[ofa-general] [PATCH RFC] libibumad: give up cpu time slice if write() falied

[Sasha Khapyorsky]

On 10:53 Thu 29 Jan     , Jason Gunthorpe wrote:
> On Thu, Jan 29, 2009 at 05:26:37PM +0200, Yevgeny Kliteynik wrote:
> 
> > Then I added 20 msec sleep after write() failure to make it give up the cpu time
> > slice, and I saw some improvement. When I added sleep of several seconds, the problem
> > disappeared completely, but then of course, the client's transaction will fail with
> > timeout unless you specifically increase the transaction timeout on the client side.
> 
> Poll or a blocking fd would be a better choice than a sleep...

True.

(However in libibumad fd is opened with O_NONBLOCK flag - I have no a
good understanding yet why it was done so (the change is from 2005).)

Sasha