xfs
[Top] [All Lists]

Re: [PATCH v2] xfs: fix the symbolic link assert in xfs_ifree

To: Mark Tinguely <tinguely@xxxxxxx>
Subject: Re: [PATCH v2] xfs: fix the symbolic link assert in xfs_ifree
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Thu, 13 Jun 2013 12:37:44 +1000
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20130612182836.141843930@xxxxxxx>
References: <20130612182818.363497890@.sgi.com> <20130612182836.141843930@xxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Wed, Jun 12, 2013 at 01:28:19PM -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>
> ---
> xfstest is coming.
> 
>  fs/xfs/xfs_symlink.c  |   24 +++++++++++++++++++++++-
>  fs/xfs/xfs_symlink.h  |    1 +
>  fs/xfs/xfs_vnodeops.c |   15 +++------------
>  3 files changed, 27 insertions(+), 13 deletions(-)
> 
> Index: b/fs/xfs/xfs_symlink.c
> ===================================================================
> --- a/fs/xfs/xfs_symlink.c
> +++ b/fs/xfs/xfs_symlink.c
> @@ -606,7 +606,7 @@ xfs_inactive_symlink_rmt(
>  
>       tp = *tpp;
>       mp = ip->i_mount;
> -     ASSERT(ip->i_d.di_size > XFS_IFORK_DSIZE(ip));
> +     ASSERT(ip->i_df.if_flags & XFS_IFEXTENTS);
>       /*
>        * We're freeing a symlink that has some
>        * blocks allocated to it.  Free the
> @@ -720,3 +720,25 @@ xfs_inactive_symlink_rmt(
>   error0:
>       return error;
>  }
> +
> +/*
> + * xfs_inactive_symlink - free a symlink
> + */
> +int
> +xfs_inactive_symlink(
> +     struct xfs_inode        *ip,
> +     struct xfs_trans        **tp)
> +{
> +

Needs to:

        - emit a tracepoint (trace_xfs_symlink_inactive(ip))
        - assert the inode is locked appropriately
        - check for a shutdown condition
        - check for zero size and return success immediately
        - check inode size is within bounds (0 < size < MAXPATHLEN)

i.e. all the same preamble that xfs_readlink() has.

> +     /*
> +      * Zero length symlinks _can_ exist.
> +      */
> +     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;
> +     } else
> +             return (xfs_inactive_symlink_rmt(ip, tp));

No need for an else here, and no need for the () around the
xfs_inactive_symlink_rmt() function call.

> +}
> Index: b/fs/xfs/xfs_symlink.h
> ===================================================================
> --- a/fs/xfs/xfs_symlink.h
> +++ b/fs/xfs/xfs_symlink.h
> @@ -61,6 +61,7 @@ int xfs_symlink(struct xfs_inode *dp, st
>               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);

xfs_inactive_symlink_rmt() is no longer exported, so it can be
removed from the header and made STATIC.

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

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