[lkml]   [2021]   [Oct]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] tools/nolibc: x86: Remove `r8`, `r9` and `r10` from the clobber list
Hello Ammar,

First, thanks for your patch. I have a few questions below.

On Mon, Oct 11, 2021 at 11:03:44AM +0700, Ammar Faizi wrote:
> Linux x86-64 syscall only clobbers rax, rcx and r11 (and "memory").
> - rax for the return value.
> - rcx to save the return address.
> - r11 to save the rflags.
> Other registers are preserved.
> Having r8, r9 and r10 in the syscall clobber list is harmless, but this
> results in a missed-optimization.
> As the syscall doesn't clobber r8-r10, GCC should be allowed to reuse
> their value after the syscall returns to userspace. But since they are
> in the clobber list, GCC will always miss this opportunity.

I agree with your conclusion about this but need to be perfectly sure
about the exact clobber list since I got a different impression when
implementing the calls. I got the impression that these ones were
clobbered by reading entry_64.S below:

* Registers on entry:
* rax system call number
* rcx return address
* r11 saved rflags (note: r11 is callee-clobbered register in C ABI)
* rdi arg0
* rsi arg1
* rdx arg2
* r10 arg3 (needs to be moved to rcx to conform to C ABI)
* r8 arg4
* r9 arg5
* (note: r12-r15, rbp, rbx are callee-preserved in C ABI)

See this last note ? Indicating that r12-15, rbp and rbx are preserved
made me think it's not the case for the other ones, but that might of
course be a misunderstanding.

And calling.h says this:

x86 function call convention, 64-bit:
arguments | callee-saved | extra caller-saved | return
[callee-clobbered] | | [callee-clobbered] |
rdi rsi rdx rcx r8-9 | rbx rbp [*] r12-15 | r10-11 | rax, rdx [**]

Note that it's indicated "function call convention", not "syscall",
leaving it open to have extra restrictions on syscalls. Later by
reading the POP_REGS macro called with pop_rdi=0 and skipr11rcx=1
on syscall leave, it seems to restore everything but r11, rcx, rax
and rdi (which is restored last with rsp).

As such, could you please add in your commit message a link to the
location where you found that authoritative information above it there
is a better place (i.e. one that doesn't require to read all the macros)?
This would significantly help anyone facing the same doubts about this
in the future.

Thank you!

 \ /
  Last update: 2021-10-12 07:29    [W:0.117 / U:6.152 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site