Merge branch 'io_uring-zerocopy-send' of git://git.kernel.org/pub/scm/linux/kernel/git/kuba/linux

Pavel Begunkov says:

====================
io_uring zerocopy send

The patchset implements io_uring zerocopy send. It works with both registered
and normal buffers, mixing is allowed but not recommended. Apart from usual
request completions, just as with MSG_ZEROCOPY, io_uring separately notifies
the userspace when buffers are freed and can be reused (see API design below),
which is delivered into io_uring's Completion Queue. Those "buffer-free"
notifications are not necessarily per request, but the userspace has control
over it and should explicitly attaching a number of requests to a single
notification. The series also adds some internal optimisations when used with
registered buffers like removing page referencing.

From the kernel networking perspective there are two main changes. The first
one is passing ubuf_info into the network layer from io_uring (inside of an
in kernel struct msghdr). This allows extra optimisations, e.g. ubuf_info
caching on the io_uring side, but also helps to avoid cross-referencing
and synchronisation problems. The second part is an optional optimisation
removing page referencing for requests with registered buffers.

Benchmarking UDP with an optimised version of the selftest (see [1]), which
sends a bunch of requests, waits for completions and repeats. "+ flush" column
posts one additional "buffer-free" notification per request, and just "zc"
doesn't post buffer notifications at all.

NIC (requests / second):
IO size | non-zc    | zc             | zc + flush
4000    | 495134    | 606420 (+22%)  | 558971 (+12%)
1500    | 551808    | 577116 (+4.5%) | 565803 (+2.5%)
1000    | 584677    | 592088 (+1.2%) | 560885 (-4%)
600     | 596292    | 598550 (+0.4%) | 555366 (-6.7%)

dummy (requests / second):
IO size | non-zc    | zc             | zc + flush
8000    | 1299916   | 2396600 (+84%) | 2224219 (+71%)
4000    | 1869230   | 2344146 (+25%) | 2170069 (+16%)
1200    | 2071617   | 2361960 (+14%) | 2203052 (+6%)
600     | 2106794   | 2381527 (+13%) | 2195295 (+4%)

Previously it also brought a massive performance speedup compared to the
msg_zerocopy tool (see [3]), which is probably not super interesting. There
is also an additional bunch of refcounting optimisations that was omitted from
the series for simplicity and as they don't change the picture drastically,
they will be sent as follow up, as well as flushing optimisations closing the
performance gap b/w two last columns.

For TCP on localhost (with hacks enabling localhost zerocopy) and including
additional overhead for receive:

IO size | non-zc    | zc
1200    | 4174      | 4148
4096    | 7597      | 11228

Using a real NIC 1200 bytes, zc is worse than non-zc ~5-10%, maybe the
omitted optimisations will somewhat help, should look better for 4000,
but couldn't test properly because of setup problems.

Links:

  liburing (benchmark + tests):
  [1] https://github.com/isilence/liburing/tree/zc_v4

  kernel repo:
  [2] https://github.com/isilence/linux/tree/zc_v4

  RFC v1:
  [3] https://lore.kernel.org/io-uring/cover.1638282789.git.asml.silence@gmail.com/

  RFC v2:
  https://lore.kernel.org/io-uring/cover.1640029579.git.asml.silence@gmail.com/

  Net patches based:
  git@github.com:isilence/linux.git zc_v4-net-base
  or
  https://github.com/isilence/linux/tree/zc_v4-net-base

API design overview:

  The series introduces an io_uring concept of notifactors. From the userspace
  perspective it's an entity to which it can bind one or more requests and then
  requesting to flush it. Flushing a notifier makes it impossible to attach new
  requests to it, and instructs the notifier to post a completion once all
  requests attached to it are completed and the kernel doesn't need the buffers
  anymore.

  Notifications are stored in notification slots, which should be registered as
  an array in io_uring. Each slot stores only one notifier at any particular
  moment. Flushing removes it from the slot and the slot automatically replaces
  it with a new notifier. All operations with notifiers are done by specifying
  an index of a slot it's currently in.

  When registering a notification the userspace specifies a u64 tag for each
  slot, which will be copied in notification completion entries as
  cqe::user_data. cqe::res is 0 and cqe::flags is equal to wrap around u32
  sequence number counting notifiers of a slot.

====================

Link: https://lore.kernel.org/r/cover.1657643355.git.asml.silence@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

2 files changed, 43 insertions, 8 deletions

diff --git a/net/core/datagram.c b/net/core/datagram.c
index 35791f86bd1a..8c702904d960 100644
--- a/net/core/datagram.c
+++ b/net/core/datagram.c
@@ -610,10 +610,16 @@ fault:
 }
 EXPORT_SYMBOL(skb_copy_datagram_from_iter);
 
-int __zerocopy_sg_from_iter(struct sock *sk, struct sk_buff *skb,
-			    struct iov_iter *from, size_t length)
+int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk,
+			    struct sk_buff *skb, struct iov_iter *from,
+			    size_t length)
 {
-	int frag = skb_shinfo(skb)->nr_frags;
+	int frag;
+
+	if (msg && msg->sg_from_iter)
+		return msg->sg_from_iter(sk, skb, from, length);
+
+	frag = skb_shinfo(skb)->nr_frags;
 
 	while (length && iov_iter_count(from)) {
 		struct page *pages[MAX_SKB_FRAGS];
@@ -699,7 +705,7 @@ int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *from)
 	if (skb_copy_datagram_from_iter(skb, 0, from, copy))
 		return -EFAULT;
 
-	return __zerocopy_sg_from_iter(NULL, skb, from, ~0U);
+	return __zerocopy_sg_from_iter(NULL, NULL, skb, from, ~0U);
 }
 EXPORT_SYMBOL(zerocopy_sg_from_iter);
 
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index c4a751781581..974bbbbe7138 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -669,11 +669,18 @@ static void skb_release_data(struct sk_buff *skb)
 			      &shinfo->dataref))
 		goto exit;
 
-	skb_zcopy_clear(skb, true);
+	if (skb_zcopy(skb)) {
+		bool skip_unref = shinfo->flags & SKBFL_MANAGED_FRAG_REFS;
+
+		skb_zcopy_clear(skb, true);
+		if (skip_unref)
+			goto free_head;
+	}
 
 	for (i = 0; i < shinfo->nr_frags; i++)
 		__skb_frag_unref(&shinfo->frags[i], skb->pp_recycle);
 
+free_head:
 	if (shinfo->frag_list)
 		kfree_skb_list(shinfo->frag_list);
 
@@ -898,7 +905,10 @@ EXPORT_SYMBOL(skb_dump);
  */
 void skb_tx_error(struct sk_buff *skb)
 {
-	skb_zcopy_clear(skb, true);
+	if (skb) {
+		skb_zcopy_downgrade_managed(skb);
+		skb_zcopy_clear(skb, true);
+	}
 }
 EXPORT_SYMBOL(skb_tx_error);
 
@@ -1196,7 +1206,7 @@ static struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size)
 	uarg->len = 1;
 	uarg->bytelen = size;
 	uarg->zerocopy = 1;
-	uarg->flags = SKBFL_ZEROCOPY_FRAG;
+	uarg->flags = SKBFL_ZEROCOPY_FRAG | SKBFL_DONT_ORPHAN;
 	refcount_set(&uarg->refcnt, 1);
 	sock_hold(sk);
 
@@ -1215,6 +1225,10 @@ struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
 		const u32 byte_limit = 1 << 19;		/* limit to a few TSO */
 		u32 bytelen, next;
 
+		/* there might be non MSG_ZEROCOPY users */
+		if (uarg->callback != msg_zerocopy_callback)
+			return NULL;
+
 		/* realloc only when socket is locked (TCP, UDP cork),
 		 * so uarg->len and sk_zckey access is serialized
 		 */
@@ -1357,7 +1371,7 @@ int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
 	if (orig_uarg && uarg != orig_uarg)
 		return -EEXIST;
 
-	err = __zerocopy_sg_from_iter(sk, skb, &msg->msg_iter, len);
+	err = __zerocopy_sg_from_iter(msg, sk, skb, &msg->msg_iter, len);
 	if (err == -EFAULT || (err == -EMSGSIZE && skb->len == orig_len)) {
 		struct sock *save_sk = skb->sk;
 
@@ -1374,6 +1388,16 @@ int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
 }
 EXPORT_SYMBOL_GPL(skb_zerocopy_iter_stream);
 
+void __skb_zcopy_downgrade_managed(struct sk_buff *skb)
+{
+	int i;
+
+	skb_shinfo(skb)->flags &= ~SKBFL_MANAGED_FRAG_REFS;
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
+		skb_frag_ref(skb, i);
+}
+EXPORT_SYMBOL_GPL(__skb_zcopy_downgrade_managed);
+
 static int skb_zerocopy_clone(struct sk_buff *nskb, struct sk_buff *orig,
 			      gfp_t gfp_mask)
 {
@@ -1692,6 +1716,8 @@ int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail,
 
 	BUG_ON(skb_shared(skb));
 
+	skb_zcopy_downgrade_managed(skb);
+
 	size = SKB_DATA_ALIGN(size);
 
 	if (skb_pfmemalloc(skb))
@@ -3486,6 +3512,8 @@ void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len)
 	int pos = skb_headlen(skb);
 	const int zc_flags = SKBFL_SHARED_FRAG | SKBFL_PURE_ZEROCOPY;
 
+	skb_zcopy_downgrade_managed(skb);
+
 	skb_shinfo(skb1)->flags |= skb_shinfo(skb)->flags & zc_flags;
 	skb_zerocopy_clone(skb1, skb, 0);
 	if (len < pos)	/* Split line is inside header. */
@@ -3834,6 +3862,7 @@ int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
 	if (skb_can_coalesce(skb, i, page, offset)) {
 		skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], size);
 	} else if (i < MAX_SKB_FRAGS) {
+		skb_zcopy_downgrade_managed(skb);
 		get_page(page);
 		skb_fill_page_desc(skb, i, page, offset, size);
 	} else {