[Samba] nmbd eats cpu for breakfast after upgrade from 3.5 to 3.6
jra at samba.org
Thu Apr 19 12:50:43 MDT 2012
On Thu, Apr 19, 2012 at 11:49:36AM -0700, Jeremy Allison wrote:
> On Thu, Apr 19, 2012 at 09:03:35AM +0300, Ander Punnar wrote:
> > On 17 April 2012 20:19, Jeremy Allison <jra at samba.org> wrote:
> > > What does strace -p <nmb-pid> say ?
> > Today, strace showed lots of following:
> > recvfrom(13, 0x7fff7c027a00, 576, 0, 0x7fff7c027c40, 0x7fff7c0279dc) =
> > -1 EAGAIN (Resource temporarily unavailable)
> > But in log.nmbd, there wasn't anything "uncommon".
> You get this when the socket is set non-blocking and a recvfrom()
> returns -1, EAGAIN.
and there's nothing available to be read (forgot to add that).
What is the syscall above the recvfrom ? We should only be
doing the recvfrom when a poll() shows the socket has data
available for reading.
More information about the samba