Re: [PATCH v4] mm/mempolicy,hugetlb: Checking hstate for hugetlbfs page in vma_migratable

From: Anshuman Khandual <anshuman.khandual@arm.com>
To: Michal Hocko <mhocko@kernel.org>, Li Xinhai <lixinhai.lxh@gmail.com>
Cc: n-horiguchi <n-horiguchi@ah.jp.nec.com>,
	"linux-mm@kvack.org" <linux-mm@kvack.org>,
	akpm <akpm@linux-foundation.org>,
	torvalds <torvalds@linux-foundation.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	Mike Kravetz <mike.kravetz@oracle.com>
Subject: Re: [PATCH v4] mm/mempolicy,hugetlb: Checking hstate for hugetlbfs page in vma_migratable
Date: Tue, 21 Jan 2020 09:12:04 +0530	[thread overview]
Message-ID: <3cef881e-8685-4917-1784-286dc3b11bf6@arm.com> (raw)
In-Reply-To: <20200120160500.GM18451@dhcp22.suse.cz>

On 01/20/2020 09:35 PM, Michal Hocko wrote:
> On Mon 20-01-20 23:37:25, Li Xinhai wrote:
> [...]
>> Changelog is updated as below, thanks for comments:
>> ---
>> mm/mempolicy: Checking hugepage migration is supported by arch in vma_migratable
>>
>> vma_migratable() is called to check if pages in vma can be migrated
>> before go ahead to further actions. Currently it is used in below code
>> path:
>> - task_numa_work
>> - mbind
>> - move_pages
>>
>> For hugetlb mapping, whether vma is migratable or not is determined by:
>> - CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
>> - arch_hugetlb_migration_supported
>>
>> Issue: current code only checks for CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION,
>> which express less accurate semantics of vma_migratable(). (note that
>> current code in vma_migratable don't cause failure or bug because
>> unmap_and_move_huge_page() will catch unsupported hugepage and handle it
>> properly)
>>
>> This patch checks the two factors for impoveing code logic and
>> robustness. It will enable early bail out of hugepage migration procedure,
>> but because currently all architecture supporting hugepage migration is able
>> to support all page size, we would not see performance gain with this patch
>> applied.
> 
> This looks definitely better than the original one. I hope it is more
> clear to you what I meant by a better description for the justification.
> I would just add that the no code should use
> CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION directly and use
> arch_hugetlb_migration_supported instead. This will be the case after
> this patch.

As I have mentioned previously on the other thread, there might be an case
to keep the existing code (just added with a comment) which will preserve
the performance. But the proposed method will do it the right way and also
get rid of CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION here. Its OK either way.

> 
> Please keep in mind that changelogs are really important and growing in
> importance as the code gets more complicated over time. It is much more
> easier to see what the patch does because reading diffs and the code is
> easy but the lack of motivation is what people usually fighting with.
> 