raid1.c

来自「Linux Kernel 2.6.9 for OMAP1710」· C语言 代码 · 共 1,426 行 · 第 1/3 页

/*
 * raid1.c : Multiple Devices driver for Linux
 *
 * Copyright (C) 1999, 2000, 2001 Ingo Molnar, Red Hat
 *
 * Copyright (C) 1996, 1997, 1998 Ingo Molnar, Miguel de Icaza, Gadi Oxman
 *
 * RAID-1 management functions.
 *
 * Better read-balancing code written by Mika Kuoppala <[e-mail]>, 2000
 *
 * Fixes to reconstruction by Jakob 豷tergaard" <[e-mail]>
 * Various fixes by Neil Brown <[e-mail]>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2, or (at your option)
 * any later version.
 *
 * You should have received a copy of the GNU General Public License
 * (for example /usr/src/linux/COPYING); if not, write to the Free
 * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

#include <linux/raid/raid1.h>

/*
 * Number of guaranteed r1bios in case of extreme VM load:
 */
#define	NR_RAID1_BIOS 256

static mdk_personality_t raid1_personality;
static spinlock_t retry_list_lock = SPIN_LOCK_UNLOCKED;
static LIST_HEAD(retry_list_head);

static void unplug_slaves(mddev_t *mddev);


static void * r1bio_pool_alloc(int gfp_flags, void *data)
{
	struct pool_info *pi = data;
	r1bio_t *r1_bio;
	int size = offsetof(r1bio_t, bios[pi->raid_disks]);

	/* allocate a r1bio with room for raid_disks entries in the bios array */
	r1_bio = kmalloc(size, gfp_flags);
	if (r1_bio)
		memset(r1_bio, 0, size);
	else
		unplug_slaves(pi->mddev);

	return r1_bio;
}

static void r1bio_pool_free(void *r1_bio, void *data)
{
	kfree(r1_bio);
}

#define RESYNC_BLOCK_SIZE (64*1024)
//#define RESYNC_BLOCK_SIZE PAGE_SIZE
#define RESYNC_SECTORS (RESYNC_BLOCK_SIZE >> 9)
#define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
#define RESYNC_WINDOW (2048*1024)

static void * r1buf_pool_alloc(int gfp_flags, void *data)
{
	struct pool_info *pi = data;
	struct page *page;
	r1bio_t *r1_bio;
	struct bio *bio;
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
	if (!r1_bio) {
		unplug_slaves(pi->mddev);
		return NULL;
	}

	/*
	 * Allocate bios : 1 for reading, n-1 for writing
	 */
	for (j = pi->raid_disks ; j-- ; ) {
		bio = bio_alloc(gfp_flags, RESYNC_PAGES);
		if (!bio)
			goto out_free_bio;
		r1_bio->bios[j] = bio;
	}
	/*
	 * Allocate RESYNC_PAGES data pages and attach them to
	 * the first bio;
	 */
	bio = r1_bio->bios[0];
	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;
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

out_free_pages:
	for ( ; i > 0 ; i--)
		__free_page(bio->bi_io_vec[i-1].bv_page);
out_free_bio:
	while ( ++j < pi->raid_disks )
		bio_put(r1_bio->bios[j]);
	r1bio_pool_free(r1_bio, data);
	return NULL;
}

static void r1buf_pool_free(void *__r1_bio, void *data)
{
	struct pool_info *pi = data;
	int i;
	r1bio_t *r1bio = __r1_bio;
	struct bio *bio = r1bio->bios[0];

	for (i = 0; i < RESYNC_PAGES; i++) {
		__free_page(bio->bi_io_vec[i].bv_page);
		bio->bi_io_vec[i].bv_page = NULL;
	}
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

static void put_all_bios(conf_t *conf, r1bio_t *r1_bio)
{
	int i;

	for (i = 0; i < conf->raid_disks; i++) {
		struct bio **bio = r1_bio->bios + i;
		if (*bio)
			bio_put(*bio);
		*bio = NULL;
	}
}

static inline void free_r1bio(r1bio_t *r1_bio)
{
	unsigned long flags;

	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	/*
	 * Wake up any possible resync thread that waits for the device
	 * to go idle.
	 */
	spin_lock_irqsave(&conf->resync_lock, flags);
	if (!--conf->nr_pending) {
		wake_up(&conf->wait_idle);
		wake_up(&conf->wait_resume);
	}
	spin_unlock_irqrestore(&conf->resync_lock, flags);

	put_all_bios(conf, r1_bio);
	mempool_free(r1_bio, conf->r1bio_pool);
}

static inline void put_buf(r1bio_t *r1_bio)
{
	conf_t *conf = mddev_to_conf(r1_bio->mddev);
	unsigned long flags;

	mempool_free(r1_bio, conf->r1buf_pool);

	spin_lock_irqsave(&conf->resync_lock, flags);
	if (!conf->barrier)
		BUG();
	--conf->barrier;
	wake_up(&conf->wait_resume);
	wake_up(&conf->wait_idle);

	if (!--conf->nr_pending) {
		wake_up(&conf->wait_idle);
		wake_up(&conf->wait_resume);
	}
	spin_unlock_irqrestore(&conf->resync_lock, flags);
}

static void reschedule_retry(r1bio_t *r1_bio)
{
	unsigned long flags;
	mddev_t *mddev = r1_bio->mddev;

	spin_lock_irqsave(&retry_list_lock, flags);
	list_add(&r1_bio->retry_list, &retry_list_head);
	spin_unlock_irqrestore(&retry_list_lock, flags);

	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(r1bio_t *r1_bio)
{
	struct bio *bio = r1_bio->master_bio;

	bio_endio(bio, bio->bi_size,
		test_bit(R1BIO_Uptodate, &r1_bio->state) ? 0 : -EIO);
	free_r1bio(r1_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
static inline void update_head_pos(int disk, r1bio_t *r1_bio)
{
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	conf->mirrors[disk].head_position =
		r1_bio->sector + (r1_bio->sectors);
}

static int raid1_end_read_request(struct bio *bio, unsigned int bytes_done, int error)
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
	int mirror;
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	if (bio->bi_size)
		return 1;
	
	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
	if (!uptodate)
		md_error(r1_bio->mddev, conf->mirrors[mirror].rdev);
	else
		/*
		 * Set R1BIO_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(R1BIO_Uptodate, &r1_bio->state);

	update_head_pos(mirror, r1_bio);

	/*
	 * we have only one bio on the read side
	 */
	if (uptodate)
		raid_end_bio_io(r1_bio);
	else {
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
		if (printk_ratelimit())
			printk(KERN_ERR "raid1: %s: rescheduling sector %llu\n",
			       bdevname(conf->mirrors[mirror].rdev->bdev,b), (unsigned long long)r1_bio->sector);
		reschedule_retry(r1_bio);
	}

	rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
	return 0;
}

static int raid1_end_write_request(struct bio *bio, unsigned int bytes_done, int error)
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
	int mirror;
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	if (bio->bi_size)
		return 1;

	for (mirror = 0; mirror < conf->raid_disks; mirror++)
		if (r1_bio->bios[mirror] == bio)
			break;

	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
	if (!uptodate)
		md_error(r1_bio->mddev, conf->mirrors[mirror].rdev);
	else
		/*
		 * Set R1BIO_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(R1BIO_Uptodate, &r1_bio->state);

	update_head_pos(mirror, r1_bio);

	/*
	 *
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
	if (atomic_dec_and_test(&r1_bio->remaining)) {
		md_write_end(r1_bio->mddev);
		raid_end_bio_io(r1_bio);
	}

	rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
	return 0;
}


/*
 * 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.
 */
static int read_balance(conf_t *conf, r1bio_t *r1_bio)
{
	const unsigned long this_sector = r1_bio->sector;
	int new_disk = conf->last_used, disk = new_disk;
	const int sectors = r1_bio->sectors;
	sector_t new_distance, current_distance;

	spin_lock_irq(&conf->device_lock);
	/*
	 * Check if it if we can balance. We can balance on the whole
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
	if (conf->mddev->recovery_cp < MaxSector &&
	    (this_sector + sectors >= conf->next_resync)) {
		/* Choose the first operation device, for consistancy */
		new_disk = 0;

		while (!conf->mirrors[new_disk].rdev ||
		       !conf->mirrors[new_disk].rdev->in_sync) {
			new_disk++;
			if (new_disk == conf->raid_disks) {
				new_disk = -1;
				break;
			}
		}
		goto rb_out;
	}


	/* make sure the disk is operational */
	while (!conf->mirrors[new_disk].rdev ||
	       !conf->mirrors[new_disk].rdev->in_sync) {
		if (new_disk <= 0)
			new_disk = conf->raid_disks;
		new_disk--;
		if (new_disk == disk) {
			new_disk = -1;
			goto rb_out;
		}
	}
	disk = new_disk;
	/* now disk == new_disk == starting point for search */

	/*
	 * Don't change to another disk for sequential reads:
	 */
	if (conf->next_seq_sect == this_sector)
		goto rb_out;
	if (this_sector == conf->mirrors[new_disk].head_position)
		goto rb_out;

	current_distance = abs(this_sector - conf->mirrors[disk].head_position);

	/* Find the disk whose head is closest */

	do {
		if (disk <= 0)
			disk = conf->raid_disks;
		disk--;

		if (!conf->mirrors[disk].rdev ||
		    !conf->mirrors[disk].rdev->in_sync)
			continue;

		if (!atomic_read(&conf->mirrors[disk].rdev->nr_pending)) {
			new_disk = disk;
			break;
		}
		new_distance = abs(this_sector - conf->mirrors[disk].head_position);
		if (new_distance < current_distance) {
			current_distance = new_distance;
			new_disk = disk;
		}
	} while (disk != conf->last_used);

rb_out:


	if (new_disk >= 0) {
		conf->next_seq_sect = this_sector + sectors;
		conf->last_used = new_disk;
		atomic_inc(&conf->mirrors[new_disk].rdev->nr_pending);
	}
	spin_unlock_irq(&conf->device_lock);

	return new_disk;
}

static void unplug_slaves(mddev_t *mddev)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i;
	unsigned long flags;

	spin_lock_irqsave(&conf->device_lock, flags);
	for (i=0; i<mddev->raid_disks; i++) {
		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
		if (rdev && atomic_read(&rdev->nr_pending)) {
			request_queue_t *r_queue = bdev_get_queue(rdev->bdev);

			atomic_inc(&rdev->nr_pending);
			spin_unlock_irqrestore(&conf->device_lock, flags);

			if (r_queue->unplug_fn)
				r_queue->unplug_fn(r_queue);

			spin_lock_irqsave(&conf->device_lock, flags);
			atomic_dec(&rdev->nr_pending);
		}
	}
	spin_unlock_irqrestore(&conf->device_lock, flags);
}
static void raid1_unplug(request_queue_t *q)
{
	unplug_slaves(q->queuedata);
}

static int raid1_issue_flush(request_queue_t *q, struct gendisk *disk,
			     sector_t *error_sector)
{
	mddev_t *mddev = q->queuedata;
	conf_t *conf = mddev_to_conf(mddev);
	unsigned long flags;
	int i, ret = 0;

	spin_lock_irqsave(&conf->device_lock, flags);
	for (i=0; i<mddev->raid_disks; i++) {
		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
		if (rdev && !rdev->faulty) {
			struct block_device *bdev = rdev->bdev;
			request_queue_t *r_queue = bdev_get_queue(bdev);

			if (r_queue->issue_flush_fn) {
				ret = r_queue->issue_flush_fn(r_queue, bdev->bd_disk, error_sector);
				if (ret)
					break;
			}
		}
	}
	spin_unlock_irqrestore(&conf->device_lock, flags);