Return-Path: Received: from smtp-o-1.desy.de ([131.169.56.154]:43903 "EHLO smtp-o-1.desy.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725879AbeHQRnp (ORCPT ); Fri, 17 Aug 2018 13:43:45 -0400 Received: from smtp-buf-1.desy.de (smtp-buf-1.desy.de [IPv6:2001:638:700:1038::1:a4]) by smtp-o-1.desy.de (DESY-O-1) with ESMTP id A4C6B280E5C for ; Fri, 17 Aug 2018 16:40:05 +0200 (CEST) Date: Fri, 17 Aug 2018 16:40:04 +0200 (CEST) From: "Mkrtchyan, Tigran" To: Trond Myklebust Cc: Anna Schumaker , linux-nfs Message-ID: <1474233718.35838144.1534516804037.JavaMail.zimbra@desy.de> In-Reply-To: <20180817143108.6504-1-trond.myklebust@hammerspace.com> References: <20180817143108.6504-1-trond.myklebust@hammerspace.com> Subject: Re: [PATCH] pNFS/flexfiles: ff_layout_pg_init_read should exit on error MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Sender: linux-nfs-owner@vger.kernel.org List-ID: Hi Trond, shouldn't this go to stable as well? Tigran. ----- Original Message ----- > From: "Trond Myklebust" > To: "Anna Schumaker" > Cc: "linux-nfs" > Sent: Friday, August 17, 2018 4:31:08 PM > Subject: [PATCH] pNFS/flexfiles: ff_layout_pg_init_read should exit on error > From: Trond Myklebust > > If we get an error while retrieving the layout, then we should > report it rather than falling back to I/O through the MDS. > > Signed-off-by: Trond Myklebust > --- > fs/nfs/flexfilelayout/flexfilelayout.c | 28 ++++++++++---------------- > 1 file changed, 11 insertions(+), 17 deletions(-) > > diff --git a/fs/nfs/flexfilelayout/flexfilelayout.c > b/fs/nfs/flexfilelayout/flexfilelayout.c > index 280b6b060d53..cae43333ef16 100644 > --- a/fs/nfs/flexfilelayout/flexfilelayout.c > +++ b/fs/nfs/flexfilelayout/flexfilelayout.c > @@ -812,7 +812,6 @@ ff_layout_pg_get_read(struct nfs_pageio_descriptor *pgio, > struct nfs_page *req, > bool strict_iomode) > { > -retry_strict: > pnfs_put_lseg(pgio->pg_lseg); > pgio->pg_lseg = pnfs_update_layout(pgio->pg_inode, > req->wb_context, > @@ -825,16 +824,6 @@ ff_layout_pg_get_read(struct nfs_pageio_descriptor *pgio, > pgio->pg_error = PTR_ERR(pgio->pg_lseg); > pgio->pg_lseg = NULL; > } > - > - /* If we don't have checking, do get a IOMODE_RW > - * segment, and the server wants to avoid READs > - * there, then retry! > - */ > - if (pgio->pg_lseg && !strict_iomode && > - ff_layout_avoid_read_on_rw(pgio->pg_lseg)) { > - strict_iomode = true; > - goto retry_strict; > - } > } > > static void > @@ -849,14 +838,16 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, > retry: > pnfs_generic_pg_check_layout(pgio); > /* Use full layout for now */ > - if (!pgio->pg_lseg) > + if (!pgio->pg_lseg) { > ff_layout_pg_get_read(pgio, req, false); > - else if (ff_layout_avoid_read_on_rw(pgio->pg_lseg)) > + if (!pgio->pg_lseg) > + goto out_nolseg; > + } > + if (ff_layout_avoid_read_on_rw(pgio->pg_lseg)) { > ff_layout_pg_get_read(pgio, req, true); > - > - /* If no lseg, fall back to read through mds */ > - if (pgio->pg_lseg == NULL) > - goto out_mds; > + if (!pgio->pg_lseg) > + goto out_nolseg; > + } > > ds = ff_layout_choose_best_ds_for_read(pgio->pg_lseg, 0, &ds_idx); > if (!ds) { > @@ -878,6 +869,9 @@ ff_layout_pg_init_read(struct nfs_pageio_descriptor *pgio, > pgm->pg_bsize = mirror->mirror_ds->ds_versions[0].rsize; > > return; > +out_nolseg: > + if (pgio->pg_error < 0) > + return; > out_mds: > pnfs_put_lseg(pgio->pg_lseg); > pgio->pg_lseg = NULL; > -- > 2.17.1