Re: [PATCH V2 12/23] xfs_db: Add support to read from external log device

[Chandan Babu R <chandan.babu@oracle.com>]

On Wed, Jul 12, 2023 at 10:35:29 AM -0700, Darrick J. Wong wrote:
> On Tue, Jun 06, 2023 at 02:57:55PM +0530, Chandan Babu R wrote:
>> This commit introduces a new function set_log_cur() allowing xfs_db to read
>> from an external log device. This is required by a future commit which will
>> add the ability to dump metadata from external log devices.
>> 
>> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com>
>> ---
>>  db/io.c | 57 ++++++++++++++++++++++++++++++++++++++++++++-------------
>>  1 file changed, 44 insertions(+), 13 deletions(-)
>> 
>> diff --git a/db/io.c b/db/io.c
>> index 3d257236..a6feaba3 100644
>> --- a/db/io.c
>> +++ b/db/io.c
>> @@ -508,18 +508,19 @@ write_cur(void)
>>  
>>  }
>>  
>> -void
>> -set_cur(
>> -	const typ_t	*type,
>> -	xfs_daddr_t	blknum,
>> -	int		len,
>> -	int		ring_flag,
>> -	bbmap_t		*bbmap)
>> +static void
>> +__set_cur(
>> +	struct xfs_buftarg	*btargp,
>> +	const typ_t		*type,
>> +	xfs_daddr_t		 blknum,
>> +	int			 len,
>> +	int			 ring_flag,
>> +	bbmap_t			*bbmap)
>>  {
>> -	struct xfs_buf	*bp;
>> -	xfs_ino_t	dirino;
>> -	xfs_ino_t	ino;
>> -	uint16_t	mode;
>> +	struct xfs_buf		*bp;
>> +	xfs_ino_t		dirino;
>> +	xfs_ino_t		ino;
>> +	uint16_t		mode;
>>  	const struct xfs_buf_ops *ops = type ? type->bops : NULL;
>>  	int		error;
>>  
>> @@ -548,11 +549,11 @@ set_cur(
>>  		if (!iocur_top->bbmap)
>>  			return;
>>  		memcpy(iocur_top->bbmap, bbmap, sizeof(struct bbmap));
>> -		error = -libxfs_buf_read_map(mp->m_ddev_targp, bbmap->b,
>> +		error = -libxfs_buf_read_map(btargp, bbmap->b,
>>  				bbmap->nmaps, LIBXFS_READBUF_SALVAGE, &bp,
>>  				ops);
>>  	} else {
>> -		error = -libxfs_buf_read(mp->m_ddev_targp, blknum, len,
>> +		error = -libxfs_buf_read(btargp, blknum, len,
>>  				LIBXFS_READBUF_SALVAGE, &bp, ops);
>>  		iocur_top->bbmap = NULL;
>>  	}
>> @@ -589,6 +590,36 @@ set_cur(
>>  		ring_add();
>>  }
>>  
>> +void
>> +set_cur(
>> +	const typ_t	*type,
>> +	xfs_daddr_t	blknum,
>> +	int		len,
>> +	int		ring_flag,
>> +	bbmap_t		*bbmap)
>> +{
>> +	__set_cur(mp->m_ddev_targp, type, blknum, len, ring_flag, bbmap);
>> +}
>> +
>> +void
>> +set_log_cur(
>> +	const typ_t	*type,
>> +	xfs_daddr_t	blknum,
>> +	int		len,
>> +	int		ring_flag,
>> +	bbmap_t		*bbmap)
>> +{
>> +	struct xfs_buftarg	*btargp = mp->m_ddev_targp;
>> +
>> +	if (mp->m_logdev_targp->bt_bdev != mp->m_ddev_targp->bt_bdev) {
>> +		ASSERT(mp->m_sb.sb_logstart == 0);
>> +		btargp = mp->m_logdev_targp;
>> +	}
>
> I think this should print an error message if there isn't an external
> log device and then leave iocur_top unchanged:
>
> 	if (mp->m_logdev_targp->bt_bdev == mp->m_ddev_targp->bt_bdev) {
> 		fprintf(stderr, "no external log specified\n");
> 		exitcode = 1;
> 		return;
> 	}
>
> 	__set_cur(mp->m_logdev_targp, type, blknum, len, ring_flag, bbmap);
>
> because metadump shouldn't crash if there's an external log device but
> sb_logstart is zero.  The superblock fields /could/ be corrupt, or other
> weird shenanigans are going on.
>
> (Also it's a layering violation for the io cursors to know anything at
> all about the filesystem stored above them.)
>

I agree with your review comment. I will make the required changes.

-- 
chandan