I wrote a little Bourne shell script to automatically configure
and deconfigure binfmt_misc. One of the things it does before
deconfiguring it is to make sure that /proc/sys/fs/binfmt_misc/status is
writable, and then echo's -1 to it and rmmod's the module. It works fine.
But.. if I run it again (after it's already off and the module is
unloaded), it causes a kernel Oops. It seems the pathnames are lasting in
/proc as ghosts, even though they should be gone. If I turn off
binfmt_misc and then rmmod the module, an ls /proc/sys/fs shows empty
(like you'd expect). But if you ls /proc/sys/fs/binfmt_misc/status, it is
there! And heaven forbid if you try to read it; it locks the machine up
solid. Perhaps this is a dcache problem?
Even with these bugs, I am very enthusiastic about this new binfmt_misc
technology.
thanks,
-bp
-- B. James Phillippe <bryan@Terran.ORG> UNIX, Linux, networks, programming, etc. NIC:BJP4 # http://w3.terran.org/~bryan