[09:42] <Quozl> Hello.  I'm trying to solve a radeon regression, and having trouble rebuilding the Ubuntu kernel on focal.  The build fails and mentions x86 but I'm on x86_64.  Log: http://dev.laptop.org/~quozl/z/1l9lyf.txt  I'm using https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel as instructions.
[22:13] <Quozl> Hello.  I'm trying to solve a radeon regression, and having trouble rebuilding the Ubuntu kernel on focal.  The build fails and mentions x86 but I'm on x86_64.  Log: http://dev.laptop.org/~quozl/z/1l9lyf.txt  I'm using https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel as instructions.
[22:14] <jeremy31> Quozl: Does the patch involve changing anything in the headers?
[22:15] <Quozl> jeremy31: no, just radeon_display.c deep down.
[22:15] <jeremy31> Quozl: I would just try to build the needed modules then
[22:16] <Quozl> hmm, i'd appreciate a pointer to how to do that.  i do recall a linux.conf.au talk about it ten years ago, but things will have changed.
[22:17] <jeremy31> Quozl: Just cd into the directory where the changes are made in terminal and do> make -C /lib/modules/$(uname -r)/build M=$(pwd) modules
[22:18] <jeremy31> Then copy the module to where the original is after renaming the original
[22:18] <Quozl> (for reference, the regression is incorrect pll1 setup for my system, possibly because of bad edid. https://gitlab.freedesktop.org/xorg/driver/xf86-video-ati/issues/164) 
[22:18] <Quozl> thanks.  i'll look into that.
[22:19] <jeremy31> Quozl: It is what I usually do for wireless changes I make.  I can't see building an entire kernel for a change to one module 
[22:23] <Quozl> jeremy31: oddly, it fails, not able to find an include file, as if it can't reach the existing file in the include/ directory of the kernel tree ...  http://dev.laptop.org/~quozl/y/1l9xs3.txt
[22:30]  * Quozl hops on his bike
[22:31] <jeremy31> Ok, do a> make -C /lib/modules/$(uname -r)/build M=$(pwd) clean
[22:32] <jeremy31> then try the make -C /lib/modules/$(uname -r)/build M=$(pwd) modules from /kernel/drivers/gpu/drm/
[23:18] <Quozl> jeremy31: thanks, a substantial improvement, i'll debug further why some things are undeclared ... http://dev.laptop.org/~quozl/z/1l9yjk.txt
[23:20] <jeremy31> Quozl: Might have to remove some things from the Makefile in the drm directory to ignore the i2c
[23:20] <jeremy31> or you might have to run the make command from /drivers/gpu
[23:22] <Quozl> clean, make modules in drivers/gpu, same undeclared.  did i do something wrong after copying /boot/config-$(uname -r) ?  make oldconfig asks questions, which is a surprise.
[23:23] <jeremy31> Quozl: I haven't copied any files from /boot/config or used make oldconfig when doing wifi drivers
[23:24] <Quozl> jeremy31: okay, thanks.  how have you configured when doing wifi drivers?
[23:24] <jeremy31> Did you clone from git.ubuntu.com or enable source code repos?
[23:25] <Quozl> a kernel repo clone with a remote git://kernel.ubuntu.com/ubuntu/ubuntu-focal.git and i'm on a branch of my own based on ubuntu-focal/master ... and i've "git clean" to get rid of everything else.  so in theory it's a clone from git.ubuntu.com
[23:26] <jeremy31> I normally enable the source code repos, and do> apt-get source linux-modules-extra-$(uname -r)
[23:26] <jeremy31> Then I make my changes and most times it works
[23:27] <Quozl> 'k, thanks.
[23:27] <Quozl> must mean there's something magic between a git clone and the source package.  i'll give it a try.
[23:33] <jeremy31> Quozl: I think your original problem with building an entire kernel was related to the directory name as it must have been an exact match to a currently installed version
[23:37] <Quozl> ah.
[23:48] <Quozl> jeremy31: certainly much better with the source package, but still an odd compilation failure, claim of a missing pblk-trace.h header file, yet it is certainly present; drivers/lightnvm/pblk-trace.h   log:http://dev.laptop.org/~quozl/z/1l9zCF.txt  
[23:49] <Quozl> it is not, however, present in /usr/src/linux-headers-5.8.0-43-generic ... hmm.
[23:53] <jeremy31> I don't know what to do to fix that, that must have been for an entire kernel build
[23:53] <Quozl> i had climbed the source tree each time i hit a snag.
[23:55] <jeremy31> Quozl: Any chance the fix is in the 5.10 kernels?
[23:56] <Quozl> i can test.  i've reviewed all changes since, and none have a commit message that explains the problem, so not likely.  the bugs in the trackers are still open.
[23:58] <jeremy31> Quozl: have a bug report filed?