On 9/28/2023 3:47 AM, Paolo Abeni wrote:

On Tue, 2023-09-19 at 15:27 -0700, Amritha Nambiar wrote:

quoted

Add the napi pointer in netdev queue for tracking the napi
instance for each queue. This achieves the queue<->napi mapping.

Signed-off-by: Amritha Nambiar <redacted>
Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
---
  include/linux/netdevice.h     |    5 +++++
  include/net/netdev_rx_queue.h |    2 ++
  net/core/dev.c                |   34 ++++++++++++++++++++++++++++++++++
  3 files changed, 41 insertions(+)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index db3d8429d50d..69e363918e4b 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h

@@ -657,6 +657,8 @@ struct netdev_queue {
  
  	unsigned long		state;
  
+	/* NAPI instance for the queue */
+	struct napi_struct      *napi;

This should be probably moved into the 'read-mostly' section, before
the '_xmit_lock' field,

Agree. Will fix.

quoted

  +/**
+ * netif_napi_add_queue - Associate queue with the napi
+ * @napi: NAPI context
+ * @queue_index: Index of queue
+ * @type: queue type as RX or TX
+ *
+ * Add queue with its corresponding napi context
+ */
+int netif_napi_add_queue(struct napi_struct *napi, unsigned int queue_index,
+			 enum netdev_queue_type type)

Very minor nit and others may have different opinion, but I feel like
this function name is misleading, since at this point both the rx and
tx queue should already exist. Perhaps 'netif_napi_link_queue' ?

Sounds right. Since, we are basically just setting the napi for the 
queue, and this function may probably be used by both 'queue-get' and 
'queue-set' commands in future, does 'netif_queue_set_napi' suit better ?

Cheers,

Paolo