Re: Live Patching Microconference at Linux Plumbers

[Josh Poimboeuf <jpoimboe@kernel.org>]

On Fri, Apr 14, 2023 at 09:30:13PM +0200, Peter Zijlstra wrote:
> On Fri, Apr 14, 2023 at 10:12:55AM -0700, Josh Poimboeuf wrote:
> 
> > > * How to avoid/minimize the need to reverse-engineer control flow for things
> > >   like ORC generation.
> > > 
> > >   On the arm64 side we're pretty averse to doing this to generate metadata for
> > >   unwinding (and we might not need to), but there are things objtool does today
> > >   that requires awareness of control-flow (e.g. forward-edge checks for noinstr
> > >   safety).
> > > 
> > >   Hopefully without a flamewar about DWARF...
> > 
> > If objtool is going to be doing control-flow anyway then it could just
> > validate DWARF/SFrame.  Then everybody's happy?
> 
> Right; so per another recent thread somewhere; you can't rely on
> DWARF/Sframe or any other compiler generated thing simply because it
> doesn't cover .S files and inline asm -- and this being a kernel, we've
> got quite a bit of that.
> 
> At best it could use DWARF to help reconstruct code flow and then
> validate Sframe for the bits that got sframe.

I wasn't (necessarily) suggesting that objtool use DWARF as an input to
help it construct the control-flow graph (CFG).

Instead, the idea is for objtool to continue to reverse-engineer the CFG
as it does today (albeit with a little help from the compiler in
specific problematic areas, e.g. noreturns and jump tables).

Then, it could use that independently-developed CFG to read the
compiler-generated metadata (DWARF/SFrame/whatever), and report any
warnings if the DWARF doesn't match objtool's CFG.

In other words, DWARF validation becomes just another optional objtool
feature, similar to its other unwinding-related features like frame
pointer validation and ORC generation.

That way, regardless of which philosophy [1] you subscribe to, if
something is amiss with reliable unwinding, the end result is the same:
a warning.

Then a human can look at the warning and investigate whether it's
objtool or DWARF/whatever that needs to be fixed.

[1] "reverse engineering is risky" vs "reverse engineering is reliable"

-- 
Josh