On Mon, 2020-05-18 at 18:47 +0100, Luis Henriques wrote:
> Similarly to commit 03f219041fdb ("ceph: check i_nlink while converting
> a file handle to dentry"), this fixes another corner case with
> name_to_handle_at/open_by_handle_at.  The issue has been detected by
> xfstest generic/467, when doing:
> 
>  - name_to_handle_at("/cephfs/myfile")
>  - open("/cephfs/myfile")
>  - unlink("/cephfs/myfile")
>  - sync; sync;
>  - drop caches
>  - open_by_handle_at()
> 
> The call to open_by_handle_at should not fail because the file hasn't been
> deleted yet (only unlinked) and we do have a valid handle to it.  -ESTALE
> shall be returned only if i_nlink is 0 *and* i_count is 1.
> 
> This patch also makes sure we have LINK caps before checking i_nlink.
> 
> Signed-off-by: Luis Henriques <lhenriq...@suse.com>
> ---
> Hi!
> 
> (and sorry for the delay in my reply!)
> 
> So, from the discussion thread and some IRC chat with Jeff, I'm sending
> v2.  What changed?  Everything! :-)
> 
> - Use i_count instead of __ceph_is_file_opened to check for open files
> - Add call to ceph_do_getattr to make sure we have LINK caps before
>   accessing i_nlink
> 
> Cheers,
> --
> Luis
> 
>  fs/ceph/export.c | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/ceph/export.c b/fs/ceph/export.c
> index 79dc06881e78..e088843a7734 100644
> --- a/fs/ceph/export.c
> +++ b/fs/ceph/export.c
> @@ -172,9 +172,16 @@ struct inode *ceph_lookup_inode(struct super_block *sb, 
> u64 ino)
>  static struct dentry *__fh_to_dentry(struct super_block *sb, u64 ino)
>  {
>       struct inode *inode = __lookup_inode(sb, ino);
> +     int err;
> +
>       if (IS_ERR(inode))
>               return ERR_CAST(inode);
> -     if (inode->i_nlink == 0) {
> +     /* We need LINK caps to reliably check i_nlink */
> +     err = ceph_do_getattr(inode, CEPH_CAP_LINK_SHARED, false);
> +     if (err)
> +             return ERR_PTR(err);
> +     /* -ESTALE if inode as been unlinked and no file is open */
> +     if ((inode->i_nlink == 0) && (atomic_read(&inode->i_count) == 1)) {
>               iput(inode);
>               return ERR_PTR(-ESTALE);
>       }
> 

Looks good. Merged into testing branch.

Thanks!
-- 
Jeff Layton <jlay...@kernel.org>

Reply via email to