Re: kern/49744 console is blank drm2/radeon

["Andrius V via gnats" <gnats-admin%NetBSD.org@localhost>]

The following reply was made to PR kern/49744; it has been noted by GNATS.

From: Andrius V <vezhlys%gmail.com@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: christos%netbsd.org@localhost, mrg%netbsd.org@localhost
Subject: Re: kern/49744 console is blank drm2/radeon
Date: Sat, 6 Jun 2026 12:04:47 +0300

 On Mon, Oct 21, 2024 at 8:45=E2=80=AFAM matthew green via gnats
 <gnats-admin%netbsd.org@localhost> wrote:
 >
 > The following reply was made to PR kern/49744; it has been noted by GNATS=
 .
 >
 > From: matthew green <mrg%eterna23.net@localhost>
 > To: Rin Okuyama <rokuyama.rk%gmail.com@localhost>
 > Cc: Taylor R Campbell <riastradh%NetBSD.org@localhost>,
 >     Martin Neitzel <neitzel%hackett.marshlabs.gaertner.de@localhost>,
 >     gnats-bugs%NetBSD.org@localhost, netbsd-bugs%NetBSD.org@localhost
 > Subject: re: kern/49744 console is blank drm2/radeon
 > Date: Mon, 21 Oct 2024 16:44:15 +1100
 >
 >  Rin Okuyama writes:
 >  > Martin Neitzel kindly reports that X11 works just fine on
 >  > HP Proliant DL360-Gen6 with ATI ES1000, if the workaround for
 >  > this PR is disabled:
 >  >
 >  > https://mail-index.netbsd.org/current-users/2024/10/10/msg045709.html
 >  >
 >  > Note that DL380 with ATI ES1000 did not work at that time.
 >  >
 >  > Can we switch `radeon_pci_ignore_r100_r200` to false by default,
 >  > or just revert the workaround itself?
 >  >
 >  > http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/external/bsd/drm2/radeon/r=
 ad=3D
 >  eon_pci.c.diff?r1=3D3D1.9&r2=3D3D1.10&f=3D3Dh
 >  >
 >  > Can someone still test recent -current or netbsd-10 on the
 >  > affected machines?
 >
 >  unfortunately, nothing has changed for me on my affected old
 >  system.  X11 hangs when starting up, and there's this spew in
 >  dmesg every second:
 >
 >  [  3358.002691] radeon0: error: error: ring 0 stalled for more than 7201=
 0m=3D
 >  sec
 >  [  3358.002691] radeon0: warn: GPU lockup (current fence id 0x0000000000=
 00=3D
 >  0003 last fence id 0x0000000000000005 on ring 0)
 >  [  3358.502708] radeon0: error: error: ring 0 stalled for more than 7251=
 0m=3D
 >  sec
 >  [  3358.502708] radeon0: warn: GPU lockup (current fence id 0x0000000000=
 00=3D
 >  0003 last fence id 0x0000000000000005 on ring 0)
 >
 >  wsfb works.  perhaps we could enable r100/r200, but make the
 >  default driver for these in X itself be wsfb not radeon, and
 >  then it becomes much easier to test with a basic xorg.conf
 >  vs. a patched kernel.
 >
 >  actually, this userland is older, and i only upgraded the
 >  kernel, so i'll try that next.
 >
 >
 >  .mrg.
 >
 
 Hi,
 
 I suggest to apply
 https://github.com/torvalds/linux/commit/ba806f98f868ce107aa9c453fef751de99=
 80e4af
 to our radeon driver, which forces PCI mode for all radeon(4) cards.
 
 Linux users were reporting similar issues, thus devs just gave up and
 forced PCI mode for all cards. It probably may impact performance on
 some cards, but gives stability others.
 
 I faced same stalls on Radeon 9550 AGP card, forcing PCI mode makes it
 work pretty stable.
 
 diff --git a/sys/external/bsd/drm2/dist/drm/radeon/radeon_drv.c
 b/sys/external/bsd/drm2/dist/drm/radeon/radeon_drv.c
 index df38fd1defc..a2b6c3ba636 100644
 --- a/sys/external/bsd/drm2/dist/drm/radeon/radeon_drv.c
 +++ b/sys/external/bsd/drm2/dist/drm/radeon/radeon_drv.c
 @@ -185,12 +185,7 @@ int radeon_no_wb;
  int radeon_modeset =3D -1;
  int radeon_dynclks =3D -1;
  int radeon_r4xx_atom =3D 0;
 -#ifdef __powerpc__
 -/* Default to PCI on PowerPC (fdo #95017) */
  int radeon_agpmode =3D -1;
 -#else
 -int radeon_agpmode =3D 0;
 -#endif
  int radeon_vram_limit =3D 0;
  int radeon_gart_size =3D -1; /* auto */
  int radeon_benchmarking =3D 0;
 
 
 Regards,
 Andrius V