raid1.c

linux和2410结合开发 用他可以生成2410所需的zImage文件

		bh_req = &r1_bh->bh_req;
		memcpy(bh_req, bh, sizeof(*bh));
		bh_req->b_blocknr = bh->b_rsector;
		bh_req->b_dev = mirror->dev;
		bh_req->b_rdev = mirror->dev;
	/*	bh_req->b_rsector = bh->n_rsector; */
		bh_req->b_end_io = raid1_end_request;
		bh_req->b_private = r1_bh;
		generic_make_request (rw, bh_req);
		return 0;
	}

	/*
	 * WRITE:
	 */

	bhl = raid1_alloc_bh(conf, conf->raid_disks);
	for (i = 0; i < disks; i++) {
		struct buffer_head *mbh;
		if (!conf->mirrors[i].operational) 
			continue;
 
	/*
	 * We should use a private pool (size depending on NR_REQUEST),
	 * to avoid writes filling up the memory with bhs
	 *
 	 * Such pools are much faster than kmalloc anyways (so we waste
 	 * almost nothing by not using the master bh when writing and
 	 * win alot of cleanness) but for now we are cool enough. --mingo
 	 *
	 * It's safe to sleep here, buffer heads cannot be used in a shared
 	 * manner in the write branch. Look how we lock the buffer at the
 	 * beginning of this function to grok the difference ;)
	 */
 		mbh = bhl;
		if (mbh == NULL) {
			MD_BUG();
			break;
		}
		bhl = mbh->b_next;
		mbh->b_next = NULL;
		mbh->b_this_page = (struct buffer_head *)1;
		
 	/*
 	 * prepare mirrored mbh (fields ordered for max mem throughput):
 	 */
		mbh->b_blocknr    = bh->b_rsector;
		mbh->b_dev        = conf->mirrors[i].dev;
		mbh->b_rdev	  = conf->mirrors[i].dev;
		mbh->b_rsector	  = bh->b_rsector;
		mbh->b_state      = (1<<BH_Req) | (1<<BH_Dirty) |
						(1<<BH_Mapped) | (1<<BH_Lock);

		atomic_set(&mbh->b_count, 1);
 		mbh->b_size       = bh->b_size;
 		mbh->b_page	  = bh->b_page;
 		mbh->b_data	  = bh->b_data;
 		mbh->b_list       = BUF_LOCKED;
 		mbh->b_end_io     = raid1_end_request;
 		mbh->b_private    = r1_bh;

		mbh->b_next = r1_bh->mirror_bh_list;
		r1_bh->mirror_bh_list = mbh;
		sum_bhs++;
	}
	if (bhl) raid1_free_bh(conf,bhl);
	if (!sum_bhs) {
		/* Gag - all mirrors non-operational.. */
		raid1_end_bh_io(r1_bh, 0);
		return 0;
	}
	md_atomic_set(&r1_bh->remaining, sum_bhs);

	/*
	 * We have to be a bit careful about the semaphore above, thats
	 * why we start the requests separately. Since kmalloc() could
	 * fail, sleep and make_request() can sleep too, this is the
	 * safer solution. Imagine, end_request decreasing the semaphore
	 * before we could have set it up ... We could play tricks with
	 * the semaphore (presetting it and correcting at the end if
	 * sum_bhs is not 'n' but we have to do end_request by hand if
	 * all requests finish until we had a chance to set up the
	 * semaphore correctly ... lots of races).
	 */
	bh = r1_bh->mirror_bh_list;
	while(bh) {
		struct buffer_head *bh2 = bh;
		bh = bh->b_next;
		generic_make_request(rw, bh2);
	}
	return (0);
}

static int raid1_status (char *page, mddev_t *mddev)
{
	raid1_conf_t *conf = mddev_to_conf(mddev);
	int sz = 0, i;
	
	sz += sprintf (page+sz, " [%d/%d] [", conf->raid_disks,
						 conf->working_disks);
	for (i = 0; i < conf->raid_disks; i++)
		sz += sprintf (page+sz, "%s",
			conf->mirrors[i].operational ? "U" : "_");
	sz += sprintf (page+sz, "]");
	return sz;
}

#define LAST_DISK KERN_ALERT \
"raid1: only one disk left and IO error.\n"

#define NO_SPARE_DISK KERN_ALERT \
"raid1: no spare disk left, degrading mirror level by one.\n"

#define DISK_FAILED KERN_ALERT \
"raid1: Disk failure on %s, disabling device. \n" \
"	Operation continuing on %d devices\n"

#define START_SYNCING KERN_ALERT \
"raid1: start syncing spare disk.\n"

#define ALREADY_SYNCING KERN_INFO \
"raid1: syncing already in progress.\n"

static void mark_disk_bad (mddev_t *mddev, int failed)
{
	raid1_conf_t *conf = mddev_to_conf(mddev);
	struct mirror_info *mirror = conf->mirrors+failed;
	mdp_super_t *sb = mddev->sb;

	mirror->operational = 0;
	mark_disk_faulty(sb->disks+mirror->number);
	mark_disk_nonsync(sb->disks+mirror->number);
	mark_disk_inactive(sb->disks+mirror->number);
	if (!mirror->write_only)
		sb->active_disks--;
	sb->working_disks--;
	sb->failed_disks++;
	mddev->sb_dirty = 1;
	md_wakeup_thread(conf->thread);
	if (!mirror->write_only)
		conf->working_disks--;
	printk (DISK_FAILED, partition_name (mirror->dev),
				 conf->working_disks);
}

static int raid1_error (mddev_t *mddev, kdev_t dev)
{
	raid1_conf_t *conf = mddev_to_conf(mddev);
	struct mirror_info * mirrors = conf->mirrors;
	int disks = MD_SB_DISKS;
	int i;

	/* Find the drive.
	 * 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
	 */

	for (i = 0; i < disks; i++)
		if (mirrors[i].dev==dev && mirrors[i].operational)
			break;
	if (i == disks)
		return 0;

	if (i < conf->raid_disks && conf->working_disks == 1) {
		/* Don't fail the drive, act as though we were just a
		 * normal single drive
		 */

		return 1;
	}
	mark_disk_bad(mddev, i);
	return 0;
}

#undef LAST_DISK
#undef NO_SPARE_DISK
#undef DISK_FAILED
#undef START_SYNCING


static void print_raid1_conf (raid1_conf_t *conf)
{
	int i;
	struct mirror_info *tmp;

	printk("RAID1 conf printout:\n");
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
	printk(" --- wd:%d rd:%d nd:%d\n", conf->working_disks,
			 conf->raid_disks, conf->nr_disks);

	for (i = 0; i < MD_SB_DISKS; i++) {
		tmp = conf->mirrors + i;
		printk(" disk %d, s:%d, o:%d, n:%d rd:%d us:%d dev:%s\n",
			i, tmp->spare,tmp->operational,
			tmp->number,tmp->raid_disk,tmp->used_slot,
			partition_name(tmp->dev));
	}
}

static void close_sync(raid1_conf_t *conf)
{
	mddev_t *mddev = conf->mddev;
	/* If reconstruction was interrupted, we need to close the "active" and "pending"
	 * holes.
	 * we know that there are no active rebuild requests, os cnt_active == cnt_ready ==0
	 */
	/* this is really needed when recovery stops too... */
	spin_lock_irq(&conf->segment_lock);
	conf->start_active = conf->start_pending;
	conf->start_ready = conf->start_pending;
	wait_event_lock_irq(conf->wait_ready, !conf->cnt_pending, conf->segment_lock);
	conf->start_active =conf->start_ready = conf->start_pending = conf->start_future;
	conf->start_future = mddev->sb->size+1;
	conf->cnt_pending = conf->cnt_future;
	conf->cnt_future = 0;
	conf->phase = conf->phase ^1;
	wait_event_lock_irq(conf->wait_ready, !conf->cnt_pending, conf->segment_lock);
	conf->start_active = conf->start_ready = conf->start_pending = conf->start_future = 0;
	conf->phase = 0;
	conf->cnt_future = conf->cnt_done;;
	conf->cnt_done = 0;
	spin_unlock_irq(&conf->segment_lock);
	wake_up(&conf->wait_done);
}

static int raid1_diskop(mddev_t *mddev, mdp_disk_t **d, int state)
{
	int err = 0;
	int i, failed_disk=-1, spare_disk=-1, removed_disk=-1, added_disk=-1;
	raid1_conf_t *conf = mddev->private;
	struct mirror_info *tmp, *sdisk, *fdisk, *rdisk, *adisk;
	mdp_super_t *sb = mddev->sb;
	mdp_disk_t *failed_desc, *spare_desc, *added_desc;
	mdk_rdev_t *spare_rdev, *failed_rdev;

	print_raid1_conf(conf);
	md_spin_lock_irq(&conf->device_lock);
	/*
	 * find the disk ...
	 */
	switch (state) {

	case DISKOP_SPARE_ACTIVE:

		/*
		 * Find the failed disk within the RAID1 configuration ...
		 * (this can only be in the first conf->working_disks part)
		 */
		for (i = 0; i < conf->raid_disks; i++) {
			tmp = conf->mirrors + i;
			if ((!tmp->operational && !tmp->spare) ||
					!tmp->used_slot) {
				failed_disk = i;
				break;
			}
		}
		/*
		 * When we activate a spare disk we _must_ have a disk in
		 * the lower (active) part of the array to replace. 
		 */
		if ((failed_disk == -1) || (failed_disk >= conf->raid_disks)) {
			MD_BUG();
			err = 1;
			goto abort;
		}
		/* fall through */

	case DISKOP_SPARE_WRITE:
	case DISKOP_SPARE_INACTIVE:

		/*
		 * Find the spare disk ... (can only be in the 'high'
		 * area of the array)
		 */
		for (i = conf->raid_disks; i < MD_SB_DISKS; i++) {
			tmp = conf->mirrors + i;
			if (tmp->spare && tmp->number == (*d)->number) {
				spare_disk = i;
				break;
			}
		}
		if (spare_disk == -1) {
			MD_BUG();
			err = 1;
			goto abort;
		}
		break;

	case DISKOP_HOT_REMOVE_DISK:

		for (i = 0; i < MD_SB_DISKS; i++) {
			tmp = conf->mirrors + i;
			if (tmp->used_slot && (tmp->number == (*d)->number)) {
				if (tmp->operational) {
					err = -EBUSY;
					goto abort;
				}
				removed_disk = i;
				break;
			}
		}
		if (removed_disk == -1) {
			MD_BUG();
			err = 1;
			goto abort;
		}
		break;

	case DISKOP_HOT_ADD_DISK:

		for (i = conf->raid_disks; i < MD_SB_DISKS; i++) {
			tmp = conf->mirrors + i;
			if (!tmp->used_slot) {
				added_disk = i;
				break;
			}
		}
		if (added_disk == -1) {
			MD_BUG();
			err = 1;
			goto abort;
		}
		break;
	}

	switch (state) {
	/*
	 * Switch the spare disk to write-only mode:
	 */
	case DISKOP_SPARE_WRITE:
		sdisk = conf->mirrors + spare_disk;
		sdisk->operational = 1;
		sdisk->write_only = 1;
		break;
	/*
	 * Deactivate a spare disk:
	 */
	case DISKOP_SPARE_INACTIVE:
		close_sync(conf);
		sdisk = conf->mirrors + spare_disk;
		sdisk->operational = 0;
		sdisk->write_only = 0;
		break;
	/*
	 * Activate (mark read-write) the (now sync) spare disk,
	 * which means we switch it's 'raid position' (->raid_disk)
	 * with the failed disk. (only the first 'conf->nr_disks'
	 * slots are used for 'real' disks and we must preserve this
	 * property)
	 */
	case DISKOP_SPARE_ACTIVE:
		close_sync(conf);
		sdisk = conf->mirrors + spare_disk;
		fdisk = conf->mirrors + failed_disk;

		spare_desc = &sb->disks[sdisk->number];
		failed_desc = &sb->disks[fdisk->number];

		if (spare_desc != *d) {
			MD_BUG();
			err = 1;
			goto abort;
		}

		if (spare_desc->raid_disk != sdisk->raid_disk) {
			MD_BUG();
			err = 1;
			goto abort;
		}
			
		if (sdisk->raid_disk != spare_disk) {
			MD_BUG();
			err = 1;
			goto abort;
		}

		if (failed_desc->raid_disk != fdisk->raid_disk) {
			MD_BUG();
			err = 1;
			goto abort;
		}

		if (fdisk->raid_disk != failed_disk) {
			MD_BUG();
			err = 1;
			goto abort;
		}

		/*
		 * do the switch finally
		 */
		spare_rdev = find_rdev_nr(mddev, spare_desc->number);
		failed_rdev = find_rdev_nr(mddev, failed_desc->number);

		/* There must be a spare_rdev, but there may not be a
		 * failed_rdev.  That slot might be empty...
		 */
		spare_rdev->desc_nr = failed_desc->number;
		if (failed_rdev)
			failed_rdev->desc_nr = spare_desc->number;
		
		xchg_values(*spare_desc, *failed_desc);
		xchg_values(*fdisk, *sdisk);

		/*
		 * (careful, 'failed' and 'spare' are switched from now on)
		 *
		 * we want to preserve linear numbering and we want to
		 * give the proper raid_disk number to the now activated
		 * disk. (this means we switch back these values)
		 */
	
		xchg_values(spare_desc->raid_disk, failed_desc->raid_disk);
		xchg_values(sdisk->raid_disk, fdisk->raid_disk);
		xchg_values(spare_desc->number, failed_desc->number);
		xchg_values(sdisk->number, fdisk->number);

		*d = failed_desc;

		if (sdisk->dev == MKDEV(0,0))
			sdisk->used_slot = 0;
		/*
		 * this really activates the spare.
		 */
		fdisk->spare = 0;
		fdisk->write_only = 0;

		/*
		 * if we activate a spare, we definitely replace a
		 * non-operational disk slot in the 'low' area of
		 * the disk array.
		 */

		conf->working_disks++;

		break;

	case DISKOP_HOT_REMOVE_DISK:
		rdisk = conf->mirrors + removed_disk;

		if (rdisk->spare && (removed_disk < conf->raid_disks)) {
			MD_BUG();	
			err = 1;
			goto abort;
		}
		rdisk->dev = MKDEV(0,0);
		rdisk->used_slot = 0;
		conf->nr_disks--;
		break;

	case DISKOP_HOT_ADD_DISK:
		adisk = conf->mirrors + added_disk;
		added_desc = *d;

		if (added_disk != added_desc->number) {
			MD_BUG();	
			err = 1;
			goto abort;
		}

		adisk->number = added_desc->number;
		adisk->raid_disk = added_desc->raid_disk;
		adisk->dev = MKDEV(added_desc->major,added_desc->minor);

		adisk->operational = 0;
		adisk->write_only = 0;
		adisk->spare = 1;
		adisk->used_slot = 1;
		adisk->head_position = 0;
		conf->nr_disks++;

		break;

	default:
		MD_BUG();	
		err = 1;
		goto abort;
	}
abort:
	md_spin_unlock_irq(&conf->device_lock);
	if (state == DISKOP_SPARE_ACTIVE || state == DISKOP_SPARE_INACTIVE)
		/* should move to "END_REBUILD" when such exists */
		raid1_shrink_buffers(conf);

	print_raid1_conf(conf);
	return err;
}


#define IO_ERROR KERN_ALERT \
"raid1: %s: unrecoverable I/O read error for block %lu\n"

#define REDIRECT_SECTOR KERN_ERR \
"raid1: %s: redirecting sector %lu to another mirror\n"

/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
 *	3.	Performs writes following reads for array syncronising.
 */
static void end_sync_write(struct buffer_head *bh, int uptodate);
static void end_sync_read(struct buffer_head *bh, int uptodate);

static void raid1d (void *data)
{
	struct raid1_bh *r1_bh;
	struct buffer_head *bh;
	unsigned long flags;
	mddev_t *mddev;
	kdev_t dev;


	for (;;) {
		md_spin_lock_irqsave(&retry_list_lock, flags);
		r1_bh = raid1_retry_list;
		if (!r1_bh)
			break;
		raid1_retry_list = r1_bh->next_r1;
		md_spin_unlock_irqrestore(&retry_list_lock, flags);

		mddev = r1_bh->mddev;
		if (mddev->sb_dirty)
			md_update_sb(mddev);
		bh = &r1_bh->bh_req;
		switch(r1_bh->cmd) {
		case SPECIAL:
			/* have to allocate lots of bh structures and
			 * schedule writes
			 */
			if (test_bit(R1BH_Uptodate, &r1_bh->state)) {
				int i, sum_bhs = 0;
				int disks = MD_SB_DISKS;
				struct buffer_head *bhl, *mbh;
				raid1_conf_t *conf;
				
				conf = mddev_to_conf(mddev);
				bhl = raid1_alloc_bh(conf, conf->raid_disks); /* don't really need this many */
				for (i = 0; i < disks ; i++) {
					if (!conf->mirrors[i].operational)
						continue;
					if (i==conf->last_used)
						/* we read from here, no need to write */
						continue;
					if (i < conf->raid_disks
					    && !conf->resync_mirrors)
						/* don't need to write this,
						 * we are just rebuilding */
						continue;
					mbh = bhl;
					if (!mbh) {
						MD_BUG();
						break;
					}
					bhl = mbh->b_next;
					mbh->b_this_page = (struct buffer_head *)1;

						
				/*
				 * prepare mirrored bh (fields ordered for max mem throughput):
				 */
					mbh->b_blocknr    = bh->b_blocknr;
					mbh->b_dev        = conf->mirrors[i].dev;
					mbh->b_rdev	  = conf->mirrors[i].dev;
					mbh->b_rsector	  = bh->b_blocknr;
					mbh->b_state      = (1<<BH_Req) | (1<<BH_Dirty) |
						(1<<BH_Mapped) | (1<<BH_Lock);
					atomic_set(&mbh->b_count, 1);
					mbh->b_size       = bh->b_size;
					mbh->b_page	  = bh->b_page;
					mbh->b_data	  = bh->b_data;
					mbh->b_list       = BUF_LOCKED;
					mbh->b_end_io     = end_sync_write;
					mbh->b_private    = r1_bh;

					mbh->b_next = r1_bh->mirror_bh_list;
					r1_bh->mirror_bh_list = mbh;

					sum_bhs++;
				}
				md_atomic_set(&r1_bh->remaining, sum_bhs);
				if (bhl) raid1_free_bh(conf, bhl);
				mbh = r1_bh->mirror_bh_list;

				if (!sum_bhs) {
					/* nowhere to write this too... I guess we
					 * must be done
					 */
					sync_request_done(bh->b_blocknr, conf);
					md_done_sync(mddev, bh->b_size>>9, 0);
					raid1_free_buf(r1_bh);
				} else
				while (mbh) {
					struct buffer_head *bh1 = mbh;
					mbh = mbh->b_next;
					generic_make_request(WRITE, bh1);
					md_sync_acct(bh1->b_dev, bh1->b_size/512);
				}
			} else {
				/* There is no point trying a read-for-reconstruct
				 * as reconstruct is about to be aborted
				 */

				printk (IO_ERROR, partition_name(bh->b_dev), bh->b_blocknr);