Sluggish text scrolling, 100%CPU after upgrade from 2.7.1 (UXA bug)

[Colin Watson]

Binary package hint: xserver-xorg-video-intel

Versions of the Intel driver later than 2:2.7.1-1ubuntu1 are unusably slow for me. The first thing I do after starting X is normally to start a pterm and maximise it; the first sign of trouble is that the scrollbar takes ages to be replaced by normal black terminal area on maximisation. After that, Xorg is using nearly 100% CPU, text scrolling is very slow, etc. Typing this report is excruciatingly painful because it can take up to a second for each character to appear, especially when the typeahead buffer gets at all full.

EXA vs. UXA makes no obvious difference.

There are some rather odd visual effects as well: lots of letters are the wrong colour. This may have only started after switching VT away from X and back; I'm not quite sure. Bizarrely, this is visible in a screenshot (I'm not using compiz or anything fancy like that), which I'll attach to this bug.

Downgrading to 2:2.7.1-1ubuntu1 puts everything back to normal.

ProblemType: Bug
Architecture: i386
Date: Wed Jun 17 10:47:29 2009
DistroRelease: Ubuntu 9.10
MachineType: Dell Inc. Latitude D830
Package: xserver-xorg-video-intel 2:2.7.99.1+git20090602.ec2fde7c-0ubuntu2
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.30-9-generic root=UUID=6df978ed-c024-49d7-9157-bdd7bd05153d ro quiet splash
ProcEnviron:
 LC_COLLATE=C
 PATH=(custom, user)
 LANG=en_GB.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.30-9.10-generic
RelatedPackageVersions:
 xserver-xorg 1:7.4~5ubuntu21
 libgl1-mesa-glx 7.4.1-1ubuntu2
 libdrm2 2.4.11-0ubuntu1
 xserver-xorg-video-intel 2:2.7.99.1+git20090602.ec2fde7c-0ubuntu2
 xserver-xorg-video-ati 1:6.12.2-2ubuntu1
SourcePackage: xserver-xorg-video-intel
Uname: Linux 2.6.30-9-generic i686
dmi.bios.date: 06/07/2007
dmi.bios.vendor: Dell Inc.
dmi.bios.version: A02
dmi.board.name: 0HN341
dmi.board.vendor: Dell Inc.
dmi.chassis.type: 8
dmi.chassis.vendor: Dell Inc.
dmi.modalias: dmi:bvnDellInc.:bvrA02:bd06/07/2007:svnDellInc.:pnLatitudeD830:pvr:rvnDellInc.:rn0HN341:rvr:cvnDellInc.:ct8:cvr:
dmi.product.name: Latitude D830
dmi.sys.vendor: Dell Inc.
fglrx: Not loaded
system:
 distro: Ubuntu
 architecture: i686kernel: 2.6.30-9-generic

[Bryce Harrington (bryce)]

Created an attachment (id=26934)
Screenshot.png

Forwarding this Ubuntu bug report:
https://bugs.edge.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/388368

[Problem]
Poor performance experienced with text scrolling and text typing when using 2.7.99.x, along with incorrectly colored letters. Issue goes away downgrading to 2.7.1.

[lspci]
00:00.0 Host bridge [0600]: Intel Corporation Mobile PM965/GM965/GL960 Memory Controller Hub [8086:2a00] (rev 0c)
 Subsystem: Dell Device [1028:01fe]
00:02.0 VGA compatible controller [0300]: Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller [8086:2a02] (rev 0c)
 Subsystem: Dell Device [1028:01fe]

[RelatedPackageVersions]
 xserver-xorg 1:7.4~5ubuntu21
 libgl1-mesa-glx 7.4.1-1ubuntu2
 libdrm2 2.4.11-0ubuntu1
 xserver-xorg-video-intel 2:2.7.99.1+git20090602.ec2fde7c-0ubuntu2
 xserver-xorg-video-ati 1:6.12.2-2ubuntu1

[System]
 Architecture: i386
 Date: Wed Jun 17 10:47:29 2009
 DistroRelease: Ubuntu 9.10
 MachineType: Dell Inc. Latitude D830
 ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.30-9-generic root=UUID=6df978ed-c024-49d7-9157-bdd7bd05153d ro quiet splash
 distro: Ubuntu
 architecture: i686kernel: 2.6.30-9-generic
 dmi.bios.date: 06/07/2007
 dmi.bios.vendor: Dell Inc.
 dmi.bios.version: A02
 dmi.board.name: 0HN341
 dmi.board.vendor: Dell Inc.
 dmi.chassis.type: 8
 dmi.chassis.vendor: Dell Inc.
 dmi.modalias: dmi:bvnDellInc.:bvrA02:bd06/07/2007:svnDellInc.:pnLatitudeD830:pvr:rvnDellInc.:rn0HN341:rvr:cvnDellInc.:ct8:cvr:
dmi.product.name: Latitude D830
 dmi.sys.vendor: Dell Inc.

[Original Report]
Versions of the Intel driver later than 2:2.7.1-1ubuntu1 are unusably slow for me. The first thing I do after starting X is normally to start a pterm and maximise it; the first sign of trouble is that the scrollbar takes ages to be replaced by normal black terminal area on maximisation. After that, Xorg is using nearly 100% CPU, text scrolling is very slow, etc. Typing this report is excruciatingly painful because it can take up to a second for each character to appear, especially when the typeahead buffer gets at all full.

EXA vs. UXA makes no obvious difference.

There are some rather odd visual effects as well: lots of letters are the wrong colour. This may have only started after switching VT away from X and back; I'm not quite sure. Bizarrely, this is visible in a screenshot (I'm not using compiz or anything fancy like that), which I'll attach to this bug.

Downgrading to 2:2.7.1-1ubuntu1 puts everything back to normal.

[Bryce Harrington (bryce)]

Created an attachment (id=26935)
BootDmesg.txt

[Bryce Harrington (bryce)]

Created an attachment (id=26936)
CurrentDmesg.txt

[Bryce Harrington (bryce)]

Created an attachment (id=26937)
XorgLog.gz

[Eric Anholt (eric-anholt)]

Tried pterm here, and it's slow because it's using core text (lame software), but not outrageously slow like you describe.

From the dmesg, your kernel config appears to be broken and including the DRM built in, but not AGP support. Sadly, Kconfig appears to allow this at the moment.

[Bryce Harrington (bryce)]

Hi Colin,

I've upstreamed your bug to https://bugs.freedesktop.org/show_bug.cgi?id=22358 - please subscribe to this bug in case upstream needs further information or wishes you to test something. Thanks ahead of time.

[Bryce Harrington (bryce)]

This seems to be the relevant issue (see comment on upstream bug)

[ 1.825763] [drm:drm_fill_in_dev] *ERROR* Cannot initialize the agpgart module.

[Colin Watson (cjwatson)]

You might be using the last officially-released version of pterm. It does have support for client-side fonts in the version I'm using (from svn; they need to get round to making a release); I'm just used to server-side fonts and haven't got round to changing over yet.

If missing agpgart is the problem, then I suspect that it might actually be more directly the fault of my /etc/modules, which had 'i915 modeset=1' but no drm or agpgart. I've adjusted this per some more current directions in https://wiki.ubuntu.com/X/KernelModeSetting, and will give that a go at my next reboot. Thanks!

[Colin Watson (cjwatson)]

Making sure that intel_agp and drm (thereby agpgart) are loaded before i915 definitely seems to have cleared this up so far. Thanks for the suggestion.

[Colin Watson (cjwatson)]

On Fri, Jun 19, 2009 at 01:44:52AM -0000, Bryce Harrington wrote:
> - uses 100% CPU after upgrade from 2.7.1
> + Sluggish text scrolling, 100%CPU after upgrade from 2.7.1 (UXA bug)

I don't know if this is relevant, but I'm not sure I see how this is a
UXA bug - as I said in my original report, the same thing happens with
Option "AccelMethod" "EXA". Or has EXA support been removed?

[Bryce Harrington (bryce)]

Yes, EXA support has been removed as of 2.7.99.x, and specifying that in
xorg.conf has no effect.

The main reason I included "(UXA bug)" is so it would show up on my
weekly UXA bug list report I send to Intel for bug escalation purposes.

Bryce

On Fri, Jun 19, 2009 at 11:17:06AM -0000, Colin Watson wrote:
> On Fri, Jun 19, 2009 at 01:44:52AM -0000, Bryce Harrington wrote:
> > - uses 100% CPU after upgrade from 2.7.1
> > + Sluggish text scrolling, 100%CPU after upgrade from 2.7.1 (UXA bug)
>
> I don't know if this is relevant, but I'm not sure I see how this is a
> UXA bug - as I said in my original report, the same thing happens with
> Option "AccelMethod" "EXA". Or has EXA support been removed?
>
> --
> Sluggish text scrolling, 100%CPU after upgrade from 2.7.1 (UXA bug)
> https://bugs.launchpad.net/bugs/388368
> You received this bug notification because you are a direct subscriber
> of the bug.
>
> Status in X.org xf86-video-intel: Confirmed
> Status in ???xserver-xorg-video-intel??? source package in Ubuntu: Triaged
>
> Bug description:
> Binary package hint: xserver-xorg-video-intel
>
> Versions of the Intel driver later than 2:2.7.1-1ubuntu1 are unusably slow for me. The first thing I do after starting X is normally to start a pterm and maximise it; the first sign of trouble is that the scrollbar takes ages to be replaced by normal black terminal area on maximisation. After that, Xorg is using nearly 100% CPU, text scrolling is very slow, etc. Typing this report is excruciatingly painful because it can take up to a second for each character to appear, especially when the typeahead buffer gets at all full.
>
> EXA vs. UXA makes no obvious difference.
>
> There are some rather odd visual effects as well: lots of letters are the wrong colour. This may have only started after switching VT away from X and back; I'm not quite sure. Bizarrely, this is visible in a screenshot (I'm not using compiz or anything fancy like that), which I'll attach to this bug.
>
> Downgrading to 2:2.7.1-1ubuntu1 puts everything back to normal.
>
> ProblemType: Bug
> Architecture: i386
> Date: Wed Jun 17 10:47:29 2009
> DistroRelease: Ubuntu 9.10
> MachineType: Dell Inc. Latitude D830
> Package: xserver-xorg-video-intel 2:2.7.99.1+git20090602.ec2fde7c-0ubuntu2
> ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.30-9-generic root=UUID=6df978ed-c024-49d7-9157-bdd7bd05153d ro quiet splash
> ProcEnviron:
> LC_COLLATE=C
> PATH=(custom, user)
> LANG=en_GB.UTF-8
> SHELL=/bin/bash
> ProcVersionSignature: Ubuntu 2.6.30-9.10-generic
> RelatedPackageVersions:
> xserver-xorg 1:7.4~5ubuntu21
> libgl1-mesa-glx 7.4.1-1ubuntu2
> libdrm2 2.4.11-0ubuntu1
> xserver-xorg-video-intel 2:2.7.99.1+git20090602.ec2fde7c-0ubuntu2
> xserver-xorg-video-ati 1:6.12.2-2ubuntu1
> SourcePackage: xserver-xorg-video-intel
> Uname: Linux 2.6.30-9-generic i686
> dmi.bios.date: 06/07/2007
> dmi.bios.vendor: Dell Inc.
> dmi.bios.version: A02
> dmi.board.name: 0HN341
> dmi.board.vendor: Dell Inc.
> dmi.chassis.type: 8
> dmi.chassis.vendor: Dell Inc.
> dmi.modalias: dmi:bvnDellInc.:bvrA02:bd06/07/2007:svnDellInc.:pnLatitudeD830:pvr:rvnDellInc.:rn0HN341:rvr:cvnDellInc.:ct8:cvr:
> dmi.product.nam...

[Bryce Harrington (bryce)]

From the upstream bug it sounds like everything's working now? Was it purely just config issues, or are there any adjustments needed for X?

[Colin Watson (cjwatson)]

On Fri, Jun 19, 2009 at 06:27:14PM -0000, Bryce Harrington wrote:
> From the upstream bug it sounds like everything's working now? Was it
> purely just config issues, or are there any adjustments needed for X?

Well, from my point of view it was primarily a configuration issue; I've
been poking at initramfs-tools and usplash to see how to get them to
interact with KMS in a vaguely sensible way, and had loaded i915 without
also loading the AGP implementation for my hardware (intel_agp).

However, it does seem rather too easy to shoot yourself in the foot in
really subtle ways with this stuff right now. It seems to me that early
userspace that's trying to bring up KMS ahead of other module loading in
order that it can have graphics support in place from the beginning has
a bit of a hard time of it. As far as I can see, you need to load any
AGP driver that matches something in modules.alias (or I suppose you
could just try them all but that'll be slow), and then try the various
KMS-capable DRM drivers.

Thinking about it, while X is a rather spectacular victim of this, I
don't know if there's anything it can do about it itself. I assume that
it's just trying to innocently use the DRM driver, and it's really the
DRM driver that's got itself hideously confused?

[Bryce Harrington (bryce)]

On Fri, Jun 19, 2009 at 11:59:06PM -0000, Colin Watson wrote:
> Well, from my point of view it was primarily a configuration issue.
>
> However, it does seem rather too easy to shoot yourself in the foot in
> really subtle ways with this stuff right now.
>
> Thinking about it, while X is a rather spectacular victim of this, I
> don't know if there's anything it can do about it itself. I assume that
> it's just trying to innocently use the DRM driver, and it's really the
> DRM driver that's got itself hideously confused?

Alright well I guess we can close it as an X issue. Sounds like most of
the trouble occurs kernel-side, but you can let the kernel team know if
there's changes needed there.

On the foot shooting, I agree, it's probably going to take some time to
digest all these changes, and having so much of the code in the kernel
makes bug analysis a bit more intricate. But so far so good, at least
with Intel; we'll see what happens when we switch KMS on by default.

Bryce

[Eric Anholt (eric-anholt)]

commit 8c52da503b7e4cf961807f11824e3258ef9f7f1c
Author: Eric Anholt <email address hidden>
Date: Thu Jun 18 20:22:19 2009 -0700

    drm/i915: Add missing dependency on Intel AGP support.

    Users could accidentally enable AGP but not the Intel AGP support, and get
    a DRM that doesn't probe as a result.

    Bug #22358.

    Signed-off-by: Eric Anholt <email address hidden>