You are not logged in.
I'm running the kernel in [testing], and the system does boot correctly, and everything is fine, but on booting, after running uevents, this message is spitted out.
liblpf: failed to find valid kernel BTF
liblpf: Error loading vmlinux BTF -3
liblpf: failed to load object 'iterators_bpf': -3
Failed load could be due to wrong endianess
The same thing occurs when I use the non-testing linux-zen kernel.
I'm not sure what is happening here, is this expected?
Nothing else is wrong with the system.
Last edited by mistersmee (2021-02-03 13:26:37)
Offline
Sounds pretty similar:
https://bbs.archlinux.org/viewtopic.php?id=262019
The other one sounds funny:
https://github.com/iovisor/bcc/issues/2872
Why should a debug utility change the way a kernel is compiled, but yeah, who knows, lately kernel guys have really strange ideas. :-)
Offline
I see.
While not harmful to the system, these error messages pop up.
They go away by disabling
CONFIG_BPF_PRELOAD
in the kernel config.
And looking at the commit log here, we disabled BPF on the 5.9 version of the kernel.
On the archlinux forum thread link you gave, someone mentioned that even though BPF was disabled, 5.9 did not spit out this message, while 5.10 started to.
So disabling the option I mentioned should do the trick.
This applies for any kernel version above 5.9, i.e linux-zen.
linux-pae is fine, the option I mentioned is disabled.
I don't know about linux-olpc-xo1.
Last edited by mistersmee (2021-02-02 12:49:18)
Offline
@mistersmee: If you send in a patch of our configs (preferably in the format of `git format-patch`), we can easily adopt your idea of disabling BPF in the kernels. (Also, this is a great way for you (and us) to test, how contributions from non-devs could work )
Offline
I have noticed errors about BPF configs coming up on my machine before it starts the kernel, which obviously aren't logged and are soon scrolled off screen
Architecture: pentium4, Testing repos: Yes, Hardware: EeePC 901+2GB RAM+OS half on the SD card.
Offline
I thought, nftables was.
btw: removing bpf from the config breaks the linux-zen kernel. I'm not sure, how to fix it.
Offline
I thought, nftables was.
btw: removing bpf from the config breaks the linux-zen kernel. I'm not sure, how to fix it.
I see. Maybe, we will try with less options disabled? Ideally the CONFIG_BPF_PRELOAD one should be enough, but the linux-ck dev turned some other BPF options off, which work for that kernel, but not for us.
I'll get such a patch ready.
Offline
I thought, nftables was.
btw: removing bpf from the config breaks the linux-zen kernel. I'm not sure, how to fix it.
Is our main kernel core/linux borked as well from the changes I made?
Edit: Nope, it is not.
Just did a pacman -Syu, got the 5.10.11-1.1 build, and it works fine.
Most importantly, the liblpf error messages are gone!
Hurrah, mission accomplished!
Last edited by mistersmee (2021-02-03 13:26:11)
Offline
I'm not even sure, that the linux-zen failure is due to your patches - it looks, like we didn't build the last couple kernels, there, anyways.
Offline
When looking through our vs. upstream's config of linux-zen, there are quite some differences. Probably, we should reduce that some time.
Offline
I looked through those a little bit, and most of them are purely related to 32-bit compilation, and replacing all the 64-bit bits. I didn't go over every one of the differences, the diff is nearly 2000 lines long. But a vim search yielded nothing related to the build error log, which was, a few modules in "net/ipv4/bbr.ko" failing to build, or something like that.
Of course, I know nothing related to configuring the Linux kernel, in fact, this is the first time I've come into contact with it.
A quick google-fu of the errors yesterday yielded nothing. I'm not sure where to go from here.
If linux-zen fails to build on 5.10.12, I'll grab the errors from the log and ask upstream. That should at least shed light on what/why it is failing.
Last edited by mistersmee (2021-02-04 09:02:50)
Offline
I'm not even sure, that the linux-zen failure is due to your patches - it looks, like we didn't build the last couple kernels, there, anyways.
Yeah, I looked at the build logs before and after I sent the second linux-zen patch, and they were the same errors, no change due to what I disabled.
Offline