On Fri, Mar 26, 2010 at 06:54:06PM +0100, Johannes Weiner wrote: > Oh, shoot, the cast needs to be renamed to (union split_pmd *) as well. I only verified x86 builds without PAE, I fixed that bit now and verified it gets past it. However it still doesn't build because PG_compound_lock exceeds the 32bit page->flags (with zone and other stuff). I don't immediately see PG_ bits to reuse, but I think PageBuddy can be converted to page->_count = -1. For now the only way to build PAE on x86 with hugepages enabled, it is to set PAT=n. Considering that the only way to disable transparent hugepages is to enable the embedded settings in the config, it isn't such great but again I'm optimistic we can drop PG_buddy. PG_reclaim is very dubious too but as long as we want to retain that functionality of trying to free pages after swapout I/O completion immediately, there's no way to keep functionality removing PG_reclaim. Others are similar... -- 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>