raid10.c

来自「Linux Kernel 2.6.9 for OMAP1710」· C语言 代码 · 共 1,781 行 · 第 1/4 页

/*
 * 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.
 */

#include <linux/raid/raid10.h>

/*
 * 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)
 *
 * 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.
 * Thus there are (near_copies*far_copies) of each chunk, and each is on a different
 * drive.
 * near_copies and far_copies must be at least one, and there product is at most
 * raid_disks.
 */

/*
 * Number of guaranteed r10bios in case of extreme VM load:
 */
#define	NR_RAID10_BIOS 256

static void unplug_slaves(mddev_t *mddev);

static void * r10bio_pool_alloc(int gfp_flags, void *data)
{
	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 */
	r10_bio = kmalloc(size, gfp_flags);
	if (r10_bio)
		memset(r10_bio, 0, size);
	else
		unplug_slaves(conf->mddev);

	return r10_bio;
}

static void r10bio_pool_free(void *r10_bio, void *data)
{
	kfree(r10_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)

/*
 * 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)
 *
 */
static void * r10buf_pool_alloc(int gfp_flags, void *data)
{
	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--)
		__free_page(bio->bi_io_vec[i-1].bv_page);
	while (j--)
		for (i = 0; i < RESYNC_PAGES ; i++)
			__free_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page);
	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++) {
				__free_page(bio->bi_io_vec[i].bv_page);
				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;
		if (*bio)
			bio_put(*bio);
		*bio = NULL;
	}
}

static inline void free_r10bio(r10bio_t *r10_bio)
{
	unsigned long flags;

	conf_t *conf = mddev_to_conf(r10_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, r10_bio);
	mempool_free(r10_bio, conf->r10bio_pool);
}

static inline void put_buf(r10bio_t *r10_bio)
{
	conf_t *conf = mddev_to_conf(r10_bio->mddev);
	unsigned long flags;

	mempool_free(r10_bio, conf->r10buf_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(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);
	spin_unlock_irqrestore(&conf->device_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(r10bio_t *r10_bio)
{
	struct bio *bio = r10_bio->master_bio;

	bio_endio(bio, bio->bi_size,
		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);
}

static int raid10_end_read_request(struct bio *bio, unsigned int bytes_done, int error)
{
	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);

	if (bio->bi_size)
		return 1;

	slot = r10_bio->read_slot;
	dev = r10_bio->devs[slot].devnum;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
	if (!uptodate)
		md_error(r10_bio->mddev, conf->mirrors[dev].rdev);
	else
		/*
		 * 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);

	update_head_pos(slot, r10_bio);

	/*
	 * we have only one bio on the read side
	 */
	if (uptodate)
		raid_end_bio_io(r10_bio);
	else {
		/*
		 * 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);
	return 0;
}

static int raid10_end_write_request(struct bio *bio, unsigned int bytes_done, int error)
{
	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);

	if (bio->bi_size)
		return 1;

	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:
	 */
	if (!uptodate)
		md_error(r10_bio->mddev, conf->mirrors[dev].rdev);
	else
		/*
		 * 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)) {
		md_write_end(r10_bio->mddev);
		raid_end_bio_io(r10_bio);
	}

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


/*
 * 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
 * on each device that it is on. If a block isn't on a device,
 * that entry in the array is set to MaxSector.
 *
 * 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);

	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;

	while (sector > conf->stride) {
		sector -= conf->stride;
		if (dev < conf->near_copies)
			dev += conf->raid_disks - conf->near_copies;
		else
			dev -= conf->near_copies;
	}

	offset = sector & conf->chunk_mask;
	chunk = sector >> conf->chunk_shift;
	vchunk = chunk * conf->raid_disks + dev;
	sector_div(vchunk, conf->near_copies);
	return (vchunk << conf->chunk_shift) + offset;
}