Add support to share a umem between queue ids on the same
device. This mode can be invoked with the XDP_SHARED_UMEM bind
flag. Previously, sharing was only supported within the same
queue id and device, and you shared one set of fill and
completion rings. However, note that when sharing a umem between
queue ids, you need to create a fill ring and a completion ring
and tie them to the socket before you do the bind with the
XDP_SHARED_UMEM flag. This so that the single-producer
single-consumer semantics can be upheld.
Signed-off-by: Magnus Karlsson <magnus.karlsson@xxxxxxxxx>
---
include/net/xsk_buff_pool.h | 3 +++
net/xdp/xsk.c | 51 +++++++++++++++++++++++++++++----------------
net/xdp/xsk_buff_pool.c | 27 ++++++++++++++++++++++--
3 files changed, 61 insertions(+), 20 deletions(-)
diff --git a/include/net/xsk_buff_pool.h b/include/net/xsk_buff_pool.h
index 7513a17..844901c 100644
--- a/include/net/xsk_buff_pool.h
+++ b/include/net/xsk_buff_pool.h
@@ -76,6 +76,9 @@ struct xsk_buff_pool *xp_assign_umem(struct xsk_buff_pool *pool,
struct xdp_umem *umem);
int xp_assign_dev(struct xsk_buff_pool *pool, struct xdp_sock *xs,
struct net_device *dev, u16 queue_id, u16 flags);
+int xp_assign_dev_shared(struct xsk_buff_pool *pool, struct xdp_sock *xs,
+ struct xdp_umem *umem, struct net_device *dev,
+ u16 queue_id);
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);
diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index 4d0028c..1abc222 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -627,6 +627,7 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
struct sockaddr_xdp *sxdp = (struct sockaddr_xdp *)addr;
struct sock *sk = sock->sk;
struct xdp_sock *xs = xdp_sk(sk);
+ struct xsk_buff_pool *new_pool;
struct net_device *dev;
u32 flags, qid;
int err = 0;
@@ -679,12 +680,6 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
goto out_unlock;
}
- if (xs->pool->fq || xs->pool->cq) {
- /* 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);
@@ -697,17 +692,43 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
sockfd_put(sock);
goto out_unlock;
}
- if (umem_xs->dev != dev || umem_xs->queue_id != qid) {
+ if (umem_xs->dev != dev) {
err = -EINVAL;
sockfd_put(sock);
goto out_unlock;
}
- /* Share the buffer pool with the other socket. */
- xp_get_pool(umem_xs->pool);
- curr_pool = xs->pool;
- xs->pool = umem_xs->pool;
- xp_destroy(curr_pool);
+ if (umem_xs->queue_id != qid) {
+ /* Share the umem with another socket on another qid */
+ new_pool = xp_assign_umem(xs->pool, umem_xs->umem);
+ if (!new_pool) {
+ sockfd_put(sock);
+ goto out_unlock;
+ }
+
+ err = xp_assign_dev_shared(new_pool, xs, umem_xs->umem,
+ dev, qid);
+ if (err) {
+ xp_destroy(new_pool);
+ sockfd_put(sock);
+ goto out_unlock;
+ }
+ xs->pool = new_pool;
+ } else {
+ /* Share the buffer pool with the other socket. */
+ if (xs->pool->fq || xs->pool->cq) {
+ /* Do not allow setting your own fq or cq. */
+ err = -EINVAL;
+ sockfd_put(sock);
+ goto out_unlock;
+ }
+
+ xp_get_pool(umem_xs->pool);
+ curr_pool = xs->pool;
+ xs->pool = umem_xs->pool;
+ xp_destroy(curr_pool);
+ }
+
xdp_get_umem(umem_xs->umem);
WRITE_ONCE(xs->umem, umem_xs->umem);
sockfd_put(sock);
@@ -715,8 +736,6 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
err = -EINVAL;
goto out_unlock;
} else {
- struct xsk_buff_pool *new_pool;
-
/* This xsk has its own umem. */
new_pool = xp_assign_umem(xs->pool, xs->umem);
if (!new_pool) {
@@ -841,10 +860,6 @@ static int xsk_setsockopt(struct socket *sock, int level, int optname,
mutex_unlock(&xs->mutex);
return -EBUSY;
}
- if (!xs->umem) {
- mutex_unlock(&xs->mutex);
- return -EINVAL;
- }
q = (optname == XDP_UMEM_FILL_RING) ? &xs->pool->fq :
&xs->pool->cq;
diff --git a/net/xdp/xsk_buff_pool.c b/net/xdp/xsk_buff_pool.c
index 3c58d76..7987c17 100644
--- a/net/xdp/xsk_buff_pool.c
+++ b/net/xdp/xsk_buff_pool.c
@@ -126,8 +126,8 @@ static void xp_disable_drv_zc(struct xsk_buff_pool *pool)
}
}
-int xp_assign_dev(struct xsk_buff_pool *pool, struct xdp_sock *xs,
- struct net_device *netdev, u16 queue_id, u16 flags)
+static int __xp_assign_dev(struct xsk_buff_pool *pool, struct xdp_sock *xs,
+ struct net_device *netdev, u16 queue_id, u16 flags)
{
bool force_zc, force_copy;
struct netdev_bpf bpf;
@@ -196,6 +196,29 @@ int xp_assign_dev(struct xsk_buff_pool *pool, struct xdp_sock *xs,
return err;
}
+int xp_assign_dev(struct xsk_buff_pool *pool, struct xdp_sock *xs,
+ struct net_device *dev, u16 queue_id, u16 flags)
+{
+ return __xp_assign_dev(pool, xs, dev, queue_id, flags);
+}
+
+int xp_assign_dev_shared(struct xsk_buff_pool *pool, struct xdp_sock *xs,
+ struct xdp_umem *umem, struct net_device *dev,
+ u16 queue_id)
+{
+ u16 flags;
+
+ /* One fill and completion ring required for each queue id. */
+ if (!pool->fq || !pool->cq)
+ return -EINVAL;
+
+ flags = umem->zc ? XDP_ZEROCOPY : XDP_COPY;
+ if (pool->uses_need_wakeup)
+ flags |= XDP_USE_NEED_WAKEUP;
+
+ return __xp_assign_dev(pool, xs, dev, queue_id, flags);
+}
+
void xp_clear_dev(struct xsk_buff_pool *pool)
{
if (!pool->netdev)
--
2.7.4
