inpcb: rename functions to inpcb_*

Inspired by rmind-smpnet patches.

inpcb: separate inpcb again to reduce the size of PCB for IPv4

The data size of PCB for IPv4 increased because of the merge of
struct in6pcb. The change decreases the size to the original size by
separating struct inpcb (again). struct in4pcb and in6pcb that embed
struct inpcb are introduced.

Even after the separation, users don't need to realize the separation
and only have to use some macros to access dedicated data. For example,
inp->inp_laddr is now accessed through in4p_laddr(inp).

inpcb: integrate data structures of PCB into one

Data structures of network protocol control blocks (PCBs), i.e.,
struct inpcb, in6pcb and inpcb_hdr, are not organized well. Users of
the data structures have to handle them separately and thus the code
is cluttered and duplicated.

The commit integrates the data structures into one, struct inpcb. As a
result, users of PCBs only have to handle just one data structure, so
the code becomes simple.

One drawback is that the data size of PCB for IPv4 increases by 40 bytes
(from 248 bytes to 288 bytes).

Add some randomness to the iss offset

branches: 1.95.6; 1.95.14;
Remove now unused tcpip.h includes. Some were already unused before.

Remove #ifdef INET. Same as tcp_input.c. Makes the code easier to
understand.

Also make tcp6_mtudisc() static in tcp_subr.c.

branches: 1.93.2;
Run tcp_slowtimo in workqueue if NET_MPSAFE

If NET_MPSAFE is enabled, we have to avoid taking softnet_lock in softint as
much as possible to prevent any softint handlers including callout handlers
such as tcp_slowtimo from sticking on softnet_lock because it results in
undesired delays of executing subsequent softint handlers.

NFCI for !NET_MPSAFE

Remove TCP_COMPAT_42. This feature is a workaround for a bug in the TCP
stack of BSD4.2. Having such features just does not make any sense, and
looking at the code, I'm not sure it actually works.

branches: 1.91.8;
fix: unlock in reverse order

branches: 1.90.2;
Sweep unnecessary route.h inclusions

sprinkle _KERNEL_OPT

branches: 1.88.2;
Do not uselessly include <sys/malloc.h>.

branches: 1.87.4;
Allow kernels compiled with INET+INET6 to be booted as IPv4-only or IPv6-only.

branches: 1.86.2; 1.86.12; 1.86.16;
NULL does not need a cast

Rewrite comments about TCP RTO calculations.

Long ago, the storage representations of srtt and rttvar were changed
from the 4.4BSD scheme, and the comments are out of sync with the
code. This commit rewrites most of the comments that explain the RTO
calculations, and points out some issues in the code.

Joint work with Bev Schwartz of BBN (original analysis and comments),
but I have rewritten and extended them, so errors are mine.

This material is based upon work supported by the Defense Advanced
Research Projects Agency and Space and Naval Warfare Systems Center,
Pacific, under Contract No. N66001-09-C-2073. Approved for Public
Release, Distribution Unlimited

branches: 1.84.8; 1.84.10;
Whitespace.

Fix kern/39769: race condition in TCP timers
When a TCP timer is disarmed (with callout_stop()) in the general case
callout_invoking() isn't checked, so the timer handler could be called run
when the current interrupt handler exits, athough the timer is disarmed.
This case cause bad things like TCPT_REXMT and TCPT_PERSIST being both pending,
causing a panic (see the PR for details).
Close the issue by aborting the handler if the timer is not callout_expired().
(the EXPIRED flag being cleared by callout_stop()).

branches: 1.82.2; 1.82.4;
tcp_delack: test for TF_DELACK.

branches: 1.81.2; 1.81.6;
Remove clause 3 and 4 from TNF licenses

branches: 1.80.2;
Merge the socket locking patch:

- Socket layer becomes MP safe.
- Unix protocols become MP safe.
- Allows protocol processing interrupts to safely block on locks.
- Fixes a number of race conditions.

With much feedback from matt@ and plunky@.

branches: 1.79.2;
Make IP, TCP, UDP, and ICMP statistics per-CPU. The stats are collated
when the user requests them via sysctl.

Change TCP stats from a structure to an array of uint64_t's.

Note: This is ABI-compatible with the old tcpstat structure; old netstat
binaries will continue to work properly.

branches: 1.77.28;
- per socket keepalive settings
- settable connection establishment timeout

branches: 1.76.2; 1.76.8; 1.76.10;
Modular (I tried ;-) TCP congestion control API. Whenever certain conditions
happen in the TCP stack, this interface calls the specified callback to
handle the situation according to the currently selected congestion
control algorithm.
A new sysctl node was created: net.inet.tcp.congctl.{available,selected}
with obvious meanings.
The old net.inet.tcp.newreno MIB was removed.
The API is discussed in tcp_congctl(9).

In the near future, it will be possible to selected a congestion control
algorithm on a per-socket basis.

Discussed on tech-net and reviewed by <yamt>.

branches: 1.75.8; 1.75.10;
integrate kauth.

Coverity CID 1153: Add KASSERT before deref.

branches: 1.73.4; 1.73.6; 1.73.8; 1.73.10; 1.73.12;
merge ktrace-lwp.

Implement PMTU checks from:

http://www.gont.com.ar/drafts/icmp-attacks-against-tcp.html

1. Don't act on ICMP-need-frag immediately if adhoc checks on the
advertised MTU fail. The MTU update is delayed until a TCP retransmit
happens.
2. Ignore ICMP Source Quench messages meant for TCP connections.

From OpenBSD.

branches: 1.71.2; 1.71.4;
Copyright maintenance.

Commit TCP SACK patches from Kentaro A. Karahone's patch at:
http://www.sigusr1.org/~kurahone/tcp-sack-netbsd-02152005.diff.gz

Fixes in that patch for pre-existing TCP pcb initializations were already
committed to NetBSD-current, so are not included in this commit.

The SACK patch has been observed to correctly negotiate and respond,
to SACKs in wide-area traffic.

There are two indepenently-observed, as-yet-unresolved anomalies:
First, seeing unexplained delays between in fast retransmission
(potentially explainable by an 0.2sec RTT between adjacent
ethernet/wifi NICs); and second, peculiar and unepxlained TCP
retransmits observed over an ath0 card.

After discussion with several interested developers, I'm committing
this now, as-is, for more eyes to use and look over. Current hypothesis
is that the anomalies above may in fact be due to link/level (hardware,
driver, HAL, firmware) abberations in the test setup, affecting both
Kentaro's wired-Ethernet NIC and in my two (different) WiFi NICs.

ANSIfy function declarations

Whoops. Exit fast recovery when handling a timeout.

Fix two problems in our TCP stack:

1) If an echoed RFC 1323 time stamp appears to be later than the current time,
ignore it and fall back to old-style RTT calculation. This prevents ending
up with a negative RTT and panicking later.

2) Fix NewReno. This involves a few changes:

a) Implement the send_high variable in RFC 2582. Our implementation is
subtly different; it is one *past* the last sequence number transmitted
rather than being equal to it. This simplifies some logic and makes
the code smaller. Additional logic was required to prevent sequence
number wraparound problems; this is not mentioned in RFC 2582.

b) Make sure we reset t_dupacks on new acks, but *not* on a partial ack.
All of the new ack code is pushed out into tcp_newreno(). (Later this
will probably be a pluggable function.) Thus t_dupacks keeps track of
whether we're in fast recovery all the time, with Reno or NewReno, which
keeps some logic simpler.

c) We do not need to update snd_recover when we're not in fast recovery.
See tech-net for an explanation of this.

d) In the gratuitous fast retransmit prevention case, do not send a packet.
RFC 2582 specifically says that we should "do nothing".

e) Do not inflate the congestion window on a partial ack. (This is done by
testing t_dupacks to see whether we're still in fast recovery.)

This brings the performance of NewReno back up to the same as Reno in a few
random test cases (e.g. transferring peer-to-peer over my wireless network).
I have not concocted a good test case for the behavior specific to NewReno.

branches: 1.66.8; 1.66.10;
some corrections from markus@openbsd;
- callout_ack() was called with wrong argument

make it compilable with TCP_DEBUG defined

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

As a temporary workaround, apply the fix from PR#20390, thereby
cooperating with the callout code in working around the race
condition caused by the TCP code's use of the callout facility.

Instead of unconditionally releasing memory in tcp_close() and
SYN_CACHE_PUT(), check whether any of the related callout handlers
are about to be invoked (but have not yet done callout_ack()), and
if so, just mark the associated data structure (tcpcb or syn cache
entry) as "dead", and test for this (and release storage) in the
callout handler functions.

branches: 1.62.2;
Test callout_pending(), not callout_active(), and eliminate now-unnecessary
callout_deactivate() calls.

Quell an uninitialised variable warning.

Guard use of "so" in tcp_timer_persist() and tcp_timer_2msl() with
#ifdef TCP_DEBUG.

whitespace

path MTU discovery blackhole detection.
PR 12790 (sorry for not committing it for a long time)

branches: 1.57.8; 1.57.10;
add RCSIDs

Change a few variable/tables to const since they are read-only.

branches: 1.55.2;
Use callouts for SYN cache timers, rather than traversing time queues
in tcp_slowtimo().

Update copyrights.

Use callouts for TCP timers, rather than traversing the list of
all open TCP connections in tcp_slowtimo() (which is called 2x
per second). It's fairly rare for TCP timers to actually fire,
so saving this list traversal is good, especially if you want
to scale to thousands of open connections.

Initialize TCP timer variables in a new function, tcp_timer_init().

Split tcp_timers() into multiple functions, one for each timer,
and call it directly from tcp_slowtimo() (via a table) rather
than going through tcp_userreq().

This will allow us to call TCP timers directly from callouts,
in a future revision.

Change the way receive idle time and round trip time are measured.
Instead of incrementing t_idle and t_rtt in tcp_slowtimo(), we now
take a timstamp (via tcp_now) and use subtraction to compute the
delta when we actually need it (using unsigned arithmetic so that
tcp_now wrapping is handled correctly).

Based on similar changes in FreeBSD.

Use a callout for the delayed ACK timer, and delete tcp_fasttimo().
Expose the delayed ACK timer as net.inet.tcp.delack_ticks.

branches: 1.48.2; 1.48.4; 1.48.6;
remove #ifdef TCP6. it is not likely for us to bring in sys/netinet6/tcp6*.c
(separate TCP/IPv6 stack) into netbsd-current.

be more friendly with INET-less build.
XXX we need to do more to do a working INET-less build

Remove register declarations.

branches: 1.45.2;
Use proper ip protocol # field and tcp hdr on sending RST against SYN,
when ip header and tcp header are not adjacent to each other
(i.e. when ip6 options are attached).

To test this, try
telnet @::1@::1 port
toward a port without responding server. Prior to the fix, the kernel will
generate broken RST packet.

IPv6 kernel code, based on KAME/NetBSD 1.4, SNAP kit 19990628.
(Sorry for a big commit, I can't separate this into several pieces...)
Pls check sys/netinet6/TODO and sys/netinet6/IMPLEMENTATION for details.

- sys/kern: do not assume single mbuf, accept chained mbuf on passing
data from userland to kernel (or other way round).
- "midway" ATM card: ATM PVC pseudo device support, like those done in ALTQ
package (ftp://ftp.csl.sony.co.jp/pub/kjc/).
- sys/netinet/tcp*: IPv4/v6 dual stack tcp support.
- sys/netinet/{ip6,icmp6}.h, sys/net/pfkeyv2.h: IETF document assumes those
file to be there so we patch it up.
- sys/netinet: IPsec additions are here and there.
- sys/netinet6/*: most of IPv6 code sits here.
- sys/netkey: IPsec key management code
- dev/pci/pcidevs: regen

In my understanding no code here is subject to export control so it
should be safe.

branches: 1.43.8; 1.43.10;
Create tcp.keepidle, tcp.keepintvl, tcp.keepcnt, tcp.slowhz sysctls.

Fix a couple of bogons related to tcp_new_iss():
* Don't add tcp_iss_seq when creating a new ISS from TIME-WAIT state.
* Do the clock increment even when using the rnd device.

Comment where we use the Loss Window.

Loss window MUST be one segment, per draft-floyd-incr-init-win-03.

Make sure a timer is marked "disarmed" once it has expired.

Nuke TUBA per my note to tech-net; there's no reason to keep it around.

Define all TCP timers in terms of PRT timers.

Use macros from tcp_timer.h to manipulate TCP timers, so that their
implementation can be changed easily.

Remove bogus black hole discovery code

Make use of the work-arounds for ancient broken TCP peers run-time
conditional (tcp_compat_42). The kernel config option TCP_COMPAT_42
will still enable this by default, or disable this by default if the
option is not included (i.e. current behavior). This will be made a
sysctl soon.

Add support for deletion of routes added by path MTU discovery;
uses new generic route timeout code. Add sysctl for timeout period.

Fix a potential-congestion case in the larger initial congestion window
code, as clarified in the TCPIMPL WG meeting at IETF #41: If the SYN
(active open) or SYN,ACK (passive open) was retransmitted, the initial
congestion window for the first slow start of that connection must be
one segment.

Fix a retransmission bug introduced by the Brakmo and Peterson
RTO estimation changes. Under some circumstances it would return a value
of 0, while the old Van Jacobson RTO code would return a minimum of 3.
This would result in 12 retransmissions, each 1 second apart.
This takes care of those instances, and ensures that t_rttmin is
used everywhere as a lower bound.

Update copyright (sigh, should have done this long ago).

Use option header file for TCP_COMPAT_42

Finishing merging 4.4BSD-Lite2 netinet. At this point, the only changes
left were SCCS IDs and Copyright dates.

Implement a queue for delayed ACK processing. This queue is used in
tcp_fasttimo() in lieu of scanning all open TCP connections.

From 4.4BSD-Lite2:
- When running the slow timers, skip PCBs in LISTEN state.
- When processing the persist timer, drop the connection if the connection
idle time exceeds the maximum backoff for retransmit. Part of
kern/2335 (pete@daemon.net).

Implement an infrastructure to allow larger initial congestion windows.
The sysctl'able variable "tcp_init_win", when set to 0, selects an
auto-tuning algorithm for selecting the initial window, based on transmit
segment size, per discussion in the IETF tcpimpl working group.

Default initial window is still 1 segment, but will soon become 2 segments,
per discussion in tcpimpl.

In tcp_fasttimo(), don't clear TF_DELACK; we need it to count delayed ACKs
in tcp_output(), and it will only be cleared in tcp_output() if the ACK was
transmitted sucessfully. Also, don't count delayed ACKs here, let tcp_output()
count them.

Costmetic change: use intotcpcb() in tcp_fasttimo().

TCP MSS fixes to provide cleaner slow-start and recovery.

branches: 1.21.2;
o Make usage of /dev/random dependant on
pseudo-device rnd # /dev/random and in-kernel generator
in config files.

o Add declaration to all architectures.

o Clean up copyright message in rnd.c, rnd.h, and rndpool.c to include
that this code is derived in part from Ted Tyso's linux code.

Add hooks to use the kernel random system to generate TCP sequence numbers.

branches: 1.19.2;
Garbage-collect some "extern"s.

Pull SYN_cache_branch down into the main line.

branches: 1.17.8;
Fix RTT scaling problems introduced with Brakmo and Peterson changes.

Add in_nullhost() and in_hosteq() macros, to hide some protocol
details. Also, fix a bug in TCP wrt SYN+URG packets.

Pass a proc pointer down to the usrreq and pcbbind functions for PRU_ATTACH, PRU_BIND and
PRU_CONTROL. The usrreq interface really needs to be split up, but this will have to wait.
Remove SS_PRIV completely.

branches: 1.14.4;
netinet prototypes

splnet --> splsoftnet

convert pcb lists to CIRCLEQs, so that the end can be looked at more
easily, and so that the original (insque/remque) logic can be effectively
mimiced. (This fixes a bug in the previous set of list changes.)
also (since terminator is no longer null) reinstate uninitted list checks,
but mark them XXX.

Various cleanup, including:
* Convert several data structures to use queue.h.
* Split in_pcbnotify() into two parts; one for notifying a specific PCB, and
one for notifying all PCBs for a particular foreign address.

be a bit more careful and explicit with types. (basically a large no-op.)

Don't return received data to the user until the initial handshake is complete.
Also use TCPS_HAVEESTABLISHED() in a few other places.

New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'

Update to 4.4-Lite networking code, with a few local changes.

Prototypes.

Fix some inconsistent spacing; spaces at the end of lines, etc.

Canonicalize all #includes.

branches: 1.3.4;
add include of select.h if necessary for protos, or delete if extraneous

make kernel select interface be one-stop shopping & clean it all up.

branches: 1.1.1;
Initial revision

inpcb: separate inpcb again to reduce the size of PCB for IPv4

The data size of PCB for IPv4 increased because of the merge of
struct in6pcb. The change decreases the size to the original size by
separating struct inpcb (again). struct in4pcb and in6pcb that embed
struct inpcb are introduced.

Even after the separation, users don't need to realize the separation
and only have to use some macros to access dedicated data. For example,
inp->inp_laddr is now accessed through in4p_laddr(inp).

inpcb: integrate data structures of PCB into one

Data structures of network protocol control blocks (PCBs), i.e.,
struct inpcb, in6pcb and inpcb_hdr, are not organized well. Users of
the data structures have to handle them separately and thus the code
is cluttered and duplicated.

The commit integrates the data structures into one, struct inpcb. As a
result, users of PCBs only have to handle just one data structure, so
the code becomes simple.

One drawback is that the data size of PCB for IPv4 increases by 40 bytes
(from 248 bytes to 288 bytes).

Add some randomness to the iss offset

branches: 1.95.6; 1.95.14;
Remove now unused tcpip.h includes. Some were already unused before.

Remove #ifdef INET. Same as tcp_input.c. Makes the code easier to
understand.

Also make tcp6_mtudisc() static in tcp_subr.c.

branches: 1.93.2;
Run tcp_slowtimo in workqueue if NET_MPSAFE

If NET_MPSAFE is enabled, we have to avoid taking softnet_lock in softint as
much as possible to prevent any softint handlers including callout handlers
such as tcp_slowtimo from sticking on softnet_lock because it results in
undesired delays of executing subsequent softint handlers.

NFCI for !NET_MPSAFE

Remove TCP_COMPAT_42. This feature is a workaround for a bug in the TCP
stack of BSD4.2. Having such features just does not make any sense, and
looking at the code, I'm not sure it actually works.

branches: 1.91.8;
fix: unlock in reverse order

branches: 1.90.2;
Sweep unnecessary route.h inclusions

sprinkle _KERNEL_OPT

branches: 1.88.2;
Do not uselessly include <sys/malloc.h>.

branches: 1.87.4;
Allow kernels compiled with INET+INET6 to be booted as IPv4-only or IPv6-only.

branches: 1.86.2; 1.86.12; 1.86.16;
NULL does not need a cast

Rewrite comments about TCP RTO calculations.

Long ago, the storage representations of srtt and rttvar were changed
from the 4.4BSD scheme, and the comments are out of sync with the
code. This commit rewrites most of the comments that explain the RTO
calculations, and points out some issues in the code.

Joint work with Bev Schwartz of BBN (original analysis and comments),
but I have rewritten and extended them, so errors are mine.

This material is based upon work supported by the Defense Advanced
Research Projects Agency and Space and Naval Warfare Systems Center,
Pacific, under Contract No. N66001-09-C-2073. Approved for Public
Release, Distribution Unlimited

branches: 1.84.8; 1.84.10;
Whitespace.

Fix kern/39769: race condition in TCP timers
When a TCP timer is disarmed (with callout_stop()) in the general case
callout_invoking() isn't checked, so the timer handler could be called run
when the current interrupt handler exits, athough the timer is disarmed.
This case cause bad things like TCPT_REXMT and TCPT_PERSIST being both pending,
causing a panic (see the PR for details).
Close the issue by aborting the handler if the timer is not callout_expired().
(the EXPIRED flag being cleared by callout_stop()).

branches: 1.82.2; 1.82.4;
tcp_delack: test for TF_DELACK.

branches: 1.81.2; 1.81.6;
Remove clause 3 and 4 from TNF licenses

branches: 1.80.2;
Merge the socket locking patch:

- Socket layer becomes MP safe.
- Unix protocols become MP safe.
- Allows protocol processing interrupts to safely block on locks.
- Fixes a number of race conditions.

With much feedback from matt@ and plunky@.

branches: 1.79.2;
Make IP, TCP, UDP, and ICMP statistics per-CPU. The stats are collated
when the user requests them via sysctl.

Change TCP stats from a structure to an array of uint64_t's.

Note: This is ABI-compatible with the old tcpstat structure; old netstat
binaries will continue to work properly.

branches: 1.77.28;
- per socket keepalive settings
- settable connection establishment timeout

branches: 1.76.2; 1.76.8; 1.76.10;
Modular (I tried ;-) TCP congestion control API. Whenever certain conditions
happen in the TCP stack, this interface calls the specified callback to
handle the situation according to the currently selected congestion
control algorithm.
A new sysctl node was created: net.inet.tcp.congctl.{available,selected}
with obvious meanings.
The old net.inet.tcp.newreno MIB was removed.
The API is discussed in tcp_congctl(9).

In the near future, it will be possible to selected a congestion control
algorithm on a per-socket basis.

Discussed on tech-net and reviewed by <yamt>.

branches: 1.75.8; 1.75.10;
integrate kauth.

Coverity CID 1153: Add KASSERT before deref.

branches: 1.73.4; 1.73.6; 1.73.8; 1.73.10; 1.73.12;
merge ktrace-lwp.

Implement PMTU checks from:

http://www.gont.com.ar/drafts/icmp-attacks-against-tcp.html

1. Don't act on ICMP-need-frag immediately if adhoc checks on the
advertised MTU fail. The MTU update is delayed until a TCP retransmit
happens.
2. Ignore ICMP Source Quench messages meant for TCP connections.

From OpenBSD.

branches: 1.71.2; 1.71.4;
Copyright maintenance.

Commit TCP SACK patches from Kentaro A. Karahone's patch at:
http://www.sigusr1.org/~kurahone/tcp-sack-netbsd-02152005.diff.gz

Fixes in that patch for pre-existing TCP pcb initializations were already
committed to NetBSD-current, so are not included in this commit.

The SACK patch has been observed to correctly negotiate and respond,
to SACKs in wide-area traffic.

There are two indepenently-observed, as-yet-unresolved anomalies:
First, seeing unexplained delays between in fast retransmission
(potentially explainable by an 0.2sec RTT between adjacent
ethernet/wifi NICs); and second, peculiar and unepxlained TCP
retransmits observed over an ath0 card.

After discussion with several interested developers, I'm committing
this now, as-is, for more eyes to use and look over. Current hypothesis
is that the anomalies above may in fact be due to link/level (hardware,
driver, HAL, firmware) abberations in the test setup, affecting both
Kentaro's wired-Ethernet NIC and in my two (different) WiFi NICs.

ANSIfy function declarations

Whoops. Exit fast recovery when handling a timeout.

Fix two problems in our TCP stack:

1) If an echoed RFC 1323 time stamp appears to be later than the current time,
ignore it and fall back to old-style RTT calculation. This prevents ending
up with a negative RTT and panicking later.

2) Fix NewReno. This involves a few changes:

a) Implement the send_high variable in RFC 2582. Our implementation is
subtly different; it is one *past* the last sequence number transmitted
rather than being equal to it. This simplifies some logic and makes
the code smaller. Additional logic was required to prevent sequence
number wraparound problems; this is not mentioned in RFC 2582.

b) Make sure we reset t_dupacks on new acks, but *not* on a partial ack.
All of the new ack code is pushed out into tcp_newreno(). (Later this
will probably be a pluggable function.) Thus t_dupacks keeps track of
whether we're in fast recovery all the time, with Reno or NewReno, which
keeps some logic simpler.

c) We do not need to update snd_recover when we're not in fast recovery.
See tech-net for an explanation of this.

d) In the gratuitous fast retransmit prevention case, do not send a packet.
RFC 2582 specifically says that we should "do nothing".

e) Do not inflate the congestion window on a partial ack. (This is done by
testing t_dupacks to see whether we're still in fast recovery.)

This brings the performance of NewReno back up to the same as Reno in a few
random test cases (e.g. transferring peer-to-peer over my wireless network).
I have not concocted a good test case for the behavior specific to NewReno.

branches: 1.66.8; 1.66.10;
some corrections from markus@openbsd;
- callout_ack() was called with wrong argument

make it compilable with TCP_DEBUG defined

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

As a temporary workaround, apply the fix from PR#20390, thereby
cooperating with the callout code in working around the race
condition caused by the TCP code's use of the callout facility.

Instead of unconditionally releasing memory in tcp_close() and
SYN_CACHE_PUT(), check whether any of the related callout handlers
are about to be invoked (but have not yet done callout_ack()), and
if so, just mark the associated data structure (tcpcb or syn cache
entry) as "dead", and test for this (and release storage) in the
callout handler functions.

branches: 1.62.2;
Test callout_pending(), not callout_active(), and eliminate now-unnecessary
callout_deactivate() calls.

Quell an uninitialised variable warning.

Guard use of "so" in tcp_timer_persist() and tcp_timer_2msl() with
#ifdef TCP_DEBUG.

whitespace

path MTU discovery blackhole detection.
PR 12790 (sorry for not committing it for a long time)

branches: 1.57.8; 1.57.10;
add RCSIDs

Change a few variable/tables to const since they are read-only.

branches: 1.55.2;
Use callouts for SYN cache timers, rather than traversing time queues
in tcp_slowtimo().

Update copyrights.

Use callouts for TCP timers, rather than traversing the list of
all open TCP connections in tcp_slowtimo() (which is called 2x
per second). It's fairly rare for TCP timers to actually fire,
so saving this list traversal is good, especially if you want
to scale to thousands of open connections.

Initialize TCP timer variables in a new function, tcp_timer_init().

Split tcp_timers() into multiple functions, one for each timer,
and call it directly from tcp_slowtimo() (via a table) rather
than going through tcp_userreq().

This will allow us to call TCP timers directly from callouts,
in a future revision.

Change the way receive idle time and round trip time are measured.
Instead of incrementing t_idle and t_rtt in tcp_slowtimo(), we now
take a timstamp (via tcp_now) and use subtraction to compute the
delta when we actually need it (using unsigned arithmetic so that
tcp_now wrapping is handled correctly).

Based on similar changes in FreeBSD.

Use a callout for the delayed ACK timer, and delete tcp_fasttimo().
Expose the delayed ACK timer as net.inet.tcp.delack_ticks.

branches: 1.48.2; 1.48.4; 1.48.6;
remove #ifdef TCP6. it is not likely for us to bring in sys/netinet6/tcp6*.c
(separate TCP/IPv6 stack) into netbsd-current.

be more friendly with INET-less build.
XXX we need to do more to do a working INET-less build

Remove register declarations.

branches: 1.45.2;
Use proper ip protocol # field and tcp hdr on sending RST against SYN,
when ip header and tcp header are not adjacent to each other
(i.e. when ip6 options are attached).

To test this, try
telnet @::1@::1 port
toward a port without responding server. Prior to the fix, the kernel will
generate broken RST packet.

IPv6 kernel code, based on KAME/NetBSD 1.4, SNAP kit 19990628.
(Sorry for a big commit, I can't separate this into several pieces...)
Pls check sys/netinet6/TODO and sys/netinet6/IMPLEMENTATION for details.

- sys/kern: do not assume single mbuf, accept chained mbuf on passing
data from userland to kernel (or other way round).
- "midway" ATM card: ATM PVC pseudo device support, like those done in ALTQ
package (ftp://ftp.csl.sony.co.jp/pub/kjc/).
- sys/netinet/tcp*: IPv4/v6 dual stack tcp support.
- sys/netinet/{ip6,icmp6}.h, sys/net/pfkeyv2.h: IETF document assumes those
file to be there so we patch it up.
- sys/netinet: IPsec additions are here and there.
- sys/netinet6/*: most of IPv6 code sits here.
- sys/netkey: IPsec key management code
- dev/pci/pcidevs: regen

In my understanding no code here is subject to export control so it
should be safe.

branches: 1.43.8; 1.43.10;
Create tcp.keepidle, tcp.keepintvl, tcp.keepcnt, tcp.slowhz sysctls.

Fix a couple of bogons related to tcp_new_iss():
* Don't add tcp_iss_seq when creating a new ISS from TIME-WAIT state.
* Do the clock increment even when using the rnd device.

Comment where we use the Loss Window.

Loss window MUST be one segment, per draft-floyd-incr-init-win-03.

Make sure a timer is marked "disarmed" once it has expired.

Nuke TUBA per my note to tech-net; there's no reason to keep it around.

Define all TCP timers in terms of PRT timers.

Use macros from tcp_timer.h to manipulate TCP timers, so that their
implementation can be changed easily.

Remove bogus black hole discovery code

Make use of the work-arounds for ancient broken TCP peers run-time
conditional (tcp_compat_42). The kernel config option TCP_COMPAT_42
will still enable this by default, or disable this by default if the
option is not included (i.e. current behavior). This will be made a
sysctl soon.

Add support for deletion of routes added by path MTU discovery;
uses new generic route timeout code. Add sysctl for timeout period.

Fix a potential-congestion case in the larger initial congestion window
code, as clarified in the TCPIMPL WG meeting at IETF #41: If the SYN
(active open) or SYN,ACK (passive open) was retransmitted, the initial
congestion window for the first slow start of that connection must be
one segment.

Fix a retransmission bug introduced by the Brakmo and Peterson
RTO estimation changes. Under some circumstances it would return a value
of 0, while the old Van Jacobson RTO code would return a minimum of 3.
This would result in 12 retransmissions, each 1 second apart.
This takes care of those instances, and ensures that t_rttmin is
used everywhere as a lower bound.

Update copyright (sigh, should have done this long ago).

Use option header file for TCP_COMPAT_42

Finishing merging 4.4BSD-Lite2 netinet. At this point, the only changes
left were SCCS IDs and Copyright dates.

Implement a queue for delayed ACK processing. This queue is used in
tcp_fasttimo() in lieu of scanning all open TCP connections.

From 4.4BSD-Lite2:
- When running the slow timers, skip PCBs in LISTEN state.
- When processing the persist timer, drop the connection if the connection
idle time exceeds the maximum backoff for retransmit. Part of
kern/2335 (pete@daemon.net).

Implement an infrastructure to allow larger initial congestion windows.
The sysctl'able variable "tcp_init_win", when set to 0, selects an
auto-tuning algorithm for selecting the initial window, based on transmit
segment size, per discussion in the IETF tcpimpl working group.

Default initial window is still 1 segment, but will soon become 2 segments,
per discussion in tcpimpl.

In tcp_fasttimo(), don't clear TF_DELACK; we need it to count delayed ACKs
in tcp_output(), and it will only be cleared in tcp_output() if the ACK was
transmitted sucessfully. Also, don't count delayed ACKs here, let tcp_output()
count them.

Costmetic change: use intotcpcb() in tcp_fasttimo().

TCP MSS fixes to provide cleaner slow-start and recovery.

branches: 1.21.2;
o Make usage of /dev/random dependant on
pseudo-device rnd # /dev/random and in-kernel generator
in config files.

o Add declaration to all architectures.

o Clean up copyright message in rnd.c, rnd.h, and rndpool.c to include
that this code is derived in part from Ted Tyso's linux code.

Add hooks to use the kernel random system to generate TCP sequence numbers.

branches: 1.19.2;
Garbage-collect some "extern"s.

Pull SYN_cache_branch down into the main line.

branches: 1.17.8;
Fix RTT scaling problems introduced with Brakmo and Peterson changes.

Add in_nullhost() and in_hosteq() macros, to hide some protocol
details. Also, fix a bug in TCP wrt SYN+URG packets.

Pass a proc pointer down to the usrreq and pcbbind functions for PRU_ATTACH, PRU_BIND and
PRU_CONTROL. The usrreq interface really needs to be split up, but this will have to wait.
Remove SS_PRIV completely.

branches: 1.14.4;
netinet prototypes

splnet --> splsoftnet

convert pcb lists to CIRCLEQs, so that the end can be looked at more
easily, and so that the original (insque/remque) logic can be effectively
mimiced. (This fixes a bug in the previous set of list changes.)
also (since terminator is no longer null) reinstate uninitted list checks,
but mark them XXX.

Various cleanup, including:
* Convert several data structures to use queue.h.
* Split in_pcbnotify() into two parts; one for notifying a specific PCB, and
one for notifying all PCBs for a particular foreign address.

be a bit more careful and explicit with types. (basically a large no-op.)

Don't return received data to the user until the initial handshake is complete.
Also use TCPS_HAVEESTABLISHED() in a few other places.

New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'

Update to 4.4-Lite networking code, with a few local changes.

Prototypes.

Fix some inconsistent spacing; spaces at the end of lines, etc.

Canonicalize all #includes.

branches: 1.3.4;
add include of select.h if necessary for protos, or delete if extraneous

make kernel select interface be one-stop shopping & clean it all up.

branches: 1.1.1;
Initial revision

Add some randomness to the iss offset

Remove now unused tcpip.h includes. Some were already unused before.

Remove #ifdef INET. Same as tcp_input.c. Makes the code easier to
understand.

Also make tcp6_mtudisc() static in tcp_subr.c.

branches: 1.93.2;
Run tcp_slowtimo in workqueue if NET_MPSAFE

If NET_MPSAFE is enabled, we have to avoid taking softnet_lock in softint as
much as possible to prevent any softint handlers including callout handlers
such as tcp_slowtimo from sticking on softnet_lock because it results in
undesired delays of executing subsequent softint handlers.

NFCI for !NET_MPSAFE

Remove TCP_COMPAT_42. This feature is a workaround for a bug in the TCP
stack of BSD4.2. Having such features just does not make any sense, and
looking at the code, I'm not sure it actually works.

branches: 1.91.8;
fix: unlock in reverse order

branches: 1.90.2;
Sweep unnecessary route.h inclusions

sprinkle _KERNEL_OPT

branches: 1.88.2;
Do not uselessly include <sys/malloc.h>.

branches: 1.87.4;
Allow kernels compiled with INET+INET6 to be booted as IPv4-only or IPv6-only.

branches: 1.86.2; 1.86.12; 1.86.16;
NULL does not need a cast

Rewrite comments about TCP RTO calculations.

Long ago, the storage representations of srtt and rttvar were changed
from the 4.4BSD scheme, and the comments are out of sync with the
code. This commit rewrites most of the comments that explain the RTO
calculations, and points out some issues in the code.

Joint work with Bev Schwartz of BBN (original analysis and comments),
but I have rewritten and extended them, so errors are mine.

This material is based upon work supported by the Defense Advanced
Research Projects Agency and Space and Naval Warfare Systems Center,
Pacific, under Contract No. N66001-09-C-2073. Approved for Public
Release, Distribution Unlimited

branches: 1.84.8; 1.84.10;
Whitespace.

Fix kern/39769: race condition in TCP timers
When a TCP timer is disarmed (with callout_stop()) in the general case
callout_invoking() isn't checked, so the timer handler could be called run
when the current interrupt handler exits, athough the timer is disarmed.
This case cause bad things like TCPT_REXMT and TCPT_PERSIST being both pending,
causing a panic (see the PR for details).
Close the issue by aborting the handler if the timer is not callout_expired().
(the EXPIRED flag being cleared by callout_stop()).

branches: 1.82.2; 1.82.4;
tcp_delack: test for TF_DELACK.

branches: 1.81.2; 1.81.6;
Remove clause 3 and 4 from TNF licenses

branches: 1.80.2;
Merge the socket locking patch:

- Socket layer becomes MP safe.
- Unix protocols become MP safe.
- Allows protocol processing interrupts to safely block on locks.
- Fixes a number of race conditions.

With much feedback from matt@ and plunky@.

branches: 1.79.2;
Make IP, TCP, UDP, and ICMP statistics per-CPU. The stats are collated
when the user requests them via sysctl.

Change TCP stats from a structure to an array of uint64_t's.

Note: This is ABI-compatible with the old tcpstat structure; old netstat
binaries will continue to work properly.

branches: 1.77.28;
- per socket keepalive settings
- settable connection establishment timeout

branches: 1.76.2; 1.76.8; 1.76.10;
Modular (I tried ;-) TCP congestion control API. Whenever certain conditions
happen in the TCP stack, this interface calls the specified callback to
handle the situation according to the currently selected congestion
control algorithm.
A new sysctl node was created: net.inet.tcp.congctl.{available,selected}
with obvious meanings.
The old net.inet.tcp.newreno MIB was removed.
The API is discussed in tcp_congctl(9).

In the near future, it will be possible to selected a congestion control
algorithm on a per-socket basis.

Discussed on tech-net and reviewed by <yamt>.

branches: 1.75.8; 1.75.10;
integrate kauth.

Coverity CID 1153: Add KASSERT before deref.

branches: 1.73.4; 1.73.6; 1.73.8; 1.73.10; 1.73.12;
merge ktrace-lwp.

Implement PMTU checks from:

http://www.gont.com.ar/drafts/icmp-attacks-against-tcp.html

1. Don't act on ICMP-need-frag immediately if adhoc checks on the
advertised MTU fail. The MTU update is delayed until a TCP retransmit
happens.
2. Ignore ICMP Source Quench messages meant for TCP connections.

From OpenBSD.

branches: 1.71.2; 1.71.4;
Copyright maintenance.

Commit TCP SACK patches from Kentaro A. Karahone's patch at:
http://www.sigusr1.org/~kurahone/tcp-sack-netbsd-02152005.diff.gz

Fixes in that patch for pre-existing TCP pcb initializations were already
committed to NetBSD-current, so are not included in this commit.

The SACK patch has been observed to correctly negotiate and respond,
to SACKs in wide-area traffic.

There are two indepenently-observed, as-yet-unresolved anomalies:
First, seeing unexplained delays between in fast retransmission
(potentially explainable by an 0.2sec RTT between adjacent
ethernet/wifi NICs); and second, peculiar and unepxlained TCP
retransmits observed over an ath0 card.

After discussion with several interested developers, I'm committing
this now, as-is, for more eyes to use and look over. Current hypothesis
is that the anomalies above may in fact be due to link/level (hardware,
driver, HAL, firmware) abberations in the test setup, affecting both
Kentaro's wired-Ethernet NIC and in my two (different) WiFi NICs.

ANSIfy function declarations

Whoops. Exit fast recovery when handling a timeout.

Fix two problems in our TCP stack:

1) If an echoed RFC 1323 time stamp appears to be later than the current time,
ignore it and fall back to old-style RTT calculation. This prevents ending
up with a negative RTT and panicking later.

2) Fix NewReno. This involves a few changes:

a) Implement the send_high variable in RFC 2582. Our implementation is
subtly different; it is one *past* the last sequence number transmitted
rather than being equal to it. This simplifies some logic and makes
the code smaller. Additional logic was required to prevent sequence
number wraparound problems; this is not mentioned in RFC 2582.

b) Make sure we reset t_dupacks on new acks, but *not* on a partial ack.
All of the new ack code is pushed out into tcp_newreno(). (Later this
will probably be a pluggable function.) Thus t_dupacks keeps track of
whether we're in fast recovery all the time, with Reno or NewReno, which
keeps some logic simpler.

c) We do not need to update snd_recover when we're not in fast recovery.
See tech-net for an explanation of this.

d) In the gratuitous fast retransmit prevention case, do not send a packet.
RFC 2582 specifically says that we should "do nothing".

e) Do not inflate the congestion window on a partial ack. (This is done by
testing t_dupacks to see whether we're still in fast recovery.)

This brings the performance of NewReno back up to the same as Reno in a few
random test cases (e.g. transferring peer-to-peer over my wireless network).
I have not concocted a good test case for the behavior specific to NewReno.

branches: 1.66.8; 1.66.10;
some corrections from markus@openbsd;
- callout_ack() was called with wrong argument

make it compilable with TCP_DEBUG defined

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

As a temporary workaround, apply the fix from PR#20390, thereby
cooperating with the callout code in working around the race
condition caused by the TCP code's use of the callout facility.

Instead of unconditionally releasing memory in tcp_close() and
SYN_CACHE_PUT(), check whether any of the related callout handlers
are about to be invoked (but have not yet done callout_ack()), and
if so, just mark the associated data structure (tcpcb or syn cache
entry) as "dead", and test for this (and release storage) in the
callout handler functions.

branches: 1.62.2;
Test callout_pending(), not callout_active(), and eliminate now-unnecessary
callout_deactivate() calls.

Quell an uninitialised variable warning.

Guard use of "so" in tcp_timer_persist() and tcp_timer_2msl() with
#ifdef TCP_DEBUG.

whitespace

path MTU discovery blackhole detection.
PR 12790 (sorry for not committing it for a long time)

branches: 1.57.8; 1.57.10;
add RCSIDs

Change a few variable/tables to const since they are read-only.

branches: 1.55.2;
Use callouts for SYN cache timers, rather than traversing time queues
in tcp_slowtimo().

Update copyrights.

Use callouts for TCP timers, rather than traversing the list of
all open TCP connections in tcp_slowtimo() (which is called 2x
per second). It's fairly rare for TCP timers to actually fire,
so saving this list traversal is good, especially if you want
to scale to thousands of open connections.

Initialize TCP timer variables in a new function, tcp_timer_init().

Split tcp_timers() into multiple functions, one for each timer,
and call it directly from tcp_slowtimo() (via a table) rather
than going through tcp_userreq().

This will allow us to call TCP timers directly from callouts,
in a future revision.

Change the way receive idle time and round trip time are measured.
Instead of incrementing t_idle and t_rtt in tcp_slowtimo(), we now
take a timstamp (via tcp_now) and use subtraction to compute the
delta when we actually need it (using unsigned arithmetic so that
tcp_now wrapping is handled correctly).

Based on similar changes in FreeBSD.

Use a callout for the delayed ACK timer, and delete tcp_fasttimo().
Expose the delayed ACK timer as net.inet.tcp.delack_ticks.

branches: 1.48.2; 1.48.4; 1.48.6;
remove #ifdef TCP6. it is not likely for us to bring in sys/netinet6/tcp6*.c
(separate TCP/IPv6 stack) into netbsd-current.

be more friendly with INET-less build.
XXX we need to do more to do a working INET-less build

Remove register declarations.

branches: 1.45.2;
Use proper ip protocol # field and tcp hdr on sending RST against SYN,
when ip header and tcp header are not adjacent to each other
(i.e. when ip6 options are attached).

To test this, try
telnet @::1@::1 port
toward a port without responding server. Prior to the fix, the kernel will
generate broken RST packet.

IPv6 kernel code, based on KAME/NetBSD 1.4, SNAP kit 19990628.
(Sorry for a big commit, I can't separate this into several pieces...)
Pls check sys/netinet6/TODO and sys/netinet6/IMPLEMENTATION for details.

- sys/kern: do not assume single mbuf, accept chained mbuf on passing
data from userland to kernel (or other way round).
- "midway" ATM card: ATM PVC pseudo device support, like those done in ALTQ
package (ftp://ftp.csl.sony.co.jp/pub/kjc/).
- sys/netinet/tcp*: IPv4/v6 dual stack tcp support.
- sys/netinet/{ip6,icmp6}.h, sys/net/pfkeyv2.h: IETF document assumes those
file to be there so we patch it up.
- sys/netinet: IPsec additions are here and there.
- sys/netinet6/*: most of IPv6 code sits here.
- sys/netkey: IPsec key management code
- dev/pci/pcidevs: regen

In my understanding no code here is subject to export control so it
should be safe.

branches: 1.43.8; 1.43.10;
Create tcp.keepidle, tcp.keepintvl, tcp.keepcnt, tcp.slowhz sysctls.

Fix a couple of bogons related to tcp_new_iss():
* Don't add tcp_iss_seq when creating a new ISS from TIME-WAIT state.
* Do the clock increment even when using the rnd device.

Comment where we use the Loss Window.

Loss window MUST be one segment, per draft-floyd-incr-init-win-03.

Make sure a timer is marked "disarmed" once it has expired.

Nuke TUBA per my note to tech-net; there's no reason to keep it around.

Define all TCP timers in terms of PRT timers.

Use macros from tcp_timer.h to manipulate TCP timers, so that their
implementation can be changed easily.

Remove bogus black hole discovery code

Make use of the work-arounds for ancient broken TCP peers run-time
conditional (tcp_compat_42). The kernel config option TCP_COMPAT_42
will still enable this by default, or disable this by default if the
option is not included (i.e. current behavior). This will be made a
sysctl soon.

Add support for deletion of routes added by path MTU discovery;
uses new generic route timeout code. Add sysctl for timeout period.

Fix a potential-congestion case in the larger initial congestion window
code, as clarified in the TCPIMPL WG meeting at IETF #41: If the SYN
(active open) or SYN,ACK (passive open) was retransmitted, the initial
congestion window for the first slow start of that connection must be
one segment.

Fix a retransmission bug introduced by the Brakmo and Peterson
RTO estimation changes. Under some circumstances it would return a value
of 0, while the old Van Jacobson RTO code would return a minimum of 3.
This would result in 12 retransmissions, each 1 second apart.
This takes care of those instances, and ensures that t_rttmin is
used everywhere as a lower bound.

Update copyright (sigh, should have done this long ago).

Use option header file for TCP_COMPAT_42

Finishing merging 4.4BSD-Lite2 netinet. At this point, the only changes
left were SCCS IDs and Copyright dates.

Implement a queue for delayed ACK processing. This queue is used in
tcp_fasttimo() in lieu of scanning all open TCP connections.

From 4.4BSD-Lite2:
- When running the slow timers, skip PCBs in LISTEN state.
- When processing the persist timer, drop the connection if the connection
idle time exceeds the maximum backoff for retransmit. Part of
kern/2335 (pete@daemon.net).

Implement an infrastructure to allow larger initial congestion windows.
The sysctl'able variable "tcp_init_win", when set to 0, selects an
auto-tuning algorithm for selecting the initial window, based on transmit
segment size, per discussion in the IETF tcpimpl working group.

Default initial window is still 1 segment, but will soon become 2 segments,
per discussion in tcpimpl.

In tcp_fasttimo(), don't clear TF_DELACK; we need it to count delayed ACKs
in tcp_output(), and it will only be cleared in tcp_output() if the ACK was
transmitted sucessfully. Also, don't count delayed ACKs here, let tcp_output()
count them.

Costmetic change: use intotcpcb() in tcp_fasttimo().

TCP MSS fixes to provide cleaner slow-start and recovery.

branches: 1.21.2;
o Make usage of /dev/random dependant on
pseudo-device rnd # /dev/random and in-kernel generator
in config files.

o Add declaration to all architectures.

o Clean up copyright message in rnd.c, rnd.h, and rndpool.c to include
that this code is derived in part from Ted Tyso's linux code.

Add hooks to use the kernel random system to generate TCP sequence numbers.

branches: 1.19.2;
Garbage-collect some "extern"s.

Pull SYN_cache_branch down into the main line.

branches: 1.17.8;
Fix RTT scaling problems introduced with Brakmo and Peterson changes.

Add in_nullhost() and in_hosteq() macros, to hide some protocol
details. Also, fix a bug in TCP wrt SYN+URG packets.

Pass a proc pointer down to the usrreq and pcbbind functions for PRU_ATTACH, PRU_BIND and
PRU_CONTROL. The usrreq interface really needs to be split up, but this will have to wait.
Remove SS_PRIV completely.

branches: 1.14.4;
netinet prototypes

splnet --> splsoftnet

convert pcb lists to CIRCLEQs, so that the end can be looked at more
easily, and so that the original (insque/remque) logic can be effectively
mimiced. (This fixes a bug in the previous set of list changes.)
also (since terminator is no longer null) reinstate uninitted list checks,
but mark them XXX.

Various cleanup, including:
* Convert several data structures to use queue.h.
* Split in_pcbnotify() into two parts; one for notifying a specific PCB, and
one for notifying all PCBs for a particular foreign address.

be a bit more careful and explicit with types. (basically a large no-op.)

Don't return received data to the user until the initial handshake is complete.
Also use TCPS_HAVEESTABLISHED() in a few other places.

New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'

Update to 4.4-Lite networking code, with a few local changes.

Prototypes.

Fix some inconsistent spacing; spaces at the end of lines, etc.

Canonicalize all #includes.

branches: 1.3.4;
add include of select.h if necessary for protos, or delete if extraneous

make kernel select interface be one-stop shopping & clean it all up.

branches: 1.1.1;
Initial revision

Remove now unused tcpip.h includes. Some were already unused before.

Remove #ifdef INET. Same as tcp_input.c. Makes the code easier to
understand.

Also make tcp6_mtudisc() static in tcp_subr.c.

branches: 1.93.2;
Run tcp_slowtimo in workqueue if NET_MPSAFE

If NET_MPSAFE is enabled, we have to avoid taking softnet_lock in softint as
much as possible to prevent any softint handlers including callout handlers
such as tcp_slowtimo from sticking on softnet_lock because it results in
undesired delays of executing subsequent softint handlers.

NFCI for !NET_MPSAFE

Remove TCP_COMPAT_42. This feature is a workaround for a bug in the TCP
stack of BSD4.2. Having such features just does not make any sense, and
looking at the code, I'm not sure it actually works.

branches: 1.91.8;
fix: unlock in reverse order

branches: 1.90.2;
Sweep unnecessary route.h inclusions

sprinkle _KERNEL_OPT

branches: 1.88.2;
Do not uselessly include <sys/malloc.h>.

branches: 1.87.4;
Allow kernels compiled with INET+INET6 to be booted as IPv4-only or IPv6-only.

branches: 1.86.2; 1.86.12; 1.86.16;
NULL does not need a cast

Rewrite comments about TCP RTO calculations.

Long ago, the storage representations of srtt and rttvar were changed
from the 4.4BSD scheme, and the comments are out of sync with the
code. This commit rewrites most of the comments that explain the RTO
calculations, and points out some issues in the code.

Joint work with Bev Schwartz of BBN (original analysis and comments),
but I have rewritten and extended them, so errors are mine.

This material is based upon work supported by the Defense Advanced
Research Projects Agency and Space and Naval Warfare Systems Center,
Pacific, under Contract No. N66001-09-C-2073. Approved for Public
Release, Distribution Unlimited

branches: 1.84.8; 1.84.10;
Whitespace.

Fix kern/39769: race condition in TCP timers
When a TCP timer is disarmed (with callout_stop()) in the general case
callout_invoking() isn't checked, so the timer handler could be called run
when the current interrupt handler exits, athough the timer is disarmed.
This case cause bad things like TCPT_REXMT and TCPT_PERSIST being both pending,
causing a panic (see the PR for details).
Close the issue by aborting the handler if the timer is not callout_expired().
(the EXPIRED flag being cleared by callout_stop()).

branches: 1.82.2; 1.82.4;
tcp_delack: test for TF_DELACK.

branches: 1.81.2; 1.81.6;
Remove clause 3 and 4 from TNF licenses

branches: 1.80.2;
Merge the socket locking patch:

- Socket layer becomes MP safe.
- Unix protocols become MP safe.
- Allows protocol processing interrupts to safely block on locks.
- Fixes a number of race conditions.

With much feedback from matt@ and plunky@.

branches: 1.79.2;
Make IP, TCP, UDP, and ICMP statistics per-CPU. The stats are collated
when the user requests them via sysctl.

Change TCP stats from a structure to an array of uint64_t's.

Note: This is ABI-compatible with the old tcpstat structure; old netstat
binaries will continue to work properly.

branches: 1.77.28;
- per socket keepalive settings
- settable connection establishment timeout

branches: 1.76.2; 1.76.8; 1.76.10;
Modular (I tried ;-) TCP congestion control API. Whenever certain conditions
happen in the TCP stack, this interface calls the specified callback to
handle the situation according to the currently selected congestion
control algorithm.
A new sysctl node was created: net.inet.tcp.congctl.{available,selected}
with obvious meanings.
The old net.inet.tcp.newreno MIB was removed.
The API is discussed in tcp_congctl(9).

In the near future, it will be possible to selected a congestion control
algorithm on a per-socket basis.

Discussed on tech-net and reviewed by <yamt>.

branches: 1.75.8; 1.75.10;
integrate kauth.

Coverity CID 1153: Add KASSERT before deref.

branches: 1.73.4; 1.73.6; 1.73.8; 1.73.10; 1.73.12;
merge ktrace-lwp.

Implement PMTU checks from:

http://www.gont.com.ar/drafts/icmp-attacks-against-tcp.html

1. Don't act on ICMP-need-frag immediately if adhoc checks on the
advertised MTU fail. The MTU update is delayed until a TCP retransmit
happens.
2. Ignore ICMP Source Quench messages meant for TCP connections.

From OpenBSD.

branches: 1.71.2; 1.71.4;
Copyright maintenance.

Commit TCP SACK patches from Kentaro A. Karahone's patch at:
http://www.sigusr1.org/~kurahone/tcp-sack-netbsd-02152005.diff.gz

Fixes in that patch for pre-existing TCP pcb initializations were already
committed to NetBSD-current, so are not included in this commit.

The SACK patch has been observed to correctly negotiate and respond,
to SACKs in wide-area traffic.

There are two indepenently-observed, as-yet-unresolved anomalies:
First, seeing unexplained delays between in fast retransmission
(potentially explainable by an 0.2sec RTT between adjacent
ethernet/wifi NICs); and second, peculiar and unepxlained TCP
retransmits observed over an ath0 card.

After discussion with several interested developers, I'm committing
this now, as-is, for more eyes to use and look over. Current hypothesis
is that the anomalies above may in fact be due to link/level (hardware,
driver, HAL, firmware) abberations in the test setup, affecting both
Kentaro's wired-Ethernet NIC and in my two (different) WiFi NICs.

ANSIfy function declarations

Whoops. Exit fast recovery when handling a timeout.

Fix two problems in our TCP stack:

1) If an echoed RFC 1323 time stamp appears to be later than the current time,
ignore it and fall back to old-style RTT calculation. This prevents ending
up with a negative RTT and panicking later.

2) Fix NewReno. This involves a few changes:

a) Implement the send_high variable in RFC 2582. Our implementation is
subtly different; it is one *past* the last sequence number transmitted
rather than being equal to it. This simplifies some logic and makes
the code smaller. Additional logic was required to prevent sequence
number wraparound problems; this is not mentioned in RFC 2582.

b) Make sure we reset t_dupacks on new acks, but *not* on a partial ack.
All of the new ack code is pushed out into tcp_newreno(). (Later this
will probably be a pluggable function.) Thus t_dupacks keeps track of
whether we're in fast recovery all the time, with Reno or NewReno, which
keeps some logic simpler.

c) We do not need to update snd_recover when we're not in fast recovery.
See tech-net for an explanation of this.

d) In the gratuitous fast retransmit prevention case, do not send a packet.
RFC 2582 specifically says that we should "do nothing".

e) Do not inflate the congestion window on a partial ack. (This is done by
testing t_dupacks to see whether we're still in fast recovery.)

This brings the performance of NewReno back up to the same as Reno in a few
random test cases (e.g. transferring peer-to-peer over my wireless network).
I have not concocted a good test case for the behavior specific to NewReno.

branches: 1.66.8; 1.66.10;
some corrections from markus@openbsd;
- callout_ack() was called with wrong argument

make it compilable with TCP_DEBUG defined

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22364, verified by myself.

As a temporary workaround, apply the fix from PR#20390, thereby
cooperating with the callout code in working around the race
condition caused by the TCP code's use of the callout facility.

Instead of unconditionally releasing memory in tcp_close() and
SYN_CACHE_PUT(), check whether any of the related callout handlers
are about to be invoked (but have not yet done callout_ack()), and
if so, just mark the associated data structure (tcpcb or syn cache
entry) as "dead", and test for this (and release storage) in the
callout handler functions.

branches: 1.62.2;
Test callout_pending(), not callout_active(), and eliminate now-unnecessary
callout_deactivate() calls.

Quell an uninitialised variable warning.

Guard use of "so" in tcp_timer_persist() and tcp_timer_2msl() with
#ifdef TCP_DEBUG.

whitespace

path MTU discovery blackhole detection.
PR 12790 (sorry for not committing it for a long time)

branches: 1.57.8; 1.57.10;
add RCSIDs

Change a few variable/tables to const since they are read-only.

branches: 1.55.2;
Use callouts for SYN cache timers, rather than traversing time queues
in tcp_slowtimo().

Update copyrights.

Use callouts for TCP timers, rather than traversing the list of
all open TCP connections in tcp_slowtimo() (which is called 2x
per second). It's fairly rare for TCP timers to actually fire,
so saving this list traversal is good, especially if you want
to scale to thousands of open connections.

Initialize TCP timer variables in a new function, tcp_timer_init().

Split tcp_timers() into multiple functions, one for each timer,
and call it directly from tcp_slowtimo() (via a table) rather
than going through tcp_userreq().

This will allow us to call TCP timers directly from callouts,
in a future revision.

Change the way receive idle time and round trip time are measured.
Instead of incrementing t_idle and t_rtt in tcp_slowtimo(), we now
take a timstamp (via tcp_now) and use subtraction to compute the
delta when we actually need it (using unsigned arithmetic so that
tcp_now wrapping is handled correctly).

Based on similar changes in FreeBSD.

Use a callout for the delayed ACK timer, and delete tcp_fasttimo().
Expose the delayed ACK timer as net.inet.tcp.delack_ticks.

branches: 1.48.2; 1.48.4; 1.48.6;
remove #ifdef TCP6. it is not likely for us to bring in sys/netinet6/tcp6*.c
(separate TCP/IPv6 stack) into netbsd-current.

be more friendly with INET-less build.
XXX we need to do more to do a working INET-less build

Remove register declarations.

branches: 1.45.2;
Use proper ip protocol # field and tcp hdr on sending RST against SYN,
when ip header and tcp header are not adjacent to each other
(i.e. when ip6 options are attached).

To test this, try
telnet @::1@::1 port
toward a port without responding server. Prior to the fix, the kernel will
generate broken RST packet.

IPv6 kernel code, based on KAME/NetBSD 1.4, SNAP kit 19990628.
(Sorry for a big commit, I can't separate this into several pieces...)
Pls check sys/netinet6/TODO and sys/netinet6/IMPLEMENTATION for details.

- sys/kern: do not assume single mbuf, accept chained mbuf on passing
data from userland to kernel (or other way round).
- "midway" ATM card: ATM PVC pseudo device support, like those done in ALTQ
package (ftp://ftp.csl.sony.co.jp/pub/kjc/).
- sys/netinet/tcp*: IPv4/v6 dual stack tcp support.
- sys/netinet/{ip6,icmp6}.h, sys/net/pfkeyv2.h: IETF document assumes those
file to be there so we patch it up.
- sys/netinet: IPsec additions are here and there.
- sys/netinet6/*: most of IPv6 code sits here.
- sys/netkey: IPsec key management code
- dev/pci/pcidevs: regen

In my understanding no code here is subject to export control so it
should be safe.

branches: 1.43.8; 1.43.10;
Create tcp.keepidle, tcp.keepintvl, tcp.keepcnt, tcp.slowhz sysctls.

Fix a couple of bogons related to tcp_new_iss():
* Don't add tcp_iss_seq when creating a new ISS from TIME-WAIT state.
* Do the clock increment even when using the rnd device.

Comment where we use the Loss Window.

Loss window MUST be one segment, per draft-floyd-incr-init-win-03.

Make sure a timer is marked "disarmed" once it has expired.

Nuke TUBA per my note to tech-net; there's no reason to keep it around.

Define all TCP timers in terms of PRT timers.

Use macros from tcp_timer.h to manipulate TCP timers, so that their
implementation can be changed easily.

Remove bogus black hole discovery code

Make use of the work-arounds for ancient broken TCP peers run-time
conditional (tcp_compat_42). The kernel config option TCP_COMPAT_42
will still enable this by default, or disable this by default if the
option is not included (i.e. current behavior). This will be made a
sysctl soon.

Add support for deletion of routes added by path MTU discovery;
uses new generic route timeout code. Add sysctl for timeout period.

Fix a potential-congestion case in the larger initial congestion window
code, as clarified in the TCPIMPL WG meeting at IETF #41: If the SYN
(active open) or SYN,ACK (passive open) was retransmitted, the initial
congestion window for the first slow start of that connection must be
one segment.

Fix a retransmission bug introduced by the Brakmo and Peterson
RTO estimation changes. Under some circumstances it would return a value
of 0, while the old Van Jacobson RTO code would return a minimum of 3.
This would result in 12 retransmissions, each 1 second apart.
This takes care of those instances, and ensures that t_rttmin is
used everywhere as a lower bound.

Update copyright (sigh, should have done this long ago).

Use option header file for TCP_COMPAT_42

Finishing merging 4.4BSD-Lite2 netinet. At this point, the only changes
left were SCCS IDs and Copyright dates.

Implement a queue for delayed ACK processing. This queue is used in
tcp_fasttimo() in lieu of scanning all open TCP connections.

From 4.4BSD-Lite2:
- When running the slow timers, skip PCBs in LISTEN state.
- When processing the persist timer, drop the connection if the connection
idle time exceeds the maximum backoff for retransmit. Part of
kern/2335 (pete@daemon.net).

Implement an infrastructure to allow larger initial congestion windows.
The sysctl'able variable "tcp_init_win", when set to 0, selects an
auto-tuning algorithm for selecting the initial window, based on transmit
segment size, per discussion in the IETF tcpimpl working group.

Default initial window is still 1 segment, but will soon become 2 segments,
per discussion in tcpimpl.

In tcp_fasttimo(), don't clear TF_DELACK; we need it to count delayed ACKs
in tcp_output(), and it will only be cleared in tcp_output() if the ACK was
transmitted sucessfully. Also, don't count delayed ACKs here, let tcp_output()
count them.

Costmetic change: use intotcpcb() in tcp_fasttimo().

TCP MSS fixes to provide cleaner slow-start and recovery.

branches: 1.21.2;
o Make usage of /dev/random dependant on
pseudo-device rnd # /dev/random and in-kernel generator
in config files.

o Add declaration to all architectures.

o Clean up copyright message in rnd.c, rnd.h, and rndpool.c to include
that this code is derived in part from Ted Tyso's linux code.

Add hooks to use the kernel random system to generate TCP sequence numbers.

branches: 1.19.2;
Garbage-collect some "extern"s.

Pull SYN_cache_branch down into the main line.

branches: 1.17.8;
Fix RTT scaling problems introduced with Brakmo and Peterson changes.

Add in_nullhost() and in_hosteq() macros, to hide some protocol
details. Also, fix a bug in TCP wrt SYN+URG packets.

Pass a proc pointer down to the usrreq and pcbbind functions for PRU_ATTACH, PRU_BIND and
PRU_CONTROL. The usrreq interface really needs to be split up, but this will have to wait.
Remove SS_PRIV completely.

branches: 1.14.4;
netinet prototypes

splnet --> splsoftnet

convert pcb lists to CIRCLEQs, so that the end can be looked at more
easily, and so that the original (insque/remque) logic can be effectively
mimiced. (This fixes a bug in the previous set of list changes.)
also (since terminator is no longer null) reinstate uninitted list checks,
but mark them XXX.

Various cleanup, including:
* Convert several data structures to use queue.h.
* Split in_pcbnotify() into two parts; one for notifying a specific PCB, and
one for notifying all PCBs for a particular foreign address.

be a bit more careful and explicit with types. (basically a large no-op.)

Don't return received data to the user until the initial handshake is complete.
Also use TCPS_HAVEESTABLISHED() in a few other places.

New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'

Update to 4.4-Lite networking code, with a few local changes.

Prototypes.

Fix some inconsistent spacing; spaces at the end of lines, etc.

Canonicalize all #includes.

branches: 1.3.4;
add include of select.h if necessary for protos, or delete if extraneous

make kernel select interface be one-stop shopping & clean it all up.

branches: 1.1.1;
Initial revision

Run tcp_slowtimo in workqueue if NET_MPSAFE

If NET_MPSAFE is enabled, we have to avoid taking softnet_lock in softint as
much as possible to prevent any softint handlers including callout handlers
such as tcp_slowtimo from sticking on softnet_lock because it results in
undesired delays of executing subsequent softint handlers.

NFCI for !NET_MPSAFE

Remove TCP_COMPAT_42. This feature is a workaround for a bug in the TCP
stack of BSD4.2. Having such features just does not make any sense, and
looking at the code, I'm not sure it actually works.

fix: unlock in reverse order

branches: 1.90.2;
Sweep unnecessary route.h inclusions

sprinkle _KERNEL_OPT

branches: 1.88.2;
Do not uselessly include <sys/malloc.h>.

branches: 1.87.4;
Allow kernels compiled with INET+INET6 to be booted as IPv4-only or IPv6-only.

branches: 1.86.2; 1.86.12; 1.86.16;
NULL does not need a cast

Rewrite comments about TCP RTO calculations.

Long ago, the storage representations of srtt and rttvar were changed
from the 4.4BSD scheme, and the comments are out of sync with the
code. This commit rewrites most of the comments that explain the RTO
calculations, and points out some issues in the code.

Joint work with Bev Schwartz of BBN (original analysis and comments),
but I have rewritten and extended them, so errors are mine.

This material is based upon work supported by the Defense Advanced
Research Projects Agency and Space and Naval Warfare Systems Center,
Pacific, under Contract No. N66001-09-C-2073. Approved for Public
Release, Distribution Unlimited

branches: 1.84.8; 1.84.10;
Whitespace.

Fix kern/39769: race condition in TCP timers
When a TCP timer is disarmed (with callout_stop()) in the general case
callout_invoking() isn't checked, so the timer handler could be called run
when the current interrupt handler exits, athough the timer is disarmed.
This case cause bad things like TCPT_REXMT and TCPT_PERSIST being both pending,
causing a panic (see the PR for details).
Close the issue by aborting the handler if the timer is not callout_expired().
(the EXPIRED flag being cleared by callout_stop()).

branches: 1.82.2; 1.82.4;
tcp_delack: test for TF_DELACK.

branches: 1.81.2; 1.81.6;
Remove clause 3 and 4 from TNF licenses

branches: 1.80.2;
Merge the socket locking patch:

- Socket layer becomes MP safe.
- Unix protocols become MP safe.
- Allows protocol processing interrupts to safely block on locks.
- Fixes a number of race conditions.

With much feedback from matt@ and plunky@.

branches: 1.79.2;
Make IP, TCP, UDP, and ICMP statistics per-CPU. The stats are collated
when the user requests them via sysctl.

Change TCP stats from a structure to an array of uint64_t's.

Note: This is ABI-compatible with the old tcpstat structure; old netstat
binaries will continue to work properly.

branches: 1.77.28;
- per socket keepalive settings
- settable connection establishment timeout