Re: [ofa-general] [PATCH] opensm/osm_qos_policy.c: matching PR query to QoS level with pkey
Yevgeny Kliteynik wrote: Nope, just the other way around. Yevgeny, we want to do some testing/validation to understand better what's goes on here, will get back to you soon Or. ___ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [ofa-general] [PATCH] opensm/osm_qos_policy.c: matching PR query to QoS level with pkey
Or Gerlitz wrote: Yevgeny Kliteynik wrote: False negatives. PR queries with PKeys (e.g. IPoIB interfaces) weren't matched to their rules. Yevgeny, Our understanding is that the bug comes into play only for queries done on a partial membership pkey, do you agree? Nope, just the other way around. When some pkey is defined in QoS policy, it is stored internally w/o the MSB. When query comes with a full member pkey (such as 0x for IPoIB), this pkey is not matched to the stored QoS policy rule. The fix was to treat any pkey that comes from request as partial membership pkey. Note that this is done for the QoS policy rules matching only. The two sides of this PR query still have to comply to the usual IB spec pkey rules. -- Yevgeny Or. ___ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [ofa-general] [PATCH] opensm/osm_qos_policy.c: matching PR query to QoS level with pkey
Yevgeny Kliteynik wrote: False negatives. PR queries with PKeys (e.g. IPoIB interfaces) weren't matched to their rules. Yevgeny, Our understanding is that the bug comes into play only for queries done on a partial membership pkey, do you agree? Or. ___ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [ofa-general] [PATCH] opensm/osm_qos_policy.c: matching PR query to QoS level with pkey
Or Gerlitz wrote: Yevgeny Kliteynik wrote: Fixing a bug in matching PR query to QoS levels when pkey specified - pkeys in QoS policy are held w/o the MSB. Hi Yevgeny, so what's the impact of this bug in the field? does it create false positives or false negatives? False negatives. PR queries with PKeys (e.g. IPoIB interfaces) weren't matched to their rules. -- Yevgeny Or. ___ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
Re: [ofa-general] [PATCH] opensm/osm_qos_policy.c: matching PR query to QoS level with pkey
Yevgeny Kliteynik wrote: Fixing a bug in matching PR query to QoS levels when pkey specified - pkeys in QoS policy are held w/o the MSB. Hi Yevgeny, so what's the impact of this bug in the field? does it create false positives or false negatives? Or. ___ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
