Re: [PATCH v4 15/16] module: Move where we mark modules RO,X

[Josh Poimboeuf]

On Wed, Oct 23, 2019 at 01:48:35PM +0200, Peter Zijlstra wrote:
> Now sadly that commit missed all the useful information, luckily I could
> find the patch in my LKML folder, more sad, that thread still didn't
> contain the actual useful information, for that I was directed to
> github:
> 
>   https://github.com/dynup/kpatch/issues/580
> 
> Now, someone is owning me a beer for having to look at github for this.

Deal.  And you probably deserve a few more for fixing our crap.

The github thing is supposed to be temporary, at least in theory we'll
eventually have all klp patch module building code in the kernel tree.

> That finally explained that what happens is that the RELA was trying to
> fix up the paravirt indirect call to 'local_irq_disable', which
> apply_paravirt() will have overwritten with 'CLI; NOP'. This then
> obviously goes *bang*.
> 
> This then raises a number of questions:
> 
>  1) why is that RELA (that obviously does not depend on any module)
>     applied so late?

Good question.  The 'pv_ops' symbol is exported by the core kernel, so I
can't see any reason why we'd need to apply that rela late.  In theory,
kpatch-build isn't supposed to convert that to a klp rela.  Maybe
something went wrong in the patch creation code.

I'm also questioning why we even need to apply the parainstructions
section late.  Maybe we can remove that apply_paravirt() call
altogether, along with .klp.arch.parainstruction sections.

I'll need to look into it...

>  2) why can't we unconditionally skip RELA's to paravirt sites?

We could, but I don't think it's needed if we fix #1.

>  3) Is there ever a possible module-dependent RELA to a paravirt /
>     alternative site?

Good question...

> Now, for 1), I would propose '.klp.rela.${mod}' sections only contain
> RELAs that depend on symbols in ${mod} (or modules in general).

That was already the goal, but we've apparently failed at that.

> We can fix up RELAs that depend on core kernel early without problems.
> Let them be in the normal .rela sections and be fixed up on loading
> the patch-module as per usual.

If such symbols aren't exported, then they still need to be in
.klp.rela.vmlinux sections, since normal relas won't work.

> This should also deal with 2, paravirt should always have RELAs into the
> core kernel.
> 
> Then for 3) we only have alternatives left, and I _think_ it unlikely to
> be the case, but I'll have to have a hard look at that.

I'm not sure about alternatives, but maybe we can enforce such
limitations with tooling and/or kernel checks.

-- 
Josh