]> pilppa.com Git - linux-2.6-omap-h63xx.git/commit
fix serial buffer memory leak
authorAlan Cox <alan@lxorguk.ukuu.org.uk>
Fri, 10 Aug 2007 20:01:05 +0000 (13:01 -0700)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Sat, 11 Aug 2007 22:47:41 +0000 (15:47 -0700)
commit42fd552e8647316757ded0176466c41d17934dcf
tree07b0f4ed1934c05c6495c62ba943ab90bde6be84
parentf8a745942b1b7f052cb76bb8a893d12cb6329c84
fix serial buffer memory leak

Patch c5c34d4862e18ef07c1276d233507f540fb5a532 (tty: flush flip buffer on
ldisc input queue flush) introduces a race condition which can lead to memory
leaks.

The problem can be triggered when tcflush() is called when data are being
pushed to the line discipline driver by flush_to_ldisc().

flush_to_ldisc() releases tty->buf.lock when calling the line discipline
receive_buf function. At that poing tty_buffer_flush() kicks in and sets both
tty->buf.head and tty->buf.tail to NULL. When flush_to_ldisc() finishes, it
restores tty->buf.head but doesn't touch tty->buf.tail. This corrups the
buffer queue, and the next call to tty_buffer_request_room() will allocate a
new buffer and overwrite tty->buf.head. The previous buffer is then lost
forever without being released.

(Thanks to Laurent for the above text, for finding, disgnosing and reporting
the bug)

- Use tty->flags bits for the flush status.

- Wait for the flag to clear again before returning

- Fix the doc error noted

- Fix flush of empty queue leaving stale flushpending

[akpm@linux-foundation.org: cleanup]
Signed-off-by: Alan Cox <alan@redhat.com>
Acked-by: Paul Fulghum <paulkf@microgate.com>
Cc: Laurent Pinchart <laurentp@cse-semaphore.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
drivers/char/tty_io.c
include/linux/tty.h