On Thu 07-01-21 19:38:00, Muchun Song wrote: > On Thu, Jan 7, 2021 at 7:18 PM Michal Hocko <mho...@suse.com> wrote: > > > > On Thu 07-01-21 16:53:13, Muchun Song wrote: > > > On Thu, Jan 7, 2021 at 4:41 PM Michal Hocko <mho...@suse.com> wrote: > > > > > > > > On Thu 07-01-21 13:39:38, Muchun Song wrote: > > > > > On Thu, Jan 7, 2021 at 12:56 AM Michal Hocko <mho...@suse.com> wrote: > > > > > > > > > > > > On Wed 06-01-21 16:47:36, Muchun Song wrote: > > > > > > > There is a race condition between __free_huge_page() > > > > > > > and dissolve_free_huge_page(). > > > > > > > > > > > > > > CPU0: CPU1: > > > > > > > > > > > > > > // page_count(page) == 1 > > > > > > > put_page(page) > > > > > > > __free_huge_page(page) > > > > > > > dissolve_free_huge_page(page) > > > > > > > spin_lock(&hugetlb_lock) > > > > > > > // PageHuge(page) && > > > > > > > !page_count(page) > > > > > > > update_and_free_page(page) > > > > > > > // page is freed to the buddy > > > > > > > spin_unlock(&hugetlb_lock) > > > > > > > spin_lock(&hugetlb_lock) > > > > > > > clear_page_huge_active(page) > > > > > > > enqueue_huge_page(page) > > > > > > > // It is wrong, the page is already freed > > > > > > > spin_unlock(&hugetlb_lock) > > > > > > > > > > > > > > The race windows is between put_page() and spin_lock() which > > > > > > > is in the __free_huge_page(). > > > > > > > > > > > > The race window reall is between put_page and > > > > > > dissolve_free_huge_page. > > > > > > And the result is that the put_page path would clobber an unrelated > > > > > > page > > > > > > (either free or already reused page) which is quite serious. > > > > > > Fortunatelly pages are dissolved very rarely. I believe that user > > > > > > would > > > > > > require to be privileged to hit this by intention. > > > > > > > > > > > > > We should make sure that the page is already on the free list > > > > > > > when it is dissolved. > > > > > > > > > > > > Another option would be to check for PageHuge in __free_huge_page. > > > > > > Have > > > > > > you considered that rather than add yet another state? The scope of > > > > > > the > > > > > > spinlock would have to be extended. If that sounds more tricky then > > > > > > can > > > > > > we check the page->lru in the dissolve path? If the page is still > > > > > > PageHuge and reference count 0 then there shouldn't be many options > > > > > > where it can be queued, right? > > > > > > > > > > Did you mean that we iterate over the free list to check whether > > > > > the page is on the free list? > > > > > > > > No I meant to check that the page is enqueued which along with ref count > > > > = 0 should mean it has been released to the pool unless I am missing > > > > something. > > > > > > The page can be on the free list or active list or empty when it > > > is freed to the pool. How to check whether it is on the free list? > > > > As I've said, I might be missing something here. But if the page is > > freed why does it matter whether it is on a active list or free list > > from the dissolve operation POV? > > As you said "check the page->lru". I have a question. > How to check the page->lru in the dissolve path?
list_empty? > BTW, dissolve_free_huge_page aims to free the page > to buddy allocator. put_page (for HugeTLB page) aims > to free the page to the hugepage pool. Right. Can we simply back off in the dissolving path when ref count is 0 && PageHuge() if list_empty(page->lru)? Is there any other scenario when the all above is true and the page is not being freed? -- Michal Hocko SUSE Labs