* This file is released under the GPL.
*/
-#include "dm.h"
+#include <linux/device-mapper.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/slab.h>
#include <linux/time.h>
#include <linux/workqueue.h>
+#include <scsi/scsi_dh.h>
#include <asm/atomic.h>
#define DM_MSG_PREFIX "multipath"
struct list_head list;
struct priority_group *pg; /* Owning PG */
+ unsigned is_active; /* Path status */
unsigned fail_count; /* Cumulative failure count */
struct dm_path path;
+ struct work_struct deactivate_path;
};
#define path_to_pgpath(__pgp) container_of((__pgp), struct pgpath, path)
spinlock_t lock;
- struct hw_handler hw_handler;
+ const char *hw_handler_name;
+ struct work_struct activate_path;
+ struct pgpath *pgpath_to_activate;
unsigned nr_priority_groups;
struct list_head priority_groups;
unsigned pg_init_required; /* pg_init needs calling? */
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;
static struct kmem_cache *_mpio_cache;
-struct workqueue_struct *kmultipathd;
+static struct workqueue_struct *kmultipathd, *kmpath_handlerd;
static void process_queued_ios(struct work_struct *work);
static void trigger_event(struct work_struct *work);
+static void activate_path(struct work_struct *work);
+static void deactivate_path(struct work_struct *work);
/*-----------------------------------------------
{
struct pgpath *pgpath = kzalloc(sizeof(*pgpath), GFP_KERNEL);
- if (pgpath)
- pgpath->path.is_active = 1;
+ if (pgpath) {
+ pgpath->is_active = 1;
+ INIT_WORK(&pgpath->deactivate_path, deactivate_path);
+ }
return pgpath;
}
kfree(pgpath);
}
+static void deactivate_path(struct work_struct *work)
+{
+ struct pgpath *pgpath =
+ container_of(work, struct pgpath, deactivate_path);
+
+ blk_abort_queue(pgpath->path.dev->bdev->bd_disk->queue);
+}
+
static struct priority_group *alloc_priority_group(void)
{
struct priority_group *pg;
static void free_pgpaths(struct list_head *pgpaths, struct dm_target *ti)
{
+ unsigned long flags;
struct pgpath *pgpath, *tmp;
+ struct multipath *m = ti->private;
list_for_each_entry_safe(pgpath, tmp, pgpaths, list) {
list_del(&pgpath->list);
+ if (m->hw_handler_name)
+ scsi_dh_detach(bdev_get_queue(pgpath->path.dev->bdev));
dm_put_device(ti, pgpath->path.dev);
+ spin_lock_irqsave(&m->lock, flags);
+ if (m->pgpath_to_activate == pgpath)
+ m->pgpath_to_activate = NULL;
+ spin_unlock_irqrestore(&m->lock, flags);
free_pgpath(pgpath);
}
}
m->queue_io = 1;
INIT_WORK(&m->process_queued_ios, process_queued_ios);
INIT_WORK(&m->trigger_event, trigger_event);
+ INIT_WORK(&m->activate_path, activate_path);
m->mpio_pool = mempool_create_slab_pool(MIN_IOS, _mpio_cache);
if (!m->mpio_pool) {
kfree(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);
- }
-
+ kfree(m->hw_handler_name);
mempool_destroy(m->mpio_pool);
kfree(m);
}
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) {
+ if (m->hw_handler_name) {
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)
r = map_io(m, bio, mpio, 1);
if (r < 0)
- bio_endio(bio, bio->bi_size, r);
+ bio_endio(bio, r);
else if (r == DM_MAPIO_REMAPPED)
generic_make_request(bio);
else if (r == DM_MAPIO_REQUEUE)
- bio_endio(bio, bio->bi_size, -EIO);
+ bio_endio(bio, -EIO);
bio = next;
}
{
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;
(!pgpath && !m->queue_if_no_path))
must_queue = 0;
- if (m->pg_init_required && !m->pg_init_in_progress) {
+ if (m->pg_init_required && !m->pg_init_in_progress && pgpath) {
+ m->pgpath_to_activate = pgpath;
+ m->pg_init_count++;
m->pg_init_required = 0;
m->pg_init_in_progress = 1;
init_required = 1;
spin_unlock_irqrestore(&m->lock, flags);
if (init_required)
- hwh->type->pg_init(hwh, pgpath->pg->bypassed, &pgpath->path);
+ queue_work(kmpath_handlerd, &m->activate_path);
if (!must_queue)
dispatch_queued_ios(m);
}
r = read_param(_params, shift(as), &ps_argc, &ti->error);
- if (r)
+ if (r) {
+ dm_put_path_selector(pst);
return -EINVAL;
+ }
r = pst->create(&pg->ps, ps_argc, as->argv);
if (r) {
{
int r;
struct pgpath *p;
+ struct multipath *m = ti->private;
/* we need at least a path arg */
if (as->argc < 1) {
ti->error = "no device given";
- return NULL;
+ return ERR_PTR(-EINVAL);
}
p = alloc_pgpath();
if (!p)
- return NULL;
+ return ERR_PTR(-ENOMEM);
r = dm_get_device(ti, shift(as), ti->begin, ti->len,
dm_table_get_mode(ti->table), &p->path.dev);
goto bad;
}
+ if (m->hw_handler_name) {
+ r = scsi_dh_attach(bdev_get_queue(p->path.dev->bdev),
+ m->hw_handler_name);
+ if (r < 0) {
+ dm_put_device(ti, p->path.dev);
+ goto bad;
+ }
+ }
+
r = ps->type->add_path(ps, &p->path, as->argc, as->argv, &ti->error);
if (r) {
dm_put_device(ti, p->path.dev);
bad:
free_pgpath(p);
- return NULL;
+ return ERR_PTR(r);
}
static struct priority_group *parse_priority_group(struct arg_set *as,
if (as->argc < 2) {
as->argc = 0;
- ti->error = "not enough priority group aruments";
- return NULL;
+ ti->error = "not enough priority group arguments";
+ return ERR_PTR(-EINVAL);
}
pg = alloc_priority_group();
if (!pg) {
ti->error = "couldn't allocate priority group";
- return NULL;
+ return ERR_PTR(-ENOMEM);
}
pg->m = m;
struct pgpath *pgpath;
struct arg_set path_args;
- if (as->argc < nr_params)
+ if (as->argc < nr_params) {
+ ti->error = "not enough path parameters";
goto bad;
+ }
path_args.argc = nr_params;
path_args.argv = as->argv;
pgpath = parse_path(&path_args, &pg->ps, ti);
- if (!pgpath)
+ if (IS_ERR(pgpath)) {
+ r = PTR_ERR(pgpath);
goto bad;
+ }
pgpath->pg = pg;
list_add_tail(&pgpath->list, &pg->pgpaths);
bad:
free_priority_group(pg, ti);
- return NULL;
+ return ERR_PTR(r);
}
static int parse_hw_handler(struct arg_set *as, struct multipath *m)
{
- int r;
- struct hw_handler_type *hwht;
unsigned hw_argc;
struct dm_target *ti = m->ti;
{0, 1024, "invalid number of hardware handler args"},
};
- r = read_param(_params, shift(as), &hw_argc, &ti->error);
- if (r)
+ if (read_param(_params, shift(as), &hw_argc, &ti->error))
return -EINVAL;
if (!hw_argc)
return 0;
- hwht = dm_get_hw_handler(shift(as));
- if (!hwht) {
+ m->hw_handler_name = kstrdup(shift(as), GFP_KERNEL);
+ request_module("scsi_dh_%s", m->hw_handler_name);
+ if (scsi_dh_handler_exist(m->hw_handler_name) == 0) {
ti->error = "unknown hardware handler type";
+ kfree(m->hw_handler_name);
+ m->hw_handler_name = NULL;
return -EINVAL;
}
- m->hw_handler.md = dm_table_get_md(ti->table);
- dm_put(m->hw_handler.md);
-
- r = hwht->create(&m->hw_handler, hw_argc - 1, as->argv);
- if (r) {
- dm_put_hw_handler(hwht);
- ti->error = "hardware handler constructor failed";
- return r;
- }
-
- m->hw_handler.type = hwht;
+ if (hw_argc > 1)
+ DMWARN("Ignoring user-specified arguments for "
+ "hardware handler \"%s\"", m->hw_handler_name);
consume(as, hw_argc - 1);
return 0;
int r;
unsigned argc;
struct dm_target *ti = m->ti;
+ const char *param_name;
static struct param _params[] = {
- {0, 1, "invalid number of feature args"},
+ {0, 3, "invalid number of feature args"},
+ {1, 50, "pg_init_retries must be between 1 and 50"},
};
r = read_param(_params, shift(as), &argc, &ti->error);
if (!argc)
return 0;
- if (!strnicmp(shift(as), MESG_STR("queue_if_no_path")))
- return queue_if_no_path(m, 1, 0);
- else {
+ do {
+ param_name = shift(as);
+ argc--;
+
+ if (!strnicmp(param_name, MESG_STR("queue_if_no_path"))) {
+ r = queue_if_no_path(m, 1, 0);
+ continue;
+ }
+
+ if (!strnicmp(param_name, MESG_STR("pg_init_retries")) &&
+ (argc >= 1)) {
+ r = read_param(_params + 1, shift(as),
+ &m->pg_init_retries, &ti->error);
+ argc--;
+ continue;
+ }
+
ti->error = "Unrecognised multipath feature request";
- return -EINVAL;
- }
+ r = -EINVAL;
+ } while (argc && !r);
+
+ return r;
}
static int multipath_ctr(struct dm_target *ti, unsigned int argc,
struct priority_group *pg;
pg = parse_priority_group(&as, m);
- if (!pg) {
- r = -EINVAL;
+ if (IS_ERR(pg)) {
+ r = PTR_ERR(pg);
goto bad;
}
{
struct multipath *m = (struct multipath *) ti->private;
+ flush_workqueue(kmpath_handlerd);
flush_workqueue(kmultipathd);
free_multipath(m);
}
struct dm_mpath_io *mpio;
struct multipath *m = (struct multipath *) ti->private;
- if (bio_barrier(bio))
- return -EOPNOTSUPP;
-
mpio = mempool_alloc(m->mpio_pool, GFP_NOIO);
dm_bio_record(&mpio->details, bio);
map_context->ptr = mpio;
- bio->bi_rw |= (1 << BIO_RW_FAILFAST);
+ bio->bi_rw |= (1 << BIO_RW_FAILFAST_TRANSPORT);
r = map_io(m, bio, mpio, 0);
if (r < 0 || r == DM_MAPIO_REQUEUE)
mempool_free(mpio, m->mpio_pool);
spin_lock_irqsave(&m->lock, flags);
- if (!pgpath->path.is_active)
+ if (!pgpath->is_active)
goto out;
DMWARN("Failing path %s.", pgpath->path.dev->name);
pgpath->pg->ps.type->fail_path(&pgpath->pg->ps, &pgpath->path);
- pgpath->path.is_active = 0;
+ pgpath->is_active = 0;
pgpath->fail_count++;
m->nr_valid_paths--;
if (pgpath == m->current_pgpath)
m->current_pgpath = NULL;
- queue_work(kmultipathd, &m->trigger_event);
+ dm_path_uevent(DM_UEVENT_PATH_FAILED, m->ti,
+ pgpath->path.dev->name, m->nr_valid_paths);
+
+ schedule_work(&m->trigger_event);
+ queue_work(kmultipathd, &pgpath->deactivate_path);
out:
spin_unlock_irqrestore(&m->lock, flags);
spin_lock_irqsave(&m->lock, flags);
- if (pgpath->path.is_active)
+ if (pgpath->is_active)
goto out;
- if (!pgpath->pg->ps.type) {
+ if (!pgpath->pg->ps.type->reinstate_path) {
DMWARN("Reinstate path not supported by path selector %s",
pgpath->pg->ps.type->name);
r = -EINVAL;
if (r)
goto out;
- pgpath->path.is_active = 1;
+ pgpath->is_active = 1;
m->current_pgpath = NULL;
if (!m->nr_valid_paths++ && m->queue_size)
queue_work(kmultipathd, &m->process_queued_ios);
- queue_work(kmultipathd, &m->trigger_event);
+ dm_path_uevent(DM_UEVENT_PATH_REINSTATED, m->ti,
+ pgpath->path.dev->name, m->nr_valid_paths);
+
+ schedule_work(&m->trigger_event);
out:
spin_unlock_irqrestore(&m->lock, flags);
spin_unlock_irqrestore(&m->lock, flags);
- queue_work(kmultipathd, &m->trigger_event);
+ schedule_work(&m->trigger_event);
}
/*
}
spin_unlock_irqrestore(&m->lock, flags);
- queue_work(kmultipathd, &m->trigger_event);
+ schedule_work(&m->trigger_event);
return 0;
}
}
/*
- * pg_init must call this when it has completed its initialisation
+ * Should we retry pg_init immediately?
*/
-void dm_pg_init_complete(struct dm_path *path, unsigned err_flags)
+static int pg_init_limit_reached(struct multipath *m, struct pgpath *pgpath)
+{
+ unsigned long flags;
+ int limit_reached = 0;
+
+ spin_lock_irqsave(&m->lock, flags);
+
+ if (m->pg_init_count <= m->pg_init_retries)
+ m->pg_init_required = 1;
+ else
+ limit_reached = 1;
+
+ spin_unlock_irqrestore(&m->lock, flags);
+
+ return limit_reached;
+}
+
+static void pg_init_done(struct dm_path *path, int errors)
{
struct pgpath *pgpath = path_to_pgpath(path);
struct priority_group *pg = pgpath->pg;
struct multipath *m = pg->m;
unsigned long flags;
- /* We insist on failing the path if the PG is already bypassed. */
- if (err_flags && pg->bypassed)
- err_flags |= MP_FAIL_PATH;
-
- if (err_flags & MP_FAIL_PATH)
+ /* device or driver problems */
+ switch (errors) {
+ case SCSI_DH_OK:
+ break;
+ case SCSI_DH_NOSYS:
+ if (!m->hw_handler_name) {
+ errors = 0;
+ break;
+ }
+ DMERR("Cannot failover device because scsi_dh_%s was not "
+ "loaded.", m->hw_handler_name);
+ /*
+ * Fail path for now, so we do not ping pong
+ */
fail_path(pgpath);
-
- if (err_flags & MP_BYPASS_PG)
+ break;
+ case SCSI_DH_DEV_TEMP_BUSY:
+ /*
+ * Probably doing something like FW upgrade on the
+ * controller so try the other pg.
+ */
bypass_pg(m, pg, 1);
+ break;
+ /* TODO: For SCSI_DH_RETRY we should wait a couple seconds */
+ case SCSI_DH_RETRY:
+ case SCSI_DH_IMM_RETRY:
+ case SCSI_DH_RES_TEMP_UNAVAIL:
+ if (pg_init_limit_reached(m, pgpath))
+ fail_path(pgpath);
+ errors = 0;
+ break;
+ default:
+ /*
+ * We probably do not want to fail the path for a device
+ * error, but this is what the old dm did. In future
+ * patches we can do more advanced handling.
+ */
+ fail_path(pgpath);
+ }
spin_lock_irqsave(&m->lock, flags);
- if (err_flags) {
+ if (errors) {
+ DMERR("Could not failover device. Error %d.", errors);
m->current_pgpath = NULL;
m->current_pg = NULL;
- } else if (!m->pg_init_required)
+ } else if (!m->pg_init_required) {
m->queue_io = 0;
+ pg->bypassed = 0;
+ }
m->pg_init_in_progress = 0;
queue_work(kmultipathd, &m->process_queued_ios);
spin_unlock_irqrestore(&m->lock, flags);
}
+static void activate_path(struct work_struct *work)
+{
+ int ret;
+ struct multipath *m =
+ container_of(work, struct multipath, activate_path);
+ struct dm_path *path;
+ unsigned long flags;
+
+ spin_lock_irqsave(&m->lock, flags);
+ path = &m->pgpath_to_activate->path;
+ m->pgpath_to_activate = NULL;
+ spin_unlock_irqrestore(&m->lock, flags);
+ if (!path)
+ return;
+ ret = scsi_dh_activate(bdev_get_queue(path->dev->bdev));
+ pg_init_done(path, ret);
+}
+
/*
* end_io handling
*/
static int do_end_io(struct multipath *m, struct bio *bio,
int error, struct dm_mpath_io *mpio)
{
- struct hw_handler *hwh = &m->hw_handler;
- unsigned err_flags = MP_FAIL_PATH; /* Default behavior */
unsigned long flags;
if (!error)
}
spin_unlock_irqrestore(&m->lock, flags);
- if (hwh->type && hwh->type->error)
- err_flags = hwh->type->error(hwh, bio);
-
- if (mpio->pgpath) {
- if (err_flags & MP_FAIL_PATH)
- fail_path(mpio->pgpath);
-
- if (err_flags & MP_BYPASS_PG)
- bypass_pg(m, mpio->pgpath->pg, 1);
- }
-
- if (err_flags & MP_ERROR_IO)
- return -EIO;
+ if (mpio->pgpath)
+ fail_path(mpio->pgpath);
requeue:
dm_bio_restore(&mpio->details, bio);
int sz = 0;
unsigned long flags;
struct multipath *m = (struct multipath *) ti->private;
- struct hw_handler *hwh = &m->hw_handler;
struct priority_group *pg;
struct pgpath *p;
unsigned pg_num;
/* Features */
if (type == STATUSTYPE_INFO)
- DMEMIT("1 %u ", m->queue_size);
- else if (m->queue_if_no_path)
- DMEMIT("1 queue_if_no_path ");
- else
- DMEMIT("0 ");
+ DMEMIT("2 %u %u ", m->queue_size, m->pg_init_count);
+ else {
+ DMEMIT("%u ", m->queue_if_no_path +
+ (m->pg_init_retries > 0) * 2);
+ if (m->queue_if_no_path)
+ DMEMIT("queue_if_no_path ");
+ if (m->pg_init_retries)
+ DMEMIT("pg_init_retries %u ", m->pg_init_retries);
+ }
- if (hwh->type && hwh->type->status)
- sz += hwh->type->status(hwh, type, result + sz, maxlen - sz);
- else if (!hwh->type || type == STATUSTYPE_INFO)
+ if (!m->hw_handler_name || type == STATUSTYPE_INFO)
DMEMIT("0 ");
else
- DMEMIT("1 %s ", hwh->type->name);
+ DMEMIT("1 %s ", m->hw_handler_name);
DMEMIT("%u ", m->nr_priority_groups);
list_for_each_entry(p, &pg->pgpaths, list) {
DMEMIT("%s %s %u ", p->path.dev->name,
- p->path.is_active ? "A" : "F",
+ p->is_active ? "A" : "F",
p->fail_count);
if (pg->ps.type->status)
sz += pg->ps.type->status(&pg->ps,
return -EINVAL;
}
-static int multipath_ioctl(struct dm_target *ti, struct inode *inode,
- struct file *filp, unsigned int cmd,
+static int multipath_ioctl(struct dm_target *ti, unsigned int cmd,
unsigned long arg)
{
struct multipath *m = (struct multipath *) ti->private;
struct block_device *bdev = NULL;
+ fmode_t mode = 0;
unsigned long flags;
- struct file fake_file = {};
- struct dentry fake_dentry = {};
int r = 0;
- fake_file.f_path.dentry = &fake_dentry;
-
spin_lock_irqsave(&m->lock, flags);
if (!m->current_pgpath)
if (m->current_pgpath) {
bdev = m->current_pgpath->path.dev->bdev;
- fake_dentry.d_inode = bdev->bd_inode;
- fake_file.f_mode = m->current_pgpath->path.dev->mode;
+ mode = m->current_pgpath->path.dev->mode;
}
if (m->queue_io)
spin_unlock_irqrestore(&m->lock, flags);
- return r ? : blkdev_driver_ioctl(bdev->bd_inode, &fake_file,
- bdev->bd_disk, cmd, arg);
+ return r ? : __blkdev_driver_ioctl(bdev, mode, cmd, arg);
}
/*-----------------------------------------------------------------
return -ENOMEM;
}
+ /*
+ * A separate workqueue is used to handle the device handlers
+ * to avoid overloading existing workqueue. Overloading the
+ * old workqueue would also create a bottleneck in the
+ * path of the storage hardware device activation.
+ */
+ kmpath_handlerd = create_singlethread_workqueue("kmpath_handlerd");
+ if (!kmpath_handlerd) {
+ DMERR("failed to create workqueue kmpath_handlerd");
+ destroy_workqueue(kmultipathd);
+ dm_unregister_target(&multipath_target);
+ kmem_cache_destroy(_mpio_cache);
+ return -ENOMEM;
+ }
+
DMINFO("version %u.%u.%u loaded",
multipath_target.version[0], multipath_target.version[1],
multipath_target.version[2]);
static void __exit dm_multipath_exit(void)
{
- int r;
-
+ destroy_workqueue(kmpath_handlerd);
destroy_workqueue(kmultipathd);
- r = dm_unregister_target(&multipath_target);
- if (r < 0)
- DMERR("target unregister failed %d", r);
+ dm_unregister_target(&multipath_target);
kmem_cache_destroy(_mpio_cache);
}
-EXPORT_SYMBOL_GPL(dm_pg_init_complete);
-
module_init(dm_multipath_init);
module_exit(dm_multipath_exit);
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff