USB4BSD release candidate number 3 - request for review

[Hans Petter Selasky]

On Saturday 08 November 2008, Paul B. Mahol wrote:
> On 11/7/08, Hans Petter Selasky <hselasky at c2i.net> wrote:
> > On Friday 07 November 2008, Paul B. Mahol wrote:
> >> On 11/7/08, M. Warner Losh <imp at bsdimp.com> wrote:
> >> > : > After some time it will appear but will start attaching and
> >> > : > dettaching all the time:
> >
> > How quick is this happening? There has been some changes to the
> > sys/dev/usb2/core/uhub2.c file recently, which control how HUB is
> > responding to events.
>
> ~3 seconds.
>
> I got also this messages:
>
> uhci1: LegSup = 0x2f00
                   ^^^^ Legacy support is complaining about something!

> uhci_interrupt: host controller halted
> uhci_dumpregs:705: usbus1 regs: cmd=0000, sts=0020, intr=0000,
> frnum=0000, flbase=3f7fa000, sof=0040, portsc1=0080, portsc2=0080
> uhci_dump_qh:779: QH(0xc627e900) at 0x05a7e902: h_next=0x05a7db82
> e_next=0x00000001
> uhci_dump_qh:779: QH(0xc627db80) at 0x05a7db82: h_next=0x05a7dc02
> e_next=0x00000001
> uhci_dump_qh:779: QH(0xc627dc00) at 0x05a7dc02: h_next=0x05a7dc82
> e_next=0x00000001
> uhci_dump_qh:779: QH(0xc627dc80) at 0x05a7dc82: h_next=0x00000001
> e_next=0x05a7dd00
> uhci_interrupt: host controller halted
> uhci_dumpregs:705: usbus1 regs: cmd=0000, sts=0020, intr=0000,
> frnum=0000, flbase=3f7fa000, sof=0040, portsc1=0080, portsc2=0080
> uhci_dump_qh:779: QH(0xc627e900) at 0x05a7e902: h_next=0x05a7db82
> e_next=0x00000001
> uhci_dump_qh:779: QH(0xc627db80) at 0x05a7db82: h_next=0x05a7dc02
> e_next=0x00000001
> uhci_dump_qh:779: QH(0xc627dc00) at 0x05a7dc02: h_next=0x05a7dc82
> e_next=0x00000001
> uhci_dump_qh:779: QH(0xc627dc80) at 0x05a7dc82: h_next=0x00000001
> e_next=0x05a7dd00
>
> Doesnt looks like hub is source of fault, one from perforce with
> 152460 comitt is not available
> in CURRENT and with latest version of usb2_hub.c from perforce problem
> is still there.
>
> > Maybe you can get the date of that file from the build where it works and
> > the
> > build where it doesn't work, and I will check the differences.
>
> Last time I tested usb2 from svn was before huge modularization.

Yes, then it probably has something to do with Legacy USB support.

Some suggestions:

0) Have you tried putting the USB modules in the kernel ?

1) Try not loading the EHCI driver. Does the UHCI driver work without the host 
controller halted error?

2) sys/dev/usb2/controller/uhci2_pci.c

Try moving the pci_read/pci_write lines before the IRQ is setup:

#if (__FreeBSD_version >= 700031)
        err = bus_setup_intr(self, sc->sc_irq_res, INTR_TYPE_BIO | 
INTR_MPSAFE,
            NULL, (void *)(void *)uhci_interrupt, sc, &sc->sc_intr_hdl);
#else
        err = bus_setup_intr(self, sc->sc_irq_res, INTR_TYPE_BIO | 
INTR_MPSAFE,
            (void *)(void *)uhci_interrupt, sc, &sc->sc_intr_hdl);
#endif

        if (err) {
                device_printf(self, "Could not setup irq, %d\n", err);
                sc->sc_intr_hdl = NULL;
                goto error;
        }
        /*
         * Set the PIRQD enable bit and switch off all the others. We don't
         * want legacy support to interfere with us XXX Does this also mean
         * that the BIOS won't touch the keyboard anymore if it is connected
         * to the ports of the root hub?
         */
#if USB_DEBUG
        if (pci_read_config(self, PCI_LEGSUP, 2) != PCI_LEGSUP_USBPIRQDEN) {
                device_printf(self, "LegSup = 0x%04x\n",
                    pci_read_config(self, PCI_LEGSUP, 2));
        }
#endif
        pci_write_config(self, PCI_LEGSUP, PCI_LEGSUP_USBPIRQDEN, 2);

3) There has been some changes to the EHCI legacy support code recently. Try 
the same in EHCI: Move "ehci_pci_takecontroller" before the IRQ is setup:

sys/dev/usb2/controller/ehci2_pci.c

#if (__FreeBSD_version >= 700031)
        err = bus_setup_intr(self, sc->sc_irq_res, INTR_TYPE_BIO | 
INTR_MPSAFE,
            NULL, (void *)(void *)ehci_interrupt, sc, &sc->sc_intr_hdl);
#else
        err = bus_setup_intr(self, sc->sc_irq_res, INTR_TYPE_BIO | 
INTR_MPSAFE,
            (void *)(void *)ehci_interrupt, sc, &sc->sc_intr_hdl);
#endif
        if (err) {
                device_printf(self, "Could not setup irq, %d\n", err);
                sc->sc_intr_hdl = NULL;
                goto error;
        }
        ehci_pci_takecontroller(self);

You only need to recompile and reinstall the USB2 modules after these changes.

Try one suggestion at a time so that we can do some elimination.

--HPS