procctl: require debug privileges over the target for most of the _CTL commands

[kib]

Summary

Also make the operations consistently interpret P_PID:0 as curproc.

Per commit view is available at https://kib.kiev.ua/git/gitweb.cgi?p=deviant3.git;a=shortlog;h=refs/heads/procctl

[kib]

Fixes

[markj]

sys/kern/kern_procctl.c

660

"single_proc" is too cryptic IMO, at first I thought it was referring to thread_single().

664

This looks like it should be named need_candebug.

731

The code below uses both data and &x to refer to the same thing, I think it should be consistent one way or the other.

740

I think this can copy out uninitialized bytes in some cases, e.g. for PROC_REAP_KILL I don't see any zeroing. I would zero out &x here and remove zeroing from the individual handlers where appropriate.

807

I wonder why we bother passing td when most handlers don't use it, and the rest expect td == curthread.

[kib]

sys/kern/kern_procctl.c

807

Same as for any syscall. I find it sometimes convenient, in fact.

[kib]

single_proc->one_proc
candebug->need_candebug
zero x
remove data

[markj]

sys/kern/kern_procctl.c

138

as in trace_ctl().

155

762

913

Why do we check p_state == PRS_NEW below but not here?

Is there any reason to permit operations on exiting processes here?

[markj]

sys/kern/kern_procctl.c

662

copyout_on_error seems more precise.

[kib]

sys/kern/kern_procctl.c

913

PRS_NEW is checked by pfind(), but not by the p_pglist iteration.
Also pfind() checks for zombies, but not for processes being exited.

After some rounds of changing my opinion, I think that there is no reason to disable looking at the exiting processes. On the other hand, p_pglist iteration needs to check for PRS_ZOMBIE

[kib]

s/9/0
s/always_copyout/copyout_on_error
Filter out zombies for group op
Remove uses of curproc

[markj]

sys/kern/kern_procctl.c

913

Ah, I missed that pfind() excludes PRS_NEW. I still wonder about the utility of permitting operations on exiting process, especially since some commands need to exclude them already, but ok.

[kib]

sys/kern/kern_procctl.c

913

For P_WEXIT but not yet PRS_ZOMBIE processes, it might be useful to be able to query some of the parameters reported by _STATUS commands. Not too useful, but might be interesting, e.g. reaper might want to query PDEATHSIG value, for something that could be relatively realistic.

[emaste]

Presumably need a clarification in procctl.2 also

[kib]

Update man page for id zero and debug privs.

[emaste]

manpage text LGTM