3rd-party/linux-kernel-module-cheat
1
0
Fork 0
mirror of https://github.com/cirosantilli/linux-kernel-module-cheat.git synced 2026-01-28 04:24:26 +01:00
Code Issues Releases Wiki Activity

libdrm works on x86_64!

Browse Source
This commit is contained in:
Ciro Santilli
2018-06-01 07:56:27 +01:00
parent 3e626ccf8f
commit 93e383902e
5 changed files with 800 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

87
README.adoc
View File

@@ -388,20 +388,14 @@ flooding the screen with colors. See also: https://superuser.com/questions/22309
==== Graphic mode arm
....
./run -aa -x -- -vnc :0
./run -aa -x
....
and on another shell:
....
vinagre :5900
....
Outcome: you see the penguin, and some boot messages, but don't get a shell.
Outcome: a window opens, and you see the penguin, and some boot messages there, but don't get a shell there, only on the host terminal.
TODO:
* why is the SDL window not opening, which forces us to use VNC? https://lists.gnu.org/archive/html/qemu-discuss/2018-05/msg00037.html
* sometimes the SDL window was not opening, a workaround was to use VNC: https://lists.gnu.org/archive/html/qemu-discuss/2018-05/msg00037.html
* how to get a shell on that graphic window? I don't think `CONFIG_FB_CONSOLE` works since we are not using fbdev but <<drm>>. Maybe try: https://github.com/dvdhrm/kmscon and then obviously: https://github.com/robclark/kmscube
This relies on the QEMU CLI option:
@@ -425,7 +419,7 @@ It is also fun to pass the `-device` multiple times and see that multiple screen
./run -aA -x -- -device virtio-gpu-pci
....
Multiple `/dev/dri/cardN` are also created accordingly.
Multiple `/dev/dri/cardN` are also created accordingly. More details at: <<drm>>.
See also https://wiki.qemu.org/Documentation/Platforms/ARM (recently edited and corrected by yours truly... :-)).
@@ -3776,11 +3770,25 @@ Looks like a recompile is needed to modify the image...
=== DRM
TODO: get something working. Anything. How can an interface be so under-documented and under-exemplified? It boggles my mind.
DRM / DRI is the new interface that supersedes `fbdev`:
DRM / DRI is the new interface that supersedes `fbdev`.
....
./build -B 'BR2_PACKAGE_LIBDRM=y' -k
./run -F '/libdrm_modeset.out' -x
....
We set `CONFIG_DRM=y` on our default kernel configuration, and it creates:
Source: link:kernel_module/user/libdrm_modeset.c[]
Outcome: for a few seconds, the screen that contains the terminal gets taken over by changing colors of the rainbow.
TODO not working for `aarch64`, it takes over the screen for a few seconds and the kernel messages disappear, but the screen stays black all the time.
....
./build -B 'BR2_PACKAGE_LIBDRM=y' -k
./run -F '/libdrm_modeset.out' -x
....
We set `CONFIG_DRM=y` on our default kernel configuration, and it creates one device file for each display:
....
# ls -l /dev/dri
@@ -3791,44 +3799,19 @@ crw------- 1 root root 226, 0 May 28 09:41 card0
# ls /sys/module/drm /sys/module/drm_kms_helper/
....
Some failed attempts: libdrm:
Try creating new displays:
....
./build -B 'BR2_PACKAGE_LIBDRM=y'
wget -O kernel_module/user/modeset.c https://raw.githubusercontent.com/dvdhrm/docs/fad7c3203b14e67053e0fc41d8490138b8ff47dd/drm-howto/modeset.c
./build -k
./run -aA -x -- -device virtio-gpu-pci
....
fails with:
to see multiple `/dev/dri/cardN`, and then use a different display with:
....
In file included from modeset.c:40:0:
/home/ciro/bak/git/linux-kernel-module-cheat/out/x86_64/buildroot/host/x86_64-buildroot-linux-uclibc/sysroot/usr/include/xf86drm.h:40:17: fatal error: drm.h: No such file or directory
#include <drm.h>
^
./run -F '/libdrm_modeset.out' -x
....
Report: https://bugs.busybox.net/show_bug.cgi?id=11056
This file appears to come from the Linux kernel tree:
....
linux/include/uapi/drm/drm.h
....
e.g. for `drm_context_t`. It does appear under `out/*/build/linux-headers`, but does not show up on `target/`.
libdri2 attempt:
....
./build -B 'BR2_PACKABE_LIBDRI2=y'
wget -O kernel_module/user/dri2test.c https://raw.githubusercontent.com/robclark/libdri2/master/test/dri2test.c
./build -k
....
but then I noticed that that example requires multiple files, and I don't feel like integrating it into our build. And when I build it on Ubuntu 18.04 host, it does not generate any executable, so I'm confused.
Ubuntu 18.04 host attempts:
Ubuntu 18.04 host attempts with an NVIDIA graphics cards failed, it seems that NVIDIA takes over:
* https://github.com/dvdhrm/kmscon/issues/113 `driver does not support dumb buffers`
* https://github.com/Aetf/kmscon/issues/2#issuecomment-392484043 kmscon attempt
@@ -3847,6 +3830,22 @@ Bibliography:
* https://en.wikipedia.org/wiki/Mode_setting KMS
* https://github.com/dvdhrm/docs
==== libdri2
TODO get working.
Looks like a more raw alternative to libdrm:
....
./build -B 'BR2_PACKABE_LIBDRI2=y'
wget -O kernel_module/user/dri2test.c https://raw.githubusercontent.com/robclark/libdri2/master/test/dri2test.c
./build -k
....
but then I noticed that that example requires multiple files, and I don't feel like integrating it into our build.
When I build it on Ubuntu 18.04 host, it does not generate any executable, so I'm confused.
=== Linux kernel testing
https://stackoverflow.com/questions/3177338/how-is-the-linux-kernel-tested
@@ -5479,7 +5478,7 @@ Header only linear algebra library supported by Buildroot:
....
./build -B 'BR2_PACKAGE_EIGEN=y' -k
./run -F '/eigen.out;/eigen_svd.out'
./run -F '/eigen_hello.out;/eigen_svd.out'
....
Source: