Re: [PATCH v11 6/9] x86/cet: Add PTRACE interface for CET - Yu, Yu-cheng

From: "Yu, Yu-cheng" <yu-cheng.yu@intel.com>
To: Dave Hansen <dave.hansen@intel.com>, Andy Lutomirski <luto@kernel.org>
Cc: Jann Horn <jannh@google.com>,
	the arch/x86 maintainers <x86@kernel.org>,
	"H. Peter Anvin" <hpa@zytor.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>,
	kernel list <linux-kernel@vger.kernel.org>,
	"open list:DOCUMENTATION" <linux-doc@vger.kernel.org>,
	Linux-MM <linux-mm@kvack.org>,
	linux-arch <linux-arch@vger.kernel.org>,
	Linux API <linux-api@vger.kernel.org>,
	Arnd Bergmann <arnd@arndb.de>,
	Balbir Singh <bsingharora@gmail.com>,
	Borislav Petkov <bp@alien8.de>,
	Cyrill Gorcunov <gorcunov@gmail.com>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Eugene Syromiatnikov <esyr@redhat.com>,
	Florian Weimer <fweimer@redhat.com>,
	"H.J. Lu" <hjl.tools@gmail.com>, Jonathan Corbet <corbet@lwn.net>,
	Kees Cook <keescook@chromium.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 <peterz@infradead.org>,
	Randy Dunlap <rdunlap@infradead.org>,
	"Ravi V. Shankar" <ravi.v.shankar@intel.com>,
	Vedvyas Shanbhogue <vedvyas.shanbhogue@intel.com>,
	Dave Martin <Dave.Martin@arm.com>,
	Weijiang Yang <weijiang.yang@intel.com>
Subject: Re: [PATCH v11 6/9] x86/cet: Add PTRACE interface for CET
Date: Thu, 3 Sep 2020 10:59:14 -0700	[thread overview]
Message-ID: <2a58982b-8a69-1280-86ec-d0b70ede4453@intel.com> (raw)
In-Reply-To: <8fcde9bb-284f-f089-96d3-702f501a6258@intel.com>

On 9/3/2020 9:42 AM, Dave Hansen wrote:
> On 9/3/20 9:32 AM, Andy Lutomirski wrote:
>>> Taking the config register out of the init state is illogical, as is
>>> writing to SSP while the config register is in its init state.
>> What's so special about the INIT state?  It's optimized by XSAVES, but
>> it's just a number, right?  So taking the register out of the INIT
>> state is kind of like saying "gdb wanted to set xmm0 to (0,0,0,1), but
>> it was in the INIT state to begin with", right?
> 
> Yeah, that's a good point.  The init state shouldn't be special, as the
> hardware is within its right to choose not to use the init optimization
> at any time.
> 
Then, I would suggest changing get_xsave_addr() to return non-null for 
the INIT state case.  For the other two cases, it still returns NULL. 
But this also requires any write to INIT states to set xstate_bv bits 
properly.  This would be a pitfall for any code addition later on.

Looking at this another way.  Would it be better for the debugger to get 
an error and then to set the MSR directly first (vs. changing the XSAVES 
INIT state first)?