Re: [RFC PATCH 0/3] kallsyms: add option to include relative filepaths into kallsyms

From: Alexander Lobakin
Date: Thu Aug 18 2022 - 09:42:47 EST


From: Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx>
Date: Thu, 18 Aug 2022 14:20:06 +0200

> On Thu, Aug 18, 2022 at 01:53:03PM +0200, Alexander Lobakin wrote:
> > This is an early RFC to not rewrite stuff one more time later on if
> > the implementation is not acceptable or any major design changes are
> > required. For the TODO list, please scroll to the end.
> >
> > Make kallsyms independent of symbols positions in vmlinux (or module)
> > by including relative filepath in each symbol's kallsyms value. I.e.
> >
> > dev_gro_receive -> net/core/gro.c:dev_gro_receive
> >
> > For the implementation details, please look at the patch 3/3.
> > Patch 2/3 is just a stub, I plan to reuse kallsyms enhancement from
> > the Rust series for it.
> > Patch 1/3 is a fix of one modpost macro straight from 2.6.12-rc2.
> >
> > A nice side effect is that it's now easier to debug the kernel, as
> > stacktraces will now tell every call's place in the file tree:
>
> That's a side effect, but that does not explain _why_ you want this
> change.
>
> What is this good for? Who can use it? Why is it worth added build
> times?

I think I wrote that we need to get rid of this positioned-based
search for kallsyms, at least for livepatching and probes, didn't
I?
OTOH I didn't write that originally that was a prereq for FG-KASLR,
but then I decided it deserves a separate series =\ Thanks for
mentioning this, so I wrote it here now and will not forget this
time to mention it in the cover letter for v2.

>
> You don't tell us anything except what this does :(
>
> thanks,
>
> greg k-h

Thanks,
Olek