Re: [PATCH] mlx4: Add support for EEH error recovery

On 7/24/2012 4:12 PM, Kleber Sacilotto de Souza wrote:

On 07/23/2012 06:26 PM, Or Gerlitz wrote:

quoted

Kleber Sacilotto de Souza [off-list ref] wrote:

quoted

quoted

For powerpc we have an IBM internal user space tool that injects the
error on the bus with the aid of the system firmware. The kernel used
was built with the option:
CONFIG_EEH=y
and without the AER options. I will run some more tests with the AER
options activated.

I tested the powerpc error injection with

CONFIG_EEH=y
CONFIG_PCIEAER=y
CONFIG_PCIEAER_INJECT=m

and with the aer_inject module loaded and it didn't affect the EEH
recovery, the adapter recovered as expected.

I wasn't sure to follow what did you mean by "it didn't affect the EEH
recovery", how did you use the aer_inject module, is that through
user-space tool which is available for us?

I wanted to say that I was testing before only with the EEH option
activated, then I activated the AER options on my powerpc system just to
make sure these options when activate wouldn't affect the EEH recovery.
I haven't injected and AER error since I don't have a system with
hardware support for it.


Thanks,

Hi

Using a special extender card I've powered down the card.
None of the callbacks were called (I added printks to be sure).
Shouldn't one on the callbacks be called?

Shlomo Pongratz.