Re: [PATCH v3 12/14] mm, swap: cap xswap growth at nr_clusters

[Chris Li]

On Wed, Sep 16, 2026 at 12:21 AM Baoquan He <hebaoquan@xxxxxxxxxx> wrote:
>
> Add si->nr_clusters as the ceiling for the cluster_info array's growth.
> It starts at the whole address space, so nothing changes yet.
>
> Signed-off-by: Baoquan He <hebaoquan@xxxxxxxxxx>
> ---
>  include/linux/swap.h |  1 +
>  mm/swapfile.c        | 19 +++++++++++--------
>  2 files changed, 12 insertions(+), 8 deletions(-)
>
> diff --git a/include/linux/swap.h b/include/linux/swap.h
> index 30642bb481df..9fe82d0f1740 100644
> --- a/include/linux/swap.h
> +++ b/include/linux/swap.h
> @@ -245,6 +245,7 @@ struct swap_info_struct {
>  #ifdef CONFIG_XSWAP
>         struct vm_struct        *cluster_vm;    /* VM_SPARSE area for cluster_info */
>         unsigned long           nr_clusters_max;/* total clusters in the xswap address space */
> +       unsigned long           nr_clusters;    /* how far the array may grow */

Do you mean the current usable size of the cluster array?

>         unsigned long           nr_clusters_mapped; /* currently mapped cluster count */
>         struct work_struct      xswap_shrink_work; /* deferred shrink trigger */
>         struct mutex            xswap_lock;     /* serialize map/unmap operations */
> diff --git a/mm/swapfile.c b/mm/swapfile.c
> index cac084bc72d6..2cf6ba0bd0c0 100644
> --- a/mm/swapfile.c
> +++ b/mm/swapfile.c
> @@ -1319,16 +1319,18 @@ static unsigned long cluster_alloc_swap_entry(struct swap_info_struct *si,
>
>  #ifdef CONFIG_XSWAP
>         /* For xswap: grow the cluster_info array, then retry. */
> -       if (!found && (si->flags & SWP_XSWAP) &&
> -           READ_ONCE(si->nr_clusters_mapped) < READ_ONCE(si->nr_clusters_max) &&
> -           list_empty(&si->free_clusters)) {
> -               unsigned long nr_new = min(READ_ONCE(si->nr_clusters_max) -
> -                                         READ_ONCE(si->nr_clusters_mapped),
> -                                         XSWAP_GROW_CLUSTERS);
> -               unsigned long start = READ_ONCE(si->nr_clusters_mapped);
> -               unsigned long i;
> +       if (!found && (si->flags & SWP_XSWAP) && list_empty(&si->free_clusters)) {
> +               unsigned long ceiling = READ_ONCE(si->nr_clusters);
> +               unsigned long mapped = READ_ONCE(si->nr_clusters_mapped);

Can you help me understand what this change is doing here?
The patch description does not help me. What is the difference between
"ceiling" and "max"?

Chris


> +               unsigned long nr_new, start, i;
>                 int ret;
>
> +               if (mapped >= ceiling)
> +                       goto done;
> +
> +               nr_new = min(ceiling - mapped, XSWAP_GROW_CLUSTERS);
> +               start = mapped;
> +
>                 /*
>                  * Mapping pages can sleep.  The lock only guards the per-cpu
>                  * cluster cache, which this path does not touch.
> @@ -4294,6 +4296,7 @@ static int setup_swap_clusters_info(struct swap_info_struct *si,
>                 cluster_info = vm->addr;
>                 si->cluster_vm = vm;
>                 si->nr_clusters_max = nr_clusters;
> +               si->nr_clusters = nr_clusters;
>                 si->cluster_info = cluster_info;
>
>                 /* Must be initialized before xswap_map_clusters() locks it. */
> --
> 2.54.0
>