2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* raid10.c : Multiple Devices driver for Linux
|
|
|
|
*
|
|
|
|
* Copyright (C) 2000-2004 Neil Brown
|
|
|
|
*
|
|
|
|
* RAID-10 support for md.
|
|
|
|
*
|
|
|
|
* Base on code in raid1.c. See raid1.c for futher copyright information.
|
|
|
|
*
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
|
2008-10-15 00:09:21 +02:00
|
|
|
#include <linux/delay.h>
|
2009-03-31 05:33:13 +02:00
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/seq_file.h>
|
2009-03-31 05:33:13 +02:00
|
|
|
#include "md.h"
|
2009-03-31 05:33:13 +02:00
|
|
|
#include "dm-bio-list.h"
|
2009-03-31 05:27:03 +02:00
|
|
|
#include "raid10.h"
|
|
|
|
#include "bitmap.h"
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* RAID10 provides a combination of RAID0 and RAID1 functionality.
|
|
|
|
* The layout of data is defined by
|
|
|
|
* chunk_size
|
|
|
|
* raid_disks
|
|
|
|
* near_copies (stored in low byte of layout)
|
|
|
|
* far_copies (stored in second byte of layout)
|
2006-06-26 09:27:41 +02:00
|
|
|
* far_offset (stored in bit 16 of layout )
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
|
|
|
* The data to be stored is divided into chunks using chunksize.
|
|
|
|
* Each device is divided into far_copies sections.
|
|
|
|
* In each section, chunks are laid out in a style similar to raid0, but
|
|
|
|
* near_copies copies of each chunk is stored (each on a different drive).
|
|
|
|
* The starting device for each section is offset near_copies from the starting
|
|
|
|
* device of the previous section.
|
2006-06-26 09:27:41 +02:00
|
|
|
* Thus they are (near_copies*far_copies) of each chunk, and each is on a different
|
2005-04-17 00:20:36 +02:00
|
|
|
* drive.
|
|
|
|
* near_copies and far_copies must be at least one, and their product is at most
|
|
|
|
* raid_disks.
|
2006-06-26 09:27:41 +02:00
|
|
|
*
|
|
|
|
* If far_offset is true, then the far_copies are handled a bit differently.
|
|
|
|
* The copies are still in different stripes, but instead of be very far apart
|
|
|
|
* on disk, there are adjacent stripes.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Number of guaranteed r10bios in case of extreme VM load:
|
|
|
|
*/
|
|
|
|
#define NR_RAID10_BIOS 256
|
|
|
|
|
|
|
|
static void unplug_slaves(mddev_t *mddev);
|
|
|
|
|
2006-01-06 09:20:13 +01:00
|
|
|
static void allow_barrier(conf_t *conf);
|
|
|
|
static void lower_barrier(conf_t *conf);
|
|
|
|
|
2005-10-07 08:46:04 +02:00
|
|
|
static void * r10bio_pool_alloc(gfp_t gfp_flags, void *data)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
conf_t *conf = data;
|
|
|
|
r10bio_t *r10_bio;
|
|
|
|
int size = offsetof(struct r10bio_s, devs[conf->copies]);
|
|
|
|
|
|
|
|
/* allocate a r10bio with room for raid_disks entries in the bios array */
|
2006-01-06 09:20:32 +01:00
|
|
|
r10_bio = kzalloc(size, gfp_flags);
|
|
|
|
if (!r10_bio)
|
2005-04-17 00:20:36 +02:00
|
|
|
unplug_slaves(conf->mddev);
|
|
|
|
|
|
|
|
return r10_bio;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void r10bio_pool_free(void *r10_bio, void *data)
|
|
|
|
{
|
|
|
|
kfree(r10_bio);
|
|
|
|
}
|
|
|
|
|
2008-08-05 07:54:14 +02:00
|
|
|
/* Maximum size of each resync request */
|
2005-04-17 00:20:36 +02:00
|
|
|
#define RESYNC_BLOCK_SIZE (64*1024)
|
|
|
|
#define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
|
2008-08-05 07:54:14 +02:00
|
|
|
/* amount of memory to reserve for resync requests */
|
|
|
|
#define RESYNC_WINDOW (1024*1024)
|
|
|
|
/* maximum number of concurrent requests, memory permitting */
|
|
|
|
#define RESYNC_DEPTH (32*1024*1024/RESYNC_BLOCK_SIZE)
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* When performing a resync, we need to read and compare, so
|
|
|
|
* we need as many pages are there are copies.
|
|
|
|
* When performing a recovery, we need 2 bios, one for read,
|
|
|
|
* one for write (we recover only one drive per r10buf)
|
|
|
|
*
|
|
|
|
*/
|
2005-10-07 08:46:04 +02:00
|
|
|
static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
conf_t *conf = data;
|
|
|
|
struct page *page;
|
|
|
|
r10bio_t *r10_bio;
|
|
|
|
struct bio *bio;
|
|
|
|
int i, j;
|
|
|
|
int nalloc;
|
|
|
|
|
|
|
|
r10_bio = r10bio_pool_alloc(gfp_flags, conf);
|
|
|
|
if (!r10_bio) {
|
|
|
|
unplug_slaves(conf->mddev);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery))
|
|
|
|
nalloc = conf->copies; /* resync */
|
|
|
|
else
|
|
|
|
nalloc = 2; /* recovery */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate bios.
|
|
|
|
*/
|
|
|
|
for (j = nalloc ; j-- ; ) {
|
|
|
|
bio = bio_alloc(gfp_flags, RESYNC_PAGES);
|
|
|
|
if (!bio)
|
|
|
|
goto out_free_bio;
|
|
|
|
r10_bio->devs[j].bio = bio;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Allocate RESYNC_PAGES data pages and attach them
|
|
|
|
* where needed.
|
|
|
|
*/
|
|
|
|
for (j = 0 ; j < nalloc; j++) {
|
|
|
|
bio = r10_bio->devs[j].bio;
|
|
|
|
for (i = 0; i < RESYNC_PAGES; i++) {
|
|
|
|
page = alloc_page(gfp_flags);
|
|
|
|
if (unlikely(!page))
|
|
|
|
goto out_free_pages;
|
|
|
|
|
|
|
|
bio->bi_io_vec[i].bv_page = page;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return r10_bio;
|
|
|
|
|
|
|
|
out_free_pages:
|
|
|
|
for ( ; i > 0 ; i--)
|
2006-01-06 09:20:40 +01:00
|
|
|
safe_put_page(bio->bi_io_vec[i-1].bv_page);
|
2005-04-17 00:20:36 +02:00
|
|
|
while (j--)
|
|
|
|
for (i = 0; i < RESYNC_PAGES ; i++)
|
2006-01-06 09:20:40 +01:00
|
|
|
safe_put_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page);
|
2005-04-17 00:20:36 +02:00
|
|
|
j = -1;
|
|
|
|
out_free_bio:
|
|
|
|
while ( ++j < nalloc )
|
|
|
|
bio_put(r10_bio->devs[j].bio);
|
|
|
|
r10bio_pool_free(r10_bio, conf);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void r10buf_pool_free(void *__r10_bio, void *data)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
conf_t *conf = data;
|
|
|
|
r10bio_t *r10bio = __r10_bio;
|
|
|
|
int j;
|
|
|
|
|
|
|
|
for (j=0; j < conf->copies; j++) {
|
|
|
|
struct bio *bio = r10bio->devs[j].bio;
|
|
|
|
if (bio) {
|
|
|
|
for (i = 0; i < RESYNC_PAGES; i++) {
|
2006-01-06 09:20:40 +01:00
|
|
|
safe_put_page(bio->bi_io_vec[i].bv_page);
|
2005-04-17 00:20:36 +02:00
|
|
|
bio->bi_io_vec[i].bv_page = NULL;
|
|
|
|
}
|
|
|
|
bio_put(bio);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
r10bio_pool_free(r10bio, conf);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void put_all_bios(conf_t *conf, r10bio_t *r10_bio)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < conf->copies; i++) {
|
|
|
|
struct bio **bio = & r10_bio->devs[i].bio;
|
2006-01-06 09:20:29 +01:00
|
|
|
if (*bio && *bio != IO_BLOCKED)
|
2005-04-17 00:20:36 +02:00
|
|
|
bio_put(*bio);
|
|
|
|
*bio = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2006-01-14 22:20:43 +01:00
|
|
|
static void free_r10bio(r10bio_t *r10_bio)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(r10_bio->mddev);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Wake up any possible resync thread that waits for the device
|
|
|
|
* to go idle.
|
|
|
|
*/
|
2006-01-06 09:20:13 +01:00
|
|
|
allow_barrier(conf);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
put_all_bios(conf, r10_bio);
|
|
|
|
mempool_free(r10_bio, conf->r10bio_pool);
|
|
|
|
}
|
|
|
|
|
2006-01-14 22:20:43 +01:00
|
|
|
static void put_buf(r10bio_t *r10_bio)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(r10_bio->mddev);
|
|
|
|
|
|
|
|
mempool_free(r10_bio, conf->r10buf_pool);
|
|
|
|
|
2006-01-06 09:20:13 +01:00
|
|
|
lower_barrier(conf);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void reschedule_retry(r10bio_t *r10_bio)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
mddev_t *mddev = r10_bio->mddev;
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
|
|
|
|
spin_lock_irqsave(&conf->device_lock, flags);
|
|
|
|
list_add(&r10_bio->retry_list, &conf->retry_list);
|
2006-01-06 09:20:28 +01:00
|
|
|
conf->nr_queued ++;
|
2005-04-17 00:20:36 +02:00
|
|
|
spin_unlock_irqrestore(&conf->device_lock, flags);
|
|
|
|
|
2008-07-25 21:03:38 +02:00
|
|
|
/* wake up frozen array... */
|
|
|
|
wake_up(&conf->wait_barrier);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
md_wakeup_thread(mddev->thread);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* raid_end_bio_io() is called when we have finished servicing a mirrored
|
|
|
|
* operation and are ready to return a success/failure code to the buffer
|
|
|
|
* cache layer.
|
|
|
|
*/
|
|
|
|
static void raid_end_bio_io(r10bio_t *r10_bio)
|
|
|
|
{
|
|
|
|
struct bio *bio = r10_bio->master_bio;
|
|
|
|
|
2007-09-27 12:47:43 +02:00
|
|
|
bio_endio(bio,
|
2005-04-17 00:20:36 +02:00
|
|
|
test_bit(R10BIO_Uptodate, &r10_bio->state) ? 0 : -EIO);
|
|
|
|
free_r10bio(r10_bio);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update disk head position estimator based on IRQ completion info.
|
|
|
|
*/
|
|
|
|
static inline void update_head_pos(int slot, r10bio_t *r10_bio)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(r10_bio->mddev);
|
|
|
|
|
|
|
|
conf->mirrors[r10_bio->devs[slot].devnum].head_position =
|
|
|
|
r10_bio->devs[slot].addr + (r10_bio->sectors);
|
|
|
|
}
|
|
|
|
|
2007-09-27 12:47:43 +02:00
|
|
|
static void raid10_end_read_request(struct bio *bio, int error)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
|
|
|
|
r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
|
|
|
|
int slot, dev;
|
|
|
|
conf_t *conf = mddev_to_conf(r10_bio->mddev);
|
|
|
|
|
|
|
|
|
|
|
|
slot = r10_bio->read_slot;
|
|
|
|
dev = r10_bio->devs[slot].devnum;
|
|
|
|
/*
|
|
|
|
* this branch is our 'one mirror IO has finished' event handler:
|
|
|
|
*/
|
2006-01-06 09:20:28 +01:00
|
|
|
update_head_pos(slot, r10_bio);
|
|
|
|
|
|
|
|
if (uptodate) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Set R10BIO_Uptodate in our master bio, so that
|
|
|
|
* we will return a good error code to the higher
|
|
|
|
* levels even if IO on some other mirrored buffer fails.
|
|
|
|
*
|
|
|
|
* The 'master' represents the composite IO operation to
|
|
|
|
* user-side. So if something waits for IO, then it will
|
|
|
|
* wait for the 'master' bio.
|
|
|
|
*/
|
|
|
|
set_bit(R10BIO_Uptodate, &r10_bio->state);
|
|
|
|
raid_end_bio_io(r10_bio);
|
2006-01-06 09:20:28 +01:00
|
|
|
} else {
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* oops, read error:
|
|
|
|
*/
|
|
|
|
char b[BDEVNAME_SIZE];
|
|
|
|
if (printk_ratelimit())
|
|
|
|
printk(KERN_ERR "raid10: %s: rescheduling sector %llu\n",
|
|
|
|
bdevname(conf->mirrors[dev].rdev->bdev,b), (unsigned long long)r10_bio->sector);
|
|
|
|
reschedule_retry(r10_bio);
|
|
|
|
}
|
|
|
|
|
|
|
|
rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
|
|
|
|
}
|
|
|
|
|
2007-09-27 12:47:43 +02:00
|
|
|
static void raid10_end_write_request(struct bio *bio, int error)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
|
|
|
|
r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
|
|
|
|
int slot, dev;
|
|
|
|
conf_t *conf = mddev_to_conf(r10_bio->mddev);
|
|
|
|
|
|
|
|
for (slot = 0; slot < conf->copies; slot++)
|
|
|
|
if (r10_bio->devs[slot].bio == bio)
|
|
|
|
break;
|
|
|
|
dev = r10_bio->devs[slot].devnum;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* this branch is our 'one mirror IO has finished' event handler:
|
|
|
|
*/
|
2006-01-06 09:20:16 +01:00
|
|
|
if (!uptodate) {
|
2005-04-17 00:20:36 +02:00
|
|
|
md_error(r10_bio->mddev, conf->mirrors[dev].rdev);
|
2006-01-06 09:20:16 +01:00
|
|
|
/* an I/O failed, we can't clear the bitmap */
|
|
|
|
set_bit(R10BIO_Degraded, &r10_bio->state);
|
|
|
|
} else
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Set R10BIO_Uptodate in our master bio, so that
|
|
|
|
* we will return a good error code for to the higher
|
|
|
|
* levels even if IO on some other mirrored buffer fails.
|
|
|
|
*
|
|
|
|
* The 'master' represents the composite IO operation to
|
|
|
|
* user-side. So if something waits for IO, then it will
|
|
|
|
* wait for the 'master' bio.
|
|
|
|
*/
|
|
|
|
set_bit(R10BIO_Uptodate, &r10_bio->state);
|
|
|
|
|
|
|
|
update_head_pos(slot, r10_bio);
|
|
|
|
|
|
|
|
/*
|
|
|
|
*
|
|
|
|
* Let's see if all mirrored write operations have finished
|
|
|
|
* already.
|
|
|
|
*/
|
|
|
|
if (atomic_dec_and_test(&r10_bio->remaining)) {
|
2006-01-06 09:20:16 +01:00
|
|
|
/* clear the bitmap if all writes complete successfully */
|
|
|
|
bitmap_endwrite(r10_bio->mddev->bitmap, r10_bio->sector,
|
|
|
|
r10_bio->sectors,
|
|
|
|
!test_bit(R10BIO_Degraded, &r10_bio->state),
|
|
|
|
0);
|
2005-04-17 00:20:36 +02:00
|
|
|
md_write_end(r10_bio->mddev);
|
|
|
|
raid_end_bio_io(r10_bio);
|
|
|
|
}
|
|
|
|
|
|
|
|
rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* RAID10 layout manager
|
|
|
|
* Aswell as the chunksize and raid_disks count, there are two
|
|
|
|
* parameters: near_copies and far_copies.
|
|
|
|
* near_copies * far_copies must be <= raid_disks.
|
|
|
|
* Normally one of these will be 1.
|
|
|
|
* If both are 1, we get raid0.
|
|
|
|
* If near_copies == raid_disks, we get raid1.
|
|
|
|
*
|
|
|
|
* Chunks are layed out in raid0 style with near_copies copies of the
|
|
|
|
* first chunk, followed by near_copies copies of the next chunk and
|
|
|
|
* so on.
|
|
|
|
* If far_copies > 1, then after 1/far_copies of the array has been assigned
|
|
|
|
* as described above, we start again with a device offset of near_copies.
|
|
|
|
* So we effectively have another copy of the whole array further down all
|
|
|
|
* the drives, but with blocks on different drives.
|
|
|
|
* With this layout, and block is never stored twice on the one device.
|
|
|
|
*
|
|
|
|
* raid10_find_phys finds the sector offset of a given virtual sector
|
2006-06-26 09:27:41 +02:00
|
|
|
* on each device that it is on.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
|
|
|
* raid10_find_virt does the reverse mapping, from a device and a
|
|
|
|
* sector offset to a virtual address
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void raid10_find_phys(conf_t *conf, r10bio_t *r10bio)
|
|
|
|
{
|
|
|
|
int n,f;
|
|
|
|
sector_t sector;
|
|
|
|
sector_t chunk;
|
|
|
|
sector_t stripe;
|
|
|
|
int dev;
|
|
|
|
|
|
|
|
int slot = 0;
|
|
|
|
|
|
|
|
/* now calculate first sector/dev */
|
|
|
|
chunk = r10bio->sector >> conf->chunk_shift;
|
|
|
|
sector = r10bio->sector & conf->chunk_mask;
|
|
|
|
|
|
|
|
chunk *= conf->near_copies;
|
|
|
|
stripe = chunk;
|
|
|
|
dev = sector_div(stripe, conf->raid_disks);
|
2006-06-26 09:27:41 +02:00
|
|
|
if (conf->far_offset)
|
|
|
|
stripe *= conf->far_copies;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
sector += stripe << conf->chunk_shift;
|
|
|
|
|
|
|
|
/* and calculate all the others */
|
|
|
|
for (n=0; n < conf->near_copies; n++) {
|
|
|
|
int d = dev;
|
|
|
|
sector_t s = sector;
|
|
|
|
r10bio->devs[slot].addr = sector;
|
|
|
|
r10bio->devs[slot].devnum = d;
|
|
|
|
slot++;
|
|
|
|
|
|
|
|
for (f = 1; f < conf->far_copies; f++) {
|
|
|
|
d += conf->near_copies;
|
|
|
|
if (d >= conf->raid_disks)
|
|
|
|
d -= conf->raid_disks;
|
|
|
|
s += conf->stride;
|
|
|
|
r10bio->devs[slot].devnum = d;
|
|
|
|
r10bio->devs[slot].addr = s;
|
|
|
|
slot++;
|
|
|
|
}
|
|
|
|
dev++;
|
|
|
|
if (dev >= conf->raid_disks) {
|
|
|
|
dev = 0;
|
|
|
|
sector += (conf->chunk_mask + 1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
BUG_ON(slot != conf->copies);
|
|
|
|
}
|
|
|
|
|
|
|
|
static sector_t raid10_find_virt(conf_t *conf, sector_t sector, int dev)
|
|
|
|
{
|
|
|
|
sector_t offset, chunk, vchunk;
|
|
|
|
|
|
|
|
offset = sector & conf->chunk_mask;
|
2006-06-26 09:27:41 +02:00
|
|
|
if (conf->far_offset) {
|
|
|
|
int fc;
|
|
|
|
chunk = sector >> conf->chunk_shift;
|
|
|
|
fc = sector_div(chunk, conf->far_copies);
|
|
|
|
dev -= fc * conf->near_copies;
|
|
|
|
if (dev < 0)
|
|
|
|
dev += conf->raid_disks;
|
|
|
|
} else {
|
2007-03-01 05:11:18 +01:00
|
|
|
while (sector >= conf->stride) {
|
2006-06-26 09:27:41 +02:00
|
|
|
sector -= conf->stride;
|
|
|
|
if (dev < conf->near_copies)
|
|
|
|
dev += conf->raid_disks - conf->near_copies;
|
|
|
|
else
|
|
|
|
dev -= conf->near_copies;
|
|
|
|
}
|
|
|
|
chunk = sector >> conf->chunk_shift;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
vchunk = chunk * conf->raid_disks + dev;
|
|
|
|
sector_div(vchunk, conf->near_copies);
|
|
|
|
return (vchunk << conf->chunk_shift) + offset;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* raid10_mergeable_bvec -- tell bio layer if a two requests can be merged
|
|
|
|
* @q: request queue
|
2008-07-03 09:53:43 +02:00
|
|
|
* @bvm: properties of new bio
|
2005-04-17 00:20:36 +02:00
|
|
|
* @biovec: the request that could be merged to it.
|
|
|
|
*
|
|
|
|
* Return amount of bytes we can accept at this offset
|
|
|
|
* If near_copies == raid_disk, there are no striping issues,
|
|
|
|
* but in that case, the function isn't called at all.
|
|
|
|
*/
|
2008-07-03 09:53:43 +02:00
|
|
|
static int raid10_mergeable_bvec(struct request_queue *q,
|
|
|
|
struct bvec_merge_data *bvm,
|
|
|
|
struct bio_vec *biovec)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
mddev_t *mddev = q->queuedata;
|
2008-07-03 09:53:43 +02:00
|
|
|
sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
|
2005-04-17 00:20:36 +02:00
|
|
|
int max;
|
|
|
|
unsigned int chunk_sectors = mddev->chunk_size >> 9;
|
2008-07-03 09:53:43 +02:00
|
|
|
unsigned int bio_sectors = bvm->bi_size >> 9;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
max = (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
|
|
|
|
if (max < 0) max = 0; /* bio_add cannot handle a negative return */
|
2008-07-03 09:53:43 +02:00
|
|
|
if (max <= biovec->bv_len && bio_sectors == 0)
|
|
|
|
return biovec->bv_len;
|
2005-04-17 00:20:36 +02:00
|
|
|
else
|
|
|
|
return max;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This routine returns the disk from which the requested read should
|
|
|
|
* be done. There is a per-array 'next expected sequential IO' sector
|
|
|
|
* number - if this matches on the next IO then we use the last disk.
|
|
|
|
* There is also a per-disk 'last know head position' sector that is
|
|
|
|
* maintained from IRQ contexts, both the normal and the resync IO
|
|
|
|
* completion handlers update this position correctly. If there is no
|
|
|
|
* perfect sequential match then we pick the disk whose head is closest.
|
|
|
|
*
|
|
|
|
* If there are 2 mirrors in the same 2 devices, performance degrades
|
|
|
|
* because position is mirror, not device based.
|
|
|
|
*
|
|
|
|
* The rdev for the device selected will have nr_pending incremented.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* FIXME: possibly should rethink readbalancing and do it differently
|
|
|
|
* depending on near_copies / far_copies geometry.
|
|
|
|
*/
|
|
|
|
static int read_balance(conf_t *conf, r10bio_t *r10_bio)
|
|
|
|
{
|
|
|
|
const unsigned long this_sector = r10_bio->sector;
|
|
|
|
int disk, slot, nslot;
|
|
|
|
const int sectors = r10_bio->sectors;
|
|
|
|
sector_t new_distance, current_distance;
|
2005-11-09 06:39:27 +01:00
|
|
|
mdk_rdev_t *rdev;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
raid10_find_phys(conf, r10_bio);
|
|
|
|
rcu_read_lock();
|
|
|
|
/*
|
|
|
|
* Check if we can balance. We can balance on the whole
|
2006-01-06 09:20:16 +01:00
|
|
|
* device if no resync is going on (recovery is ok), or below
|
|
|
|
* the resync window. We take the first readable disk when
|
|
|
|
* above the resync window.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
if (conf->mddev->recovery_cp < MaxSector
|
|
|
|
&& (this_sector + sectors >= conf->next_resync)) {
|
|
|
|
/* make sure that disk is operational */
|
|
|
|
slot = 0;
|
|
|
|
disk = r10_bio->devs[slot].devnum;
|
|
|
|
|
2005-11-09 06:39:27 +01:00
|
|
|
while ((rdev = rcu_dereference(conf->mirrors[disk].rdev)) == NULL ||
|
2006-01-06 09:20:29 +01:00
|
|
|
r10_bio->devs[slot].bio == IO_BLOCKED ||
|
2005-11-09 06:39:31 +01:00
|
|
|
!test_bit(In_sync, &rdev->flags)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
slot++;
|
|
|
|
if (slot == conf->copies) {
|
|
|
|
slot = 0;
|
|
|
|
disk = -1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
disk = r10_bio->devs[slot].devnum;
|
|
|
|
}
|
|
|
|
goto rb_out;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* make sure the disk is operational */
|
|
|
|
slot = 0;
|
|
|
|
disk = r10_bio->devs[slot].devnum;
|
2005-11-09 06:39:27 +01:00
|
|
|
while ((rdev=rcu_dereference(conf->mirrors[disk].rdev)) == NULL ||
|
2006-01-06 09:20:29 +01:00
|
|
|
r10_bio->devs[slot].bio == IO_BLOCKED ||
|
2005-11-09 06:39:31 +01:00
|
|
|
!test_bit(In_sync, &rdev->flags)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
slot ++;
|
|
|
|
if (slot == conf->copies) {
|
|
|
|
disk = -1;
|
|
|
|
goto rb_out;
|
|
|
|
}
|
|
|
|
disk = r10_bio->devs[slot].devnum;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-09-10 01:23:40 +02:00
|
|
|
current_distance = abs(r10_bio->devs[slot].addr -
|
|
|
|
conf->mirrors[disk].head_position);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-03-04 23:29:34 +01:00
|
|
|
/* Find the disk whose head is closest,
|
|
|
|
* or - for far > 1 - find the closest to partition beginning */
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
for (nslot = slot; nslot < conf->copies; nslot++) {
|
|
|
|
int ndisk = r10_bio->devs[nslot].devnum;
|
|
|
|
|
|
|
|
|
2005-11-09 06:39:27 +01:00
|
|
|
if ((rdev=rcu_dereference(conf->mirrors[ndisk].rdev)) == NULL ||
|
2006-01-06 09:20:29 +01:00
|
|
|
r10_bio->devs[nslot].bio == IO_BLOCKED ||
|
2005-11-09 06:39:31 +01:00
|
|
|
!test_bit(In_sync, &rdev->flags))
|
2005-04-17 00:20:36 +02:00
|
|
|
continue;
|
|
|
|
|
2005-11-28 22:44:09 +01:00
|
|
|
/* This optimisation is debatable, and completely destroys
|
|
|
|
* sequential read speed for 'far copies' arrays. So only
|
|
|
|
* keep it for 'near' arrays, and review those later.
|
|
|
|
*/
|
|
|
|
if (conf->near_copies > 1 && !atomic_read(&rdev->nr_pending)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
disk = ndisk;
|
|
|
|
slot = nslot;
|
|
|
|
break;
|
|
|
|
}
|
2008-03-04 23:29:34 +01:00
|
|
|
|
|
|
|
/* for far > 1 always use the lowest address */
|
|
|
|
if (conf->far_copies > 1)
|
|
|
|
new_distance = r10_bio->devs[nslot].addr;
|
|
|
|
else
|
|
|
|
new_distance = abs(r10_bio->devs[nslot].addr -
|
|
|
|
conf->mirrors[ndisk].head_position);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (new_distance < current_distance) {
|
|
|
|
current_distance = new_distance;
|
|
|
|
disk = ndisk;
|
|
|
|
slot = nslot;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
rb_out:
|
|
|
|
r10_bio->read_slot = slot;
|
|
|
|
/* conf->next_seq_sect = this_sector + sectors;*/
|
|
|
|
|
2005-11-09 06:39:27 +01:00
|
|
|
if (disk >= 0 && (rdev=rcu_dereference(conf->mirrors[disk].rdev))!= NULL)
|
2005-04-17 00:20:36 +02:00
|
|
|
atomic_inc(&conf->mirrors[disk].rdev->nr_pending);
|
2006-02-03 12:03:41 +01:00
|
|
|
else
|
|
|
|
disk = -1;
|
2005-04-17 00:20:36 +02:00
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return disk;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void unplug_slaves(mddev_t *mddev)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
int i;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
for (i=0; i<mddev->raid_disks; i++) {
|
2005-11-09 06:39:27 +01:00
|
|
|
mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
|
2005-11-09 06:39:31 +01:00
|
|
|
if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
|
2007-07-24 09:28:11 +02:00
|
|
|
struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
atomic_inc(&rdev->nr_pending);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
2007-11-07 20:26:56 +01:00
|
|
|
blk_unplug(r_queue);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
rdev_dec_pending(rdev, mddev);
|
|
|
|
rcu_read_lock();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
}
|
|
|
|
|
2007-07-24 09:28:11 +02:00
|
|
|
static void raid10_unplug(struct request_queue *q)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-01-06 09:20:16 +01:00
|
|
|
mddev_t *mddev = q->queuedata;
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
unplug_slaves(q->queuedata);
|
2006-01-06 09:20:16 +01:00
|
|
|
md_wakeup_thread(mddev->thread);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-10-03 10:15:54 +02:00
|
|
|
static int raid10_congested(void *data, int bits)
|
|
|
|
{
|
|
|
|
mddev_t *mddev = data;
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
int i, ret = 0;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
for (i = 0; i < mddev->raid_disks && ret == 0; i++) {
|
|
|
|
mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
|
|
|
|
if (rdev && !test_bit(Faulty, &rdev->flags)) {
|
2007-07-24 09:28:11 +02:00
|
|
|
struct request_queue *q = bdev_get_queue(rdev->bdev);
|
2006-10-03 10:15:54 +02:00
|
|
|
|
|
|
|
ret |= bdi_congested(&q->backing_dev_info, bits);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-03-04 23:29:29 +01:00
|
|
|
static int flush_pending_writes(conf_t *conf)
|
|
|
|
{
|
|
|
|
/* Any writes that have been queued but are awaiting
|
|
|
|
* bitmap updates get flushed here.
|
|
|
|
* We return 1 if any requests were actually submitted.
|
|
|
|
*/
|
|
|
|
int rv = 0;
|
|
|
|
|
|
|
|
spin_lock_irq(&conf->device_lock);
|
|
|
|
|
|
|
|
if (conf->pending_bio_list.head) {
|
|
|
|
struct bio *bio;
|
|
|
|
bio = bio_list_get(&conf->pending_bio_list);
|
|
|
|
blk_remove_plug(conf->mddev->queue);
|
|
|
|
spin_unlock_irq(&conf->device_lock);
|
|
|
|
/* flush any pending bitmap writes to disk
|
|
|
|
* before proceeding w/ I/O */
|
|
|
|
bitmap_unplug(conf->mddev->bitmap);
|
|
|
|
|
|
|
|
while (bio) { /* submit pending writes */
|
|
|
|
struct bio *next = bio->bi_next;
|
|
|
|
bio->bi_next = NULL;
|
|
|
|
generic_make_request(bio);
|
|
|
|
bio = next;
|
|
|
|
}
|
|
|
|
rv = 1;
|
|
|
|
} else
|
|
|
|
spin_unlock_irq(&conf->device_lock);
|
|
|
|
return rv;
|
|
|
|
}
|
2006-01-06 09:20:13 +01:00
|
|
|
/* Barriers....
|
|
|
|
* Sometimes we need to suspend IO while we do something else,
|
|
|
|
* either some resync/recovery, or reconfigure the array.
|
|
|
|
* To do this we raise a 'barrier'.
|
|
|
|
* The 'barrier' is a counter that can be raised multiple times
|
|
|
|
* to count how many activities are happening which preclude
|
|
|
|
* normal IO.
|
|
|
|
* We can only raise the barrier if there is no pending IO.
|
|
|
|
* i.e. if nr_pending == 0.
|
|
|
|
* We choose only to raise the barrier if no-one is waiting for the
|
|
|
|
* barrier to go down. This means that as soon as an IO request
|
|
|
|
* is ready, no other operations which require a barrier will start
|
|
|
|
* until the IO request has had a chance.
|
|
|
|
*
|
|
|
|
* So: regular IO calls 'wait_barrier'. When that returns there
|
|
|
|
* is no backgroup IO happening, It must arrange to call
|
|
|
|
* allow_barrier when it has finished its IO.
|
|
|
|
* backgroup IO calls must call raise_barrier. Once that returns
|
|
|
|
* there is no normal IO happeing. It must arrange to call
|
|
|
|
* lower_barrier when the particular background IO completes.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
static void raise_barrier(conf_t *conf, int force)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-01-06 09:20:16 +01:00
|
|
|
BUG_ON(force && !conf->barrier);
|
2005-04-17 00:20:36 +02:00
|
|
|
spin_lock_irq(&conf->resync_lock);
|
2006-01-06 09:20:13 +01:00
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
/* Wait until no block IO is waiting (unless 'force') */
|
|
|
|
wait_event_lock_irq(conf->wait_barrier, force || !conf->nr_waiting,
|
2006-01-06 09:20:13 +01:00
|
|
|
conf->resync_lock,
|
|
|
|
raid10_unplug(conf->mddev->queue));
|
|
|
|
|
|
|
|
/* block any new IO from starting */
|
|
|
|
conf->barrier++;
|
|
|
|
|
|
|
|
/* No wait for all pending IO to complete */
|
|
|
|
wait_event_lock_irq(conf->wait_barrier,
|
|
|
|
!conf->nr_pending && conf->barrier < RESYNC_DEPTH,
|
|
|
|
conf->resync_lock,
|
|
|
|
raid10_unplug(conf->mddev->queue));
|
|
|
|
|
|
|
|
spin_unlock_irq(&conf->resync_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void lower_barrier(conf_t *conf)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
spin_lock_irqsave(&conf->resync_lock, flags);
|
|
|
|
conf->barrier--;
|
|
|
|
spin_unlock_irqrestore(&conf->resync_lock, flags);
|
|
|
|
wake_up(&conf->wait_barrier);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void wait_barrier(conf_t *conf)
|
|
|
|
{
|
|
|
|
spin_lock_irq(&conf->resync_lock);
|
|
|
|
if (conf->barrier) {
|
|
|
|
conf->nr_waiting++;
|
|
|
|
wait_event_lock_irq(conf->wait_barrier, !conf->barrier,
|
|
|
|
conf->resync_lock,
|
|
|
|
raid10_unplug(conf->mddev->queue));
|
|
|
|
conf->nr_waiting--;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2006-01-06 09:20:13 +01:00
|
|
|
conf->nr_pending++;
|
2005-04-17 00:20:36 +02:00
|
|
|
spin_unlock_irq(&conf->resync_lock);
|
|
|
|
}
|
|
|
|
|
2006-01-06 09:20:13 +01:00
|
|
|
static void allow_barrier(conf_t *conf)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
spin_lock_irqsave(&conf->resync_lock, flags);
|
|
|
|
conf->nr_pending--;
|
|
|
|
spin_unlock_irqrestore(&conf->resync_lock, flags);
|
|
|
|
wake_up(&conf->wait_barrier);
|
|
|
|
}
|
|
|
|
|
2006-01-06 09:20:28 +01:00
|
|
|
static void freeze_array(conf_t *conf)
|
|
|
|
{
|
|
|
|
/* stop syncio and normal IO and wait for everything to
|
2006-01-06 09:20:42 +01:00
|
|
|
* go quiet.
|
2006-01-06 09:20:28 +01:00
|
|
|
* We increment barrier and nr_waiting, and then
|
2008-03-04 23:29:35 +01:00
|
|
|
* wait until nr_pending match nr_queued+1
|
|
|
|
* This is called in the context of one normal IO request
|
|
|
|
* that has failed. Thus any sync request that might be pending
|
|
|
|
* will be blocked by nr_pending, and we need to wait for
|
|
|
|
* pending IO requests to complete or be queued for re-try.
|
|
|
|
* Thus the number queued (nr_queued) plus this request (1)
|
|
|
|
* must match the number of pending IOs (nr_pending) before
|
|
|
|
* we continue.
|
2006-01-06 09:20:28 +01:00
|
|
|
*/
|
|
|
|
spin_lock_irq(&conf->resync_lock);
|
|
|
|
conf->barrier++;
|
|
|
|
conf->nr_waiting++;
|
|
|
|
wait_event_lock_irq(conf->wait_barrier,
|
2008-03-04 23:29:35 +01:00
|
|
|
conf->nr_pending == conf->nr_queued+1,
|
2006-01-06 09:20:28 +01:00
|
|
|
conf->resync_lock,
|
2008-03-04 23:29:29 +01:00
|
|
|
({ flush_pending_writes(conf);
|
|
|
|
raid10_unplug(conf->mddev->queue); }));
|
2006-01-06 09:20:28 +01:00
|
|
|
spin_unlock_irq(&conf->resync_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void unfreeze_array(conf_t *conf)
|
|
|
|
{
|
|
|
|
/* reverse the effect of the freeze */
|
|
|
|
spin_lock_irq(&conf->resync_lock);
|
|
|
|
conf->barrier--;
|
|
|
|
conf->nr_waiting--;
|
|
|
|
wake_up(&conf->wait_barrier);
|
|
|
|
spin_unlock_irq(&conf->resync_lock);
|
|
|
|
}
|
|
|
|
|
2007-07-24 09:28:11 +02:00
|
|
|
static int make_request(struct request_queue *q, struct bio * bio)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
mddev_t *mddev = q->queuedata;
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
mirror_info_t *mirror;
|
|
|
|
r10bio_t *r10_bio;
|
|
|
|
struct bio *read_bio;
|
2008-08-25 12:47:21 +02:00
|
|
|
int cpu;
|
2005-04-17 00:20:36 +02:00
|
|
|
int i;
|
|
|
|
int chunk_sects = conf->chunk_mask + 1;
|
2005-11-01 09:26:16 +01:00
|
|
|
const int rw = bio_data_dir(bio);
|
2007-01-11 08:15:37 +01:00
|
|
|
const int do_sync = bio_sync(bio);
|
2006-01-06 09:20:16 +01:00
|
|
|
struct bio_list bl;
|
|
|
|
unsigned long flags;
|
2008-04-30 09:52:32 +02:00
|
|
|
mdk_rdev_t *blocked_rdev;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-09-10 01:23:41 +02:00
|
|
|
if (unlikely(bio_barrier(bio))) {
|
2007-09-27 12:47:43 +02:00
|
|
|
bio_endio(bio, -EOPNOTSUPP);
|
2005-09-10 01:23:41 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/* If this request crosses a chunk boundary, we need to
|
|
|
|
* split it. This will only happen for 1 PAGE (or less) requests.
|
|
|
|
*/
|
|
|
|
if (unlikely( (bio->bi_sector & conf->chunk_mask) + (bio->bi_size >> 9)
|
|
|
|
> chunk_sects &&
|
|
|
|
conf->near_copies < conf->raid_disks)) {
|
|
|
|
struct bio_pair *bp;
|
|
|
|
/* Sanity check -- queue functions should prevent this happening */
|
|
|
|
if (bio->bi_vcnt != 1 ||
|
|
|
|
bio->bi_idx != 0)
|
|
|
|
goto bad_map;
|
|
|
|
/* This is a one page bio that upper layers
|
|
|
|
* refuse to split for us, so we need to split it.
|
|
|
|
*/
|
2008-10-09 08:57:05 +02:00
|
|
|
bp = bio_split(bio,
|
2005-04-17 00:20:36 +02:00
|
|
|
chunk_sects - (bio->bi_sector & (chunk_sects - 1)) );
|
|
|
|
if (make_request(q, &bp->bio1))
|
|
|
|
generic_make_request(&bp->bio1);
|
|
|
|
if (make_request(q, &bp->bio2))
|
|
|
|
generic_make_request(&bp->bio2);
|
|
|
|
|
|
|
|
bio_pair_release(bp);
|
|
|
|
return 0;
|
|
|
|
bad_map:
|
|
|
|
printk("raid10_make_request bug: can't convert block across chunks"
|
|
|
|
" or bigger than %dk %llu %d\n", chunk_sects/2,
|
|
|
|
(unsigned long long)bio->bi_sector, bio->bi_size >> 10);
|
|
|
|
|
2007-09-27 12:47:43 +02:00
|
|
|
bio_io_error(bio);
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2005-06-22 02:17:26 +02:00
|
|
|
md_write_start(mddev, bio);
|
2005-06-22 02:17:12 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Register the new request and wait if the reconstruction
|
|
|
|
* thread has put up a bar for new requests.
|
|
|
|
* Continue immediately if no resync is active currently.
|
|
|
|
*/
|
2006-01-06 09:20:13 +01:00
|
|
|
wait_barrier(conf);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-08-25 12:56:14 +02:00
|
|
|
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();
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
r10_bio = mempool_alloc(conf->r10bio_pool, GFP_NOIO);
|
|
|
|
|
|
|
|
r10_bio->master_bio = bio;
|
|
|
|
r10_bio->sectors = bio->bi_size >> 9;
|
|
|
|
|
|
|
|
r10_bio->mddev = mddev;
|
|
|
|
r10_bio->sector = bio->bi_sector;
|
2006-01-06 09:20:16 +01:00
|
|
|
r10_bio->state = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-11-01 09:26:16 +01:00
|
|
|
if (rw == READ) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* read balancing logic:
|
|
|
|
*/
|
|
|
|
int disk = read_balance(conf, r10_bio);
|
|
|
|
int slot = r10_bio->read_slot;
|
|
|
|
if (disk < 0) {
|
|
|
|
raid_end_bio_io(r10_bio);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
mirror = conf->mirrors + disk;
|
|
|
|
|
|
|
|
read_bio = bio_clone(bio, GFP_NOIO);
|
|
|
|
|
|
|
|
r10_bio->devs[slot].bio = read_bio;
|
|
|
|
|
|
|
|
read_bio->bi_sector = r10_bio->devs[slot].addr +
|
|
|
|
mirror->rdev->data_offset;
|
|
|
|
read_bio->bi_bdev = mirror->rdev->bdev;
|
|
|
|
read_bio->bi_end_io = raid10_end_read_request;
|
2007-01-11 08:15:37 +01:00
|
|
|
read_bio->bi_rw = READ | do_sync;
|
2005-04-17 00:20:36 +02:00
|
|
|
read_bio->bi_private = r10_bio;
|
|
|
|
|
|
|
|
generic_make_request(read_bio);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* WRITE:
|
|
|
|
*/
|
2008-04-30 09:52:32 +02:00
|
|
|
/* first select target devices under rcu_lock and
|
2005-04-17 00:20:36 +02:00
|
|
|
* inc refcount on their rdev. Record them by setting
|
|
|
|
* bios[x] to bio
|
|
|
|
*/
|
|
|
|
raid10_find_phys(conf, r10_bio);
|
2008-04-30 09:52:32 +02:00
|
|
|
retry_write:
|
2008-05-07 05:42:32 +02:00
|
|
|
blocked_rdev = NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
rcu_read_lock();
|
|
|
|
for (i = 0; i < conf->copies; i++) {
|
|
|
|
int d = r10_bio->devs[i].devnum;
|
2005-11-09 06:39:27 +01:00
|
|
|
mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[d].rdev);
|
2008-04-30 09:52:32 +02:00
|
|
|
if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
|
|
|
|
atomic_inc(&rdev->nr_pending);
|
|
|
|
blocked_rdev = rdev;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (rdev && !test_bit(Faulty, &rdev->flags)) {
|
2005-11-09 06:39:27 +01:00
|
|
|
atomic_inc(&rdev->nr_pending);
|
2005-04-17 00:20:36 +02:00
|
|
|
r10_bio->devs[i].bio = bio;
|
2006-01-06 09:20:16 +01:00
|
|
|
} else {
|
2005-04-17 00:20:36 +02:00
|
|
|
r10_bio->devs[i].bio = NULL;
|
2006-01-06 09:20:16 +01:00
|
|
|
set_bit(R10BIO_Degraded, &r10_bio->state);
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
2008-04-30 09:52:32 +02:00
|
|
|
if (unlikely(blocked_rdev)) {
|
|
|
|
/* Have to wait for this device to get unblocked, then retry */
|
|
|
|
int j;
|
|
|
|
int d;
|
|
|
|
|
|
|
|
for (j = 0; j < i; j++)
|
|
|
|
if (r10_bio->devs[j].bio) {
|
|
|
|
d = r10_bio->devs[j].devnum;
|
|
|
|
rdev_dec_pending(conf->mirrors[d].rdev, mddev);
|
|
|
|
}
|
|
|
|
allow_barrier(conf);
|
|
|
|
md_wait_for_blocked_rdev(blocked_rdev, mddev);
|
|
|
|
wait_barrier(conf);
|
|
|
|
goto retry_write;
|
|
|
|
}
|
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
atomic_set(&r10_bio->remaining, 0);
|
2005-06-22 02:17:12 +02:00
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
bio_list_init(&bl);
|
2005-04-17 00:20:36 +02:00
|
|
|
for (i = 0; i < conf->copies; i++) {
|
|
|
|
struct bio *mbio;
|
|
|
|
int d = r10_bio->devs[i].devnum;
|
|
|
|
if (!r10_bio->devs[i].bio)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
mbio = bio_clone(bio, GFP_NOIO);
|
|
|
|
r10_bio->devs[i].bio = mbio;
|
|
|
|
|
|
|
|
mbio->bi_sector = r10_bio->devs[i].addr+
|
|
|
|
conf->mirrors[d].rdev->data_offset;
|
|
|
|
mbio->bi_bdev = conf->mirrors[d].rdev->bdev;
|
|
|
|
mbio->bi_end_io = raid10_end_write_request;
|
2007-01-11 08:15:37 +01:00
|
|
|
mbio->bi_rw = WRITE | do_sync;
|
2005-04-17 00:20:36 +02:00
|
|
|
mbio->bi_private = r10_bio;
|
|
|
|
|
|
|
|
atomic_inc(&r10_bio->remaining);
|
2006-01-06 09:20:16 +01:00
|
|
|
bio_list_add(&bl, mbio);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2007-07-31 09:37:57 +02:00
|
|
|
if (unlikely(!atomic_read(&r10_bio->remaining))) {
|
|
|
|
/* the array is dead */
|
|
|
|
md_write_end(mddev);
|
|
|
|
raid_end_bio_io(r10_bio);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
bitmap_startwrite(mddev->bitmap, bio->bi_sector, r10_bio->sectors, 0);
|
|
|
|
spin_lock_irqsave(&conf->device_lock, flags);
|
|
|
|
bio_list_merge(&conf->pending_bio_list, &bl);
|
|
|
|
blk_plug_device(mddev->queue);
|
|
|
|
spin_unlock_irqrestore(&conf->device_lock, flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-03-04 23:29:29 +01:00
|
|
|
/* In case raid10d snuck in to freeze_array */
|
|
|
|
wake_up(&conf->wait_barrier);
|
|
|
|
|
2007-01-11 08:15:37 +01:00
|
|
|
if (do_sync)
|
|
|
|
md_wakeup_thread(mddev->thread);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void status(struct seq_file *seq, mddev_t *mddev)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (conf->near_copies < conf->raid_disks)
|
|
|
|
seq_printf(seq, " %dK chunks", mddev->chunk_size/1024);
|
|
|
|
if (conf->near_copies > 1)
|
|
|
|
seq_printf(seq, " %d near-copies", conf->near_copies);
|
2006-06-26 09:27:41 +02:00
|
|
|
if (conf->far_copies > 1) {
|
|
|
|
if (conf->far_offset)
|
|
|
|
seq_printf(seq, " %d offset-copies", conf->far_copies);
|
|
|
|
else
|
|
|
|
seq_printf(seq, " %d far-copies", conf->far_copies);
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
seq_printf(seq, " [%d/%d] [", conf->raid_disks,
|
2006-10-03 10:15:48 +02:00
|
|
|
conf->raid_disks - mddev->degraded);
|
2005-04-17 00:20:36 +02:00
|
|
|
for (i = 0; i < conf->raid_disks; i++)
|
|
|
|
seq_printf(seq, "%s",
|
|
|
|
conf->mirrors[i].rdev &&
|
2005-11-09 06:39:31 +01:00
|
|
|
test_bit(In_sync, &conf->mirrors[i].rdev->flags) ? "U" : "_");
|
2005-04-17 00:20:36 +02:00
|
|
|
seq_printf(seq, "]");
|
|
|
|
}
|
|
|
|
|
|
|
|
static void error(mddev_t *mddev, mdk_rdev_t *rdev)
|
|
|
|
{
|
|
|
|
char b[BDEVNAME_SIZE];
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If it is not operational, then we have already marked it as dead
|
|
|
|
* else if it is the last working disks, ignore the error, let the
|
|
|
|
* next level up know.
|
|
|
|
* else mark the drive as failed
|
|
|
|
*/
|
2005-11-09 06:39:31 +01:00
|
|
|
if (test_bit(In_sync, &rdev->flags)
|
2006-10-03 10:15:48 +02:00
|
|
|
&& conf->raid_disks-mddev->degraded == 1)
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Don't fail the drive, just return an IO error.
|
|
|
|
* The test should really be more sophisticated than
|
|
|
|
* "working_disks == 1", but it isn't critical, and
|
|
|
|
* can wait until we do more sophisticated "is the drive
|
|
|
|
* really dead" tests...
|
|
|
|
*/
|
|
|
|
return;
|
2006-10-03 10:15:53 +02:00
|
|
|
if (test_and_clear_bit(In_sync, &rdev->flags)) {
|
|
|
|
unsigned long flags;
|
|
|
|
spin_lock_irqsave(&conf->device_lock, flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
mddev->degraded++;
|
2006-10-03 10:15:53 +02:00
|
|
|
spin_unlock_irqrestore(&conf->device_lock, flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* if recovery is running, make sure it aborts.
|
|
|
|
*/
|
md: restart recovery cleanly after device failure.
When we get any IO error during a recovery (rebuilding a spare), we abort
the recovery and restart it.
For RAID6 (and multi-drive RAID1) it may not be best to restart at the
beginning: when multiple failures can be tolerated, the recovery may be
able to continue and re-doing all that has already been done doesn't make
sense.
We already have the infrastructure to record where a recovery is up to
and restart from there, but it is not being used properly.
This is because:
- We sometimes abort with MD_RECOVERY_ERR rather than just MD_RECOVERY_INTR,
which causes the recovery not be be checkpointed.
- We remove spares and then re-added them which loses important state
information.
The distinction between MD_RECOVERY_ERR and MD_RECOVERY_INTR really isn't
needed. If there is an error, the relevant drive will be marked as
Faulty, and that is enough to ensure correct handling of the error. So we
first remove MD_RECOVERY_ERR, changing some of the uses of it to
MD_RECOVERY_INTR.
Then we cause the attempt to remove a non-faulty device from an array to
fail (unless recovery is impossible as the array is too degraded). Then
when remove_and_add_spares attempts to remove the devices on which
recovery can continue, it will fail, they will remain in place, and
recovery will continue on them as desired.
Issue: If we are halfway through rebuilding a spare and another drive
fails, and a new spare is immediately available, do we want to:
1/ complete the current rebuild, then go back and rebuild the new spare or
2/ restart the rebuild from the start and rebuild both devices in
parallel.
Both options can be argued for. The code currently takes option 2 as
a/ this requires least code change
b/ this results in a minimally-degraded array in minimal time.
Cc: "Eivind Sarto" <ivan@kasenna.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-05-23 22:04:39 +02:00
|
|
|
set_bit(MD_RECOVERY_INTR, &mddev->recovery);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2005-11-09 06:39:31 +01:00
|
|
|
set_bit(Faulty, &rdev->flags);
|
2006-10-03 10:15:46 +02:00
|
|
|
set_bit(MD_CHANGE_DEVS, &mddev->flags);
|
2008-04-28 11:15:55 +02:00
|
|
|
printk(KERN_ALERT "raid10: Disk failure on %s, disabling device.\n"
|
|
|
|
"raid10: Operation continuing on %d devices.\n",
|
2006-10-03 10:15:48 +02:00
|
|
|
bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void print_conf(conf_t *conf)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
mirror_info_t *tmp;
|
|
|
|
|
|
|
|
printk("RAID10 conf printout:\n");
|
|
|
|
if (!conf) {
|
|
|
|
printk("(!conf)\n");
|
|
|
|
return;
|
|
|
|
}
|
2006-10-03 10:15:48 +02:00
|
|
|
printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
|
2005-04-17 00:20:36 +02:00
|
|
|
conf->raid_disks);
|
|
|
|
|
|
|
|
for (i = 0; i < conf->raid_disks; i++) {
|
|
|
|
char b[BDEVNAME_SIZE];
|
|
|
|
tmp = conf->mirrors + i;
|
|
|
|
if (tmp->rdev)
|
|
|
|
printk(" disk %d, wo:%d, o:%d, dev:%s\n",
|
2005-11-09 06:39:31 +01:00
|
|
|
i, !test_bit(In_sync, &tmp->rdev->flags),
|
|
|
|
!test_bit(Faulty, &tmp->rdev->flags),
|
2005-04-17 00:20:36 +02:00
|
|
|
bdevname(tmp->rdev->bdev,b));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void close_sync(conf_t *conf)
|
|
|
|
{
|
2006-01-06 09:20:13 +01:00
|
|
|
wait_barrier(conf);
|
|
|
|
allow_barrier(conf);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
mempool_destroy(conf->r10buf_pool);
|
|
|
|
conf->r10buf_pool = NULL;
|
|
|
|
}
|
|
|
|
|
2005-09-10 01:24:03 +02:00
|
|
|
/* check if there are enough drives for
|
|
|
|
* every block to appear on atleast one
|
|
|
|
*/
|
|
|
|
static int enough(conf_t *conf)
|
|
|
|
{
|
|
|
|
int first = 0;
|
|
|
|
|
|
|
|
do {
|
|
|
|
int n = conf->copies;
|
|
|
|
int cnt = 0;
|
|
|
|
while (n--) {
|
|
|
|
if (conf->mirrors[first].rdev)
|
|
|
|
cnt++;
|
|
|
|
first = (first+1) % conf->raid_disks;
|
|
|
|
}
|
|
|
|
if (cnt == 0)
|
|
|
|
return 0;
|
|
|
|
} while (first != 0);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
static int raid10_spare_active(mddev_t *mddev)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
conf_t *conf = mddev->private;
|
|
|
|
mirror_info_t *tmp;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find all non-in_sync disks within the RAID10 configuration
|
|
|
|
* and mark them in_sync
|
|
|
|
*/
|
|
|
|
for (i = 0; i < conf->raid_disks; i++) {
|
|
|
|
tmp = conf->mirrors + i;
|
|
|
|
if (tmp->rdev
|
2005-11-09 06:39:31 +01:00
|
|
|
&& !test_bit(Faulty, &tmp->rdev->flags)
|
2006-10-03 10:15:53 +02:00
|
|
|
&& !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
|
|
|
|
unsigned long flags;
|
|
|
|
spin_lock_irqsave(&conf->device_lock, flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
mddev->degraded--;
|
2006-10-03 10:15:53 +02:00
|
|
|
spin_unlock_irqrestore(&conf->device_lock, flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
print_conf(conf);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int raid10_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev->private;
|
2008-06-28 00:31:33 +02:00
|
|
|
int err = -EEXIST;
|
2005-04-17 00:20:36 +02:00
|
|
|
int mirror;
|
|
|
|
mirror_info_t *p;
|
2008-06-28 00:31:31 +02:00
|
|
|
int first = 0;
|
|
|
|
int last = mddev->raid_disks - 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (mddev->recovery_cp < MaxSector)
|
|
|
|
/* only hot-add to in-sync arrays, as recovery is
|
|
|
|
* very different from resync
|
|
|
|
*/
|
2008-06-28 00:31:33 +02:00
|
|
|
return -EBUSY;
|
2005-09-10 01:24:03 +02:00
|
|
|
if (!enough(conf))
|
2008-06-28 00:31:33 +02:00
|
|
|
return -EINVAL;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2008-11-06 07:28:20 +01:00
|
|
|
if (rdev->raid_disk >= 0)
|
2008-06-28 00:31:31 +02:00
|
|
|
first = last = rdev->raid_disk;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
if (rdev->saved_raid_disk >= 0 &&
|
2008-06-28 00:31:31 +02:00
|
|
|
rdev->saved_raid_disk >= first &&
|
2006-01-06 09:20:16 +01:00
|
|
|
conf->mirrors[rdev->saved_raid_disk].rdev == NULL)
|
|
|
|
mirror = rdev->saved_raid_disk;
|
|
|
|
else
|
2008-06-28 00:31:31 +02:00
|
|
|
mirror = first;
|
|
|
|
for ( ; mirror <= last ; mirror++)
|
2005-04-17 00:20:36 +02:00
|
|
|
if ( !(p=conf->mirrors+mirror)->rdev) {
|
|
|
|
|
|
|
|
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 &&
|
|
|
|
mddev->queue->max_sectors > (PAGE_SIZE>>9))
|
|
|
|
mddev->queue->max_sectors = (PAGE_SIZE>>9);
|
|
|
|
|
|
|
|
p->head_position = 0;
|
|
|
|
rdev->raid_disk = mirror;
|
2008-06-28 00:31:33 +02:00
|
|
|
err = 0;
|
2006-01-06 09:20:16 +01:00
|
|
|
if (rdev->saved_raid_disk != mirror)
|
|
|
|
conf->fullsync = 1;
|
2005-11-09 06:39:27 +01:00
|
|
|
rcu_assign_pointer(p->rdev, rdev);
|
2005-04-17 00:20:36 +02:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
print_conf(conf);
|
2008-06-28 00:31:33 +02:00
|
|
|
return err;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int raid10_remove_disk(mddev_t *mddev, int number)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev->private;
|
|
|
|
int err = 0;
|
|
|
|
mdk_rdev_t *rdev;
|
|
|
|
mirror_info_t *p = conf->mirrors+ number;
|
|
|
|
|
|
|
|
print_conf(conf);
|
|
|
|
rdev = p->rdev;
|
|
|
|
if (rdev) {
|
2005-11-09 06:39:31 +01:00
|
|
|
if (test_bit(In_sync, &rdev->flags) ||
|
2005-04-17 00:20:36 +02:00
|
|
|
atomic_read(&rdev->nr_pending)) {
|
|
|
|
err = -EBUSY;
|
|
|
|
goto abort;
|
|
|
|
}
|
md: restart recovery cleanly after device failure.
When we get any IO error during a recovery (rebuilding a spare), we abort
the recovery and restart it.
For RAID6 (and multi-drive RAID1) it may not be best to restart at the
beginning: when multiple failures can be tolerated, the recovery may be
able to continue and re-doing all that has already been done doesn't make
sense.
We already have the infrastructure to record where a recovery is up to
and restart from there, but it is not being used properly.
This is because:
- We sometimes abort with MD_RECOVERY_ERR rather than just MD_RECOVERY_INTR,
which causes the recovery not be be checkpointed.
- We remove spares and then re-added them which loses important state
information.
The distinction between MD_RECOVERY_ERR and MD_RECOVERY_INTR really isn't
needed. If there is an error, the relevant drive will be marked as
Faulty, and that is enough to ensure correct handling of the error. So we
first remove MD_RECOVERY_ERR, changing some of the uses of it to
MD_RECOVERY_INTR.
Then we cause the attempt to remove a non-faulty device from an array to
fail (unless recovery is impossible as the array is too degraded). Then
when remove_and_add_spares attempts to remove the devices on which
recovery can continue, it will fail, they will remain in place, and
recovery will continue on them as desired.
Issue: If we are halfway through rebuilding a spare and another drive
fails, and a new spare is immediately available, do we want to:
1/ complete the current rebuild, then go back and rebuild the new spare or
2/ restart the rebuild from the start and rebuild both devices in
parallel.
Both options can be argued for. The code currently takes option 2 as
a/ this requires least code change
b/ this results in a minimally-degraded array in minimal time.
Cc: "Eivind Sarto" <ivan@kasenna.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-05-23 22:04:39 +02:00
|
|
|
/* Only remove faulty devices in recovery
|
|
|
|
* is not possible.
|
|
|
|
*/
|
|
|
|
if (!test_bit(Faulty, &rdev->flags) &&
|
|
|
|
enough(conf)) {
|
|
|
|
err = -EBUSY;
|
|
|
|
goto abort;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
p->rdev = NULL;
|
2005-05-01 17:59:04 +02:00
|
|
|
synchronize_rcu();
|
2005-04-17 00:20:36 +02:00
|
|
|
if (atomic_read(&rdev->nr_pending)) {
|
|
|
|
/* lost the race, try later */
|
|
|
|
err = -EBUSY;
|
|
|
|
p->rdev = rdev;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
abort:
|
|
|
|
|
|
|
|
print_conf(conf);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2007-09-27 12:47:43 +02:00
|
|
|
static void end_sync_read(struct bio *bio, int error)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
|
|
|
|
conf_t *conf = mddev_to_conf(r10_bio->mddev);
|
|
|
|
int i,d;
|
|
|
|
|
|
|
|
for (i=0; i<conf->copies; i++)
|
|
|
|
if (r10_bio->devs[i].bio == bio)
|
|
|
|
break;
|
2006-04-02 13:34:29 +02:00
|
|
|
BUG_ON(i == conf->copies);
|
2005-04-17 00:20:36 +02:00
|
|
|
update_head_pos(i, r10_bio);
|
|
|
|
d = r10_bio->devs[i].devnum;
|
2006-01-06 09:20:29 +01:00
|
|
|
|
|
|
|
if (test_bit(BIO_UPTODATE, &bio->bi_flags))
|
|
|
|
set_bit(R10BIO_Uptodate, &r10_bio->state);
|
2006-01-06 09:20:52 +01:00
|
|
|
else {
|
|
|
|
atomic_add(r10_bio->sectors,
|
|
|
|
&conf->mirrors[d].rdev->corrected_errors);
|
|
|
|
if (!test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery))
|
|
|
|
md_error(r10_bio->mddev,
|
|
|
|
conf->mirrors[d].rdev);
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/* for reconstruct, we always reschedule after a read.
|
|
|
|
* for resync, only after all reads
|
|
|
|
*/
|
2009-02-25 03:18:47 +01:00
|
|
|
rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (test_bit(R10BIO_IsRecover, &r10_bio->state) ||
|
|
|
|
atomic_dec_and_test(&r10_bio->remaining)) {
|
|
|
|
/* we have read all the blocks,
|
|
|
|
* do the comparison in process context in raid10d
|
|
|
|
*/
|
|
|
|
reschedule_retry(r10_bio);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-09-27 12:47:43 +02:00
|
|
|
static void end_sync_write(struct bio *bio, int error)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
|
|
|
|
r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
|
|
|
|
mddev_t *mddev = r10_bio->mddev;
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
int i,d;
|
|
|
|
|
|
|
|
for (i = 0; i < conf->copies; i++)
|
|
|
|
if (r10_bio->devs[i].bio == bio)
|
|
|
|
break;
|
|
|
|
d = r10_bio->devs[i].devnum;
|
|
|
|
|
|
|
|
if (!uptodate)
|
|
|
|
md_error(mddev, conf->mirrors[d].rdev);
|
md: restart recovery cleanly after device failure.
When we get any IO error during a recovery (rebuilding a spare), we abort
the recovery and restart it.
For RAID6 (and multi-drive RAID1) it may not be best to restart at the
beginning: when multiple failures can be tolerated, the recovery may be
able to continue and re-doing all that has already been done doesn't make
sense.
We already have the infrastructure to record where a recovery is up to
and restart from there, but it is not being used properly.
This is because:
- We sometimes abort with MD_RECOVERY_ERR rather than just MD_RECOVERY_INTR,
which causes the recovery not be be checkpointed.
- We remove spares and then re-added them which loses important state
information.
The distinction between MD_RECOVERY_ERR and MD_RECOVERY_INTR really isn't
needed. If there is an error, the relevant drive will be marked as
Faulty, and that is enough to ensure correct handling of the error. So we
first remove MD_RECOVERY_ERR, changing some of the uses of it to
MD_RECOVERY_INTR.
Then we cause the attempt to remove a non-faulty device from an array to
fail (unless recovery is impossible as the array is too degraded). Then
when remove_and_add_spares attempts to remove the devices on which
recovery can continue, it will fail, they will remain in place, and
recovery will continue on them as desired.
Issue: If we are halfway through rebuilding a spare and another drive
fails, and a new spare is immediately available, do we want to:
1/ complete the current rebuild, then go back and rebuild the new spare or
2/ restart the rebuild from the start and rebuild both devices in
parallel.
Both options can be argued for. The code currently takes option 2 as
a/ this requires least code change
b/ this results in a minimally-degraded array in minimal time.
Cc: "Eivind Sarto" <ivan@kasenna.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-05-23 22:04:39 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
update_head_pos(i, r10_bio);
|
|
|
|
|
2009-02-25 03:18:47 +01:00
|
|
|
rdev_dec_pending(conf->mirrors[d].rdev, mddev);
|
2005-04-17 00:20:36 +02:00
|
|
|
while (atomic_dec_and_test(&r10_bio->remaining)) {
|
|
|
|
if (r10_bio->master_bio == NULL) {
|
|
|
|
/* the primary of several recovery bios */
|
2009-02-25 03:18:47 +01:00
|
|
|
sector_t s = r10_bio->sectors;
|
2005-04-17 00:20:36 +02:00
|
|
|
put_buf(r10_bio);
|
2009-02-25 03:18:47 +01:00
|
|
|
md_done_sync(mddev, s, 1);
|
2005-04-17 00:20:36 +02:00
|
|
|
break;
|
|
|
|
} else {
|
|
|
|
r10bio_t *r10_bio2 = (r10bio_t *)r10_bio->master_bio;
|
|
|
|
put_buf(r10_bio);
|
|
|
|
r10_bio = r10_bio2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Note: sync and recover and handled very differently for raid10
|
|
|
|
* This code is for resync.
|
|
|
|
* For resync, we read through virtual addresses and read all blocks.
|
|
|
|
* If there is any error, we schedule a write. The lowest numbered
|
|
|
|
* drive is authoritative.
|
|
|
|
* However requests come for physical address, so we need to map.
|
|
|
|
* For every physical address there are raid_disks/copies virtual addresses,
|
|
|
|
* which is always are least one, but is not necessarly an integer.
|
|
|
|
* This means that a physical address can span multiple chunks, so we may
|
|
|
|
* have to submit multiple io requests for a single sync request.
|
|
|
|
*/
|
|
|
|
/*
|
|
|
|
* We check if all blocks are in-sync and only write to blocks that
|
|
|
|
* aren't in sync
|
|
|
|
*/
|
|
|
|
static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
int i, first;
|
|
|
|
struct bio *tbio, *fbio;
|
|
|
|
|
|
|
|
atomic_set(&r10_bio->remaining, 1);
|
|
|
|
|
|
|
|
/* find the first device with a block */
|
|
|
|
for (i=0; i<conf->copies; i++)
|
|
|
|
if (test_bit(BIO_UPTODATE, &r10_bio->devs[i].bio->bi_flags))
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (i == conf->copies)
|
|
|
|
goto done;
|
|
|
|
|
|
|
|
first = i;
|
|
|
|
fbio = r10_bio->devs[i].bio;
|
|
|
|
|
|
|
|
/* now find blocks with errors */
|
2006-01-06 09:20:29 +01:00
|
|
|
for (i=0 ; i < conf->copies ; i++) {
|
|
|
|
int j, d;
|
|
|
|
int vcnt = r10_bio->sectors >> (PAGE_SHIFT-9);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
tbio = r10_bio->devs[i].bio;
|
2006-01-06 09:20:29 +01:00
|
|
|
|
|
|
|
if (tbio->bi_end_io != end_sync_read)
|
|
|
|
continue;
|
|
|
|
if (i == first)
|
2005-04-17 00:20:36 +02:00
|
|
|
continue;
|
2006-01-06 09:20:29 +01:00
|
|
|
if (test_bit(BIO_UPTODATE, &r10_bio->devs[i].bio->bi_flags)) {
|
|
|
|
/* We know that the bi_io_vec layout is the same for
|
|
|
|
* both 'first' and 'i', so we just compare them.
|
|
|
|
* All vec entries are PAGE_SIZE;
|
|
|
|
*/
|
|
|
|
for (j = 0; j < vcnt; j++)
|
|
|
|
if (memcmp(page_address(fbio->bi_io_vec[j].bv_page),
|
|
|
|
page_address(tbio->bi_io_vec[j].bv_page),
|
|
|
|
PAGE_SIZE))
|
|
|
|
break;
|
|
|
|
if (j == vcnt)
|
|
|
|
continue;
|
|
|
|
mddev->resync_mismatches += r10_bio->sectors;
|
|
|
|
}
|
2006-01-06 09:20:25 +01:00
|
|
|
if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
|
|
|
|
/* Don't fix anything. */
|
|
|
|
continue;
|
2005-04-17 00:20:36 +02:00
|
|
|
/* Ok, we need to write this bio
|
|
|
|
* First we need to fixup bv_offset, bv_len and
|
|
|
|
* bi_vecs, as the read request might have corrupted these
|
|
|
|
*/
|
|
|
|
tbio->bi_vcnt = vcnt;
|
|
|
|
tbio->bi_size = r10_bio->sectors << 9;
|
|
|
|
tbio->bi_idx = 0;
|
|
|
|
tbio->bi_phys_segments = 0;
|
|
|
|
tbio->bi_flags &= ~(BIO_POOL_MASK - 1);
|
|
|
|
tbio->bi_flags |= 1 << BIO_UPTODATE;
|
|
|
|
tbio->bi_next = NULL;
|
|
|
|
tbio->bi_rw = WRITE;
|
|
|
|
tbio->bi_private = r10_bio;
|
|
|
|
tbio->bi_sector = r10_bio->devs[i].addr;
|
|
|
|
|
|
|
|
for (j=0; j < vcnt ; j++) {
|
|
|
|
tbio->bi_io_vec[j].bv_offset = 0;
|
|
|
|
tbio->bi_io_vec[j].bv_len = PAGE_SIZE;
|
|
|
|
|
|
|
|
memcpy(page_address(tbio->bi_io_vec[j].bv_page),
|
|
|
|
page_address(fbio->bi_io_vec[j].bv_page),
|
|
|
|
PAGE_SIZE);
|
|
|
|
}
|
|
|
|
tbio->bi_end_io = end_sync_write;
|
|
|
|
|
|
|
|
d = r10_bio->devs[i].devnum;
|
|
|
|
atomic_inc(&conf->mirrors[d].rdev->nr_pending);
|
|
|
|
atomic_inc(&r10_bio->remaining);
|
|
|
|
md_sync_acct(conf->mirrors[d].rdev->bdev, tbio->bi_size >> 9);
|
|
|
|
|
|
|
|
tbio->bi_sector += conf->mirrors[d].rdev->data_offset;
|
|
|
|
tbio->bi_bdev = conf->mirrors[d].rdev->bdev;
|
|
|
|
generic_make_request(tbio);
|
|
|
|
}
|
|
|
|
|
|
|
|
done:
|
|
|
|
if (atomic_dec_and_test(&r10_bio->remaining)) {
|
|
|
|
md_done_sync(mddev, r10_bio->sectors, 1);
|
|
|
|
put_buf(r10_bio);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Now for the recovery code.
|
|
|
|
* Recovery happens across physical sectors.
|
|
|
|
* We recover all non-is_sync drives by finding the virtual address of
|
|
|
|
* each, and then choose a working drive that also has that virt address.
|
|
|
|
* There is a separate r10_bio for each non-in_sync drive.
|
|
|
|
* Only the first two slots are in use. The first for reading,
|
|
|
|
* The second for writing.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void recovery_request_write(mddev_t *mddev, r10bio_t *r10_bio)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
int i, d;
|
|
|
|
struct bio *bio, *wbio;
|
|
|
|
|
|
|
|
|
|
|
|
/* move the pages across to the second bio
|
|
|
|
* and submit the write request
|
|
|
|
*/
|
|
|
|
bio = r10_bio->devs[0].bio;
|
|
|
|
wbio = r10_bio->devs[1].bio;
|
|
|
|
for (i=0; i < wbio->bi_vcnt; i++) {
|
|
|
|
struct page *p = bio->bi_io_vec[i].bv_page;
|
|
|
|
bio->bi_io_vec[i].bv_page = wbio->bi_io_vec[i].bv_page;
|
|
|
|
wbio->bi_io_vec[i].bv_page = p;
|
|
|
|
}
|
|
|
|
d = r10_bio->devs[1].devnum;
|
|
|
|
|
|
|
|
atomic_inc(&conf->mirrors[d].rdev->nr_pending);
|
|
|
|
md_sync_acct(conf->mirrors[d].rdev->bdev, wbio->bi_size >> 9);
|
2006-01-06 09:20:29 +01:00
|
|
|
if (test_bit(R10BIO_Uptodate, &r10_bio->state))
|
|
|
|
generic_make_request(wbio);
|
|
|
|
else
|
2007-09-27 12:47:43 +02:00
|
|
|
bio_endio(wbio, -EIO);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This is a kernel thread which:
|
|
|
|
*
|
|
|
|
* 1. Retries failed read operations on working mirrors.
|
|
|
|
* 2. Updates the raid superblock when problems encounter.
|
2006-10-03 10:15:45 +02:00
|
|
|
* 3. Performs writes following reads for array synchronising.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
|
2006-10-03 10:15:45 +02:00
|
|
|
static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
|
|
|
|
{
|
|
|
|
int sect = 0; /* Offset from r10_bio->sector */
|
|
|
|
int sectors = r10_bio->sectors;
|
|
|
|
mdk_rdev_t*rdev;
|
|
|
|
while(sectors) {
|
|
|
|
int s = sectors;
|
|
|
|
int sl = r10_bio->read_slot;
|
|
|
|
int success = 0;
|
|
|
|
int start;
|
|
|
|
|
|
|
|
if (s > (PAGE_SIZE>>9))
|
|
|
|
s = PAGE_SIZE >> 9;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
do {
|
|
|
|
int d = r10_bio->devs[sl].devnum;
|
|
|
|
rdev = rcu_dereference(conf->mirrors[d].rdev);
|
|
|
|
if (rdev &&
|
|
|
|
test_bit(In_sync, &rdev->flags)) {
|
|
|
|
atomic_inc(&rdev->nr_pending);
|
|
|
|
rcu_read_unlock();
|
|
|
|
success = sync_page_io(rdev->bdev,
|
|
|
|
r10_bio->devs[sl].addr +
|
|
|
|
sect + rdev->data_offset,
|
|
|
|
s<<9,
|
|
|
|
conf->tmppage, READ);
|
|
|
|
rdev_dec_pending(rdev, mddev);
|
|
|
|
rcu_read_lock();
|
|
|
|
if (success)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
sl++;
|
|
|
|
if (sl == conf->copies)
|
|
|
|
sl = 0;
|
|
|
|
} while (!success && sl != r10_bio->read_slot);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
if (!success) {
|
|
|
|
/* Cannot read from anywhere -- bye bye array */
|
|
|
|
int dn = r10_bio->devs[r10_bio->read_slot].devnum;
|
|
|
|
md_error(mddev, conf->mirrors[dn].rdev);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
start = sl;
|
|
|
|
/* write it back and re-read */
|
|
|
|
rcu_read_lock();
|
|
|
|
while (sl != r10_bio->read_slot) {
|
|
|
|
int d;
|
|
|
|
if (sl==0)
|
|
|
|
sl = conf->copies;
|
|
|
|
sl--;
|
|
|
|
d = r10_bio->devs[sl].devnum;
|
|
|
|
rdev = rcu_dereference(conf->mirrors[d].rdev);
|
|
|
|
if (rdev &&
|
|
|
|
test_bit(In_sync, &rdev->flags)) {
|
|
|
|
atomic_inc(&rdev->nr_pending);
|
|
|
|
rcu_read_unlock();
|
|
|
|
atomic_add(s, &rdev->corrected_errors);
|
|
|
|
if (sync_page_io(rdev->bdev,
|
|
|
|
r10_bio->devs[sl].addr +
|
|
|
|
sect + rdev->data_offset,
|
|
|
|
s<<9, conf->tmppage, WRITE)
|
|
|
|
== 0)
|
|
|
|
/* Well, this device is dead */
|
|
|
|
md_error(mddev, rdev);
|
|
|
|
rdev_dec_pending(rdev, mddev);
|
|
|
|
rcu_read_lock();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
sl = start;
|
|
|
|
while (sl != r10_bio->read_slot) {
|
|
|
|
int d;
|
|
|
|
if (sl==0)
|
|
|
|
sl = conf->copies;
|
|
|
|
sl--;
|
|
|
|
d = r10_bio->devs[sl].devnum;
|
|
|
|
rdev = rcu_dereference(conf->mirrors[d].rdev);
|
|
|
|
if (rdev &&
|
|
|
|
test_bit(In_sync, &rdev->flags)) {
|
|
|
|
char b[BDEVNAME_SIZE];
|
|
|
|
atomic_inc(&rdev->nr_pending);
|
|
|
|
rcu_read_unlock();
|
|
|
|
if (sync_page_io(rdev->bdev,
|
|
|
|
r10_bio->devs[sl].addr +
|
|
|
|
sect + rdev->data_offset,
|
|
|
|
s<<9, conf->tmppage, READ) == 0)
|
|
|
|
/* Well, this device is dead */
|
|
|
|
md_error(mddev, rdev);
|
|
|
|
else
|
|
|
|
printk(KERN_INFO
|
|
|
|
"raid10:%s: read error corrected"
|
|
|
|
" (%d sectors at %llu on %s)\n",
|
|
|
|
mdname(mddev), s,
|
2006-10-28 19:38:32 +02:00
|
|
|
(unsigned long long)(sect+
|
|
|
|
rdev->data_offset),
|
2006-10-03 10:15:45 +02:00
|
|
|
bdevname(rdev->bdev, b));
|
|
|
|
|
|
|
|
rdev_dec_pending(rdev, mddev);
|
|
|
|
rcu_read_lock();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
sectors -= s;
|
|
|
|
sect += s;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
static void raid10d(mddev_t *mddev)
|
|
|
|
{
|
|
|
|
r10bio_t *r10_bio;
|
|
|
|
struct bio *bio;
|
|
|
|
unsigned long flags;
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
struct list_head *head = &conf->retry_list;
|
|
|
|
int unplug=0;
|
|
|
|
mdk_rdev_t *rdev;
|
|
|
|
|
|
|
|
md_check_recovery(mddev);
|
|
|
|
|
|
|
|
for (;;) {
|
|
|
|
char b[BDEVNAME_SIZE];
|
2006-01-06 09:20:16 +01:00
|
|
|
|
2008-03-04 23:29:29 +01:00
|
|
|
unplug += flush_pending_writes(conf);
|
2006-01-06 09:20:16 +01:00
|
|
|
|
2008-03-04 23:29:29 +01:00
|
|
|
spin_lock_irqsave(&conf->device_lock, flags);
|
|
|
|
if (list_empty(head)) {
|
|
|
|
spin_unlock_irqrestore(&conf->device_lock, flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
break;
|
2008-03-04 23:29:29 +01:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
r10_bio = list_entry(head->prev, r10bio_t, retry_list);
|
|
|
|
list_del(head->prev);
|
2006-01-06 09:20:28 +01:00
|
|
|
conf->nr_queued--;
|
2005-04-17 00:20:36 +02:00
|
|
|
spin_unlock_irqrestore(&conf->device_lock, flags);
|
|
|
|
|
|
|
|
mddev = r10_bio->mddev;
|
|
|
|
conf = mddev_to_conf(mddev);
|
|
|
|
if (test_bit(R10BIO_IsSync, &r10_bio->state)) {
|
|
|
|
sync_request_write(mddev, r10_bio);
|
|
|
|
unplug = 1;
|
|
|
|
} else if (test_bit(R10BIO_IsRecover, &r10_bio->state)) {
|
|
|
|
recovery_request_write(mddev, r10_bio);
|
|
|
|
unplug = 1;
|
|
|
|
} else {
|
|
|
|
int mirror;
|
2006-01-06 09:20:28 +01:00
|
|
|
/* we got a read error. Maybe the drive is bad. Maybe just
|
|
|
|
* the block and we can fix it.
|
|
|
|
* We freeze all other IO, and try reading the block from
|
|
|
|
* other devices. When we find one, we re-write
|
|
|
|
* and check it that fixes the read error.
|
|
|
|
* This is all done synchronously while the array is
|
|
|
|
* frozen.
|
|
|
|
*/
|
2006-10-03 10:15:45 +02:00
|
|
|
if (mddev->ro == 0) {
|
|
|
|
freeze_array(conf);
|
|
|
|
fix_read_error(conf, mddev, r10_bio);
|
|
|
|
unfreeze_array(conf);
|
2006-01-06 09:20:28 +01:00
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
bio = r10_bio->devs[r10_bio->read_slot].bio;
|
2006-01-06 09:20:29 +01:00
|
|
|
r10_bio->devs[r10_bio->read_slot].bio =
|
|
|
|
mddev->ro ? IO_BLOCKED : NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
mirror = read_balance(conf, r10_bio);
|
|
|
|
if (mirror == -1) {
|
|
|
|
printk(KERN_ALERT "raid10: %s: unrecoverable I/O"
|
|
|
|
" read error for block %llu\n",
|
|
|
|
bdevname(bio->bi_bdev,b),
|
|
|
|
(unsigned long long)r10_bio->sector);
|
|
|
|
raid_end_bio_io(r10_bio);
|
2007-07-31 09:37:57 +02:00
|
|
|
bio_put(bio);
|
2005-04-17 00:20:36 +02:00
|
|
|
} else {
|
2007-01-11 08:15:37 +01:00
|
|
|
const int do_sync = bio_sync(r10_bio->master_bio);
|
2007-07-31 09:37:57 +02:00
|
|
|
bio_put(bio);
|
2005-04-17 00:20:36 +02:00
|
|
|
rdev = conf->mirrors[mirror].rdev;
|
|
|
|
if (printk_ratelimit())
|
|
|
|
printk(KERN_ERR "raid10: %s: redirecting sector %llu to"
|
|
|
|
" another mirror\n",
|
|
|
|
bdevname(rdev->bdev,b),
|
|
|
|
(unsigned long long)r10_bio->sector);
|
|
|
|
bio = bio_clone(r10_bio->master_bio, GFP_NOIO);
|
|
|
|
r10_bio->devs[r10_bio->read_slot].bio = bio;
|
|
|
|
bio->bi_sector = r10_bio->devs[r10_bio->read_slot].addr
|
|
|
|
+ rdev->data_offset;
|
|
|
|
bio->bi_bdev = rdev->bdev;
|
2007-01-11 08:15:37 +01:00
|
|
|
bio->bi_rw = READ | do_sync;
|
2005-04-17 00:20:36 +02:00
|
|
|
bio->bi_private = r10_bio;
|
|
|
|
bio->bi_end_io = raid10_end_read_request;
|
|
|
|
unplug = 1;
|
|
|
|
generic_make_request(bio);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (unplug)
|
|
|
|
unplug_slaves(mddev);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static int init_resync(conf_t *conf)
|
|
|
|
{
|
|
|
|
int buffs;
|
|
|
|
|
|
|
|
buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
|
2006-04-02 13:34:29 +02:00
|
|
|
BUG_ON(conf->r10buf_pool);
|
2005-04-17 00:20:36 +02:00
|
|
|
conf->r10buf_pool = mempool_create(buffs, r10buf_pool_alloc, r10buf_pool_free, conf);
|
|
|
|
if (!conf->r10buf_pool)
|
|
|
|
return -ENOMEM;
|
|
|
|
conf->next_resync = 0;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* perform a "sync" on one "block"
|
|
|
|
*
|
|
|
|
* We need to make sure that no normal I/O request - particularly write
|
|
|
|
* requests - conflict with active sync requests.
|
|
|
|
*
|
|
|
|
* This is achieved by tracking pending requests and a 'barrier' concept
|
|
|
|
* that can be installed to exclude normal IO requests.
|
|
|
|
*
|
|
|
|
* Resync and recovery are handled very differently.
|
|
|
|
* We differentiate by looking at MD_RECOVERY_SYNC in mddev->recovery.
|
|
|
|
*
|
|
|
|
* For resync, we iterate over virtual addresses, read all copies,
|
|
|
|
* and update if there are differences. If only one copy is live,
|
|
|
|
* skip it.
|
|
|
|
* For recovery, we iterate over physical addresses, read a good
|
|
|
|
* value for each non-in_sync drive, and over-write.
|
|
|
|
*
|
|
|
|
* So, for recovery we may have several outstanding complex requests for a
|
|
|
|
* given address, one for each out-of-sync device. We model this by allocating
|
|
|
|
* a number of r10_bio structures, one for each out-of-sync device.
|
|
|
|
* As we setup these structures, we collect all bio's together into a list
|
|
|
|
* which we then process collectively to add pages, and then process again
|
|
|
|
* to pass to generic_make_request.
|
|
|
|
*
|
|
|
|
* The r10_bio structures are linked using a borrowed master_bio pointer.
|
|
|
|
* This link is counted in ->remaining. When the r10_bio that points to NULL
|
|
|
|
* has its remaining count decremented to 0, the whole complex operation
|
|
|
|
* is complete.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
|
2005-06-22 02:17:13 +02:00
|
|
|
static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
r10bio_t *r10_bio;
|
|
|
|
struct bio *biolist = NULL, *bio;
|
|
|
|
sector_t max_sector, nr_sectors;
|
|
|
|
int disk;
|
|
|
|
int i;
|
2006-01-06 09:20:16 +01:00
|
|
|
int max_sync;
|
|
|
|
int sync_blocks;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
sector_t sectors_skipped = 0;
|
|
|
|
int chunks_skipped = 0;
|
|
|
|
|
|
|
|
if (!conf->r10buf_pool)
|
|
|
|
if (init_resync(conf))
|
2005-06-22 02:17:13 +02:00
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
skipped:
|
|
|
|
max_sector = mddev->size << 1;
|
|
|
|
if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
|
|
|
|
max_sector = mddev->resync_max_sectors;
|
|
|
|
if (sector_nr >= max_sector) {
|
2006-01-06 09:20:16 +01:00
|
|
|
/* If we aborted, we need to abort the
|
|
|
|
* sync on the 'current' bitmap chucks (there can
|
|
|
|
* be several when recovering multiple devices).
|
|
|
|
* as we may have started syncing it but not finished.
|
|
|
|
* We can find the current address in
|
|
|
|
* mddev->curr_resync, but for recovery,
|
|
|
|
* we need to convert that to several
|
|
|
|
* virtual addresses.
|
|
|
|
*/
|
|
|
|
if (mddev->curr_resync < max_sector) { /* aborted */
|
|
|
|
if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
|
|
|
|
bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
|
|
|
|
&sync_blocks, 1);
|
|
|
|
else for (i=0; i<conf->raid_disks; i++) {
|
|
|
|
sector_t sect =
|
|
|
|
raid10_find_virt(conf, mddev->curr_resync, i);
|
|
|
|
bitmap_end_sync(mddev->bitmap, sect,
|
|
|
|
&sync_blocks, 1);
|
|
|
|
}
|
|
|
|
} else /* completed sync */
|
|
|
|
conf->fullsync = 0;
|
|
|
|
|
|
|
|
bitmap_close_sync(mddev->bitmap);
|
2005-04-17 00:20:36 +02:00
|
|
|
close_sync(conf);
|
2005-06-22 02:17:13 +02:00
|
|
|
*skipped = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
return sectors_skipped;
|
|
|
|
}
|
|
|
|
if (chunks_skipped >= conf->raid_disks) {
|
|
|
|
/* if there has been nothing to do on any drive,
|
|
|
|
* then there is nothing to do at all..
|
|
|
|
*/
|
2005-06-22 02:17:13 +02:00
|
|
|
*skipped = 1;
|
|
|
|
return (max_sector - sector_nr) + sectors_skipped;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2008-02-06 10:39:52 +01:00
|
|
|
if (max_sector > mddev->resync_max)
|
|
|
|
max_sector = mddev->resync_max; /* Don't do IO beyond here */
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/* make sure whole request will fit in a chunk - if chunks
|
|
|
|
* are meaningful
|
|
|
|
*/
|
|
|
|
if (conf->near_copies < conf->raid_disks &&
|
|
|
|
max_sector > (sector_nr | conf->chunk_mask))
|
|
|
|
max_sector = (sector_nr | conf->chunk_mask) + 1;
|
|
|
|
/*
|
|
|
|
* If there is non-resync activity waiting for us then
|
|
|
|
* put in a delay to throttle resync.
|
|
|
|
*/
|
2006-01-06 09:20:13 +01:00
|
|
|
if (!go_faster && conf->nr_waiting)
|
2005-04-17 00:20:36 +02:00
|
|
|
msleep_interruptible(1000);
|
|
|
|
|
|
|
|
/* Again, very different code for resync and recovery.
|
|
|
|
* Both must result in an r10bio with a list of bios that
|
|
|
|
* have bi_end_io, bi_sector, bi_bdev set,
|
|
|
|
* and bi_private set to the r10bio.
|
|
|
|
* For recovery, we may actually create several r10bios
|
|
|
|
* with 2 bios in each, that correspond to the bios in the main one.
|
|
|
|
* In this case, the subordinate r10bios link back through a
|
|
|
|
* borrowed master_bio pointer, and the counter in the master
|
|
|
|
* includes a ref from each subordinate.
|
|
|
|
*/
|
|
|
|
/* First, we decide what to do and set ->bi_end_io
|
|
|
|
* To end_sync_read if we want to read, and
|
|
|
|
* end_sync_write if we will want to write.
|
|
|
|
*/
|
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
max_sync = RESYNC_PAGES << (PAGE_SHIFT-9);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (!test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
|
|
|
|
/* recovery... the complicated one */
|
|
|
|
int i, j, k;
|
|
|
|
r10_bio = NULL;
|
|
|
|
|
|
|
|
for (i=0 ; i<conf->raid_disks; i++)
|
|
|
|
if (conf->mirrors[i].rdev &&
|
2005-11-09 06:39:31 +01:00
|
|
|
!test_bit(In_sync, &conf->mirrors[i].rdev->flags)) {
|
2006-01-06 09:20:16 +01:00
|
|
|
int still_degraded = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
/* want to reconstruct this device */
|
|
|
|
r10bio_t *rb2 = r10_bio;
|
2006-01-06 09:20:16 +01:00
|
|
|
sector_t sect = raid10_find_virt(conf, sector_nr, i);
|
|
|
|
int must_sync;
|
|
|
|
/* Unless we are doing a full sync, we only need
|
|
|
|
* to recover the block if it is set in the bitmap
|
|
|
|
*/
|
|
|
|
must_sync = bitmap_start_sync(mddev->bitmap, sect,
|
|
|
|
&sync_blocks, 1);
|
|
|
|
if (sync_blocks < max_sync)
|
|
|
|
max_sync = sync_blocks;
|
|
|
|
if (!must_sync &&
|
|
|
|
!conf->fullsync) {
|
|
|
|
/* yep, skip the sync_blocks here, but don't assume
|
|
|
|
* that there will never be anything to do here
|
|
|
|
*/
|
|
|
|
chunks_skipped = -1;
|
|
|
|
continue;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO);
|
2006-01-06 09:20:16 +01:00
|
|
|
raise_barrier(conf, rb2 != NULL);
|
2005-04-17 00:20:36 +02:00
|
|
|
atomic_set(&r10_bio->remaining, 0);
|
|
|
|
|
|
|
|
r10_bio->master_bio = (struct bio*)rb2;
|
|
|
|
if (rb2)
|
|
|
|
atomic_inc(&rb2->remaining);
|
|
|
|
r10_bio->mddev = mddev;
|
|
|
|
set_bit(R10BIO_IsRecover, &r10_bio->state);
|
2006-01-06 09:20:16 +01:00
|
|
|
r10_bio->sector = sect;
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
raid10_find_phys(conf, r10_bio);
|
2006-01-06 09:20:16 +01:00
|
|
|
/* Need to check if this section will still be
|
|
|
|
* degraded
|
|
|
|
*/
|
|
|
|
for (j=0; j<conf->copies;j++) {
|
|
|
|
int d = r10_bio->devs[j].devnum;
|
|
|
|
if (conf->mirrors[d].rdev == NULL ||
|
2006-01-06 09:20:35 +01:00
|
|
|
test_bit(Faulty, &conf->mirrors[d].rdev->flags)) {
|
2006-01-06 09:20:16 +01:00
|
|
|
still_degraded = 1;
|
2006-01-06 09:20:35 +01:00
|
|
|
break;
|
|
|
|
}
|
2006-01-06 09:20:16 +01:00
|
|
|
}
|
|
|
|
must_sync = bitmap_start_sync(mddev->bitmap, sect,
|
|
|
|
&sync_blocks, still_degraded);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
for (j=0; j<conf->copies;j++) {
|
|
|
|
int d = r10_bio->devs[j].devnum;
|
|
|
|
if (conf->mirrors[d].rdev &&
|
2005-11-09 06:39:31 +01:00
|
|
|
test_bit(In_sync, &conf->mirrors[d].rdev->flags)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
/* This is where we read from */
|
|
|
|
bio = r10_bio->devs[0].bio;
|
|
|
|
bio->bi_next = biolist;
|
|
|
|
biolist = bio;
|
|
|
|
bio->bi_private = r10_bio;
|
|
|
|
bio->bi_end_io = end_sync_read;
|
2006-12-13 09:34:13 +01:00
|
|
|
bio->bi_rw = READ;
|
2005-04-17 00:20:36 +02:00
|
|
|
bio->bi_sector = r10_bio->devs[j].addr +
|
|
|
|
conf->mirrors[d].rdev->data_offset;
|
|
|
|
bio->bi_bdev = conf->mirrors[d].rdev->bdev;
|
|
|
|
atomic_inc(&conf->mirrors[d].rdev->nr_pending);
|
|
|
|
atomic_inc(&r10_bio->remaining);
|
|
|
|
/* and we write to 'i' */
|
|
|
|
|
|
|
|
for (k=0; k<conf->copies; k++)
|
|
|
|
if (r10_bio->devs[k].devnum == i)
|
|
|
|
break;
|
2007-03-01 05:11:18 +01:00
|
|
|
BUG_ON(k == conf->copies);
|
2005-04-17 00:20:36 +02:00
|
|
|
bio = r10_bio->devs[1].bio;
|
|
|
|
bio->bi_next = biolist;
|
|
|
|
biolist = bio;
|
|
|
|
bio->bi_private = r10_bio;
|
|
|
|
bio->bi_end_io = end_sync_write;
|
2006-12-13 09:34:13 +01:00
|
|
|
bio->bi_rw = WRITE;
|
2005-04-17 00:20:36 +02:00
|
|
|
bio->bi_sector = r10_bio->devs[k].addr +
|
|
|
|
conf->mirrors[i].rdev->data_offset;
|
|
|
|
bio->bi_bdev = conf->mirrors[i].rdev->bdev;
|
|
|
|
|
|
|
|
r10_bio->devs[0].devnum = d;
|
|
|
|
r10_bio->devs[1].devnum = i;
|
|
|
|
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (j == conf->copies) {
|
2005-09-10 01:24:04 +02:00
|
|
|
/* Cannot recover, so abort the recovery */
|
|
|
|
put_buf(r10_bio);
|
md: the md RAID10 resync thread could cause a md RAID10 array deadlock
This message describes another issue about md RAID10 found by testing the
2.6.24 md RAID10 using new scsi fault injection framework.
Abstract:
When a scsi error results in disabling a disk during RAID10 recovery, the
resync threads of md RAID10 could stall.
This case, the raid array has already been broken and it may not matter. But
I think stall is not preferable. If it occurs, even shutdown or reboot will
fail because of resource busy.
The deadlock mechanism:
The r10bio_s structure has a "remaining" member to keep track of BIOs yet to
be handled when recovering. The "remaining" counter is incremented when
building a BIO in sync_request() and is decremented when finish a BIO in
end_sync_write().
If building a BIO fails for some reasons in sync_request(), the "remaining"
should be decremented if it has already been incremented. I found a case
where this decrement is forgotten. This causes a md_do_sync() deadlock
because md_do_sync() waits for md_done_sync() called by end_sync_write(), but
end_sync_write() never calls md_done_sync() because of the "remaining" counter
mismatch.
For example, this problem would be reproduced in the following case:
Personalities : [raid10]
md0 : active raid10 sdf1[4] sde1[5](F) sdd1[2] sdc1[1] sdb1[6](F)
3919616 blocks 64K chunks 2 near-copies [4/2] [_UU_]
[>....................] recovery = 2.2% (45376/1959808) finish=0.7min speed=45376K/sec
This case, sdf1 is recovering, sdb1 and sde1 are disabled.
An additional error with detaching sdd will cause a deadlock.
md0 : active raid10 sdf1[4] sde1[5](F) sdd1[6](F) sdc1[1] sdb1[7](F)
3919616 blocks 64K chunks 2 near-copies [4/1] [_U__]
[=>...................] recovery = 5.0% (99520/1959808) finish=5.9min speed=5237K/sec
2739 ? S< 0:17 [md0_raid10]
28608 ? D< 0:00 [md0_resync]
28629 pts/1 Ss 0:00 bash
28830 pts/1 R+ 0:00 ps ax
31819 ? D< 0:00 [kjournald]
The resync thread keeps working, but actually it is deadlocked.
Patch:
By this patch, the remaining counter will be decremented if needed.
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-03-04 23:29:37 +01:00
|
|
|
if (rb2)
|
|
|
|
atomic_dec(&rb2->remaining);
|
2005-09-10 01:24:04 +02:00
|
|
|
r10_bio = rb2;
|
md: restart recovery cleanly after device failure.
When we get any IO error during a recovery (rebuilding a spare), we abort
the recovery and restart it.
For RAID6 (and multi-drive RAID1) it may not be best to restart at the
beginning: when multiple failures can be tolerated, the recovery may be
able to continue and re-doing all that has already been done doesn't make
sense.
We already have the infrastructure to record where a recovery is up to
and restart from there, but it is not being used properly.
This is because:
- We sometimes abort with MD_RECOVERY_ERR rather than just MD_RECOVERY_INTR,
which causes the recovery not be be checkpointed.
- We remove spares and then re-added them which loses important state
information.
The distinction between MD_RECOVERY_ERR and MD_RECOVERY_INTR really isn't
needed. If there is an error, the relevant drive will be marked as
Faulty, and that is enough to ensure correct handling of the error. So we
first remove MD_RECOVERY_ERR, changing some of the uses of it to
MD_RECOVERY_INTR.
Then we cause the attempt to remove a non-faulty device from an array to
fail (unless recovery is impossible as the array is too degraded). Then
when remove_and_add_spares attempts to remove the devices on which
recovery can continue, it will fail, they will remain in place, and
recovery will continue on them as desired.
Issue: If we are halfway through rebuilding a spare and another drive
fails, and a new spare is immediately available, do we want to:
1/ complete the current rebuild, then go back and rebuild the new spare or
2/ restart the rebuild from the start and rebuild both devices in
parallel.
Both options can be argued for. The code currently takes option 2 as
a/ this requires least code change
b/ this results in a minimally-degraded array in minimal time.
Cc: "Eivind Sarto" <ivan@kasenna.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-05-23 22:04:39 +02:00
|
|
|
if (!test_and_set_bit(MD_RECOVERY_INTR,
|
|
|
|
&mddev->recovery))
|
2005-09-10 01:24:04 +02:00
|
|
|
printk(KERN_INFO "raid10: %s: insufficient working devices for recovery.\n",
|
|
|
|
mdname(mddev));
|
|
|
|
break;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
if (biolist == NULL) {
|
|
|
|
while (r10_bio) {
|
|
|
|
r10bio_t *rb2 = r10_bio;
|
|
|
|
r10_bio = (r10bio_t*) rb2->master_bio;
|
|
|
|
rb2->master_bio = NULL;
|
|
|
|
put_buf(rb2);
|
|
|
|
}
|
|
|
|
goto giveup;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/* resync. Schedule a read for every block at this virt offset */
|
|
|
|
int count = 0;
|
2006-01-06 09:20:16 +01:00
|
|
|
|
2009-02-25 03:18:47 +01:00
|
|
|
bitmap_cond_end_sync(mddev->bitmap, sector_nr);
|
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
if (!bitmap_start_sync(mddev->bitmap, sector_nr,
|
|
|
|
&sync_blocks, mddev->degraded) &&
|
|
|
|
!conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
|
|
|
|
/* We can skip this block */
|
|
|
|
*skipped = 1;
|
|
|
|
return sync_blocks + sectors_skipped;
|
|
|
|
}
|
|
|
|
if (sync_blocks < max_sync)
|
|
|
|
max_sync = sync_blocks;
|
2005-04-17 00:20:36 +02:00
|
|
|
r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO);
|
|
|
|
|
|
|
|
r10_bio->mddev = mddev;
|
|
|
|
atomic_set(&r10_bio->remaining, 0);
|
2006-01-06 09:20:16 +01:00
|
|
|
raise_barrier(conf, 0);
|
|
|
|
conf->next_resync = sector_nr;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
r10_bio->master_bio = NULL;
|
|
|
|
r10_bio->sector = sector_nr;
|
|
|
|
set_bit(R10BIO_IsSync, &r10_bio->state);
|
|
|
|
raid10_find_phys(conf, r10_bio);
|
|
|
|
r10_bio->sectors = (sector_nr | conf->chunk_mask) - sector_nr +1;
|
|
|
|
|
|
|
|
for (i=0; i<conf->copies; i++) {
|
|
|
|
int d = r10_bio->devs[i].devnum;
|
|
|
|
bio = r10_bio->devs[i].bio;
|
|
|
|
bio->bi_end_io = NULL;
|
2007-06-16 19:16:06 +02:00
|
|
|
clear_bit(BIO_UPTODATE, &bio->bi_flags);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (conf->mirrors[d].rdev == NULL ||
|
2005-11-09 06:39:31 +01:00
|
|
|
test_bit(Faulty, &conf->mirrors[d].rdev->flags))
|
2005-04-17 00:20:36 +02:00
|
|
|
continue;
|
|
|
|
atomic_inc(&conf->mirrors[d].rdev->nr_pending);
|
|
|
|
atomic_inc(&r10_bio->remaining);
|
|
|
|
bio->bi_next = biolist;
|
|
|
|
biolist = bio;
|
|
|
|
bio->bi_private = r10_bio;
|
|
|
|
bio->bi_end_io = end_sync_read;
|
2006-12-13 09:34:13 +01:00
|
|
|
bio->bi_rw = READ;
|
2005-04-17 00:20:36 +02:00
|
|
|
bio->bi_sector = r10_bio->devs[i].addr +
|
|
|
|
conf->mirrors[d].rdev->data_offset;
|
|
|
|
bio->bi_bdev = conf->mirrors[d].rdev->bdev;
|
|
|
|
count++;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (count < 2) {
|
|
|
|
for (i=0; i<conf->copies; i++) {
|
|
|
|
int d = r10_bio->devs[i].devnum;
|
|
|
|
if (r10_bio->devs[i].bio->bi_end_io)
|
|
|
|
rdev_dec_pending(conf->mirrors[d].rdev, mddev);
|
|
|
|
}
|
|
|
|
put_buf(r10_bio);
|
|
|
|
biolist = NULL;
|
|
|
|
goto giveup;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
for (bio = biolist; bio ; bio=bio->bi_next) {
|
|
|
|
|
|
|
|
bio->bi_flags &= ~(BIO_POOL_MASK - 1);
|
|
|
|
if (bio->bi_end_io)
|
|
|
|
bio->bi_flags |= 1 << BIO_UPTODATE;
|
|
|
|
bio->bi_vcnt = 0;
|
|
|
|
bio->bi_idx = 0;
|
|
|
|
bio->bi_phys_segments = 0;
|
|
|
|
bio->bi_size = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
nr_sectors = 0;
|
2006-01-06 09:20:16 +01:00
|
|
|
if (sector_nr + max_sync < max_sector)
|
|
|
|
max_sector = sector_nr + max_sync;
|
2005-04-17 00:20:36 +02:00
|
|
|
do {
|
|
|
|
struct page *page;
|
|
|
|
int len = PAGE_SIZE;
|
|
|
|
disk = 0;
|
|
|
|
if (sector_nr + (len>>9) > max_sector)
|
|
|
|
len = (max_sector - sector_nr) << 9;
|
|
|
|
if (len == 0)
|
|
|
|
break;
|
|
|
|
for (bio= biolist ; bio ; bio=bio->bi_next) {
|
|
|
|
page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
|
|
|
|
if (bio_add_page(bio, page, len, 0) == 0) {
|
|
|
|
/* stop here */
|
|
|
|
struct bio *bio2;
|
|
|
|
bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
|
|
|
|
for (bio2 = biolist; bio2 && bio2 != bio; bio2 = bio2->bi_next) {
|
|
|
|
/* remove last page from this bio */
|
|
|
|
bio2->bi_vcnt--;
|
|
|
|
bio2->bi_size -= len;
|
|
|
|
bio2->bi_flags &= ~(1<< BIO_SEG_VALID);
|
|
|
|
}
|
|
|
|
goto bio_full;
|
|
|
|
}
|
|
|
|
disk = i;
|
|
|
|
}
|
|
|
|
nr_sectors += len>>9;
|
|
|
|
sector_nr += len>>9;
|
|
|
|
} while (biolist->bi_vcnt < RESYNC_PAGES);
|
|
|
|
bio_full:
|
|
|
|
r10_bio->sectors = nr_sectors;
|
|
|
|
|
|
|
|
while (biolist) {
|
|
|
|
bio = biolist;
|
|
|
|
biolist = biolist->bi_next;
|
|
|
|
|
|
|
|
bio->bi_next = NULL;
|
|
|
|
r10_bio = bio->bi_private;
|
|
|
|
r10_bio->sectors = nr_sectors;
|
|
|
|
|
|
|
|
if (bio->bi_end_io == end_sync_read) {
|
|
|
|
md_sync_acct(bio->bi_bdev, nr_sectors);
|
|
|
|
generic_make_request(bio);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-06-22 02:17:13 +02:00
|
|
|
if (sectors_skipped)
|
|
|
|
/* pretend they weren't skipped, it makes
|
|
|
|
* no important difference in this case
|
|
|
|
*/
|
|
|
|
md_done_sync(mddev, sectors_skipped, 1);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
return sectors_skipped + nr_sectors;
|
|
|
|
giveup:
|
|
|
|
/* There is nowhere to write, so all non-sync
|
|
|
|
* drives must be failed, so try the next chunk...
|
|
|
|
*/
|
2009-02-25 03:18:47 +01:00
|
|
|
if (sector_nr + max_sync < max_sector)
|
|
|
|
max_sector = sector_nr + max_sync;
|
|
|
|
|
|
|
|
sectors_skipped += (max_sector - sector_nr);
|
2005-04-17 00:20:36 +02:00
|
|
|
chunks_skipped ++;
|
|
|
|
sector_nr = max_sector;
|
|
|
|
goto skipped;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int run(mddev_t *mddev)
|
|
|
|
{
|
|
|
|
conf_t *conf;
|
|
|
|
int i, disk_idx;
|
|
|
|
mirror_info_t *disk;
|
|
|
|
mdk_rdev_t *rdev;
|
2006-06-26 09:27:41 +02:00
|
|
|
int nc, fc, fo;
|
2005-04-17 00:20:36 +02:00
|
|
|
sector_t stride, size;
|
|
|
|
|
2008-10-13 02:55:12 +02:00
|
|
|
if (mddev->chunk_size < PAGE_SIZE) {
|
|
|
|
printk(KERN_ERR "md/raid10: chunk size must be "
|
|
|
|
"at least PAGE_SIZE(%ld).\n", PAGE_SIZE);
|
2006-01-06 09:20:36 +01:00
|
|
|
return -EINVAL;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2006-01-06 09:20:36 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
nc = mddev->layout & 255;
|
|
|
|
fc = (mddev->layout >> 8) & 255;
|
2006-06-26 09:27:41 +02:00
|
|
|
fo = mddev->layout & (1<<16);
|
2005-04-17 00:20:36 +02:00
|
|
|
if ((nc*fc) <2 || (nc*fc) > mddev->raid_disks ||
|
2006-06-26 09:27:41 +02:00
|
|
|
(mddev->layout >> 17)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
printk(KERN_ERR "raid10: %s: unsupported raid10 layout: 0x%8x\n",
|
|
|
|
mdname(mddev), mddev->layout);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* copy the already verified devices into our private RAID10
|
|
|
|
* bookkeeping area. [whatever we allocate in run(),
|
|
|
|
* should be freed in stop()]
|
|
|
|
*/
|
2006-01-06 09:20:28 +01:00
|
|
|
conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
|
2005-04-17 00:20:36 +02:00
|
|
|
mddev->private = conf;
|
|
|
|
if (!conf) {
|
|
|
|
printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
|
|
|
|
mdname(mddev));
|
|
|
|
goto out;
|
|
|
|
}
|
2006-01-06 09:20:28 +01:00
|
|
|
conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
|
2005-04-17 00:20:36 +02:00
|
|
|
GFP_KERNEL);
|
|
|
|
if (!conf->mirrors) {
|
|
|
|
printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
|
|
|
|
mdname(mddev));
|
|
|
|
goto out_free_conf;
|
|
|
|
}
|
2006-01-06 09:20:28 +01:00
|
|
|
|
|
|
|
conf->tmppage = alloc_page(GFP_KERNEL);
|
|
|
|
if (!conf->tmppage)
|
|
|
|
goto out_free_conf;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-03-01 05:11:18 +01:00
|
|
|
conf->mddev = mddev;
|
|
|
|
conf->raid_disks = mddev->raid_disks;
|
2005-04-17 00:20:36 +02:00
|
|
|
conf->near_copies = nc;
|
|
|
|
conf->far_copies = fc;
|
|
|
|
conf->copies = nc*fc;
|
2006-06-26 09:27:41 +02:00
|
|
|
conf->far_offset = fo;
|
2005-04-17 00:20:36 +02:00
|
|
|
conf->chunk_mask = (sector_t)(mddev->chunk_size>>9)-1;
|
|
|
|
conf->chunk_shift = ffz(~mddev->chunk_size) - 9;
|
2007-03-01 05:11:18 +01:00
|
|
|
size = mddev->size >> (conf->chunk_shift-1);
|
|
|
|
sector_div(size, fc);
|
|
|
|
size = size * conf->raid_disks;
|
|
|
|
sector_div(size, nc);
|
|
|
|
/* 'size' is now the number of chunks in the array */
|
|
|
|
/* calculate "used chunks per device" in 'stride' */
|
|
|
|
stride = size * conf->copies;
|
2007-06-16 19:16:06 +02:00
|
|
|
|
|
|
|
/* We need to round up when dividing by raid_disks to
|
|
|
|
* get the stride size.
|
|
|
|
*/
|
|
|
|
stride += conf->raid_disks - 1;
|
2007-03-01 05:11:18 +01:00
|
|
|
sector_div(stride, conf->raid_disks);
|
|
|
|
mddev->size = stride << (conf->chunk_shift-1);
|
|
|
|
|
2006-06-26 09:27:41 +02:00
|
|
|
if (fo)
|
2007-03-01 05:11:18 +01:00
|
|
|
stride = 1;
|
|
|
|
else
|
2006-06-26 09:27:41 +02:00
|
|
|
sector_div(stride, fc);
|
2007-03-01 05:11:18 +01:00
|
|
|
conf->stride = stride << conf->chunk_shift;
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
conf->r10bio_pool = mempool_create(NR_RAID10_BIOS, r10bio_pool_alloc,
|
|
|
|
r10bio_pool_free, conf);
|
|
|
|
if (!conf->r10bio_pool) {
|
|
|
|
printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
|
|
|
|
mdname(mddev));
|
|
|
|
goto out_free_conf;
|
|
|
|
}
|
|
|
|
|
2008-05-15 01:05:54 +02:00
|
|
|
spin_lock_init(&conf->device_lock);
|
|
|
|
mddev->queue->queue_lock = &conf->device_lock;
|
|
|
|
|
2009-01-08 22:31:08 +01:00
|
|
|
list_for_each_entry(rdev, &mddev->disks, same_set) {
|
2005-04-17 00:20:36 +02:00
|
|
|
disk_idx = rdev->raid_disk;
|
|
|
|
if (disk_idx >= mddev->raid_disks
|
|
|
|
|| disk_idx < 0)
|
|
|
|
continue;
|
|
|
|
disk = conf->mirrors + disk_idx;
|
|
|
|
|
|
|
|
disk->rdev = rdev;
|
|
|
|
|
|
|
|
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 &&
|
|
|
|
mddev->queue->max_sectors > (PAGE_SIZE>>9))
|
|
|
|
mddev->queue->max_sectors = (PAGE_SIZE>>9);
|
|
|
|
|
|
|
|
disk->head_position = 0;
|
|
|
|
}
|
|
|
|
INIT_LIST_HEAD(&conf->retry_list);
|
|
|
|
|
|
|
|
spin_lock_init(&conf->resync_lock);
|
2006-01-06 09:20:13 +01:00
|
|
|
init_waitqueue_head(&conf->wait_barrier);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-09-10 01:24:03 +02:00
|
|
|
/* need to check that every block has at least one working mirror */
|
|
|
|
if (!enough(conf)) {
|
|
|
|
printk(KERN_ERR "raid10: not enough operational mirrors for %s\n",
|
|
|
|
mdname(mddev));
|
2005-04-17 00:20:36 +02:00
|
|
|
goto out_free_conf;
|
|
|
|
}
|
|
|
|
|
|
|
|
mddev->degraded = 0;
|
|
|
|
for (i = 0; i < conf->raid_disks; i++) {
|
|
|
|
|
|
|
|
disk = conf->mirrors + i;
|
|
|
|
|
2006-06-26 09:27:40 +02:00
|
|
|
if (!disk->rdev ||
|
2006-10-21 19:24:07 +02:00
|
|
|
!test_bit(In_sync, &disk->rdev->flags)) {
|
2005-04-17 00:20:36 +02:00
|
|
|
disk->head_position = 0;
|
|
|
|
mddev->degraded++;
|
Ensure interrupted recovery completed properly (v1 metadata plus bitmap)
If, while assembling an array, we find a device which is not fully
in-sync with the array, it is important to set the "fullsync" flags.
This is an exact analog to the setting of this flag in hot_add_disk
methods.
Currently, only v1.x metadata supports having devices in an array
which are not fully in-sync (it keep track of how in sync they are).
The 'fullsync' flag only makes a difference when a write-intent bitmap
is being used. In this case it tells recovery to ignore the bitmap
and recovery all blocks.
This fix is already in place for raid1, but not raid5/6 or raid10.
So without this fix, a raid1 ir raid4/5/6 array with version 1.x
metadata and a write intent bitmaps, that is stopped in the middle
of a recovery, will appear to complete the recovery instantly
after it is reassembled, but the recovery will not be correct.
If you might have an array like that, issueing
echo repair > /sys/block/mdXX/md/sync_action
will make sure recovery completes properly.
Cc: <stable@kernel.org>
Signed-off-by: Neil Brown <neilb@suse.de>
2008-06-28 00:30:52 +02:00
|
|
|
if (disk->rdev)
|
|
|
|
conf->fullsync = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
mddev->thread = md_register_thread(raid10d, mddev, "%s_raid10");
|
|
|
|
if (!mddev->thread) {
|
|
|
|
printk(KERN_ERR
|
|
|
|
"raid10: couldn't allocate thread for %s\n",
|
|
|
|
mdname(mddev));
|
|
|
|
goto out_free_conf;
|
|
|
|
}
|
|
|
|
|
|
|
|
printk(KERN_INFO
|
|
|
|
"raid10: raid set %s active with %d out of %d devices\n",
|
|
|
|
mdname(mddev), mddev->raid_disks - mddev->degraded,
|
|
|
|
mddev->raid_disks);
|
|
|
|
/*
|
|
|
|
* Ok, everything is just fine now
|
|
|
|
*/
|
2008-07-21 09:05:22 +02:00
|
|
|
mddev->array_sectors = size << conf->chunk_shift;
|
2007-03-01 05:11:18 +01:00
|
|
|
mddev->resync_max_sectors = size << conf->chunk_shift;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-05-17 06:53:16 +02:00
|
|
|
mddev->queue->unplug_fn = raid10_unplug;
|
2006-10-03 10:15:54 +02:00
|
|
|
mddev->queue->backing_dev_info.congested_fn = raid10_congested;
|
|
|
|
mddev->queue->backing_dev_info.congested_data = mddev;
|
2005-05-17 06:53:16 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/* Calculate max read-ahead size.
|
|
|
|
* We need to readahead at least twice a whole stripe....
|
|
|
|
* maybe...
|
|
|
|
*/
|
|
|
|
{
|
2006-06-26 09:27:36 +02:00
|
|
|
int stripe = conf->raid_disks * (mddev->chunk_size / PAGE_SIZE);
|
2005-04-17 00:20:36 +02:00
|
|
|
stripe /= conf->near_copies;
|
|
|
|
if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
|
|
|
|
mddev->queue->backing_dev_info.ra_pages = 2* stripe;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (conf->near_copies < mddev->raid_disks)
|
|
|
|
blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_free_conf:
|
|
|
|
if (conf->r10bio_pool)
|
|
|
|
mempool_destroy(conf->r10bio_pool);
|
2006-01-06 09:20:40 +01:00
|
|
|
safe_put_page(conf->tmppage);
|
2005-06-22 02:17:30 +02:00
|
|
|
kfree(conf->mirrors);
|
2005-04-17 00:20:36 +02:00
|
|
|
kfree(conf);
|
|
|
|
mddev->private = NULL;
|
|
|
|
out:
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int stop(mddev_t *mddev)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
|
|
|
|
md_unregister_thread(mddev->thread);
|
|
|
|
mddev->thread = NULL;
|
|
|
|
blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
|
|
|
|
if (conf->r10bio_pool)
|
|
|
|
mempool_destroy(conf->r10bio_pool);
|
2005-06-22 02:17:30 +02:00
|
|
|
kfree(conf->mirrors);
|
2005-04-17 00:20:36 +02:00
|
|
|
kfree(conf);
|
|
|
|
mddev->private = NULL;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-01-06 09:20:16 +01:00
|
|
|
static void raid10_quiesce(mddev_t *mddev, int state)
|
|
|
|
{
|
|
|
|
conf_t *conf = mddev_to_conf(mddev);
|
|
|
|
|
|
|
|
switch(state) {
|
|
|
|
case 1:
|
|
|
|
raise_barrier(conf, 0);
|
|
|
|
break;
|
|
|
|
case 0:
|
|
|
|
lower_barrier(conf);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (mddev->thread) {
|
|
|
|
if (mddev->bitmap)
|
|
|
|
mddev->thread->timeout = mddev->bitmap->daemon_sleep * HZ;
|
|
|
|
else
|
|
|
|
mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
|
|
|
|
md_wakeup_thread(mddev->thread);
|
|
|
|
}
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-01-06 09:20:36 +01:00
|
|
|
static struct mdk_personality raid10_personality =
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
.name = "raid10",
|
2006-01-06 09:20:36 +01:00
|
|
|
.level = 10,
|
2005-04-17 00:20:36 +02:00
|
|
|
.owner = THIS_MODULE,
|
|
|
|
.make_request = make_request,
|
|
|
|
.run = run,
|
|
|
|
.stop = stop,
|
|
|
|
.status = status,
|
|
|
|
.error_handler = error,
|
|
|
|
.hot_add_disk = raid10_add_disk,
|
|
|
|
.hot_remove_disk= raid10_remove_disk,
|
|
|
|
.spare_active = raid10_spare_active,
|
|
|
|
.sync_request = sync_request,
|
2006-01-06 09:20:16 +01:00
|
|
|
.quiesce = raid10_quiesce,
|
2005-04-17 00:20:36 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
static int __init raid_init(void)
|
|
|
|
{
|
2006-01-06 09:20:36 +01:00
|
|
|
return register_md_personality(&raid10_personality);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void raid_exit(void)
|
|
|
|
{
|
2006-01-06 09:20:36 +01:00
|
|
|
unregister_md_personality(&raid10_personality);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
module_init(raid_init);
|
|
|
|
module_exit(raid_exit);
|
|
|
|
MODULE_LICENSE("GPL");
|
|
|
|
MODULE_ALIAS("md-personality-9"); /* RAID10 */
|
2006-01-06 09:20:51 +01:00
|
|
|
MODULE_ALIAS("md-raid10");
|
2006-01-06 09:20:36 +01:00
|
|
|
MODULE_ALIAS("md-level-10");
|