usb/da vs sata geometry calculations (was Re: Switchover to CAM ATA?)

Andrew Reilly areilly at
Sun Apr 25 08:01:28 UTC 2010

Hi all,

Sorry to interrupt this thread with an off-topic question, but
it seems vaguely related, and you folk seem to be the right ones
to ask:

I've recently done a drive upgrade in a 1U rack machine that
only had space for the two active drives that were in it, and I
couldn't afford the down-time that it would take to install from
scratch.  So I formatted and populated the first replacement
drive in an external USB cradle, and when it was looking like
a good replacement for the (gmirror'd) image that was running,
I did the physical swap, and all was good, as expected.  All
except that that the identical drive that I inserted as
the second element of the mirror would *not* accept a copy
of the first disk's MBR block (with fdisk).  It said that the
calculated geometry was incompatible.  Luckily for me (I think)
the calculated geometry was a megabyte or so *larger* than the
first drive, so I was still able to bsdlabel it to match, and
slot it into the gmirror as planned.

Was this the result of the umass/da driver having a different
synthetic geometry calculation routine than the SATA driver?

This was all on an 8-STABLE system about 400 days old, fwiw.

Should I expect any on-going badness as a result of this
difference in "geometry" between two identical drives?



More information about the freebsd-current mailing list