Re: [RFC PATCH 2/2] livepatch: Clear relocation targets on a module removal
From: Josh Poimboeuf
Date: Thu Sep 05 2019 - 09:42:50 EST
On Thu, Sep 05, 2019 at 03:31:56PM +0200, Jiri Kosina wrote:
> On Thu, 5 Sep 2019, Josh Poimboeuf wrote:
> > > All the indirect jumps are turned into alternatives when retpolines
> > > are in place.
> > Actually in C code those are done by the compiler as calls/jumps to
> > __x86_indirect_thunk_*.
> Sure, and the thunks do the redirection via JMP_NOSPEC / CALL_NOSPEC,
> which has alternative in it.
But the thunks are isolated to arch/x86/lib/retpoline.S. We can't patch
that code anyway.