Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <38574ed3-ea96-a72e-00dd-4e6204413a86@redhat.com>
Date:   Wed, 14 Jun 2023 17:47:31 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.12.0
Subject: Re: [PATCH 4/7] mm/hugetlb: Prepare hugetlb_follow_page_mask() for
 FOLL_PIN
Content-Language: en-US
To:     Peter Xu <peterx@redhat.com>
Cc:     linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        Matthew Wilcox <willy@infradead.org>,
        Andrea Arcangeli <aarcange@redhat.com>,
        John Hubbard <jhubbard@nvidia.com>,
        Mike Rapoport <rppt@kernel.org>,
        Vlastimil Babka <vbabka@suse.cz>,
        "Kirill A . Shutemov" <kirill@shutemov.name>,
        Andrew Morton <akpm@linux-foundation.org>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        James Houghton <jthoughton@google.com>,
        Hugh Dickins <hughd@google.com>
References: <20230613215346.1022773-1-peterx@redhat.com>
 <20230613215346.1022773-5-peterx@redhat.com>
 <533c32cf-9a18-1590-4d29-f076d6bd58c1@redhat.com> <ZInYh3cVUil9R/cf@x1n>
 <ef0f8e0e-cbce-7c7b-1b0e-c9d52ede7e0e@redhat.com> <ZIndN9isc4pTp2zK@x1n>
From:   David Hildenbrand <david@redhat.com>
Organization: Red Hat
In-Reply-To: <ZIndN9isc4pTp2zK@x1n>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 14.06.23 17:31, Peter Xu wrote:
> On Wed, Jun 14, 2023 at 05:17:13PM +0200, David Hildenbrand wrote:
>> On 14.06.23 17:11, Peter Xu wrote:
>>> On Wed, Jun 14, 2023 at 04:57:37PM +0200, David Hildenbrand wrote:
>>>> On 13.06.23 23:53, Peter Xu wrote:
>>>>> It's coming, not yet, but soon.  Loose the restriction.
>>>>>
>>>>> Signed-off-by: Peter Xu <peterx@redhat.com>
>>>>> ---
>>>>>     mm/hugetlb.c | 7 -------
>>>>>     1 file changed, 7 deletions(-)
>>>>>
>>>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>>>> index f037eaf9d819..31d8f18bc2e4 100644
>>>>> --- a/mm/hugetlb.c
>>>>> +++ b/mm/hugetlb.c
>>>>> @@ -6467,13 +6467,6 @@ struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma,
>>>>>     	spinlock_t *ptl;
>>>>>     	pte_t *pte, entry;
>>>>> -	/*
>>>>> -	 * FOLL_PIN is not supported for follow_page(). Ordinary GUP goes via
>>>>> -	 * follow_hugetlb_page().
>>>>> -	 */
>>>>> -	if (WARN_ON_ONCE(flags & FOLL_PIN))
>>>>> -		return NULL;
>>>>> -
>>>>>     	hugetlb_vma_lock_read(vma);
>>>>>     	pte = hugetlb_walk(vma, haddr, huge_page_size(h));
>>>>>     	if (!pte)
>>>> Did you fix why the warning was placed there in the first place? (IIRC, at
>>>> least unsharing support needs to be added, maybe more)
>>>
>>> Feel free to have a look at patch 2 - it should be done there, hopefully in
>>> the right way.  And IIUC it could be a bug to not do that before (besides
>>> CoR there was also the pgtable permission checks that was missing).  More
>>> details in patch 2's commit message.  Thanks,
>>
>> Oh, that slipped my eyes (unsharing is not really a permission check) -- and
> 
> I think it is still a "permission check"?  It means, we forbid anyone R/O
> taking the page if it's not exclusively owned, just like we forbid anyone
> RW taking the page if it's not writable?

Agreed, just not in the traditional PTE-protection case.

> 
> It's just that the permission check only applies to PIN which follow_page()
> doesn't yet care, so it won't ever trigger.
> 
>> the patch description could have been more explicit about why we can now
>> lift the restrictions.
>>
>> For the records: we don't use CoR terminology upstream. As suggested by
>> John, we use "GUP-triggered unsharing".
> 
> Sure.
> 
>>
>> As unsharing only applies to FOLL_PIN, it doesn't quite fit into patch #2.
>> Either move that to this patch or squash both.
> 
> Sure, no strong opinions here.
> 
> The plan is _if_ someone wants to backport patch 2, this patch should not
> be part of it.  But then maybe it makes more sense to move the CoR change
> there into this one, not because "it's not permission check", but because
> CoR is not relevant in follow_page(), so not relevant to a backport.

Right. Then just call patch #2 "Add missing write-permission check" and 
this patch "Support FOLL_PIN in hugetlb_follow_page_mask()" or sth. like 
that.

Regarding the backport, I really wonder if patch #2 is required at all, 
because I didn't sport any applicable FOLL_WRITE users. Maybe there were 
some? Hm. If it's not applicable, a single "Support FOLL_PIN in 
hugetlb_follow_page_mask()" patch might be cleanest.

-- 
Cheers,

David / dhildenb