On 2018/9/26 1:15, Luck, Tony wrote:
> On Tue, Sep 25, 2018 at 12:11:07PM +0200, Borislav Petkov wrote:
>> On Tue, Sep 25, 2018 at 02:36:04AM +0000, YueHaibing wrote:
>>> if (uncorrected_error) {
>>> - if (ripv) {
>>> - type = "FATAL";
>>> + if (ripv)
>>> tp_event = HW_EVENT_ERR_FATAL;
>>> - } else {
>>> - type = "NON_FATAL";
>>> + else
>>> tp_event = HW_EVENT_ERR_UNCORRECTED;
>>> - }
>>> } else {
>>> - type = "CORRECTED";
>>> tp_event = HW_EVENT_ERR_CORRECTED;
>>> }
>>
>> Looks like this @type thing was never used. Tony, did you want to use it
>> for something?
>
> That block of code looks to have been pasted from sb_edac.c, so likely
> need the same fix there.
>
> It also looks to be wrong. Saying "FATAL" if mcgstatus.ripv is set
> looks wrong (and incomplete). If RIPV is *NOT* set, then this is fatal.
> But if it is set, then we should be looking at other bits (like PCC
> and AR) ... or maybe just calling mce_severity()?
>
> Perhaps a moot point though. If the error is fatal, then we'll panic
> instead of calling the EDAC driver.
Thank you for comment
>
> -Tony
>
> .
>
