Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20201031200518.4178786-1-harshadshirwadkar@gmail.com>
 <20201031200518.4178786-3-harshadshirwadkar@gmail.com> <20201103141331.GF3440@quack2.suse.cz>
In-Reply-To: <20201103141331.GF3440@quack2.suse.cz>
From:   harshad shirwadkar <harshadshirwadkar@gmail.com>
Date:   Tue, 3 Nov 2020 10:33:47 -0800
Message-ID: <CAD+ocbyXyjA9AKS-us4dFmA=ExdFQttYeXH2bJ8bQUAm0qYRDg@mail.gmail.com>
Subject: Re: [PATCH 02/10] ext4: mark fc ineligible if inode gets evictied due
 to mem pressure
To:     Jan Kara <jack@suse.cz>
Cc:     Ext4 Developers List <linux-ext4@vger.kernel.org>,
        "Theodore Y. Ts'o" <tytso@mit.edu>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Nov 3, 2020 at 6:13 AM Jan Kara <jack@suse.cz> wrote:
>
> On Sat 31-10-20 13:05:10, Harshad Shirwadkar wrote:
> > If inode gets evicted due to memory pressure, we have to remove it
> > from the fast commit list. However, that inode may have uncommitted
> > changes that fast commits will lose. So, just fall back to full
> > commits in this case. Also, rename the fast commit ineligiblity reason
> > from "EXT4_FC_REASON_MEM" to "EXT4_FC_REASON_MEM_CRUNCH" for better
> > expression.
> >
> > Suggested-by: Jan Kara <jack@suse.cz>
> > Signed-off-by: Harshad Shirwadkar <harshadshirwadkar@gmail.com>
>
> ...
>
> > diff --git a/fs/ext4/fast_commit.h b/fs/ext4/fast_commit.h
> > index 06907d485989..cde86747faf8 100644
> > --- a/fs/ext4/fast_commit.h
> > +++ b/fs/ext4/fast_commit.h
> > @@ -100,7 +100,7 @@ enum {
> >       EXT4_FC_REASON_XATTR = 0,
> >       EXT4_FC_REASON_CROSS_RENAME,
> >       EXT4_FC_REASON_JOURNAL_FLAG_CHANGE,
> > -     EXT4_FC_REASON_MEM,
> > +     EXT4_FC_REASON_MEM_CRUNCH,
>
> Well MEM_CRUNCH doesn't really sound more understandable to me :). I'd
> rather call it MEM_RECLAIM or ENOMEM or something like that...
Okay, ENOMEM sounds good, since this is also used in case of memory
allocation failures.
>
> > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> > index b96a18679a27..52ff71236290 100644
> > --- a/fs/ext4/inode.c
> > +++ b/fs/ext4/inode.c
> > @@ -327,6 +327,7 @@ void ext4_evict_inode(struct inode *inode)
> >       ext4_xattr_inode_array_free(ea_inode_array);
> >       return;
> >  no_delete:
> > +     ext4_fc_mark_ineligible(inode->i_sb, EXT4_FC_REASON_MEM_CRUNCH);
> >       ext4_clear_inode(inode);        /* We must guarantee clearing of inode... */
> >  }
>
> This will make fs ineligible on every inode reclaim. Even if the inode was
> clean, not part of any FC. I guess this is too aggressive...
Right, I missed that, so first checking if the inode is on FC list and
then marking the FS as ineligible should suffice?

Thanks,
Harshad
>
>                                                                 Honza
> --
> Jan Kara <jack@suse.com>
> SUSE Labs, CR