Mdadm may incorrecty fail to assemble a clean raid5 array if the array had a failed drive with a dirty superblock at the time the array was last stopped. While assembling, mdadm will use the superblock of the failed drive during assembly, if the failed drive is probed first. If the failed drive's superblock is active/dirty, mdadm may think that the entire array is dirty-degraded instead of clean-degraded. Mdadm will then not try and run the array unless a --force is specified. This patch ensures that mdadm will always use the most recent superblock before making desicions about running the array. This bug appears in all recent version of mdadm. -eivind
Attachment:
mdadm.patch
Description: Binary data
