Re: [PATCH v6 5/6] tracing: Show inode and device major:minor in deferred user space stacktrace

[Sam James <sam@gentoo.org>]

Arnaldo Carvalho de Melo <arnaldo.melo@gmail.com> writes:

> On August 28, 2025 5:51:39 PM GMT-03:00, Steven Rostedt <rostedt@kernel.org> wrote:
>>On Thu, 28 Aug 2025 17:27:37 -0300
>>Arnaldo Carvalho de Melo <arnaldo.melo@gmail.com> wrote:
>>
>>> >I would love to have a hash to use. The next patch does the mapping
>>> >of the inode numbers to their path name. It can  
>>> 
>>> The path name is a nice to have detail, but a content based hash is
>>> what we want, no?
>>> 
>>> Tracing/profiling has to be about contents of files later used for
>>> analysis, and filenames provide no guarantee about that.
>>
>>I could add the build id to the inode_cache as well (which I'll rename
>>to file_cache).
>>
>>Thus, the user stack trace will just have the offset and a hash value
>>that will be match the output of the file_cache event which will have
>>the path name and a build id (if one exists).
>>
>>Would that work?
>
> Probably.
>
> This "if it is available" question is valid, but since 2016 it's is more of a "did developers disabled it explicitly?"
>
> If my "googling" isn't wrong, GNU LD defaults to generating a build ID in ELF images since 2011 and clang's companion since 2016.

GNU ld doesn't ever default to generating build IDs, and I don't *think*
LLVM does either (either in Clang, or LLD).

GCC, on the other hand, has a configure arg to control this, but it's
default-off. Clang generally prefers to have defaults like this done via
user/sysadmin specified configuration files rather than adding
build-time configure flags.

Now, is it a reasonable ask to say "we require build IDs for this
feature"? Yeah, it probably is, but it's not default-on right now, and
indeed we in Gentoo aren't using them yet (but I'm working on enabling
them).

>
> So making it even more available than what the BPF guys did long ago
> and perf piggybacked on at some point, by having it cached, on
> request?, in some 20 bytes alignment hole in task_struct that would be
> only used when profiling/tracing may be amenable.

thanks,
sam