linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Dmitry Monakhov <dmonakhov@openvz.org>
To: linux-fsdevel@vger.kernel.org
Cc: jack@suse.cz, hch@infradead.org, Dmitry Monakhov <dmonakhov@openvz.org>
Subject: [PATCH 14/19] quota: protect dqget() from parallels quotaoff via RCU
Date: Fri, 22 Oct 2010 21:34:59 +0400	[thread overview]
Message-ID: <1287768904-27810-15-git-send-email-dmonakhov@openvz.org> (raw)
In-Reply-To: <1287768904-27810-1-git-send-email-dmonakhov@openvz.org>

This allow us to remove dq_state_lock from dqget quota state test.
synchronize_rcu() in dquot_disable() is a bit strage usage for RCU,
but rcu_read_lock will be used in dqget() more properly later to
get make lookup lock-less.

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
---
 fs/quota/dquot.c |   32 ++++++++++++++++++++++----------
 1 files changed, 22 insertions(+), 10 deletions(-)

diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c
index 771aaab..0d755ef 100644
--- a/fs/quota/dquot.c
+++ b/fs/quota/dquot.c
@@ -878,7 +878,7 @@ static struct dquot *get_empty_dquot(struct super_block *sb, int type)
 /*
  * Get reference to dquot
  *
- * Locking is slightly tricky here. We are guarded from parallel quotaoff()
+ * We are guarded from parallel quotaoff() by holding rcu_read_lock
  * destroying our dquot by:
  *   a) checking for quota flags under dq_list_lock and
  *   b) getting a reference to dquot before we release dq_list_lock
@@ -888,17 +888,15 @@ struct dquot *dqget(struct super_block *sb, unsigned int id, int type)
 	struct dquot *dquot = NULL, *empty = NULL;
 	struct quota_info *dqopt = dqopts(sb);
 
-        if (!sb_has_quota_active(sb, type))
-		return NULL;
 we_slept:
-	spin_lock(&dqopt->dq_list_lock);
-	spin_lock(&dqopt->dq_state_lock);
+	rcu_read_lock();
 	if (!sb_has_quota_active(sb, type)) {
-		spin_unlock(&dqopt->dq_state_lock);
-		spin_unlock(&dqopt->dq_list_lock);
-		goto out;
+		rcu_read_unlock();
+		return NULL;
 	}
-	spin_unlock(&dqopt->dq_state_lock);
+	spin_lock(&dqopt->dq_list_lock);
+	/* XXX: Currently RCU used only for synchronization with quotaoff */
+	rcu_read_unlock();
 
 	dquot = find_dquot(sb, id, type);
 	if (!dquot) {
@@ -2113,6 +2111,21 @@ int dquot_disable(struct super_block *sb, int type, unsigned int flags)
 		if (sb_has_quota_loaded(sb, cnt) && !(flags & DQUOT_SUSPENDED))
 			continue;
 
+		toputinode[cnt] = dqopt->files[cnt];
+	}
+	/*
+	 * Wait for all dqget() callers to finish.
+	 */
+	synchronize_rcu();
+
+	/*
+	 * At this moment all quota functions disabled, is is now safe to
+	 * perform final cleanup.
+	 */
+	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
+
+		if (!toputinode[cnt])
+			continue;
 		/* Note: these are blocking operations */
 		drop_dquot_ref(sb, cnt);
 		invalidate_dquots(sb, cnt);
@@ -2126,7 +2139,6 @@ int dquot_disable(struct super_block *sb, int type, unsigned int flags)
 			dqopt->fmt_ops[cnt]->free_file_info(sb, cnt);
 		put_quota_format(dqopt->info[cnt].dqi_format);
 
-		toputinode[cnt] = dqopt->files[cnt];
 		if (!sb_has_quota_loaded(sb, cnt))
 			dqopt->files[cnt] = NULL;
 		dqopt->info[cnt].dqi_flags = 0;
-- 
1.6.5.2


  parent reply	other threads:[~2010-10-22 17:36 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-10-22 17:34 [PATCH 00/19] quota: RFC SMP improvements for generic quota V2 Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 01/19] quota: protect getfmt call with dqonoff_mutex lock Dmitry Monakhov
2010-10-28 14:27   ` Christoph Hellwig
2010-10-22 17:34 ` [PATCH 02/19] quota: Wrap common expression to helper function Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 03/19] quota: mode quota internals from sb to quota_info Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 04/19] quota: Convert dq_state_lock to per-sb dq_state_lock Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 05/19] quota: add quota format lock Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 06/19] quota: make dquot lists per-sb Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 07/19] quota: make per-sb hash array Dmitry Monakhov
2010-10-27 19:31   ` Al Viro
2010-10-28 10:58     ` Dmitry
2010-10-22 17:34 ` [PATCH 08/19] quota: remove global dq_list_lock Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 09/19] quota: rename dq_lock Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 10/19] quota: make per-sb dq_data_lock Dmitry Monakhov
2010-10-26 17:38   ` Dmitry
2010-10-22 17:34 ` [PATCH 11/19] quota: protect dquot mem info with object's lock Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 12/19] quota: drop dq_data_lock where possible Dmitry Monakhov
2010-10-22 17:34 ` [PATCH 13/19] quota: relax dq_data_lock dq_lock locking consistency Dmitry Monakhov
2010-10-22 17:34 ` Dmitry Monakhov [this message]
2010-10-22 17:35 ` [PATCH 15/19] quota: remove dq_state_lock Dmitry Monakhov
2010-10-22 17:35 ` [PATCH 16/19] fs: add unlocked helpers Dmitry Monakhov
2010-10-22 17:35 ` [PATCH 17/19] quota: Some stylistic cleanup for dquot interface Dmitry Monakhov
2010-10-22 17:35 ` [PATCH 18/19] quota: remove dqptr_sem Dmitry Monakhov
2010-10-22 17:35 ` [PATCH 19/19] quota: redesign dquot reference counting Dmitry Monakhov

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=1287768904-27810-15-git-send-email-dmonakhov@openvz.org \
    --to=dmonakhov@openvz.org \
    --cc=hch@infradead.org \
    --cc=jack@suse.cz \
    --cc=linux-fsdevel@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;
as well as URLs for NNTP newsgroup(s).