On Mon, Dec 21, 2020 at 10:14 PM Oscar Salvador [off-list ref] wrote:

On Mon, Dec 21, 2020 at 07:07:18PM +0800, Muchun Song wrote:

quoted

quoted

The above implies that update_and_free_page() is __always__ called from a
non-task context, but that is not always the case?

IIUC, here is always the case.

I might be missing something obvious, so bear with me.

I guess you are refering to the call __free_huge_page()->update_and_free_page().
AFAICS, free_huge_page might call __free_huge_page right away when in task
context, and so, we would be calling update_and_free in a task context as well.

Yeah. You are right. I mean the call __free_huge_page()->update_and_free_page().
Because update_and_free_page is called under hugetlb_lock, it is
non-task context,
right?

Or are you referring to the other callers?

--
Oscar Salvador
SUSE L3

-- 
Yours,
Muchun