From: "Yan, Zheng" <zheng.z.yan@intel.com>
To: Sage Weil <sage@inktank.com>
Cc: ceph-devel <ceph-devel@vger.kernel.org>
Subject: Re: [PATCH 3/3] ceph: handle race between cap reconnect and cap release
Date: Tue, 24 Sep 2013 09:05:51 +0800 [thread overview]
Message-ID: <5240E56F.9030508@intel.com> (raw)
In-Reply-To: <CAAM7YA=apJ-jJy3iHVrW3O-_nhXvkYFUTKBv73kD-B-RoY47Dg@mail.gmail.com>
updated patch. add comments for the locking, no function changes.
---
From d40facb3df480994ab6575a1ce90b645123b1f92 Mon Sep 17 00:00:00 2001
From: "Yan, Zheng" <zheng.z.yan@intel.com>
Date: Sun, 22 Sep 2013 11:08:14 +0800
Subject: [PATCH] ceph: handle race between cap reconnect and cap release
When a cap get released while composing the cap reconnect message.
We should skip queuing the release message if the cap hasn't been
added to the cap reconnect message.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
---
fs/ceph/caps.c | 8 +++++++-
fs/ceph/mds_client.c | 21 ++++++++++++++++++---
fs/ceph/mds_client.h | 1 +
3 files changed, 26 insertions(+), 4 deletions(-)
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index d2d6e40..3c0a4bd 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -909,7 +909,13 @@ void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
/* remove from session list */
spin_lock(&session->s_cap_lock);
- if (queue_release)
+ /*
+ * s_cap_reconnect is protected by s_cap_lock. no one changes
+ * s_cap_gen while session is in the reconnect state.
+ */
+ if (queue_release &&
+ (!session->s_cap_reconnect ||
+ cap->cap_gen == session->s_cap_gen))
__queue_cap_release(session, ci->i_vino.ino, cap->cap_id,
cap->mseq, cap->issue_seq);
diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 6b4ed35..30397a6 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -444,6 +444,7 @@ static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
INIT_LIST_HEAD(&s->s_waiting);
INIT_LIST_HEAD(&s->s_unsafe);
s->s_num_cap_releases = 0;
+ s->s_cap_reconnect = 0;
s->s_cap_iterator = NULL;
INIT_LIST_HEAD(&s->s_cap_releases);
INIT_LIST_HEAD(&s->s_cap_releases_done);
@@ -1415,7 +1416,6 @@ static void discard_cap_releases(struct ceph_mds_client *mdsc,
unsigned num;
dout("discard_cap_releases mds%d\n", session->s_mds);
- spin_lock(&session->s_cap_lock);
/* zero out the in-progress message */
msg = list_first_entry(&session->s_cap_releases,
@@ -1442,8 +1442,6 @@ static void discard_cap_releases(struct ceph_mds_client *mdsc,
msg->front.iov_len = sizeof(*head);
list_add(&msg->list_head, &session->s_cap_releases);
}
-
- spin_unlock(&session->s_cap_lock);
}
/*
@@ -2488,6 +2486,7 @@ static int encode_caps_cb(struct inode *inode, struct ceph_cap *cap,
cap->seq = 0; /* reset cap seq */
cap->issue_seq = 0; /* and issue_seq */
cap->mseq = 0; /* and migrate_seq */
+ cap->cap_gen = cap->session->s_cap_gen;
if (recon_state->flock) {
rec.v2.cap_id = cpu_to_le64(cap->cap_id);
@@ -2611,8 +2610,20 @@ static void send_mds_reconnect(struct ceph_mds_client *mdsc,
dout("session %p state %s\n", session,
session_state_name(session->s_state));
+ spin_lock(&session->s_gen_ttl_lock);
+ session->s_cap_gen++;
+ spin_unlock(&session->s_gen_ttl_lock);
+
+ spin_lock(&session->s_cap_lock);
+ /*
+ * notify __ceph_remove_cap() that we are composing cap reconnect.
+ * If a cap get released before being added to the cap reconnect,
+ * __ceph_remove_cap() should skip queuing cap release.
+ */
+ session->s_cap_reconnect = 1;
/* drop old cap expires; we're about to reestablish that state */
discard_cap_releases(mdsc, session);
+ spin_unlock(&session->s_cap_lock);
/* traverse this session's caps */
s_nr_caps = session->s_nr_caps;
@@ -2627,6 +2638,10 @@ static void send_mds_reconnect(struct ceph_mds_client *mdsc,
if (err < 0)
goto fail;
+ spin_lock(&session->s_cap_lock);
+ session->s_cap_reconnect = 0;
+ spin_unlock(&session->s_cap_lock);
+
/*
* snaprealms. we provide mds with the ino, seq (version), and
* parent for all of our realms. If the mds has any newer info,
diff --git a/fs/ceph/mds_client.h b/fs/ceph/mds_client.h
index c2a19fb..4c053d0 100644
--- a/fs/ceph/mds_client.h
+++ b/fs/ceph/mds_client.h
@@ -132,6 +132,7 @@ struct ceph_mds_session {
struct list_head s_caps; /* all caps issued by this session */
int s_nr_caps, s_trim_caps;
int s_num_cap_releases;
+ int s_cap_reconnect;
struct list_head s_cap_releases; /* waiting cap_release messages */
struct list_head s_cap_releases_done; /* ready to send */
struct ceph_cap *s_cap_iterator;
--
1.8.1.4
next prev parent reply other threads:[~2013-09-24 1:05 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-09-22 4:46 [PATCH 1/3] ceph: queue cap release in __ceph_remove_cap() Yan, Zheng
2013-09-22 4:46 ` [PATCH 2/3] ceph: set caps count after composing cap reconnect message Yan, Zheng
2013-09-22 4:46 ` [PATCH 3/3] ceph: handle race between cap reconnect and cap release Yan, Zheng
2013-09-23 14:31 ` Sage Weil
2013-09-23 16:08 ` Yan, Zheng
2013-09-24 1:05 ` Yan, Zheng [this message]
2013-09-23 3:48 ` [PATCH 1/3] ceph: queue cap release in __ceph_remove_cap() Sage Weil
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=5240E56F.9030508@intel.com \
--to=zheng.z.yan@intel.com \
--cc=ceph-devel@vger.kernel.org \
--cc=sage@inktank.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.