linux-kernel - Re: [PATCH v2 4/7] ghes_edac: avoid multiple calls to dmi

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1502128177.2042.131.camel@hpe.com>
Date:   Mon, 7 Aug 2017 17:59:15 +0000
From:   "Kani, Toshimitsu" <toshi.kani@....com>
To:     "bp@...en8.de" <bp@...en8.de>
CC:     "linux-acpi@...r.kernel.org" <linux-acpi@...r.kernel.org>,
        "lenb@...nel.org" <lenb@...nel.org>,
        "mchehab@...nel.org" <mchehab@...nel.org>,
        "tony.luck@...el.com" <tony.luck@...el.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-edac@...r.kernel.org" <linux-edac@...r.kernel.org>,
        "rjw@...ysocki.net" <rjw@...ysocki.net>
Subject: Re: [PATCH v2 4/7] ghes_edac: avoid multiple calls to dmi_walk()

On Sat, 2017-08-05 at 07:16 +0200, Borislav Petkov wrote:
> On Fri, Aug 04, 2017 at 09:02:17PM +0000, Kani, Toshimitsu wrote:
> > GHES platform devices correspond to GHES entries, which define
> > firmware interfaces to report generic memory errors to the OS, such
> > as NMI and SCI.  These devices are associated with all DIMMs, not a
> > particular DIMM.
> 
> And? Stating the obvious brings you what exactly?
> 
> IOW, you still haven't answered my question.

Sorry about that.  Let me copy your original comments to make sure I
answer your questions this time.

> So the problem is that ghes_edac_register() gets called multiple
> times depending on how many GHES platform devices are on the system.
> But yet they all scan *all* DIMMs. 

Right, and this patch changes ghes_edac_register() to scan all DIMMs at
the first time and do not scan them next times.

> So instead you should return if
> the DIMMs have been counted already and not register a second time.
> Which makes that whole mc counting kinda useless. So you could rip
>
that out too.

Oh, I see.  So, ghes_edac_register() should return no-op a second time,
and does not call edac_mc_add_mc() to register with a separate mci.

I think we should keep the current scheme, which registers an mci for
each GHES entry.  ghes_edac_report_mem_error() expects that error-
reporting is serialized per a GHES entry.  Sharing a single mci among
all GHES entries / error interfaces might lead to a race condition.

Thanks,
-Toshi