On 8/16/21 12:47 AM, Dmitry Vyukov wrote: > On Sun, 15 Aug 2021 at 20:17, Mike Kravetz <mike.kravetz@xxxxxxxxxx> wrote: >> >> On 8/14/21 4:24 PM, Matthew Wilcox wrote: >>> On Sat, Aug 14, 2021 at 03:06:52PM -0700, Andrew Morton wrote: >>>> On Sat, 14 Aug 2021 04:07:23 -0700 syzbot <syzbot+57876cfc2dca0c6b2e6a@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote: >>>>> HEAD commit: 92d00774360d Add linux-next specific files for 20210810 >>>>> RIP: 0010:__list_add include/linux/list.h:71 [inline] >>>>> RIP: 0010:list_add include/linux/list.h:86 [inline] >>>>> RIP: 0010:alloc_huge_page+0xdcb/0x11b0 mm/hugetlb.c:2821 >>>> >>>> That appears to be >>>> >>>> list_add(&page->lru, &h->hugepage_activelist); >>>> >>>> I'm suspecting Mike's surplus page changes? >>> >>> Is this not the same as >>> https://lore.kernel.org/linux-mm/846c4502-3332-0d25-87f5-cb3b71afc38f@xxxxxxxxxxx/ >>> which is already patched in your tree? >> >> Yes, I believe it is the same, and already fixed. RCX: fffffffffffffffc >> makes me 99.9% confident. > > Let's tell syzbot about the fix: > > #syz fix: > mm/hugetlb: Initialize page to NULL in alloc_buddy_huge_page_with_mpol() > Thanks! I verified and removed my .01% of doubt. -- Mike Kravetz
