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;
MIME-Version: 1.0
References: <20220222155345.138861-1-tsbogend@alpha.franken.de>
In-Reply-To: <20220222155345.138861-1-tsbogend@alpha.franken.de>
From:   Arnd Bergmann <arnd@kernel.org>
Date:   Tue, 22 Feb 2022 18:04:07 +0100
Message-ID: <CAK8P3a0QV7y_gFv=VHGKVWjXyYmFFZRrXj3m52d21Fyydib4NQ@mail.gmail.com>
Subject: Re: [PATCH] MIPS: Handle address errors for accesses above CPU max
 virtual user address
To:     Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc:     linux-mips <linux-mips@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Feb 22, 2022 at 4:53 PM Thomas Bogendoerfer
<tsbogend@alpha.franken.de> wrote:
>
> Address errors have always been treated as unaliged accesses and handled
> as such. But address errors are also issued for illegal accesses like
> user to kernel space or accesses outside of implemented spaces. This
> change implements Linux exception handling for accesses to the illegal
> space above the CPU implemented maximum virtual user address and the
> MIPS 64bit architecture maximum. With this we can now use a fixed value
> for the maximum task size on every MIPS CPU and get a more optimized
> access_ok().
>
> Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>

Thank you for addressing this. Should I add this patch to my series
ahead of "mips: use simpler access_ok()"? That way I can keep it all
in my asm-generic tree as a series for 5.18.

>  arch/mips/kernel/unaligned.c | 17 +++++++++++++++++
>  1 file changed, 17 insertions(+)
>
> diff --git a/arch/mips/kernel/unaligned.c b/arch/mips/kernel/unaligned.c
> index df4b708c04a9..7b5aba5df02e 100644
> --- a/arch/mips/kernel/unaligned.c
> +++ b/arch/mips/kernel/unaligned.c
> @@ -1480,6 +1480,23 @@ asmlinkage void do_ade(struct pt_regs *regs)
>         prev_state = exception_enter();
>         perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS,
>                         1, regs, regs->cp0_badvaddr);
> +
> +#ifdef CONFIG_64BIT
> +       /*
> +        * check, if we are hitting space between CPU implemented maximum
> +        * virtual user address and 64bit maximum virtual user address
> +        * and do exception handling to get EFAULTs for get_user/put_user
> +        */
> +       if ((regs->cp0_badvaddr >= (1UL << cpu_vmbits)) &&
> +           (regs->cp0_badvaddr < XKSSEG)) {

It might be clearer to use TASK_SIZE_MAX here instead of XKSSEG,
to match the check in access_ok(). If you like, I can change that while
applying.

        Arnd