[Intel-wired-lan] [PATCH v3 next-queue 00/10] ixgbe: Add ipsec offload

Shannon Nelson shannon.nelson at oracle.com
Thu Dec 21 17:55:51 UTC 2017


On 12/20/2017 11:09 PM, Yanjun Zhu wrote:
> On 2017/12/21 14:39, Yanjun Zhu wrote:
>> On 2017/12/20 7:59, Shannon Nelson wrote:
>>> This is an implementation of the ipsec hardware offload feature for
>>> the ixgbe driver and Intel's 10Gbe series NICs: x540, x550, 82599.
>> Hi, Nelson
>>
>> I notice that the ipsec feature is based on x540, x550, 82599. But 
>> this ixgbe driver
>> will also work with 82598.
>>
>> Does this ipsec feature also work with 82598?
> Sorry. I mean, after these ipsec patches are applied, whether ipsec 
> offload enabled or not,
> can this ixgbe driver still work well with 82598?

Hmm... I don't have one to test on, but I suspect the 82598 might not be 
happy with this.  I'll send a followup patch to catch this case.

Thanks!
sln


> 
> Zhu Yanjun
>>
>> Thanks a lot.
>> Zhu Yanjun
>>> These patches apply to net-next v4.14 as well as Jeff Kirsher's 
>>> next-queue
>>> v4.15-rc1-206-ge47375b.
>>>
>>> The ixgbe NICs support ipsec offload for 1024 Rx and 1024 Tx Security
>>> Associations (SAs), using up to 128 inbound IP addresses, and using the
>>> rfc4106(gcm(aes)) encryption.  This code does not yet support IPv6,
>>> checksum offload, or TSO in conjunction with the ipsec offload - those
>>> will be added in the future.
>>>
>>> This code shows improvements in both packet throughput and CPU 
>>> utilization.
>>> For example, here are some quicky numbers that show the magnitude of the
>>> performance gain on a single run of "iperf -c <dest>" with the ipsec
>>> offload on both ends of a point-to-point connection:
>>>
>>>     9.4 Gbps - normal case
>>>     7.6 Gbps - ipsec with offload
>>>     343 Mbps - ipsec no offload
>>>
>>> To set up a similar test case, you first need to be sure you have a 
>>> recent
>>> version of iproute2 that supports the ipsec offload tag, probably 
>>> something
>>> from ip 4.12 or newer would be best.  I have a shell script that builds
>>> up the appropriate commands for me, but here are the resulting commands
>>> for all tcp traffic between 14.0.0.52 and 14.0.0.70:
>>>
>>> For the left side (14.0.0.52):
>>>    ip x p add dir out src 14.0.0.52/24 dst 14.0.0.70/24 proto tcp tmpl \
>>>       proto esp src 14.0.0.52 dst 14.0.0.70 spi 0x07 mode transport 
>>> reqid 0x07
>>>    ip x p add dir in src 14.0.0.70/24 dst 14.0.0.52/24 proto tcp tmpl \
>>>       proto esp dst 14.0.0.52 src 14.0.0.70 spi 0x07 mode transport 
>>> reqid 0x07
>>>    ip x s add proto esp src 14.0.0.52 dst 14.0.0.70 spi 0x07 mode 
>>> transport \
>>>       reqid 0x07 replay-window 32 \
>>>       aead 'rfc4106(gcm(aes))' 
>>> 0x44434241343332312423222114131211f4f3f2f1 128 \
>>>       sel src 14.0.0.52/24 dst 14.0.0.70/24 proto tcp offload dev 
>>> eth4 dir out
>>>    ip x s add proto esp dst 14.0.0.52 src 14.0.0.70 spi 0x07 mode 
>>> transport \
>>>       reqid 0x07 replay-window 32 \
>>>       aead 'rfc4106(gcm(aes))' 
>>> 0x44434241343332312423222114131211f4f3f2f1 128 \
>>>       sel src 14.0.0.70/24 dst 14.0.0.52/24 proto tcp offload dev 
>>> eth4 dir in
>>>   For the right side (14.0.0.70):
>>>    ip x p add dir out src 14.0.0.70/24 dst 14.0.0.52/24 proto tcp tmpl \
>>>       proto esp src 14.0.0.70 dst 14.0.0.52 spi 0x07 mode transport 
>>> reqid 0x07
>>>    ip x p add dir in src 14.0.0.52/24 dst 14.0.0.70/24 proto tcp tmpl \
>>>       proto esp dst 14.0.0.70 src 14.0.0.52 spi 0x07 mode transport 
>>> reqid 0x07
>>>    ip x s add proto esp src 14.0.0.70 dst 14.0.0.52 spi 0x07 mode 
>>> transport \
>>>       reqid 0x07 replay-window 32 \
>>>       aead 'rfc4106(gcm(aes))' 
>>> 0x44434241343332312423222114131211f4f3f2f1 128 \
>>>       sel src 14.0.0.70/24 dst 14.0.0.52/24 proto tcp offload dev 
>>> eth4 dir out
>>>    ip x s add proto esp dst 14.0.0.70 src 14.0.0.52 spi 0x07 mode 
>>> transport \
>>>       reqid 0x07 replay-window 32 \
>>>       aead 'rfc4106(gcm(aes))' 
>>> 0x44434241343332312423222114131211f4f3f2f1 128 \
>>>       sel src 14.0.0.52/24 dst 14.0.0.70/24 proto tcp offload dev 
>>> eth4 dir in
>>>
>>> In both cases, the command "ip x s flush ; ip x p flush" will clean
>>> it all out and remove the offloads.
>>>
>>> Lastly, thanks to Alex Duyck for his early comments.
>>>
>>> Please see the individual patches for specific update info.
>>>
>>> v3: fixes after comments from those wonderfully pesky kbuild robots
>>> v2: fixes after comments from Alex
>>>
>>> Shannon Nelson (10):
>>>    ixgbe: clean up ipsec defines
>>>    ixgbe: add ipsec register access routines
>>>    ixgbe: add ipsec engine start and stop routines
>>>    ixgbe: add ipsec data structures
>>>    ixgbe: add ipsec offload add and remove SA
>>>    ixgbe: restore offloaded SAs after a reset
>>>    ixgbe: process the Rx ipsec offload
>>>    ixgbe: process the Tx ipsec offload
>>>    ixgbe: ipsec offload stats
>>>    ixgbe: register ipsec offload with the xfrm subsystem
>>>
>>>   drivers/net/ethernet/intel/ixgbe/Makefile        |   1 +
>>>   drivers/net/ethernet/intel/ixgbe/ixgbe.h         |  33 +-
>>>   drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c |   2 +
>>>   drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c   | 923 
>>> +++++++++++++++++++++++
>>>   drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.h   |  92 +++
>>>   drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c     |   4 +-
>>>   drivers/net/ethernet/intel/ixgbe/ixgbe_main.c    |  39 +-
>>>   drivers/net/ethernet/intel/ixgbe/ixgbe_type.h    |  22 +-
>>>   8 files changed, 1093 insertions(+), 23 deletions(-)
>>>   create mode 100644 drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
>>>   create mode 100644 drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.h
>>>
>>
>>
> 


More information about the Intel-wired-lan mailing list