[PATCH net 0/9] net/smc: bug fixes 2017-09-20

STALE3270d

6 messages, 2 authors, 2017-09-21 · open the first message on its own page

[PATCH net 0/9] net/smc: bug fixes 2017-09-20

From: Ursula Braun <hidden>
Date: 2017-09-21 07:16:42

Hi Dave,

here is a collection of small smc-patches built for net fixing
smc problems in different areas.

Thanks,
Ursula
*** BLURB HERE ***

Hans Wippel (2):
  net/smc: add missing dev_put
  net/smc: add receive timeout check

Ursula Braun (7):
  net/smc: take RCU read lock for routing cache lookup
  net/smc: adjust net_device refcount
  net/smc: adapt send request completion notification
  net/smc: longer delay for client link group removal
  net/smc: terminate link group if out-of-sync is received
  net/smc: introduce a delay
  net/smc: no close wait in case of process shut down

 net/smc/af_smc.c    | 16 +++++++++-------
 net/smc/smc.h       |  2 +-
 net/smc/smc_clc.c   | 10 +++++-----
 net/smc/smc_clc.h   |  3 +--
 net/smc/smc_close.c | 27 +++++++++++++++------------
 net/smc/smc_core.c  | 16 ++++++++++++----
 net/smc/smc_ib.c    |  1 +
 net/smc/smc_pnet.c  |  4 +++-
 net/smc/smc_rx.c    |  2 ++
 net/smc/smc_tx.c    | 12 ++++++++----
 net/smc/smc_wr.c    |  2 +-
 11 files changed, 58 insertions(+), 37 deletions(-)

-- 
2.13.5

[PATCH net 3/9] net/smc: take RCU read lock for routing cache lookup

From: Ursula Braun <hidden>
Date: 2017-09-21 07:16:46

smc_netinfo_by_tcpsk() looks up the routing cache. Such a lookup requires
protection by an RCU read lock.

Signed-off-by: Ursula Braun <redacted>
---
 net/smc/af_smc.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index 8c6d24b2995d..2e8d2dabac0c 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c

@@ -282,6 +282,7 @@ int smc_netinfo_by_tcpsk(struct socket *clcsock,
 			 __be32 *subnet, u8 *prefix_len)
 {
 	struct dst_entry *dst = sk_dst_get(clcsock->sk);
+	struct in_device *in_dev;
 	struct sockaddr_in addr;
 	int rc = -ENOENT;
 	int len;

@@ -298,14 +299,17 @@ int smc_netinfo_by_tcpsk(struct socket *clcsock,
 	/* get address to which the internal TCP socket is bound */
 	kernel_getsockname(clcsock, (struct sockaddr *)&addr, &len);
 	/* analyze IPv4 specific data of net_device belonging to TCP socket */
-	for_ifa(dst->dev->ip_ptr) {
-		if (ifa->ifa_address != addr.sin_addr.s_addr)
+	rcu_read_lock();
+	in_dev = __in_dev_get_rcu(dst->dev);
+	for_ifa(in_dev) {
+		if (!inet_ifa_match(addr.sin_addr.s_addr, ifa))
 			continue;
 		*prefix_len = inet_mask_len(ifa->ifa_mask);
 		*subnet = ifa->ifa_address & ifa->ifa_mask;
 		rc = 0;
 		break;
-	} endfor_ifa(dst->dev->ip_ptr);
+	} endfor_ifa(in_dev);
+	rcu_read_unlock();
 
 out_rel:
 	dst_release(dst);

-- 
2.13.5

[PATCH net 2/9] net/smc: add receive timeout check

From: Ursula Braun <hidden>
Date: 2017-09-21 07:16:46

From: Hans Wippel <redacted>

The SMC receive function currently lacks a timeout check under the
condition that no data were received and no data are available. This
patch adds such a check.

Signed-off-by: Hans Wippel <redacted>
Signed-off-by: Ursula Braun <redacted>
---
 net/smc/smc_rx.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/smc/smc_rx.c b/net/smc/smc_rx.c
index b17a333e9bb0..3e631ae4b6b6 100644
--- a/net/smc/smc_rx.c
+++ b/net/smc/smc_rx.c

@@ -148,6 +148,8 @@ int smc_rx_recvmsg(struct smc_sock *smc, struct msghdr *msg, size_t len,
 				read_done = sock_intr_errno(timeo);
 				break;
 			}
+			if (!timeo)
+				return -EAGAIN;
 		}
 
 		if (!atomic_read(&conn->bytes_to_rcv)) {

-- 
2.13.5

[PATCH net 4/9] net/smc: adjust net_device refcount

From: Ursula Braun <hidden>
Date: 2017-09-21 07:16:48

smc_pnet_fill_entry() uses dev_get_by_name() adding a refcount to ndev.
The following smc_pnet_enter() has to reduce the refcount if the entry
to be added exists already in the pnet table.

Signed-off-by: Ursula Braun <redacted>
---
 net/smc/smc_pnet.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c
index 78f7af28ae4f..31f8453c25c5 100644
--- a/net/smc/smc_pnet.c
+++ b/net/smc/smc_pnet.c

@@ -181,8 +181,10 @@ static int smc_pnet_enter(struct smc_pnetentry *new_pnetelem)
 			     sizeof(new_pnetelem->ndev->name)) ||
 		    smc_pnet_same_ibname(pnetelem,
 					 new_pnetelem->smcibdev->ibdev->name,
-					 new_pnetelem->ib_port))
+					 new_pnetelem->ib_port)) {
+			dev_put(pnetelem->ndev);
 			goto found;
+		}
 	}
 	list_add_tail(&new_pnetelem->list, &smc_pnettable.pnetlist);
 	rc = 0;

-- 
2.13.5

[PATCH net 8/9] net/smc: introduce a delay

From: Ursula Braun <hidden>
Date: 2017-09-21 07:16:52

The number of outstanding work requests is limited. If all work
requests are in use, tx processing is postponed to another scheduling
of the tx worker. Switch to a delayed worker to have a gap for tx
completion queue events before the next retry.

Signed-off-by: Ursula Braun <redacted>
---
 net/smc/smc.h       |  2 +-
 net/smc/smc_close.c | 12 +++++++-----
 net/smc/smc_tx.c    | 12 ++++++++----
 3 files changed, 16 insertions(+), 10 deletions(-)

diff --git a/net/smc/smc.h b/net/smc/smc.h
index 6e44313e4467..0ccd6fa387ad 100644
--- a/net/smc/smc.h
+++ b/net/smc/smc.h

@@ -149,7 +149,7 @@ struct smc_connection {
 	atomic_t		sndbuf_space;	/* remaining space in sndbuf */
 	u16			tx_cdc_seq;	/* sequence # for CDC send */
 	spinlock_t		send_lock;	/* protect wr_sends */
-	struct work_struct	tx_work;	/* retry of smc_cdc_msg_send */
+	struct delayed_work	tx_work;	/* retry of smc_cdc_msg_send */
 
 	struct smc_host_cdc_msg	local_rx_ctrl;	/* filled during event_handl.
 						 * .prod cf. TCP rcv_nxt

diff --git a/net/smc/smc_close.c b/net/smc/smc_close.c
index 3c2e166b5d22..5201bc103bd8 100644
--- a/net/smc/smc_close.c
+++ b/net/smc/smc_close.c

@@ -208,7 +208,7 @@ int smc_close_active(struct smc_sock *smc)
 	case SMC_ACTIVE:
 		smc_close_stream_wait(smc, timeout);
 		release_sock(sk);
-		cancel_work_sync(&conn->tx_work);
+		cancel_delayed_work_sync(&conn->tx_work);
 		lock_sock(sk);
 		if (sk->sk_state == SMC_ACTIVE) {
 			/* send close request */

@@ -234,7 +234,7 @@ int smc_close_active(struct smc_sock *smc)
 		if (!smc_cdc_rxed_any_close(conn))
 			smc_close_stream_wait(smc, timeout);
 		release_sock(sk);
-		cancel_work_sync(&conn->tx_work);
+		cancel_delayed_work_sync(&conn->tx_work);
 		lock_sock(sk);
 		if (sk->sk_err != ECONNABORTED) {
 			/* confirm close from peer */

@@ -263,7 +263,9 @@ int smc_close_active(struct smc_sock *smc)
 		/* peer sending PeerConnectionClosed will cause transition */
 		break;
 	case SMC_PROCESSABORT:
-		cancel_work_sync(&conn->tx_work);
+		release_sock(sk);
+		cancel_delayed_work_sync(&conn->tx_work);
+		lock_sock(sk);
 		smc_close_abort(conn);
 		sk->sk_state = SMC_CLOSED;
 		smc_close_wait_tx_pends(smc);

@@ -425,7 +427,7 @@ int smc_close_shutdown_write(struct smc_sock *smc)
 	case SMC_ACTIVE:
 		smc_close_stream_wait(smc, timeout);
 		release_sock(sk);
-		cancel_work_sync(&conn->tx_work);
+		cancel_delayed_work_sync(&conn->tx_work);
 		lock_sock(sk);
 		/* send close wr request */
 		rc = smc_close_wr(conn);

@@ -439,7 +441,7 @@ int smc_close_shutdown_write(struct smc_sock *smc)
 		if (!smc_cdc_rxed_any_close(conn))
 			smc_close_stream_wait(smc, timeout);
 		release_sock(sk);
-		cancel_work_sync(&conn->tx_work);
+		cancel_delayed_work_sync(&conn->tx_work);
 		lock_sock(sk);
 		/* confirm close from peer */
 		rc = smc_close_wr(conn);

diff --git a/net/smc/smc_tx.c b/net/smc/smc_tx.c
index 3c656beb8820..3866573288dd 100644
--- a/net/smc/smc_tx.c
+++ b/net/smc/smc_tx.c

@@ -24,6 +24,8 @@
 #include "smc_cdc.h"
 #include "smc_tx.h"
 
+#define SMC_TX_WORK_DELAY	HZ
+
 /***************************** sndbuf producer *******************************/
 
 /* callback implementation for sk.sk_write_space()

@@ -406,7 +408,8 @@ int smc_tx_sndbuf_nonempty(struct smc_connection *conn)
 				goto out_unlock;
 			}
 			rc = 0;
-			schedule_work(&conn->tx_work);
+			schedule_delayed_work(&conn->tx_work,
+					      SMC_TX_WORK_DELAY);
 		}
 		goto out_unlock;
 	}

@@ -430,7 +433,7 @@ int smc_tx_sndbuf_nonempty(struct smc_connection *conn)
  */
 static void smc_tx_work(struct work_struct *work)
 {
-	struct smc_connection *conn = container_of(work,
+	struct smc_connection *conn = container_of(to_delayed_work(work),
 						   struct smc_connection,
 						   tx_work);
 	struct smc_sock *smc = container_of(conn, struct smc_sock, conn);

@@ -468,7 +471,8 @@ void smc_tx_consumer_update(struct smc_connection *conn)
 		if (!rc)
 			rc = smc_cdc_msg_send(conn, wr_buf, pend);
 		if (rc < 0) {
-			schedule_work(&conn->tx_work);
+			schedule_delayed_work(&conn->tx_work,
+					      SMC_TX_WORK_DELAY);
 			return;
 		}
 		smc_curs_write(&conn->rx_curs_confirmed,

@@ -487,6 +491,6 @@ void smc_tx_consumer_update(struct smc_connection *conn)
 void smc_tx_init(struct smc_sock *smc)
 {
 	smc->sk.sk_write_space = smc_tx_write_space;
-	INIT_WORK(&smc->conn.tx_work, smc_tx_work);
+	INIT_DELAYED_WORK(&smc->conn.tx_work, smc_tx_work);
 	spin_lock_init(&smc->conn.send_lock);
 }

-- 
2.13.5

Re: [PATCH net 0/9] net/smc: bug fixes 2017-09-20

From: David Miller <davem@davemloft.net>
Date: 2017-09-21 22:32:01

From: Ursula Braun <redacted>
Date: Thu, 21 Sep 2017 09:16:25 +0200

here is a collection of small smc-patches built for net fixing
smc problems in different areas.

Series applied, thanks.

`h`	back out one level
`j`	next message in thread
`k`	previous message in thread
`l`	drill in
`Esc`	close help / fold thread tree
`?`	toggle this help