Revision d7831a0bdf06b9f722b947bb0c205ff7d77cebd8 authored by Richard Kennedy on 30 June 2009, 18:41:35 UTC, committed by Linus Torvalds on 01 July 2009, 01:56:01 UTC
balance_dirty_pages can overreact and move all of the dirty pages to writeback unnecessarily. balance_dirty_pages makes its decision to throttle based on the number of dirty plus writeback pages that are over the calculated limit,so it will continue to move pages even when there are plenty of pages in writeback and less than the threshold still dirty. This allows it to overshoot its limits and move all the dirty pages to writeback while waiting for the drives to catch up and empty the writeback list. A simple fio test easily demonstrates this problem. fio --name=f1 --directory=/disk1 --size=2G -rw=write --name=f2 --directory=/disk2 --size=1G --rw=write --startdelay=10 This is the simplest fix I could find, but I'm not entirely sure that it alone will be enough for all cases. But it certainly is an improvement on my desktop machine writing to 2 disks. Do we need something more for machines with large arrays where bdi_threshold * number_of_drives is greater than the dirty_ratio ? Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Jens Axboe <jens.axboe@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent df279ca
linear.c
/*
linear.c : Multiple Devices driver for Linux
Copyright (C) 1994-96 Marc ZYNGIER
<zyngier@ufr-info-p7.ibp.fr> or
<maz@gloups.fdn.fr>
Linear mode management functions.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2, or (at your option)
any later version.
You should have received a copy of the GNU General Public License
(for example /usr/src/linux/COPYING); if not, write to the Free
Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include <linux/blkdev.h>
#include <linux/raid/md_u.h>
#include <linux/seq_file.h>
#include "md.h"
#include "linear.h"
/*
* find which device holds a particular offset
*/
static inline dev_info_t *which_dev(mddev_t *mddev, sector_t sector)
{
int lo, mid, hi;
linear_conf_t *conf;
lo = 0;
hi = mddev->raid_disks - 1;
conf = rcu_dereference(mddev->private);
/*
* Binary Search
*/
while (hi > lo) {
mid = (hi + lo) / 2;
if (sector < conf->disks[mid].end_sector)
hi = mid;
else
lo = mid + 1;
}
return conf->disks + lo;
}
/**
* linear_mergeable_bvec -- tell bio layer if two requests can be merged
* @q: request queue
* @bvm: properties of new bio
* @biovec: the request that could be merged to it.
*
* Return amount of bytes we can take at this offset
*/
static int linear_mergeable_bvec(struct request_queue *q,
struct bvec_merge_data *bvm,
struct bio_vec *biovec)
{
mddev_t *mddev = q->queuedata;
dev_info_t *dev0;
unsigned long maxsectors, bio_sectors = bvm->bi_size >> 9;
sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
rcu_read_lock();
dev0 = which_dev(mddev, sector);
maxsectors = dev0->end_sector - sector;
rcu_read_unlock();
if (maxsectors < bio_sectors)
maxsectors = 0;
else
maxsectors -= bio_sectors;
if (maxsectors <= (PAGE_SIZE >> 9 ) && bio_sectors == 0)
return biovec->bv_len;
/* The bytes available at this offset could be really big,
* so we cap at 2^31 to avoid overflow */
if (maxsectors > (1 << (31-9)))
return 1<<31;
return maxsectors << 9;
}
static void linear_unplug(struct request_queue *q)
{
mddev_t *mddev = q->queuedata;
linear_conf_t *conf;
int i;
rcu_read_lock();
conf = rcu_dereference(mddev->private);
for (i=0; i < mddev->raid_disks; i++) {
struct request_queue *r_queue = bdev_get_queue(conf->disks[i].rdev->bdev);
blk_unplug(r_queue);
}
rcu_read_unlock();
}
static int linear_congested(void *data, int bits)
{
mddev_t *mddev = data;
linear_conf_t *conf;
int i, ret = 0;
rcu_read_lock();
conf = rcu_dereference(mddev->private);
for (i = 0; i < mddev->raid_disks && !ret ; i++) {
struct request_queue *q = bdev_get_queue(conf->disks[i].rdev->bdev);
ret |= bdi_congested(&q->backing_dev_info, bits);
}
rcu_read_unlock();
return ret;
}
static sector_t linear_size(mddev_t *mddev, sector_t sectors, int raid_disks)
{
linear_conf_t *conf;
sector_t array_sectors;
rcu_read_lock();
conf = rcu_dereference(mddev->private);
WARN_ONCE(sectors || raid_disks,
"%s does not support generic reshape\n", __func__);
array_sectors = conf->array_sectors;
rcu_read_unlock();
return array_sectors;
}
static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
{
linear_conf_t *conf;
mdk_rdev_t *rdev;
int i, cnt;
conf = kzalloc (sizeof (*conf) + raid_disks*sizeof(dev_info_t),
GFP_KERNEL);
if (!conf)
return NULL;
cnt = 0;
conf->array_sectors = 0;
list_for_each_entry(rdev, &mddev->disks, same_set) {
int j = rdev->raid_disk;
dev_info_t *disk = conf->disks + j;
sector_t sectors;
if (j < 0 || j >= raid_disks || disk->rdev) {
printk("linear: disk numbering problem. Aborting!\n");
goto out;
}
disk->rdev = rdev;
if (mddev->chunk_sectors) {
sectors = rdev->sectors;
sector_div(sectors, mddev->chunk_sectors);
rdev->sectors = sectors * mddev->chunk_sectors;
}
blk_queue_stack_limits(mddev->queue,
rdev->bdev->bd_disk->queue);
/* as we don't honour merge_bvec_fn, we must never risk
* violating it, so limit ->max_sector to one PAGE, as
* a one page request is never in violation.
*/
if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
queue_max_sectors(mddev->queue) > (PAGE_SIZE>>9))
blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
conf->array_sectors += rdev->sectors;
cnt++;
}
if (cnt != raid_disks) {
printk("linear: not enough drives present. Aborting!\n");
goto out;
}
/*
* Here we calculate the device offsets.
*/
conf->disks[0].end_sector = conf->disks[0].rdev->sectors;
for (i = 1; i < raid_disks; i++)
conf->disks[i].end_sector =
conf->disks[i-1].end_sector +
conf->disks[i].rdev->sectors;
return conf;
out:
kfree(conf);
return NULL;
}
static int linear_run (mddev_t *mddev)
{
linear_conf_t *conf;
if (md_check_no_bitmap(mddev))
return -EINVAL;
mddev->queue->queue_lock = &mddev->queue->__queue_lock;
conf = linear_conf(mddev, mddev->raid_disks);
if (!conf)
return 1;
mddev->private = conf;
md_set_array_sectors(mddev, linear_size(mddev, 0, 0));
blk_queue_merge_bvec(mddev->queue, linear_mergeable_bvec);
mddev->queue->unplug_fn = linear_unplug;
mddev->queue->backing_dev_info.congested_fn = linear_congested;
mddev->queue->backing_dev_info.congested_data = mddev;
return 0;
}
static void free_conf(struct rcu_head *head)
{
linear_conf_t *conf = container_of(head, linear_conf_t, rcu);
kfree(conf);
}
static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev)
{
/* Adding a drive to a linear array allows the array to grow.
* It is permitted if the new drive has a matching superblock
* already on it, with raid_disk equal to raid_disks.
* It is achieved by creating a new linear_private_data structure
* and swapping it in in-place of the current one.
* The current one is never freed until the array is stopped.
* This avoids races.
*/
linear_conf_t *newconf, *oldconf;
if (rdev->saved_raid_disk != mddev->raid_disks)
return -EINVAL;
rdev->raid_disk = rdev->saved_raid_disk;
newconf = linear_conf(mddev,mddev->raid_disks+1);
if (!newconf)
return -ENOMEM;
oldconf = rcu_dereference(mddev->private);
mddev->raid_disks++;
rcu_assign_pointer(mddev->private, newconf);
md_set_array_sectors(mddev, linear_size(mddev, 0, 0));
set_capacity(mddev->gendisk, mddev->array_sectors);
call_rcu(&oldconf->rcu, free_conf);
return 0;
}
static int linear_stop (mddev_t *mddev)
{
linear_conf_t *conf = mddev->private;
/*
* We do not require rcu protection here since
* we hold reconfig_mutex for both linear_add and
* linear_stop, so they cannot race.
* We should make sure any old 'conf's are properly
* freed though.
*/
rcu_barrier();
blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
kfree(conf);
return 0;
}
static int linear_make_request (struct request_queue *q, struct bio *bio)
{
const int rw = bio_data_dir(bio);
mddev_t *mddev = q->queuedata;
dev_info_t *tmp_dev;
sector_t start_sector;
int cpu;
if (unlikely(bio_barrier(bio))) {
bio_endio(bio, -EOPNOTSUPP);
return 0;
}
cpu = part_stat_lock();
part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
bio_sectors(bio));
part_stat_unlock();
rcu_read_lock();
tmp_dev = which_dev(mddev, bio->bi_sector);
start_sector = tmp_dev->end_sector - tmp_dev->rdev->sectors;
if (unlikely(bio->bi_sector >= (tmp_dev->end_sector)
|| (bio->bi_sector < start_sector))) {
char b[BDEVNAME_SIZE];
printk("linear_make_request: Sector %llu out of bounds on "
"dev %s: %llu sectors, offset %llu\n",
(unsigned long long)bio->bi_sector,
bdevname(tmp_dev->rdev->bdev, b),
(unsigned long long)tmp_dev->rdev->sectors,
(unsigned long long)start_sector);
rcu_read_unlock();
bio_io_error(bio);
return 0;
}
if (unlikely(bio->bi_sector + (bio->bi_size >> 9) >
tmp_dev->end_sector)) {
/* This bio crosses a device boundary, so we have to
* split it.
*/
struct bio_pair *bp;
sector_t end_sector = tmp_dev->end_sector;
rcu_read_unlock();
bp = bio_split(bio, end_sector - bio->bi_sector);
if (linear_make_request(q, &bp->bio1))
generic_make_request(&bp->bio1);
if (linear_make_request(q, &bp->bio2))
generic_make_request(&bp->bio2);
bio_pair_release(bp);
return 0;
}
bio->bi_bdev = tmp_dev->rdev->bdev;
bio->bi_sector = bio->bi_sector - start_sector
+ tmp_dev->rdev->data_offset;
rcu_read_unlock();
return 1;
}
static void linear_status (struct seq_file *seq, mddev_t *mddev)
{
seq_printf(seq, " %dk rounding", mddev->chunk_sectors / 2);
}
static struct mdk_personality linear_personality =
{
.name = "linear",
.level = LEVEL_LINEAR,
.owner = THIS_MODULE,
.make_request = linear_make_request,
.run = linear_run,
.stop = linear_stop,
.status = linear_status,
.hot_add_disk = linear_add,
.size = linear_size,
};
static int __init linear_init (void)
{
return register_md_personality (&linear_personality);
}
static void linear_exit (void)
{
unregister_md_personality (&linear_personality);
}
module_init(linear_init);
module_exit(linear_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("md-personality-1"); /* LINEAR - deprecated*/
MODULE_ALIAS("md-linear");
MODULE_ALIAS("md-level--1");
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...