<_florent_>
bl0x_: Thanks, that's indeed probably a copy&paste leftover that I missed when merging the cmod a7 support, I'll merge your PR
<_florent_>
david-sawatzke[m: Thanks, interesting. I'll havea closer look at what is does. I'm indeed not sure if this could be integrated, but we could definitely have a link to it in the wiki
_whitelogger has joined #litex
gatecat_ is now known as gatecat
nats` has joined #litex
rektide_ has quit [Ping timeout: 256 seconds]
rektide has joined #litex
linear_cannon has quit [Ping timeout: 256 seconds]
linear_cannon has joined #litex
gruetze_ is now known as gruetzkopf
trabucay1e is now known as trabucayre
cr1901_ is now known as cr1901
<Peanut>
When trying to build the literoot for linux-on-litex-vexriscv, I always end up with "Incorrect selection of kernel headers: expected 5.16.x, got 5.14.x" - which is doubly puzzling, because the BR2_LINUX_KERNEL_CUSTOM_REPO_VERSION points to a 5.13 kernel, according to the git commit stamp.
<_florent_>
Peanut: This need to be updated, could you send a PR if you have the updated sha1?
<Peanut>
_florent_: I'm not quite at the level that I understand how to fix this. I -think- that the sha1 points to a particular commit on the litex-rebase branch of the kernel. However, if I try e.g. the current commit, it fails to build the kernel due to an error in drivers/fpga/litex-fpga.c
<Peanut>
In line 168, devm_fpga_mgr_create() is not known when building.
<_florent_>
Peanut: ok, I'll need to rebuild the kernel and update. I'll look at this in the next days
<Peanut>
Thanks
<Peanut>
Should I open an issue?
<geertu>
Peanut: Commit 4ba0b2c294fe6919 ("fpga: mgr: Use standard dev_release for class driver") didn't update drivers/fpga/litex-fpga.c (it's not upstream), and somlo forgot to update it when rebasing?
<geertu>
somlo: ^
<geertu>
Peanut: I always build my Linux images from a separate Linux source tree (i.e. linux-on-litex-vexriscv/images/Image is a symlink to /path/to/linux-riscv-orangecrab/arch/riscv/boot/Image)
cr1901 has joined #litex
<Peanut>
geertu: thanks for that. I can see in the commit that you found that devm_fpga_mgr_create() gets removed/renamed, so that matches the error I'm seeing.
<somlo>
(I don't think I "lost" anything during rebase, so my guess is it was never part of the stuff I'm keeping track of -- but I might be wrong, happens fairly regularly :)
peepsalot has quit [Quit: Connection reset by peep]
<tpb>
Title: Config.in.host « linux-headers « package - buildroot - Buildroot: Making Embedded Linux easy (at git.buildroot.net)
<Peanut>
trabucayre: That particular version doesn't exist in the file you linked?
<trabucayre>
use BR2_KERNEL_HEADERS_VERSION and set BR2_DEFAULT_KERNEL_VERSION with something like 5.14.0
<trabucayre>
or BR2_KERNEL_HEADERS_AS_KERNEL=y
<Peanut>
That last option seems the most convenient, testing that first..
FabM has quit [Ping timeout: 268 seconds]
zjason`` has joined #litex
zjason` has quit [Ping timeout: 240 seconds]
<trabucayre>
yep
<Peanut>
Took a while, but with BR2_LINUX_KERNEL_CUSTOM_REPO_VERSION="53b46d10f9a438a29c061cac05fb250568d1d21b" and BR2_KERNEL_HEADERS_AS_KERNEL=y, it still results in "Incorrect selection of kernel headers: expected 5.16.x, got 5.14.x"
<geertu>
somlo: You have to make a similar change to drivers/fpga/litex-fpga.c
<trabucayre>
Peanut: you have to do 'make clean && make'
<trabucayre>
when you change something in toolchain you have to do a full rebuild
<somlo>
geertu: similar to what? also, I drivers/fpga/litex-fpga.c (driver for ICAPBitstream fpga manager) isn't something I actually use with Rocket, so while I'm happy to apply patches on top of it in the "curated" tree, I am not set up to test them myself
<Peanut>
Yes, of course I did. I did 'make clean', 'make BR2_EXTERNAL=' and then 'make'.
<somlo>
this is a vexriscv 32-bit thing I don't deal with directly on my end
<somlo>
"patches welcome" and all that happy stuff :)
<trabucayre>
you have modified your defconfig or your current config? If defconfig you have to do 'make xxx_defconfig' again
<Peanut>
I made the change in linux-on-litex-rexriscv/buildroot/configs/litex_vexriscv_defconfig, then in buildroot did 'make clean', 'make BR2_EXTERNAL=../linux-on-litex-vexriscv/buildroot litex_vexriscv_defconfig', then 'make'.
<trabucayre>
looks good
<trabucayre>
you have BR2_DEFAULT_KERNEL_VERSION="5.14.0" or something like this?
<Peanut>
No, I don't. (I did a test with it earlier, but it's properly commented out).
<somlo>
geertu: if I change something in https://github.com/litex-hub/linux/tree/litex-rebase I would prefer to either 1. build it and test it myself (I can only do that easily on rocket), or 2. to have it come as a pre-tested patch from someone I trust (e.g., you :) ), in case it's specific to vexriscv or mor1k or whatever