public inbox for linux-xfs@vger.kernel.org
 help / color / mirror / Atom feed
From: Dave Chinner <david@fromorbit.com>
To: xfs@oss.sgi.com
Subject: ***** SUSPECTED SPAM ***** [PATCH 45/50] xfs: avoid CIL allocation during insert
Date: Mon, 12 Aug 2013 20:50:06 +1000	[thread overview]
Message-ID: <1376304611-22994-46-git-send-email-david@fromorbit.com> (raw)
In-Reply-To: <1376304611-22994-1-git-send-email-david@fromorbit.com>

From: Dave Chinner <dchinner@redhat.com>

Now that we have the size of the log vector that has been allocated,
we can determine if we need to allocate a new log vector for
formatting and insertion. We only need to allocate a new vector if
it won't fit into the existing buffer.

However, we need to hold the CIL context lock while we do this so
that we can't race with a push draining the currently queued log
vectors. It is safe to do this as long as we do GFP_NOFS allocation
to avoid avoid memory allocation recursing into the filesystem.
Hence we can safely overwrite the existing log vector on the CIL if
it is large enough to hold all the dirty regions of the current
item.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
 fs/xfs/xfs_log_cil.c | 52 +++++++++++++++++++++++++++++++++-------------------
 fs/xfs/xfs_trans.h   |  1 -
 2 files changed, 33 insertions(+), 20 deletions(-)

diff --git a/fs/xfs/xfs_log_cil.c b/fs/xfs/xfs_log_cil.c
index 423ceaf..b20b157 100644
--- a/fs/xfs/xfs_log_cil.c
+++ b/fs/xfs/xfs_log_cil.c
@@ -185,6 +185,22 @@ xlog_cil_prepare_log_vecs(
 		buf_size = sizeof(struct xfs_log_vec) + nbytes +
 				niovecs * sizeof(struct xfs_log_iovec);
 
+		/* compare to existing item size */
+		if (lip->li_lv && buf_size <= lip->li_lv->lv_size) {
+			/* same or smaller, optimise common overwrite case */
+			lv = lip->li_lv;
+			lv->lv_next = NULL;
+
+			if (ordered)
+				goto insert;
+
+			/* Ensure the lv is set up according to ->iop_size */
+			lv->lv_niovecs = niovecs;
+			lv->lv_buf = (char *)lv + buf_size - nbytes;
+			lv->lv_buf_len = xlog_cil_lv_item_format(lip, lv);
+			goto insert;
+		}
+
 		/* allocate new data chunk */
 		lv = kmem_zalloc(buf_size, KM_SLEEP|KM_NOFS);
 		lv->lv_item = lip;
@@ -204,8 +220,8 @@ xlog_cil_prepare_log_vecs(
 		lv->lv_buf = (char *)lv + buf_size - nbytes;
 
 		lv->lv_buf_len = xlog_cil_lv_item_format(lip, lv);
-		ASSERT(lv->lv_buf_len <= nbytes);
 insert:
+		ASSERT(lv->lv_buf_len <= nbytes);
 		if (!ret_lv)
 			ret_lv = lv;
 		else
@@ -230,7 +246,17 @@ xfs_cil_prepare_item(
 {
 	struct xfs_log_vec	*old = lv->lv_item->li_lv;
 
-	if (old) {
+	if (!old) {
+		/* new lv, must pin the log item */
+		ASSERT(!lv->lv_item->li_lv);
+
+		if (lv->lv_buf_len != XFS_LOG_VEC_ORDERED) {
+			*len += lv->lv_buf_len;
+			*diff_iovecs += lv->lv_niovecs;
+		}
+		lv->lv_item->li_ops->iop_pin(lv->lv_item);
+
+	} else if (old != lv) {
 		/* existing lv on log item, space used is a delta */
 		ASSERT((old->lv_buf && old->lv_buf_len && old->lv_niovecs) ||
 			old->lv_buf_len == XFS_LOG_VEC_ORDERED);
@@ -249,15 +275,8 @@ xfs_cil_prepare_item(
 		*diff_iovecs += lv->lv_niovecs - old->lv_niovecs;
 		kmem_free(old);
 	} else {
-		/* new lv, must pin the log item */
-		ASSERT(!lv->lv_item->li_lv);
-
-		if (lv->lv_buf_len != XFS_LOG_VEC_ORDERED) {
-			*len += lv->lv_buf_len;
-			*diff_iovecs += lv->lv_niovecs;
-		}
-		IOP_PIN(lv->lv_item);
-
+		/* re-used lv */
+		/* XXX: can't account for len/diff_iovecs yet */
 	}
 
 	/* attach new log vector to log item */
@@ -733,18 +752,13 @@ xfs_log_commit_cil(
 	if (flags & XFS_TRANS_RELEASE_LOG_RES)
 		log_flags = XFS_LOG_REL_PERM_RESERV;
 
-	/*
-	 * Do all the hard work of formatting items (including memory
-	 * allocation) outside the CIL context lock. This prevents stalling CIL
-	 * pushes when we are low on memory and a transaction commit spends a
-	 * lot of time in memory reclaim.
-	 */
+	/* lock out background commit */
+	down_read(&log->l_cilp->xc_ctx_lock);
+
 	log_vector = xlog_cil_prepare_log_vecs(tp);
 	if (!log_vector)
 		return ENOMEM;
 
-	/* lock out background commit */
-	down_read(&log->l_cilp->xc_ctx_lock);
 	if (commit_lsn)
 		*commit_lsn = log->l_cilp->xc_ctx->sequence;
 
diff --git a/fs/xfs/xfs_trans.h b/fs/xfs/xfs_trans.h
index 24bbdcd..4786170 100644
--- a/fs/xfs/xfs_trans.h
+++ b/fs/xfs/xfs_trans.h
@@ -77,7 +77,6 @@ struct xfs_item_ops {
 	void (*iop_committing)(xfs_log_item_t *, xfs_lsn_t);
 };
 
-#define IOP_PIN(ip)		(*(ip)->li_ops->iop_pin)(ip)
 #define IOP_UNPIN(ip, remove)	(*(ip)->li_ops->iop_unpin)(ip, remove)
 #define IOP_PUSH(ip, list)	(*(ip)->li_ops->iop_push)(ip, list)
 #define IOP_UNLOCK(ip)		(*(ip)->li_ops->iop_unlock)(ip)
-- 
1.8.3.2

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

  parent reply	other threads:[~2013-08-12 10:50 UTC|newest]

Thread overview: 86+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-08-12 10:49 ***** SUSPECTED SPAM ***** [PATCH 00/50] xfs: patches for 3.12 Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 01/50] xfs: separate out log format definitions Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 02/50] xfs: split out inode log item format definition Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 03/50] xfs: split out buf log item format definitions Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 04/50] xfs: split out EFI/EFD log item format definition Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 05/50] xfs: separate dquot on disk format definitions out of xfs_quota.h Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 06/50] xfs: separate icreate log format definitions from xfs_icreate_item.h Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 07/50] xfs: split out on-disk transaction definitions Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 08/50] xfs: introduce xfs_rtalloc_defs.h Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 09/50] xfs: introduce xfs_quota_defs.h Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 10/50] xfs: sync minor header differences needed by userspace Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 11/50] xfs: split out transaction reservation code Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 12/50] xfs: move inode fork definitions to a new header file Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 13/50] xfs: move unrelated definitions out of xfs_inode.h Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 14/50] xfs: introduce xfs_inode_buf.c for inode buffer operations Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 15/50] xfs: move getdents code into it's own file Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 16/50] xfs: reshuffle dir2 definitions around for userspace Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 17/50] xfs: split out attribute listing code into separate file Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 18/50] xfs: split out attribute fork truncation " Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 19/50] xfs: split out the remote symlink handling Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 20/50] xfs: introduce xfs_sb.c for sharing with libxfs Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 21/50] xfs: create xfs_bmap_util.[ch] Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 22/50] xfs: minor cleanups Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 23/50] xfs: fix issues that cause userspace warnings Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 24/50] xfs: kill xfs_vnodeops.[ch] Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 25/50] xfs: consolidate xfs_rename.c Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 26/50] xfs: consolidate xfs_utils.c Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 27/50] xfs: consolidate extent swap code Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 28/50] xfs: don't special case shared superblock mounts Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 29/50] xfs: kill __KERNEL__ check for debug code in allocation code Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 30/50] xfs: remove __KERNEL__ from debug code Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 31/50] xfs: remove __KERNEL__ check from xfs_dir2_leaf.c Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 32/50] xfs: xfs_filestreams.h doesn't need __KERNEL__ Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 33/50] xfs: move kernel specific type definitions to xfs.h Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 34/50] xfs: make struct xfs_perag kernel only Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 35/50] xfs: Introduce a new structure to hold transaction reservation items Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 36/50] xfs: Introduce tr_fsyncts to m_reservation Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 37/50] xfs: Make writeid transaction use tr_writeid Dave Chinner
2013-08-12 10:49 ` ***** SUSPECTED SPAM ***** [PATCH 38/50] xfs: refactor xfs_trans_reserve() interface Dave Chinner
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 39/50] xfs: Get rid of all XFS_XXX_LOG_RES() macro Dave Chinner
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 40/50] xfs: Refactor xfs_ticket_alloc() to extract a new helper Dave Chinner
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 41/50] xfs: Add xfs_log_rlimit.c Dave Chinner
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 42/50] xfs: Validate log space at mount time Dave Chinner
2013-08-12 18:46   ` Mark Tinguely
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 43/50] xfs: return log item size in IOP_SIZE Dave Chinner
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 44/50] xfs: Reduce allocations during CIL insertion Dave Chinner
2013-08-13 13:28   ` Mark Tinguely
2013-08-12 10:50 ` Dave Chinner [this message]
2013-08-13 13:37   ` [PATCH 45/50] xfs: avoid CIL allocation during insert Mark Tinguely
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 46/50] xfs: Combine CIL insert and prepare passes Dave Chinner
2013-08-13 14:02   ` Mark Tinguely
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 47/50] xfs: split the CIL lock Dave Chinner
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 48/50] xfs: Add read-only support for dirent filetype field Dave Chinner
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 49/50] xfs: Add write " Dave Chinner
2013-08-12 10:50 ` ***** SUSPECTED SPAM ***** [PATCH 50/50] xfs: use reference counts to free clean buffer items Dave Chinner
2013-08-13 15:03   ` Mark Tinguely
2013-08-13 21:46     ` Dave Chinner
2013-08-13 22:00       ` Mark Tinguely
2013-08-14  3:57         ` Dave Chinner
2013-08-14  4:12           ` Zhi Yong Wu
2013-08-14  6:41             ` Dave Chinner
2013-08-14 13:26           ` Mark Tinguely
2013-08-14 17:49             ` Mark Tinguely
2013-08-15  0:48               ` Dave Chinner
2013-08-15 21:43   ` ***** SUSPECTED SPAM ***** " Ben Myers
2013-08-12 22:55 ` ***** SUSPECTED SPAM ***** [PATCH 00/50] xfs: patches for 3.12 Ben Myers
2013-08-13 21:28   ` Ben Myers
2013-08-19 20:19 ` [PATCH 51/50] xfs: add xfs sb v4 support for dirent filetype field Mark Tinguely
2013-08-19 23:28   ` Eric Sandeen
2013-08-20 14:29     ` Mark Tinguely
2013-08-20 14:45       ` Eric Sandeen
2013-08-20 18:50         ` Ben Myers
2013-08-20 21:00           ` Mark Tinguely
2013-08-20 21:05             ` Ben Myers
2013-08-20 23:19       ` Dave Chinner
2013-08-21  0:06       ` Dave Chinner
2013-08-21 17:03         ` Ben Myers
2013-08-22  2:02           ` Dave Chinner
2013-08-22 16:14             ` Geoffrey Wehrman
2013-08-22 18:19               ` Ben Myers
2013-08-25  5:18                 ` Michael L. Semon
2013-08-25 23:21                   ` Michael L. Semon
2013-08-26 15:40                   ` Mark Tinguely
2013-08-19 23:40   ` Dave Chinner
2013-08-20 19:57   ` Geoffrey Wehrman
2013-08-22 15:59 ` ***** SUSPECTED SPAM ***** [PATCH 00/50] xfs: patches for 3.12 Ben Myers

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=1376304611-22994-46-git-send-email-david@fromorbit.com \
    --to=david@fromorbit.com \
    --cc=xfs@oss.sgi.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