[flashrom] flashrom -p gfxnvidia detect my atheros ath9k wifi card

[Carl-Daniel Hailfinger]

Ping?

IMHO this patch fixes a few structural problems, and although it
probably isn't the final desired result, it is a step in the right
direction.

Regards,
Carl-Daniel

Am 22.07.2012 04:39 schrieb Carl-Daniel Hailfinger:
> Am 21.07.2012 16:02 schrieb Michael Karcher:
>> Am Dienstag, den 17.07.2012, 09:17 +0200 schrieb Carl-Daniel Hailfinger:
>>
>>> Second patch in the series. Error checking has not changed at all, and I
>>> intend to fix that in a second spin of this patch. Right now I just want
>>> to check if we're on the same page.
> Actually, I plan to commit the error checking change separately because
> Niklas Söderlund already sent a patch doing that.
>
>
>>> Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006 at gmx.net>
>>> -struct pci_dev *pcidev_dev = NULL;
>> Yeah!
>>
>>>  			if ((addr = pcidev_readbar(dev, bar)) != 0) {
>>> -				curaddr = addr;
>>> -				pcidev_dev = dev;
>>> +				found_dev = dev;
>>>  				found++;
>>>  			}
>> I wonder what to do about this: While the goal of my suggestion was to
>> decouple BAR access from the PCI device scan, it is an integral part of
>> the scan loop, probably to avoid disabled chips (e.g. for onboard
>> components that are not used).
> Intel dual port NICs are represented by two PCI devices, but only one of
> them has an active BAR for our purposes. It would be unfair of us to
> have users guess the right PCI device.
>
>
>> The nice property of the code as-is is
>> that you can be sure reading the "primary" BAR will not fail.
>>
>> As we sometimes need two BARs, having one valid BAR does not mean the
>> device is necessarily usable for us, so this check is only half of what
>> we need. As already discussed on IRC, passing a set of BARs into this
>> function is not really the direction we want to head to, so client code
>> needs to be prepared to find unusable BARs anyway. Still, we like
>> autoskip. Several ideas come to my mind
>> [...]
>> All of these approaches of course complicate pcidev_init, but it seems
>> like the only choices we have is:
>>  - do the BAR check the right way (TM)
>>  - lose the autoskip of disabled PCI devices

-- 
http://www.hailfinger.org/