lib/llist.c: fix data race in llist_del_first
authorDmitry Vyukov <dvyukov@google.com>
Sat, 7 Nov 2015 00:31:17 +0000 (16:31 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 7 Nov 2015 01:50:42 +0000 (17:50 -0800)
commit2cf12f821cd4f996bfabeec23d8f25e7a2052a28
tree489e2ae74b1640db6732076da04fb854e9167ccb
parent943ba6503802a46318cb9f5ab45be31d42e6f884
lib/llist.c: fix data race in llist_del_first

llist_del_first reads entry->next, but it did not acquire visibility over
the entry node.  As the result it can get a stale value of entry->next
(e.g.  NULL or whatever garbage was there before the appending thread
wrote correct value).  And then commit that value as llist head with
cmpxchg.  That will corrupt llist.

Note there is a control-dependency between read of head->first and read of
entry->next, but it does not make the code correct.  Kernel memory model
unambiguously says: "A load-load control dependency requires a full read
memory barrier".

Use smp_load_acquire to acquire visibility over the entry node.

The data race was found with KernelThreadSanitizer (KTSAN).

Here is an example of KTSAN report:

ThreadSanitizer: data-race in llist_del_first

Read of size 1 by thread T389 (K2630, CPU0):
 [<ffffffff8156b8a9>] llist_del_first+0x39/0x70 lib/llist.c:74
 [<     inlined    >] tty_buffer_alloc drivers/tty/tty_buffer.c:181
 [<ffffffff81664af4>] __tty_buffer_request_room+0xb4/0x250 drivers/tty/tty_buffer.c:292
 [<ffffffff81664e6c>] tty_insert_flip_string_fixed_flag+0x6c/0x150 drivers/tty/tty_buffer.c:337
 [<     inlined    >] tty_insert_flip_string include/linux/tty_flip.h:35
 [<ffffffff81667422>] pty_write+0x72/0xc0 drivers/tty/pty.c:110
 [<     inlined    >] process_output_block drivers/tty/n_tty.c:611
 [<ffffffff8165c016>] n_tty_write+0x346/0x7f0 drivers/tty/n_tty.c:2401
 [<     inlined    >] do_tty_write drivers/tty/tty_io.c:1159
 [<ffffffff816568df>] tty_write+0x21f/0x3f0 drivers/tty/tty_io.c:1245
 [<ffffffff8125f00f>] __vfs_write+0x5f/0x1f0 fs/read_write.c:489
 [<ffffffff8125ff8f>] vfs_write+0xef/0x280 fs/read_write.c:538
 [<     inlined    >] SYSC_write fs/read_write.c:585
 [<ffffffff81261390>] SyS_write+0x70/0xe0 fs/read_write.c:577
 [<ffffffff81ee862e>] entry_SYSCALL_64_fastpath+0x12/0x71 arch/x86/entry/entry_64.S:186

Previous write of size 8 by thread T226 (K761, CPU0):
 [<ffffffff8156b832>] llist_add_batch+0x32/0x70 lib/llist.c:44 (discriminator 16)
 [<     inlined    >] llist_add include/linux/llist.h:180
 [<ffffffff816649fc>] tty_buffer_free+0x6c/0xb0 drivers/tty/tty_buffer.c:221
 [<ffffffff816651e7>] flush_to_ldisc+0x107/0x300 drivers/tty/tty_buffer.c:514
 [<ffffffff810b20ee>] process_one_work+0x47e/0x930 kernel/workqueue.c:2036
 [<ffffffff810b2650>] worker_thread+0xb0/0x900 kernel/workqueue.c:2170
 [<ffffffff810bbe20>] kthread+0x150/0x170 kernel/kthread.c:209
 [<ffffffff81ee8a1f>] ret_from_fork+0x3f/0x70 arch/x86/entry/entry_64.S:526

Signed-off-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Konstantin Serebryany <kcc@google.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Alexander Potapenko <glider@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
lib/llist.c