Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 2 Sep 2022 11:22:45 +0200
From:   Andrew Jones <ajones@ventanamicro.com>
To:     Al Viro <viro@zeniv.linux.org.uk>
Cc:     linux-riscv@lists.infradead.org,
        Linus Torvalds <torvalds@linux-foundation.org>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] riscv: fix a nasty sigreturn bug...
Message-ID: <20220902092245.ande6fvievnbn35h@kamzik>
References: <YU0wDzeS/jXwkAca@zeniv-ca.linux.org.uk>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YU0wDzeS/jXwkAca@zeniv-ca.linux.org.uk>
Precedence: bulk

On Fri, Sep 24, 2021 at 01:55:27AM +0000, Al Viro wrote:
> riscv has an equivalent of arm bug fixed by 653d48b22166; if signal
> gets caught by an interrupt that hits when we have the right value
> in a0 (-513), *and* another signal gets delivered upon sigreturn()
> (e.g. included into the blocked mask for the first signal and posted
> while the handler had been running), the syscall restart logics will
> see regs->cause equal to EXC_SYSCALL (we are in a syscall, after all)
> and a0 already restored to its original value (-513, which happens to
> be -ERESTARTNOINTR) and assume that we need to apply the usual
> syscall restart logics.
>     
> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
> ---
> diff --git a/arch/riscv/kernel/signal.c b/arch/riscv/kernel/signal.c
> index c2d5ecbe55264..f8fb85dc94b7a 100644
> --- a/arch/riscv/kernel/signal.c
> +++ b/arch/riscv/kernel/signal.c
> @@ -121,6 +121,8 @@ SYSCALL_DEFINE0(rt_sigreturn)
>  	if (restore_altstack(&frame->uc.uc_stack))
>  		goto badframe;
>  
> +	regs->cause = -1UL;
> +
>  	return regs->a0;
>  
>  badframe:
>

This looks good to me based on what other architectures do.

For example, arm64 does

  rt_sigreturn
    restore_sigframe
      forget_syscall
        regs->syscallno = NO_SYSCALL

  which results in do_signal avoiding syscall restarting

And x86 does

  rt_sigreturn
    restore_sigcontext
      regs->orig_ax = -1

  where its handle_signal only restarts syscalls when regs->orig_ax != -1

So, for riscv, where in do_signal and handle_signal syscall restarting
is avoided when regs->cause != EXC_SYSCALL and it's common to set cause
to -1 to avoid it, then it makes sense to set regs->cause != EXEC_SYSCALL
in rt_sigreturn or in restore_sigcontext, which rt_sigreturn calls, as
well.

So the only question I have is whether or not the cause assignment
is better in restore_sigcontext() like other architectures? At least,
since rt_sigreturn is the only caller of restore_sigcontext, it can't
break anything putting it there atm...

Anyway,

Reviewed-by: Andrew Jones <ajones@ventanamicro.com>

BTW, I ran the testcase from 653d48b22166 with the asm modified for
riscv for a while over QEMU. It didn't reproduce, but I suppose that
doesn't mean much.

Thanks,
drew