Re: tty related lockdep trace during bootup on 3.2-rc2

[Cong Wang]

On Wed, Nov 23, 2011 at 3:28 PM, Havard Skinnemoen
<hskinnemoen@xxxxxxxxxx> wrote:
> On Tue, Nov 22, 2011 at 7:38 PM, Dave Jones <davej@xxxxxxxxxx> wrote:
>> From Linus' current tree...
>>
>> related to 5dc2470c602da8851907ec18942cd876c3b4ecc1 maybe ?
>
> Yes, probably. I did have a bad feeling about the locking, but it
> seemed to behave well during testing. Wonder why I didn't see this.
>
> So what's happening is that tty_open() holds big_tty_mutex while
> calling acm_tty_open which takes open_lock, and acm_tty_close holds
> open_lock while calling tty_port_close_start which takes
> big_tty_mutex?

Yes, exactly.

>
> Not sure how to solve this. Not taking the lock before calling
> tty_port_close_start means the tty_port may get freed before it
> returns.

But...

static void acm_tty_hangup(struct tty_struct *tty)
{
        struct acm *acm = tty->driver_data;
        tty_port_hangup(&acm->port); //<---------------- No open_mutex
protected neither...
        mutex_lock(&open_mutex);
        acm_port_down(acm);
        mutex_unlock(&open_mutex);
}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/