tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

re: partitionSizeHi in raidframe component label, take 2 (wasRe: partitionSizeHi in raidframe component label)

this patch causes the raid on my ultra60 to no longer autoconfigure
attach.  it took me a while to figure it out, because i also broke
accesses to sd0 in the same update, but after some testing i've
determined that without this change in -current, my u60 sees the
RAID1 at boot time fine and boots normally.

i'm still investigating exactly what is wrong, but i suspect that
this is probably it:

        - this system is one of the few that has random bytes in
        the new the new sizeHi member of the label.

        - it has never run -current since my original change to
        enable >2TiB autoconf support

        - that causes this test to now fail:

      if (!raidread_component_label(secsize, dev, vp, clabel)) {
              /* Got the label.  Does it look reasonable? */
              if (rf_reasonable_label(clabel) &&
>>>               (rf_component_label_partitionsize(clabel) <= size)) {
                      rf_fix_old_label_size(clabel, numsecs);
#ifdef DEBUG
                      printf("Component on: %s: %llu\n",

since this part of the change is actually the major bug fix of what
you're trying to fix (everything else looks like clean up?), i'm not
really sure what the right answer is now.  and since i've booted my
system with a kernel with my >2TiB changes now, it has probably fixed
my label such that it won't have a problem anymore.

however, it mostly makes my rf_fix_old_label_size() useless.

any ideas anyone?


Home | Main Index | Thread Index | Old Index