From: Jan Kara Subject: Re: [PATCH 3/8] ext3: remove lock/unlock super Date: Thu, 16 Aug 2012 18:39:04 +0200 Message-ID: <20120816163904.GA17526@quack.suse.cz> References: <502CC4BA.3040702@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: bharrosh@panasas.com, bhalevy@tonian.com, jack@suse.cz, Andrew Morton , adilger.kernel@dilger.ca, tytso@mit.edu, hirofumi@mail.parknet.co.jp, mikulas@artax.karlin.mff.cuni.cz, Al Viro , hch@infradead.org, dushistov@mail.ru, osd-dev@open-osd.org, Linux Kernel , linux-ext4@vger.kernel.org, Linux FS Devel To: Marco Stornelli Return-path: Content-Disposition: inline In-Reply-To: <502CC4BA.3040702@gmail.com> Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-ext4.vger.kernel.org On Thu 16-08-12 12:00:26, Marco Stornelli wrote: > From: Marco Stornelli > > Replaced lock and unlock super with a new fs mutex s_lock. Hum, is the lock needed at all? Remount & unfreeze both run with s_umount held for writing. Thus we already have exclusion between these two calls. The same seems to hold for ext4 BTW. Honza > > Signed-off-by: Marco Stornelli > --- > > diff -Nurp linux-3.6-rc1-orig/fs/ext3/ext3.h linux-3.6-rc1/fs/ext3/ext3.h > --- linux-3.6-rc1-orig/fs/ext3/ext3.h 2012-08-16 09:37:31.000000000 +0200 > +++ linux-3.6-rc1/fs/ext3/ext3.h 2012-08-16 09:46:39.000000000 +0200 > @@ -672,6 +672,7 @@ struct ext3_sb_info { > char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */ > int s_jquota_fmt; /* Format of quota to use */ > #endif > + struct mutex s_lock; > }; > static inline spinlock_t * > diff -Nurp linux-3.6-rc1-orig/fs/ext3/super.c linux-3.6-rc1/fs/ext3/super.c > --- linux-3.6-rc1-orig/fs/ext3/super.c 2012-08-16 09:37:31.000000000 +0200 > +++ linux-3.6-rc1/fs/ext3/super.c 2012-08-16 09:49:11.000000000 +0200 > @@ -1939,6 +1939,7 @@ static int ext3_fill_super (struct super > sbi->s_gdb_count = db_count; > get_random_bytes(&sbi->s_next_generation, sizeof(u32)); > spin_lock_init(&sbi->s_next_gen_lock); > + mutex_init(&sbi->s_lock); > /* per fileystem reservation list head & lock */ > spin_lock_init(&sbi->s_rsv_window_lock); > @@ -2582,11 +2583,11 @@ out: > static int ext3_unfreeze(struct super_block *sb) > { > if (!(sb->s_flags & MS_RDONLY)) { > - lock_super(sb); > + mutex_lock(&EXT3_SB(sb)->s_lock); > /* Reser the needs_recovery flag before the fs is unlocked. */ > EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); > ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1); > - unlock_super(sb); > + mutex_unlock(&EXT3_SB(sb)->s_lock); > journal_unlock_updates(EXT3_SB(sb)->s_journal); > } > return 0; > @@ -2606,7 +2607,7 @@ static int ext3_remount (struct super_bl > #endif > /* Store the original options */ > - lock_super(sb); > + mutex_lock(&EXT3_SB(sb)->s_lock); > old_sb_flags = sb->s_flags; > old_opts.s_mount_opt = sbi->s_mount_opt; > old_opts.s_resuid = sbi->s_resuid; > @@ -2712,7 +2713,7 @@ static int ext3_remount (struct super_bl > old_opts.s_qf_names[i] != sbi->s_qf_names[i]) > kfree(old_opts.s_qf_names[i]); > #endif > - unlock_super(sb); > + mutex_unlock(&EXT3_SB(sb)->s_lock); > if (enable_quota) > dquot_resume(sb, -1); > @@ -2732,7 +2733,7 @@ restore_opts: > sbi->s_qf_names[i] = old_opts.s_qf_names[i]; > } > #endif > - unlock_super(sb); > + mutex_unlock(&EXT3_SB(sb)->s_lock); > return err; > } > -- Jan Kara SUSE Labs, CR