by I have tried multiple times to get OSS449F to load under OSE 5.0.0. This
weekend I attempted it again and I also tried to load OSS468A. While
OSS448E works fine, the following are the results when trying to load the
above patches:
OSS449F:
--------
a) Kernel Link Failed: Unresolved Symbols in the "str" driver:
mblk_pool_size, mblk_pool_release, mblk_pool_mem
b) I noticed that these symbols were in the Space.c file (upper case
'S') in the /opt/K/ tree and not in the /etc/conf/pack.d/space.c file.
So I tried copying /opt/K/SCO/Unix/5.0.0Cl/etc/conf/pack.d/str/Space.c
to /etc/conf/pack.d/str/space.c.
c) With the copy in (b) performed, the kernel linked successfully.
d) After a reboot, just about every networking service generated
"Protocol not available" errors to syslog. Examples:
portmap[390]: svctcp_create - cannot getsockname or listen: Protocol
not available
named[419]: listen(vs, 5): Protocol not available
sendmail[492]: NOQUEUE: SYSERR(root): opendaemonsocket: cannot listen:
Protocol not available
Strangely enough, ftp/telnet/ping to other systems were OK, no inbound
local service was operating.
e) I backed out these changes and went back to OSS449E and all is
working again now.
f) Along the way (backing it out), there was an error replacing the
OSS449F /shlib/protlib_s with the OSS449E version- leaving the
OSS449F version in there although ALL portions of OSS449F showed
removed. I believe this error generated some Kernel panics the
first time networking started up in multiuser mode. The panic
appeared to go away once the OSS449E version was restored again.
g) A note concerning OSS449E: I reported to SCO (a while back) that I was
experiencing a hang in RLOGIN just about every time I hit ^C (INTR).
While I never heard back from SCO that there was a fix out there...
The OSS449F release notes do indicate a potential fix to this problem
and to others I am experiencing. That's why I'm trying to get
OSS449F up, rather than staying at OSS449E.
OSS468A:
--------
a) The UNIX portion of this patch "Had nothing to load", perhaps since no
SMP?
b) The TCP portion of this patch generated a kernel link error on the
Symbol: ip_spinreport.
c) The NFS portion of this patch loaded successfully.
d) Given the errors in this patch and with OSS449F, I did back the entire
patch out.
Custom Errors:
--------------
Note that it appears that custom (scoadmin software) often does not clean
up the installed patches when it gets kernel link errors. I used the
CAncel option as well as the IGnore option when prompted. Most times I
had to manually unroll all the individual pieces of the patches after
errors.
Questions:
----------
- Has anyone been able to load these patches successfully under 5.0.0?
- Does anyone know what is the cause of these problems?
- Is there another fix to the "Hit INTR and rlogin hangs" problem
with OSS449E? (ie. find a different version of rlogin?)
System Details:
---------------
The list of installed patches to the SCO TCP/IP subsystem as seen in Patch
Rollback are:
Level Patch
*5 OSS449E:TCP Network Performance SLS
4 TCP Component of SCO Internet Supplement for OpenServer 5.0
3 TCP/IP component of Internet Supplement
2 TCP/IP Version 2.0.0: Networking Supplement Release 1.0
1 TCP/IP Version 2.0.0 Release Supplement
System Specifics: SCO Release 3.2v5.0.0 (KernelID 95/08/08):
sco5.0.0, devsys, merge 3.2.1f, rs 5.0.0d, net 100, ahs 5.1, oss408a,
oss409a, oss415a, oss416a, oss422a, ahs 5.2, oss430a, oss434a, oss436a,
rs25c, rs39a, oss414a, oss437a, oss438a, oss441a, oss424a, tuneup 6.0.0g,
oss449a, oss443a, oss448a, afps 3.5.2n, arcserve 2.2.0f, netscape 3.0.1c,
smc sme0 v1.24 lli 3.4, eqnx 3.30, oss452a, oss456b, oss465a, arcserve
2.2.0 maint sup 1.0.0, intel pro/100b v5.0.4e, alad 325 d2.11, rs40b,
oss465b, oss470a, oss473a, app477a, oss449e
Thanks in advance!
-Rob
--
Robert Weiner / Programming Plus Hardware & Software Consulting