openbsd
2 years agoFix format string: use %zu for size_t, not %lu.
tb [Wed, 22 Jun 2022 09:54:19 +0000 (09:54 +0000)]
Fix format string: use %zu for size_t, not %lu.

2 years agoError out on negative shifts in BN_{r,l}shift()
tb [Wed, 22 Jun 2022 09:03:06 +0000 (09:03 +0000)]
Error out on negative shifts in BN_{r,l}shift()

Without these checks in both functions nw = n / BN_BITS2 will be negative
and this leads to out-of-bounds accesses via negative array indices and
memset with a negative size.

Pointed out by cheloha

ok jsing

2 years agoRevert part of r1.212, which caused a regression in a previously working
bket [Tue, 21 Jun 2022 17:33:21 +0000 (17:33 +0000)]
Revert part of r1.212, which caused a regression in a previously working
setup

OK dlg@

2 years agoMake sure not to fclose() the same fd twice in case of an error.
tobhe [Tue, 21 Jun 2022 14:52:13 +0000 (14:52 +0000)]
Make sure not to fclose() the same fd twice in case of an error.

ok dtucker@

2 years agoAvoid saying that running netstart manually post-boot will "reset an
sthen [Tue, 21 Jun 2022 11:55:55 +0000 (11:55 +0000)]
Avoid saying that running netstart manually post-boot will "reset an
existing interface to its default state", because it doesn't.  ok jmc@

2 years agodrm/i915/adl_p: Add ddc pin mapping
jsg [Tue, 21 Jun 2022 11:21:22 +0000 (11:21 +0000)]
drm/i915/adl_p: Add ddc pin mapping

From Tejas Upadhyay
af10ec31a81b2f8b9a3b2d1ef05f553cc9495701 in mainline linux

2 years agoUse applymask() instead of inetXapplymask() since this code operates on
claudio [Tue, 21 Jun 2022 10:05:48 +0000 (10:05 +0000)]
Use applymask() instead of inetXapplymask() since this code operates on
struct bgpd_addr.
OK tb@

2 years agodrm/i915/adlp: Implement workaround 16013190616
jsg [Tue, 21 Jun 2022 09:46:33 +0000 (09:46 +0000)]
drm/i915/adlp: Implement workaround 16013190616

From Jose Roberto de Souza
9556829ce4d0618ae4295af8e4b3dd7e38f43598 in mainline linux

2 years agoStore time lines are scrolled into history and display in copy mode.
nicm [Tue, 21 Jun 2022 09:30:01 +0000 (09:30 +0000)]
Store time lines are scrolled into history and display in copy mode.

2 years agosync
jsg [Tue, 21 Jun 2022 07:11:18 +0000 (07:11 +0000)]
sync

2 years agodrm/i915/adlp: Remove require_force_probe protection
jsg [Tue, 21 Jun 2022 07:08:53 +0000 (07:08 +0000)]
drm/i915/adlp: Remove require_force_probe protection

From Clint Taylor
dfb924e33927ffa51e197acd58ad1c5edf1b5ff5 in mainline linux

2 years agodrm/i915/adl_s: Remove require_force_probe protection
jsg [Tue, 21 Jun 2022 07:05:55 +0000 (07:05 +0000)]
drm/i915/adl_s: Remove require_force_probe protection

From ravitejax
f5392e5f8ef300c5d8fb97fb441aad217e44f394 in mainline linux

2 years agounstub intel_guc_submission.c
jsg [Tue, 21 Jun 2022 06:17:32 +0000 (06:17 +0000)]
unstub intel_guc_submission.c

2 years agoadd Intel 600 Series LP
jsg [Tue, 21 Jun 2022 04:17:21 +0000 (04:17 +0000)]
add Intel 600 Series LP

2 years agoregen
jsg [Tue, 21 Jun 2022 04:09:28 +0000 (04:09 +0000)]
regen

2 years agoadd some Intel ADL-P and 600 Series LP ids
jsg [Tue, 21 Jun 2022 04:08:53 +0000 (04:08 +0000)]
add some Intel ADL-P and 600 Series LP ids

2 years agoTweak a comment
tb [Mon, 20 Jun 2022 19:42:58 +0000 (19:42 +0000)]
Tweak a comment

2 years agoFlip roles of lowercase and uppercase A and B.
tb [Mon, 20 Jun 2022 19:38:25 +0000 (19:38 +0000)]
Flip roles of lowercase and uppercase A and B.

This matches Cohen's text better and makes the entire thing easier to
read.

suggested by jsing

2 years agoClean up BN_kronecker()
tb [Mon, 20 Jun 2022 19:32:35 +0000 (19:32 +0000)]
Clean up BN_kronecker()

Instead of "Cohen's step N" explain in words what is being done. Things
such as (A & B & 2) != 0 being equivalent to (-1)^((A-1)(B-1)/4) being
negative are not entirely obvious...  Remove the strange error dance and
adjust variable names to what Cohen's book uses. Simplify various curly
bits.

ok jsing

2 years agoFix console screen blanker setting timeout back to zero
gnezdo [Mon, 20 Jun 2022 16:28:42 +0000 (16:28 +0000)]
Fix console screen blanker setting timeout back to zero

Tested locally. "looks correct" to miod@

From Crystal Kolipe kolipe DOT c AT exoticsilicon DOT com

2 years agoFix some bizarre indentation and line breaks.
tb [Mon, 20 Jun 2022 15:02:21 +0000 (15:02 +0000)]
Fix some bizarre indentation and line breaks.

2 years agoDo not display configuration file errors in a pane when in control mode,
nicm [Mon, 20 Jun 2022 07:59:37 +0000 (07:59 +0000)]
Do not display configuration file errors in a pane when in control mode,
instead report them with a %config-error notification. GitHub issue 3193.

2 years agoFixup comment
jca [Mon, 20 Jun 2022 06:45:31 +0000 (06:45 +0000)]
Fixup comment

2 years agoWhen there are multiple GOPs, try to select first one from which we
yasuoka [Mon, 20 Jun 2022 02:22:05 +0000 (02:22 +0000)]
When there are multiple GOPs, try to select first one from which we
can get a handle of DevicePath protocol if any, otherwise select first
one.  This is expected to workaround the garbled screen problem when
booting with efiboot happened on some machines.

origin diff from Alexei K.

2 years agoRemove unused struct fileops field fo_poll and callbacks.
visa [Mon, 20 Jun 2022 01:39:44 +0000 (01:39 +0000)]
Remove unused struct fileops field fo_poll and callbacks.

OK mpi@

2 years agoRevert previous iwn(4) commit. Later testing done by Christian Schulte
stsp [Sun, 19 Jun 2022 18:27:06 +0000 (18:27 +0000)]
Revert previous iwn(4) commit. Later testing done by Christian Schulte
suggests that this fix did not in fact help with the issue. And this
change broke wireless for other iwn(4) users, as reported by Lucas on bugs@

2 years agoUse uppercase for SUCCESS for consistency
tb [Sun, 19 Jun 2022 17:10:02 +0000 (17:10 +0000)]
Use uppercase for SUCCESS for consistency

2 years agoNone of these tests needs to link statically.
tb [Sun, 19 Jun 2022 17:08:19 +0000 (17:08 +0000)]
None of these tests needs to link statically.

2 years agoDrop bogus DPADD += ${LIBSSL}
tb [Sun, 19 Jun 2022 17:05:31 +0000 (17:05 +0000)]
Drop bogus DPADD += ${LIBSSL}

2 years agoMake expected output match reality again.
tb [Sun, 19 Jun 2022 14:02:44 +0000 (14:02 +0000)]
Make expected output match reality again.

2 years agoThere is no reason to print the RTF_DONE bit in the routing table output.
claudio [Sun, 19 Jun 2022 13:59:22 +0000 (13:59 +0000)]
There is no reason to print the RTF_DONE bit in the routing table output.
It is now set on all routes (and before it was never set).
OK tb@ anton@

2 years agopluart(4) does not report the correct baudrate for the hardware console
anton [Sun, 19 Jun 2022 12:52:19 +0000 (12:52 +0000)]
pluart(4) does not report the correct baudrate for the hardware console
but instead defaults to 38400. This in turn causes the same baudrate to
end up in /etc/ttys during installation. Note that this is not a problem
as of now since pluart does not support changing the baudrate just yet.

Instead, honor and propagate the baudrate given to pluartcnattach()
while attaching the hardware console.

ok visa@

2 years agoSince rtsock.c -r1.329 routes sent via sysctl(2) are now marked RTF_DONE.
tb [Sun, 19 Jun 2022 10:58:22 +0000 (10:58 +0000)]
Since rtsock.c -r1.329 routes sent via sysctl(2) are now marked RTF_DONE.
Adjust expected output.

2 years agoFix rttest output after rtsock.c r1.329 that RTF_DONE to routes sent
tb [Sun, 19 Jun 2022 10:56:25 +0000 (10:56 +0000)]
Fix rttest output after rtsock.c r1.329 that RTF_DONE to routes sent
via sysctl(2)

2 years agoImplement a applymask() function that works on bgpd_addr structs.
claudio [Sun, 19 Jun 2022 10:30:09 +0000 (10:30 +0000)]
Implement a applymask() function that works on bgpd_addr structs.
Use this function in kroute so that kroute_find and kroute6_find can switch
to use struct bgpd_addr as argument.
OK tb@

2 years agogrammar fix; from S M
jmc [Sun, 19 Jun 2022 05:49:51 +0000 (05:49 +0000)]
grammar fix; from S M

2 years agoQuick regression test that checks that BN_is_prime_fasttest_ex()
tb [Sat, 18 Jun 2022 19:53:19 +0000 (19:53 +0000)]
Quick regression test that checks that BN_is_prime_fasttest_ex()
recognizes the primes in the primes[] table with and without trial
division. Would have caught the bug fixed in bn_primes.c r1.9.

2 years agosync to libunbound 1.16.0; heavy lifting by sthen
florian [Sat, 18 Jun 2022 16:20:13 +0000 (16:20 +0000)]
sync to libunbound 1.16.0; heavy lifting by sthen

2 years agoWe were a bit too enthusiastic when we copied libunbound and some other
florian [Sat, 18 Jun 2022 16:17:29 +0000 (16:17 +0000)]
We were a bit too enthusiastic when we copied libunbound and some other
helper functions from unbound(8).
Delete unused shared memory functions that the unbound(8) daemon uses
for statistics gathering. While here get rid of daemon.h which is
unused, too.
With this we can remove the hack in config.h to pretend we don't have
shmget to prevent linker errors which makes syncing with unbound(8)
easier.

2 years agoFix prime recognition when doing trial divisions
tb [Sat, 18 Jun 2022 15:52:35 +0000 (15:52 +0000)]
Fix prime recognition when doing trial divisions

If gcd(a, primes[i]) == 0 then a could still be a prime, namely in the
case that a == primes[i], so check for that case as well.

Problem noted by Martin Grenouilloux

ok jsing

2 years agoFix gzip byte counts with 32-bit integers
gkoehler [Sat, 18 Jun 2022 03:23:19 +0000 (03:23 +0000)]
Fix gzip byte counts with 32-bit integers

If zlib is without our local change in lib/libz/zlib.h r1.7, then
s->z_stream.total_in and s->z_stream.total_out might overflow on
architectures where uLong has 32 bits.  After overflow, the total
would be 4G below the correct total.

Calculate our own 64-bit totals.  When decompressing, take
(uLong)s->z_stream.total_in as a total modulo ULONG_MAX + 1.

ok tb@

2 years agoDon't attempt to fprintf a null identity comment. From Martin Vahlensieck
dtucker [Sat, 18 Jun 2022 02:17:16 +0000 (02:17 +0000)]
Don't attempt to fprintf a null identity comment.  From Martin Vahlensieck
via tech@.

2 years agoThe timeout for ipsec acquire does not decrement the reference
bluhm [Fri, 17 Jun 2022 13:40:21 +0000 (13:40 +0000)]
The timeout for ipsec acquire does not decrement the reference
counter to 0 properly.  We have one reference count for the lists,
and one for the timeout handler.  When the timout fires, it has to
decrement the reference to itself.  Then the ipa is removed from
the lists and decremented again.
from Stefan Butz; OK tobhe@ mvs@

2 years agoOnly probe device 0 on PCI busses corresponding to a PCIe root port or a
kettenis [Fri, 17 Jun 2022 10:08:36 +0000 (10:08 +0000)]
Only probe device 0 on PCI busses corresponding to a PCIe root port or a
PCIe switch/bridge downstream port.  This fixes an issue where a device would
show up 32 times if the PCIe port does not terminate configuration requests
targeted at device numbers 1-31 as required by the spec and the PCIe device
does not filter requests based on its own device number.  Linux, NetBSD and
FreeBSD (to some extent) already do this.

This should fix an issue seen on Ampere Altra where the BMC's PCI bridge
and USB controller would show up multiple times.

ok jsg@

2 years agoAdapt the KAME hack in sa2addr() to be the same as in route/show.c
claudio [Fri, 17 Jun 2022 09:12:06 +0000 (09:12 +0000)]
Adapt the KAME hack in sa2addr() to be the same as in route/show.c
Also fix the function by moving the memcpy() after the embedded scope
has been removed.
OK tb@

2 years agoCheck cursor options when a pane is created, not just when they are changed.
nicm [Fri, 17 Jun 2022 07:28:05 +0000 (07:28 +0000)]
Check cursor options when a pane is created, not just when they are changed.

2 years agoLog an error if pipe() fails while accepting a connection. bz#3447, from
dtucker [Fri, 17 Jun 2022 01:00:03 +0000 (01:00 +0000)]
Log an error if pipe() fails while accepting a connection.  bz#3447, from
vincent-openssh at vinc17 net, ok djm@

2 years agoAdd boundary checks to hidms_setup.
bru [Thu, 16 Jun 2022 20:52:38 +0000 (20:52 +0000)]
Add boundary checks to hidms_setup.

Thanks to Sven M. Hallberg.

2 years agopfctl reports existing table as being added. glitch has
sashan [Thu, 16 Jun 2022 20:47:26 +0000 (20:47 +0000)]
pfctl reports existing table as being added. glitch has
been spotted and reported by jmc@

OK kn@

2 years agoBump MAXCPUS to 256, which is the maximum number of cores on a dual socket
kettenis [Thu, 16 Jun 2022 20:45:42 +0000 (20:45 +0000)]
Bump MAXCPUS to 256, which is the maximum number of cores on a dual socket
machine with Ampere Altra Max CPUs.  OpenBSD should run on such a machine
now.

ok patrick@, deraadt@

2 years agoThe 64-core versions of the Ampere Altra SoC still have 80 redistributors.
kettenis [Thu, 16 Jun 2022 20:44:09 +0000 (20:44 +0000)]
The 64-core versions of the Ampere Altra SoC still have 80 redistributors.
Presumably these are the same silicon and the 64-core version just has some
of the cores disabled.  This means that the redistributors don't necessarily
all match up with a core that is actually enabled.  So remove the panic
that would happen if we found a redistributor without a matching core.

ok patrick@, deraadt@

2 years agoOpenBSD syslogd(8) escapes binary data with vis(3). Use the
bluhm [Thu, 16 Jun 2022 18:44:43 +0000 (18:44 +0000)]
OpenBSD syslogd(8) escapes binary data with vis(3).  Use the
VIS_NOSLASH option to avoid additional backslashes.  Although this
option prevents to decode binaries, it makes automatic post processing
easier.  Formats like JSON use backslash escaping themself, additional
escaping from syslogd would break SIEM systems.  vis protection was
introduced to avoid evil characters in log files and not to make
binary logging possible.
from Matthias Pitzl; OK deraadt@

2 years agoRRDP serial numbers should only increase.
claudio [Thu, 16 Jun 2022 16:09:56 +0000 (16:09 +0000)]
RRDP serial numbers should only increase.
Warn if the serial number decreases between syncs.

On top of this only allow a small window of up to 2 deltas from the
current one to consider our cache to be in sync.
The number 2 is probably to conservative and should be adjusted once
some data points got collected.

It seems to happen that CAs restore RRDP snapshots instead of building
a fresh snapshot with a new session-id. Which results in rpki-client to
ignore the repo until the serial number is bigger again.
OK tb@

2 years agoNeither inet4allone nor inet6allone need global scope. Make them static.
claudio [Thu, 16 Jun 2022 15:36:36 +0000 (15:36 +0000)]
Neither inet4allone nor inet6allone need global scope. Make them static.
Noticed by tb@

2 years agoRemove prefixlen2mask() uasge outside of util.c. Instead use inet4applymask().
claudio [Thu, 16 Jun 2022 15:34:07 +0000 (15:34 +0000)]
Remove prefixlen2mask() uasge outside of util.c. Instead use inet4applymask().
Neccessary after similar change in bgpd.
OK tb@

2 years agoRemove prefixlen2mask() uasge outside of util.c. Instead use inet4applymask().
claudio [Thu, 16 Jun 2022 15:33:05 +0000 (15:33 +0000)]
Remove prefixlen2mask() uasge outside of util.c. Instead use inet4applymask().

Using inet4applymask() makes the code similar to the IPv6 version.
Also switch kroute{,6}_match() to use a struct bgpd_addr * as argument.
OK tb@

2 years agoIf anything in name2id conversion fails then just return 0.
claudio [Thu, 16 Jun 2022 15:30:12 +0000 (15:30 +0000)]
If anything in name2id conversion fails then just return 0.

Do not set any errno. So if a non empty string is used in name2id
conversion and the return value is 0 then that would be an error.
Callers in most cases do not care and accept that a label may be
lost because of conversion failure.

Noticed by and OK tb@

2 years agoshmaddr is a pointer, so use NULL instead 0f
deraadt [Thu, 16 Jun 2022 13:54:39 +0000 (13:54 +0000)]
shmaddr is a pointer, so use NULL instead 0f

2 years agoKeep cursor on selected item on menu (useful for blind people), GitHub
nicm [Thu, 16 Jun 2022 13:27:39 +0000 (13:27 +0000)]
Keep cursor on selected item on menu (useful for blind people), GitHub
issue 3225.

2 years agoMark routes sent via sysctl(2) with RTF_DONE like it is done on the
claudio [Thu, 16 Jun 2022 10:35:45 +0000 (10:35 +0000)]
Mark routes sent via sysctl(2) with RTF_DONE like it is done on the
route socket. All messages passed are by definition done. This may
allow to share more code between sysctl and route socket parsers.
OK mpi@

2 years agoRemove __func__ from warning messages where it does not add anything
claudio [Thu, 16 Jun 2022 09:51:07 +0000 (09:51 +0000)]
Remove __func__ from warning messages where it does not add anything
to the warning. The messages are clear enough and unique.
OK tb@

2 years agomake sure that UseDNS hostname lookup happens in the monitor and
djm [Wed, 15 Jun 2022 16:08:25 +0000 (16:08 +0000)]
make sure that UseDNS hostname lookup happens in the monitor and
not in the pledge(2)'d unprivileged process; fixes regression
caused by recent refactoring spotted by henning@

2 years agoPrevent busy poll loop on pfkey message error.
claudio [Wed, 15 Jun 2022 16:00:37 +0000 (16:00 +0000)]
Prevent busy poll loop on pfkey message error.

When pfkey_reply() detects an error in a message it needs to flush that
message out of the buffer because pfkey_read() only did a peak read.
OK tb@

2 years agoNeed to continue to the next message for route messages with errors
claudio [Wed, 15 Jun 2022 15:06:25 +0000 (15:06 +0000)]
Need to continue to the next message for route messages with errors
or which are not marked RTF_DONE. Possible fix for bgpd exiting for
no obvious reason. Found with help from florian@
OK florian@

2 years agoDo not use defines from pfkeyv2.h in portable code.
claudio [Wed, 15 Jun 2022 14:09:30 +0000 (14:09 +0000)]
Do not use defines from pfkeyv2.h in portable code.

Instead define our own algorithm enums for the IPsec code.
OK tb@ sthen@

2 years agoAdjust after chaning F_BGPD_INSERTED to F_BGPD and removal of F_DYNAMIC
claudio [Wed, 15 Jun 2022 10:10:50 +0000 (10:10 +0000)]
Adjust after chaning F_BGPD_INSERTED to F_BGPD and removal of F_DYNAMIC

2 years agoRename F_BGPD_INSERTED to F_BGPD and use F_BGPD_INSERTED as a flag that
claudio [Wed, 15 Jun 2022 10:10:03 +0000 (10:10 +0000)]
Rename F_BGPD_INSERTED to F_BGPD and use F_BGPD_INSERTED as a flag that
indicates that the route was successfully added to the FIB.

Filter out dynamic routes, like it is already done for ARP and ND routes) and
kill F_DYNAMIC.

Also remove the protect_lo() bits. Adding dummy kroute entries does no longer
prevent bad routes to hit the FIB. Also loopback IPs are checked in a few
other places to prevent bad routes to be installed into the FIB.

OK tb@

2 years agoIn iwn(4), delay 40MHz RXON config until we have negotiated 11n mode.
stsp [Wed, 15 Jun 2022 08:43:17 +0000 (08:43 +0000)]
In iwn(4), delay 40MHz RXON config until we have negotiated 11n mode.

Enabling 40MHz too early can cause firmware sysassert 0x05 on 5300 devices
when we send the initial AUTH frame to kick off association. At this point
we haven't negotiated 11n mode yet (this happens later during the association
sequence), so perhaps firmware saw an inconsistent configuration state?
It is unclear why this problem did not seem to affect any other devices.

Reported by Christian Schulte on misc@

2 years agodrm/amdgpu: update VCN codec support for Yellow Carp
jsg [Wed, 15 Jun 2022 07:49:55 +0000 (07:49 +0000)]
drm/amdgpu: update VCN codec support for Yellow Carp

From Alex Deucher
2b7d9fd0f3725135b40571d6a44bbbaeead132fe in linux 5.15.y/5.15.47
97e50305542f384741a5b45699aba349fe9fca73 in mainline linux

2 years agodrm/atomic: Force bridge self-refresh-exit on CRTC switch
jsg [Wed, 15 Jun 2022 07:47:22 +0000 (07:47 +0000)]
drm/atomic: Force bridge self-refresh-exit on CRTC switch

From Brian Norris
cab7cd86f9e81834c407e043844779f21653ff3c in linux 5.15.y/5.15.47
e54a4424925a27ed94dff046db3ce5caf4b1e748 in mainline linux

2 years agodrm/amd/pm: use bitmap_{from,to}_arr32 where appropriate
jsg [Wed, 15 Jun 2022 07:44:36 +0000 (07:44 +0000)]
drm/amd/pm: use bitmap_{from,to}_arr32 where appropriate

From Yury Norov
b2d359f095883d10b33e74a4a404363d2312830d in linux 5.15.y/5.15.47
525d6515604eb1373ce5e6372a6b6640953b2d6a in mainline linux

2 years agodrm/amd/pm: Fix missing thermal throttler status
jsg [Wed, 15 Jun 2022 07:40:33 +0000 (07:40 +0000)]
drm/amd/pm: Fix missing thermal throttler status

From Lijo Lazar
e0199ce728fb98a96a20136a5edb11c160d3151f in linux 5.15.y/5.15.47
b0f4d663fce6a4232d3c20ce820f919111b1c60b in mainline linux

2 years agodrm/radeon: fix a possible null pointer dereference
jsg [Wed, 15 Jun 2022 07:37:40 +0000 (07:37 +0000)]
drm/radeon: fix a possible null pointer dereference

From Gong Yuanjun
7b7fba107b2c4ec7673d0f45bdbb9d1af697d9b9 in linux 5.15.y/5.15.47
a2b28708b645c5632dc93669ab06e97874c8244f in mainline linux

2 years agodrm/amd/display: Check if modulo is 0 before dividing.
jsg [Wed, 15 Jun 2022 07:34:21 +0000 (07:34 +0000)]
drm/amd/display: Check if modulo is 0 before dividing.

From David Galiffi
10ef82d6e0af5536ec64770c07f6bbabfdd6977c in linux 5.15.y/5.15.47
49947b906a6bd9668eaf4f9cf691973c25c26955 in mainline linux

2 years agoadd bitmap_to_arr32() for 5.15.47 drm
jsg [Wed, 15 Jun 2022 07:04:09 +0000 (07:04 +0000)]
add bitmap_to_arr32() for 5.15.47 drm

discussed with and partly from kettenis@

2 years agoSwitch to using TLS_INT instead of handrolling it
tb [Wed, 15 Jun 2022 06:14:59 +0000 (06:14 +0000)]
Switch to using TLS_INT instead of handrolling it

2 years agom4: use ASCII quotes `...' in man page example
naddy [Tue, 14 Jun 2022 21:31:45 +0000 (21:31 +0000)]
m4: use ASCII quotes `...' in man page example

The m4 syntax specifically requires `...' and not the pretty Unicode
quotes that the Sq macro can produce.  Use of escape sequences
clarified by schwarze@

2 years agoThe curly braces are tokens in the described language,
schwarze [Tue, 14 Jun 2022 15:49:58 +0000 (15:49 +0000)]
The curly braces are tokens in the described language,
so add the missing .Cm macros; markup error reported by espie@;
OK florian@ espie@

2 years agomy email should be @openbsd.org everywhere, and nothing else
espie [Tue, 14 Jun 2022 15:15:17 +0000 (15:15 +0000)]
my email should be @openbsd.org everywhere, and nothing else

2 years agoAdd my copyright, I think I added enough code to justify it.
claudio [Tue, 14 Jun 2022 14:06:48 +0000 (14:06 +0000)]
Add my copyright, I think I added enough code to justify it.

2 years agoDocument search order
kn [Tue, 14 Jun 2022 09:56:07 +0000 (09:56 +0000)]
Document search order

Explain when and how LD_LIBRARY_PATH, DT_RUNPATH and DT_RPATH are used.

Input OK guenther

2 years agoAdd Longreach airport (LRE), which I visited so long ago I had to check with
jmatthew [Tue, 14 Jun 2022 09:03:46 +0000 (09:03 +0000)]
Add Longreach airport (LRE), which I visited so long ago I had to check with
my parents.

2 years agoBump version to 7.4
claudio [Tue, 14 Jun 2022 07:33:58 +0000 (07:33 +0000)]
Bump version to 7.4

2 years agokf* terminfo capabilities are poorly defined and rxvt uses them in a
nicm [Tue, 14 Jun 2022 07:29:00 +0000 (07:29 +0000)]
kf* terminfo capabilities are poorly defined and rxvt uses them in a
different way from xterm, so add a feature flag for rxvt to make tmux
ignore the capabilities and instead rely on its builtin definitions.

2 years agofix logic bug in pf_find_state()
henning [Mon, 13 Jun 2022 12:48:00 +0000 (12:48 +0000)]
fix logic bug in pf_find_state()
a state in PFTM_PURGE could potentially hide another state on the same state
key that is active and we'd incorrectly block the packet
I believe that cannot happen as things are now.
ok sashan

2 years agoFix compatible strings to match the official device tree binding.
kettenis [Mon, 13 Jun 2022 12:40:30 +0000 (12:40 +0000)]
Fix compatible strings to match the official device tree binding.

ok patrick@, dlg@, jsg@

2 years agoSplit dispatch_rtmsg_addr() into a function parsing a route message
claudio [Mon, 13 Jun 2022 09:57:44 +0000 (09:57 +0000)]
Split dispatch_rtmsg_addr() into a function parsing a route message
into a struct kroute_full and then two functions to add/change or delete
that route from the kroute tree.
Cleanup some additional things and use the same parse funtion in fetchtable.
OK tb@

2 years agosync
deraadt [Mon, 13 Jun 2022 03:47:12 +0000 (03:47 +0000)]
sync

2 years agovmm(4): remove vm teardown from vcpu run path
dv [Sun, 12 Jun 2022 19:48:12 +0000 (19:48 +0000)]
vmm(4): remove vm teardown from vcpu run path

Remove the calling of vm_teardown from the code path in vm_run after
vmm has exited the vm/vcpu and is on its way back to userland/vmd(8).
vmd(8) and other vmm(4) users should use the VMM_IOC_TERM ioctl(2)
instead.

Help testing from Mischa Peters, OK mlarkin@

2 years agoaplsart(4)
kettenis [Sun, 12 Jun 2022 16:02:33 +0000 (16:02 +0000)]
aplsart(4)

2 years agoThe NVMe firmware distributed with the macOS 13 developer beta does not
kettenis [Sun, 12 Jun 2022 16:00:12 +0000 (16:00 +0000)]
The NVMe firmware distributed with the macOS 13 developer beta does not
like some of the shortcuts we've taken in the code that talks to RTKit.
In particular it does not like our NAK of the request for ioreport buffers.
So allocate the requested buffers and ACK the request instead.

However, the hardware implements a address filter between the coprocessor
that the firmware runs on and main memory.  So we have to add a mapping
into this address filter such that the firmware can access these buffers.
Support for this address filter is implemented in a new aplsart(4) driver.

ok dlg@

2 years agoAdd stftemp(4), a driver for the temperature sensor integrated on the
kettenis [Sun, 12 Jun 2022 10:51:55 +0000 (10:51 +0000)]
Add stftemp(4), a driver for the temperature sensor integrated on the
StarFive JH7100 SoC.

ok jsg@

2 years agoAllow sleeping while clearing a sleep timeout
visa [Sun, 12 Jun 2022 10:36:04 +0000 (10:36 +0000)]
Allow sleeping while clearing a sleep timeout

Since sys/kern/kern_timeout.c r1.84, timeout_barrier() has used sleeping
with soft-interrupt-driven timeouts. Adjust the sleep machinery so that
the timeout clearing can block in sleep_finish().

This adds one step of recursion inside sleep_finish(). However, the
sleep queue handling does not recurse because sleep_finish() completes
it before calling timeout_del_barrier().

This fixes the following panic:

panic: kernel diagnostic assertion "(p->p_flag & P_TIMEOUT) == 0" failed: file "sys/kern/kern_synch.c", line 373
Stopped at db_enter+0x10:  popq    %rbp
db_enter() at db_enter+0x10
panic() at panic+0xbf
__assert() at __assert+0x25
sleep_setup() at sleep_setup+0x1d8
cond_wait() at cond_wait+0x46
timeout_barrier() at timeout_barrier+0x109
timeout_del_barrier() at timeout_del_barrier+0xa2
sleep_finish() at sleep_finish+0x16d
tsleep() at tsleep+0xb2
sys_nanosleep() at sys_nanosleep+0x12d
syscall() at syscall+0x374

OK mpi@ dlg@

2 years agokqueue: Fix missing wakeup
visa [Sun, 12 Jun 2022 10:34:36 +0000 (10:34 +0000)]
kqueue: Fix missing wakeup

While one thread is running kqueue_scan(), another thread can begin
scanning the same kqueue, observe that the event queue is empty, and
go to sleep. If the first thread re-inserts a knote for re-processing,
the second thread can miss the newly pending event. Wake up the kqueue
after a re-insert to correct this.

This fixes a Go test hang that jsing@ tracked down to kqueue.

Tested in snaps for a week.

OK jsing@ mpi@

2 years agoDocument ext2fs traversal by locate.updatedb(8)
gnezdo [Sun, 12 Jun 2022 05:36:20 +0000 (05:36 +0000)]
Document ext2fs traversal by locate.updatedb(8)

From Larry Hynes

2 years agowhitespace
kettenis [Sat, 11 Jun 2022 19:12:59 +0000 (19:12 +0000)]
whitespace

2 years agoFix size of flags output buffer.
nicm [Sat, 11 Jun 2022 16:59:33 +0000 (16:59 +0000)]
Fix size of flags output buffer.

2 years agoBring back PL011 UART FIFO support to pluart(4). This time with working
anton [Sat, 11 Jun 2022 05:29:24 +0000 (05:29 +0000)]
Bring back PL011 UART FIFO support to pluart(4). This time with working
tx interrupts.

Special treatment of Server Base System Architecture (SBSA) generic UART
devices is required as presence of the interrupt trigger level register
is not guaranteed. Therefore treat such devices of having a 1-byte FIFO.

ok visa@

2 years agoMore %i vs %d cleanup
tb [Fri, 10 Jun 2022 22:00:15 +0000 (22:00 +0000)]
More %i vs %d cleanup

2 years agoImplement CPU_BUSY_CYCLE() with the ZiHintPause extension
jca [Fri, 10 Jun 2022 21:34:15 +0000 (21:34 +0000)]
Implement CPU_BUSY_CYCLE() with the ZiHintPause extension

This may help save power but also improves performance in contended
workloads.

kettenis@ helped a lot:
- pointing out the error in my newest diff
- pointing out that the ZiHintPause had been ratified in early 2021
- testing this on his visionfive board

ok kettenis@

libc make -j4 time before
17m39.06s real    22m13.85s user    34m27.88s system
17m35.60s real    22m15.01s user    34m28.51s system
17m40.78s real    22m14.89s user    34m42.91s system
17m40.61s real    22m06.99s user    34m29.58s system
17m42.79s real    22m08.04s user    34m33.19s system
17m39.21s real    22m11.27s user    34m23.18s system
17m32.81s real    22m13.84s user    34m09.67s system

libc make -j4 time after
16m35.48s real    21m57.94s user    32m05.50s system
16m49.10s real    21m58.47s user    32m42.37s system
16m42.16s real    22m00.51s user    32m22.18s system
16m41.54s real    21m53.32s user    32m19.53s system
16m45.42s real    21m52.14s user    32m37.78s system
16m46.55s real    21m58.98s user    32m36.70s system