From 0bca4a5b78ad922624e9a76a7e86453f1fec7239 Mon Sep 17 00:00:00 2001 From: Mikulas Patocka Date: Thu, 10 Dec 2009 23:51:52 +0000 Subject: dm: avoid _hash_lock deadlock commit 6076905b5ef39e0ea58db32583c9e0036c05e47b upstream. Fix a reported deadlock if there are still unprocessed multipath events on a device that is being removed. _hash_lock is held during dev_remove while trying to send the outstanding events. Sending the events requests the _hash_lock again in dm_copy_name_and_uuid. This patch introduces a separate lock around regions that modify the link to the hash table (dm_set_mdptr) or the name or uuid so that dm_copy_name_and_uuid no longer needs _hash_lock. Additionally, dm_copy_name_and_uuid can only be called if md exists so we can drop the dm_get() and dm_put() which can lead to a BUG() while md is being freed. The deadlock: #0 [ffff8106298dfb48] schedule at ffffffff80063035 #1 [ffff8106298dfc20] __down_read at ffffffff8006475d #2 [ffff8106298dfc60] dm_copy_name_and_uuid at ffffffff8824f740 #3 [ffff8106298dfc90] dm_send_uevents at ffffffff88252685 #4 [ffff8106298dfcd0] event_callback at ffffffff8824c678 #5 [ffff8106298dfd00] dm_table_event at ffffffff8824dd01 #6 [ffff8106298dfd10] __hash_remove at ffffffff882507ad #7 [ffff8106298dfd30] dev_remove at ffffffff88250865 #8 [ffff8106298dfd60] ctl_ioctl at ffffffff88250d80 #9 [ffff8106298dfee0] do_ioctl at ffffffff800418c4 #10 [ffff8106298dff00] vfs_ioctl at ffffffff8002fab9 #11 [ffff8106298dff40] sys_ioctl at ffffffff8004bdaf #12 [ffff8106298dff80] tracesys at ffffffff8005d28d (via system_call) Reported-by: guy keren Signed-off-by: Mikulas Patocka Signed-off-by: Alasdair G Kergon Signed-off-by: Greg Kroah-Hartman --- drivers/md/dm-ioctl.c | 17 +++++++++++++---- drivers/md/dm-uevent.c | 9 ++++----- 2 files changed, 17 insertions(+), 9 deletions(-) diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c index a67942931582..d19854c98184 100644 --- a/drivers/md/dm-ioctl.c +++ b/drivers/md/dm-ioctl.c @@ -56,6 +56,11 @@ static void dm_hash_remove_all(int keep_open_devices); */ static DECLARE_RWSEM(_hash_lock); +/* + * Protects use of mdptr to obtain hash cell name and uuid from mapped device. + */ +static DEFINE_MUTEX(dm_hash_cells_mutex); + static void init_buckets(struct list_head *buckets) { unsigned int i; @@ -206,7 +211,9 @@ static int dm_hash_insert(const char *name, const char *uuid, struct mapped_devi list_add(&cell->uuid_list, _uuid_buckets + hash_str(uuid)); } dm_get(md); + mutex_lock(&dm_hash_cells_mutex); dm_set_mdptr(md, cell); + mutex_unlock(&dm_hash_cells_mutex); up_write(&_hash_lock); return 0; @@ -224,7 +231,9 @@ static void __hash_remove(struct hash_cell *hc) /* remove from the dev hash */ list_del(&hc->uuid_list); list_del(&hc->name_list); + mutex_lock(&dm_hash_cells_mutex); dm_set_mdptr(hc->md, NULL); + mutex_unlock(&dm_hash_cells_mutex); table = dm_get_table(hc->md); if (table) { @@ -321,7 +330,9 @@ static int dm_hash_rename(uint32_t cookie, const char *old, const char *new) */ list_del(&hc->name_list); old_name = hc->name; + mutex_lock(&dm_hash_cells_mutex); hc->name = new_name; + mutex_unlock(&dm_hash_cells_mutex); list_add(&hc->name_list, _name_buckets + hash_str(new_name)); /* @@ -1582,8 +1593,7 @@ int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid) if (!md) return -ENXIO; - dm_get(md); - down_read(&_hash_lock); + mutex_lock(&dm_hash_cells_mutex); hc = dm_get_mdptr(md); if (!hc || hc->md != md) { r = -ENXIO; @@ -1596,8 +1606,7 @@ int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid) strcpy(uuid, hc->uuid ? : ""); out: - up_read(&_hash_lock); - dm_put(md); + mutex_unlock(&dm_hash_cells_mutex); return r; } diff --git a/drivers/md/dm-uevent.c b/drivers/md/dm-uevent.c index 6f65883aef12..c7c555a8c7b2 100644 --- a/drivers/md/dm-uevent.c +++ b/drivers/md/dm-uevent.c @@ -139,14 +139,13 @@ void dm_send_uevents(struct list_head *events, struct kobject *kobj) list_del_init(&event->elist); /* - * Need to call dm_copy_name_and_uuid from here for now. - * Context of previous var adds and locking used for - * hash_cell not compatable. + * When a device is being removed this copy fails and we + * discard these unsent events. */ if (dm_copy_name_and_uuid(event->md, event->name, event->uuid)) { - DMERR("%s: dm_copy_name_and_uuid() failed", - __func__); + DMINFO("%s: skipping sending uevent for lost device", + __func__); goto uevent_free; } -- cgit v1.2.3