io_uring/kbuf: Use slab for struct io_buffer objects

The allocation of struct io_buffer for metadata of provided buffers is
done through a custom allocator that directly gets pages and
fragments them.  But, slab would do just fine, as this is not a hot path
(in fact, it is a deprecated feature) and, by keeping a custom allocator
implementation we lose benefits like tracking, poisoning,
sanitizers. Finally, the custom code is more complex and requires
keeping the list of pages in struct ctx for no good reason.  This patch
cleans this path up and just uses slab.

I microbenchmarked it by forcing the allocation of a large number of
objects with the least number of io_uring commands possible (keeping
nbufs=USHRT_MAX), with and without the patch.  There is a slight
increase in time spent in the allocation with slab, of course, but even
when allocating to system resources exhaustion, which is not very
realistic and happened around 1/2 billion provided buffers for me, it
wasn't a significant hit in system time.  Specially if we think of a
real-world scenario, an application doing register/unregister of
provided buffers will hit ctx->io_buffers_cache more often than actually
going to slab.

Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
Link: https://lore.kernel.org/r/20231005000531.30800-4-krisman@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 26 insertions, 21 deletions

diff --git a/io_uring/kbuf.c b/io_uring/kbuf.c
index 12a357348733..d5a04467666f 100644
--- a/io_uring/kbuf.c
+++ b/io_uring/kbuf.c
@@ -22,6 +22,8 @@
 /* BIDs are addressed by a 16-bit field in a CQE */
 #define MAX_BIDS_PER_BGID (1 << 16)
 
+struct kmem_cache *io_buf_cachep;
+
 struct io_provide_buf {
 	struct file			*file;
 	__u64				addr;
@@ -258,6 +260,8 @@ static int __io_remove_buffers(struct io_ring_ctx *ctx,
 void io_destroy_buffers(struct io_ring_ctx *ctx)
 {
 	struct io_buffer_list *bl;
+	struct list_head *item, *tmp;
+	struct io_buffer *buf;
 	unsigned long index;
 	int i;
 
@@ -273,12 +277,9 @@ void io_destroy_buffers(struct io_ring_ctx *ctx)
 		kfree(bl);
 	}
 
-	while (!list_empty(&ctx->io_buffers_pages)) {
-		struct page *page;
-
-		page = list_first_entry(&ctx->io_buffers_pages, struct page, lru);
-		list_del_init(&page->lru);
-		__free_page(page);
+	list_for_each_safe(item, tmp, &ctx->io_buffers_cache) {
+		buf = list_entry(item, struct io_buffer, list);
+		kmem_cache_free(io_buf_cachep, buf);
 	}
 }
 
@@ -361,11 +362,12 @@ int io_provide_buffers_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe
 	return 0;
 }
 
+#define IO_BUFFER_ALLOC_BATCH 64
+
 static int io_refill_buffer_cache(struct io_ring_ctx *ctx)
 {
-	struct io_buffer *buf;
-	struct page *page;
-	int bufs_in_page;
+	struct io_buffer *bufs[IO_BUFFER_ALLOC_BATCH];
+	int allocated;
 
 	/*
 	 * Completions that don't happen inline (eg not under uring_lock) will
@@ -385,22 +387,25 @@ static int io_refill_buffer_cache(struct io_ring_ctx *ctx)
 
 	/*
 	 * No free buffers and no completion entries either. Allocate a new
-	 * page worth of buffer entries and add those to our freelist.
+	 * batch of buffer entries and add those to our freelist.
 	 */
-	page = alloc_page(GFP_KERNEL_ACCOUNT);
-	if (!page)
-		return -ENOMEM;
 
-	list_add(&page->lru, &ctx->io_buffers_pages);
-
-	buf = page_address(page);
-	bufs_in_page = PAGE_SIZE / sizeof(*buf);
-	while (bufs_in_page) {
-		list_add_tail(&buf->list, &ctx->io_buffers_cache);
-		buf++;
-		bufs_in_page--;
+	allocated = kmem_cache_alloc_bulk(io_buf_cachep, GFP_KERNEL_ACCOUNT,
+					  ARRAY_SIZE(bufs), (void **) bufs);
+	if (unlikely(!allocated)) {
+		/*
+		 * Bulk alloc is all-or-nothing. If we fail to get a batch,
+		 * retry single alloc to be on the safe side.
+		 */
+		bufs[0] = kmem_cache_alloc(io_buf_cachep, GFP_KERNEL);
+		if (!bufs[0])
+			return -ENOMEM;
+		allocated = 1;
 	}
 
+	while (allocated)
+		list_add_tail(&bufs[--allocated]->list, &ctx->io_buffers_cache);
+
 	return 0;
 }