In our recent cpu/memory hotadd testing, with multiple nodes hotadded, kernel easily panics under stress workload like kernel building. The root cause is that the new populated zones of hotadded nodes are sharing same per_cpu_pageset, i.e. boot strapping boot_pageset, which finally causes page state wrong. The following three patches will setup the pagesets for hotadded nodes with dynamically allocated per_cpu_pageset struct. --- include/linux/memory_hotplug.h | 8 +++++ include/linux/mmzone.h | 2 +- init/main.c | 2 +- mm/memory_hotplug.c | 27 +++++++++++----- mm/page_alloc.c | 66 +++++++++++++++++++++++++++------------- 5 files changed, 74 insertions(+), 31 deletions(-) -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>