Mailing List Archive

Update on irq patch for kernel 2.6
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

An update, as subject says.

U60(2x300) seems stable. But it's uptime averages 4 weeks, so 6 days is
not very significant.

U2(2x400) seems stable, and the daily updatedb had been a killer for it.

U60)(2x450) completely unstable.

All were built from the same .config, and I have verified that the patch
is indeed installed (yes, I can mess that sort of thing up, but in this
case, I haven't.)

I note in passing, and I think it is significant, that the unstable U60
also still has -you-need-to-unplug-the-mouse problems. The other two
systems did, but no longer do. (Genrally, I run these systems very hard
with -X-, so they are not idle at all.)

By the way, the SB1000 is doing very nicely with 2.6.11-r7, but java
applications (like firefox, thunderbird) really suck. Random Bus errors
get, shall we say, tedious.

- --
Ferris McCormick (P44646, MI) <fmccor@gentoo.org>
Developer, Gentoo Linux (sparc)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCgliQQa6M3+I///cRAgPzAJ4jljVmx8cxu0D0Taqb5lc5u+ezHACfZLz7
FezGLfyQlBejrkRbN3/ybjg=
=uHPB
-----END PGP SIGNATURE-----
--
gentoo-sparc@gentoo.org mailing list
Re: Update on irq patch for kernel 2.6 [ In reply to ]
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wed, 11 May 2005, Ferris McCormick wrote:

> By the way, the SB1000 is doing very nicely with 2.6.11-r7, but java
> applications (like firefox, thunderbird) really suck. Random Bus errors
> get, shall we say, tedious.

Are you using -mcpu=ultrasparc3 as part of your CFLAGS? I get a lot of
java crashes on my Blade 1000 as well and am thinking that blackdown may
have trouble dealing with dealing with libs built with certain
optimizations.

Cheers,
- --
Jason Wever
Gentoo/Sparc Co-Team Lead
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCgmHVdKvgdVioq28RAv+kAKCxbNKPtsgEO3KuwfQBNYvK7COawQCdE8hx
ReqASSUXWcY4mYrk1fx2qqI=
=mMND
-----END PGP SIGNATURE-----
--
gentoo-sparc@gentoo.org mailing list
Re: Update on irq patch for kernel 2.6 [ In reply to ]
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wed, 11 May 2005, Jason Wever wrote:

> --[PinePGP]--------------------------------------------------[begin]--
> On Wed, 11 May 2005, Ferris McCormick wrote:
>
>> By the way, the SB1000 is doing very nicely with 2.6.11-r7, but java
>> applications (like firefox, thunderbird) really suck. Random Bus errors
>> get, shall we say, tedious.
>
> Are you using -mcpu=ultrasparc3 as part of your CFLAGS? I get a lot of
> java crashes on my Blade 1000 as well and am thinking that blackdown may
> have trouble dealing with dealing with libs built with certain
> optimizations.
>

Yes, I am. You had said earlier that it might be a problem, so I'm
pushing it. I think firefox & friends force -mcpu=v8, but I can't confirm
off hand. And I know that everything else has been rebuilt aggressively
optimized, because it's supposed to work, after all. (Although c++ +
hardened remanins unusalbe.) (But I hate java,
and so maybe it's just fighting back at us?)

> Cheers,
> --
> Jason Wever
> Gentoo/Sparc Co-Team Lead
> --[PinePGP]-----------------------------------------------------------
> gpg: Signature made Wed May 11 19:49:41 2005 UTC using DSA key ID 58A8AB6F
> gpg: Good signature from "Jason Wever <weeve@gentoo.org>"
> --[PinePGP]----------------------------------------------------[end]--
>
>

- --
Ferris McCormick (P44646, MI) <fmccor@gentoo.org>
Developer, Gentoo Linux (sparc)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCgmvKQa6M3+I///cRAqf8AKCf/dywRy11T66D0TQRbvGxdbdaqwCeKnXh
loGQALWgLIm8RsgEL05wZOA=
=MhtA
-----END PGP SIGNATURE-----
--
gentoo-sparc@gentoo.org mailing list
Re: Update on irq patch for kernel 2.6 [ In reply to ]
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wed, 11 May 2005, Ferris McCormick wrote:

> Yes, I am. You had said earlier that it might be a problem, so I'm
> pushing it. I think firefox & friends force -mcpu=v8, but I can't confirm
> off hand. And I know that everything else has been rebuilt aggressively
> optimized, because it's supposed to work, after all. (Although c++ +
> hardened remanins unusalbe.) (But I hate java,
> and so maybe it's just fighting back at us?)

Currently the "replace-sparc64-flags" function from the flag-o-matic
eclass (used by mozilla* as well as others) doesn't handle changing
- -mcpu=ultrasparc3, only -mcpu=ultrasparc or -mcpu=v9.

If you have the resources and feel up to it, it might be worth
experimenting by adding it into the eclass locally and rebuilding mozilla*
or any other packages that use it. For that matter, it's probably still
worthwhile to see if replace-sparc64-flags is still applicable in the
ebuilds that call it.

Cheers,
- --
Jason Wever
Gentoo/Sparc Co-Team Lead
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCgnE8dKvgdVioq28RAp4uAJ9wT1Hp2qh0HpLyaZ+KmeSLgJRU6ACeMazF
vY6SaeKdnObfHwfrpamN8v0=
=oR9/
-----END PGP SIGNATURE-----
--
gentoo-sparc@gentoo.org mailing list
Re: Update on irq patch for kernel 2.6 [ In reply to ]
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wed, 11 May 2005, Jason Wever wrote:

> --[PinePGP]--------------------------------------------------[begin]--
> On Wed, 11 May 2005, Ferris McCormick wrote:
>
>> Yes, I am. You had said earlier that it might be a problem, so I'm
>> pushing it. I think firefox & friends force -mcpu=v8, but I can't confirm
>> off hand. And I know that everything else has been rebuilt aggressively
>> optimized, because it's supposed to work, after all. (Although c++ +
>> hardened remanins unusalbe.) (But I hate java,
>> and so maybe it's just fighting back at us?)
>
> Currently the "replace-sparc64-flags" function from the flag-o-matic
> eclass (used by mozilla* as well as others) doesn't handle changing
> -mcpu=ultrasparc3, only -mcpu=ultrasparc or -mcpu=v9.
>
> If you have the resources and feel up to it, it might be worth
> experimenting by adding it into the eclass locally and rebuilding mozilla*
> or any other packages that use it. For that matter, it's probably still
> worthwhile to see if replace-sparc64-flags is still applicable in the
> ebuilds that call it.
>

Will do. It'll take me a couple of days, though.


> Cheers,
> --
> Jason Wever
> Gentoo/Sparc Co-Team Lead
> --[PinePGP]-----------------------------------------------------------
> gpg: Signature made Wed May 11 20:55:24 2005 UTC using DSA key ID 58A8AB6F
> gpg: Good signature from "Jason Wever <weeve@gentoo.org>"
> --[PinePGP]----------------------------------------------------[end]--
>
>

- --
Ferris McCormick (P44646, MI) <fmccor@gentoo.org>
Developer, Gentoo Linux (sparc)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFCgnQLQa6M3+I///cRAhLNAKCsXgDjPpVWpsCOGfezfS/E1uc64QCgvA3G
x2oR52t0Fu9dkR/jv7tPw9w=
=avY6
-----END PGP SIGNATURE-----
--
gentoo-sparc@gentoo.org mailing list
Re: Update on irq patch for kernel 2.6 [ In reply to ]
On Wed, May 11, 2005 at 02:55:22PM -0600, Jason Wever wrote:
> On Wed, 11 May 2005, Ferris McCormick wrote:
>
> Currently the "replace-sparc64-flags" function from the flag-o-matic
> eclass (used by mozilla* as well as others) doesn't handle changing
> - -mcpu=ultrasparc3, only -mcpu=ultrasparc or -mcpu=v9.
>

When i implemented replace-sparc64-flags(), i defined the
possible replace candidates in a local variable. If you want
to have it also replace ultrasparc3 just add the CPU at the
following line in replace-sparc64-flags():

local SPARC64_CPUS="ultrasparc v9"

Or tell me so i will do it. :)


Christian
--
gentoo-sparc@gentoo.org mailing list