On Tue, Jun 03, 2025 at 08:19:26AM -0600, Theo de Raadt wrote:
> Can the operations in nmbclust_update() be done by sysctl storing a single
> new value, which is observed later in a codepath (which already needs locking
> against subsystems below), as being different from the current variable, and
> then it changes what it needs?
> 
> 

We don't need to serialize `mbuf_mem_limit' and the following
pool_wakeup() calls, they are independent.