Patch-ID# 100120-04
Keywords: 4.1 TLI Patch 
Synopsis: SunOS 4.1: TLI Jumbo patch.
Date: 12/June/91

SunOS RELEASE: 4.1

Topic: TLI patches

BugID's fixed with this patch: 1032950 1035527 1035529 1037715 1039287 1049125 1057481 1059282

Architectures for which this patch is available: sun3, sun3x, sun4, sun4c

Obsolete By: 5.0

Problem Description:

BUGID:1032950
	copy of a 10M file over RFS hangs the RFS server.
	copy of a  REMOTE_RESOURCE_FILE  to  local_disk_filename
	causes the machine which owns the remote resource to
	HANG such that a <stop> cannot halt the CPU but
	copy of a local_disk_filename to REMOTE_RESOURCE_FILE 
	works correctly.


BUGID:1035527
	When a user specifies a zero length address, tcptli
	returns the default address but fails to return the
	length of the default address (which is sockaddr_in)
	to the user.


BUGID:1035529
	When a close is issued, tcptli loops thru its event queue
	and attempts to remove any outstanding event on the
	pending queue. A bug in the logic that it fails to 
	remove if that connection instance is the first on
	the queue.

	Also, if a timeout is issued for that device instance,
	the close routine fails to do an untimeout. Later, when
	the aux proc wakes up, it removes the connection instance
	from the queue and it had already been closed,
	results in panic.

BUGID:1037715
	TCPTLI mistakely indicated ISO orderly release is supported,
	but actually, it is not. The fix now prompts the error
	of feature not supported.

BUGID:1039287
	When more than 4052 bytes are sent on the TCP TLI transport, 
	the transport fails to correctly send the data, and the
	connection may hang.

BUGID:1049125
 	When a server attempts to reject an incoming connection
	request via the t_snddis(3) TLI call, the system panic
	with the error msg: "panic: soclose: NOFDREF".


BUGID:1057481
	A bug in "tcp_tli.c" 'tcptli_senddata()' routine.  When the
	data upstream comes down faster than the underlying networking
	code can handle, it has to recover from the partial send
	situation.  Without this fix TLI will cause the system to
	hang.  The bug fix corrects the code which puts data on the
	read-queue, instead of the write-queue.

BUGID:1059282
	The customer is developing TLI programs.  The TLI client program
	connects to the customer's TLI server program using TCP/IP.
	Following are the problems:

	1)  Occasionally, an abortive termination (CTL-C) on the client
	    causes the system to re-boot.

        2)  If the cusomter's TLI server program issues an exit statement
	    after sending out all messages, the client reboots with the
	    following message when attempting to issue a 't_snd()':

	      panic: assertion failed: q->q_next, file ../../os/str_io_buf.c,
			line: 1707

INSTALL:
	File param.c should be placed in your /usr/kvm/sys/conf.common
	directory.
As root:
# cp /usr/kvm/sys/conf.common/param.c /usr/kvm/sys/conf.common/param.c.FCS
# cp /sys/sun{3,3x,4,4c}/OBJ/tcp_tli.o /sys/sun{3,3x,4,4c}/OBJ/tcp_tli.o.FCS
# cp /sys/sun{3,3x,4,4c}/OBJ/tcp_tliaux.o /sys/sun{3,3x,4,4c}/OBJ/tcp_tliaux.o.FCS
# cp /sys/sun{3,3x,4,4c}/OBJ/tcp_tlisubr.o /sys/sun{3,3x,4,4c}/OBJ/tcp_tlisubr.o.FCS
# cp /sys/sun{3,3x,4,4c}/OBJ/ti_mod.o /sys/sun{3,3x,4,4c}/OBJ/ti_mod.o.FCS
# cp /sys/sun{3,3x,4,4c}/OBJ/ti_rdwr.o /sys/sun{3,3x,4,4c}/OBJ/ti_rdwr.o.FCS

# cp param.c /usr/kvm/sys/conf.common/param.c
# cp sun{3,3x,4,4c}/tcp_tli.o /sys/sun{3,3x,4,4c}/OBJ/tcp_tli.o
# cp sun{3,3x,4,4c}/tcp_tliaux.o /sys/sun{3,3x,4,4c}/OBJ/tcp_tliaux.o
# cp sun{3,3x,4,4c}/tcp_tlisubr.o /sys/sun{3,3x,4,4c}/OBJ/tcp_tlisubr.o
# cp sun{3,3x,4,4c}/ti_mod.o /sys/sun{3,3x,4,4c}/OBJ/ti_mod.o
# cp sun{3,3x,4,4c}/ti_rdwr.o /sys/sun{3,3x,4,4c}/OBJ/ti_rdwr.o

rerun /etc/config <kernel-name>
and make and install the new kernel.
Please refer to the system and networking administration manual
for details on building and installing a new kernel