On Tue, 2010-08-31 at 18:26 -0700, Robert Love wrote: > On Wed, 2010-08-25 at 01:31 +0000, Bhanu Gollapudi wrote: > > On systems with with higher nr_cpu_ids, per cpu exchange pool will > > have very limited exchange resources. This can cause certain > > operations such as discovery to fail even with finite amount > > of retires. This problem is handled by dividing the entire > > Hi Bhanu, > > Can you tell me a bit more about your scenario and how many CPUs > you're dealing with? Is it an offload EM or non-offload EM that's > running out of resources?
Hi Robert, I described the scenario in one of the earlier emails http://www.mail-archive.com/[email protected]/msg07738.html It is not an offload EM, and nr_cpu_ids in that system is 32, which leaves only 128 xids per cpu pool. > > The (non-offload) EM starts with 4k XIDs. It must be an extreme case if > it is running out of per-CPU XIDs. Especially if you're failing > discovery, when there shouldn't be that many exchanges in use. Please note that this is an extreme case, where there were 255 NPIV ports and the issue was triggered when a switch port shutdown/no-shutdown is performed. (There is another open issue in disc_stop which tries to cancel itself, described in the email) > > > exchange pool into two, one half to be used for cpu pool, and the > > other half to be used as common pool. If the exchange resource > > is not available in per cpu pool, the exchange resouce will > > be obtained from the common pool. > > I'm not sure the offload EM would want this common pool since the OEM > may have a much smaller set of XIDs. Doesnt the offload EM run into similar issues? > > Thanks, //Rob > > _______________________________________________ devel mailing list [email protected] http://www.open-fcoe.org/mailman/listinfo/devel
