Re: [PATCH] jbd2: move more common code into journal_init_common()

From: Geliang Tang
Date: Sat Sep 17 2016 - 22:39:36 EST


On Thu, Sep 15, 2016 at 03:58:18PM -0400, Theodore Ts'o wrote:
> On Thu, Sep 15, 2016 at 12:03:09PM -0400, Theodore Ts'o wrote:
> > On Wed, Sep 07, 2016 at 03:16:24PM +0200, Jan Kara wrote:
> > > On Wed 07-09-16 20:41:13, Geliang Tang wrote:
> > > > There are some repetitive code in jbd2_journal_init_dev() and
> > > > jbd2_journal_init_inode(). So this patch moves the common code into
> > > > journal_init_common() helper to simplify the code. And fix the coding
> > > > style warnings reported by checkpatch.pl by the way.
> > > >
> > > > Signed-off-by: Geliang Tang <geliangtang@xxxxxxxxx>
> > >
> > > The patch looks good to me. You can add:
> > >
> > > Reviewed-by: Jan Kara <jack@xxxxxxx>
> >
> > Applied, thanks.
> >
>
> Hi Geiliang,
>
> This patch is causing a WARN_ON:
>
> [ 13.923139] ------------[ cut here ]------------
> [ 13.924644] WARNING: CPU: 0 PID: 2534 at /usr/projects/linux/ext4/fs/proc/generic.c:369 __proc_create+0xe1/0x156
> [ 13.926751] name len 0
> [ 13.927283] Modules linked in:
> [ 13.927954] CPU: 0 PID: 2534 Comm: mount Tainted: G W 4.8.0-rc4-00139-g52c4278 #685
> [ 13.929809] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Debian-1.8.2-1 04/01/2014
> [ 13.931643] 00000000 00000246 f3c1dd3c c13faa3e f3c1dd68 c11cc9c1 f3c1dd54 c1087d00
> [ 13.933425] 00000171 f4727b0c f3c1ddac f5183bc0 f3c1dd70 c1087d44 00000009 00000000
> [ 13.935199] f3c1dd68 c1969704 f3c1dd84 f3c1dda0 c11cc9c1 c19696d9 00000171 c1969704
> [ 13.936987] Call Trace:
> [ 13.937507] [<c13faa3e>] dump_stack+0x73/0xa5
> [ 13.938438] [<c11cc9c1>] ? __proc_create+0xe1/0x156
> [ 13.939503] [<c1087d00>] __warn+0xbc/0xd3
> [ 13.940404] [<c1087d44>] warn_slowpath_fmt+0x2d/0x32
> [ 13.941470] [<c11cc9c1>] __proc_create+0xe1/0x156
> [ 13.942461] [<c11ccc8d>] proc_mkdir_data+0x2c/0x6e
> [ 13.943484] [<c11cccf6>] proc_mkdir+0x13/0x15
> [ 13.944425] [<c122a597>] journal_init_common+0x1a8/0x26a
> [ 13.945539] [<c122a753>] jbd2_journal_init_inode+0xa9/0xfd
> [ 13.946702] [<c11fa5cf>] ext4_fill_super+0x18e5/0x2a92
> [ 13.947794] [<c1402d01>] ? bitmap_string.isra.6+0xa9/0xc1
> [ 13.948926] [<c117ec1f>] mount_bdev+0x114/0x15f
> [ 13.950333] [<c11f48c4>] ext4_mount+0x15/0x17
> [ 13.951985] [<c11f8cea>] ? ext4_calculate_overhead+0x30e/0x30e
> [ 13.954172] [<c117f49d>] mount_fs+0x58/0x115
> [ 13.955789] [<c11943cb>] vfs_kern_mount+0x4c/0xae
> [ 13.957588] [<c11966dc>] do_mount+0x6b0/0x8d7
> [ 13.959270] [<c1405620>] ? _copy_from_user+0x44/0x57
> [ 13.961140] [<c1150873>] ? strndup_user+0x31/0x42
> [ 13.962919] [<c1196aab>] SyS_mount+0x57/0x7b
> [ 13.964609] [<c10015b2>] do_int80_syscall_32+0x4d/0x5f
> [ 13.966555] [<c16f6ccb>] entry_INT80_32+0x2f/0x2f
> [ 13.968402] ---[ end trace 2eb7cc6d9a94f309 ]---
> [ 14.017482] EXT4-fs (vdc): mounted filesystem with ordered data mode. Opts: (null)
>
> The problem is that journal->j_devname isn't initialized until *after*
> journal_init_common(), and it's used by jbd2_stats_proc_init(), which
> is called by journal_init_common().
>
> To fix it I applied the following fix on top of your patch.
>
> - Ted
>
> diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
> index 07e14ef..927da49 100644
> --- a/fs/jbd2/journal.c
> +++ b/fs/jbd2/journal.c
> @@ -1141,13 +1141,11 @@ static journal_t *journal_init_common(struct block_device *bdev,
> journal->j_fs_dev = fs_dev;
> journal->j_blk_offset = start;
> journal->j_maxlen = len;
> - jbd2_stats_proc_init(journal);
> n = journal->j_blocksize / sizeof(journal_block_tag_t);
> journal->j_wbufsize = n;
> journal->j_wbuf = kmalloc_array(n, sizeof(struct buffer_head *),
> GFP_KERNEL);
> if (!journal->j_wbuf) {
> - jbd2_stats_proc_exit(journal);
> kfree(journal);
> return NULL;
> }
> @@ -1157,7 +1155,6 @@ static journal_t *journal_init_common(struct block_device *bdev,
> pr_err("%s: Cannot get buffer for journal superblock\n",
> __func__);
> kfree(journal->j_wbuf);
> - jbd2_stats_proc_exit(journal);
> kfree(journal);
> return NULL;
> }
> @@ -1202,6 +1199,7 @@ journal_t *jbd2_journal_init_dev(struct block_device *bdev,
>
> bdevname(journal->j_dev, journal->j_devname);
> strreplace(journal->j_devname, '/', '!');
> + jbd2_stats_proc_init(journal);
>
> return journal;
> }
> @@ -1241,6 +1239,7 @@ journal_t *jbd2_journal_init_inode(struct inode *inode)
> bdevname(journal->j_dev, journal->j_devname);
> p = strreplace(journal->j_devname, '/', '!');
> sprintf(p, "-%lu", journal->j_inode->i_ino);
> + jbd2_stats_proc_init(journal);
>
> return journal;
> }

Thanks Ted, this looks fine to me.

-Geliang