SunSolve Internal

Infodoc ID		Synopsis		Date
12571		kernel config troubleshooting and common error messages		26 Jan 1996

Description

Top

Troubleshooting Kernel Config Problems
		======================================

Disclaimer: These notes are considered "unofficial" 

This provides some tips for resolving a wide variety of kernel config problems.
Refer to the SunOS 4.1.x System and Network Administration manual, part number
800-3805-10  27-March 1990,  Chapter 9, esp. p.243 and Ch.22, esp. p.700-704. 

You should be able to build  GENERIC  (unless ex or make is damaged!)  by:
# cd /sys/`arch -k`/conf;config GENERIC;cd ../GENERIC;make ; mv vmunix /;reboot
If not, the most basic "fix" is to reload /sys, using add_services,  or by:
# cd /usr/kvm ; /usr/etc/install/extract_files st0 Sys   (or sr0 for CD-ROM)

vfork failed  OR  process table full  is caused by hitting current "maxusers".
pstat -T | grep process   will show you  <current_useage>/<(maxusers*16+10)>
Edit   maxusers   in /sys/`arch -k`/conf/YOUR_SYS_NAME    (GENERIC has an 8)
where `arch -k` inserts your kernel sub-architecture: sun4/sun4c/sun4m/... and
YOUR_SYSTEM_NAME refers to a particular config file [or] build directory.

If System V IPC (sem,shm,msg) syscalls fail, it may be because you are running
GENERIC_SMALL.  ipcs  will say "...facility not in system."   A quick-fix is:
# cp  /usr/kvm/stand/vmunix  /    ;    reboot
See also Sys&Net Adm man. p.702-703 and {/usr/include,/sys}/sys/{sem,shm,msg}.h

To find undefined symbols (caused by 3rd party .o's or patch conflicts) try:
nm -o /sys/`arch -k`/OBJ/*.o | grep the_missing_symbol
nm -o /sys/`arch -k`/YOUR_SYSTEM_NAME/*.o | grep the_missing_symbol
Also, check for NON-original 4.1.1 & 4.1.2 object files (identify patches) by:
ls -l /sys/`arch -k`/OBJ | grep -v "Oct 1[12]  1990" | grep -v "Oct 23  1991"

ps fails with "...read error on ktextseg..." IF you booted OTHER THAN /vmunix
Certain error messages are reported in /sys/`arch -k`/YOUR_SYS_NAME/makedeperrs
The following EXTREME step will usually uncover the most obscure make problems:
config -n YOUR_SYSTEM_NAME; cd ../YOUR_SYSTEM_NAME; make -dd -DD
FAX us ONLY the last 50 lines (1 page). 508-794-5350  Include Service Call#.
Miscellaneous:  You can verify a message is from the GENERIC SunOS kernel by:
strings /usr/kvm/stand/vmunix | grep "whatever message"
For heterogeneous clients, sys is in /export/exec/kvm/client_arch-k (on
server).
Note various things shown/controlled by the   csh   limit   command.

		    Other common (unrelated) errors:

"out of memory" means not enough swap space (swap "IS" virtual memory!) Check:
pstat -T | tail -1
IMPORTANT corrections to adding swap, Sys&Net Adm manual p.247:
Do NOT use  -n  with mkfile!! (-v is OK). Secondary swap is entered in fstab
as:
/whatever  swap  swap  rw  0  0     (that page has several typo's on this!)

vmunix: sendsig: bad signal stack  is strictly an application program "bug"
(a user-written program, itself, corrupted its stack)

silo overflow  simply indicates too many characters came into some serial port.
zs0/1 is ttya/b;  zs2 is the keyboard;  zs3 is the mouse.

out of ptys (from rlogin): See OSS SRDB#2988 [modem#: 800-477-4768
online/info4u]

if_snd/mbuf map full  are ethernet problems.  Check   netstat -i   for a max.
of a few I/O errors per 10,000 I/O pkts; collisions  <  a few % of Opkts.
x{m,d}balloc: out of {m,d}blks are streams overload.  Check   netstat -m

For panic: ifree/alloc/iinactive/..., check dkinfo for overlapping partitions.
These also might be from hardware failure, but they do not indicate an OS bug.

SOLUTION SUMMARY: