* [PATCH net 0/9] net/smc: bug fixes 2017-09-20
@ 2017-09-21 7:16 Ursula Braun
2017-09-21 7:16 ` [PATCH net 2/9] net/smc: add receive timeout check Ursula Braun
` (4 more replies)
0 siblings, 5 replies; 6+ messages in thread
From: Ursula Braun @ 2017-09-21 7:16 UTC (permalink / raw)
To: davem
Cc: netdev, linux-rdma, linux-s390, jwi, schwidefsky, heiko.carstens,
raspl, ubraun
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
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH net 2/9] net/smc: add receive timeout check
2017-09-21 7:16 [PATCH net 0/9] net/smc: bug fixes 2017-09-20 Ursula Braun
@ 2017-09-21 7:16 ` Ursula Braun
2017-09-21 7:16 ` [PATCH net 3/9] net/smc: take RCU read lock for routing cache lookup Ursula Braun
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Ursula Braun @ 2017-09-21 7:16 UTC (permalink / raw)
To: davem
Cc: netdev, linux-rdma, linux-s390, jwi, schwidefsky, heiko.carstens,
raspl, ubraun
From: Hans Wippel <hwippel@linux.vnet.ibm.com>
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 <hwippel@linux.vnet.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
---
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
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH net 3/9] net/smc: take RCU read lock for routing cache lookup
2017-09-21 7:16 [PATCH net 0/9] net/smc: bug fixes 2017-09-20 Ursula Braun
2017-09-21 7:16 ` [PATCH net 2/9] net/smc: add receive timeout check Ursula Braun
@ 2017-09-21 7:16 ` Ursula Braun
2017-09-21 7:16 ` [PATCH net 4/9] net/smc: adjust net_device refcount Ursula Braun
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Ursula Braun @ 2017-09-21 7:16 UTC (permalink / raw)
To: davem
Cc: netdev, linux-rdma, linux-s390, jwi, schwidefsky, heiko.carstens,
raspl, ubraun
smc_netinfo_by_tcpsk() looks up the routing cache. Such a lookup requires
protection by an RCU read lock.
Signed-off-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
---
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
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH net 4/9] net/smc: adjust net_device refcount
2017-09-21 7:16 [PATCH net 0/9] net/smc: bug fixes 2017-09-20 Ursula Braun
2017-09-21 7:16 ` [PATCH net 2/9] net/smc: add receive timeout check Ursula Braun
2017-09-21 7:16 ` [PATCH net 3/9] net/smc: take RCU read lock for routing cache lookup Ursula Braun
@ 2017-09-21 7:16 ` Ursula Braun
2017-09-21 7:16 ` [PATCH net 8/9] net/smc: introduce a delay Ursula Braun
2017-09-21 22:31 ` [PATCH net 0/9] net/smc: bug fixes 2017-09-20 David Miller
4 siblings, 0 replies; 6+ messages in thread
From: Ursula Braun @ 2017-09-21 7:16 UTC (permalink / raw)
To: davem
Cc: netdev, linux-rdma, linux-s390, jwi, schwidefsky, heiko.carstens,
raspl, ubraun
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 <ubraun@linux.vnet.ibm.com>
---
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
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH net 8/9] net/smc: introduce a delay
2017-09-21 7:16 [PATCH net 0/9] net/smc: bug fixes 2017-09-20 Ursula Braun
` (2 preceding siblings ...)
2017-09-21 7:16 ` [PATCH net 4/9] net/smc: adjust net_device refcount Ursula Braun
@ 2017-09-21 7:16 ` Ursula Braun
2017-09-21 22:31 ` [PATCH net 0/9] net/smc: bug fixes 2017-09-20 David Miller
4 siblings, 0 replies; 6+ messages in thread
From: Ursula Braun @ 2017-09-21 7:16 UTC (permalink / raw)
To: davem
Cc: netdev, linux-rdma, linux-s390, jwi, schwidefsky, heiko.carstens,
raspl, ubraun
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 <ubraun@linux.vnet.ibm.com>
---
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
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH net 0/9] net/smc: bug fixes 2017-09-20
2017-09-21 7:16 [PATCH net 0/9] net/smc: bug fixes 2017-09-20 Ursula Braun
` (3 preceding siblings ...)
2017-09-21 7:16 ` [PATCH net 8/9] net/smc: introduce a delay Ursula Braun
@ 2017-09-21 22:31 ` David Miller
4 siblings, 0 replies; 6+ messages in thread
From: David Miller @ 2017-09-21 22:31 UTC (permalink / raw)
To: ubraun
Cc: netdev, linux-rdma, linux-s390, jwi, schwidefsky, heiko.carstens,
raspl
From: Ursula Braun <ubraun@linux.vnet.ibm.com>
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.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2017-09-21 22:31 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2017-09-21 7:16 [PATCH net 0/9] net/smc: bug fixes 2017-09-20 Ursula Braun
2017-09-21 7:16 ` [PATCH net 2/9] net/smc: add receive timeout check Ursula Braun
2017-09-21 7:16 ` [PATCH net 3/9] net/smc: take RCU read lock for routing cache lookup Ursula Braun
2017-09-21 7:16 ` [PATCH net 4/9] net/smc: adjust net_device refcount Ursula Braun
2017-09-21 7:16 ` [PATCH net 8/9] net/smc: introduce a delay Ursula Braun
2017-09-21 22:31 ` [PATCH net 0/9] net/smc: bug fixes 2017-09-20 David Miller
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox