Re: armv5 userland broken

[Rin Okuyama <rokuyama.rk%gmail.com@localhost>]

Nick, thank you for committing your patch.

For my side, I've confirmed there's no regression for ATF on
evbarmv[67]{,hf} and evbarmv5.

It seems that LLVM does not accept pull requests from their
GitHub account.

I will examine how to, and submit the patch, this weekend.

Thanks,
rin

On 2021/06/30 18:20, Rin Okuyama wrote:
> Hi,
>
> On 2021/06/30 16:39, Nick Hudson wrote:
> > On 30/06/2021 07:22, Nick Hudson wrote:
> > > Great work!
> > >
> > > On 29/06/2021 08:13, Rin Okuyama wrote:
> > > > I've examined assembler codes for userland libraries for arm:
> > > > common/lib/libc, csu, libc, compiler_rt, and libm, and found that
> > > > some routines in compiler_rt do not align sp correctly:
> > > >
> > > > http://www.netbsd.org/~rin/arm_stack_align_compiler_rt.patch
> > > >
> > > > Also, some routines in libc do not also in Thumb-mode userland:
> > > >
> > > > http://www.netbsd.org/~rin/arm_stack_align_libc_thumb.patch
> > > I think it's better / less ugly to push even numbers of registers rather
> > > than adjusting sp (especially for thumb)
> >
> > How about this diff? (untested)
> >
> > I've dropped callee saved registers from the compiler_rt code as (I
> > think) they're unused.
>
> Thank you very much for the patch. Much cleaner than before!
>
> As you pointed out, r7 is unused for compiler_rt codes at the moment.
>
> The reason why I hesitated to drop r7 is: the codes will be strangely
> broken when the upstream changes their mind to use r7 again.
>
> But your version seems much nicer. I will submit the patch to the
> upstream immediately after you commit it to our tree.
>
> I will carry out ATF with your patch for some architectures.
>
> Thanks,
> ri