Package: mdadm Version: 1.9.0-4 Severity: important This was with the 20050322 Binary-1 minimal install CD, so the version might now be right. This might also be considered a Debian-Installer problem, or the RAID udeb...
When using alternative partitioning schemes (Sun disklabel here), the userspace RAID partition detection incorrectly identifies existing partitions. Specifically, the userspace detection can misdetect the existance of a RAID setup on the whole-disk slices if the last partition on disk has RAID labels. output of fdisk -l: Disk /dev/sd (Sun disk label): 64 heads, 32 sectors, 2060 cylinders Units = cylinders of 2048 * 512 bytes Device Flag Start End Blocks Id System /dev/sd1 0 183 187392 fd Linux raid autodetect /dev/sd2 1159 1403 249856 fd Linux raid autodetect /dev/sd3 0 2060 2109440 5 Whole disk /dev/sd4 183 915 749568 fd Linux raid autodetect /dev/sd5 1403 1525 124928 fd Linux raid autodetect /dev/sd6 915 1159 249856 fd Linux raid autodetect /dev/sd7 1525 2060 547840 fd Linux raid autodetect Would result in detection of md0 on sd?1, md1 on sd?2, md3 on sd?4, md4 on sd?5, and md5 on sd?6. md2 also gets detected on sd?3, while it fails to detect the set on sd?7. If the RAID levels for md0 and md6 disagree, this can result in massive data loss on one or both filesystems. Specifically running e2fsck on md0 prior to telling the kernel md2 (RAID-5) is incorrect will destroy md0 (RAID-1). BSD disklabels and SGI disklabels can almost certainly result in similar problems. The correct thing would be to check the partition type, but looking for overlapping partitions would also work. I do not feel I can justify a higher severity due to the relative rarity of the partition type, but this is a data destruction bug (backups save the day). -- (\___(\___(\______ --=> 8-) EHM <=-- ______/)___/)___/) \BS ( | [EMAIL PROTECTED] PGP 8881EF59 | ) / \_CS\ | _____ -O #include <stddisclaimer.h> O- _____ | / _/ \___\_|_/82 04 A1 3C C7 B1 37 2A*E3 6E 84 DA 97 4C 40 E6\_|_/___/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]