[dpdk-dev] [PATCH] examples/kni: fix crash while handling userspace request

[dpdk-dev] [PATCH] examples/kni: fix crash while handling userspace request

[Ferruh Yigit]

When KNI interface receives RTE_KNI_REQ_CFG_NETWORK_IF request, it
stap/start the physical device which as a result of stop() can free all
the mbufs in its queue.
Meanwhile sample application continues to read from KNI interface queues
and push into device queues. This simultaneous access may cause a crash,
crash log can be found at defect description.

As a solution KNI sample application can do the proper synchronization,
and stop transfer between KNI interface and physical interface while
physical device stop/started.

Bugzilla ID: 116
Fixes: 3fc5ca2f6352 ("kni: initial import")
Cc: stable@dpdk.org

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
Cc: he.qiao17@zte.com.cn
I am not able to re-produce the defect can you please test with this
patch?

And another concern is performance affect of this patch, new check is
added into fast path. Cc'ed a few more people for comment:
Cc: Igor Ryzhov <iryzhov@nfware.com>
Cc: Phil Yang <phil.yang@arm.com>
Cc: Dan Gora <dg@adax.com>
---
 examples/kni/main.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/examples/kni/main.c b/examples/kni/main.c
index c468ffd19..a8ce07e2a 100644
--- a/examples/kni/main.c
+++ b/examples/kni/main.c
@@ -132,6 +132,7 @@ static int kni_config_network_interface(uint16_t port_id, uint8_t if_up);
 static int kni_config_mac_address(uint16_t port_id, uint8_t mac_addr[]);
 
 static rte_atomic32_t kni_stop = RTE_ATOMIC32_INIT(0);
+static rte_atomic32_t kni_pause = RTE_ATOMIC32_INIT(0);
 
 /* Print out statistics on packets handled */
 static void
@@ -276,6 +277,7 @@ main_loop(__rte_unused void *arg)
 {
 	uint16_t i;
 	int32_t f_stop;
+	int32_t f_pause;
 	const unsigned lcore_id = rte_lcore_id();
 	enum lcore_rxtx {
 		LCORE_NONE,
@@ -304,8 +306,11 @@ main_loop(__rte_unused void *arg)
 					kni_port_params_array[i]->port_id);
 		while (1) {
 			f_stop = rte_atomic32_read(&kni_stop);
+			f_pause = rte_atomic32_read(&kni_pause);
 			if (f_stop)
 				break;
+			if (f_pause)
+				continue;
 			kni_ingress(kni_port_params_array[i]);
 		}
 	} else if (flag == LCORE_TX) {
@@ -314,8 +319,11 @@ main_loop(__rte_unused void *arg)
 					kni_port_params_array[i]->port_id);
 		while (1) {
 			f_stop = rte_atomic32_read(&kni_stop);
+			f_pause = rte_atomic32_read(&kni_pause);
 			if (f_stop)
 				break;
+			if (f_pause)
+				continue;
 			kni_egress(kni_port_params_array[i]);
 		}
 	} else
@@ -807,12 +815,16 @@ kni_config_network_interface(uint16_t port_id, uint8_t if_up)
 	RTE_LOG(INFO, APP, "Configure network interface of %d %s\n",
 					port_id, if_up ? "up" : "down");
 
+	rte_atomic32_inc(&kni_pause);
+
 	if (if_up != 0) { /* Configure network interface up */
 		rte_eth_dev_stop(port_id);
 		ret = rte_eth_dev_start(port_id);
 	} else /* Configure network interface down */
 		rte_eth_dev_stop(port_id);
 
+	rte_atomic32_dec(&kni_pause);
+
 	if (ret < 0)
 		RTE_LOG(ERR, APP, "Failed to start port %d\n", port_id);
 
-- 
2.17.2

Re: [dpdk-dev] [PATCH] examples/kni: fix crash while handling userspace request

[Phil Yang (Arm Technology China)]

Hi Ferruh,

> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@intel.com>
> Sent: Wednesday, January 16, 2019 1:28 AM
> To: dev@dpdk.org
> Cc: stable@dpdk.org; he.qiao17@zte.com.cn; Igor Ryzhov
> <iryzhov@nfware.com>; Phil Yang (Arm Technology China)
> <Phil.Yang@arm.com>; Dan Gora <dg@adax.com>
> Subject: [PATCH] examples/kni: fix crash while handling userspace request
>
> When KNI interface receives RTE_KNI_REQ_CFG_NETWORK_IF request, it
> stap/start the physical device which as a result of stop() can free all the mbufs in
> its queue.
> Meanwhile sample application continues to read from KNI interface queues and
> push into device queues. This simultaneous access may cause a crash, crash log
> can be found at defect description.
>
> As a solution KNI sample application can do the proper synchronization, and
> stop transfer between KNI interface and physical interface while physical device
> stop/started.
>
> Bugzilla ID: 116
> Fixes: 3fc5ca2f6352 ("kni: initial import")
> Cc: stable@dpdk.org
>
> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> ---
> Cc: he.qiao17@zte.com.cn
> I am not able to re-produce the defect can you please test with this patch?
>
> And another concern is performance affect of this patch, new check is added
> into fast path. Cc'ed a few more people for comment:
> Cc: Igor Ryzhov <iryzhov@nfware.com>
> Cc: Phil Yang <phil.yang@arm.com>
> Cc: Dan Gora <dg@adax.com>
> ---
>  examples/kni/main.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
>
> diff --git a/examples/kni/main.c b/examples/kni/main.c index
> c468ffd19..a8ce07e2a 100644
> --- a/examples/kni/main.c
> +++ b/examples/kni/main.c
> @@ -132,6 +132,7 @@ static int kni_config_network_interface(uint16_t port_id,
> uint8_t if_up);  static int kni_config_mac_address(uint16_t port_id, uint8_t
> mac_addr[]);
>
>  static rte_atomic32_t kni_stop = RTE_ATOMIC32_INIT(0);
> +static rte_atomic32_t kni_pause = RTE_ATOMIC32_INIT(0);
>
>  /* Print out statistics on packets handled */  static void @@ -276,6 +277,7 @@
> main_loop(__rte_unused void *arg)  {
>  uint16_t i;
>  int32_t f_stop;
> +int32_t f_pause;
>  const unsigned lcore_id = rte_lcore_id();
>  enum lcore_rxtx {
>  LCORE_NONE,
> @@ -304,8 +306,11 @@ main_loop(__rte_unused void *arg)
>  kni_port_params_array[i]->port_id);
>  while (1) {
>  f_stop = rte_atomic32_read(&kni_stop);
> +f_pause = rte_atomic32_read(&kni_pause);
>  if (f_stop)
>  break;
> +if (f_pause)
> +continue;
>  kni_ingress(kni_port_params_array[i]);
>  }
>  } else if (flag == LCORE_TX) {
> @@ -314,8 +319,11 @@ main_loop(__rte_unused void *arg)
>  kni_port_params_array[i]->port_id);
>  while (1) {
>  f_stop = rte_atomic32_read(&kni_stop);
> +f_pause = rte_atomic32_read(&kni_pause);
>  if (f_stop)
>  break;
> +if (f_pause)
> +continue;
>  kni_egress(kni_port_params_array[i]);
>  }
>  } else
> @@ -807,12 +815,16 @@ kni_config_network_interface(uint16_t port_id,
> uint8_t if_up)
>  RTE_LOG(INFO, APP, "Configure network interface of %d %s\n",
>  port_id, if_up ? "up" : "down");
>
> +rte_atomic32_inc(&kni_pause);
> +
>  if (if_up != 0) { /* Configure network interface up */
>  rte_eth_dev_stop(port_id);
>  ret = rte_eth_dev_start(port_id);
>  } else /* Configure network interface down */
>  rte_eth_dev_stop(port_id);
>
> +rte_atomic32_dec(&kni_pause);
> +
>  if (ret < 0)
>  RTE_LOG(ERR, APP, "Failed to start port %d\n", port_id);
>
> --
> 2.17.2

Acked-by: Phil Yang <phil.yang@arm.com>

Thanks,
Phil Yang
IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.

Re: [dpdk-dev] [PATCH] examples/kni: fix crash while handling userspace request

[Thomas Monjalon]

> > When KNI interface receives RTE_KNI_REQ_CFG_NETWORK_IF request, it
> > stap/start the physical device which as a result of stop() can free all the mbufs in
> > its queue.
> > Meanwhile sample application continues to read from KNI interface queues and
> > push into device queues. This simultaneous access may cause a crash, crash log
> > can be found at defect description.
> >
> > As a solution KNI sample application can do the proper synchronization, and
> > stop transfer between KNI interface and physical interface while physical device
> > stop/started.
> >
> > Bugzilla ID: 116
> > Fixes: 3fc5ca2f6352 ("kni: initial import")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> 
> Acked-by: Phil Yang <phil.yang@arm.com>

Applied, thanks