Mike Walter wrote:
An IBMer from Omaha told us that one should always format with CMS or
ICKDSF before giving an MDISK to a Linux guest before the guest
formats the MDISK with "dasdfmt". No reason for the CMS or ICKDSF
format was given.
Formatting an MDISK twice really, really (pun intended) rubs me the
wrong way.
Except you are not really formatting it twice. You are formatting the
hardware disk in CMS format the first time, and then "formatting" the
area withing that space for Linux's use (dasdfmt). just laying down data
within the MDISK. It' kinda like the pre-ata hard drives that had to be
hardware formatted (with a servo track or some other) the first time,
before you could DOS format them. Think of the ICKDSF format as a
hardware format and the dasdfmt as the software format, even though both
are software generated. The CMS format is just arranging the ECKD dasd
in 4k blocks with the appropriate pseudo-IRGs betweeen them. Linux is
then putting down it's formatting on the disk so that ext2/3 can find
the eyecatchers it's looking fo on the disk. Remember, the file system
doesn't care what is on the disk, but is expecting information that
dasdfmt has left behind, perhaps in control blocks in the first block or
two of the "partition". That partition has to look the same on my laptop
as on a SCSI disk, as on a USB keyfob drive, as on a 3390...
But we have experienced inconsistent problems with MDISKs given to
Linux guests after formatting with only "dasdfmt" (i.e. not with
ICKDSF or CMS FORMAT). Sometimes the MDISK worked with only dasdfmt,
other times Linux would not recognize it and we had to format it with
CMS, then dasdfmt. When at a Linux workshop in Poughkeepsie, the IBM
instructor related that Linux mdisks should always be formatted by CMS
or ICKDSF before giving them to the Linux guest to beformatted with
"dasdfmt" - having experienced random failures himself.
I suspect that the problem is related to the MDISK needing a volser or
dummy VTOC, but have never had time to research it. My guess is that
skipping the CMS format works when the newly defined MDISK happens to
align with a previously used MDISK beginning at that same cylinder
extent.
Does anyone have a verifiable explanation for the requirement to
format the MDISK before giving it to Linux for a dasdfmt? It may
explain Loren's problem, too.
See above. I hope I've cleared it up and not confused it for you.
Mike Walter
Hewitt Associates
The opinions expressed herein are mine alone, not my employer's.
Kim Goldenberg
State of NJ - OIT
The opinions expressed herein are mine alone, not my employer's.
