Re: e1000_netpoll() , BUG: sleeping function called from invalid context

[Gabriel C]

On 01.03.2017 23:38, Thomas Gleixner wrote:
> On Wed, 1 Mar 2017, Gabriel C wrote:
> > On 01.03.2017 18:13, Thomas Gleixner wrote:
> > > On Sat, 18 Feb 2017, Gabriel C wrote:
> > > > That got changed in commit 3111912971251 which got into Linus tree during
> > > > the 4.10 merge window. So it is in 4.10-rc8.
> > > >
> > > > Confused.
> > >
> > > I'm still confused. Gabriel, can you please try to reproduce with 4.10
> > > final?
> >
> > Sure , do you want plain 4.10.0 or 4.10.1 ?
> >
> > Btw is possible to rename e1000_netpoll() from e1000e driver
> > to e1000e_netpoll() so we know what driver is in use ?
>
> Grr. yes. That would be definitely helpful. I was staring into the wrong
> one of course.
>
> > My card uses the e1000e driver..
>
> Does the patch below fix it?

I'll test you patch in a bit and let you know.

> Thanks,
>
> 	tglx
> 	
> 8<--------------
>
> --- a/drivers/net/ethernet/intel/e1000e/netdev.c
> +++ b/drivers/net/ethernet/intel/e1000e/netdev.c
> @@ -6716,19 +6716,19 @@ static irqreturn_t e1000_intr_msix(int _
>
>  		vector = 0;
>  		msix_irq = adapter->msix_entries[vector].vector;
> -		disable_irq(msix_irq);
> +		disable_hardirq(msix_irq);
>  		e1000_intr_msix_rx(msix_irq, netdev);
>  		enable_irq(msix_irq);
>
>  		vector++;
>  		msix_irq = adapter->msix_entries[vector].vector;
> -		disable_irq(msix_irq);
> +		disable_hardirq(msix_irq);
>  		e1000_intr_msix_tx(msix_irq, netdev);
>  		enable_irq(msix_irq);
>
>  		vector++;
>  		msix_irq = adapter->msix_entries[vector].vector;
> -		disable_irq(msix_irq);
> +		disable_hardirq(msix_irq);
>  		e1000_msix_other(msix_irq, netdev);
>  		enable_irq(msix_irq);
>  	}