Message-ID: <4996fd95-49bf-71bb-7071-6050b60529c1@intel.com>
Date:   Mon, 20 Feb 2023 07:38:49 -0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.1
Subject: Re: [PATCH RESEND] x86/speculation: Fix user-mode spectre-v2
 protection with KERNEL_IBRS
Content-Language: en-US
To:     Borislav Petkov <bp@alien8.de>, KP Singh <kpsingh@kernel.org>,
        Andrew Cooper <andrew.cooper3@citrix.com>
Cc:     Josh Poimboeuf <jpoimboe@kernel.org>, linux-kernel@vger.kernel.org,
        pjt@google.com, evn@google.com, tglx@linutronix.de,
        mingo@redhat.com, dave.hansen@linux.intel.com, x86@kernel.org,
        hpa@zytor.com, peterz@infradead.org,
        pawan.kumar.gupta@linux.intel.com, kim.phillips@amd.com,
        alexandre.chartre@oracle.com, daniel.sneddon@linux.intel.com,
        =?UTF-8?Q?Jos=c3=a9_Oliveira?= <joseloliveira11@gmail.com>,
        Rodrigo Branco <rodrigo@kernelhacking.com>,
        Alexandra Sandulescu <aesa@google.com>,
        Jim Mattson <jmattson@google.com>
References: <20230220120127.1975241-1-kpsingh@kernel.org>
 <20230220121350.aidsipw3kd4rsyss@treble>
 <CACYkzJ5L9MLuE5Jz+z5-NJCCrUqTbgKQkXSqnQnCfTD_WNA7_Q@mail.gmail.com>
 <CACYkzJ6n=-tobhX0ONQhjHSgmnNjWnNe_dZnEOGtD8Y6S3RHbA@mail.gmail.com>
 <Y/OETPFTJbGtCADM@zn.tnic>
From:   Dave Hansen <dave.hansen@intel.com>
In-Reply-To: <Y/OETPFTJbGtCADM@zn.tnic>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2/20/23 06:31, Borislav Petkov wrote:
> This above probably wants to say that you need to write 1 on CPL change
> because it has a flushing behavior of killing user prediction entries.

Right.  The naive way of looking at IBRS is that setting IBRS=1
mitigates spectre-v2.  But, as the documentation says, just _leaving_ it
set to 1 is not good enough.  It must be actively rewritten in order to
get the strongest semantics.

It's still rather early in the morning, but I'm also quite confused
about what exactly the problem is here.  The patch and the changelog
aren't especially clear.  I'll need to stare at it with a cup of coffee
before I can give any coherent better suggestions, though.