On 23/03/20 20:30, Nick Desaulniers wrote:
> <ndesaulniers@xxxxxxxxxx> wrote:
>> So maybe we can find why
>> commit 76b043848fd2 ("x86/retpoline: Add initial retpoline support")
>> added THUNK_TARGET with and without "m" constraint, and either:
>> - remove "m" from THUNK_TARGET. (Maybe this doesn't compile somewhere)
>> or
>> - use my above recommendation locally avoiding THUNK_TARGET. We can
>> use "r" rather than "a" (what Clang would have picked) or "b (what GCC
>> would have picked) to give the compilers maximal flexibility.
> So I've sent a patch for the latter; my reason for not pursuing the former is:
> 1. I assume that the thunk target could be spilled, or a pointer, and
> we'd like to keep flexibility for the general case of inline asm that
> doesn't modify the stack pointer.
> 2. `entry` is local to `handle_external_interrupt_irqoff`; it's not
> being passed in via pointer as a function parameter.
> 3. register pressure is irrelevant if the resulting code is incorrect.
Yes, this is fair enough. I've queued your patch and will send it
shortly to Linus.
Paolo
