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_mergeable_bvec -- tell bio layer if a two requests can be merged
 *	@q: request queue
 *	@bio: the buffer head that's been built up so far
 *	@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.
 */
static int raid10_mergeable_bvec(request_queue_t *q, struct bio *bio,
				struct bio_vec *bio_vec)
{
	mddev_t *mddev = q->queuedata;
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
	int max;
	unsigned int chunk_sectors = mddev->chunk_size >> 9;
	unsigned int bio_sectors = bio->bi_size >> 9;

	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
	if (max <= bio_vec->bv_len && bio_sectors == 0)
		return bio_vec->bv_len;
	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;

	raid10_find_phys(conf, r10_bio);
	spin_lock_irq(&conf->device_lock);
	/*
	 * Check 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)) {
		/* make sure that disk is operational */
		slot = 0;
		disk = r10_bio->devs[slot].devnum;

		while (!conf->mirrors[disk].rdev ||
		       !conf->mirrors[disk].rdev->in_sync) {
			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;
	while (!conf->mirrors[disk].rdev ||
	       !conf->mirrors[disk].rdev->in_sync) {
		slot ++;
		if (slot == conf->copies) {
			disk = -1;
			goto rb_out;
		}
		disk = r10_bio->devs[slot].devnum;
	}


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

	/* Find the disk whose head is closest */

	for (nslot = slot; nslot < conf->copies; nslot++) {
		int ndisk = r10_bio->devs[nslot].devnum;


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

		if (!atomic_read(&conf->mirrors[ndisk].rdev->nr_pending)) {
			disk = ndisk;
			slot = nslot;
			break;
		}
		new_distance = abs(r10_bio->devs[nslot].addr -
				   conf->mirrors[ndisk].head_position);
		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;*/

	if (disk >= 0 && conf->mirrors[disk].rdev)
		atomic_inc(&conf->mirrors[disk].rdev->nr_pending);
	spin_unlock_irq(&conf->device_lock);

	return 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 raid10_unplug(request_queue_t *q)
{
	unplug_slaves(q->queuedata);
}

static int raid10_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);
	return ret;
}

/*
 * Throttle resync depth, so that we can both get proper overlapping of
 * requests, but are still able to handle normal requests quickly.
 */
#define RESYNC_DEPTH 32

static void device_barrier(conf_t *conf, sector_t sect)
{
	spin_lock_irq(&conf->resync_lock);
	wait_event_lock_irq(conf->wait_idle, !waitqueue_active(&conf->wait_resume),
			    conf->resync_lock, unplug_slaves(conf->mddev));

	if (!conf->barrier++) {
		wait_event_lock_irq(conf->wait_idle, !conf->nr_pending,
				    conf->resync_lock, unplug_slaves(conf->mddev));
		if (conf->nr_pending)
			BUG();
	}
	wait_event_lock_irq(conf->wait_resume, conf->barrier < RESYNC_DEPTH,
			    conf->resync_lock, unplug_slaves(conf->mddev));
	conf->next_resync = sect;
	spin_unlock_irq(&conf->resync_lock);
}

static int make_request(request_queue_t *q, struct bio * bio)
{
	mddev_t *mddev = q->queuedata;
	conf_t *conf = mddev_to_conf(mddev);
	mirror_info_t *mirror;
	r10bio_t *r10_bio;
	struct bio *read_bio;
	int i;
	int chunk_sects = conf->chunk_mask + 1;

	/* 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.
		 */
		bp = bio_split(bio, bio_split_pool,
			       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);

		bio_io_error(bio, bio->bi_size);
		return 0;
	}

	/*
	 * 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.
	 */
	spin_lock_irq(&conf->resync_lock);
	wait_event_lock_irq(conf->wait_resume, !conf->barrier, conf->resync_lock, );
	conf->nr_pending++;
	spin_unlock_irq(&conf->resync_lock);

	if (bio_data_dir(bio)==WRITE) {
		disk_stat_inc(mddev->gendisk, writes);
		disk_stat_add(mddev->gendisk, write_sectors, bio_sectors(bio));
	} else {
		disk_stat_inc(mddev->gendisk, reads);
		disk_stat_add(mddev->gendisk, read_sectors, bio_sectors(bio));
	}

	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;

	if (bio_data_dir(bio) == READ) {
		/*
		 * 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;
		read_bio->bi_rw = READ;
		read_bio->bi_private = r10_bio;

		generic_make_request(read_bio);
		return 0;
	}

	/*
	 * WRITE:
	 */
	/* first select target devices under spinlock and
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
	 */
	raid10_find_phys(conf, r10_bio);
	spin_lock_irq(&conf->device_lock);
	for (i = 0;  i < conf->copies; i++) {
		int d = r10_bio->devs[i].devnum;
		if (conf->mirrors[d].rdev &&
		    !conf->mirrors[d].rdev->faulty) {
			atomic_inc(&conf->mirrors[d].rdev->nr_pending);
			r10_bio->devs[i].bio = bio;
		} else
			r10_bio->devs[i].bio = NULL;
	}
	spin_unlock_irq(&conf->device_lock);

	atomic_set(&r10_bio->remaining, 1);
	md_write_start(mddev);
	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;
		mbio->bi_rw = WRITE;
		mbio->bi_private = r10_bio;

		atomic_inc(&r10_bio->remaining);
		generic_make_request(mbio);
	}

	if (atomic_dec_and_test(&r10_bio->remaining)) {
		md_write_end(mddev);
		raid_end_bio_io(r10_bio);
	}

	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);
	if (conf->far_copies > 1)
		seq_printf(seq, " %d far-copies", conf->far_copies);

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
						conf->working_disks);
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf(seq, "%s",
			      conf->mirrors[i].rdev &&
			      conf->mirrors[i].rdev->in_sync ? "U" : "_");
	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
	 */
	if (rdev->in_sync
	    && conf->working_disks == 1)
		/*
		 * 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;
	if (rdev->in_sync) {
		mddev->degraded++;
		conf->working_disks--;
		/*
		 * if recovery is running, make sure it aborts.
		 */
		set_bit(MD_RECOVERY_ERR, &mddev->recovery);
	}
	rdev->in_sync = 0;
	rdev->faulty = 1;
	mddev->sb_dirty = 1;
	printk(KERN_ALERT "raid10: Disk failure on %s, disabling device. \n"
		"	Operation continuing on %d devices\n",
		bdevname(rdev->bdev,b), conf->working_disks);
}

static void print_conf(conf_t *conf)
{
	int i;
	mirror_info_t *tmp;

	printk("RAID10 conf printout:\n");
	if (!conf) {
		printk("(!conf)\n");
		return;
	}
	printk(" --- wd:%d rd:%d\n", conf->working_disks,
		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",
				i, !tmp->rdev->in_sync, !tmp->rdev->faulty,
				bdevname(tmp->rdev->bdev,b));
	}
}

static void close_sync(conf_t *conf)
{
	spin_lock_irq(&conf->resync_lock);
	wait_event_lock_irq(conf->wait_resume, !conf->barrier,
			    conf->resync_lock, 	unplug_slaves(conf->mddev));
	spin_unlock_irq(&conf->resync_lock);

	if (conf->barrier) BUG();
	if (waitqueue_active(&conf->wait_idle)) BUG();