i915 module with modeset=1 freezes machine (kernels 2.6.33 and 2.6.36)

i915 module with modeset=1 freezes machine (kernels 2.6.33 and 2.6.36)

[wzabolot@elektron.elka.pw.edu.pl]

Hi,

I have two machines with intel graphics chipsets.
Below is the output from the lspci:
1st machine:
00:02.0 VGA compatible controller: Intel Corporation 82865G Integrated 
Graphics Controller (rev 02)
2nd machine:
00:02.0 VGA compatible controller: Intel Corporation Mobile 945GM/GMS, 
943/940GML Express Integrated Graphics Controller (rev 03)

Both machines worked a few years with intel Xorg xserver, with 
/etc/modprobe.d/i915-kms.conf
containing the line "options i915 modeset=0"
Unfortunately last time Xorg removed support for usermode mode setting 
in intel xserver,
and I had to change the above line to "options i915 modeset=1".
After that, both machines during boot simply freeze (screen is blank, no 
access through ssh, so probably the machine is frozen completely).

I tried to recompile the kernel with support for Intel agpgart and i915 
compiled in the kernel
but then the machine freezes even faster (no boot messages appear, 
screen gets black almost immediately after loading of the kernel).

I've tried to investigate the logs, but no error messages are logged.

Well, there is one warning in the logs, but the system worked with that 
warning for at least one year and it didn't cause any additional 
problems...:
Oct 26 21:10:51 wzab kernel: Pid: 1010, comm: modprobe Not tainted 2.6.36 #1
Oct 26 21:10:51 wzab kernel: Call Trace:
Oct 26 21:10:51 wzab kernel:  [<c0133cb8>] ? warn_slowpath_common+0x78/0xb0
Oct 26 21:10:51 wzab kernel:  [<f8392c18>] ? 
intelfbhw_active_pipe+0x58/0x60 [intelfb]
Oct 26 21:10:51 wzab kernel:  [<f8392c18>] ? 
intelfbhw_active_pipe+0x58/0x60 [intelfb]
Oct 26 21:10:51 wzab kernel:  [<c0133d0b>] ? warn_slowpath_null+0x1b/0x20
Oct 26 21:10:51 wzab kernel:  [<f8392c18>] ? 
intelfbhw_active_pipe+0x58/0x60 [intelfb]
Oct 26 21:10:51 wzab kernel:  [<f83954e9>] ? 
intelfb_pci_register+0xc27/0xdc3 [intelfb]
Oct 26 21:10:51 wzab kernel:  [<f83948c2>] ? 
intelfb_pci_register+0x0/0xdc3 [intelfb]
Oct 26 21:10:51 wzab kernel:  [<c026e54b>] ? local_pci_probe+0xb/0x10
Oct 26 21:10:51 wzab kernel:  [<c026ed71>] ? pci_device_probe+0x61/0x80
Oct 26 21:10:51 wzab kernel:  [<c02ccfb5>] ? driver_probe_device+0x65/0x170
Oct 26 21:10:51 wzab kernel:  [<c026e6a0>] ? pci_match_device+0xa0/0xc0
Oct 26 21:10:51 wzab kernel:  [<c02cd139>] ? __driver_attach+0x79/0x80
Oct 26 21:10:51 wzab kernel:  [<c02cd0c0>] ? __driver_attach+0x0/0x80
Oct 26 21:10:51 wzab kernel:  [<c02cc8a2>] ? bus_for_each_dev+0x52/0x80
Oct 26 21:10:51 wzab kernel:  [<c02cce56>] ? driver_attach+0x16/0x20
Oct 26 21:10:51 wzab kernel:  [<c02cd0c0>] ? __driver_attach+0x0/0x80
Oct 26 21:10:51 wzab kernel:  [<c02cc2a7>] ? bus_add_driver+0x177/0x230
Oct 26 21:10:51 wzab kernel:  [<c026ecb0>] ? pci_device_remove+0x0/0x40
Oct 26 21:10:51 wzab kernel:  [<c02cd3c3>] ? driver_register+0x63/0x120
Oct 26 21:10:51 wzab kernel:  [<c026ef9d>] ? __pci_register_driver+0x3d/0xb0
Oct 26 21:10:51 wzab kernel:  [<c0101132>] ? do_one_initcall+0x32/0x160
Oct 26 21:10:51 wzab kernel:  [<c0197808>] ? __vunmap+0xa8/0xe0
Oct 26 21:10:51 wzab kernel:  [<f82ea000>] ? intelfb_init+0x0/0x42 [intelfb]
Oct 26 21:10:51 wzab kernel:  [<c0162893>] ? sys_init_module+0x123/0xfd0
Oct 26 21:10:51 wzab kernel:  [<c0102c10>] ? sysenter_do_call+0x12/0x26
Oct 26 21:10:51 wzab kernel: ---[ end trace ec82370e7b0b3abd ]---

More details - Re: i915 module with modeset=1 freezes machine (kernels 2.6.33 and 2.6.36)

[Wojtek Zabolotny]

I have tried to boot my desktop machine (chipset 82865G with agpgart and 
i915 compiled into kernel) with i915.modeset=0
In this case machine works, but X work only with vesa driver.

Browsing reports about similar problems I have found suggestion to use 
i915.modeset=1 i915.powersave=0
unfortunately this also leads to black screen and freeze of machine.

In my laptop machine (chipset 945GM with agpgart and intel_agp and i915 
compiled as modules), when I reload i915 with modeset=1 it causes 
immediate black screen and freeze (the value of powersave parameter does 
not change the behaviour of my system).

I have also found some info that the problem may be related to detection 
of lid switch (in laptop) - I tried to press and release the lid switch 
in my laptop machine - this switched the backlight on, but system was 
still frozen and unaccessible.

Re: More details - Re: i915 module with modeset=1 freezes machine (kernels 2.6.33 and 2.6.36)

[wzabolot@elektron.elka.pw.edu.pl]

The 82865G based system boots correctly with debian distribution kernel 
linux-image-2.6.32-5-686 .
I tried to copy settings related to DRM, framebuffers and other graphics 
related options from 2.6.32-5 to my self-compiled 2.6.36 kernel, but it 
still hangs.

SOLVED? Re: i915 module with modeset=1 freezes machine (kernels 2.6.33 and 2.6.36)

[wzabolot@elektron.elka.pw.edu.pl]

I ve set two options in my kernel configuration:
CONFIG_FRAMEBUFFER_CONSOLE=y
CONFIG_FRAMEBUFFER_CONSOLE_ROTATION=y

Now the system boots correctly with KMS switched on.

Kernel configuration should warn when KMS is switched on but framebuffer console is not?

[wzabolot@elektron.elka.pw.edu.pl]

To avoid the problems described in the above bug reports, maybe the kernel
configuration should check and warn when CONFIG_DRM_I915_KMS is set to "y"
while CONFIG_FRAMEBUFFER_CONSOLE is not set to "y"?

Re: Kernel configuration should warn when KMS is switched on but framebuffer console is not?

[Randy Dunlap]

On Sat, 30 Oct 2010 15:34:43 +0200 wzabolot@elektron.elka.pw.edu.pl wrote:

> To avoid the problems described in the above bug reports, maybe the kernel
> configuration should check and warn when CONFIG_DRM_I915_KMS is set to "y"
> while CONFIG_FRAMEBUFFER_CONSOLE is not set to "y"?

What "above bug reports" are you referring to, please?

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

Re: Kernel configuration should warn when KMS is switched on but framebuffer console is not?

[wzabolot@elektron.elka.pw.edu.pl]

W dniu 31.10.2010 18:29, Randy Dunlap pisze:
> On Sat, 30 Oct 2010 15:34:43 +0200 wzabolot@elektron.elka.pw.edu.pl wrote:
>
>    
>> To avoid the problems described in the above bug reports, maybe the kernel
>> configuration should check and warn when CONFIG_DRM_I915_KMS is set to "y"
>> while CONFIG_FRAMEBUFFER_CONSOLE is not set to "y"?
>>      
> What "above bug reports" are you referring to, please?
>
>    
I mean my report archived as http://lkml.org/lkml/2010/10/26/466
and following informations about attempts to resolve it (next messages 
in the thread).
-- 
Regards,
Wojtek Zabolotny