Received-SPF: pass (google.com: domain of linux-kernel+bounces-38109-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Subject: Re: [PATCH bpf 3/3] selftest/bpf: Test the read of vsyscall page
 under x86-64
To: Sohil Mehta <sohil.mehta@intel.com>, x86@kernel.org, bpf@vger.kernel.org
Cc: Dave Hansen <dave.hansen@linux.intel.com>,
 Andy Lutomirski <luto@kernel.org>, Peter Zijlstra <peterz@infradead.org>,
 Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@redhat.com>,
 Borislav Petkov <bp@alien8.de>, "H . Peter Anvin" <hpa@zytor.com>,
 linux-kernel@vger.kernel.org, xingwei lee <xrivendell7@gmail.com>,
 Jann Horn <jannh@google.com>, houtao1@huawei.com
References: <20240119073019.1528573-1-houtao@huaweicloud.com>
 <20240119073019.1528573-4-houtao@huaweicloud.com>
 <1c260ab1-d4c0-48e6-bf2b-df69bf083d27@intel.com>
From: Hou Tao <houtao@huaweicloud.com>
Message-ID: <01ce6947-8b46-9e27-6ce2-ce5763e3c5a9@huaweicloud.com>
Date: Thu, 25 Jan 2024 15:54:21 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101
 Thunderbird/78.6.0
Precedence: bulk
MIME-Version: 1.0
In-Reply-To: <1c260ab1-d4c0-48e6-bf2b-df69bf083d27@intel.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Content-Language: en-US


On 1/23/2024 8:25 AM, Sohil Mehta wrote:
> On 1/18/2024 11:30 PM, Hou Tao wrote:
>
>> vsyscall page could be disabled by CONFIG_LEGACY_VSYSCALL_NONE or
>> vsyscall=none boot cmd-line, but it doesn't affect the reproduce of the
>> problem and the returned error codes.
>>
> With vsyscall=emulate a direct read of the vsyscall address from
> userspace is expected to go through. This is mode deprecated so maybe it
> wouldn't matter much. Without the fix in patch 2/3, do you see the same
> behavior with vsyscall=emulate set in the cmdline?

Er, I think it depends on whether or not SMAP [1] feature is available.
When SMAP feature is enabled, even the vsyscall page is populated,
reading the vsyscall page through bpf_read_kernel() will trigger a page
fault and then oops. But when there is not SMAP, bpf_read_kernel() will
succeed. So I think the test may need to be skipped if vsyscall_mode is
emulate.

[1]: https://en.wikipedia.org/wiki/Supervisor_Mode_Access_Prevention
>
> Sohil