2019-05-19 15:08:55 +03:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-only
|
2006-10-20 10:28:16 +04:00
|
|
|
|
2022-04-20 07:27:19 +03:00
|
|
|
#include <linux/blkdev.h>
|
2006-10-20 10:28:16 +04:00
|
|
|
#include <linux/wait.h>
|
2019-08-26 19:06:53 +03:00
|
|
|
#include <linux/rbtree.h>
|
2009-09-09 11:08:54 +04:00
|
|
|
#include <linux/kthread.h>
|
2021-09-20 15:33:13 +03:00
|
|
|
#include <linux/backing-dev.h>
|
|
|
|
#include <linux/blk-cgroup.h>
|
2009-09-09 11:08:54 +04:00
|
|
|
#include <linux/freezer.h>
|
2006-10-20 10:28:16 +04:00
|
|
|
#include <linux/fs.h>
|
2009-03-17 11:35:06 +03:00
|
|
|
#include <linux/pagemap.h>
|
2009-09-09 11:08:54 +04:00
|
|
|
#include <linux/mm.h>
|
2021-02-26 04:18:45 +03:00
|
|
|
#include <linux/sched/mm.h>
|
2006-10-20 10:28:16 +04:00
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/module.h>
|
2008-04-30 11:54:32 +04:00
|
|
|
#include <linux/writeback.h>
|
|
|
|
#include <linux/device.h>
|
2010-07-07 07:24:06 +04:00
|
|
|
#include <trace/events/writeback.h>
|
2008-04-30 11:54:32 +04:00
|
|
|
|
2020-09-24 09:51:40 +03:00
|
|
|
struct backing_dev_info noop_backing_dev_info;
|
2015-05-23 00:13:33 +03:00
|
|
|
EXPORT_SYMBOL_GPL(noop_backing_dev_info);
|
2010-04-25 10:54:42 +04:00
|
|
|
|
2008-04-30 11:54:32 +04:00
|
|
|
static struct class *bdi_class;
|
2020-05-04 15:47:54 +03:00
|
|
|
static const char *bdi_unknown_name = "(unknown)";
|
2009-09-14 15:12:40 +04:00
|
|
|
|
|
|
|
/*
|
2019-08-26 19:06:53 +03:00
|
|
|
* bdi_lock protects bdi_tree and updates to bdi_list. bdi_list has RCU
|
|
|
|
* reader side locking.
|
2009-09-14 15:12:40 +04:00
|
|
|
*/
|
2009-09-09 11:08:54 +04:00
|
|
|
DEFINE_SPINLOCK(bdi_lock);
|
2019-08-26 19:06:53 +03:00
|
|
|
static u64 bdi_id_cursor;
|
|
|
|
static struct rb_root bdi_tree = RB_ROOT;
|
2009-09-02 11:19:46 +04:00
|
|
|
LIST_HEAD(bdi_list);
|
2009-09-09 11:08:54 +04:00
|
|
|
|
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>
2013-04-02 06:08:06 +04:00
|
|
|
/* bdi_wq serves all asynchronous writeback tasks */
|
|
|
|
struct workqueue_struct *bdi_wq;
|
|
|
|
|
2021-02-24 23:02:52 +03:00
|
|
|
#define K(x) ((x) << (PAGE_SHIFT - 10))
|
|
|
|
|
2008-04-30 11:54:36 +04:00
|
|
|
#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;
|
2010-06-20 01:08:06 +04:00
|
|
|
struct bdi_writeback *wb = &bdi->wb;
|
2009-01-07 01:39:29 +03:00
|
|
|
unsigned long background_thresh;
|
|
|
|
unsigned long dirty_thresh;
|
2015-05-23 01:23:19 +03:00
|
|
|
unsigned long wb_thresh;
|
2015-02-02 08:37:00 +03:00
|
|
|
unsigned long nr_dirty, nr_io, nr_more_io, nr_dirty_time;
|
2009-05-25 11:08:21 +04:00
|
|
|
struct inode *inode;
|
|
|
|
|
2015-02-02 08:37:00 +03:00
|
|
|
nr_dirty = nr_io = nr_more_io = nr_dirty_time = 0;
|
2011-04-22 04:19:44 +04:00
|
|
|
spin_lock(&wb->list_lock);
|
2015-03-04 22:07:22 +03:00
|
|
|
list_for_each_entry(inode, &wb->b_dirty, i_io_list)
|
2010-06-20 01:08:06 +04:00
|
|
|
nr_dirty++;
|
2015-03-04 22:07:22 +03:00
|
|
|
list_for_each_entry(inode, &wb->b_io, i_io_list)
|
2010-06-20 01:08:06 +04:00
|
|
|
nr_io++;
|
2015-03-04 22:07:22 +03:00
|
|
|
list_for_each_entry(inode, &wb->b_more_io, i_io_list)
|
2010-06-20 01:08:06 +04:00
|
|
|
nr_more_io++;
|
2015-03-04 22:07:22 +03:00
|
|
|
list_for_each_entry(inode, &wb->b_dirty_time, i_io_list)
|
2015-02-02 08:37:00 +03:00
|
|
|
if (inode->i_state & I_DIRTY_TIME)
|
|
|
|
nr_dirty_time++;
|
2011-04-22 04:19:44 +04:00
|
|
|
spin_unlock(&wb->list_lock);
|
2008-04-30 11:54:36 +04:00
|
|
|
|
2010-08-12 01:17:39 +04:00
|
|
|
global_dirty_limits(&background_thresh, &dirty_thresh);
|
2015-05-23 01:23:19 +03:00
|
|
|
wb_thresh = wb_calc_thresh(wb, dirty_thresh);
|
2008-04-30 11:54:36 +04:00
|
|
|
|
|
|
|
seq_printf(m,
|
2010-08-29 21:28:45 +04:00
|
|
|
"BdiWriteback: %10lu kB\n"
|
|
|
|
"BdiReclaimable: %10lu kB\n"
|
|
|
|
"BdiDirtyThresh: %10lu kB\n"
|
|
|
|
"DirtyThresh: %10lu kB\n"
|
|
|
|
"BackgroundThresh: %10lu kB\n"
|
2011-01-23 19:07:47 +03:00
|
|
|
"BdiDirtied: %10lu kB\n"
|
2010-08-29 21:28:45 +04:00
|
|
|
"BdiWritten: %10lu kB\n"
|
|
|
|
"BdiWriteBandwidth: %10lu kBps\n"
|
|
|
|
"b_dirty: %10lu\n"
|
|
|
|
"b_io: %10lu\n"
|
|
|
|
"b_more_io: %10lu\n"
|
2015-02-02 08:37:00 +03:00
|
|
|
"b_dirty_time: %10lu\n"
|
2010-08-29 21:28:45 +04:00
|
|
|
"bdi_list: %10u\n"
|
|
|
|
"state: %10lx\n",
|
2015-05-23 00:13:27 +03:00
|
|
|
(unsigned long) K(wb_stat(wb, WB_WRITEBACK)),
|
|
|
|
(unsigned long) K(wb_stat(wb, WB_RECLAIMABLE)),
|
2015-05-23 01:23:19 +03:00
|
|
|
K(wb_thresh),
|
2010-12-09 07:44:24 +03:00
|
|
|
K(dirty_thresh),
|
|
|
|
K(background_thresh),
|
2015-05-23 00:13:27 +03:00
|
|
|
(unsigned long) K(wb_stat(wb, WB_DIRTIED)),
|
|
|
|
(unsigned long) K(wb_stat(wb, WB_WRITTEN)),
|
writeback: move bandwidth related fields from backing_dev_info into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bandwidth related fields from backing_dev_info into
bdi_writeback.
* The moved fields are: bw_time_stamp, dirtied_stamp, written_stamp,
write_bandwidth, avg_write_bandwidth, dirty_ratelimit,
balanced_dirty_ratelimit, completions and dirty_exceeded.
* writeback_chunk_size() and over_bground_thresh() now take @wb
instead of @bdi.
* bdi_writeout_fraction(bdi, ...) -> wb_writeout_fraction(wb, ...)
bdi_dirty_limit(bdi, ...) -> wb_dirty_limit(wb, ...)
bdi_position_ration(bdi, ...) -> wb_position_ratio(wb, ...)
bdi_update_writebandwidth(bdi, ...) -> wb_update_write_bandwidth(wb, ...)
[__]bdi_update_bandwidth(bdi, ...) -> [__]wb_update_bandwidth(wb, ...)
bdi_{max|min}_pause(bdi, ...) -> wb_{max|min}_pause(wb, ...)
bdi_dirty_limits(bdi, ...) -> wb_dirty_limits(wb, ...)
* Init/exits of the relocated fields are moved to bdi_wb_init/exit()
respectively. Note that explicit zeroing is dropped in the process
as wb's are cleared in entirety anyway.
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->stat[] are mechanically replaced with bdi->wb.stat[]
introducing no behavior changes.
v2: Typo in description fixed as suggested by Jan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:28 +03:00
|
|
|
(unsigned long) K(wb->write_bandwidth),
|
2010-12-09 07:44:24 +03:00
|
|
|
nr_dirty,
|
|
|
|
nr_io,
|
|
|
|
nr_more_io,
|
2015-02-02 08:37:00 +03:00
|
|
|
nr_dirty_time,
|
2015-05-23 00:13:26 +03:00
|
|
|
!list_empty(&bdi->bdi_list), bdi->wb.state);
|
2008-04-30 11:54:36 +04:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2018-04-06 02:23:16 +03:00
|
|
|
DEFINE_SHOW_ATTRIBUTE(bdi_debug_stats);
|
2008-04-30 11:54:36 +04:00
|
|
|
|
2019-01-22 18:21:07 +03:00
|
|
|
static void bdi_debug_register(struct backing_dev_info *bdi, const char *name)
|
2008-04-30 11:54:36 +04:00
|
|
|
{
|
|
|
|
bdi->debug_dir = debugfs_create_dir(name, bdi_debug_root);
|
2017-10-31 13:37:54 +03:00
|
|
|
|
2019-01-22 18:21:07 +03:00
|
|
|
debugfs_create_file("stats", 0444, bdi->debug_dir, bdi,
|
|
|
|
&bdi_debug_stats_fops);
|
2008-04-30 11:54:36 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
static void bdi_debug_unregister(struct backing_dev_info *bdi)
|
|
|
|
{
|
2019-01-22 18:21:07 +03:00
|
|
|
debugfs_remove_recursive(bdi->debug_dir);
|
2008-04-30 11:54:36 +04:00
|
|
|
}
|
|
|
|
#else
|
|
|
|
static inline void bdi_debug_init(void)
|
|
|
|
{
|
|
|
|
}
|
2019-01-22 18:21:07 +03:00
|
|
|
static inline void bdi_debug_register(struct backing_dev_info *bdi,
|
2008-04-30 11:54:36 +04:00
|
|
|
const char *name)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
static inline void bdi_debug_unregister(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-04-30 11:54:32 +04:00
|
|
|
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 12:57:27 +04:00
|
|
|
ssize_t ret;
|
2008-04-30 11:54:32 +04:00
|
|
|
|
2012-08-25 12:57:27 +04:00
|
|
|
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 11:54:32 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
#define BDI_SHOW(name, expr) \
|
|
|
|
static ssize_t name##_show(struct device *dev, \
|
2020-12-15 06:14:50 +03:00
|
|
|
struct device_attribute *attr, char *buf) \
|
2008-04-30 11:54:32 +04:00
|
|
|
{ \
|
|
|
|
struct backing_dev_info *bdi = dev_get_drvdata(dev); \
|
|
|
|
\
|
2020-12-15 06:14:50 +03:00
|
|
|
return sysfs_emit(buf, "%lld\n", (long long)expr); \
|
2013-07-25 02:05:26 +04:00
|
|
|
} \
|
|
|
|
static DEVICE_ATTR_RW(name);
|
2008-04-30 11:54:32 +04:00
|
|
|
|
|
|
|
BDI_SHOW(read_ahead_kb, K(bdi->ra_pages))
|
|
|
|
|
2008-04-30 11:54:35 +04:00
|
|
|
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 12:57:27 +04:00
|
|
|
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 11:54:35 +04:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
BDI_SHOW(min_ratio, bdi->min_ratio)
|
|
|
|
|
2008-04-30 11:54:36 +04:00
|
|
|
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 12:57:27 +04:00
|
|
|
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 11:54:36 +04:00
|
|
|
|
|
|
|
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>
2013-02-22 04:42:48 +04:00
|
|
|
static ssize_t stable_pages_required_show(struct device *dev,
|
|
|
|
struct device_attribute *attr,
|
2020-12-15 06:14:50 +03:00
|
|
|
char *buf)
|
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>
2013-02-22 04:42:48 +04:00
|
|
|
{
|
2020-09-24 09:51:38 +03:00
|
|
|
dev_warn_once(dev,
|
|
|
|
"the stable_pages_required attribute has been removed. Use the stable_writes queue attribute instead.\n");
|
2020-12-15 06:14:50 +03:00
|
|
|
return sysfs_emit(buf, "%d\n", 0);
|
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>
2013-02-22 04:42:48 +04:00
|
|
|
}
|
2013-07-25 02:05:26 +04:00
|
|
|
static DEVICE_ATTR_RO(stable_pages_required);
|
|
|
|
|
|
|
|
static struct attribute *bdi_dev_attrs[] = {
|
|
|
|
&dev_attr_read_ahead_kb.attr,
|
|
|
|
&dev_attr_min_ratio.attr,
|
|
|
|
&dev_attr_max_ratio.attr,
|
|
|
|
&dev_attr_stable_pages_required.attr,
|
|
|
|
NULL,
|
2008-04-30 11:54:32 +04:00
|
|
|
};
|
2013-07-25 02:05:26 +04:00
|
|
|
ATTRIBUTE_GROUPS(bdi_dev);
|
2008-04-30 11:54:32 +04:00
|
|
|
|
|
|
|
static __init int bdi_class_init(void)
|
|
|
|
{
|
|
|
|
bdi_class = class_create(THIS_MODULE, "bdi");
|
2010-04-02 11:46:55 +04:00
|
|
|
if (IS_ERR(bdi_class))
|
|
|
|
return PTR_ERR(bdi_class);
|
|
|
|
|
2013-07-25 02:05:26 +04:00
|
|
|
bdi_class->dev_groups = bdi_dev_groups;
|
2008-04-30 11:54:36 +04:00
|
|
|
bdi_debug_init();
|
2017-02-02 17:56:51 +03:00
|
|
|
|
2008-04-30 11:54:32 +04:00
|
|
|
return 0;
|
|
|
|
}
|
2008-04-30 11:54:36 +04:00
|
|
|
postcore_initcall(bdi_class_init);
|
2008-04-30 11:54:32 +04:00
|
|
|
|
2009-03-17 11:35:06 +03:00
|
|
|
static int __init default_bdi_init(void)
|
|
|
|
{
|
2019-10-04 13:00:24 +03:00
|
|
|
bdi_wq = alloc_workqueue("writeback", WQ_MEM_RECLAIM | 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>
2013-04-02 06:08:06 +04:00
|
|
|
if (!bdi_wq)
|
|
|
|
return -ENOMEM;
|
2022-06-15 16:22:29 +03:00
|
|
|
return 0;
|
2009-03-17 11:35:06 +03:00
|
|
|
}
|
|
|
|
subsys_initcall(default_bdi_init);
|
|
|
|
|
2010-07-25 15:29:22 +04:00
|
|
|
/*
|
writeback: move backing_dev_info->wb_lock and ->worklist into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bdi->wb_lock and ->worklist into wb.
* The lock protects bdi->worklist and bdi->wb.dwork scheduling. While
moving, rename it to wb->work_lock as wb->wb_lock is confusing.
Also, move wb->dwork downwards so that it's colocated with the new
->work_lock and ->work_list fields.
* bdi_writeback_workfn() -> wb_workfn()
bdi_wakeup_thread_delayed(bdi) -> wb_wakeup_delayed(wb)
bdi_wakeup_thread(bdi) -> wb_wakeup(wb)
bdi_queue_work(bdi, ...) -> wb_queue_work(wb, ...)
__bdi_start_writeback(bdi, ...) -> __wb_start_writeback(wb, ...)
get_next_work_item(bdi) -> get_next_work_item(wb)
* bdi_wb_shutdown() is renamed to wb_shutdown() and now takes @wb.
The function contained parts which belong to the containing bdi
rather than the wb itself - testing cap_writeback_dirty and
bdi_remove_from_list() invocation. Those are moved to
bdi_unregister().
* bdi_wb_{init|exit}() are renamed to wb_{init|exit}().
Initializations of the moved bdi->wb_lock and ->work_list are
relocated from bdi_init() to wb_init().
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->state are mechanically replaced with bdi->wb.state
introducing no behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:30 +03:00
|
|
|
* This function is used when the first inode for this wb is marked dirty. It
|
2010-07-25 15:29:22 +04:00
|
|
|
* 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.
|
2014-04-04 01:46:22 +04:00
|
|
|
*
|
|
|
|
* We have to be careful not to postpone flush work if it is scheduled for
|
|
|
|
* earlier. Thus we use queue_delayed_work().
|
2010-07-25 15:29:22 +04:00
|
|
|
*/
|
writeback: move backing_dev_info->wb_lock and ->worklist into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bdi->wb_lock and ->worklist into wb.
* The lock protects bdi->worklist and bdi->wb.dwork scheduling. While
moving, rename it to wb->work_lock as wb->wb_lock is confusing.
Also, move wb->dwork downwards so that it's colocated with the new
->work_lock and ->work_list fields.
* bdi_writeback_workfn() -> wb_workfn()
bdi_wakeup_thread_delayed(bdi) -> wb_wakeup_delayed(wb)
bdi_wakeup_thread(bdi) -> wb_wakeup(wb)
bdi_queue_work(bdi, ...) -> wb_queue_work(wb, ...)
__bdi_start_writeback(bdi, ...) -> __wb_start_writeback(wb, ...)
get_next_work_item(bdi) -> get_next_work_item(wb)
* bdi_wb_shutdown() is renamed to wb_shutdown() and now takes @wb.
The function contained parts which belong to the containing bdi
rather than the wb itself - testing cap_writeback_dirty and
bdi_remove_from_list() invocation. Those are moved to
bdi_unregister().
* bdi_wb_{init|exit}() are renamed to wb_{init|exit}().
Initializations of the moved bdi->wb_lock and ->work_list are
relocated from bdi_init() to wb_init().
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->state are mechanically replaced with bdi->wb.state
introducing no behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:30 +03:00
|
|
|
void wb_wakeup_delayed(struct bdi_writeback *wb)
|
2010-07-25 15:29:22 +04:00
|
|
|
{
|
|
|
|
unsigned long timeout;
|
|
|
|
|
|
|
|
timeout = msecs_to_jiffies(dirty_writeback_interval * 10);
|
writeback: move backing_dev_info->wb_lock and ->worklist into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bdi->wb_lock and ->worklist into wb.
* The lock protects bdi->worklist and bdi->wb.dwork scheduling. While
moving, rename it to wb->work_lock as wb->wb_lock is confusing.
Also, move wb->dwork downwards so that it's colocated with the new
->work_lock and ->work_list fields.
* bdi_writeback_workfn() -> wb_workfn()
bdi_wakeup_thread_delayed(bdi) -> wb_wakeup_delayed(wb)
bdi_wakeup_thread(bdi) -> wb_wakeup(wb)
bdi_queue_work(bdi, ...) -> wb_queue_work(wb, ...)
__bdi_start_writeback(bdi, ...) -> __wb_start_writeback(wb, ...)
get_next_work_item(bdi) -> get_next_work_item(wb)
* bdi_wb_shutdown() is renamed to wb_shutdown() and now takes @wb.
The function contained parts which belong to the containing bdi
rather than the wb itself - testing cap_writeback_dirty and
bdi_remove_from_list() invocation. Those are moved to
bdi_unregister().
* bdi_wb_{init|exit}() are renamed to wb_{init|exit}().
Initializations of the moved bdi->wb_lock and ->work_list are
relocated from bdi_init() to wb_init().
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->state are mechanically replaced with bdi->wb.state
introducing no behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:30 +03:00
|
|
|
spin_lock_bh(&wb->work_lock);
|
|
|
|
if (test_bit(WB_registered, &wb->state))
|
|
|
|
queue_delayed_work(bdi_wq, &wb->dwork, timeout);
|
|
|
|
spin_unlock_bh(&wb->work_lock);
|
2009-09-09 11:08:54 +04:00
|
|
|
}
|
|
|
|
|
2021-09-03 00:53:09 +03:00
|
|
|
static void wb_update_bandwidth_workfn(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct bdi_writeback *wb = container_of(to_delayed_work(work),
|
|
|
|
struct bdi_writeback, bw_dwork);
|
|
|
|
|
|
|
|
wb_update_bandwidth(wb);
|
|
|
|
}
|
|
|
|
|
2009-09-14 15:12:40 +04:00
|
|
|
/*
|
writeback: move bandwidth related fields from backing_dev_info into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bandwidth related fields from backing_dev_info into
bdi_writeback.
* The moved fields are: bw_time_stamp, dirtied_stamp, written_stamp,
write_bandwidth, avg_write_bandwidth, dirty_ratelimit,
balanced_dirty_ratelimit, completions and dirty_exceeded.
* writeback_chunk_size() and over_bground_thresh() now take @wb
instead of @bdi.
* bdi_writeout_fraction(bdi, ...) -> wb_writeout_fraction(wb, ...)
bdi_dirty_limit(bdi, ...) -> wb_dirty_limit(wb, ...)
bdi_position_ration(bdi, ...) -> wb_position_ratio(wb, ...)
bdi_update_writebandwidth(bdi, ...) -> wb_update_write_bandwidth(wb, ...)
[__]bdi_update_bandwidth(bdi, ...) -> [__]wb_update_bandwidth(wb, ...)
bdi_{max|min}_pause(bdi, ...) -> wb_{max|min}_pause(wb, ...)
bdi_dirty_limits(bdi, ...) -> wb_dirty_limits(wb, ...)
* Init/exits of the relocated fields are moved to bdi_wb_init/exit()
respectively. Note that explicit zeroing is dropped in the process
as wb's are cleared in entirety anyway.
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->stat[] are mechanically replaced with bdi->wb.stat[]
introducing no behavior changes.
v2: Typo in description fixed as suggested by Jan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:28 +03:00
|
|
|
* Initial write bandwidth: 100 MB/s
|
2009-09-14 15:12:40 +04:00
|
|
|
*/
|
writeback: move bandwidth related fields from backing_dev_info into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bandwidth related fields from backing_dev_info into
bdi_writeback.
* The moved fields are: bw_time_stamp, dirtied_stamp, written_stamp,
write_bandwidth, avg_write_bandwidth, dirty_ratelimit,
balanced_dirty_ratelimit, completions and dirty_exceeded.
* writeback_chunk_size() and over_bground_thresh() now take @wb
instead of @bdi.
* bdi_writeout_fraction(bdi, ...) -> wb_writeout_fraction(wb, ...)
bdi_dirty_limit(bdi, ...) -> wb_dirty_limit(wb, ...)
bdi_position_ration(bdi, ...) -> wb_position_ratio(wb, ...)
bdi_update_writebandwidth(bdi, ...) -> wb_update_write_bandwidth(wb, ...)
[__]bdi_update_bandwidth(bdi, ...) -> [__]wb_update_bandwidth(wb, ...)
bdi_{max|min}_pause(bdi, ...) -> wb_{max|min}_pause(wb, ...)
bdi_dirty_limits(bdi, ...) -> wb_dirty_limits(wb, ...)
* Init/exits of the relocated fields are moved to bdi_wb_init/exit()
respectively. Note that explicit zeroing is dropped in the process
as wb's are cleared in entirety anyway.
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->stat[] are mechanically replaced with bdi->wb.stat[]
introducing no behavior changes.
v2: Typo in description fixed as suggested by Jan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:28 +03:00
|
|
|
#define INIT_BW (100 << (20 - PAGE_SHIFT))
|
2009-09-14 15:12:40 +04:00
|
|
|
|
2015-05-23 00:13:34 +03:00
|
|
|
static int wb_init(struct bdi_writeback *wb, struct backing_dev_info *bdi,
|
2020-07-01 12:06:21 +03:00
|
|
|
gfp_t gfp)
|
2008-04-30 11:54:32 +04:00
|
|
|
{
|
2015-05-23 00:13:27 +03:00
|
|
|
int i, err;
|
2008-04-30 11:54:32 +04:00
|
|
|
|
2010-07-25 15:29:22 +04:00
|
|
|
memset(wb, 0, sizeof(*wb));
|
2008-12-02 21:31:50 +03:00
|
|
|
|
2010-07-25 15:29:22 +04:00
|
|
|
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);
|
2015-02-02 08:37:00 +03:00
|
|
|
INIT_LIST_HEAD(&wb->b_dirty_time);
|
2011-04-22 04:19:44 +04:00
|
|
|
spin_lock_init(&wb->list_lock);
|
2009-09-02 11:19:46 +04:00
|
|
|
|
2021-09-03 00:53:04 +03:00
|
|
|
atomic_set(&wb->writeback_inodes, 0);
|
writeback: move bandwidth related fields from backing_dev_info into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bandwidth related fields from backing_dev_info into
bdi_writeback.
* The moved fields are: bw_time_stamp, dirtied_stamp, written_stamp,
write_bandwidth, avg_write_bandwidth, dirty_ratelimit,
balanced_dirty_ratelimit, completions and dirty_exceeded.
* writeback_chunk_size() and over_bground_thresh() now take @wb
instead of @bdi.
* bdi_writeout_fraction(bdi, ...) -> wb_writeout_fraction(wb, ...)
bdi_dirty_limit(bdi, ...) -> wb_dirty_limit(wb, ...)
bdi_position_ration(bdi, ...) -> wb_position_ratio(wb, ...)
bdi_update_writebandwidth(bdi, ...) -> wb_update_write_bandwidth(wb, ...)
[__]bdi_update_bandwidth(bdi, ...) -> [__]wb_update_bandwidth(wb, ...)
bdi_{max|min}_pause(bdi, ...) -> wb_{max|min}_pause(wb, ...)
bdi_dirty_limits(bdi, ...) -> wb_dirty_limits(wb, ...)
* Init/exits of the relocated fields are moved to bdi_wb_init/exit()
respectively. Note that explicit zeroing is dropped in the process
as wb's are cleared in entirety anyway.
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->stat[] are mechanically replaced with bdi->wb.stat[]
introducing no behavior changes.
v2: Typo in description fixed as suggested by Jan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:28 +03:00
|
|
|
wb->bw_time_stamp = jiffies;
|
|
|
|
wb->balanced_dirty_ratelimit = INIT_BW;
|
|
|
|
wb->dirty_ratelimit = INIT_BW;
|
|
|
|
wb->write_bandwidth = INIT_BW;
|
|
|
|
wb->avg_write_bandwidth = INIT_BW;
|
2008-04-30 11:54:32 +04:00
|
|
|
|
writeback: move backing_dev_info->wb_lock and ->worklist into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bdi->wb_lock and ->worklist into wb.
* The lock protects bdi->worklist and bdi->wb.dwork scheduling. While
moving, rename it to wb->work_lock as wb->wb_lock is confusing.
Also, move wb->dwork downwards so that it's colocated with the new
->work_lock and ->work_list fields.
* bdi_writeback_workfn() -> wb_workfn()
bdi_wakeup_thread_delayed(bdi) -> wb_wakeup_delayed(wb)
bdi_wakeup_thread(bdi) -> wb_wakeup(wb)
bdi_queue_work(bdi, ...) -> wb_queue_work(wb, ...)
__bdi_start_writeback(bdi, ...) -> __wb_start_writeback(wb, ...)
get_next_work_item(bdi) -> get_next_work_item(wb)
* bdi_wb_shutdown() is renamed to wb_shutdown() and now takes @wb.
The function contained parts which belong to the containing bdi
rather than the wb itself - testing cap_writeback_dirty and
bdi_remove_from_list() invocation. Those are moved to
bdi_unregister().
* bdi_wb_{init|exit}() are renamed to wb_{init|exit}().
Initializations of the moved bdi->wb_lock and ->work_list are
relocated from bdi_init() to wb_init().
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->state are mechanically replaced with bdi->wb.state
introducing no behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:30 +03:00
|
|
|
spin_lock_init(&wb->work_lock);
|
|
|
|
INIT_LIST_HEAD(&wb->work_list);
|
|
|
|
INIT_DELAYED_WORK(&wb->dwork, wb_workfn);
|
2021-09-03 00:53:09 +03:00
|
|
|
INIT_DELAYED_WORK(&wb->bw_dwork, wb_update_bandwidth_workfn);
|
2016-09-01 19:20:33 +03:00
|
|
|
wb->dirty_sleep = jiffies;
|
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 15:29:25 +04:00
|
|
|
|
2015-05-23 00:13:34 +03:00
|
|
|
err = fprop_local_init_percpu(&wb->completions, gfp);
|
writeback: move bandwidth related fields from backing_dev_info into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bandwidth related fields from backing_dev_info into
bdi_writeback.
* The moved fields are: bw_time_stamp, dirtied_stamp, written_stamp,
write_bandwidth, avg_write_bandwidth, dirty_ratelimit,
balanced_dirty_ratelimit, completions and dirty_exceeded.
* writeback_chunk_size() and over_bground_thresh() now take @wb
instead of @bdi.
* bdi_writeout_fraction(bdi, ...) -> wb_writeout_fraction(wb, ...)
bdi_dirty_limit(bdi, ...) -> wb_dirty_limit(wb, ...)
bdi_position_ration(bdi, ...) -> wb_position_ratio(wb, ...)
bdi_update_writebandwidth(bdi, ...) -> wb_update_write_bandwidth(wb, ...)
[__]bdi_update_bandwidth(bdi, ...) -> [__]wb_update_bandwidth(wb, ...)
bdi_{max|min}_pause(bdi, ...) -> wb_{max|min}_pause(wb, ...)
bdi_dirty_limits(bdi, ...) -> wb_dirty_limits(wb, ...)
* Init/exits of the relocated fields are moved to bdi_wb_init/exit()
respectively. Note that explicit zeroing is dropped in the process
as wb's are cleared in entirety anyway.
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->stat[] are mechanically replaced with bdi->wb.stat[]
introducing no behavior changes.
v2: Typo in description fixed as suggested by Jan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:28 +03:00
|
|
|
if (err)
|
2021-11-05 23:37:04 +03:00
|
|
|
return err;
|
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 15:29:25 +04:00
|
|
|
|
2015-05-23 00:13:27 +03:00
|
|
|
for (i = 0; i < NR_WB_STAT_ITEMS; i++) {
|
2015-05-23 00:13:34 +03:00
|
|
|
err = percpu_counter_init(&wb->stat[i], 0, gfp);
|
2015-07-02 17:44:34 +03:00
|
|
|
if (err)
|
|
|
|
goto out_destroy_stat;
|
2015-05-23 00:13:27 +03:00
|
|
|
}
|
2008-04-30 11:54:32 +04:00
|
|
|
|
2015-05-23 00:13:27 +03:00
|
|
|
return 0;
|
2015-07-02 17:44:34 +03:00
|
|
|
|
|
|
|
out_destroy_stat:
|
2016-02-12 03:13:06 +03:00
|
|
|
while (i--)
|
2015-07-02 17:44:34 +03:00
|
|
|
percpu_counter_destroy(&wb->stat[i]);
|
|
|
|
fprop_local_destroy_percpu(&wb->completions);
|
|
|
|
return err;
|
2008-04-30 11:54:32 +04:00
|
|
|
}
|
|
|
|
|
2017-03-23 03:36:56 +03:00
|
|
|
static void cgwb_remove_from_bdi_list(struct bdi_writeback *wb);
|
|
|
|
|
2009-09-09 11:08:54 +04:00
|
|
|
/*
|
|
|
|
* Remove bdi from the global list and shutdown any threads we have running
|
|
|
|
*/
|
2015-05-23 00:13:31 +03:00
|
|
|
static void wb_shutdown(struct bdi_writeback *wb)
|
2009-09-02 11:19:46 +04:00
|
|
|
{
|
2015-01-21 00:05:00 +03:00
|
|
|
/* Make sure nobody queues further work */
|
2015-05-23 00:13:31 +03:00
|
|
|
spin_lock_bh(&wb->work_lock);
|
|
|
|
if (!test_and_clear_bit(WB_registered, &wb->state)) {
|
|
|
|
spin_unlock_bh(&wb->work_lock);
|
2009-09-09 11:08:54 +04:00
|
|
|
return;
|
2015-01-21 00:05:00 +03:00
|
|
|
}
|
2015-05-23 00:13:31 +03:00
|
|
|
spin_unlock_bh(&wb->work_lock);
|
2009-09-09 11:08:54 +04:00
|
|
|
|
2017-03-23 03:36:56 +03:00
|
|
|
cgwb_remove_from_bdi_list(wb);
|
2009-09-09 11:08:54 +04:00
|
|
|
/*
|
2015-05-23 00:13:31 +03:00
|
|
|
* Drain work list and shutdown the delayed_work. !WB_registered
|
|
|
|
* tells wb_workfn() that @wb is dying and its work_list needs to
|
|
|
|
* be drained no matter what.
|
2009-09-09 11:08:54 +04:00
|
|
|
*/
|
2015-05-23 00:13:31 +03:00
|
|
|
mod_delayed_work(bdi_wq, &wb->dwork, 0);
|
|
|
|
flush_delayed_work(&wb->dwork);
|
|
|
|
WARN_ON(!list_empty(&wb->work_list));
|
2021-09-03 00:53:09 +03:00
|
|
|
flush_delayed_work(&wb->bw_dwork);
|
2015-05-23 00:13:31 +03:00
|
|
|
}
|
|
|
|
|
writeback: move backing_dev_info->wb_lock and ->worklist into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bdi->wb_lock and ->worklist into wb.
* The lock protects bdi->worklist and bdi->wb.dwork scheduling. While
moving, rename it to wb->work_lock as wb->wb_lock is confusing.
Also, move wb->dwork downwards so that it's colocated with the new
->work_lock and ->work_list fields.
* bdi_writeback_workfn() -> wb_workfn()
bdi_wakeup_thread_delayed(bdi) -> wb_wakeup_delayed(wb)
bdi_wakeup_thread(bdi) -> wb_wakeup(wb)
bdi_queue_work(bdi, ...) -> wb_queue_work(wb, ...)
__bdi_start_writeback(bdi, ...) -> __wb_start_writeback(wb, ...)
get_next_work_item(bdi) -> get_next_work_item(wb)
* bdi_wb_shutdown() is renamed to wb_shutdown() and now takes @wb.
The function contained parts which belong to the containing bdi
rather than the wb itself - testing cap_writeback_dirty and
bdi_remove_from_list() invocation. Those are moved to
bdi_unregister().
* bdi_wb_{init|exit}() are renamed to wb_{init|exit}().
Initializations of the moved bdi->wb_lock and ->work_list are
relocated from bdi_init() to wb_init().
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->state are mechanically replaced with bdi->wb.state
introducing no behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:30 +03:00
|
|
|
static void wb_exit(struct bdi_writeback *wb)
|
2015-05-23 00:13:27 +03:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
WARN_ON(delayed_work_pending(&wb->dwork));
|
|
|
|
|
|
|
|
for (i = 0; i < NR_WB_STAT_ITEMS; i++)
|
|
|
|
percpu_counter_destroy(&wb->stat[i]);
|
2010-07-25 15:29:22 +04:00
|
|
|
|
writeback: move bandwidth related fields from backing_dev_info into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bandwidth related fields from backing_dev_info into
bdi_writeback.
* The moved fields are: bw_time_stamp, dirtied_stamp, written_stamp,
write_bandwidth, avg_write_bandwidth, dirty_ratelimit,
balanced_dirty_ratelimit, completions and dirty_exceeded.
* writeback_chunk_size() and over_bground_thresh() now take @wb
instead of @bdi.
* bdi_writeout_fraction(bdi, ...) -> wb_writeout_fraction(wb, ...)
bdi_dirty_limit(bdi, ...) -> wb_dirty_limit(wb, ...)
bdi_position_ration(bdi, ...) -> wb_position_ratio(wb, ...)
bdi_update_writebandwidth(bdi, ...) -> wb_update_write_bandwidth(wb, ...)
[__]bdi_update_bandwidth(bdi, ...) -> [__]wb_update_bandwidth(wb, ...)
bdi_{max|min}_pause(bdi, ...) -> wb_{max|min}_pause(wb, ...)
bdi_dirty_limits(bdi, ...) -> wb_dirty_limits(wb, ...)
* Init/exits of the relocated fields are moved to bdi_wb_init/exit()
respectively. Note that explicit zeroing is dropped in the process
as wb's are cleared in entirety anyway.
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->stat[] are mechanically replaced with bdi->wb.stat[]
introducing no behavior changes.
v2: Typo in description fixed as suggested by Jan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:28 +03:00
|
|
|
fprop_local_destroy_percpu(&wb->completions);
|
|
|
|
}
|
2010-08-29 21:22:30 +04:00
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
#ifdef CONFIG_CGROUP_WRITEBACK
|
|
|
|
|
|
|
|
#include <linux/memcontrol.h>
|
|
|
|
|
|
|
|
/*
|
2021-06-29 05:36:03 +03:00
|
|
|
* cgwb_lock protects bdi->cgwb_tree, blkcg->cgwb_list, offline_cgwbs and
|
|
|
|
* memcg->cgwb_list. bdi->cgwb_tree is also RCU protected.
|
2015-05-23 00:13:37 +03:00
|
|
|
*/
|
|
|
|
static DEFINE_SPINLOCK(cgwb_lock);
|
2018-05-23 20:56:32 +03:00
|
|
|
static struct workqueue_struct *cgwb_release_wq;
|
2015-05-23 00:13:37 +03:00
|
|
|
|
2021-06-29 05:36:03 +03:00
|
|
|
static LIST_HEAD(offline_cgwbs);
|
|
|
|
static void cleanup_offline_cgwbs_workfn(struct work_struct *work);
|
|
|
|
static DECLARE_WORK(cleanup_offline_cgwbs_work, cleanup_offline_cgwbs_workfn);
|
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
static void cgwb_release_workfn(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct bdi_writeback *wb = container_of(work, struct bdi_writeback,
|
|
|
|
release_work);
|
2021-11-05 23:37:04 +03:00
|
|
|
struct backing_dev_info *bdi = wb->bdi;
|
2015-05-23 00:13:37 +03:00
|
|
|
|
2018-06-18 16:46:58 +03:00
|
|
|
mutex_lock(&wb->bdi->cgwb_release_mutex);
|
2015-05-23 00:13:37 +03:00
|
|
|
wb_shutdown(wb);
|
|
|
|
|
|
|
|
css_put(wb->memcg_css);
|
|
|
|
css_put(wb->blkcg_css);
|
2018-06-18 16:46:58 +03:00
|
|
|
mutex_unlock(&wb->bdi->cgwb_release_mutex);
|
2015-05-23 00:13:37 +03:00
|
|
|
|
2019-07-24 20:37:22 +03:00
|
|
|
/* triggers blkg destruction if no online users left */
|
2022-04-20 07:27:14 +03:00
|
|
|
blkcg_unpin_online(wb->blkcg_css);
|
2018-08-31 23:22:43 +03:00
|
|
|
|
2015-05-23 01:23:33 +03:00
|
|
|
fprop_local_destroy_percpu(&wb->memcg_completions);
|
2021-06-29 05:36:03 +03:00
|
|
|
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
|
|
|
list_del(&wb->offline_node);
|
|
|
|
spin_unlock_irq(&cgwb_lock);
|
|
|
|
|
writeback, cgroup: remove wb from offline list before releasing refcnt
Boyang reported that the commit c22d70a162d3 ("writeback, cgroup:
release dying cgwbs by switching attached inodes") causes the kernel to
crash while running xfstests generic/256 on ext4 on aarch64 and ppc64le.
run fstests generic/256 at 2021-07-12 05:41:40
EXT4-fs (vda3): mounted filesystem with ordered data mode. Opts: . Quota mode: none.
Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
Mem abort info:
ESR = 0x96000005
EC = 0x25: DABT (current EL), IL = 32 bits
SET = 0, FnV = 0
EA = 0, S1PTW = 0
FSC = 0x05: level 1 translation fault
Data abort info:
ISV = 0, ISS = 0x00000005
CM = 0, WnR = 0
user pgtable: 64k pages, 48-bit VAs, pgdp=00000000b0502000
[0000000000000000] pgd=0000000000000000, p4d=0000000000000000, pud=0000000000000000
Internal error: Oops: 96000005 [#1] SMP
Modules linked in: dm_flakey dm_snapshot dm_bufio dm_zero dm_mod loop tls rpcsec_gss_krb5 auth_rpcgss nfsv4 dns_resolver nfs lockd grace fscache netfs rfkill sunrpc ext4 vfat fat mbcache jbd2 drm fuse xfs libcrc32c crct10dif_ce ghash_ce sha2_ce sha256_arm64 sha1_ce virtio_blk virtio_net net_failover virtio_console failover virtio_mmio aes_neon_bs [last unloaded: scsi_debug]
CPU: 0 PID: 408468 Comm: kworker/u8:5 Tainted: G X --------- --- 5.14.0-0.rc1.15.bx.el9.aarch64 #1
Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0 02/06/2015
Workqueue: events_unbound cleanup_offline_cgwbs_workfn
pstate: 004000c5 (nzcv daIF +PAN -UAO -TCO BTYPE=--)
pc : cleanup_offline_cgwbs_workfn+0x320/0x394
lr : cleanup_offline_cgwbs_workfn+0xe0/0x394
sp : ffff80001554fd10
x29: ffff80001554fd10 x28: 0000000000000000 x27: 0000000000000001
x26: 0000000000000000 x25: 00000000000000e0 x24: ffffd2a2fbe671a8
x23: ffff80001554fd88 x22: ffffd2a2fbe67198 x21: ffffd2a2fc25a730
x20: ffff210412bc3000 x19: ffff210412bc3280 x18: 0000000000000000
x17: 0000000000000000 x16: 0000000000000000 x15: 0000000000000000
x14: 0000000000000000 x13: 0000000000000030 x12: 0000000000000040
x11: ffff210481572238 x10: ffff21048157223a x9 : ffffd2a2fa276c60
x8 : ffff210484106b60 x7 : 0000000000000000 x6 : 000000000007d18a
x5 : ffff210416a86400 x4 : ffff210412bc0280 x3 : 0000000000000000
x2 : ffff80001554fd88 x1 : ffff210412bc0280 x0 : 0000000000000003
Call trace:
cleanup_offline_cgwbs_workfn+0x320/0x394
process_one_work+0x1f4/0x4b0
worker_thread+0x184/0x540
kthread+0x114/0x120
ret_from_fork+0x10/0x18
Code: d63f0020 97f99963 17ffffa6 f8588263 (f9400061)
---[ end trace e250fe289272792a ]---
Kernel panic - not syncing: Oops: Fatal exception
SMP: stopping secondary CPUs
SMP: failed to stop secondary CPUs 0-2
Kernel Offset: 0x52a2e9fa0000 from 0xffff800010000000
PHYS_OFFSET: 0xfff0defca0000000
CPU features: 0x00200251,23200840
Memory Limit: none
---[ end Kernel panic - not syncing: Oops: Fatal exception ]---
The problem happens when cgwb_release_workfn() races with
cleanup_offline_cgwbs_workfn(): wb_tryget() in
cleanup_offline_cgwbs_workfn() can be called after percpu_ref_exit() is
cgwb_release_workfn(), which is basically a use-after-free error.
Fix the problem by making removing the writeback structure from the
offline list before releasing the percpu reference counter. It will
guarantee that cleanup_offline_cgwbs_workfn() will not see and not
access writeback structures which are about to be released.
Link: https://lkml.kernel.org/r/20210716201039.3762203-1-guro@fb.com
Fixes: c22d70a162d3 ("writeback, cgroup: release dying cgwbs by switching attached inodes")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reported-by: Boyang Xue <bxue@redhat.com>
Suggested-by: Jan Kara <jack@suse.cz>
Tested-by: Darrick J. Wong <djwong@kernel.org>
Cc: Will Deacon <will@kernel.org>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Murphy Zhou <jencce.kernel@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2021-07-24 01:50:29 +03:00
|
|
|
percpu_ref_exit(&wb->refcnt);
|
2015-05-23 00:13:37 +03:00
|
|
|
wb_exit(wb);
|
2021-11-05 23:37:04 +03:00
|
|
|
bdi_put(bdi);
|
2021-06-29 05:35:53 +03:00
|
|
|
WARN_ON_ONCE(!list_empty(&wb->b_attached));
|
2015-05-23 00:13:37 +03:00
|
|
|
kfree_rcu(wb, rcu);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void cgwb_release(struct percpu_ref *refcnt)
|
|
|
|
{
|
|
|
|
struct bdi_writeback *wb = container_of(refcnt, struct bdi_writeback,
|
|
|
|
refcnt);
|
2018-05-23 20:56:32 +03:00
|
|
|
queue_work(cgwb_release_wq, &wb->release_work);
|
2015-05-23 00:13:37 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
static void cgwb_kill(struct bdi_writeback *wb)
|
|
|
|
{
|
|
|
|
lockdep_assert_held(&cgwb_lock);
|
|
|
|
|
|
|
|
WARN_ON(!radix_tree_delete(&wb->bdi->cgwb_tree, wb->memcg_css->id));
|
|
|
|
list_del(&wb->memcg_node);
|
|
|
|
list_del(&wb->blkcg_node);
|
2021-06-29 05:36:03 +03:00
|
|
|
list_add(&wb->offline_node, &offline_cgwbs);
|
2015-05-23 00:13:37 +03:00
|
|
|
percpu_ref_kill(&wb->refcnt);
|
|
|
|
}
|
|
|
|
|
2017-03-23 03:36:56 +03:00
|
|
|
static void cgwb_remove_from_bdi_list(struct bdi_writeback *wb)
|
|
|
|
{
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
|
|
|
list_del_rcu(&wb->bdi_node);
|
|
|
|
spin_unlock_irq(&cgwb_lock);
|
|
|
|
}
|
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
static int cgwb_create(struct backing_dev_info *bdi,
|
|
|
|
struct cgroup_subsys_state *memcg_css, gfp_t gfp)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *memcg;
|
|
|
|
struct cgroup_subsys_state *blkcg_css;
|
|
|
|
struct list_head *memcg_cgwb_list, *blkcg_cgwb_list;
|
|
|
|
struct bdi_writeback *wb;
|
|
|
|
unsigned long flags;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
memcg = mem_cgroup_from_css(memcg_css);
|
blkcg: rename subsystem name from blkio to io
blkio interface has become messy over time and is currently the
largest. In addition to the inconsistent naming scheme, it has
multiple stat files which report more or less the same thing, a number
of debug stat files which expose internal details which shouldn't have
been part of the public interface in the first place, recursive and
non-recursive stats and leaf and non-leaf knobs.
Both recursive vs. non-recursive and leaf vs. non-leaf distinctions
don't make any sense on the unified hierarchy as only leaf cgroups can
contain processes. cgroups is going through a major interface
revision with the unified hierarchy involving significant fundamental
usage changes and given that a significant portion of the interface
doesn't make sense anymore, it's a good time to reorganize the
interface.
As the first step, this patch renames the external visible subsystem
name from "blkio" to "io". This is more concise, matches the other
two major subsystem names, "cpu" and "memory", and better suited as
blkcg will be involved in anything writeback related too whether an
actual block device is involved or not.
As the subsystem legacy_name is set to "blkio", the only userland
visible change outside the unified hierarchy is that blkcg is reported
as "io" instead of "blkio" in the subsystem initialized message during
boot. On the unified hierarchy, blkcg now appears as "io".
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: cgroups@vger.kernel.org
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-08-19 00:55:29 +03:00
|
|
|
blkcg_css = cgroup_get_e_css(memcg_css->cgroup, &io_cgrp_subsys);
|
2018-06-08 03:07:19 +03:00
|
|
|
memcg_cgwb_list = &memcg->cgwb_list;
|
2022-04-20 07:27:15 +03:00
|
|
|
blkcg_cgwb_list = blkcg_get_cgwb_list(blkcg_css);
|
2015-05-23 00:13:37 +03:00
|
|
|
|
|
|
|
/* look up again under lock and discard on blkcg mismatch */
|
|
|
|
spin_lock_irqsave(&cgwb_lock, flags);
|
|
|
|
wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
|
|
|
|
if (wb && wb->blkcg_css != blkcg_css) {
|
|
|
|
cgwb_kill(wb);
|
|
|
|
wb = NULL;
|
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&cgwb_lock, flags);
|
|
|
|
if (wb)
|
|
|
|
goto out_put;
|
|
|
|
|
|
|
|
/* need to create a new one */
|
|
|
|
wb = kmalloc(sizeof(*wb), gfp);
|
2017-09-11 22:43:23 +03:00
|
|
|
if (!wb) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out_put;
|
|
|
|
}
|
2015-05-23 00:13:37 +03:00
|
|
|
|
2020-07-01 12:06:21 +03:00
|
|
|
ret = wb_init(wb, bdi, gfp);
|
2015-05-23 00:13:37 +03:00
|
|
|
if (ret)
|
|
|
|
goto err_free;
|
|
|
|
|
|
|
|
ret = percpu_ref_init(&wb->refcnt, cgwb_release, 0, gfp);
|
|
|
|
if (ret)
|
|
|
|
goto err_wb_exit;
|
|
|
|
|
2015-05-23 01:23:33 +03:00
|
|
|
ret = fprop_local_init_percpu(&wb->memcg_completions, gfp);
|
|
|
|
if (ret)
|
|
|
|
goto err_ref_exit;
|
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
wb->memcg_css = memcg_css;
|
|
|
|
wb->blkcg_css = blkcg_css;
|
2021-06-29 05:35:53 +03:00
|
|
|
INIT_LIST_HEAD(&wb->b_attached);
|
2015-05-23 00:13:37 +03:00
|
|
|
INIT_WORK(&wb->release_work, cgwb_release_workfn);
|
|
|
|
set_bit(WB_registered, &wb->state);
|
2021-11-05 23:37:04 +03:00
|
|
|
bdi_get(bdi);
|
2009-09-09 11:08:54 +04:00
|
|
|
|
|
|
|
/*
|
2015-05-23 00:13:37 +03:00
|
|
|
* The root wb determines the registered state of the whole bdi and
|
|
|
|
* memcg_cgwb_list and blkcg_cgwb_list's next pointers indicate
|
|
|
|
* whether they're still online. Don't link @wb if any is dead.
|
|
|
|
* See wb_memcg_offline() and wb_blkcg_offline().
|
2009-09-09 11:08:54 +04:00
|
|
|
*/
|
2015-05-23 00:13:37 +03:00
|
|
|
ret = -ENODEV;
|
|
|
|
spin_lock_irqsave(&cgwb_lock, flags);
|
|
|
|
if (test_bit(WB_registered, &bdi->wb.state) &&
|
|
|
|
blkcg_cgwb_list->next && memcg_cgwb_list->next) {
|
|
|
|
/* we might have raced another instance of this function */
|
|
|
|
ret = radix_tree_insert(&bdi->cgwb_tree, memcg_css->id, wb);
|
|
|
|
if (!ret) {
|
2015-10-02 21:47:05 +03:00
|
|
|
list_add_tail_rcu(&wb->bdi_node, &bdi->wb_list);
|
2015-05-23 00:13:37 +03:00
|
|
|
list_add(&wb->memcg_node, memcg_cgwb_list);
|
|
|
|
list_add(&wb->blkcg_node, blkcg_cgwb_list);
|
2022-04-20 07:27:14 +03:00
|
|
|
blkcg_pin_online(blkcg_css);
|
2015-05-23 00:13:37 +03:00
|
|
|
css_get(memcg_css);
|
|
|
|
css_get(blkcg_css);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&cgwb_lock, flags);
|
|
|
|
if (ret) {
|
|
|
|
if (ret == -EEXIST)
|
|
|
|
ret = 0;
|
2015-07-02 17:44:34 +03:00
|
|
|
goto err_fprop_exit;
|
2015-05-23 00:13:37 +03:00
|
|
|
}
|
|
|
|
goto out_put;
|
|
|
|
|
2015-05-23 01:23:33 +03:00
|
|
|
err_fprop_exit:
|
2021-11-05 23:37:04 +03:00
|
|
|
bdi_put(bdi);
|
2015-05-23 01:23:33 +03:00
|
|
|
fprop_local_destroy_percpu(&wb->memcg_completions);
|
2015-05-23 00:13:37 +03:00
|
|
|
err_ref_exit:
|
|
|
|
percpu_ref_exit(&wb->refcnt);
|
|
|
|
err_wb_exit:
|
|
|
|
wb_exit(wb);
|
|
|
|
err_free:
|
|
|
|
kfree(wb);
|
|
|
|
out_put:
|
|
|
|
css_put(blkcg_css);
|
|
|
|
return ret;
|
2009-09-02 11:19:46 +04:00
|
|
|
}
|
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
/**
|
2019-08-26 19:06:54 +03:00
|
|
|
* wb_get_lookup - get wb for a given memcg
|
2015-05-23 00:13:37 +03:00
|
|
|
* @bdi: target bdi
|
|
|
|
* @memcg_css: cgroup_subsys_state of the target memcg (must have positive ref)
|
|
|
|
*
|
2019-08-26 19:06:54 +03:00
|
|
|
* Try to get the wb for @memcg_css on @bdi. The returned wb has its
|
|
|
|
* refcount incremented.
|
2015-05-23 00:13:37 +03:00
|
|
|
*
|
|
|
|
* This function uses css_get() on @memcg_css and thus expects its refcnt
|
|
|
|
* to be positive on invocation. IOW, rcu_read_lock() protection on
|
|
|
|
* @memcg_css isn't enough. try_get it before calling this function.
|
|
|
|
*
|
|
|
|
* A wb is keyed by its associated memcg. As blkcg implicitly enables
|
|
|
|
* memcg on the default hierarchy, memcg association is guaranteed to be
|
|
|
|
* more specific (equal or descendant to the associated blkcg) and thus can
|
|
|
|
* identify both the memcg and blkcg associations.
|
|
|
|
*
|
|
|
|
* Because the blkcg associated with a memcg may change as blkcg is enabled
|
|
|
|
* and disabled closer to root in the hierarchy, each wb keeps track of
|
|
|
|
* both the memcg and blkcg associated with it and verifies the blkcg on
|
|
|
|
* each lookup. On mismatch, the existing wb is discarded and a new one is
|
|
|
|
* created.
|
|
|
|
*/
|
2019-08-26 19:06:54 +03:00
|
|
|
struct bdi_writeback *wb_get_lookup(struct backing_dev_info *bdi,
|
|
|
|
struct cgroup_subsys_state *memcg_css)
|
|
|
|
{
|
|
|
|
struct bdi_writeback *wb;
|
|
|
|
|
|
|
|
if (!memcg_css->parent)
|
|
|
|
return &bdi->wb;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
|
|
|
|
if (wb) {
|
|
|
|
struct cgroup_subsys_state *blkcg_css;
|
|
|
|
|
|
|
|
/* see whether the blkcg association has changed */
|
|
|
|
blkcg_css = cgroup_get_e_css(memcg_css->cgroup, &io_cgrp_subsys);
|
|
|
|
if (unlikely(wb->blkcg_css != blkcg_css || !wb_tryget(wb)))
|
|
|
|
wb = NULL;
|
|
|
|
css_put(blkcg_css);
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return wb;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* wb_get_create - get wb for a given memcg, create if necessary
|
|
|
|
* @bdi: target bdi
|
|
|
|
* @memcg_css: cgroup_subsys_state of the target memcg (must have positive ref)
|
|
|
|
* @gfp: allocation mask to use
|
|
|
|
*
|
|
|
|
* Try to get the wb for @memcg_css on @bdi. If it doesn't exist, try to
|
|
|
|
* create one. See wb_get_lookup() for more details.
|
|
|
|
*/
|
2015-05-23 00:13:37 +03:00
|
|
|
struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi,
|
|
|
|
struct cgroup_subsys_state *memcg_css,
|
|
|
|
gfp_t gfp)
|
2010-07-25 15:29:22 +04:00
|
|
|
{
|
2015-05-23 00:13:37 +03:00
|
|
|
struct bdi_writeback *wb;
|
|
|
|
|
2021-02-26 04:18:45 +03:00
|
|
|
might_alloc(gfp);
|
2015-05-23 00:13:37 +03:00
|
|
|
|
|
|
|
if (!memcg_css->parent)
|
|
|
|
return &bdi->wb;
|
|
|
|
|
|
|
|
do {
|
2019-08-26 19:06:54 +03:00
|
|
|
wb = wb_get_lookup(bdi, memcg_css);
|
2015-05-23 00:13:37 +03:00
|
|
|
} while (!wb && !cgwb_create(bdi, memcg_css, gfp));
|
|
|
|
|
|
|
|
return wb;
|
|
|
|
}
|
2010-07-25 15:29:22 +04:00
|
|
|
|
2015-07-02 17:44:34 +03:00
|
|
|
static int cgwb_bdi_init(struct backing_dev_info *bdi)
|
2015-05-23 00:13:37 +03:00
|
|
|
{
|
2015-07-02 17:44:34 +03:00
|
|
|
int ret;
|
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
INIT_RADIX_TREE(&bdi->cgwb_tree, GFP_ATOMIC);
|
2018-06-18 16:46:58 +03:00
|
|
|
mutex_init(&bdi->cgwb_release_mutex);
|
2017-12-12 19:38:30 +03:00
|
|
|
init_rwsem(&bdi->wb_switch_rwsem);
|
2015-07-02 17:44:34 +03:00
|
|
|
|
2020-07-01 12:06:21 +03:00
|
|
|
ret = wb_init(&bdi->wb, bdi, GFP_KERNEL);
|
2015-07-02 17:44:34 +03:00
|
|
|
if (!ret) {
|
2016-01-15 02:20:56 +03:00
|
|
|
bdi->wb.memcg_css = &root_mem_cgroup->css;
|
2015-07-02 17:44:34 +03:00
|
|
|
bdi->wb.blkcg_css = blkcg_root_css;
|
|
|
|
}
|
|
|
|
return ret;
|
2010-07-25 15:29:22 +04:00
|
|
|
}
|
|
|
|
|
2017-03-23 03:36:59 +03:00
|
|
|
static void cgwb_bdi_unregister(struct backing_dev_info *bdi)
|
2015-05-23 00:13:37 +03:00
|
|
|
{
|
|
|
|
struct radix_tree_iter iter;
|
|
|
|
void **slot;
|
2017-03-23 03:36:57 +03:00
|
|
|
struct bdi_writeback *wb;
|
2015-05-23 00:13:37 +03:00
|
|
|
|
|
|
|
WARN_ON(test_bit(WB_registered, &bdi->wb.state));
|
|
|
|
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
|
|
|
radix_tree_for_each_slot(slot, &bdi->cgwb_tree, &iter, 0)
|
|
|
|
cgwb_kill(*slot);
|
2018-06-18 16:46:58 +03:00
|
|
|
spin_unlock_irq(&cgwb_lock);
|
2017-03-23 03:36:57 +03:00
|
|
|
|
2018-06-18 16:46:58 +03:00
|
|
|
mutex_lock(&bdi->cgwb_release_mutex);
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
2017-03-23 03:36:57 +03:00
|
|
|
while (!list_empty(&bdi->wb_list)) {
|
|
|
|
wb = list_first_entry(&bdi->wb_list, struct bdi_writeback,
|
|
|
|
bdi_node);
|
|
|
|
spin_unlock_irq(&cgwb_lock);
|
|
|
|
wb_shutdown(wb);
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
|
|
|
}
|
2015-05-23 00:13:37 +03:00
|
|
|
spin_unlock_irq(&cgwb_lock);
|
2018-06-18 16:46:58 +03:00
|
|
|
mutex_unlock(&bdi->cgwb_release_mutex);
|
2015-05-23 00:13:37 +03:00
|
|
|
}
|
|
|
|
|
2021-06-29 05:36:03 +03:00
|
|
|
/*
|
|
|
|
* cleanup_offline_cgwbs_workfn - try to release dying cgwbs
|
|
|
|
*
|
|
|
|
* Try to release dying cgwbs by switching attached inodes to the nearest
|
|
|
|
* living ancestor's writeback. Processed wbs are placed at the end
|
|
|
|
* of the list to guarantee the forward progress.
|
|
|
|
*/
|
|
|
|
static void cleanup_offline_cgwbs_workfn(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct bdi_writeback *wb;
|
|
|
|
LIST_HEAD(processed);
|
|
|
|
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
|
|
|
|
|
|
|
while (!list_empty(&offline_cgwbs)) {
|
|
|
|
wb = list_first_entry(&offline_cgwbs, struct bdi_writeback,
|
|
|
|
offline_node);
|
|
|
|
list_move(&wb->offline_node, &processed);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If wb is dirty, cleaning up the writeback by switching
|
|
|
|
* attached inodes will result in an effective removal of any
|
|
|
|
* bandwidth restrictions, which isn't the goal. Instead,
|
|
|
|
* it can be postponed until the next time, when all io
|
|
|
|
* will be likely completed. If in the meantime some inodes
|
|
|
|
* will get re-dirtied, they should be eventually switched to
|
|
|
|
* a new cgwb.
|
|
|
|
*/
|
|
|
|
if (wb_has_dirty_io(wb))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (!wb_tryget(wb))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
spin_unlock_irq(&cgwb_lock);
|
|
|
|
while (cleanup_offline_cgwb(wb))
|
|
|
|
cond_resched();
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
|
|
|
|
|
|
|
wb_put(wb);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!list_empty(&processed))
|
|
|
|
list_splice_tail(&processed, &offline_cgwbs);
|
|
|
|
|
|
|
|
spin_unlock_irq(&cgwb_lock);
|
|
|
|
}
|
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
/**
|
|
|
|
* wb_memcg_offline - kill all wb's associated with a memcg being offlined
|
|
|
|
* @memcg: memcg being offlined
|
|
|
|
*
|
|
|
|
* Also prevents creation of any new wb's associated with @memcg.
|
2010-08-29 21:22:30 +04:00
|
|
|
*/
|
2015-05-23 00:13:37 +03:00
|
|
|
void wb_memcg_offline(struct mem_cgroup *memcg)
|
|
|
|
{
|
2018-06-08 03:07:19 +03:00
|
|
|
struct list_head *memcg_cgwb_list = &memcg->cgwb_list;
|
2015-05-23 00:13:37 +03:00
|
|
|
struct bdi_writeback *wb, *next;
|
|
|
|
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
|
|
|
list_for_each_entry_safe(wb, next, memcg_cgwb_list, memcg_node)
|
|
|
|
cgwb_kill(wb);
|
|
|
|
memcg_cgwb_list->next = NULL; /* prevent new wb's */
|
|
|
|
spin_unlock_irq(&cgwb_lock);
|
2021-06-29 05:36:03 +03:00
|
|
|
|
|
|
|
queue_work(system_unbound_wq, &cleanup_offline_cgwbs_work);
|
2015-05-23 00:13:37 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* wb_blkcg_offline - kill all wb's associated with a blkcg being offlined
|
2022-04-20 07:27:15 +03:00
|
|
|
* @css: blkcg being offlined
|
2015-05-23 00:13:37 +03:00
|
|
|
*
|
|
|
|
* Also prevents creation of any new wb's associated with @blkcg.
|
|
|
|
*/
|
2022-04-20 07:27:15 +03:00
|
|
|
void wb_blkcg_offline(struct cgroup_subsys_state *css)
|
2015-05-23 00:13:37 +03:00
|
|
|
{
|
|
|
|
struct bdi_writeback *wb, *next;
|
2022-04-20 07:27:15 +03:00
|
|
|
struct list_head *list = blkcg_get_cgwb_list(css);
|
2015-05-23 00:13:37 +03:00
|
|
|
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
2022-04-20 07:27:15 +03:00
|
|
|
list_for_each_entry_safe(wb, next, list, blkcg_node)
|
2015-05-23 00:13:37 +03:00
|
|
|
cgwb_kill(wb);
|
2022-04-20 07:27:15 +03:00
|
|
|
list->next = NULL; /* prevent new wb's */
|
2015-05-23 00:13:37 +03:00
|
|
|
spin_unlock_irq(&cgwb_lock);
|
|
|
|
}
|
|
|
|
|
2017-03-23 03:36:56 +03:00
|
|
|
static void cgwb_bdi_register(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
spin_lock_irq(&cgwb_lock);
|
|
|
|
list_add_tail_rcu(&bdi->wb.bdi_node, &bdi->wb_list);
|
|
|
|
spin_unlock_irq(&cgwb_lock);
|
|
|
|
}
|
|
|
|
|
2018-05-23 20:56:32 +03:00
|
|
|
static int __init cgwb_init(void)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* There can be many concurrent release work items overwhelming
|
|
|
|
* system_wq. Put them in a separate wq and limit concurrency.
|
|
|
|
* There's no point in executing many of these in parallel.
|
|
|
|
*/
|
|
|
|
cgwb_release_wq = alloc_workqueue("cgwb_release", 0, 1);
|
|
|
|
if (!cgwb_release_wq)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
subsys_initcall(cgwb_init);
|
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
#else /* CONFIG_CGROUP_WRITEBACK */
|
|
|
|
|
2015-07-02 17:44:34 +03:00
|
|
|
static int cgwb_bdi_init(struct backing_dev_info *bdi)
|
|
|
|
{
|
2020-07-01 12:06:21 +03:00
|
|
|
return wb_init(&bdi->wb, bdi, GFP_KERNEL);
|
2015-07-02 17:44:34 +03:00
|
|
|
}
|
|
|
|
|
2017-03-23 03:36:59 +03:00
|
|
|
static void cgwb_bdi_unregister(struct backing_dev_info *bdi) { }
|
2017-03-08 19:48:32 +03:00
|
|
|
|
2017-03-23 03:36:56 +03:00
|
|
|
static void cgwb_bdi_register(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
list_add_tail_rcu(&bdi->wb.bdi_node, &bdi->wb_list);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void cgwb_remove_from_bdi_list(struct bdi_writeback *wb)
|
|
|
|
{
|
|
|
|
list_del_rcu(&wb->bdi_node);
|
|
|
|
}
|
|
|
|
|
2015-05-23 00:13:37 +03:00
|
|
|
#endif /* CONFIG_CGROUP_WRITEBACK */
|
2010-08-29 21:22:30 +04:00
|
|
|
|
2022-06-15 16:22:29 +03:00
|
|
|
int bdi_init(struct backing_dev_info *bdi)
|
2007-10-17 10:25:47 +04:00
|
|
|
{
|
2015-10-02 21:47:05 +03:00
|
|
|
int ret;
|
|
|
|
|
2008-04-30 11:54:32 +04:00
|
|
|
bdi->dev = NULL;
|
|
|
|
|
2017-02-02 17:56:51 +03:00
|
|
|
kref_init(&bdi->refcnt);
|
2008-04-30 11:54:35 +04:00
|
|
|
bdi->min_ratio = 0;
|
2008-04-30 11:54:36 +04:00
|
|
|
bdi->max_ratio = 100;
|
2012-05-24 20:59:11 +04:00
|
|
|
bdi->max_prop_frac = FPROP_FRAC_BASE;
|
2009-09-02 11:19:46 +04:00
|
|
|
INIT_LIST_HEAD(&bdi->bdi_list);
|
2015-10-02 21:47:05 +03:00
|
|
|
INIT_LIST_HEAD(&bdi->wb_list);
|
2015-05-23 00:13:58 +03:00
|
|
|
init_waitqueue_head(&bdi->wb_waitq);
|
2009-09-09 11:08:54 +04:00
|
|
|
|
2015-10-02 21:47:05 +03:00
|
|
|
ret = cgwb_bdi_init(bdi);
|
|
|
|
|
|
|
|
return ret;
|
2007-10-17 10:25:47 +04:00
|
|
|
}
|
2010-08-29 21:22:30 +04:00
|
|
|
|
2020-05-04 15:48:00 +03:00
|
|
|
struct backing_dev_info *bdi_alloc(int node_id)
|
2017-02-02 17:56:51 +03:00
|
|
|
{
|
|
|
|
struct backing_dev_info *bdi;
|
|
|
|
|
2020-05-04 15:48:00 +03:00
|
|
|
bdi = kzalloc_node(sizeof(*bdi), GFP_KERNEL, node_id);
|
2017-02-02 17:56:51 +03:00
|
|
|
if (!bdi)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (bdi_init(bdi)) {
|
|
|
|
kfree(bdi);
|
|
|
|
return NULL;
|
|
|
|
}
|
2020-09-24 09:51:40 +03:00
|
|
|
bdi->capabilities = BDI_CAP_WRITEBACK | BDI_CAP_WRITEBACK_ACCT;
|
2020-09-24 09:51:32 +03:00
|
|
|
bdi->ra_pages = VM_READAHEAD_PAGES;
|
|
|
|
bdi->io_pages = VM_READAHEAD_PAGES;
|
2021-08-09 17:17:40 +03:00
|
|
|
timer_setup(&bdi->laptop_mode_wb_timer, laptop_mode_timer_fn, 0);
|
2017-02-02 17:56:51 +03:00
|
|
|
return bdi;
|
|
|
|
}
|
2020-05-04 15:48:00 +03:00
|
|
|
EXPORT_SYMBOL(bdi_alloc);
|
2017-02-02 17:56:51 +03:00
|
|
|
|
2019-08-26 19:06:53 +03:00
|
|
|
static struct rb_node **bdi_lookup_rb_node(u64 id, struct rb_node **parentp)
|
|
|
|
{
|
|
|
|
struct rb_node **p = &bdi_tree.rb_node;
|
|
|
|
struct rb_node *parent = NULL;
|
|
|
|
struct backing_dev_info *bdi;
|
|
|
|
|
|
|
|
lockdep_assert_held(&bdi_lock);
|
|
|
|
|
|
|
|
while (*p) {
|
|
|
|
parent = *p;
|
|
|
|
bdi = rb_entry(parent, struct backing_dev_info, rb_node);
|
|
|
|
|
|
|
|
if (bdi->id > id)
|
|
|
|
p = &(*p)->rb_left;
|
|
|
|
else if (bdi->id < id)
|
|
|
|
p = &(*p)->rb_right;
|
|
|
|
else
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (parentp)
|
|
|
|
*parentp = parent;
|
|
|
|
return p;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* bdi_get_by_id - lookup and get bdi from its id
|
|
|
|
* @id: bdi id to lookup
|
|
|
|
*
|
|
|
|
* Find bdi matching @id and get it. Returns NULL if the matching bdi
|
|
|
|
* doesn't exist or is already unregistered.
|
|
|
|
*/
|
|
|
|
struct backing_dev_info *bdi_get_by_id(u64 id)
|
|
|
|
{
|
|
|
|
struct backing_dev_info *bdi = NULL;
|
|
|
|
struct rb_node **p;
|
|
|
|
|
|
|
|
spin_lock_bh(&bdi_lock);
|
|
|
|
p = bdi_lookup_rb_node(id, NULL);
|
|
|
|
if (*p) {
|
|
|
|
bdi = rb_entry(*p, struct backing_dev_info, rb_node);
|
|
|
|
bdi_get(bdi);
|
|
|
|
}
|
|
|
|
spin_unlock_bh(&bdi_lock);
|
|
|
|
|
|
|
|
return bdi;
|
|
|
|
}
|
|
|
|
|
2017-04-12 13:24:49 +03:00
|
|
|
int bdi_register_va(struct backing_dev_info *bdi, const char *fmt, va_list args)
|
2015-05-23 00:13:31 +03:00
|
|
|
{
|
|
|
|
struct device *dev;
|
2019-08-26 19:06:53 +03:00
|
|
|
struct rb_node *parent, **p;
|
2010-08-29 21:22:30 +04:00
|
|
|
|
2015-05-23 00:13:31 +03:00
|
|
|
if (bdi->dev) /* The driver needs to use separate queues per device */
|
|
|
|
return 0;
|
2010-08-29 21:22:30 +04:00
|
|
|
|
2020-05-04 15:47:56 +03:00
|
|
|
vsnprintf(bdi->dev_name, sizeof(bdi->dev_name), fmt, args);
|
|
|
|
dev = device_create(bdi_class, NULL, MKDEV(0, 0), bdi, bdi->dev_name);
|
2015-05-23 00:13:31 +03:00
|
|
|
if (IS_ERR(dev))
|
|
|
|
return PTR_ERR(dev);
|
2007-10-17 10:25:50 +04:00
|
|
|
|
2017-03-23 03:36:56 +03:00
|
|
|
cgwb_bdi_register(bdi);
|
2015-05-23 00:13:31 +03:00
|
|
|
bdi->dev = dev;
|
2007-10-17 10:25:47 +04:00
|
|
|
|
2017-12-21 20:01:30 +03:00
|
|
|
bdi_debug_register(bdi, dev_name(dev));
|
2015-05-23 00:13:31 +03:00
|
|
|
set_bit(WB_registered, &bdi->wb.state);
|
|
|
|
|
|
|
|
spin_lock_bh(&bdi_lock);
|
2019-08-26 19:06:53 +03:00
|
|
|
|
|
|
|
bdi->id = ++bdi_id_cursor;
|
|
|
|
|
|
|
|
p = bdi_lookup_rb_node(bdi->id, &parent);
|
|
|
|
rb_link_node(&bdi->rb_node, parent, p);
|
|
|
|
rb_insert_color(&bdi->rb_node, &bdi_tree);
|
|
|
|
|
2015-05-23 00:13:31 +03:00
|
|
|
list_add_tail_rcu(&bdi->bdi_list, &bdi_list);
|
2019-08-26 19:06:53 +03:00
|
|
|
|
2015-05-23 00:13:31 +03:00
|
|
|
spin_unlock_bh(&bdi_lock);
|
|
|
|
|
|
|
|
trace_writeback_bdi_register(bdi);
|
|
|
|
return 0;
|
2007-10-17 10:25:47 +04:00
|
|
|
}
|
2017-04-12 13:24:25 +03:00
|
|
|
|
2017-04-12 13:24:49 +03:00
|
|
|
int bdi_register(struct backing_dev_info *bdi, const char *fmt, ...)
|
2017-04-12 13:24:25 +03:00
|
|
|
{
|
|
|
|
va_list args;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
va_start(args, fmt);
|
2017-04-12 13:24:49 +03:00
|
|
|
ret = bdi_register_va(bdi, fmt, args);
|
2017-04-12 13:24:25 +03:00
|
|
|
va_end(args);
|
|
|
|
return ret;
|
|
|
|
}
|
2015-05-23 00:13:31 +03:00
|
|
|
EXPORT_SYMBOL(bdi_register);
|
2007-10-17 10:25:47 +04:00
|
|
|
|
2020-05-04 15:47:59 +03:00
|
|
|
void bdi_set_owner(struct backing_dev_info *bdi, struct device *owner)
|
2016-07-31 21:15:13 +03:00
|
|
|
{
|
2020-05-04 15:47:59 +03:00
|
|
|
WARN_ON_ONCE(bdi->owner);
|
2016-07-31 21:15:13 +03:00
|
|
|
bdi->owner = owner;
|
|
|
|
get_device(owner);
|
|
|
|
}
|
|
|
|
|
2015-05-23 00:13:31 +03:00
|
|
|
/*
|
|
|
|
* 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);
|
2019-08-26 19:06:53 +03:00
|
|
|
rb_erase(&bdi->rb_node, &bdi_tree);
|
2015-05-23 00:13:31 +03:00
|
|
|
list_del_rcu(&bdi->bdi_list);
|
|
|
|
spin_unlock_bh(&bdi_lock);
|
2007-10-17 10:25:47 +04:00
|
|
|
|
2015-05-23 00:13:31 +03:00
|
|
|
synchronize_rcu_expedited();
|
|
|
|
}
|
2008-04-30 11:54:32 +04:00
|
|
|
|
block: don't release bdi while request_queue has live references
bdi's are initialized in two steps, bdi_init() and bdi_register(), but
destroyed in a single step by bdi_destroy() which, for a bdi embedded
in a request_queue, is called during blk_cleanup_queue() which makes
the queue invisible and starts the draining of remaining usages.
A request_queue's user can access the congestion state of the embedded
bdi as long as it holds a reference to the queue. As such, it may
access the congested state of a queue which finished
blk_cleanup_queue() but hasn't reached blk_release_queue() yet.
Because the congested state was embedded in backing_dev_info which in
turn is embedded in request_queue, accessing the congested state after
bdi_destroy() was called was fine. The bdi was destroyed but the
memory region for the congested state remained accessible till the
queue got released.
a13f35e87140 ("writeback: don't embed root bdi_writeback_congested in
bdi_writeback") changed the situation. Now, the root congested state
which is expected to be pinned while request_queue remains accessible
is separately reference counted and the base ref is put during
bdi_destroy(). This means that the root congested state may go away
prematurely while the queue is between bdi_dstroy() and
blk_cleanup_queue(), which was detected by Andrey's KASAN tests.
The root cause of this problem is that bdi doesn't distinguish the two
steps of destruction, unregistration and release, and now the root
congested state actually requires a separate release step. To fix the
issue, this patch separates out bdi_unregister() and bdi_exit() from
bdi_destroy(). bdi_unregister() is called from blk_cleanup_queue()
and bdi_exit() from blk_release_queue(). bdi_destroy() is now just a
simple wrapper calling the two steps back-to-back.
While at it, the prototype of bdi_destroy() is moved right below
bdi_setup_and_register() so that the counterpart operations are
located together.
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: a13f35e87140 ("writeback: don't embed root bdi_writeback_congested in bdi_writeback")
Cc: stable@vger.kernel.org # v4.2+
Reported-and-tested-by: Andrey Konovalov <andreyknvl@google.com>
Link: http://lkml.kernel.org/g/CAAeHK+zUJ74Zn17=rOyxacHU18SgCfC6bsYW=6kCY5GXJBwGfQ@mail.gmail.com
Reviewed-by: Jan Kara <jack@suse.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-09-08 19:20:22 +03:00
|
|
|
void bdi_unregister(struct backing_dev_info *bdi)
|
2007-10-17 10:25:47 +04:00
|
|
|
{
|
2021-08-09 17:17:40 +03:00
|
|
|
del_timer_sync(&bdi->laptop_mode_wb_timer);
|
|
|
|
|
writeback: move backing_dev_info->wb_lock and ->worklist into bdi_writeback
Currently, a bdi (backing_dev_info) embeds single wb (bdi_writeback)
and the role of the separation is unclear. For cgroup support for
writeback IOs, a bdi will be updated to host multiple wb's where each
wb serves writeback IOs of a different cgroup on the bdi. To achieve
that, a wb should carry all states necessary for servicing writeback
IOs for a cgroup independently.
This patch moves bdi->wb_lock and ->worklist into wb.
* The lock protects bdi->worklist and bdi->wb.dwork scheduling. While
moving, rename it to wb->work_lock as wb->wb_lock is confusing.
Also, move wb->dwork downwards so that it's colocated with the new
->work_lock and ->work_list fields.
* bdi_writeback_workfn() -> wb_workfn()
bdi_wakeup_thread_delayed(bdi) -> wb_wakeup_delayed(wb)
bdi_wakeup_thread(bdi) -> wb_wakeup(wb)
bdi_queue_work(bdi, ...) -> wb_queue_work(wb, ...)
__bdi_start_writeback(bdi, ...) -> __wb_start_writeback(wb, ...)
get_next_work_item(bdi) -> get_next_work_item(wb)
* bdi_wb_shutdown() is renamed to wb_shutdown() and now takes @wb.
The function contained parts which belong to the containing bdi
rather than the wb itself - testing cap_writeback_dirty and
bdi_remove_from_list() invocation. Those are moved to
bdi_unregister().
* bdi_wb_{init|exit}() are renamed to wb_{init|exit}().
Initializations of the moved bdi->wb_lock and ->work_list are
relocated from bdi_init() to wb_init().
* As there's still only one bdi_writeback per backing_dev_info, all
uses of bdi->state are mechanically replaced with bdi->wb.state
introducing no behavior changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-05-23 00:13:30 +03:00
|
|
|
/* make sure nobody finds us on the bdi_list anymore */
|
|
|
|
bdi_remove_from_list(bdi);
|
|
|
|
wb_shutdown(&bdi->wb);
|
2017-03-23 03:36:59 +03:00
|
|
|
cgwb_bdi_unregister(bdi);
|
2011-11-11 16:29:04 +04:00
|
|
|
|
2021-12-11 01:47:11 +03:00
|
|
|
/*
|
|
|
|
* If this BDI's min ratio has been set, use bdi_set_min_ratio() to
|
|
|
|
* update the global bdi_min_ratio.
|
|
|
|
*/
|
|
|
|
if (bdi->min_ratio)
|
|
|
|
bdi_set_min_ratio(bdi, 0);
|
|
|
|
|
2015-01-21 00:05:00 +03:00
|
|
|
if (bdi->dev) {
|
|
|
|
bdi_debug_unregister(bdi);
|
|
|
|
device_unregister(bdi->dev);
|
|
|
|
bdi->dev = NULL;
|
|
|
|
}
|
2016-07-31 21:15:13 +03:00
|
|
|
|
|
|
|
if (bdi->owner) {
|
|
|
|
put_device(bdi->owner);
|
|
|
|
bdi->owner = NULL;
|
|
|
|
}
|
block: don't release bdi while request_queue has live references
bdi's are initialized in two steps, bdi_init() and bdi_register(), but
destroyed in a single step by bdi_destroy() which, for a bdi embedded
in a request_queue, is called during blk_cleanup_queue() which makes
the queue invisible and starts the draining of remaining usages.
A request_queue's user can access the congestion state of the embedded
bdi as long as it holds a reference to the queue. As such, it may
access the congested state of a queue which finished
blk_cleanup_queue() but hasn't reached blk_release_queue() yet.
Because the congested state was embedded in backing_dev_info which in
turn is embedded in request_queue, accessing the congested state after
bdi_destroy() was called was fine. The bdi was destroyed but the
memory region for the congested state remained accessible till the
queue got released.
a13f35e87140 ("writeback: don't embed root bdi_writeback_congested in
bdi_writeback") changed the situation. Now, the root congested state
which is expected to be pinned while request_queue remains accessible
is separately reference counted and the base ref is put during
bdi_destroy(). This means that the root congested state may go away
prematurely while the queue is between bdi_dstroy() and
blk_cleanup_queue(), which was detected by Andrey's KASAN tests.
The root cause of this problem is that bdi doesn't distinguish the two
steps of destruction, unregistration and release, and now the root
congested state actually requires a separate release step. To fix the
issue, this patch separates out bdi_unregister() and bdi_exit() from
bdi_destroy(). bdi_unregister() is called from blk_cleanup_queue()
and bdi_exit() from blk_release_queue(). bdi_destroy() is now just a
simple wrapper calling the two steps back-to-back.
While at it, the prototype of bdi_destroy() is moved right below
bdi_setup_and_register() so that the counterpart operations are
located together.
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: a13f35e87140 ("writeback: don't embed root bdi_writeback_congested in bdi_writeback")
Cc: stable@vger.kernel.org # v4.2+
Reported-and-tested-by: Andrey Konovalov <andreyknvl@google.com>
Link: http://lkml.kernel.org/g/CAAeHK+zUJ74Zn17=rOyxacHU18SgCfC6bsYW=6kCY5GXJBwGfQ@mail.gmail.com
Reviewed-by: Jan Kara <jack@suse.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-09-08 19:20:22 +03:00
|
|
|
}
|
2021-11-05 23:36:52 +03:00
|
|
|
EXPORT_SYMBOL(bdi_unregister);
|
2015-01-21 00:05:00 +03:00
|
|
|
|
2017-02-02 17:56:51 +03:00
|
|
|
static void release_bdi(struct kref *ref)
|
|
|
|
{
|
|
|
|
struct backing_dev_info *bdi =
|
|
|
|
container_of(ref, struct backing_dev_info, refcnt);
|
|
|
|
|
2021-11-05 23:37:01 +03:00
|
|
|
WARN_ON_ONCE(test_bit(WB_registered, &bdi->wb.state));
|
2017-04-12 13:24:48 +03:00
|
|
|
WARN_ON_ONCE(bdi->dev);
|
|
|
|
wb_exit(&bdi->wb);
|
2017-02-02 17:56:51 +03:00
|
|
|
kfree(bdi);
|
|
|
|
}
|
|
|
|
|
|
|
|
void bdi_put(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
kref_put(&bdi->refcnt, release_bdi);
|
|
|
|
}
|
2017-04-12 13:24:27 +03:00
|
|
|
EXPORT_SYMBOL(bdi_put);
|
2017-02-02 17:56:51 +03:00
|
|
|
|
2021-09-20 15:33:14 +03:00
|
|
|
struct backing_dev_info *inode_to_bdi(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct super_block *sb;
|
|
|
|
|
|
|
|
if (!inode)
|
|
|
|
return &noop_backing_dev_info;
|
|
|
|
|
|
|
|
sb = inode->i_sb;
|
|
|
|
#ifdef CONFIG_BLOCK
|
|
|
|
if (sb_is_blkdev_sb(sb))
|
|
|
|
return I_BDEV(inode)->bd_disk->bdi;
|
|
|
|
#endif
|
|
|
|
return sb->s_bdi;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(inode_to_bdi);
|
|
|
|
|
2020-05-04 15:47:54 +03:00
|
|
|
const char *bdi_dev_name(struct backing_dev_info *bdi)
|
|
|
|
{
|
|
|
|
if (!bdi || !bdi->dev)
|
|
|
|
return bdi_unknown_name;
|
2020-05-04 15:47:56 +03:00
|
|
|
return bdi->dev_name;
|
2020-05-04 15:47:54 +03:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(bdi_dev_name);
|