Re: [PATCH V10 RESEND] EDAC: Add EDAC driver for loongson memory controller

[Zhao Qunqin]

在 2025/1/4 上午12:19, Borislav Petkov 写道:
> On Tue, Dec 17, 2024 at 10:25:25AM +0800, Zhao Qunqin wrote:
> > 在 2024/12/16 下午7:55, Borislav Petkov 写道:
> > > On Mon, Dec 16, 2024 at 09:33:51AM +0800, Zhao Qunqin wrote:
> > > > +LOONGSON EDAC DRIVER
> > > > +M:	Zhao Qunqin <zhaoqunqin@xxxxxxxxxxx>
> > > > +L:	linux-edac@xxxxxxxxxxxxxxx
> > > > +S:	Maintained
> > > > +F:	drivers/edac/loongson_edac.c
> > > If you add yourself as a maintainer, I'd expect you to review and/or ack
> > > patches for your driver so that I can pick them up.
> > OK. I can review the patches for this driver.
> Not "can" - I would need you to. I don't have the hardware to test and don't
> have the time to review them.
Let me review the patches for this driver.
> > > > +	new = read_ecc(mci);
> > > > +	add = new - pvt->last_ce_count;
> > > > +	pvt->last_ce_count = new;
> > > That last_ce_count is just silly. Kill it.
> > Then  I  can't calculate the error count added since the last check,  cause
> > what record in Loongson's ECC register is  the error count  from reset of
> > the memory controller.
> Can that ECC register be zeroed on each check function run?
That ECC register can not be zeroed.
> Because this should work this way: every time you have to issue an error
> message about how many errors were logged, you simply issue the error count
> you've logged this time.
>
> Userspace tools like rasdaemon etc can then aggregate the counts.