Re: [PATCH] xen/mm: avoid watchdog timeout in dump_numa() on large domains

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 02/06/2026 9:49 am, Bernhard Kaindl wrote:
> diff --git a/xen/arch/x86/mm/mem_sharing.c b/xen/arch/x86/mm/mem_sharing.c
> index 5c7a0ff30e8b..89ae418be0ae 100644
> --- a/xen/arch/x86/mm/mem_sharing.c
> +++ b/xen/arch/x86/mm/mem_sharing.c
> @@ -769,7 +769,7 @@ static int page_make_private(struct domain *d, struct 
> page_info *page)
>      ASSERT(page_get_owner(page) == dom_cow);
>      page_set_owner(page, d);
>  
> -    if ( domain_adjust_tot_pages(d, 1) == 1 )
> +    if ( domain_adjust_tot_pages(d, page_to_nid(page), 1) == 1 )
>          get_knownalive_domain(d);
>      page_list_add_tail(page, &d->page_list);
>      nrspin_unlock(&d->page_alloc_lock);
> diff --git a/xen/common/domain.c b/xen/common/domain.c
> index 8cb4241b0511..0b6afba2acdb 100644
> --- a/xen/common/domain.c
> +++ b/xen/common/domain.c
> @@ -1558,6 +1558,15 @@ void domain_destroy(struct domain *d)
>      /* Remove from the domlist/hash. */
>      domlist_remove(d);
>  
> +    /*
> +     * Final invariant check: all pages still owned by the dying domain must
> +     * be accounted for per-node before complete_domain_destroy() reclaims
> +     * them.  Debug-only; expands to a no-op in production builds.
> +     */
> +    nrspin_lock(&d->page_alloc_lock);
> +    assert_numa_page_count(d);
> +    nrspin_unlock(&d->page_alloc_lock);
> +
>      /* Schedule RCU asynchronous completion of domain destroy. */
>      call_rcu(&d->rcu, complete_domain_destroy);
>  }

Why have you re-merged these two patches?   I explicitly split them in
the XenServer patchqueue, and they should remain separate.

~Andrew