I haven't tried SystemRescueCd yet, but from the init script in the initrd, it appears that the initrd tries to find the sysrcd.dat squashfs image by mounting every block device in the system in a brute-force fashion. While it uses the -r mount option to avoid modifying data on disk, journaled filesystems like ext3/ext4 will still replay the journal to disk even when mounting read-only, so the -r option is insufficient to prevent disk writes. To make matters worse, as far as I can tell, no attempt is made to exclude component drives of RAID arrays from this probing process, so it seems likely that probing might cause data loss by replaying journals directly to individual mirrors instead of through the md device.
Am I missing something here? If not, it seems like it would be a good idea to do probing through the loopback device, using "losetup -r" to ensure no writes take place to the probed block devices until the correct one is found. -Nathan
All other questions
1 post • Page 1 of 1
Who is online
Users browsing this forum: Google [Bot] and 2 guests