[CFT][PATCH] procfs cleanups - round 4
Alexander Viro (viro@math.psu.edu)
Sat, 31 Jul 1999 09:01:19 -0400 (EDT)
Folks, please help to test this one. It touches a lot of places
and while I think that I've caught all stupid typos... well, I can be
wrong. Please, try it with different configs and see if something in
/proc will break. Patch (against 2.3.13-pre1) lives on
ftp://ftp.math.psu.edu/pub/viro/proc-patch-12.gz
Description:
* /proc/h8 and proc/net/bmac didn't release their entries upon
rmmod. Result: oopsen. Fixed.
* /proc/net/raw6 and /proc/net/udp6 shared the inumber. Fixed.
* rmmod ax25 released /proc/net/ax25_route twice. Fixed.
* symbolectomy: proc_net_file_operations is no more.
Rationale: it's identical to proc_file_file_operations, which
is what proc_register defaults to. So proc_net_file_operations
in the intializers had been replaced with NULL.
* new functions:
create_proc_info_entry(name, mode, base, get_info)
creates an entry with given name and mode under
the directory base and sets ->get_info() method.
mode defaults to S_IFREG|S_IRUSR. base defaults to
root of procfs.
create_proc_read_entry(name, mode, base, read_proc, data)
same as above, but sets ->read_proc() and ->data.
proc_net_create(name, mode, get_info)
same as create_proc_info_entry, uses /proc/net and base;
the most common case in net/*/*.c
proc_net_remove(name)
removes name from /proc/net
* lots of places switched from explicit calls of proc_{un}register()
to those functions.
Rationale: cleanus up a lot of mess, reduces the number of places that
have to know proc_dir_entry layout.
* proc_array_file_operations are used only for per-process stuff. Old
system-wide stuff that used to use it moved to generic
scheme.
Rationale: those entries were the last remnants of old scheme and
they gave a lot of unneeded ad-hackery. They are not different
from /proc/net/ stuff, etc. And their initialization gave
a lot of clutter in fs/proc/root.c
* fs/proc/net.c is no more. It was dead since 2.1.21, time to bury it.
* fs/proc/array.c is only for per-process stuff now. The rest had been
moved to fs/proc/proc_misc.c, together with the initialization.
* proc_root_fs, proc_mca and proc_sys_root became pointers.
Rationale: no reason to keep them static; other /proc subdirectories
are dynamic now.
* Most of the stuff in procfs doesn't have fixed inumbers now. Exceptions:
SCSI stuff, MCA stuff and PROC_NET_IP_MASQ_APP (the latter can
go right now, the rest is slightly trickier) + per-process
stuff.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/