Linux-2.6.12-rc2v2.6.12-rc2

Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.

Let it rip!

1 files changed, 1787 insertions, 0 deletions

diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
new file mode 100644
index 00000000000..b100bfe4fdc
--- /dev/null
+++ b/drivers/md/raid10.c
@@ -0,0 +1,1787 @@
+/*
+ * 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 their 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(unsigned int __nocast 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(unsigned int __nocast 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;
+}
+
+/**
+ *	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);
+	rcu_read_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);
+	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++) {
+		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
+		if (rdev && !rdev->faulty && atomic_read(&rdev->nr_pending)) {
+			request_queue_t *r_queue = bdev_get_queue(rdev->bdev);
+
+			atomic_inc(&rdev->nr_pending);
+			rcu_read_unlock();
+
+			if (r_queue->unplug_fn)
+				r_queue->unplug_fn(r_queue);
+
+			rdev_dec_pending(rdev, mddev);
+			rcu_read_lock();
+		}
+	}
+	rcu_read_unlock();
+}
+
+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);
+	int i, ret = 0;
+
+	rcu_read_lock();
+	for (i=0; i<mddev->raid_disks && ret == 0; 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 = -EOPNOTSUPP;
+			else {
+				atomic_inc(&rdev->nr_pending);
+				rcu_read_unlock();
+				ret = r_queue->issue_flush_fn(r_queue, bdev->bd_disk,
+							      error_sector);
+				rdev_dec_pending(rdev, mddev);
+				rcu_read_lock();
+			}
+		}
+	}
+	rcu_read_unlock();
+	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);
+	rcu_read_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;
+	}
+	rcu_read_unlock();
+
+	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();
+
+	mempool_destroy(conf->r10buf_pool);
+	conf->r10buf_pool = NULL;
+}
+
+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
+		    && !tmp->rdev->faulty
+		    && !tmp->rdev->in_sync) {
+			conf->working_disks++;
+			mddev->degraded--;
+			tmp->rdev->in_sync = 1;
+		}
+	}
+
+	print_conf(conf);
+	return 0;
+}
+
+
+static int raid10_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
+{
+	conf_t *conf = mddev->private;
+	int found = 0;
+	int mirror;
+	mirror_info_t *p;
+
+	if (mddev->recovery_cp < MaxSector)
+		/* only hot-add to in-sync arrays, as recovery is
+		 * very different from resync
+		 */
+		return 0;
+
+	for (mirror=0; mirror < mddev->raid_disks; mirror++)
+		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;
+			found = 1;
+			p->rdev = rdev;
+			break;
+		}
+
+	print_conf(conf);
+	return found;
+}
+
+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) {
+		if (rdev->in_sync ||
+		    atomic_read(&rdev->nr_pending)) {
+			err = -EBUSY;
+			goto abort;
+		}
+		p->rdev = NULL;
+		synchronize_kernel();
+		if (atomic_read(&rdev->nr_pending)) {
+			/* lost the race, try later */
+			err = -EBUSY;
+			p->rdev = rdev;
+		}
+	}
+abort:
+
+	print_conf(conf);
+	return err;
+}
+
+
+static int end_sync_read(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);
+	conf_t *conf = mddev_to_conf(r10_bio->mddev);
+	int i,d;
+
+	if (bio->bi_size)
+		return 1;
+
+	for (i=0; i<conf->copies; i++)
+		if (r10_bio->devs[i].bio == bio)
+			break;
+	if (i == conf->copies)
+		BUG();
+	update_head_pos(i, r10_bio);
+	d = r10_bio->devs[i].devnum;
+	if (!uptodate)
+		md_error(r10_bio->mddev,
+			 conf->mirrors[d].rdev);
+
+	/* for reconstruct, we always reschedule after a read.
+	 * for resync, only after all reads
+	 */
+	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);
+	}
+	rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev);
+	return 0;
+}
+
+static int end_sync_write(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);
+	mddev_t *mddev = r10_bio->mddev;
+	conf_t *conf = mddev_to_conf(mddev);
+	int i,d;
+
+	if (bio->bi_size)
+		return 1;
+
+	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);
+	update_head_pos(i, r10_bio);
+
+	while (atomic_dec_and_test(&r10_bio->remaining)) {
+		if (r10_bio->master_bio == NULL) {
+			/* the primary of several recovery bios */
+			md_done_sync(mddev, r10_bio->sectors, 1);
+			put_buf(r10_bio);
+			break;
+		} else {
+			r10bio_t *r10_bio2 = (r10bio_t *)r10_bio->master_bio;
+			put_buf(r10_bio);
+			r10_bio = r10_bio2;
+		}
+	}
+	rdev_dec_pending(conf->mirrors[d].rdev, mddev);
+	return 0;
+}
+
+/*
+ * 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 */
+	for (i=first+1 ; i < conf->copies ; i++) {
+		int vcnt, j, d;
+
+		if (!test_bit(BIO_UPTODATE, &r10_bio->devs[i].bio->bi_flags))
+			continue;
+		/* 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;
+		 */
+		tbio = r10_bio->devs[i].bio;
+		vcnt = r10_bio->sectors >> (PAGE_SHIFT-9);
+		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;
+		/* 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_hw_segments = 0;
+		tbio->bi_hw_front_size = 0;
+		tbio->bi_hw_back_size = 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);