On 10/11/2015 20:25, Måns Rullgård wrote:
> Mason writes:
>
>> On 10/11/2015 17:14, Mans Rullgard wrote:
>>
>>> This adds a driver for the Aurora VLSI NB8800 Ethernet controller.
>>> It is an almost complete rewrite of a driver originally found in
>>> a Sigma Designs 2.6.22 tree.
>>>
>>>
Mason writes:
> On 10/11/2015 20:25, Måns Rullgård wrote:
>
>> Mason writes:
>>
>>> On 10/11/2015 17:14, Mans Rullgard wrote:
>>>
This adds a driver for the Aurora VLSI NB8800 Ethernet controller.
It is an almost complete rewrite of a driver originally found in
Mason writes:
> [ CCing a few knowledgeable people ]
>
> Despite the subject, this is about an Atheros 8035 PHY :-)
>
> On 12/11/2015 15:04, Måns Rullgård wrote:
>
>> Mason wrote:
>>
>>> BTW, you're not using the PHY IRQ, right? I think I remember you saying
>>> it didn't
Måns Rullgård writes:
> Mason writes:
>
>> [ CCing a few knowledgeable people ]
>>
>> Despite the subject, this is about an Atheros 8035 PHY :-)
>>
>> On 12/11/2015 15:04, Måns Rullgård wrote:
>>
>>> Mason wrote:
>>>
BTW, you're not using the PHY IRQ,
[ CCing a few knowledgeable people ]
Despite the subject, this is about an Atheros 8035 PHY :-)
On 12/11/2015 15:04, Måns Rullgård wrote:
> Mason wrote:
>
>> BTW, you're not using the PHY IRQ, right? I think I remember you saying
>> it didn't work reliably?
>
> It doesn't seem to be wired up
David Miller writes:
> From: Måns Rullgård
> Date: Wed, 11 Nov 2015 13:04:07 +
>
>> Måns Rullgård writes:
>>
>>> David Miller writes:
>>>
From: Måns Rullgård
Date: Wed, 11 Nov 2015
On Wed, 2015-11-11 at 14:15 +, Måns Rullgård wrote:
> FWIW, with UDP I get 650 Mbps.
Have you tried pktgen ? It should give you raw numbers, without the
stack overhead, especially when using the clone operator.
(It also has xmit_more support)
If the NIC can not get more than 650 Mbps, not
From: Måns Rullgård
Date: Wed, 11 Nov 2015 13:04:07 +
> Måns Rullgård writes:
>
>> David Miller writes:
>>
>>> From: Måns Rullgård
>>> Date: Wed, 11 Nov 2015 00:40:09 +
>>>
When the DMA complete interrupt
On Wed, 2015-11-11 at 13:04 +, Måns Rullgård wrote:
> I tried that, and throughput (as measured by iperf3) dropped by 2%.
> Maybe I did something wrong.
What link speed have you used, what was the throughput you got,
and is the receiver using the same NIC ?
--
To unsubscribe from this
Mason writes:
> On 10/11/2015 22:51, Eric Dumazet wrote:
>
>> On Tue, 2015-11-10 at 21:21 +, Måns Rullgård wrote:
>>
>>> Even ixgbe uses napi_complete() while netdevice.h says one should
>>> "consider using napi_complete_done() instead." Did the author consider
>>> it
On Wed, 2015-11-11 at 13:54 +, Måns Rullgård wrote:
> Mason writes:
>
> > On 10/11/2015 22:51, Eric Dumazet wrote:
> >
> >> On Tue, 2015-11-10 at 21:21 +, Måns Rullgård wrote:
> >>
> >>> Even ixgbe uses napi_complete() while netdevice.h says one should
> >>> "consider
Eric Dumazet writes:
> On Wed, 2015-11-11 at 13:48 +, Måns Rullgård wrote:
>> Eric Dumazet writes:
>>
>> > On Wed, 2015-11-11 at 13:04 +, Måns Rullgård wrote:
>> >
>> >> I tried that, and throughput (as measured by iperf3) dropped by 2%.
Måns Rullgård writes:
> Eric Dumazet writes:
>
>> On Wed, 2015-11-11 at 13:48 +, Måns Rullgård wrote:
>>> Eric Dumazet writes:
>>>
>>> > On Wed, 2015-11-11 at 13:04 +, Måns Rullgård wrote:
>>> >
>>> >> I tried that, and
From: Måns Rullgård
Date: Wed, 11 Nov 2015 18:25:05 +
> If the TX DMA channel is idle when start_xmit is called, it can be
> started immediately. Checking the DMA status and starting it if
> idle has to be done atomically somehow.
->ndo_start_xmit() is guaranteed to be
From: Måns Rullgård
Date: Wed, 11 Nov 2015 19:35:05 +
>> I don't think it's silly at all.
>
> I'm sure I read somewhere that the time spent spinning on a lock should
> be kept as small as possible.
>
>> And unless you can measure it making a difference, don't knock the
David Miller writes:
> From: Måns Rullgård
> Date: Wed, 11 Nov 2015 18:25:05 +
>
>> If the TX DMA channel is idle when start_xmit is called, it can be
>> started immediately. Checking the DMA status and starting it if
>> idle has to be done atomically
David Miller writes:
> From: Måns Rullgård
> Date: Wed, 11 Nov 2015 19:09:19 +
>
>> David Miller writes:
>>
>>> From: Måns Rullgård
>>> Date: Wed, 11 Nov 2015 18:25:05 +
>>>
If the TX DMA channel is idle
From: Måns Rullgård
Date: Wed, 11 Nov 2015 19:25:46 +
> David Miller writes:
>
>> From: Måns Rullgård
>> Date: Wed, 11 Nov 2015 19:17:07 +
>>
>>> David Miller writes:
>>>
From: Måns Rullgård
From: Måns Rullgård
Date: Wed, 11 Nov 2015 19:09:19 +
> David Miller writes:
>
>> From: Måns Rullgård
>> Date: Wed, 11 Nov 2015 18:25:05 +
>>
>>> If the TX DMA channel is idle when start_xmit is called, it can be
>>> started
From: Måns Rullgård
Date: Wed, 11 Nov 2015 19:17:07 +
> David Miller writes:
>
>> From: Måns Rullgård
>> Date: Wed, 11 Nov 2015 19:09:19 +
>>
>>> David Miller writes:
>>>
From: Måns Rullgård
David Miller writes:
> From: Måns Rullgård
> Date: Wed, 11 Nov 2015 19:17:07 +
>
>> David Miller writes:
>>
>>> From: Måns Rullgård
>>> Date: Wed, 11 Nov 2015 19:09:19 +
>>>
David Miller
David Miller writes:
> From: Måns Rullgård
> Date: Wed, 11 Nov 2015 19:35:05 +
>
>>> I don't think it's silly at all.
>>
>> I'm sure I read somewhere that the time spent spinning on a lock should
>> be kept as small as possible.
>>
>>> And unless you
Eric Dumazet writes:
> On Tue, 2015-11-10 at 16:14 +, Mans Rullgard wrote:
>> This adds a driver for the Aurora VLSI NB8800 Ethernet controller.
>> It is an almost complete rewrite of a driver originally found in
>> a Sigma Designs 2.6.22 tree.
>
> ...
>
>> +
>>
On 10/11/2015 17:14, Mans Rullgard wrote:
> This adds a driver for the Aurora VLSI NB8800 Ethernet controller.
> It is an almost complete rewrite of a driver originally found in
> a Sigma Designs 2.6.22 tree.
>
> Signed-off-by: Mans Rullgard
> ---
> Changes:
> - Refactored mdio
Eric Dumazet writes:
> On Tue, 2015-11-10 at 16:14 +, Mans Rullgard wrote:
>> This adds a driver for the Aurora VLSI NB8800 Ethernet controller.
>> It is an almost complete rewrite of a driver originally found in
>> a Sigma Designs 2.6.22 tree.
>
> ...
>
>> +
>>
Mason writes:
> On 10/11/2015 17:14, Mans Rullgard wrote:
>
>> This adds a driver for the Aurora VLSI NB8800 Ethernet controller.
>> It is an almost complete rewrite of a driver originally found in
>> a Sigma Designs 2.6.22 tree.
>>
>> Signed-off-by: Mans Rullgard
David Miller writes:
> From: Måns Rullgård
> Date: Tue, 10 Nov 2015 18:05:15 +
>
>> Because I haven't been following the netdev list closely for the last
>> five years, and no documentation I read mentioned this function. I can
>> certainly change it.
>
From: Måns Rullgård
Date: Tue, 10 Nov 2015 20:53:19 +
> David Miller writes:
>
>> From: Måns Rullgård
>> Date: Tue, 10 Nov 2015 18:05:15 +
>>
>>> Because I haven't been following the netdev list closely for the last
>>> five years,
From: Måns Rullgård
Date: Tue, 10 Nov 2015 18:05:15 +
> Because I haven't been following the netdev list closely for the last
> five years, and no documentation I read mentioned this function. I can
> certainly change it.
It is always advisable to mimick what other drivers
David Miller writes:
> From: Måns Rullgård
> Date: Tue, 10 Nov 2015 20:53:19 +
>
>> David Miller writes:
>>
>>> From: Måns Rullgård
>>> Date: Tue, 10 Nov 2015 18:05:15 +
>>>
Because I haven't been
On Tue, 2015-11-10 at 21:21 +, Måns Rullgård wrote:
> Even ixgbe uses napi_complete() while netdevice.h says one should
> "consider using napi_complete_done() instead." Did the author consider
> it and decide not to, or has the driver simply not been updated?
napi_complete_done() is quite
Andy Shevchenko writes:
> On Wed, Nov 11, 2015 at 12:34 AM, Måns Rullgård wrote:
>> Andy Shevchenko writes:
>>
+static inline void nb8800_maskb(struct nb8800_priv *priv, int reg,
+ u32
On Tue, Nov 10, 2015 at 6:14 PM, Mans Rullgard wrote:
> This adds a driver for the Aurora VLSI NB8800 Ethernet controller.
> It is an almost complete rewrite of a driver originally found in
> a Sigma Designs 2.6.22 tree.
Few nitpicks below.
>
> Signed-off-by: Mans Rullgard
Andy Shevchenko writes:
>> +static inline void nb8800_maskb(struct nb8800_priv *priv, int reg,
>> + u32 mask, u32 val)
>> +{
>> + u32 old = nb8800_readb(priv, reg);
>> + u32 new = (old & ~mask) | val;
>
> Shoudn't be "… | (val
On Wed, Nov 11, 2015 at 1:07 AM, Måns Rullgård wrote:
> Andy Shevchenko writes:
> + nb8800_writel(priv, NB8800_TX_DESC_ADDR, txb->dma_desc);
> + wmb(); /* ensure desc addr is written before starting
> DMA */
From: Måns Rullgård
Date: Wed, 11 Nov 2015 00:40:09 +
> When the DMA complete interrupt arrives, the next chain should be
> kicked off as quickly as possible, and I don't see why that would
> benefit from being done in napi context.
NAPI isn't about low latency, it's about
Francois Romieu writes:
> Mans Rullgard :
>> diff --git a/drivers/net/ethernet/aurora/nb8800.c
>> b/drivers/net/ethernet/aurora/nb8800.c
>> new file mode 100644
>> index 000..11cd389
>> --- /dev/null
>> +++ b/drivers/net/ethernet/aurora/nb8800.c
>
37 matches
Mail list logo