Message-ID: <7fe36ce0-be09-828d-16e5-776add0a182d@oracle.com>
Date:   Tue, 12 Dec 2017 21:45:13 -0800
From:   Shannon Nelson <shannon.nelson@...cle.com>
To:     Alexander Duyck <alexander.duyck@...il.com>
Cc:     intel-wired-lan <intel-wired-lan@...ts.osuosl.org>,
        Jeff Kirsher <jeffrey.t.kirsher@...el.com>,
        Steffen Klassert <steffen.klassert@...unet.com>,
        Netdev <netdev@...r.kernel.org>,
        Sowmini Varadhan <sowmini.varadhan@...cle.com>
Subject: Re: [Intel-wired-lan] [PATCH v2 next-queue 08/10] ixgbe: process the
 Tx ipsec offload

On 12/12/2017 5:59 PM, Alexander Duyck wrote:
> On Tue, Dec 12, 2017 at 3:37 PM, Shannon Nelson
> <shannon.nelson@...cle.com> wrote:
>> If the skb has a security association referenced in the skb, then
>> set up the Tx descriptor with the ipsec offload bits.  While we're
>> here, we fix an oddly named field in the context descriptor struct.
>>
>> Signed-off-by: Shannon Nelson <shannon.nelson@...cle.com>
>> ---
> 
> So I was thinking about it and I have a pretty good idea of the things
> that might be messed up with the combination of ipsec offload and TSO
> or Tx checksum offload.

Before we go much further with this, you should know that I'm finding 
that at least the transport_header and csum_start look like they are off 
by multiples of 8 bytes, which happens to be the length of the ESP 
header.  It looks like the transport_header didn't get incremented when 
the esp was inserted, but the csum_start was incremented twice.  I don't 
believe that ixgbe is changing them, so I'm pretty sure they are given 
to us incorrectly.  I'm not sure yet why the mlx5 didn't run into this 
issue with their ipsec offload, but I don't see them using the 
csum_start and I suspect their use of transport_header is benign. 
Obviously, I need to dig more to get this worked out.

I decided to push out these patches now with ESP_TX_CSUM disabled while 
I try to track down where things are getting messed up.  Once I get that 
worked out, then I can worry about getting the rest of this sorted out. 
For now, these patches seem useful and give a significant boost to 
throughput.

> 
> First in the case of TSO we need to not do the IP checksum update that
> we do for GSO_PARTIAL. We could probably just wrap that code in a
> check that says to do that only if we are performing a GSO_PARTIAL
> type offload, or we just don't do it when performing an IPSEC
> offfload.
> 
> Second we are reporting the wrong IPv4 header length. I am assuming
> that the esp header doesn't occupy skb_checksum_start. As a result we

I'm not sure what you mean by "doesn't occupy"; however, the ESP header 
comes in between the IP and the TCP headers, and the csum_start should 
be moved 8 bytes accordingly.

> are computing the IPv4 header and including the AH/ESP header in it
> and then writing that to vlan_macip_lens. What we may want to do is

Actually, this is correct.  In describing IPLEN the datasheet 
specifically says (7.2.3.2.3) "For IPsec packets, it is the sum of IP 
header length plus IPsec header length."  This is repeated again at the 
end of 7.12.4.2.

> just update the code so that if the TX_FLAGS_IPSEC is set we report an
> IP length of 20, otherwise we use the computed value. As a result we
> cannot do GSO_PARTIAL offload and IPSEC at the same time but I doubt
> that would be a common combination.
> 
>> v2: use ihl != 5
>>      move the ixgbe_ipsec_tx() call to near the call to ixgbe_tso()
>>      drop the ipsec packet if the tx offload setup fails
>>      simplify the ixgbe_ipsec_tx() parameters by using 'first'
>>      leave out the ixgbe_tso() changes since we don't support TSO
>>         with ipsec yet.
>>
>>   drivers/net/ethernet/intel/ixgbe/ixgbe.h       | 10 +++-
>>   drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c | 79 ++++++++++++++++++++++++++
>>   drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c   |  4 +-
>>   drivers/net/ethernet/intel/ixgbe/ixgbe_main.c  | 24 ++++++--
>>   drivers/net/ethernet/intel/ixgbe/ixgbe_type.h  |  2 +-
>>   5 files changed, 110 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
>> index a094b23..3d2b7bf 100644
>> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
>> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
>> @@ -171,10 +171,11 @@ enum ixgbe_tx_flags {
>>          IXGBE_TX_FLAGS_CC       = 0x08,
>>          IXGBE_TX_FLAGS_IPV4     = 0x10,
>>          IXGBE_TX_FLAGS_CSUM     = 0x20,
>> +       IXGBE_TX_FLAGS_IPSEC    = 0x40,
>>
>>          /* software defined flags */
>> -       IXGBE_TX_FLAGS_SW_VLAN  = 0x40,
>> -       IXGBE_TX_FLAGS_FCOE     = 0x80,
>> +       IXGBE_TX_FLAGS_SW_VLAN  = 0x80,
>> +       IXGBE_TX_FLAGS_FCOE     = 0x100,
>>   };
>>
>>   /* VLAN info */
>> @@ -1014,6 +1015,8 @@ void ixgbe_ipsec_restore(struct ixgbe_adapter *adapter);
>>   void ixgbe_ipsec_rx(struct ixgbe_ring *rx_ring,
>>                      union ixgbe_adv_rx_desc *rx_desc,
>>                      struct sk_buff *skb);
>> +int ixgbe_ipsec_tx(struct ixgbe_ring *tx_ring, struct ixgbe_tx_buffer *first,
>> +                  struct ixgbe_ipsec_tx_data *itd);
>>   #else
>>   static inline void ixgbe_init_ipsec_offload(struct ixgbe_adapter *adapter) { };
>>   static inline void ixgbe_stop_ipsec_offload(struct ixgbe_adapter *adapter) { };
>> @@ -1021,5 +1024,8 @@ static inline void ixgbe_ipsec_restore(struct ixgbe_adapter *adapter) { };
>>   static inline void ixgbe_ipsec_rx(struct ixgbe_ring *rx_ring,
>>                                    union ixgbe_adv_rx_desc *rx_desc,
>>                                    struct sk_buff *skb) { };
>> +static inline int ixgbe_ipsec_tx(struct ixgbe_ring *tx_ring,
>> +                                struct ixgbe_tx_buffer *first,
>> +                                struct ixgbe_ipsec_tx_data *itd) { return 0; };
>>   #endif /* CONFIG_XFRM_OFFLOAD */
>>   #endif /* _IXGBE_H_ */
>> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
>> index 7e421b8..5ed8a4f 100644
>> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
>> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
>> @@ -695,12 +695,91 @@ static void ixgbe_ipsec_del_sa(struct xfrm_state *xs)
>>          }
>>   }
>>
>> +/**
>> + * ixgbe_ipsec_offload_ok - can this packet use the xfrm hw offload
>> + * @skb: current data packet
>> + * @xs: pointer to transformer state struct
>> + **/
>> +static bool ixgbe_ipsec_offload_ok(struct sk_buff *skb, struct xfrm_state *xs)
>> +{
>> +       if (xs->props.family == AF_INET) {
>> +               /* Offload with IPv4 options is not supported yet */
>> +               if (ip_hdr(skb)->ihl != 5)
>> +                       return false;
>> +       } else {
>> +               /* Offload with IPv6 extension headers is not support yet */
>> +               if (ipv6_ext_hdr(ipv6_hdr(skb)->nexthdr))
>> +                       return false;
>> +       }
>> +
>> +       return true;
>> +}
>> +
>>   static const struct xfrmdev_ops ixgbe_xfrmdev_ops = {
>>          .xdo_dev_state_add = ixgbe_ipsec_add_sa,
>>          .xdo_dev_state_delete = ixgbe_ipsec_del_sa,
>> +       .xdo_dev_offload_ok = ixgbe_ipsec_offload_ok,
>>   };
>>
>>   /**
>> + * ixgbe_ipsec_tx - setup Tx flags for ipsec offload
>> + * @tx_ring: outgoing context
>> + * @first: current data packet
>> + * @itd: ipsec Tx data for later use in building context descriptor
>> + **/
>> +int ixgbe_ipsec_tx(struct ixgbe_ring *tx_ring,
>> +                  struct ixgbe_tx_buffer *first,
>> +                  struct ixgbe_ipsec_tx_data *itd)
>> +{
>> +       struct ixgbe_adapter *adapter = netdev_priv(tx_ring->netdev);
>> +       struct ixgbe_ipsec *ipsec = adapter->ipsec;
>> +       struct xfrm_state *xs;
>> +       struct tx_sa *tsa;
>> +
>> +       if (!first->skb->sp->len) {
>> +               netdev_err(tx_ring->netdev, "%s: no xfrm state len = %d\n",
>> +                          __func__, first->skb->sp->len);
>> +               return 0;
>> +       }
>> +
>> +       xs = xfrm_input_state(first->skb);
>> +       if (!xs) {
>> +               netdev_err(tx_ring->netdev, "%s: no xfrm_input_state() xs = %p\n",
>> +                          __func__, xs);
>> +               return 0;
>> +       }
>> +
>> +       itd->sa_idx = xs->xso.offload_handle - IXGBE_IPSEC_BASE_TX_INDEX;
>> +       if (itd->sa_idx > IXGBE_IPSEC_MAX_SA_COUNT) {
>> +               netdev_err(tx_ring->netdev, "%s: bad sa_idx=%d handle=%lu\n",
>> +                          __func__, itd->sa_idx, xs->xso.offload_handle);
>> +               return 0;
>> +       }
>> +
>> +       tsa = &ipsec->tx_tbl[itd->sa_idx];
>> +       if (!tsa->used) {
>> +               netdev_err(tx_ring->netdev, "%s: unused sa_idx=%d\n",
>> +                          __func__, itd->sa_idx);
>> +               return 0;
>> +       }
>> +
>> +       first->tx_flags |= IXGBE_TX_FLAGS_IPSEC | IXGBE_TX_FLAGS_CC;
>> +
>> +       itd->flags = 0;
>> +       if (xs->id.proto == IPPROTO_ESP) {
>> +               itd->flags |= IXGBE_ADVTXD_TUCMD_IPSEC_TYPE_ESP |
>> +                             IXGBE_ADVTXD_TUCMD_L4T_TCP;
>> +               if (first->protocol == htons(ETH_P_IP))
>> +                       itd->flags |= IXGBE_ADVTXD_TUCMD_IPV4;
>> +               itd->trailer_len = xs->props.trailer_len;
>> +       }
> 
> I still have some concerns about these bits. First I would say you
> could move the TUCMD_IPV4 portion into the end of the tx_csum logic
> just after the no_csum label. No point in carrying this separately if
> we can just take care of this there. If it is safe here I would assume
> it would be safe to perform generically.

Sure, the TUCMD_IPV4 portion could get moved, but I'd like to keep the 
bit flipping logic for ipsec on one place, and simple add in the 
collection in the calling function.

> 
> Also you are still setting the TCP flag for these packets. I'm
> wondering if the TCP flag is really needed. I get that your current
> setup only runs on TCP flows but what about other types of flows?
> Couldn't UDP be encapsulated via ESP?

That L4T_TCP bit annoys me - if I don't set it, the offload doesn't 
work, whether doing csum offload or not.

The UDP messages can be encapsulated, but they don't seem to go down the 
offload route.  I haven't looked into why yet.

> 
>> +       if (tsa->encrypt)
>> +               itd->flags |= IXGBE_ADVTXD_TUCMD_IPSEC_ENCRYPT_EN;
>> +
>> +       return 1;
>> +}
>> +
>> +/**
>>    * ixgbe_ipsec_rx - decode ipsec bits from Rx descriptor
>>    * @rx_ring: receiving ring
>>    * @rx_desc: receive data descriptor
>> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c
>> index f1bfae0..d7875b3 100644
>> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c
>> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c
>> @@ -1261,7 +1261,7 @@ void ixgbe_clear_interrupt_scheme(struct ixgbe_adapter *adapter)
>>   }
>>
>>   void ixgbe_tx_ctxtdesc(struct ixgbe_ring *tx_ring, u32 vlan_macip_lens,
>> -                      u32 fcoe_sof_eof, u32 type_tucmd, u32 mss_l4len_idx)
>> +                      u32 fceof_saidx, u32 type_tucmd, u32 mss_l4len_idx)
>>   {
>>          struct ixgbe_adv_tx_context_desc *context_desc;
>>          u16 i = tx_ring->next_to_use;
>> @@ -1275,7 +1275,7 @@ void ixgbe_tx_ctxtdesc(struct ixgbe_ring *tx_ring, u32 vlan_macip_lens,
>>          type_tucmd |= IXGBE_TXD_CMD_DEXT | IXGBE_ADVTXD_DTYP_CTXT;
>>
>>          context_desc->vlan_macip_lens   = cpu_to_le32(vlan_macip_lens);
>> -       context_desc->seqnum_seed       = cpu_to_le32(fcoe_sof_eof);
>> +       context_desc->fceof_saidx       = cpu_to_le32(fceof_saidx);
>>          context_desc->type_tucmd_mlhl   = cpu_to_le32(type_tucmd);
>>          context_desc->mss_l4len_idx     = cpu_to_le32(mss_l4len_idx);
>>   }
>> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
>> index 0ee1e5e..84fbfb9 100644
>> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
>> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
>> @@ -7756,10 +7756,12 @@ static inline bool ixgbe_ipv6_csum_is_sctp(struct sk_buff *skb)
>>   }
>>
>>   static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
>> -                         struct ixgbe_tx_buffer *first)
>> +                         struct ixgbe_tx_buffer *first,
>> +                         struct ixgbe_ipsec_tx_data *itd)
>>   {
>>          struct sk_buff *skb = first->skb;
>>          u32 vlan_macip_lens = 0;
>> +       u32 fceof_saidx = 0;
>>          u32 type_tucmd = 0;
>>
>>          if (skb->ip_summed != CHECKSUM_PARTIAL) {
>> @@ -7800,7 +7802,12 @@ static void ixgbe_tx_csum(struct ixgbe_ring *tx_ring,
>>          vlan_macip_lens |= skb_network_offset(skb) << IXGBE_ADVTXD_MACLEN_SHIFT;
>>          vlan_macip_lens |= first->tx_flags & IXGBE_TX_FLAGS_VLAN_MASK;
>>
>> -       ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, 0, type_tucmd, 0);
>> +       if (first->tx_flags & IXGBE_TX_FLAGS_IPSEC) {
>> +               fceof_saidx |= itd->sa_idx;
>> +               type_tucmd |= itd->flags | itd->trailer_len;
> 
> These would all be 0 if the IPSEC flag isn't set, so I would say just
> write them in all cases and avoid the conditional jump.

Sure.

sln

> 
>> +       }
>> +
>> +       ixgbe_tx_ctxtdesc(tx_ring, vlan_macip_lens, fceof_saidx, type_tucmd, 0);
>>   }
>>
>>   #define IXGBE_SET_FLAG(_input, _flag, _result) \
>> @@ -7843,11 +7850,16 @@ static void ixgbe_tx_olinfo_status(union ixgbe_adv_tx_desc *tx_desc,
>>                                          IXGBE_TX_FLAGS_CSUM,
>>                                          IXGBE_ADVTXD_POPTS_TXSM);
>>
>> -       /* enble IPv4 checksum for TSO */
>> +       /* enable IPv4 checksum for TSO */
>>          olinfo_status |= IXGBE_SET_FLAG(tx_flags,
>>                                          IXGBE_TX_FLAGS_IPV4,
>>                                          IXGBE_ADVTXD_POPTS_IXSM);
>>
>> +       /* enable IPsec */
>> +       olinfo_status |= IXGBE_SET_FLAG(tx_flags,
>> +                                       IXGBE_TX_FLAGS_IPSEC,
>> +                                       IXGBE_ADVTXD_POPTS_IPSEC);
>> +
>>          /*
>>           * Check Context must be set if Tx switch is enabled, which it
>>           * always is for case where virtual functions are running
>> @@ -8306,6 +8318,7 @@ netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
>>          u32 tx_flags = 0;
>>          unsigned short f;
>>          u16 count = TXD_USE_COUNT(skb_headlen(skb));
>> +       struct ixgbe_ipsec_tx_data ipsec_tx = { 0 };
>>          __be16 protocol = skb->protocol;
>>          u8 hdr_len = 0;
>>
>> @@ -8410,11 +8423,14 @@ netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
>>          }
>>
>>   #endif /* IXGBE_FCOE */
>> +
>> +       if (skb->sp && !ixgbe_ipsec_tx(tx_ring, first, &ipsec_tx))
>> +               goto out_drop;
>>          tso = ixgbe_tso(tx_ring, first, &hdr_len);
>>          if (tso < 0)
>>                  goto out_drop;
>>          else if (!tso)
>> -               ixgbe_tx_csum(tx_ring, first);
>> +               ixgbe_tx_csum(tx_ring, first, &ipsec_tx);
>>
>>          /* add the ATR filter if ATR is on */
>>          if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
>> diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h b/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
>> index 3df0763..0ac725fa 100644
>> --- a/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
>> +++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
>> @@ -2856,7 +2856,7 @@ union ixgbe_adv_rx_desc {
>>   /* Context descriptors */
>>   struct ixgbe_adv_tx_context_desc {
>>          __le32 vlan_macip_lens;
>> -       __le32 seqnum_seed;
>> +       __le32 fceof_saidx;
>>          __le32 type_tucmd_mlhl;
>>          __le32 mss_l4len_idx;
>>   };
>> --
>> 2.7.4
>>
>> _______________________________________________
>> Intel-wired-lan mailing list
>> Intel-wired-lan@...osl.org
>> https://lists.osuosl.org/mailman/listinfo/intel-wired-lan

Powered by blists - more mailing lists