Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH v3 1/5] mm,memory_hotplug: Allocate memmap from the added
 memory range
To:     Oscar Salvador <osalvador@suse.de>, Zi Yan <ziy@nvidia.com>
Cc:     Andrew Morton <akpm@linux-foundation.org>,
        Michal Hocko <mhocko@kernel.org>,
        Anshuman Khandual <anshuman.khandual@arm.com>,
        Vlastimil Babka <vbabka@suse.cz>,
        Pavel Tatashin <pasha.tatashin@soleen.com>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org,
        Mike Kravetz <mike.kravetz@oracle.com>
References: <20210304100002.7740-1-osalvador@suse.de>
 <20210304100002.7740-2-osalvador@suse.de>
 <830F715B-82B4-4A13-861F-B3A89327F722@nvidia.com>
 <YEYvAzPoaOkrsEaf@localhost.localdomain>
From:   David Hildenbrand <david@redhat.com>
Organization: Red Hat GmbH
Message-ID: <51662b87-35bd-ff18-17dd-b2a99d1095d7@redhat.com>
Date:   Mon, 8 Mar 2021 15:06:49 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.8.0
MIME-Version: 1.0
In-Reply-To: <YEYvAzPoaOkrsEaf@localhost.localdomain>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 08.03.21 15:04, Oscar Salvador wrote:
> On Sun, Mar 07, 2021 at 10:16:36PM -0500, Zi Yan wrote:
>> +Mike for hugetlb discussion.
>>
>> Just thinking about how it might impact gigantic page allocation like hugetlb.
>> When MHP_MEMMAP_ON_MEMORY is on, memmap pages are placed at the beginning
>> of each hot added memory block, so available PFNs from two consecutive
>> hot added memory blocks are not all contiguous, separated by memmap pages.
>> If the memory block size is <= 1GB, there is no way of reserving gigantic
>> pages for hugetlb during runtime using alloc_contig_pages from any hot
>> added memory. Am I getting this right?
> 
> Yes, that is why it is stated both in boot parameter documentation and
> patch changelog that this feature does not play well in those setups
> where your workload is in need of large contiguous chunks of memory,
> that being gigantic hugetlb or just normal memory.
> 
>> I see this implication is documented at the high level in patch 3. Just
>> wonder if we want to be more specific. Or hugetlb is rarely used along
>> with hot-add memory.
> 
> I think it is quite normal to see hugetlb and hotplug operations in the
> same environment.
> One thing excludes the other, just need to be careful when it comes to
> potential pitfalls during offline operations.
> 
> I guess we could mention hugetlb pages in the documentation, if it feels
> it is necesary.

Runtime allocation of gigantic pages without CMA is absolutely 
unreliable either way. IMHO, a tunable for the admin is good enough.

-- 
Thanks,

David / dhildenb