xserver-xorg-video-intel does not work with 845G

[Shofi Islam]

Binary package hint: xserver-xorg-video-intel

Package affected: xserver-xorg-video-intel
Kernel: 2.6.20-16-generic
Ubuntu version: Feisty (07.04)
CPU: Intel Celeron PIV (tualatin) 2.0GHz
Memory: 512MB

I have an i-friend nc202i4 laptop with an onboard intel 845G graphics chipset. The xserver-xorg-video-i810 driver works providing both 2D and limited 3D performance, but I believe this is being phased out in preference for xserver-xorg-video-intel. To test, I intalled it and rebooted my laptop. However X fails with the resulting xserver output stating another chipset was found. I have altered my xorg.conf changing the video driver from i810 to intel with no joy.

To recover, I reinstall the i810 driver and set xorg.conf to use i180. Now everything is working again.

I've included the xorg.conf and Xorg.0.log files used during the test. Please let me know if you need anymore information.

[Shofi Islam (shofi-islam)]

Binary package hint: xserver-xorg-video-intel

Package affected: xserver-xorg-video-intel
Kernel: 2.6.20-16-generic
Ubuntu version: Feisty (07.04)
CPU: Intel Celeron PIV (tualatin) 2.0GHz
Memory: 512MB

I have an i-friend nc202i4 laptop with an onboard intel 845G graphics chipset. The xserver-xorg-video-i810 driver works providing both 2D and limited 3D performance, but I believe this is being phased out in preference for xserver-xorg-video-intel. To test, I intalled it and rebooted my laptop. However X fails with the resulting xserver output stating another chipset was found. I have altered my xorg.conf changing the video driver from i810 to intel with no joy.

To recover, I reinstall the i810 driver and set xorg.conf to use i180. Now everything is working again.

I've included the xorg.conf and Xorg.0.log files used during the test. Please let me know if you need anymore information.

[Timo Aaltonen (tjaalton)]

Try commenting out the Options for the driver.

Also, AIGLX and Composite are already enabled in the server, so no need to specify them in the config.

[Shofi Islam (shofi-islam)]

Done as requested, but same error occurs. According to /var/log/Xorg.0.log, driver thinks I have a SIL card installed and mentions not enough RAM. I have 512MB installed, but no BIOS option to directly reserve RAM for graphics.

Is there anything else I can do to provide more information? Is a trace-enabled driver available.

Please Advise.

[Aaron Whitehouse (aaron-whitehouse)]

It looks like everything asked for is there.

[Gordon Jin (gordon-jin)]

Are you still seeing this problem with the latest driver? Please attach xorg.conf and Xorg.0.log.

[Shofi Islam (shofi-islam)]

The intel driver also doesn't work on the Gutsy live CD and the Gutsy alternate install. Work around is to edit xorg.conf changing the driver from 'intel' to 'i810'.

Please advise.

[Peter Clifton (pcjc2)]

Can you try removing the VideoRAM line from the xorg.conf and re-try with the intel driver please.

Can you post the xorg.log from it not working with the Live CD pelase, does this fail in the same way?

Ignore the errores about sil for now, the sil164 is a laptop LCD driver chip which may / may not be used in conjunction with the 845 graphics chip.

You might like to have a look in the BIOS settings and see if there are any options relating to how much video memory is allocated.

[Bryce Harrington (bryce)]

These errors appear in the Xorg.0.log; sounds like VideoRAM is set incorrectly; Peter's right - you should test with the VideoRAM line removed (just let Xorg figure it out itself).

(II) intel(0): Allocating 5273 scanlines for pixmap cache
(EE) intel(0): Failed to allocate framebuffer. Is your VideoRAM set too low?
(WW) intel(0): Not enough video memory. Disabling DRI.
(II) intel(0): Allocating 5273 scanlines for pixmap cache
(EE) intel(0): Failed to allocate framebuffer. Is your VideoRAM set too low?
(EE) intel(0): Couldn't allocate video memory

[Shofi Islam (shofi-islam)]

Have commented out the VIDEORAM line and the other options with no success. I performed 2 tests, 1 using the i810 module in xorg.conf and then using the intel module in xorg.conf.

Please see attached xorg.conf, Xorg.0.log, and dmesg output for each test.

[Shofi Islam (shofi-islam)]

2nd test, commenting out VIDEORAM line and using intel module instead of i810.

[Shofi Islam (shofi-islam)]

Test repeated on Gutsy with same result. Xorg.conf used and the resulting Xorg.0.log included.

[unggnu (unggnu)]

I can confirm this with Gutsy and -intel driver 2:2.1.1-0ubuntu9.

[unggnu (unggnu)]

I can confirm this. It doesn't work with intel driver and it only works with i810 driver with 640x480.
Even the standard Gutsy xorg.conf doesn't work with i810 since there is no 640x480 listed by default afaik.

[unggnu (unggnu)]

Btw. with new Intel driver the monitor only flickers very often but never show a desktop. Live CD boots fine with Feisty because of i810 driver but only with 640x480.

[unggnu (unggnu)]

Usplash seems to work fine with 1024x768 in full screen.

[unggnu (unggnu)]

1024x768 works fine with i810 driver if I add "HorizSync 40-51" to Monitor section. Seems to be a bug of the Gericom Hummer 2430 Notebook.

[unggnu (unggnu)]

Created an attachment (id=12331)
Xorg.0.log with new Intel driver

[unggnu (unggnu)]

Created an attachment (id=12332)
lspci -vvnn

[unggnu (unggnu)]

Created an attachment (id=12333)
xorg.conf

[unggnu (unggnu)]

Another interesting thing with the 845g chipset and Gericom Hummer 2430. If you start the laptop the first time without a battery it boots fine until X but after X start the mouse is watchable for some seconds, then disappears and the whole laptop hangs. After a hard reboot X starts fine.
It seems to have something to do with direct rendering or something like that. The i810 driver is used because of this bug.

[Bryce Harrington (bryce)]

Subscribing Intel Hardware Certification to help push this forward.

[unggnu (unggnu)]

The DDC problem seems to be known for this chipset (Intel Corporation 82845G/GL[Brookdale-G]/GE/PE) according to http://ubuntuforums.org/showthread.php?p=1900323 and Bug #29436 so it doesn't seem to be laptop specific.

[unggnu (unggnu)]

Btw. the 640x480 problem could be fixed since Gutsy with "Screen and Graphics" through choosing the Generic 1024x768 LCD Panel.

[Hong-liu (hong-liu)]

Would you please try the following patch to see if the problem is fixed?

Thanks,
Hong

diff --git a/src/i830_crt.c b/src/i830_crt.c
index d7762a0..7def610 100644
--- a/src/i830_crt.c
+++ b/src/i830_crt.c
@@ -162,9 +162,12 @@ i830_crt_detect_hotplug(xf86OutputPtr output)
     const int timeout_ms = 1000;
     int starttime, curtime;

- temp = INREG(PORT_HOTPLUG_EN);
+ /* clear the CRT hotplug status bits first */
+ OUTREG(PORT_HOTPLUG_STAT, CRT_HOTPLUG_MONITOR_MASK);

- OUTREG(PORT_HOTPLUG_EN, temp | CRT_HOTPLUG_FORCE_DETECT | (1 << 5));
+ temp = INREG(PORT_HOTPLUG_EN);
+ OUTREG(PORT_HOTPLUG_EN,
+ temp | CRT_HOTPLUG_FORCE_DETECT | (1 << 5) | CRT_HOTPLUG_INT_EN);

     for (curtime = starttime = GetTimeInMillis();
         (curtime - starttime) < timeout_ms; curtime = GetTimeInMillis())

[unggnu (unggnu)]

Created an attachment (id=12687)
Xorg.0.log witch patched Ubuntu Gutsy Gibbon Intel driver after xorg.conf removal

Many thanks for your patch but problem still appears.
Still gdm starts with 640x480 instead of 1024x768.
If I use my normal xorg.conf and replace i810 with intel X crashes. I am going to attach a gdb report soon.

[Hong-liu (hong-liu)]

(In reply to comment #7)
> Created an attachment (id=12687) [details]
> Xorg.0.log witch patched Ubuntu Gutsy Gibbon Intel driver after xorg.conf
> removal
>
> Many thanks for your patch but problem still appears.

Sorry, the patch only works on new chips (945 and 965..).
It seems we can't detect the connected monitors, will work on this to see if we can find any solution...

> Still gdm starts with 640x480 instead of 1024x768.
> If I use my normal xorg.conf and replace i810 with intel X crashes. I am going
> to attach a gdb report soon.
>

For i810 driver, it filters out the 1024x768 mode because of the hysnc range:

(II) I810(0): Builtin Default Monitor: Using default hsync range of 31.50-37.90 kHz

You can try to add HorizSync to your monitor section to enlarge the hysnc range.

Thanks,
Hong

[unggnu (unggnu)]

Created an attachment (id=12696)
Xorg.0.log with patched Ubuntu Intel driver

[unggnu (unggnu)]

Created an attachment (id=12697)
xorg.conf

[unggnu (unggnu)]

Created an attachment (id=12698)
Xorg.0.log with i810 driver (HorizSync have to been activated in xorg.conf to get 1024x768)

[unggnu (unggnu)]

How it is possible to get a gdb backtrace of Xorg which crashes directly after start so it is not possible to get a pid and attach it?
But according to log it only seems to be a ddc/resolution since it shutdown normally.

[unggnu (unggnu)]

Thanks for your reply. Manual adding of HorizSync only works with i810 driver. The Intel driver seems to ignore these because of modesetting I guess.

[Hong-liu (hong-liu)]

(In reply to comment #13)
> Thanks for your reply. Manual adding of HorizSync only works with i810 driver.
> The Intel driver seems to ignore these because of modesetting I guess.
>

Yes, you have to stick with the i810 driver currently :(

With the Intel driver, we can't detect the connected CRT monitor:
(II) intel(0): Output VGA disconnected
(II) intel(0): EDID for output VGA
(II) intel(0): Output VGA disconnected
(EE) intel(0): No valid modes.

So X server exits without success. We need to figure out a better solution for monitor detection for this chipset.

[Hong-liu (hong-liu)]

Created an attachment (id=12731)
2nd try

Please have a try on the attached patch.

Thanks,
Hong

[unggnu (unggnu)]

Created an attachment (id=12763)
Xorg.0.log.old with 2nd try patched Ubuntu Intel driver

[unggnu (unggnu)]

Many thanks for your patch but I got still the same or a similar error.

[unggnu (unggnu)]

Btw. is HorizSync needed with your 2nd try Patch?

[Hong-liu (hong-liu)]

(In reply to comment #18)
> Btw. is HorizSync needed with your 2nd try Patch?
>

Should not, I am not sure why we can't correctly detect your VGA connection status. This caused your problem.

Thanks,
Hong

[unggnu (unggnu)]

It seems to be a general problem with the 845 chipset since there are some duplicates for and posts under the Ubuntu bug report https://bugs.launchpad.net/bugs/135093 .

[unggnu (unggnu)]

I don't know if there are more but all of the reporters seem to have the Brookdale-G version.

[Bryce Harrington (bryce)]

Matthew Garrett reports that -intel does not support chipsets older than 855. "Before 855, it's possible to attach basically arbitrary LVDS encoders to the chipset - the BIOS knows how to talk to them, but -intel doesn't use the BIOS so needs to have code for each of them individually. As far as I know, the entire set isn't currently supported."

So if this is the case, we need to know if -i810 should be retained permanently for use with older chipsets, or if there are any plans/efforts to incorporate the code into -intel.

[Bryce Harrington (bryce)]

I suspect the core issue may be that chipsets older than 855 are not supported on -intel yet (see my reply on the upstream bug). I think we need direction from upstream regarding which direction to go (continue to rely on -i810 in this case, or learn if support in -intel is desired/planned).

[wasim2050 (mukadam-wasim)]

My Motherboard is with Intel 845 G and my problem is little strange then others. If I use i810 driver then everything is fine but I do not get my desired resolution so when I install intel driver I get my desired resolution but whenever I start the PC I am greeted with a low resolution login screen so when I login the resolution automatically jumps from low resolution to high resolution and sometimes when this happens the screen goes blank I have no choice but to login in the terminal mode and restart the PC.

[Eric Anholt (eric-anholt)]

No, Bryce. The released packages have supported the ivch which is in every 830 laptop we've seen so far. The 845 laptops almost all have ch7017, which is also supported at least upstream (not sure if it was in 2.1 or not). The remaining chip, which we've seen one instance of, is a natsemi device which we can't find specs for. And the bug you're commenting on appears to be about a desktop anyway.

unggnu: please open your own bug for your own issue, and include Option "ModeDebug" "YES" in your device section. No additional patches.

Original submitter: Could we get your log with option "ModeDebug" "YES"?

[Rolla Selbak (rolla-n-selbak)]

Just a quick note about the upstream direction for Intel gfx...

The -intel driver has support for i810 and forward chips. Some laptops with i845 chips will need a fairly recent driver update to get support for their panels, but at least some people have reported that these work today.

There are no plans to keep up with -i810. All pre-855 support should be present in -intel (and therefore any bugs found should be filed against -intel to get correct attention).

[Shofi Islam (shofi-islam)]

Created an attachment (id=13414)
Testcase with MODE DEBUG

Test case with the MODE DEBUG option.

[Michael Fu (michael-fu-intel)]

Created an attachment (id=13564)
xorg.0.log with modedebug turned on.

extracted from comment# 24.

[Hong-liu (hong-liu)]

(In reply to comment #24)
> Created an attachment (id=13414) [details]
> Testcase with MODE DEBUG
>
> Test case with the MODE DEBUG option.
>

Would you please use the intel driver instead of the vesa driver?

Thanks,
Hong

[Shofi Islam (shofi-islam)]

Created an attachment (id=13972)
Xorg.conf used for MODEDEBUG test

In response to comment 25. Test case is with Intel driver, not VESA. However Xorg is reporting VESA for some reason. Will repeat test and publish results

[Shofi Islam (shofi-islam)]

Created an attachment (id=13973)
Xorg log with MODEBUG and xorg.conf above

[Shofi Islam (shofi-islam)]

Created an attachment (id=13974)
13972: Xorg.conf used for MODEDEBUG test

[Shofi Islam (shofi-islam)]

Created an attachment (id=13975)
13972: Xorg.conf used for MODEDEBUG test

[Shofi Islam (shofi-islam)]

Created an attachment (id=13976)
Xorg log with MODEBUG and xorg.conf above

In response to comment 26, the Intel driver is being used despite the VESA messages shown. I have attached the xorg.conf and xorg.0.log.

[Shofi Islam (shofi-islam)]

Created an attachment (id=13977)
lspci output from above test

Note Unknown device mentioned at line 9

[Shofi Islam (shofi-islam)]

Created an attachment (id=13978)
ddcprobe

[Michael Fu (michael-fu-intel)]

(In reply to comment #31)
> Created an attachment (id=13976) [details]
> Xorg log with MODEBUG and xorg.conf above
>
> In response to comment 26, the Intel driver is being used despite the VESA
> messages shown. I have attached the xorg.conf and xorg.0.log.
>

Weird... it shouldn't be...

[Shofi Islam (shofi-islam)]

Created an attachment (id=14076)
Updated Xorg.0.log with MODEDEBUG and xorg.conf

Refreshed Xorg.conf, no changes made but VESA messages gone replaced with intel error messages.

[Michael Fu (michael-fu-intel)]

Shofi, would you please also attach you kernel dmesg?

[Hong-liu (hong-liu)]

For all who have problem with 845G, you are saying a 845G laptop?
If so, it seems I have made a big mistake for I am trying to find out why VGA can't be detected while it is not connected.

For all who have i810 work on your machine, would you please try to dump the registers when it is working. The tool is in the xf86-video-intel/src/reg_dump, you will need libpciaccess to compile this tool.

Thanks,
Hong

[Hong-liu (hong-liu)]

It seems 845G laptop uses ch7017 module to drive the LVDS output.
But I didn't see any ch7017 related info in your xorg log.

Would you please ensure that ch7017 module is existed in your distribution (usually it is on /usr/lib/xorg/modules/drivers/)?

Thanks,
Hong

[Shofi Islam (shofi-islam)]

Created an attachment (id=14116)
dmesg log as requested

dmesg log as requested

[Shofi Islam (shofi-islam)]

Created an attachment (id=14117)
location of module ch7017

Module ch7017 present within filesystem, but is not visible using lsmod. I cannot insert it using modprobe or insmod, but I assume that is because I cannot use these commands.

[Hong-liu (hong-liu)]

(In reply to comment #40)
> Created an attachment (id=14117) [details]
> location of module ch7017
>
> Module ch7017 present within filesystem, but is not visible using lsmod. I
> cannot insert it using modprobe or insmod, but I assume that is because I
> cannot use these commands.
>

Sorry for the confusion, ch7017 is a module for X server (not for kernel).
Our drive will try all the compiled modules one by one to try to detect which
output is active.

f.e this is when we are trying the sil164 module (in xorg log).
(II) Loading sub module "sil164"
(II) LoadModule: "sil164"
(II) Loading /usr/lib/xorg/modules/drivers//sil164.so
(II) Module sil164: vendor="X.Org Foundation"
 compiled for 1.3.0, module version = 1.0.0
 ABI class: X.Org Video Driver, version 1.2
(II) intel(0): I2C bus "DVOI2C_E" initialized.
(EE) intel(0): detecting sil164
(EE) intel(0): Unable to read from DVOI2C_E Slave 112.

We have an array in i830_dvo.c file (i830_dvo_drivers[]), and we will try one by one until we find one active output.
The problem is in your log, the last module we tried is tfp410 (which is not active), but the driver didn't try ch7017 module which is after the tfp410.

Which driver version are you using? Would you please give the latest driver a try?

Or would you please try to move the following block in i830_dvo_drivers[] to a front position to have a try.
   {
        .type = I830_OUTPUT_DVO_LVDS,
        .modulename = "ch7017",
        .fntablename = "ch7017_methods",
        .dvo_reg = DVOC,
        .address = 0xea,
        .symbols = ch7017_symbols,
        .gpio = GPIOE,
    }

Thanks,
Hong

[Shofi Islam (shofi-islam)]

Created an attachment (id=14140)
i830_drv.c from ubntu Gutsy (2:2.1.1-0ubuntu9)

Am using Ubuntu Gutsy with driver version 2:2.1.1-0ubuntu9(gutsy). This has been raised as a downstream bug in https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/135093

The code block mentioned previously does not appear in the file; it only references a module called ch7xxx. Would you review the module source (attached) and comment if this is correct.

[Hong-liu (hong-liu)]

(In reply to comment #42)
> Created an attachment (id=14140) [details]
> i830_drv.c from ubntu Gutsy (2:2.1.1-0ubuntu9)
>
The ch7017 related code has been commented out. Would you please try the upstream driver. You can get it by
$ git-clone git://anongit.freedesktop.org/git/xorg/driver/xf86-video-intel

Thanks,
Hong

[Timo Aaltonen (tjaalton)]

Could someone try Hardy alpha4 livecd to see if the issue with 845G is fixed? Testing this is very important, since this bug has five dupes by now and marked as high priority.

[Timo Aaltonen (tjaalton)]

Note that the driver in alpha4 has all upstream commits from the stable branch up to commit 103dc34f3afc17cea7e086927e124a30c605a686 which is ~week old.

[Shofi Islam (shofi-islam)]

Created an attachment (id=14179)
Xorg.0.log using recompiled driver

The latest driver works, but all 3D ability is now gone. glxinfo reports no direct rendering.

[Shofi Islam (shofi-islam)]

Created an attachment (id=14180)
glxinfo

libGL error: drmMap of framebuffer failed (Invalid argument)
libGL error: reverting to (slow) indirect rendering

Apologies if it seems I'm spamming this bug. I'm keen to help fix this.

[Shofi Islam (shofi-islam)]

I've recompiled the latest driver from git (see upstream bug) and the driver now works. Looks like some code relating to the ch7017 chipset was removed in the gutsy version of the driver.

Although the recompiled driver works, 3D is not available. Hopefully this can be fixed soon.

I will try Hardy Alpha 4 over the weekend.

[Shofi Islam (shofi-islam)]

Tested liveCD of Ubuntu Hardy alpha 4 with the latest drivers. Intel driver now working with ch7017 chipset and acceleration.

[Shofi Islam (shofi-islam)]

Live CD of Hardy alpha 4 works. Xorg reports intel driver being used and the ch7017 chipset module present. Direct rendering active.

However xorg appears to fail occasionally after resuming from screensaver. Not sure if this is related or a new bug to look into.

Would an updated driver for Feisty (or a backport) be the solution for this bug?

[Gordon Jin (gordon-jin)]

Good to know this works now. So I'm closing this one.

[unggnu (unggnu)]

@Shofi Islam
Your new issue might be related to bug #160309.
I guess you mean Gutsy instead of Feisty.

It would be great if an updated driver would at least make it to Gutsy proposed so people which are affected could fix this easily.

[Shofi Islam (shofi-islam)]

Confirm I mean Gusty...and Hardy eventually.

Upstream have marked their bug as fixed so it's up to the Devs now.

[Timo Aaltonen (tjaalton)]

Closing as fixed in Hardy. Thank you for testing.

[Dan Allen (dan.j.allen)]

I can confirm that the display now renders properly in Hardy.

Sorry for the delay...I finally got my hands on the computer again and was able to test it out.

The support for older graphics cards has an important impact on adoption since those trying out Ubuntu tend to try it out first on "an old computer just lying around". When it doesn't work then, they lost faith in it...especially since display issues prevent them from even getting to the desktop.

We should feel good about the fact that Hardy will impress.