Re: [PATCH bpf-next v2 0/5] Subskeleton support for BPF libraries

[Yonghong Song <yhs@fb.com>]

On 3/11/22 10:18 AM, Delyan Kratunov wrote:
> On Thu, 2022-03-10 at 21:10 -0800, Yonghong Song wrote:
>>
>> On 3/10/22 4:11 PM, Delyan Kratunov wrote:
> [..]
>>
>> When I tried to build the patch set with parallel mode (-j),
>>      make -C tools/testing/selftests/bpf -j
>> I hit the following errors:
>>
>> /bin/sh: line 1: 3484984 Bus error               (core dumped)
>> /home/yhs/work/bpf-next/tools/testing/selftests/bpf/tools/sbin/bpftool
>> gen skeleton
>> /home/yhs/work/bpf-next/tools/testing/selftests/bpf/test_ksyms_weak.linked3.o
>> name test_ksyms_weak >
>> /home/yhs/work/bpf-next/tools/testing/selftests/bpf/test_ksyms_weak.skel.h
>> make: *** [Makefile:496:
>> /home/yhs/work/bpf-next/tools/testing/selftests/bpf/test_ksyms_weak.skel.h]
>> Error 135
>> make: *** Deleting file
>> '/home/yhs/work/bpf-next/tools/testing/selftests/bpf/test_ksyms_weak.skel.h'
>> make: *** Waiting for unfinished jobs....
>> make: Leaving directory
>> '/home/yhs/work/bpf-next/tools/testing/selftests/bpf'
>>
>> Probably some make file related issues.
>> I didn't hit this issue before without this patch set.
> 
> Hm, that's interesting, can you reproduce it? I build everything with -j and
> have not seen any bpftool issues. I also use ASAN for bpftool and that's not
> complaining about anything either.

I can reproduce it in 50% of tries with command line:
   make -C tools/testing/selftests/bpf -j

> 
> SIGBUS suggests a memory mapped file was not there. I'll try and come up with
> ways that can happen, especially given that it's a `gen skeleton` invocation,
> which I haven't changed at all.
> 
> --Delyan