Re: sb600/sb700 ohci experimental patch

on 25/09/2009 10:28 Hans Petter Selasky said the following:
> On Friday 25 September 2009 08:34:21 Andriy Gapon wrote:
>> Not sure how to interpret this.
> 
> In ohci_controller_init() try to disable the 
> 
> DPRINTF("SMM active, request owner change\n");
> 
> part of the code for !(ohci_unit == 0) or (ohci_unit == 2) and see what 
> happens.
> 
> Your clue might also indicate that we should request owner change for all 
> OHCI's before resetting any of them. Possibly a BIOS bug!

Haven't tried the suggested changes yet, but here is some more info.
This is register dump of ohci0 just _before_ we start doing anything with it:

ohci0: <OHCI (generic) USB controller> mem 0xfe02e000-0xfe02efff irq 16 at
device 18.0 on pci0
ohci0: [ITHREAD]
ohci_dumpregs:567: ohci_dumpregs: rev=0x00000110 control=0x00000184
command=0x00000000
ohci_dumpregs:571:                intrstat=0x00000024 intre=0xc0000042
intrd=0xc0000042
ohci_dumpregs:575:                hcca=0xbfdf1f00 percur=0x00000000
ctrlhd=0xbfdf1c50
ohci_dumpregs:579:                ctrlcur=0x00000000 bulkhd=0x00000000
bulkcur=0x00000000
ohci_dumpregs:583:                done=0xbfdf1ca0 fmival=0x27782edf fmrem=0x000009bd
ohci_dumpregs:587:                fmnum=0x00008e3a perst=0x00002a27
lsthrs=0x00000628
ohci_dumpregs:591:                desca=0x02000b03 descb=0x00000000 stat=0x00000000
ohci_dumpregs:594:                port1=0x00000303 port2=0x00000100
ohci_dumpregs:600:          HCCA: frame_number=0x0000 done_head=0x00000000

This is dump just after we programmed it:
ohci_controller_init:308: rewrite head regs
ohci_dumpregs:567: ohci_dumpregs: rev=0x00000110 control=0x000000af
command=0x00000000
ohci_dumpregs:571:                intrstat=0x00000044 intre=0x8000005a
intrd=0x8000005a
ohci_dumpregs:575:                hcca=0x06647000 percur=0x00000000
ctrlhd=0x06692000
ohci_dumpregs:579:                ctrlcur=0x00000000 bulkhd=0x06693000
bulkcur=0x00000000
ohci_dumpregs:583:                done=0x00000000 fmival=0xa7782edf fmrem=0x80001096
ohci_dumpregs:587:                fmnum=0x0000000d perst=0x00002a2f
lsthrs=0x00000628
ohci_dumpregs:591:                desca=0x02000b03 descb=0x00000000 stat=0x00000000
ohci_dumpregs:594:                port1=0x00010301 port2=0x00000100
ohci_dumpregs:600:          HCCA: frame_number=0x000e done_head=0x00000000

This is dump of ohci0 registers just before we run takeover code of ohci1:
ohci1: <OHCI (generic) USB controller> mem 0xfe02d000-0xfe02dfff irq 16 at
device 18.1 on pci0
ohci1: [ITHREAD]
ohci_controller_init:185: reread ohci0 regs:
ohci_dumpregs:567: ohci_dumpregs: rev=0x00000110 control=0x000000af
command=0x00000000
ohci_dumpregs:571:                intrstat=0x00000044 intre=0x8000005a
intrd=0x8000005a
ohci_dumpregs:575:                hcca=0x06647000 percur=0x00000000
ctrlhd=0x06692000
ohci_dumpregs:579:                ctrlcur=0x00000000 bulkhd=0x06693000
bulkcur=0x00000000
ohci_dumpregs:583:                done=0x00000000 fmival=0xa7782edf fmrem=0x800003b5
ohci_dumpregs:587:                fmnum=0x00000012 perst=0x00002a2f
lsthrs=0x00000628
ohci_dumpregs:591:                desca=0x02000b03 descb=0x00000000 stat=0x00000000
ohci_dumpregs:594:                port1=0x00010301 port2=0x00000100
ohci_dumpregs:600:          HCCA: frame_number=0x0012 done_head=0x00000000

And this is dump of ohci0 right after we've taken over ohci1:
ohci_controller_init:195: SMM active, request owner change
ohci_controller_init:219: usbus1: resetting
ohci_controller_init:246: reread ohci0 regs:
ohci_dumpregs:567: ohci_dumpregs: rev=0x00000110 control=0x000000af
command=0x00000000
ohci_dumpregs:571:                intrstat=0x00000004 intre=0x8000005a
intrd=0x8000005a
ohci_dumpregs:575:                hcca=0x06647000 percur=0x00000000
ctrlhd=0xbfdf1c50
ohci_dumpregs:579:                ctrlcur=0x00000000 bulkhd=0x06693000
bulkcur=0x00000000
ohci_dumpregs:583:                done=0xbfdf1ca0 fmival=0xa7782edf fmrem=0x80002122
ohci_dumpregs:587:                fmnum=0x00000192 perst=0x00002a2f
lsthrs=0x00000628
ohci_dumpregs:591:                desca=0x02000b03 descb=0x00000000 stat=0x00000000
ohci_dumpregs:594:                port1=0x00000303 port2=0x00000100
ohci_dumpregs:600:          HCCA: frame_number=0x0192 done_head=0x00000000

As you can see, indeed, the register gets over-written right when we take over
ohci1.
Some additional observations:
1. frame_number seems to grow quite a lot for ohci0
2. before we touch ohci0 it has port1=0x00000303, after reset port1=0x00010301,
after ohci1 takeover port1=0x00000303 again.

I'd say that this is a pretty strong evidence that BIOS does something to ohci0
after we took over it and while we are taking over ohci1.

Another idea of working around this:
1) in pci fixup code disable USB SMI for these chipsets
2) (optional) in ohci code skip takeover step
Sounds messy.

-- 
Andriy Gapon