Hello, I have been working on a coherent device memory (CDM) representation in the kernel. Last year, I had posted these two RFCs in this regard which attempts to represent non-system RAM coherent device memory as a NUMA node with some allocation restrictions. https://lkml.org/lkml/2016/10/24/19 (CDM with modified zonelists) https://lkml.org/lkml/2016/11/22/339 (CDM with modified cpusets) During the course of this work, I had experimented with existing device memory frameworks like ZONE_DEVICE, HMM etc. In my quest for isolation, also looked into how early system boot allocated CMA region can be used in this regard. Thought I would like to participate primarily on the device memory management infrastructure discussions (detailed TOPICs list can be found below) but would also like to contribute in those areas where I had faced some limitations or the other during the CDM work. I would like to attend to discuss on the listed topics on the mailing list as mentioned below which I have already replied in their respective threads as well. (1) [LSF/MM TOPIC/ATTEND] Memory Types (2) [LSF/MM TOPIC] Un-addressable device memory and block/fs implications (3) [LSF/MM TOPIC] Memory hotplug, ZONE_DEVICE, and the future of struct page (4) [LSF/MM ATTEND] HMM, CDM and other infrastructure for device memory management Apart from this, would like to discuss on the following generic topics as I had mentioned earlier. (1) Support for memory hotplug as CMA regions (2) Seamless migration between LRU pages and device managed non LRU CMA pages (3) Explore the possibility of enforcing CDM node isolation through a changed buddy page allocator specifically how it currently handles requested nodemask_t along with cpuset based nodemask at various phases of fast path and slow path Regards Anshuman -- 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>