LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: "Luck, Tony" <tony.luck@intel.com>
To: Borislav Petkov <bp@alien8.de>
Cc: x86@kernel.org, linux-edac@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	Sumanth Kamatala <skamatala@juniper.net>
Subject: Re: [PATCH] x86/mce/dev-mcelog: Call mce_register_decode_chain() much earlier
Date: Mon, 23 Aug 2021 11:45:47 -0700	[thread overview]
Message-ID: <20210823184547.GA1638691@agluck-desk2.amr.corp.intel.com> (raw)
In-Reply-To: <YR/Oxark0bhLlona@zn.tnic>

On Fri, Aug 20, 2021 at 05:48:21PM +0200, Borislav Petkov wrote:
> On Fri, Aug 20, 2021 at 07:43:14AM -0700, Luck, Tony wrote:
> > How can the kernel tell that all consumers have registered? Is there
> > some new kernel crystal ball functionality that can predict that an
> > EDAC driver module is going to be loaded at some point in the future
> > when user space is up and running :-)
> 
> The crystal ball is called mcheck_late_init(). There's even:
> 
>         /*
>          * Flush out everything that has been logged during early boot, now that
>          * everything has been initialized (workqueues, decoders, ...).
>          */
>         mce_schedule_work();
> 
> in there. That thing is late_initcall() and by that time mcelog should
> have been registered. And I wonder why isn't that working as expected...
> 
> > I think the best we could do would be to set a timer for some point
> > far enough out (one minute?, two minutes?) to give a chance for
> > modules to load.
> 
> Forget modules - only the built-in stuff. We cannot be waiting
> indefinitely until someone loads mcelog for decoding.

I added some traces:

$ dmesg | grep mce:
[    0.033648] mce: mce_register_decode_chain fn=mce_early_notifier+0x0/0x50 pri=6
[    0.033655] mce: mce_register_decode_chain fn=uc_decode_notifier+0x0/0xd0 pri=5
[    0.033659] mce: mce_register_decode_chain fn=mce_default_notifier+0x0/0x30 pri=0
[    4.392631] mce: [Hardware Error]: Machine check events logged
[    4.393356] mce: [Hardware Error]: CPU 0: Machine Check: 0 Bank 0: a000000000004321
[    4.395352] mce: [Hardware Error]: TSC 0
[    4.396352] mce: [Hardware Error]: PROCESSOR 0:406f1 TIME 1629743651 SOCKET 0 APIC 0 microcode b000019
[   15.172861] mce: mce_register_decode_chain fn=dev_mce_log+0x0/0x110 pri=1
[   15.192101] mce: mcheck_late_init: calling mce_schedule_work()
[   31.618245] mce: mce_register_decode_chain fn=sbridge_mce_check_error+0x0/0x92 [sb_edac] pri=2

So you are right that mcheck_late_init() is called after dev_mce_log()
is registered.  But it seems someone kicks the queue long before that
happens.  Probably this:

void mce_log(struct mce *m)
{
        if (!mce_gen_pool_add(m))
                irq_work_queue(&mce_irq_work);
}

Could we add a flag:

static bool mce_ready_to_rock; // better name needed :-)

Which gets set in mcheck_late_init(). Then mce_log() becomes:

void mce_log(struct mce *m)
{
        if (!mce_gen_pool_add(m) && mce_ready_to_rock)
                irq_work_queue(&mce_irq_work);
}

-Tony

  reply	other threads:[~2021-08-23 18:45 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-19 22:44 Tony Luck
2021-08-20 12:28 ` Borislav Petkov
2021-08-20 14:43   ` Luck, Tony
2021-08-20 15:48     ` Borislav Petkov
2021-08-23 18:45       ` Luck, Tony [this message]
2021-08-23 20:41         ` [PATCH v2] x86/mce: Defer processing early errors until mcheck_late_init() Luck, Tony
2021-08-23 20:51           ` Borislav Petkov
2021-08-23 21:41             ` Luck, Tony
2021-08-24  0:31             ` [PATCH v3] x86/mce: Defer processing of early errors Luck, Tony
2021-08-24  8:44               ` [tip: ras/core] " tip-bot2 for Borislav Petkov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210823184547.GA1638691@agluck-desk2.amr.corp.intel.com \
    --to=tony.luck@intel.com \
    --cc=bp@alien8.de \
    --cc=linux-edac@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=skamatala@juniper.net \
    --cc=x86@kernel.org \
    --subject='Re: [PATCH] x86/mce/dev-mcelog: Call mce_register_decode_chain() much earlier' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).