summaryrefslogtreecommitdiff
path: root/drivers/infiniband/sw
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/infiniband/sw')
-rw-r--r--drivers/infiniband/sw/rdmavt/cq.c3
-rw-r--r--drivers/infiniband/sw/rxe/rxe_mr.c1
-rw-r--r--drivers/infiniband/sw/rxe/rxe_net.c81
-rw-r--r--drivers/infiniband/sw/rxe/rxe_odp.c5
-rw-r--r--drivers/infiniband/sw/rxe/rxe_qp.c49
-rw-r--r--drivers/infiniband/sw/rxe/rxe_srq.c7
-rw-r--r--drivers/infiniband/sw/siw/siw_cm.c59
7 files changed, 191 insertions, 14 deletions
diff --git a/drivers/infiniband/sw/rdmavt/cq.c b/drivers/infiniband/sw/rdmavt/cq.c
index 0ca2743f1075..e7835ca70e2b 100644
--- a/drivers/infiniband/sw/rdmavt/cq.c
+++ b/drivers/infiniband/sw/rdmavt/cq.c
@@ -518,7 +518,8 @@ int rvt_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry)
*/
int rvt_driver_cq_init(void)
{
- comp_vector_wq = alloc_workqueue("%s", WQ_HIGHPRI | WQ_CPU_INTENSIVE,
+ comp_vector_wq = alloc_workqueue("%s",
+ WQ_HIGHPRI | WQ_CPU_INTENSIVE | WQ_PERCPU,
0, "rdmavt_cq");
if (!comp_vector_wq)
return -ENOMEM;
diff --git a/drivers/infiniband/sw/rxe/rxe_mr.c b/drivers/infiniband/sw/rxe/rxe_mr.c
index bcb97b3ea58a..b1df05238848 100644
--- a/drivers/infiniband/sw/rxe/rxe_mr.c
+++ b/drivers/infiniband/sw/rxe/rxe_mr.c
@@ -452,7 +452,6 @@ static int rxe_mr_flush_pmem_iova(struct rxe_mr *mr, u64 iova, unsigned int leng
length -= bytes;
iova += bytes;
- page_offset = 0;
}
return 0;
diff --git a/drivers/infiniband/sw/rxe/rxe_net.c b/drivers/infiniband/sw/rxe/rxe_net.c
index ac0183a2ff7a..0bd0902b11f7 100644
--- a/drivers/infiniband/sw/rxe/rxe_net.c
+++ b/drivers/infiniband/sw/rxe/rxe_net.c
@@ -20,6 +20,86 @@
static struct rxe_recv_sockets recv_sockets;
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+/*
+ * lockdep can detect false positive circular dependencies
+ * when there are user-space socket API users or in kernel
+ * users switching between a tcp and rdma transport.
+ * Maybe also switching between siw and rxe may cause
+ * problems as per default sockets are only classified
+ * by family and not by ip protocol. And there might
+ * be different locks used between the application
+ * and the low level sockets.
+ *
+ * Problems were seen with ksmbd.ko and cifs.ko,
+ * switching transports, use git blame to find
+ * more details.
+ */
+static struct lock_class_key rxe_recv_sk_key[2];
+static struct lock_class_key rxe_recv_slock_key[2];
+#endif /* CONFIG_DEBUG_LOCK_ALLOC */
+
+static inline void rxe_reclassify_recv_socket(struct socket *sock)
+{
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+ struct sock *sk = sock->sk;
+
+ if (WARN_ON_ONCE(!sock_allow_reclassification(sk)))
+ return;
+
+ switch (sk->sk_family) {
+ case AF_INET:
+ sock_lock_init_class_and_name(sk,
+ "slock-AF_INET-RDMA-RXE-RECV",
+ &rxe_recv_slock_key[0],
+ "sk_lock-AF_INET-RDMA-RXE-RECV",
+ &rxe_recv_sk_key[0]);
+ break;
+ case AF_INET6:
+ sock_lock_init_class_and_name(sk,
+ "slock-AF_INET6-RDMA-RXE-RECV",
+ &rxe_recv_slock_key[1],
+ "sk_lock-AF_INET6-RDMA-RXE-RECV",
+ &rxe_recv_sk_key[1]);
+ break;
+ default:
+ WARN_ON_ONCE(1);
+ return;
+ }
+ /*
+ * sock_lock_init_class_and_name() calls
+ * sk_owner_set(sk, THIS_MODULE); in order
+ * to make sure the referenced global
+ * variables rxe_recv_slock_key and
+ * rxe_recv_sk_key are not removed
+ * before the socket is closed.
+ *
+ * However this prevents rxe_net_exit()
+ * from being called and 'rmmod rdma_rxe'
+ * is refused because of the references.
+ *
+ * For the global sockets in recv_sockets,
+ * we are sure that rxe_net_exit() will call
+ * rxe_release_udp_tunnel -> udp_tunnel_sock_release.
+ *
+ * So we don't need the additional reference to
+ * our own (THIS_MODULE).
+ */
+ sk_owner_put(sk);
+ /*
+ * We also call sk_owner_clear() otherwise
+ * sk_owner_put(sk) in sk_prot_free will
+ * fail, which is called via
+ * sk_free -> __sk_free -> sk_destruct
+ * and sk_destruct calls __sk_destruct
+ * directly or via call_rcu()
+ * so sk_prot_free() might be called
+ * after rxe_net_exit().
+ */
+ sk_owner_clear(sk);
+#endif /* CONFIG_DEBUG_LOCK_ALLOC */
+}
+
static struct dst_entry *rxe_find_route4(struct rxe_qp *qp,
struct net_device *ndev,
struct in_addr *saddr,
@@ -192,6 +272,7 @@ static struct socket *rxe_setup_udp_tunnel(struct net *net, __be16 port,
err = udp_sock_create(net, &udp_cfg, &sock);
if (err < 0)
return ERR_PTR(err);
+ rxe_reclassify_recv_socket(sock);
tnl_cfg.encap_type = 1;
tnl_cfg.encap_rcv = rxe_udp_encap_recv;
diff --git a/drivers/infiniband/sw/rxe/rxe_odp.c b/drivers/infiniband/sw/rxe/rxe_odp.c
index f58e3ec6252f..c928cbf2e35f 100644
--- a/drivers/infiniband/sw/rxe/rxe_odp.c
+++ b/drivers/infiniband/sw/rxe/rxe_odp.c
@@ -179,8 +179,10 @@ static int rxe_odp_map_range_and_lock(struct rxe_mr *mr, u64 iova, int length, u
return err;
need_fault = rxe_check_pagefault(umem_odp, iova, length);
- if (need_fault)
+ if (need_fault) {
+ mutex_unlock(&umem_odp->umem_mutex);
return -EFAULT;
+ }
}
return 0;
@@ -358,7 +360,6 @@ int rxe_odp_flush_pmem_iova(struct rxe_mr *mr, u64 iova,
length -= bytes;
iova += bytes;
- page_offset = 0;
}
mutex_unlock(&umem_odp->umem_mutex);
diff --git a/drivers/infiniband/sw/rxe/rxe_qp.c b/drivers/infiniband/sw/rxe/rxe_qp.c
index 95f1c1c2949d..845bdd03ca28 100644
--- a/drivers/infiniband/sw/rxe/rxe_qp.c
+++ b/drivers/infiniband/sw/rxe/rxe_qp.c
@@ -15,6 +15,54 @@
#include "rxe_queue.h"
#include "rxe_task.h"
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+/*
+ * lockdep can detect false positive circular dependencies
+ * when there are user-space socket API users or in kernel
+ * users switching between a tcp and rdma transport.
+ * Maybe also switching between siw and rxe may cause
+ * problems as per default sockets are only classified
+ * by family and not by ip protocol. And there might
+ * be different locks used between the application
+ * and the low level sockets.
+ *
+ * Problems were seen with ksmbd.ko and cifs.ko,
+ * switching transports, use git blame to find
+ * more details.
+ */
+static struct lock_class_key rxe_send_sk_key[2];
+static struct lock_class_key rxe_send_slock_key[2];
+#endif /* CONFIG_DEBUG_LOCK_ALLOC */
+
+static inline void rxe_reclassify_send_socket(struct socket *sock)
+{
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+ struct sock *sk = sock->sk;
+
+ if (WARN_ON_ONCE(!sock_allow_reclassification(sk)))
+ return;
+
+ switch (sk->sk_family) {
+ case AF_INET:
+ sock_lock_init_class_and_name(sk,
+ "slock-AF_INET-RDMA-RXE-SEND",
+ &rxe_send_slock_key[0],
+ "sk_lock-AF_INET-RDMA-RXE-SEND",
+ &rxe_send_sk_key[0]);
+ break;
+ case AF_INET6:
+ sock_lock_init_class_and_name(sk,
+ "slock-AF_INET6-RDMA-RXE-SEND",
+ &rxe_send_slock_key[1],
+ "sk_lock-AF_INET6-RDMA-RXE-SEND",
+ &rxe_send_sk_key[1]);
+ break;
+ default:
+ WARN_ON_ONCE(1);
+ }
+#endif /* CONFIG_DEBUG_LOCK_ALLOC */
+}
+
static int rxe_qp_chk_cap(struct rxe_dev *rxe, struct ib_qp_cap *cap,
int has_srq)
{
@@ -244,6 +292,7 @@ static int rxe_qp_init_req(struct rxe_dev *rxe, struct rxe_qp *qp,
err = sock_create_kern(&init_net, AF_INET, SOCK_DGRAM, 0, &qp->sk);
if (err < 0)
return err;
+ rxe_reclassify_send_socket(qp->sk);
qp->sk->sk->sk_user_data = qp;
/* pick a source UDP port number for this QP based on
diff --git a/drivers/infiniband/sw/rxe/rxe_srq.c b/drivers/infiniband/sw/rxe/rxe_srq.c
index 3661cb627d28..2a234f26ac10 100644
--- a/drivers/infiniband/sw/rxe/rxe_srq.c
+++ b/drivers/infiniband/sw/rxe/rxe_srq.c
@@ -171,7 +171,7 @@ int rxe_srq_from_attr(struct rxe_dev *rxe, struct rxe_srq *srq,
udata, mi, &srq->rq.producer_lock,
&srq->rq.consumer_lock);
if (err)
- goto err_free;
+ return err;
srq->rq.max_wr = attr->max_wr;
}
@@ -180,11 +180,6 @@ int rxe_srq_from_attr(struct rxe_dev *rxe, struct rxe_srq *srq,
srq->limit = attr->srq_limit;
return 0;
-
-err_free:
- rxe_queue_cleanup(q);
- srq->rq.queue = NULL;
- return err;
}
void rxe_srq_cleanup(struct rxe_pool_elem *elem)
diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index 708b13993fdf..1d3de8209bfa 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -39,6 +39,55 @@ static void siw_cm_llp_error_report(struct sock *s);
static int siw_cm_upcall(struct siw_cep *cep, enum iw_cm_event_type reason,
int status);
+
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+/*
+ * lockdep can detect false positive circular dependencies
+ * when there are user-space socket API users or in kernel
+ * users switching between a tcp and rdma transport.
+ * Maybe also switching between siw and rxe may cause
+ * problems as per default sockets are only classified
+ * by family and not by ip protocol. And there might
+ * be different locks used between the application
+ * and the low level sockets.
+ *
+ * Problems were seen with ksmbd.ko and cifs.ko,
+ * switching transports, use git blame to find
+ * more details.
+ */
+static struct lock_class_key siw_sk_key[2];
+static struct lock_class_key siw_slock_key[2];
+#endif /* CONFIG_DEBUG_LOCK_ALLOC */
+
+static inline void siw_reclassify_socket(struct socket *sock)
+{
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+ struct sock *sk = sock->sk;
+
+ if (WARN_ON_ONCE(!sock_allow_reclassification(sk)))
+ return;
+
+ switch (sk->sk_family) {
+ case AF_INET:
+ sock_lock_init_class_and_name(sk,
+ "slock-AF_INET-RDMA-SIW",
+ &siw_slock_key[0],
+ "sk_lock-AF_INET-RDMA-SIW",
+ &siw_sk_key[0]);
+ break;
+ case AF_INET6:
+ sock_lock_init_class_and_name(sk,
+ "slock-AF_INET6-RDMA-SIW",
+ &siw_slock_key[1],
+ "sk_lock-AF_INET6-RDMA-SIW",
+ &siw_sk_key[1]);
+ break;
+ default:
+ WARN_ON_ONCE(1);
+ }
+#endif /* CONFIG_DEBUG_LOCK_ALLOC */
+}
+
static void siw_sk_assign_cm_upcalls(struct sock *sk)
{
struct siw_cep *cep = sk_to_cep(sk);
@@ -1340,11 +1389,11 @@ static int kernel_bindconnect(struct socket *s, struct sockaddr *laddr,
return rv;
}
- rv = s->ops->bind(s, laddr, size);
+ rv = s->ops->bind(s, (struct sockaddr_unsized *)laddr, size);
if (rv < 0)
return rv;
- rv = s->ops->connect(s, raddr, size, flags);
+ rv = s->ops->connect(s, (struct sockaddr_unsized *)raddr, size, flags);
return rv < 0 ? rv : 0;
}
@@ -1394,6 +1443,7 @@ int siw_connect(struct iw_cm_id *id, struct iw_cm_conn_param *params)
rv = sock_create(v4 ? AF_INET : AF_INET6, SOCK_STREAM, IPPROTO_TCP, &s);
if (rv < 0)
goto error;
+ siw_reclassify_socket(s);
/*
* NOTE: For simplification, connect() is called in blocking
@@ -1770,6 +1820,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
rv = sock_create(addr_family, SOCK_STREAM, IPPROTO_TCP, &s);
if (rv < 0)
return rv;
+ siw_reclassify_socket(s);
/*
* Allow binding local port when still in TIME_WAIT from last close.
@@ -1789,7 +1840,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
goto error;
}
}
- rv = s->ops->bind(s, (struct sockaddr *)laddr,
+ rv = s->ops->bind(s, (struct sockaddr_unsized *)laddr,
sizeof(struct sockaddr_in));
} else {
struct sockaddr_in6 *laddr = &to_sockaddr_in6(id->local_addr);
@@ -1813,7 +1864,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
goto error;
}
}
- rv = s->ops->bind(s, (struct sockaddr *)laddr,
+ rv = s->ops->bind(s, (struct sockaddr_unsized *)laddr,
sizeof(struct sockaddr_in6));
}
if (rv) {
generated by cgit (git 2.34.1) at 2026-01-03 17:07:56 +0000