The code for reading DASD partition tables is so generic
that it will pretend to find a partition even on a totally
zeroed disk. Prevent that by recognizing dasd partition tables
only on DASD disks. Such a check was already present for DM
mappings on DASD, but (strangely) not for DASD itself.
Without this, kpartx will (try to) create a partition mapping
on a loop device with no partition table.
Found this during testing because test-kpartx unexpectedly
succeeded without my "fix part deletion without partition table" fix
when run on a loop device. That was caused by this bug pretending an
existing partition table although there was none.
Signed-off-by: Martin Wilck <mwilck@xxxxxxxx>
---
kpartx/dasd.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/kpartx/dasd.c b/kpartx/dasd.c
index f50c1bdb..e418d5ac 100644
--- a/kpartx/dasd.c
+++ b/kpartx/dasd.c
@@ -133,6 +133,9 @@ read_dasd_pt(int fd, struct slice all, struct slice *sp, int ns)
/* Couldn't open the device */
return -1;
}
+ } else if ((unsigned int)major(sbuf.st_rdev) != 94) {
+ /* Not a DASD */
+ return -1;
} else {
fd_dasd = fd;
}
--
2.14.0
--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel
