openbsd
9 years agoMissing t at end of response, from Vincent Bernat.
nicm [Mon, 1 Jun 2015 13:59:57 +0000 (13:59 +0000)]
Missing t at end of response, from Vincent Bernat.

9 years agoFix memory leak in error case. OK bluhm@ benno@
claudio [Mon, 1 Jun 2015 09:54:16 +0000 (09:54 +0000)]
Fix memory leak in error case. OK bluhm@ benno@

9 years agoPut END() matching _ENTRY() (== ENTRY() w/o prof).
uebayasi [Mon, 1 Jun 2015 09:31:40 +0000 (09:31 +0000)]
Put END() matching _ENTRY() (== ENTRY() w/o prof).

9 years agoPut END() matching ENTRY().
uebayasi [Mon, 1 Jun 2015 09:27:53 +0000 (09:27 +0000)]
Put END() matching ENTRY().

9 years agoClear signal handlers before event_reinit as apparently it can otherwise
nicm [Mon, 1 Jun 2015 09:20:19 +0000 (09:20 +0000)]
Clear signal handlers before event_reinit as apparently it can otherwise
cause libevent to go strange.

9 years agoConvert tun(4) to if_input().
mpi [Mon, 1 Jun 2015 07:48:04 +0000 (07:48 +0000)]
Convert tun(4) to if_input().

Tested by Norman Golisz and <mxb AT alumni DOT chalmers DOT se>, thanks!

ok bluhm@

9 years agoSimple macro fixes (Fd -> In, Nm -> Fn) from Kaspars at Bankovskis dot net.
schwarze [Sun, 31 May 2015 23:54:25 +0000 (23:54 +0000)]
Simple macro fixes (Fd -> In, Nm -> Fn) from Kaspars at Bankovskis dot net.
No content change.

9 years agodoes not need syslog.h
deraadt [Sun, 31 May 2015 23:27:06 +0000 (23:27 +0000)]
does not need syslog.h

9 years agoImplement the roff(7) `r' (register exists) conditional.
schwarze [Sun, 31 May 2015 23:12:16 +0000 (23:12 +0000)]
Implement the roff(7) `r' (register exists) conditional.
Missing feature found by Markus <Waldeck at gmx dot de>
in Debian's bash(1) manual page.

9 years agoFix typos.
rpe [Sun, 31 May 2015 21:21:10 +0000 (21:21 +0000)]
Fix typos.

OK miod@

9 years agoIf the first list was empty, ml_join() did not not clear the second
bluhm [Sun, 31 May 2015 20:10:44 +0000 (20:10 +0000)]
If the first list was empty, ml_join() did not not clear the second
list after transferring all elements away.  Reorder the conditionals
to make sure that ml_init() is always called for a non empty second
list.  This makes all cases consistent and is less surprising.
OK dlg@

9 years agoReplace identical code in the MD scripts of the installer with a new
rpe [Sun, 31 May 2015 19:40:10 +0000 (19:40 +0000)]
Replace identical code in the MD scripts of the installer with a new
MI function disklabel_autolayout() which now handles all aspects of
the disklabel auto-layout and autopartitioning case for the root disk.
Remove get_disklabel_template() and merge it with the new function.

"move forward" deraadt@

9 years agoBuild with -mbwx rather than -Wa,-mev56 so that upcoming changes to the compiler
miod [Sat, 30 May 2015 19:53:02 +0000 (19:53 +0000)]
Build with -mbwx rather than -Wa,-mev56 so that upcoming changes to the compiler
will output proper .arch statements in its output.
No functional change.

9 years agoRemove commented out DRMDEBUG/DRM_DEBUG lines. Other drivers don't have
jsg [Sat, 30 May 2015 18:14:08 +0000 (18:14 +0000)]
Remove commented out DRMDEBUG/DRM_DEBUG lines.  Other drivers don't have
similiar lines so drm shouldn't either.

9 years agoDRMDEBUG not DRM_DEBUG. From frantisek holop.
jsg [Sat, 30 May 2015 18:09:26 +0000 (18:09 +0000)]
DRMDEBUG not DRM_DEBUG.  From frantisek holop.

9 years agoPlug a memory leak by simplifying kv_free(). By checking the type and
claudio [Sat, 30 May 2015 09:47:25 +0000 (09:47 +0000)]
Plug a memory leak by simplifying kv_free(). By checking the type and
returning early the key and value memory got leaked on HTTP header kvs
since their type was never set.
OK benno@

9 years agoRemove fields from struct that are not used. OK benno@
claudio [Sat, 30 May 2015 09:38:11 +0000 (09:38 +0000)]
Remove fields from struct that are not used. OK benno@

9 years agoNative atomic operations for i386.
kettenis [Sat, 30 May 2015 08:41:30 +0000 (08:41 +0000)]
Native atomic operations for i386.

ok deraadt@, guenther@, dlg@

9 years agoAcquire/release the i2c bus before/after reading the temperature register.
kettenis [Sat, 30 May 2015 08:39:05 +0000 (08:39 +0000)]
Acquire/release the i2c bus before/after reading the temperature register.
Prevents concurrent access to the smu(4) microcontroller on Apple G5
machines, which would result in errors reading the RTC.

ok deraadt@, mpi@

9 years agoSet the usb otg port on the cubox to host mode and attach ehci to it.
jsg [Sat, 30 May 2015 08:09:19 +0000 (08:09 +0000)]
Set the usb otg port on the cubox to host mode and attach ehci to it.
While ehci attaches, devices don't seem to be recognised in the otg port.
From Patrick Wildt in bitrig.

9 years agoSupport - to read from stdin, from Sebastien Marie.
nicm [Sat, 30 May 2015 06:25:35 +0000 (06:25 +0000)]
Support - to read from stdin, from Sebastien Marie.

9 years agofix usage();
jmc [Sat, 30 May 2015 06:24:59 +0000 (06:24 +0000)]
fix usage();

9 years agoSet the sdmmc(4) emmc highspeed flag if the capability register claims
jsg [Sat, 30 May 2015 03:20:54 +0000 (03:20 +0000)]
Set the sdmmc(4) emmc highspeed flag if the capability register claims
it is supported.

9 years agocleanup attach message
deraadt [Sat, 30 May 2015 02:49:23 +0000 (02:49 +0000)]
cleanup attach message

9 years agofix the build when SDHC_DEBUG is defined
jsg [Sat, 30 May 2015 02:17:36 +0000 (02:17 +0000)]
fix the build when SDHC_DEBUG is defined

9 years agoDon't use special strings if #() commands fail, just remove the format
nicm [Fri, 29 May 2015 23:26:52 +0000 (23:26 +0000)]
Don't use special strings if #() commands fail, just remove the format
(as if the command produced nothing). Makes constructions that can fail
like '#(test whatever && echo foo)' work as they did before.

9 years agoExpand formats again inside #(), and free the temporaries.
nicm [Fri, 29 May 2015 23:12:38 +0000 (23:12 +0000)]
Expand formats again inside #(), and free the temporaries.

9 years agoUse RB_MIN to get the lowest index for the current window when creating
nicm [Fri, 29 May 2015 23:02:27 +0000 (23:02 +0000)]
Use RB_MIN to get the lowest index for the current window when creating
grouped sessions, rather than using RB_ROOT.

9 years agoSwitch vax to bunnytils 2.17.
miod [Fri, 29 May 2015 20:30:49 +0000 (20:30 +0000)]
Switch vax to bunnytils 2.17.

9 years agoMake ld.so on alpha cope with binaries built with secureplt (where secure here
miod [Fri, 29 May 2015 19:12:26 +0000 (19:12 +0000)]
Make ld.so on alpha cope with binaries built with secureplt (where secure here
means read-only, which our ld.so already enforced, but a smaller plt section).

9 years agomillert points out I can just call tzset() early to avoid having to
nicm [Fri, 29 May 2015 15:58:34 +0000 (15:58 +0000)]
millert points out I can just call tzset() early to avoid having to
permit access() in the child.

9 years agoUse gmtime not localtime when we want UTC, I got it right in one place
nicm [Fri, 29 May 2015 15:58:01 +0000 (15:58 +0000)]
Use gmtime not localtime when we want UTC, I got it right in one place
but not in another two. I expect I was confused by the fact that the
original file(1) had them the wrong way round (so, for example, leqldate
was actually UTC).

9 years agodo not need non-NULL-check before free(p), other minor refactoring
deraadt [Fri, 29 May 2015 15:57:36 +0000 (15:57 +0000)]
do not need non-NULL-check before free(p), other minor refactoring
from Benjamin Baier

9 years agoremove duplicate assignments; from Benjamin Baier
deraadt [Fri, 29 May 2015 15:53:46 +0000 (15:53 +0000)]
remove duplicate assignments; from Benjamin Baier

9 years agoUpdate file(1) regress (less verbose ELF output).
nicm [Fri, 29 May 2015 15:12:11 +0000 (15:12 +0000)]
Update file(1) regress (less verbose ELF output).

9 years agoDon't stop at the first successful test unless it has generated some
nicm [Fri, 29 May 2015 15:11:08 +0000 (15:11 +0000)]
Don't stop at the first successful test unless it has generated some
output.

9 years agoThe case-insensitive flag (c) for string and search tests doesn't make
nicm [Fri, 29 May 2015 14:52:10 +0000 (14:52 +0000)]
The case-insensitive flag (c) for string and search tests doesn't make
the entire match case sensitive, only lower case characters.

9 years agoI got confused and made the strength multiplier 20, it should be 10.
nicm [Fri, 29 May 2015 14:15:41 +0000 (14:15 +0000)]
I got confused and made the strength multiplier 20, it should be 10.

9 years agoNuke annoying whitespace nits to shrink some future diffs.
krw [Fri, 29 May 2015 13:54:26 +0000 (13:54 +0000)]
Nuke annoying whitespace nits to shrink some future diffs.

9 years agoNuke annoying whitespace nits to shrink some future diffs.
krw [Fri, 29 May 2015 13:48:45 +0000 (13:48 +0000)]
Nuke annoying whitespace nits to shrink some future diffs.

9 years agofile(1) needs access(/etc/localtime) for localtime().
nicm [Fri, 29 May 2015 12:38:28 +0000 (12:38 +0000)]
file(1) needs access(/etc/localtime) for localtime().

9 years agoOnly mmap() on regular files; mmap() on /dev/stdin will happily map as
nicm [Fri, 29 May 2015 12:33:41 +0000 (12:33 +0000)]
Only mmap() on regular files; mmap() on /dev/stdin will happily map as
much as we ask for but only the first page will be usable. (We could get
the actual size with ioctl(FIONREAD) and mmap() that but it would need
to be done in the parent - I think just using read() is simpler.)

9 years agoMake things a little less confusing by making fill_buffer self-contained
nicm [Fri, 29 May 2015 11:59:01 +0000 (11:59 +0000)]
Make things a little less confusing by making fill_buffer self-contained
and return both the buffer and amount filled rather than having it
modify the input struct and return a value.

9 years agoIf reading into buffer, correct the stored file size at EOF, we don't
nicm [Fri, 29 May 2015 11:03:37 +0000 (11:03 +0000)]
If reading into buffer, correct the stored file size at EOF, we don't
want to look at any garbage that might already be in the buffer after
that. From Sebastien Marie.

9 years agoPut END() where appropriate.
uebayasi [Fri, 29 May 2015 09:25:28 +0000 (09:25 +0000)]
Put END() where appropriate.

While here, kill redundant use of _C_LABEL() in ENTRY().

9 years agoPut obvious END() macros that match ENTRY() entries.
uebayasi [Fri, 29 May 2015 08:50:12 +0000 (08:50 +0000)]
Put obvious END() macros that match ENTRY() entries.

9 years agofix a possible off-by-one when reading /etc/hosts if it doesn't end
eric [Fri, 29 May 2015 08:49:37 +0000 (08:49 +0000)]
fix a possible off-by-one when reading /etc/hosts if it doesn't end
with a newline.

ok jca@

9 years agoSilently fall back to /etc/magic if ~/.magic can't be opened, ok deraadt stsp
nicm [Fri, 29 May 2015 07:30:28 +0000 (07:30 +0000)]
Silently fall back to /etc/magic if ~/.magic can't be opened, ok deraadt stsp

9 years agoSprinkle END() in some straightforward *.S files that have ENTRY(). The
uebayasi [Fri, 29 May 2015 07:21:09 +0000 (07:21 +0000)]
Sprinkle END() in some straightforward *.S files that have ENTRY().  The
resulting *.o have "FUNC" symbols with size set.

9 years agoput back SSH1 for now - we want to make sure we don't regress
djm [Fri, 29 May 2015 06:47:43 +0000 (06:47 +0000)]
put back SSH1 for now - we want to make sure we don't regress
it for the 6.9 release; ok deraadt@

9 years agoadd some more cortex A ids
jsg [Fri, 29 May 2015 05:48:07 +0000 (05:48 +0000)]
add some more cortex A ids

9 years agoNeed to operate of CXXFLAGS now.
miod [Fri, 29 May 2015 04:17:11 +0000 (04:17 +0000)]
Need to operate of CXXFLAGS now.

9 years agomake this work without SUDO set; ok dtucker@
djm [Fri, 29 May 2015 03:05:13 +0000 (03:05 +0000)]
make this work without SUDO set; ok dtucker@

9 years agoenable agtimer required for cortex a15
jsg [Fri, 29 May 2015 02:44:14 +0000 (02:44 +0000)]
enable agtimer required for cortex a15

9 years agoRemove unused bus space tags/handles. The generic timer uses the cp15
jsg [Fri, 29 May 2015 02:35:43 +0000 (02:35 +0000)]
Remove unused bus space tags/handles.  The generic timer uses the cp15
coprocessor space.

9 years agoRevert unrelated changes in previous.
uebayasi [Fri, 29 May 2015 00:37:10 +0000 (00:37 +0000)]
Revert unrelated changes in previous.

9 years agoInitial addition of ``Patrol Read'' support in bio(4), biocto(8), and
uebayasi [Fri, 29 May 2015 00:33:37 +0000 (00:33 +0000)]
Initial addition of ``Patrol Read'' support in bio(4), biocto(8), and
mfi(4).  Based on FreeBSD, but done without mfiutil(8).

OK deraadt@

9 years agodetect crashes from constraint sub-processes, instead of ignoring them.
deraadt [Thu, 28 May 2015 21:34:36 +0000 (21:34 +0000)]
detect crashes from constraint sub-processes, instead of ignoring them.
if this happens, we want to tear down all of ntpd, so that people will
report it, any such bug can be found, and fixed.
ok bcook

9 years agowhen machdep.allowaperture sysctl is set to 3, allow concurrent access
jcs [Thu, 28 May 2015 20:53:05 +0000 (20:53 +0000)]
when machdep.allowaperture sysctl is set to 3, allow concurrent access

9 years agoSave the cpuid(6) eax bits in the cpu_info and report the SENSOR and ARAT
guenther [Thu, 28 May 2015 20:10:58 +0000 (20:10 +0000)]
Save the cpuid(6) eax bits in the cpu_info and report the SENSOR and ARAT
bits from it.

ok krw@ kettenis@

9 years agouse "uri"; from yegor timoschenko
jmc [Thu, 28 May 2015 19:29:40 +0000 (19:29 +0000)]
use "uri"; from yegor timoschenko

9 years agoprobably makes more sense to show "file" as being mandatory;
jmc [Thu, 28 May 2015 19:26:37 +0000 (19:26 +0000)]
probably makes more sense to show "file" as being mandatory;
from sebastien marie

9 years agoInvert the binutils 2.17 logic to only list non-2.17 platforms.
miod [Thu, 28 May 2015 18:23:11 +0000 (18:23 +0000)]
Invert the binutils 2.17 logic to only list non-2.17 platforms.

9 years agoSwitch alpha, arm, sh and sparc to binutils 2.17.
miod [Thu, 28 May 2015 18:16:41 +0000 (18:16 +0000)]
Switch alpha, arm, sh and sparc to binutils 2.17.

9 years agoDo not try to unlink the control socket in an unprivileged child
florian [Thu, 28 May 2015 17:09:18 +0000 (17:09 +0000)]
Do not try to unlink the control socket in an unprivileged child
process on shutdown.
Found while working on tame(2).
OK gilles@

9 years agoDo not try to unlink the control socket in an unprivileged child
florian [Thu, 28 May 2015 17:08:08 +0000 (17:08 +0000)]
Do not try to unlink the control socket in an unprivileged child
process on shutdown.
Found while working on tame(2).
OK benno@

9 years agoKill unused IF_INPUT_ENQUEUE().
mpi [Thu, 28 May 2015 11:57:33 +0000 (11:57 +0000)]
Kill unused IF_INPUT_ENQUEUE().

ok dlg@

9 years agowrap all moduli-related code in #ifdef WITH_OPENSSL.
djm [Thu, 28 May 2015 07:37:31 +0000 (07:37 +0000)]
wrap all moduli-related code in #ifdef WITH_OPENSSL.
based on patch from Reuben Hawkins; bz#2388
feedback and ok dtucker@

9 years agoIncrease the allowed length of the known host file name in the log
dtucker [Thu, 28 May 2015 05:41:29 +0000 (05:41 +0000)]
Increase the allowed length of the known host file name in the log
message to be consistent with other cases.  Part of bz#1993, ok deraadt.

9 years agoFix typo (keywork->keyword)
dtucker [Thu, 28 May 2015 05:09:45 +0000 (05:09 +0000)]
Fix typo (keywork->keyword)

9 years agoadd error message on ftruncate failure; bz#2176
djm [Thu, 28 May 2015 04:50:53 +0000 (04:50 +0000)]
add error message on ftruncate failure; bz#2176

9 years agomake ssh-keygen default to ed25519 keys when compiled without
djm [Thu, 28 May 2015 04:40:13 +0000 (04:40 +0000)]
make ssh-keygen default to ed25519 keys when compiled without
OpenSSL; bz#2388, ok dtucker@

9 years agoRemove intermediate copy of the moduli file. "make" will rebuild any
dtucker [Thu, 28 May 2015 00:54:01 +0000 (00:54 +0000)]
Remove intermediate copy of the moduli file.  "make" will rebuild any
missing moduli fragment files (which will be a no-op unless you have
explicitly deleted them) and "make update-moduli" will update the
main file under ${BSDSRCDIR}/etc/moduli.  ok djm@

9 years agoRemove 1k bit groups. ok deraadt@, markus@
dtucker [Thu, 28 May 2015 00:03:06 +0000 (00:03 +0000)]
Remove 1k bit groups.  ok deraadt@, markus@

9 years agoRemove 1k bit groups. ok deraadt@, markus@
dtucker [Thu, 28 May 2015 00:00:24 +0000 (00:00 +0000)]
Remove 1k bit groups.  ok deraadt@, markus@

9 years agoReorder client proposal to prefer diffie-hellman-group-exchange-sha1 over
dtucker [Wed, 27 May 2015 23:51:10 +0000 (23:51 +0000)]
Reorder client proposal to prefer diffie-hellman-group-exchange-sha1 over
diffie-hellman-group14-sha1.  ok djm@

9 years agoAdd a stronger (4k bit) fallback group that sshd can use when the moduli
dtucker [Wed, 27 May 2015 23:39:18 +0000 (23:39 +0000)]
Add a stronger (4k bit) fallback group that sshd can use when the moduli
file is missing or broken, sourced from RFC3526.  bz#2302, ok markus@
(earlier version), djm@

9 years agoUse m_defrag(9) instead of rolling our own inlined version.
kettenis [Wed, 27 May 2015 22:10:52 +0000 (22:10 +0000)]
Use m_defrag(9) instead of rolling our own inlined version.

ok mikeb@

9 years agoMove the jobs output cache into the formats code so that #() work more
nicm [Wed, 27 May 2015 13:28:04 +0000 (13:28 +0000)]
Move the jobs output cache into the formats code so that #() work more
generally (for example, again working in set-titles-string).

9 years agocheck if the packet is for us or if we're promisc before we cut the
dlg [Wed, 27 May 2015 12:23:44 +0000 (12:23 +0000)]
check if the packet is for us or if we're promisc before we cut the
vlan header out of the packet.

fixes rx on top of trunks on top of nics that dont do hw vlan tagging.

ok mpi@

9 years agoImprove the controller state check in xhci_reset
mikeb [Wed, 27 May 2015 11:13:34 +0000 (11:13 +0000)]
Improve the controller state check in xhci_reset

From FreeBSD, OK mpi

9 years agocall exynos_platform_match() if the kernel is compiled with exynos devices
jsg [Wed, 27 May 2015 08:03:43 +0000 (08:03 +0000)]
call exynos_platform_match() if the kernel is compiled with exynos devices

9 years agosupport PKCS#11 devices with external PIN entry devices
djm [Wed, 27 May 2015 05:15:02 +0000 (05:15 +0000)]
support PKCS#11 devices with external PIN entry devices
bz#2240, based on patch from Dirk-Willem van Gulik;
feedback and ok dtucker@

9 years agomake the exynos code compile without fdt
jsg [Wed, 27 May 2015 00:06:14 +0000 (00:06 +0000)]
make the exynos code compile without fdt
ok bmercer@

9 years agomake this build with WSDISPLAY_COMPAT_RAWKBD defined
jsg [Tue, 26 May 2015 23:47:04 +0000 (23:47 +0000)]
make this build with WSDISPLAY_COMPAT_RAWKBD defined
ok bmercer@

9 years agoCap DH-GEX group size at 4kbits for Cisco implementations. Some of them
dtucker [Tue, 26 May 2015 23:23:40 +0000 (23:23 +0000)]
Cap DH-GEX group size at 4kbits for Cisco implementations.  Some of them
will choke when asked for preferred sizes >4k instead of returning the 4k
group that they do have.  bz#2209, ok djm@

9 years agoBring back `1:' label removed by mistake in 1.7. Complained loudly about by gas
miod [Tue, 26 May 2015 20:23:39 +0000 (20:23 +0000)]
Bring back `1:' label removed by mistake in 1.7. Complained loudly about by gas
2.17, while 2.15 didn't care, which just caused a cold shiver accross my spine.

9 years agoBetter cross-compilation check
miod [Tue, 26 May 2015 20:15:54 +0000 (20:15 +0000)]
Better cross-compilation check

9 years agoUse a relative path against BSDOBJDIR to pick libcrypto; makes cross-lib
miod [Tue, 26 May 2015 19:38:13 +0000 (19:38 +0000)]
Use a relative path against BSDOBJDIR to pick libcrypto; makes cross-lib
work again.

9 years agosimply use _PATH_HOSTS where appropriate
eric [Tue, 26 May 2015 19:28:57 +0000 (19:28 +0000)]
simply use _PATH_HOSTS where appropriate

9 years agoFix missing vlan.h include in if_vio.c
sf [Tue, 26 May 2015 19:12:24 +0000 (19:12 +0000)]
Fix missing vlan.h include in if_vio.c

patch by jsg@

9 years agoImprove diffability against m68k.
miod [Tue, 26 May 2015 18:57:08 +0000 (18:57 +0000)]
Improve diffability against m68k.

9 years agoExposes number of bytes processed by the sound card (as returned
ratchov [Tue, 26 May 2015 18:17:12 +0000 (18:17 +0000)]
Exposes number of bytes processed by the sound card (as returned
by AUDIO_GET{I,O}OFFS) instead of the number of bytes processed
minus the xruns (as returned by AUDIO_GETINFO).

ok armani

9 years agoDon't create ICMP states on reply packets unless tracking states sloppy
mikeb [Tue, 26 May 2015 16:17:51 +0000 (16:17 +0000)]
Don't create ICMP states on reply packets unless tracking states sloppy

Since we've strengthened the ICMP state matching procedure during lookup
to only match packets against states set up in a particular direction, we
need to make sure we don't create states on packets that would otherwise
be flowing in the direction opposite to the direction of the state and
prevent further packets from matching the created state due to strict
rules imposed by the ICMP direction check.

Problem reported by Alexandr Nedvedicky, alexandr.nedvedicky-at-oracle.com.
Discussed with reyk@;  OK henning

9 years agoUse if_output() instead of rerolling it.
mpi [Tue, 26 May 2015 15:34:00 +0000 (15:34 +0000)]
Use if_output() instead of rerolling it.

ok stsp@

9 years agoCreate aliases.db from the installed aliases file, so we get the correct
naddy [Tue, 26 May 2015 14:34:59 +0000 (14:34 +0000)]
Create aliases.db from the installed aliases file, so we get the correct
owner and group.  Reported by Mark Patruck.  ok deraadt@ miod@

9 years agobump the number of tx and rx descriptors from 128 up to 512.
dlg [Tue, 26 May 2015 12:29:42 +0000 (12:29 +0000)]
bump the number of tx and rx descriptors from 128 up to 512.

9 years agoStore the IP address of the corresponding ifa in the rt_gateway field
mpi [Tue, 26 May 2015 12:19:51 +0000 (12:19 +0000)]
Store the IP address of the corresponding ifa in the rt_gateway field
of RTF_CLONING and RTF_BROASCAST routes to not create MPATH conflicts
when IP address aliases are used.

This change makes it possible to have multiple RTF_CLONING routes with
the same priority.  Note that any of the existing RTF_CLONING route
might be used by the kernel to create a RTF_CLONED route which should
not be a problem with aliases since they are attached to the same ifp.

This unbreak address aliases since the kernel supports multiple connected
routes for a subnet.

Found the hardway by djm@, ok claudio@

9 years agoNormalize route destination before checking for MPATH conflicts.
mpi [Tue, 26 May 2015 12:02:11 +0000 (12:02 +0000)]
Normalize route destination before checking for MPATH conflicts.

This makes rt_mpath_conflict() work as expected when adding routes
with the same destination and the same netmask.

With and ok claudio@

9 years agoDo not create ARP entries for RTF_BROADCAST routes.
mpi [Tue, 26 May 2015 11:55:34 +0000 (11:55 +0000)]
Do not create ARP entries for RTF_BROADCAST routes.

This has been done because historically routes to broadcast addresses
were cloned like any ARP entry.  But for obvious reasons, no matching
Ethernet address could ever be resolved.  That's why we played tricks
with the expire timer.

Now that a RTF_BROADCAST route is created per configured IPv4 address,
we need to differenciate duplicated one.  And by not creating an ARP
entry we are allowed to write the IP address in the rt_gateway field,
which prevents MPATH conflicts.

This change is part of a fix to unbreak aliases since the kernel support
multiple connected routes for a subnet.

Found the hardway by djm@, ok claudio@