Re: [PATCH] user_ns: use correct check for single-threadedness

From: ebiederm@xmission.com (Eric W. Biederman)
To: Oleg Nesterov <oleg@redhat.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>,
	Andrew Morton <akpm@linux-foundation.org>,
	Kees Cook <keescook@chromium.org>,
	David Howells <dhowells@redhat.com>,
	linux-kernel@vger.kernel.org,
	Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@kernel.org>,
	"Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
	Rik van Riel <riel@redhat.com>,
	Vladimir Davydov <vdavydov@parallels.com>,
	Ricky Zhou <rickyz@chromium.org>, Julien Tinnes <jln@google.com>
Subject: Re: [PATCH] user_ns: use correct check for single-threadedness
Date: Tue, 11 Aug 2015 20:17:48 -0500	[thread overview]
Message-ID: <87egj9cnk3.fsf@x220.int.ebiederm.org> (raw)
In-Reply-To: <20150806134426.GA6843@redhat.com> (Oleg Nesterov's message of "Thu, 6 Aug 2015 15:44:26 +0200")

Oleg Nesterov <oleg@redhat.com> writes:

> On 08/06, Oleg Nesterov wrote:
>>
>> On 08/05, Eric W. Biederman wrote:
>> >
>> > So I have to ask.
>>
>> I hope you are asking someone else, not me ;) I never understood what
>> exactly we try to restrict and why.
>>
>> > Is it possible to rework these checks such that we
>> > look at the sighand struct and signal sharing handling sharing instead
>> > of the count on the mm_struct?
>>
>> Then why we can't simply check thread_group_empty() == T ? Why should we
>> worry about CLONE_SIGHAND at all?
>
> The same for clone() actually... I forgot why we decided to check
> CLONE_SIGHAND, iirc I suggested CLONE_THREAD initially then we switched
> to CLONE_SIGHAND "just in case", to make it as strict as possible.

I do agree that making the test be for CLONE_THREAD is safe, makes
sense, and is less confusing than what we have now.x

> How about the patch below?
>
> (note that the "or parent" part of the comment is wrong in any case).

It was correct.  You failed to removed it when you removed CLONE_PARENT
from that test.

Eric