Change Details

In order to not hold on to the hostcache hashbucket locks excessively, coordinate the repeated drain of the sbuf in between iterating over all hashbuckets, and if need be, also while processing a more used bucket. This may change the output to not give a complete snapshot in time (it didn't before between hash buckets either), but seems to be a reasonable trade off between visibility and excessive locking. The latter may prevent the use of stored TCPCB data for subsequent sessions as long as the hostcache dump lasts - either stalling the session setup, or proceeding with non- optimal parameters.

Explicitly drain the sbuf after completing each hash bucket to minimize the work performed while holding the hash bucket lock. This may change the output to not give a complete snapshot in time (it didn't before between hash buckets either), but seems to be a reasonable trade off between visibility and excessive locking. The latter may prevent the use of stored TCPCB data for subsequent sessions as long as the hostcache dump lasts - either stalling the session setup, or proceeding with non- optimal parameters.

In order to not hold on to the hostcacheExplicitly drain the sbuf after completing each hash bucket locks excessively, coordinate the repeated drain ofto minimize the work performed while holding the sbufhash in between iterating over all hashbuckets, and if need be, also while processing a more used bucket lock. This may change the output to not give a complete snapshot in time (it didn't before between hash buckets either), but seems to be a reasonable trade off between visibility and excessive locking. The latter may prevent the use of stored TCPCB data for subsequent sessions as long as the hostcache dump lasts - either stalling the session setup, or proceeding with non- optimal parameters.