On Mon, Jun 19, 2023 at 10:35:32AM +0800, zhujun2 wrote:
> file system, the disk by-partlabel probability loss problem
>
> The superblock of ceph_bluestore exists in the first 1024 bytes, and the
> real data is stored after 8K (that is, 16 sectors). By checking the source
> code of blkid, it is found to check whether it is the xfs_external_log file
> system, which is to cycle through the first 256k data range (the first
> 512 sectors ). That is to say, when blkid judges whether the partition is
> an xfs_external_log file system, it will mistakenly judge the real data of
> ceph_bluestore. This means that if the user's data itself conforms to
> the xfs_external_log format and is stored in sectors 17~512, it will be mistakenly
> considered as the xfs_external_log file system.
Not sure if I understand. According to docs BlueStore uses XFS
partition to store metadata and another device to store raw data. It
sounds like the XFS should be visible as a normal filesystem.
>From your description it sounds like the both is on the same device
and we should not detect it as XFS log. If yes, then I do not
understand why anyone use XFS log header for this purpose.
XFS log detection seems robust, so I guess if libblkid detects the
superblock then it's really the FS.
Karel
> Signed-off-by: zhujun2 <zhujun2@xxxxxxxxxxxxxxxxxxxx>
> ---
> libblkid/src/superblocks/xfs.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/libblkid/src/superblocks/xfs.c b/libblkid/src/superblocks/xfs.c
> index d8c6fb6d4..3686bd52b 100644
> --- a/libblkid/src/superblocks/xfs.c
> +++ b/libblkid/src/superblocks/xfs.c
> @@ -259,6 +259,11 @@ static int probe_xfs_log(blkid_probe pr,
> if (memcmp(&buf[i*512], "XFSB", 4) == 0)
> return 1;
>
> + if (memcmp(&buf[i*512], "bluestore block device", 22) == 0) {
> + DBG(LOWPROBE, ul_debug("\t device has ceph_bluestore ambivalent"));
> + return 1;
> + }
> +
> rhead = (struct xlog_rec_header *)&buf[i*512];
>
> if (xlog_valid_rec_header(rhead)) {
> --
> 2.20.1
>
>
>
--
Karel Zak <kzak@xxxxxxxxxx>
http://karelzak.blogspot.com
