Fragment Size with large Raid Arrays

Question

No, I am not a new user. I've done newfs probably hundreds
of times. But, I've never had to futz around with obscure
parameters because the device geometry was insane.
I want to get a file system made that
doesn't require 4096 byte fragments. Is that too much to ask?
I would think that there is some combination of cylinders,
heads, and sectors that should give that result, but I don't
have time to search for it.
tahoe# newfs -m 1 -o time /dev/rdsk/c2t8d0s2
newfs: construct a new file system /dev/rdsk/c2t8d0s2: (y/n)? y
With 16129 sectors per cylinder, minimum cylinders per group is 16
and the fragment size to be changed from 0 to 4096
With these parameters, it returns immediately, seemingly doing
nothing -f 4096 to specify the fragment size, like:
tahoe# newfs -f 4096 -m 1 -o time /dev/rdsk/c2t8d0s2
This worked, or seemed to.
The problem now is that it I am losing a lot of space when I have
directories with lots of little files.

Answer 1

The relocation you show is a relocation for a 32-bit value
relative to a linker section, in this case the .lbss section.
I would not expect such a relocation to be generated
unless the object file was produced by compiling for the
small memory model.
What type of symbol is maillage_curv.curv_rank_from_index_?
From the name, I would expect it to be a routine name, but
routine names should not be defined in the .lbss section.
Bob Corbett