Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.32 as permitted sender) client-ip=23.128.96.32;
From:   "Huang, Ying" <ying.huang@intel.com>
To:     Kairui Song <ryncsn@gmail.com>
Cc:     linux-mm@kvack.org, Andrew Morton <akpm@linux-foundation.org>,
        David Hildenbrand <david@redhat.com>,
        Hugh Dickins <hughd@google.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Matthew Wilcox <willy@infradead.org>,
        Michal Hocko <mhocko@suse.com>, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 16/24] mm/swap: reduce scope of get_swap_device in
 swapin path
In-Reply-To: <CAMgjq7Ba9inFHCqoZSNY_gFUv=UEVEC8FvG+V7-7Qpv027tEmQ@mail.gmail.com>
        (Kairui Song's message of "Thu, 23 Nov 2023 19:13:27 +0800")
References: <20231119194740.94101-1-ryncsn@gmail.com>
        <20231119194740.94101-17-ryncsn@gmail.com>
        <87sf4yaajv.fsf@yhuang6-desk2.ccr.corp.intel.com>
        <CAMgjq7Ba9inFHCqoZSNY_gFUv=UEVEC8FvG+V7-7Qpv027tEmQ@mail.gmail.com>
Date:   Fri, 24 Nov 2023 08:40:56 +0800
Message-ID: <875y1s7zl3.fsf@yhuang6-desk2.ccr.corp.intel.com>
User-Agent: Gnus/5.13 (Gnus v5.13)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Kairui Song <ryncsn@gmail.com> writes:

> Huang, Ying <ying.huang@intel.com> =E4=BA=8E2023=E5=B9=B411=E6=9C=8822=E6=
=97=A5=E5=91=A8=E4=B8=89 08:38=E5=86=99=E9=81=93=EF=BC=9A
>>
>> Kairui Song <ryncsn@gmail.com> writes:
>>
>> > From: Kairui Song <kasong@tencent.com>
>> >
>> > Move get_swap_device into swapin_readahead, simplify the code
>> > and prepare for follow up commits.
>>
>> No.  Please don't do this.  Please check the get/put_swap_device() usage
>> rule in the comments of get_swap_device().
>>
>> "
>>  * When we get a swap entry, if there aren't some other ways to
>>  * prevent swapoff, such as the folio in swap cache is locked, page
>>  * table lock is held, etc., the swap entry may become invalid because
>>  * of swapoff.  Then, we need to enclose all swap related functions
>>  * with get_swap_device() and put_swap_device(), unless the swap
>>  * functions call get/put_swap_device() by themselves.
>> "
>>
>> This is to simplify the reasoning about swapoff and swap entry.
>>
>> Why does it bother you?
>
> Hi Ying,
>
> This is trying to reduce LOC, avoid a trivial si read, and make error
> checking logic easier to refactor in later commits.

The race with swapoff isn't considered by many developers usually.  So,
we should use a simple rule as much as possible.  For example, if you
get a swap entry, use get/put_swap_device() to enclose all code that
operate on the swap entry.  This makes code easier to be maintained in
the long run.  Yes.  Sometimes we break the rule a little, but only if
we have enough benefit, such as improving performance in some practical
use cases.

> And besides there is one trivial change I forgot to include in this
> commit, get_swap_device can be put after swap_cache_get_folio in
> swapin_readahead, since swap_cache_get_folio doesn't need to hold the
> swap device, so in cache hit case this get/put_swap_device() can be
> saved.

swapoff is rare operation, it's OK to delay it a little to make the code
easier to be understood.

> The comment also mentioned:
>
> "Then, we need to enclose all swap related functions with
> get_swap_device() and put_swap_device(), unless the swap functions
> call get/put_swap_device() by themselves"
>
> So I think it should be OK to do this.

No.  You should change the code with a good enough reason.  Compared
with complexity it introduced, the benefit isn't enough for me so far.

--
Best Regards,
Huang, Ying