From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: with ECARTIS (v1.0.0; list xfs); Sat, 13 Sep 2008 07:56:50 -0700 (PDT) Received: from cuda.sgi.com (cuda1.sgi.com [192.48.168.28]) by oss.sgi.com (8.12.11.20060308/8.12.11/SuSE Linux 0.7) with ESMTP id m8DEuZ5U022246 for ; Sat, 13 Sep 2008 07:56:35 -0700 Received: from ipmail05.adl2.internode.on.net (localhost [127.0.0.1]) by cuda.sgi.com (Spam Firewall) with ESMTP id 3DC31126C1AD for ; Sat, 13 Sep 2008 07:58:05 -0700 (PDT) Received: from ipmail05.adl2.internode.on.net (ipmail05.adl2.internode.on.net [203.16.214.145]) by cuda.sgi.com with ESMTP id 7md9VGCoOAqH24X2 for ; Sat, 13 Sep 2008 07:58:05 -0700 (PDT) Received: from dave by disturbed with local (Exim 4.69) (envelope-from ) id 1KeWZS-0002Nv-2Z for xfs@oss.sgi.com; Sun, 14 Sep 2008 00:57:58 +1000 From: Dave Chinner Subject: [PATCH 4/8] XFS: Allow 64 bit machines to avoid the AIL lock during flushes Date: Sun, 14 Sep 2008 00:57:53 +1000 Message-Id: <1221317877-8333-5-git-send-email-david@fromorbit.com> In-Reply-To: <1221317877-8333-1-git-send-email-david@fromorbit.com> References: <1221317877-8333-1-git-send-email-david@fromorbit.com> Sender: xfs-bounce@oss.sgi.com Errors-to: xfs-bounce@oss.sgi.com List-Id: xfs To: xfs@oss.sgi.com When copying lsn's from the log item to the inode or dquot flush lsn, we currently grab the AIL lock. We do this because the LSN is a 64 bit quantity and it needs to be read atomically. The lock is used to guarantee atomicity for 32 bit platforms. Make the LSN copying a small function, and make the function used conditional on BITS_PER_LONG so that 64 bit machines don't need to take the AIL lock in these places. Signed-off-by: Dave Chinner --- fs/xfs/quota/xfs_dquot.c | 6 ++---- fs/xfs/xfs_inode.c | 17 +++++++---------- fs/xfs/xfs_trans_priv.h | 23 +++++++++++++++++++++++ 3 files changed, 32 insertions(+), 14 deletions(-) diff --git a/fs/xfs/quota/xfs_dquot.c b/fs/xfs/quota/xfs_dquot.c index f2705f2..8e30a10 100644 --- a/fs/xfs/quota/xfs_dquot.c +++ b/fs/xfs/quota/xfs_dquot.c @@ -1274,10 +1274,8 @@ xfs_qm_dqflush( dqp->dq_flags &= ~(XFS_DQ_DIRTY); mp = dqp->q_mount; - /* lsn is 64 bits */ - spin_lock(&mp->m_ail_lock); - dqp->q_logitem.qli_flush_lsn = dqp->q_logitem.qli_item.li_lsn; - spin_unlock(&mp->m_ail_lock); + xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn, + &dqp->q_logitem.qli_item.li_lsn); /* * Attach an iodone routine so that we can remove this dquot from the diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 401b92c..448f507 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -2213,9 +2213,9 @@ xfs_ifree_cluster( iip = (xfs_inode_log_item_t *)lip; ASSERT(iip->ili_logged == 1); lip->li_cb = (void(*)(xfs_buf_t*,xfs_log_item_t*)) xfs_istale_done; - spin_lock(&mp->m_ail_lock); - iip->ili_flush_lsn = iip->ili_item.li_lsn; - spin_unlock(&mp->m_ail_lock); + xfs_trans_ail_copy_lsn(mp->m_ail, + &iip->ili_flush_lsn, + &iip->ili_item.li_lsn); xfs_iflags_set(iip->ili_inode, XFS_ISTALE); pre_flushed++; } @@ -2236,9 +2236,8 @@ xfs_ifree_cluster( iip->ili_last_fields = iip->ili_format.ilf_fields; iip->ili_format.ilf_fields = 0; iip->ili_logged = 1; - spin_lock(&mp->m_ail_lock); - iip->ili_flush_lsn = iip->ili_item.li_lsn; - spin_unlock(&mp->m_ail_lock); + xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn, + &iip->ili_item.li_lsn); xfs_buf_attach_iodone(bp, (void(*)(xfs_buf_t*,xfs_log_item_t*)) @@ -3479,10 +3478,8 @@ xfs_iflush_int( iip->ili_format.ilf_fields = 0; iip->ili_logged = 1; - ASSERT(sizeof(xfs_lsn_t) == 8); /* don't lock if it shrinks */ - spin_lock(&mp->m_ail_lock); - iip->ili_flush_lsn = iip->ili_item.li_lsn; - spin_unlock(&mp->m_ail_lock); + xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn, + &iip->ili_item.li_lsn); /* * Attach the function xfs_iflush_done to the inode's diff --git a/fs/xfs/xfs_trans_priv.h b/fs/xfs/xfs_trans_priv.h index 8b84167..c6ea7fe 100644 --- a/fs/xfs/xfs_trans_priv.h +++ b/fs/xfs/xfs_trans_priv.h @@ -103,4 +103,27 @@ void xfsaild_wakeup(struct xfs_ail *, xfs_lsn_t); int xfsaild_start(struct xfs_ail *); void xfsaild_stop(struct xfs_ail *); +#if BITS_PER_LONG != 64 +static inline void +xfs_trans_ail_copy_lsn( + struct xfs_ail *ailp, + xfs_lsn_t *dst, + xfs_lsn_t *src) +{ + ASSERT(sizeof(xfs_lsn_t) == 8); /* don't lock if it shrinks */ + spin_lock(&ailp->xa_mount->m_ail_lock); + *dst = *src; + spin_unlock(&ailp->xa_mount->m_ail_lock); +} +#else +static inline void +xfs_trans_ail_copy_lsn( + struct xfs_ail *ailp, + xfs_lsn_t *dst, + xfs_lsn_t *src) +{ + ASSERT(sizeof(xfs_lsn_t) == 8); + *dst = *src; +} +#endif #endif /* __XFS_TRANS_PRIV_H__ */ -- 1.5.6