[PATCH] net: EMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's

[PATCH] net: EMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's

[Stefan Roese]

Currently, all non TAH equipped 4xx PPC's call emac_start_xmit() upon
xmit. This routine doesn't check if the frame length exceeds the max.
MAL buffer size.

This patch now changes the driver to call emac_start_xmit_sg() on all
platforms and not only the TAH equipped ones (440GX). This enables an
MTU of 9000 instead 4080.

Tested on Kilauea (405EX) with gbit link -> jumbo frames enabled.

Signed-off-by: Stefan Roese <sr@denx.de>
---
Eugene & Ben, do you see any problems with this patch? If not, then I'll
send another version for the newemac driver too.

Thanks.

 drivers/net/ibm_emac/ibm_emac_core.c |   10 ++--------
 1 files changed, 2 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ibm_emac/ibm_emac_core.c b/drivers/net/ibm_emac/ibm_emac_core.c
index 73664f2..198de44 100644
--- a/drivers/net/ibm_emac/ibm_emac_core.c
+++ b/drivers/net/ibm_emac/ibm_emac_core.c
@@ -1089,7 +1089,6 @@ static int emac_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 	return emac_xmit_finish(dev, len);
 }
 
-#if defined(CONFIG_IBM_EMAC_TAH)
 static inline int emac_xmit_split(struct ocp_enet_private *dev, int slot,
 				  u32 pd, int len, int last, u16 base_ctrl)
 {
@@ -1203,9 +1202,6 @@ static int emac_start_xmit_sg(struct sk_buff *skb, struct net_device *ndev)
 	DBG2("%d: stopped TX queue" NL, dev->def->index);
 	return 1;
 }
-#else
-# define emac_start_xmit_sg	emac_start_xmit
-#endif	/* !defined(CONFIG_IBM_EMAC_TAH) */
 
 /* BHs disabled */
 static void emac_parse_tx_error(struct ocp_enet_private *dev, u16 ctrl)
@@ -2163,11 +2159,9 @@ static int __init emac_probe(struct ocp_device *ocpdev)
 
 	/* Fill in the driver function table */
 	ndev->open = &emac_open;
-	if (dev->tah_dev) {
-		ndev->hard_start_xmit = &emac_start_xmit_sg;
+	ndev->hard_start_xmit = &emac_start_xmit_sg;
+	if (dev->tah_dev)
 		ndev->features |= NETIF_F_IP_CSUM | NETIF_F_SG;
-	} else
-		ndev->hard_start_xmit = &emac_start_xmit;
 	ndev->tx_timeout = &emac_full_tx_reset;
 	ndev->watchdog_timeo = 5 * HZ;
 	ndev->stop = &emac_close;
-- 
1.5.4.rc3

Re: [PATCH] net: EMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's

[Eugene Surovegin]

On Tue, Jan 15, 2008 at 01:40:09PM +0100, Stefan Roese wrote:
> Currently, all non TAH equipped 4xx PPC's call emac_start_xmit() upon
> xmit. This routine doesn't check if the frame length exceeds the max.
> MAL buffer size.
> 
> This patch now changes the driver to call emac_start_xmit_sg() on all
> platforms and not only the TAH equipped ones (440GX). This enables an
> MTU of 9000 instead 4080.
> 
> Tested on Kilauea (405EX) with gbit link -> jumbo frames enabled.
> 
> Signed-off-by: Stefan Roese <sr@denx.de>
> ---
> Eugene & Ben, do you see any problems with this patch? If not, then I'll
> send another version for the newemac driver too.

Hmm, so why not make GigE support a condition to hook SG version of 
xmit then? I don't like when you change behaviour for chips where it 
perefectly legal not to do this check because you cannot change MTU 
anyways.

-- 
Eugene

Re: [PATCH] net: EMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's

[Stefan Roese]

On Tuesday 15 January 2008, Eugene Surovegin wrote:
> On Tue, Jan 15, 2008 at 01:40:09PM +0100, Stefan Roese wrote:
> > Currently, all non TAH equipped 4xx PPC's call emac_start_xmit() upon
> > xmit. This routine doesn't check if the frame length exceeds the max.
> > MAL buffer size.
> >
> > This patch now changes the driver to call emac_start_xmit_sg() on all
> > platforms and not only the TAH equipped ones (440GX). This enables an
> > MTU of 9000 instead 4080.
> >
> > Tested on Kilauea (405EX) with gbit link -> jumbo frames enabled.
> >
> > Signed-off-by: Stefan Roese <sr@denx.de>
> > ---
> > Eugene & Ben, do you see any problems with this patch? If not, then I'll
> > send another version for the newemac driver too.
>
> Hmm, so why not make GigE support a condition to hook SG version of
> xmit then? I don't like when you change behaviour for chips where it
> perefectly legal not to do this check because you cannot change MTU
> anyways.

OK. But how do we detect GigE support? Seems like GigE enabled devices have 
CONFIG_IBM_EMAC4 defined. If nobody objects I'll fix up another version 
tomorrow.

Thanks.

Best regards,
Stefan

Re: [PATCH] net: EMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's

[Eugene Surovegin]

On Tue, Jan 15, 2008 at 08:46:01PM +0100, Stefan Roese wrote:
> On Tuesday 15 January 2008, Eugene Surovegin wrote:
> > On Tue, Jan 15, 2008 at 01:40:09PM +0100, Stefan Roese wrote:
> > > Currently, all non TAH equipped 4xx PPC's call emac_start_xmit() upon
> > > xmit. This routine doesn't check if the frame length exceeds the max.
> > > MAL buffer size.
> > >
> > > This patch now changes the driver to call emac_start_xmit_sg() on all
> > > platforms and not only the TAH equipped ones (440GX). This enables an
> > > MTU of 9000 instead 4080.
> > >
> > > Tested on Kilauea (405EX) with gbit link -> jumbo frames enabled.
> > >
> > > Signed-off-by: Stefan Roese <sr@denx.de>
> > > ---
> > > Eugene & Ben, do you see any problems with this patch? If not, then I'll
> > > send another version for the newemac driver too.
> >
> > Hmm, so why not make GigE support a condition to hook SG version of
> > xmit then? I don't like when you change behaviour for chips where it
> > perefectly legal not to do this check because you cannot change MTU
> > anyways.
> 
> OK. But how do we detect GigE support? Seems like GigE enabled devices have 
> CONFIG_IBM_EMAC4 defined. If nobody objects I'll fix up another version 
> tomorrow.

Look couple of lines down where I set MTU changing hook. If you cannot 
change MTU you cannot get big frames.

-- 
Eugene

Re: [PATCH] net: EMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's

[Stefan Roese]

On Tuesday 15 January 2008, Eugene Surovegin wrote:
> > OK. But how do we detect GigE support? Seems like GigE enabled devices
> > have CONFIG_IBM_EMAC4 defined. If nobody objects I'll fix up another
> > version tomorrow.
>
> Look couple of lines down where I set MTU changing hook. If you cannot
> change MTU you cannot get big frames.

Ahhh, I must have been blind. Thanks for pointing out.

I'll send new patches for ibm_emac and ibm_newemac in a short while.

Best regards,
Stefan

[PATCH] net: EMAC: Fix problem with mtu > 4080 on non TAH equipped 4xx PPC's

[Stefan Roese]

Currently, all non TAH equipped 4xx PPC's call emac_start_xmit() upon
xmit. This routine doesn't check if the frame length exceeds the max.
MAL buffer size.

This patch now changes the driver to call emac_start_xmit_sg() on all
GigE platforms and not only the TAH equipped ones (440GX). This enables
an MTU of 9000 instead 4080.

Tested on Kilauea (405EX) with GigE link -> jumbo frames enabled.

Signed-off-by: Stefan Roese <sr@denx.de>
---
 drivers/net/ibm_emac/ibm_emac_core.c |   12 ++++--------
 1 files changed, 4 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ibm_emac/ibm_emac_core.c b/drivers/net/ibm_emac/ibm_emac_core.c
index 73664f2..8d1901e 100644
--- a/drivers/net/ibm_emac/ibm_emac_core.c
+++ b/drivers/net/ibm_emac/ibm_emac_core.c
@@ -1089,7 +1089,6 @@ static int emac_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 	return emac_xmit_finish(dev, len);
 }
 
-#if defined(CONFIG_IBM_EMAC_TAH)
 static inline int emac_xmit_split(struct ocp_enet_private *dev, int slot,
 				  u32 pd, int len, int last, u16 base_ctrl)
 {
@@ -1203,9 +1202,6 @@ static int emac_start_xmit_sg(struct sk_buff *skb, struct net_device *ndev)
 	DBG2("%d: stopped TX queue" NL, dev->def->index);
 	return 1;
 }
-#else
-# define emac_start_xmit_sg	emac_start_xmit
-#endif	/* !defined(CONFIG_IBM_EMAC_TAH) */
 
 /* BHs disabled */
 static void emac_parse_tx_error(struct ocp_enet_private *dev, u16 ctrl)
@@ -2163,11 +2159,8 @@ static int __init emac_probe(struct ocp_device *ocpdev)
 
 	/* Fill in the driver function table */
 	ndev->open = &emac_open;
-	if (dev->tah_dev) {
-		ndev->hard_start_xmit = &emac_start_xmit_sg;
+	if (dev->tah_dev)
 		ndev->features |= NETIF_F_IP_CSUM | NETIF_F_SG;
-	} else
-		ndev->hard_start_xmit = &emac_start_xmit;
 	ndev->tx_timeout = &emac_full_tx_reset;
 	ndev->watchdog_timeo = 5 * HZ;
 	ndev->stop = &emac_close;
@@ -2175,8 +2168,11 @@ static int __init emac_probe(struct ocp_device *ocpdev)
 	ndev->set_multicast_list = &emac_set_multicast_list;
 	ndev->do_ioctl = &emac_ioctl;
 	if (emac_phy_supports_gige(emacdata->phy_mode)) {
+		ndev->hard_start_xmit = &emac_start_xmit_sg;
 		ndev->change_mtu = &emac_change_mtu;
 		dev->commac.ops = &emac_commac_sg_ops;
+	} else {
+		ndev->hard_start_xmit = &emac_start_xmit;
 	}
 	SET_ETHTOOL_OPS(ndev, &emac_ethtool_ops);
 
-- 
1.5.4.rc3