On Tue, 4 Feb 2020 15:24:38 +0800 Shile Zhang <shile.zhang@xxxxxxxxxxxxxxxxx> wrote: > Hi, Andrew, > > Sorry for ping, could you please also help to check this issue? > Any comments from you is welcome! > Kirill's suggestion isn't pretty but it appears that it will work OK - redo deferred_init_memmap() so it releases 1024(?) pages at a time, releasing and retaking pgdat_resize_lock/unlock each time around?
