jul-05 04:07 UTC
Re: [PATCH 1/2] Btrfs: fix space leak when skipping small extents during trimming
On Tuesday, June 21, 2011 01:49:19 PM Li Zefan wrote:> We''re taking a free space extent out of the free space cache, trimming > it and then putting it back into the cache. >sorry for the late reply, I can hardly to find time look at this.> However for an extent that is smaller than the specified minimum length, > it''s taken out but won''t be put back, which causes space leak.yes, you are correct, and the fix looks good to me, Thanks> > Signed-off-by: Li Zefan <lizf@cn.fujitsu.com> > --- > > Unfortunately I have no trim-able device to test the patch. > > --- > fs/btrfs/free-space-cache.c | 34 +++++++++++++++++----------------- > 1 files changed, 17 insertions(+), 17 deletions(-) > > diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c > index 9f985a4..292c0d9 100644 > --- a/fs/btrfs/free-space-cache.c > +++ b/fs/btrfs/free-space-cache.c > @@ -2460,6 +2460,7 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group, > u64 bytes = 0; > u64 actually_trimmed; > int ret = 0; > + int update_ret; > > *trimmed = 0; > > @@ -2483,6 +2484,7 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group, > } > > if (entry->bitmap) { > + bytes = 0; > ret = search_bitmap(ctl, entry, &start, &bytes); > if (!ret) { > if (start >= end) { > @@ -2490,6 +2492,8 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group, > break; > } > bytes = min(bytes, end - start); > + if (bytes < minlen) > + goto next; > bitmap_clear_bits(ctl, entry, start, bytes); > if (entry->bytes == 0) > free_bitmap(ctl, entry); > @@ -2503,33 +2507,29 @@ int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group, > } else { > start = entry->offset; > bytes = min(entry->bytes, end - start); > + if (bytes < minlen) > + goto next; > unlink_free_space(ctl, entry); > kmem_cache_free(btrfs_free_space_cachep, entry); > } > > spin_unlock(&ctl->tree_lock); > > - if (bytes >= minlen) { > - int update_ret; > - update_ret = btrfs_update_reserved_bytes(block_group, > - bytes, 1, 1); > + update_ret = btrfs_update_reserved_bytes(block_group, > + bytes, 1, 1); > > - ret = btrfs_error_discard_extent(fs_info->extent_root, > - start, > - bytes, > - &actually_trimmed); > + ret = btrfs_error_discard_extent(fs_info->extent_root, start, > + bytes, &actually_trimmed); > > - btrfs_add_free_space(block_group, start, bytes); > - if (!update_ret) > - btrfs_update_reserved_bytes(block_group, > - bytes, 0, 1); > + btrfs_add_free_space(block_group, start, bytes); > + if (!update_ret) > + btrfs_update_reserved_bytes(block_group, bytes, 0, 1); > > - if (ret) > - break; > - *trimmed += actually_trimmed; > - } > + if (ret) > + break; > + *trimmed += actually_trimmed; > +next: > start += bytes; > - bytes = 0; > > if (fatal_signal_pending(current)) { > ret = -ERESTARTSYS; >-- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html