On Tue, 2017-02-28 at 08:17 -0800, Stephen Hemminger wrote:
> Maybe just as simple as using irqsave/irqrestore in driver.
CPU can be differents.
irqsave will not help.
On Mon, 27 Feb 2017 12:18:31 -0800
Eric Dumazet wrote:
> This can happen with busy polling users, or if gro_flush_timeout is
> used. But some other uses of napi_schedule() in drivers can cause this
> as well.
Where were IRQ's re-enabled?
> thread 1 thread 2 (cou
On Tue, 2017-02-28 at 10:14 +, David Laight wrote:
> From: Eric Dumazet
> > Sent: 27 February 2017 22:35
> > On Mon, 2017-02-27 at 14:14 -0800, Stephen Hemminger wrote:
> >
> > > The original design (as Davem mentioned) was that IRQ's must be disabled
> > > during device polling. If that was t
On Tue, 2017-02-28 at 10:14 +, David Laight wrote:
> From: Eric Dumazet
> > Sent: 27 February 2017 22:35
> > On Mon, 2017-02-27 at 14:14 -0800, Stephen Hemminger wrote:
> >
> > > The original design (as Davem mentioned) was that IRQ's must be disabled
> > > during device polling. If that was t
From: Eric Dumazet
> Sent: 27 February 2017 22:35
> On Mon, 2017-02-27 at 14:14 -0800, Stephen Hemminger wrote:
>
> > The original design (as Davem mentioned) was that IRQ's must be disabled
> > during device polling. If that was true, then the race above
> > would be impossible.
>
> I would love
On Mon, 27 Feb 2017 17:48:54 -0500 (EST)
David Miller wrote:
> From: Stephen Hemminger
> Date: Mon, 27 Feb 2017 14:44:55 -0800
>
> > On Mon, 27 Feb 2017 14:35:17 -0800
> > Eric Dumazet wrote:
> >
> >> On Mon, 2017-02-27 at 14:14 -0800, Stephen Hemminger wrote:
> >>
> >> > The original de
From: Stephen Hemminger
Date: Mon, 27 Feb 2017 14:44:55 -0800
> On Mon, 27 Feb 2017 14:35:17 -0800
> Eric Dumazet wrote:
>
>> On Mon, 2017-02-27 at 14:14 -0800, Stephen Hemminger wrote:
>>
>> > The original design (as Davem mentioned) was that IRQ's must be disabled
>> > during device polling.
On Mon, 27 Feb 2017 14:35:17 -0800
Eric Dumazet wrote:
> On Mon, 2017-02-27 at 14:14 -0800, Stephen Hemminger wrote:
>
> > The original design (as Davem mentioned) was that IRQ's must be disabled
> > during device polling. If that was true, then the race above
> > would be impossible.
>
> I w
On Mon, 2017-02-27 at 14:14 -0800, Stephen Hemminger wrote:
> The original design (as Davem mentioned) was that IRQ's must be disabled
> during device polling. If that was true, then the race above
> would be impossible.
I would love to see an alternative patch.
On Mon, 27 Feb 2017 12:18:31 -0800
Eric Dumazet wrote:
> thread 1 thread 2 (could be on same cpu)
>
> // busy polling or napi_watchdog()
> napi_schedule();
> ...
> napi->poll()
>
> device polling:
> read 2 packets from ring buffer
>
From: Eric Dumazet
While playing with mlx4 hardware timestamping of RX packets, I found
that some packets were received by TCP stack with a ~200 ms delay...
Since the timestamp was provided by the NIC, and my probe was added
in tcp_v4_rcv() while in BH handler, I was confident it was not
a sende
11 matches
Mail list logo
