Re: [PATCH v27 24/31] x86/cet/shstk: Handle thread shadow stack

From: Dave Hansen <dave.hansen@intel.com>
To: John Allen <john.allen@amd.com>, Yu-cheng Yu <yu-cheng.yu@intel.com>
Cc: 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@kernel.org>,
	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>, Jann Horn <jannh@google.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>,
	Pengfei Xu <pengfei.xu@intel.com>,
	Haitao Huang <haitao.huang@intel.com>
Subject: Re: [PATCH v27 24/31] x86/cet/shstk: Handle thread shadow stack
Date: Wed, 21 Jul 2021 11:37:59 -0700	[thread overview]
Message-ID: <cd064202-2c5f-b1d5-2970-9bff0a762a95@intel.com> (raw)
In-Reply-To: <YPhkIHJ0guc4UNoO@AUS-LX-JohALLEN.amd.com>

On 7/21/21 11:14 AM, John Allen wrote:
>> +int shstk_alloc_thread_stack(struct task_struct *tsk, unsigned long clone_flags,
>> +			     unsigned long stack_size)
>> +{
>> +	struct thread_shstk *shstk = &tsk->thread.shstk;
>> +	struct cet_user_state *state;
>> +	unsigned long addr;
>> +
>> +	if (!stack_size)
>> +		return -EINVAL;
> I've been doing some light testing on AMD hardware and I've found that
> this version of the patchset doesn't boot for me. It appears that when
> systemd processes start spawning, they hit the above case, return
> -EINVAL, and the fork fails. In these cases, copy_thread has been passed
> 0 for both sp and stack_size.

A few tangential things I noticed:

This hunk is not mentioned in the version changelog at all.  I also
don't see any feedback that might have prompted it.  This is one reason
per-patch changelogs are preferred.

As a general rule, new features should strive to be implemented in a way
that it's *obvious* that they won't break old code.
shstk_alloc_thread_stack() fails that test for me.  If it had:

	if (!cpu_feature_enabled(X86_FEATURE_SHSTK)) // or whatever
		return 0;

in the function, it would be obviously harmless.  Better yet would be
doing the feature check at the shstk_alloc_thread_stack() call site,
that way even the function call can be optimized out.

Further, this confused me because the changelog didn't even mention the
arg -> stack_size rename.  That would have been nice for another patch,
or an extra sentence in the changelog.