linux/fs
Arjan van de Ven 5b6509aa8c [PATCH] inotify: fix deadlock found by lockdep
This is a real deadlock, a nice complex one:
(warning: long explanation follows so that Andrew can have a complete
patch description)

it's an ABCDA deadlock:

A iprune_mutex
B inode->inotify_mutex
C ih->mutex
D dev->ev_mutex

The AB relationship comes straight from invalidate_inodes()

int invalidate_inodes(struct super_block * sb)
{
        int busy;
        LIST_HEAD(throw_away);

        mutex_lock(&iprune_mutex);
        spin_lock(&inode_lock);
        inotify_unmount_inodes(&sb->s_inodes);

where inotify_umount_inodes() takes the
                mutex_lock(&inode->inotify_mutex);

The BC relationship comes directly from inotify_find_update_watch():
s32 inotify_find_update_watch(struct inotify_handle *ih, struct inode *inode,
                              u32 mask)
{
   ...
        mutex_lock(&inode->inotify_mutex);
        mutex_lock(&ih->mutex);

The CD relationship comes from inotify_rm_wd:
inotify_rm_wd does
        mutex_lock(&inode->inotify_mutex);
        mutex_lock(&ih->mutex)
and then calls inotify_remove_watch_locked() which calls
notify_dev_queue_event() which does
	        mutex_lock(&dev->ev_mutex);

(this strictly is a BCD relationship)

The DA relationship comes from the most interesting part:

  [<ffffffff8022d9f2>] shrink_icache_memory+0x42/0x270
  [<ffffffff80240dc4>] shrink_slab+0x11d/0x1c9
  [<ffffffff802b5104>] try_to_free_pages+0x187/0x244
  [<ffffffff8020efed>] __alloc_pages+0x1cd/0x2e0
  [<ffffffff8025e1f8>] cache_alloc_refill+0x3f8/0x821
  [<ffffffff8020a5e5>] kmem_cache_alloc+0x85/0xcb
  [<ffffffff802db027>] kernel_event+0x2e/0x122
  [<ffffffff8021d61c>] inotify_dev_queue_event+0xcc/0x140

inotify_dev_queue_event schedules a kernel_event which does a
kmem_cache_alloc( , GFP_KERNEL) which may try to shrink slabs, including
the inode cache .. which then takes iprune_mutex.

And voila, there is an AB, a BC, a CD relationship (even a direct BCD),
and also now a DA relationship -> a circular type AB-BA deadlock but
involving 4 locks.

The solution is simple: kernel_event() is NOT allowed to use GFP_KERNEL,
but must use GFP_NOFS to not cause recursion into the VFS.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Robert Love <rml@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-07-31 13:28:41 -07:00
..
9p Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
adfs Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
affs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
afs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
autofs [PATCH] VFS: Permit filesystem to override root dentry on mount 2006-06-23 07:42:45 -07:00
autofs4 [PATCH] fs: use list_move() 2006-06-26 09:58:18 -07:00
befs typo fixes: aquire -> acquire 2006-06-30 18:23:04 +02:00
bfs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
cifs Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
coda Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
configfs configfs: Clear up a few extra spaces where there should be TABs. 2006-06-29 14:43:01 -07:00
cramfs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
debugfs Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
devpts [PATCH] VFS: Permit filesystem to override root dentry on mount 2006-06-23 07:42:45 -07:00
efs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
exportfs [PATCH] NFS server subtree_check returns dubious value 2006-05-21 12:59:16 -07:00
ext2 [PATCH] lockdep: annotate the quota code 2006-07-03 15:27:08 -07:00
ext3 [PATCH] ext3: avoid triggering ext3_error on bad NFS file handle 2006-07-31 13:28:36 -07:00
fat [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
freevxfs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
fuse [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
hfs Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
hfsplus Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
hostfs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
hpfs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
hppfs [PATCH] VFS: Permit filesystem to perform statfs with a known root dentry 2006-06-23 07:42:45 -07:00
hugetlbfs [PATCH] mmap zero-length hugetlb file with PROT_NONE to protect a hugetlb virtual area 2006-07-10 13:24:21 -07:00
isofs Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
jbd [PATCH] add poison.h and patch primary users 2006-06-27 17:32:38 -07:00
jffs Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
jffs2 Merge git://git.infradead.org/mtd-2.6 2006-07-03 21:29:08 -07:00
jfs Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/shaggy/jfs-2.6 2006-07-15 14:43:30 -07:00
lockd NLM,NFSv4: Wait on local locks before we put RPC calls on the wire 2006-07-05 13:13:18 -04:00
minix [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
msdos [PATCH] VFS: Permit filesystem to override root dentry on mount 2006-06-23 07:42:45 -07:00
ncpfs Merge git://git.kernel.org/pub/scm/linux/kernel/git/bunk/trivial 2006-06-30 15:39:30 -07:00
nfs Merge branch 'locks' 2006-07-05 13:19:25 -04:00
nfs_common
nfsd [PATCH] knfsd: Fix stale file handle problem with subtree_checking. 2006-07-31 13:28:37 -07:00
nls Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
ntfs [PATCH] lockdep: annotate NTFS locking rules 2006-07-03 15:27:08 -07:00
ocfs2 ocfs2: remove redundant NULL checks in ocfs2_direct_IO_get_blocks() 2006-06-29 16:13:35 -07:00
openpromfs [OPENPROMFS]: Rewrite using in-kernel device tree and seq_file. 2006-06-25 23:19:14 -07:00
partitions [PATCH] enable mac partition label per default on pmac 2006-07-31 13:28:41 -07:00
proc Don't allow chmod() on the /proc/<pid>/ files 2006-07-15 12:26:45 -07:00
qnx4 Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
ramfs [PATCH] NOMMU: Fix execution off of ramfs with mmap() 2006-07-10 13:24:21 -07:00
reiserfs [PATCH] reiserfs: fix handling of device names with /'s in them 2006-07-14 21:53:54 -07:00
romfs [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
smbfs Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
sysfs [PATCH] lockdep: annotate the sysfs i_mutex to be a separate class 2006-07-12 12:52:54 -07:00
sysv [PATCH] mark address_space_operations const 2006-06-28 14:59:04 -07:00
udf Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
ufs [PATCH] lockdep: annotate the quota code 2006-07-03 15:27:08 -07:00
vfat [PATCH] VFS: Permit filesystem to override root dentry on mount 2006-06-23 07:42:45 -07:00
xfs [XFS] Ensure bulkstat from an invalid inode number gets caught always with 2006-07-28 17:05:51 +10:00
aio.c spelling fixes 2006-06-26 18:35:02 +02:00
attr.c [PATCH] capable/capability.h (fs/) 2006-01-11 18:42:13 -08:00
bad_inode.c [PATCH] Make most file operations structs in fs/ const 2006-03-28 09:16:06 -08:00
binfmt_aout.c [PATCH] dump_thread() cleanup 2006-01-10 08:01:25 -08:00
binfmt_elf_fdpic.c [PATCH] FDPIC: Add coredump capability for the ELF-FDPIC binfmt 2006-07-10 13:24:22 -07:00
binfmt_elf.c [PATCH] FDPIC: Move roundup() into linux/kernel.h 2006-07-10 13:24:22 -07:00
binfmt_em86.c
binfmt_flat.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
binfmt_misc.c Merge branch 'master' of /home/trondmy/kernel/linux-2.6/ 2006-06-24 13:07:53 -04:00
binfmt_script.c
binfmt_som.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
bio.c [PATCH] Fix missing ret assignment in __bio_map_user() error path 2006-06-17 10:52:12 -07:00
block_dev.c [PATCH] lockdep: annotate blkdev nesting 2006-07-03 15:27:10 -07:00
buffer.c [PATCH] invalidate_bdev() speedup 2006-07-31 13:28:38 -07:00
char_dev.c [PATCH] add function documentation for register_chrdev() 2006-07-14 21:53:54 -07:00
compat_ioctl.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
compat.c [PATCH] x86_64: Add compat_printk and sysctl to turn off compat layer warnings 2006-06-26 10:48:16 -07:00
dcache.c [PATCH] lockdep: annotate dcache 2006-07-03 15:27:06 -07:00
dcookies.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
direct-io.c [PATCH] lockdep: annotate direct io 2006-07-03 15:27:06 -07:00
dnotify.c [PATCH] Use __read_mostly on some hot fs variables 2006-03-26 08:56:56 -08:00
dquot.c [PATCH] use list_add_tail() instead of list_add() 2006-06-26 09:58:17 -07:00
drop_caches.c [PATCH] drop-pagecache 2006-01-08 20:12:40 -08:00
eventpoll.c [PATCH] sched: cleanup, remove task_t, convert to struct task_struct 2006-07-03 15:27:11 -07:00
exec.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
fcntl.c BUG_ON() Conversion in fs/fcntl.c 2006-04-02 13:37:19 +02:00
fifo.c [PATCH] pipe.c/fifo.c code cleanups 2006-04-11 13:53:33 +02:00
file_table.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
file.c [PATCH] alloc_fdtable() expansion fix 2006-07-12 12:52:55 -07:00
filesystems.c
fs-writeback.c [PATCH] zoned vm counters: conversion of nr_unstable to per zone counter 2006-06-30 11:25:36 -07:00
inode.c Merge git://git.kernel.org/pub/scm/linux/kernel/git/bunk/trivial 2006-06-30 15:39:30 -07:00
inotify_user.c [PATCH] inotify: fix deadlock found by lockdep 2006-07-31 13:28:41 -07:00
inotify.c [PATCH] inotify (4/5): allow watch removal from event handler 2006-06-20 05:25:19 -04:00
ioctl.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
ioprio.c [PATCH] SELinux: Add security hook definition for getioprio and insert hooks 2006-06-30 11:25:37 -07:00
Kconfig Merge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6 2006-07-13 16:38:58 -07:00
Kconfig.binfmt [PATCH] frv: suppress configuration of certain features for FRV 2006-01-08 20:13:36 -08:00
libfs.c [PATCH] core: use list_move() 2006-06-26 09:58:17 -07:00
locks.c VFS: Add support for the FL_ACCESS flag to flock_lock_file() 2006-07-05 13:13:18 -04:00
Makefile [PATCH] devfs: Remove devfs from the kernel tree 2006-06-26 12:25:05 -07:00
mbcache.c [PATCH] Typo fixes 2006-03-28 09:16:08 -08:00
mpage.c [PATCH] writeback: fix range handling 2006-06-23 07:42:49 -07:00
namei.c [PATCH] struct file leakage 2006-07-14 21:53:50 -07:00
namespace.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
nfsctl.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
open.c [PATCH] ftruncate does not always update m/ctime 2006-06-25 10:01:15 -07:00
pipe.c [PATCH] VFS: Permit filesystem to override root dentry on mount 2006-06-23 07:42:45 -07:00
pnode.c [PATCH] core: use list_move() 2006-06-26 09:58:17 -07:00
pnode.h
posix_acl.c
quota_v1.c
quota_v2.c [PATCH] sem2mutex: quota 2006-03-23 07:38:11 -08:00
quota.c [PATCH] sem2mutex: quota 2006-03-23 07:38:11 -08:00
read_write.c [PATCH] fs/read_write.c: EXPORT_UNUSED_SYMBOL 2006-07-10 13:24:18 -07:00
readdir.c [PATCH] mutex subsystem, semaphore to mutex: VFS, ->i_sem 2006-01-09 15:59:24 -08:00
select.c [PATCH] fs: sys_poll with timeout -1 bug fix 2006-06-25 10:01:22 -07:00
seq_file.c [PATCH] sem2mutex: fs/seq_file.c 2006-03-23 07:38:12 -08:00
splice.c [PATCH] splice: fix problems with sys_tee() 2006-07-10 11:00:01 +02:00
stat.c Remove obsolete #include <linux/config.h> 2006-06-30 19:25:36 +02:00
super.c [PATCH] lockdep: annotate sb ->s_umount 2006-07-03 15:27:09 -07:00
sync.c [PATCH] writeback: fix range handling 2006-06-23 07:42:49 -07:00
xattr_acl.c
xattr.c [PATCH] log more info for directory entry change events 2006-06-20 05:25:28 -04:00