On Thu, Jul 19 2007, Vasily Tarasov wrote: > On Wed, 2007-07-18 at 20:51 +0200, Jens Axboe wrote: > > On Wed, Jul 18 2007, Vasily Tarasov wrote: > > > Jens, I think the last patch, that makes queues allocation per priority, > > > has a problem. > > > > > > If we have two processes with different ioprio_class, but the same > > > ioprio_data, their async requests will fall into the same queue. I guess > > > such behavior is not expected, because it's not right to put real-time > > > requests and best-effort requests in the same queue. > > > > > > The attached patch fixes the problem by introducing additional *cfqq > > > fields on cfqd, pointing to per-(class,priority) async queues. > > > > Ugh yes. I'm pretty tempted just to reinstate the cfqq hash again, it > > used to be a clean up but now the it's not stacking up so well. > > > > Hello, Jens, > > From my humble point of view cfqq hash has two problems: > > 1. It is excess data structure. All needed information can be obtained > from other structures easily, so the presence of hash is a bit > strange... I mean that it's aim is not obvious :) > > 2. Hash hides from a developer a pretty important concept of CFQ: there > are shared between processes per-priority async queues. I think the code > is the best documentation, so the explicit async cfqq pointers at cfqd > structure reveal this concept greatly. > > Summary: > > IMHO the hash revival is not very good way. However, this is of course > fully in your competence to choose the right decision! ;)
Yeah, it's probably still better off without the hash. I'll play with it a bit and see what comes of it. -- Jens Axboe _______________________________________________ Devel mailing list [email protected] https://openvz.org/mailman/listinfo/devel
