xsk: Move fill and completion rings to buffer pool

Move the fill and completion rings from the umem to the buffer
pool. This so that we in a later commit can share the umem
between multiple HW queue ids. In this case, we need one fill and
completion ring per queue id. As the buffer pool is per queue id
and napi id this is a natural place for it and one umem
struture can be shared between these buffer pools.

Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Link: https://lore.kernel.org/bpf/1598603189-32145-5-git-send-email-magnus.karlsson@intel.com

include/net/xdp_sock.h

@@ -18,8 +18,6 @@ struct xsk_queue;
 struct xdp_buff;
 
 struct xdp_umem {
-	struct xsk_queue *fq;
-	struct xsk_queue *cq;
 	u64 size;
 	u32 headroom;
 	u32 chunk_size;
@@ -77,6 +75,8 @@ struct xdp_sock {
 	struct list_head map_list;
 	/* Protects map_list */
 	spinlock_t map_list_lock;
+	struct xsk_queue *fq_tmp; /* Only as tmp storage before bind */
+	struct xsk_queue *cq_tmp; /* Only as tmp storage before bind */
 };
 
 #ifdef CONFIG_XDP_SOCKETS

include/net/xsk_buff_pool.h

@@ -30,6 +30,7 @@ struct xdp_buff_xsk {
 
 struct xsk_buff_pool {
 	struct xsk_queue *fq;
+	struct xsk_queue *cq;
 	struct list_head free_list;
 	dma_addr_t *dma_pages;
 	struct xdp_buff_xsk *heads;
@@ -57,7 +58,6 @@ struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs,
 						struct xdp_umem *umem);
 int xp_assign_dev(struct xsk_buff_pool *pool, struct net_device *dev,
 		  u16 queue_id, u16 flags);
-void xp_set_fq(struct xsk_buff_pool *pool, struct xsk_queue *fq);
 void xp_destroy(struct xsk_buff_pool *pool);
 void xp_release(struct xdp_buff_xsk *xskb);
 void xp_get_pool(struct xsk_buff_pool *pool);

net/xdp/xdp_umem.c

@@ -85,16 +85,6 @@ static void xdp_umem_release(struct xdp_umem *umem)
 
 	ida_simple_remove(&umem_ida, umem->id);
 
-	if (umem->fq) {
-		xskq_destroy(umem->fq);
-		umem->fq = NULL;
-	}
-
-	if (umem->cq) {
-		xskq_destroy(umem->cq);
-		umem->cq = NULL;
-	}
-
 	xdp_umem_unpin_pages(umem);
 
 	xdp_umem_unaccount_pages(umem);
@@ -278,8 +268,3 @@ struct xdp_umem *xdp_umem_create(struct xdp_umem_reg *mr)
 
 	return umem;
 }
-
-bool xdp_umem_validate_queues(struct xdp_umem *umem)
-{
-	return umem->fq && umem->cq;
-}

net/xdp/xsk.c

@@ -36,7 +36,7 @@ static DEFINE_PER_CPU(struct list_head, xskmap_flush_list);
 bool xsk_is_setup_for_bpf_map(struct xdp_sock *xs)
 {
 	return READ_ONCE(xs->rx) &&  READ_ONCE(xs->umem) &&
-		READ_ONCE(xs->umem->fq);
+		(xs->pool->fq || READ_ONCE(xs->fq_tmp));
 }
 
 void xsk_set_rx_need_wakeup(struct xsk_buff_pool *pool)
@@ -46,7 +46,7 @@ void xsk_set_rx_need_wakeup(struct xsk_buff_pool *pool)
 	if (umem->need_wakeup & XDP_WAKEUP_RX)
 		return;
 
-	umem->fq->ring->flags |= XDP_RING_NEED_WAKEUP;
+	pool->fq->ring->flags |= XDP_RING_NEED_WAKEUP;
 	umem->need_wakeup |= XDP_WAKEUP_RX;
 }
 EXPORT_SYMBOL(xsk_set_rx_need_wakeup);
@@ -76,7 +76,7 @@ void xsk_clear_rx_need_wakeup(struct xsk_buff_pool *pool)
 	if (!(umem->need_wakeup & XDP_WAKEUP_RX))
 		return;
 
-	umem->fq->ring->flags &= ~XDP_RING_NEED_WAKEUP;
+	pool->fq->ring->flags &= ~XDP_RING_NEED_WAKEUP;
 	umem->need_wakeup &= ~XDP_WAKEUP_RX;
 }
 EXPORT_SYMBOL(xsk_clear_rx_need_wakeup);
@@ -254,7 +254,7 @@ static int xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp,
 static void xsk_flush(struct xdp_sock *xs)
 {
 	xskq_prod_submit(xs->rx);
-	__xskq_cons_release(xs->umem->fq);
+	__xskq_cons_release(xs->pool->fq);
 	sock_def_readable(&xs->sk);
 }
 
@@ -297,7 +297,7 @@ void __xsk_map_flush(void)
 
 void xsk_tx_completed(struct xsk_buff_pool *pool, u32 nb_entries)
 {
-	xskq_prod_submit_n(pool->umem->cq, nb_entries);
+	xskq_prod_submit_n(pool->cq, nb_entries);
 }
 EXPORT_SYMBOL(xsk_tx_completed);
 
@@ -331,7 +331,7 @@ bool xsk_tx_peek_desc(struct xsk_buff_pool *pool, struct xdp_desc *desc)
 		 * if there is space in it. This avoids having to implement
 		 * any buffering in the Tx path.
 		 */
-		if (xskq_prod_reserve_addr(umem->cq, desc->addr))
+		if (xskq_prod_reserve_addr(pool->cq, desc->addr))
 			goto out;
 
 		xskq_cons_release(xs->tx);
@@ -369,7 +369,7 @@ static void xsk_destruct_skb(struct sk_buff *skb)
 	unsigned long flags;
 
 	spin_lock_irqsave(&xs->tx_completion_lock, flags);
-	xskq_prod_submit_addr(xs->umem->cq, addr);
+	xskq_prod_submit_addr(xs->pool->cq, addr);
 	spin_unlock_irqrestore(&xs->tx_completion_lock, flags);
 
 	sock_wfree(skb);
@@ -413,7 +413,7 @@ static int xsk_generic_xmit(struct sock *sk)
 		 * if there is space in it. This avoids having to implement
 		 * any buffering in the Tx path.
 		 */
-		if (unlikely(err) || xskq_prod_reserve(xs->umem->cq)) {
+		if (unlikely(err) || xskq_prod_reserve(xs->pool->cq)) {
 			kfree_skb(skb);
 			goto out;
 		}
@@ -606,6 +606,8 @@ static int xsk_release(struct socket *sock)
 
 	xskq_destroy(xs->rx);
 	xskq_destroy(xs->tx);
+	xskq_destroy(xs->fq_tmp);
+	xskq_destroy(xs->cq_tmp);
 
 	sock_orphan(sk);
 	sock->sk = NULL;
@@ -633,6 +635,11 @@ static struct socket *xsk_lookup_xsk_from_fd(int fd)
 	return sock;
 }
 
+static bool xsk_validate_queues(struct xdp_sock *xs)
+{
+	return xs->fq_tmp && xs->cq_tmp;
+}
+
 static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
 {
 	struct sockaddr_xdp *sxdp = (struct sockaddr_xdp *)addr;
@@ -689,6 +696,12 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
 			goto out_unlock;
 		}
 
+		if (xs->fq_tmp || xs->cq_tmp) {
+			/* Do not allow setting your own fq or cq. */
+			err = -EINVAL;
+			goto out_unlock;
+		}
+
 		sock = xsk_lookup_xsk_from_fd(sxdp->sxdp_shared_umem_fd);
 		if (IS_ERR(sock)) {
 			err = PTR_ERR(sock);
@@ -713,7 +726,7 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
 		xdp_get_umem(umem_xs->umem);
 		WRITE_ONCE(xs->umem, umem_xs->umem);
 		sockfd_put(sock);
-	} else if (!xs->umem || !xdp_umem_validate_queues(xs->umem)) {
+	} else if (!xs->umem || !xsk_validate_queues(xs)) {
 		err = -EINVAL;
 		goto out_unlock;
 	} else {
@@ -849,11 +862,9 @@ static int xsk_setsockopt(struct socket *sock, int level, int optname,
 			return -EINVAL;
 		}
 
-		q = (optname == XDP_UMEM_FILL_RING) ? &xs->umem->fq :
+		q = (optname == XDP_UMEM_FILL_RING) ? &xs->fq_tmp :
-			&xs->umem->cq;
+			&xs->cq_tmp;
 		err = xsk_init_queue(entries, q, true);
-		if (optname == XDP_UMEM_FILL_RING)
-			xp_set_fq(xs->pool, *q);
 		mutex_unlock(&xs->mutex);
 		return err;
 	}
@@ -920,7 +931,7 @@ static int xsk_getsockopt(struct socket *sock, int level, int optname,
 		if (extra_stats) {
 			stats.rx_ring_full = xs->rx_queue_full;
 			stats.rx_fill_ring_empty_descs =
-				xs->umem ? xskq_nb_queue_empty_descs(xs->umem->fq) : 0;
+				xs->pool ? xskq_nb_queue_empty_descs(xs->pool->fq) : 0;
 			stats.tx_ring_empty_descs = xskq_nb_queue_empty_descs(xs->tx);
 		} else {
 			stats.rx_dropped += xs->rx_queue_full;
@@ -1022,7 +1033,6 @@ static int xsk_mmap(struct file *file, struct socket *sock,
 	unsigned long size = vma->vm_end - vma->vm_start;
 	struct xdp_sock *xs = xdp_sk(sock->sk);
 	struct xsk_queue *q = NULL;
-	struct xdp_umem *umem;
 	unsigned long pfn;
 	struct page *qpg;
 
@@ -1034,16 +1044,12 @@ static int xsk_mmap(struct file *file, struct socket *sock,
 	} else if (offset == XDP_PGOFF_TX_RING) {
 		q = READ_ONCE(xs->tx);
 	} else {
-		umem = READ_ONCE(xs->umem);
-		if (!umem)
-			return -EINVAL;
-
 		/* Matches the smp_wmb() in XDP_UMEM_REG */
 		smp_rmb();
 		if (offset == XDP_UMEM_PGOFF_FILL_RING)
-			q = READ_ONCE(umem->fq);
+			q = READ_ONCE(xs->fq_tmp);
 		else if (offset == XDP_UMEM_PGOFF_COMPLETION_RING)
-			q = READ_ONCE(umem->cq);
+			q = READ_ONCE(xs->cq_tmp);
 	}
 
 	if (!q)

net/xdp/xsk_buff_pool.c

@@ -65,6 +65,11 @@ struct xsk_buff_pool *xp_create_and_assign_umem(struct xdp_sock *xs,
 	INIT_LIST_HEAD(&pool->free_list);
 	refcount_set(&pool->users, 1);
 
+	pool->fq = xs->fq_tmp;
+	pool->cq = xs->cq_tmp;
+	xs->fq_tmp = NULL;
+	xs->cq_tmp = NULL;
+
 	for (i = 0; i < pool->free_heads_cnt; i++) {
 		xskb = &pool->heads[i];
 		xskb->pool = pool;
@@ -81,11 +86,6 @@ out:
 	return NULL;
 }
 
-void xp_set_fq(struct xsk_buff_pool *pool, struct xsk_queue *fq)
-{
-	pool->fq = fq;
-}
-
 void xp_set_rxq_info(struct xsk_buff_pool *pool, struct xdp_rxq_info *rxq)
 {
 	u32 i;
@@ -189,6 +189,16 @@ static void xp_release_deferred(struct work_struct *work)
 	xp_clear_dev(pool);
 	rtnl_unlock();
 
+	if (pool->fq) {
+		xskq_destroy(pool->fq);
+		pool->fq = NULL;
+	}
+
+	if (pool->cq) {
+		xskq_destroy(pool->cq);
+		pool->cq = NULL;
+	}
+
 	xdp_put_umem(pool->umem);
 	xp_destroy(pool);
 }

net/xdp/xsk_diag.c

@@ -46,6 +46,7 @@ static int xsk_diag_put_rings_cfg(const struct xdp_sock *xs,
 
 static int xsk_diag_put_umem(const struct xdp_sock *xs, struct sk_buff *nlskb)
 {
+	struct xsk_buff_pool *pool = xs->pool;
 	struct xdp_umem *umem = xs->umem;
 	struct xdp_diag_umem du = {};
 	int err;
@@ -67,10 +68,11 @@ static int xsk_diag_put_umem(const struct xdp_sock *xs, struct sk_buff *nlskb)
 
 	err = nla_put(nlskb, XDP_DIAG_UMEM, sizeof(du), &du);
 
-	if (!err && umem->fq)
+	if (!err && pool->fq)
-		err = xsk_diag_put_ring(umem->fq, XDP_DIAG_UMEM_FILL_RING, nlskb);
+		err = xsk_diag_put_ring(pool->fq,
-	if (!err && umem->cq) {
+					XDP_DIAG_UMEM_FILL_RING, nlskb);
-		err = xsk_diag_put_ring(umem->cq, XDP_DIAG_UMEM_COMPLETION_RING,
+	if (!err && pool->cq) {
+		err = xsk_diag_put_ring(pool->cq, XDP_DIAG_UMEM_COMPLETION_RING,
 					nlskb);
 	}
 	return err;
@@ -83,7 +85,7 @@ static int xsk_diag_put_stats(const struct xdp_sock *xs, struct sk_buff *nlskb)
 	du.n_rx_dropped = xs->rx_dropped;
 	du.n_rx_invalid = xskq_nb_invalid_descs(xs->rx);
 	du.n_rx_full = xs->rx_queue_full;
-	du.n_fill_ring_empty = xs->umem ? xskq_nb_queue_empty_descs(xs->umem->fq) : 0;
+	du.n_fill_ring_empty = xs->pool ? xskq_nb_queue_empty_descs(xs->pool->fq) : 0;
 	du.n_tx_invalid = xskq_nb_invalid_descs(xs->tx);
 	du.n_tx_ring_empty = xskq_nb_queue_empty_descs(xs->tx);
 	return nla_put(nlskb, XDP_DIAG_STATS, sizeof(du), &du);