lkml.org 
[lkml]   [2006]   [Aug]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH] NFS: Replace null dentries that appear in readdir's list
From
Date
On Wed, 2006-08-16 at 17:15 +0100, David Howells wrote:
> Trond Myklebust <trond.myklebust@fys.uio.no> wrote:
>
> > Better still, in the case of a negative dentry: just call d_drop().

Yes, unhashing the dentry will make it invisible but I thought your
point was that the server hadn't actually been contacted when I
suggested it before. Is that a concern?

>
> How about this then?
>
> David
> ---
> NFS: Replace null dentries that appear in readdir's list
>
> From: David Howells <dhowells@redhat.com>
>
> Have nfs_readdir_lookup() drop and replace any null dentry when it
> that gets listed by a READDIR RPC call.
>
> This can be caused by an optimisation in nfs_lookup() that causes a dentry to
> be incorrectly left as negative when mkdir() or similar is aborted by SELinux
> mid-procedure.
>
> This can be triggered by mounting through autofs4 a server:/ NFS share for
> which there are other exports available on that server. SELinux also has to
> be turned on in enforcing mode to abort mid-flow the mkdir operation performed
> by autofs4.
>
> The problematic sequence of events is this:
>
> (1) nfs_lookup() is called by sys_mkdirat() -> lookup_create() ->
> __lookup_hash() with intent to create exclusively set in the nameidata:
>
> nd->flags == LOOKUP_CREATE
> nd->intent.open.flags == O_EXCL
>
> (2) nfs_lookup() has an optimisation to avoid going to the server in this
> case, presumably since the nfs_mkdir() op or whatever will deal with the
> conflict.
>
> (3) nfs_lookup() returns successfully, leaving the dentry in a negative state,
> but attached to the parent directory.
>
> (4) sys_mkdirat() calls vfs_mkdir() which calls may_create(). may_create()
> checks that the directory has MAY_WRITE and MAY_EXEC permissions.
>
> (5) may_create() calls nfs_permission(), which grants permission.
>
> (6) may_create() calls security_inode_permission(), which calls SELinux, which
> then _DENIES_ permission.
>
> (7) may_create() fails, and vfs_mkdir() then fails and sys_mkdirat() then
> fails (as does sys_mkdir).
>
> _However_, the new dentry is left in the negative state, with no
> consultation of the server.
>
> (8) The parent directory is listed, and the name of the new dentry is
> returned.
>
> (9) stat on the new dentry fails (because it's negative), and "ls -l" returns
> "?---------" as the file type and mode.
>
> This fix makes step (8) replace the dentry looked up in steps (1) - (3).
>
> Signed-Off-By: David Howells <dhowells@redhat.com>
> ---
>
> fs/nfs/dir.c | 9 +++++++--
> 1 files changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> index e746ed1..bb8b5f0 100644
> --- a/fs/nfs/dir.c
> +++ b/fs/nfs/dir.c
> @@ -1105,8 +1105,13 @@ static struct dentry *nfs_readdir_lookup
> }
> name.hash = full_name_hash(name.name, name.len);
> dentry = d_lookup(parent, &name);
> - if (dentry != NULL)
> - return dentry;
> + if (dentry != NULL) {
> + /* negative dentries must be reconsidered */
> + if (!IS_ERR(dentry) && !dentry->d_inode)
> + d_drop(dentry);
> + else
> + return dentry;
> + }
> if (!desc->plus || !(entry->fattr->valid & NFS_ATTR_FATTR))
> return NULL;
> /* Note: caller is already holding the dir->i_mutex! */
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-08-17 07:17    [from the cache]
©2003-2011 Jasper Spaans