* [dpdk-dev] [PATCHv2 1/4] e1000: do not release queue on alloc error
2014-06-09 17:26 [dpdk-dev] [PATCHv2 0/4] fix for 2 consecutive rte_eth_dev_start() can cause a SIGSEGV Konstantin Ananyev
@ 2014-06-09 17:26 ` Konstantin Ananyev
2014-06-09 17:26 ` [dpdk-dev] [PATCHv2 2/4] igb/ixgbe: reset queue pointers after releasing Konstantin Ananyev
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Konstantin Ananyev @ 2014-06-09 17:26 UTC (permalink / raw)
To: dev, dev
If igb_alloc_rx_queue_mbufs() would fail to allocate an mbuf for RX queue,
it calls igb_rx_queue_release(rxq).
That causes rxq to be silently freed, without updating
dev->data->rx_queues[].
So any further reference to it will trigger the SIGSEGV.
Same thing in em PMD too.
To fix: igb_alloc_rx_queue_mbufs() should just return an error to the
caller and let upper layer to deal with the probem.
That's what ixgbe PMD is doing right now.
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
---
lib/librte_pmd_e1000/em_rxtx.c | 1 -
lib/librte_pmd_e1000/igb_rxtx.c | 1 -
2 files changed, 2 deletions(-)
diff --git a/lib/librte_pmd_e1000/em_rxtx.c b/lib/librte_pmd_e1000/em_rxtx.c
index 4f98a3f..5ed4def 100644
--- a/lib/librte_pmd_e1000/em_rxtx.c
+++ b/lib/librte_pmd_e1000/em_rxtx.c
@@ -1580,7 +1580,6 @@ em_alloc_rx_queue_mbufs(struct em_rx_queue *rxq)
if (mbuf == NULL) {
PMD_INIT_LOG(ERR, "RX mbuf alloc failed "
"queue_id=%hu\n", rxq->queue_id);
- em_rx_queue_release(rxq);
return (-ENOMEM);
}
diff --git a/lib/librte_pmd_e1000/igb_rxtx.c b/lib/librte_pmd_e1000/igb_rxtx.c
index ae53428..e35649f 100644
--- a/lib/librte_pmd_e1000/igb_rxtx.c
+++ b/lib/librte_pmd_e1000/igb_rxtx.c
@@ -1818,7 +1818,6 @@ igb_alloc_rx_queue_mbufs(struct igb_rx_queue *rxq)
if (mbuf == NULL) {
PMD_INIT_LOG(ERR, "RX mbuf alloc failed "
"queue_id=%hu\n", rxq->queue_id);
- igb_rx_queue_release(rxq);
return (-ENOMEM);
}
dma_addr =
--
1.8.3.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* [dpdk-dev] [PATCHv2 2/4] igb/ixgbe: reset queue pointers after releasing
2014-06-09 17:26 [dpdk-dev] [PATCHv2 0/4] fix for 2 consecutive rte_eth_dev_start() can cause a SIGSEGV Konstantin Ananyev
2014-06-09 17:26 ` [dpdk-dev] [PATCHv2 1/4] e1000: do not release queue on alloc error Konstantin Ananyev
@ 2014-06-09 17:26 ` Konstantin Ananyev
2014-06-09 17:26 ` [dpdk-dev] [PATCHv2 3/4] ethdev: prevent from starting/stopping already started/stopped device Konstantin Ananyev
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Konstantin Ananyev @ 2014-06-09 17:26 UTC (permalink / raw)
To: dev, dev
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
---
lib/librte_pmd_e1000/igb_rxtx.c | 4 +++-
lib/librte_pmd_ixgbe/ixgbe_rxtx.c | 8 ++++++--
2 files changed, 9 insertions(+), 3 deletions(-)
diff --git a/lib/librte_pmd_e1000/igb_rxtx.c b/lib/librte_pmd_e1000/igb_rxtx.c
index e35649f..712086a 100644
--- a/lib/librte_pmd_e1000/igb_rxtx.c
+++ b/lib/librte_pmd_e1000/igb_rxtx.c
@@ -1212,8 +1212,10 @@ eth_igb_tx_queue_setup(struct rte_eth_dev *dev,
"the TX WTHRESH value to 4, 8, or 16.\n");
/* Free memory prior to re-allocation if needed */
- if (dev->data->tx_queues[queue_idx] != NULL)
+ if (dev->data->tx_queues[queue_idx] != NULL) {
igb_tx_queue_release(dev->data->tx_queues[queue_idx]);
+ dev->data->tx_queues[queue_idx] = NULL;
+ }
/* First allocate the tx queue data structure */
txq = rte_zmalloc("ethdev TX queue", sizeof(struct igb_tx_queue),
diff --git a/lib/librte_pmd_ixgbe/ixgbe_rxtx.c b/lib/librte_pmd_ixgbe/ixgbe_rxtx.c
index dc79c4b..99584b0 100644
--- a/lib/librte_pmd_ixgbe/ixgbe_rxtx.c
+++ b/lib/librte_pmd_ixgbe/ixgbe_rxtx.c
@@ -1802,8 +1802,10 @@ ixgbe_dev_tx_queue_setup(struct rte_eth_dev *dev,
}
/* Free memory prior to re-allocation if needed... */
- if (dev->data->tx_queues[queue_idx] != NULL)
+ if (dev->data->tx_queues[queue_idx] != NULL) {
ixgbe_tx_queue_release(dev->data->tx_queues[queue_idx]);
+ dev->data->tx_queues[queue_idx] = NULL;
+ }
/* First allocate the tx queue data structure */
txq = rte_zmalloc_socket("ethdev TX queue", sizeof(struct igb_tx_queue),
@@ -2061,8 +2063,10 @@ ixgbe_dev_rx_queue_setup(struct rte_eth_dev *dev,
}
/* Free memory prior to re-allocation if needed... */
- if (dev->data->rx_queues[queue_idx] != NULL)
+ if (dev->data->rx_queues[queue_idx] != NULL) {
ixgbe_rx_queue_release(dev->data->rx_queues[queue_idx]);
+ dev->data->rx_queues[queue_idx] = NULL;
+ }
/* First allocate the rx queue data structure */
rxq = rte_zmalloc_socket("ethdev RX queue", sizeof(struct igb_rx_queue),
--
1.8.3.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* [dpdk-dev] [PATCHv2 3/4] ethdev: prevent from starting/stopping already started/stopped device
2014-06-09 17:26 [dpdk-dev] [PATCHv2 0/4] fix for 2 consecutive rte_eth_dev_start() can cause a SIGSEGV Konstantin Ananyev
2014-06-09 17:26 ` [dpdk-dev] [PATCHv2 1/4] e1000: do not release queue on alloc error Konstantin Ananyev
2014-06-09 17:26 ` [dpdk-dev] [PATCHv2 2/4] igb/ixgbe: reset queue pointers after releasing Konstantin Ananyev
@ 2014-06-09 17:26 ` Konstantin Ananyev
2014-06-09 17:26 ` [dpdk-dev] [PATCHv2 4/4] ethdev: fix compiler warning on PMD_DEBUG_TRACE formats Konstantin Ananyev
2014-06-11 10:06 ` [dpdk-dev] [PATCHv2 0/4] fix for 2 consecutive rte_eth_dev_start() can cause a SIGSEGV Thomas Monjalon
4 siblings, 0 replies; 6+ messages in thread
From: Konstantin Ananyev @ 2014-06-09 17:26 UTC (permalink / raw)
To: dev, dev
From: Konstantin Ananyev <konstantin.ananyev@intel.com>
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
---
lib/librte_ether/rte_ethdev.c | 16 ++++++++++++++++
1 file changed, 16 insertions(+)
diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
index 11e877b..47bcee1 100644
--- a/lib/librte_ether/rte_ethdev.c
+++ b/lib/librte_ether/rte_ethdev.c
@@ -764,6 +764,14 @@ rte_eth_dev_start(uint8_t port_id)
dev = &rte_eth_devices[port_id];
FUNC_PTR_OR_ERR_RET(*dev->dev_ops->dev_start, -ENOTSUP);
+
+ if (dev->data->dev_started != 0) {
+ PMD_DEBUG_TRACE("Device with port_id=%" PRIu8
+ " already started\n",
+ port_id);
+ return (0);
+ }
+
diag = (*dev->dev_ops->dev_start)(dev);
if (diag == 0)
dev->data->dev_started = 1;
@@ -791,6 +799,14 @@ rte_eth_dev_stop(uint8_t port_id)
dev = &rte_eth_devices[port_id];
FUNC_PTR_OR_RET(*dev->dev_ops->dev_stop);
+
+ if (dev->data->dev_started == 0) {
+ PMD_DEBUG_TRACE("Device with port_id=%" PRIu8
+ " already stopped\n",
+ port_id);
+ return;
+ }
+
dev->data->dev_started = 0;
(*dev->dev_ops->dev_stop)(dev);
}
--
1.8.3.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* [dpdk-dev] [PATCHv2 4/4] ethdev: fix compiler warning on PMD_DEBUG_TRACE formats
2014-06-09 17:26 [dpdk-dev] [PATCHv2 0/4] fix for 2 consecutive rte_eth_dev_start() can cause a SIGSEGV Konstantin Ananyev
` (2 preceding siblings ...)
2014-06-09 17:26 ` [dpdk-dev] [PATCHv2 3/4] ethdev: prevent from starting/stopping already started/stopped device Konstantin Ananyev
@ 2014-06-09 17:26 ` Konstantin Ananyev
2014-06-11 10:06 ` [dpdk-dev] [PATCHv2 0/4] fix for 2 consecutive rte_eth_dev_start() can cause a SIGSEGV Thomas Monjalon
4 siblings, 0 replies; 6+ messages in thread
From: Konstantin Ananyev @ 2014-06-09 17:26 UTC (permalink / raw)
To: dev, dev
icc 12.1 complains about RTE_LOG() format:
"argument is incompatible with corresponding format string conversion"
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
---
lib/librte_ether/rte_ethdev.c | 13 ++++++++-----
1 file changed, 8 insertions(+), 5 deletions(-)
diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
index 47bcee1..301ff61 100644
--- a/lib/librte_ether/rte_ethdev.c
+++ b/lib/librte_ether/rte_ethdev.c
@@ -447,7 +447,8 @@ rte_eth_dev_check_mq_mode(uint8_t port_id, uint16_t nb_rx_q, uint16_t nb_tx_q,
(dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB_RSS) ||
(dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB)) {
/* SRIOV only works in VMDq enable mode */
- PMD_DEBUG_TRACE("ethdev port_id=%hhu SRIOV active, "
+ PMD_DEBUG_TRACE("ethdev port_id=%" PRIu8
+ " SRIOV active, "
"wrong VMDQ mq_mode rx %u tx %u\n",
port_id,
dev_conf->rxmode.mq_mode,
@@ -460,7 +461,8 @@ rte_eth_dev_check_mq_mode(uint8_t port_id, uint16_t nb_rx_q, uint16_t nb_tx_q,
case ETH_MQ_RX_VMDQ_DCB:
case ETH_MQ_RX_VMDQ_DCB_RSS:
/* DCB/RSS VMDQ in SRIOV mode, not implement yet */
- PMD_DEBUG_TRACE("ethdev port_id=%hhu SRIOV active, "
+ PMD_DEBUG_TRACE("ethdev port_id=%" PRIu8
+ " SRIOV active, "
"unsupported VMDQ mq_mode rx %u\n",
port_id, dev_conf->rxmode.mq_mode);
return (-EINVAL);
@@ -475,7 +477,8 @@ rte_eth_dev_check_mq_mode(uint8_t port_id, uint16_t nb_rx_q, uint16_t nb_tx_q,
switch (dev_conf->txmode.mq_mode) {
case ETH_MQ_TX_VMDQ_DCB:
/* DCB VMDQ in SRIOV mode, not implement yet */
- PMD_DEBUG_TRACE("ethdev port_id=%hhu SRIOV active, "
+ PMD_DEBUG_TRACE("ethdev port_id=%" PRIu8
+ " SRIOV active, "
"unsupported VMDQ mq_mode tx %u\n",
port_id, dev_conf->txmode.mq_mode);
return (-EINVAL);
@@ -758,7 +761,7 @@ rte_eth_dev_start(uint8_t port_id)
PROC_PRIMARY_OR_ERR_RET(-E_RTE_SECONDARY);
if (port_id >= nb_ports) {
- PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
+ PMD_DEBUG_TRACE("Invalid port_id=%" PRIu8 "\n", port_id);
return (-EINVAL);
}
dev = &rte_eth_devices[port_id];
@@ -793,7 +796,7 @@ rte_eth_dev_stop(uint8_t port_id)
PROC_PRIMARY_OR_RET();
if (port_id >= nb_ports) {
- PMD_DEBUG_TRACE("Invalid port_id=%d\n", port_id);
+ PMD_DEBUG_TRACE("Invalid port_id=%" PRIu8 "\n", port_id);
return;
}
dev = &rte_eth_devices[port_id];
--
1.8.3.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [dpdk-dev] [PATCHv2 0/4] fix for 2 consecutive rte_eth_dev_start() can cause a SIGSEGV
2014-06-09 17:26 [dpdk-dev] [PATCHv2 0/4] fix for 2 consecutive rte_eth_dev_start() can cause a SIGSEGV Konstantin Ananyev
` (3 preceding siblings ...)
2014-06-09 17:26 ` [dpdk-dev] [PATCHv2 4/4] ethdev: fix compiler warning on PMD_DEBUG_TRACE formats Konstantin Ananyev
@ 2014-06-11 10:06 ` Thomas Monjalon
4 siblings, 0 replies; 6+ messages in thread
From: Thomas Monjalon @ 2014-06-11 10:06 UTC (permalink / raw)
To: Konstantin Ananyev; +Cc: dev
> Konstantin Ananyev (4):
> e1000: do not release queue on alloc error
> igb/ixgbe: reset queue pointers after releasing
> ethdev: fix compiler warning on PMD_DEBUG_TRACE formats
> ethdev: prevent from starting/stopping already started/stopped
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Applied for version 1.7.0.
Thanks
--
Thomas
^ permalink raw reply [flat|nested] 6+ messages in thread