Commit Graph

2551 Commits

Author SHA1 Message Date
Yaroslav Tykhiy
4b97d7affd There is a consensus that ifaddr.ifa_addr should never be NULL,
except in places dealing with ifaddr creation or destruction; and
in such special places incomplete ifaddrs should never be linked
to system-wide data structures.  Therefore we can eliminate all the
superfluous checks for "ifa->ifa_addr != NULL" and get ready
to the system crashing honestly instead of masking possible bugs.

Suggested by:	glebius, jhb, ru
2006-06-29 19:22:05 +00:00
Yaroslav Tykhiy
ad67537233 Use TAILQ_FOREACH consistently. 2006-06-29 17:09:47 +00:00
Gleb Smirnoff
4d09f5a030 Fix URL to Bellovin's paper.
Submitted by:	Anton Yuzhaninov <citrin rambler-co.ru>
2006-06-29 13:38:36 +00:00
Bjoern A. Zeeb
333ad3bc40 Eliminate the offset argument from send_reject. It's not been
used since FreeBSD-SA-06:04.ipfw.
Adopt send_reject6 to what had been done for legacy IP: no longer
send or permit sending rejects for any but the first fragment.

Discussed with: oleg, csjp (some weeks ago)
2006-06-29 11:17:16 +00:00
Bjoern A. Zeeb
421d8aa603 Use INPLOOKUP_WILDCARD instead of just 1 more consistently.
OKed by: rwatson (some weeks ago)
2006-06-29 10:49:49 +00:00
Pawel Jakub Dawidek
835d4b8924 - Use suser_cred(9) instead of directly checking cr_uid.
- Change the order of conditions to first verify that we actually need
  to check for privileges and then eventually check them.

Reviewed by:	rwatson
2006-06-27 11:35:53 +00:00
Andre Oppermann
cc477a6347 In syncache_respond() do not reply with a MSS that is larger than what
the peer announced to us but make it at least tcp_minmss in size.

Sponsored by:	TCP/IP Optimization Fundraise 2005
2006-06-26 17:54:53 +00:00
Andre Oppermann
8bfb19180d Some cleanups and janitorial work to tcp_syncache:
o don't assign remote/local host/port information manually between provided
   struct in_conninfo and struct syncache, bcopy() it instead
 o rename sc_tsrecent to sc_tsreflect in struct syncache to better capture
   the purpose of this field
 o rename sc_request_r_scale to sc_requested_r_scale for ditto reasons
 o fix IPSEC error case printf's to report correct function name
 o in syncache_socket() only transpose enhanced tcp options parameters to
   struct tcpcb when the inpcb doesn't has TF_NOOPT set
 o in syncache_respond() reorder stack variables
 o in syncache_respond() remove bogus KASSERT()

No functional changes.

Sponsored by:	TCP/IP Optimization Fundraise 2005
2006-06-26 16:14:19 +00:00
Andre Oppermann
f72167f4d1 Some cleanups and janitorial work to tcp_dooptions():
o redefine the parameter 'is_syn' to 'flags', add TO_SYN flag and adjust its
   usage accordingly
 o update the comments to the tcp_dooptions() invocation in
   tcp_input():after_listen to reflect reality
 o move the logic checking the echoed timestamp out of tcp_dooptions() to the
   only place that uses it next to the invocation described in the previous
   item
 o adjust parsing of TCPOPT_SACK_PERMITTED to use the same style as the others
 o add comments in to struct tcpopt.to_flags #defines

No functional changes.

Sponsored by:	TCP/IP Optimization Fundraise 2005
2006-06-26 15:35:25 +00:00
Andre Oppermann
dfabcc1d29 Reverse the source/destination parameters to in[6]_pcblookup_hash() in
syncache_respond() for the #ifdef MAC case.

Submitted by:	Tai-hwa Liang <avatar-at-mmlab.cse.yzu.edu.tw>
2006-06-26 09:43:55 +00:00
Robert Watson
b4470c1639 In tcp6_usr_attach(), return immediately if SS_ISDISCONNECTED, to
avoid dereferencing an uninitialized inp variable.

Submitted by:	Michiel Boland <michiel at boland dot org>
MFC after:	1 month
2006-06-26 09:38:08 +00:00
Andre Oppermann
a846263567 Decrement the global syncache counter in syncache_expand() when the entry
is removed from the bucket.  This fixes the syncache statistics.
2006-06-25 11:11:33 +00:00
Andre Oppermann
649ac0ce5f Move the syncookie MD5 context from globals to the stack to make it MP safe. 2006-06-22 15:07:45 +00:00
Hajimu UMEMOTO
a0a59ae4af - Pullup even when the extention header is unknown, to prevent
infinite loop with net.inet6.ip6.fw.deny_unknown_exthdrs=0.
- Teach ipv6 and ipencap as they appear in an IPv4/IPv6 over IPv6
  tunnel.
- Test the next extention header even when the routing header type
  is unknown with net.inet6.ip6.fw.deny_unknown_exthdrs=0.

Found by:	xcast-fan-club
MFC after:	1 week
2006-06-22 13:22:54 +00:00
Andre Oppermann
c9f7b0ad5b Allocate a zero'ed syncache hashtable. mtx_init() tests the supplied
memory location for already existing/initialized mutexes.  With random
data in the memory location this fails (ie. after a soft reboot).

Reported by:	brueffer, YAMAMOTO Shigeru
Submitted by:	YAMAMOTO Shigeru <shigeru-at-iij.ad.jp>
2006-06-20 08:11:30 +00:00
David Malone
5e1aa27995 When we receive an out-of-window SYN for an "ESTABLISHED" connection,
ACK the SYN as required by RFC793, rather than ignoring it. NetBSD
have had a similar change since 1999.

PR:		93236
Submitted by:	Grant Edwards <grante@visi.com>
MFC after:	1 month
2006-06-19 12:33:52 +00:00
Andre Oppermann
6593a94979 Remove T/TCP RFC1644 Connection Count comparison macros. They are no longer
used and needed.

Sponsored by:   TCP/IP Optimization Fundraise 2005
2006-06-18 14:24:12 +00:00
Andre Oppermann
2f1a4ccfc1 Do not access syncache entry before it was allocated for the TF_NOOPT case
in syncache_add().

Found by:	Coverity Prevent
CID:		1473
2006-06-18 13:03:42 +00:00
Andre Oppermann
8411d000a1 Move all syncache related structures to tcp_syncache.c. They are only used
there.

This unbreaks userland programs that include tcp_var.h.

Discussed with:	rwatson
2006-06-18 12:26:11 +00:00
Andre Oppermann
bdfbf1e203 Remove double lock acquisition in syncookie_lookup() which came from last
minute conversions to macros.

Pointy hat to:	andre
2006-06-18 11:48:03 +00:00
Andre Oppermann
ee2e4c1d4e Fix the !INET6 compile.
Reported by:	alc
2006-06-17 18:42:07 +00:00
Andre Oppermann
93f0d0c5bf Rearrange fields in struct syncache and syncache_head to make them more
cache line friendly.

Sponsored by:   TCP/IP Optimization Fundraise 2005
2006-06-17 17:57:36 +00:00
Andre Oppermann
0c529372f0 ANSIfy and tidy up comments.
Sponsored by:   TCP/IP Optimization Fundraise 2005
2006-06-17 17:49:11 +00:00
Andre Oppermann
351630c40d Add locking to TCP syncache and drop the global tcpinfo lock as early
as possible for the syncache_add() case.  The syncache timer no longer
aquires the tcpinfo lock and timeout/retransmit runs can happen in
parallel with bucket granularity.

On a P4 the additional locks cause a slight degression of 0.7% in tcp
connections per second.  When IP and TCP input are deserialized and
can run in parallel this little overhead can be neglected. The syncookie
handling still leaves room for improvement and its random salts may be
moved to the syncache bucket head structures to remove the second lock
operation currently required for it.  However this would be a more
involved change from the way syncookies work at the moment.

Reviewed by:	rwatson
Tested by:	rwatson, ps (earlier version)
Sponsored by:	TCP/IP Optimization Fundraise 2005
2006-06-17 17:32:38 +00:00
Oleg Bulyzhin
254c472561 Add support of 'tablearg' feature for:
- 'tag' & 'untag' action parameters.
- 'tagged' & 'limit' rule options.
Rule examples:
	pipe 1 tag tablearg ip from table(1) to any
	allow ip from any to table(2) tagged tablearg
	allow tcp from table(3) to any 25 setup limit src-addr tablearg

sbin/ipfw/ipfw2.c:
1) new macros
   GET_UINT_ARG - support of 'tablearg' keyword, argument range checking.
   PRINT_UINT_ARG - support of 'tablearg' keyword.
2) strtoport(): do not silently truncate/accept invalid port list expressions
   like: '1,2-abc' or '1,2-3-4' or '1,2-3x4'. style(9) cleanup.

Approved by:	glebius (mentor)
MFC after:	1 month
2006-06-15 09:39:22 +00:00
Oleg Bulyzhin
58a0fab73f install_state(): style(9) cleanup
Approved by:	glebius (mentor)
MFC after:	1 month
2006-06-15 08:54:29 +00:00
Andrew Thompson
5feebeeb53 Enable proxy ARP answers on any of the bridged interfaces if proxy record
belongs to another interface within the bridge group.

PR:		kern/94408
Submitted by:	Eygene A. Ryabinkin
MFC after:	1 month
2006-06-09 00:33:30 +00:00
Oleg Bulyzhin
458009ae93 install_state() should properly initialize 'addr_type' field of newly created
flows for O_LIMIT rules.  Otherwise 'ipfw -d show' is unable to display
PARENT rules properly.
(This bug was exposed by ipfw2.c rev.1.90)

Approved by:	glebius (mentor)
MFC after:	2 weeks
2006-06-08 11:27:45 +00:00
Oleg Bulyzhin
d2dc1907e8 Fix following rules: pipe X (tag|altq) Y ...
Approved by:	glebius (mentor)
MFC after:	2 weeks
2006-06-08 11:13:23 +00:00
Robert Watson
f2de87fec4 Push acquisition of pcbinfo lock out of tcp_usr_attach() into
tcp_attach() after the call to soreserve(), as it doesn't require
the global lock.  Rearrange inpcb locking here also.

MFC after:	1 month
2006-06-04 09:31:34 +00:00
Robert Watson
d8ab0ec661 When entering a timer on a tcpcb, don't continue processing if it has been
dropped.  This prevents a bug introduced during the socket/pcb refcounting
work from occuring, in which occasionally the retransmit timer may fire
after a connection has been reset, resulting in the resulting R|A TCP
packet having a source port of 0, as the port reservation has been
released.

While here, fixing up some RUNLOCK->WUNLOCK bugs.

MFC after:	1 month
2006-06-03 19:37:08 +00:00
Robert Watson
f24618aaf0 Acquire udbinfo lock after call to soreserve() rather than before, as it
is not required.  This simplifies error-handling, and reduces the time
that this lock is held.

MFC after:	1 month
2006-06-03 19:29:26 +00:00
Christian S.J. Peron
16d878cc99 Fix the following bpf(4) race condition which can result in a panic:
(1) bpf peer attaches to interface netif0
	(2) Packet is received by netif0
	(3) ifp->if_bpf pointer is checked and handed off to bpf
	(4) bpf peer detaches from netif0 resulting in ifp->if_bpf being
	    initialized to NULL.
	(5) ifp->if_bpf is dereferenced by bpf machinery
	(6) Kaboom

This race condition likely explains the various different kernel panics
reported around sending SIGINT to tcpdump or dhclient processes. But really
this race can result in kernel panics anywhere you have frequent bpf attach
and detach operations with high packet per second load.

Summary of changes:

- Remove the bpf interface's "driverp" member
- When we attach bpf interfaces, we now set the ifp->if_bpf member to the
  bpf interface structure. Once this is done, ifp->if_bpf should never be
  NULL. [1]
- Introduce bpf_peers_present function, an inline operation which will do
  a lockless read bpf peer list associated with the interface. It should
  be noted that the bpf code will pickup the bpf_interface lock before adding
  or removing bpf peers. This should serialize the access to the bpf descriptor
  list, removing the race.
- Expose the bpf_if structure in bpf.h so that the bpf_peers_present function
  can use it. This also removes the struct bpf_if; hack that was there.
- Adjust all consumers of the raw if_bpf structure to use bpf_peers_present

Now what happens is:

	(1) Packet is received by netif0
	(2) Check to see if bpf descriptor list is empty
	(3) Pickup the bpf interface lock
	(4) Hand packet off to process

From the attach/detach side:

	(1) Pickup the bpf interface lock
	(2) Add/remove from bpf descriptor list

Now that we are storing the bpf interface structure with the ifnet, there is
is no need to walk the bpf interface list to locate the correct bpf interface.
We now simply look up the interface, and initialize the pointer. This has a
nice side effect of changing a bpf interface attach operation from O(N) (where
N is the number of bpf interfaces), to O(1).

[1] From now on, we can no longer check ifp->if_bpf to tell us whether or
    not we have any bpf peers that might be interested in receiving packets.

In collaboration with:	sam@
MFC after:	1 month
2006-06-02 19:59:33 +00:00
Robert Watson
ad3a630f7e Minor restyling and cleanup around ipport_tick().
MFC after:	1 month
2006-06-02 08:18:27 +00:00
Oleg Bulyzhin
6a7d5cb645 Implement internal (i.e. inside kernel) packet tagging using mbuf_tags(9).
Since tags are kept while packet resides in kernelspace, it's possible to
use other kernel facilities (like netgraph nodes) for altering those tags.

Submitted by:	Andrey Elsukov <bu7cher at yandex dot ru>
Submitted by:	Vadim Goncharov <vadimnuclight at tpu dot ru>
Approved by:	glebius (mentor)
Idea from:	OpenBSD PF
MFC after:	1 month
2006-05-24 13:09:55 +00:00
Maxim Konovalov
d45e4f9945 o In udp|rip_disconnect() acquire a socket lock before the socket
state modification.  To prevent races do that while holding inpcb
lock.

Reviewed by:	rwatson
2006-05-21 19:28:46 +00:00
Maxim Konovalov
635354c446 o Add missed error check: in ip_ctloutput() sooptcopyin() returns a
result but we never examine it.

Reviewed by:	rwatson
MFC after:	2 weeks
2006-05-21 17:52:08 +00:00
Bruce M Simpson
8d7d85149e Initialize the new members of struct ip_moptions as
a defensive programming measure.

Note that whilst these members are not used by the ip_output()
path, we are passing an instance of struct ip_moptions here
which is declared on the stack (which could be considered a
bad thing).

ip_output() does not consume struct ip_moptions, but in case it
does in future, declare an in_multi vector on the stack too to
behave more like ip_findmoptions() does.
2006-05-18 19:51:08 +00:00
Gleb Smirnoff
e5f88c4492 Since m_pullup() can return a new mbuf, change gre_input2() to
return mbuf back to gre_input(). If the former returns mbuf back
to the latter, then pass it to raw_input().

Coverity ID:	829
2006-05-16 11:15:22 +00:00
Gleb Smirnoff
ffb761f624 - Backout one line from 1.78. The tp can be freed by tcp_drop().
- Style next line.

Coverity ID:	912
2006-05-16 10:51:26 +00:00
Maxim Konovalov
eb16472f74 o In rip_disconnect() do not call rip_abort(), just mark a socket
as not connected.  In soclose() case rip_detach() will kill inpcb for
us later.

It makes rawconnect regression test do not panic a system.

Reviewed by:	rwatson
X-MFC after:	with all 1th April inpcb changes
2006-05-15 09:28:57 +00:00
Max Laier
0e7185f6e7 Use only lower 64bit of src/dest (and src/dest port) for hashing of IPv6
connections and get rid of the flow_id as it is not guaranteed to be stable
some (most?) current implementations seem to just zero it out.

PR:		kern/88664
Reported by:	jylefort
Submitted by:	Joost Bekkers (w/ changes)
Tested by	"regisr" <regisrApoboxDcom>
2006-05-14 23:42:24 +00:00
Bruce M Simpson
3548bfc964 Fix a long-standing limitation in IPv4 multicast group membership.
By making the imo_membership array a dynamically allocated vector,
this minimizes disruption to existing IPv4 multicast code. This
change breaks the ABI for the kernel module ip_mroute.ko, and may
cause a small amount of churn for folks working on the IGMPv3 merge.

Previously, sockets were subject to a compile-time limitation on
the number of IPv4 group memberships, which was hard-coded to 20.
The imo_membership relationship, however, is 1:1 with regards to
a tuple of multicast group address and interface address. Users who
ran routing protocols such as OSPF ran into this limitation on machines
with a large system interface tree.
2006-05-14 14:22:49 +00:00
Max Laier
656faadcb8 Remove ip6fw. Since ipfw has full functional IPv6 support now and - in
contrast to ip6fw - is properly lockes, it is time to retire ip6fw.
2006-05-12 20:39:23 +00:00
Max Laier
e93187482d Reintroduce net.inet6.ip6.fw.enable sysctl to dis/enable the ipv6 processing
seperately.  Also use pfil hook/unhook instead of keeping the check
functions in pfil just to return there based on the sysctl.  While here fix
some whitespace on a nearby SYSCTL_ macro.
2006-05-12 04:41:27 +00:00
Max Laier
432288dcb6 Don't claim "(+ipv6)" if we didn't build with INET6. 2006-05-11 15:22:38 +00:00
Robert Watson
59b8854eee Modify UDP to use sosend_dgram() instead of sosend(). This allows
for signicantly optimized UDP socket I/O when using a single UDP
socket from many threads or processes that share it, by avoiding
significant locking and other overhead in the general sosend()
path that isn't necessary for simple datagram sockets.  Specifically,
this change results in a significant performance improvement for
threaded name service in BIND9 under load.

Suggested by:	Jinmei_Tatsuya at isc dot org
2006-05-06 11:24:59 +00:00
Bjoern A. Zeeb
91b309a1c4 Make sure the ip data pointer is correct before touching it again
after ipsec4_output processing else KAME IPSec using the handbook
configuration with gif(4) will panic the kernel.

Problem reported by:    t. patterson <tp lot.org>
Tested by:              t. patterson <tp lot.org>
2006-05-05 07:31:03 +00:00
Robert Watson
3127286870 Only return (tw) from tcp_twclose() if reuse is passed, otherwise
return NULL.  In principle this shouldn't change the behavior, but
avoids returning a potentially invalid/inappropriate pointer to
the caller.

Found with:	Coverity Prevent (tm)
Submitted by:	pjd
MFC after:	3 months
2006-05-05 06:50:23 +00:00
Pawel Jakub Dawidek
1d7d0bfe5e /tmp/cvsTXPIwQ 2006-05-05 06:24:34 +00:00