>http://bugs.grommit.com/show_bug.cgi?id=645
>
>           Summary: with BSM enabled, suspending a login shell (started with
>                    `su -`) hangs (SIGSTOP)
>           Product: ksh93-integration
>           Version: unspecified
>          Platform: i386 (64bit/AMD64)
>        OS/Version: Solaris 11/Nevada
>            Status: NEW
>          Severity: normal
>          Priority: P2
>         Component: App/ksh93
>        AssignedTo: ksh93-integration-discuss at opensolaris.org
>        ReportedBy: rlawhead at ucsd.edu
>
>
>Without BSM enabled I frequently become root as `su -`, and intermittantly
>suspend and resume the su session.  ptree shows something like:
># ptree $$
>658   /usr/lib/ssh/sshd
>  5473  /usr/lib/ssh/sshd
>    5474  /usr/lib/ssh/sshd
>      5477  -ksh
>        5506  /bin/ksh
>          5541  ptree 5506
>
>With BSM enabled, the same procedure results in a shell that must be signaled
>(CONT or TERM, usually) from a different shell, to revive the one that was
>being suspended.  For this situation, ptree shows:
># ptree $$
>568   /usr/lib/ssh/sshd
>  973   /usr/lib/ssh/sshd
>    974   /usr/lib/ssh/sshd
>      1017  -ksh
>        1326  su -
>          1327  -ksh
>            1337  ptree 1327
>
>Note that with audition enabled the `su` hasn't been replaced by ksh.
>The process tree is similar for other shells (bash, csh, tcsh were tested), but
>they refuse to suspend a login shell, so don't become hung.
>
>Is the behavior described expected to preserve the audit trail?  If so,
>could ksh be made aware that a login shell shouldn't be suspended when its
>parent isn't a shell?


No, I think this is an actual bug.  "su" must stop when the child stops.

Casper

Reply via email to