Commit Graph

24 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
liyi
391b501639 net: extract l3 header build code into new functions
Signed-off-by: liyi <liyi25@xiaomi.com>
2022-11-29 18:36:15 +08:00
chao an
a8d3286258 net: move device buffer define to common header
Signed-off-by: chao an <anchao@xiaomi.com>
2022-10-28 00:32:16 -04:00
Petro Karashchenko
3ccb657dc2 nuttx: remove space befone newline in logs
Signed-off-by: Petro Karashchenko <petro.karashchenko@gmail.com>
2021-12-27 21:01:19 -06:00
Xiang Xiao
2e54df0f35 Don't include assert.h from public header file
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
2021-06-03 08:36:03 -07:00
Alin Jerpelea
37d5c1b0d9 net: Author Gregory Nutt: update licenses to Apache
Gregory Nutt has submitted the SGA and we can migrate the licenses
 to Apache.

Signed-off-by: Alin Jerpelea <alin.jerpelea@sony.com>
2021-02-20 00:38:18 -08:00
Alin Jerpelea
7dc7ef45f1 net: nxstyle fixes
Nxstyle fixes to pass CI.

Signed-off-by: Alin Jerpelea <alin.jerpelea@sony.com>
2021-02-20 00:38:18 -08:00
Gregory Nutt
3b275fcf4e net/: Run nxstyle against all C files. 2019-10-25 11:31:42 -06:00
Gregory Nutt
8a3fc26b74 Squashed commit of the following:
The MLD implementation did not follow the RFC correctly when it is the Querier.  The Querier should use a general query and get query messages from all members of all groups.  This would be driven by a single timer per sub-nset since all groups are queried at once. Instead, the design used a Multicast Address Specific Query with one timer per group and ignores groups that we are not members of.

    Similary, the MLDv1 compatibility timer should be a single, separate timer, not a per-group timer.

    net/mld: Group may be NULL when sending a general query
2018-11-11 11:38:29 -06:00
Gregory Nutt
1a56229386 net/mld: Resolve and issue with sending reports from multiple groups in the same polling cycle. 2018-11-11 07:31:59 -06:00
Gregory Nutt
f8d39e2c84 net/MLD: The MAS query should be sent on the group multicast address. 2018-11-10 11:41:50 -06:00
Gregory Nutt
854046a931 /net/devif/ipv6_input.c: Correct handling of IPv6 extension headers. The main confusion was that the payload length in the IPv6 header does not include its extension headers.
net/icmpv6/icmpv6_input.c:  Correct caculation of the ICMPv6 header address
    net/mld/mld_query.c:  Correct back test for group found.
    net/mld/mld_report.c: Fix host vs. network order problem.
    net/mld/mld_send.c: Correct the address used in sending the General Query.  It should be the unspecified address in that case.
    net/mld:  Querying workaround.  The MLD implementation does not follow the RFC correct when it is the Querier.  The Querier should use a general query and get query messages from all members of all groups.  This would be driven by a single timer since all groups are queried at once.  Instead, the design currently uses a Multicast Address Specific Query with one timer per group and ignores groups that we are not members of.
2018-11-10 11:29:02 -06:00
Gregory Nutt
02f83334d5 Squashed commit of the following:
net/mld:  Fix a couple of places where I forgot to unlock the network in the previous commit.

    net/mld:  Implement 'Other Querier Present Timer'.  This timer is used to revert to Querier mode if there is no other querier on the network.  Also, fix some naming:  The Done message is not just Version 1 but is used with Version 2 as well.

    net/igmp:  Back out some blind, backported improvements to IGMP from MLD.  There are too many subtle differences in the protocols for this to be safe.
2018-11-07 10:40:30 -06:00
Gregory Nutt
62e4a41209 net/mld: Add reference counting on the group joins. If there are multiple joins from the same group on this, the don't really leave the group until the matching number of leaves happen. Also add to hooks for the router case: Keep track of the number members NOT on this host. Integrate this with the Leave logic to that the group structure can persist while there no non-local members of the group as well. 2018-11-06 14:46:03 -06:00
Gregory Nutt
f58e7976df net/procfs: Add support to provide MLD stats at /proc/net/mld. net/mld: Fix a couple more bugs found in further MLD testing. 2018-11-06 10:42:53 -06:00
Gregory Nutt
3db3c3952b net/mld: Standard usage of debug output macros. Add configuration options to enable MLD debug even when there network debug is disabled. Add configuration option to dump the content of outgoing MLD packets. 2018-11-05 18:39:55 -06:00
Gregory Nutt
33ed27ae65 Squashed commit of the following:
net/mld: Checksum calculation needs to know the full size of the IPv6 header, including the size of the extension headers.  The payload size in the IPv64 header must include the extension headers (not considered part of the header in this case).  Fixes a few additional errors in size, endian-ness and checksum calculations.  Wireshark now reports the the outgoing Report has a good checksum.

    net/mld/mld_query.c:  Add a cast to assure that the left shift does not overflow.
2018-11-05 08:56:07 -06:00
Gregory Nutt
c6156818b0 net/mld: Fix some compile errors fill debug info is enabled. Fix some errors in formatting the Listerner report. Still has a checksum error. 2018-11-04 18:18:14 -06:00
Gregory Nutt
9a75a14800 net/mld: Implement MLDv1 compatibility mode. This completes coding of MLD. MLD is, howeer, still completely untested. 2018-11-04 14:07:27 -06:00
Gregory Nutt
b1a61834d9 net/mld: Review and update all MLD message receipt logic. Also ripple changs that are, hopefully, improvements back to IGMP. 2018-11-04 10:57:21 -06:00
Gregory Nutt
1c15aa3b2d Squashed commit of the following:
net/mld:  More updates from comparison with roughly leveraged code and the MDL RFCs 2710 and 3810.

    net/mld:  More updated from comparison with roughly leveraged code and the MDL RFCs 2710 and 3810.

    net/mld:  Beginning comparison with roughly leveraged code and the MDL RFCs 2710 and 3810.
2018-11-03 13:45:33 -06:00
Gregory Nutt
6c0ab0e077 Squashed commit of the following:
include/nuttx/net/ipv6ext.h:  Fix some terminology:  Hop-by-hop, not Hop-to-hop or Hop2hop.

    net/devif/ipv6_input.c and net/icmpv6/icmpv6_input.c:  Add logic to skiip over the variable number of IPv6 extension headers that may be present between the IPv6 header and the transport layer header.  The extension headers are simply ignored.  This is necessary because with MLD, certain incoming messages may have, at a mimimum, a Router Alert Hop-by-hop extension header.

    net/inet/ipv6_setsockopt.c:  Implement the IPV6_JOIN_GROUP and IPV6_LEAVE_GROUP socket options.
2018-11-02 12:06:57 -06:00
Gregory Nutt
9f2cb42918 include/nuttx/net/ipv6ext.h: Create header file with all of the definitions of IPv6 extension headers that I could find. net/mld/mld_send.c: Correct the size and format of the Hop-to-hop, Router alert extension header. 2018-11-01 17:51:47 -06:00
Gregory Nutt
5b7ef856a0 Squashed commit of the following:
net/mld:  The MLD logic now compiles and is much less toxic.  It still is not a proper MLD implementation:  (1) It is basically a port of IGMP, tweaked to work with IPv6 and ICMPv6 MLD messages, (2) it needs a proper analysis and comparison with RFC 3810, and (3) it is completely untested.  For this reason, it will remain EXPERIMENTAL for some time.
    net/mld:  Add some missing macros, more fixes related to IPv6 vs IPv4 types,
    net/mld:  More compilation cleaning.  Most fixups for IPv6 vs IPv4 types.
    net/mld:  Hook crudely converted .c files into build system and resolve a few of the many, many compilation/design problems.
    net/mld:  Add support for MLD statistics.
    net/mld:  Hook in MLD poll and packet transmission logic.
    net/mld:  Change references to IPv4 definitions to IPv6 definitions; Remove mld_input() since MLD piggybacks on ICMPv6 input.  Add functions to catch MLD messages dispatched by ICMPv6 input logic.
    net/mld:  As a starting point, copy all net/igmp/*.c files to net/mld/. and change all occurrences of igmp (or IGMP) to mld (or MLD).
    net/mld:  More compilation cleaning.  Most fixups for IPv6 vs IPv4 types.
    net/mld:  Hook crudely converted .c files into build system and resolve a few of the many, many compilation/design problems.
    net/mld:  Add support for MLD statistics.
    net/mld:  Hook in MLD poll and packet transmission logic.
    net/mld:  Change references to IPv4 definitions to IPv6 definitions; Remove mld_input() since MLD piggybacks on ICMPv6 input.  Add functions to catch MLD messages dispatched by ICMPv6 input logic.
    net/mld:  As a starting point, copy all net/igmp/*.c files to net/mld/. and change all occurrences of igmp (or IGMP) to mld (or MLD).
2018-11-01 15:19:47 -06:00