Replace the now deprecated kmem_atomic() with kmem_local_page().

Optimize for the non-highmem cases and avoid disabling preemption and
pagefaults, the caller's context is atomic anyway, but that is irrelevant
to kmem. The memcpy itself does not require any such semantics and the
mapping would hold valid across context switches anyway. Further, highmem
is planned to to be removed[1].

[1] https://lore.kernel.org/all/4ff89b72-03ff-4447-9d21-dd6a5fe1550f@app.fastmail.com/
Signed-off-by: Davidlohr Bueso <dave@stgolabs.net>
---
 kernel/watch_queue.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/watch_queue.c b/kernel/watch_queue.c
index 7e45559521af..52f89f1137da 100644
--- a/kernel/watch_queue.c
+++ b/kernel/watch_queue.c
@@ -119,9 +119,9 @@ static bool post_one_notification(struct watch_queue *wqueue,
 	offset = note % WATCH_QUEUE_NOTES_PER_PAGE * WATCH_QUEUE_NOTE_SIZE;
 	get_page(page);
 	len = n->info & WATCH_INFO_LENGTH;
-	p = kmap_atomic(page);
+	p = kmap_local_page(page);
 	memcpy(p + offset, n, len);
-	kunmap_atomic(p);
+	kunmap_local(p);
 
 	buf = pipe_buf(pipe, head);
 	buf->page = page;
-- 
2.39.5

On Tue, 18 Nov 2025, Davidlohr Bueso wrote:

>Replace the now deprecated kmem_atomic() with kmem_local_page().

heh spell checker fail.

Will send a v2 with s/kmem/kmap for the whole changelog, sorry for the noise.