[dpdk-dev] [PATCH 1/2] net/virtio: do not re-enter clean up routines

[Maxime Coquelin]

On 11/1/18 3:45 PM, Luca Boccassi wrote:
> On Mon, 2017-07-17 at 19:05 -0400, Charles (Chas) Williams wrote:
>> .dev_uninit calls .dev_stop and .dev_close.  The work that is done in
>> those routines doesn't need repeated.  Use started and opened to
>> track
>> the adapter's status.
>>
>> Signed-off-by: Chas Williams <ciwillia at brocade.com>
>> ---
>>   drivers/net/virtio/virtio_ethdev.c | 15 ++++++++++++---
>>   drivers/net/virtio/virtio_pci.h    |  4 +++-
>>   2 files changed, 15 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/net/virtio/virtio_ethdev.c
>> b/drivers/net/virtio/virtio_ethdev.c
>> index 00a3122..eff0545 100644
>> --- a/drivers/net/virtio/virtio_ethdev.c
>> +++ b/drivers/net/virtio/virtio_ethdev.c
>> @@ -609,6 +609,10 @@ virtio_dev_close(struct rte_eth_dev *dev)
>>   
>>   	PMD_INIT_LOG(DEBUG, "virtio_dev_close");
>>   
>> +	if (!hw->opened)
>> +		return;
>> +	hw->opened = false;
>> +
>>   	/* reset the NIC */
>>   	if (dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC)
>>   		VTPCI_OPS(hw)->set_config_irq(hw,
>> VIRTIO_MSI_NO_VECTOR);
>> @@ -1696,6 +1700,8 @@ virtio_dev_configure(struct rte_eth_dev *dev)
>>   			return -EBUSY;
>>   		}
>>   
>> +	hw->opened = true;
>> +
>>   	return 0;
>>   }
>>   
>> @@ -1759,7 +1765,7 @@ virtio_dev_start(struct rte_eth_dev *dev)
>>   		VIRTQUEUE_DUMP(txvq->vq);
>>   	}
>>   
>> -	hw->started = 1;
>> +	hw->started = true;
>>   
>>   	/* Initialize Link state */
>>   	virtio_dev_link_update(dev, 0);
>> @@ -1824,10 +1830,13 @@ virtio_dev_stop(struct rte_eth_dev *dev)
>>   
>>   	PMD_INIT_LOG(DEBUG, "stop");
>>   
>> +	if (!hw->started)
>> +		return;
>> +	hw->started = false;
>> +
>>   	if (intr_conf->lsc || intr_conf->rxq)
>>   		rte_intr_disable(dev->intr_handle);
>>   
>> -	hw->started = 0;
>>   	memset(&link, 0, sizeof(link));
>>   	virtio_dev_atomic_write_link_status(dev, &link);
>>   }
>> @@ -1844,7 +1853,7 @@ virtio_dev_link_update(struct rte_eth_dev *dev,
>> __rte_unused int wait_to_complet
>>   	link.link_duplex = ETH_LINK_FULL_DUPLEX;
>>   	link.link_speed  = SPEED_10G;
>>   
>> -	if (hw->started == 0) {
>> +	if (!hw->started) {
>>   		link.link_status = ETH_LINK_DOWN;
>>   	} else if (vtpci_with_feature(hw, VIRTIO_NET_F_STATUS)) {
>>   		PMD_INIT_LOG(DEBUG, "Get link status from hw");
>> diff --git a/drivers/net/virtio/virtio_pci.h
>> b/drivers/net/virtio/virtio_pci.h
>> index 18caebd..65bad2d 100644
>> --- a/drivers/net/virtio/virtio_pci.h
>> +++ b/drivers/net/virtio/virtio_pci.h
>> @@ -35,6 +35,7 @@
>>   #define _VIRTIO_PCI_H_
>>   
>>   #include <stdint.h>
>> +#include <stdbool.h>
>>   
>>   #include <rte_pci.h>
>>   #include <rte_ethdev.h>
>> @@ -253,7 +254,7 @@ struct virtio_hw {
>>   	uint64_t    req_guest_features;
>>   	uint64_t    guest_features;
>>   	uint32_t    max_queue_pairs;
>> -	uint16_t    started;
>> +	bool        started;
>>   	uint16_t	max_mtu;
>>   	uint16_t    vtnet_hdr_size;
>>   	uint8_t	    vlan_strip;
>> @@ -268,6 +269,7 @@ struct virtio_hw {
>>   	struct virtio_pci_common_cfg *common_cfg;
>>   	struct virtio_net_config *dev_cfg;
>>   	void	    *virtio_user_dev;
>> +	bool        opened;
>>   
>>   	struct virtqueue **vqs;
>>   };
> 
> Hi Maxime,
> 
> Any chance this could be reviewed and taken care of, please? It's been
> forgotten in the queue for a year, but we still use it.
> I have sent 2/2 separately so ignore that (I didn't notice this series
> was still pending review).
> 
> Thanks!
> 


Reviewed-by: Maxime Coquelin <maxime.coquelin at redhat.com>

I'll certainly have to rebase it, but it should be trivial.
And as discussed on IRC, it fixes a real issue, so will have to be
backported to stable. I'll take care of it.

Regards,
Maxime