dm-mpath.c

linux 内核源代码

/*
 * Copyright (C) 2003 Sistina Software Limited.
 * Copyright (C) 2004-2005 Red Hat, Inc. All rights reserved.
 *
 * This file is released under the GPL.
 */

#include "dm.h"
#include "dm-path-selector.h"
#include "dm-hw-handler.h"
#include "dm-bio-list.h"
#include "dm-bio-record.h"
#include "dm-uevent.h"

#include <linux/ctype.h>
#include <linux/init.h>
#include <linux/mempool.h>
#include <linux/module.h>
#include <linux/pagemap.h>
#include <linux/slab.h>
#include <linux/time.h>
#include <linux/workqueue.h>
#include <asm/atomic.h>

#define DM_MSG_PREFIX "multipath"
#define MESG_STR(x) x, sizeof(x)

/* Path properties */
struct pgpath {
	struct list_head list;

	struct priority_group *pg;	/* Owning PG */
	unsigned fail_count;		/* Cumulative failure count */

	struct dm_path path;
};

#define path_to_pgpath(__pgp) container_of((__pgp), struct pgpath, path)

/*
 * Paths are grouped into Priority Groups and numbered from 1 upwards.
 * Each has a path selector which controls which path gets used.
 */
struct priority_group {
	struct list_head list;

	struct multipath *m;		/* Owning multipath instance */
	struct path_selector ps;

	unsigned pg_num;		/* Reference number */
	unsigned bypassed;		/* Temporarily bypass this PG? */

	unsigned nr_pgpaths;		/* Number of paths in PG */
	struct list_head pgpaths;
};

/* Multipath context */
struct multipath {
	struct list_head list;
	struct dm_target *ti;

	spinlock_t lock;

	struct hw_handler hw_handler;
	unsigned nr_priority_groups;
	struct list_head priority_groups;
	unsigned pg_init_required;	/* pg_init needs calling? */
	unsigned pg_init_in_progress;	/* Only one pg_init allowed at once */

	unsigned nr_valid_paths;	/* Total number of usable paths */
	struct pgpath *current_pgpath;
	struct priority_group *current_pg;
	struct priority_group *next_pg;	/* Switch to this PG if set */
	unsigned repeat_count;		/* I/Os left before calling PS again */

	unsigned queue_io;		/* Must we queue all I/O? */
	unsigned queue_if_no_path;	/* Queue I/O if last path fails? */
	unsigned saved_queue_if_no_path;/* Saved state during suspension */
	unsigned pg_init_retries;	/* Number of times to retry pg_init */
	unsigned pg_init_count;		/* Number of times pg_init called */

	struct work_struct process_queued_ios;
	struct bio_list queued_ios;
	unsigned queue_size;

	struct work_struct trigger_event;

	/*
	 * We must use a mempool of dm_mpath_io structs so that we
	 * can resubmit bios on error.
	 */
	mempool_t *mpio_pool;
};

/*
 * Context information attached to each bio we process.
 */
struct dm_mpath_io {
	struct pgpath *pgpath;
	struct dm_bio_details details;
};

typedef int (*action_fn) (struct pgpath *pgpath);

#define MIN_IOS 256	/* Mempool size */

static struct kmem_cache *_mpio_cache;

struct workqueue_struct *kmultipathd;
static void process_queued_ios(struct work_struct *work);
static void trigger_event(struct work_struct *work);


/*-----------------------------------------------
 * Allocation routines
 *-----------------------------------------------*/

static struct pgpath *alloc_pgpath(void)
{
	struct pgpath *pgpath = kzalloc(sizeof(*pgpath), GFP_KERNEL);

	if (pgpath)
		pgpath->path.is_active = 1;

	return pgpath;
}

static void free_pgpath(struct pgpath *pgpath)
{
	kfree(pgpath);
}

static struct priority_group *alloc_priority_group(void)
{
	struct priority_group *pg;

	pg = kzalloc(sizeof(*pg), GFP_KERNEL);

	if (pg)
		INIT_LIST_HEAD(&pg->pgpaths);

	return pg;
}

static void free_pgpaths(struct list_head *pgpaths, struct dm_target *ti)
{
	struct pgpath *pgpath, *tmp;

	list_for_each_entry_safe(pgpath, tmp, pgpaths, list) {
		list_del(&pgpath->list);
		dm_put_device(ti, pgpath->path.dev);
		free_pgpath(pgpath);
	}
}

static void free_priority_group(struct priority_group *pg,
				struct dm_target *ti)
{
	struct path_selector *ps = &pg->ps;

	if (ps->type) {
		ps->type->destroy(ps);
		dm_put_path_selector(ps->type);
	}

	free_pgpaths(&pg->pgpaths, ti);
	kfree(pg);
}

static struct multipath *alloc_multipath(struct dm_target *ti)
{
	struct multipath *m;

	m = kzalloc(sizeof(*m), GFP_KERNEL);
	if (m) {
		INIT_LIST_HEAD(&m->priority_groups);
		spin_lock_init(&m->lock);
		m->queue_io = 1;
		INIT_WORK(&m->process_queued_ios, process_queued_ios);
		INIT_WORK(&m->trigger_event, trigger_event);
		m->mpio_pool = mempool_create_slab_pool(MIN_IOS, _mpio_cache);
		if (!m->mpio_pool) {
			kfree(m);
			return NULL;
		}
		m->ti = ti;
		ti->private = m;
	}

	return m;
}

static void free_multipath(struct multipath *m)
{
	struct priority_group *pg, *tmp;
	struct hw_handler *hwh = &m->hw_handler;

	list_for_each_entry_safe(pg, tmp, &m->priority_groups, list) {
		list_del(&pg->list);
		free_priority_group(pg, m->ti);
	}

	if (hwh->type) {
		hwh->type->destroy(hwh);
		dm_put_hw_handler(hwh->type);
	}

	mempool_destroy(m->mpio_pool);
	kfree(m);
}


/*-----------------------------------------------
 * Path selection
 *-----------------------------------------------*/

static void __switch_pg(struct multipath *m, struct pgpath *pgpath)
{
	struct hw_handler *hwh = &m->hw_handler;

	m->current_pg = pgpath->pg;

	/* Must we initialise the PG first, and queue I/O till it's ready? */
	if (hwh->type && hwh->type->pg_init) {
		m->pg_init_required = 1;
		m->queue_io = 1;
	} else {
		m->pg_init_required = 0;
		m->queue_io = 0;
	}

	m->pg_init_count = 0;
}

static int __choose_path_in_pg(struct multipath *m, struct priority_group *pg)
{
	struct dm_path *path;

	path = pg->ps.type->select_path(&pg->ps, &m->repeat_count);
	if (!path)
		return -ENXIO;

	m->current_pgpath = path_to_pgpath(path);

	if (m->current_pg != pg)
		__switch_pg(m, m->current_pgpath);

	return 0;
}

static void __choose_pgpath(struct multipath *m)
{
	struct priority_group *pg;
	unsigned bypassed = 1;

	if (!m->nr_valid_paths)
		goto failed;

	/* Were we instructed to switch PG? */
	if (m->next_pg) {
		pg = m->next_pg;
		m->next_pg = NULL;
		if (!__choose_path_in_pg(m, pg))
			return;
	}

	/* Don't change PG until it has no remaining paths */
	if (m->current_pg && !__choose_path_in_pg(m, m->current_pg))
		return;

	/*
	 * Loop through priority groups until we find a valid path.
	 * First time we skip PGs marked 'bypassed'.
	 * Second time we only try the ones we skipped.
	 */
	do {
		list_for_each_entry(pg, &m->priority_groups, list) {
			if (pg->bypassed == bypassed)
				continue;
			if (!__choose_path_in_pg(m, pg))
				return;
		}
	} while (bypassed--);

failed:
	m->current_pgpath = NULL;
	m->current_pg = NULL;
}

/*
 * Check whether bios must be queued in the device-mapper core rather
 * than here in the target.
 *
 * m->lock must be held on entry.
 *
 * If m->queue_if_no_path and m->saved_queue_if_no_path hold the
 * same value then we are not between multipath_presuspend()
 * and multipath_resume() calls and we have no need to check
 * for the DMF_NOFLUSH_SUSPENDING flag.
 */
static int __must_push_back(struct multipath *m)
{
	return (m->queue_if_no_path != m->saved_queue_if_no_path &&
		dm_noflush_suspending(m->ti));
}

static int map_io(struct multipath *m, struct bio *bio,
		  struct dm_mpath_io *mpio, unsigned was_queued)
{
	int r = DM_MAPIO_REMAPPED;
	unsigned long flags;
	struct pgpath *pgpath;

	spin_lock_irqsave(&m->lock, flags);

	/* Do we need to select a new pgpath? */
	if (!m->current_pgpath ||
	    (!m->queue_io && (m->repeat_count && --m->repeat_count == 0)))
		__choose_pgpath(m);

	pgpath = m->current_pgpath;

	if (was_queued)
		m->queue_size--;

	if ((pgpath && m->queue_io) ||
	    (!pgpath && m->queue_if_no_path)) {
		/* Queue for the daemon to resubmit */
		bio_list_add(&m->queued_ios, bio);
		m->queue_size++;
		if ((m->pg_init_required && !m->pg_init_in_progress) ||
		    !m->queue_io)
			queue_work(kmultipathd, &m->process_queued_ios);
		pgpath = NULL;
		r = DM_MAPIO_SUBMITTED;
	} else if (pgpath)
		bio->bi_bdev = pgpath->path.dev->bdev;
	else if (__must_push_back(m))
		r = DM_MAPIO_REQUEUE;
	else
		r = -EIO;	/* Failed */

	mpio->pgpath = pgpath;

	spin_unlock_irqrestore(&m->lock, flags);

	return r;
}

/*
 * If we run out of usable paths, should we queue I/O or error it?
 */
static int queue_if_no_path(struct multipath *m, unsigned queue_if_no_path,
			    unsigned save_old_value)
{
	unsigned long flags;

	spin_lock_irqsave(&m->lock, flags);

	if (save_old_value)
		m->saved_queue_if_no_path = m->queue_if_no_path;
	else
		m->saved_queue_if_no_path = queue_if_no_path;
	m->queue_if_no_path = queue_if_no_path;
	if (!m->queue_if_no_path && m->queue_size)
		queue_work(kmultipathd, &m->process_queued_ios);

	spin_unlock_irqrestore(&m->lock, flags);

	return 0;
}

/*-----------------------------------------------------------------
 * The multipath daemon is responsible for resubmitting queued ios.
 *---------------------------------------------------------------*/

static void dispatch_queued_ios(struct multipath *m)
{
	int r;
	unsigned long flags;
	struct bio *bio = NULL, *next;
	struct dm_mpath_io *mpio;
	union map_info *info;

	spin_lock_irqsave(&m->lock, flags);
	bio = bio_list_get(&m->queued_ios);
	spin_unlock_irqrestore(&m->lock, flags);

	while (bio) {
		next = bio->bi_next;
		bio->bi_next = NULL;

		info = dm_get_mapinfo(bio);
		mpio = info->ptr;

		r = map_io(m, bio, mpio, 1);
		if (r < 0)
			bio_endio(bio, r);
		else if (r == DM_MAPIO_REMAPPED)
			generic_make_request(bio);
		else if (r == DM_MAPIO_REQUEUE)
			bio_endio(bio, -EIO);

		bio = next;
	}
}

static void process_queued_ios(struct work_struct *work)
{
	struct multipath *m =
		container_of(work, struct multipath, process_queued_ios);
	struct hw_handler *hwh = &m->hw_handler;
	struct pgpath *pgpath = NULL;
	unsigned init_required = 0, must_queue = 1;
	unsigned long flags;

	spin_lock_irqsave(&m->lock, flags);

	if (!m->queue_size)
		goto out;

	if (!m->current_pgpath)
		__choose_pgpath(m);

	pgpath = m->current_pgpath;

	if ((pgpath && !m->queue_io) ||
	    (!pgpath && !m->queue_if_no_path))
		must_queue = 0;

	if (m->pg_init_required && !m->pg_init_in_progress) {
		m->pg_init_count++;
		m->pg_init_required = 0;
		m->pg_init_in_progress = 1;
		init_required = 1;
	}

out:
	spin_unlock_irqrestore(&m->lock, flags);

	if (init_required)
		hwh->type->pg_init(hwh, pgpath->pg->bypassed, &pgpath->path);

	if (!must_queue)
		dispatch_queued_ios(m);
}

/*
 * An event is triggered whenever a path is taken out of use.
 * Includes path failure and PG bypass.
 */
static void trigger_event(struct work_struct *work)
{
	struct multipath *m =
		container_of(work, struct multipath, trigger_event);

	dm_table_event(m->ti->table);
}

/*-----------------------------------------------------------------
 * Constructor/argument parsing:
 * <#multipath feature args> [<arg>]*
 * <#hw_handler args> [hw_handler [<arg>]*]
 * <#priority groups>
 * <initial priority group>
 *     [<selector> <#selector args> [<arg>]*
 *      <#paths> <#per-path selector args>
 *         [<path> [<arg>]* ]+ ]+
 *---------------------------------------------------------------*/
struct param {
	unsigned min;
	unsigned max;
	char *error;
};

static int read_param(struct param *param, char *str, unsigned *v, char **error)
{
	if (!str ||
	    (sscanf(str, "%u", v) != 1) ||
	    (*v < param->min) ||
	    (*v > param->max)) {
		*error = param->error;
		return -EINVAL;
	}