possible locking bug in tty_open
Arnd Bergmann
maintainerWhile playing some more with my TTY BKL patches, I stumbled over what looks like a bug in tty_open, introduced in e8c6210 "tty: push the BKL down into the handlers a bit":
After the "retry_open:" label, we first get the tty_mutex and then the BKL. However a the end of tty_open, we jump back to retry_open with the BKL still held. If we run into this case, the tty_open function will be left with the BKL still held.
It may be impossible to actually trigger this bug, because the path is only taken if a tty driver open function returns -ERESTARTSYS without setting signal_pending().
Arnd
Quoted 10 lines · On Sun, 2 May 2010 22:47:33 +0200 Arnd Bergmann <[email protected]> wrote:
While playing some more with my TTY BKL patches, I stumbled over what looks like a bug in
tty_open, introduced in e8c6210 "tty: push the BKL down into the handlers a bit":After the "
retry_open:" label, we first get thetty_mutexand then the BKL. However a the end oftty_open, we jump back toretry_openwith the BKL still held. If we run into this case, thetty_openfunction will be left with the BKL still held.
Quoted 3 lines
It may be impossible to actually trigger this bug, because the path is only taken if a tty driver open function returns
-ERESTARTSYSwithout settingsignal_pending().
It looks pretty much impossible to trigger but it's definitely a bug. I'll send Greg a patch
Quoted 10 lines · On Sun, 2 May 2010 22:47:33 +0200 Arnd Bergmann <[email protected]> wrote:
While playing some more with my TTY BKL patches, I stumbled over what looks like a bug in
tty_open, introduced in e8c6210 "tty: push the BKL down into the handlers a bit":After the "
retry_open:" label, we first get thetty_mutexand then the BKL. However a the end oftty_open, we jump back toretry_openwith the BKL still held. If we run into this case, thetty_openfunction will be left with the BKL still held.
Quoted 5 lines
It may be impossible to actually trigger this bug, because the path is only taken if a tty driver open function returns
-ERESTARTSYSwithout settingsignal_pending().Arnd
I think all we need is probably this
tty: Fix unbalanced BKL handling in error path
Arnd noted:
After the "retry_open:" label, we first get the tty_mutex and then the BKL. However a the end of tty_open, we jump back to retry_open with the BKL still held. If we run into this case, the tty_open function will be left with the BKL still held.
drivers/char/tty_io.c +1 -0
@@ -1875,6 +1875,7 @@ got_driver:*/ if (filp->f_op == &hung_up_tty_fops) filp->f_op = &tty_fops; +unlock_kernel (); goto retry_open; }unlock_kernel ();
Quoted 13 lines · On Tuesday 04 May 2010 21:42:36 Alan Cox wrote:
Quoted 5 lines · On Sun, 2 May 2010 22:47:33 +0200 Arnd Bergmann <[email protected]> wrote:
After the "
retry_open:" label, we first get thetty_mutexand then the BKL. However a the end oftty_open, we jump back toretry_openwith the BKL still held. If we run into this case, thetty_openfunction will be left with the BKL still held.I think all we need is probably this
tty: Fix unbalanced BKL handling in error path
Quoted 7 lines
Arnd noted:
After the "
retry_open:" label, we first get thetty_mutexand then the BKL. However a the end oftty_open, we jump back toretry_openwith the BKL still held. If we run into this case, thetty_openfunction will be left with the BKL still held.
Yes, looks good.
Quoted 2 lines
Signed-off-by Alan Cox <[email protected]>