(2012/11/15 11:54), Glauber Costa wrote: > The original reason for the existence of this test, was that > memcg_kmem_cache_create could be called from either softirq context > (where memcg_stop/resume_account is not needed), or process context, > (where memcg_stop/resume_account is needed). Just skipping it > in-function was the cleanest way to merge both behaviors. The reason for > that is that we would try to create caches right away through > memcg_kmem_cache_create if the context would allow us to. > > However, the final version of the code that merged did not have this > behavior and we always queue up new cache creation. Thus, instead of a > comment explaining why current->mm test is needed, my proposal in this > patch is to remove memcg_stop/resume_account from the worker thread and > make sure all callers have a valid mm context. > > Signed-off-by: Glauber Costa <glommer@xxxxxxxxxxxxx> > CC: Michal Hocko <mhocko@xxxxxxx> > CC: Kamezawa Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> > CC: Johannes Weiner <hannes@xxxxxxxxxxx> > CC: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> seems ok to me. But do we need VM_BUG_ON() ? It seems functions called under memcg_stop_kmem_account() doesn't access current->mm... Anyway. Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>