raid1.c

linux和2410结合开发 用他可以生成2410所需的zImage文件

/*
 * raid1.c : Multiple Devices driver for Linux
 *
 * Copyright (C) 1999, 2000 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/module.h>
#include <linux/slab.h>
#include <linux/raid/raid1.h>
#include <asm/atomic.h>

#define MAJOR_NR MD_MAJOR
#define MD_DRIVER
#define MD_PERSONALITY

#define MAX_WORK_PER_DISK 128

#define	NR_RESERVED_BUFS	32


/*
 * The following can be used to debug the driver
 */
#define RAID1_DEBUG	0

#if RAID1_DEBUG
#define PRINTK(x...)   printk(x)
#define inline
#define __inline__
#else
#define PRINTK(x...)  do { } while (0)
#endif


static mdk_personality_t raid1_personality;
static md_spinlock_t retry_list_lock = MD_SPIN_LOCK_UNLOCKED;
struct raid1_bh *raid1_retry_list = NULL, **raid1_retry_tail;

static struct buffer_head *raid1_alloc_bh(raid1_conf_t *conf, int cnt)
{
	/* return a linked list of "cnt" struct buffer_heads.
	 * don't take any off the free list unless we know we can
	 * get all we need, otherwise we could deadlock
	 */
	struct buffer_head *bh=NULL;

	while(cnt) {
		struct buffer_head *t;
		md_spin_lock_irq(&conf->device_lock);
		if (!conf->freebh_blocked && conf->freebh_cnt >= cnt)
			while (cnt) {
				t = conf->freebh;
				conf->freebh = t->b_next;
				t->b_next = bh;
				bh = t;
				t->b_state = 0;
				conf->freebh_cnt--;
				cnt--;
			}
		md_spin_unlock_irq(&conf->device_lock);
		if (cnt == 0)
			break;
		t = kmem_cache_alloc(bh_cachep, SLAB_NOIO);
		if (t) {
			t->b_next = bh;
			bh = t;
			cnt--;
		} else {
			PRINTK("raid1: waiting for %d bh\n", cnt);
			conf->freebh_blocked = 1;
			wait_disk_event(conf->wait_buffer,
					!conf->freebh_blocked ||
					conf->freebh_cnt > conf->raid_disks * NR_RESERVED_BUFS/2);
			conf->freebh_blocked = 0;
		}
	}
	return bh;
}

static inline void raid1_free_bh(raid1_conf_t *conf, struct buffer_head *bh)
{
	unsigned long flags;
	spin_lock_irqsave(&conf->device_lock, flags);
	while (bh) {
		struct buffer_head *t = bh;
		bh=bh->b_next;
		if (t->b_pprev == NULL)
			kmem_cache_free(bh_cachep, t);
		else {
			t->b_next= conf->freebh;
			conf->freebh = t;
			conf->freebh_cnt++;
		}
	}
	spin_unlock_irqrestore(&conf->device_lock, flags);
	wake_up(&conf->wait_buffer);
}

static int raid1_grow_bh(raid1_conf_t *conf, int cnt)
{
	/* allocate cnt buffer_heads, possibly less if kmalloc fails */
	int i = 0;

	while (i < cnt) {
		struct buffer_head *bh;
		bh = kmem_cache_alloc(bh_cachep, SLAB_KERNEL);
		if (!bh) break;

		md_spin_lock_irq(&conf->device_lock);
		bh->b_pprev = &conf->freebh;
		bh->b_next = conf->freebh;
		conf->freebh = bh;
		conf->freebh_cnt++;
		md_spin_unlock_irq(&conf->device_lock);

		i++;
	}
	return i;
}

static void raid1_shrink_bh(raid1_conf_t *conf)
{
	/* discard all buffer_heads */

	md_spin_lock_irq(&conf->device_lock);
	while (conf->freebh) {
		struct buffer_head *bh = conf->freebh;
		conf->freebh = bh->b_next;
		kmem_cache_free(bh_cachep, bh);
		conf->freebh_cnt--;
	}
	md_spin_unlock_irq(&conf->device_lock);
}
		

static struct raid1_bh *raid1_alloc_r1bh(raid1_conf_t *conf)
{
	struct raid1_bh *r1_bh = NULL;

	do {
		md_spin_lock_irq(&conf->device_lock);
		if (!conf->freer1_blocked && conf->freer1) {
			r1_bh = conf->freer1;
			conf->freer1 = r1_bh->next_r1;
			conf->freer1_cnt--;
			r1_bh->next_r1 = NULL;
			r1_bh->state = (1 << R1BH_PreAlloc);
			r1_bh->bh_req.b_state = 0;
		}
		md_spin_unlock_irq(&conf->device_lock);
		if (r1_bh)
			return r1_bh;
		r1_bh = (struct raid1_bh *) kmalloc(sizeof(struct raid1_bh), GFP_NOIO);
		if (r1_bh) {
			memset(r1_bh, 0, sizeof(*r1_bh));
			return r1_bh;
		}
		conf->freer1_blocked = 1;
		wait_disk_event(conf->wait_buffer,
				!conf->freer1_blocked ||
				conf->freer1_cnt > NR_RESERVED_BUFS/2
			);
		conf->freer1_blocked = 0;
	} while (1);
}

static inline void raid1_free_r1bh(struct raid1_bh *r1_bh)
{
	struct buffer_head *bh = r1_bh->mirror_bh_list;
	raid1_conf_t *conf = mddev_to_conf(r1_bh->mddev);

	r1_bh->mirror_bh_list = NULL;

	if (test_bit(R1BH_PreAlloc, &r1_bh->state)) {
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
		r1_bh->next_r1 = conf->freer1;
		conf->freer1 = r1_bh;
		conf->freer1_cnt++;
		spin_unlock_irqrestore(&conf->device_lock, flags);
		/* don't need to wakeup wait_buffer because
		 *  raid1_free_bh below will do that
		 */
	} else {
		kfree(r1_bh);
	}
	raid1_free_bh(conf, bh);
}

static int raid1_grow_r1bh (raid1_conf_t *conf, int cnt)
{
	int i = 0;

	while (i < cnt) {
		struct raid1_bh *r1_bh;
		r1_bh = (struct raid1_bh*)kmalloc(sizeof(*r1_bh), GFP_KERNEL);
		if (!r1_bh)
			break;
		memset(r1_bh, 0, sizeof(*r1_bh));
		set_bit(R1BH_PreAlloc, &r1_bh->state);
		r1_bh->mddev = conf->mddev;

		raid1_free_r1bh(r1_bh);
		i++;
	}
	return i;
}

static void raid1_shrink_r1bh(raid1_conf_t *conf)
{
	md_spin_lock_irq(&conf->device_lock);
	while (conf->freer1) {
		struct raid1_bh *r1_bh = conf->freer1;
		conf->freer1 = r1_bh->next_r1;
		conf->freer1_cnt--;
		kfree(r1_bh);
	}
	md_spin_unlock_irq(&conf->device_lock);
}



static inline void raid1_free_buf(struct raid1_bh *r1_bh)
{
	unsigned long flags;
	struct buffer_head *bh = r1_bh->mirror_bh_list;
	raid1_conf_t *conf = mddev_to_conf(r1_bh->mddev);
	r1_bh->mirror_bh_list = NULL;
	
	spin_lock_irqsave(&conf->device_lock, flags);
	r1_bh->next_r1 = conf->freebuf;
	conf->freebuf = r1_bh;
	spin_unlock_irqrestore(&conf->device_lock, flags);
	raid1_free_bh(conf, bh);
}

static struct raid1_bh *raid1_alloc_buf(raid1_conf_t *conf)
{
	struct raid1_bh *r1_bh;

	md_spin_lock_irq(&conf->device_lock);
	wait_event_lock_irq(conf->wait_buffer, conf->freebuf, conf->device_lock);
	r1_bh = conf->freebuf;
	conf->freebuf = r1_bh->next_r1;
	r1_bh->next_r1= NULL;
	md_spin_unlock_irq(&conf->device_lock);

	return r1_bh;
}

static int raid1_grow_buffers (raid1_conf_t *conf, int cnt)
{
	int i = 0;

	md_spin_lock_irq(&conf->device_lock);
	while (i < cnt) {
		struct raid1_bh *r1_bh;
		struct page *page;

		page = alloc_page(GFP_KERNEL);
		if (!page)
			break;

		r1_bh = (struct raid1_bh *) kmalloc(sizeof(*r1_bh), GFP_KERNEL);
		if (!r1_bh) {
			__free_page(page);
			break;
		}
		memset(r1_bh, 0, sizeof(*r1_bh));
		r1_bh->bh_req.b_page = page;
		r1_bh->bh_req.b_data = page_address(page);
		r1_bh->next_r1 = conf->freebuf;
		conf->freebuf = r1_bh;
		i++;
	}
	md_spin_unlock_irq(&conf->device_lock);
	return i;
}

static void raid1_shrink_buffers (raid1_conf_t *conf)
{
	md_spin_lock_irq(&conf->device_lock);
	while (conf->freebuf) {
		struct raid1_bh *r1_bh = conf->freebuf;
		conf->freebuf = r1_bh->next_r1;
		__free_page(r1_bh->bh_req.b_page);
		kfree(r1_bh);
	}
	md_spin_unlock_irq(&conf->device_lock);
}

static int raid1_map (mddev_t *mddev, kdev_t *rdev)
{
	raid1_conf_t *conf = mddev_to_conf(mddev);
	int i, disks = MD_SB_DISKS;

	/*
	 * Later we do read balancing on the read side 
	 * now we use the first available disk.
	 */

	for (i = 0; i < disks; i++) {
		if (conf->mirrors[i].operational) {
			*rdev = conf->mirrors[i].dev;
			return (0);
		}
	}

	printk (KERN_ERR "raid1_map(): huh, no more operational devices?\n");
	return (-1);
}

static void raid1_reschedule_retry (struct raid1_bh *r1_bh)
{
	unsigned long flags;
	mddev_t *mddev = r1_bh->mddev;
	raid1_conf_t *conf = mddev_to_conf(mddev);

	md_spin_lock_irqsave(&retry_list_lock, flags);
	if (raid1_retry_list == NULL)
		raid1_retry_tail = &raid1_retry_list;
	*raid1_retry_tail = r1_bh;
	raid1_retry_tail = &r1_bh->next_r1;
	r1_bh->next_r1 = NULL;
	md_spin_unlock_irqrestore(&retry_list_lock, flags);
	md_wakeup_thread(conf->thread);
}


static void inline io_request_done(unsigned long sector, raid1_conf_t *conf, int phase)
{
	unsigned long flags;
	spin_lock_irqsave(&conf->segment_lock, flags);
	if (sector < conf->start_active)
		conf->cnt_done--;
	else if (sector >= conf->start_future && conf->phase == phase)
		conf->cnt_future--;
	else if (!--conf->cnt_pending)
		wake_up(&conf->wait_ready);

	spin_unlock_irqrestore(&conf->segment_lock, flags);
}

static void inline sync_request_done (unsigned long sector, raid1_conf_t *conf)
{
	unsigned long flags;
	spin_lock_irqsave(&conf->segment_lock, flags);
	if (sector >= conf->start_ready)
		--conf->cnt_ready;
	else if (sector >= conf->start_active) {
		if (!--conf->cnt_active) {
			conf->start_active = conf->start_ready;
			wake_up(&conf->wait_done);
		}
	}
	spin_unlock_irqrestore(&conf->segment_lock, flags);
}

/*
 * raid1_end_bh_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 raid1_end_bh_io (struct raid1_bh *r1_bh, int uptodate)
{
	struct buffer_head *bh = r1_bh->master_bh;

	io_request_done(bh->b_rsector, mddev_to_conf(r1_bh->mddev),
			test_bit(R1BH_SyncPhase, &r1_bh->state));

	bh->b_end_io(bh, uptodate);
	raid1_free_r1bh(r1_bh);
}
void raid1_end_request (struct buffer_head *bh, int uptodate)
{
	struct raid1_bh * r1_bh = (struct raid1_bh *)(bh->b_private);

	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
	if (!uptodate)
		md_error (r1_bh->mddev, bh->b_dev);
	else
		/*
		 * Set R1BH_Uptodate in our master buffer_head, 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 complex operation to 
		 * user-side. So if something waits for IO, then it will
		 * wait for the 'master' buffer_head.
		 */
		set_bit (R1BH_Uptodate, &r1_bh->state);

	/*
	 * We split up the read and write side, imho they are 
	 * conceptually different.
	 */

	if ( (r1_bh->cmd == READ) || (r1_bh->cmd == READA) ) {
		/*
		 * we have only one buffer_head on the read side
		 */
		
		if (uptodate) {
			raid1_end_bh_io(r1_bh, uptodate);
			return;
		}
		/*
		 * oops, read error:
		 */
		printk(KERN_ERR "raid1: %s: rescheduling block %lu\n", 
			 partition_name(bh->b_dev), bh->b_blocknr);
		raid1_reschedule_retry(r1_bh);
		return;
	}

	/*
	 * WRITE:
	 *
	 * Let's see if all mirrored write operations have finished 
	 * already.
	 */

	if (atomic_dec_and_test(&r1_bh->remaining))
		raid1_end_bh_io(r1_bh, test_bit(R1BH_Uptodate, &r1_bh->state));
}

/*
 * This routine returns the disk from which the requested read should
 * be done. It bookkeeps the last read position for every disk
 * in array and when new read requests come, the disk which last
 * position is nearest to the request, is chosen.
 *
 * TODO: now if there are 2 mirrors in the same 2 devices, performance
 * degrades dramatically because position is mirror, not device based.
 * This should be changed to be device based. Also atomic sequential
 * reads should be somehow balanced.
 */

static int raid1_read_balance (raid1_conf_t *conf, struct buffer_head *bh)
{
	int new_disk = conf->last_used;
	const int sectors = bh->b_size >> 9;
	const unsigned long this_sector = bh->b_rsector;
	int disk = new_disk;
	unsigned long new_distance;
	unsigned long current_distance;
	
	/*
	 * Check if it is sane at all to balance
	 */
	
	if (conf->resync_mirrors)
		goto rb_out;
	

	/* make sure that disk is operational */
	while( !conf->mirrors[new_disk].operational) {
		if (new_disk <= 0) new_disk = conf->raid_disks;
		new_disk--;
		if (new_disk == disk) {
			/*
			 * This means no working disk was found
			 * Nothing much to do, lets not change anything
			 * and hope for the best...
			 */
			
			new_disk = conf->last_used;

			goto rb_out;
		}
	}
	disk = new_disk;
	/* now disk == new_disk == starting point for search */
	
	/*
	 * Don't touch anything for sequential reads.
	 */

	if (this_sector == conf->mirrors[new_disk].head_position)
		goto rb_out;
	
	/*
	 * If reads have been done only on a single disk
	 * for a time, lets give another disk a change.
	 * This is for kicking those idling disks so that
	 * they would find work near some hotspot.
	 */
	
	if (conf->sect_count >= conf->mirrors[new_disk].sect_limit) {
		conf->sect_count = 0;

		do {
			if (new_disk<=0)
				new_disk = conf->raid_disks;
			new_disk--;
			if (new_disk == disk)
				break;
		} while ((conf->mirrors[new_disk].write_only) ||
			 (!conf->mirrors[new_disk].operational));

		goto rb_out;
	}
	
	current_distance = abs(this_sector -
				conf->mirrors[disk].head_position);
	
	/* Find the disk which is closest */
	
	do {
		if (disk <= 0)
			disk = conf->raid_disks;
		disk--;
		
		if ((conf->mirrors[disk].write_only) ||
				(!conf->mirrors[disk].operational))
			continue;
		
		new_distance = abs(this_sector -
					conf->mirrors[disk].head_position);
		
		if (new_distance < current_distance) {
			conf->sect_count = 0;
			current_distance = new_distance;
			new_disk = disk;
		}
	} while (disk != conf->last_used);

rb_out:
	conf->mirrors[new_disk].head_position = this_sector + sectors;

	conf->last_used = new_disk;
	conf->sect_count += sectors;

	return new_disk;
}

static int raid1_make_request (mddev_t *mddev, int rw,
			       struct buffer_head * bh)
{
	raid1_conf_t *conf = mddev_to_conf(mddev);
	struct buffer_head *bh_req, *bhl;
	struct raid1_bh * r1_bh;
	int disks = MD_SB_DISKS;
	int i, sum_bhs = 0;
	struct mirror_info *mirror;

	if (!buffer_locked(bh))
		BUG();
	
/*
 * make_request() can abort the operation when READA is being
 * used and no empty request is available.
 *
 * Currently, just replace the command with READ/WRITE.
 */
	if (rw == READA)
		rw = READ;

	r1_bh = raid1_alloc_r1bh (conf);

	spin_lock_irq(&conf->segment_lock);
	wait_event_lock_irq(conf->wait_done,
			bh->b_rsector < conf->start_active ||
			bh->b_rsector >= conf->start_future,
			conf->segment_lock);
	if (bh->b_rsector < conf->start_active) 
		conf->cnt_done++;
	else {
		conf->cnt_future++;
		if (conf->phase)
			set_bit(R1BH_SyncPhase, &r1_bh->state);
	}
	spin_unlock_irq(&conf->segment_lock);
	
	/*
	 * i think the read and write branch should be separated completely,
	 * since we want to do read balancing on the read side for example.
	 * Alternative implementations? :) --mingo
	 */

	r1_bh->master_bh = bh;
	r1_bh->mddev = mddev;
	r1_bh->cmd = rw;

	if (rw == READ) {
		/*
		 * read balancing logic:
		 */
		mirror = conf->mirrors + raid1_read_balance(conf, bh);