* [patch] Use BIO_META tags for metadata rather than BIO_SYNC
@ 2008-01-21 4:18 David Chinner
2008-01-22 3:50 ` Lachlan McIlroy
2008-07-23 3:49 ` Eric Sandeen
0 siblings, 2 replies; 3+ messages in thread
From: David Chinner @ 2008-01-21 4:18 UTC (permalink / raw)
To: xfs-dev; +Cc: xfs-oss
Improve metadata I/O merging in the elevator
Change all async metadata buffers to use [READ|WRITE]_META I/O types
so that the I/O doesn't get issued immediately. This allows merging
of adjacent metadata requests but still prioritises them over bulk
data. This shows a 10-15% improvement in sequential create speed of
small files.
Don't include the log buffers in this classification - leave them
as sync types so they are issued immediately.
Signed-off-by: Dave Chinner <dgc@sgi.com>
---
fs/xfs/linux-2.6/xfs_buf.c | 6 +++++-
include/linux/fs.h | 1 +
2 files changed, 6 insertions(+), 1 deletion(-)
Index: 2.6.x-xfs-new/fs/xfs/linux-2.6/xfs_buf.c
===================================================================
--- 2.6.x-xfs-new.orig/fs/xfs/linux-2.6/xfs_buf.c 2007-12-19 11:56:47.977477993 +1100
+++ 2.6.x-xfs-new/fs/xfs/linux-2.6/xfs_buf.c 2007-12-19 11:56:52.200932303 +1100
@@ -1136,10 +1136,14 @@ _xfs_buf_ioapply(
if (bp->b_flags & XBF_ORDERED) {
ASSERT(!(bp->b_flags & XBF_READ));
rw = WRITE_BARRIER;
- } else if (bp->b_flags & _XBF_RUN_QUEUES) {
+ } else if (bp->b_flags & XBF_LOG_BUFFER) {
ASSERT(!(bp->b_flags & XBF_READ_AHEAD));
bp->b_flags &= ~_XBF_RUN_QUEUES;
rw = (bp->b_flags & XBF_WRITE) ? WRITE_SYNC : READ_SYNC;
+ } else if (bp->b_flags & _XBF_RUN_QUEUES) {
+ ASSERT(!(bp->b_flags & XBF_READ_AHEAD));
+ bp->b_flags &= ~_XBF_RUN_QUEUES;
+ rw = (bp->b_flags & XBF_WRITE) ? WRITE_META : READ_META;
} else {
rw = (bp->b_flags & XBF_WRITE) ? WRITE :
(bp->b_flags & XBF_READ_AHEAD) ? READA : READ;
Index: 2.6.x-xfs-new/include/linux/fs.h
===================================================================
--- 2.6.x-xfs-new.orig/include/linux/fs.h 2007-12-19 01:11:16.495229881 +1100
+++ 2.6.x-xfs-new/include/linux/fs.h 2007-12-19 11:56:52.352912667 +1100
@@ -83,6 +83,7 @@ extern int dir_notify_enable;
#define READ_SYNC (READ | (1 << BIO_RW_SYNC))
#define READ_META (READ | (1 << BIO_RW_META))
#define WRITE_SYNC (WRITE | (1 << BIO_RW_SYNC))
+#define WRITE_META (WRITE | (1 << BIO_RW_META))
#define WRITE_BARRIER ((1 << BIO_RW) | (1 << BIO_RW_BARRIER))
#define SEL_IN 1
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [patch] Use BIO_META tags for metadata rather than BIO_SYNC
2008-01-21 4:18 [patch] Use BIO_META tags for metadata rather than BIO_SYNC David Chinner
@ 2008-01-22 3:50 ` Lachlan McIlroy
2008-07-23 3:49 ` Eric Sandeen
1 sibling, 0 replies; 3+ messages in thread
From: Lachlan McIlroy @ 2008-01-22 3:50 UTC (permalink / raw)
To: David Chinner; +Cc: xfs-dev, xfs-oss
Looks good.
David Chinner wrote:
> Improve metadata I/O merging in the elevator
>
> Change all async metadata buffers to use [READ|WRITE]_META I/O types
> so that the I/O doesn't get issued immediately. This allows merging
> of adjacent metadata requests but still prioritises them over bulk
> data. This shows a 10-15% improvement in sequential create speed of
> small files.
>
> Don't include the log buffers in this classification - leave them
> as sync types so they are issued immediately.
>
> Signed-off-by: Dave Chinner <dgc@sgi.com>
> ---
> fs/xfs/linux-2.6/xfs_buf.c | 6 +++++-
> include/linux/fs.h | 1 +
> 2 files changed, 6 insertions(+), 1 deletion(-)
>
> Index: 2.6.x-xfs-new/fs/xfs/linux-2.6/xfs_buf.c
> ===================================================================
> --- 2.6.x-xfs-new.orig/fs/xfs/linux-2.6/xfs_buf.c 2007-12-19 11:56:47.977477993 +1100
> +++ 2.6.x-xfs-new/fs/xfs/linux-2.6/xfs_buf.c 2007-12-19 11:56:52.200932303 +1100
> @@ -1136,10 +1136,14 @@ _xfs_buf_ioapply(
> if (bp->b_flags & XBF_ORDERED) {
> ASSERT(!(bp->b_flags & XBF_READ));
> rw = WRITE_BARRIER;
> - } else if (bp->b_flags & _XBF_RUN_QUEUES) {
> + } else if (bp->b_flags & XBF_LOG_BUFFER) {
> ASSERT(!(bp->b_flags & XBF_READ_AHEAD));
> bp->b_flags &= ~_XBF_RUN_QUEUES;
> rw = (bp->b_flags & XBF_WRITE) ? WRITE_SYNC : READ_SYNC;
> + } else if (bp->b_flags & _XBF_RUN_QUEUES) {
> + ASSERT(!(bp->b_flags & XBF_READ_AHEAD));
> + bp->b_flags &= ~_XBF_RUN_QUEUES;
> + rw = (bp->b_flags & XBF_WRITE) ? WRITE_META : READ_META;
> } else {
> rw = (bp->b_flags & XBF_WRITE) ? WRITE :
> (bp->b_flags & XBF_READ_AHEAD) ? READA : READ;
> Index: 2.6.x-xfs-new/include/linux/fs.h
> ===================================================================
> --- 2.6.x-xfs-new.orig/include/linux/fs.h 2007-12-19 01:11:16.495229881 +1100
> +++ 2.6.x-xfs-new/include/linux/fs.h 2007-12-19 11:56:52.352912667 +1100
> @@ -83,6 +83,7 @@ extern int dir_notify_enable;
> #define READ_SYNC (READ | (1 << BIO_RW_SYNC))
> #define READ_META (READ | (1 << BIO_RW_META))
> #define WRITE_SYNC (WRITE | (1 << BIO_RW_SYNC))
> +#define WRITE_META (WRITE | (1 << BIO_RW_META))
> #define WRITE_BARRIER ((1 << BIO_RW) | (1 << BIO_RW_BARRIER))
>
> #define SEL_IN 1
>
>
>
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [patch] Use BIO_META tags for metadata rather than BIO_SYNC
2008-01-21 4:18 [patch] Use BIO_META tags for metadata rather than BIO_SYNC David Chinner
2008-01-22 3:50 ` Lachlan McIlroy
@ 2008-07-23 3:49 ` Eric Sandeen
1 sibling, 0 replies; 3+ messages in thread
From: Eric Sandeen @ 2008-07-23 3:49 UTC (permalink / raw)
To: David Chinner; +Cc: xfs-dev, xfs-oss
David Chinner wrote:
> Improve metadata I/O merging in the elevator
>
> Change all async metadata buffers to use [READ|WRITE]_META I/O types
> so that the I/O doesn't get issued immediately. This allows merging
> of adjacent metadata requests but still prioritises them over bulk
> data. This shows a 10-15% improvement in sequential create speed of
> small files.
That sounds awesome. I'll test it on a similar workload I've been
playing with. Lachlan reviewed this, I think - any plans to add it?
I'll run my test & post results.
Thanks,
-Eric
> Don't include the log buffers in this classification - leave them
> as sync types so they are issued immediately.
>
> Signed-off-by: Dave Chinner <dgc@sgi.com>
> ---
> fs/xfs/linux-2.6/xfs_buf.c | 6 +++++-
> include/linux/fs.h | 1 +
> 2 files changed, 6 insertions(+), 1 deletion(-)
>
> Index: 2.6.x-xfs-new/fs/xfs/linux-2.6/xfs_buf.c
> ===================================================================
> --- 2.6.x-xfs-new.orig/fs/xfs/linux-2.6/xfs_buf.c 2007-12-19 11:56:47.977477993 +1100
> +++ 2.6.x-xfs-new/fs/xfs/linux-2.6/xfs_buf.c 2007-12-19 11:56:52.200932303 +1100
> @@ -1136,10 +1136,14 @@ _xfs_buf_ioapply(
> if (bp->b_flags & XBF_ORDERED) {
> ASSERT(!(bp->b_flags & XBF_READ));
> rw = WRITE_BARRIER;
> - } else if (bp->b_flags & _XBF_RUN_QUEUES) {
> + } else if (bp->b_flags & XBF_LOG_BUFFER) {
> ASSERT(!(bp->b_flags & XBF_READ_AHEAD));
> bp->b_flags &= ~_XBF_RUN_QUEUES;
> rw = (bp->b_flags & XBF_WRITE) ? WRITE_SYNC : READ_SYNC;
> + } else if (bp->b_flags & _XBF_RUN_QUEUES) {
> + ASSERT(!(bp->b_flags & XBF_READ_AHEAD));
> + bp->b_flags &= ~_XBF_RUN_QUEUES;
> + rw = (bp->b_flags & XBF_WRITE) ? WRITE_META : READ_META;
> } else {
> rw = (bp->b_flags & XBF_WRITE) ? WRITE :
> (bp->b_flags & XBF_READ_AHEAD) ? READA : READ;
> Index: 2.6.x-xfs-new/include/linux/fs.h
> ===================================================================
> --- 2.6.x-xfs-new.orig/include/linux/fs.h 2007-12-19 01:11:16.495229881 +1100
> +++ 2.6.x-xfs-new/include/linux/fs.h 2007-12-19 11:56:52.352912667 +1100
> @@ -83,6 +83,7 @@ extern int dir_notify_enable;
> #define READ_SYNC (READ | (1 << BIO_RW_SYNC))
> #define READ_META (READ | (1 << BIO_RW_META))
> #define WRITE_SYNC (WRITE | (1 << BIO_RW_SYNC))
> +#define WRITE_META (WRITE | (1 << BIO_RW_META))
> #define WRITE_BARRIER ((1 << BIO_RW) | (1 << BIO_RW_BARRIER))
>
> #define SEL_IN 1
>
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2008-07-23 3:48 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-01-21 4:18 [patch] Use BIO_META tags for metadata rather than BIO_SYNC David Chinner
2008-01-22 3:50 ` Lachlan McIlroy
2008-07-23 3:49 ` Eric Sandeen
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox