X hard locks machine with a Radeon 9200

[Thom May]

From one of my sounders:
I installed Ubuntu last night. Unfortunately, when gdm attempted to invoke X,
the machine hard-froze. It worked with the vga and vesa X drivers, however. So I
guess the Radeon 9200 is verboten
And it was a hard crash - couldn't ping it from another machine. Tried it with
and without fbdev.
What's funny, though, is that Caps Lock and Num Lock worked initially, but
trying, blindly, to switch back to a virtual console, or ctrl+alt+backspace
froze even those.
He'll provide dmesg, lspci, and an strace of gdm starting this evening (GMT)

[Fabio Massimo Di Nitto (fabbione)]

I will need all possible information. config files (working and not working),
possibly a gdb session using xserver-xfree86-dbg, lspci -n and so on.

[Fabio Massimo Di Nitto (fabbione)]

Thom, I still didn't receive any information about this problem. Can you kindly
shake the guy who did the report?

I am downgrading the severity in the meanwhile.

[Thom May (thombot)]

Nick added vga=792 to the kernel parameters and the card now works:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=118860
https://freedesktop.org/bugzilla/show_bug.cgi?id=1109

I'll ask for the logs again on monday.

[Fabio Massimo Di Nitto (fabbione)]

Ok we updated the ati driver to a certain degree with X ubuntu20. Please let us
know if that fixes the problem. Otherwise i am more tempt to close this bug for
lack of information.

[Fabio Massimo Di Nitto (fabbione)]

This bug won't be fixed for warty. tagging it as LATER.

[Fabio Massimo Di Nitto (fabbione)]

As agreed with Matt on irc.

[Daniel Stone (daniels)]

Wickity word -- I fixed this bug upstream about a month ago when I was looking
at that particular bit of code, but didn't realise that it was affecting this
bug. Now, looking at it, I'm almost certain it does. I proposed it for 6.8.2 a
few hours ago, with patch, and it will be in 000_stolen_from_6.8_branch.diff in
the next xorg upload.

Isn't it great when you have an epiphany? I was sorting through the bug list
and suddenly got very happy when I realised what this bug was.

[Daniel Stone (daniels)]

xorg (6.8.1-1ubuntu4) hoary; urgency=low

  * Remove debian/patches/099d*, as it broke UTF-8 locale handling and didn't
    actually fix the xterm manpage at all (closes: Ubuntu#3645).

  * Add patches stolen from the X.Org 6.8.x branch, including:
    + Perform soft resets right on non-r3xx (fd.o#988).
    + Prevent segfault in GLX on certain GL versions (fd.o#1672).
    + Fix server crash when unsetting screensaver attributes (fd.o#1695).
      (fd.o#591).
    + Prevent conflict with certain #defined constants (bool) in xf86cfg
      (fd.o#557).
    + Drag nVidia driver back from XFree86 (via X.Org HEAD) with more hardware
      support, fix for DPMS issues, PCI-Xpress support, DPMS on DVI, cursor
      fixes for NV11, proper console restoration, and Xv syncing on vblank
      (fd.o#1777).
    + Fix a stack overflow in XKB handling within libX11 (fd.o#1459).
    + Fix xmodmap stack overflow (fd.o#1818).
    + Fix from myself to fix panel size derivation when stretching is enabled
      (fd.o#1881, Ubuntu#1218).
    + Include server control keys in macintosh/us XKB definition (fd.o#1872).
    + Actually reset DPMS timer on input (fd.o#792, Ubuntu#3614).
    + Move all the fixes from #000 Fedora to this patch, as they have since
      been approved for 6.8.2.
    - New: debian/patches/000_stolen_from_6.8_branch.diff
    - Removed: debian/patches/000_stolen_from_fedora.diff
    + Large patch from Ben Herrenschmidt rejigging the detection logic for
      many ATI cards; see annotation in patch. Note that this is not yet
      committed, so this patch is subject to change and hasn't yet been folded
      into _6.8_branch.diff. This patch obsoletes #099m.
      - New: debian/patches/000_stolen_from_6.8_branch_benh_ati.diff
      - Updated:
      debian/patches/024_ati_r128_and_radeon_enable_build_without_vgahw.diff
      - Removed: debian/patches/099m_radeon_fwpll_support.diff

  * Remove interrupt handler completely on exit so we don't get lockups
    (closes: Ubuntu#2411).
    + New: debian/patches/025_r128_remove_interrupt_handler.diff

  * Add 1440x900 @ 100Hz modeline for 17" Apple iMacs (closes: Ubuntu#3712).
    + Updated: debian/xserver-xorg.templates,
      debian/989_ubuntu_add_extra_modelines_from_xorg.diff

  * Stop writing out HorizSync and VertRefresh lines to xorg.conf; if it can
    be detected by DDC, then let the driver do that instead of tying us to a
    specific monitor, and if we need to calculate it from the resolution, then
    it's a driver bug that needs fixing.
    + Updated: debian/local/dexconf.

  * First cut at proper ia64 control files (closes: Ubuntu#3817).
    + Updated: debian/MANIFEST.ia64.in, debian/xserver-xorg.install.ia64,
      debian/xlibmesa-dri.install.ia64, debian/xlibmesa-dri-dbg.install.ia64,
      debian/xserver-xorg.docs.ia64

  * Stop shipping damage.h, which is an old preliminary cut at the Damage
    client API; superseded by Xdamage.h.
    + Updated: MANIFEST.*.in, libxdamage-dev.install.

  * Fix link name (Xres->XRes) in libxres1.

  * Make libx11-dev Depend on libxkbfile-dev, and libxp-dev on libxau-dev, as
    they use files from these libraries.

  * Add dmxext.h and dmxproto.h to libdmx...