HPET on ICH4 and -hrt7 patch

[Andre Müller]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

Carlos R. Mafra wrote:
>> Evgeni Golov wrote:
>> > Hi,
>> >
>> > On Tue, 12 Jun 2007 12:58:50 -0300 Carlos R. Mafra wrote:
>> >
>> >> Maybe some people may find it useful to know that the -hrt7 patch [1]
>> >> made my ICH4 chipset use its hidden HPET sucesfully and that
>> >> suspend to ram is not broken.
>> >
>> > Well, it enables HPET for me, but s2ram is still broken (rc4 resumes
>> > fine without hrt patches). What hardware do you have? It's an IBM X31
>> > here...
> 
> I have a ThinkPad R50e, 2000 MHz, ICH4 chipset.

I've got an R50e, 1834-8RG, Celeron M on an ICH4M chipset.

> 
>> and here, -hrt7 does not force hpet in, using tsc instead.
>> I am on linux-2.1.22-rc4-hrt7, an ICH-4M chipset.
>> using -hrt2, all is well.
> 
> Hmm...as the <extra timer interrupt> has vanished after -hrt2 and
> -hrt7 I suppose
> the patch made hpet work. There is also a 30 wakeups drop from a non
> -hrt kernel.
> Here is the relevant lines from my dmesg (using -hrt7):
> 
> hpet clockevent registered
> hpet0: at MMIO 0xfed00000, IRQs 2, 8, 0
> hpet0: 3 64-bit timers, 14318180 Hz
> 
> But I also have this line:
> 
> Time: tsc clocksource has been installed.
> 
> Now I wonder which one my laptop is using...

you've got
# cat/sys/devices/system/clocksource/clocksource0/current_clocksource
to tell. also powertop won't work properly with tsc, anouncing
the cpu is all busy and never sleeps, so you would know.

i can switch clocksources on the -hrt2-kernel, as far as i can tell.
hpet is listed as available on -hrt7, but enabling hpet
on-the-fly on -hrt7 did not do either.

best, andre

>
> After using 2.6.22-rc4-hrt7 for more time I realized that there is a
> very bad problem
> with it. The power consume has raised almost 3 Watts! And powertop no
> longer
> shows cpu C-states info when running, saying that the cpu was at 100%
> (which
> is not the case, as my system was idle).
> 
> Here is the output of Powertop 1.6 on 2.6.22-rc4-hrt7 on a idle X:
> (it is in Portuguese)
> 
> ***************************** 2.6.22-rc4-hrt7
> **************************************
> < A CPU estava 100% ocupada; não se entrou em nenhum estado-C >
> 
> Ativações por segundo: 16,7
> Power usage (ACPI estimate): 11,1W (2,1 hours) (long term: 11,2W,/2,1h)
> Maiores causas de ativações:
>  32,1% (  3,5)              Xorg : do_setitimer (it_real_fn)
>  26,6% (  2,9)              WindowMaker : schedule_timeout
> (process_timeout)
>  18,3% (  2,0)              rc.local : queue_delayed_work_on
> (delayed_work_timer_fn)
>   9,2% (  1,0)               Xorg : schedule_timeout (process_timeout)
>   4,6% (  0,5)               gpm : do_nanosleep (hrtimer_wakeup)
>   4,6% (  0,5)               <kernel core> : queue_delayed_work_on
> (delayed_work_timer_fn)
>   1,8% (  0,2)               init : schedule_timeout (process_timeout)
>   1,8% (  0,2)               <kernel core> :
> neigh_table_init_no_netlink (neigh_periodic_timer)
>   0,9% (  0,1)               syslogd : do_setitimer (it_real_fn)
> **************************************************************************************
> 
> 
> And here it is the output for -hrt2, under the same conditions:
> 
> ************************** 2.6.22-rc4-hrt2
> *************************************
> Cn          Residência média (20s)      Residência média a longo prazo
> 
> C0 (cpu ocupada)        ( 0,3%)
> C1                0,0ms ( 0,0%)                   0,0ms
> C2                0,0ms ( 0,0%)                   0,2ms
> C3               37,0ms ( 2,2%)                  43,7ms
> C4               83,6ms (97,5%)                  80,2ms
> 
> Ativações por segundo: 12,2
> Power usage (ACPI estimate): 8,4W (2,7 hours)
> Maiores causas de ativações:
>  30,6% (  3,5)              Xorg : do_setitimer (it_real_fn)
>  28,8% (  3,3)              WindowMaker : schedule_timeout
> (process_timeout)
>  17,5% (  2,0)              rc.local : queue_delayed_work_on
> (delayed_work_timer_fn)
>   8,7% (  1,0)               Xorg : schedule_timeout (process_timeout)
>   4,4% (  0,5)               <kernel core> : queue_delayed_work_on
> (delayed_work_timer_fn)
>   4,4% (  0,5)               gpm : do_nanosleep (hrtimer_wakeup)
>   2,2% (  0,2)               <kernel core> :
> neigh_table_init_no_netlink (neigh_periodic_timer)
>   1,7% (  0,2)                init : schedule_timeout (process_timeout)
>   0,9% (  0,1)           <kernel module>
> ieee80211_node_attach(ieee80211_node_timeout)
>   0,4% (  0,1)               <kernel core> : addrconf_verify
> (addrconf_verify)
>   0,4% (  0,1)                syslogd : do_setitimer (it_real_fn)
> 
> But in -hrt2 the s2ram does not work, and in -hrt7 it works. But now I
> will come back to -hrt2, because 3 Watts is more important than s2ram
> for me :-)
> 
> Carlos R. Mafra
> 
> PS: Does somebody out there use WindowMaker? I think wmaker has a bug,
> because
> it is waking up at 3 Hz without any reason. The wm-dev list is down,
> so I can´t ask anyone
> to take a look at it :-(
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)

iD8DBQFGbwNghYOgL/M0VS0RCm3JAJ92jYL+M2KH1FggtTn1XhYZDBpLYQCeOsNQ
G+TCufzxBKM1DadV9Pti0gU=
=JplM
-----END PGP SIGNATURE-----