Merge branch 'shrink-struct-ubuf_info'

Pavel Begunkov says:

====================
shrink struct ubuf_info

struct ubuf_info is large but not all fields are needed for all
cases. We have limited space in io_uring for it and large ubuf_info
prevents some struct embedding, even though we use only a subset
of the fields. It's also not very clean trying to use this typeless
extra space.

Shrink struct ubuf_info to only necessary fields used in generic paths,
namely ->callback, ->refcnt and ->flags, which take only 16 bytes. And
make MSG_ZEROCOPY and some other users to embed it into a larger struct
ubuf_info_msgzc mimicking the former ubuf_info.

Note, xen/vhost may also have some cleaning on top by creating
new structs containing ubuf_info but with proper types.
====================

Link: https://lore.kernel.org/r/cover.1663892211.git.asml.silence@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

9 files changed, 48 insertions, 35 deletions

diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-netback/common.h
index 8174d7b2966c..1545cbee77a4 100644
--- a/drivers/net/xen-netback/common.h
+++ b/drivers/net/xen-netback/common.h
@@ -62,7 +62,7 @@ struct pending_tx_info {
 	 * ubuf_to_vif is a helper which finds the struct xenvif from a pointer
 	 * to this field.
 	 */
-	struct ubuf_info callback_struct;
+	struct ubuf_info_msgzc callback_struct;
 };
 
 #define XEN_NETIF_TX_RING_SIZE __CONST_RING_SIZE(xen_netif_tx, XEN_PAGE_SIZE)
diff --git a/drivers/net/xen-netback/interface.c b/drivers/net/xen-netback/interface.c
index fb32ae82d9b0..e579ecd40b74 100644
--- a/drivers/net/xen-netback/interface.c
+++ b/drivers/net/xen-netback/interface.c
@@ -591,8 +591,8 @@ int xenvif_init_queue(struct xenvif_queue *queue)
 	}
 
 	for (i = 0; i < MAX_PENDING_REQS; i++) {
-		queue->pending_tx_info[i].callback_struct = (struct ubuf_info)
-			{ .callback = xenvif_zerocopy_callback,
+		queue->pending_tx_info[i].callback_struct = (struct ubuf_info_msgzc)
+			{ { .callback = xenvif_zerocopy_callback },
 			  { { .ctx = NULL,
 			      .desc = i } } };
 		queue->grant_tx_handle[i] = NETBACK_INVALID_HANDLE;
diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c
index a256695fc89e..3d2081bbbc86 100644
--- a/drivers/net/xen-netback/netback.c
+++ b/drivers/net/xen-netback/netback.c
@@ -133,7 +133,7 @@ static inline unsigned long idx_to_kaddr(struct xenvif_queue *queue,
 
 /* Find the containing VIF's structure from a pointer in pending_tx_info array
  */
-static inline struct xenvif_queue *ubuf_to_queue(const struct ubuf_info *ubuf)
+static inline struct xenvif_queue *ubuf_to_queue(const struct ubuf_info_msgzc *ubuf)
 {
 	u16 pending_idx = ubuf->desc;
 	struct pending_tx_info *temp =
@@ -1228,11 +1228,12 @@ static int xenvif_tx_submit(struct xenvif_queue *queue)
 	return work_done;
 }
 
-void xenvif_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *ubuf,
+void xenvif_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *ubuf_base,
 			      bool zerocopy_success)
 {
 	unsigned long flags;
 	pending_ring_idx_t index;
+	struct ubuf_info_msgzc *ubuf = uarg_to_msgzc(ubuf_base);
 	struct xenvif_queue *queue = ubuf_to_queue(ubuf);
 
 	/* This is the only place where we grab this lock, to protect callbacks
@@ -1241,7 +1242,7 @@ void xenvif_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *ubuf,
 	spin_lock_irqsave(&queue->callback_lock, flags);
 	do {
 		u16 pending_idx = ubuf->desc;
-		ubuf = (struct ubuf_info *) ubuf->ctx;
+		ubuf = (struct ubuf_info_msgzc *) ubuf->ctx;
 		BUG_ON(queue->dealloc_prod - queue->dealloc_cons >=
 			MAX_PENDING_REQS);
 		index = pending_index(queue->dealloc_prod);
diff --git a/drivers/vhost/net.c b/drivers/vhost/net.c
index 68e4ecd1cc0e..d7a04d573988 100644
--- a/drivers/vhost/net.c
+++ b/drivers/vhost/net.c
@@ -118,7 +118,7 @@ struct vhost_net_virtqueue {
 	/* Number of XDP frames batched */
 	int batched_xdp;
 	/* an array of userspace buffers info */
-	struct ubuf_info *ubuf_info;
+	struct ubuf_info_msgzc *ubuf_info;
 	/* Reference counting for outstanding ubufs.
 	 * Protected by vq mutex. Writers must also take device mutex. */
 	struct vhost_net_ubuf_ref *ubufs;
@@ -382,8 +382,9 @@ static void vhost_zerocopy_signal_used(struct vhost_net *net,
 }
 
 static void vhost_zerocopy_callback(struct sk_buff *skb,
-				    struct ubuf_info *ubuf, bool success)
+				    struct ubuf_info *ubuf_base, bool success)
 {
+	struct ubuf_info_msgzc *ubuf = uarg_to_msgzc(ubuf_base);
 	struct vhost_net_ubuf_ref *ubufs = ubuf->ctx;
 	struct vhost_virtqueue *vq = ubufs->vq;
 	int cnt;
@@ -871,7 +872,7 @@ static void handle_tx_zerocopy(struct vhost_net *net, struct socket *sock)
 	size_t len, total_len = 0;
 	int err;
 	struct vhost_net_ubuf_ref *ubufs;
-	struct ubuf_info *ubuf;
+	struct ubuf_info_msgzc *ubuf;
 	bool zcopy_used;
 	int sent_pkts = 0;
 
@@ -907,14 +908,14 @@ static void handle_tx_zerocopy(struct vhost_net *net, struct socket *sock)
 			ubuf = nvq->ubuf_info + nvq->upend_idx;
 			vq->heads[nvq->upend_idx].id = cpu_to_vhost32(vq, head);
 			vq->heads[nvq->upend_idx].len = VHOST_DMA_IN_PROGRESS;
-			ubuf->callback = vhost_zerocopy_callback;
 			ubuf->ctx = nvq->ubufs;
 			ubuf->desc = nvq->upend_idx;
-			ubuf->flags = SKBFL_ZEROCOPY_FRAG;
-			refcount_set(&ubuf->refcnt, 1);
+			ubuf->ubuf.callback = vhost_zerocopy_callback;
+			ubuf->ubuf.flags = SKBFL_ZEROCOPY_FRAG;
+			refcount_set(&ubuf->ubuf.refcnt, 1);
 			msg.msg_control = &ctl;
 			ctl.type = TUN_MSG_UBUF;
-			ctl.ptr = ubuf;
+			ctl.ptr = &ubuf->ubuf;
 			msg.msg_controllen = sizeof(ctl);
 			ubufs = nvq->ubufs;
 			atomic_inc(&ubufs->refcount);
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index f15d5b62539b..920eb6413fee 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -533,6 +533,13 @@ enum {
 struct ubuf_info {
 	void (*callback)(struct sk_buff *, struct ubuf_info *,
 			 bool zerocopy_success);
+	refcount_t refcnt;
+	u8 flags;
+};
+
+struct ubuf_info_msgzc {
+	struct ubuf_info ubuf;
+
 	union {
 		struct {
 			unsigned long desc;
@@ -545,8 +552,6 @@ struct ubuf_info {
 			u32 bytelen;
 		};
 	};
-	refcount_t refcnt;
-	u8 flags;
 
 	struct mmpin {
 		struct user_struct *user;
@@ -555,6 +560,8 @@ struct ubuf_info {
 };
 
 #define skb_uarg(SKB)	((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
+#define uarg_to_msgzc(ubuf_ptr)	container_of((ubuf_ptr), struct ubuf_info_msgzc, \
+					     ubuf)
 
 int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
 void mm_unaccount_pinned_pages(struct mmpin *mmp);
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index f1b8b20fc20b..bbcfb1c7f59e 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -1188,7 +1188,7 @@ EXPORT_SYMBOL_GPL(mm_unaccount_pinned_pages);
 
 static struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size)
 {
-	struct ubuf_info *uarg;
+	struct ubuf_info_msgzc *uarg;
 	struct sk_buff *skb;
 
 	WARN_ON_ONCE(!in_task());
@@ -1206,19 +1206,19 @@ static struct ubuf_info *msg_zerocopy_alloc(struct sock *sk, size_t size)
 		return NULL;
 	}
 
-	uarg->callback = msg_zerocopy_callback;
+	uarg->ubuf.callback = msg_zerocopy_callback;
 	uarg->id = ((u32)atomic_inc_return(&sk->sk_zckey)) - 1;
 	uarg->len = 1;
 	uarg->bytelen = size;
 	uarg->zerocopy = 1;
-	uarg->flags = SKBFL_ZEROCOPY_FRAG | SKBFL_DONT_ORPHAN;
-	refcount_set(&uarg->refcnt, 1);
+	uarg->ubuf.flags = SKBFL_ZEROCOPY_FRAG | SKBFL_DONT_ORPHAN;
+	refcount_set(&uarg->ubuf.refcnt, 1);
 	sock_hold(sk);
 
-	return uarg;
+	return &uarg->ubuf;
 }
 
-static inline struct sk_buff *skb_from_uarg(struct ubuf_info *uarg)
+static inline struct sk_buff *skb_from_uarg(struct ubuf_info_msgzc *uarg)
 {
 	return container_of((void *)uarg, struct sk_buff, cb);
 }
@@ -1227,6 +1227,7 @@ struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
 				       struct ubuf_info *uarg)
 {
 	if (uarg) {
+		struct ubuf_info_msgzc *uarg_zc;
 		const u32 byte_limit = 1 << 19;		/* limit to a few TSO */
 		u32 bytelen, next;
 
@@ -1242,8 +1243,9 @@ struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
 			return NULL;
 		}
 
-		bytelen = uarg->bytelen + size;
-		if (uarg->len == USHRT_MAX - 1 || bytelen > byte_limit) {
+		uarg_zc = uarg_to_msgzc(uarg);
+		bytelen = uarg_zc->bytelen + size;
+		if (uarg_zc->len == USHRT_MAX - 1 || bytelen > byte_limit) {
 			/* TCP can create new skb to attach new uarg */
 			if (sk->sk_type == SOCK_STREAM)
 				goto new_alloc;
@@ -1251,11 +1253,11 @@ struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size,
 		}
 
 		next = (u32)atomic_read(&sk->sk_zckey);
-		if ((u32)(uarg->id + uarg->len) == next) {
-			if (mm_account_pinned_pages(&uarg->mmp, size))
+		if ((u32)(uarg_zc->id + uarg_zc->len) == next) {
+			if (mm_account_pinned_pages(&uarg_zc->mmp, size))
 				return NULL;
-			uarg->len++;
-			uarg->bytelen = bytelen;
+			uarg_zc->len++;
+			uarg_zc->bytelen = bytelen;
 			atomic_set(&sk->sk_zckey, ++next);
 
 			/* no extra ref when appending to datagram (MSG_MORE) */
@@ -1291,7 +1293,7 @@ static bool skb_zerocopy_notify_extend(struct sk_buff *skb, u32 lo, u16 len)
 	return true;
 }
 
-static void __msg_zerocopy_callback(struct ubuf_info *uarg)
+static void __msg_zerocopy_callback(struct ubuf_info_msgzc *uarg)
 {
 	struct sk_buff *tail, *skb = skb_from_uarg(uarg);
 	struct sock_exterr_skb *serr;
@@ -1344,19 +1346,21 @@ release:
 void msg_zerocopy_callback(struct sk_buff *skb, struct ubuf_info *uarg,
 			   bool success)
 {
-	uarg->zerocopy = uarg->zerocopy & success;
+	struct ubuf_info_msgzc *uarg_zc = uarg_to_msgzc(uarg);
+
+	uarg_zc->zerocopy = uarg_zc->zerocopy & success;
 
 	if (refcount_dec_and_test(&uarg->refcnt))
-		__msg_zerocopy_callback(uarg);
+		__msg_zerocopy_callback(uarg_zc);
 }
 EXPORT_SYMBOL_GPL(msg_zerocopy_callback);
 
 void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref)
 {
-	struct sock *sk = skb_from_uarg(uarg)->sk;
+	struct sock *sk = skb_from_uarg(uarg_to_msgzc(uarg))->sk;
 
 	atomic_dec(&sk->sk_zckey);
-	uarg->len--;
+	uarg_to_msgzc(uarg)->len--;
 
 	if (have_uref)
 		msg_zerocopy_callback(NULL, uarg, true);
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index 8201cd423ff9..1ae83ad629b2 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -1043,7 +1043,7 @@ static int __ip_append_data(struct sock *sk,
 				paged = true;
 				zc = true;
 			} else {
-				uarg->zerocopy = 0;
+				uarg_to_msgzc(uarg)->zerocopy = 0;
 				skb_zcopy_set(skb, uarg, &extra_uref);
 			}
 		}
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 5702ca9b952d..a109a4cb1e47 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -1239,7 +1239,7 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
 			}
 			zc = sk->sk_route_caps & NETIF_F_SG;
 			if (!zc)
-				uarg->zerocopy = 0;
+				uarg_to_msgzc(uarg)->zerocopy = 0;
 		}
 	}
 
diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
index a60176e913a8..e19507614f64 100644
--- a/net/ipv6/ip6_output.c
+++ b/net/ipv6/ip6_output.c
@@ -1567,7 +1567,7 @@ emsgsize:
 				paged = true;
 				zc = true;
 			} else {
-				uarg->zerocopy = 0;
+				uarg_to_msgzc(uarg)->zerocopy = 0;
 				skb_zcopy_set(skb, uarg, &extra_uref);
 			}
 		}