openbsd
7 years agotweak previous;
jmc [Tue, 30 May 2017 12:24:43 +0000 (12:24 +0000)]
tweak previous;

7 years agoremove an extra .Ed;
jmc [Tue, 30 May 2017 12:17:17 +0000 (12:17 +0000)]
remove an extra .Ed;

7 years agotweak previous; ok bluhm
jmc [Tue, 30 May 2017 12:14:38 +0000 (12:14 +0000)]
tweak previous; ok bluhm

7 years agofix a comment
henning [Tue, 30 May 2017 12:13:04 +0000 (12:13 +0000)]
fix a comment

7 years agoCarp balancing ip does not work since there is a mac filter in
friehm [Tue, 30 May 2017 12:09:27 +0000 (12:09 +0000)]
Carp balancing ip does not work since there is a mac filter in
ether_input(). Now we use mbuf tags instead of modifying the MAC
address.

ok mpi@

7 years agoIntroduce a scary rc.conf(8) knob library_aslr=(YES|NO) to turn off the
tb [Tue, 30 May 2017 12:04:26 +0000 (12:04 +0000)]
Introduce a scary rc.conf(8) knob library_aslr=(YES|NO) to turn off the
reordering of libraries by rc(8). This way machines with very slow disk I/O
have a chance of booting within reasonable time now that libcrypto is also
randomized.

Discussed with various;
input & ok from deraadt ajacoutot

7 years agoLink iked test to build so that obj directories are created.
bluhm [Tue, 30 May 2017 12:03:11 +0000 (12:03 +0000)]
Link iked test to build so that obj directories are created.

7 years agoRemove the .INTERRUPT target, it prevented to terminate the test
bluhm [Tue, 30 May 2017 11:57:59 +0000 (11:57 +0000)]
Remove the .INTERRUPT target, it prevented to terminate the test
with Ctrl-C in some situations.

7 years agovirtio: Put entries in freelist in reverse order
sf [Tue, 30 May 2017 11:02:39 +0000 (11:02 +0000)]
virtio: Put entries in freelist in reverse order

This avoids breaking krw@'s ongoing work on improving virtio_enqueue_trim().

7 years agoUse ieee80211_{min,max}_basic_rate() in iwn(4).
stsp [Tue, 30 May 2017 11:01:38 +0000 (11:01 +0000)]
Use ieee80211_{min,max}_basic_rate() in iwn(4).
ok mpi@ phessler@

7 years agoIntroduce ieee80211_min_basic_rate() and ieee80211_max_basic_rate().
stsp [Tue, 30 May 2017 11:00:38 +0000 (11:00 +0000)]
Introduce ieee80211_min_basic_rate() and ieee80211_max_basic_rate().

These helpers can be used by drivers to improve compatibility with APs
that disable some mandatory PHY rates in the basic rate set.

For instance, many of our drivers hard-code 11b rates on 2 Ghz and run
into problems when APs disable them. Since 11b rates are being disabled
by default by some vendors, hardcoding them is not a good idea anymore.

ok mpi@ phessler@

7 years agoAlways set the link state DOWN once we enter ieee80211_newstate(), regardless
stsp [Tue, 30 May 2017 10:50:33 +0000 (10:50 +0000)]
Always set the link state DOWN once we enter ieee80211_newstate(), regardless
of whether the wifi interface happens to be leaving RUN state. The interface
is never usable during state transitions so setting the link DOWN is the only
reasonable option when any transition is triggered.

Fixes a problem where, at boot time, the link state of wifi interfaces was
reported to userland as UNKNOWN (which, curiously, has value 0). dhclient's
link detection logic was recently changed from ifmedia to getifaddrs which
exposed the UNKOWN link state. Since dhclient assumes an UNKNOWN link state
means UP it would start trying to negotiate a lease too early during boot.

Problem reported by tb@
ok krw@

7 years agoext2fs: Mark superblock as not modified when written
sf [Tue, 30 May 2017 10:32:53 +0000 (10:32 +0000)]
ext2fs: Mark superblock as not modified when written

I have seen spurious "file system not clean; please fsck(8)" warnings
during "mount -ur". Set e2fs_fmod = 0 when writing the superblock (as
ffs does).

"Makes sense" deraadt@

7 years agoreduce lies
florian [Tue, 30 May 2017 10:24:05 +0000 (10:24 +0000)]
reduce lies

7 years agosync
fcambus [Tue, 30 May 2017 10:20:55 +0000 (10:20 +0000)]
sync

7 years agoRemove lucida16x29.h and omron12x20 fonts.
fcambus [Tue, 30 May 2017 10:19:22 +0000 (10:19 +0000)]
Remove lucida16x29.h and omron12x20 fonts.

- lucida16x29 is a bad conversion from a truetype font, it is too bold
which makes it unreadable.
- omron12x20 is a serif font, and is very difficult to read due to
inconsistencies in character spacing, alignement, and width.

Moreover, both fonts only have 96 glyphs.

No objection from mlarkin@

OK mpi@, jcs@

7 years agoEnable more error codes for BGP. These were commited some time ago, but
akfaew [Tue, 30 May 2017 09:40:08 +0000 (09:40 +0000)]
Enable more error codes for BGP. These were commited some time ago, but
never enabled. Also add error codes for FSM.

With input from Job Snijders.

OK phessler@ (previous version), claudio@, deraadt@

7 years agoTeach ypldap to use tls when connecting to ldap servers.
jmatthew [Tue, 30 May 2017 09:33:31 +0000 (09:33 +0000)]
Teach ypldap to use tls when connecting to ldap servers.
libtls help from jsing@, linker help from deraadt@

7 years agoKeep ipv6_input() definition inside _KERNEL.
mpi [Tue, 30 May 2017 09:10:49 +0000 (09:10 +0000)]
Keep ipv6_input() definition inside _KERNEL.

Unbreak userland, found the hardway by deraadt@

7 years agoEnable slaacd(8) by default and disable router solicitation and
florian [Tue, 30 May 2017 08:58:34 +0000 (08:58 +0000)]
Enable slaacd(8) by default and disable router solicitation and
advertisement processing in the kernel.
Go for it!!! deraadt@
additional encouragement to push forward from at least mpi and henning
special thanks to naddy for being an early adopter and finding bugs.

7 years agoswitch from Key typedef with struct sshkey; ok djm@
markus [Tue, 30 May 2017 08:52:19 +0000 (08:52 +0000)]
switch from Key typedef with struct sshkey; ok djm@

7 years agoremove ssh1 references; ok djm@
markus [Tue, 30 May 2017 08:49:58 +0000 (08:49 +0000)]
remove ssh1 references; ok djm@

7 years agorevise sshkey_load_public(): remove ssh1 related comments, remove extra
markus [Tue, 30 May 2017 08:49:32 +0000 (08:49 +0000)]
revise sshkey_load_public(): remove ssh1 related comments, remove extra
open()/close() on keyfile, prevent leak of 'pub' if 'keyp' is NULL,
replace strlcpy+cat with asprintf; ok djm@

7 years agofileops test: Also run on ext2fs
sf [Tue, 30 May 2017 08:44:58 +0000 (08:44 +0000)]
fileops test: Also run on ext2fs

Also sort subdirs in makefile

7 years agovirtio: Do LIFO in the freelist
sf [Tue, 30 May 2017 08:35:32 +0000 (08:35 +0000)]
virtio: Do LIFO in the freelist

Use a SLIST instead of a SIMPLEQ and use LIFO instead of FIFO. This should
improve cache usage.

7 years agoRemove the unused pcdisplay_mapchar_simple function in pcdisplay(4).
fcambus [Tue, 30 May 2017 08:24:56 +0000 (08:24 +0000)]
Remove the unused pcdisplay_mapchar_simple function in pcdisplay(4).

It has been unused since this code was imported from NetBSD.

OK mlarkin@

7 years agoPreserve search string when entering prompt again.
nicm [Tue, 30 May 2017 08:13:48 +0000 (08:13 +0000)]
Preserve search string when entering prompt again.

7 years agoteach pf_build_tcp() about SACK, ok & with sashan
henning [Tue, 30 May 2017 08:10:01 +0000 (08:10 +0000)]
teach pf_build_tcp() about SACK, ok & with sashan

7 years agoIntroduce ipv{4,6}_input(), two wrappers around IP queues.
mpi [Tue, 30 May 2017 07:50:37 +0000 (07:50 +0000)]
Introduce ipv{4,6}_input(), two wrappers around IP queues.

This will help transitionning to an un-KERNEL_LOCK()ed IP
forwarding path.

Disucssed with bluhm@, ok claudio@

7 years agoUse memmove() instead of memcpy(), buffers are overlapping.
mpi [Tue, 30 May 2017 07:40:24 +0000 (07:40 +0000)]
Use memmove() instead of memcpy(), buffers are overlapping.

From Coverity via NetBSD via miod@

ok claudio@, tedu@, deraadt@, krw@

7 years agoThe is open-file command is actually called find-file. Typo pointed out
tb [Tue, 30 May 2017 07:11:40 +0000 (07:11 +0000)]
The is open-file command is actually called find-file.  Typo pointed out
by "Choose a display name <thelocals_job_applicant protonmail com>".

7 years agofix -Wshadow warnings.
florian [Tue, 30 May 2017 07:05:22 +0000 (07:05 +0000)]
fix -Wshadow warnings.
Found by bcallah, thanks!

7 years agoUnbreak previous for git diffs that do not have the a/ prefix.
florian [Tue, 30 May 2017 06:55:40 +0000 (06:55 +0000)]
Unbreak previous for git diffs that do not have the a/ prefix.
Found the hard way be me, OK tedu

7 years agoClear the internal table used by if_get(9) and sleep until all
mpi [Tue, 30 May 2017 06:42:13 +0000 (06:42 +0000)]
Clear the internal table used by if_get(9) and sleep until all
remaining ifp references are released *without* the NET_LOCK().

It's safe to do so because the KERNEL_LOCK() serializes accesses
to ``if_map''.

More importantly this fix possible deadlocks between if_get() and
the NET_LOCK().  It is now possible to call them in whatever order.

ok visa@, dlg@

7 years agoFix a race by always modifying the 'counter' variable while holding the
mpi [Tue, 30 May 2017 06:38:10 +0000 (06:38 +0000)]
Fix a race by always modifying the 'counter' variable while holding the
mutex.

ok tedu@

7 years agoavoid some shadow warnings. from Brian Callahan
tedu [Tue, 30 May 2017 06:01:30 +0000 (06:01 +0000)]
avoid some shadow warnings. from Brian Callahan

7 years agofix a shadow warning. from Brian Callahan
tedu [Tue, 30 May 2017 05:58:44 +0000 (05:58 +0000)]
fix a shadow warning. from Brian Callahan

7 years agorename variable to avoid shadow. from Brian Callahan
tedu [Tue, 30 May 2017 05:57:46 +0000 (05:57 +0000)]
rename variable to avoid shadow. from Brian Callahan

7 years agotypo - extra semicolon
mlarkin [Tue, 30 May 2017 04:45:38 +0000 (04:45 +0000)]
typo - extra semicolon

7 years agomissing .El;
jmc [Mon, 29 May 2017 21:27:36 +0000 (21:27 +0000)]
missing .El;

7 years agoEnable futex-based mutex and condvar.
mpi [Mon, 29 May 2017 21:19:30 +0000 (21:19 +0000)]
Enable futex-based mutex and condvar.

ok everybody

7 years ago*** empty log message ***
markus [Mon, 29 May 2017 20:57:21 +0000 (20:57 +0000)]
*** empty log message ***

7 years agofuzz the iked payload parser using the openssh unit-test framework
markus [Mon, 29 May 2017 20:59:28 +0000 (20:59 +0000)]
fuzz the iked payload parser using the openssh unit-test framework
from hshoexer

7 years agoAdd a flag to stop the prompt input being expanded.
nicm [Mon, 29 May 2017 20:42:53 +0000 (20:42 +0000)]
Add a flag to stop the prompt input being expanded.

7 years agoStore a copy of the old status line, will be needed soon for new choose mode.
nicm [Mon, 29 May 2017 20:41:29 +0000 (20:41 +0000)]
Store a copy of the old status line, will be needed soon for new choose mode.

7 years agoFunction to count clients.
nicm [Mon, 29 May 2017 20:37:30 +0000 (20:37 +0000)]
Function to count clients.

7 years agoAdd more regress guidelines to bsd.regress.mk(5) man page.
bluhm [Mon, 29 May 2017 20:35:47 +0000 (20:35 +0000)]
Add more regress guidelines to bsd.regress.mk(5) man page.
input schwarze@ jmc@; OK henning@ benno@

7 years agoThe kernel sets IN6_IFF_AUTOCONF for autoconf addresses, also make
florian [Mon, 29 May 2017 20:31:31 +0000 (20:31 +0000)]
The kernel sets IN6_IFF_AUTOCONF for autoconf addresses, also make
slaacd(8) do that. At least you can identify them in ifconfig output.
Difference pointed out by & OK naddy

7 years agoReplace handrolled list with LIST_* macros. OK mpi@
claudio [Mon, 29 May 2017 20:31:12 +0000 (20:31 +0000)]
Replace handrolled list with LIST_* macros. OK mpi@

7 years agoAllow passing in IN6_IFF_AUTOCONF from userland. Needed by slaacd(8)
florian [Mon, 29 May 2017 20:28:57 +0000 (20:28 +0000)]
Allow passing in IN6_IFF_AUTOCONF from userland. Needed by slaacd(8)
OK naddy

7 years agochange section title in bgpd.conf talking about MPLS,
benno [Mon, 29 May 2017 20:23:18 +0000 (20:23 +0000)]
change section title in bgpd.conf talking about MPLS,
add a bit about working with multiple bgpds in rdomains in bgpctl.
shouting from claudio@ and help from jmc@

7 years agoAdd ||, && format operators and C: to search pane content.
nicm [Mon, 29 May 2017 18:06:34 +0000 (18:06 +0000)]
Add ||, && format operators and C: to search pane content.

7 years agocvs_file_classify() was making the mistake of only taking over a sticky
joris [Mon, 29 May 2017 17:18:56 +0000 (17:18 +0000)]
cvs_file_classify() was making the mistake of only taking over a sticky
tag from Entries for a given file if the sticky tag is present.

while changing this we can simplify some logic in update.c on how it
decides what tag to use for file classification.

7 years agoAdd m: for fnmatch(3) format matching.
nicm [Mon, 29 May 2017 15:43:48 +0000 (15:43 +0000)]
Add m: for fnmatch(3) format matching.

7 years agosync
deraadt [Mon, 29 May 2017 15:21:12 +0000 (15:21 +0000)]
sync

7 years agoSPINLOCK_SPIN_HOOK is no more, define our own set of macros.
mpi [Mon, 29 May 2017 14:47:54 +0000 (14:47 +0000)]
SPINLOCK_SPIN_HOOK is no more, define our own set of macros.

Prodded by kettenis@ and tedu@

7 years agoRemove unused zclose function.
fcambus [Mon, 29 May 2017 14:41:16 +0000 (14:41 +0000)]
Remove unused zclose function.

From Brian Callahan, thanks!

OK tb@

7 years agoopenbsd does not use nonexec stack segment header hints.
tedu [Mon, 29 May 2017 14:37:51 +0000 (14:37 +0000)]
openbsd does not use nonexec stack segment header hints.
in fact, false is a much better default for UsesNonexecutableStackSection.
platforms that require it can reenable, instead, saving the rest of us an
unnecessary program header (causes trouble for some special binaries).
ok kettenis

7 years agoPer-interface list of addresses, both multicast and unicast, are
mpi [Mon, 29 May 2017 14:36:22 +0000 (14:36 +0000)]
Per-interface list of addresses, both multicast and unicast, are
currently protected by the NET_LOCK().

They are not accessed in the hot path, so protecting them with a
mutex could be an option.  However since we're now going to run
with a NET_LOCK() for some time, assert that it is held.

IPsec is not yet ready to run without KERNEL_LOCK(), so assert it
is held, even in the forwarding path.

Tested by sthen@, ok visa@, claudio@, bluhm@

7 years agoKill struct pfkey_version and move struct pfkeyv2_socket & dump_state
claudio [Mon, 29 May 2017 14:28:01 +0000 (14:28 +0000)]
Kill struct pfkey_version and move struct pfkeyv2_socket & dump_state
to pfkeyv2.c. These structs are nowhere else needed.
OK gcc

7 years agofix "depend on" when carp interface goes from MASTER into state INIT
benno [Mon, 29 May 2017 14:22:51 +0000 (14:22 +0000)]
fix "depend on" when carp interface goes from MASTER into state INIT
aka UNKNOWN.
ok claudio@ phessler@ and henning helped along too

7 years agoKill SPINLOCK_SPIN_HOOK, use CPU_BUSY_CYCLE() instead.
mpi [Mon, 29 May 2017 14:19:49 +0000 (14:19 +0000)]
Kill SPINLOCK_SPIN_HOOK, use CPU_BUSY_CYCLE() instead.

ok visa@, kettenis@

7 years agoexport_pflow() is no longer grabbing the NET_LOCK(), so no need to
mpi [Mon, 29 May 2017 14:18:32 +0000 (14:18 +0000)]
export_pflow() is no longer grabbing the NET_LOCK(), so no need to
release it beforehand.

ok henning@, benno@

7 years agosd: flush cache when closing writable FD
sf [Mon, 29 May 2017 14:08:49 +0000 (14:08 +0000)]
sd: flush cache when closing writable FD

Don't skip the cache flush until the last opening of the device is
closed. Otherwise, when umounting a writable partition while a different
partition is still mounted read-only, the necessary disk flush may be
delayed for a very long time.

ok krw@ deraadt@

7 years agomsdosfs & ffs: flush cache if updating mount from r/w to r/o
sf [Mon, 29 May 2017 14:07:16 +0000 (14:07 +0000)]
msdosfs & ffs: flush cache if updating mount from r/w to r/o

ok deraadt@

7 years agoAdd an ioctl to tell storage devices to flush their internal caches
sf [Mon, 29 May 2017 14:05:31 +0000 (14:05 +0000)]
Add an ioctl to tell storage devices to flush their internal caches

Currently implemented for wd and sd.
Initially ported from netbsd by pedro@

ok deraadt@

7 years agoReduce code duplication. The local-CPU-only case is now handled
visa [Mon, 29 May 2017 14:03:46 +0000 (14:03 +0000)]
Reduce code duplication. The local-CPU-only case is now handled
without a lock in smp_rendezvous_cpus().

7 years agoMinor regress test cleanup.
bluhm [Mon, 29 May 2017 13:49:40 +0000 (13:49 +0000)]
Minor regress test cleanup.
OK sf@

7 years agoImplement VFS read clustering for MSDOSFS
sf [Mon, 29 May 2017 13:48:12 +0000 (13:48 +0000)]
Implement VFS read clustering for MSDOSFS

This is the reverted commit by mpi@ from msdosfs_vnops.c 1.105 plus some
additional tweaks to fix some cluster/block number confusion that lead
to regressions when seeking past the end of a file.

The original commit message was:

  The logic used in msdosfs_bmap() to loop calling pcbmap() comes from
  FreeBSD and is not really efficient but it is good enough since it is
  only called when generating I/O.

  With this diff I get a 100% improvement when reading big files from a
  crappy USB stick.

  With this and bread_cluster(9) modified to not re-fetch B_CACHED buffers,
  reading large contiguous files with chunk sizes of MAXPHYS is almost as
  fast as physio(9) on the same device.

  For a 'real world' example, when copying music files from a USB stick I
  see a speed jump from 15MB/s on -current to 24Mb/s with this diff.

  While here rename some 'lbn' variables into 'cn' to better reflect what
  we're dealing with.

  Tested by Mathieu, with support from deraadt@

ok mpi@

7 years agozap scanio.h, found by deraadt
tedu [Mon, 29 May 2017 13:30:35 +0000 (13:30 +0000)]
zap scanio.h, found by deraadt

7 years agoAfter some consideration, revert last commit and don't abuse rde_dump_done.
claudio [Mon, 29 May 2017 13:10:40 +0000 (13:10 +0000)]
After some consideration, revert last commit and don't abuse rde_dump_done.
IMSG_CTL_SHOW_RIB_PREFIX case is not using rib_dump_r and so should not use
rde_dump_done, instead send the IMSG_CTL_END msg and free the ctx directly.
This is easier to understand.

7 years agoChange the mmap(2)-based binary history file with lots of magic and a
tb [Mon, 29 May 2017 13:09:17 +0000 (13:09 +0000)]
Change the mmap(2)-based binary history file with lots of magic and a
tendency for corruption to a simpler plaintext version.

To convert your current ksh history to plaintext, issue

fc -ln 1 | sed 's/^ //' > ~/ksh_hist.txt

before upgrading and use ksh_hist.txt as HISTFILE after the upgrade.

Original patch by marco in 2011. Ported to current during g2k16 by me.
Testing, bugfixes and improvements in joint work with natano.

Additional testing by anton and mestre. Includes some tweaks by anton.
Committing now to shake out remaining bugs before 6.2 is cut.

ok deraadt, mestre, anton, sthen

7 years agoAdd hibernate support for nvme(4). This creates a separate IO queue that is
jmatthew [Mon, 29 May 2017 12:58:37 +0000 (12:58 +0000)]
Add hibernate support for nvme(4).  This creates a separate IO queue that is
only active during hibernation, preallocating the the submission and completion
queues during attach as they need to be page-aligned.

tested on an x270 that successfully hibernates now.
ok dlg@

7 years agosync proc.c changes from relayd/httpd
benno [Mon, 29 May 2017 12:56:26 +0000 (12:56 +0000)]
sync proc.c changes from relayd/httpd

7 years agodo not whine if installpath is undef.
espie [Mon, 29 May 2017 12:54:05 +0000 (12:54 +0000)]
do not whine if installpath is undef.

7 years agohook up slaacd and slaacctl.
florian [Mon, 29 May 2017 12:51:56 +0000 (12:51 +0000)]
hook up slaacd and slaacctl.
OK naddy, deraadt

7 years agotweak previous to read a little better;
jmc [Mon, 29 May 2017 12:50:33 +0000 (12:50 +0000)]
tweak previous to read a little better;

7 years agoFix crash in IMSG_CTL_SHOW_RIB_PREFIX handling. The ctx is not added to
claudio [Mon, 29 May 2017 12:48:11 +0000 (12:48 +0000)]
Fix crash in IMSG_CTL_SHOW_RIB_PREFIX handling. The ctx is not added to
the list in that case so don't try to LIST_REMOVE it.
Problem found by benno@

7 years agoavoid Xr to self;
jmc [Mon, 29 May 2017 12:41:57 +0000 (12:41 +0000)]
avoid Xr to self;

7 years agoRefactor locator code so it's a proper object.
espie [Mon, 29 May 2017 12:28:54 +0000 (12:28 +0000)]
Refactor locator code so it's a proper object.
allow  fw_update to short-circuit entirely, thus the specific
treatment of paths is actually less of a special case,
and the code that reaches into /etc/installurl is no longer used
at all in that case.

problem noticed by mlarkin@

7 years agorm xr mailaddr.7
tedu [Mon, 29 May 2017 12:16:50 +0000 (12:16 +0000)]
rm xr mailaddr.7

7 years agoremove mention of Berknet and mailaddr.7
tedu [Mon, 29 May 2017 12:15:39 +0000 (12:15 +0000)]
remove mention of Berknet and mailaddr.7

7 years agoremove mailaddr.7 and its many lies.
tedu [Mon, 29 May 2017 12:13:50 +0000 (12:13 +0000)]
remove mailaddr.7 and its many lies.
ok gilles jmc mpi

7 years agoclang has builtin_memmove. ok deraadt
tedu [Mon, 29 May 2017 12:12:35 +0000 (12:12 +0000)]
clang has builtin_memmove. ok deraadt

7 years agoscanner support died some time ago, the header can be removed too.
tedu [Mon, 29 May 2017 12:10:51 +0000 (12:10 +0000)]
scanner support died some time ago, the header can be removed too.
a gift from miod

7 years agoRe-add line continuation \ that went missing in the previous commit
sf [Mon, 29 May 2017 11:59:32 +0000 (11:59 +0000)]
Re-add line continuation \ that went missing in the previous commit

7 years agoInvalidate OCTEON icache after TLB ASID wraparound, otherwise userspace
visa [Mon, 29 May 2017 11:46:49 +0000 (11:46 +0000)]
Invalidate OCTEON icache after TLB ASID wraparound, otherwise userspace
might execute stale instructions. The cache is virtually tagged and does
not seem to pick up TLB alterations automatically.

7 years agosort SEE ALSO;
jmc [Mon, 29 May 2017 11:13:09 +0000 (11:13 +0000)]
sort SEE ALSO;

7 years agosort SEE ALSO;
jmc [Mon, 29 May 2017 11:10:29 +0000 (11:10 +0000)]
sort SEE ALSO;

7 years agoRegress test for sparse files & mmap with different FSs
sf [Mon, 29 May 2017 11:01:16 +0000 (11:01 +0000)]
Regress test for sparse files & mmap with different FSs

This test creates a file with a hole in the middle and then checks that
when reading or mmaping the whole file, the data is correct.

This catches the bug introduced by the
"Implement VFS read clustering for MSDOSFS" commit that hat to be
reverted.

For now, we run the test with FAT16, FAT32, FFS

With much advice from bluhm@

7 years agoRegress test for sparse files & mmap with different FSs
sf [Mon, 29 May 2017 11:00:04 +0000 (11:00 +0000)]
Regress test for sparse files & mmap with different FSs

This test creates a file with a hole in the middle and then checks that
when reading or mmaping the whole file, the data is correct.

This catches the bug introduced by the
"Implement VFS read clustering for MSDOSFS" commit that hat to be
reverted.

For now, we run the test with FAT16, FAT32, FFS

With much advice from bluhm@

7 years agoPFKEY version 2 is the only pfkey version supported. No need for extra
claudio [Mon, 29 May 2017 10:55:34 +0000 (10:55 +0000)]
PFKEY version 2 is the only pfkey version supported. No need for extra
abstraction. First step of making PF_KEY a bit more like PF_ROUTE.
OK mpi@

7 years agorc.d(8) for slaacd
florian [Mon, 29 May 2017 10:24:06 +0000 (10:24 +0000)]
rc.d(8) for slaacd
OK phessler, deraadt

7 years agochange bgpd's control socket from /var/run/bgpd.sock to
benno [Mon, 29 May 2017 09:56:33 +0000 (09:56 +0000)]
change bgpd's control socket from /var/run/bgpd.sock to
/var/run/bgpd.sock.<rdomain> in both bgpd and bgpctl. makes life
easier for admins.
ok phessler, claudio, henning, feedback from Job Snijders

7 years agoRandomize link-order of libcrypto as we do with libc. This library
deraadt [Mon, 29 May 2017 09:44:01 +0000 (09:44 +0000)]
Randomize link-order of libcrypto as we do with libc.  This library
has many small functions without significant local storage, therefore
less tail protection from -fstack-protector-strong to prevent their use
as ROP gadgets.  It is used in security contexts.  Also many functions
dribble pointers onto the stack, allowing discovery of gadgets via the
fixed relative addresses, so let's randomly bias those.
ok tedu jsing

The rc script will soon need a strategy for skipping this step on
machines with poor IO performance.  Or maybe do it less often?  However,
I don't see many more libraries we'll do this with, these are the two
most important ones.

7 years agosync
deraadt [Mon, 29 May 2017 09:40:13 +0000 (09:40 +0000)]
sync

7 years agoIt is distasteful to have manual pages which don't refer to real
deraadt [Mon, 29 May 2017 09:40:02 +0000 (09:40 +0000)]
It is distasteful to have manual pages which don't refer to real
function calls, but instead a "class" of functions like "sigsetops".
Rename to sigaddset", and while at it improve documentation in sigprocmask(2)
to point to it.
ok tedu

7 years agosync
deraadt [Mon, 29 May 2017 09:37:33 +0000 (09:37 +0000)]
sync

7 years agoclang points out that the first argument to setproctitle is a format
florian [Mon, 29 May 2017 08:59:42 +0000 (08:59 +0000)]
clang points out that the first argument to setproctitle is a format
string. Not really an issue since we are eventually passing in a
string literal...
While here don't set the proctitle for the main process to play nicer
with rc.d(8)

7 years agoCompare the l2 address from which the slaac address was formed with
florian [Mon, 29 May 2017 08:15:38 +0000 (08:15 +0000)]
Compare the l2 address from which the slaac address was formed with
the current interface l2 address. Only update the address lifetime if
they are the same. This way we get a new address on ifconfig lladdr
random.