Re: [dpdk-dev] [PATCH] net/e1000: enable MSI-x for linkup and other events in 82574

[Luca Boccassi <bluca@debian.org>]

On Tue, 2018-10-09 at 10:16 +0000, Zhao1, Wei wrote:
> Hi,  Luca Boccassi
> 
>    A question for this feature, in em code, function
> eth_em_interrupt_handler() we have no process for handle other cause
> interrupt.
> So, why do we enable this OTHER cause interrupt in DPDK?
> You mean some application may take advance of this interrupt for some
> unknown purpose when using DPDK?

Hi,

This MSI-x interrupt is for link state change, which is already handled
by the driver.
As mentioned, without this link state change for the 82574 NIC does not
work.

> 
>  
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Luca Boccassi
> > Sent: Saturday, July 28, 2018 2:13 AM
> > To: dev@dpdk.org
> > Cc: Lu, Wenzhuo <wenzhuo.lu@intel.com>; Luca Boccassi
> > <bluca@debian.org>
> > Subject: [dpdk-dev] [PATCH] net/e1000: enable MSI-x for linkup and
> > other
> > events in 82574
> > 
> > Linux supports 82574 MSI-x interrupt for linkup and other link
> > conditions.
> > Enable the same feature in the e1000 PMD in order to allow this
> > card to work
> > properly, as it requires interrupt handling for link negotiations
> > among other
> > things.
> > 
> > Signed-off-by: Luca Boccassi <bluca@debian.org>
> > ---
> > 
> > Hi Wenzhuo - just to make it clear, this patch and the previous are
> > not
> > targeted for 18.08, just clearing my queue.
> > 
> >  drivers/net/e1000/base/e1000_82571.c | 5 +++++
> >  drivers/net/e1000/em_ethdev.c        | 5 +++--
> >  2 files changed, 8 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/net/e1000/base/e1000_82571.c
> > b/drivers/net/e1000/base/e1000_82571.c
> > index 7c279dbb3..397dd9482 100644
> > --- a/drivers/net/e1000/base/e1000_82571.c
> > +++ b/drivers/net/e1000/base/e1000_82571.c
> > @@ -1257,6 +1257,11 @@ STATIC s32 e1000_init_hw_82571(struct
> > e1000_hw
> > *hw)
> >  	 */
> >  	e1000_clear_hw_cntrs_82571(hw);
> > 
> > +	/* MSI-X configure for 82574 */
> > +	if (mac->type == e1000_82574)
> > +		E1000_WRITE_REG(hw, E1000_IVAR,
> > +				(E1000_IVAR_INT_ALLOC_VALID <<
> > 16));
> > +
> >  	return ret_val;
> >  }
> > 
> > diff --git a/drivers/net/e1000/em_ethdev.c
> > b/drivers/net/e1000/em_ethdev.c index 7039dc100..bd05e4de0 100644
> > --- a/drivers/net/e1000/em_ethdev.c
> > +++ b/drivers/net/e1000/em_ethdev.c
> > @@ -1444,7 +1444,8 @@ eth_em_interrupt_setup(struct rte_eth_dev
> > *dev)
> >  	/* clear interrupt */
> >  	E1000_READ_REG(hw, E1000_ICR);
> >  	regval = E1000_READ_REG(hw, E1000_IMS);
> > -	E1000_WRITE_REG(hw, E1000_IMS, regval | E1000_ICR_LSC);
> > +	E1000_WRITE_REG(hw, E1000_IMS,
> > +			regval | E1000_ICR_LSC | E1000_ICR_OTHER);
> >  	return 0;
> >  }
> > 
> > @@ -1494,7 +1495,7 @@ em_rxq_intr_enable(struct e1000_hw
> > *hw)  static
> > void  em_lsc_intr_disable(struct e1000_hw *hw)  {
> > -	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC);
> > +	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC |
> > E1000_IMS_OTHER);
> >  	E1000_WRITE_FLUSH(hw);
> >  }
> > 
> > --
> > 2.18.0
> 
> 

-- 
Kind regards,
Luca Boccassi