Prev: [07/11] USB: add missing delay during remote wakeup
Next: clocksource: Prevent potential kgdb dead lock
From: Eric W. Biederman on 26 Jan 2010 15:00
Greg KH <gregkh(a)suse.de> writes:

> 2.6.27-stable review patch. If anyone has any objections, please let us know.

Only that __f_setown by way of f_modown unconditionally enables interrupts. So
without touching f_modown as well in mainline we have nasty sounding lockdep warnings.

Eric


> ------------------
>
> From: Greg Kroah-Hartman <gregkh(a)suse.de>
>
> commit 703625118069f9f8960d356676662d3db5a9d116 upstream.
>
> We need to keep the lock held over the call to __f_setown() to
> prevent a PID race.
>
> Thanks to Al Viro for pointing out the problem, and to Travis for
> making us look here in the first place.
>
> Cc: Eric W. Biederman <ebiederm(a)xmission.com>
> Cc: Al Viro <viro(a)ZenIV.linux.org.uk>
> Cc: Alan Cox <alan(a)lxorguk.ukuu.org.uk>
> Cc: Linus Torvalds <torvalds(a)linux-foundation.org>
> Cc: Tavis Ormandy <taviso(a)google.com>
> Cc: Jeff Dike <jdike(a)addtoit.com>
> Cc: Julien Tinnes <jln(a)google.com>
> Cc: Matt Mackall <mpm(a)selenic.com>
> Signed-off-by: Greg Kroah-Hartman <gregkh(a)suse.de>
>
> ---
> drivers/char/tty_io.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> --- a/drivers/char/tty_io.c
> +++ b/drivers/char/tty_io.c
> @@ -2437,8 +2437,8 @@ static int tty_fasync(int fd, struct fil
> pid = task_pid(current);
> type = PIDTYPE_PID;
> }
> - spin_unlock_irqrestore(&tty->ctrl_lock, flags);
> retval = __f_setown(filp, pid, type, 0);
> + spin_unlock_irqrestore(&tty->ctrl_lock, flags);
> if (retval)
> goto out;
> } else {
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Linus Torvalds on 26 Jan 2010 17:20


On Tue, 26 Jan 2010, Eric W. Biederman wrote:

> Greg KH <gregkh(a)suse.de> writes:
>
> > 2.6.27-stable review patch. If anyone has any objections, please let us know.
>
> Only that __f_setown by way of f_modown unconditionally enables interrupts. So
> without touching f_modown as well in mainline we have nasty sounding lockdep warnings.

Hmm. That seems to be true in mainline too, isn't it?

So now we have:
- tty_fasync() gets tty->ctrl_lock, with spin_lock_irqsave()

- it then calls __f_setown()

- which calls f_modown(),

- which does

write_lock_irq(&filp->f_owner.lock);
..
write_unlock_irq(&filp->f_owner.lock);

which in turn enables interrupts while we still hold ctrl_lock.

so that whole commit 70362511806 was/is buggy in mainline too.

The minimal fix is likely to just make f_modown() use
write_lock_irqsave/restore. Greg?

Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Eric W. Biederman on 26 Jan 2010 18:10
Linus Torvalds <torvalds(a)linux-foundation.org> writes:

> On Tue, 26 Jan 2010, Eric W. Biederman wrote:
>
>> Greg KH <gregkh(a)suse.de> writes:
>>
>> > 2.6.27-stable review patch. If anyone has any objections, please let us know.
>>
>> Only that __f_setown by way of f_modown unconditionally enables interrupts. So
>> without touching f_modown as well in mainline we have nasty sounding lockdep warnings.
>
> Hmm. That seems to be true in mainline too, isn't it?

Yes. We are having that conversation in the thread:
"[2.6.33-rc5] starting emacs makes lockdep warning"

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Greg KH on 26 Jan 2010 18:10
On Tue, Jan 26, 2010 at 02:11:28PM -0800, Linus Torvalds wrote:
>
>
> On Tue, 26 Jan 2010, Eric W. Biederman wrote:
>
> > Greg KH <gregkh(a)suse.de> writes:
> >
> > > 2.6.27-stable review patch. If anyone has any objections, please let us know.
> >
> > Only that __f_setown by way of f_modown unconditionally enables interrupts. So
> > without touching f_modown as well in mainline we have nasty sounding lockdep warnings.
>
> Hmm. That seems to be true in mainline too, isn't it?
>
> So now we have:
> - tty_fasync() gets tty->ctrl_lock, with spin_lock_irqsave()
>
> - it then calls __f_setown()
>
> - which calls f_modown(),
>
> - which does
>
> write_lock_irq(&filp->f_owner.lock);
> ..
> write_unlock_irq(&filp->f_owner.lock);
>
> which in turn enables interrupts while we still hold ctrl_lock.
>
> so that whole commit 70362511806 was/is buggy in mainline too.
>
> The minimal fix is likely to just make f_modown() use
> write_lock_irqsave/restore. Greg?

Yes, that looks correct.

Here's a patch that does just that:
---------

From: Greg Kroah-Hartman <gregkh(a)suse.de>
Subject: fnctl: f_modown should call write_lock_irqsave/restore

Commit 703625118069f9f8960d356676662d3db5a9d116 exposed that f_modown()
should call write_lock_irqsave instead of just write_lock_irq so that
because a caller could have a spinlock held and it would not be good to
renable interrupts.

Cc: Eric W. Biederman <ebiederm(a)xmission.com>
Cc: Al Viro <viro(a)ZenIV.linux.org.uk>
Cc: Alan Cox <alan(a)lxorguk.ukuu.org.uk>
Cc: Tavis Ormandy <taviso(a)google.com>
Cc: stable <stable(a)kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh(a)suse.de>

diff --git a/fs/fcntl.c b/fs/fcntl.c
index 97e01dc..5ef953e 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -199,7 +199,9 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
int force)
{
- write_lock_irq(&filp->f_owner.lock);
+ unsigned long flags;
+
+ write_lock_irqsave(&filp->f_owner.lock, flags);
if (force || !filp->f_owner.pid) {
put_pid(filp->f_owner.pid);
filp->f_owner.pid = get_pid(pid);
@@ -211,7 +213,7 @@ static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
filp->f_owner.euid = cred->euid;
}
}
- write_unlock_irq(&filp->f_owner.lock);
+ write_unlock_irqrestore(&filp->f_owner.lock, flags);
}

int __f_setown(struct file *filp, struct pid *pid, enum pid_type type,
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Linus Torvalds on 26 Jan 2010 20:40


On Tue, 26 Jan 2010, Greg KH wrote:
> From: Greg Kroah-Hartman <gregkh(a)suse.de>
> Subject: fnctl: f_modown should call write_lock_irqsave/restore

Ok, applied and pushed out as b04da8b, so you can add it to the -stable
list.

Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 |  Next  |  Last
Pages: 1 2
Prev: [07/11] USB: add missing delay during remote wakeup
Next: clocksource: Prevent potential kgdb dead lock