large file system with RAIDframe

[Steven Bellovin <smb%cs.columbia.edu@localhost>]

I'm trying to create a large file system on an amd64/5.x RAID5 system.  The 
RAID device in question is not the root partition.  I can't get the disklabel 
to say anything sane.

I have three 2TB disks in a RAID5 configuration.  This should, as I understand 
things, give me 4TB of usable space.

Each disk individually has a sane label:

5 partitions:
#        size    offset     fstype [fsize bsize cpg/sgs]
 c: 3907029105        63     unused      0     0        # (Cyl.      0*- 
3876020)
 d: 3907029168         0     unused      0     0        # (Cyl.      0 - 
3876020)
 e: 3907029105        63       RAID                     # (Cyl.      0*- 
3876020)


Per raidctl(8), I used -C, -c, -I, and -i to initialize things.  The problem 
occurs when I try to label the raid device.  The initial default label shows

#        size    offset     fstype [fsize bsize cpg/sgs]
 a: 3519090752         0     4.2BSD      0     0     0  # (Cyl.      0 - 
4582149*)
 d: 3519090752         0     unused      0     0        # (Cyl.      0 - 
4582149*)

If I try to create an 'e' partition that's larger, it gets truncated to 
2147483648.  Even the kernel knows that that's not right; I see

raid1: WARNING: raid1: total sector size in disklabel (3519090752) != the size 
of raid (7814058048)

in dmesg.  I found a message from last year 
(http://mail-index.netbsd.org/current-users/2009/02/22/msg008222.html) about 
similar problems; however, the solution involved wedges on a single, large 
drive.  It is not clear to me how that translates to RAIDframe.

Suggestions welcome.

                --Steve Bellovin, http://www.cs.columbia.edu/~smb