On Thu, Mar 28, 2019 at 04:47:02PM -0700, Mike Kravetz wrote: > I stumbled on these two hugetlbfs issues while looking at other things: > - The 'restore reserve' functionality at page free time should not > be adjusting subpool counts. > - A BUG can be triggered (not easily) due to temporarily mapping a > page before doing a COW. > > Both are described in detail in the commit message of the patches. > I would appreciate comments from Davidlohr Bueso as one patch is > directly related to code he added in commit 8382d914ebf7. > > I did not cc stable as the first problem has been around since reserves > were added to hugetlbfs and nobody has noticed. The second is very hard > to hit/reproduce. > > v2 - Update definition and all callers of hugetlb_fault_mutex_hash as > the arguments mm and vma are no longer used or necessary. > > Mike Kravetz (2): > huegtlbfs: on restore reserve error path retain subpool reservation > hugetlb: use same fault hash key for shared and private mappings > > fs/hugetlbfs/inode.c | 7 ++----- > include/linux/hugetlb.h | 4 +--- > mm/hugetlb.c | 43 +++++++++++++++++++++-------------------- > mm/userfaultfd.c | 3 +-- > 4 files changed, 26 insertions(+), 31 deletions(-) Both fixes look fine to me. Reviewed-by: Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx>
