Bug 167947 - [setfib] [patch] arpresolve checks only the default FIB for the interface route
Summary: [setfib] [patch] arpresolve checks only the default FIB for the interface route
Status: Closed FIXED
Alias: None
Product: Base System
Classification: Unclassified
Component: kern (show other bugs)
Version: 8.3-STABLE
Hardware: Any Any
: Normal Affects Some People
Assignee: Alan Somers
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-05-16 04:50 UTC by Nikolay Denev
Modified: 2014-09-18 20:35 UTC (History)
0 users

See Also:


Attachments
file.diff (806 bytes, patch)
2012-05-16 04:50 UTC, Nikolay Denev
no flags Details | Diff
167947.diff (1.05 KB, patch)
2014-03-25 15:15 UTC, Alan Somers
no flags Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Nikolay Denev 2012-05-16 04:50:01 UTC
arpresolve() checks if a connected route exists before adding ARP entries to the arp table.
The problem is that it only checks the default FIB 0.
When one does have interface in non default FIB with it's directly attached routes removed from the default FIB, arpresolve fails and no ARP entries can be added :

arpresolve: can't allocate llinfo for x.x.x.x

Fix: Apply the included patch.

Patch attached with submission follows:
How-To-Repeat: put these in /boot/loader.conf :

net.add_addr_allfibs="0"
net.fibs="8"

Then :

/usr/sbin/setfib 1 /sbin/ifconfig em1 10.180.0.2 netmask 255.255.255.0 fib 1
/usr/sbin/setfib 1 /sbin/route add default 10.180.0.254

Now then I try to ping the host from its gw 10.180.0.254, the host generates
"arpresolve: can't allocate llinfo for 10.180.0.254" messages.
Comment 1 Glen Barber freebsd_committer freebsd_triage 2012-05-16 04:57:09 UTC
Responsible Changed
From-To: freebsd-bugs->freebsd-net

Fix synopsis and assign.
Comment 2 Rudolf Polzer 2012-06-12 08:34:23 UTC
Hi,

I had the problem too, and could confirm the fix.

FreeBSD ********.arcor.net 8.3-RELEASE-p2 FreeBSD 8.3-RELEASE-p2 #1: Tue
Jun 12 09:06:31 CEST 2012  =20
root@********.arcor.net:/usr/src/sys/amd64/compile/GENERICfib  amd64

However, the fix ONLY works if using ifconfig ... fib N to set the fib
for an interface, and not the FreeBSD 7.x method to assign the FIB via
ipfw rules. Just thought I should mention this - in our case, the
ifconfig way is the better choice anyway.

Best regards,

Rudolf Polzer=
Comment 3 Alexander V. Chernikov freebsd_committer freebsd_triage 2013-10-17 18:52:40 UTC
Responsible Changed
From-To: freebsd-net->melifaro

Take.
Comment 4 Alan Somers freebsd_committer freebsd_triage 2014-03-11 19:17:59 UTC
Melifaro,
     I'm working in this area at the moment.  I even think that I can
write an ATF regression test for this.  Do you mind if I take the bug
myself?

-Alan
Comment 5 Alexander V. Chernikov freebsd_committer freebsd_triage 2014-03-11 19:37:07 UTC
>  Melifaro,
>       I'm working in this area at the moment.  I even think that I can
>  write an ATF regression test for this.  Do you mind if I take the bug
>  myself?
Well, probably we should discuss what exactly you're planning to do first.
Proper fix is quite tricky.

(1) Current fib code has an awful hack for interface routes which are
added to all fibs (net.add_addr_allfibs). We need to remove that first
which involves some route/rtsock code cleanup first (which I stared to
do some time ago). We also need to handle interface fib changes properly
and so on.

(2) Then, we have to do per-interface LLINFO for IPv4/IPv6 (since it is
currently global).

(3) After that, we can think of changing arpresolve to honor fib, but we
should handle cases like 'route leaking' or something similar which
involves GW mac from GRT (fib 0) being proxied somehow to fib X
Comment 6 Alan Somers freebsd_committer freebsd_triage 2014-03-11 20:27:07 UTC
On Tue, Mar 11, 2014 at 1:37 PM, Alexander V. Chernikov
<melifaro@freebsd.org> wrote:
>>  Melifaro,
>>       I'm working in this area at the moment.  I even think that I can
>>  write an ATF regression test for this.  Do you mind if I take the bug
>>  myself?
> Well, probably we should discuss what exactly you're planning to do first.
> Proper fix is quite tricky.
>
> (1) Current fib code has an awful hack for interface routes which are
> added to all fibs (net.add_addr_allfibs). We need to remove that first
> which involves some route/rtsock code cleanup first (which I stared to
> do some time ago). We also need to handle interface fib changes properly
> and so on.

I'm currently doing all my work with net.add_addr_allfibs=0.  To get
it working, I had to apply Nikolay's patch for this bug, and fix
several other bugs too.  I haven't opened GNATS issues for all of them
yet, but I intend to.

>
> (2) Then, we have to do per-interface LLINFO for IPv4/IPv6 (since it is
> currently global).

Why do you want separate tables per interface?  Are you concerned that
physically separate networks, connected to different interfaces of the
same host, might each contain hosts with the same IP address but
different MACs?  I don't think that's an issue for IPv4.  IPv4 doesn't
have any concept of scoping, so I don't think there's any sane way to
construct such a network.  IPv6 might be a different story.  But I
would consider that a separate issue from kern/167947.

>
> (3) After that, we can think of changing arpresolve to honor fib, but we
> should handle cases like 'route leaking' or something similar which
> involves GW mac from GRT (fib 0) being proxied somehow to fib X

I don't understand.  Are you talking about a situation where both FIBs
would have direct access to the same subnet?  I've tested that
situation extensively (and had to fix another bug there).  With my
patches, I can have two interfaces, each with separate FIBs, attached
to the same subnet and it works fine.

-Alan

>
>
Comment 7 Alexander V. Chernikov freebsd_committer freebsd_triage 2014-03-12 07:36:07 UTC
On 12.03.2014 00:27, Alan Somers wrote:
>> (2) Then, we have to do per-interface LLINFO for IPv4/IPv6 (since it is
>> currently global).
> Why do you want separate tables per interface?  Are you concerned that
Sorry, actually I mean 'per-fib', but this doesn't differ much.
> physically separate networks, connected to different interfaces of the
> same host, might each contain hosts with the same IP address but
> different MACs?  I don't think that's an issue for IPv4.  IPv4 doesn't
Well, it is. RFC1918 was always popular. And it is even more popular 
nowadays :)
> have any concept of scoping, so I don't think there's any sane way to
> construct such a network.  IPv6 might be a different story.  But I
> would consider that a separate issue from kern/167947.
>
>> (3) After that, we can think of changing arpresolve to honor fib, but we
>> should handle cases like 'route leaking' or something similar which
>> involves GW mac from GRT (fib 0) being proxied somehow to fib X
> I don't understand.  Are you talking about a situation where both FIBs
> would have direct access to the same subnet?  I've tested that
Sort of.
> situation extensively (and had to fix another bug there).  With my
> patches, I can have two interfaces, each with separate FIBs, attached
> to the same subnet and it works fine.
Can you send me these patches?
>
> -Alan
>
>>
Comment 8 Alan Somers freebsd_committer freebsd_triage 2014-03-12 15:32:54 UTC
On Wed, Mar 12, 2014 at 1:36 AM, Alexander V. Chernikov
<melifaro@freebsd.org> wrote:
> On 12.03.2014 00:27, Alan Somers wrote:
>>>
>>> (2) Then, we have to do per-interface LLINFO for IPv4/IPv6 (since it is
>>> currently global).
>>
>> Why do you want separate tables per interface?  Are you concerned that
>
> Sorry, actually I mean 'per-fib', but this doesn't differ much.
>
>> physically separate networks, connected to different interfaces of the
>> same host, might each contain hosts with the same IP address but
>> different MACs?  I don't think that's an issue for IPv4.  IPv4 doesn't
>
> Well, it is. RFC1918 was always popular. And it is even more popular
> nowadays :)

Sure, private networks are popular.  But it's impossible for traffic
to pass from one to another if they reuse addresses.  Given that,
under what circumstances would a FreeBSD box ever need to be able to
talk to multiple such networks?  I can think of a bad reason or two,
but no good reasons.

>
>> have any concept of scoping, so I don't think there's any sane way to
>> construct such a network.  IPv6 might be a different story.  But I
>> would consider that a separate issue from kern/167947.
>>
>>> (3) After that, we can think of changing arpresolve to honor fib, but we
>>> should handle cases like 'route leaking' or something similar which
>>> involves GW mac from GRT (fib 0) being proxied somehow to fib X
>>
>> I don't understand.  Are you talking about a situation where both FIBs
>> would have direct access to the same subnet?  I've tested that
>
> Sort of.
>
>> situation extensively (and had to fix another bug there).  With my
>> patches, I can have two interfaces, each with separate FIBs, attached
>> to the same subnet and it works fine.
>
> Can you send me these patches?

I'll open new bugs for each one.  First, I need to make ATF test cases
for them.  I've got about another day's worth of work there.

-Alan
Comment 9 dfilter service freebsd_committer freebsd_triage 2014-03-20 20:39:48 UTC
Author: asomers
Date: Thu Mar 20 20:39:41 2014
New Revision: 263445
URL: http://svnweb.freebsd.org/changeset/base/263445

Log:
  Add several ATF tests that deal with multiple fibs.  They're described in
  several different PRs, but the tests share some common code, so I'm
  committing them together.
  
  sbin/ifconfig/tests
  sbin/ifconfig/tests/fibs_test.sh
  sbin/ifconfig/tests/Makefile
  sbin/ifconfig/Makefile
  	Add fibs_test.sh, which regresses bin/187551
  
  tests/sys/netinet
  tests/sys/netinet/fibs_test.sh
  tests/sys/netinet/udp_dontroute.c
  tests/sys/netinet/Makefile
  tests/sys/Makefile
  	Add fibs_test.sh, which regresses kern/167947, kern/187552
  	kern/187549, kern/187550, and kern/187553
  
  etc/mtree/BSD.tests.dist
  	Add newly created directories
  
  PR:		bin/187551
  PR:		kern/167947
  PR:		kern/187552
  PR:		kern/187549
  PR:		kern/187550
  PR:		kern/187553
  Discussed with:	melifaro
  MFC after:	3 weeks
  Sponsored by:	Spectra Logic Corporation

Added:
  head/sbin/ifconfig/tests/
  head/sbin/ifconfig/tests/Makefile   (contents, props changed)
  head/sbin/ifconfig/tests/fibs_test.sh   (contents, props changed)
  head/tests/sys/netinet/
  head/tests/sys/netinet/Makefile   (contents, props changed)
  head/tests/sys/netinet/fibs_test.sh   (contents, props changed)
  head/tests/sys/netinet/udp_dontroute.c   (contents, props changed)
Modified:
  head/etc/mtree/BSD.tests.dist
  head/sbin/ifconfig/Makefile
  head/tests/sys/Makefile

Modified: head/etc/mtree/BSD.tests.dist
==============================================================================
--- head/etc/mtree/BSD.tests.dist	Thu Mar 20 20:33:18 2014	(r263444)
+++ head/etc/mtree/BSD.tests.dist	Thu Mar 20 20:39:41 2014	(r263445)
@@ -75,6 +75,8 @@
             ..
             growfs
             ..
+            ifconfig
+            ..
             mdconfig
             ..
         ..
@@ -91,6 +93,8 @@
         sys
             kern
             ..
+            netinet
+            ..
         ..
         usr.bin
             apply

Modified: head/sbin/ifconfig/Makefile
==============================================================================
--- head/sbin/ifconfig/Makefile	Thu Mar 20 20:33:18 2014	(r263444)
+++ head/sbin/ifconfig/Makefile	Thu Mar 20 20:39:41 2014	(r263445)
@@ -63,4 +63,8 @@ MAN=	ifconfig.8
 CFLAGS+= -Wall -Wmissing-prototypes -Wcast-qual -Wwrite-strings -Wnested-externs
 WARNS?=	2
 
+.if ${MK_TESTS} != "no"
+SUBDIR+=    tests
+.endif
+
 .include <bsd.prog.mk>

Added: head/sbin/ifconfig/tests/Makefile
==============================================================================
--- /dev/null	00:00:00 1970	(empty, because file is newly added)
+++ head/sbin/ifconfig/tests/Makefile	Thu Mar 20 20:39:41 2014	(r263445)
@@ -0,0 +1,9 @@
+# $FreeBSD$
+
+.include <bsd.own.mk>
+
+TESTSDIR=	${TESTSBASE}/sbin/ifconfig
+
+ATF_TESTS_SH=	fibs_test
+
+.include <bsd.test.mk>

Added: head/sbin/ifconfig/tests/fibs_test.sh
==============================================================================
--- /dev/null	00:00:00 1970	(empty, because file is newly added)
+++ head/sbin/ifconfig/tests/fibs_test.sh	Thu Mar 20 20:39:41 2014	(r263445)
@@ -0,0 +1,126 @@
+#
+#  Copyright (c) 2014 Spectra Logic Corporation
+#  All rights reserved.
+# 
+#  Redistribution and use in source and binary forms, with or without
+#  modification, are permitted provided that the following conditions
+#  are met:
+#  1. Redistributions of source code must retain the above copyright
+#     notice, this list of conditions, and the following disclaimer,
+#     without modification.
+#  2. Redistributions in binary form must reproduce at minimum a disclaimer
+#     substantially similar to the "NO WARRANTY" disclaimer below
+#     ("Disclaimer") and any redistribution must be conditioned upon
+#     including a substantially similar Disclaimer requirement for further
+#     binary redistribution.
+# 
+#  NO WARRANTY
+#  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
+#  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#  HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+#  DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+#  OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+#  HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+#  STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
+#  IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+#  POSSIBILITY OF SUCH DAMAGES.
+# 
+#  Authors: Alan Somers         (Spectra Logic Corporation)
+#
+# $FreeBSD$
+
+
+# Regression test for bin/187551
+atf_test_case process_fib cleanup
+process_fib_head()
+{
+	atf_set "descr" "ifconfig will set its process fib whenever configuring an interface with nondefault fib"
+	atf_set "require.user" "root"
+	atf_set "require.config" "fibs"
+}
+process_fib_body()
+{
+	atf_expect_fail "bin/187551 ifconfig should change its process fib when configuring an interface with nondefault fib"
+	# Configure the TAP interface to use an RFC5737 nonrouteable address
+	# and a non-default fib
+	ADDR="192.0.2.2"
+	SUBNET="192.0.2.0"
+	MASK="24"
+
+	# Check system configuration
+	if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then
+		atf_skip "This test requires net.add_addr_allfibs=0"
+	fi
+	get_fibs 1
+
+	# Configure a TAP interface
+	get_tap
+	ktrace ifconfig $TAP ${ADDR}/${MASK} fib $FIB0
+	if kdump -s | egrep -q 'CALL[[:space:]]+setfib'; then
+		atf_pass
+	else
+		atf_fail "ifconfig never called setfib(2)"
+	fi
+}
+
+process_fib_cleanup()
+{
+	cleanup_tap
+}
+
+atf_init_test_cases()
+{
+	atf_add_test_case process_fib
+}
+
+
+# parameter numfibs	The number of fibs to lookup
+get_fibs()
+{
+	NUMFIBS=$1
+	net_fibs=`sysctl -n net.fibs`
+	i=0
+	while [ $i -lt "$NUMFIBS" ]; do
+		fib=`atf_config_get "fibs" | \
+			awk -v i=$(( i + 1 )) '{print $i}'`
+		echo "fib is ${fib}"
+		eval FIB${i}=${fib}
+		if [ "$fib" -ge "$net_fibs" ]; then
+			atf_skip "The ${i}th configured fib is ${fib}, which is not less than net.fibs, which is ${net_fibs}"
+		fi
+		i=$(( $i + 1 ))
+	done
+}
+
+
+
+# Creates a new tap(4) interface, registers it for cleanup, and returns the
+# name via the environment variable TAP
+get_tap()
+{
+	local TAPN=0
+	while ! ifconfig tap${TAPN} create > /dev/null 2>&1; do
+		if [ "$TAPN" -ge 8 ]; then
+			atf_skip "Could not create a tap(4) interface"
+		else
+			TAPN=$(($TAPN + 1))
+		fi
+	done
+	local TAPD=tap${TAPN}
+	# Record the TAP device so we can clean it up later
+	echo ${TAPD} >> "tap_devices_to_cleanup"
+	TAP=${TAPD}
+}
+
+
+
+
+cleanup_tap()
+{
+	for TAPD in `cat "tap_devices_to_cleanup"`; do
+		ifconfig ${TAPD} destroy
+	done
+}
+

Modified: head/tests/sys/Makefile
==============================================================================
--- head/tests/sys/Makefile	Thu Mar 20 20:33:18 2014	(r263444)
+++ head/tests/sys/Makefile	Thu Mar 20 20:39:41 2014	(r263445)
@@ -5,6 +5,7 @@
 .PATH: ${.CURDIR}/..
 
 TESTS_SUBDIRS+=		kern
+TESTS_SUBDIRS+=		netinet
 TESTSDIR= ${TESTSBASE}/sys
 
 KYUAFILE= yes

Added: head/tests/sys/netinet/Makefile
==============================================================================
--- /dev/null	00:00:00 1970	(empty, because file is newly added)
+++ head/tests/sys/netinet/Makefile	Thu Mar 20 20:39:41 2014	(r263445)
@@ -0,0 +1,12 @@
+# $FreeBSD$
+
+TESTSDIR=	${TESTSBASE}/sys/netinet
+BINDIR=		${TESTSDIR}
+
+ATF_TESTS_SH+=	fibs_test
+PROG=	udp_dontroute
+SRCS=	udp_dontroute.c
+NO_MAN=
+WARNS?=	6
+
+.include <bsd.test.mk>

Added: head/tests/sys/netinet/fibs_test.sh
==============================================================================
--- /dev/null	00:00:00 1970	(empty, because file is newly added)
+++ head/tests/sys/netinet/fibs_test.sh	Thu Mar 20 20:39:41 2014	(r263445)
@@ -0,0 +1,371 @@
+#
+#  Copyright (c) 2014 Spectra Logic Corporation
+#  All rights reserved.
+# 
+#  Redistribution and use in source and binary forms, with or without
+#  modification, are permitted provided that the following conditions
+#  are met:
+#  1. Redistributions of source code must retain the above copyright
+#     notice, this list of conditions, and the following disclaimer,
+#     without modification.
+#  2. Redistributions in binary form must reproduce at minimum a disclaimer
+#     substantially similar to the "NO WARRANTY" disclaimer below
+#     ("Disclaimer") and any redistribution must be conditioned upon
+#     including a substantially similar Disclaimer requirement for further
+#     binary redistribution.
+# 
+#  NO WARRANTY
+#  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
+#  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#  HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+#  DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+#  OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+#  HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+#  STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
+#  IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+#  POSSIBILITY OF SUCH DAMAGES.
+# 
+#  Authors: Alan Somers         (Spectra Logic Corporation)
+#
+# $FreeBSD$
+
+# All of the tests in this file requires the test-suite config variable "fibs"
+# to be defined to a space-delimited list of FIBs that may be used for testing.
+
+# arpresolve should check the interface fib for routes to a target when
+# creating an ARP table entry.  This is a regression for kern/167947, where
+# arpresolve only checked the default route.
+#
+# Outline:
+# Create two tap(4) interfaces
+# Simulate a crossover cable between them by using net/socat
+# Use nping (from security/nmap) to send an ICMP echo request from one
+# interface to the other, spoofing the source IP.  The source IP must be
+# spoofed, or else it will already have an entry in the arp table. 
+# Check whether an arp entry exists for the spoofed IP
+atf_test_case arpresolve_checks_interface_fib cleanup
+arpresolve_checks_interface_fib_head()
+{
+	atf_set "descr" "arpresolve should check the interface fib, not the default fib, for routes"
+	atf_set "require.user" "root"
+	atf_set "require.config" "fibs"
+	atf_set "require.progs" "socat nping"
+}
+arpresolve_checks_interface_fib_body()
+{
+	atf_expect_fail "kern/167947 arpresolve checks only the default FIB for the interface route"
+	# Configure the TAP interfaces to use a RFC5737 nonrouteable addresses
+	# and a non-default fib
+	ADDR0="192.0.2.2"
+	ADDR1="192.0.2.3"
+	SUBNET="192.0.2.0"
+	# Due to bug TBD (regressed by multiple_fibs_on_same_subnet) we need
+	# diffferent subnet masks, or FIB1 won't have a subnet route.
+	MASK0="24"
+	MASK1="25"
+	# Spoof a MAC that is reserved per RFC7042
+	SPOOF_ADDR="192.0.2.4"
+	SPOOF_MAC="00:00:5E:00:53:00"
+
+	# Check system configuration
+	if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then
+		atf_skip "This test requires net.add_addr_allfibs=0"
+	fi
+	get_fibs 2
+
+	# Configure TAP interfaces
+	setup_tap "$FIB0" ${ADDR0} ${MASK0}
+	TAP0=$TAP
+	setup_tap "$FIB1" ${ADDR1} ${MASK1}
+	TAP1=$TAP
+
+	# Simulate a crossover cable
+	socat /dev/${TAP0} /dev/${TAP1} &
+	SOCAT_PID=$!
+	echo ${SOCAT_PID} >> "processes_to_kill"
+	
+	# Send an ICMP echo request with a spoofed source IP
+	setfib 2 nping -c 1 -e ${TAP0} -S ${SPOOF_ADDR} \
+		--source-mac ${SPOOF_MAC} --icmp --icmp-type "echo-request" \
+		--icmp-code 0 --icmp-id 0xdead --icmp-seq 1 --data 0xbeef \
+		${ADDR1}
+	# For informational and debugging purposes only, look for the
+	# characteristic error message
+	dmesg | grep "llinfo.*${SPOOF_ADDR}"
+	# Check that the ARP entry exists
+	atf_check -o match:"${SPOOF_ADDR}.*expires" setfib 3 arp ${SPOOF_ADDR}
+}
+arpresolve_checks_interface_fib_cleanup()
+{
+	for PID in `cat "processes_to_kill"`; do
+		kill $PID
+	done
+	cleanup_tap
+}
+
+
+# Regression test for kern/187549
+atf_test_case loopback_and_network_routes_on_nondefault_fib cleanup
+loopback_and_network_routes_on_nondefault_fib_head()
+{
+	atf_set "descr" "When creating and deleting loopback routes, use the interface's fib"
+	atf_set "require.user" "root"
+	atf_set "require.config" "fibs"
+}
+
+loopback_and_network_routes_on_nondefault_fib_body()
+{
+	atf_expect_fail "kern/187549 Host and network routes for a new interface appear in the wrong FIB"
+	# Configure the TAP interface to use an RFC5737 nonrouteable address
+	# and a non-default fib
+	ADDR="192.0.2.2"
+	SUBNET="192.0.2.0"
+	MASK="24"
+
+	# Check system configuration
+	if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then
+		atf_skip "This test requires net.add_addr_allfibs=0"
+	fi
+	get_fibs 1
+
+	# Configure a TAP interface
+	setup_tap ${FIB0} ${ADDR} ${MASK}
+
+	# Check whether the host route exists in only the correct FIB
+	setfib ${FIB0} netstat -rn -f inet | grep -q "^${ADDR}.*UHS.*lo0"
+	if [ 0 -ne $? ]; then
+		setfib ${FIB0} netstat -rn -f inet
+		atf_fail "Host route did not appear in the correct FIB"
+	fi
+	setfib 0 netstat -rn -f inet | grep -q "^${ADDR}.*UHS.*lo0"
+	if [ 0 -eq $? ]; then
+		setfib 0 netstat -rn -f inet
+		atf_fail "Host route appeared in the wrong FIB"
+	fi
+
+	# Check whether the network route exists in only the correct FIB
+	setfib ${FIB0} netstat -rn -f inet | \
+		grep -q "^${SUBNET}/${MASK}.*${TAPD}"
+	if [ 0 -ne $? ]; then
+		setfib ${FIB0} netstat -rn -f inet
+		atf_fail "Network route did not appear in the correct FIB"
+	fi
+	setfib 0 netstat -rn -f inet | \
+		grep -q "^${SUBNET}/${MASK}.*${TAPD}"
+	if [ 0 -eq $? ]; then
+		setfib ${FIB0} netstat -rn -f inet
+		atf_fail "Network route appeared in the wrong FIB"
+	fi
+}
+
+loopback_and_network_routes_on_nondefault_fib_cleanup()
+{
+	cleanup_tap
+}
+
+
+# Regression test for kern/187552
+atf_test_case default_route_with_multiple_fibs_on_same_subnet cleanup
+default_route_with_multiple_fibs_on_same_subnet_head()
+{
+	atf_set "descr" "Multiple interfaces on the same subnet but with different fibs can both have default routes"
+	atf_set "require.user" "root"
+	atf_set "require.config" "fibs"
+}
+
+default_route_with_multiple_fibs_on_same_subnet_body()
+{
+	atf_expect_fail "kern/187552 default route uses the wrong interface when multiple interfaces have the same subnet but different fibs"
+	# Configure the TAP interfaces to use a RFC5737 nonrouteable addresses
+	# and a non-default fib
+	ADDR0="192.0.2.2"
+	ADDR1="192.0.2.3"
+	GATEWAY="192.0.2.1"
+	SUBNET="192.0.2.0"
+	MASK="24"
+
+	# Check system configuration
+	if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then
+		atf_skip "This test requires net.add_addr_allfibs=0"
+	fi
+	get_fibs 2
+
+	# Configure TAP interfaces
+	setup_tap "$FIB0" ${ADDR0} ${MASK}
+	TAP0=$TAP
+	setup_tap "$FIB1" ${ADDR1} ${MASK}
+	TAP1=$TAP
+
+	# Attempt to add default routes
+	setfib ${FIB0} route add default ${GATEWAY}
+	setfib ${FIB1} route add default ${GATEWAY}
+
+	# Verify that the default route exists for both fibs, with their
+	# respective interfaces.
+	atf_check -o match:"^default.*${TAP0}$" \
+		setfib ${FIB0} netstat -rn -f inet
+	atf_check -o match:"^default.*${TAP1}$" \
+		setfib ${FIB1} netstat -rn -f inet
+}
+
+default_route_with_multiple_fibs_on_same_subnet_cleanup()
+{
+	cleanup_tap
+}
+
+
+# Regression test for kern/187550
+atf_test_case subnet_route_with_multiple_fibs_on_same_subnet cleanup
+subnet_route_with_multiple_fibs_on_same_subnet_head()
+{
+	atf_set "descr" "Multiple FIBs can have subnet routes for the same subnet"
+	atf_set "require.user" "root"
+	atf_set "require.config" "fibs"
+}
+
+subnet_route_with_multiple_fibs_on_same_subnet_body()
+{
+	atf_expect_fail "kern/187550 Multiple interfaces on different FIBs but the same subnet don't all have a subnet route"
+	# Configure the TAP interfaces to use a RFC5737 nonrouteable addresses
+	# and a non-default fib
+	ADDR0="192.0.2.2"
+	ADDR1="192.0.2.3"
+	SUBNET="192.0.2.0"
+	MASK="24"
+
+	# Check system configuration
+	if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then
+		atf_skip "This test requires net.add_addr_allfibs=0"
+	fi
+	get_fibs 2
+
+	# Configure TAP interfaces
+	setup_tap "$FIB0" ${ADDR0} ${MASK}
+	setup_tap "$FIB1" ${ADDR1} ${MASK}
+
+	# Check that a subnet route exists on both fibs
+	atf_check -o ignore setfib "$FIB0" route get $ADDR1
+	atf_check -o ignore setfib "$FIB1" route get $ADDR0
+}
+
+subnet_route_with_multiple_fibs_on_same_subnet_cleanup()
+{
+	cleanup_tap
+}
+
+# Test that source address selection works correctly for UDP packets with
+# SO_DONTROUTE set that are sent on non-default FIBs.
+# This bug was discovered with "setfib 1 netperf -t UDP_STREAM -H some_host"
+# Regression test for kern/187553
+atf_test_case udp_dontroute cleanup
+udp_dontroute_head()
+{
+	atf_set "descr" "Source address selection for UDP packets with SO_DONTROUTE on non-default FIBs works"
+	atf_set "require.user" "root"
+	atf_set "require.config" "fibs"
+}
+
+udp_dontroute_body()
+{
+	atf_expect_fail "kern/187553 Source address selection for UDP packets with SO_DONTROUTE uses the default FIB"
+	# Configure the TAP interface to use an RFC5737 nonrouteable address
+	# and a non-default fib
+	ADDR="192.0.2.2"
+	SUBNET="192.0.2.0"
+	MASK="24"
+	# Use a different IP on the same subnet as the target
+	TARGET="192.0.2.100"
+
+	# Check system configuration
+	if [ 0 != `sysctl -n net.add_addr_allfibs` ]; then
+		atf_skip "This test requires net.add_addr_allfibs=0"
+	fi
+	get_fibs 1
+
+	# Configure a TAP interface
+	setup_tap ${FIB0} ${ADDR} ${MASK}
+
+	# Send a UDP packet with SO_DONTROUTE.  In the failure case, it will
+	# return ENETUNREACH
+	SRCDIR=`atf_get_srcdir`
+	atf_check -o ignore setfib ${FIB0} ${SRCDIR}/udp_dontroute ${TARGET}
+}
+
+udp_dontroute_cleanup()
+{
+	cleanup_tap
+}
+
+
+atf_init_test_cases()
+{
+	atf_add_test_case arpresolve_checks_interface_fib
+	atf_add_test_case loopback_and_network_routes_on_nondefault_fib 
+	atf_add_test_case default_route_with_multiple_fibs_on_same_subnet 
+	atf_add_test_case subnet_route_with_multiple_fibs_on_same_subnet 
+	atf_add_test_case udp_dontroute
+}
+
+# Looks up one or more fibs from the configuration data and validates them.
+# Returns the results in the env varilables FIB0, FIB1, etc.
+
+# parameter numfibs	The number of fibs to lookup
+get_fibs()
+{
+	NUMFIBS=$1
+	net_fibs=`sysctl -n net.fibs`
+	i=0
+	while [ $i -lt "$NUMFIBS" ]; do
+		fib=`atf_config_get "fibs" | \
+			awk -v i=$(( i + 1 )) '{print $i}'`
+		echo "fib is ${fib}"
+		eval FIB${i}=${fib}
+		if [ "$fib" -ge "$net_fibs" ]; then
+			atf_skip "The ${i}th configured fib is ${fib}, which is not less than net.fibs, which is ${net_fibs}"
+		fi
+		i=$(( $i + 1 ))
+	done
+}
+
+# Creates a new tap(4) interface, registers it for cleanup, and returns the
+# name via the environment variable TAP
+get_tap()
+{
+	local TAPN=0
+	while ! ifconfig tap${TAPN} create > /dev/null 2>&1; do
+		if [ "$TAPN" -ge 8 ]; then
+			atf_skip "Could not create a tap(4) interface"
+		else
+			TAPN=$(($TAPN + 1))
+		fi
+	done
+	local TAPD=tap${TAPN}
+	# Record the TAP device so we can clean it up later
+	echo ${TAPD} >> "tap_devices_to_cleanup"
+	TAP=${TAPD}
+}
+
+# Create a tap(4) interface, configure it, and register it for cleanup.
+# parameters:
+# fib
+# IP address
+# Netmask in number of bits (eg 24 or 8)
+# Return: the tap interface name as the env variable TAP
+setup_tap()
+{
+	local FIB=$1
+	local ADDR=$2
+	local MASK=$3
+	get_tap
+	echo setfib ${FIB} ifconfig $TAP ${ADDR}/${MASK} fib $FIB
+	setfib ${FIB} ifconfig $TAP ${ADDR}/${MASK} fib $FIB
+}
+
+cleanup_tap()
+{
+	for TAPD in `cat "tap_devices_to_cleanup"`; do
+		ifconfig ${TAPD} destroy
+	done
+}

Added: head/tests/sys/netinet/udp_dontroute.c
==============================================================================
--- /dev/null	00:00:00 1970	(empty, because file is newly added)
+++ head/tests/sys/netinet/udp_dontroute.c	Thu Mar 20 20:39:41 2014	(r263445)
@@ -0,0 +1,85 @@
+/*
+ *  Copyright (c) 2014 Spectra Logic Corporation
+ *  All rights reserved.
+ * 
+ *  Redistribution and use in source and binary forms, with or without
+ *  modification, are permitted provided that the following conditions
+ *  are met:
+ *  1. Redistributions of source code must retain the above copyright
+ *     notice, this list of conditions, and the following disclaimer,
+ *     without modification.
+ *  2. Redistributions in binary form must reproduce at minimum a disclaimer
+ *     substantially similar to the "NO WARRANTY" disclaimer below
+ *     ("Disclaimer") and any redistribution must be conditioned upon
+ *     including a substantially similar Disclaimer requirement for further
+ *     binary redistribution.
+ * 
+ *  NO WARRANTY
+ *  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
+ *  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *  HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ *  DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ *  OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ *  HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ *  STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
+ *  IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ *  POSSIBILITY OF SUCH DAMAGES.
+ * 
+ *  Authors: Alan Somers         (Spectra Logic Corporation)
+ *
+ * $FreeBSD$
+ */
+
+#include <arpa/inet.h>
+#include <netinet/in.h>
+#include <sys/types.h>
+#include <sys/socket.h>
+
+#include <err.h>
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* 
+ * Sends a single UDP packet to the provided address, with SO_DONTROUTE set
+ * I couldn't find a way to do this with builtin utilities like nc(1)
+ */
+int main(int argc, char **argv)
+{
+	struct sockaddr_in dst;
+	int s;
+	int opt;
+	int ret;
+	const char* buf = "Hello, World!";
+
+	if (argc != 2) {
+		fprintf(stderr, "Usage: %s ip_address\n", argv[0]);
+		exit(2);
+	}
+	s = socket(PF_INET, SOCK_DGRAM, 0);
+	if (s < 0)
+		err(errno, "socket");
+	opt = 1;
+
+	ret = setsockopt(s, SOL_SOCKET, SO_DONTROUTE, &opt, sizeof(opt));
+	if (ret == -1)
+		err(errno, "setsockopt(SO_DONTROUTE)");
+
+	dst.sin_len = sizeof(dst);
+	dst.sin_family = AF_INET;
+	dst.sin_port = htons(46120);
+	dst.sin_addr.s_addr = inet_addr(argv[1]);
+	if (dst.sin_addr.s_addr == htonl(INADDR_NONE)) {
+		fprintf(stderr, "Invalid address: %s\n", argv[1]);
+		exit(2);
+	}
+	ret = sendto(s, buf, strlen(buf), 0, (struct sockaddr*)&dst,
+	    dst.sin_len);
+	if (ret == -1)
+		err(errno, "sendto");
+	
+	return (0);
+}
_______________________________________________
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscribe@freebsd.org"
Comment 10 Alan Somers freebsd_committer freebsd_triage 2014-03-25 15:15:39 UTC
Here is an updated patch for this bug.  The test case is
sys/netinet/fibs_test:arpresolve_checks_interface_fib .  BTW, the
other FIB bugs I've opened are kern/187549, kern/187550, bin/187551,
kern/187552, kern/187553, and bin/187699.
Comment 11 dfilter service freebsd_committer freebsd_triage 2014-03-26 22:46:08 UTC
Author: asomers
Date: Wed Mar 26 22:46:03 2014
New Revision: 263779
URL: http://svnweb.freebsd.org/changeset/base/263779

Log:
  Correct ARP update handling when the routes for network interfaces are
  restricted to a single FIB in a multifib system.
  
  Restricting an interface's routes to the FIB to which it is assigned (by
  setting net.add_addr_allfibs=0) causes ARP updates to fail with "arpresolve:
  can't allocate llinfo for x.x.x.x".  This is due to the ARP update code hard
  coding it's lookup for existing routing entries to FIB 0.
  
  sys/netinet/in.c:
  	When dealing with RTM_ADD (add route) requests for an interface, use
  	the interface's assigned FIB instead of the default (FIB 0).
  
  sys/netinet/if_ether.c:
  	In arpresolve(), enhance error message generated when an
  	lla_lookup() fails so that the interface causing the error is
  	visible in logs.
  
  tests/sys/netinet/fibs_test.sh
  	Clear ATF expected error.
  
  PR:		kern/167947
  Submitted by:	Nikolay Denev <ndenev@gmail.com> (previous version)
  Reviewed by:	melifaro
  MFC after:	3 weeks
  Sponsored by:	Spectra Logic Corporation

Modified:
  head/sys/netinet/if_ether.c
  head/sys/netinet/in.c
  head/tests/sys/netinet/fibs_test.sh

Modified: head/sys/netinet/if_ether.c
==============================================================================
--- head/sys/netinet/if_ether.c	Wed Mar 26 22:30:38 2014	(r263778)
+++ head/sys/netinet/if_ether.c	Wed Mar 26 22:46:03 2014	(r263779)
@@ -334,8 +334,8 @@ retry:
 	if (la == NULL) {
 		if (flags & LLE_CREATE)
 			log(LOG_DEBUG,
-			    "arpresolve: can't allocate llinfo for %s\n",
-			    inet_ntoa(SIN(dst)->sin_addr));
+			    "arpresolve: can't allocate llinfo for %s on %s\n",
+			    inet_ntoa(SIN(dst)->sin_addr), ifp->if_xname);
 		m_freem(m);
 		return (EINVAL);
 	}

Modified: head/sys/netinet/in.c
==============================================================================
--- head/sys/netinet/in.c	Wed Mar 26 22:30:38 2014	(r263778)
+++ head/sys/netinet/in.c	Wed Mar 26 22:46:03 2014	(r263779)
@@ -994,8 +994,9 @@ in_lltable_rtcheck(struct ifnet *ifp, u_
 	KASSERT(l3addr->sa_family == AF_INET,
 	    ("sin_family %d", l3addr->sa_family));
 
-	/* XXX rtalloc1 should take a const param */
-	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
+	/* XXX rtalloc1_fib should take a const param */
+	rt = rtalloc1_fib(__DECONST(struct sockaddr *, l3addr), 0, 0,
+	    ifp->if_fib);
 
 	if (rt == NULL)
 		return (EINVAL);

Modified: head/tests/sys/netinet/fibs_test.sh
==============================================================================
--- head/tests/sys/netinet/fibs_test.sh	Wed Mar 26 22:30:38 2014	(r263778)
+++ head/tests/sys/netinet/fibs_test.sh	Wed Mar 26 22:46:03 2014	(r263779)
@@ -55,7 +55,6 @@ arpresolve_checks_interface_fib_head()
 }
 arpresolve_checks_interface_fib_body()
 {
-	atf_expect_fail "kern/167947 arpresolve checks only the default FIB for the interface route"
 	# Configure the TAP interfaces to use a RFC5737 nonrouteable addresses
 	# and a non-default fib
 	ADDR0="192.0.2.2"
_______________________________________________
svn-src-all@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/svn-src-all
To unsubscribe, send any mail to "svn-src-all-unsubscribe@freebsd.org"
Comment 12 Mark Linimon freebsd_committer freebsd_triage 2014-04-20 23:06:46 UTC
State Changed
From-To: open->closed

see kern/187549. 


Comment 13 Mark Linimon freebsd_committer freebsd_triage 2014-04-20 23:06:46 UTC
Responsible Changed
From-To: melifaro->asomers
Comment 14 Alan Somers freebsd_committer freebsd_triage 2014-04-21 17:14:39 UTC
State Changed
From-To: closed->patched

This is merely patched, not closed, because I haven't MFCed the fix
Comment 15 commit-hook freebsd_committer freebsd_triage 2014-09-18 20:18:02 UTC
A commit references this bug:

Author: asomers
Date: Thu Sep 18 20:17:42 UTC 2014
New revision: 271809
URL: http://svnweb.freebsd.org/changeset/base/271809

Log:
  MFC 263779, except for the ATF test change.

  Correct ARP update handling when the routes for network interfaces are
  restricted to a single FIB in a multifib system.

  Restricting an interface's routes to the FIB to which it is assigned (by
  setting net.add_addr_allfibs=0) causes ARP updates to fail with "arpresolve:
  can't allocate llinfo for x.x.x.x".  This is due to the ARP update code hard
  coding it's lookup for existing routing entries to FIB 0.

  sys/netinet/in.c:
          When dealing with RTM_ADD (add route) requests for an interface, use
          the interface's assigned FIB instead of the default (FIB 0).

  sys/netinet/if_ether.c:
          In arpresolve(), enhance error message generated when an
          lla_lookup() fails so that the interface causing the error is
          visible in logs.

  PR:             kern/167947

Changes:
_U  stable/9/
_U  stable/9/sys/
  stable/9/sys/netinet/if_ether.c
  stable/9/sys/netinet/in.c
Comment 16 Alan Somers freebsd_committer freebsd_triage 2014-09-18 20:35:13 UTC
MFCed to stable/9 by change 271809 and to stable/10 by change 267175.