Roland Mainz wrote:
> bugmail-sender at Sun.COM wrote:
> > *Synopsis*: pkill fails on native and sn1 branded zones
[snip]
> > === *Public Comments*
> > ========================================================
> > Appears to be due to ksh93 update:
> >
> > # ptree 101659
> > 100557 /usr/lib/ssh/sshd
> > 100924 /usr/lib/ssh/sshd
> > 100925 /usr/lib/ssh/sshd
> > 100945 -ksh
> > 100951 sh
> > 101659 zlogin -l root nv /usr/bin/sleep 3600
> > 101660 sh -c /usr/bin/sleep 3600
> > 101663 /usr/bin/ksh93 /usr/bin/sleep 3600
> >
> > sleep is a ksh93 script, so it's "-f" full process name is:
> >
> > /usr/bin/ksh93 /usr/bin/sleep 3600
> >
> > This happens for the global zone as well:
> >
> > $ /usr/bin/sleep 3600&
> > $ pkill sleep
> >
> > THe above does not kill the sleep, which appears as a process named
> >
> > /usr/bin/ksh93 /usr/bin/sleep 3600
> >
> > Same goes for
> > $ sleep 3600 &
> > $ pkill sleep
> >
> > /usr/bin/sleep is a ksh93 script which calls it's own sleep builtin.
> > function.
> >
> > For scripts which implement "sleep-for-syncronization" hackary, this could
> > easily break them.
> >
> > *** (#1 of 1): 2009-01-14 00:47:44 GMT+00:00 stephen.lawrence at sun.com
>
> Stephen: Does the testcase work if you change $ pkill -z 2 -f
> "^/usr/bin/sleep 128301" # to $ pkill -z 2 -f "/usr/bin/sleep 128301" #
> (e.g. remove the leading '^' character from the process name pattern) ?
BTW: The "real" fix for this problem is to compile the shell scripts -
in that case the compiled shell script /usr/bin/sleep will just appear
as "just another binary application" in the $ ps -ef # output ...
... the trouble is that the matching kernel module ("shbinexec") was
added to OS/Net in B106 and we need some builds before we're able to
request a flag day and make the availablity of the kernel module a
requirement... ;-/
How fast is a fix needed ?
----
Bye,
Roland
--
__ . . __
(o.\ \/ /.o) roland.mainz at nrubsig.org
\__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer
/O /==\ O\ TEL +49 641 3992797
(;O/ \/ \O;)
