From: Wu Fengguang <fengguang.wu@intel.com>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Theodore Tso <tytso@mit.edu>,
Christoph Hellwig <hch@infradead.org>,
Dave Chinner <david@fromorbit.com>,
Chris Mason <chris.mason@oracle.com>,
Peter Zijlstra <a.p.zijlstra@chello.nl>,
"Li Shaohua" <shaohua.li@intel.com>,
"Myklebust Trond" <Trond.Myklebust@netapp.com>,
"jens.axboe@oracle.com" <jens.axboe@oracle.com>,
Jan Kara <jack@suse.cz>, Nick Piggin <npiggin@suse.de>,
<linux-fsdevel@vger.kernel.org>,
Wu Fengguang <fengguang.wu@intel.com>
Cc: LKML <linux-kernel@vger.kernel.org>
Subject: [PATCH 21/45] writeback: estimate bdi write bandwidth
Date: Wed, 07 Oct 2009 15:38:39 +0800 [thread overview]
Message-ID: <20091007074903.867171763@intel.com> (raw)
In-Reply-To: 20091007073818.318088777@intel.com
[-- Attachment #1: writeback-throughput-estimation.patch --]
[-- Type: text/plain, Size: 5410 bytes --]
Estimate bdi write bandwidth in bdi_writeback_wakeup(), at which time
the queue is not starved due to the dirtying process and the associated
background writeback.
The estimation should be able to reflect the max device capability,
unless there are busy reads, in which case we need lower nr_to_write
anyway.
CC: Theodore Ts'o <tytso@mit.edu>
CC: Jan Kara <jack@suse.cz>
CC: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
---
TODO: the estimated write bandwidth (~30MB/s) is mysteriously only half the real
throughput (~60MB/s). Here are some debug printk from bdi_calc_write_bandwidth():
printk("write_bandwidth: comm=%s pages=%lu time=%lums\n",
current->comm, nr_pages, time * 1000 / HZ);
[ 1093.397700] write_bandwidth: comm=swapper pages=1536 time=204ms
[ 1093.594319] write_bandwidth: comm=swapper pages=1536 time=196ms
[ 1093.796642] write_bandwidth: comm=swapper pages=1536 time=200ms
[ 1093.986128] write_bandwidth: comm=swapper pages=1536 time=192ms
[ 1094.179983] write_bandwidth: comm=swapper pages=1536 time=192ms
[ 1094.374021] write_bandwidth: comm=swapper pages=1536 time=196ms
[ 1094.570611] write_bandwidth: comm=swapper pages=1536 time=196ms
[ 1094.771847] write_bandwidth: comm=swapper pages=1536 time=200ms
[ 1094.961981] write_bandwidth: comm=swapper pages=1536 time=192ms
Workload is several concurrent copies.
fs/fs-writeback.c | 30 +++++++++++++++++++++---------
include/linux/backing-dev.h | 2 ++
include/linux/writeback.h | 10 ++++++++++
mm/backing-dev.c | 2 ++
4 files changed, 35 insertions(+), 9 deletions(-)
--- linux.orig/fs/fs-writeback.c 2009-10-06 23:38:43.000000000 +0800
+++ linux/fs/fs-writeback.c 2009-10-06 23:38:44.000000000 +0800
@@ -266,7 +266,23 @@ void bdi_start_writeback(struct backing_
bdi_alloc_queue_work(bdi, &args);
}
+static int bdi_writeback_chunk(struct backing_dev_info *bdi)
+{
+ return max(MIN_WRITEBACK_PAGES, bdi->write_bandwidth);
+}
+
+static void bdi_calc_write_bandwidth(struct backing_dev_info *bdi,
+ unsigned long nr_pages,
+ unsigned long time)
+{
+ unsigned long bw;
+
+ bw = HZ * nr_pages / (time | 1);
+ bdi->write_bandwidth = (bdi->write_bandwidth * 63 + bw) / 64;
+}
+
struct dirty_throttle_task {
+ unsigned long start_time;
long nr_pages;
struct list_head list;
struct completion complete;
@@ -275,6 +291,7 @@ struct dirty_throttle_task {
void bdi_writeback_wait(struct backing_dev_info *bdi, long nr_pages)
{
struct dirty_throttle_task tt = {
+ .start_time = jiffies,
.nr_pages = nr_pages,
.complete = COMPLETION_INITIALIZER_ONSTACK(tt.complete),
};
@@ -314,6 +331,9 @@ int bdi_writeback_wakeup(struct backing_
tt = list_entry(bdi->throttle_list.prev,
struct dirty_throttle_task, list);
list_del(&tt->list);
+ if (atomic_read(&bdi->throttle_pages) == 0)
+ bdi_calc_write_bandwidth(bdi, tt->nr_pages,
+ jiffies - tt->start_time);
complete(&tt->complete);
}
/*
@@ -323,6 +343,7 @@ int bdi_writeback_wakeup(struct backing_
tt = list_entry(bdi->throttle_list.prev,
struct dirty_throttle_task, list);
atomic_set(&bdi->throttle_pages, tt->nr_pages);
+ tt->start_time = jiffies;
} else {
tt = NULL;
atomic_set(&bdi->throttle_pages, DIRTY_THROTTLE_PAGES_STOP * 2);
@@ -717,15 +738,6 @@ void writeback_inodes_wbc(struct writeba
writeback_inodes_wb(&bdi->wb, wbc);
}
-/*
- * The maximum number of pages to writeout in a single bdi flush/kupdate
- * operation. We do this so we don't hold I_SYNC against an inode for
- * enormous amounts of time, which would block a userspace task which has
- * been forced to throttle against that inode. Also, the code reevaluates
- * the dirty each time it has written this many pages.
- */
-#define MAX_WRITEBACK_PAGES 1024
-
static inline bool over_bground_thresh(void)
{
unsigned long background_thresh, dirty_thresh;
--- linux.orig/include/linux/writeback.h 2009-10-06 23:37:46.000000000 +0800
+++ linux/include/linux/writeback.h 2009-10-06 23:38:44.000000000 +0800
@@ -14,6 +14,16 @@ extern struct list_head inode_in_use;
extern struct list_head inode_unused;
/*
+ * The max number of pages to writeout for each inode.
+ *
+ * We honor each inode a nr_to_write that will take about 1 second
+ * to finish, based on dynamic estimation of the bdi's write bandwidth.
+ * MAX_ serves as initial bandwidth value; MIN_ serves as low boundary.
+ */
+#define MAX_WRITEBACK_PAGES (128 << (20 - PAGE_CACHE_SHIFT))
+#define MIN_WRITEBACK_PAGES ( 16 << (20 - PAGE_CACHE_SHIFT))
+
+/*
* fs/fs-writeback.c
*/
enum writeback_sync_modes {
--- linux.orig/include/linux/backing-dev.h 2009-10-06 23:38:43.000000000 +0800
+++ linux/include/linux/backing-dev.h 2009-10-06 23:38:44.000000000 +0800
@@ -86,6 +86,8 @@ struct backing_dev_info {
struct list_head work_list;
+ int write_bandwidth; /* pages per second */
+
/*
* dirtier process throttling
*/
--- linux.orig/mm/backing-dev.c 2009-10-06 23:38:43.000000000 +0800
+++ linux/mm/backing-dev.c 2009-10-06 23:38:44.000000000 +0800
@@ -646,6 +646,8 @@ int bdi_init(struct backing_dev_info *bd
bdi->wb_mask = 1;
bdi->wb_cnt = 1;
+ bdi->write_bandwidth = MAX_WRITEBACK_PAGES;
+
spin_lock_init(&bdi->throttle_lock);
INIT_LIST_HEAD(&bdi->throttle_list);
atomic_set(&bdi->throttle_pages, DIRTY_THROTTLE_PAGES_STOP * 2);
next prev parent reply other threads:[~2009-10-07 7:38 UTC|newest]
Thread overview: 116+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-10-07 7:38 [PATCH 00/45] some writeback experiments Wu Fengguang
2009-10-07 7:38 ` [PATCH 01/45] writeback: reduce calls to global_page_state in balance_dirty_pages() Wu Fengguang
2009-10-09 15:12 ` Jan Kara
2009-10-09 15:18 ` Peter Zijlstra
2009-10-09 15:47 ` Jan Kara
2009-10-11 2:28 ` Wu Fengguang
2009-10-11 7:44 ` Peter Zijlstra
2009-10-11 10:50 ` Wu Fengguang
2009-10-11 10:58 ` Peter Zijlstra
2009-10-11 11:25 ` Peter Zijlstra
2009-10-12 1:26 ` Wu Fengguang
2009-10-12 9:07 ` Peter Zijlstra
2009-10-12 9:24 ` Wu Fengguang
2009-10-10 21:33 ` Wu Fengguang
2009-10-12 21:18 ` Jan Kara
2009-10-13 3:24 ` Wu Fengguang
2009-10-13 8:41 ` Peter Zijlstra
2009-10-13 18:12 ` Jan Kara
2009-10-13 18:28 ` Peter Zijlstra
2009-10-14 1:38 ` Wu Fengguang
2009-10-14 11:22 ` Peter Zijlstra
2009-10-17 5:30 ` Wu Fengguang
2009-10-07 7:38 ` [PATCH 02/45] writeback: reduce calculation of bdi dirty thresholds Wu Fengguang
2009-10-07 7:38 ` [PATCH 03/45] ext4: remove unused parameter wbc from __ext4_journalled_writepage() Wu Fengguang
2009-10-07 7:38 ` [PATCH 04/45] writeback: remove unused nonblocking and congestion checks Wu Fengguang
2009-10-09 15:26 ` Jan Kara
2009-10-10 13:47 ` Wu Fengguang
2009-10-07 7:38 ` [PATCH 05/45] writeback: remove the always false bdi_cap_writeback_dirty() test Wu Fengguang
2009-10-07 7:38 ` [PATCH 06/45] writeback: use larger ratelimit when dirty_exceeded Wu Fengguang
2009-10-07 8:53 ` Peter Zijlstra
2009-10-07 9:17 ` Wu Fengguang
2009-10-07 7:38 ` [PATCH 07/45] writeback: dont redirty tail an inode with dirty pages Wu Fengguang
2009-10-09 15:45 ` Jan Kara
2009-10-07 7:38 ` [PATCH 08/45] writeback: quit on wrap for .range_cyclic (write_cache_pages) Wu Fengguang
2009-10-07 7:38 ` [PATCH 09/45] writeback: quit on wrap for .range_cyclic (pohmelfs) Wu Fengguang
2009-10-07 12:32 ` Evgeniy Polyakov
2009-10-07 14:23 ` Wu Fengguang
2009-10-07 7:38 ` [PATCH 10/45] writeback: quit on wrap for .range_cyclic (btrfs) Wu Fengguang
2009-10-07 7:38 ` [PATCH 11/45] writeback: quit on wrap for .range_cyclic (cifs) Wu Fengguang
2009-10-07 7:38 ` [PATCH 12/45] writeback: quit on wrap for .range_cyclic (ext4) Wu Fengguang
2009-10-07 7:38 ` [PATCH 13/45] writeback: quit on wrap for .range_cyclic (gfs2) Wu Fengguang
2009-10-07 7:38 ` [PATCH 14/45] writeback: quit on wrap for .range_cyclic (afs) Wu Fengguang
2009-10-07 7:38 ` [PATCH 15/45] writeback: fix queue_io() ordering Wu Fengguang
2009-10-07 7:38 ` [PATCH 16/45] writeback: merge for_kupdate and !for_kupdate cases Wu Fengguang
2009-10-07 7:38 ` [PATCH 17/45] writeback: only allow two background writeback works Wu Fengguang
2009-10-07 7:38 ` [PATCH 18/45] writeback: introduce wait queue for balance_dirty_pages() Wu Fengguang
2009-10-08 1:01 ` KAMEZAWA Hiroyuki
2009-10-08 1:58 ` Wu Fengguang
2009-10-08 2:40 ` KAMEZAWA Hiroyuki
2009-10-08 4:01 ` Wu Fengguang
2009-10-08 5:59 ` KAMEZAWA Hiroyuki
2009-10-08 6:07 ` Wu Fengguang
2009-10-08 6:28 ` Wu Fengguang
2009-10-08 6:39 ` KAMEZAWA Hiroyuki
2009-10-08 8:08 ` Peter Zijlstra
2009-10-08 8:11 ` KAMEZAWA Hiroyuki
2009-10-08 8:36 ` Jens Axboe
2009-10-09 2:52 ` [PATCH] writeback: account IO throttling wait as iowait Wu Fengguang
2009-10-09 10:41 ` Jens Axboe
2009-10-09 10:58 ` Wu Fengguang
2009-10-09 11:01 ` Jens Axboe
2009-10-08 8:05 ` [PATCH 18/45] writeback: introduce wait queue for balance_dirty_pages() Peter Zijlstra
2009-10-07 7:38 ` [PATCH 19/45] writeback: remove the loop in balance_dirty_pages() Wu Fengguang
2009-10-07 7:38 ` [PATCH 20/45] NFS: introduce writeback wait queue Wu Fengguang
2009-10-07 8:53 ` Peter Zijlstra
2009-10-07 9:07 ` Wu Fengguang
2009-10-07 9:15 ` Peter Zijlstra
2009-10-07 9:19 ` Wu Fengguang
2009-10-07 9:17 ` Nick Piggin
2009-10-07 9:52 ` Wu Fengguang
2009-10-07 7:38 ` Wu Fengguang [this message]
2009-10-07 8:53 ` [PATCH 21/45] writeback: estimate bdi write bandwidth Peter Zijlstra
2009-10-07 9:39 ` Wu Fengguang
2009-10-07 7:38 ` [PATCH 22/45] writeback: show bdi write bandwidth in debugfs Wu Fengguang
2009-10-07 7:38 ` [PATCH 23/45] writeback: kill space in debugfs item name Wu Fengguang
2009-10-07 7:38 ` [PATCH 24/45] writeback: remove global nr_to_write and use timeout instead Wu Fengguang
2009-10-07 7:38 ` [PATCH 25/45] writeback: convert wbc.nr_to_write to per-file parameter Wu Fengguang
2009-10-07 7:38 ` [PATCH 26/45] block: pass the non-rotational queue flag to backing_dev_info Wu Fengguang
2009-10-07 7:38 ` [PATCH 27/45] writeback: introduce wbc.for_background Wu Fengguang
2009-10-07 7:38 ` [PATCH 28/45] writeback: introduce wbc.nr_segments Wu Fengguang
2009-10-07 7:38 ` [PATCH 29/45] writeback: fix the shmem AOP_WRITEPAGE_ACTIVATE case Wu Fengguang
2009-10-07 11:57 ` Hugh Dickins
2009-10-07 14:00 ` Wu Fengguang
2009-10-07 7:38 ` [PATCH 30/45] vmscan: lumpy pageout Wu Fengguang
2009-10-07 7:38 ` [PATCH 31/45] writeback: sync old inodes first in background writeback Wu Fengguang
2010-07-12 3:01 ` Christoph Hellwig
2010-07-12 15:24 ` Wu Fengguang
2009-10-07 7:38 ` [PATCH 32/45] writeback: update kupdate expire timestamp on each scan of b_io Wu Fengguang
2009-10-07 7:38 ` [PATCH 34/45] writeback: sync livelock - kick background writeback Wu Fengguang
2009-10-07 7:38 ` [PATCH 35/45] writeback: sync livelock - use single timestamp for whole sync work Wu Fengguang
2009-10-07 7:38 ` [PATCH 36/45] writeback: sync livelock - curb dirty speed for inodes to be synced Wu Fengguang
2009-10-07 7:38 ` [PATCH 37/45] writeback: use timestamp to indicate dirty exceeded Wu Fengguang
2009-10-07 7:38 ` [PATCH 38/45] writeback: introduce queue b_more_io_wait Wu Fengguang
2009-10-07 7:38 ` [PATCH 39/45] writeback: remove wbc.more_io Wu Fengguang
2009-10-07 7:38 ` [PATCH 40/45] writeback: requeue_io_wait() on I_SYNC locked inode Wu Fengguang
2009-10-07 7:38 ` [PATCH 41/45] writeback: requeue_io_wait() on pages_skipped inode Wu Fengguang
2009-10-07 7:39 ` [PATCH 42/45] writeback: requeue_io_wait() on blocked inode Wu Fengguang
2009-10-07 7:39 ` [PATCH 43/45] writeback: requeue_io_wait() on fs redirtied inode Wu Fengguang
2009-10-07 7:39 ` [PATCH 44/45] NFS: remove NFS_INO_FLUSHING lock Wu Fengguang
2009-10-07 13:11 ` Peter Staubach
2009-10-07 13:32 ` Wu Fengguang
2009-10-07 13:59 ` Peter Staubach
2009-10-08 1:44 ` Wu Fengguang
2009-10-07 7:39 ` [PATCH 45/45] btrfs: fix race on syncing the btree inode Wu Fengguang
2009-10-07 8:53 ` [PATCH 00/45] some writeback experiments Peter Zijlstra
2009-10-07 10:17 ` [PATCH 14/45] writeback: quit on wrap for .range_cyclic (afs) David Howells
2009-10-07 10:21 ` Nick Piggin
2009-10-07 10:47 ` Wu Fengguang
2009-10-07 11:23 ` Nick Piggin
2009-10-07 12:21 ` Wu Fengguang
2009-10-07 13:47 ` [PATCH 00/45] some writeback experiments Peter Staubach
2009-10-07 15:18 ` Wu Fengguang
2009-10-08 5:33 ` Wu Fengguang
2009-10-08 5:44 ` Wu Fengguang
2009-10-07 14:26 ` Theodore Tso
2009-10-07 14:45 ` Wu Fengguang
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=20091007074903.867171763@intel.com \
--to=fengguang.wu@intel.com \
--cc=Trond.Myklebust@netapp.com \
--cc=a.p.zijlstra@chello.nl \
--cc=akpm@linux-foundation.org \
--cc=chris.mason@oracle.com \
--cc=david@fromorbit.com \
--cc=hch@infradead.org \
--cc=jack@suse.cz \
--cc=jens.axboe@oracle.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=npiggin@suse.de \
--cc=shaohua.li@intel.com \
--cc=tytso@mit.edu \
/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).