From mboxrd@z Thu Jan  1 00:00:00 1970
From: Dave Hansen <dave.hansen@linux.intel.com>
Subject: Re: [RFC PATCH v2 18/27] x86/cet/shstk: Introduce WRUSS instruction
Date: Fri, 13 Jul 2018 05:12:02 -0700
Message-ID: <166536e2-b296-7be5-d1b7-982cf56f1f9b@linux.intel.com>
References: <20180710222639.8241-1-yu-cheng.yu@intel.com>
 <20180710222639.8241-19-yu-cheng.yu@intel.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <20180710222639.8241-19-yu-cheng.yu@intel.com>
Content-Language: en-US
Sender: linux-kernel-owner@vger.kernel.org
To: Yu-cheng Yu <yu-cheng.yu@intel.com>, x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>, Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@redhat.com>, linux-kernel@vger.kernel.org, linux-doc@vger.kernel.org, linux-mm@kvack.org, linux-arch@vger.kernel.org, linux-api@vger.kernel.org, Arnd Bergmann <arnd@arndb.de>, Andy Lutomirski <luto@amacapital.net>, Balbir Singh <bsingharora@gmail.com>, Cyrill Gorcunov <gorcunov@gmail.com>, Florian Weimer <fweimer@redhat.com>, "H.J. Lu" <hjl.tools@gmail.com>, Jann Horn <jannh@google.com>, Jonathan Corbet <corbet@lwn.net>, Kees Cook <keescook@chromiun.org>, Mike Kravetz <mike.kravetz@oracle.com>, Nadav Amit <nadav.amit@gmail.com>, Oleg Nesterov <oleg@redhat.com>, Pavel Machek <pavel@ucw.cz>, Peter Zijlstra <pet>
List-Id: linux-api@vger.kernel.org

On 07/10/2018 03:26 PM, Yu-cheng Yu wrote:
> +static int is_wruss(struct pt_regs *regs, unsigned long error_code)
> +{
> +	return (((error_code & (X86_PF_USER | X86_PF_SHSTK)) ==
> +		(X86_PF_USER | X86_PF_SHSTK)) && !user_mode(regs));
> +}
> +
>  static void
>  show_fault_oops(struct pt_regs *regs, unsigned long error_code,
>  		unsigned long address)
> @@ -848,7 +859,7 @@ __bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
>  	struct task_struct *tsk = current;
>  
>  	/* User mode accesses just cause a SIGSEGV */
> -	if (error_code & X86_PF_USER) {
> +	if ((error_code & X86_PF_USER) && !is_wruss(regs, error_code)) {
>  		/*
>  		 * It's possible to have interrupts off here:
>  		 */

Please don't do it this way.

We have two styles of page fault:
1. User page faults: find a VMA, try to handle (allocate memory et al.),
   kill process if we can't handle.
2. Kernel page faults: search for a *discrete* set of conditions that
   can be handled, including faults in instructions marked in exception
   tables.

X86_PF_USER *means*: do user page fault handling.  In the places where
the hardware doesn't set it, but we still want user page fault handling,
we manually set it, like this where we "downgrade" an implicit
supervisor access to a user access:

        if (user_mode(regs)) {
                local_irq_enable();
                error_code |= X86_PF_USER;
                flags |= FAULT_FLAG_USER;

So, just please *clear* X86_PF_USER if !user_mode(regs) and X86_PF_SS.
We do not want user page fault handling, thus we should not keep the bit
set.