Commit Graph

51 Commits

Author SHA1 Message Date
Andre Oppermann
97d8d152c2 Introduce tcp_hostcache and remove the tcp specific metrics from
the routing table.  Move all usage and references in the tcp stack
from the routing table metrics to the tcp hostcache.

It caches measured parameters of past tcp sessions to provide better
initial start values for following connections from or to the same
source or destination.  Depending on the network parameters to/from
the remote host this can lead to significant speedups for new tcp
connections after the first one because they inherit and shortcut
the learning curve.

tcp_hostcache is designed for multiple concurrent access in SMP
environments with high contention and is hash indexed by remote
ip address.

It removes significant locking requirements from the tcp stack with
regard to the routing table.

Reviewed by:	sam (mentor), bms
Reviewed by:	-net, -current, core@kame.net (IPv6 parts)
Approved by:	re (scottl)
2003-11-20 20:07:39 +00:00
Andre Oppermann
26d02ca7ba Remove RTF_PRCLONING from routing table and adjust users of it
accordingly.  The define is left intact for ABI compatibility
with userland.

This is a pre-step for the introduction of tcp_hostcache.  The
network stack remains fully useable with this change.

Reviewed by:	sam (mentor), bms
Reviewed by:	-net, -current, core@kame.net (IPv6 parts)
Approved by:	re (scottl)
2003-11-20 19:47:31 +00:00
Andre Oppermann
02c1c7070e Remove the global one-level rtcache variable and associated
complex locking and rework ip_rtaddr() to do its own rtlookup.
Adopt all its callers to this and make ip_output() callable
with NULL rt pointer.

Reviewed by:	sam (mentor)
2003-11-14 21:48:57 +00:00
Sam Leffler
04df2fbbb8 Remove bogus RTFREE that was added in rev 1.47. The rmx code operates
directly on the radix tree and does not hold any routing table refernces.
This fixes the reference counting problems that manifested itself as a
panic during unmount of filesystems that were mounted by NFS over an
interface that had been removed.

Supported by:	FreeBSD Foundation
2003-11-03 06:11:44 +00:00
Sam Leffler
9c63e9dbd7 Overhaul routing table entry cleanup by introducing a new rtexpunge
routine that takes a locked routing table reference and removes all
references to the entry in the various data structures. This
eliminates instances of recursive locking and also closes races
where the lock on the entry had to be dropped prior to calling
rtrequest(RTM_DELETE).  This also cleans up confusion where the
caller held a reference to an entry that might have been reclaimed
(and in some cases used that reference).

Supported by:	FreeBSD Foundation
2003-10-30 23:02:51 +00:00
Sam Leffler
929b31ddab Lock ip forwarding route cache. While we're at it, remove the global
variable ipforward_rt by introducing an ip_forward_cacheinval() call
to use to invalidate the cache.

Supported by:	FreeBSD Foundation
2003-10-14 19:19:12 +00:00
Sam Leffler
d1dd20be6e Locking for updates to routing table entries. Each rtentry gets a mutex
that covers updates to the contents.  Note this is separate from holding
a reference and/or locking the routing table itself.

Other/related changes:

o rtredirect loses the final parameter by which an rtentry reference
  may be returned; this was never used and added unwarranted complexity
  for locking.
o minor style cleanups to routing code (e.g. ansi-fy function decls)
o remove the logic to bump the refcnt on the parent of cloned routes,
  we assume the parent will remain as long as the clone; doing this avoids
  a circularity in locking during delete
o convert some timeouts to MPSAFE callouts

Notes:

1. rt_mtx in struct rtentry is guarded by #ifdef _KERNEL as user-level
   applications cannot/do-no know about mutex's.  Doing this requires
   that the mutex be the last element in the structure.  A better solution
   is to introduce an externalized version of struct rtentry but this is
   a major task because of the intertwining of rtentry and other data
   structures that are visible to user applications.
2. There are known LOR's that are expected to go away with forthcoming
   work to eliminate many held references.  If not these will be resolved
   prior to release.
3. ATM changes are untested.

Sponsored by:	FreeBSD Foundation
Obtained from:	BSD/OS (partly)
2003-10-04 03:44:50 +00:00
Jeffrey Hsu
24652ff6e1 Get cosmetic changes out of the way before I add routing table SMP locks. 2003-02-10 22:01:34 +00:00
Jeffrey Hsu
956b0b653c SMP locking for radix nodes. 2002-12-24 03:03:39 +00:00
Alfred Perlstein
4d77a549fe Remove __P. 2002-03-19 21:25:46 +00:00
Jonathan Lemon
9a10980e2a Centralize satosin(), sintosa() and ifatoia() macros in <netinet/in.h>
Remove local definitions.
2001-09-29 03:23:44 +00:00
Ruslan Ermilov
9185426827 In in_ifadown(), differentiate between whether the interface goes
down or interface address is deleted.  Only delete static routes
in the latter case.

Reported by:	Alexander Leidinger <Alexander@leidinger.net>
2001-05-11 14:37:34 +00:00
Ruslan Ermilov
1e3d5af041 Invalidate cached forwarding route (ipforward_rt) whenever a new route
is added to the routing table, otherwise we may end up using the wrong
route when forwarding.

PR:		kern/10778
Reviewed by:	silence on -net
2001-03-19 09:16:16 +00:00
Ruslan Ermilov
089cdfad78 net/route.c:
A route generated from an RTF_CLONING route had the RTF_WASCLONED flag
  set but did not have a reference to the parent route, as documented in
  the rtentry(9) manpage.  This prevented such routes from being deleted
  when their parent route is deleted.

  Now, for example, if you delete an IP address from a network interface,
  all ARP entries that were cloned from this interface route are flushed.

  This also has an impact on netstat(1) output.  Previously, dynamically
  created ARP cache entries (RTF_STATIC flag is unset) were displayed as
  part of the routing table display (-r).  Now, they are only printed if
  the -a option is given.

netinet/in.c, netinet/in_rmx.c:

  When address is removed from an interface, also delete all routes that
  point to this interface and address.  Previously, for example, if you
  changed the address on an interface, outgoing IP datagrams might still
  use the old address.  The only solution was to delete and re-add some
  routes.  (The problem is easily observed with the route(8) command.)

  Note, that if the socket was already bound to the local address before
  this address is removed, new datagrams generated from this socket will
  still be sent from the old address.

PR:		kern/20785, kern/21914
Reviewed by:	wollman (the idea)
2001-03-15 14:52:12 +00:00
Peter Wemm
c3aac50f28 $Id$ -> $FreeBSD$ 1999-08-28 01:08:13 +00:00
Bill Fumerola
3d177f465a Add sysctl descriptions to many SYSCTL_XXXs
PR:		kern/11197
Submitted by:	Adrian Chadd <adrian@FreeBSD.org>
Reviewed by:	billf(spelling/style/minor nits)
Looked at by:	bde(style)
1999-05-03 23:57:32 +00:00
Bruce Evans
db176bbac6 Converted the last instance of hzto() to tvtohz(). 1998-08-05 16:59:20 +00:00
Poul-Henning Kamp
227ee8a188 Eradicate the variable "time" from the kernel, using various measures.
"time" wasn't a atomic variable, so splfoo() protection were needed
around any access to it, unless you just wanted the seconds part.

Most uses of time.tv_sec now uses the new variable time_second instead.

gettime() changed to getmicrotime(0.

Remove a couple of unneeded splfoo() protections, the new getmicrotime()
is atomic, (until Bruce sets a breakpoint in it).

A couple of places needed random data, so use read_random() instead
of mucking about with time which isn't random.

Add a new nfs_curusec() function.

Mark a couple of bogosities involving the now disappeard time variable.

Update ffs_update() to avoid the weird "== &time" checks, by fixing the
one remaining call that passwd &time as args.

Change profiling in ncr.c to use ticks instead of time.  Resolution is
the same.

Add new function "tvtohz()" to avoid the bogus "splfoo(), add time, call
hzto() which subtracts time" sequences.

Reviewed by:	bde
1998-03-30 09:56:58 +00:00
Peter Wemm
5c3a5f4a20 A fix for a link down route cleanup panic, when the route cleanup
pulls the rug out from underneath itself.

Obtained from: wollman (a few months ago, I've been using this for ages)
1998-03-27 14:30:18 +00:00
Eivind Eklund
0b08f5f737 Back out DIAGNOSTIC changes. 1998-02-06 12:14:30 +00:00
Eivind Eklund
47cfdb166d Turn DIAGNOSTIC into a new-style option. 1998-02-04 22:34:03 +00:00
Bruce Evans
1fd0b0588f Removed unused #includes. 1997-08-02 14:33:27 +00:00
Garrett Wollman
d0390e0570 Fix the mechanism for choosing wehether to save the slow-start threshold
in the route.  This allows us to remove the unconditional setting of the
pipesize in the route, which should mean that SO_SNDBUF and SO_RCVBUF
should actually work again.  While we're at it:

- Convert udp_usrreq from `mondo switch statement from Hell' to new-style.
- Delete old TCP mondo switch statement from Hell, which had previously
  been diked out.
1997-02-14 18:15:53 +00:00
Garrett Wollman
39191c8eb8 Provide PRC_IFDOWN and PRC_IFUP support for IP. Now, when an interface
is administratively downed, all routes to that interface (including the
interface route itself) which are not static will be deleted.  When
it comes back up, and addresses remaining will have their interface routes
re-added.  This solves the problem where, for example, an Ethernet interface
is downed by traffic continues to flow by way of ARP entries.
1997-02-13 19:46:45 +00:00
Jordan K. Hubbard
1130b656e5 Make the long-awaited change from $Id$ to $FreeBSD$
This will make a number of things easier in the future, as well as (finally!)
avoiding the Id-smashing problem which has plagued developers for so long.

Boy, I'm glad we're not using sup anymore.  This update would have been
insane otherwise.
1997-01-14 07:20:47 +00:00
Nate Williams
17040b78cd Put the 'debug' messages of the type:
/kernel: in_rtqtimo: adjusted rtq_reallyold to 1066
  /kernel: in_rtqtimo: adjusted rtq_reallyold to 710
inside of #ifdef DIAGNOSTIC to avoid the support questions from folks
asking what this means.
1996-06-20 15:41:23 +00:00
Garrett Wollman
9f9b3dc4ae Add three new route flags to help determine what sort of address
the destination represents.  For IP:

- Iff it is a host route, RTF_LOCAL and RTF_BROADCAST indicate local
  (belongs to this host) and broadcast addresses, respectively.

- For all routes, RTF_MULTICAST is set if the destination is multicast.

The RTF_BROADCAST flag is used by ip_output() to eliminate a call to
in_broadcast() in a common case; this gives about 1% in our packet-generation
experiments.  All three flags might be used (although they aren't now)
to determine whether a packet can be forwarded; a given host route can
represent a forwardable address if:

	(rt->rt_flags & (RTF_HOST | RTF_LOCAL | RTF_BROADCAST | RTF_MULTICAST))
	== RTF_HOST

Obviously, one still has to do all the work if a host route is not present,
but this code allows one to cache the results of such a lookup if rtalloc1()
is called without masking RTF_PRCLONING.
1996-05-06 17:42:13 +00:00
Garrett Wollman
93902408b8 Delete #if 0 block containing remnants of pre-MTU discovery rmx_mtu
initialization.
1996-04-26 18:31:41 +00:00
Bill Fenner
4dde4205e1 First piece of fixing ppp/proxy arp problem:
If an attempt to add a route fails because an "ARP table" entry is in
the way, remove the ARP entry and retry the add.

Reviewed by:	nate
1996-01-23 05:15:30 +00:00
Garrett Wollman
ce29ab3ac6 Actually call in_rtqdrain()as was originally intended. 1995-12-19 20:46:15 +00:00
Garrett Wollman
b7a44e3486 Path MTU Discovery is now standard. 1995-12-05 17:46:50 +00:00
Bruce Evans
ce7609a49c Completed function declarations and/or added prototypes. 1995-12-02 19:38:06 +00:00
Poul-Henning Kamp
0312fbe97d New style sysctl & staticize alot of stuff. 1995-11-14 20:34:56 +00:00
Poul-Henning Kamp
98163b98dc Start adding new style sysctl here too. 1995-11-09 20:23:09 +00:00
Poul-Henning Kamp
a98ca4699e Second batch of cleanup changes.
This time mostly making a lot of things static and some unused
variables here and there.
1995-10-29 15:33:36 +00:00
Garrett Wollman
5cbf3e086c Initial back-end support for IP MTU discovery, gated on MTUDISC. The support
for TCP has yet to be written.
1995-09-18 15:51:40 +00:00
Garrett Wollman
dd22498271 tcp_input.c - keep track of how many times a route contained a cached rtt
or ssthresh that we were able to use

tcp_var.h - declare tcpstat entries for above; declare tcp_{send,recv}space

in_rmx.c - fill in the MTU and pipe sizes with the defaults TCP would have
	used anyway in the absence of values here
1995-07-10 15:39:16 +00:00
Garrett Wollman
43bed81d64 Fix an error in the comparison direction of the ap->updating case of
in_rtqkill().

Submitted by: W. Richard Stevens
1995-06-21 19:48:53 +00:00
Rodney W. Grimes
9b2e535452 Remove trailing whitespace. 1995-05-30 08:16:23 +00:00
David Greenman
edf8a81561 Removed redundant newlines that were in some panic strings. 1995-03-19 14:29:26 +00:00
David Greenman
38aa9fc375 Added missing newlines to calls to log(). 1995-02-20 15:48:46 +00:00
Garrett Wollman
628e3e3401 After dynamically reducing rtq_reallyold, have in_rtqkill() reduce the
expiration timer of anything which would expire later than that.  (There
should be a way to call this from ip_sysctl() as well, but there currently
isn't.)
1995-02-14 23:11:26 +00:00
Garrett Wollman
ea80aed1a9 Attempt to make the host route cache a bit smarter under conditions of
high load:

	1) If there ever get to be more than net.inet.ip.rtmaxcache entries
	   in the cache, in_rtqtimo() will reduce net.inet.ip.rtexpire by
	   1/3 and do another round, unles net.inet.ip.rtexpire is less than
	   net.inet.ip.rtminexpire, and never more than once in ten minutes
	   (rtq_timeout).

	2) If net.inet.ip.rtexpire is set to zero, don't bother to cache
	   anything.
1995-02-14 23:04:52 +00:00
Garrett Wollman
bbb7c7109b Change caching strategy somewhat:
1) Don't clone routes to multicast destinations; there is nothing useful
   to be gained in this case.
2) Reduce default expiration timer to one hour.  Busy sites will still
   likely want to reduce this, but for ordinary users this is a reasonable
   value to use.
1995-01-23 02:02:50 +00:00
Garrett Wollman
f811ab0233 Avoid a serious race by blocking netisrs while walking the route tree.
(IWBRNI we could just block IP netisrs...)
1994-12-21 17:25:52 +00:00
Garrett Wollman
31246bc25f Update calls to rtalloc1(). Also merge rt_prflags with rt_flags. 1994-12-13 22:32:45 +00:00
Garrett Wollman
114a506d1a Delete old, confusing comment. 1994-12-02 23:10:32 +00:00
Garrett Wollman
73579c400f Add a check to make sure that we don't fiddle with the NFS routing tables
as well (bleah!).  Also, increase the interval to the real-life value and
eliminate debugging printfs.  This will be standard once tested by others.
1994-12-02 03:32:24 +00:00
Garrett Wollman
fa1f0e90fb Add latest version of ``advanced route metric management'' :-)
As before, this is currently conditionalized on options IN_RMX until
I'm sure it's working.
1994-12-01 23:19:48 +00:00
Garrett Wollman
bb0135e920 Completely replace JTW's idea with my (incompletely implemented) original
idea.  This is les likely to crash your machine.  As before, this code is only
enabled under `options IN_RMX'.
1994-11-03 01:05:34 +00:00