On 7/28/2023 9:05 PM, Stephen Hemminger wrote:

On Fri, 28 Jul 2023 17:47:28 -0700
Amritha Nambiar [off-list ref] wrote:

quoted

Add support to associate the interrupt vector number for a
NAPI instance.

Signed-off-by: Amritha Nambiar <redacted>
---
  drivers/net/ethernet/intel/ice/ice_lib.c |    3 +++
  include/linux/netdevice.h                |    6 ++++++
  net/core/dev.c                           |    1 +
  net/core/netdev-genl.c                   |    4 ++++
  4 files changed, 14 insertions(+)

diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 171177db8fb4..1ebd293ca7de 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c

@@ -2975,6 +2975,9 @@ int ice_q_vector_add_napi_queues(struct ice_q_vector *q_vector)
  			return ret;
  	}
  
+	/* Also set the interrupt number for the NAPI */
+	napi_set_irq(&q_vector->napi, q_vector->irq.virq);
+
  	return ret;
  }

Doing this for only one device seems like a potential problem.

For devices that does not call napi_set_irq(), irq will be initialized 
to -1 as part of netif_napi_add_weight().

Also, there are some weird devices where there may not be a 1:1:1 mapping
between IRQ, NAPI, and netdev.

IIUC, there's a 1:1 mapping between IRQ and NAPI, and need not be mapped 
:1 with netdev.