According to RFC 5681 (3.2) the TCP Fast Retransmit algorithm should start
if the threshold of 3 duplicate ACKs is reached.
Thus the threshold should be a constant, not an integer option.
While it's a neat idea, it doesn't work well in reality.
* Many of modern tcp stacks don't obey the "ack every other packet"
rule these days. (Linux, macOS, ...)
* Even if a traditional TCP implementation is assumed, we can't
predict/control which packets are acked reliably. For example,
window updates can easily mess up our strategy.
RFC2001: TCP Slow Start, Congestion Avoidance, Fast Retransmit,
and Fast Recovery Algorithms
...
3. Fast Retransmit
Modifications to the congestion avoidance algorithm were proposed in
1990 [3]. Before describing the change, realize that TCP may
generate an immediate acknowledgment (a duplicate ACK) when an out-
of-order segment is received (Section 4.2.2.21 of [1], with a note
that one reason for doing so was for the experimental fast-
retransmit algorithm). This duplicate ACK should not be delayed.
The purpose of this duplicate ACK is to let the other end know that a
segment was received out of order, and to tell it what sequence
number is expected.
Since TCP does not know whether a duplicate ACK is caused by a lost
segment or just a reordering of segments, it waits for a small number
of duplicate ACKs to be received. It is assumed that if there is
just a reordering of the segments, there will be only one or two
duplicate ACKs before the reordered segment is processed, which will
then generate a new ACK. If three or more duplicate ACKs are
received in a row, it is a strong indication that a segment has been
lost. TCP then performs a retransmission of what appears to be the
missing segment, without waiting for a retransmission timer to
expire.
Change-Id: Ie2cbcecab507c3d831f74390a6a85e0c5c8e0652
Signed-off-by: chao.an <anchao@xiaomi.com>
Here is the email loop talk about why it is better to remove the option:
https://groups.google.com/forum/#!topic/nuttx/AaNkS7oU6R0
Change-Id: Ib66c037752149ad4b2787ef447f966c77aa12aad
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Author: Gregory Nutt <gnutt@nuttx.org>
Run all .h and .c files modified in last PR through nxstyle.
Author: Xiang Xiao <xiaoxiang@xiaomi.com>
Net cleanup (#17)
* Fix the semaphore usage issue found in tcp/udp
1. The count semaphore need disable priority inheritance
2. Loop again if net_lockedwait return -EINTR
3. Call nxsem_trywait to avoid the race condition
4. Call nxsem_post instead of sem_post
* Put the work notifier into free list to avoid the heap fragment in the long run. Since the allocation strategy is encapsulated internally, we can even refine the implementation later.
* Network stack shouldn't allocate memory in the poll implementation to avoid the heap fragment in the long run, other modification include:
1. Select MM_IOB automatically since ICMP[v6] socket can't work without the read ahead buffer
2. Remove the net lock since xxx_callback_free already do the same thing
3. TCP/UDP poll should work even the read ahead buffer isn't enabled at all
* Add NET_ prefix for UDP_NOTIFIER and TCP_NOTIFIER option to align with other UDP/TCP option convention
* Remove the unused _SF_[IDLE|ACCEPT|SEND|RECV|MASK] flags since there are code to set/clear these flags, but nobody check them.
Squashed commit of the following:
net/tmp: Rename the unacked field of the tcp connection structure to tx_unacked. Too confusing with the implementation of delayed RX ACKs.
net/tcp: Initial implementation of TCP delayed ACKs.
net/tcp: Add delayed ACK configuration selection. Rename tcp_ack() to tcp_synack(). It may or may not send a ACK. It will always send SYN or SYN/ACK.
Squashed commit of the following:
Fix up some final compile isses.
net/netdev: Convert the network down notification logic to use the new wqueue-based notification factility.
net/udp: Convert the UDP readahead notification logic to use the new wqueue-based notification factility.
net/tcp: Convert the TCP readahead notification logic to use the new wqueue-based notification factility.
mm/iob: Convert the IOB notification logic to use the new wqueue-based notification factility.
sched/wqueue: Signals are not good IPCs to support the target poll functionality for several reasons including the amount of data that can be passed with a signal and in the fact that in protected and kernel modes, user threads executing signal handlers in protected, kernel memory is problematic. Instead, convert the same logic to perform the notifications via function callback on the high priority work queue.
Squashed commit of the following:
net/tcp: Add signal notification for the case when UDP read-ahead data is buffered. This is basically of clone of the TCP notification logic with naming adapted for UDP.
net/tcp: Add signal notification for the case when TCP read-ahead data is buffered.
Fix a few typo/compilation problems.
net/: Remove all CONFIG_NET_xxx_TCP_RECVWNDO configuration variables. They were used only to initialize the d_recwndo of the network device structure which no longer exists.
net/: Remove the device TCP receive window field (d_recvwndo) from the device structure. That value is no longer retained, but is calculated dynamically.
Remove some dangling references to CONFIG_NET_TCP_RWND_CONTROL.
net/tcp: Take read-ahead throttling into account when calculating the TCP receive window size.
net/tcp: tcp_get_recvwindow() now returns the receive window size directly (vs. indirectly via the device structure).
net/tcp: Remove CONFIG_NET_TCP_RWND_CONTROL. TCP window algorithm is now trigged only by CONFIG_NET_TCP_READAHEAD.
net/tcp: Add logic to send probes when SO_KEEPALIVE is enabled.
net/tcp: TCP socket should not have to be connected to configure KeepAlive.
net/: Add a separate configuration to enable/disable KEEPALIVE socket options.
net/tcp: Arguments to TCP keep-alive timing functions probably should be struct timeval as are the times for other time-related socket options.
net/tcp: Fix a backward conditional
net/tcp: Add some more checks and debug output to TCP-protocol socket options.
net/tcp: Cosmetic changes to some alignment.
net/: Adds socket options needed to manage TCP-keepalive and TCP state machine logic to detect if that the remote peer is alive. Still missing the timer poll logic to send the keep-alive probes and the state machine logic to respond to probes.
Squashed commit of the following:
net/udp: Address most of the issues with UDP write buffering. There is a remaining issue with handling one network going down in a multi-network environment. None of this has been test but it is certainly ready for test. Hence, the feature is marked EXPERIMENTAL.
net/udp: Some baby steps toward a corrected write buffering design.
net/udp: Remove pesky write buffer macros.
Eliminate trailing space at the end of lines.
net/udp: A little more UDP write buffering logic. Still at least on big gaping hole in the design.
net/udp: Undefined CONFIG_NET_SENDTO_TIMEOUT.
net/udp: Crude, naive port of the TCP write buffering logic into UDP. This commit is certainly non-functional and is simply a starting point for the implementatin of UDP write buffering.
net/udp: Rename udp/udp_psock_sendto.c udp/udp_psock_sendto_unbuffered.c.
net/tcp: Introduce tcp receive window control based on I/O buffer
NOTE: The algorithm is still experimental but useful for http streaming.
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Approved-by: Gregory Nutt <gnutt@nuttx.org>
User-space networking stack API allows user-space daemon to
provide TCP/IP stack implementation for NuttX network.
Main use for this is to allow use and seamless integration of
HW-provided TCP/IP stacks to NuttX.
For example, user-space daemon can translate /dev/usrsock
API requests to HW TCP/IP API requests while rest of the
user-space can access standard socket API, with socket
descriptors that can be used with NuttX system calls.
