sergiotarxz/nuttx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Networking: Move relevant parts of net/Kconfig to tcp/Kconfig, udp/Kconfig, and pkt/Kconfig

Browse Source
This commit is contained in:
Gregory Nutt 2014-06-18 11:45:55 -06:00
parent e46ff63eb8
commit b9c6b9b425
6 changed files with 289 additions and 249 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

252
net/Kconfig
View File

@ -102,255 +102,9 @@ config NET_BUFSIZE
IPv6 hosts are required to be able to handle an MSS of 1220 octets,
resulting in a minimum buffer size of of 1220+20+40+14 = 1294
config NET_TCPURGDATA
bool "Urgent data"
default n
---help---
Determines if support for TCP urgent data notification should be
compiled in. Urgent data (out-of-band data) is a rarely used TCP feature
that is very seldom would be required.
menu "Raw Socket Support"
config NET_PKT
bool "Socket packet socket support"
default n
---help---
Enable or disable support for packet sockets.
Packet sockets allow receiving and transmitting frames without
a transport protocol in between. Frames received are copied into
a packet socket tap before they enter uIP. Data written into a
packet socket will bypass uIP altogether and be placed in the
transmission buffer of the network interface driver.
if NET_PKT
config NET_PKT_CONNS
int "Max packet sockets"
default 1
endif # NET_PKT
endmenu # Packet Sockets
menu "TCP/IP Networking"
config NET_TCP
bool "TCP/IP Networking"
default n
---help---
TCP support on or off
if NET_TCP
config NET_TCP_CONNS
int "Number of TCP/IP connections"
default 8
---help---
Maximum number of TCP/IP connections (all tasks)
config NET_MAX_LISTENPORTS
int "Number of listening ports"
default 20
---help---
Maximum number of listening TCP/IP ports (all tasks). Default: 20
config NET_TCP_READAHEAD
bool "Enable TCP/IP read-ahead buffering"
default y
---help---
Read-ahead buffers allows buffering of TCP/IP packets when there is no
receive in place to catch the TCP packet. In that case, the packet
will be retained in the NuttX read-ahead buffers.
You might want to disable TCP/IP read-ahead buffering on a highly
memory constrained system that does not have any TCP/IP packet rate
issues.
if NET_TCP_READAHEAD
config NET_TCP_READAHEAD_BUFSIZE
int "TCP/IP read-ahead buffer size"
default 1220 if !NET_SLIP && NET_IPv6
default 536 if !NET_SLIP && !NET_IPv6
default 256 if NET_SLIP && !NET_IPv6
---help---
Read-ahead buffers allows buffering of TCP/IP packets when there is no
receive in place to catch the TCP packet. In that case, the packet
will be retained in the NuttX read-ahead buffers.
This setting specifies the size of one TCP/IP read-ahead buffer.
This should best be a equal to the maximum packet payload size (MSS).
Optimally, the size of the read-ahead buffer will should the maximum
size of an incoming TCP packet payload (MSS). This MSS value is
determined by NET_BUFSIZE minus the size of TCP, IP, and Ethernet
headers (assuming you are using the Ethernet transport). IPv4 hosts
are required to be able to handle an MSS of 536 octets and IPv6 hosts
are required to be able to handle an MSS of 1220 octets.
config NET_NTCP_READAHEAD_BUFFERS
int "Number of TCP/IP read-ahead buffers"
default 8
---help---
Read-ahead buffers allows buffering of TCP/IP packets when there is no
receive in place to catch the TCP packet. In that case, the packet
will be retained in the NuttX read-ahead buffers.
This setting specifies the number of TCP/IP read-ahead buffers.
endif # NET_TCP_READAHEAD
config NET_TCP_WRITE_BUFFERS
bool "Enable TCP/IP write buffering"
default n
---help---
Write buffers allows buffering of ongoing TCP/IP packets, providing
for higher performance, streamed output.
You might want to disable TCP/IP write buffering on a highly memory
memory constrained system where there are no performance issues.
if NET_TCP_WRITE_BUFFERS
config NET_TCP_WRITE_BUFSIZE
int "TCP/IP write buffer size"
default 1220 if !NET_SLIP && NET_IPv6
default 536 if !NET_SLIP && !NET_IPv6
default 256 if NET_SLIP && !NET_IPv6
---help---
Write buffers allows buffering of ongoing TCP/IP packets, providing
for higher performance, streamed output.
The size of the write buffer will determine the maximum size of an
outgoing TCP packet payload (MSS). This value should NOT exceed the
maximum MSS which is determined by NET_BUFSIZE minus the size of
TCP, IP, and Ethernet headers (assuming you are using the Ethernet
transport). IPv4 hosts are required to be able to handle an MSS
of 536 octets and IPv6 hosts are required to be able to handle an
MSS of 1220 octets.
This setting specifies the size of one TCP/IP write buffer. This
should best be a equal to the maximum packet size (NET_BUFSIZE).
config NET_NTCP_WRITE_BUFFERS
int "Number of TCP/IP write buffers"
default 8
---help---
Write buffers allows buffering of ongoing TCP/IP packets, providing
for higher performance, streamed output.
This setting specifies the number of TCP/IP write buffers.
endif # NET_TCP_WRITE_BUFFERS
config NET_TCP_RECVDELAY
int "TCP Rx delay"
default 0
---help---
If NET_TCP_READAHEAD_BUFFERS is undefined, then there will be no buffering
of TCP/IP packets: Any TCP/IP packet received will be ACKed, but its contents
will be dropped in the bit-bucket.
One low-performance option is delay for a short period of time after a
TCP/IP packet is received to see if another comes right behind it. Then
the packet data from both can be combined. This option only makes since
if performance is not an issue and you need to handle short bursts of
small, back-to-back packets. The delay is in units of deciseconds.
config NET_TCPBACKLOG
bool "TCP/IP backlog support"
default n
---help---
Incoming connections pend in a backlog until accept() is called.
The size of the backlog is selected when listen() is called.
config NET_TCP_SPLIT
bool "Enable packet splitting"
default n
depends on !NET_TCP_WRITE_BUFFERS
---help---
send() will not return until the transfer has been ACKed by the
recipient. But under RFC 1122, the host need not ACK each packet
immediately; the host may wait for 500 MS before ACKing. This
combination can cause very slow performance with small transfers are
made to an RFC 1122 client. However, the RFC 1122 must ACK at least
every second (odd) packet.
This option enables logic to trick the RFC 1122 host be exploiting
this last RFC 1122 requirement: If an odd number of packets were to
be sent, then send() will split the last even packet to guarantee
that an even number of packets will be sent and the RFC 1122 host
will ACK the final packet immediately.
if NET_TCP_SPLIT
config NET_TCP_SPLIT_SIZE
int "Split size threshold"
default 40
---help---
Packets of this size or smaller than this will not be split.
endif # NET_TCP_SPLIT
config NET_SENDFILE
bool "Optimized network sendfile()"
default n
---help---
Support larger, higher performance sendfile() for transferring
files out a TCP connection.
endif # NET_TCP
endmenu # TCP/IP Networking
menu "UDP Networking"
config NET_UDP
bool "UDP Networking"
default n
depends on NET
---help---
Enable or disable UDP networking support.
if NET_UDP
config NET_UDP_CHECKSUMS
bool "UDP checksums"
default n
---help---
Enable/disable UDP checksum support
config NET_UDP_CONNS
int "Number of UDP sockets"
default 8
---help---
The maximum amount of open concurrent UDP sockets
config NET_BROADCAST
bool "UDP broadcast Rx support"
default n
---help---
Incoming UDP broadcast support
config NET_RXAVAIL
bool "Driver-based UDP backlog"
default n
---help---
One problem with UDP communications is that, unlike TCP/IP, there is
no backlog of UDP packets. So if you are listening at the precise
moment that the UDP packet is sent, it will not be received. This
is not incompatible with the properties of UDP, but can result in
bad performance if packets are missed, time out, and are resent.
Some Ethernet controllers have built-in RAM and the drivers can
support retention of UDP packets in that RAM. If the drivers
supports such a capability, this option may be enabled to use it.
NOTE: If this option is enabled, the driver must support the
rxavail() method in the uip_driver_s structure.
endif # NET_UDP
endmenu
source "net/pkt/Kconfig"
source "net/tcp/Kconfig"
source "net/udp/Kconfig"
source "net/icmp/Kconfig"
source "net/igmp/Kconfig"

7
net/net_send_buffered.c
View File

@ -43,6 +43,13 @@
#if defined(CONFIG_NET) && defined(CONFIG_NET_TCP) && \
defined(CONFIG_NET_TCP_WRITE_BUFFERS)
#if defined(CONFIG_DEBUG) && defined(CONFIG_NET_TCP_WRBUFFER_DEBUG)
/* Force debug output (from this file only) */
# undef CONFIG_DEBUG_NET
# define CONFIG_DEBUG_NET 1
#endif
#include <sys/types.h>
#include <sys/socket.h>

27
net/pkt/Kconfig Normal file
View File

@ -0,0 +1,27 @@
#
# For a description of the syntax of this configuration file,
# see misc/tools/kconfig-language.txt.
#
menu "Raw Socket Support"
config NET_PKT
bool "Socket packet socket support"
default n
---help---
Enable or disable support for packet sockets.
Packet sockets allow receiving and transmitting frames without
a transport protocol in between. Frames received are copied into
a packet socket tap before they enter uIP. Data written into a
packet socket will bypass uIP altogether and be placed in the
transmission buffer of the network interface driver.
if NET_PKT
config NET_PKT_CONNS
int "Max packet sockets"
default 1
endif # NET_PKT
endmenu # PRaw Socket Support

193
net/tcp/Kconfig Normal file
View File

@ -0,0 +1,193 @@
#
# For a description of the syntax of this configuration file,
# see misc/tools/kconfig-language.txt.
#
menu "TCP/IP Networking"
config NET_TCP
bool "TCP/IP Networking"
default n
---help---
TCP support on or off
if NET_TCP
config NET_TCPURGDATA
bool "Urgent data"
default n
---help---
Determines if support for TCP urgent data notification should be
compiled in. Urgent data (out-of-band data) is a rarely used TCP feature
that is very seldom would be required.
config NET_TCP_CONNS
int "Number of TCP/IP connections"
default 8
---help---
Maximum number of TCP/IP connections (all tasks)
config NET_MAX_LISTENPORTS
int "Number of listening ports"
default 20
---help---
Maximum number of listening TCP/IP ports (all tasks). Default: 20
config NET_TCP_READAHEAD
bool "Enable TCP/IP read-ahead buffering"
default y
---help---
Read-ahead buffers allows buffering of TCP/IP packets when there is no
receive in place to catch the TCP packet. In that case, the packet
will be retained in the NuttX read-ahead buffers.
You might want to disable TCP/IP read-ahead buffering on a highly
memory constrained system that does not have any TCP/IP packet rate
issues.
if NET_TCP_READAHEAD
config NET_TCP_READAHEAD_BUFSIZE
int "TCP/IP read-ahead buffer size"
default 1220 if !NET_SLIP && NET_IPv6
default 536 if !NET_SLIP && !NET_IPv6
default 256 if NET_SLIP && !NET_IPv6
---help---
Read-ahead buffers allows buffering of TCP/IP packets when there is no
receive in place to catch the TCP packet. In that case, the packet
will be retained in the NuttX read-ahead buffers.
This setting specifies the size of one TCP/IP read-ahead buffer.
This should best be a equal to the maximum packet payload size (MSS).
Optimally, the size of the read-ahead buffer will should the maximum
size of an incoming TCP packet payload (MSS). This MSS value is
determined by NET_BUFSIZE minus the size of TCP, IP, and Ethernet
headers (assuming you are using the Ethernet transport). IPv4 hosts
are required to be able to handle an MSS of 536 octets and IPv6 hosts
are required to be able to handle an MSS of 1220 octets.
config NET_NTCP_READAHEAD_BUFFERS
int "Number of TCP/IP read-ahead buffers"
default 8
---help---
Read-ahead buffers allows buffering of TCP/IP packets when there is no
receive in place to catch the TCP packet. In that case, the packet
will be retained in the NuttX read-ahead buffers.
This setting specifies the number of TCP/IP read-ahead buffers.
endif # NET_TCP_READAHEAD
config NET_TCP_WRITE_BUFFERS
bool "Enable TCP/IP write buffering"
default n
---help---
Write buffers allows buffering of ongoing TCP/IP packets, providing
for higher performance, streamed output.
You might want to disable TCP/IP write buffering on a highly memory
memory constrained system where there are no performance issues.
if NET_TCP_WRITE_BUFFERS
config NET_TCP_WRITE_BUFSIZE
int "TCP/IP write buffer size"
default 1220 if !NET_SLIP && NET_IPv6
default 536 if !NET_SLIP && !NET_IPv6
default 256 if NET_SLIP && !NET_IPv6
---help---
Write buffers allows buffering of ongoing TCP/IP packets, providing
for higher performance, streamed output.
The size of the write buffer will determine the maximum size of an
outgoing TCP packet payload (MSS). This value should NOT exceed the
maximum MSS which is determined by NET_BUFSIZE minus the size of
TCP, IP, and Ethernet headers (assuming you are using the Ethernet
transport). IPv4 hosts are required to be able to handle an MSS
of 536 octets and IPv6 hosts are required to be able to handle an
MSS of 1220 octets.
This setting specifies the size of one TCP/IP write buffer. This
should best be a equal to the maximum packet size (NET_BUFSIZE).
config NET_NTCP_WRITE_BUFFERS
int "Number of TCP/IP write buffers"
default 8
---help---
Write buffers allows buffering of ongoing TCP/IP packets, providing
for higher performance, streamed output.
This setting specifies the number of TCP/IP write buffers.
config NET_TCP_WRBUFFER_DEBUG
bool "Force write buffer debug"
default n
depends on DEBUG
---help---
This option will force debug output from TCP write buffer logic,
even with network debug output. This is not normally something
that would want to do but is convenient if you are debugging the
write buffer logic and do not want to get overloaded with other
network-related debug output.
endif # NET_TCP_WRITE_BUFFERS
config NET_TCP_RECVDELAY
int "TCP Rx delay"
default 0
---help---
If NET_TCP_READAHEAD_BUFFERS is undefined, then there will be no buffering
of TCP/IP packets: Any TCP/IP packet received will be ACKed, but its contents
will be dropped in the bit-bucket.
One low-performance option is delay for a short period of time after a
TCP/IP packet is received to see if another comes right behind it. Then
the packet data from both can be combined. This option only makes since
if performance is not an issue and you need to handle short bursts of
small, back-to-back packets. The delay is in units of deciseconds.
config NET_TCPBACKLOG
bool "TCP/IP backlog support"
default n
---help---
Incoming connections pend in a backlog until accept() is called.
The size of the backlog is selected when listen() is called.
config NET_TCP_SPLIT
bool "Enable packet splitting"
default n
depends on !NET_TCP_WRITE_BUFFERS
---help---
send() will not return until the transfer has been ACKed by the
recipient. But under RFC 1122, the host need not ACK each packet
immediately; the host may wait for 500 MS before ACKing. This
combination can cause very slow performance with small transfers are
made to an RFC 1122 client. However, the RFC 1122 must ACK at least
every second (odd) packet.
This option enables logic to trick the RFC 1122 host be exploiting
this last RFC 1122 requirement: If an odd number of packets were to
be sent, then send() will split the last even packet to guarantee
that an even number of packets will be sent and the RFC 1122 host
will ACK the final packet immediately.
if NET_TCP_SPLIT
config NET_TCP_SPLIT_SIZE
int "Split size threshold"
default 40
---help---
Packets of this size or smaller than this will not be split.
endif # NET_TCP_SPLIT
config NET_SENDFILE
bool "Optimized network sendfile()"
default n
---help---
Support larger, higher performance sendfile() for transferring
files out a TCP connection.
endif # NET_TCP
endmenu # TCP/IP Networking

7
net/tcp/tcp_wrbuffer.c
View File

@ -41,6 +41,13 @@
#include <nuttx/net/uip/uipopt.h>
#if defined(CONFIG_NET) && defined(CONFIG_NET_TCP) && defined(CONFIG_NET_TCP_WRITE_BUFFERS)
#if defined(CONFIG_DEBUG) && defined(CONFIG_NET_TCP_WRBUFFER_DEBUG)
/* Force debug output (from this file only) */
# undef CONFIG_DEBUG_NET
# define CONFIG_DEBUG_NET 1
#endif
#include <queue.h>
#include <semaphore.h>
#include <debug.h>

52
net/udp/Kconfig Normal file
View File

@ -0,0 +1,52 @@
#
# For a description of the syntax of this configuration file,
# see misc/tools/kconfig-language.txt.
#
menu "UDP Networking"
config NET_UDP
bool "UDP Networking"
default n
depends on NET
---help---
Enable or disable UDP networking support.
if NET_UDP
config NET_UDP_CHECKSUMS
bool "UDP checksums"
default n
---help---
Enable/disable UDP checksum support
config NET_UDP_CONNS
int "Number of UDP sockets"
default 8
---help---
The maximum amount of open concurrent UDP sockets
config NET_BROADCAST
bool "UDP broadcast Rx support"
default n
---help---
Incoming UDP broadcast support
config NET_RXAVAIL
bool "Driver-based UDP backlog"
default n
---help---
One problem with UDP communications is that, unlike TCP/IP, there is
no backlog of UDP packets. So if you are listening at the precise
moment that the UDP packet is sent, it will not be received. This
is not incompatible with the properties of UDP, but can result in
bad performance if packets are missed, time out, and are resent.
Some Ethernet controllers have built-in RAM and the drivers can
support retention of UDP packets in that RAM. If the drivers
supports such a capability, this option may be enabled to use it.
NOTE: If this option is enabled, the driver must support the
rxavail() method in the uip_driver_s structure.
endif # NET_UDP
endmenu # UDP Networking