Dmitry Vyukov reported yet another VM_BUG_ON_PAGE(PageTail(page)) bug from isolate_lru_page().
The first patch relaxes VM_BUG_ON_PAGE(): no need to crash on tail pages which are not on LRU. We are not going to isolate them anyway. The second patch tries streamline logic within queue_pages_range(): no need to scan non migratable VMAs. The patch requires more careful review. Any of these patches should fix the issue. I think both should be applied. The first one is subject for 4.5 as the bogus was introduced there. Kirill A. Shutemov (2): mm: fix bogus VM_BUG_ON_PAGE() in isolate_lru_page() mempolicy: do not try to queue pages from !vma_migratable() mm/mempolicy.c | 14 +++++--------- mm/vmscan.c | 2 +- 2 files changed, 6 insertions(+), 10 deletions(-) -- 2.7.0
