> On May 3, 2019, at 4:16 PM, Linus Torvalds <torva...@linux-foundation.org>
> wrote:
>
>> On Fri, May 3, 2019 at 3:55 PM Andy Lutomirski <l...@amacapital.net> wrote:
>>
>> But I think this will end up worse than the version where the entry code
>> fixes it up. This is because, if the C code moves pt_regs, then we need
>> some way to pass the new pointer back to the asm.
>
> What? I already posted that code. Let me quote it again:
>
> Message-ID:
> <CAHk-=wh8bi5c_gkyjptdaiaxazrqtmhws30usuvs4qk_f+c...@mail.gmail.com>
>
> # args: pt_regs pointer (no error code for int3)
> movl %esp,%eax
> # allocate a bit of extra room on the stack, so that
> # 'kernel_int3' can move the pt_regs
> subl $8,%esp
> call kernel_int3
> movl %eax,%esp
>
> It's that easy (this is with the assumption that we've already applied
> the "standalone simple int3" case, but I think the above might work
> even with the current code model, just the "call do_int3" needs to
> have the kernel/not-kernel distinction and do the above for the kernel
> case)
>
> That's *MUCH* easier than your code to move entries around on the
> stack just as you return, and has the advantage of not changing any
> C-visible layout.
>
> The C interface looks like this
>
> /* Note: on x86-32, we can move 'regs' around for push/pop emulation */
> struct pt_regs *kernel_int3(struct pt_regs *regs)
> {
> ..
> .. need to pass regs to emulation functions
> .. and call emulation needs to return it
> ..
> return regs;
> }
>
> and I just posted as a response to Stephen the *trivial* do_int3()
> wrapper (so that x86-64 doesn't need to care), and the *trivial* code
> to actually emulate a call instruction.
>
> And when I say "trivial", I obviously mean "totally untested and
> probably buggy", but it sure seems *simple*.,
>
> Notice? Simple and minimal changes to entry code that only affect
> int3, and nothing else.
>
>
I can get on board with this.
