Received-SPF: pass (google.com: domain of linux-kernel+bounces-61673-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Message-ID: <e3068889-3991-429c-8a08-b23a1b2d368a@bytedance.com>
Date: Mon, 12 Feb 2024 21:20:19 +0800
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH 1/2] mm/zswap: global lru and shrinker shared by all
 zswap_pools
Content-Language: en-US
To: Nhat Pham <nphamcs@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
 Johannes Weiner <hannes@cmpxchg.org>, Yosry Ahmed <yosryahmed@google.com>,
 linux-mm@kvack.org, linux-kernel@vger.kernel.org
References: <20240210-zswap-global-lru-v1-0-853473d7b0da@bytedance.com>
 <20240210-zswap-global-lru-v1-1-853473d7b0da@bytedance.com>
 <CAKEwX=Nbbm=BZ8Rp7Knfk3P=jXK2SJnhgR_vOK9w1ovPUdF9UA@mail.gmail.com>
From: Chengming Zhou <zhouchengming@bytedance.com>
In-Reply-To: <CAKEwX=Nbbm=BZ8Rp7Knfk3P=jXK2SJnhgR_vOK9w1ovPUdF9UA@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

On 2024/2/12 05:04, Nhat Pham wrote:
> On Sun, Feb 11, 2024 at 5:57 AM Chengming Zhou
> <zhouchengming@bytedance.com> wrote:
>>
>> Dynamic zswap_pool creation may create/reuse to have multiple
>> zswap_pools in a list, only the first will be current used.
>>
>> Each zswap_pool has its own lru and shrinker, which is not
>> necessary and has its problem:
>>
>> 1. When memory has pressure, all shrinker of zswap_pools will
>>    try to shrink its own lru, there is no order between them.
>>
>> 2. When zswap limit hit, only the last zswap_pool's shrink_work
>>    will try to shrink its lru, which is inefficient.
>>
>> Anyway, having a global lru and shrinker shared by all zswap_pools
>> is better and efficient.
>>
>> Signed-off-by: Chengming Zhou <zhouchengming@bytedance.com>
> 
> I'll do a careful review later, but IMO this is a good idea :)

Ok, thanks, take your time. :)

> 
> Chris pointed out when he reviewed the zswap shrinker patch series
> that the reclaim algorithm has to decide which pool to reclaim from,
> and I have always thought that it was a bit weird that we have to do
> it at all. We should reclaim stored objects by access ordering,
> irregardless of which pool it belongs to. Having a shared LRU and
> other associated reclaim structures is sound, and saves a bit of space
> too while we're at it.

Right, agree!