[Intel-wired-lan] [PATCH net-next] net: ixgbe: Fix cls_u32 offload support for ports and fields with masks.

John Fastabend john.fastabend at gmail.com
Fri Mar 4 22:22:45 UTC 2016


On 16-03-04 01:27 PM, Samudrala, Sridhar wrote:
> On 3/4/2016 12:41 PM, John Fastabend wrote:
>> On 16-03-04 11:47 AM, Sridhar Samudrala wrote:
>>> Fix support for 16 bit source/dest port matches in ixgbe model.
>>> u32 uses a single 32-bit key value for both source and destination ports
>>> starting at offset 0. So replace the 2 functions with a single function
>>> that takes this key value/mask to program both source and dest ports.
>>>
>>> Remove the incorrect check for mask in ixgbe_configure_clsu32()
>>>
>>> Tested with the following filters:
>>>
>>>   #tc qdisc add dev p4p1 ingress
>>>   #tc filter add dev p4p1 parent ffff: protocol ip prio 99 \
>>>     handle 800:0:1 u32 ht 800: \
>>>     match ip dst 11.0.0.1/24 match ip src 11.0.0.2/24 action drop
>>>
>>>   #tc filter del dev p4p1 parent ffff: protocol ip prio 99 \
>>>     handle 800:0:1 u32
>>>   #tc filter add dev p4p1 parent ffff: protocol ip prio 99 \
>>>     handle 1: u32 divisor 1
>>>   #tc filter add dev p4p1 parent ffff: protocol ip prio 99 \
>>>     handle 800:0:10 u32 ht 800: link 1: \
>>>     offset at 0 mask 0f00 shift 6 plus 0 eat match ip protocol 6 ff
>>>   #tc filter add dev p4p1 parent ffff: protocol ip prio 99 \
>>>     handle 1:0:10 u32 ht 1: \
>>>     match tcp src 1024 ffff match tcp dst 80 ffff action drop
>>>
>>> Signed-off-by: Sridhar Samudrala <sridhar.samudrala at intel.com>
>>> ---
>> But this will break setting only dst port or only src port.
> 
> No. This will not break specifying only src or dst port. The value/mask
> for the
> unspecified port will be set to zero. So it should be fine.
> 
> For ex:
>     match tcp src 1024 ffff match tcp dst 80 ffff
>         => match 04000050/ffffffff at nexthdr+0
>     match tcp src 1024 ffff
>         => match 04000000/ffff0000 at nexthdr+0
>     match tcp dst 80 ffff
>         => match 00000050/0000ffff at nexthdr+0
> 
> 
>>   Do we
>> actually need three signatures to match? Something like,
>>
>>    static struct ixgbe_mat_field ixgbe_tcp_fields[] = {
>>     {.off = 0, .mask = 0xffffffff, .val = ixgbe_mat_prgm_ports,
>>         .type = IXGBE_ATR_FLOW_TYPE_TCPV4},
>>     {.off = 0, .mask = 0xffff0000, .val = ixgbe_mat_prgm_dport,
>>         .type = IXGBE_ATR_FLOW_TYPE_TCPV4},
>>     {.off = 0, .mask = 0x0000ffff, .val = ixgbe_mat_prgm_sport,
>>         .type = IXGBE_ATR_FLOW_TYPE_TCPV4},
>>        { .val = NULL } /* terminal node */
>>    };
>>
>> Also just a reminder if we get multiple fields in a ixgbe_mat_field
>> struct we need to abort out of the for loop in the cls_u32 configure
>> function. Actually we can probably just push that as its own patch
>> to make the core function more versatile/usable.
>>
>> Thanks,
>> John
> 

Dropped netdev no reason to push mail to netdev that has to do with our
driver internals.

OK I see so you dropped the .mask check in fact you actually put it
there in the commit message I just missed the detail/implication.

I think this is fine and it allows supporting mask entries now which
I blocked in the initial submission. But with this is there any reason
to have a mask field in ixgbe_mat_field? We could probably do this with
two patches, one to drop the 'mask' field and check and another to fix
the ixgbe_tcp_fields patch?

Thanks,
John


More information about the Intel-wired-lan mailing list