On Thu, Jan 7, 2021 at 1:07 AM Michal Hocko [off-list ref] wrote:

On Wed 06-01-21 16:47:37, Muchun Song wrote:

quoted

When dissolve_free_huge_page() races with __free_huge_page(), we can
do a retry. Because the race window is small.

Is this a bug fix or mere optimization. I have hard time to tell from
the description.

It is optimization. Thanks.

quoted

Signed-off-by: Muchun Song <redacted>
---
 mm/hugetlb.c | 26 +++++++++++++++++++++-----
 1 file changed, 21 insertions(+), 5 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c

[...]

quoted

@@ -1825,6 +1828,14 @@ int dissolve_free_huge_page(struct page *page)
      }
 out:
      spin_unlock(&hugetlb_lock);
+
+     /*
+      * If the freeing of the HugeTLB page is put on a work queue, we should
+      * flush the work before retrying.
+      */
+     if (unlikely(rc == -EAGAIN))
+             flush_work(&free_hpage_work);

Is it safe to wait for the work to finish from this context?

Yes. It is safe.

--
Michal Hocko
SUSE Labs