sergiotarxz/nuttx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
47,341 Commits 1 Branch 0 Tags 291 MiB
8fc1e524ab
Commit Graph

13 Commits

Author SHA1 Message Date
chao an
34d2cde8a8 net/l2/l3/l4: add support of iob offload 
1. Add new config CONFIG_NET_LL_GUARDSIZE to isolation of l2 stack,
   which will benefit l3(IP) layer for multi-MAC(l2) implementation,
   especially in some NICs such as celluler net driver.

new configuration options: CONFIG_NET_LL_GUARDSIZE

CONFIG_NET_LL_GUARDSIZE will reserved l2 buffer header size of
network buffer to isolate the L2/L3 (MAC/IP) data on network layer,
which will be beneficial to L3 network layer protocol transparent
transmission and forwarding

------------------------------------------------------------
Layout of frist iob entry:

        iob_data (aligned by CONFIG_IOB_ALIGNMENT)
            |
            |                  io_offset(CONFIG_NET_LL_GUARDSIZE)
            |                                |
            -------------------------------------------------
      iob   |            Reserved            |    io_len    |
            -------------------------------------------------

-------------------------------------------------------------
Layout of different NICs implementation:

        iob_data (aligned by CONFIG_IOB_ALIGNMENT)
            |
            |                 io_offset(CONFIG_NET_LL_GUARDSIZE)
            |                                |
            -------------------------------------------------
 Ethernet   |       Reserved    | ETH_HDRLEN |    io_len    |
            ---------------------------------|---------------
 8021Q      |   Reserved  | ETH_8021Q_HDRLEN |    io_len    |
            ---------------------------------|---------------
 ipforward  |            Reserved            |    io_len    |
            -------------------------------------------------

--------------------------------------------------------------------

2. Support iob offload to l2 driver to avoid unnecessary memory copy

Support send/receive iob vectors directly between the NICs and l3/l4
stack to avoid unnecessary memory copies, especially on hardware that
supports Scatter/gather, which can greatly improve performance.

new interface to support iob offload:

  ------------------------------------------
  |    IOB version     |     original      |
  |----------------------------------------|
  |  devif_iob_poll()  |   devif_poll()    |
  |       ...          |       ...         |
  ------------------------------------------

--------------------------------------------------------------------

1> NIC hardware support Scatter/gather transfer

TX:

                tcp_poll()/udp_poll()/pkt_poll()/...(l3|l4)
                           /              \
                          /                \
devif_poll_[l3|l4]_connections()     devif_iob_send() (nocopy:udp/icmp/...)
           /                                   \      (copy:tcp)
          /                                     \
  devif_iob_poll("NIC"_txpoll)                callback() // "NIC"_txpoll
                                                  |
                            dev->d_iob:           |
                                                ---------------         ---------------
                             io_data       iob1 |  |          |    iob3 |  |          |
                                    \           ---------------         ---------------
                                  ---------------  |       --------------- |
                             iob0 |  |          |  |  iob2 |  |          | |
                                  ---------------  |       --------------- |
                                     \             |          /           /
                                        \          |       /           /
                                   ----------------------------------------------
                    NICs io vector |    |    |    |    |    |    |    |    |    |
                                   ----------------------------------------------

RX:

  [tcp|udp|icmp|...]ipv[4|6]_data_handler()(iob_concat/append to readahead)
                    |
                    |
      [tcp|udp|icmp|...]_ipv[4|6]_in()/...
                    |
                    |
          pkt/ipv[4/6]_input()/...
                    |
                    |
     NICs io vector receive(iov_base to each iobs)

--------------------------------------------------------------------

2> CONFIG_IOB_BUFSIZE is greater than MTU:

TX:

"(CONFIG_IOB_BUFSIZE) > (MAX_NETDEV_PKTSIZE + CONFIG_NET_GUARDSIZE + CONFIG_NET_LL_GUARDSIZE)"

                tcp_poll()/udp_poll()/pkt_poll()/...(l3|l4)
                           /              \
                          /                \
devif_poll_[l3|l4]_connections()     devif_iob_send() (nocopy:udp/icmp/...)
           /                                   \      (copy:tcp)
          /                                     \
  devif_iob_poll("NIC"_txpoll)                callback() // "NIC"_txpoll
                                                  |
                                             "NIC"_send()
                          (dev->d_iob->io_data[CONFIG_NET_LL_GUARDSIZE - NET_LL_HDRLEN(dev)])

RX:

  [tcp|udp|icmp|...]ipv[4|6]_data_handler()(iob_concat/append to readahead)
                    |
                    |
      [tcp|udp|icmp|...]_ipv[4|6]_in()/...
                    |
                    |
          pkt/ipv[4/6]_input()/...
                    |
                    |
     NICs io vector receive(iov_base to io_data)

--------------------------------------------------------------------

3> Compatible with all old flat buffer NICs

TX:
                tcp_poll()/udp_poll()/pkt_poll()/...(l3|l4)
                           /              \
                          /                \
devif_poll_[l3|l4]_connections()     devif_iob_send() (nocopy:udp/icmp/...)
           /                                   \      (copy:tcp)
          /                                     \
  devif_iob_poll(devif_poll_callback())  devif_poll_callback() /* new interface, gather iobs to flat buffer */
       /                                           \
      /                                             \
 devif_poll("NIC"_txpoll)                     "NIC"_send()(dev->d_buf)

RX:

  [tcp|udp|icmp|...]ipv[4|6]_data_handler()(iob_concat/append to readahead)
                    |
                    |
      [tcp|udp|icmp|...]_ipv[4|6]_in()/...
                    |
                    |
               netdev_input()  /* new interface, Scatter/gather flat/iob buffer */
                    |
                    |
          pkt/ipv[4|6]_input()/...
                    |
                    |
    NICs io vector receive(Orignal flat buffer)

3. Iperf passthrough on NuttX simulator:

  -------------------------------------------------
  |  Protocol      | Server | Client |            |
  |-----------------------------------------------|
  |  TCP           |  813   |   834  |  Mbits/sec |
  |  TCP(Offload)  | 1720   |  1100  |  Mbits/sec |
  |  UDP           |   22   |   757  |  Mbits/sec |
  |  UDP(Offload)  |   25   |  1250  |  Mbits/sec |
  -------------------------------------------------

Signed-off-by: chao an <anchao@xiaomi.com>
 2022-12-03 11:47:04 +08:00
Gregory Nutt
22cd0d47fa This commit attempts remove some long standard confusion in naming and some actual problems that result from the naming confusion. The basic problem is the standard MTU does not include the size of the Ethernet header. For clarity, I changed the naming of most things called MTU to PKTSIZE. For example, CONFIG_NET_ETH_MTU is now CONFIG_NET_ETH_PKTSIZE. 
This makes the user interface a little hostile.  People thing of an MTU of 1500 bytes, but the corresponding packet is really 1514 bytes (including the 14 byte Ethernet header).  A more friendly solution would configure the MTU (as before), but then derive the packet buffer size by adding the MAC header length.  Instead, we define the packet buffer size then derive the MTU.

The MTU is not common currency in networking.  On the wire, the only real issue is the MSS which is derived from MTU by subtracting the IP header and TCP header sizes (for the case of TCP).  Now it is derived for the PKTSIZE by subtracting the IP header, the TCP header, and the MAC header sizes.  So we should be all good and without the recurring 14 byte error in MTU's and MSS's.

Squashed commit of the following:

    Trivial update to fix some spacing issues.
    net/: Rename several macros containing _MTU to _PKTSIZE.
    net/: Rename CONFIG_NET_SLIP_MTU to CONFIG_NET_SLIP_PKTSIZE and similarly for CONFIG_NET_TUN_MTU.  These are not the MTU which does not include the size of the link layer header.  These are the full size of the packet buffer memory (minus any GUARD bytes).
    net/: Rename CONFIG_NET_6LOWPAN_MTU to CONFIG_NET_6LOWPAN_PKTSIZE and similarly for CONFIG_NET_TUN_MTU.  These are not the MTU which does not include the size of the link layer header.  These are the full size of the packet buffer memory (minus any GUARD bytes).
    net/: Rename CONFIG_NET_ETH_MTU to CONFIG_NET_ETH_PKTSIZE.  This is not the MTU which does not include the size of the link layer header.  This is the full size of the packet buffer memory (minus any GUARD bytes).
    net/: Rename the file d_mtu in the network driver structure to d_pktsize.  That value saved there is not the MTU.  The packetsize is the memory large enough to hold the maximum packet PLUS the size of the link layer header.  The MTU does not include the link layer header.
 2018-07-04 14:10:40 -06:00
Gregory Nutt
88a87f8e3f 6LoWPAN: The original, Contiki-based design used only a single buffer for reassemblying larger packets. This could be a problem issue for hub configurations which really need the capability concurrently reassemble multiple incoming streams concurrently. These was also a design issue in that the reassembly buffer could be corrupted by outgoing packets. The design was extended to support multiple reassembly buffers, each associated with the reassembly tag and source address. This assures that there can be be no corruption of the reassembly once it has started. 2017-08-26 10:00:47 -06:00
Gregory Nutt
31f832d8c5 IP forwarding: Flesh out TCP, UDP, and ICMPv6 packet forwarding logic. 2017-07-05 11:01:16 -06:00
Gregory Nutt
0fb035f76b Standardize some naming in code section comments 2016-02-21 18:09:04 -06:00
Gregory Nutt
cb9e27c3b0 Standardize naming used for public data and function groupings 2015-10-02 16:30:35 -06:00
Gregory Nutt
aec4528882 Cosmetic standardization of some naming in comments 2015-01-19 16:02:56 -06:00
Gregory Nutt
88a0e82934 Networking: Remove field d_sndata from the device structure. It is the same as d_appdata and unnecessary 2015-01-17 05:59:13 -06:00
Gregory Nutt
2340d46d20 Rename NET_LL_MTU to NET_DEV_MTU; rename d_llmtu to d_mtu 2014-11-16 10:42:19 -06:00
Gregory Nutt
859748a94e Completes conversion of CONFIG_NET_BUFIZE to CONFIG_NET_ETH/SLIP_MTU 2014-11-16 09:22:38 -06:00
Gregory Nutt
0bb153b8cb Remove all inclusion of uip.h 2014-07-04 16:58:22 -06:00
Gregory Nutt
8e706eb4ff Rename many functions in net/devif from uip_* to devif_* 2014-06-28 18:36:09 -06:00
Gregory Nutt
5790c94ba3 Rename net/uip to net/devif. Rename uip/uip.h to devif/devif.h 2014-06-28 18:07:02 -06:00