On 09.02.21 14:38, Oscar Salvador wrote:
Many places expects us to pass a pageblock aligned range. E.g: memmap_init_zone() needs a pageblock aligned range in order to set the proper migrate type for it. online_pages() needs to operate on a pageblock aligned range for isolation purposes. Make sure we disable the feature in case we cannot guarantee the right alignment. Signed-off-by: Oscar Salvador <osalvador@xxxxxxx> --- mm/memory_hotplug.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c index d3fb036d33fd..1a4d5dd1a2c8 100644 --- a/mm/memory_hotplug.c +++ b/mm/memory_hotplug.c @@ -56,12 +56,16 @@ static int memmap_on_memory_show(char *buffer, const struct kernel_param *kp) static __meminit int memmap_on_memory_store(const char *val, const struct kernel_param *kp) { + unsigned long pageblock_size = PFN_PHYS(pageblock_nr_pages); + /* * Fail silently in case we cannot enable it due to system constraints. * User can always check whether it is enabled or not via /sys/module. */ if (!IS_ENABLED(CONFIG_ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE) || - (PMD_SIZE % sizeof(struct page))) + (PMD_SIZE % sizeof(struct page)) || + !(MIN_MEMORY_BLOCK_SIZE - PMD_SIZE) || + !(MIN_MEMORY_BLOCK_SIZE - PMD_SIZE) % pageblock_size) return 0;return param_set_bool(val, kp);
Dito, rather squash in #1 and add a comment explaining what's happening there.
-- Thanks, David / dhildenb
