[Letux-kernel] pyra + letux stretch lxde + kernel 4.20 = not enough memory

[H. Nikolaus Schaller]

Hi Andreas,

> Am 31.12.2018 um 16:37 schrieb Andreas Kemnade <andreas at kemnade.info>:
> 
> Hi,
> 
> now with the fixed display I played around with makesd and created a
> stretch lxde installation using makesd. After bootup I have a steady
> cursor in the upper left area of the display.

Hm. should be blinking IMHO.

> No X11.
> looking at Xorg.log
> reveals the following:
>    20.994] (EE) modeset(0): glamor initialization failed                      
> [    20.994] (II) modeset(0): ShadowFB: preferred NO, enabled NO                
> [    20.994] (II) modeset(0): Output DSI-1 has no monitor section               
> [    20.994] (II) modeset(0): Output HDMI-1 has no monitor section              
> [    20.995] (II) modeset(0): EDID for output DSI-1                             
> [    20.995] (II) modeset(0): Printing probed modes for output DSI-1            
> [    20.995] (II) modeset(0): Modeline "720x1280"x60.0   76.72  720 725 730 888 
> 1280 1330 1390 1440 (86.4 kHz eP)                                              
> [    20.995] (II) modeset(0): EDID for output HDMI-1                            
> [    20.995] (II) modeset(0): Output DSI-1 connected                            
> [    20.995] (II) modeset(0): Output HDMI-1 disconnected                        
> [    20.995] (II) modeset(0): Using exact sizes for initial modes               
> [    20.995] (II) modeset(0): Output DSI-1 using initial mode 720x1280 +0+0     
> [    20.995] (==) modeset(0): Using gamma correction (1.0, 1.0, 1.0)            
> [    20.995] (==) modeset(0): DPI set to (96, 96)  

looks good up to here...

>                             
> ...
> [    22.157] (EE) modeset(0): failed to set mode: No space left on device       

Oops.

Maybe you can first run the Xorg in strace to identify the ioctl (or however it
is done) that ends in this message and then add a printk to the memory size calculation.
To identify why it returns -ENOMEM.

> rmmod omapdrm ; modprobe omapdrm seems to solve the problem. So there
> is still something rotten in regards to the tiler.
> After that procedure, X11 is wrongly rotated, but that is fixable using
> xrandr.

Most likely the tiler memory size calculations for the rotated case are
not exactly and always correct. For Jessie with omapdrm they seem to fit.

Anyways, please cc: the pyra related findings to kernel at pyra-handheld.com
because some pyra developers are only reading there (because they seem to
have a narrower focus than we have here).

BR and a nice and happy transition to year++ to everybody,
Nikolaus