* [RFC] find_get_heads_contig
@ 2020-10-21 1:26 Matthew Wilcox
2020-10-21 4:19 ` Kent Overstreet
0 siblings, 1 reply; 2+ messages in thread
From: Matthew Wilcox @ 2020-10-21 1:26 UTC (permalink / raw)
To: linux-mm, linux-fsdevel; +Cc: Kent Overstreet, David Howells
I was going to convert find_get_pages_contig() to only return head pages,
but I want to change the API to take a pagevec like the other find_*
functions have or will have. And it'd be nice if the name of the function
reminded callers that it only returns head pages. So comments on this?
diff --git a/mm/filemap.c b/mm/filemap.c
index 31ba06409dfa..b7dd2523fe79 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2093,6 +2093,58 @@ unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t index,
}
EXPORT_SYMBOL(find_get_pages_contig);
+/**
+ * find_get_heads_contig - Return head pages for a contiguous byte range.
+ * @mapping: The address_space to search.
+ * @start: The starting page index.
+ * @end: The final page index (inclusive).
+ * @pvec: Where the resulting pages are placed.
+ *
+ * find_get_heads_contig() will return a batch of head pages from
+ * @mapping. Pages are returned with an incremented refcount. Only the
+ * head page of a THP is returned. In contrast to find_get_entries(),
+ * pages which are partially outside the range are returned. The head
+ * pages have ascending indices. The indices may not be consecutive,
+ * but the bytes represented by the pages are contiguous. If there is
+ * no page at @start, no pages will be returned.
+ *
+ * Return: The number of pages which were found.
+ */
+unsigned find_get_heads_contig(struct address_space *mapping, pgoff_t start,
+ pgoff_t end, struct pagevec *pvec)
+{
+ XA_STATE(xas, &mapping->i_pages, start);
+ struct page *page;
+
+ rcu_read_lock();
+ for (page = xas_load(&xas); page; page = xas_next(&xas)) {
+ if (xas.xa_index > end)
+ break;
+ if (xas_retry(&xas, page) || xa_is_sibling(page))
+ continue;
+ if (xa_is_value(page))
+ break;
+
+ if (!page_cache_get_speculative(page))
+ goto retry;
+
+ /* Has the page moved or been split? */
+ if (unlikely(page != xas_reload(&xas)))
+ goto put_page;
+
+ if (!pagevec_add(pvec, page))
+ break;
+ continue;
+put_page:
+ put_page(page);
+retry:
+ xas_reset(&xas);
+ }
+ rcu_read_unlock();
+ return pagevec_count(pvec);
+}
+EXPORT_SYMBOL(find_get_heads_contig);
+
/**
* find_get_pages_range_tag - Find and return head pages matching @tag.
* @mapping: the address_space to search
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [RFC] find_get_heads_contig
2020-10-21 1:26 [RFC] find_get_heads_contig Matthew Wilcox
@ 2020-10-21 4:19 ` Kent Overstreet
0 siblings, 0 replies; 2+ messages in thread
From: Kent Overstreet @ 2020-10-21 4:19 UTC (permalink / raw)
To: Matthew Wilcox; +Cc: linux-mm, linux-fsdevel, David Howells
On Wed, Oct 21, 2020 at 02:26:30AM +0100, Matthew Wilcox wrote:
> I was going to convert find_get_pages_contig() to only return head pages,
> but I want to change the API to take a pagevec like the other find_*
> functions have or will have. And it'd be nice if the name of the function
> reminded callers that it only returns head pages. So comments on this?
Perfect, looks like exactly what we need - at some point I can change my
vectorized pagecache stuff to use this.
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
>
> diff --git a/mm/filemap.c b/mm/filemap.c
> index 31ba06409dfa..b7dd2523fe79 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -2093,6 +2093,58 @@ unsigned find_get_pages_contig(struct address_space *mapping, pgoff_t index,
> }
> EXPORT_SYMBOL(find_get_pages_contig);
>
> +/**
> + * find_get_heads_contig - Return head pages for a contiguous byte range.
> + * @mapping: The address_space to search.
> + * @start: The starting page index.
> + * @end: The final page index (inclusive).
> + * @pvec: Where the resulting pages are placed.
> + *
> + * find_get_heads_contig() will return a batch of head pages from
> + * @mapping. Pages are returned with an incremented refcount. Only the
> + * head page of a THP is returned. In contrast to find_get_entries(),
> + * pages which are partially outside the range are returned. The head
> + * pages have ascending indices. The indices may not be consecutive,
> + * but the bytes represented by the pages are contiguous. If there is
> + * no page at @start, no pages will be returned.
> + *
> + * Return: The number of pages which were found.
> + */
> +unsigned find_get_heads_contig(struct address_space *mapping, pgoff_t start,
> + pgoff_t end, struct pagevec *pvec)
> +{
> + XA_STATE(xas, &mapping->i_pages, start);
> + struct page *page;
> +
> + rcu_read_lock();
> + for (page = xas_load(&xas); page; page = xas_next(&xas)) {
> + if (xas.xa_index > end)
> + break;
> + if (xas_retry(&xas, page) || xa_is_sibling(page))
> + continue;
> + if (xa_is_value(page))
> + break;
> +
> + if (!page_cache_get_speculative(page))
> + goto retry;
> +
> + /* Has the page moved or been split? */
> + if (unlikely(page != xas_reload(&xas)))
> + goto put_page;
> +
> + if (!pagevec_add(pvec, page))
> + break;
> + continue;
> +put_page:
> + put_page(page);
> +retry:
> + xas_reset(&xas);
> + }
> + rcu_read_unlock();
> + return pagevec_count(pvec);
> +}
> +EXPORT_SYMBOL(find_get_heads_contig);
> +
> /**
> * find_get_pages_range_tag - Find and return head pages matching @tag.
> * @mapping: the address_space to search
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2020-10-21 4:19 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-10-21 1:26 [RFC] find_get_heads_contig Matthew Wilcox
2020-10-21 4:19 ` Kent Overstreet
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).