[PATCH V2] arm64/mm: Intercept pfn changes in set_pte_at()

[Robin Murphy]

On 2023-01-26 13:33, Will Deacon wrote:
> On Tue, Jan 24, 2023 at 11:11:49AM +0530, Anshuman Khandual wrote:
>> On 1/9/23 10:58, Anshuman Khandual wrote:
>>> Changing pfn on a user page table mapped entry, without first going through
>>> break-before-make (BBM) procedure is unsafe. This just updates set_pte_at()
>>> to intercept such changes, via an updated pgattr_change_is_safe(). This new
>>> check happens via __check_racy_pte_update(), which has now been renamed as
>>> __check_safe_pte_update().
>>>
>>> Cc: Catalin Marinas <catalin.marinas at arm.com>
>>> Cc: Will Deacon <will at kernel.org>
>>> Cc: Mark Rutland <mark.rutland at arm.com>
>>> Cc: Andrew Morton <akpm at linux-foundation.org>
>>> Cc: linux-arm-kernel at lists.infradead.org
>>> Cc: linux-kernel at vger.kernel.org
>>> Signed-off-by: Anshuman Khandual <anshuman.khandual at arm.com>
>>> ---
>>> This applies on v6.2-rc3. This patch had some test time on an internal CI
>>> system without any issues being reported.
>>
>> Gentle ping, any updates on this patch ? Still any concerns ?
> 
> I don't think we really got to the bottom of Mark's concerns with
> unreachable ptes on the stack, did we? I also have vague recollections
> of somebody (Robin?) running into issues with the vmap code not honouring
> BBM.

Doesn't ring a bell, so either it wasn't me, or it was many years ago 
and about 5 levels deep into trying to fix something else :/

> So I think we should confirm/fix the vmap issue before we enable this check
> and also try to get some testing coverage to address Mark's worries. I think
> he has a syzkaller instance set up, so that sound like a good place to
> start.

I think we're also missing a subtlety here in that this restriction 
doesn't *always* apply. For instance if someone wants to move a page by 
making the mapping read-only, copying the contents to a new page, then 
pointing the RO mapping at that new page, that should technically not 
require BBM.

Thanks,
Robin.