Re: Revert "aio: fix aio request leak when events are reaped by user space"

[Dan Aloni <dan@kernelim.com>]

On Tue, Aug 19, 2014 at 12:54:04PM -0400, Benjamin LaHaise wrote:
> On Tue, Aug 19, 2014 at 07:37:33PM +0300, Dan Aloni wrote:
> > Some testing I've done today indicates that the original commit broke
> > AIO with regard to users that overflow the maximum number of request
> > per IO context (where -EAGAIN is returned).
> > 
> > In fact, it did worse - the attached C program can easily overrun the
> > ring buffer that is responsible for managing the completed requests,
> > and caused notification about their completion never to be returned.
> 
> Argh, that would be a problem.
> 
> ...
> > This reverts commit b34e0e1319b31202eb142dcd9688cf7145a30bf6.
> 
> Reverting isn't okay, as that reintroduces another regression.  We need 
> to come up with a fix for this issue that doesn't reintroduce the other 
> regression for events reaped in user space.  Let me have a look and see 
> what I can come up with...

About the original regression you mention, is there a program you can
indicate that reproduces it? On my setups, the regression testing in 
libaio was not able to detect the current regression too.


-- 
Dan Aloni