On Aug 19, 2015, at 1:24 PM, Peng Tao <tao.peng@xxxxxxxxxxxxxxx> wrote:
> From: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>
>
> pNFS writes don't return attributes, however that doesn't mean that we
> should ignore the fact that they may be extending the file. This patch
> ensures that if a write is seen to extend the file, then we always set
> an attribute barrier, and update the cached file size.
I think this is the same problem I see when a write delegation
is in effect. The flush-after-close does not wait, and does not
set an attribute barrier. So a subsequent append write starts
at the wrong file length.
> Signed-off-by: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>
> ---
> Please apply this to 4.0+ stable releases. It is upstream commit
> 2b83d3de4c18af49800e0b26ae013db4fcf43a4a. It fixes
> a08a8cd375db9769588257e7782f6b6b68561b88(NFS: Add attribute update barriers to NFS writebacks)
> and makes sure we pass fstest read_write/mctime.t test with delegation and inband IO.
> Without it, users might see mtime not updated after write.
>
> Cheers,
> Tao
>
> fs/nfs/write.c | 15 +++++++++------
> 1 file changed, 9 insertions(+), 6 deletions(-)
>
> diff --git a/fs/nfs/write.c b/fs/nfs/write.c
> index 359e9ad..0e6a2b8 100644
> --- a/fs/nfs/write.c
> +++ b/fs/nfs/write.c
> @@ -1378,24 +1378,27 @@ static void nfs_writeback_check_extend(struct nfs_pgio_header *hdr,
> {
> struct nfs_pgio_args *argp = &hdr->args;
> struct nfs_pgio_res *resp = &hdr->res;
> + u64 size = argp->offset + resp->count;
>
> if (!(fattr->valid & NFS_ATTR_FATTR_SIZE))
> + fattr->size = size;
> + if (nfs_size_to_loff_t(fattr->size) < i_size_read(hdr->inode)) {
> + fattr->valid &= ~NFS_ATTR_FATTR_SIZE;
> return;
> - if (argp->offset + resp->count != fattr->size)
> - return;
> - if (nfs_size_to_loff_t(fattr->size) < i_size_read(hdr->inode))
> + }
> + if (size != fattr->size)
> return;
> /* Set attribute barrier */
> nfs_fattr_set_barrier(fattr);
> + /* ...and update size */
> + fattr->valid |= NFS_ATTR_FATTR_SIZE;
> }
>
> void nfs_writeback_update_inode(struct nfs_pgio_header *hdr)
> {
> - struct nfs_fattr *fattr = hdr->res.fattr;
> + struct nfs_fattr *fattr = &hdr->fattr;
> struct inode *inode = hdr->inode;
>
> - if (fattr == NULL)
> - return;
> spin_lock(&inode->i_lock);
> nfs_writeback_check_extend(hdr, fattr);
> nfs_post_op_update_inode_force_wcc_locked(inode, fattr);
> --
> 1.8.3.1
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
Chuck Lever
--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html