Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] x86/alternatives: check int3 breakpoint physical
 addresses
To:     Thomas Gleixner <tglx@linutronix.de>
References: <1548428249-8258-1-git-send-email-alexandre.chartre@oracle.com>
 <alpine.DEB.2.21.1902102221500.8784@nanos.tec.linutronix.de>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        Ingo Molnar <mingo@kernel.org>, Borislav Petkov <bp@alien8.de>,
        "H. Peter Anvin" <hpa@zytor.com>,
        Josh Poimboeuf <jpoimboe@redhat.com>, x86@kernel.org,
        Peter Zijlstra <peterz@infradead.org>,
        Alexandre Chartre <alexandre.chartre@oracle.com>
From:   Alexandre Chartre <alexandre.chartre@oracle.com>
Organization: Oracle Corporation
Message-ID: <8995b3b4-b805-c8c7-95a8-a39ab000f289@oracle.com>
Date:   Mon, 11 Feb 2019 10:08:11 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
 Thunderbird/45.6.0
MIME-Version: 1.0
In-Reply-To: <alpine.DEB.2.21.1902102221500.8784@nanos.tec.linutronix.de>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On 02/10/2019 10:23 PM, Thomas Gleixner wrote:
> On Fri, 25 Jan 2019, Alexandre Chartre wrote:
>> Note that this issue has been observed and reproduced with a custom kernel
>> with some code mapped to different virtual addresses and using jump labels
>> As jump labels use text_poke_bp(), crashes were sometimes observed when
>> updating jump labels.
>
> Rightfully so. text_poke_bp() pokes at the kernels text mapping which is
> very well defined and unique. Why would you map the same text to different
> virtual addresses and then use a randomly chosen one to poke at it?
>

As an example, we used to have per-CPU SYSCALL entry trampoline [1] where the
entry_SYSCALL_64_trampoline function was mapped to a different virtual address
for each CPU. So, a syscall would execute entry_SYSCALL_64_trampoline() from a
different virtual address depending on the CPU being used. With that code,
adding a jump label in entry_SYSCALL_64_trampoline() causes the described issue.

This mapping was eventually removed [2]. I don't know if any other code like this
is currently present in the kernel (I couldn't find any, but I might not have
covered everything). But, as this past commit have shown, this is definitively
something that can happen.

Thanks,

alex.

---
[1] 3386bc8aed82 ("x86/entry/64: Create a per-CPU SYSCALL entry trampoline")
     https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=3386bc8aed82

[2] bf904d2762ee ("x86/pti/64: Remove the SYSCALL64 entry trampoline")
     https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=bf904d2762ee