fix bug pointed out by Stevens in his vol II book

limit process to 327.67 rather than 227 seconds timeout;
return correct timeout from getsockopt (from rstevens@noao.edu)

return types, light clean.

avoid hang when given a negative write count (from polk@bsdi.com)

For a setsockopt at the socket level, if the protocol has no ctloutput
function, an mbuf is lost. From: Mike Karels <karels@BSDI.COM>

4.4BSD snapshot (revision 8.1); add 1993 to copyright

deliver data to non-atomic protocols in regular-sized chunks even on machines
without MAPPED_MBUFS, or TCP will send 968-byte packets

If an internal-use protocol entry doesn't have a usrreq routine
bail out very early before creating a socket. (from thomas@lkg.dec.com)

stop if we reach out of band mark while peeking; otherwise
we will go into an infinite loop copying 0 bytes each time. (From
Tom Truscott <trt@cs.duke.edu>)

fix bugs in sending messages through sockets (4.4BSD-alpha/sys/6)

make kernel includes standard

multicast bugfix from arc@sgi.com

need to include systm.h

use proc pointer from uio rather than (sometimes unset curproc)

convert to use new select interface, selrecord/selwakeup

lint

upcall facilities for sorwakeup, non-blocking sblock for NFS (from Macklem)

last line of while not reached; shut up mips compiler (from Ralph Campbell);
make sosetopt inform protocols of SOL_SOCKET options being changed.

notify lower levels of some socket setopt changes, esp. buffering

possibly change conventions about where M_EOR lies (never tested)

need proc parameter; for now use curproc. u. is gone

minor fix to allow sendmsg with purely control data while confirming

Try not to initiate copyouts that are going to block
for MSG_WAITALL or watermark conditions, unless there are exceptional conditionsa
(by sklower for history purposes before changes sockbuf storage

Try not to initiate copyouts that are going to block
for MSG_WAITALL or watermark conditions, unless there are exceptional conditionsa
(by sklower for history purposes before changes sockbuf storage conventions)/

show more ...

(checked in by karels) treat MT_HEADER as MT_DATA in soreceive;
need to init len when MAPPED_MBUFS not defined

lint, rm unneeded assignment (from torek)