Re: [PATCH] show an error instead of segfaulting on grub-probe -t partmap on a unsynced raid

[Felix Zielcke]

I have told this all Robert just on IRC, but it would be really nice if
others would comment on this problem too :)
I just don't know if GRUB_UTIL should ignore the missing devices just
like real GRUB or if it's a good idea to at least warn the user that
sth. is wrong with the mdraid.

Am Mittwoch, den 30.07.2008, 12:37 +0200 schrieb Robert Millan:


> We generate that list ourselves (I think in util/raid.c), so if one device
> is linked in the list but just points to NULL that's our own fault, at least
> up to the util/raid.c layer.
> 

The list is generated by grub_raid_memberlist at disk/raid.c:72
and that doestn't handle missing devices, it adds every disk to the list

  for (i = 0; i < array->total_devs; i++)
    {
      tmp = grub_malloc (sizeof (*tmp));
      tmp->disk = array->device[i];
      tmp->next = list;
      list = tmp;
    }


grub_raid_read for raid1 can handle missing devices without problems
disk/raid.c:209

        for (i = 0; i < array->total_devs; i++)
          {
            if (array->device[i])


This is from grub_raid_scan_device after hd0,1 is added
sdb1 / hd1,1 has been removed and --zero-superblock'ed

(gdb) print *array
 $3 = {number = 0, version = 0, level = 1, layout = 0, total_devs = 1, nr_devs 
= 1, chunk_size = 0, uuid = {1191940856, 3853190239, 
                     1276382316, 2567164211}, 
   name = 0x19e1230 "md0", disk_size = 16777216, device = {0x0, 0x19e1080, 0x0 
<repeats 30 times>}, next = 0x0}