retitle 286013 xserver-xfree86: [mga] driver locks up when XAA enabled on alpha with MGA G400 AGP rev 4
tag 286013 + moreinfo upstream
thanks
On Thu, Dec 16, 2004 at 03:12:38PM -0700, Christoph Reichenbach wrote:
> Package: xserver-xfree86
> Version: 4.3.0-2
I'm sorry it has taken a long time to follow up on this report.
This is a pretty ancient version of the xfree86 packages:
xfree86 (4.3.0-2) unstable; urgency=low
[...]
-- Branden Robinson <branden@debian.org> Wed, 18 Feb 2004 00:28:20 -0500
> Severity: important
>
> While the MGA driver on XF86 4.2.1 works perfectly, it locks up very
> quickly in 4.3.0 (running xdm, or opening an xterm and dumping two
> pages of text are sufficient). The bug is also present in
> 4.3.0.dfsg.1-9 and in several versions in between; 4.3.0-2 (the first
> release of 4.3.0 for alpha) is the first release to exhibit this
> problem.
Oh, I see. In the future, please file the report against the most recent
version that has the problem, and describe the earliest known version with
the problem in the bug description.
> My machine is an UP1000 (nautilus), the G400 is attached to the AGP
> slot; the machine's firmware is an API-internal SRM release. I tried
> debugging for a bit, but couldn't make sense of what I was getting (I
> don't have MGA docs); basically, at some point the G400 stops emptying
> the (command?) fifo, causing WAITFIFO() (mga_macros.h) to wait
> indefinitely. Disabling
Okay. I wonder if some sort of engine-reset logic is needed.
Your bug title suggests, but your message body does not state, that
disabling XAA works around the problem. Can you confirm that? I realize
performance is worse with "NoAccel" set, but I would like to gather that
datum.
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... this last line is repeated a couple of thousand times;
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... repeated some 30 or 40 times ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... repeated 60 times or so ...)
>
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... repeated about 450 times ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... repeated about 800 times ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... repeated some 850 times ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... not even 70 this time ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... I know you don't care about the numbers, but I have to
> vent my frustration somewhere ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... A bit more than 300, I'd say ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... Precisely thirty repetitions omitted ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... There it is... and I forgot to count the lines ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... something 60-ish ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... It would've been easier to write a short remove-dups function ...)
>
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... int10 sounds like an IA32 thing, somehow ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... Come to think of it, EMACS probably has the remove-dups thingy
> built in already ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... repeated about 250 times ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... repeated about 200 times ...)
[...]
> Elf_RelocateEntry() Unsupported relocation type 10
>
> (... repeated a few hundred times ...)
For what it's worth, I believe the above was fixed in the very next package
release subsequent to the one you filed your bug against. That might have
spared some of your frustration.
xfree86 (4.3.0-3) unstable; urgency=medium
[...]
* Add support for R_ALPHA_SREL32 relocations in XFree86 X server's ELF
object loader; fixes "Elf_RelocateEntry() Unsupported relocation type 10"
warnings on Alpha; see #234073 (thanks, Falk Hueffner!).
- debian/patches/202_alpha_elfloader_support_R_ALPHA_SREL32.diff: new file
- debian/patches/303_arm_cache_flush.diff: resynced
[...]
-- Branden Robinson <branden@debian.org> Fri, 27 Feb 2004 15:07:25 -0500
--
G. Branden Robinson |
Debian GNU/Linux | If encryption is outlawed, only
branden@debian.org | outlaws will @goH7Ok=<q4fDj]Kz?.
http://people.debian.org/~branden/ |
Attachment:
signature.asc
Description: Digital signature