Hi, This patchset is suggesting to adjust unpoison_memory() to the new way of isolation and refcounting of hwpoisoned pages. As mentioned in [1], NR_FREE_PAGES is incremented via newly added function undoing take_page_off_buddy(), so the counter is consistent with this series. It depends on some other hwpoison-related patches (not mainlined but in linux-mm), so I listed the dependencies here: - mm/hwpoison: do not lock page again when me_huge_page() successfully recovers - mm/memory-failure: make sure wait for page writeback in memory_failure - mm,hwpoison: make get_hwpoison_page() call get_any_page() - mm,hwpoison: fix race with hugetlb page allocation - mm,hwpoison: Send SIGBUS with error virutal address - mm,hwpoison: Return -EHWPOISON to denote that the page has already been poisoned - mm/memory-failure: Use a mutex to avoid memory_failure() races Or GitHub branch I pushed for easy patch access might be helpful for reviewers. git fetch https://github.com/nhoriguchi/linux hwpoison [1] https://lore.kernel.org/linux-mm/20210527003436.GA3543069@xxxxxxxxxxxxxxxxxxxxxxxxxxx/ Thanks, Naoya Horiguchi --- Summary: Naoya Horiguchi (6): mm/hwpoison: mf_mutex for soft offline and unpoison mm/hwpoison: remove race consideration mm/hwpoison: introduce MF_MSG_PAGETABLE mm/hwpoison: remove MF_MSG_BUDDY_2ND and MF_MSG_POISONED_HUGE mm/hwpoison: make some kernel pages handlable mm/hwpoison: fix unpoison_memory() include/linux/mm.h | 4 +- include/linux/page-flags.h | 4 ++ include/ras/ras_event.h | 3 +- mm/memory-failure.c | 176 +++++++++++++++++++++++++-------------------- mm/page_alloc.c | 19 +++++ 5 files changed, 126 insertions(+), 80 deletions(-)
