mbox series

[v3,0/2] net/failsafe: support multicast MAC address set

Message ID 1537544182-17823-1-git-send-email-arybchenko@solarflare.com (mailing list archive)
Headers
Series net/failsafe: support multicast MAC address set |

Message

Andrew Rybchenko Sept. 21, 2018, 3:36 p.m. UTC 
  v3:
    - move apply on sync to fs_eth_dev_conf_apply() to apply to
      a new subdevice only
    - use ethdev API to apply to sub-device on sync
    - remove unnecessary check the same pointer from the method
      implementation in failsafe

v2:
    - fix setting of zero addresses since rte_realloc() returns NULL

Evgeny Im (2):
  net/failsafe: remove not supported multicast MAC filter
  net/failsafe: support multicast address list set

 doc/guides/rel_notes/release_18_11.rst  |  1 +
 drivers/net/failsafe/failsafe.c         |  1 +
 drivers/net/failsafe/failsafe_ether.c   | 17 +++++++++
 drivers/net/failsafe/failsafe_ops.c     | 50 +++++++++++++++++++++++++
 drivers/net/failsafe/failsafe_private.h |  2 +
 5 files changed, 71 insertions(+)

Comments

Andrew Rybchenko Sept. 21, 2018, 3:36 p.m. UTC | #1 
v3:
    - move apply on sync to fs_eth_dev_conf_apply() to apply to
      a new subdevice only
    - use ethdev API to apply to sub-device on sync
    - remove unnecessary check the same pointer from the method
      implementation in failsafe

v2:
    - fix setting of zero addresses since rte_realloc() returns NULL

Evgeny Im (2):
  net/failsafe: remove not supported multicast MAC filter
  net/failsafe: support multicast address list set

 doc/guides/rel_notes/release_18_11.rst  |  7 ++++
 drivers/net/failsafe/failsafe.c         |  1 +
 drivers/net/failsafe/failsafe_ether.c   | 17 +++++++++
 drivers/net/failsafe/failsafe_ops.c     | 50 +++++++++++++++++++++++++
 drivers/net/failsafe/failsafe_private.h |  2 +
 5 files changed, 77 insertions(+)
Gaëtan Rivet Sept. 21, 2018, 4:09 p.m. UTC | #2 
Hi,

Seems good, thanks.

Acked-by: Gaetan Rivet <gaetan.rivet@6wind.com>

On Fri, Sep 21, 2018 at 04:36:20PM +0100, Andrew Rybchenko wrote:
> v3:
>     - move apply on sync to fs_eth_dev_conf_apply() to apply to
>       a new subdevice only
>     - use ethdev API to apply to sub-device on sync
>     - remove unnecessary check the same pointer from the method
>       implementation in failsafe
> 
> v2:
>     - fix setting of zero addresses since rte_realloc() returns NULL
> 
> Evgeny Im (2):
>   net/failsafe: remove not supported multicast MAC filter
>   net/failsafe: support multicast address list set
> 
>  doc/guides/rel_notes/release_18_11.rst  |  7 ++++
>  drivers/net/failsafe/failsafe.c         |  1 +
>  drivers/net/failsafe/failsafe_ether.c   | 17 +++++++++
>  drivers/net/failsafe/failsafe_ops.c     | 50 +++++++++++++++++++++++++
>  drivers/net/failsafe/failsafe_private.h |  2 +
>  5 files changed, 77 insertions(+)
> 
> -- 
> 2.17.1
>
Stephen Hemminger Sept. 21, 2018, 4:21 p.m. UTC | #3 
On Fri, 21 Sep 2018 16:36:19 +0100
Andrew Rybchenko <arybchenko@solarflare.com> wrote:

> v3:
>     - move apply on sync to fs_eth_dev_conf_apply() to apply to
>       a new subdevice only
>     - use ethdev API to apply to sub-device on sync
>     - remove unnecessary check the same pointer from the method
>       implementation in failsafe
> 
> v2:
>     - fix setting of zero addresses since rte_realloc() returns NULL
> 
> Evgeny Im (2):
>   net/failsafe: remove not supported multicast MAC filter
>   net/failsafe: support multicast address list set
> 
>  doc/guides/rel_notes/release_18_11.rst  |  1 +
>  drivers/net/failsafe/failsafe.c         |  1 +
>  drivers/net/failsafe/failsafe_ether.c   | 17 +++++++++
>  drivers/net/failsafe/failsafe_ops.c     | 50 +++++++++++++++++++++++++
>  drivers/net/failsafe/failsafe_private.h |  2 +
>  5 files changed, 71 insertions(+)
> 

Since TAP has no filtering this doesn't really do much for the slowpath.
Andrew Rybchenko Sept. 21, 2018, 4:33 p.m. UTC | #4 
On 9/21/18 7:21 PM, Stephen Hemminger wrote:
> On Fri, 21 Sep 2018 16:36:19 +0100
> Andrew Rybchenko <arybchenko@solarflare.com> wrote:
>
>> v3:
>>      - move apply on sync to fs_eth_dev_conf_apply() to apply to
>>        a new subdevice only
>>      - use ethdev API to apply to sub-device on sync
>>      - remove unnecessary check the same pointer from the method
>>        implementation in failsafe
>>
>> v2:
>>      - fix setting of zero addresses since rte_realloc() returns NULL
>>
>> Evgeny Im (2):
>>    net/failsafe: remove not supported multicast MAC filter
>>    net/failsafe: support multicast address list set
>>
>>   doc/guides/rel_notes/release_18_11.rst  |  1 +
>>   drivers/net/failsafe/failsafe.c         |  1 +
>>   drivers/net/failsafe/failsafe_ether.c   | 17 +++++++++
>>   drivers/net/failsafe/failsafe_ops.c     | 50 +++++++++++++++++++++++++
>>   drivers/net/failsafe/failsafe_private.h |  2 +
>>   5 files changed, 71 insertions(+)
>>
> Since TAP has no filtering this doesn't really do much for the slowpath.

Yes, I know that TAP does not support it. Is failsafe so tighly bound to 
TAP to block it?
Andrew Rybchenko Sept. 21, 2018, 4:38 p.m. UTC | #5 
On 9/21/18 7:33 PM, Andrew Rybchenko wrote:
> On 9/21/18 7:21 PM, Stephen Hemminger wrote:
>> On Fri, 21 Sep 2018 16:36:19 +0100
>> Andrew Rybchenko <arybchenko@solarflare.com> wrote:
>>
>>> v3:
>>>      - move apply on sync to fs_eth_dev_conf_apply() to apply to
>>>        a new subdevice only
>>>      - use ethdev API to apply to sub-device on sync
>>>      - remove unnecessary check the same pointer from the method
>>>        implementation in failsafe
>>>
>>> v2:
>>>      - fix setting of zero addresses since rte_realloc() returns NULL
>>>
>>> Evgeny Im (2):
>>>    net/failsafe: remove not supported multicast MAC filter
>>>    net/failsafe: support multicast address list set
>>>
>>>   doc/guides/rel_notes/release_18_11.rst  |  1 +
>>>   drivers/net/failsafe/failsafe.c         |  1 +
>>>   drivers/net/failsafe/failsafe_ether.c   | 17 +++++++++
>>>   drivers/net/failsafe/failsafe_ops.c     | 50 
>>> +++++++++++++++++++++++++
>>>   drivers/net/failsafe/failsafe_private.h |  2 +
>>>   5 files changed, 71 insertions(+)
>>>
>> Since TAP has no filtering this doesn't really do much for the slowpath.
>
> Yes, I know that TAP does not support it. Is failsafe so tighly bound 
> to TAP to block it?

In fact I've put it incorrectly. TAP supports multicast address set, but 
simply ignores
it since "every packet is received". But it is still makes sense for 
other failsafe
sub-devices which really support it.
Stephen Hemminger Sept. 21, 2018, 4:43 p.m. UTC | #6 
On Fri, 21 Sep 2018 19:33:16 +0300
Andrew Rybchenko <arybchenko@solarflare.com> wrote:

> On 9/21/18 7:21 PM, Stephen Hemminger wrote:
> > On Fri, 21 Sep 2018 16:36:19 +0100
> > Andrew Rybchenko <arybchenko@solarflare.com> wrote:
> >  
> >> v3:
> >>      - move apply on sync to fs_eth_dev_conf_apply() to apply to
> >>        a new subdevice only
> >>      - use ethdev API to apply to sub-device on sync
> >>      - remove unnecessary check the same pointer from the method
> >>        implementation in failsafe
> >>
> >> v2:
> >>      - fix setting of zero addresses since rte_realloc() returns NULL
> >>
> >> Evgeny Im (2):
> >>    net/failsafe: remove not supported multicast MAC filter
> >>    net/failsafe: support multicast address list set
> >>
> >>   doc/guides/rel_notes/release_18_11.rst  |  1 +
> >>   drivers/net/failsafe/failsafe.c         |  1 +
> >>   drivers/net/failsafe/failsafe_ether.c   | 17 +++++++++
> >>   drivers/net/failsafe/failsafe_ops.c     | 50 +++++++++++++++++++++++++
> >>   drivers/net/failsafe/failsafe_private.h |  2 +
> >>   5 files changed, 71 insertions(+)
> >>  
> > Since TAP has no filtering this doesn't really do much for the slowpath.  
> 
> Yes, I know that TAP does not support it. Is failsafe so tighly bound to 
> TAP to block it?
> 

failsafe was built to be used with tap and mlx on Azure.
It could be used with virtio and other NIC on KVM, but this is mostly theoretical.
Ferruh Yigit Sept. 24, 2018, 3:58 p.m. UTC | #7 
On 9/21/2018 5:09 PM, Gaëtan Rivet wrote:
> Hi,
> 
> Seems good, thanks.
> 
> Acked-by: Gaetan Rivet <gaetan.rivet@6wind.com>

Series applied to dpdk-next-net/master, thanks.

> 
> On Fri, Sep 21, 2018 at 04:36:20PM +0100, Andrew Rybchenko wrote:
>> v3:
>>     - move apply on sync to fs_eth_dev_conf_apply() to apply to
>>       a new subdevice only
>>     - use ethdev API to apply to sub-device on sync
>>     - remove unnecessary check the same pointer from the method
>>       implementation in failsafe
>>
>> v2:
>>     - fix setting of zero addresses since rte_realloc() returns NULL
>>
>> Evgeny Im (2):
>>   net/failsafe: remove not supported multicast MAC filter
>>   net/failsafe: support multicast address list set
>>
>>  doc/guides/rel_notes/release_18_11.rst  |  7 ++++
>>  drivers/net/failsafe/failsafe.c         |  1 +
>>  drivers/net/failsafe/failsafe_ether.c   | 17 +++++++++
>>  drivers/net/failsafe/failsafe_ops.c     | 50 +++++++++++++++++++++++++
>>  drivers/net/failsafe/failsafe_private.h |  2 +
>>  5 files changed, 77 insertions(+)
>>
>> -- 
>> 2.17.1
>>
>