Message ID | 4C2C51E2.7030308@parrot.com |
---|---|
State | New, archived |
Headers | show |
On Thu, 2010-07-01 at 10:29 +0200, Matthieu CASTET wrote: > Hi, > > I know 2.6.27-backport is not supported anymore, but I found a bug, and > I post it for the record. > > The bug happen with a kernel oops [1]. > > After investigation it happens because of programming a timer that is > already programmed (we don't check wbuf->no_timer in new_wbuf_timer_nolock) Thanks, although I do not maintain 2.6.27 port anymore [1], I will of course push your patch a bit later, when I find time to process my MTD input queue :-) [1] http://www.linux-mtd.infradead.org/doc/ubifs.html#L_source
On Thu, 2010-07-01 at 10:29 +0200, Matthieu CASTET wrote: > Hi, > > I know 2.6.27-backport is not supported anymore, but I found a bug, and > I post it for the record. > > The bug happen with a kernel oops [1]. > > After investigation it happens because of programming a timer that is > already programmed (we don't check wbuf->no_timer in new_wbuf_timer_nolock) > Pushed. BTW, if you back ported newer stuff to ubifs-v2.6.27, I can pull it from you and we can make this tree more up-to-date.
diff --git a/fs/ubifs/io.c b/fs/ubifs/io.c index 05471ee..dfbd859 100644 --- a/fs/ubifs/io.c +++ b/fs/ubifs/io.c @@ -313,7 +313,7 @@ static void new_wbuf_timer_nolock(struct ubifs_wbuf *wbuf) { ubifs_assert(!hrtimer_active(&wbuf->timer)); - if (!ktime_to_ns(wbuf->hardlimit)) + if (wbuf->no_timer) return; dbg_io("set timer for jhead %s, %llu millisecs", dbg_jhead(wbuf->jhead),