Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 23 Jun 2010 00:59:11 +0300
From:      Alexander Motin <mav@FreeBSD.org>
To:        Doug Barton <dougb@FreeBSD.org>
Cc:        freebsd-current@FreeBSD.org
Subject:   Re: Timer panic on boot (r209434)
Message-ID:  <4C21322F.9080009@FreeBSD.org>
In-Reply-To: <4C212DB9.6030308@FreeBSD.org>
References:  <4C211538.1040808@FreeBSD.org> <4C2115D2.5080504@FreeBSD.org> <4C2118D1.3080903@FreeBSD.org> <4C212460.1070200@FreeBSD.org> <4C212873.5010807@FreeBSD.org> <4C212DB9.6030308@FreeBSD.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Doug Barton wrote:
> On 06/22/10 14:17, Alexander Motin wrote:
>> Run `sysctl kern.eventtimer.timer2=i8254`, then after few seconds check
>> messages to see if system liked this timer (it should fall back
>> automatically if it's not),
> 
> Seems ok. Here is what I got on the console, no error messages in
> /var/log/all.
> 
> sysctl kern.eventtimer.timer2=i8254
> kern.eventtimer.timer2: HPET1Starting kernel event timers: HPET @ 100Hz,
> i8254 @
>  128Hz
>   -> i8254
>  t_delta 16.01a20d312197c8b0 too long
> 
>> then check 'vmstat -ia' to see whether irq0 interrupts are arriving,
> 
> This also seems fine:
> 
> interrupt                          total       rate
> ???                                    0          0
> irq1: atkbd0                        3448          1
> stray irq1                             0          0
> irq0: attimer0                     15756          6
> stray irq0                             0          0
> irq3:                                  0          0
> stray irq3                             0          0
> 
> The total for irq0 is going up consistently.

OK, thanks. It means that your ACPI is lying for some reason. I'll
probably commit this patch tomorrow.

> Should I continue using the HPET timer? 

As you wish.

> Is it "better" in some way?

Comparing to what? Comparing to LAPIC - it is not dying in C3. Comparing
to RTC - if is faster and much more flexible. Comparing to i8254 - it
can work per-CPU and supports one-shot mode, both not very important
now, but should benefit later.

> Anything else I can do to help?

Find any more issues to fix. :)

As you have latest HEAD, you may try my latest addition (r209440) - HPET
"legacy route" support. It should allow HPET to work per-CPU on your
hardware. To enable it, add such lines to /boot/loader.conf:
hint.atrtc.0.clock=0
hint.attimer.0.clock=0
hint.hpet.0.legacy_route=1

-- 
Alexander Motin



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4C21322F.9080009>