[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [UNIKRAFT PATCH v4 05/12] plat/xen/drivers/net: Create netfront queues



Hello Costin,

I have a minor comments inline but the rest of it fine.

Reviewed-by: Sharan Santhanam <sharan.santhanam@xxxxxxxxx>


Thanks & Regards

Sharan

On 8/13/20 10:53 AM, Costin Lupu wrote:
We continue with the device configuration by retrieving the Xenstore
information regarding the number of queues and their associated event channels.
Netfront devices operate pairs of Rx/Tx queues and for notifications we can
either use a single event channel per pair or split event channels.

Signed-off-by: Costin Lupu <costin.lupu@xxxxxxxxx>
Signed-off-by: Razvan Cojocaru <razvan.cojocaru93@xxxxxxxxx>
---
  plat/xen/drivers/net/netfront.c    | 81 ++++++++++++++++++++++++++++++
  plat/xen/drivers/net/netfront.h    | 20 ++++++++
  plat/xen/drivers/net/netfront_xs.c | 18 ++++++-
  3 files changed, 117 insertions(+), 2 deletions(-)

diff --git a/plat/xen/drivers/net/netfront.c b/plat/xen/drivers/net/netfront.c
index 539e1cbc..ddbb4b70 100644
--- a/plat/xen/drivers/net/netfront.c
+++ b/plat/xen/drivers/net/netfront.c
@@ -49,6 +49,84 @@
static struct uk_alloc *drv_allocator; +static int netfront_rxtx_alloc(struct netfront_dev *nfdev,
+               const struct uk_netdev_conf *conf)
+{
+       int rc = 0;
+
+       if (conf->nb_tx_queues != conf->nb_rx_queues) {
+               uk_pr_err("Different number of queues not supported\n");
+               rc = -ENOTSUP;
+               goto err_free_txrx;
+       }
+
+       nfdev->max_queue_pairs =
+               MIN(nfdev->max_queue_pairs, conf->nb_tx_queues);

Should we change the max_queue_pair to the user input or retain the max value provided by the system and use separate variable in the structure to store this?

+
+       nfdev->txqs = uk_calloc(drv_allocator,
+               nfdev->max_queue_pairs, sizeof(*nfdev->txqs));
+       if (unlikely(!nfdev->txqs)) {
+               uk_pr_err("Failed to allocate memory for tx queues\n");
+               rc = -ENOMEM;
+               goto err_free_txrx;
+       }
+
+       nfdev->rxqs = uk_calloc(drv_allocator,
+               nfdev->max_queue_pairs, sizeof(*nfdev->rxqs));
+       if (unlikely(!nfdev->rxqs)) {
+               uk_pr_err("Failed to allocate memory for rx queues\n");
+               rc = -ENOMEM;
+               goto err_free_txrx;
+       }
+
+       return rc;
+
+err_free_txrx:
+       if (!nfdev->rxqs)
+               uk_free(drv_allocator, nfdev->rxqs);
+       if (!nfdev->txqs)
+               uk_free(drv_allocator, nfdev->txqs);
+
+       return rc;
+}
+
+static int netfront_configure(struct uk_netdev *n,
+               const struct uk_netdev_conf *conf)
+{
+       int rc;
+       struct netfront_dev *nfdev;
+
+       UK_ASSERT(n != NULL);
+       UK_ASSERT(conf != NULL);
+
+       nfdev = to_netfront_dev(n);
+
+       rc = netfront_rxtx_alloc(nfdev, conf);
+       if (rc != 0) {
+               uk_pr_err("Failed to allocate rx and tx rings %d\n", rc);
+               goto out;
+       }
+
+out:
+       return rc;
+}
+
+static void netfront_info_get(struct uk_netdev *n,
+               struct uk_netdev_info *dev_info)
+{
+       struct netfront_dev *nfdev;
+
+       UK_ASSERT(n != NULL);
+       UK_ASSERT(dev_info != NULL);
+
+       nfdev = to_netfront_dev(n);
+       dev_info->max_rx_queues = nfdev->max_queue_pairs;
+       dev_info->max_tx_queues = nfdev->max_queue_pairs;
+       dev_info->max_mtu = nfdev->mtu;
+       dev_info->nb_encap_tx = 0;
+       dev_info->nb_encap_rx = 0;
+}
+
  static const void *netfront_einfo_get(struct uk_netdev *n,
                enum uk_netdev_einfo_type einfo_type)
  {
@@ -100,6 +178,8 @@ static unsigned int netfront_promisc_get(struct uk_netdev 
*n)
  }
static const struct uk_netdev_ops netfront_ops = {
+       .configure = netfront_configure,
+       .info_get = netfront_info_get,
        .einfo_get = netfront_einfo_get,
        .hwaddr_get = netfront_mac_get,
        .mtu_get = netfront_mtu_get,
@@ -121,6 +201,7 @@ static int netfront_add_dev(struct xenbus_device *xendev)
nfdev->xendev = xendev;
        nfdev->mtu = ETH_PKT_PAYLOAD_LEN;
+       nfdev->max_queue_pairs = 1;
/* Xenbus initialization */
        rc = netfront_xb_init(nfdev, drv_allocator);
diff --git a/plat/xen/drivers/net/netfront.h b/plat/xen/drivers/net/netfront.h
index 0cc8230b..a811b092 100644
--- a/plat/xen/drivers/net/netfront.h
+++ b/plat/xen/drivers/net/netfront.h
@@ -38,6 +38,18 @@
#include <uk/netdev.h> +/**
+ * internal structure to represent the transmit queue.
+ */
+struct uk_netdev_tx_queue {
+};
+
+/**
+ * internal structure to represent the receive queue.
+ */
+struct uk_netdev_rx_queue {
+};
+
  struct xs_econf {
        char *ipv4addr;
        char *ipv4mask;
@@ -50,6 +62,14 @@ struct netfront_dev {
        /* Network device */
        struct uk_netdev netdev;
+ /* List of the Rx/Tx queues */
+       struct uk_netdev_tx_queue *txqs;
+       struct uk_netdev_rx_queue *rxqs;
+       /* Maximum number of queue pairs */
+       uint16_t  max_queue_pairs;
+       /* True if using split event channels */
+       bool split_evtchn;
+
        /* Configuration parameters */
        struct xs_econf econf;
diff --git a/plat/xen/drivers/net/netfront_xs.c b/plat/xen/drivers/net/netfront_xs.c
index 48f01b34..1bde44e0 100644
--- a/plat/xen/drivers/net/netfront_xs.c
+++ b/plat/xen/drivers/net/netfront_xs.c
@@ -122,7 +122,7 @@ out_err:
  int netfront_xb_init(struct netfront_dev *nfdev, struct uk_alloc *a)
  {
        struct xenbus_device *xendev;
-       char *mac_str, *p, *ip_str;
+       char *mac_str, *p, *ip_str, *int_str;
        int rc;
UK_ASSERT(nfdev != NULL);
@@ -173,7 +173,21 @@ int netfront_xb_init(struct netfront_dev *nfdev, struct 
uk_alloc *a)
                goto no_conf;
        free(ip_str);
- /* TODO spit event channels */
+       /* maximum queues number */
+       int_str = xs_read(XBT_NIL, xendev->otherend,
+               "multi-queue-max-queues");
+       if (!PTRISERR(int_str)) {
+               nfdev->max_queue_pairs = (uint16_t) strtoul(int_str, NULL, 10);
+               free(int_str);
+       }
+
+       /* spit event channels */
+       int_str = xs_read(XBT_NIL, xendev->otherend,
+               "feature-split-event-channels");
+       if (!PTRISERR(int_str)) {
+               nfdev->split_evtchn = (bool) strtoul(int_str, NULL, 10);
+               free(int_str);
+       }
/* TODO netmap */



 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.