2006-10-20 00:28:16 -06:00
|
|
|
|
|
|
|
#include <linux/wait.h>
|
|
|
|
#include <linux/backing-dev.h>
|
2009-09-09 01:08:54 -06:00
|
|
|
#include <linux/kthread.h>
|
|
|
|
#include <linux/freezer.h>
|
2006-10-20 00:28:16 -06:00
|
|
|
#include <linux/fs.h>
|
2009-03-17 02:35:06 -06:00
|
|
|
#include <linux/pagemap.h>
|
2009-09-09 01:08:54 -06:00
|
|
|
#include <linux/mm.h>
|
2006-10-20 00:28:16 -06:00
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/module.h>
|
2008-04-30 01:54:32 -06:00
|
|
|
#include <linux/writeback.h>
|
|
|
|
#include <linux/device.h>
|
2010-07-06 21:24:06 -06:00
|
|
|
#include <trace/events/writeback.h>
|
2008-04-30 01:54:32 -06:00
|
|
|
|
2010-04-22 03:37:01 -06:00
|
|
|
static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
|
|
|
|
|
2009-03-17 02:35:06 -06:00
|
|
|
struct backing_dev_info default_backing_dev_info = {
|
2009-06-12 06:45:52 -06:00
|
|
|
.name = "default",
|
2009-03-17 02:35:06 -06:00
|
|
|
.ra_pages = VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE,
|
|
|
|
.state = 0,
|
|
|
|
.capabilities = BDI_CAP_MAP_COPY,
|
|
|
|
};
|
|
|
|
EXPORT_SYMBOL_GPL(default_backing_dev_info);
|
2008-04-30 01:54:32 -06:00
|
|
|
|
2010-04-25 00:54:42 -06:00
|
|
|
struct backing_dev_info noop_backing_dev_info = {
|
|
|
|
.name = "noop",
|
2010-09-21 03:48:55 -06:00
|
|
|
.capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK,
|
2010-04-25 00:54:42 -06:00
|
|
|
};
|
|
|
|
EXPORT_SYMBOL_GPL(noop_backing_dev_info);
|
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
static struct class *bdi_class;
|
2009-09-14 05:12:40 -06:00
|
|
|
|
|
|
|
/*
|
2013-04-01 20:08:06 -06:00
|
|
|
* bdi_lock protects updates to bdi_list. bdi_list has RCU reader side
|
2009-09-14 05:12:40 -06:00
|
|
|
* locking.
|
|
|
|
*/
|
2009-09-09 01:08:54 -06:00
|
|
|
DEFINE_SPINLOCK(bdi_lock);
|
2009-09-02 01:19:46 -06:00
|
|
|
LIST_HEAD(bdi_list);
|
2009-09-09 01:08:54 -06:00
|
|
|
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
/* bdi_wq serves all asynchronous writeback tasks */
|
|
|
|
struct workqueue_struct *bdi_wq;
|
|
|
|
|
2011-04-21 18:19:44 -06:00
|
|
|
void bdi_lock_two(struct bdi_writeback *wb1, struct bdi_writeback *wb2)
|
|
|
|
{
|
|
|
|
if (wb1 < wb2) {
|
|
|
|
spin_lock(&wb1->list_lock);
|
|
|
|
spin_lock_nested(&wb2->list_lock, 1);
|
|
|
|
} else {
|
|
|
|
spin_lock(&wb2->list_lock);
|
|
|
|
spin_lock_nested(&wb1->list_lock, 1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-04-30 01:54:36 -06:00
|
|
|
#ifdef CONFIG_DEBUG_FS
|
|
|
|
#include <linux/debugfs.h>
|
|
|
|
#include <linux/seq_file.h>
|
|
|
|
|
|
|
|
static struct dentry *bdi_debug_root;
|
|
|
|
|
|
|
|
static void bdi_debug_init(void)
|
|
|
|
{
|
|
|
|
bdi_debug_root = debugfs_create_dir("bdi", NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int bdi_debug_stats_show(struct seq_file *m, void *v)
|
|
|
|
{
|
|
|
|
struct backing_dev_info *bdi = m->private;
|
2010-06-19 15:08:06 -06:00
|
|
|
struct bdi_writeback *wb = &bdi->wb;
|
2009-01-06 15:39:29 -07:00
|
|
|
unsigned long background_thresh;
|
|
|
|
unsigned long dirty_thresh;
|
|
|
|
unsigned long bdi_thresh;
|
2011-05-20 13:23:37 -06:00
|
|
|
unsigned long nr_dirty, nr_io, nr_more_io;
|
2009-05-25 01:08:21 -06:00
|
|
|
struct inode *inode;
|
|
|
|
|
2011-05-20 13:23:37 -06:00
|
|
|
nr_dirty = nr_io = nr_more_io = 0;
|
2011-04-21 18:19:44 -06:00
|
|
|
spin_lock(&wb->list_lock);
|
2010-10-20 18:49:30 -06:00
|
|
|
list_for_each_entry(inode, &wb->b_dirty, i_wb_list)
|
2010-06-19 15:08:06 -06:00
|
|
|
nr_dirty++;
|
2010-10-20 18:49:30 -06:00
|
|
|
list_for_each_entry(inode, &wb->b_io, i_wb_list)
|
2010-06-19 15:08:06 -06:00
|
|
|
nr_io++;
|
2010-10-20 18:49:30 -06:00
|
|
|
list_for_each_entry(inode, &wb->b_more_io, i_wb_list)
|
2010-06-19 15:08:06 -06:00
|
|
|
nr_more_io++;
|
2011-04-21 18:19:44 -06:00
|
|
|
spin_unlock(&wb->list_lock);
|
2008-04-30 01:54:36 -06:00
|
|
|
|
2010-08-11 15:17:39 -06:00
|
|
|
global_dirty_limits(&background_thresh, &dirty_thresh);
|
|
|
|
bdi_thresh = bdi_dirty_limit(bdi, dirty_thresh);
|
2008-04-30 01:54:36 -06:00
|
|
|
|
|
|
|
#define K(x) ((x) << (PAGE_SHIFT - 10))
|
|
|
|
seq_printf(m,
|
2010-08-29 11:28:45 -06:00
|
|
|
"BdiWriteback: %10lu kB\n"
|
|
|
|
"BdiReclaimable: %10lu kB\n"
|
|
|
|
"BdiDirtyThresh: %10lu kB\n"
|
|
|
|
"DirtyThresh: %10lu kB\n"
|
|
|
|
"BackgroundThresh: %10lu kB\n"
|
2011-01-23 09:07:47 -07:00
|
|
|
"BdiDirtied: %10lu kB\n"
|
2010-08-29 11:28:45 -06:00
|
|
|
"BdiWritten: %10lu kB\n"
|
|
|
|
"BdiWriteBandwidth: %10lu kBps\n"
|
|
|
|
"b_dirty: %10lu\n"
|
|
|
|
"b_io: %10lu\n"
|
|
|
|
"b_more_io: %10lu\n"
|
|
|
|
"bdi_list: %10u\n"
|
|
|
|
"state: %10lx\n",
|
2008-04-30 01:54:36 -06:00
|
|
|
(unsigned long) K(bdi_stat(bdi, BDI_WRITEBACK)),
|
|
|
|
(unsigned long) K(bdi_stat(bdi, BDI_RECLAIMABLE)),
|
2010-12-08 21:44:24 -07:00
|
|
|
K(bdi_thresh),
|
|
|
|
K(dirty_thresh),
|
|
|
|
K(background_thresh),
|
2011-01-23 09:07:47 -07:00
|
|
|
(unsigned long) K(bdi_stat(bdi, BDI_DIRTIED)),
|
2010-12-08 21:44:24 -07:00
|
|
|
(unsigned long) K(bdi_stat(bdi, BDI_WRITTEN)),
|
2010-08-29 11:28:45 -06:00
|
|
|
(unsigned long) K(bdi->write_bandwidth),
|
2010-12-08 21:44:24 -07:00
|
|
|
nr_dirty,
|
|
|
|
nr_io,
|
|
|
|
nr_more_io,
|
2010-06-19 15:08:06 -06:00
|
|
|
!list_empty(&bdi->bdi_list), bdi->state);
|
2008-04-30 01:54:36 -06:00
|
|
|
#undef K
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int bdi_debug_stats_open(struct inode *inode, struct file *file)
|
|
|
|
{
|
|
|
|
return single_open(file, bdi_debug_stats_show, inode->i_private);
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct file_operations bdi_debug_stats_fops = {
|
|
|
|
.open = bdi_debug_stats_open,
|
|
|
|
.read = seq_read,
|
|
|
|
.llseek = seq_lseek,
|
|
|
|
.release = single_release,
|
|
|
|
};
|
|
|
|
|
|
|
|
static void bdi_debug_register(struct backing_dev_info *bdi, const char *name)
|
|
|
|
{
|
|
|
|
bdi->debug_dir = debugfs_create_dir(name, bdi_debug_root);
|
|
|
|
bdi->debug_stats = debugfs_create_file("stats", 0444, bdi->debug_dir,
|
|
|
|
bdi, &bdi_debug_stats_fops);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void bdi_debug_unregister(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
debugfs_remove(bdi->debug_stats);
|
|
|
|
debugfs_remove(bdi->debug_dir);
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static inline void bdi_debug_init(void)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
static inline void bdi_debug_register(struct backing_dev_info *bdi,
|
|
|
|
const char *name)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
static inline void bdi_debug_unregister(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
static ssize_t read_ahead_kb_store(struct device *dev,
|
|
|
|
struct device_attribute *attr,
|
|
|
|
const char *buf, size_t count)
|
|
|
|
{
|
|
|
|
struct backing_dev_info *bdi = dev_get_drvdata(dev);
|
|
|
|
unsigned long read_ahead_kb;
|
2012-08-25 02:57:27 -06:00
|
|
|
ssize_t ret;
|
2008-04-30 01:54:32 -06:00
|
|
|
|
2012-08-25 02:57:27 -06:00
|
|
|
ret = kstrtoul(buf, 10, &read_ahead_kb);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
bdi->ra_pages = read_ahead_kb >> (PAGE_SHIFT - 10);
|
|
|
|
|
|
|
|
return count;
|
2008-04-30 01:54:32 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
#define K(pages) ((pages) << (PAGE_SHIFT - 10))
|
|
|
|
|
|
|
|
#define BDI_SHOW(name, expr) \
|
|
|
|
static ssize_t name##_show(struct device *dev, \
|
|
|
|
struct device_attribute *attr, char *page) \
|
|
|
|
{ \
|
|
|
|
struct backing_dev_info *bdi = dev_get_drvdata(dev); \
|
|
|
|
\
|
|
|
|
return snprintf(page, PAGE_SIZE-1, "%lld\n", (long long)expr); \
|
|
|
|
}
|
|
|
|
|
|
|
|
BDI_SHOW(read_ahead_kb, K(bdi->ra_pages))
|
|
|
|
|
2008-04-30 01:54:35 -06:00
|
|
|
static ssize_t min_ratio_store(struct device *dev,
|
|
|
|
struct device_attribute *attr, const char *buf, size_t count)
|
|
|
|
{
|
|
|
|
struct backing_dev_info *bdi = dev_get_drvdata(dev);
|
|
|
|
unsigned int ratio;
|
2012-08-25 02:57:27 -06:00
|
|
|
ssize_t ret;
|
|
|
|
|
|
|
|
ret = kstrtouint(buf, 10, &ratio);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = bdi_set_min_ratio(bdi, ratio);
|
|
|
|
if (!ret)
|
|
|
|
ret = count;
|
2008-04-30 01:54:35 -06:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
BDI_SHOW(min_ratio, bdi->min_ratio)
|
|
|
|
|
2008-04-30 01:54:36 -06:00
|
|
|
static ssize_t max_ratio_store(struct device *dev,
|
|
|
|
struct device_attribute *attr, const char *buf, size_t count)
|
|
|
|
{
|
|
|
|
struct backing_dev_info *bdi = dev_get_drvdata(dev);
|
|
|
|
unsigned int ratio;
|
2012-08-25 02:57:27 -06:00
|
|
|
ssize_t ret;
|
|
|
|
|
|
|
|
ret = kstrtouint(buf, 10, &ratio);
|
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = bdi_set_max_ratio(bdi, ratio);
|
|
|
|
if (!ret)
|
|
|
|
ret = count;
|
2008-04-30 01:54:36 -06:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
BDI_SHOW(max_ratio, bdi->max_ratio)
|
|
|
|
|
bdi: allow block devices to say that they require stable page writes
This patchset ("stable page writes, part 2") makes some key
modifications to the original 'stable page writes' patchset. First, it
provides creators (devices and filesystems) of a backing_dev_info a flag
that declares whether or not it is necessary to ensure that page
contents cannot change during writeout. It is no longer assumed that
this is true of all devices (which was never true anyway). Second, the
flag is used to relaxed the wait_on_page_writeback calls so that wait
only occurs if the device needs it. Third, it fixes up the remaining
disk-backed filesystems to use this improved conditional-wait logic to
provide stable page writes on those filesystems.
It is hoped that (for people not using checksumming devices, anyway)
this patchset will give back unnecessary performance decreases since the
original stable page write patchset went into 3.0. Sorry about not
fixing it sooner.
Complaints were registered by several people about the long write
latencies introduced by the original stable page write patchset.
Generally speaking, the kernel ought to allocate as little extra memory
as possible to facilitate writeout, but for people who simply cannot
wait, a second page stability strategy is (re)introduced: snapshotting
page contents. The waiting behavior is still the default strategy; to
enable page snapshotting, a superblock flag (MS_SNAP_STABLE) must be
set. This flag is used to bandaid^Henable stable page writeback on
ext3[1], and is not used anywhere else.
Given that there are already a few storage devices and network FSes that
have rolled their own page stability wait/page snapshot code, it would
be nice to move towards consolidating all of these. It seems possible
that iscsi and raid5 may wish to use the new stable page write support
to enable zero-copy writeout.
Thank you to Jan Kara for helping fix a couple more filesystems.
Per Andrew Morton's request, here are the result of using dbench to measure
latencies on ext2:
3.8.0-rc3:
Operation Count AvgLat MaxLat
----------------------------------------
WriteX 109347 0.028 59.817
ReadX 347180 0.004 3.391
Flush 15514 29.828 287.283
Throughput 57.429 MB/sec 4 clients 4 procs max_latency=287.290 ms
3.8.0-rc3 + patches:
WriteX 105556 0.029 4.273
ReadX 335004 0.005 4.112
Flush 14982 30.540 298.634
Throughput 55.4496 MB/sec 4 clients 4 procs max_latency=298.650 ms
As you can see, for ext2 the maximum write latency decreases from ~60ms
on a laptop hard disk to ~4ms. I'm not sure why the flush latencies
increase, though I suspect that being able to dirty pages faster gives
the flusher more work to do.
On ext4, the average write latency decreases as well as all the maximum
latencies:
3.8.0-rc3:
WriteX 85624 0.152 33.078
ReadX 272090 0.010 61.210
Flush 12129 36.219 168.260
Throughput 44.8618 MB/sec 4 clients 4 procs max_latency=168.276 ms
3.8.0-rc3 + patches:
WriteX 86082 0.141 30.928
ReadX 273358 0.010 36.124
Flush 12214 34.800 165.689
Throughput 44.9941 MB/sec 4 clients 4 procs max_latency=165.722 ms
XFS seems to exhibit similar latency improvements as ext2:
3.8.0-rc3:
WriteX 125739 0.028 104.343
ReadX 399070 0.005 4.115
Flush 17851 25.004 131.390
Throughput 66.0024 MB/sec 4 clients 4 procs max_latency=131.406 ms
3.8.0-rc3 + patches:
WriteX 123529 0.028 6.299
ReadX 392434 0.005 4.287
Flush 17549 25.120 188.687
Throughput 64.9113 MB/sec 4 clients 4 procs max_latency=188.704 ms
...and btrfs, just to round things out, also shows some latency
decreases:
3.8.0-rc3:
WriteX 67122 0.083 82.355
ReadX 212719 0.005 2.828
Flush 9547 47.561 147.418
Throughput 35.3391 MB/sec 4 clients 4 procs max_latency=147.433 ms
3.8.0-rc3 + patches:
WriteX 64898 0.101 71.631
ReadX 206673 0.005 7.123
Flush 9190 47.963 219.034
Throughput 34.0795 MB/sec 4 clients 4 procs max_latency=219.044 ms
Before this patchset, all filesystems would block, regardless of whether
or not it was necessary. ext3 would wait, but still generate occasional
checksum errors. The network filesystems were left to do their own
thing, so they'd wait too.
After this patchset, all the disk filesystems except ext3 and btrfs will
wait only if the hardware requires it. ext3 (if necessary) snapshots
pages instead of blocking, and btrfs provides its own bdi so the mm will
never wait. Network filesystems haven't been touched, so either they
provide their own wait code, or they don't block at all. The blocking
behavior is back to what it was before 3.0 if you don't have a disk
requiring stable page writes.
This patchset has been tested on 3.8.0-rc3 on x64 with ext3, ext4, and
xfs. I've spot-checked 3.8.0-rc4 and seem to be getting the same
results as -rc3.
[1] The alternative fixes to ext3 include fixing the locking order and
page bit handling like we did for ext4 (but then why not just use
ext4?), or setting PG_writeback so early that ext3 becomes extremely
slow. I tried that, but the number of write()s I could initiate dropped
by nearly an order of magnitude. That was a bit much even for the
author of the stable page series! :)
This patch:
Creates a per-backing-device flag that tracks whether or not pages must
be held immutable during writeout. Eventually it will be used to waive
wait_for_page_writeback() if nothing requires stable pages.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Artem Bityutskiy <dedekind1@gmail.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Ron Minnich <rminnich@sandia.gov>
Cc: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-21 17:42:48 -07:00
|
|
|
static ssize_t stable_pages_required_show(struct device *dev,
|
|
|
|
struct device_attribute *attr,
|
|
|
|
char *page)
|
|
|
|
{
|
|
|
|
struct backing_dev_info *bdi = dev_get_drvdata(dev);
|
|
|
|
|
|
|
|
return snprintf(page, PAGE_SIZE-1, "%d\n",
|
|
|
|
bdi_cap_stable_pages_required(bdi) ? 1 : 0);
|
|
|
|
}
|
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
#define __ATTR_RW(attr) __ATTR(attr, 0644, attr##_show, attr##_store)
|
|
|
|
|
|
|
|
static struct device_attribute bdi_dev_attrs[] = {
|
|
|
|
__ATTR_RW(read_ahead_kb),
|
2008-04-30 01:54:35 -06:00
|
|
|
__ATTR_RW(min_ratio),
|
2008-04-30 01:54:36 -06:00
|
|
|
__ATTR_RW(max_ratio),
|
bdi: allow block devices to say that they require stable page writes
This patchset ("stable page writes, part 2") makes some key
modifications to the original 'stable page writes' patchset. First, it
provides creators (devices and filesystems) of a backing_dev_info a flag
that declares whether or not it is necessary to ensure that page
contents cannot change during writeout. It is no longer assumed that
this is true of all devices (which was never true anyway). Second, the
flag is used to relaxed the wait_on_page_writeback calls so that wait
only occurs if the device needs it. Third, it fixes up the remaining
disk-backed filesystems to use this improved conditional-wait logic to
provide stable page writes on those filesystems.
It is hoped that (for people not using checksumming devices, anyway)
this patchset will give back unnecessary performance decreases since the
original stable page write patchset went into 3.0. Sorry about not
fixing it sooner.
Complaints were registered by several people about the long write
latencies introduced by the original stable page write patchset.
Generally speaking, the kernel ought to allocate as little extra memory
as possible to facilitate writeout, but for people who simply cannot
wait, a second page stability strategy is (re)introduced: snapshotting
page contents. The waiting behavior is still the default strategy; to
enable page snapshotting, a superblock flag (MS_SNAP_STABLE) must be
set. This flag is used to bandaid^Henable stable page writeback on
ext3[1], and is not used anywhere else.
Given that there are already a few storage devices and network FSes that
have rolled their own page stability wait/page snapshot code, it would
be nice to move towards consolidating all of these. It seems possible
that iscsi and raid5 may wish to use the new stable page write support
to enable zero-copy writeout.
Thank you to Jan Kara for helping fix a couple more filesystems.
Per Andrew Morton's request, here are the result of using dbench to measure
latencies on ext2:
3.8.0-rc3:
Operation Count AvgLat MaxLat
----------------------------------------
WriteX 109347 0.028 59.817
ReadX 347180 0.004 3.391
Flush 15514 29.828 287.283
Throughput 57.429 MB/sec 4 clients 4 procs max_latency=287.290 ms
3.8.0-rc3 + patches:
WriteX 105556 0.029 4.273
ReadX 335004 0.005 4.112
Flush 14982 30.540 298.634
Throughput 55.4496 MB/sec 4 clients 4 procs max_latency=298.650 ms
As you can see, for ext2 the maximum write latency decreases from ~60ms
on a laptop hard disk to ~4ms. I'm not sure why the flush latencies
increase, though I suspect that being able to dirty pages faster gives
the flusher more work to do.
On ext4, the average write latency decreases as well as all the maximum
latencies:
3.8.0-rc3:
WriteX 85624 0.152 33.078
ReadX 272090 0.010 61.210
Flush 12129 36.219 168.260
Throughput 44.8618 MB/sec 4 clients 4 procs max_latency=168.276 ms
3.8.0-rc3 + patches:
WriteX 86082 0.141 30.928
ReadX 273358 0.010 36.124
Flush 12214 34.800 165.689
Throughput 44.9941 MB/sec 4 clients 4 procs max_latency=165.722 ms
XFS seems to exhibit similar latency improvements as ext2:
3.8.0-rc3:
WriteX 125739 0.028 104.343
ReadX 399070 0.005 4.115
Flush 17851 25.004 131.390
Throughput 66.0024 MB/sec 4 clients 4 procs max_latency=131.406 ms
3.8.0-rc3 + patches:
WriteX 123529 0.028 6.299
ReadX 392434 0.005 4.287
Flush 17549 25.120 188.687
Throughput 64.9113 MB/sec 4 clients 4 procs max_latency=188.704 ms
...and btrfs, just to round things out, also shows some latency
decreases:
3.8.0-rc3:
WriteX 67122 0.083 82.355
ReadX 212719 0.005 2.828
Flush 9547 47.561 147.418
Throughput 35.3391 MB/sec 4 clients 4 procs max_latency=147.433 ms
3.8.0-rc3 + patches:
WriteX 64898 0.101 71.631
ReadX 206673 0.005 7.123
Flush 9190 47.963 219.034
Throughput 34.0795 MB/sec 4 clients 4 procs max_latency=219.044 ms
Before this patchset, all filesystems would block, regardless of whether
or not it was necessary. ext3 would wait, but still generate occasional
checksum errors. The network filesystems were left to do their own
thing, so they'd wait too.
After this patchset, all the disk filesystems except ext3 and btrfs will
wait only if the hardware requires it. ext3 (if necessary) snapshots
pages instead of blocking, and btrfs provides its own bdi so the mm will
never wait. Network filesystems haven't been touched, so either they
provide their own wait code, or they don't block at all. The blocking
behavior is back to what it was before 3.0 if you don't have a disk
requiring stable page writes.
This patchset has been tested on 3.8.0-rc3 on x64 with ext3, ext4, and
xfs. I've spot-checked 3.8.0-rc4 and seem to be getting the same
results as -rc3.
[1] The alternative fixes to ext3 include fixing the locking order and
page bit handling like we did for ext4 (but then why not just use
ext4?), or setting PG_writeback so early that ext3 becomes extremely
slow. I tried that, but the number of write()s I could initiate dropped
by nearly an order of magnitude. That was a bit much even for the
author of the stable page series! :)
This patch:
Creates a per-backing-device flag that tracks whether or not pages must
be held immutable during writeout. Eventually it will be used to waive
wait_for_page_writeback() if nothing requires stable pages.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Artem Bityutskiy <dedekind1@gmail.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Ron Minnich <rminnich@sandia.gov>
Cc: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-21 17:42:48 -07:00
|
|
|
__ATTR_RO(stable_pages_required),
|
2008-04-30 01:54:32 -06:00
|
|
|
__ATTR_NULL,
|
|
|
|
};
|
|
|
|
|
|
|
|
static __init int bdi_class_init(void)
|
|
|
|
{
|
|
|
|
bdi_class = class_create(THIS_MODULE, "bdi");
|
2010-04-02 01:46:55 -06:00
|
|
|
if (IS_ERR(bdi_class))
|
|
|
|
return PTR_ERR(bdi_class);
|
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
bdi_class->dev_attrs = bdi_dev_attrs;
|
2008-04-30 01:54:36 -06:00
|
|
|
bdi_debug_init();
|
2008-04-30 01:54:32 -06:00
|
|
|
return 0;
|
|
|
|
}
|
2008-04-30 01:54:36 -06:00
|
|
|
postcore_initcall(bdi_class_init);
|
2008-04-30 01:54:32 -06:00
|
|
|
|
2009-03-17 02:35:06 -06:00
|
|
|
static int __init default_bdi_init(void)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
bdi_wq = alloc_workqueue("writeback", WQ_MEM_RECLAIM | WQ_FREEZABLE |
|
2013-04-01 20:08:06 -06:00
|
|
|
WQ_UNBOUND | WQ_SYSFS, 0);
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
if (!bdi_wq)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2009-03-17 02:35:06 -06:00
|
|
|
err = bdi_init(&default_backing_dev_info);
|
|
|
|
if (!err)
|
|
|
|
bdi_register(&default_backing_dev_info, NULL, "default");
|
2010-09-21 03:48:55 -06:00
|
|
|
err = bdi_init(&noop_backing_dev_info);
|
2009-03-17 02:35:06 -06:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
subsys_initcall(default_bdi_init);
|
|
|
|
|
2009-09-09 01:08:54 -06:00
|
|
|
int bdi_has_dirty_io(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
return wb_has_dirty_io(&bdi->wb);
|
|
|
|
}
|
|
|
|
|
2010-07-25 05:29:22 -06:00
|
|
|
/*
|
|
|
|
* This function is used when the first inode for this bdi is marked dirty. It
|
|
|
|
* wakes-up the corresponding bdi thread which should then take care of the
|
|
|
|
* periodic background write-out of dirty inodes. Since the write-out would
|
|
|
|
* starts only 'dirty_writeback_interval' centisecs from now anyway, we just
|
|
|
|
* set up a timer which wakes the bdi thread up later.
|
|
|
|
*
|
|
|
|
* Note, we wouldn't bother setting up the timer, but this function is on the
|
|
|
|
* fast-path (used by '__mark_inode_dirty()'), so we save few context switches
|
|
|
|
* by delaying the wake-up.
|
|
|
|
*/
|
|
|
|
void bdi_wakeup_thread_delayed(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
unsigned long timeout;
|
|
|
|
|
|
|
|
timeout = msecs_to_jiffies(dirty_writeback_interval * 10);
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
mod_delayed_work(bdi_wq, &bdi->wb.dwork, timeout);
|
2009-09-09 01:08:54 -06:00
|
|
|
}
|
|
|
|
|
2009-09-14 05:12:40 -06:00
|
|
|
/*
|
|
|
|
* Remove bdi from bdi_list, and ensure that it is no longer visible
|
|
|
|
*/
|
|
|
|
static void bdi_remove_from_list(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
spin_lock_bh(&bdi_lock);
|
|
|
|
list_del_rcu(&bdi->bdi_list);
|
|
|
|
spin_unlock_bh(&bdi_lock);
|
|
|
|
|
2011-07-23 12:44:24 -06:00
|
|
|
synchronize_rcu_expedited();
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
|
|
|
|
/* bdi_list is now unused, clear it to mark @bdi dying */
|
|
|
|
INIT_LIST_HEAD(&bdi->bdi_list);
|
2009-09-14 05:12:40 -06:00
|
|
|
}
|
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
int bdi_register(struct backing_dev_info *bdi, struct device *parent,
|
|
|
|
const char *fmt, ...)
|
|
|
|
{
|
|
|
|
va_list args;
|
|
|
|
struct device *dev;
|
|
|
|
|
2008-12-09 14:14:06 -07:00
|
|
|
if (bdi->dev) /* The driver needs to use separate queues per device */
|
writeback: cleanup bdi_register
This patch makes sure we first initialize everything and set the BDI_registered
flag, and only after this we add the bdi to 'bdi_list'. Current code adds the
bdi to the list too early, and as a result I the
WARN(!test_bit(BDI_registered, &bdi->state)
in bdi forker is triggered. Also, it is in general good practice to make things
visible only when they are fully initialized.
Also, this patch does few micro clean-ups:
1. Removes the 'exit' label which does not do anything, just returns. This
allows to get rid of few braces and 'ret' variable and make the code smaller.
2. If 'kthread_run()' fails, remove the error code it returns, not hard-coded
'-ENOMEM'. Theoretically, some day 'kthread_run()' can return something
else. Also, in case of failure it is not necessary to set 'bdi->wb.task' to
NULL.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-07-25 05:29:25 -06:00
|
|
|
return 0;
|
2008-12-02 11:31:50 -07:00
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
va_start(args, fmt);
|
2008-05-15 14:44:08 -06:00
|
|
|
dev = device_create_vargs(bdi_class, parent, MKDEV(0, 0), bdi, fmt, args);
|
2008-04-30 01:54:32 -06:00
|
|
|
va_end(args);
|
writeback: cleanup bdi_register
This patch makes sure we first initialize everything and set the BDI_registered
flag, and only after this we add the bdi to 'bdi_list'. Current code adds the
bdi to the list too early, and as a result I the
WARN(!test_bit(BDI_registered, &bdi->state)
in bdi forker is triggered. Also, it is in general good practice to make things
visible only when they are fully initialized.
Also, this patch does few micro clean-ups:
1. Removes the 'exit' label which does not do anything, just returns. This
allows to get rid of few braces and 'ret' variable and make the code smaller.
2. If 'kthread_run()' fails, remove the error code it returns, not hard-coded
'-ENOMEM'. Theoretically, some day 'kthread_run()' can return something
else. Also, in case of failure it is not necessary to set 'bdi->wb.task' to
NULL.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-07-25 05:29:25 -06:00
|
|
|
if (IS_ERR(dev))
|
|
|
|
return PTR_ERR(dev);
|
2009-09-02 01:19:46 -06:00
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
bdi->dev = dev;
|
|
|
|
|
2009-09-09 01:08:54 -06:00
|
|
|
bdi_debug_register(bdi, dev_name(dev));
|
2009-09-09 01:10:25 -06:00
|
|
|
set_bit(BDI_registered, &bdi->state);
|
writeback: cleanup bdi_register
This patch makes sure we first initialize everything and set the BDI_registered
flag, and only after this we add the bdi to 'bdi_list'. Current code adds the
bdi to the list too early, and as a result I the
WARN(!test_bit(BDI_registered, &bdi->state)
in bdi forker is triggered. Also, it is in general good practice to make things
visible only when they are fully initialized.
Also, this patch does few micro clean-ups:
1. Removes the 'exit' label which does not do anything, just returns. This
allows to get rid of few braces and 'ret' variable and make the code smaller.
2. If 'kthread_run()' fails, remove the error code it returns, not hard-coded
'-ENOMEM'. Theoretically, some day 'kthread_run()' can return something
else. Also, in case of failure it is not necessary to set 'bdi->wb.task' to
NULL.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-07-25 05:29:25 -06:00
|
|
|
|
|
|
|
spin_lock_bh(&bdi_lock);
|
|
|
|
list_add_tail_rcu(&bdi->bdi_list, &bdi_list);
|
|
|
|
spin_unlock_bh(&bdi_lock);
|
|
|
|
|
2010-07-06 21:24:06 -06:00
|
|
|
trace_writeback_bdi_register(bdi);
|
writeback: cleanup bdi_register
This patch makes sure we first initialize everything and set the BDI_registered
flag, and only after this we add the bdi to 'bdi_list'. Current code adds the
bdi to the list too early, and as a result I the
WARN(!test_bit(BDI_registered, &bdi->state)
in bdi forker is triggered. Also, it is in general good practice to make things
visible only when they are fully initialized.
Also, this patch does few micro clean-ups:
1. Removes the 'exit' label which does not do anything, just returns. This
allows to get rid of few braces and 'ret' variable and make the code smaller.
2. If 'kthread_run()' fails, remove the error code it returns, not hard-coded
'-ENOMEM'. Theoretically, some day 'kthread_run()' can return something
else. Also, in case of failure it is not necessary to set 'bdi->wb.task' to
NULL.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-07-25 05:29:25 -06:00
|
|
|
return 0;
|
2008-04-30 01:54:32 -06:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(bdi_register);
|
|
|
|
|
|
|
|
int bdi_register_dev(struct backing_dev_info *bdi, dev_t dev)
|
|
|
|
{
|
|
|
|
return bdi_register(bdi, NULL, "%u:%u", MAJOR(dev), MINOR(dev));
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(bdi_register_dev);
|
|
|
|
|
2009-09-09 01:08:54 -06:00
|
|
|
/*
|
|
|
|
* Remove bdi from the global list and shutdown any threads we have running
|
|
|
|
*/
|
|
|
|
static void bdi_wb_shutdown(struct backing_dev_info *bdi)
|
2009-09-02 01:19:46 -06:00
|
|
|
{
|
2009-09-09 01:08:54 -06:00
|
|
|
if (!bdi_cap_writeback_dirty(bdi))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
writeback: move bdi threads exiting logic to the forker thread
Currently, bdi threads can decide to exit if there were no useful activities
for 5 minutes. However, this causes nasty races: we can easily oops in the
'bdi_queue_work()' if the bdi thread decides to exit while we are waking it up.
And even if we do not oops, but the bdi tread exits immediately after we wake
it up, we'd lose the wake-up event and have an unnecessary delay (up to 5 secs)
in the bdi work processing.
This patch makes the forker thread to be the central place which not only
creates bdi threads, but also kills them if they were inactive long enough.
This better design-wise.
Another reason why this change was done is to prepare for the further changes
which will prevent the bdi threads from waking up every 5 sec and wasting
power. Indeed, when the task does not wake up periodically anymore, it won't be
able to exit either.
This patch also moves the the 'wake_up_bit()' call from the bdi thread to the
forker thread as well. So now the forker thread sets the BDI_pending bit, then
forks the task or kills it, then clears the bit and wakes up the waiting
process.
The only process which may wain on the bit is 'bdi_wb_shutdown()'. This
function was changed as well - now it first removes the bdi from the
'bdi_list', then waits on the 'BDI_pending' bit. Once it wakes up, it is
guaranteed that the forker thread won't race with it, because the bdi is not
visible. Note, the forker thread sets the 'BDI_pending' bit under the
'bdi->wb_lock' which is essential for proper serialization.
And additionally, when we change 'bdi->wb.task', we now take the
'bdi->work_lock', to make sure that we do not lose wake-ups which we otherwise
would when raced with, say, 'bdi_queue_work()'.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-07-25 05:29:20 -06:00
|
|
|
* Make sure nobody finds us on the bdi_list anymore
|
2009-09-09 01:08:54 -06:00
|
|
|
*/
|
writeback: move bdi threads exiting logic to the forker thread
Currently, bdi threads can decide to exit if there were no useful activities
for 5 minutes. However, this causes nasty races: we can easily oops in the
'bdi_queue_work()' if the bdi thread decides to exit while we are waking it up.
And even if we do not oops, but the bdi tread exits immediately after we wake
it up, we'd lose the wake-up event and have an unnecessary delay (up to 5 secs)
in the bdi work processing.
This patch makes the forker thread to be the central place which not only
creates bdi threads, but also kills them if they were inactive long enough.
This better design-wise.
Another reason why this change was done is to prepare for the further changes
which will prevent the bdi threads from waking up every 5 sec and wasting
power. Indeed, when the task does not wake up periodically anymore, it won't be
able to exit either.
This patch also moves the the 'wake_up_bit()' call from the bdi thread to the
forker thread as well. So now the forker thread sets the BDI_pending bit, then
forks the task or kills it, then clears the bit and wakes up the waiting
process.
The only process which may wain on the bit is 'bdi_wb_shutdown()'. This
function was changed as well - now it first removes the bdi from the
'bdi_list', then waits on the 'BDI_pending' bit. Once it wakes up, it is
guaranteed that the forker thread won't race with it, because the bdi is not
visible. Note, the forker thread sets the 'BDI_pending' bit under the
'bdi->wb_lock' which is essential for proper serialization.
And additionally, when we change 'bdi->wb.task', we now take the
'bdi->work_lock', to make sure that we do not lose wake-ups which we otherwise
would when raced with, say, 'bdi_queue_work()'.
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-07-25 05:29:20 -06:00
|
|
|
bdi_remove_from_list(bdi);
|
2009-09-09 01:08:54 -06:00
|
|
|
|
|
|
|
/*
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
* Drain work list and shutdown the delayed_work. At this point,
|
|
|
|
* @bdi->bdi_list is empty telling bdi_Writeback_workfn() that @bdi
|
|
|
|
* is dying and its work_list needs to be drained no matter what.
|
2009-09-09 01:08:54 -06:00
|
|
|
*/
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
mod_delayed_work(bdi_wq, &bdi->wb.dwork, 0);
|
|
|
|
flush_delayed_work(&bdi->wb.dwork);
|
|
|
|
WARN_ON(!list_empty(&bdi->work_list));
|
2009-09-09 01:08:54 -06:00
|
|
|
|
|
|
|
/*
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
* This shouldn't be necessary unless @bdi for some reason has
|
|
|
|
* unflushed dirty IO after work_list is drained. Do it anyway
|
|
|
|
* just in case.
|
2009-09-09 01:08:54 -06:00
|
|
|
*/
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
cancel_delayed_work_sync(&bdi->wb.dwork);
|
2009-09-02 01:19:46 -06:00
|
|
|
}
|
|
|
|
|
2009-10-29 04:46:12 -06:00
|
|
|
/*
|
|
|
|
* This bdi is going away now, make sure that no super_blocks point to it
|
|
|
|
*/
|
|
|
|
static void bdi_prune_sb(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
struct super_block *sb;
|
|
|
|
|
|
|
|
spin_lock(&sb_lock);
|
|
|
|
list_for_each_entry(sb, &super_blocks, s_list) {
|
|
|
|
if (sb->s_bdi == bdi)
|
2011-03-17 04:13:12 -06:00
|
|
|
sb->s_bdi = &default_backing_dev_info;
|
2009-10-29 04:46:12 -06:00
|
|
|
}
|
|
|
|
spin_unlock(&sb_lock);
|
|
|
|
}
|
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
void bdi_unregister(struct backing_dev_info *bdi)
|
|
|
|
{
|
2012-01-29 11:17:33 -07:00
|
|
|
struct device *dev = bdi->dev;
|
|
|
|
|
|
|
|
if (dev) {
|
2011-07-25 18:11:57 -06:00
|
|
|
bdi_set_min_ratio(bdi, 0);
|
2010-07-06 21:24:06 -06:00
|
|
|
trace_writeback_bdi_unregister(bdi);
|
2009-11-03 12:18:44 -07:00
|
|
|
bdi_prune_sb(bdi);
|
|
|
|
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
bdi_wb_shutdown(bdi);
|
2008-04-30 01:54:36 -06:00
|
|
|
bdi_debug_unregister(bdi);
|
2012-01-29 11:17:33 -07:00
|
|
|
|
|
|
|
spin_lock_bh(&bdi->wb_lock);
|
2008-04-30 01:54:32 -06:00
|
|
|
bdi->dev = NULL;
|
2012-01-29 11:17:33 -07:00
|
|
|
spin_unlock_bh(&bdi->wb_lock);
|
|
|
|
|
|
|
|
device_unregister(dev);
|
2008-04-30 01:54:32 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(bdi_unregister);
|
2006-10-20 00:28:16 -06:00
|
|
|
|
2010-07-25 05:29:22 -06:00
|
|
|
static void bdi_wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
memset(wb, 0, sizeof(*wb));
|
|
|
|
|
|
|
|
wb->bdi = bdi;
|
|
|
|
wb->last_old_flush = jiffies;
|
|
|
|
INIT_LIST_HEAD(&wb->b_dirty);
|
|
|
|
INIT_LIST_HEAD(&wb->b_io);
|
|
|
|
INIT_LIST_HEAD(&wb->b_more_io);
|
2011-04-21 18:19:44 -06:00
|
|
|
spin_lock_init(&wb->list_lock);
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
INIT_DELAYED_WORK(&wb->dwork, bdi_writeback_workfn);
|
2010-07-25 05:29:22 -06:00
|
|
|
}
|
|
|
|
|
2010-08-29 11:22:30 -06:00
|
|
|
/*
|
|
|
|
* Initial write bandwidth: 100 MB/s
|
|
|
|
*/
|
|
|
|
#define INIT_BW (100 << (20 - PAGE_SHIFT))
|
|
|
|
|
2007-10-17 00:25:47 -06:00
|
|
|
int bdi_init(struct backing_dev_info *bdi)
|
|
|
|
{
|
2009-09-09 01:08:54 -06:00
|
|
|
int i, err;
|
2007-10-17 00:25:47 -06:00
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
bdi->dev = NULL;
|
|
|
|
|
2008-04-30 01:54:35 -06:00
|
|
|
bdi->min_ratio = 0;
|
2008-04-30 01:54:36 -06:00
|
|
|
bdi->max_ratio = 100;
|
2012-05-24 10:59:11 -06:00
|
|
|
bdi->max_prop_frac = FPROP_FRAC_BASE;
|
2009-09-09 01:08:54 -06:00
|
|
|
spin_lock_init(&bdi->wb_lock);
|
2009-09-02 01:19:46 -06:00
|
|
|
INIT_LIST_HEAD(&bdi->bdi_list);
|
2009-09-09 01:08:54 -06:00
|
|
|
INIT_LIST_HEAD(&bdi->work_list);
|
|
|
|
|
|
|
|
bdi_wb_init(&bdi->wb, bdi);
|
|
|
|
|
2007-10-17 00:25:47 -06:00
|
|
|
for (i = 0; i < NR_BDI_STAT_ITEMS; i++) {
|
2008-12-26 07:08:55 -07:00
|
|
|
err = percpu_counter_init(&bdi->bdi_stat[i], 0);
|
2007-10-17 00:25:50 -06:00
|
|
|
if (err)
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
bdi->dirty_exceeded = 0;
|
2010-08-29 11:22:30 -06:00
|
|
|
|
|
|
|
bdi->bw_time_stamp = jiffies;
|
|
|
|
bdi->written_stamp = 0;
|
|
|
|
|
writeback: stabilize bdi->dirty_ratelimit
There are some imperfections in balanced_dirty_ratelimit.
1) large fluctuations
The dirty_rate used for computing balanced_dirty_ratelimit is merely
averaged in the past 200ms (very small comparing to the 3s estimation
period for write_bw), which makes rather dispersed distribution of
balanced_dirty_ratelimit.
It's pretty hard to average out the singular points by increasing the
estimation period. Considering that the averaging technique will
introduce very undesirable time lags, I give it up totally. (btw, the 3s
write_bw averaging time lag is much more acceptable because its impact
is one-way and therefore won't lead to oscillations.)
The more practical way is filtering -- most singular
balanced_dirty_ratelimit points can be filtered out by remembering some
prev_balanced_rate and prev_prev_balanced_rate. However the more
reliable way is to guard balanced_dirty_ratelimit with task_ratelimit.
2) due to truncates and fs redirties, the (write_bw <=> dirty_rate)
match could become unbalanced, which may lead to large systematical
errors in balanced_dirty_ratelimit. The truncates, due to its possibly
bumpy nature, can hardly be compensated smoothly. So let's face it. When
some over-estimated balanced_dirty_ratelimit brings dirty_ratelimit
high, dirty pages will go higher than the setpoint. task_ratelimit will
in turn become lower than dirty_ratelimit. So if we consider both
balanced_dirty_ratelimit and task_ratelimit and update dirty_ratelimit
only when they are on the same side of dirty_ratelimit, the systematical
errors in balanced_dirty_ratelimit won't be able to bring
dirty_ratelimit far away.
The balanced_dirty_ratelimit estimation may also be inaccurate near
@limit or @freerun, however is less an issue.
3) since we ultimately want to
- keep the fluctuations of task ratelimit as small as possible
- keep the dirty pages around the setpoint as long time as possible
the update policy used for (2) also serves the above goals nicely:
if for some reason the dirty pages are high (task_ratelimit < dirty_ratelimit),
and dirty_ratelimit is low (dirty_ratelimit < balanced_dirty_ratelimit),
there is no point to bring up dirty_ratelimit in a hurry only to hurt
both the above two goals.
So, we make use of task_ratelimit to limit the update of dirty_ratelimit
in two ways:
1) avoid changing dirty rate when it's against the position control target
(the adjusted rate will slow down the progress of dirty pages going
back to setpoint).
2) limit the step size. task_ratelimit is changing values step by step,
leaving a consistent trace comparing to the randomly jumping
balanced_dirty_ratelimit. task_ratelimit also has the nice smaller
errors in stable state and typically larger errors when there are big
errors in rate. So it's a pretty good limiting factor for the step
size of dirty_ratelimit.
Note that bdi->dirty_ratelimit is always tracking balanced_dirty_ratelimit.
task_ratelimit is merely used as a limiting factor.
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
2011-08-26 15:53:24 -06:00
|
|
|
bdi->balanced_dirty_ratelimit = INIT_BW;
|
writeback: dirty rate control
It's all about bdi->dirty_ratelimit, which aims to be (write_bw / N)
when there are N dd tasks.
On write() syscall, use bdi->dirty_ratelimit
============================================
balance_dirty_pages(pages_dirtied)
{
task_ratelimit = bdi->dirty_ratelimit * bdi_position_ratio();
pause = pages_dirtied / task_ratelimit;
sleep(pause);
}
On every 200ms, update bdi->dirty_ratelimit
===========================================
bdi_update_dirty_ratelimit()
{
task_ratelimit = bdi->dirty_ratelimit * bdi_position_ratio();
balanced_dirty_ratelimit = task_ratelimit * write_bw / dirty_rate;
bdi->dirty_ratelimit = balanced_dirty_ratelimit
}
Estimation of balanced bdi->dirty_ratelimit
===========================================
balanced task_ratelimit
-----------------------
balance_dirty_pages() needs to throttle tasks dirtying pages such that
the total amount of dirty pages stays below the specified dirty limit in
order to avoid memory deadlocks. Furthermore we desire fairness in that
tasks get throttled proportionally to the amount of pages they dirty.
IOW we want to throttle tasks such that we match the dirty rate to the
writeout bandwidth, this yields a stable amount of dirty pages:
dirty_rate == write_bw (1)
The fairness requirement gives us:
task_ratelimit = balanced_dirty_ratelimit
== write_bw / N (2)
where N is the number of dd tasks. We don't know N beforehand, but
still can estimate balanced_dirty_ratelimit within 200ms.
Start by throttling each dd task at rate
task_ratelimit = task_ratelimit_0 (3)
(any non-zero initial value is OK)
After 200ms, we measured
dirty_rate = # of pages dirtied by all dd's / 200ms
write_bw = # of pages written to the disk / 200ms
For the aggressive dd dirtiers, the equality holds
dirty_rate == N * task_rate
== N * task_ratelimit_0 (4)
Or
task_ratelimit_0 == dirty_rate / N (5)
Now we conclude that the balanced task ratelimit can be estimated by
write_bw
balanced_dirty_ratelimit = task_ratelimit_0 * ---------- (6)
dirty_rate
Because with (4) and (5) we can get the desired equality (1):
write_bw
balanced_dirty_ratelimit == (dirty_rate / N) * ----------
dirty_rate
== write_bw / N
Then using the balanced task ratelimit we can compute task pause times like:
task_pause = task->nr_dirtied / task_ratelimit
task_ratelimit with position control
------------------------------------
However, while the above gives us means of matching the dirty rate to
the writeout bandwidth, it at best provides us with a stable dirty page
count (assuming a static system). In order to control the dirty page
count such that it is high enough to provide performance, but does not
exceed the specified limit we need another control.
The dirty position control works by extending (2) to
task_ratelimit = balanced_dirty_ratelimit * pos_ratio (7)
where pos_ratio is a negative feedback function that subjects to
1) f(setpoint) = 1.0
2) df/dx < 0
That is, if the dirty pages are ABOVE the setpoint, we throttle each
task a bit more HEAVY than balanced_dirty_ratelimit, so that the dirty
pages are created less fast than they are cleaned, thus DROP to the
setpoints (and the reverse).
Based on (7) and the assumption that both dirty_ratelimit and pos_ratio
remains CONSTANT for the past 200ms, we get
task_ratelimit_0 = balanced_dirty_ratelimit * pos_ratio (8)
Putting (8) into (6), we get the formula used in
bdi_update_dirty_ratelimit():
write_bw
balanced_dirty_ratelimit *= pos_ratio * ---------- (9)
dirty_rate
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
2011-06-12 10:51:31 -06:00
|
|
|
bdi->dirty_ratelimit = INIT_BW;
|
2010-08-29 11:22:30 -06:00
|
|
|
bdi->write_bandwidth = INIT_BW;
|
|
|
|
bdi->avg_write_bandwidth = INIT_BW;
|
|
|
|
|
2012-05-24 10:59:11 -06:00
|
|
|
err = fprop_local_init_percpu(&bdi->completions);
|
2007-10-17 00:25:50 -06:00
|
|
|
|
|
|
|
if (err) {
|
|
|
|
err:
|
2007-12-05 00:45:07 -07:00
|
|
|
while (i--)
|
2007-10-17 00:25:50 -06:00
|
|
|
percpu_counter_destroy(&bdi->bdi_stat[i]);
|
2007-10-17 00:25:47 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(bdi_init);
|
|
|
|
|
|
|
|
void bdi_destroy(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2009-09-14 04:57:56 -06:00
|
|
|
/*
|
|
|
|
* Splice our entries to the default_backing_dev_info, if this
|
|
|
|
* bdi disappears
|
|
|
|
*/
|
|
|
|
if (bdi_has_dirty_io(bdi)) {
|
|
|
|
struct bdi_writeback *dst = &default_backing_dev_info.wb;
|
|
|
|
|
2011-04-21 18:19:44 -06:00
|
|
|
bdi_lock_two(&bdi->wb, dst);
|
2009-09-14 04:57:56 -06:00
|
|
|
list_splice(&bdi->wb.b_dirty, &dst->b_dirty);
|
|
|
|
list_splice(&bdi->wb.b_io, &dst->b_io);
|
|
|
|
list_splice(&bdi->wb.b_more_io, &dst->b_more_io);
|
2011-04-21 18:19:44 -06:00
|
|
|
spin_unlock(&bdi->wb.list_lock);
|
|
|
|
spin_unlock(&dst->list_lock);
|
2009-09-14 04:57:56 -06:00
|
|
|
}
|
2009-09-02 01:19:46 -06:00
|
|
|
|
2008-04-30 01:54:32 -06:00
|
|
|
bdi_unregister(bdi);
|
|
|
|
|
2011-11-11 05:29:04 -07:00
|
|
|
/*
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
* If bdi_unregister() had already been called earlier, the dwork
|
|
|
|
* could still be pending because bdi_prune_sb() can race with the
|
|
|
|
* bdi_wakeup_thread_delayed() calls from __mark_inode_dirty().
|
2011-11-11 05:29:04 -07:00
|
|
|
*/
|
writeback: replace custom worker pool implementation with unbound workqueue
Writeback implements its own worker pool - each bdi can be associated
with a worker thread which is created and destroyed dynamically. The
worker thread for the default bdi is always present and serves as the
"forker" thread which forks off worker threads for other bdis.
there's no reason for writeback to implement its own worker pool when
using unbound workqueue instead is much simpler and more efficient.
This patch replaces custom worker pool implementation in writeback
with an unbound workqueue.
The conversion isn't too complicated but the followings are worth
mentioning.
* bdi_writeback->last_active, task and wakeup_timer are removed.
delayed_work ->dwork is added instead. Explicit timer handling is
no longer necessary. Everything works by either queueing / modding
/ flushing / canceling the delayed_work item.
* bdi_writeback_thread() becomes bdi_writeback_workfn() which runs off
bdi_writeback->dwork. On each execution, it processes
bdi->work_list and reschedules itself if there are more things to
do.
The function also handles low-mem condition, which used to be
handled by the forker thread. If the function is running off a
rescuer thread, it only writes out limited number of pages so that
the rescuer can serve other bdis too. This preserves the flusher
creation failure behavior of the forker thread.
* INIT_LIST_HEAD(&bdi->bdi_list) is used to tell
bdi_writeback_workfn() about on-going bdi unregistration so that it
always drains work_list even if it's running off the rescuer. Note
that the original code was broken in this regard. Under memory
pressure, a bdi could finish unregistration with non-empty
work_list.
* The default bdi is no longer special. It now is treated the same as
any other bdi and bdi_cap_flush_forker() is removed.
* BDI_pending is no longer used. Removed.
* Some tracepoints become non-applicable. The following TPs are
removed - writeback_nothread, writeback_wake_thread,
writeback_wake_forker_thread, writeback_thread_start,
writeback_thread_stop.
Everything, including devices coming and going away and rescuer
operation under simulated memory pressure, seems to work fine in my
test setup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
2013-04-01 20:08:06 -06:00
|
|
|
cancel_delayed_work_sync(&bdi->wb.dwork);
|
2011-11-11 05:29:04 -07:00
|
|
|
|
2007-10-17 00:25:47 -06:00
|
|
|
for (i = 0; i < NR_BDI_STAT_ITEMS; i++)
|
|
|
|
percpu_counter_destroy(&bdi->bdi_stat[i]);
|
2007-10-17 00:25:50 -06:00
|
|
|
|
2012-05-24 10:59:11 -06:00
|
|
|
fprop_local_destroy_percpu(&bdi->completions);
|
2007-10-17 00:25:47 -06:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(bdi_destroy);
|
|
|
|
|
2010-04-22 03:37:01 -06:00
|
|
|
/*
|
|
|
|
* For use from filesystems to quickly init and register a bdi associated
|
|
|
|
* with dirty writeback
|
|
|
|
*/
|
|
|
|
int bdi_setup_and_register(struct backing_dev_info *bdi, char *name,
|
|
|
|
unsigned int cap)
|
|
|
|
{
|
|
|
|
char tmp[32];
|
|
|
|
int err;
|
|
|
|
|
|
|
|
bdi->name = name;
|
|
|
|
bdi->capabilities = cap;
|
|
|
|
err = bdi_init(bdi);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
sprintf(tmp, "%.28s%s", name, "-%d");
|
|
|
|
err = bdi_register(bdi, NULL, tmp, atomic_long_inc_return(&bdi_seq));
|
|
|
|
if (err) {
|
|
|
|
bdi_destroy(bdi);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(bdi_setup_and_register);
|
|
|
|
|
2006-10-20 00:28:16 -06:00
|
|
|
static wait_queue_head_t congestion_wqh[2] = {
|
|
|
|
__WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[0]),
|
|
|
|
__WAIT_QUEUE_HEAD_INITIALIZER(congestion_wqh[1])
|
|
|
|
};
|
2010-10-26 15:21:45 -06:00
|
|
|
static atomic_t nr_bdi_congested[2];
|
2006-10-20 00:28:16 -06:00
|
|
|
|
2009-04-06 06:48:01 -06:00
|
|
|
void clear_bdi_congested(struct backing_dev_info *bdi, int sync)
|
2006-10-20 00:28:16 -06:00
|
|
|
{
|
|
|
|
enum bdi_state bit;
|
2009-04-06 06:48:01 -06:00
|
|
|
wait_queue_head_t *wqh = &congestion_wqh[sync];
|
2006-10-20 00:28:16 -06:00
|
|
|
|
2009-04-06 06:48:01 -06:00
|
|
|
bit = sync ? BDI_sync_congested : BDI_async_congested;
|
2010-10-26 15:21:45 -06:00
|
|
|
if (test_and_clear_bit(bit, &bdi->state))
|
|
|
|
atomic_dec(&nr_bdi_congested[sync]);
|
2006-10-20 00:28:16 -06:00
|
|
|
smp_mb__after_clear_bit();
|
|
|
|
if (waitqueue_active(wqh))
|
|
|
|
wake_up(wqh);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(clear_bdi_congested);
|
|
|
|
|
2009-04-06 06:48:01 -06:00
|
|
|
void set_bdi_congested(struct backing_dev_info *bdi, int sync)
|
2006-10-20 00:28:16 -06:00
|
|
|
{
|
|
|
|
enum bdi_state bit;
|
|
|
|
|
2009-04-06 06:48:01 -06:00
|
|
|
bit = sync ? BDI_sync_congested : BDI_async_congested;
|
2010-10-26 15:21:45 -06:00
|
|
|
if (!test_and_set_bit(bit, &bdi->state))
|
|
|
|
atomic_inc(&nr_bdi_congested[sync]);
|
2006-10-20 00:28:16 -06:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(set_bdi_congested);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* congestion_wait - wait for a backing_dev to become uncongested
|
2009-07-09 06:52:32 -06:00
|
|
|
* @sync: SYNC or ASYNC IO
|
2006-10-20 00:28:16 -06:00
|
|
|
* @timeout: timeout in jiffies
|
|
|
|
*
|
|
|
|
* Waits for up to @timeout jiffies for a backing_dev (any backing_dev) to exit
|
|
|
|
* write congestion. If no backing_devs are congested then just wait for the
|
|
|
|
* next write to be completed.
|
|
|
|
*/
|
2009-07-09 06:52:32 -06:00
|
|
|
long congestion_wait(int sync, long timeout)
|
2006-10-20 00:28:16 -06:00
|
|
|
{
|
|
|
|
long ret;
|
2010-10-26 15:21:41 -06:00
|
|
|
unsigned long start = jiffies;
|
2006-10-20 00:28:16 -06:00
|
|
|
DEFINE_WAIT(wait);
|
2009-07-09 06:52:32 -06:00
|
|
|
wait_queue_head_t *wqh = &congestion_wqh[sync];
|
2006-10-20 00:28:16 -06:00
|
|
|
|
|
|
|
prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE);
|
|
|
|
ret = io_schedule_timeout(timeout);
|
|
|
|
finish_wait(wqh, &wait);
|
2010-10-26 15:21:41 -06:00
|
|
|
|
|
|
|
trace_writeback_congestion_wait(jiffies_to_usecs(timeout),
|
|
|
|
jiffies_to_usecs(jiffies - start));
|
|
|
|
|
2006-10-20 00:28:16 -06:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(congestion_wait);
|
2007-10-17 00:25:50 -06:00
|
|
|
|
2010-10-26 15:21:45 -06:00
|
|
|
/**
|
|
|
|
* wait_iff_congested - Conditionally wait for a backing_dev to become uncongested or a zone to complete writes
|
|
|
|
* @zone: A zone to check if it is heavily congested
|
|
|
|
* @sync: SYNC or ASYNC IO
|
|
|
|
* @timeout: timeout in jiffies
|
|
|
|
*
|
|
|
|
* In the event of a congested backing_dev (any backing_dev) and the given
|
|
|
|
* @zone has experienced recent congestion, this waits for up to @timeout
|
|
|
|
* jiffies for either a BDI to exit congestion of the given @sync queue
|
|
|
|
* or a write to complete.
|
|
|
|
*
|
2011-03-30 19:57:33 -06:00
|
|
|
* In the absence of zone congestion, cond_resched() is called to yield
|
2010-10-26 15:21:45 -06:00
|
|
|
* the processor if necessary but otherwise does not sleep.
|
|
|
|
*
|
|
|
|
* The return value is 0 if the sleep is for the full timeout. Otherwise,
|
|
|
|
* it is the number of jiffies that were still remaining when the function
|
|
|
|
* returned. return_value == timeout implies the function did not sleep.
|
|
|
|
*/
|
|
|
|
long wait_iff_congested(struct zone *zone, int sync, long timeout)
|
|
|
|
{
|
|
|
|
long ret;
|
|
|
|
unsigned long start = jiffies;
|
|
|
|
DEFINE_WAIT(wait);
|
|
|
|
wait_queue_head_t *wqh = &congestion_wqh[sync];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If there is no congestion, or heavy congestion is not being
|
|
|
|
* encountered in the current zone, yield if necessary instead
|
|
|
|
* of sleeping on the congestion queue
|
|
|
|
*/
|
|
|
|
if (atomic_read(&nr_bdi_congested[sync]) == 0 ||
|
|
|
|
!zone_is_reclaim_congested(zone)) {
|
|
|
|
cond_resched();
|
|
|
|
|
|
|
|
/* In case we scheduled, work out time remaining */
|
|
|
|
ret = timeout - (jiffies - start);
|
|
|
|
if (ret < 0)
|
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Sleep until uncongested or a write happens */
|
|
|
|
prepare_to_wait(wqh, &wait, TASK_UNINTERRUPTIBLE);
|
|
|
|
ret = io_schedule_timeout(timeout);
|
|
|
|
finish_wait(wqh, &wait);
|
|
|
|
|
|
|
|
out:
|
|
|
|
trace_writeback_wait_iff_congested(jiffies_to_usecs(timeout),
|
|
|
|
jiffies_to_usecs(jiffies - start));
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(wait_iff_congested);
|
2012-07-31 17:41:52 -06:00
|
|
|
|
|
|
|
int pdflush_proc_obsolete(struct ctl_table *table, int write,
|
|
|
|
void __user *buffer, size_t *lenp, loff_t *ppos)
|
|
|
|
{
|
|
|
|
char kbuf[] = "0\n";
|
|
|
|
|
|
|
|
if (*ppos) {
|
|
|
|
*lenp = 0;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (copy_to_user(buffer, kbuf, sizeof(kbuf)))
|
|
|
|
return -EFAULT;
|
|
|
|
printk_once(KERN_WARNING "%s exported in /proc is scheduled for removal\n",
|
|
|
|
table->procname);
|
|
|
|
|
|
|
|
*lenp = 2;
|
|
|
|
*ppos += *lenp;
|
|
|
|
return 2;
|
|
|
|
}
|