[ath9k-devel] atk9k PCIe-card sometimes comes up with wrong PCI ID

[ath9k-devel] atk9k PCIe-card sometimes comes up with wrong PCI ID

[Holger Schurig]

Hi !

I'm using several Sparklan WPEA 110N cards that normally come up as
168c:002a. But sometimes they announces themself as 168c:ff1c. Even
"lspci -nn -A intel-conf2" (which goes directly to the PCI address
range) shows the wrong PCI ID in such cases.

What could be the cause of this issue?

Am I right to assume that in ath9k-based designs, there is some EEPROM
that contains the MAC address, country code, antenna configuration and
PCI ids?  Could it be the case that some clock is not working that the
PCI hardware part of the chips needs to setup, so that the hardware
defaults are reported instead?

[ath9k-devel] atk9k PCIe-card sometimes comes up with wrong PCI ID

[Michael Schwingen]

On 10/21/2013 02:51 PM, Holger Schurig wrote:
> Hi !
>
> I'm using several Sparklan WPEA 110N cards that normally come up as
> 168c:002a. But sometimes they announces themself as 168c:ff1c. Even
> "lspci -nn -A intel-conf2" (which goes directly to the PCI address
> range) shows the wrong PCI ID in such cases.
>
> What could be the cause of this issue?
I have seen such behaviour on older cards if the card sees two reset 
pulses in too short succession - ie. when the second reset pulse happens 
while the card is still loading data from the EEPROM. Newer Atheros 
cards use I2C EEPROMs, so a chip reset during communication can hang up 
the bus to the EEPROM.

cu
Michael

[ath9k-devel] atk9k PCIe-card sometimes comes up with wrong PCI ID

[Holger Schurig]

Hi Michael,

I found that PCI-RESET issue in the mailing list archive. And I asked
someone to use a storage oscilloscope to look into that.

However, this issue started on our hardware only in the last one, two
month. The PC-like platform has been in use since 4 years or so, so if
the BIOS would have been whacky, that bug would have be visible since
a long time?!?!