09:10 <invano> Hi fche! Can I ask you what rc 0 is referring to for uprobes registration errors?

09:11 <invano> I have errors like "WARNING: probe process("/opt/lib/libc-2.23.so").statement(0x5523c)? inode-offset 0005523c registration error (rc 0)" with consequent segfault of the binary under probes. I'm under ARM 32, no problems with i386 and amd64

09:57 <fche> that rc 0 part doesn't make sense if there was an error

10:03 <invano> ok I had a look at the source code and in runtime/linux/uprobes-inode.c, errors or not, no value is ever assigned to ret in stapiu_target_reg()

10:04 <invano> I'm on stap 2.9/0.165 on x86, 64 and armhf32. didn't try yet on mips

10:09 <fche> yeah, so stap probably tracks the error code incorrectly

10:09 <fche> but there is probably a kernel error anyway; not super surprising as uprobes are pretty intricate

10:12 <fche> though in current code, runtime/linux/uprobes-node.c line 393ish (in git stap), the error numbers are preserved

10:12 <fche> (that was fixed very recently)

10:15 <invano> well probably I should get latest git tree, to track the error code at least

10:17 <fche> yup

10:21 <invano> ok for now I just fixed that line to get the correct error. let's see how it goes

10:22 <invano> at least normal kernel probes are working correctly :) not the case for mips o32 w/o kernel syms but I;m gonna fix it as soon as I reach that part

10:33 <fche> righto

11:24 <invano> ok the rc now is rc -22. should be EINVAL, right?

11:46 <fche> yeah. so the kernel's uprobes code would need to be looked at