Question about IMCFGLO timeouts workaround

Question about IMCFGLO timeouts workaround

[Rafał Miłecki]

drivers/net/wireless/b43/main.c
b43_imcfglo_timeouts_workaround

Does it make sense to have here code:
case SSB_BUSTYPE_SSB:
	tmp &= ~SSB_IMCFGLO_REQTO;
	tmp &= ~SSB_IMCFGLO_SERTO;
	tmp |= 0x53;
	break;

Does SSB_BUSTYPE_SSB still have PCI core? I though SSB_BUSTYPE_SSB is
some kind of native, not attached with PCI host.

-- 
Rafa?

Question about IMCFGLO timeouts workaround

[Michael Büsch]

On Wed, 2011-05-04 at 12:41 +0200, Rafa? Mi?ecki wrote: 
> drivers/net/wireless/b43/main.c
> b43_imcfglo_timeouts_workaround
> 
> Does it make sense to have here code:
> case SSB_BUSTYPE_SSB:
> 	tmp &= ~SSB_IMCFGLO_REQTO;
> 	tmp &= ~SSB_IMCFGLO_SERTO;
> 	tmp |= 0x53;
> 	break;
> 
> Does SSB_BUSTYPE_SSB still have PCI core? I though SSB_BUSTYPE_SSB is
> some kind of native, not attached with PCI host.

Well, is REQTO and SERTO related to the PCI core only?

-- 
Greetings Michael.

Question about IMCFGLO timeouts workaround

[Rafał Miłecki]

W dniu 4 maja 2011 14:35 u?ytkownik Michael B?sch <m@bues.ch> napisa?:
> On Wed, 2011-05-04 at 12:41 +0200, Rafa? Mi?ecki wrote:
>> drivers/net/wireless/b43/main.c
>> b43_imcfglo_timeouts_workaround
>>
>> Does it make sense to have here code:
>> case SSB_BUSTYPE_SSB:
>> ? ? ? tmp &= ~SSB_IMCFGLO_REQTO;
>> ? ? ? tmp &= ~SSB_IMCFGLO_SERTO;
>> ? ? ? tmp |= 0x53;
>> ? ? ? break;
>>
>> Does SSB_BUSTYPE_SSB still have PCI core? I though SSB_BUSTYPE_SSB is
>> some kind of native, not attached with PCI host.
>
> Well, is REQTO and SERTO related to the PCI core only?

No. There are two documented cases:
1) PCI (not e) cores rev < 5 need x32
2) Some BCM4311 and BCM4312 need 0x03

This SSB_BUSTYPE_SSB with 0x53 is tricky for me and seems to be undocumented.

-- 
Rafa?

Question about IMCFGLO timeouts workaround

[Rafał Miłecki]

W dniu 4 maja 2011 14:45 u?ytkownik Rafa? Mi?ecki <zajec5@gmail.com> napisa?:
> 1) PCI (not e) cores rev < 5 need x32

This one is performed on PCI core


> 2) Some BCM4311 and BCM4312 need 0x03

This one is performed on 80211 core.


> This SSB_BUSTYPE_SSB with 0x53 is tricky for me and seems to be undocumented.

Not sure about this one. We currently perform this on 80211 core, but
this is in group with 0x32, which we were performing incorrectly on
80211 core.

-- 
Rafa?

Question about IMCFGLO timeouts workaround

[Michael Büsch]

On Wed, 2011-05-04 at 14:45 +0200, Rafa? Mi?ecki wrote: 
> > Well, is REQTO and SERTO related to the PCI core only?
> 
> No. There are two documented cases:
> 1) PCI (not e) cores rev < 5 need x32
> 2) Some BCM4311 and BCM4312 need 0x03
> 
> This SSB_BUSTYPE_SSB with 0x53 is tricky for me and seems to be undocumented.

I'm not sure where it came from. It probably was there from the
very first b43 version on. I don't remember why I put it there.
Maybe you can find hints on the old (v3) specifications.

-- 
Greetings Michael.