[PATCH] bgmac: Remove all offloading features, including GRO.

[PATCH] bgmac: Remove all offloading features, including GRO.

[Rosen Penev]

On a linksys E1200v1 (actually a crossflashed E1000v2), the offloading features give no measurable benefit to speed or latency. Furthermore, disabling GRO actually improves iperf performance by a whoppimg 3mbps. Results:

Currently:

[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52410
[ ID] Interval        Transfer    Bandwidth       Reads   Dist(bin=16.0K)
[  4] 0.00-10.02 sec  52.4 MBytes  43.8 Mbits/sec  641    75:181:12:1:1:0:0:371
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52412
[  5] 0.00-10.02 sec  52.4 MBytes  43.8 Mbits/sec  629    51:194:13:1:0:1:0:369
[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52414
[  4] 0.00-10.02 sec  51.9 MBytes  43.4 Mbits/sec  695    126:203:1:0:0:0:2:363
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52416
[  5] 0.00-10.01 sec  52.4 MBytes  43.9 Mbits/sec  626    57:186:10:0:0:0:0:373
[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52420
[  4] 0.00-10.02 sec  52.4 MBytes  43.8 Mbits/sec  605    36:179:16:1:0:1:0:372
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52422

After disabling everything - including GRO:

[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52440
[ ID] Interval        Transfer    Bandwidth       Reads   Dist(bin=16.0K)
[  4] 0.00-10.01 sec  55.1 MBytes  46.2 Mbits/sec  672    180:82:0:0:1:0:0:409
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52442
[  5] 0.00-10.01 sec  56.0 MBytes  46.9 Mbits/sec  636    117:96:0:0:1:0:0:422
[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52444
[  4] 0.00-10.01 sec  55.4 MBytes  46.4 Mbits/sec  675    172:92:0:0:1:0:0:410
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52446
[  5] 0.00-10.01 sec  56.0 MBytes  46.9 Mbits/sec  633    119:90:0:1:1:0:0:422
[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52448
[  4] 0.00-10.01 sec  55.2 MBytes  46.3 Mbits/sec  688    157:123:0:0:2:0:0:406

v2: Changed napi_gro_receive to netif_receive_skb. Seems to have an identical result.

Signed-off-by: Rosen Penev <rosenp@gmail.com>
---
 drivers/net/ethernet/broadcom/bgmac.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/broadcom/bgmac.c b/drivers/net/ethernet/broadcom/bgmac.c
index 48d672b204a4..1fb0053aeee7 100644
--- a/drivers/net/ethernet/broadcom/bgmac.c
+++ b/drivers/net/ethernet/broadcom/bgmac.c
@@ -483,7 +483,7 @@ static int bgmac_dma_rx_read(struct bgmac *bgmac, struct bgmac_dma_ring *ring,
 			skb->protocol = eth_type_trans(skb, bgmac->net_dev);
 			bgmac->net_dev->stats.rx_bytes += len;
 			bgmac->net_dev->stats.rx_packets++;
-			napi_gro_receive(&bgmac->napi, skb);
+			netif_receive_skb(skb);
 			handled++;
 		} while (0);
 
-- 
2.13.5

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[Eric Dumazet]

On Fri, 2017-09-15 at 17:23 -0700, Rosen Penev wrote:
> On a linksys E1200v1 (actually a crossflashed E1000v2), the offloading features give no measurable benefit to speed or latency. Furthermore, disabling GRO actually improves iperf performance by a whoppimg 3mbps. Results:
> 
> Currently:

> 
> v2: Changed napi_gro_receive to netif_receive_skb. Seems to have an identical result.
> 
> Signed-off-by: Rosen Penev <rosenp@gmail.com>
> ---
>  drivers/net/ethernet/broadcom/bgmac.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/broadcom/bgmac.c b/drivers/net/ethernet/broadcom/bgmac.c
> index 48d672b204a4..1fb0053aeee7 100644
> --- a/drivers/net/ethernet/broadcom/bgmac.c
> +++ b/drivers/net/ethernet/broadcom/bgmac.c
> @@ -483,7 +483,7 @@ static int bgmac_dma_rx_read(struct bgmac *bgmac, struct bgmac_dma_ring *ring,
>  			skb->protocol = eth_type_trans(skb, bgmac->net_dev);
>  			bgmac->net_dev->stats.rx_bytes += len;
>  			bgmac->net_dev->stats.rx_packets++;
> -			napi_gro_receive(&bgmac->napi, skb);
> +			netif_receive_skb(skb);
>  			handled++;
>  		} while (0);
>  

And have you tested 1Gbit link speed ?
( Or 2.5 Gbit link speed )

If you want to disable GRO on your host, fine : you can use ethtool -K

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[rosenp]

I have not. Unfortunately I own no gigabit hardware to test this on.
The MIPS CPU runs at 300MHz on my unit.

On Fri, 2017-09-15 at 17:34 -0700, Eric Dumazet wrote:
> On Fri, 2017-09-15 at 17:23 -0700, Rosen Penev wrote:
> > On a linksys E1200v1 (actually a crossflashed E1000v2), the
> > offloading features give no measurable benefit to speed or latency.
> > Furthermore, disabling GRO actually improves iperf performance by a
> > whoppimg 3mbps. Results:
> > 
> > Currently:
> > 
> > v2: Changed napi_gro_receive to netif_receive_skb. Seems to have an
> > identical result.
> > 
> > Signed-off-by: Rosen Penev <rosenp@gmail.com>
> > ---
> >  drivers/net/ethernet/broadcom/bgmac.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/net/ethernet/broadcom/bgmac.c
> > b/drivers/net/ethernet/broadcom/bgmac.c
> > index 48d672b204a4..1fb0053aeee7 100644
> > --- a/drivers/net/ethernet/broadcom/bgmac.c
> > +++ b/drivers/net/ethernet/broadcom/bgmac.c
> > @@ -483,7 +483,7 @@ static int bgmac_dma_rx_read(struct bgmac
> > *bgmac, struct bgmac_dma_ring *ring,
> >  			skb->protocol = eth_type_trans(skb, bgmac-
> > >net_dev);
> >  			bgmac->net_dev->stats.rx_bytes += len;
> >  			bgmac->net_dev->stats.rx_packets++;
> > -			napi_gro_receive(&bgmac->napi, skb);
> > +			netif_receive_skb(skb);
> >  			handled++;
> >  		} while (0);
> >  
> 
> And have you tested 1Gbit link speed ?
> ( Or 2.5 Gbit link speed )
> 
> If you want to disable GRO on your host, fine : you can use ethtool
> -K
> 
> 
> 

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[Florian Fainelli]

On September 15, 2017 5:38:42 PM PDT, rosenp@gmail.com wrote:
>I have not. Unfortunately I own no gigabit hardware to test this on.
>The MIPS CPU runs at 300MHz on my unit.
>

bgmac is used on Gigabit capable hardware, like Northstar and Northstar Plus, and others too, so unless you can get access to such HW or get confirmation from someone that your patches changes something, I would just drop this change and not bother. This is already not 100mbits/sec linerate...

>On Fri, 2017-09-15 at 17:34 -0700, Eric Dumazet wrote:
>> On Fri, 2017-09-15 at 17:23 -0700, Rosen Penev wrote:
>> > On a linksys E1200v1 (actually a crossflashed E1000v2), the
>> > offloading features give no measurable benefit to speed or latency.
>> > Furthermore, disabling GRO actually improves iperf performance by a
>> > whoppimg 3mbps. Results:
>> > 
>> > Currently:
>> > 
>> > v2: Changed napi_gro_receive to netif_receive_skb. Seems to have an
>> > identical result.
>> > 
>> > Signed-off-by: Rosen Penev <rosenp@gmail.com>
>> > ---
>> >  drivers/net/ethernet/broadcom/bgmac.c | 2 +-
>> >  1 file changed, 1 insertion(+), 1 deletion(-)
>> > 
>> > diff --git a/drivers/net/ethernet/broadcom/bgmac.c
>> > b/drivers/net/ethernet/broadcom/bgmac.c
>> > index 48d672b204a4..1fb0053aeee7 100644
>> > --- a/drivers/net/ethernet/broadcom/bgmac.c
>> > +++ b/drivers/net/ethernet/broadcom/bgmac.c
>> > @@ -483,7 +483,7 @@ static int bgmac_dma_rx_read(struct bgmac
>> > *bgmac, struct bgmac_dma_ring *ring,
>> >  			skb->protocol = eth_type_trans(skb, bgmac-
>> > >net_dev);
>> >  			bgmac->net_dev->stats.rx_bytes += len;
>> >  			bgmac->net_dev->stats.rx_packets++;
>> > -			napi_gro_receive(&bgmac->napi, skb);
>> > +			netif_receive_skb(skb);
>> >  			handled++;
>> >  		} while (0);
>> >  
>> 
>> And have you tested 1Gbit link speed ?
>> ( Or 2.5 Gbit link speed )
>> 
>> If you want to disable GRO on your host, fine : you can use ethtool
>> -K
>> 
>> 
>> 

(please don't top-post)
-- 
Florian

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[rosenp]

Sorry for the noise. After more testing I've found out that the cause
was that I had BBR enabled on my laptop. Switching back to CUBIC fixed
the issue.

In other words, this patch is detrimental.

~67mbps - gro off
~87mbps - gro on

On Fri, 2017-09-15 at 23:04 -0700, Florian Fainelli wrote:
> On September 15, 2017 5:38:42 PM PDT, rosenp@gmail.com wrote:
> > I have not. Unfortunately I own no gigabit hardware to test this
> > on.
> > The MIPS CPU runs at 300MHz on my unit.
> > 
> 
> bgmac is used on Gigabit capable hardware, like Northstar and
> Northstar Plus, and others too, so unless you can get access to such
> HW or get confirmation from someone that your patches changes
> something, I would just drop this change and not bother. This is
> already not 100mbits/sec linerate...
> 
> > On Fri, 2017-09-15 at 17:34 -0700, Eric Dumazet wrote:
> > > On Fri, 2017-09-15 at 17:23 -0700, Rosen Penev wrote:
> > > > On a linksys E1200v1 (actually a crossflashed E1000v2), the
> > > > offloading features give no measurable benefit to speed or
> > > > latency.
> > > > Furthermore, disabling GRO actually improves iperf performance
> > > > by a
> > > > whoppimg 3mbps. Results:
> > > > 
> > > > Currently:
> > > > 
> > > > v2: Changed napi_gro_receive to netif_receive_skb. Seems to
> > > > have an
> > > > identical result.
> > > > 
> > > > Signed-off-by: Rosen Penev <rosenp@gmail.com>
> > > > ---
> > > >  drivers/net/ethernet/broadcom/bgmac.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/net/ethernet/broadcom/bgmac.c
> > > > b/drivers/net/ethernet/broadcom/bgmac.c
> > > > index 48d672b204a4..1fb0053aeee7 100644
> > > > --- a/drivers/net/ethernet/broadcom/bgmac.c
> > > > +++ b/drivers/net/ethernet/broadcom/bgmac.c
> > > > @@ -483,7 +483,7 @@ static int bgmac_dma_rx_read(struct bgmac
> > > > *bgmac, struct bgmac_dma_ring *ring,
> > > >  			skb->protocol = eth_type_trans(skb,
> > > > bgmac-
> > > > > net_dev);
> > > > 
> > > >  			bgmac->net_dev->stats.rx_bytes += len;
> > > >  			bgmac->net_dev->stats.rx_packets++;
> > > > -			napi_gro_receive(&bgmac->napi, skb);
> > > > +			netif_receive_skb(skb);
> > > >  			handled++;
> > > >  		} while (0);
> > > >  
> > > 
> > > And have you tested 1Gbit link speed ?
> > > ( Or 2.5 Gbit link speed )
> > > 
> > > If you want to disable GRO on your host, fine : you can use
> > > ethtool
> > > -K
> > > 
> > > 
> > > 
> 
> (please don't top-post)

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[Florian Fainelli]

On 09/20/2017 02:27 PM, rosenp@gmail.com wrote:
> Sorry for the noise. After more testing I've found out that the cause
> was that I had BBR enabled on my laptop. Switching back to CUBIC fixed
> the issue.
> 
> In other words, this patch is detrimental.

Quite unsurprisingly, thanks for coming back, and please don't top-post
on netdev.

> 
> ~67mbps - gro off
> ~87mbps - gro on
> 
> On Fri, 2017-09-15 at 23:04 -0700, Florian Fainelli wrote:
>> On September 15, 2017 5:38:42 PM PDT, rosenp@gmail.com wrote:
>>> I have not. Unfortunately I own no gigabit hardware to test this
>>> on.
>>> The MIPS CPU runs at 300MHz on my unit.
>>>
>>
>> bgmac is used on Gigabit capable hardware, like Northstar and
>> Northstar Plus, and others too, so unless you can get access to such
>> HW or get confirmation from someone that your patches changes
>> something, I would just drop this change and not bother. This is
>> already not 100mbits/sec linerate...
>>
>>> On Fri, 2017-09-15 at 17:34 -0700, Eric Dumazet wrote:
>>>> On Fri, 2017-09-15 at 17:23 -0700, Rosen Penev wrote:
>>>>> On a linksys E1200v1 (actually a crossflashed E1000v2), the
>>>>> offloading features give no measurable benefit to speed or
>>>>> latency.
>>>>> Furthermore, disabling GRO actually improves iperf performance
>>>>> by a
>>>>> whoppimg 3mbps. Results:
>>>>>
>>>>> Currently:
>>>>>
>>>>> v2: Changed napi_gro_receive to netif_receive_skb. Seems to
>>>>> have an
>>>>> identical result.
>>>>>
>>>>> Signed-off-by: Rosen Penev <rosenp@gmail.com>
>>>>> ---
>>>>>  drivers/net/ethernet/broadcom/bgmac.c | 2 +-
>>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/drivers/net/ethernet/broadcom/bgmac.c
>>>>> b/drivers/net/ethernet/broadcom/bgmac.c
>>>>> index 48d672b204a4..1fb0053aeee7 100644
>>>>> --- a/drivers/net/ethernet/broadcom/bgmac.c
>>>>> +++ b/drivers/net/ethernet/broadcom/bgmac.c
>>>>> @@ -483,7 +483,7 @@ static int bgmac_dma_rx_read(struct bgmac
>>>>> *bgmac, struct bgmac_dma_ring *ring,
>>>>>  			skb->protocol = eth_type_trans(skb,
>>>>> bgmac-
>>>>>> net_dev);
>>>>>
>>>>>  			bgmac->net_dev->stats.rx_bytes += len;
>>>>>  			bgmac->net_dev->stats.rx_packets++;
>>>>> -			napi_gro_receive(&bgmac->napi, skb);
>>>>> +			netif_receive_skb(skb);
>>>>>  			handled++;
>>>>>  		} while (0);
>>>>>  
>>>>
>>>> And have you tested 1Gbit link speed ?
>>>> ( Or 2.5 Gbit link speed )
>>>>
>>>> If you want to disable GRO on your host, fine : you can use
>>>> ethtool
>>>> -K
>>>>
>>>>
>>>>
>>
>> (please don't top-post)


-- 
Florian

[PATCH] bgmac: Remove all offloading features, including GRO.

[Rosen Penev]

On a linksys E1200v1 (actually a crossflashed E1000v2), the offloading features give no measurable benefit to speed or latency. Furthermore, disabling GRO actually improves iperf performance by a whoppimg 3mbps. Results:

Currently:

[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52410
[ ID] Interval        Transfer    Bandwidth       Reads   Dist(bin=16.0K)
[  4] 0.00-10.02 sec  52.4 MBytes  43.8 Mbits/sec  641    75:181:12:1:1:0:0:371
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52412
[  5] 0.00-10.02 sec  52.4 MBytes  43.8 Mbits/sec  629    51:194:13:1:0:1:0:369
[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52414
[  4] 0.00-10.02 sec  51.9 MBytes  43.4 Mbits/sec  695    126:203:1:0:0:0:2:363
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52416
[  5] 0.00-10.01 sec  52.4 MBytes  43.9 Mbits/sec  626    57:186:10:0:0:0:0:373
[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52420
[  4] 0.00-10.02 sec  52.4 MBytes  43.8 Mbits/sec  605    36:179:16:1:0:1:0:372
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52422

After disabling everything - including GRO:

[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52440
[ ID] Interval        Transfer    Bandwidth       Reads   Dist(bin=16.0K)
[  4] 0.00-10.01 sec  55.1 MBytes  46.2 Mbits/sec  672    180:82:0:0:1:0:0:409
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52442
[  5] 0.00-10.01 sec  56.0 MBytes  46.9 Mbits/sec  636    117:96:0:0:1:0:0:422
[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52444
[  4] 0.00-10.01 sec  55.4 MBytes  46.4 Mbits/sec  675    172:92:0:0:1:0:0:410
[  5] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52446
[  5] 0.00-10.01 sec  56.0 MBytes  46.9 Mbits/sec  633    119:90:0:1:1:0:0:422
[  4] local 192.168.1.1 port 5001 connected with 192.168.1.100 port 52448
[  4] 0.00-10.01 sec  55.2 MBytes  46.3 Mbits/sec  688    157:123:0:0:2:0:0:406

Signed-off-by: Rosen Penev <rosenp@gmail.com>
---
 drivers/net/ethernet/broadcom/bgmac.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/broadcom/bgmac.c b/drivers/net/ethernet/broadcom/bgmac.c
index 48d672b204a4..c2db5673b073 100644
--- a/drivers/net/ethernet/broadcom/bgmac.c
+++ b/drivers/net/ethernet/broadcom/bgmac.c
@@ -1534,7 +1534,7 @@ int bgmac_enet_probe(struct bgmac *bgmac)
 		goto err_dma_free;
 	}
 
-	net_dev->features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
+	net_dev->features &= NETIF_F_GRO;
 	net_dev->hw_features = net_dev->features;
 	net_dev->vlan_features = net_dev->features;
 
-- 
2.13.5

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[Florian Fainelli]

On September 15, 2017 3:22:18 PM PDT, Rosen Penev <rosenp@gmail.com> wrote:
>On a linksys E1200v1 (actually a crossflashed E1000v2), the offloading
>features give no measurable benefit to speed or latency. Furthermore,
>disabling GRO actually improves iperf performance by a whoppimg 3mbps.

Do you have a way to generate gigabit tests and see what results you are getting? We probably are not going to see a 30% improvement just by extrapolation.

-- 
Florian

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[Eric Dumazet]

On Fri, 2017-09-15 at 15:54 -0700, Florian Fainelli wrote:
> On September 15, 2017 3:22:18 PM PDT, Rosen Penev <rosenp@gmail.com>
> wrote:
> >On a linksys E1200v1 (actually a crossflashed E1000v2), the
> offloading
> >features give no measurable benefit to speed or latency. Furthermore,
> >disabling GRO actually improves iperf performance by a whoppimg
> 3mbps.
> 
> Do you have a way to generate gigabit tests and see what results you
> are getting? We probably are not going to see a 30% improvement just
> by extrapolation.
> 
+1

It seems silly to remove NETIF_F_SG | NETIF_F_IP_CSUM |
NETIF_F_IPV6_CSUM from dev->features, yet leave the dead-code in the
driver to handle these features.

And of course GRO was not removed, meaning the bench results were non
conclusive.

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[rosenp]

Ok fair enough. Will only disable GRO in the driver.

Tests were done using "ethtool -K eth0 gro off" and on.

On Fri, 2017-09-15 at 17:03 -0700, Eric Dumazet wrote:
> On Fri, 2017-09-15 at 15:54 -0700, Florian Fainelli wrote:
> > On September 15, 2017 3:22:18 PM PDT, Rosen Penev <rosenp@gmail.com
> > >
> > wrote:
> > > On a linksys E1200v1 (actually a crossflashed E1000v2), the
> > 
> > offloading
> > > features give no measurable benefit to speed or latency.
> > > Furthermore,
> > > disabling GRO actually improves iperf performance by a whoppimg
> > 
> > 3mbps.
> > 
> > Do you have a way to generate gigabit tests and see what results
> > you
> > are getting? We probably are not going to see a 30% improvement
> > just
> > by extrapolation.
> > 
> 
> +1
> 
> It seems silly to remove NETIF_F_SG | NETIF_F_IP_CSUM |
> NETIF_F_IPV6_CSUM from dev->features, yet leave the dead-code in the
> driver to handle these features.
> 
> And of course GRO was not removed, meaning the bench results were non
> conclusive.
> 
> 
> 

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[Eric Dumazet]

On Fri, 2017-09-15 at 17:10 -0700, rosenp@gmail.com wrote:
> Ok fair enough. Will only disable GRO in the driver.

Well, do not even try.

NETIF_F_SOFT_FEATURES is set by core networking stack in
register_netdevice(), ( commit 212b573f5552c60265da721ff9ce32e3462a2cdd
)

Absolutely no driver disables GRO (excepts the ones playing with LRO)

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[rosenp]

netif_receive_skb seems to have the same effect as "ethtool -K eth0 gro
off".

On Fri, 2017-09-15 at 17:18 -0700, Eric Dumazet wrote:
> On Fri, 2017-09-15 at 17:10 -0700, rosenp@gmail.com wrote:
> > Ok fair enough. Will only disable GRO in the driver.
> 
> Well, do not even try.
> 
> NETIF_F_SOFT_FEATURES is set by core networking stack in
> register_netdevice(), ( commit
> 212b573f5552c60265da721ff9ce32e3462a2cdd
> )
> 
> Absolutely no driver disables GRO (excepts the ones playing with LRO)
> 
> 
> 
> 
> 
> 

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[Denys Fedoryshchenko]

On 2017-09-16 03:18, Eric Dumazet wrote:
> On Fri, 2017-09-15 at 17:10 -0700, rosenp@gmail.com wrote:
>> Ok fair enough. Will only disable GRO in the driver.
> 
> Well, do not even try.
> 
> NETIF_F_SOFT_FEATURES is set by core networking stack in
> register_netdevice(), ( commit 212b573f5552c60265da721ff9ce32e3462a2cdd
> )
> 
> Absolutely no driver disables GRO (excepts the ones playing with LRO)
I believe also iperf is definitely inconclusive test.
Except iperf there is lot of different workloads and configurations, 
that might have different results.

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Fri, 15 Sep 2017 17:03:30 -0700

> On Fri, 2017-09-15 at 15:54 -0700, Florian Fainelli wrote:
>> On September 15, 2017 3:22:18 PM PDT, Rosen Penev <rosenp@gmail.com>
>> wrote:
>> >On a linksys E1200v1 (actually a crossflashed E1000v2), the
>> offloading
>> >features give no measurable benefit to speed or latency. Furthermore,
>> >disabling GRO actually improves iperf performance by a whoppimg
>> 3mbps.
>> 
>> Do you have a way to generate gigabit tests and see what results you
>> are getting? We probably are not going to see a 30% improvement just
>> by extrapolation.
>> 
> +1
> 
> It seems silly to remove NETIF_F_SG | NETIF_F_IP_CSUM |
> NETIF_F_IPV6_CSUM from dev->features, yet leave the dead-code in the
> driver to handle these features.
> 
> And of course GRO was not removed, meaning the bench results were non
> conclusive.

My sentiments exactly, I think this is a completely unwise change.

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[David Miller]

From: Rosen Penev <rosenp@gmail.com>
Date: Fri, 15 Sep 2017 15:22:18 -0700

> On a linksys E1200v1 (actually a crossflashed E1000v2), the offloading features give no measurable benefit to speed or latency. Furthermore, disabling GRO actually improves iperf performance by a whoppimg 3mbps. Results:
 ...
> -	net_dev->features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
> +	net_dev->features &= NETIF_F_GRO;
>  	net_dev->hw_features = net_dev->features;

This doesn't disable GRO.

Re: [PATCH] bgmac: Remove all offloading features, including GRO.

[rosenp]

you're absolutely correct. will send an updated version shortly.

On Fri, 2017-09-15 at 16:14 -0700, David Miller wrote:
> From: Rosen Penev <rosenp@gmail.com>
> Date: Fri, 15 Sep 2017 15:22:18 -0700
> 
> > On a linksys E1200v1 (actually a crossflashed E1000v2), the
> offloading features give no measurable benefit to speed or latency.
> Furthermore, disabling GRO actually improves iperf performance by a
> whoppimg 3mbps. Results:
>  ...
> > -     net_dev->features = NETIF_F_SG | NETIF_F_IP_CSUM |
> NETIF_F_IPV6_CSUM;
> > +     net_dev->features &= NETIF_F_GRO;
> >       net_dev->hw_features = net_dev->features;
> 
> This doesn't disable GRO.