jffs2: Fix lock acquisition order bug in jffs2_write_begin
authorThomas Betker <thomas.betker@freenet.de>
Wed, 17 Oct 2012 20:59:30 +0000 (22:59 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 3 Dec 2012 19:47:06 +0000 (11:47 -0800)
commit262a70e8fbf870359ced2059becf116a249bb2d5
tree7d99a8401d476dcb650373efed0f5a23e8059801
parent11371ec0a92756cbaa84b86fbbaa3aeae7457658
jffs2: Fix lock acquisition order bug in jffs2_write_begin

commit 5ffd3412ae5536a4c57469cb8ea31887121dcb2e upstream.

jffs2_write_begin() first acquires the page lock, then f->sem. This
causes an AB-BA deadlock with jffs2_garbage_collect_live(), which first
acquires f->sem, then the page lock:

jffs2_garbage_collect_live
    mutex_lock(&f->sem)                         (A)
    jffs2_garbage_collect_dnode
        jffs2_gc_fetch_page
            read_cache_page_async
                do_read_cache_page
                    lock_page(page)             (B)

jffs2_write_begin
    grab_cache_page_write_begin
        find_lock_page
            lock_page(page)                     (B)
    mutex_lock(&f->sem)                         (A)

We fix this by restructuring jffs2_write_begin() to take f->sem before
the page lock. However, we make sure that f->sem is not held when
calling jffs2_reserve_space(), as this is not permitted by the locking
rules.

The deadlock above was observed multiple times on an SoC with a dual
ARMv7 (Cortex-A9), running the long-term 3.4.11 kernel; it occurred
when using scp to copy files from a host system to the ARM target
system. The fix was heavily tested on the same target system.

Signed-off-by: Thomas Betker <thomas.betker@rohde-schwarz.com>
Acked-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/jffs2/file.c