2011-03-04 17:13:55

by Al Viro

Subject: [RFC] a couple of i_nlink fixes in btrfs

a) rename() plays with i_nlink of old_inode; bad, since it's not
locked. I'd add a variant of btrfs_unlink_inode() that would leave
btrfs_drop_nlink()/btrfs_update_inode() to callers and use it instead.
b) btrfs_link() doesn't check for i_nlink overflows. I don't
know if there's anything preventing that many links to a file on btrfs,
but if there is, it's at least worth a comment in there...

Please, review; patches in followups or in #btrfs in vfs-2.6.git

2011-03-04 17:14:51

by Al Viro

Subject: [PATCH 1/2] btrfs: don't mess with i_nlink of unlocked inode in rename()

old_inode is not locked; it's not safe to play with its link
count. Instead of bumping it and calling btrfs_unlink_inode(),
add a variant of the latter that does not do btrfs_drop_nlink()/
btrfs_update_inode(), call it instead of btrfs_inc_nlink()/
btrfs_unlink_inode() and do btrfs_update_inode() ourselves.

Signed-off-by: Al Viro <[email protected]>
---
fs/btrfs/inode.c | 36 +++++++++++++++++++++++++-----------
1 files changed, 25 insertions(+), 11 deletions(-)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 0efdb65..099d64c 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -2634,10 +2634,10 @@ failed:
* recovery code. It remove a link in a directory with a given name, and
* also drops the back refs in the inode to the directory
*/
-int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct inode *dir, struct inode *inode,
- const char *name, int name_len)
+static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct inode *dir, struct inode *inode,
+ const char *name, int name_len)
{
struct btrfs_path *path;
int ret = 0;
@@ -2709,12 +2709,25 @@ err:
btrfs_i_size_write(dir, dir->i_size - name_len * 2);
inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME;
btrfs_update_inode(trans, root, dir);
- btrfs_drop_nlink(inode);
- ret = btrfs_update_inode(trans, root, inode);
out:
return ret;
}

+int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct inode *dir, struct inode *inode,
+ const char *name, int name_len)
+{
+ int ret;
+ ret = __btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
+ if (!ret) {
+ btrfs_drop_nlink(inode);
+ ret = btrfs_update_inode(trans, root, inode);
+ }
+ return ret;
+}
+
+
/* helper to check if there is any shared block in the path */
static int check_path_shared(struct btrfs_root *root,
struct btrfs_path *path)
@@ -6908,11 +6921,12 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
old_dentry->d_name.name,
old_dentry->d_name.len);
} else {
- btrfs_inc_nlink(old_dentry->d_inode);
- ret = btrfs_unlink_inode(trans, root, old_dir,
- old_dentry->d_inode,
- old_dentry->d_name.name,
- old_dentry->d_name.len);
+ ret = __btrfs_unlink_inode(trans, root, old_dir,
+ old_dentry->d_inode,
+ old_dentry->d_name.name,
+ old_dentry->d_name.len);
+ if (!ret)
+ ret = btrfs_update_inode(trans, root, old_inode);
}
BUG_ON(ret);

--
1.7.2.5

2011-03-04 17:15:22

by Al Viro

Subject: [PATCH 2/2] btrfs: check link counter overflow in link(2)

Signed-off-by: Al Viro <[email protected]>
---
fs/btrfs/inode.c | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 099d64c..8a6df03 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4826,6 +4826,9 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
if (root->objectid != BTRFS_I(inode)->root->objectid)
return -EPERM;

+ if (inode->i_nlink == ~0U)
+ return -EMLINK;
+
btrfs_inc_nlink(inode);
inode->i_ctime = CURRENT_TIME;

--
1.7.2.5

2011-03-07 16:59:24

by Chris Mason

Subject: Re: [RFC] a couple of i_nlink fixes in btrfs

Excerpts from Al Viro's message of 2011-03-04 12:13:53 -0500:
> a) rename() plays with i_nlink of old_inode; bad, since it's not
> locked. I'd add a variant of btrfs_unlink_inode() that would leave
> btrfs_drop_nlink()/btrfs_update_inode() to callers and use it instead.
> b) btrfs_link() doesn't check for i_nlink overflows. I don't
> know if there's anything preventing that many links to a file on btrfs,
> but if there is, it's at least worth a comment in there...
>
> Please, review; patches in followups or in #btrfs in vfs-2.6.git

Thanks, these both look good but I'll test here as well. Are you
planning on pushing for .38?

-chris

2011-03-21 05:17:27

by Al Viro

Subject: Re: [RFC] a couple of i_nlink fixes in btrfs

On Mon, Mar 07, 2011 at 11:58:13AM -0500, Chris Mason wrote:

> Thanks, these both look good but I'll test here as well. Are you
> planning on pushing for .38?

No, but .39 would be nice ;-) Do you want that to go through btrfs tree
or through vfs one?

2011-03-21 11:37:53

by Chris Mason

Subject: Re: [RFC] a couple of i_nlink fixes in btrfs

Excerpts from Al Viro's message of 2011-03-21 01:17:25 -0400:
> On Mon, Mar 07, 2011 at 11:58:13AM -0500, Chris Mason wrote:
>
> > Thanks, these both look good but I'll test here as well. Are you
> > planning on pushing for .38?
>
> No, but .39 would be nice ;-) Do you want that to go through btrfs tree
> or through vfs one?

I'll take them in mine, it'll be easier to put in with all these other
changes.

-chris