On Thu 05-01-17 17:43:38, Mike Kravetz wrote: > On 01/05/2017 07:37 AM, Michal Hocko wrote: [...] > > Another problem, somehow related, is that there are areas which have > > evolved into a really bad shape because nobody has really payed > > attention to them from the architectural POV when they were merged. To > > name one the memory hotplug doesn't seem very healthy, full of kludges, > > random hacks and fixes for fixes working for a particualr usecase > > without any longterm vision. We have allowed to (ab)use concepts like > > ZONE_MOVABLE which are finding new users because that seems to be the > > simplest way forward. Now we are left with fixing the code which has > > some fundamental issues because it is used out there. Are we going to do > > anything about those? E.g. generate a list of them, discuss how to make > > that code healthy again and do not allow new features until we sort that > > out? > > hugetlb reservation processing seems to be one of those areas. I certainly > have been guilty of stretching the limits of the current code to meet the > demands of new functionality. It has been my desire to do some rewrite or > rearchitecture in this area. I think that it would be really useful to start by a throughout design documentation of the current code before any rewrite. I believe this will already tell us a lot of the design complexities and shortcomings already. -- Michal Hocko SUSE Labs -- 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>