On Thu, Nov 24, 2011 at 12:21:44PM +0000, Mel Gorman wrote: > On Thu, Nov 24, 2011 at 02:19:43AM +0100, Andrea Arcangeli wrote: > > Yes also note, ironically this is likely to be a better test for this > > without the __GFP_MOVABLE in block_dev.c. Even if we want it fixed, > > maybe another source that reduces the non movable pages may be needed then. > > > > I'm also running other tests to avoid tuning for just this test cases. > Right now, the list looks like; > > 1. postmark with something creating anonymous mappings in the background > 2. plain USB writing while creating anonymous mappings I've been testing the original case that started this thread -- writing multiple GB to a USB attached, FAT, very slow SD card. I'm currently running 7f80850d + "mm: Do not stall in synchronous compaction for THP allocations" Mel's original patch. With this patch I cannot reproduce the hangs that I saw. I haven't retried without the patch to confirm that they're reproducible, though. someone asked about CONFIG_NUMA; I have CONFIG_NUMA=y. I can reboot this weekend; what patches should I test with next? -andy -- 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/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>