Re: [PATCH 14/19] mm: Create/affine kswapd to its preferred node

From: Vlastimil Babka
Date: Wed Aug 07 2024 - 13:06:20 EST


On 8/7/24 18:02, Frederic Weisbecker wrote:
> kswapd is dedicated to a specific node. As such it wants to be
> preferrably affine to it, memory and CPUs-wise.
>
> Use the proper kthread API to achieve that. As a bonus it takes care of
> CPU-hotplug events and CPU-isolation on its behalf.
>
> Signed-off-by: Frederic Weisbecker <frederic@xxxxxxxxxx>

Acked-by: Vlastimil Babka <vbabka@xxxxxxx>

But there's a bug to fix:

> ---
> mm/vmscan.c | 7 ++-----
> 1 file changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index cfa839284b92..b8ee3bf8fa29 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -7159,10 +7159,6 @@ static int kswapd(void *p)
> unsigned int highest_zoneidx = MAX_NR_ZONES - 1;
> pg_data_t *pgdat = (pg_data_t *)p;
> struct task_struct *tsk = current;
> - const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
> -
> - if (!cpumask_empty(cpumask))
> - set_cpus_allowed_ptr(tsk, cpumask);
>
> /*
> * Tell the memory management that we're a "memory allocator",
> @@ -7331,7 +7327,7 @@ void __meminit kswapd_run(int nid)
>
> pgdat_kswapd_lock(pgdat);
> if (!pgdat->kswapd) {
> - pgdat->kswapd = kthread_run(kswapd, pgdat, "kswapd%d", nid);
> + pgdat->kswapd = kthread_create_on_node(kswapd, pgdat, nid, "kswapd%d", nid);
> if (IS_ERR(pgdat->kswapd)) {
> /* failure at boot is fatal */
> pr_err("Failed to start kswapd on node %d,ret=%ld\n",
> @@ -7339,6 +7335,7 @@ void __meminit kswapd_run(int nid)
> BUG_ON(system_state < SYSTEM_RUNNING);
> pgdat->kswapd = NULL;
> }
> + wake_up_process(pgdat->kswapd);

should be in an else { } branch as it only ends up with BUG() during early
boot and not when hotplugging later

> }
> pgdat_kswapd_unlock(pgdat);
> }