Re: [PATCH] usb-serial regression fix

[Jim Radford]

On Mon, Mar 12, 2007 at 09:55:14PM -0400, Mark Lord wrote:
> Jim Radford wrote:
> >On Mon, Mar 12, 2007 at 05:18:19PM -0700, Greg KH wrote:
> >>On Mon, Mar 12, 2007 at 03:59:22PM -0700, Jim Radford wrote:
> >>>On Mon, Mar 12, 2007 at 03:42:35PM -0700, Jim Radford wrote:
> >>>>On Mon, Mar 12, 2007 at 01:33:31PM -0700, Greg KH wrote:
> >>>>>On Mon, Mar 12, 2007 at 04:22:22PM -0400, Mark Lord wrote:
> >>>>>>Oliver Neukum wrote:
> >>>>>>>>Mark Lord wrote:
> >>>>>>>>>Okay, from that part (above), the problem is obvious:
> >>>>>>>>>in that the "MCT U232 converter now disconnected"
> >>>>>>>>>appears, and then we continue to try and call the
> >>>>>>>>>driver's method.. Oops!

> >>>>>>>IMHO shutdown() is using serial->port[] and bombs.
> >>>>>>>Could you reverse the order here?

> >>>Do not NULL serial->port[i] since it is used in ->shutdown().
> >>>This wasn't an issue until the order or ->shutdown() and
> >>>device_unregister was corrected.

> >>>        for (i = 0; i < serial->num_ports; ++i)
> >>>                if (serial->port[i]->dev.parent != NULL) {
> >>>                        device_unregister(&serial->port[i]->dev);
> >>>-                       serial->port[i] = NULL;
> >>>                }

> >>But shouldn't you null it out somewhere?  It will be an "empty"
> >>pointer at some point in time...

> >Not as far as I can see. The serial structure that ->port[i] is in
> >gets kfree()ed soon after, in the same function, and nothing in
> >between, other than ->shutdown(), uses ->port[].  I assume it was
> >someone being overly cautious.

> So where does the memory get freed -- the structure pointed at
> by the serial->port[i] thingie ?  It's not a leak, is it?

It gets free'd through device_unregister

    for (i = 0; i < num_ports; ++i) {
       ...
       port->dev.release = &port_release;
       ...
       retval = device_register(&port->dev);

which means that until all the drivers get converted to use
->port_probe() and ->port_remove() (which gets called by
device_unregister) and stop using the ->port[] array in ->shutdown()
we need to have ->shutdown() called before device_unregister.

> > > > Look at changeset d9a7ecacac5f8274d2afce09aadcf37bdb42b93a in Linus's
> > > > tree from Jim Radford:

> > > > http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=d9a7ecacac5f8274d2afce09aadcf37bdb42b93a

So, this patch should be reverted for now.

-Jim
-
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/