Text table slots getting lost (4.3BSD)

per at erix.UUCP per at erix.UUCP
Fri Feb 27 04:27:34 AEST 1987


A while ago, we started getting "text: table is full" messages on our
system (4.3BSD on VAX 11/750), which was a bit surprising, since we
had never hit that limit before. And sure enough, pstat reported >10
entries available (when it didn't fail to execute because the text
table was full...). Close examination (w pstat -x modified to always
print the entire table) showed that a number of the entries consistently
had *all* of IPTR, FORW, and BACK =0, i.e. not used and not in the free list!

Rebooting of course removed the symptom, but unfortunately not the cause.
Three days ago, with an uptime of 11 days, I had another look at the text
table, and found 8 (out of 60) entries lost in this way. Today there
is one more missing. What gives? Is this a known bug, with a fix that
I've somehow missed, or what could possibly cause it to happen? We have
a few local kernel mods, but none of them seem relevant.

Per Hedeland

UUCP: per at erix.se  or  per at erix.uucp  or  ...{mcvax,seismo}!enea!erix!per
ARPA: per at erix.se  or  per%erix.se at seismo.css.gov



More information about the Comp.unix.wizards mailing list