public inbox for linux-nfs@vger.kernel.org
 help / color / mirror / Atom feed
From: Scott Mayhew <smayhew@redhat.com>
To: trond.myklebust@primarydata.com, anna.schumaker@netapp.com
Cc: bcodding@redhat.com, dwysocha@redhat.com, fsorenso@redhat.com,
	linux-nfs@vger.kernel.org
Subject: [PATCH] nfs/pnfs: fix nfs_direct_req ref leak when i/o falls back to the mds
Date: Wed, 13 Dec 2017 11:15:55 -0500	[thread overview]
Message-ID: <20171213161555.27993-1-smayhew@redhat.com> (raw)

Currently when falling back to doing I/O through the MDS (via
pnfs_{read|write}_through_mds), the client frees the nfs_pgio_header
without releasing the reference taken on the dreq
via pnfs_generic_pg_{read|write}pages -> nfs_pgheader_init ->
nfs_direct_pgio_init.  It then takes another reference on the dreq via
nfs_generic_pg_pgios -> nfs_pgheader_init -> nfs_direct_pgio_init and
as a result the requester will become stuck in inode_dio_wait.  Once
that happens, other processes accessing the inode will become stuck as
well.

Moving the init_hdr call down to nfs_initiate_pgio ensures we take the
reference on the dreq only once.

This can be reproduced (sometimes) by performing "storage failover
takeover" commands on NetApp filer while doing direct I/O from a client.

This can also be reproduced using SystemTap to simulate a failure while
doing direct I/O from a client (from Dave Wysochanski
<dwysocha@redhat.com>):

stap -v -g -e 'probe module("nfs_layout_nfsv41_files").function("nfs4_fl_prepare_ds").return { $return=NULL; exit(); }'

Suggested-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Scott Mayhew <smayhew@redhat.com>
---
 fs/nfs/pagelist.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c
index d0543e1..d478c19 100644
--- a/fs/nfs/pagelist.c
+++ b/fs/nfs/pagelist.c
@@ -54,9 +54,6 @@ void nfs_pgheader_init(struct nfs_pageio_descriptor *desc,
 	hdr->dreq = desc->pg_dreq;
 	hdr->release = release;
 	hdr->completion_ops = desc->pg_completion_ops;
-	if (hdr->completion_ops->init_hdr)
-		hdr->completion_ops->init_hdr(hdr);
-
 	hdr->pgio_mirror_idx = desc->pg_mirror_idx;
 }
 EXPORT_SYMBOL_GPL(nfs_pgheader_init);
@@ -607,6 +604,9 @@ int nfs_initiate_pgio(struct rpc_clnt *clnt, struct nfs_pgio_header *hdr,
 	};
 	int ret = 0;
 
+	if (hdr->completion_ops->init_hdr)
+		hdr->completion_ops->init_hdr(hdr);
+
 	hdr->rw_ops->rw_initiate(hdr, &msg, rpc_ops, &task_setup_data, how);
 
 	dprintk("NFS: initiated pgio call "
-- 
2.9.5

             reply	other threads:[~2017-12-13 16:15 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-12-13 16:15 Scott Mayhew [this message]
2017-12-13 17:35 ` [PATCH] nfs/pnfs: fix nfs_direct_req ref leak when i/o falls back to the mds Trond Myklebust
2017-12-13 18:22   ` Scott Mayhew
2017-12-15 21:12     ` [PATCH v2] " Scott Mayhew
2018-01-11 13:10       ` Scott Mayhew
2018-01-11 13:20         ` Trond Myklebust
2018-01-12 17:39           ` Scott Mayhew

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20171213161555.27993-1-smayhew@redhat.com \
    --to=smayhew@redhat.com \
    --cc=anna.schumaker@netapp.com \
    --cc=bcodding@redhat.com \
    --cc=dwysocha@redhat.com \
    --cc=fsorenso@redhat.com \
    --cc=linux-nfs@vger.kernel.org \
    --cc=trond.myklebust@primarydata.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox