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;
Date:   Wed, 14 Jun 2023 09:12:33 -0700
From:   Jaegeuk Kim <jaegeuk@kernel.org>
To:     Daejun Park <daejun7.park@samsung.com>
Cc:     "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "linux-f2fs-devel@lists.sourceforge.net" 
        <linux-f2fs-devel@lists.sourceforge.net>
Subject: Re: [f2fs-dev] [PATCH] f2fs: do not issue small discard commands
 during checkpoint
Message-ID: <ZInm8cxDnqnD9QrU@google.com>
References: <20230613203947.2745943-1-jaegeuk@kernel.org>
 <CGME20230613204109epcas2p158ecc100d3fe2db1ab2b7ee8335d01e7@epcms2p4>
 <20230614060757epcms2p4e11a8f8bf2fa44eac99a5bb7f47f8dcd@epcms2p4>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20230614060757epcms2p4e11a8f8bf2fa44eac99a5bb7f47f8dcd@epcms2p4>
Precedence: bulk

Hi Daejun,

On 06/14, Daejun Park wrote:
> Hi Jaegeuk,
> 
> > If there're huge # of small discards, this will increase checkpoint latency
> > insanely. Let's issue small discards only by trim.
> > 
> > Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
> > ---
> >  fs/f2fs/segment.c | 5 ++---
> >  1 file changed, 2 insertions(+), 3 deletions(-)
> > 
> > diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> > index 0c0c033c4bdd..ef46bb085385 100644
> > --- a/fs/f2fs/segment.c
> > +++ b/fs/f2fs/segment.c
> > @@ -2178,7 +2178,7 @@ void f2fs_clear_prefree_segments(struct f2fs_sb_info *sbi,
> > ? ? ? ? ?}
> > ? ? ? ? ?mutex_unlock(&dirty_i->seglist_lock);
> >  
> > - ? ? ? ?if (!f2fs_block_unit_discard(sbi))
> > + ? ? ? ?if (!f2fs_block_unit_discard(sbi) || !force)
> 
> If we don't handle the discard entries here, dcc->entry_list will still have them,
> so stale discard entries may be handled by trim, causing incorrect discards to be issued.
> Therefore, I think this patch should also prevent the creation of discard entries
> in add_discard_addrs(), unless it is a checkpoint caused by trim.
> This would further reduce the latency caused by the creation of a discard entry.

I found this causes some objects were not reclaimed when removing the module.
Hence I'm testing v2.

> 
> Thanks,
> Daejun
>