NETIF_F_TSO vs NETIF_F_TSO{6,_ECN}

NETIF_F_TSO vs NETIF_F_TSO{6,_ECN}

[Ben Hutchings]

According to the commit that introduced NETIF_F_TSO6
(f83ef8c0b58dac17211a4c0b6df0e2b1bd6637b1):
    
        This patch will introduce a new flag NETIF_F_TSO6 which will be used
        to check whether device supports TSO over IPv6. If device support TSO
        over IPv6 then we don't clear of NETIF_F_TSO and which will make the
        TCP layer to create TSO packets. Any device supporting TSO over IPv6
        will set NETIF_F_TSO6 flag in "dev->features" along with NETIF_F_TSO.
    
        In case when user disables TSO using ethtool, NETIF_F_TSO will get
        cleared from "dev->features". So even if we have NETIF_F_TSO6 we don't
        get TSO packets created by TCP layer.

So I think we need to either:
1. Disallow toggling NETIF_F_TSO6 (following the previous rule)
2. Disable NETIF_F_TSO6 when NETIF_F_TSO is disabled

The same presumably applies to NETIF_F_TSO_ECN.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: NETIF_F_TSO vs NETIF_F_TSO{6,_ECN}

[Michał Mirosław]

On Tue, Apr 05, 2011 at 04:50:02PM +0100, Ben Hutchings wrote:
> According to the commit that introduced NETIF_F_TSO6
> (f83ef8c0b58dac17211a4c0b6df0e2b1bd6637b1):
>     
>         This patch will introduce a new flag NETIF_F_TSO6 which will be used
>         to check whether device supports TSO over IPv6. If device support TSO
>         over IPv6 then we don't clear of NETIF_F_TSO and which will make the
>         TCP layer to create TSO packets. Any device supporting TSO over IPv6
>         will set NETIF_F_TSO6 flag in "dev->features" along with NETIF_F_TSO.
>     
>         In case when user disables TSO using ethtool, NETIF_F_TSO will get
>         cleared from "dev->features". So even if we have NETIF_F_TSO6 we don't
>         get TSO packets created by TCP layer.
> 
> So I think we need to either:
> 1. Disallow toggling NETIF_F_TSO6 (following the previous rule)
> 2. Disable NETIF_F_TSO6 when NETIF_F_TSO is disabled
> 
> The same presumably applies to NETIF_F_TSO_ECN.

There seems to be no such dependency in the networking code. I.e. TSO6
should just work with TSO4 disabled.

Best Regards,
Michał Mirosław

Re: NETIF_F_TSO vs NETIF_F_TSO{6,_ECN}

[Ben Hutchings]

On Tue, 2011-04-05 at 20:03 +0200, Michał Mirosław wrote:
> On Tue, Apr 05, 2011 at 04:50:02PM +0100, Ben Hutchings wrote:
> > According to the commit that introduced NETIF_F_TSO6
> > (f83ef8c0b58dac17211a4c0b6df0e2b1bd6637b1):
> >     
> >         This patch will introduce a new flag NETIF_F_TSO6 which will be used
> >         to check whether device supports TSO over IPv6. If device support TSO
> >         over IPv6 then we don't clear of NETIF_F_TSO and which will make the
> >         TCP layer to create TSO packets. Any device supporting TSO over IPv6
> >         will set NETIF_F_TSO6 flag in "dev->features" along with NETIF_F_TSO.
> >     
> >         In case when user disables TSO using ethtool, NETIF_F_TSO will get
> >         cleared from "dev->features". So even if we have NETIF_F_TSO6 we don't
> >         get TSO packets created by TCP layer.
> > 
> > So I think we need to either:
> > 1. Disallow toggling NETIF_F_TSO6 (following the previous rule)
> > 2. Disable NETIF_F_TSO6 when NETIF_F_TSO is disabled
> > 
> > The same presumably applies to NETIF_F_TSO_ECN.
> 
> There seems to be no such dependency in the networking code. I.e. TSO6
> should just work with TSO4 disabled.

*sigh*  So it seems the commit message was wrong... and it should have
included a change like this:

--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -5203,9 +5203,9 @@ u32 netdev_fix_features(struct net_device *dev, u32 features)
 	}
 
 	/* TSO requires that SG is present as well. */
-	if ((features & NETIF_F_TSO) && !(features & NETIF_F_SG)) {
-		netdev_info(dev, "Dropping NETIF_F_TSO since no SG feature.\n");
-		features &= ~NETIF_F_TSO;
+	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
+		netdev_info(dev, "Dropping TSO since no SG feature.\n");
+		features &= ~NETIF_F_ALL_TSO;
 	}
 
 	/* Software GSO depends on SG. */
---

Now that we're relying on these checks for dynamic changes to features,
this is pretty important.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

Re: NETIF_F_TSO vs NETIF_F_TSO{6,_ECN}

[David Miller]

From: Ben Hutchings <bhutchings@solarflare.com>
Date: Tue, 05 Apr 2011 19:31:38 +0100

> *sigh*  So it seems the commit message was wrong... and it should have
> included a change like this:

This seems correct, can you formally submit this fix?

Thanks.