net/virtio: support power monitor
authorMiao Li <miao.li@intel.com>
Mon, 25 Oct 2021 14:47:21 +0000 (14:47 +0000)
committerMaxime Coquelin <maxime.coquelin@redhat.com>
Fri, 29 Oct 2021 10:32:29 +0000 (12:32 +0200)
According to current semantics of power monitor, this commit adds a
callback function to decide whether aborts the sleep by checking
current value against the expected value and virtio_get_monitor_addr
to provide address to monitor. When no packet come in, the value of
address will not be changed and the running core will sleep. Once
packets arrive, the value of address will be changed and the running
core will wakeup.

Signed-off-by: Miao Li <miao.li@intel.com>
Reviewed-by: Chenbo Xia <chenbo.xia@intel.com>
Acked-by: David Hunt <david.hunt@intel.com>
doc/guides/rel_notes/release_21_11.rst
drivers/net/virtio/virtio_ethdev.c

index da3da47..3682d54 100644 (file)
@@ -205,6 +205,7 @@ New Features
     with the capability for the application to configure the hash key,
     the RETA and the hash types.
     Virtio hash reporting is yet to be added.
+  * Added power monitor support in virtio PMD.
 
 * **Updated Marvell cnxk crypto PMD.**
 
index 0de3406..4570bec 100644 (file)
@@ -84,6 +84,8 @@ static int virtio_mac_addr_set(struct rte_eth_dev *dev,
                                struct rte_ether_addr *mac_addr);
 
 static int virtio_intr_disable(struct rte_eth_dev *dev);
+static int virtio_get_monitor_addr(void *rx_queue,
+                               struct rte_power_monitor_cond *pmc);
 
 static int virtio_dev_queue_stats_mapping_set(
        struct rte_eth_dev *eth_dev,
@@ -1051,6 +1053,7 @@ static const struct eth_dev_ops virtio_eth_dev_ops = {
        .mac_addr_add            = virtio_mac_addr_add,
        .mac_addr_remove         = virtio_mac_addr_remove,
        .mac_addr_set            = virtio_mac_addr_set,
+       .get_monitor_addr        = virtio_get_monitor_addr,
 };
 
 /*
@@ -1382,6 +1385,59 @@ virtio_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr)
        return 0;
 }
 
+#define CLB_VAL_IDX 0
+#define CLB_MSK_IDX 1
+#define CLB_MATCH_IDX 2
+static int
+virtio_monitor_callback(const uint64_t value,
+               const uint64_t opaque[RTE_POWER_MONITOR_OPAQUE_SZ])
+{
+       const uint64_t m = opaque[CLB_MSK_IDX];
+       const uint64_t v = opaque[CLB_VAL_IDX];
+       const uint64_t c = opaque[CLB_MATCH_IDX];
+
+       if (c)
+               return (value & m) == v ? -1 : 0;
+       else
+               return (value & m) == v ? 0 : -1;
+}
+
+static int
+virtio_get_monitor_addr(void *rx_queue, struct rte_power_monitor_cond *pmc)
+{
+       struct virtnet_rx *rxvq = rx_queue;
+       struct virtqueue *vq = virtnet_rxq_to_vq(rxvq);
+       struct virtio_hw *hw;
+
+       if (vq == NULL)
+               return -EINVAL;
+
+       hw = vq->hw;
+       if (virtio_with_packed_queue(hw)) {
+               struct vring_packed_desc *desc;
+               desc = vq->vq_packed.ring.desc;
+               pmc->addr = &desc[vq->vq_used_cons_idx].flags;
+               if (vq->vq_packed.used_wrap_counter)
+                       pmc->opaque[CLB_VAL_IDX] =
+                                               VRING_PACKED_DESC_F_AVAIL_USED;
+               else
+                       pmc->opaque[CLB_VAL_IDX] = 0;
+               pmc->opaque[CLB_MSK_IDX] = VRING_PACKED_DESC_F_AVAIL_USED;
+               pmc->opaque[CLB_MATCH_IDX] = 1;
+               pmc->size = sizeof(desc[vq->vq_used_cons_idx].flags);
+       } else {
+               pmc->addr = &vq->vq_split.ring.used->idx;
+               pmc->opaque[CLB_VAL_IDX] = vq->vq_used_cons_idx
+                                       & (vq->vq_nentries - 1);
+               pmc->opaque[CLB_MSK_IDX] = vq->vq_nentries - 1;
+               pmc->opaque[CLB_MATCH_IDX] = 0;
+               pmc->size = sizeof(vq->vq_split.ring.used->idx);
+       }
+       pmc->fn = virtio_monitor_callback;
+
+       return 0;
+}
+
 static int
 virtio_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 {