[Bug 200493] Killing pid 11 (idle) wedges the disk IO

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=200493

Konstantin Belousov <kib at FreeBSD.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kib at FreeBSD.org

--- Comment #1 from Konstantin Belousov <kib at FreeBSD.org> ---
(In reply to Edward Tomasz Napierala from comment #0)
It is very _weird_ bug report from developer.  You did not invesigated what is
the state of the idle (?) process after the kill, obviously pid value does not
matter since kernel starts variable number of the kprocs during the boot (on my
kernel idle has pid 10). And last and most important, kernel processes do not
process signals: there is no place where cursig()/postsig() pair is called,
since there is no return from kernel to user mode, and no ast handler called.

E.g. after I do kill -9 10 (pid 10 is my idle process), I see
sandy% sudo procstat -i 10
  PID COMM             SIG     FLAGS
  10 idle             KILL     P--
I.e. SIGKILL was put into the queue, but nothing processed it.  And I do not
observe any weirdness in the system behaviour afterward.

If your system consumed the SIGKILL, there should be some code which called
postsig() in the context of the idle threads.

FYI, the idle loop is sched_idletd(), private for the given scheduler
implementation.

-- 
You are receiving this mail because:
You are the assignee for the bug.