Linux NFS development
 help / color / mirror / Atom feed
From: Andrew Elble <aweits@rit.edu>
To: linux-nfs@vger.kernel.org
Cc: Andrew Elble <aweits@rit.edu>
Subject: [PATCH RFC] nfsd: don't revoke delegations that a client has stated it doesn't have
Date: Thu, 15 Oct 2015 12:09:19 -0400	[thread overview]
Message-ID: <1444925359-18251-1-git-send-email-aweits@rit.edu> (raw)

Assuming a client has lost a delegation: If the server goes to recall
the delegation, an attempt is made to recall it twice separated by
a delay of 2 seconds. Both times, the client will state that it
doesn't have the delegation via -EBADHANDLE or -NFS4ERR_BAD_STATEID.

1.) Any race between a delegation grant and a recall has been
    presumably avoided by the delay and second attempt.
2.) The client doesn't have the delegation.
3.) The backchannel is responsive.

After these two attempts fail, the laundromat will eventually revoke
them and add these delegations to cl_revoked. This results in another
attempt to get the client to return the delegation via
TEST/FREE STATEID. This will also fail with no means
of resolution, and will cause the server and client to loop
indefinitely, as the client has nothing to give the server to satisfy it.

The changes here are to establish a safe way to recover by:

If the client has responded with -EBADHANDLE or -NFS4ERR_BAD_STATEID:
1.) Not failing the backchannel after two attempts at a recall.
2.) At the time revocation would normally occur: destroying the
    delegation on the server side.

Signed-off-by: Andrew Elble <aweits@rit.edu>
---
 fs/nfsd/nfs4state.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index da21df673ed9..34d928bd847d 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -3578,7 +3578,7 @@ static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
 			rpc_delay(task, 2 * HZ);
 			return 0;
 		}
-		/*FALLTHRU*/
+		return 1;
 	default:
 		return -1;
 	}
@@ -4450,8 +4450,16 @@ nfs4_laundromat(struct nfsd_net *nn)
 	while (!list_empty(&reaplist)) {
 		dp = list_first_entry(&reaplist, struct nfs4_delegation,
 					dl_recall_lru);
-		list_del_init(&dp->dl_recall_lru);
-		revoke_delegation(dp);
+		if ((dp->dl_recall.cb_status == -EBADHANDLE) ||
+			(dp->dl_recall.cb_status == -NFS4ERR_BAD_STATEID)) {
+			dprintk("nfsd: client: %.*s is losing delegations",
+				(int)dp->dl_recall.cb_clp->cl_name.len,
+				dp->dl_recall.cb_clp->cl_name.data);
+			destroy_delegation(dp);
+		} else {
+			list_del_init(&dp->dl_recall_lru);
+			revoke_delegation(dp);
+		}
 	}
 
 	spin_lock(&nn->client_lock);
-- 
2.4.6


                 reply	other threads:[~2015-10-15 16:09 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

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=1444925359-18251-1-git-send-email-aweits@rit.edu \
    --to=aweits@rit.edu \
    --cc=linux-nfs@vger.kernel.org \
    /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