Bug 67261 - [mga g400] broken vga= processing disables (analog) video output when CONFIG_X86_SYSFB=yes
Summary: [mga g400] broken vga= processing disables (analog) video output when CONFIG_...
Status: NEW
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(Other) (show other bugs)
Hardware: i386 Linux
: P1 normal
Assignee: drivers_video-other
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-12-18 00:33 UTC by Felix Miata
Modified: 2013-12-21 00:27 UTC (History)
4 users (show)

See Also:
Kernel Version: 3.12.2-1.2 vanilla, 3.12.2-1.2 default, 3.13.rc3-5.1.g5b170d7 default
Subsystem:
Regression: Yes
Bisected commit-id:


Attachments
.tgz of dmesgs from 3.11.10 (OK) and 3.12.2.1 (no video) boots (22.24 KB, application/octet-stream)
2013-12-18 15:29 UTC, Felix Miata
Details

Description Felix Miata 2013-12-18 00:33:42 UTC
To reproduce:
1-use MGA gfxchip (KMS support unavailable)
2-include any of 791, 0x317, 794, 788, etc. appended to vga= on Grub Legacy cmdline
3-include 3 on cmdline
4-boot

Actual results:
1-screen quickly blanks and stays that way
2-unusable vttys
3-vtty mode other than 80x25 impossible on unattended boot

Expected results:
1-specified vga mode in effect
2-visible text on vttys

Notes:
1-works as expected with 3.11.6-4 default and 3.11.13.g137a69e default.
2-result of specifying vga=ask and selecting any of the common framebuffer modes works as expected
3-specifying vga=normal works as expected
4-same brokenness with or without any valid video= on cmdline
5-same brokenness with or without nomodeset on cmdline
6-Plymouth not installed
7-tested only using openSUSE 13.2M0 and BS rpms
Comment 1 Alan 2013-12-18 14:08:23 UTC
Can you attach a dmesg of the relevant good/bad boots. Want to know which video driver is actually being used here (and also if this is yet more simplefb breakage)
Comment 2 Felix Miata 2013-12-18 15:29:14 UTC
Created attachment 118911 [details]
.tgz of dmesgs from 3.11.10 (OK) and 3.12.2.1 (no video) boots

Dmesg from 3.12 made by typing on a black screen.

In 3.11, Hot Plug Controller Driver @0.652203 is followed by vesafb: mode 1024x768x16 @0.652354

In 3.12, Hot Plug Controller Driver @0.654668 is followed by intel_idle does not run @0.654786 then GHES: HEST is not enabled @0.654834, with no vesa string anywhere
Comment 3 David Herrmann 2013-12-19 10:31:08 UTC
You probably enabled CONFIG_X86_SYSFB. Please see the Kconfig description for detailed information on it.

We got a regression with driver transition with sysfbs. A patch is pending on LKML:
http://thread.gmane.org/gmane.linux.kernel/1616367/focus=1616583

Can you give it a try?
Comment 4 David Herrmann 2013-12-19 10:34:00 UTC
And please, if you enable CONFIG_X86_SYSFB (which is 'n' by default), also enable CONFIG_FB_SIMPLE. The help text clearly states that if sysfb is enabled, you no longer get efi/vesa-framebuffer devices but instead the architecture generic simple-framebuffer.

So if you really want vesafb, just set CONFIG_X86_SYSFB=n, otherwise, set it to y and enable simplefb.
Comment 5 Felix Miata 2013-12-19 13:42:50 UTC
I don't build, so can't test patches directly. The comment 0 MGA system also has Fedora 21 and 3.13.rc4 with CONFIG_X86_SYSFB unset and VGA= is working as expected. openSUSE 13.2's 3.12 & 3.13 have CONFIG_X86_SYSFB=y, where VGA= works OK with Intel (KMS) but not MGA (no KMS). It looks in https://bugzilla.novell.com/show_bug.cgi?id=855821 like CONFIG_X86_SYSFB will be disabled shortly if it hasn't been already in openSUSE kernels.
Comment 6 Felix Miata 2013-12-21 00:21:13 UTC
Most recently built openSUSE 13.2 3.12.5 and 3.13.rc7 kernels have CONFIG_X86_SYSFB unset, and VGA= is functioning normally using them.
Comment 7 Felix Miata 2013-12-21 00:27:51 UTC
s/3.13.rc7/3.13.0-rc3-7/

Note You need to log in before you can comment on or make changes to this bug.