On Wed, Jun 03, 2026 at 02:08:25PM +0100, Mykyta Yatsenko wrote:
>
> For v7 I'm dropping automatic_shrinking, because it adds a risk of
> calling schedule_work() on element deletion path (__rhashtable_remove_fast_one())
> when hashtable size drops below 30% of the capacity.

Now that expansion uses irq work I think shrinking should switch
to that as well.

---8<---
Use irq work for automatic shrinking so that this may be called
in NMI context.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

diff --git a/include/linux/rhashtable.h b/include/linux/rhashtable.h
index ef5230cece36..0693bce6f890 100644
--- a/include/linux/rhashtable.h
+++ b/include/linux/rhashtable.h
@@ -1117,7 +1117,7 @@ static __always_inline int __rhashtable_remove_fast_one(
 		atomic_dec(&ht->nelems);
 		if (unlikely(ht->p.automatic_shrinking &&
 			     rht_shrink_below_30(ht, tbl)))
-			schedule_work(&ht->run_work);
+			irq_work_queue(&ht->run_irq_work);
 		err = 0;
 	}
 
Cheers,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt