Does anybody know when 6446146 will be fixed?
(DISCLAIMER: rant follows)
I can't believe that this has been in Solaris since Solaris 10 (and earlier, as
early as 1993), and is still lingering around.
I have two USB disks I bought yesterday, and hooked up to my SPARC
workstation. 2 x 500GB. I wanted to use 250GB for a zpool, and 250GB for
Oracle ASM.
And what happens? Well, I hit a bug which has made his rounds in various forms
since 1993, but was never properly corrected, and today is 2008.
So I go and apply the latest recommended patch cluster; reboot; and I still
can't properly either partition the disk with fdisk(1M), or format(1M) it. I
get:
[EMAIL PROTECTED]/]> fdisk /dev/rdsk/c2t0d0s2
fdisk: Cannot open device /dev/rdsk/c2t0d0s2.
OK, so fdisk(1M) is giving me the finger, but I've got a trick or two left to
play:
[EMAIL PROTECTED]/]> prtvtoc /dev/rdsk/c4t0d0s2 | fmthard -s '-'
/dev/rdsk/c2t0d0s2
fmthard: Partition 0 overlaps partition 2. Overlap is allowed
only on partition on the full disk partition).
I know that fmthard(1M) is lying or busted, because apparently, ZFS has no
problem:
[EMAIL PROTECTED]/]> zpool status
pool: space
state: ONLINE
scrub: scrub in progress, 15.56% done, 1h2m to go
config:
NAME STATE READ WRITE CKSUM
space ONLINE 0 0 0
mirror ONLINE 0 0 0
c2t0d0s0 ONLINE 0 0 0
c4t0d0s0 ONLINE 0 0 0
errors: No known data errors
So, I've hit a bug. This gets me on a hunt, and I find a whole string of
related bugs which are plaguing both fdisk(1M), format(1M), and sometimes
fmthard(1M). At the heart of the problem seems to be the fact that fdisk(1M)
uses disk geometry when doing slice and write calculations, except that when
it's time to write the disk label, the label gets mangled. These "mangles" have
been fixed in various forms and fixes over the years, but apparently none of
the fixes were a complete end-to-end solution.
Of course, it's kind of rediculous that I'm even forced to use fdisk(1M), just
because I'm sitting on two USB disks.
I'd like to know, why can't I, on sparc, just use format(1M) to write a VTOC
label like on any other disk?
Why are USB disks treated as "removable" is really beyond me; I mean, a SCSI
disk can in that sense be "removable", because nothing is stopping me from
quiescing the SCSI bus with cfgadm(1M), and disconnecting the disk(s), all the
more so, since in today's day and age, almost all the servers come with hot
swappable SCSI and SATA bays.
Why is USB different?
Even more so, please allow me the luxury of being stupid, and asking a really
stupid question:
how is it, that the most advanced operating system on the planet, still does
not handle USB properly? All the other operating systems do so without
problems, only Solaris, which, ironically, has the most competent of the
engineers, does not? Is my stupid question really and completely unfounded?
This message posted from opensolaris.org
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
