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: <2e20c90d-4d1f-dd83-aa63-9d8d17021263@redhat.com>
Date:   Tue, 30 Aug 2022 20:23:52 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.12.0
Subject: Re: [PATCH v1 2/3] mm/gup: use gup_can_follow_protnone() also in
 GUP-fast
Content-Language: en-US
From:   David Hildenbrand <david@redhat.com>
To:     linux-kernel@vger.kernel.org
Cc:     linux-mm@kvack.org, Andrew Morton <akpm@linux-foundation.org>,
        Mel Gorman <mgorman@suse.de>, Jason Gunthorpe <jgg@nvidia.com>,
        John Hubbard <jhubbard@nvidia.com>,
        "Matthew Wilcox (Oracle)" <willy@infradead.org>,
        Andrea Arcangeli <aarcange@redhat.com>,
        Hugh Dickins <hughd@google.com>, Peter Xu <peterx@redhat.com>
References: <20220825164659.89824-1-david@redhat.com>
 <20220825164659.89824-3-david@redhat.com>
 <1892f6de-fd22-0e8b-3ff6-4c8641e1c68e@redhat.com>
Organization: Red Hat
In-Reply-To: <1892f6de-fd22-0e8b-3ff6-4c8641e1c68e@redhat.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 26.08.22 16:59, David Hildenbrand wrote:
> On 25.08.22 18:46, David Hildenbrand wrote:
>> There seems to be no reason why FOLL_FORCE during GUP-fast would have to
>> fallback to the slow path when stumbling over a PROT_NONE mapped page. We
>> only have to trigger hinting faults in case FOLL_FORCE is not set, and any
>> kind of fault handling naturally happens from the slow path -- where
>> NUMA hinting accounting/handling would be performed.
>>
>> Note that the comment regarding THP migration is outdated:
>> commit 2b4847e73004 ("mm: numa: serialise parallel get_user_page against
>> THP migration") described that this was required for THP due to lack of PMD
>> migration entries. Nowadays, we do have proper PMD migration entries in
>> place -- see set_pmd_migration_entry(), which does a proper
>> pmdp_invalidate() when placing the migration entry.
>>
>> So let's just reuse gup_can_follow_protnone() here to make it
>> consistent and drop the somewhat outdated comments.
>>
>> Signed-off-by: David Hildenbrand <david@redhat.com>
>> ---
>>  mm/gup.c | 14 +++-----------
>>  1 file changed, 3 insertions(+), 11 deletions(-)
>>
>> diff --git a/mm/gup.c b/mm/gup.c
>> index a1355dbd848e..dfef23071dc8 100644
>> --- a/mm/gup.c
>> +++ b/mm/gup.c
>> @@ -2350,11 +2350,7 @@ static int gup_pte_range(pmd_t pmd, unsigned long addr, unsigned long end,
>>  		struct page *page;
>>  		struct folio *folio;
>>  
>> -		/*
>> -		 * Similar to the PMD case below, NUMA hinting must take slow
>> -		 * path using the pte_protnone check.
>> -		 */
>> -		if (pte_protnone(pte))
>> +		if (pte_protnone(pte) && !gup_can_follow_protnone(flags))
>>  			goto pte_unmap;
>>  
>>  		if (!pte_access_permitted(pte, flags & FOLL_WRITE))
>> @@ -2736,12 +2732,8 @@ static int gup_pmd_range(pud_t *pudp, pud_t pud, unsigned long addr, unsigned lo
>>  
>>  		if (unlikely(pmd_trans_huge(pmd) || pmd_huge(pmd) ||
>>  			     pmd_devmap(pmd))) {
>> -			/*
>> -			 * NUMA hinting faults need to be handled in the GUP
>> -			 * slowpath for accounting purposes and so that they
>> -			 * can be serialised against THP migration.
>> -			 */
>> -			if (pmd_protnone(pmd))
>> +			if (pmd_protnone(pmd) &&
>> +			    !gup_can_follow_protnone(flags))
>>  				return 0;
>>  
>>  			if (!gup_huge_pmd(pmd, pmdp, addr, next, flags,
> 
> 
> I just stumbled over something interesting. If we have a pte_protnone()
> entry, ptep_clear_flush() might not flush, because the !pte_accessible()
>  does not hold.
> 
> Consequently, we could be in trouble when using ptep_clear_flush() on a
> pte_protnone() PTE to make sure that GUP cannot run anymore.
> 
> Will give this a better thought, but most probably I'll replace this
> patch by a proper documentation update here.

... and looking into the details of TLB flush and GUP-fast interaction
nowadays, that case is no longer relevant. A TLB flush is no longer
sufficient to stop concurrent GUP-fast ever since we introduced generic
RCU GUP-fast.

-- 
Thanks,

David / dhildenb