Re: [PATCH 1/3] igb: Link IRQs to NAPI instances

[Joe Damato <jdamato@fastly.com>]

On Mon, Feb 10, 2025 at 10:25:47AM -0800, Joe Damato wrote:
> On Mon, Feb 10, 2025 at 10:19:35AM +0100, Kurt Kanzenbach wrote:
> > Link IRQs to NAPI instances via netdev-genl API. This allows users to query
> > that information via netlink:
> > 
> > |$ ./tools/net/ynl/pyynl/cli.py --spec Documentation/netlink/specs/netdev.yaml \
> > |                               --dump napi-get --json='{"ifindex": 2}'
> > |[{'defer-hard-irqs': 0,
> > |  'gro-flush-timeout': 0,
> > |  'id': 8204,
> > |  'ifindex': 2,
> > |  'irq': 127,
> > |  'irq-suspend-timeout': 0},
> > | {'defer-hard-irqs': 0,
> > |  'gro-flush-timeout': 0,
> > |  'id': 8203,
> > |  'ifindex': 2,
> > |  'irq': 126,
> > |  'irq-suspend-timeout': 0},
> > | {'defer-hard-irqs': 0,
> > |  'gro-flush-timeout': 0,
> > |  'id': 8202,
> > |  'ifindex': 2,
> > |  'irq': 125,
> > |  'irq-suspend-timeout': 0},
> > | {'defer-hard-irqs': 0,
> > |  'gro-flush-timeout': 0,
> > |  'id': 8201,
> > |  'ifindex': 2,
> > |  'irq': 124,
> > |  'irq-suspend-timeout': 0}]
> > |$ cat /proc/interrupts | grep enp2s0
> > |123:          0          1 IR-PCI-MSIX-0000:02:00.0   0-edge      enp2s0
> > |124:          0          7 IR-PCI-MSIX-0000:02:00.0   1-edge      enp2s0-TxRx-0
> > |125:          0          0 IR-PCI-MSIX-0000:02:00.0   2-edge      enp2s0-TxRx-1
> > |126:          0          5 IR-PCI-MSIX-0000:02:00.0   3-edge      enp2s0-TxRx-2
> > |127:          0          0 IR-PCI-MSIX-0000:02:00.0   4-edge      enp2s0-TxRx-3
> > 
> > Signed-off-by: Kurt Kanzenbach <kurt@linutronix.de>
> > ---
> >  drivers/net/ethernet/intel/igb/igb_main.c | 3 +++
> >  1 file changed, 3 insertions(+)
> > 
> > diff --git a/drivers/net/ethernet/intel/igb/igb_main.c b/drivers/net/ethernet/intel/igb/igb_main.c
> > index d368b753a4675d01b5dfa50dee4cd218e6a5e14b..d4128d19cc08f62f95682069bb5ed9b8bbbf10cb 100644
> > --- a/drivers/net/ethernet/intel/igb/igb_main.c
> > +++ b/drivers/net/ethernet/intel/igb/igb_main.c
> > @@ -947,6 +947,9 @@ static int igb_request_msix(struct igb_adapter *adapter)
> >  				  q_vector);
> >  		if (err)
> >  			goto err_free;
> > +
> > +		netif_napi_set_irq(&q_vector->napi,
> > +				   adapter->msix_entries[vector].vector);
> >  	}
> 
> As far as I can tell, all paths that lead here hold RTNL:

A nit on my own comment, netif_napi_set_irq doesn't ASSERT_RTNL (but
does hold net_device->lock); its the other functions in patch 2 that
ASSERT_RTNL.

My reviewed-by stands; just wanted to correct myself.