xfs
[Top] [All Lists]

Re: [PATCH v3] xfs: fix the symbolic link assert in xfs_ifree

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: [PATCH v3] xfs: fix the symbolic link assert in xfs_ifree
From: Mark Tinguely <tinguely@xxxxxxx>
Date: Fri, 14 Jun 2013 08:30:18 -0500
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20130614021720.GR29338@dastard>
References: <20130613170449.144027450@xxxxxxx> <20130614021720.GR29338@dastard>
User-agent: Mozilla/5.0 (X11; FreeBSD amd64; rv:9.0) Gecko/20120122 Thunderbird/9.0
On 06/13/13 21:17, Dave Chinner wrote:
On Thu, Jun 13, 2013 at 12:04:49PM -0500, Mark Tinguely wrote:
Adding an extended attribute to a symbolic link can force that
link to an remote extent. xfs_inactive() incorrectly assumes
that any symbolic link small enough to be in the inode core
is incore, resulting in the remote extent to not be removed.
xfs_ifree() will assert on presence of this leaked remote extent.

Signed-off-by: Mark Tinguely<tinguely@xxxxxxx>
.....
+/*
+ * xfs_inactive_symlink - free a symlink
+ */
+int
+xfs_inactive_symlink(
+       struct xfs_inode        *ip,
+       struct xfs_trans        **tp)
+{
+       struct xfs_mount *mp = ip->i_mount;
+       xfs_fsize_t     pathlen;

int will do here - it can't be more than MAXPATHLEN = 1024, and this
gets rid of the casts in the error print.

+
+       trace_xfs_readlink(ip);

trace_xfs_inactive_symlink, as you mentioned ;)

sigh, and I reverted the inactive and symlink in the trace below. sigh again.


+
+       ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
+
+       if (XFS_FORCED_SHUTDOWN(mp))
+               return XFS_ERROR(EIO);
+
+       /*
+        * Zero length symlinks _can_ exist.
+        */
+       pathlen = ip->i_d.di_size;
+       if (!pathlen)
+               return 0;
+
+       if (pathlen<  0 || pathlen>  MAXPATHLEN) {
+               xfs_alert(mp, "%s: inode (%llu) bad symlink length (%lld)",
+                        __func__, (unsigned long long) ip->i_ino,
                                                        ^ whitespace
+                        (long long) pathlen);

Can we change this to a hex-based inode number (much easier to read)
and the pathlen cast can go away if you use an int. i.e.:

                xfs_alert(mp, "%s: inode (0x%llx) bad symlink length (%d)",
                         __func__, (unsigned long long)ip->i_ino, pathlen);



Both of the items above were stole directly from xfs_readlink().
Thought they should be consistent(ly wrong :) ).


+               ASSERT(0);
+               return XFS_ERROR(EFSCORRUPTED);
+       }
+
+       if (ip->i_df.if_flags&  XFS_IFINLINE) {
+               if (ip->i_df.if_bytes>  0)
+                       xfs_idata_realloc(ip, -(ip->i_df.if_bytes),
+                                         XFS_DATA_FORK);
+               ASSERT(ip->i_df.if_bytes == 0);
+               return 0;
+       }
+
+       /* remove the remote symlink */
+       return(xfs_inactive_symlink_rmt(ip, tp));

And no () around xfs_inactive_symlink_rmt().


+}
Index: b/fs/xfs/xfs_symlink.h
===================================================================
--- a/fs/xfs/xfs_symlink.h
+++ b/fs/xfs/xfs_symlink.h
@@ -60,7 +60,7 @@ extern const struct xfs_buf_ops xfs_syml
  int xfs_symlink(struct xfs_inode *dp, struct xfs_name *link_name,
                const char *target_path, umode_t mode, struct xfs_inode **ipp);
  int xfs_readlink(struct xfs_inode *ip, char *link);
-int xfs_inactive_symlink_rmt(struct xfs_inode *ip, struct xfs_trans **tpp);
+int xfs_inactive_symlink(struct xfs_inode *ip, struct xfs_trans **tpp);

  #endif /* __KERNEL__ */
  #endif /* __XFS_SYMLINK_H */
Index: b/fs/xfs/xfs_trace.h
===================================================================
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -594,6 +594,8 @@ DEFINE_INODE_EVENT(xfs_inode_set_eofbloc
  DEFINE_INODE_EVENT(xfs_inode_clear_eofblocks_tag);
  DEFINE_INODE_EVENT(xfs_inode_free_eofblocks_invalid);

+DEFINE_INODE_EVENT(xfs_symlink_inactive);
+

No need for the extra whitespace, just put it hard up against the
other DEFINE_INODE_EVENT()s. Might be better to put it directly
after the readlink trace definition.

Cheers,

Dave.

Thanks.

--Mark.

<Prev in Thread] Current Thread [Next in Thread>