On Sat, Feb 18, 2017 at 4:46 AM, Eric Dumazet wrote:
> On Thu, 2017-02-16 at 07:30 -0800, Eric Dumazet wrote:
>> On Thu, 2017-02-16 at 14:38 +0200, Saeed Mahameed wrote:
>>
>> > Acked-by: Saeed Mahameed
>>
>> Thanks for reviewing this Saeed !
>
> Note that mlx4_add_cq_to_tasklet() is called from
On Fri, Feb 17, 2017 at 1:13 AM, Eric Dumazet wrote:
> On Thu, 2017-02-16 at 23:17 +0200, Saeed Mahameed wrote:
>
>> so i guess you are not busy polling .. and adaptive moderation decided
>> to lower down
>> rx-usecs for you, and you are looking to improve latency.
>>
>> > Interrupt moderation is
On Thu, 2017-02-16 at 07:30 -0800, Eric Dumazet wrote:
> On Thu, 2017-02-16 at 14:38 +0200, Saeed Mahameed wrote:
>
> > Acked-by: Saeed Mahameed
>
> Thanks for reviewing this Saeed !
Note that mlx4_add_cq_to_tasklet() is called from hard irq context, so
we could replace the spin_lock_irqsave()
From: Eric Dumazet
Date: Fri, 10 Feb 2017 04:27:58 -0800
> From: Eric Dumazet
>
> All rx and rx netdev interrupts are handled by respectively
> by mlx4_en_rx_irq() and mlx4_en_tx_irq() which simply schedule a NAPI.
>
> But mlx4_eq_int() also fires a tasklet to service all items that were
> que
On Thu, 2017-02-16 at 23:17 +0200, Saeed Mahameed wrote:
> so i guess you are not busy polling .. and adaptive moderation decided
> to lower down
> rx-usecs for you, and you are looking to improve latency.
>
> > Interrupt moderation is a latency killer, we want our usec back.
> >
>
> well, for R
On Thu, Feb 16, 2017 at 5:49 PM, Eric Dumazet wrote:
> On Thu, 2017-02-16 at 14:44 +0200, Saeed Mahameed wrote:
>> On Wed, Feb 15, 2017 at 4:04 PM, Eric Dumazet wrote:
>> > On Wed, 2017-02-15 at 05:29 -0800, Eric Dumazet wrote:
>> >
>> >>
>> >> mlx4_eq_int() is a hard irq handler.
>> >>
>> >> How
On Thu, 2017-02-16 at 14:44 +0200, Saeed Mahameed wrote:
> On Wed, Feb 15, 2017 at 4:04 PM, Eric Dumazet wrote:
> > On Wed, 2017-02-15 at 05:29 -0800, Eric Dumazet wrote:
> >
> >>
> >> mlx4_eq_int() is a hard irq handler.
> >>
> >> How a tasklet could run in the middle of it ?
> >>
> >> A tasklet
On Thu, 2017-02-16 at 14:38 +0200, Saeed Mahameed wrote:
> Acked-by: Saeed Mahameed
Thanks for reviewing this Saeed !
On Wed, Feb 15, 2017 at 4:04 PM, Eric Dumazet wrote:
> On Wed, 2017-02-15 at 05:29 -0800, Eric Dumazet wrote:
>
>>
>> mlx4_eq_int() is a hard irq handler.
>>
>> How a tasklet could run in the middle of it ?
>>
>> A tasklet is a softirq handler.
>
> Speaking of mlx4_eq_int() , 50% of cycles are spe
On Fri, Feb 10, 2017 at 2:27 PM, Eric Dumazet wrote:
> From: Eric Dumazet
>
> All rx and rx netdev interrupts are handled by respectively
> by mlx4_en_rx_irq() and mlx4_en_tx_irq() which simply schedule a NAPI.
>
> But mlx4_eq_int() also fires a tasklet to service all items that were
> queued via
On Wed, Feb 15, 2017 at 5:26 PM, Eric Dumazet wrote:
> On Wed, 2017-02-15 at 16:52 +0200, Matan Barak (External) wrote:
>
>> So, in case of RDMA CQs, we add some per-CQE overhead of comparing the
>> list pointers and condition upon that. Maybe we could add an
>> invoke_tasklet boolean field on mlx
On Wed, 2017-02-15 at 16:52 +0200, Matan Barak (External) wrote:
> So, in case of RDMA CQs, we add some per-CQE overhead of comparing the
> list pointers and condition upon that. Maybe we could add an
> invoke_tasklet boolean field on mlx4_cq and return its value from
> mlx4_cq_completion.
> Th
On 15/02/2017 13:10, Saeed Mahameed wrote:
On Fri, Feb 10, 2017 at 2:27 PM, Eric Dumazet wrote:
From: Eric Dumazet
All rx and rx netdev interrupts are handled by respectively
by mlx4_en_rx_irq() and mlx4_en_tx_irq() which simply schedule a NAPI.
But mlx4_eq_int() also fires a tasklet to serv
On Wed, 2017-02-15 at 15:59 +0200, Saeed Mahameed wrote:
> can the tasklet run on a different core ?
No. tasklets are scheduled on local cpu, like softirqs in general.
They are not migrated, unless cpu is removed (hotplug)
>
> the point is that if the EQ is full of CQEs from different CQs you
On Wed, 2017-02-15 at 05:29 -0800, Eric Dumazet wrote:
>
> mlx4_eq_int() is a hard irq handler.
>
> How a tasklet could run in the middle of it ?
>
> A tasklet is a softirq handler.
Speaking of mlx4_eq_int() , 50% of cycles are spent on mb() (mfence)
in eq_set_ci()
I wonder why this very expe
On Wed, Feb 15, 2017 at 3:29 PM, Eric Dumazet wrote:
> On Wed, 2017-02-15 at 13:10 +0200, Saeed Mahameed wrote:
>> On Fri, Feb 10, 2017 at 2:27 PM, Eric Dumazet wrote:
>> > From: Eric Dumazet
>> >
>> > All rx and rx netdev interrupts are handled by respectively
>> > by mlx4_en_rx_irq() and mlx4_
On Wed, 2017-02-15 at 13:10 +0200, Saeed Mahameed wrote:
> On Fri, Feb 10, 2017 at 2:27 PM, Eric Dumazet wrote:
> > From: Eric Dumazet
> >
> > All rx and rx netdev interrupts are handled by respectively
> > by mlx4_en_rx_irq() and mlx4_en_tx_irq() which simply schedule a NAPI.
> >
> > But mlx4_eq
On Fri, Feb 10, 2017 at 2:27 PM, Eric Dumazet wrote:
> From: Eric Dumazet
>
> All rx and rx netdev interrupts are handled by respectively
> by mlx4_en_rx_irq() and mlx4_en_tx_irq() which simply schedule a NAPI.
>
> But mlx4_eq_int() also fires a tasklet to service all items that were
> queued via
From: Eric Dumazet
Date: Fri, 10 Feb 2017 04:27:58 -0800
> From: Eric Dumazet
>
> All rx and rx netdev interrupts are handled by respectively
> by mlx4_en_rx_irq() and mlx4_en_tx_irq() which simply schedule a NAPI.
>
> But mlx4_eq_int() also fires a tasklet to service all items that were
> que
From: Eric Dumazet
All rx and rx netdev interrupts are handled by respectively
by mlx4_en_rx_irq() and mlx4_en_tx_irq() which simply schedule a NAPI.
But mlx4_eq_int() also fires a tasklet to service all items that were
queued via mlx4_add_cq_to_tasklet(), but this handler was not called
unless
20 matches
Mail list logo
