SIGSTOP and SIGKILL
kostikbel at gmail.com
Mon Mar 14 09:03:26 UTC 2011
On Sun, Mar 13, 2011 at 11:30:00PM -0700, Ravi Murty wrote:
> I haven't, are there specific improvements in this area of the kernel?
First, the 8.2, compared to 8.0, changed the mechanism of delivering
the process-global signal to a thread. Now, the thread to deliver is
selected at the moment of delivery, instead of the moment of post.
This was done to close a race where either thread changed signal mask
after post, or exited after post, causing signal be undefinitely
delayed or ignored, respectively.
Second, 8.2 less often stops the threads in the sleep state, more often
trying to move the thread to safer user<->kernel boundary, to take a stop.
Sleeping thread might indicate the unsafe sleep (from the POV of stopping)
by using PBDRY flag to msleep.
Both the changes are not directly address the behaviour you described,
but they are explicitely in the area of your interest.
BTW, it is not clear to me, does the process in the state as you
described, able to unwind from the stop state using SIGCONT ? I.e.,
is it 'weird but recoverable' state, or is it 'weird and unrecoverable'
> On Mar 13, 2011, at 10:30 PM, Erich Dollansky <erich at alogreentechnologies.com> wrote:
> > Hi,
> > On Monday 14 March 2011 11:41:21 Ravi Murty wrote:
> >> Any good solutions to this problem?
> > did you try your program on 8.2?
> > Erich
> freebsd-hackers at freebsd.org mailing list
> To unsubscribe, send any mail to "freebsd-hackers-unsubscribe at freebsd.org"
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 196 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-hackers/attachments/20110314/8744c057/attachment.pgp
More information about the freebsd-hackers