lots of cleanup

idecref cwd in exit

forbid create or write of existing directory
mkdir("d1/d2/d3"), .. should refer to d2, not cwd
mkdir increase parent link count

fix race in mknod / creat
use last component in dirent in mknod, not path

clean up test program

nit in sbrk
indirect block
fix dup to share fd struct

cd in shell
nits in mkdir, ls, etc.

mkdir
ls shows stat info for each dir entry

don't hold fd table lock across idecref() (latter does block i/o)
idecref calls iput() in case last ref -> freeing inode
dir size is 512 * # blocks, so readi/writei &c work
unlink deletes dirent even if ip->nlink > 0

multi-block directories
track size of directory (size = number entries in use)
should namei (and other code that scans through directories) scan through all blocks of a directory and not use size?

don't increase length of directory

read problem, but untested

usertests for concurrent create/delete, and read() after unlink()

primitive ls

remove console init code from userfs

working unlink, but doesn't free dir blocks that become empty
remove out-of-date comment in ioapic

sh accepts 0-argument commands (like userfs)
reads from console

so fast interrupts overflow the kernel stack
fix: cli() before lapic_eoi()

fix acquire() to cli() *before* incrementing nlock
make T_SYSCALL a trap gate, not an interrupt gate
sadly, various crashes if you hold down a keyboard key...