Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH 3/6] mremap: Don't allow MREMAP_DONTUNMAP on
 special_mappings and aio
To:     Brian Geffon <bgeffon@google.com>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        Dmitry Safonov <0x7f454c46@gmail.com>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Andrew Morton <akpm@linux-foundation.org>,
        Andy Lutomirski <luto@kernel.org>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Dan Williams <dan.j.williams@intel.com>,
        Dave Jiang <dave.jiang@intel.com>,
        Hugh Dickins <hughd@google.com>,
        Ingo Molnar <mingo@redhat.com>,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        Minchan Kim <minchan@kernel.org>,
        Russell King <linux@armlinux.org.uk>,
        Thomas Bogendoerfer <tsbogend@alpha.franken.de>,
        Thomas Gleixner <tglx@linutronix.de>,
        Vishal Verma <vishal.l.verma@intel.com>,
        Vlastimil Babka <vbabka@suse.cz>,
        Will Deacon <will@kernel.org>, linux-aio@kvack.org,
        linux-fsdevel@vger.kernel.org, linux-mm <linux-mm@kvack.org>
References: <20201013013416.390574-1-dima@arista.com>
 <20201013013416.390574-4-dima@arista.com>
 <CADyq12ww2SB=x16pdH4LBZJJxMakOWgkR0qX-maUe-RzYZ491Q@mail.gmail.com>
From:   Dmitry Safonov <dima@arista.com>
Message-ID: <d5d608ba-a25c-d3a2-b0f4-c97437a6dab1@arista.com>
Date:   Mon, 28 Dec 2020 19:33:13 +0000
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.12.0
MIME-Version: 1.0
In-Reply-To: <CADyq12ww2SB=x16pdH4LBZJJxMakOWgkR0qX-maUe-RzYZ491Q@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

[I moved your reply to avoid top-posting]

On 12/28/20 6:03 PM, Brian Geffon wrote:
> On Mon, Oct 12, 2020 at 6:34 PM Dmitry Safonov <dima@arista.com> wrote:
>>
>> As kernel expect to see only one of such mappings, any further
>> operations on the VMA-copy may be unexpected by the kernel.
>> Maybe it's being on the safe side, but there doesn't seem to be any
>> expected use-case for this, so restrict it now.
>>
>> Fixes: commit e346b3813067 ("mm/mremap: add MREMAP_DONTUNMAP to mremap()")
>> Signed-off-by: Dmitry Safonov <dima@arista.com>
>
> I don't think this situation can ever happen MREMAP_DONTUNMAP is
> already restricted to anonymous mappings (defined as not having
> vm_ops) and vma_to_resize checks that the mapping is anonymous before
> move_vma is called.

I've looked again now, I think it is possible. One can call
MREMAP_DONTUNMAP without MREMAP_FIXED and without resizing. So that the
old VMA is copied at some free address.

The calltrace would be: mremap()=>move_vma()
[under if (flags & MREMAP_MAYMOVE)].

On the other side I agree with you that the fix could have been better
if I realized the semantics that MREMAP_DONTUNMAP should only work with
anonymous mappings.

Probably, a better fix would be to move
:       if (flags & MREMAP_DONTUNMAP && (!vma_is_anonymous(vma) ||
:                       vma->vm_flags & VM_SHARED))
:               return ERR_PTR(-EINVAL);

from vma_to_resize() into the mremap() syscall directly.
What do you think?

-- 
          Dima