Kdavid@gizzmo.UUCP (David Solan) (05/08/88)
I'd like to add my thoughts on the mkfs command that a few
people have been commenting on recently.
Mkfs is a forgiving command in terms of its error output -- but
it cannot do the impossible! If you execute it with numeric
parameters which exceed the number of available blocks on the
diskette (or hard disk) in ANY way, or ask it to create any sort of
contradiction on the file system, and then use that floppy as a
mountable file system -- you will be sorrrrrrryyy! It will appear
to work, but in fact you will be losing and rewriting data all over
the place, not to mention getting I/O errors. A complete mess will
ensue. Of course this mess will NOT ensue if all you do is use your
floppies for cpio -- cpio overwrites the inode table of contents
that mkfs creates.
If you just use mkfs by itself it KNOWS the maximums and correct
values and plugs them in for you automatically. If it's any help,
the "physical" "block number" in mkfs is in 512 byte "blocks", which
hold 8 inodes, so adding 16 inodes means losing 2 "physical" data
blocks and vice versa (you can only go up or down 2 "blocks" or 16
inodes at a time). However, keep in mind that the second argument
to mkfs ("blocks[:inodes]") is in a form where "blocks" means the
TOTAL blocks of the system, not just those used for data in the
filesystem (as opposed to those used for inodes). Thus, this number
should STAY THE SAME no matter how many inodes you are creating.
You will see the difference in a df -t command, but for mkfs, the
number is always the same.
Furthermore, the "gap" and "blockock = 1024
total logical blocks = 415
total inodes = 144
gap (physical blocks) = 2
cylinder size (physical blocks) = 20
In the first case above, a df -t command (on the mounted file system)
results in:
/mnt (/dev/fp021): 812 blocks 94 i-nodes
total: 830 blocks 96 i-nodes
And, in the second case, a df -t command gives:
/mnt (/dev/fp021): 806 blocks 142 i-nodes
total: 830 blocks 144 i-nodes
Keep It Short and Simple.
The 3.51 Utilities package has a detailed description of the mkfs
command in the UNIX System V User's Manual, Volume 1 -- look it
up!!! ("when all else fails, read the directions").
David Solan
Objective Programming Incorporated
Post Office Box 123
Norwalk, CT 06856
Voice: (203) 866-6900
attmail: dsolan
USENET: gizzmo!kdavid
--
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
{codas,u1100a}-----\
David Solan rutgers!rochester!pcid!kodak!gizzmo!mark
{lazlo,ethos,fthood}-----/Kdavid@gizzmo.UUCP (David Solan) (05/08/88)
Oooopps. Somehow the end of my "random thoughts on mkfs" got
censored by USENET. Here they are again:
Here is an example of what I mean. After you have initialized a
floppy diskette to be 42 cylinders(tracks) and 10 physical sectors
per track, doing an mkfs by itself on this floppy results in the
following:
mkfs /dev/fp021
---- ----------
slice 1
bytes per logical block = 1024
total logical blocks = 415
total inodes = 96
gap (physical blocks) = 2
cylinder size (physical blocks) = 20
The "415" above means 830 "physical blocks", as far as mkfs is
concerned. So, to get more inodes, do the following:
mkfs /dev/fp021 830:144 2 20
---- ---------- ------- - --
bytes per logical block = 1024
total logical blocks = 415
total inodes = 144
gap (physical blocks) = 2
cylinder size (physical blocks) = 20
In the first case above, a df -t command (on the mounted file system)
results in:
/mnt (/dev/fp021): 812 blocks 94 i-nodes
total: 830 blocks 96 i-nodes
And, in the second case, a df -t command gives:
/mnt (/dev/fp021): 806 blocks 142 i-nodes
total: 830 blocks 144 i-nodes
Keep It Short and Simple.
The 3.51 Utilities package has a detailed description of the mkfs
command in the UNIX System V User's Manual, Volume 1 -- look it
up!!! ("when all else fails, read the directions").
David Solan
Objective Programming Incorporated
Post Office Box 123
Norwalk, CT 06856
Voice: (203) 866-6900
attmail: dsolan
USENET: gizzmo!kdavid
--
::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
{codas,u1100a}-----\
David Solan rutgers!rochester!pcid!kodak!gizzmo!mark
{lazlo,ethos,fthood}-----/