Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 9 Nov 2007 00:05:18 +0100
From:      Benjamin Lutz <mail@maxlor.com>
To:        Simon Barner <barner@freebsd.org>
Cc:        Matus Harvan <mharvan@inf.ethz.ch>, freebsd-current@freebsd.org
Subject:   Re: powerd adaptive mode latching
Message-ID:  <200711090005.21842.mail@maxlor.com>
In-Reply-To: <20071108224313.GA1927@dose.local.invalid>
References:  <472E9D0B.5080409@csub.edu> <47334C71.1010102@nortel.com> <20071108224313.GA1927@dose.local.invalid>

next in thread | previous in thread | raw e-mail | index | archive | help
--nextPart3315018.dH3f6CITOK
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

On Thursday 08 November 2007 23:43:13 Simon Barner wrote:
> > > Please see kern/114722. The patch from the PR works fine with my
> > > T61 (T7300).
> > >
> > > Funny enough, I contacted re@ to get this into 7.0 only two
> > > minutes ago.
> > >
> > > For the archives, the similar bug described in bin/117375 already
> > > seems to be adressed in RELENG_7.
> >
> > both pr's are open .. and
> >
> > releng_7 and head are both at v 1.26 of acpi_perf.c
> >
> > so, no it's not fixed, *anywhere*.  :)
>
> It's true that both PRs are still open, but:
>
> 1) kern/114722 should fix your problem (CPUFREQ_CMP takes care of
> almost identical frequencies). Have you already had a chance to
> verify that?
>
> 2) bin/117375 talks about exactly identical frequencies, which should
> be handled by acpi_perf.c (1.26, line 303-306). However, the reporter
> (Cc'ed) of that PR runs FreeBSD 6.2-p8 which already contains the
> removal of duplicate entries (MFC from acpi_perf.c 1.24).
>
>    @Benjamin Lutz: Could you please check if the problem still
> exists, and if so, whether the patch from kern/114722 fixes it?

Before patch (still on 6.2-RELEASE-p8):
  dev.cpu.0.freq_levels: 2100/15000 2100/13720 1890/11360 1050/5531
So yes, the problem still exists.

After patch:
  dev.cpu.0.freq_levels: 2100/15000 2100/13720 1890/11360 1050/5531
And it seems the patch doesn't fix it.

Btw,  looking at that part (the for loop around line 303) of acpi_perf.c=20
in isolation, it just occurred to me that the check for duplicate=20
entries fails if the duplicate entries are the last in the list, which=20
would probably prevent powerd from scaling the CPU back up. Or maybe=20
I'm wrong, I haven't really looked at the rest of the code.

Thanks for working on this!

Cheers
Benjamin

--nextPart3315018.dH3f6CITOK
Content-Type: application/pgp-signature; name=signature.asc 
Content-Description: This is a digitally signed message part.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)

iD8DBQBHM5YuzZEjpyKHuQwRAqfDAJ4s1tx0KGvjcjKEdOI8nv9RFo9bowCfeW2F
YztcXXUWi0bPIUj7UYHB6/g=
=juXD
-----END PGP SIGNATURE-----

--nextPart3315018.dH3f6CITOK--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200711090005.21842.mail>