Re: BPF relocation 'perf test' entries failing was: Re: [GIT PULL 00/27] perf/core improvements and fixes

From: Arnaldo Carvalho de Melo
Date: Thu Jul 26 2018 - 15:14:42 EST

Em Thu, Jul 26, 2018 at 03:58:05PM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Thu, Jul 26, 2018 at 11:04:08PM +0530, Sandipan Das escreveu:
> > I came across the same problem. Does the following patch fix it?

> >

> Oh my, that one was subtle... Checking...

Right, when it said it was running:

40.3: BPF prologue generation

In fact it was running:

40.4: BPF relocation checker

That error message in the 'perf test -v' output:

libbpf: Program 'func=sys_write' contains non-map related relo data pointing to section 65522

Was the _expected_ one for "40.4: BPF relocation checker", then it
failed when it couldn't run the off by one "last" subtest, duh.

Thomas, this was a problem introduced by a patch from you, just for
reference, this one:

Fixes: 9ef0112442bd ("perf test: Fix subtest number when showing results")

I should have caught this by running 'perf test' before/after applying
that patch, I'll now make sure this is done before sending pull reqs

In fact I did after, and thought, hey, some BPF related regression, I
must've updated clang/llvm and something new appeared on the radar, so I
went that direction and went nowhere, well, now I have an uptodate
llvm/clang combo to play with BTF, pahole, etc 8-)


- Arnaldo