sio / puc wedging on both -current and -stable
John Baldwin
jhb at FreeBSD.org
Mon Dec 27 14:11:13 PST 2004
On Wednesday 08 December 2004 02:52 am, Bruce Evans wrote:
> Long ago, On Thu, 3 Jun 2004, Mike Tancsa wrote:
> > Just a quick recap. I can fairly easily trigger an interrupt storm on
> > these machines with USB enabled in the BIOS. If I disable it, I dont
> > have a problem and all works well.... However, what I accidently came
> > across today, was that if I load the USB drivers as a kld, I can *not*
> > wedge the machine. Note the bottom of the following diff
>
> I can now explain this. When usb is in the kernel proper, it normally
> gets the interrupt first and exposes bugs in sio (see other mail --
> there is a conflict but sio ignores the error). When usb is in a
> module, sio normally gets the interrupt first. There is again a
> conflict but usb doesn't ignore the error.
>
> > diff dmesg.kld dmesg.static
> >
> > < uhci2: <Intel 82801DB (ICH4) USB controller USB-C> port 0xb400-0xb41f
> > irq 12 at device 29.2 on pci0
> > < uhci2: Could not allocate irq
> > < device_probe_and_attach: uhci2 attach returned 6
> > < uhci2: <Intel 82801DB (ICH4) USB controller USB-C> port 0xb400-0xb41f
> > irq 12 at device 29.2 on pci0
> > < uhci2: Could not allocate irq
> > < device_probe_and_attach: uhci2 attach returned 6
>
> This shows uhci2 not ignoring the error.
I'd like to force sio(4) devices attached via PCI to just always share the
interrupt and not use INTR_FAST for now. That would allow sio and usb to
play well in both situations.
--
John Baldwin <jhb at FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve" = http://www.FreeBSD.org
More information about the freebsd-current
mailing list