On Wednesday 13 February 2008 09:04:53 Or Gerlitz wrote:
> Pradeep Satyanarayana wrote:
> > This patch fixes -fail to destroy ipoib rx QP 
> > (https://bugs.openfabrics.org/show_bug.cgi?id=906)
> > Hence the usecnt issue reported previously on ehca is solved and allows the 
> > qp to be destroyed.
> > 
> > As per Eli's request, I am splitting up the patches. This is first portion 
> > of yesterday's patch.
> > Tested on ppc64 machines with ehca and mthca.
> 
> Also here, does this problem exist in the 2.6.25-rc1 upstream code as 
> well? from the change log I don't understand the source of the problem 
> (only the symptom of failing to destroy ipoib/cm rx QP) and the solution.
> 
> Or.

Hi,
yes this problem does also exist in 2.6.25-rc1. It was introduced by a patch 
from roland:
http://git.kernel.org/?p=linux/kernel/git/roland/infiniband.git;a=commitdiff;h=efcd99717f76c6d19dd81203c60fe198480de522

In function ipoib_cm_dev_stop() the error-,drain- and flush lists are put into 
a local list after a timeout.
In the past there was a list_for_each_entry loop iterating over this local list 
and destroyed all added QPs. 
With the patch above the list_for_each_entry call is moved to function 
ipoib_cm_free_rx_reap_list(),
which does not iterate the former local list, but device's reap_list.
Pradeeps patch puts now all QPs after a timeout from error, drain and flush 
lists into the reap_list so that they were all freed in 
poib_cm_free_rx_reap_list().

Stefan
_______________________________________________
ewg mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg

Reply via email to