Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 23 Mar 2009 03:21:00 -0500
From:      Robert Noland <rnoland@FreeBSD.org>
To:        Anonymous <swell.k@gmail.com>
Cc:        freebsd-x11 <freebsd-x11@freebsd.org>
Subject:   Re: [PREVIEW] Nouveau on FreeBSD (Take 2)
Message-ID:  <1237796460.2110.4.camel@balrog.2hip.net>
In-Reply-To: <86ljqwsp7o.fsf@gmail.com>
References:  <1237680263.1938.10.camel@balrog.2hip.net> <86r60pp8c0.fsf@gmail.com> <1237763230.1694.0.camel@balrog.2hip.net> <86ab7dxf5c.fsf@gmail.com> <1237768535.1712.3.camel@balrog.2hip.net> <86ljqwsp7o.fsf@gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help

--=-6mTyHTy6bn1sxav0ZxLJ
Content-Type: text/plain
Content-Transfer-Encoding: quoted-printable

On Mon, 2009-03-23 at 09:55 +0300, Anonymous wrote:
> Robert Noland <rnoland@FreeBSD.org> writes:
>=20
> > On Mon, 2009-03-23 at 03:21 +0300, Anonymous wrote:
> >> Robert Noland <rnoland@FreeBSD.org> writes:
> >>=20
> >> > On Mon, 2009-03-23 at 00:15 +0300, Anonymous wrote:
> >> >> Robert Noland <rnoland@FreeBSD.org> writes:
> >> >>=20
> >> >> > Ok, this patch should work on NV50 chips also.
> >> >> >
> >> >> > What you get is EXA and Xv.
> >> >> >
> >> >> > You still need:
> >> >> >
> >> >> > A recent -CURRENT or -STABLE.
> >> >> >
> >> >> > git master of libdrm and xf86-video-nouveau.
> >> >> >
> >> >> > This patch.
> >> >> >
> >> >> > Things I've figured out since the last patch...
> >> >> >
> >> >> > On NV50 class hardware you need to have a compositing manager run=
ning
> >> >> > for Xv to work.  That means xcompmgr, metacity with composite ena=
bled,
> >> >> > xfce (rumored to work as well, haven't tried).  If your running G=
nome
> >> >> > with metacity, open gconf-editor and go to apps->metacity->genera=
l and
> >> >> > check the composite box.
> >> >> [...]
> >> >> >
> >> >> > http://people.freebsd.org/~rnoland/drm-nouveau-032109.patch
> >> >> >
> >> >> > robert.
> >> >>=20
> >> >> - This error in Xorg.log is still present
> >> >>=20
> >> >>       (II) NOUVEAU(0): [DRI] installation complete
> >> >>       (EE) NOUVEAU(0): [dri] unable to reference front buffer: -19
> >> >
> >> > Ok, update your libdrm... this was fixed in the last few days.
> >> >
> >> > robert.
> >> >
> >>=20
> >> Oops, looks like libdrm installed here wasn't latest. Updated and this
> >> error is gone now.
> >>=20
> >> >>   Should I ignore it?
> >> >>=20
> >> >>=20
> >> >>       info: [drm] PGRAPH_ERROR - nSource:info: [drm] PROTECTION_ERR=
ORinfo: [drm] DMA_R_PROTECTIONinfo: [drm] , nStatus:info: [drm]=20
> >> >>       info: [drm] PGRAPH_ERROR - Ch 2/2 Class 0x8297 Mthd 0x15e0 Da=
ta 0x00000000:0x00000000
> >> >>       error: [drm:pid11:nv50_pgraph_irq_handler] *ERROR* magic set =
1:
> >> >>       error: [drm:pid11:nv50_pgraph_irq_handler] *ERROR* 	0x0040890=
0: 0x80000000
> >> >>       error: [drm:pid11:nv50_pgraph_irq_handler] *ERROR* 	0x0040890=
4: 0xfdb76b7b
> >> >>=20
> >> >>   Should I ignore them?
> >>=20
> >> Looks like this one is not fixed in r190296M. Look into logs below.
> >
> > This is a fencing issue, I haven't ported the drm fence manager.  Do yo=
u
> > know what you are doing to trigger it?  I can't seem to make it occur
> > here...
>=20
> I can reproduce it on xserver-1.6 just by `X -config test.conf', where
> test.conf contains nothing but Device section which specifies to use
> nouveau ddx. I guess it's related to the problem with xcompmgr. They
> usually appear together.

I've seen this a couple of times, but never figured out what is
triggering it yet.  I'll look through the code some more and see if I
can find it...

> >
> >> >>=20
> >> >> - Launching `xcompmgr -a' is tricky. Most of the time it just leave=
s
> >> >>   screen in unusable state, it's not possible to switch to console =
or
> >> >>   move pointer. I want to help debug this one. Here are logs:
> >> >>   http://pastebin.com/m1ca3fc2f
> >> >>   http://pastebin.com/m579d358e
> >
> > FWIW, I don't seem to have any trouble enabling / disabling composite,
> > so this may be a bug in git xserver or libraries.
>=20
> It's same for xserver-1.6. I guess it's because I usually try to use
> vt switch when Xserver hangs on restart and then reboot. After reboot
> xcompmgr doesn't work fine and leaves screen unusable.
>=20
> Workaround is to set hw.drm.msi=3D0 once and next boot should work fine
> without it with MSI enabled.

Yeah, we are probably the only ones running with msi enabled...  We are
the only ones running msi on radeons and I did have to modify the
interrupt handling a little to support it, but I had info from amd on
how to do it...  I need to poke around the registers and see if i see
anything interesting...

robert.

> >
> > robert.
> >
> >> rebuild kernel, libdrm, xf86-video-nouveau, xserver just in case and
> >> reproduced the problem again
> >> http://pastebin.com/m2be24e75
> >> http://pastebin.com/m6c80e1e
--=20
Robert Noland <rnoland@FreeBSD.org>
FreeBSD

--=-6mTyHTy6bn1sxav0ZxLJ
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: This is a digitally signed message part

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.11 (FreeBSD)

iEYEABECAAYFAknHRmwACgkQM4TrQ4qfRONzYACfWY4lBw1KgBLRoUW+G7tV+2L9
cfEAnjWzrtWjLFbtxWgA7Q6muxtCzdHA
=7fYy
-----END PGP SIGNATURE-----

--=-6mTyHTy6bn1sxav0ZxLJ--




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1237796460.2110.4.camel>