Bug 37545

Please attach your xorg log, xrandr output, and dmesg output.

Created attachment 47103 [details]
Xorg log after xrandr run

Created attachment 47104 [details]
dmesg

Created attachment 47105 [details]
xrandr ouput

The problem in this case is vram fragmentation.  When you start X, the driver allocates a buffer large enough for both dualhead (1024x768 + 1600x1200).  When you turn off the LVDS port, the driver tried to allocate a new 1600x1200 buffer for just the VGA monitor before freeing the previous dualhead allocation.  since there isn't enough VRAM available for both the dualhead allocation and the VGA only allocation, the modeset fails.

Does it work if you start X without tha VGA monitor connected?  Then connect it after you've started X, something like:
xrandr --output VGA-0 --off
xrandr --output VGA-0 --auto --pos 0x0
xrandr --output LVDS --off

That does succeed, although not really a workable solution.

Is there any way to tell X to not enable the LVDS output on startup?

Section "Monitor"
        Identifier "LVDS"
        Option "Disable" "TRUE"
EndSection

Yeah, but I'm assuming that would still deactivate the LVDS output even if it was the only output available.  I need to conditionally deactivate it if the lid is closed.

Created attachment 47201 [details] [review]
possible fix

Does this patch fix the issue?

No that doesn't help.  It appears that info->front_bo is non-null at that point, so your new code doesn't run.

Breakpoint 2, drmmode_xf86crtc_resize (scrn=0xa09a880, width=1600, height=1200)
    at drmmode_display.c:1288
1288            if (!info->front_bo) {
(gdb) print info->front_bo
$2 = (struct radeon_bo *) 0xa0a68d0

At this point the log has:
[   516.548] (II) RADEON(0): Allocate new frame buffer 1600x1200 stride 1600

The invalid argument error comes later.

The invalid argument error comes in this code:

1337            for (i = 0; i < xf86_config->num_crtc; i++) {
(gdb) 
1343                    drmmode_set_mode_major(crtc, &crtc->mode,

Interesting that at this point there is no fail handling - which must be why the VGA output dies and goes into power save.

Created attachment 47403 [details]
dmesg with drm.debug=4

Mass closure: This bug has been untouched for more than six years, and is not
obviously still valid. Please reopen this bug or file a new report if you continue to experience issues with current releases.