[RFC PATCH v2 03/12] Add *kill, pidfd_send_signal syscalls to %process class

[Dmitry V. Levin]

On Fri, Jun 12, 2020 at 09:46:11PM +0200, Ákos Uzonyi wrote:
> On Fri, 12 Jun 2020 at 21:28, Dmitry V. Levin wrote:
> > On Fri, Jun 12, 2020 at 08:23:57PM +0200, Ákos Uzonyi wrote:
> > > Since rt_tgsigqueueinfo is part of the %process class, it is reasonable
> > > to include all other syscalls that send a signal to a process.
> >
> > It's still unclear what is the criteria for including syscalls in %process
> > class.  For example, what about rt_sigqueueinfo?
> 
> Oh yes, I forgot about rt_sigqueueinfo, thanks. I think after adding
> rt_sigqueueinfo the criteria I wrote is fulfilled.

The manual page describes %process class as "all system calls which
involve process management.  This is useful for watching the fork, wait,
and exec steps of a process."  This description is rather vague,
and the class currently includes the following system calls:
- arch_prctl
- clone
- clone2
- clone3
- execv
- execve
- execveat
- exit
- exit_group
- fork
- rt_tgsigqueueinfo
- unshare
- vfork
- wait4
- waitid
- waitpid

It's not obvious why rt_tgsigqueueinfo or any other syscall that sends
signals to processes should belong to "involve process management" class.


-- 
ldv