If we only enable one of `CONFIG_NET_ARP_SEND` and
`CONFIG_NET_ICMPv6_NEIGHBOR`, both IPv4 and IPv6 traffic will send
ARP or NDP, which causes problem.
Example:
`CONFIG_NET_ARP_SEND=n`
`CONFIG_NET_ICMPv6_NEIGHBOR=y`
Wrong:
IPv4 traffic (`PF_INET`) goes into `icmpv6_neighbor`, which
definitely causes problem.
Correct:
IPv4 traffic doesn't call anything, IPv6 traffic calls `icmpv6_neighbor`
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
When CONFIG_NET_TCP_WRITE_BUFFERS is enabled, iobs are used for
both queuing data from application, and for assembling packets
for sending. If there is a system-wide shortage of iobs, it could
happen that there is not enough free space to form any packets
to send. The buffers allocated for TCP data also can't be released
until the packet is sent.
Normally this should be avoided by setting suitable values for
CONFIG_IOB_NBUFFERS and CONFIG_IOB_THROTTLE. The default values
are ok for light usage, but can run out when using multiple
simultaneous TCP streams.
Before this commit, iob shortage would cause TCP connections to
get stuck and eventually timeout. With this change, TCP stack
sends smaller packets, eventually freeing some buffers from the
write queue.
Previously ipv6 multi-address support decided packet source
address based on its destination. This doesn't work if NuttX
device has multiple addresses within same subnet.
Instead when a packet is a response to existing connection,
the source address should be based on the destination address
used in the received packet.
Note that user-space related code, like procfs and lifreq related ioctl commands, are not touched in this commit.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
Some use cases, such as VPN, use both the device's network
stack with the Usrsock daemon and the Kernel's network stack.
Therefore, remove NET_TCP_NO_STACK/NET_UDP_NO_STACK select
from Usrsock's Kconfig.
The previous iob_trimhead added dev->iob->io_offset, so if the
input frame is not merged into the ofo segment, we need to reset
dev->iob->io_offset so that the subsequent tcp_send can properly
assemble packets.
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
We may just free some TCP connections before monitor stopped, e.g.
sacrificie a TCP conn in `tcp_alloc` will just call `tcp_free` and reuse
the connection. But we noticed that the TCP monitor is not released in
`tcp_free` because it is mounted on `conn->connevents` instead of
`conn->sconn.list` while `tcp_free` only release the latter.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
with nonblocking sockets, POLLOUT is returned when the application
calls poll even if the send buffer is full
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
Normally, `SO_ERROR` for disconn events will be set by `tcp_poll_eventhandler`, but when the socket is closed before poll, we should also set the `SO_ERROR`.
On Linux, `tcp_poll` returns `EPOLLERR` event when `sk->sk_err` has value but doesn't let `poll` fail (doesn't set `errno`). https://github.com/torvalds/linux/blob/v6.5/net/ipv4/tcp.c#L594-L596
Note: `sk->sk_err` can be get by socket option `SO_ERROR` on Linux, so `POLLERR` will always be together with `SO_ERROR`.
Common libs like curl may try to read `SO_ERROR` on `POLLERR`.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
After the sack is enabled and the ofosegs has gap, tcp cannot update the
tx_unacked, so the peer received packets are retransmitted after the
timer timeout.
So update tx_unacked first.
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
Non-blocking sockets are almost never in the connected state when app try to
get result of connect success through poll, to avoid confusion because of
this error print, so downgrade the print level to warning.
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
Using the macro places the buffers into .data section which means they
will consume the full buffer size of flash / read only memory as well.
Place the buffers into .bss to avoid this case.
When do poll operation and the tcp connection state is TCP_ALLOCATED, eventset(POLLERR|POLLHUP) is return, causing the libuv poll_multiple_handles to fail.
Verification: Use the libuv test case ` uv_run_tests poll_multiple_handles`.
Signed-off-by: liqinhui <liqinhui@xiaomi.com>
The error log is as follows:
tcp/tcp_send_buffered.c:376:57: runtime error: member access within misaligned
address 0x10075942 for type 'struct tcp_sack_s', which requires 4 byte alignment
Signed-off-by: liqinhui <liqinhui@xiaomi.com>
The netdev of link-local address cannot be auto decided, and the link-local address should always be reguarded as address on local network.
The problem we met:
When using `icmpv6_autoconfig` with multiple netdev, the `icmpv6_neighbor` may take out wrong netdev with ip address already set, then it may send solicitation with wrong address (`dev->d_ipv6draddr`) on wrong device, and regard the link-local address as conflict (because `dev->d_ipv6draddr` exists on this network).
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
When using IOB queue to store readahead data, we use one IOB for each
UDP packet. Then if the packets are very small, like 10Bytes per packet,
we'll use ~1600 IOBs just for 16KB recv buffer size, which is wasteful
and dangerous. So change conn->readahead to a single IOB chain like TCP.
Benefits:
- Using memory and IOBs more efficiently (small packets are common in
UDP)
Side effects:
- UDP recv buffer size may count the overhead
- A little bit drop in performance (<1%, more seek & copy)
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
1. Update all CMakeLists.txt to adapt to new layout
2. Fix cmake build break
3. Update all new file license
4. Fully compatible with current compilation environment(use configure.sh or cmake as you choose)
------------------
How to test
From within nuttx/. Configure:
cmake -B build -DBOARD_CONFIG=sim/nsh -GNinja
cmake -B build -DBOARD_CONFIG=sim:nsh -GNinja
cmake -B build -DBOARD_CONFIG=sabre-6quad/smp -GNinja
cmake -B build -DBOARD_CONFIG=lm3s6965-ek/qemu-flat -GNinja
(or full path in custom board) :
cmake -B build -DBOARD_CONFIG=$PWD/boards/sim/sim/sim/configs/nsh -GNinja
This uses ninja generator (install with sudo apt install ninja-build). To build:
$ cmake --build build
menuconfig:
$ cmake --build build -t menuconfig
--------------------------
2. cmake/build: reformat the cmake style by cmake-format
https://github.com/cheshirekow/cmake_format
$ pip install cmakelang
$ for i in `find -name CMakeLists.txt`;do cmake-format $i -o $i;done
$ for i in `find -name *\.cmake`;do cmake-format $i -o $i;done
Co-authored-by: Matias N <matias@protobits.dev>
Signed-off-by: chao an <anchao@xiaomi.com>
When a task needs to send data, a callback is allocated and the
transmission is happening in a worker task through devif_send.
Synchronization between the two tasks (sender & worker) is
achieved by a semaphore.
If devif_send fails, this semaphore was never posted, leaving
the sending task blocked indefinitely. This commit fixes this
by checking the return code of netif_send, and posting this
semaphore in case of failure.
Polling then stops, and execution is resumed on the sending
task.
Problem:
When tcp_close_work starts to run in LPWORK, if another event comes and
calls tcp_free before tcp_close_work takes net_lock, the tcp_free will
be called twice and cause double free.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
When do socket bind, if the connection domain is not equal to the bound address type, this will cause the stack-buffer-overflow.
Signed-off-by: liqinhui <liqinhui@xiaomi.com>
- NewReno congestion control algorithm is used to solve the problem
of network congestion breakdown. NewReno congestion control includes
slow start, collision avoidance, fast retransmission, and fast
recovery. The implementation refers to RFC6582 and RFC5681.
- In addition, we optimize the congestion algorithm. In the conflict
avoidance stage, the maximum congestion window max_cwnd is used to
limit the excessive growth of cwnd and prevent network jitter
caused by congestion. Maximum congestion window max_cwnd is updated
with the current congestion window cwnd and the update weight is
0.875 when an RTO timeout occurs.
Signed-off-by: liqinhui <liqinhui@xiaomi.com>
If TCP_FIN is received before the user calls accept, the later accept will
report an error, then the previously received data cannot be read.
operation flow:
tcp_server_listen -> tcp_client_connect -> tcp_client_send ->
tcp_client_close -> tcp_server_accept(fix this error) -> tcp_server_recv
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
We have a case that an http server gives out-of-ordered ACKs, and NuttX client makes `ofoseg`s with length 0, trying to rebuild / put them into `ofosegs` array, which is not intended (no available data and should be skipped). This breaks later logic and finally crashed in `tcp_ofoseg_bufsize` (`ofosegs[i].data` is `NULL`, which should never happen in normal logic).
Note:
- `iob_trimhead` won't return `NULL` when it's applying on normal IOB.
- Keep `dev->d_iob == NULL` to avoid `iob_trimhead` changed.
- `iob_free_chain` will do nothing when applied to `NULL`.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
According to RFC 2525, Section 2.17:
"When an application closes a connection in such a way that it can no longer read any received data, the TCP SHOULD, per section 4.2.2.13 of RFC 1122, send a RST if there is any unread received data, or if any new data is received."
When our TCP socket is closed (even when the thread has exited), the peer can keep sending data and NuttX keeps replying ACK (we've tried for ~12h). This is not a good behavior (also different from Linux), so send RST instead of ACK for data receiving in FIN_WAIT.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
zhanghongyu