Re: [PATCH 1/2] xen: add helpers for caching grant mapping pages

From: boris . ostrovsky
Date: Mon Dec 07 2020 - 18:53:55 EST



On 12/7/20 8:30 AM, Juergen Gross wrote:
> Instead of having similar helpers in multiple backend drivers use
> common helpers for caching pages allocated via gnttab_alloc_pages().
>
> Make use of those helpers in blkback and scsiback.
>
> Signed-off-by: Juergen Gross <jgross@xxxxxxxx>


Reviewed-by: Boris Ostrovsky <boris.ostrovksy@xxxxxxxxxx>


> +
> +void gnttab_page_cache_shrink(struct gnttab_page_cache *cache, unsigned int num)
> +{
> + struct page *page[10];
> + unsigned int i = 0;
> + unsigned long flags;
> +
> + spin_lock_irqsave(&cache->lock, flags);
> +
> + while (cache->num_pages > num) {
> + page[i] = list_first_entry(&cache->pages, struct page, lru);
> + list_del(&page[i]->lru);
> + cache->num_pages--;
> + if (++i == ARRAY_SIZE(page)) {
> + spin_unlock_irqrestore(&cache->lock, flags);
> + gnttab_free_pages(i, page);
> + i = 0;
> + spin_lock_irqsave(&cache->lock, flags);
> + }
> + }
> +
> + spin_unlock_irqrestore(&cache->lock, flags);
> +
> + if (i != 0)
> + gnttab_free_pages(i, page);
> +}


How about splitting cache->pages list into two lists (one @num long and the other holding the rest) and then batching gntab_free_pages() from that first list? Then you won't have to deal with locking.


In fact, I am not even sure batching gains us much, I'd consider going one-by-one.


-boris