Re: port-i386/46808: 6.0_BETA2 install crashes on VIA VT6202 detection

["jc.lynx" <jc.lynx%laposte.net@localhost>]

OK, I set maxncomp to 1 since it seems to be this value on my other NetBSD machine : 1 companion which has 4 ports. Anyway, better initializing not enough peripherals than trying to initialize non-existing ones, I think.
It didn't get very far, here is the end of dmesg :

maxncomp = 15
EREAD4(&sc->sc, EHCI_HCSPARAMS) = ffffffff
forcing maxncomp to 1
panic : kernel diagnostic assertion "(reg & 0x3) == 0" failed: file "/usr/sys/arch/x86/pci/pci_machdep.c", line 476

 fatal breakpoint (...)

The function it crashes in is pci_conf_read(pci_chipset_tag_t pc, pcitag_t tag, int reg)

I haven't inspected this further yet.

> > Message du 26/08/12 09:19
> > De : "Christos Zoulas" 
> > A : "jc.lynx" , gnats-bugs%NetBSD.org@localhost, port-i386-maintainer%netbsd.org@localhost, gnats-admin%netbsd.org@localhost, netbsd-bugs%netbsd.org@localhost
> > Copie à : 
> > Objet : Re: port-i386/46808: 6.0_BETA2 install crashes on VIA VT6202 detection
> >
> > On Aug 25, 11:21pm, jc.lynx%laposte.net@localhost ("jc.lynx") wrote:
> > -- Subject: Re: port-i386/46808: 6.0_BETA2 install crashes on VIA VT6202 dete
> > 
> > | OK. I've installed NetBSD on another machine and done the following alterat=
> > | ions to ehci_pci.c
> > | 
> > | Replaced :
> > | 
> > | const u_int maxncomp =3D EHCI_HCS_N_CC(EREAD4(&sc->sc, EHCI_HCSPARAMS));
> > | 
> > | with :
> > | 
> > | const u_int vovo =3D EREAD4(&sc->sc, EHCI_HCSPARAMS);
> > | const u_int maxncomp =3D EHCI_HCS_N_CC(vovo);
> > | printf("maxncomp =3D %u\n", maxncomp);
> > | printf("EREAD4(&sc->sc, EHCI_HCSPARAMS) =3D %x, vovo);
> > | 
> > | I hope I got the types right.
> > | Anyway, I burned and booted boot.iso and got this :
> > | 
> > | maxncomp =3D 15
> > | EREAD4(&sc->sc, EHCI_HCSPARAMS) =3D ffffffff
> > | 
> > | So, yeah, EHCI_HCS_N_CC(ffffffff), i.e. (ffffffff >> 12) & 0xf is 15, indee=
> > | d greater than EHCI_COMPANION_MAX which is 8 ; but I guess that this ffffff=
> > |  is not the data we were looking for anyway.
> > | I could not find code related to "special" VT6202 handling in other BSD's c=
> > | ode, except an unrelated issue with EHCI polling frequency.
> > | I suppose we now have to dig into how sc was initialised, or whether EREAD4=
> > | (&sc->sc, EHCI_HCS_PARAMS) is the way to get what we were looking for on th=
> > | is particular device.
> > 
> > Thsat it a good way to debug it, but you can just try to clamp it to
> > EHCI_COMPANION_MAX after you print a warning and see if it gets you
> > further. My guess is that something else is wrong and the device is
> > not mapped or initialized properly. Perhaps there is something in
> > the bios you can change?
> > 
> > christos
> >