RE: [PATCH v4 3/5] bpf: add a test for BPF ELF load

[Marat Khalili <marat.khalili@huawei.com>]

> > RTE>>bpf_elf_load_autotest
> > BPF: resolve_xsym(40): EBPF_PSEUDO_CALL to external function: dummy_func1
> > ../../src/lib/bpf/bpf_validate.c:247:24: runtime error: signed integer overflow: 9223372036854775807
> + 2147483647 cannot be represented in type 'long int'
> >     #0 0x2b1fdca in eval_add ../../src/lib/bpf/bpf_validate.c:247
> >     #1 0x2b2b9dc in eval_alu ../../src/lib/bpf/bpf_validate.c:691
> >     #2 0x2b3b187 in evaluate ../../src/lib/bpf/bpf_validate.c:2446
> >     #3 0x2b3bf0f in __rte_bpf_validate ../../src/lib/bpf/bpf_validate.c:2537
> >     #4 0x2b19b33 in rte_bpf_load ../../src/lib/bpf/bpf_load.c:113
> >     #5 0x2b47d6d in bpf_load_elf ../../src/lib/bpf/bpf_load_elf.c:288
> >     #6 0x2b47fe8 in rte_bpf_elf_load ../../src/lib/bpf/bpf_load_elf.c:320
> >     #7 0x66b4b8 in test_bpf_elf_load ../../src/app/test/test_bpf.c:3382
> 
> Hmm. looks like a pre-existing BPF library bug since this code was never
> tested before.

Yes, that's correct. Sent some fixes for this and some other issues to the mailing list just now, maybe we should treat these problems separately.