From: Daniel Ritz <daniel.ritz@gmx.ch>
- thread_done should only be completed when the wait_queue is installed.
- all wake up conditions should be checked before schedule()
this fixes a hang of rmmod in the sequence modprobe yenta_socket; rmmod
yenta_socket as reported by Andreas Steinmetz. w/o this rmmod yenta_socket
can hang on wait_for_completion() in pcmcia_unregister_socket()
Signed-off-by: Daniel Ritz <daniel.ritz@gmx.ch>
Cc: Dominik Brodowski <linux@brodo.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---
drivers/pcmcia/cs.c | 9 +++++----
1 files changed, 5 insertions(+), 4 deletions(-)
diff -puN drivers/pcmcia/cs.c~pcmcia-cs-fix-possible-missed-wakeup drivers/pcmcia/cs.c
--- 25/drivers/pcmcia/cs.c~pcmcia-cs-fix-possible-missed-wakeup Wed Aug 17 16:06:45 2005
+++ 25-akpm/drivers/pcmcia/cs.c Wed Aug 17 16:06:45 2005
@@ -654,9 +654,10 @@ static int pccardd(void *__skt)
skt->thread = NULL;
complete_and_exit(&skt->thread_done, 0);
}
- complete(&skt->thread_done);
add_wait_queue(&skt->thread_wait, &wait);
+ complete(&skt->thread_done);
+
for (;;) {
unsigned long flags;
unsigned int events;
@@ -682,11 +683,11 @@ static int pccardd(void *__skt)
continue;
}
- schedule();
- try_to_freeze();
-
if (!skt->thread)
break;
+
+ schedule();
+ try_to_freeze();
}
remove_wait_queue(&skt->thread_wait, &wait);
_