* SUN Neptune byte order bug in TCAM filter/ethtool
@ 2009-05-06 16:16 Robert Olsson
2009-06-10 8:26 ` David Miller
0 siblings, 1 reply; 2+ messages in thread
From: Robert Olsson @ 2009-05-06 16:16 UTC (permalink / raw)
To: netdev; +Cc: David, Miller <davem, Matheos.Worku, Jens.Laas, robert
David Miller writes:
> > FYI byte order seems reversed on Intel meaning you have to reverse
> > byte order for address and ports when adding TCAM entries via
> > ethtool.
>
> Please report this formally with a reproduction script/commands/etc.
> with netdev CC:'d as this is a very serious bug and exactly what I
> tried to avoid have happen when I was so anal about the endian
> oriented typing of the ethtool structures. :-//////////
First a disclaimer this our production system 2.6.29-rc2 but I haven't
seen any patches since.
# This rule makes the TCAM drop pkts with IP-src 10.10.10.1
# Note the reversed byte-order. Machine is Opteron little Endian.
ethtool -N eth0 rx-class-rule-add ip4 udp sip 1.10.10.10 drop
# The rule listed
ethtool -n eth0 rx-class-rule-all
Total 1 rules
IPv4 Rule: ID[84] Target[Discard]
IP src addr[1.10.10.10] mask[255.255.255.255]
IP dst addr[0.0.0.0] mask[0.0.0.0]
L4 proto[UDP]
L4 src port[0] mask[0x0]
L4 dst port[0] mask[0x0]
IP TOS[0x0] mask[0x0]
# The rule deleted
ethtool -N eth0 rx-class-rule-del 84
Cheers
--ro
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: SUN Neptune byte order bug in TCAM filter/ethtool
2009-05-06 16:16 SUN Neptune byte order bug in TCAM filter/ethtool Robert Olsson
@ 2009-06-10 8:26 ` David Miller
0 siblings, 0 replies; 2+ messages in thread
From: David Miller @ 2009-06-10 8:26 UTC (permalink / raw)
To: robert; +Cc: netdev, davem, Matheos.Worku, Jens.Laas
From: Robert Olsson <robert@robur.slu.se>
Date: Wed, 6 May 2009 18:16:07 +0200
> # This rule makes the TCAM drop pkts with IP-src 10.10.10.1
> # Note the reversed byte-order. Machine is Opteron little Endian.
>
> ethtool -N eth0 rx-class-rule-add ip4 udp sip 1.10.10.10 drop
>
> # The rule listed
>
> ethtool -n eth0 rx-class-rule-all
> Total 1 rules
>
> IPv4 Rule: ID[84] Target[Discard]
> IP src addr[1.10.10.10] mask[255.255.255.255]
> IP dst addr[0.0.0.0] mask[0.0.0.0]
> L4 proto[UDP]
> L4 src port[0] mask[0x0]
> L4 dst port[0] mask[0x0]
> IP TOS[0x0] mask[0x0]
I suspect that ethtool is probably to blame, the kernel seems
sane and treats the IP addresses, ports, etc. as big-endian.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2009-06-10 16:58 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-05-06 16:16 SUN Neptune byte order bug in TCAM filter/ethtool Robert Olsson
2009-06-10 8:26 ` David Miller
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.