An even more convincing (to me :-) data point:
- 2.4.0-test1-ac7 on that laptop got OOPSen as below.
- Adding just this patch made them go away.
So far, all the evidence is strongly in favor of this
going into the 2.4 kernel ASAP.
Cyrille, thanks!
- Dave
----- Original Message -----
From: "David Brownell" <[EMAIL PROTECTED]>
To: "Cyrille Chepelov (home)" <[EMAIL PROTECTED]>; <[EMAIL PROTECTED]>
Cc: <[EMAIL PROTECTED]>
Sent: Saturday, 3 June, 2000 11:06 AM
Subject: Re: [linux-usb] (OHCI) nasty bug in sohci_unlink_urb() !
> One data point:
>
> - 2.4.0-test1 on my laptop got OOPSen in OHCI if I sneezed.
> (As in, simple usbdevfs calls to look at devices Oopsed.)
>
> - 2.4.0-test1-ac7 with this patch and Johannes' hub patch
> works reliably, like pre9 or so (light testing, though).
>
> So far, so good; but I've not tried ac7 ohci without this patch.
> And it does look like it'd affect code that doesn't use async
> unlink.
>
> Roman, do you know of anything else that'd have made the OHCI
> driver become unstable?
>
> - Dave
>
>
> ----- Original Message -----
> From: "Cyrille Chepelov (home)" <[EMAIL PROTECTED]>
> To: <[EMAIL PROTECTED]>
> Cc: <[EMAIL PROTECTED]>
> Sent: Thursday, 1 June, 2000 2:25 PM
> Subject: [linux-usb] (OHCI) nasty bug in sohci_unlink_urb() !
>
>
> >
> > This patchlet fixes a BUG() when using async unlink on OHCI. Basically, an
> > error in refcounting led to a schedule()-hungry sohci_free_dev call, even
> > when in_interrupt().
> >
> > It's been against 2.4.0-test1-ac7, which includes Roman's last
> > visible patch (22/5).
> >
> > -- Cyrille
> >
> > ------------------------------------------------------------------------------
> > Grumpf.
> >
> >
>
>
> --------------------------------------------------------------------------------
>
>
> > --- drivers/usb/usb-ohci.romans.c Thu Jun 1 22:12:35 2000
> > +++ drivers/usb/usb-ohci.c Thu Jun 1 22:19:18 2000
> > @@ -527,12 +527,12 @@
> > #ifdef DEBUG
> > urb_print (urb, "UNLINK", 1);
> > #endif
> > -
> > - usb_dec_dev_use (urb->dev);
> >
> > - if (usb_pipedevice (urb->pipe) == ohci->rh.devnum)
> > + if (usb_pipedevice (urb->pipe) == ohci->rh.devnum) {
> > + usb_dec_dev_use(urb->dev);
> > return rh_unlink_urb (urb); /* a request to the virtual root hub */
> > -
> > + }
> > +
> > if (urb->hcpriv) {
> > /* URB active? */
> > if (urb->status == USB_ST_URB_PENDING && !ohci->disabled) {
> > @@ -548,15 +548,19 @@
> > urb_priv->ed->state |= ED_URB_DEL;
> > spin_unlock_irqrestore (&usb_ed_lock, flags);
> > if (!(urb->transfer_flags & USB_ASYNC_UNLINK)) {
> > + usb_dec_dev_use (urb->dev);
> > add_wait_queue (&op_wakeup, &wait);
> > current->state = TASK_UNINTERRUPTIBLE;
> > if (!schedule_timeout (HZ / 10)) /* wait until all TDs are deleted */
> > err("unlink URB timeout!");
> > remove_wait_queue (&op_wakeup, &wait);
> > urb->status = -ENOENT;
> > - } else
> > + } else {
> > + /* usb_dec_dev_use done in dl_del_list() */
> > urb->status = -EINPROGRESS;
> > + }
> > } else {
> > + usb_dec_dev_use (urb->dev);
> > urb_rm_priv (urb);
> > if (urb->complete && (urb->transfer_flags & USB_ASYNC_UNLINK)) {
> > urb->complete (urb);
> >
>
>
> --------------------------------------------------------------------------------
>
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
