Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2019-12-27

The following pull-request contains BPF updates for your *net-next* tree.

We've added 127 non-merge commits during the last 17 day(s) which contain
a total of 110 files changed, 6901 insertions(+), 2721 deletions(-).

There are three merge conflicts. Conflicts and resolution looks as follows:

1) Merge conflict in net/bpf/test_run.c:

There was a tree-wide cleanup c593642c8be0 ("treewide: Use sizeof_field() macro")
which gets in the way with b590cb5f802d ("bpf: Switch to offsetofend in
BPF_PROG_TEST_RUN"):

  <<<<<<< HEAD
          if (!range_is_zero(__skb, offsetof(struct __sk_buff, priority) +
                             sizeof_field(struct __sk_buff, priority),
  =======
          if (!range_is_zero(__skb, offsetofend(struct __sk_buff, priority),
  >>>>>>> 7c8dce4b166113743adad131b5a24c4acc12f92c

There are a few occasions that look similar to this. Always take the chunk with
offsetofend(). Note that there is one where the fields differ in here:

  <<<<<<< HEAD
          if (!range_is_zero(__skb, offsetof(struct __sk_buff, tstamp) +
                             sizeof_field(struct __sk_buff, tstamp),
  =======
          if (!range_is_zero(__skb, offsetofend(struct __sk_buff, gso_segs),
  >>>>>>> 7c8dce4b166113743adad131b5a24c4acc12f92c

Just take the one with offsetofend() /and/ gso_segs. Latter is correct due to
850a88cc4096 ("bpf: Expose __sk_buff wire_len/gso_segs to BPF_PROG_TEST_RUN").

2) Merge conflict in arch/riscv/net/bpf_jit_comp.c:

(I'm keeping Bjorn in Cc here for a double-check in case I got it wrong.)

  <<<<<<< HEAD
          if (is_13b_check(off, insn))
                  return -1;
          emit(rv_blt(tcc, RV_REG_ZERO, off >> 1), ctx);
  =======
          emit_branch(BPF_JSLT, RV_REG_T1, RV_REG_ZERO, off, ctx);
  >>>>>>> 7c8dce4b166113743adad131b5a24c4acc12f92c

Result should look like:

          emit_branch(BPF_JSLT, tcc, RV_REG_ZERO, off, ctx);

3) Merge conflict in arch/riscv/include/asm/pgtable.h:

  <<<<<<< HEAD
  =======
  #define VMALLOC_SIZE     (KERN_VIRT_SIZE >> 1)
  #define VMALLOC_END      (PAGE_OFFSET - 1)
  #define VMALLOC_START    (PAGE_OFFSET - VMALLOC_SIZE)

  #define BPF_JIT_REGION_SIZE     (SZ_128M)
  #define BPF_JIT_REGION_START    (PAGE_OFFSET - BPF_JIT_REGION_SIZE)
  #define BPF_JIT_REGION_END      (VMALLOC_END)

  /*
   * Roughly size the vmemmap space to be large enough to fit enough
   * struct pages to map half the virtual address space. Then
   * position vmemmap directly below the VMALLOC region.
   */
  #define VMEMMAP_SHIFT \
          (CONFIG_VA_BITS - PAGE_SHIFT - 1 + STRUCT_PAGE_MAX_SHIFT)
  #define VMEMMAP_SIZE    BIT(VMEMMAP_SHIFT)
  #define VMEMMAP_END     (VMALLOC_START - 1)
  #define VMEMMAP_START   (VMALLOC_START - VMEMMAP_SIZE)

  #define vmemmap         ((struct page *)VMEMMAP_START)

  >>>>>>> 7c8dce4b166113743adad131b5a24c4acc12f92c

Only take the BPF_* defines from there and move them higher up in the
same file. Remove the rest from the chunk. The VMALLOC_* etc defines
got moved via 01f52e16b868 ("riscv: define vmemmap before pfn_to_page
calls"). Result:

  [...]
  #define __S101  PAGE_READ_EXEC
  #define __S110  PAGE_SHARED_EXEC
  #define __S111  PAGE_SHARED_EXEC

  #define VMALLOC_SIZE     (KERN_VIRT_SIZE >> 1)
  #define VMALLOC_END      (PAGE_OFFSET - 1)
  #define VMALLOC_START    (PAGE_OFFSET - VMALLOC_SIZE)

  #define BPF_JIT_REGION_SIZE     (SZ_128M)
  #define BPF_JIT_REGION_START    (PAGE_OFFSET - BPF_JIT_REGION_SIZE)
  #define BPF_JIT_REGION_END      (VMALLOC_END)

  /*
   * Roughly size the vmemmap space to be large enough to fit enough
   * struct pages to map half the virtual address space. Then
   * position vmemmap directly below the VMALLOC region.
   */
  #define VMEMMAP_SHIFT \
          (CONFIG_VA_BITS - PAGE_SHIFT - 1 + STRUCT_PAGE_MAX_SHIFT)
  #define VMEMMAP_SIZE    BIT(VMEMMAP_SHIFT)
  #define VMEMMAP_END     (VMALLOC_START - 1)
  #define VMEMMAP_START   (VMALLOC_START - VMEMMAP_SIZE)

  [...]

Let me know if there are any other issues.

Anyway, the main changes are:

1) Extend bpftool to produce a struct (aka "skeleton") tailored and specific
   to a provided BPF object file. This provides an alternative, simplified API
   compared to standard libbpf interaction. Also, add libbpf extern variable
   resolution for .kconfig section to import Kconfig data, from Andrii Nakryiko.

2) Add BPF dispatcher for XDP which is a mechanism to avoid indirect calls by
   generating a branch funnel as discussed back in bpfconf'19 at LSF/MM. Also,
   add various BPF riscv JIT improvements, from Björn Töpel.

3) Extend bpftool to allow matching BPF programs and maps by name,
   from Paul Chaignon.

4) Support for replacing cgroup BPF programs attached with BPF_F_ALLOW_MULTI
   flag for allowing updates without service interruption, from Andrey Ignatov.

5) Cleanup and simplification of ring access functions for AF_XDP with a
   bonus of 0-5% performance improvement, from Magnus Karlsson.

6) Enable BPF JITs for x86-64 and arm64 by default. Also, final version of
   audit support for BPF, from Daniel Borkmann and latter with Jiri Olsa.

7) Move and extend test_select_reuseport into BPF program tests under
   BPF selftests, from Jakub Sitnicki.

8) Various BPF sample improvements for xdpsock for customizing parameters
   to set up and benchmark AF_XDP, from Jay Jayatheerthan.

9) Improve libbpf to provide a ulimit hint on permission denied errors.
   Also change XDP sample programs to attach in driver mode by default,
   from Toke Høiland-Jørgensen.

10) Extend BPF test infrastructure to allow changing skb mark from tc BPF
    programs, from Nikita V. Shirokov.

11) Optimize prologue code sequence in BPF arm32 JIT, from Russell King.

12) Fix xdp_redirect_cpu BPF sample to manually attach to tracepoints after
    libbpf conversion, from Jesper Dangaard Brouer.

13) Minor misc improvements from various others.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 241 insertions, 224 deletions

diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index 328f661b83b2..02ada7ab8c6e 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -31,6 +31,8 @@
 
 #define TX_BATCH_SIZE 16
 
+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) &&
@@ -39,21 +41,21 @@ bool xsk_is_setup_for_bpf_map(struct xdp_sock *xs)
 
 bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt)
 {
-	return xskq_has_addrs(umem->fq, cnt);
+	return xskq_cons_has_entries(umem->fq, cnt);
 }
 EXPORT_SYMBOL(xsk_umem_has_addrs);
 
-u64 *xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr)
+bool xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr)
 {
-	return xskq_peek_addr(umem->fq, addr, umem);
+	return xskq_cons_peek_addr(umem->fq, addr, umem);
 }
 EXPORT_SYMBOL(xsk_umem_peek_addr);
 
-void xsk_umem_discard_addr(struct xdp_umem *umem)
+void xsk_umem_release_addr(struct xdp_umem *umem)
 {
-	xskq_discard_addr(umem->fq);
+	xskq_cons_release(umem->fq);
 }
-EXPORT_SYMBOL(xsk_umem_discard_addr);
+EXPORT_SYMBOL(xsk_umem_release_addr);
 
 void xsk_set_rx_need_wakeup(struct xdp_umem *umem)
 {
@@ -124,7 +126,7 @@ static void __xsk_rcv_memcpy(struct xdp_umem *umem, u64 addr, void *from_buf,
 	void *to_buf = xdp_umem_get_data(umem, addr);
 
 	addr = xsk_umem_add_offset_to_addr(addr);
-	if (xskq_crosses_non_contig_pg(umem, addr, len + metalen)) {
+	if (xskq_cons_crosses_non_contig_pg(umem, addr, len + metalen)) {
 		void *next_pg_addr = umem->pages[(addr >> PAGE_SHIFT) + 1].addr;
 		u64 page_start = addr & ~(PAGE_SIZE - 1);
 		u64 first_len = PAGE_SIZE - (addr - page_start);
@@ -146,7 +148,7 @@ static int __xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len)
 	u32 metalen;
 	int err;
 
-	if (!xskq_peek_addr(xs->umem->fq, &addr, xs->umem) ||
+	if (!xskq_cons_peek_addr(xs->umem->fq, &addr, xs->umem) ||
 	    len > xs->umem->chunk_size_nohr - XDP_PACKET_HEADROOM) {
 		xs->rx_dropped++;
 		return -ENOSPC;
@@ -165,9 +167,9 @@ static int __xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len)
 
 	offset += metalen;
 	addr = xsk_umem_adjust_offset(xs->umem, addr, offset);
-	err = xskq_produce_batch_desc(xs->rx, addr, len);
+	err = xskq_prod_reserve_desc(xs->rx, addr, len);
 	if (!err) {
-		xskq_discard_addr(xs->umem->fq);
+		xskq_cons_release(xs->umem->fq);
 		xdp_return_buff(xdp);
 		return 0;
 	}
@@ -178,7 +180,7 @@ static int __xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len)
 
 static int __xsk_rcv_zc(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len)
 {
-	int err = xskq_produce_batch_desc(xs->rx, (u64)xdp->handle, len);
+	int err = xskq_prod_reserve_desc(xs->rx, xdp->handle, len);
 
 	if (err)
 		xs->rx_dropped++;
@@ -214,7 +216,7 @@ static int xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
 
 static void xsk_flush(struct xdp_sock *xs)
 {
-	xskq_produce_flush_desc(xs->rx);
+	xskq_prod_submit(xs->rx);
 	xs->sk.sk_data_ready(&xs->sk);
 }
 
@@ -234,7 +236,7 @@ int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
 		goto out_unlock;
 	}
 
-	if (!xskq_peek_addr(xs->umem->fq, &addr, xs->umem) ||
+	if (!xskq_cons_peek_addr(xs->umem->fq, &addr, xs->umem) ||
 	    len > xs->umem->chunk_size_nohr - XDP_PACKET_HEADROOM) {
 		err = -ENOSPC;
 		goto out_drop;
@@ -245,12 +247,12 @@ int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
 	memcpy(buffer, xdp->data_meta, len + metalen);
 
 	addr = xsk_umem_adjust_offset(xs->umem, addr, metalen);
-	err = xskq_produce_batch_desc(xs->rx, addr, len);
+	err = xskq_prod_reserve_desc(xs->rx, addr, len);
 	if (err)
 		goto out_drop;
 
-	xskq_discard_addr(xs->umem->fq);
-	xskq_produce_flush_desc(xs->rx);
+	xskq_cons_release(xs->umem->fq);
+	xskq_prod_submit(xs->rx);
 
 	spin_unlock_bh(&xs->rx_lock);
 
@@ -264,11 +266,9 @@ out_unlock:
 	return err;
 }
 
-int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp,
-		       struct xdp_sock *xs)
+int __xsk_map_redirect(struct xdp_sock *xs, struct xdp_buff *xdp)
 {
-	struct xsk_map *m = container_of(map, struct xsk_map, map);
-	struct list_head *flush_list = this_cpu_ptr(m->flush_list);
+	struct list_head *flush_list = this_cpu_ptr(&xskmap_flush_list);
 	int err;
 
 	err = xsk_rcv(xs, xdp);
@@ -281,10 +281,9 @@ int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp,
 	return 0;
 }
 
-void __xsk_map_flush(struct bpf_map *map)
+void __xsk_map_flush(void)
 {
-	struct xsk_map *m = container_of(map, struct xsk_map, map);
-	struct list_head *flush_list = this_cpu_ptr(m->flush_list);
+	struct list_head *flush_list = this_cpu_ptr(&xskmap_flush_list);
 	struct xdp_sock *xs, *tmp;
 
 	list_for_each_entry_safe(xs, tmp, flush_list, flush_node) {
@@ -295,7 +294,7 @@ void __xsk_map_flush(struct bpf_map *map)
 
 void xsk_umem_complete_tx(struct xdp_umem *umem, u32 nb_entries)
 {
-	xskq_produce_flush_addr_n(umem->cq, nb_entries);
+	xskq_prod_submit_n(umem->cq, nb_entries);
 }
 EXPORT_SYMBOL(xsk_umem_complete_tx);
 
@@ -317,13 +316,18 @@ bool xsk_umem_consume_tx(struct xdp_umem *umem, struct xdp_desc *desc)
 
 	rcu_read_lock();
 	list_for_each_entry_rcu(xs, &umem->xsk_list, list) {
-		if (!xskq_peek_desc(xs->tx, desc, umem))
+		if (!xskq_cons_peek_desc(xs->tx, desc, umem))
 			continue;
 
-		if (xskq_produce_addr_lazy(umem->cq, desc->addr))
+		/* This is the backpreassure mechanism for the Tx path.
+		 * Reserve space in the completion queue and only proceed
+		 * 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))
 			goto out;
 
-		xskq_discard_desc(xs->tx);
+		xskq_cons_release(xs->tx);
 		rcu_read_unlock();
 		return true;
 	}
@@ -358,7 +362,7 @@ static void xsk_destruct_skb(struct sk_buff *skb)
 	unsigned long flags;
 
 	spin_lock_irqsave(&xs->tx_completion_lock, flags);
-	WARN_ON_ONCE(xskq_produce_addr(xs->umem->cq, addr));
+	xskq_prod_submit_addr(xs->umem->cq, addr);
 	spin_unlock_irqrestore(&xs->tx_completion_lock, flags);
 
 	sock_wfree(skb);
@@ -378,7 +382,7 @@ static int xsk_generic_xmit(struct sock *sk)
 	if (xs->queue_id >= xs->dev->real_num_tx_queues)
 		goto out;
 
-	while (xskq_peek_desc(xs->tx, &desc, xs->umem)) {
+	while (xskq_cons_peek_desc(xs->tx, &desc, xs->umem)) {
 		char *buffer;
 		u64 addr;
 		u32 len;
@@ -399,7 +403,12 @@ static int xsk_generic_xmit(struct sock *sk)
 		addr = desc.addr;
 		buffer = xdp_umem_get_data(xs->umem, addr);
 		err = skb_store_bits(skb, 0, buffer, len);
-		if (unlikely(err) || xskq_reserve_addr(xs->umem->cq)) {
+		/* This is the backpreassure mechanism for the Tx path.
+		 * Reserve space in the completion queue and only proceed
+		 * 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)) {
 			kfree_skb(skb);
 			goto out;
 		}
@@ -411,7 +420,7 @@ static int xsk_generic_xmit(struct sock *sk)
 		skb->destructor = xsk_destruct_skb;
 
 		err = dev_direct_xmit(skb, xs->queue_id);
-		xskq_discard_desc(xs->tx);
+		xskq_cons_release(xs->tx);
 		/* Ignore NET_XMIT_CN as packet might have been sent */
 		if (err == NET_XMIT_DROP || err == NETDEV_TX_BUSY) {
 			/* SKB completed but not sent */
@@ -477,9 +486,9 @@ static __poll_t xsk_poll(struct file *file, struct socket *sock,
 			__xsk_sendmsg(sk);
 	}
 
-	if (xs->rx && !xskq_empty_desc(xs->rx))
+	if (xs->rx && !xskq_prod_is_empty(xs->rx))
 		mask |= EPOLLIN | EPOLLRDNORM;
-	if (xs->tx && !xskq_full_desc(xs->tx))
+	if (xs->tx && !xskq_cons_is_full(xs->tx))
 		mask |= EPOLLOUT | EPOLLWRNORM;
 
 	return mask;
@@ -1183,7 +1192,7 @@ static struct pernet_operations xsk_net_ops = {
 
 static int __init xsk_init(void)
 {
-	int err;
+	int err, cpu;
 
 	err = proto_register(&xsk_proto, 0 /* no slab */);
 	if (err)
@@ -1201,6 +1210,8 @@ static int __init xsk_init(void)
 	if (err)
 		goto out_pernet;
 
+	for_each_possible_cpu(cpu)
+		INIT_LIST_HEAD(&per_cpu(xskmap_flush_list, cpu));
 	return 0;
 
 out_pernet:
diff --git a/net/xdp/xsk_queue.c b/net/xdp/xsk_queue.c
index b66504592d9b..c90e9c1e3c63 100644
--- a/net/xdp/xsk_queue.c
+++ b/net/xdp/xsk_queue.c
@@ -18,14 +18,14 @@ void xskq_set_umem(struct xsk_queue *q, u64 size, u64 chunk_mask)
 	q->chunk_mask = chunk_mask;
 }
 
-static u32 xskq_umem_get_ring_size(struct xsk_queue *q)
+static size_t xskq_get_ring_size(struct xsk_queue *q, bool umem_queue)
 {
-	return sizeof(struct xdp_umem_ring) + q->nentries * sizeof(u64);
-}
+	struct xdp_umem_ring *umem_ring;
+	struct xdp_rxtx_ring *rxtx_ring;
 
-static u32 xskq_rxtx_get_ring_size(struct xsk_queue *q)
-{
-	return sizeof(struct xdp_ring) + q->nentries * sizeof(struct xdp_desc);
+	if (umem_queue)
+		return struct_size(umem_ring, desc, q->nentries);
+	return struct_size(rxtx_ring, desc, q->nentries);
 }
 
 struct xsk_queue *xskq_create(u32 nentries, bool umem_queue)
@@ -43,8 +43,7 @@ struct xsk_queue *xskq_create(u32 nentries, bool umem_queue)
 
 	gfp_flags = GFP_KERNEL | __GFP_ZERO | __GFP_NOWARN |
 		    __GFP_COMP  | __GFP_NORETRY;
-	size = umem_queue ? xskq_umem_get_ring_size(q) :
-	       xskq_rxtx_get_ring_size(q);
+	size = xskq_get_ring_size(q, umem_queue);
 
 	q->ring = (struct xdp_ring *)__get_free_pages(gfp_flags,
 						      get_order(size));
diff --git a/net/xdp/xsk_queue.h b/net/xdp/xsk_queue.h
index eddae4688862..bec2af11853a 100644
--- a/net/xdp/xsk_queue.h
+++ b/net/xdp/xsk_queue.h
@@ -10,9 +10,6 @@
 #include <linux/if_xdp.h>
 #include <net/xdp_sock.h>
 
-#define RX_BATCH_SIZE 16
-#define LAZY_UPDATE_THRESHOLD 128
-
 struct xdp_ring {
 	u32 producer ____cacheline_aligned_in_smp;
 	u32 consumer ____cacheline_aligned_in_smp;
@@ -36,10 +33,8 @@ struct xsk_queue {
 	u64 size;
 	u32 ring_mask;
 	u32 nentries;
-	u32 prod_head;
-	u32 prod_tail;
-	u32 cons_head;
-	u32 cons_tail;
+	u32 cached_prod;
+	u32 cached_cons;
 	struct xdp_ring *ring;
 	u64 invalid_descs;
 };
@@ -86,56 +81,31 @@ struct xsk_queue {
  * now and again after circling through the ring.
  */
 
-/* Common functions operating for both RXTX and umem queues */
-
-static inline u64 xskq_nb_invalid_descs(struct xsk_queue *q)
-{
-	return q ? q->invalid_descs : 0;
-}
-
-static inline u32 xskq_nb_avail(struct xsk_queue *q, u32 dcnt)
-{
-	u32 entries = q->prod_tail - q->cons_tail;
-
-	if (entries == 0) {
-		/* Refresh the local pointer */
-		q->prod_tail = READ_ONCE(q->ring->producer);
-		entries = q->prod_tail - q->cons_tail;
-	}
-
-	return (entries > dcnt) ? dcnt : entries;
-}
-
-static inline u32 xskq_nb_free(struct xsk_queue *q, u32 producer, u32 dcnt)
-{
-	u32 free_entries = q->nentries - (producer - q->cons_tail);
-
-	if (free_entries >= dcnt)
-		return free_entries;
-
-	/* Refresh the local tail pointer */
-	q->cons_tail = READ_ONCE(q->ring->consumer);
-	return q->nentries - (producer - q->cons_tail);
-}
-
-static inline bool xskq_has_addrs(struct xsk_queue *q, u32 cnt)
-{
-	u32 entries = q->prod_tail - q->cons_tail;
-
-	if (entries >= cnt)
-		return true;
-
-	/* Refresh the local pointer. */
-	q->prod_tail = READ_ONCE(q->ring->producer);
-	entries = q->prod_tail - q->cons_tail;
-
-	return entries >= cnt;
-}
+/* The operations on the rings are the following:
+ *
+ * producer                           consumer
+ *
+ * RESERVE entries                    PEEK in the ring for entries
+ * WRITE data into the ring           READ data from the ring
+ * SUBMIT entries                     RELEASE entries
+ *
+ * The producer reserves one or more entries in the ring. It can then
+ * fill in these entries and finally submit them so that they can be
+ * seen and read by the consumer.
+ *
+ * The consumer peeks into the ring to see if the producer has written
+ * any new entries. If so, the producer can then read these entries
+ * and when it is done reading them release them back to the producer
+ * so that the producer can use these slots to fill in new entries.
+ *
+ * The function names below reflect these operations.
+ */
 
-/* UMEM queue */
+/* Functions that read and validate content from consumer rings. */
 
-static inline bool xskq_crosses_non_contig_pg(struct xdp_umem *umem, u64 addr,
-					      u64 length)
+static inline bool xskq_cons_crosses_non_contig_pg(struct xdp_umem *umem,
+						   u64 addr,
+						   u64 length)
 {
 	bool cross_pg = (addr & (PAGE_SIZE - 1)) + length > PAGE_SIZE;
 	bool next_pg_contig =
@@ -145,9 +115,16 @@ static inline bool xskq_crosses_non_contig_pg(struct xdp_umem *umem, u64 addr,
 	return cross_pg && !next_pg_contig;
 }
 
-static inline bool xskq_is_valid_addr(struct xsk_queue *q, u64 addr)
+static inline bool xskq_cons_is_valid_unaligned(struct xsk_queue *q,
+						u64 addr,
+						u64 length,
+						struct xdp_umem *umem)
 {
-	if (addr >= q->size) {
+	u64 base_addr = xsk_umem_extract_addr(addr);
+
+	addr = xsk_umem_add_offset_to_addr(addr);
+	if (base_addr >= q->size || addr >= q->size ||
+	    xskq_cons_crosses_non_contig_pg(umem, addr, length)) {
 		q->invalid_descs++;
 		return false;
 	}
@@ -155,15 +132,9 @@ static inline bool xskq_is_valid_addr(struct xsk_queue *q, u64 addr)
 	return true;
 }
 
-static inline bool xskq_is_valid_addr_unaligned(struct xsk_queue *q, u64 addr,
-						u64 length,
-						struct xdp_umem *umem)
+static inline bool xskq_cons_is_valid_addr(struct xsk_queue *q, u64 addr)
 {
-	u64 base_addr = xsk_umem_extract_addr(addr);
-
-	addr = xsk_umem_add_offset_to_addr(addr);
-	if (base_addr >= q->size || addr >= q->size ||
-	    xskq_crosses_non_contig_pg(umem, addr, length)) {
+	if (addr >= q->size) {
 		q->invalid_descs++;
 		return false;
 	}
@@ -171,204 +142,240 @@ static inline bool xskq_is_valid_addr_unaligned(struct xsk_queue *q, u64 addr,
 	return true;
 }
 
-static inline u64 *xskq_validate_addr(struct xsk_queue *q, u64 *addr,
-				      struct xdp_umem *umem)
+static inline bool xskq_cons_read_addr(struct xsk_queue *q, u64 *addr,
+				       struct xdp_umem *umem)
 {
-	while (q->cons_tail != q->cons_head) {
-		struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
-		unsigned int idx = q->cons_tail & q->ring_mask;
+	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
 
-		*addr = READ_ONCE(ring->desc[idx]) & q->chunk_mask;
+	while (q->cached_cons != q->cached_prod) {
+		u32 idx = q->cached_cons & q->ring_mask;
+
+		*addr = ring->desc[idx] & q->chunk_mask;
 
 		if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG) {
-			if (xskq_is_valid_addr_unaligned(q, *addr,
+			if (xskq_cons_is_valid_unaligned(q, *addr,
 							 umem->chunk_size_nohr,
 							 umem))
-				return addr;
+				return true;
 			goto out;
 		}
 
-		if (xskq_is_valid_addr(q, *addr))
-			return addr;
+		if (xskq_cons_is_valid_addr(q, *addr))
+			return true;
 
 out:
-		q->cons_tail++;
+		q->cached_cons++;
 	}
 
-	return NULL;
+	return false;
 }
 
-static inline u64 *xskq_peek_addr(struct xsk_queue *q, u64 *addr,
-				  struct xdp_umem *umem)
+static inline bool xskq_cons_is_valid_desc(struct xsk_queue *q,
+					   struct xdp_desc *d,
+					   struct xdp_umem *umem)
 {
-	if (q->cons_tail == q->cons_head) {
-		smp_mb(); /* D, matches A */
-		WRITE_ONCE(q->ring->consumer, q->cons_tail);
-		q->cons_head = q->cons_tail + xskq_nb_avail(q, RX_BATCH_SIZE);
+	if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG) {
+		if (!xskq_cons_is_valid_unaligned(q, d->addr, d->len, umem))
+			return false;
+
+		if (d->len > umem->chunk_size_nohr || d->options) {
+			q->invalid_descs++;
+			return false;
+		}
 
-		/* Order consumer and data */
-		smp_rmb();
+		return true;
 	}
 
-	return xskq_validate_addr(q, addr, umem);
-}
+	if (!xskq_cons_is_valid_addr(q, d->addr))
+		return false;
 
-static inline void xskq_discard_addr(struct xsk_queue *q)
-{
-	q->cons_tail++;
+	if (((d->addr + d->len) & q->chunk_mask) != (d->addr & q->chunk_mask) ||
+	    d->options) {
+		q->invalid_descs++;
+		return false;
+	}
+
+	return true;
 }
 
-static inline int xskq_produce_addr(struct xsk_queue *q, u64 addr)
+static inline bool xskq_cons_read_desc(struct xsk_queue *q,
+				       struct xdp_desc *desc,
+				       struct xdp_umem *umem)
 {
-	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
-
-	if (xskq_nb_free(q, q->prod_tail, 1) == 0)
-		return -ENOSPC;
+	while (q->cached_cons != q->cached_prod) {
+		struct xdp_rxtx_ring *ring = (struct xdp_rxtx_ring *)q->ring;
+		u32 idx = q->cached_cons & q->ring_mask;
 
-	/* A, matches D */
-	ring->desc[q->prod_tail++ & q->ring_mask] = addr;
+		*desc = ring->desc[idx];
+		if (xskq_cons_is_valid_desc(q, desc, umem))
+			return true;
 
-	/* Order producer and data */
-	smp_wmb(); /* B, matches C */
+		q->cached_cons++;
+	}
 
-	WRITE_ONCE(q->ring->producer, q->prod_tail);
-	return 0;
+	return false;
 }
 
-static inline int xskq_produce_addr_lazy(struct xsk_queue *q, u64 addr)
+/* Functions for consumers */
+
+static inline void __xskq_cons_release(struct xsk_queue *q)
 {
-	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
+	smp_mb(); /* D, matches A */
+	WRITE_ONCE(q->ring->consumer, q->cached_cons);
+}
 
-	if (xskq_nb_free(q, q->prod_head, LAZY_UPDATE_THRESHOLD) == 0)
-		return -ENOSPC;
+static inline void __xskq_cons_peek(struct xsk_queue *q)
+{
+	/* Refresh the local pointer */
+	q->cached_prod = READ_ONCE(q->ring->producer);
+	smp_rmb(); /* C, matches B */
+}
 
-	/* A, matches D */
-	ring->desc[q->prod_head++ & q->ring_mask] = addr;
-	return 0;
+static inline void xskq_cons_get_entries(struct xsk_queue *q)
+{
+	__xskq_cons_release(q);
+	__xskq_cons_peek(q);
 }
 
-static inline void xskq_produce_flush_addr_n(struct xsk_queue *q,
-					     u32 nb_entries)
+static inline bool xskq_cons_has_entries(struct xsk_queue *q, u32 cnt)
 {
-	/* Order producer and data */
-	smp_wmb(); /* B, matches C */
+	u32 entries = q->cached_prod - q->cached_cons;
 
-	q->prod_tail += nb_entries;
-	WRITE_ONCE(q->ring->producer, q->prod_tail);
+	if (entries >= cnt)
+		return true;
+
+	__xskq_cons_peek(q);
+	entries = q->cached_prod - q->cached_cons;
+
+	return entries >= cnt;
 }
 
-static inline int xskq_reserve_addr(struct xsk_queue *q)
+static inline bool xskq_cons_peek_addr(struct xsk_queue *q, u64 *addr,
+				       struct xdp_umem *umem)
 {
-	if (xskq_nb_free(q, q->prod_head, 1) == 0)
-		return -ENOSPC;
+	if (q->cached_prod == q->cached_cons)
+		xskq_cons_get_entries(q);
+	return xskq_cons_read_addr(q, addr, umem);
+}
 
-	/* A, matches D */
-	q->prod_head++;
-	return 0;
+static inline bool xskq_cons_peek_desc(struct xsk_queue *q,
+				       struct xdp_desc *desc,
+				       struct xdp_umem *umem)
+{
+	if (q->cached_prod == q->cached_cons)
+		xskq_cons_get_entries(q);
+	return xskq_cons_read_desc(q, desc, umem);
 }
 
-/* Rx/Tx queue */
+static inline void xskq_cons_release(struct xsk_queue *q)
+{
+	/* To improve performance, only update local state here.
+	 * Reflect this to global state when we get new entries
+	 * from the ring in xskq_cons_get_entries().
+	 */
+	q->cached_cons++;
+}
 
-static inline bool xskq_is_valid_desc(struct xsk_queue *q, struct xdp_desc *d,
-				      struct xdp_umem *umem)
+static inline bool xskq_cons_is_full(struct xsk_queue *q)
 {
-	if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG) {
-		if (!xskq_is_valid_addr_unaligned(q, d->addr, d->len, umem))
-			return false;
+	/* No barriers needed since data is not accessed */
+	return READ_ONCE(q->ring->producer) - READ_ONCE(q->ring->consumer) ==
+		q->nentries;
+}
 
-		if (d->len > umem->chunk_size_nohr || d->options) {
-			q->invalid_descs++;
-			return false;
-		}
+/* Functions for producers */
 
-		return true;
-	}
+static inline bool xskq_prod_is_full(struct xsk_queue *q)
+{
+	u32 free_entries = q->nentries - (q->cached_prod - q->cached_cons);
 
-	if (!xskq_is_valid_addr(q, d->addr))
+	if (free_entries)
 		return false;
 
-	if (((d->addr + d->len) & q->chunk_mask) != (d->addr & q->chunk_mask) ||
-	    d->options) {
-		q->invalid_descs++;
-		return false;
-	}
+	/* Refresh the local tail pointer */
+	q->cached_cons = READ_ONCE(q->ring->consumer);
+	free_entries = q->nentries - (q->cached_prod - q->cached_cons);
 
-	return true;
+	return !free_entries;
 }
 
-static inline struct xdp_desc *xskq_validate_desc(struct xsk_queue *q,
-						  struct xdp_desc *desc,
-						  struct xdp_umem *umem)
+static inline int xskq_prod_reserve(struct xsk_queue *q)
 {
-	while (q->cons_tail != q->cons_head) {
-		struct xdp_rxtx_ring *ring = (struct xdp_rxtx_ring *)q->ring;
-		unsigned int idx = q->cons_tail & q->ring_mask;
-
-		*desc = READ_ONCE(ring->desc[idx]);
-		if (xskq_is_valid_desc(q, desc, umem))
-			return desc;
-
-		q->cons_tail++;
-	}
+	if (xskq_prod_is_full(q))
+		return -ENOSPC;
 
-	return NULL;
+	/* A, matches D */
+	q->cached_prod++;
+	return 0;
 }
 
-static inline struct xdp_desc *xskq_peek_desc(struct xsk_queue *q,
-					      struct xdp_desc *desc,
-					      struct xdp_umem *umem)
+static inline int xskq_prod_reserve_addr(struct xsk_queue *q, u64 addr)
 {
-	if (q->cons_tail == q->cons_head) {
-		smp_mb(); /* D, matches A */
-		WRITE_ONCE(q->ring->consumer, q->cons_tail);
-		q->cons_head = q->cons_tail + xskq_nb_avail(q, RX_BATCH_SIZE);
-
-		/* Order consumer and data */
-		smp_rmb(); /* C, matches B */
-	}
+	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
 
-	return xskq_validate_desc(q, desc, umem);
-}
+	if (xskq_prod_is_full(q))
+		return -ENOSPC;
 
-static inline void xskq_discard_desc(struct xsk_queue *q)
-{
-	q->cons_tail++;
+	/* A, matches D */
+	ring->desc[q->cached_prod++ & q->ring_mask] = addr;
+	return 0;
 }
 
-static inline int xskq_produce_batch_desc(struct xsk_queue *q,
-					  u64 addr, u32 len)
+static inline int xskq_prod_reserve_desc(struct xsk_queue *q,
+					 u64 addr, u32 len)
 {
 	struct xdp_rxtx_ring *ring = (struct xdp_rxtx_ring *)q->ring;
-	unsigned int idx;
+	u32 idx;
 
-	if (xskq_nb_free(q, q->prod_head, 1) == 0)
+	if (xskq_prod_is_full(q))
 		return -ENOSPC;
 
 	/* A, matches D */
-	idx = (q->prod_head++) & q->ring_mask;
+	idx = q->cached_prod++ & q->ring_mask;
 	ring->desc[idx].addr = addr;
 	ring->desc[idx].len = len;
 
 	return 0;
 }
 
-static inline void xskq_produce_flush_desc(struct xsk_queue *q)
+static inline void __xskq_prod_submit(struct xsk_queue *q, u32 idx)
 {
-	/* Order producer and data */
 	smp_wmb(); /* B, matches C */
 
-	q->prod_tail = q->prod_head;
-	WRITE_ONCE(q->ring->producer, q->prod_tail);
+	WRITE_ONCE(q->ring->producer, idx);
+}
+
+static inline void xskq_prod_submit(struct xsk_queue *q)
+{
+	__xskq_prod_submit(q, q->cached_prod);
+}
+
+static inline void xskq_prod_submit_addr(struct xsk_queue *q, u64 addr)
+{
+	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
+	u32 idx = q->ring->producer;
+
+	ring->desc[idx++ & q->ring_mask] = addr;
+
+	__xskq_prod_submit(q, idx);
 }
 
-static inline bool xskq_full_desc(struct xsk_queue *q)
+static inline void xskq_prod_submit_n(struct xsk_queue *q, u32 nb_entries)
 {
-	return xskq_nb_avail(q, q->nentries) == q->nentries;
+	__xskq_prod_submit(q, q->ring->producer + nb_entries);
 }
 
-static inline bool xskq_empty_desc(struct xsk_queue *q)
+static inline bool xskq_prod_is_empty(struct xsk_queue *q)
 {
-	return xskq_nb_free(q, q->prod_tail, q->nentries) == q->nentries;
+	/* No barriers needed since data is not accessed */
+	return READ_ONCE(q->ring->consumer) == READ_ONCE(q->ring->producer);
+}
+
+/* For both producers and consumers */
+
+static inline u64 xskq_nb_invalid_descs(struct xsk_queue *q)
+{
+	return q ? q->invalid_descs : 0;
 }
 
 void xskq_set_umem(struct xsk_queue *q, u64 size, u64 chunk_mask);