Hi,
On 1/7/2022 6:00 AM, Daniel Borkmann wrote:
> On 1/4/22 2:42 AM, Hou Tao wrote:
>> The bpf line info for arm64 is broken due to two reasons:
>> (1) insn_to_jit_off passed to bpf_prog_fill_jited_linfo() is
>> calculated in instruction granularity instead of bytes
>> granularity.
>> (2) insn_to_jit_off only considers the body itself and ignores
>> prologue before the body.
>>
>> So fix it by calculating offset as byte-offset and do build_prologue()
>> first in the first JIT pass.
>>
[snip]
>> - /* Fake pass to fill in ctx->offset. */
>> - if (build_body(&ctx, extra_pass)) {
>> + /*
>> + * 1. Initial fake pass to compute ctx->idx and ctx->offset.
>> + *
>> + * BPF line info needs ctx->offset[i] to be the byte offset
>> + * of instruction[i] in jited image, so build prologue first.
>> + */
>> + if (build_prologue(&ctx, was_classic)) {
>> prog = orig_prog;
>> goto out_off;
>> }
>> - if (build_prologue(&ctx, was_classic)) {
>> + if (build_body(&ctx, extra_pass)) {
>> prog = orig_prog;
>> goto out_off;
>
> Could you split this into two logical patches? Both 1/2 seem independent
> of each other and should have been rather 2 patches instead of 1.
>
Sorry for the later reply. Splitting into two patches make sense for me. Will do
it in v2.
> Did you check if also other JITs could be affected?
It seems sparc also doesn't represent offset by bytes and I can check other
arches as well,
but it is sad that I don't have the environments for these arches.
> Thanks,
> Daniel
> .
