linux-kernel - Re: [PATCH 4/5] mm/mseal: separate out and simplify VMA gap check

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <d0b345c8-29be-4ca4-8243-de16583c93e2@redhat.com>
Date: Mon, 14 Jul 2025 17:25:59 +0200
From: David Hildenbrand <david@...hat.com>
To: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>,
 Pedro Falcato <pfalcato@...e.de>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
 "Liam R . Howlett" <Liam.Howlett@...cle.com>,
 Vlastimil Babka <vbabka@...e.cz>, Jann Horn <jannh@...gle.com>,
 linux-mm@...ck.org, linux-kernel@...r.kernel.org,
 Jeff Xu <jeffxu@...omium.org>
Subject: Re: [PATCH 4/5] mm/mseal: separate out and simplify VMA gap check

On 14.07.25 17:23, Lorenzo Stoakes wrote:
> On Mon, Jul 14, 2025 at 04:17:23PM +0100, Pedro Falcato wrote:
>> On Mon, Jul 14, 2025 at 02:00:39PM +0100, Lorenzo Stoakes wrote:
>>> The check_mm_seal() function is doing something general - checking whether
>>> a range contains only VMAs (or rather that it does NOT contain any unmapped
>>> regions).
>>>
>>> Generalise this and put the logic in mm/vma.c - introducing
>>> range_contains_unmapped(). Additionally we can simplify the logic, we are
>>> simply checking whether the last vma->vm_end has either a VMA starting
>>> after it or ends before the end parameter.
>>>
>>
>> I don't like this. Unless you have any other user for this in mind,
>> we'll proliferate this awful behavior (and add this into core vma code).
> 
> I'm not sure how putting it in an internal-only mm file perpetuates
> anything.
> 
> I'm naming the function by what it does, and putting it where it belongs in
> the VMA logic, and additionally making the function less horrible.
> 
> Let's not please get stuck on the isues with mseal implementation which
> will catch-22 us into not being able to refactor.
> 
> We can do the refactoring first and it's fine to just yank this if it's not
> used.
> 
> I'm not having a function like this sat in mm/mseal.c when it has
> absolutely nothing to do with mseal specifically though.
> 
>>
>> I have some patches locally to fully remove this upfront check, and AFAIK
>> we're somewhat in agreement that we can simply nuke this check (for
>> various reasons, including that we *still* don't have a man page for the
>> syscall). I can send them for proper discussion after your series lands.
> 
> Yes I agree this check is odd, I don't really see why on earth we're
> concerned with whether there are gaps or not, you'd surely want to just
> seal whatever VMAs exist in the range?

Probably because GAPs cannot be sealed. So user space could assume that 
in fact, nothing in that area can change after a successful mseal, while 
it can ...

Not sure, though ...

-- 
Cheers,

David / dhildenb