Abhinav Anil Sharma
Abhinav Anil Sharma
On the same system where the test failures reproduce, I tried using perf to trace the same test app (suite/tests/bin/simple_app), and it didn't fail (see https://perf.wiki.kernel.org/index.php/Perf_tools_support_for_Intel%C2%AE_Processor_Trace#Kernel-only_tracing for the detailed steps)....
Here's the documented logic used by perf for copying kcore: https://github.com/torvalds/linux/blob/6d0dc8559c847e2dcd66c5dd93dbab3d3d887ff5/tools/perf/util/symbol-elf.c#L2473. Particularly, ``` The kernel map starts at _stext or the lowest function symbol, and ends at _etext or the...
Following up on our offline discussion: @dolanzhao Can you provide details of the kernel version you were able to reproduce this issue with? Do you have a fix ready that...
> Based on the contents of my /proc/kallsyms, I see that if we also considered the address of the lowest and highest t,w,T,W symbols, the unmapped address reported by libipt...
As a side note: perf also has this logic to copy the "entry trampolines" that we don't do[4]. I didn't observe any failures in our tracing because of missing this,...
I tried using DR's dumped kcore with `perf script` by copying it to the perf.data dir of a perf-collected trace, and perf was still able to decode its trace. I...
I was able to find the symbol for the missing IP in /proc/kallsyms after relaxing /proc/sys/net/core/bpf_jit_harden. One possible workaround is that we special-case bpf, and copy the /proc/kcore regions that...
Re the aliases: this came up again in discussion internally when an instruction that objdump shows as OP_mov showed up as an OP_orr in DR's view tool. Would be nice...
I see more changes were made since Derek's last comment. When this PR is ready for a re-review, please use the Github UI to signal that (see list of reviewers...
It does look like only the first mov instr is mismatching between the two versions: Provided ilist: ``` ilist for recreation: TAG 0x00007f1da7115000 +0 m4 @0x00007f1b66bf4af0 48 a1 48 54...