There is a serious bug somewhere in the ffs code.  I have a filesystem
that fsck passes without complaint, but when used, panicks the kernel
with "ffs_alloccgblk: can't find blk in cyl".  It's almost certainly a
very long-standing bug; SunOS 4.x suffers from it too.

It's probably possible to construct a truly tiny example of this, but I
don't understand it well enough to do so.  However, I do have a 9728000
byte (a little under nine and a half megs) image of a filesystem that
provokes it, and it contains nothing private, only copies of fragments
of the 1.2 kernel source tree plus a few tiny files I added to fill it
up during part of the process.  I will be glad to give a copy of this
filesystem to anyone who wants.  After gzip --best, it's 1956875 bytes;
a second gzip --best brings it down to 1947156 bytes.  (It's a
big-endian filesystem.)  I'll even drop a copy on homeworld if desired,
but I hesitate to dump this big a file there without clear approval
first.

Should I send-pr this?  I'm hesitant to put a two-meg gzipped
filesystem image into a PR, but I'm also hesitant to file a PR without
providing the example.

					der Mouse

			       mouse@rodents.montreal.qc.ca
		     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B