linux-kernel - Re: [PATCH] acpi: apei: call into AER handling regardless of severity

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: <20170829221932.ojkvr4y6s76hcpkj@pd.tnic>
Date:   Wed, 30 Aug 2017 00:19:32 +0200
From:   Borislav Petkov <bp@...e.de>
To:     "Baicar, Tyler" <tbaicar@...eaurora.org>,
        Tony Luck <tony.luck@...el.com>
Cc:     rjw@...ysocki.net, lenb@...nel.org, will.deacon@....com,
        james.morse@....com, prarit@...hat.com, punit.agrawal@....com,
        shiju.jose@...wei.com, andriy.shevchenko@...ux.intel.com,
        linux-acpi@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] acpi: apei: call into AER handling regardless of severity

On Tue, Aug 29, 2017 at 03:27:42PM -0600, Baicar, Tyler wrote:
> To avoid calling the
> do_recovery() function for correctable errors I created
> https://patchwork.kernel.org/patch/9925877/

enum {
        GHES_SEV_NO = 0x0,
        GHES_SEV_CORRECTED = 0x1,
        GHES_SEV_RECOVERABLE = 0x2,
        GHES_SEV_PANIC = 0x3,
};

>From all those severity types above, you want to do recovery for
GHES_SEV_RECOVERABLE but print *all* severities. Yes? I mean, this is
what makes most sense: you want to dump all errors but try to recover
from those from which you *actually* have the possibility to do so.

Looking at the severities conversion, GHES_SEV_RECOVERABLE is
CPER_SEV_RECOVERABLE. cper_severity_to_aer() converts then
CPER_SEV_RECOVERABLE to AER_NONFATAL.

  [ Btw, this is the dumbest sh*t ever. Three different severities!!!
    Looks like someone has won a contest of how to design something as
    needlessly complex as possible. ]

So it looks to me like you want to do rather:

	if (entry.severity == AER_NONFATAL)
		do_recovery(pdev, entry.severity);

which should correspond to the GHES_SEV_RECOVERABLE. And this would be
the straight-forward thing and that would be fine but...

... that is still not 100% equivalent because the check is:

	if (sev == GHES_SEV_RECOVERABLE && sec_sev == GHES_SEV_RECOVERABLE...

so there's the severity of the estatus block and then the severity of
each section successively.

And I have no idea why we're doing this.

Because if we have to keep this, then the above simplification won't work and
you'll have to pass in a separate argument to aer_recover_queue():

	aer_recover_queue( ..., sev == GHES_SEV_RECOVERABLE &&
				sec_sev == GHES_SEV_RECOVERABLE, ...

which, if true, would mean, do recovery.

So let's find out first why do we have to look at both severities.

Tony, any ideas?

-- 
Regards/Gruss,
    Boris.

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
--