mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Vineet Gupta <vineetg@rivosinc.com>
To: "Xi Ruoyao" <xry111@xry111.site>,
	"Alexandre Ghiti" <alex@ghiti.fr>,
	"Thomas Weißschuh" <thomas.weissschuh@linutronix.de>,
	"Nathan Chancellor" <nathan@kernel.org>
Cc: "Jason A . Donenfeld" <Jason@zx2c4.com>,
	Paul Walmsley <paul.walmsley@sifive.com>,
	Palmer Dabbelt <palmer@dabbelt.com>, Guo Ren <guoren@kernel.org>,
	linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] RISC-V: vDSO: Correct inline assembly constraints in the getrandom syscall wrapper
Date: Fri, 6 Jun 2025 15:01:27 -0700	[thread overview]
Message-ID: <94f9af73-0b2d-484c-ba1d-d4435908336b@rivosinc.com> (raw)
In-Reply-To: <20250606092443.73650-2-xry111@xry111.site>

On 6/6/25 02:24, Xi Ruoyao wrote:
> As recently pointed out by Thomas, if a register is forced for two
> different register variables, among them one is used as "+" (both input
> and output) and another is only used as input, Clang would treat the
> conflicting input parameters as undefined behaviour and optimize away
> the argument assignment.
>
> Per an example in the GCC documentation, for this purpose we can use "="
> (only output) for the output, and "0" for the input for that we must
> reuse the same register as the output.  And GCC developers have
> confirmed using a simple "r" (that we use for most vDSO implementations)
> instead of "0" is also fine.
>
> Link: https://lore.kernel.org/all/20250603-loongarch-vdso-syscall-v1-1-6d12d6dfbdd0@linutronix.de/
> Link: https://gcc.gnu.org/onlinedocs/gcc-15.1.0/gcc/Local-Register-Variables.html
> Link: https://gcc.gnu.org/pipermail/gcc-help/2025-June/144266.html
> Cc: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
> Cc: Nathan Chancellor <nathan@kernel.org>
> Signed-off-by: Xi Ruoyao <xry111@xry111.site>
> ---
>
> v1 -> v2: Keep using "r" for buffer to follow the existing convention
> (that the GCC developers have confirmed fine).
>
>  arch/riscv/include/asm/vdso/getrandom.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/riscv/include/asm/vdso/getrandom.h b/arch/riscv/include/asm/vdso/getrandom.h
> index 8dc92441702a..c6d66895c1f5 100644
> --- a/arch/riscv/include/asm/vdso/getrandom.h
> +++ b/arch/riscv/include/asm/vdso/getrandom.h
> @@ -18,7 +18,7 @@ static __always_inline ssize_t getrandom_syscall(void *_buffer, size_t _len, uns
>  	register unsigned int flags asm("a2") = _flags;
>  
>  	asm volatile ("ecall\n"
> -		      : "+r" (ret)
> +		      : "=r" (ret)
>  		      : "r" (nr), "r" (buffer), "r" (len), "r" (flags)
>  		      : "memory");

My 2 cents as I've dabbled into this for ARC glibc syscall macros [1] where r0
is both the first syscall/function arg and also the function/syscall return.

The v2 approach still keeps 2 different variables in same local reg which has
potential for any future compiler shenanigans.
Segher's example avoided specifying the same reg.
What about something like the following: seems to generate the right code (with
gcc 15)

       register long ret asm("a0");
       register long nr asm("a7") = __NR_getrandom;
       register size_t len asm("a1") = _len;
       register unsigned int flags asm("a2") = _flags;
       ret = (unsigned long) _buffer;

       asm volatile ("ecall\n"
                     : "+r" (ret)                                 // keep "+r"
for input _buffer / output ret
                     : "r" (nr), "r" (len), "r" (flags)
                     : "memory");

       return ret;

Thx,
-Vineet

[1] https://github.com/bminor/glibc/blob/master/sysdeps/unix/sysv/linux/arc/sysdep.h

  parent reply	other threads:[~2025-06-06 22:01 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-04-11  2:46 [PATCH v2] RISC-V: vDSO: Wire up getrandom() vDSO implementation Xi Ruoyao
2025-04-11  8:04 ` Thomas Weißschuh
2025-05-23  8:01   ` Alexandre Ghiti
2025-05-23  8:02     ` Xi Ruoyao
2025-05-23 10:06       ` Alexandre Ghiti
2025-06-03 12:48         ` Xi Ruoyao
2025-06-04 11:45           ` Alexandre Ghiti
2025-06-04 15:09             ` [PATCH] RISC-V: vDSO: Correct inline assembly constraints in the getrandom syscall wrapper Xi Ruoyao
2025-06-05  7:15               ` Thomas Weißschuh
2025-06-06  9:24                 ` [PATCH v2] " Xi Ruoyao
2025-06-06  9:50                   ` Thomas Weißschuh
2025-06-06 22:01                   ` Vineet Gupta [this message]
2025-06-07 14:16                     ` Xi Ruoyao
2025-06-10  7:11                       ` Thomas Weißschuh
2025-06-10  8:15                         ` Alexandre Ghiti
2025-06-12 20:10                   ` patchwork-bot+linux-riscv

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=94f9af73-0b2d-484c-ba1d-d4435908336b@rivosinc.com \
    --to=vineetg@rivosinc.com \
    --cc=Jason@zx2c4.com \
    --cc=alex@ghiti.fr \
    --cc=guoren@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-riscv@lists.infradead.org \
    --cc=nathan@kernel.org \
    --cc=palmer@dabbelt.com \
    --cc=paul.walmsley@sifive.com \
    --cc=thomas.weissschuh@linutronix.de \
    --cc=xry111@xry111.site \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®