From: Amit Kumar Salecha <hidden> Date: 2011-03-21 10:37:23
netxen_nic_set_flags() rejects data if other flag than ETH_FLAG_LRO is set.
Driver also supports NETIF_F_HW_VLAN_TX.
Now compare data with ethtool_op_get_flags(), to get all supported features.
Reported-by: Jesper Dangaard Brouer <redacted>
Signed-off-by: Rajesh Borundia <redacted>
Signed-off-by: Amit Kumar Salecha <redacted>
---
drivers/net/netxen/netxen_nic_ethtool.c | 3 ++-
1 files changed, 2 insertions(+), 1 deletions(-)
From: Eric Dumazet <hidden> Date: 2011-03-21 10:47:58
Le lundi 21 mars 2011 à 03:37 -0700, Amit Kumar Salecha a écrit :
netxen_nic_set_flags() rejects data if other flag than ETH_FLAG_LRO is set.
Driver also supports NETIF_F_HW_VLAN_TX.
Now compare data with ethtool_op_get_flags(), to get all supported features.
Reported-by: Jesper Dangaard Brouer <redacted>
Signed-off-by: Rajesh Borundia <redacted>
Signed-off-by: Amit Kumar Salecha <redacted>
---
Hmm, it would be really nice you provide more information for stable
teams.
If I am not mistaken, bug was introduced in 2.6.36 ?
Hmm, it would be really nice you provide more information for stable
teams.
If I am not mistaken, bug was introduced in 2.6.36 ?
No, It was introduced in 2.6.37, when ETH_FLAG_TXVLAN and ETH_FLAG_RXVLAN introduced.
Yes it should also make into 2.6.37.4 stable kernel.
This message and any attached documents contain information from QLogic Corporation or its wholly-owned subsidiaries that may be confidential. If you are not the intended recipient, you may not read, copy, distribute, or use this information. If you have received this transmission in error, please notify the sender immediately by reply e-mail and then delete this message.
Hmm, it would be really nice you provide more information for stable
teams.
If I am not mistaken, bug was introduced in 2.6.36 ?
No, It was introduced in 2.6.37, when ETH_FLAG_TXVLAN and
ETH_FLAG_RXVLAN introduced.
quoted
Yes it should also make into 2.6.37.4 stable kernel.
Applied, but 2.6.37.x maintainence has ceased so there is no point
worrying about this patch for that stable branch any longer.
I thought you will apply patch from Stanislaw Gruszka (net: fix ethtool->set_flags not intended -EINVAL return value), which is more generic and takes care of all drivers.
David, you are not cc in that patch. I will forward you, in case you miss it.
-Amit
This message and any attached documents contain information from QLogic Corporation or its wholly-owned subsidiaries that may be confidential. If you are not the intended recipient, you may not read, copy, distribute, or use this information. If you have received this transmission in error, please notify the sender immediately by reply e-mail and then delete this message.
Hmm, it would be really nice you provide more information for stable
teams.
If I am not mistaken, bug was introduced in 2.6.36 ?
No, It was introduced in 2.6.37, when ETH_FLAG_TXVLAN and
ETH_FLAG_RXVLAN introduced.
quoted
Yes it should also make into 2.6.37.4 stable kernel.
Applied, but 2.6.37.x maintainence has ceased so there is no point
worrying about this patch for that stable branch any longer.
2.6.38 needs that fix as well ...
I thought you will apply patch from Stanislaw Gruszka (net: fix ethtool->set_flags not intended -EINVAL return value), which is more generic and takes care of all drivers.
David, you are not cc in that patch. I will forward you, in case you miss it.
From: David Miller <davem@davemloft.net> Date: 2011-03-28 06:29:51
From: Amit Salecha <redacted>
Date: Mon, 28 Mar 2011 00:33:43 -0500
I thought you will apply patch from Stanislaw Gruszka (net: fix
ethtool->set_flags not intended -EINVAL return value), which is more
generic and takes care of all drivers.
David, you are not cc in that patch. I will forward you, in case you miss it.
Ok, I was wondering about that, thanks for explaining I'll fix this up.
From: Marc Haber <hidden> Date: 2011-04-03 19:14:21
Hi,
On Mon, Mar 21, 2011 at 03:37:08AM -0700, Amit Kumar Salecha wrote:
netxen_nic_set_flags() rejects data if other flag than ETH_FLAG_LRO is set.
Driver also supports NETIF_F_HW_VLAN_TX.
Now compare data with ethtool_op_get_flags(), to get all supported features.
Could that be the cause for packet loss on kernel 2.6.38.2 if:
- receiving card is NX3031 [4040:0100]
- frames are received with VLAN tags
- large received offload is on.
Packet Loss of this kind is noticed when doing TCP data transfers
towards the host with the Netxen Interface and the TCP session is
terminated on the Netxen host itself. TCP sessions routed through the
Netxen host are not affected.
My ethtool doesn't allow me to influence the LRO setting alone - it is
disabled when I set rx off but doesn't come on again when rx is set to
on again. So, ethtool -K rx off, ethtool -K rx on fixes the issue.
Is this a known bug, maybe with an available patch?
Greetings
Marc
--
-----------------------------------------------------------------------------
Marc Haber | "I don't trust Computers. They | Mailadresse im Header
Mannheim, Germany | lose things." Winona Ryder | Fon: *49 621 72739834
Nordisch by Nature | How to make an American Quilt | Fax: *49 3221 2323190
From: Amit Salecha <hidden> Date: 2011-04-05 05:38:20
Hi,
On Mon, Mar 21, 2011 at 03:37:08AM -0700, Amit Kumar Salecha wrote:
quoted
netxen_nic_set_flags() rejects data if other flag than ETH_FLAG_LRO
is set.
quoted
Driver also supports NETIF_F_HW_VLAN_TX.
Now compare data with ethtool_op_get_flags(), to get all supported
features.
Could that be the cause for packet loss on kernel 2.6.38.2 if:
- receiving card is NX3031 [4040:0100]
- frames are received with VLAN tags
- large received offload is on.
If ip_forwarding or routing is enable ....then you may see packet loss.
Packet Loss of this kind is noticed when doing TCP data transfers
towards the host with the Netxen Interface and the TCP session is
terminated on the Netxen host itself. TCP sessions routed through the
Netxen host are not affected.
My ethtool doesn't allow me to influence the LRO setting alone - it is
disabled when I set rx off but doesn't come on again when rx is set to
on again. So, ethtool -K rx off, ethtool -K rx on fixes the issue.
If rx csum is disabled, LRO will be disable. LRO won't be enabled automatically if you enable rx csum.
You need to explicitly enable LRO.
Is this a known bug, maybe with an available patch?
You need to retest with this patch http://patchwork.ozlabs.org/patch/88060/. This patch got applied instead of mine.
This message and any attached documents contain information from QLogic Corporation or its wholly-owned subsidiaries that may be confidential. If you are not the intended recipient, you may not read, copy, distribute, or use this information. If you have received this transmission in error, please notify the sender immediately by reply e-mail and then delete this message.
From: Michał Mirosław <hidden> Date: 2011-04-05 10:38:29
2011/4/5 Amit Salecha [off-list ref]:
quoted
On Mon, Mar 21, 2011 at 03:37:08AM -0700, Amit Kumar Salecha wrote:
quoted
netxen_nic_set_flags() rejects data if other flag than ETH_FLAG_LRO
is set.
quoted
Driver also supports NETIF_F_HW_VLAN_TX.
Now compare data with ethtool_op_get_flags(), to get all supported
features.
[...]
quoted
My ethtool doesn't allow me to influence the LRO setting alone - it is
disabled when I set rx off but doesn't come on again when rx is set to
on again. So, ethtool -K rx off, ethtool -K rx on fixes the issue.
If rx csum is disabled, LRO will be disable. LRO won't be enabled automatically if you enable rx csum.
You need to explicitly enable LRO.
This will change once the driver is converted to hw_features.
Best Regards,
Michał Mirosław
From: Marc Haber <hidden> Date: 2011-04-05 12:41:59
On Tue, Apr 05, 2011 at 12:38:11AM -0500, Amit Salecha wrote:
quoted
On Mon, Mar 21, 2011 at 03:37:08AM -0700, Amit Kumar Salecha wrote:
quoted
netxen_nic_set_flags() rejects data if other flag than ETH_FLAG_LRO
is set.
quoted
Driver also supports NETIF_F_HW_VLAN_TX.
Now compare data with ethtool_op_get_flags(), to get all supported
features.
Could that be the cause for packet loss on kernel 2.6.38.2 if:
- receiving card is NX3031 [4040:0100]
- frames are received with VLAN tags
- large received offload is on.
If ip_forwarding or routing is enable ....then you may see packet loss.
The box is intended to route, so disabling routing is
contraproductive. I just would like to download software updates to
the box with decent speed as well.
quoted
Packet Loss of this kind is noticed when doing TCP data transfers
towards the host with the Netxen Interface and the TCP session is
terminated on the Netxen host itself. TCP sessions routed through the
Netxen host are not affected.
My ethtool doesn't allow me to influence the LRO setting alone - it is
disabled when I set rx off but doesn't come on again when rx is set to
on again. So, ethtool -K rx off, ethtool -K rx on fixes the issue.
If rx csum is disabled, LRO will be disable. LRO won't be enabled automatically if you enable rx csum.
You need to explicitly enable LRO.
Explicitly enabling LRO does not work ("invalid argument", if I recall
correctly).
quoted
Is this a known bug, maybe with an available patch?
Will that patch fix the behavior of the interface regarding the packet
loss, or only its connection to ethtool?
Greetings
Marc
--
-----------------------------------------------------------------------------
Marc Haber | "I don't trust Computers. They | Mailadresse im Header
Mannheim, Germany | lose things." Winona Ryder | Fon: *49 621 72739834
Nordisch by Nature | How to make an American Quilt | Fax: *49 3221 2323190
From: Amit Salecha <hidden> Date: 2011-04-05 13:15:55
On Tue, Apr 05, 2011 at 12:38:11AM -0500, Amit Salecha wrote:
quoted
quoted
On Mon, Mar 21, 2011 at 03:37:08AM -0700, Amit Kumar Salecha wrote:
quoted
netxen_nic_set_flags() rejects data if other flag than
ETH_FLAG_LRO
quoted
quoted
is set.
quoted
Driver also supports NETIF_F_HW_VLAN_TX.
Now compare data with ethtool_op_get_flags(), to get all
supported
quoted
quoted
features.
Could that be the cause for packet loss on kernel 2.6.38.2 if:
- receiving card is NX3031 [4040:0100]
- frames are received with VLAN tags
- large received offload is on.
If ip_forwarding or routing is enable ....then you may see packet
loss.
The box is intended to route, so disabling routing is
contraproductive. I just would like to download software updates to
the box with decent speed as well.
What I meant, with LRO enable and routing, you may see packet loss.
quoted
quoted
Packet Loss of this kind is noticed when doing TCP data transfers
towards the host with the Netxen Interface and the TCP session is
terminated on the Netxen host itself. TCP sessions routed through
the
quoted
quoted
Netxen host are not affected.
My ethtool doesn't allow me to influence the LRO setting alone - it
is
quoted
quoted
disabled when I set rx off but doesn't come on again when rx is set
to
quoted
quoted
on again. So, ethtool -K rx off, ethtool -K rx on fixes the issue.
If rx csum is disabled, LRO will be disable. LRO won't be enabled
automatically if you enable rx csum.
quoted
You need to explicitly enable LRO.
Explicitly enabling LRO does not work ("invalid argument", if I recall
correctly).
With below patch enabling/disabling LRO will work.
quoted
quoted
Is this a known bug, maybe with an available patch?
Will that patch fix the behavior of the interface regarding the packet
loss, or only its connection to ethtool?
This will fix LRO configuration problem. Do you see packet loss with LRO disable ?
This message and any attached documents contain information from QLogic Corporation or its wholly-owned subsidiaries that may be confidential. If you are not the intended recipient, you may not read, copy, distribute, or use this information. If you have received this transmission in error, please notify the sender immediately by reply e-mail and then delete this message.
From: Marc Haber <hidden> Date: 2011-04-05 15:18:55
On Tue, Apr 05, 2011 at 08:15:16AM -0500, Amit Salecha wrote:
This will fix LRO configuration problem. Do you see packet loss with
LRO disable ?
No, with LRO disabled things seem to be fine. LRO gets enabled by
default though.
Greetings
Marc
--
-----------------------------------------------------------------------------
Marc Haber | "I don't trust Computers. They | Mailadresse im Header
Mannheim, Germany | lose things." Winona Ryder | Fon: *49 621 72739834
Nordisch by Nature | How to make an American Quilt | Fax: *49 3221 2323190
From: Stephen Hemminger <hidden> Date: 2011-04-05 15:46:57
On Tue, 5 Apr 2011 17:18:54 +0200
Marc Haber [off-list ref] wrote:
On Tue, Apr 05, 2011 at 08:15:16AM -0500, Amit Salecha wrote:
quoted
This will fix LRO configuration problem. Do you see packet loss with
LRO disable ?
No, with LRO disabled things seem to be fine. LRO gets enabled by
default though.
LRO and routing are fundamentally incompatible, that is why
the kernel attempts to turn it off. When forwarding packets
should not be combined (end-to-end principle) and that is what
LRO does.
Therefore if doing anything like bridging or forwarding kernel
attempts to get driver to turn LRO off.