openbsd
3 years agoQuirk-compatibility with GNU tbl(1):
schwarze [Fri, 10 Sep 2021 13:23:44 +0000 (13:23 +0000)]
Quirk-compatibility with GNU tbl(1):
With the "nospaces" option, skip space characters before and after "T{",
in addition to skipping those at the beginning and end of data cells.

Minor issue reported by <Oliver dot Corff at email dot de>.

3 years agoProperly handle keep-alive for HTTP/1.1. If the server uses HTTP/1.1
claudio [Fri, 10 Sep 2021 13:20:03 +0000 (13:20 +0000)]
Properly handle keep-alive for HTTP/1.1. If the server uses HTTP/1.1
keep-alive is the default. Check this early on and disable keep-alive
if a Connection: closed header is sent. Fixes the keep-alive issues
I have seen.
OK sthen@

3 years agoIn a tbl(7) having the "nospaces" option, skip space characters
schwarze [Fri, 10 Sep 2021 12:06:29 +0000 (12:06 +0000)]
In a tbl(7) having the "nospaces" option, skip space characters
not only at the end of data cells, but also after "T}",
aligning the behaviour of the parser with GNU tbl(1).

Issue reported by <Oliver dot Corff at email dot de>.

3 years agoDocument that non-interactive commands are run via the user's shell
dtucker [Fri, 10 Sep 2021 11:38:38 +0000 (11:38 +0000)]
Document that non-interactive commands are run via the user's shell
using the -c flag.  ok jmc@

3 years agoDocument behaviour of arguments following non-interactive commands.
dtucker [Fri, 10 Sep 2021 10:26:02 +0000 (10:26 +0000)]
Document behaviour of arguments following non-interactive commands.
Prompted by github PR#139 from EvanTheB, feedback & ok djm@ jmc@

3 years agoDo not ignore SSL_TLSEXT_ERR_FATAL from the ALPN callback
tb [Fri, 10 Sep 2021 09:25:29 +0000 (09:25 +0000)]
Do not ignore SSL_TLSEXT_ERR_FATAL from the ALPN callback

As reported by Jeremy Harris, we inherited a strange behavior from
OpenSSL, in that we ignore the SSL_TLSEXT_ERR_FATAL return from the
ALPN callback. RFC 7301, 3.2 states: 'In the event that the server
supports no protocols that the client advertises, then the server
SHALL respond with a fatal "no_application_protocol" alert.'

Honor this requirement and succeed only on SSL_TLSEXT_ERR_{OK,NOACK}
which is the current behavior of OpenSSL. The documentation change
is taken from OpenSSL 1.1.1 as well.

As pointed out by jsing, there is more to be fixed here:
- ensure that the same protocol is selected on session resumption
- should the callback be called even if no ALPN extension was sent?
- ensure for TLSv1.2 and earlier that the SNI has already been processed

ok beck jsing

3 years agoPrepare to provide BN_RAND_* flags for BN_rand_range()
tb [Fri, 10 Sep 2021 09:08:03 +0000 (09:08 +0000)]
Prepare to provide BN_RAND_* flags for BN_rand_range()

ok beck jsing

3 years agoPrepare to provide SSL_CTX_get0_privatekey()
tb [Fri, 10 Sep 2021 08:59:56 +0000 (08:59 +0000)]
Prepare to provide SSL_CTX_get0_privatekey()

ok beck

3 years agoDisable aliases inside aliases for the moment.
nicm [Fri, 10 Sep 2021 08:52:46 +0000 (08:52 +0000)]
Disable aliases inside aliases for the moment.

3 years agoClarify which file's attributes -p preserves, and that it's specifically
dtucker [Fri, 10 Sep 2021 07:11:11 +0000 (07:11 +0000)]
Clarify which file's attributes -p preserves, and that it's specifically
the file mode bits. bz#3340 from calestyo at scientia.net, ok djm@ jmc@

3 years agoMinor KNF nit, align struct field.
anton [Fri, 10 Sep 2021 05:48:43 +0000 (05:48 +0000)]
Minor KNF nit, align struct field.

3 years agoInstead of letting uhidev drivers get the report sizes, do it once in
anton [Fri, 10 Sep 2021 05:47:38 +0000 (05:47 +0000)]
Instead of letting uhidev drivers get the report sizes, do it once in
uhidev and pass the same sizes as part of the attach arguments. Makes
the uhidev drivers a bit less repetitive.

It might look tempting to let uhidev assign the sizes after a driver has
attached, removing the need to repeat this logic in each driver. This
does however not work since the input size must be known while calling
uhidev_open() in order to open the interrupt pipe; and uhidev_open() is
called from several attach routines.

Note that this change only works and applies to when attaching to a
single report ID.

ok jcs@

3 years agoopenssh-7.4 was incorrectly listed twice; spotted by Dmitry
djm [Fri, 10 Sep 2021 05:46:09 +0000 (05:46 +0000)]
openssh-7.4 was incorrectly listed twice; spotted by Dmitry
Belyavskiy, ok dtucker@

3 years agoRemove unused repsizes array.
anton [Fri, 10 Sep 2021 05:46:01 +0000 (05:46 +0000)]
Remove unused repsizes array.

ok jcs@ as part of a larger diff

3 years agoannotate what symbols are used from sys/param.h lines, or delete them
deraadt [Fri, 10 Sep 2021 00:02:43 +0000 (00:02 +0000)]
annotate what symbols are used from sys/param.h lines, or delete them
if not required.  when deleting, add sys/signal.h or other lines which
were not being pulled in

3 years agonothing from sys/param.h is used
deraadt [Fri, 10 Sep 2021 00:01:13 +0000 (00:01 +0000)]
nothing from sys/param.h is used

3 years agothe SunOS lseek 4G wraparound workaround is not needed, consequently
deraadt [Fri, 10 Sep 2021 00:00:55 +0000 (00:00 +0000)]
the SunOS lseek 4G wraparound workaround is not needed, consequently
pulling BSD from sys/param.h is not needed either

3 years agoAdjust for DT binding changes. Add some temporary backwards compatibility
kettenis [Thu, 9 Sep 2021 22:46:03 +0000 (22:46 +0000)]
Adjust for DT binding changes.  Add some temporary backwards compatibility
code to help making the transition.  This will be removed in a few weeks.

3 years agoKeep -? as usage.
nicm [Thu, 9 Sep 2021 21:55:03 +0000 (21:55 +0000)]
Keep -? as usage.

3 years agoadd test for printing empty arguments
jasper [Thu, 9 Sep 2021 20:08:15 +0000 (20:08 +0000)]
add test for printing empty arguments

ok mpi@

3 years agofix crash when passing empty cli arguments as B_AT_NIL wasn't handled as a valid...
jasper [Thu, 9 Sep 2021 20:07:49 +0000 (20:07 +0000)]
fix crash when passing empty cli arguments as B_AT_NIL wasn't handled as a valid argument type

found with afl++
ok mpi@

3 years agoTurn on both button and all mouse modes for menus since some terminals
nicm [Thu, 9 Sep 2021 19:37:17 +0000 (19:37 +0000)]
Turn on both button and all mouse modes for menus since some terminals
only support the former.

3 years agoadd hist() tests similar to mapempty.bt
jasper [Thu, 9 Sep 2021 19:02:50 +0000 (19:02 +0000)]
add hist() tests similar to mapempty.bt

ok mpi@

3 years agoAdd THREAD_PID_OFFSET to tracepoint arguments that pass a TID to userland.
mpi [Thu, 9 Sep 2021 18:41:39 +0000 (18:41 +0000)]
Add THREAD_PID_OFFSET to tracepoint arguments that pass a TID to userland.

Bring these values in sync with the `tid' builtin which already include
the offset.  This is necessary to build script comparing them, like:

tracepoint:sched:enqueue
{
@ts[arg0] = nsecs;
}

tracepoint:sched:on__cpu
/@ts[tid]/
{
latency = nsecs - @ts[tid];
}

Discussed with and ok bluhm@

3 years agoMove a check to avoid panicing on contended rwlock(9) outside of DIAGNOSTIC.
mpi [Thu, 9 Sep 2021 18:23:31 +0000 (18:23 +0000)]
Move a check to avoid panicing on contended rwlock(9) outside of DIAGNOSTIC.

ok kettenis@

3 years agoEnsure that the kill signal undergoing testing is not ignored.
anton [Thu, 9 Sep 2021 17:36:34 +0000 (17:36 +0000)]
Ensure that the kill signal undergoing testing is not ignored.

ok bluhm@

3 years agoIn HTML output, in cells with an "n" (number) layout, pad numbers
schwarze [Thu, 9 Sep 2021 16:50:57 +0000 (16:50 +0000)]
In HTML output, in cells with an "n" (number) layout, pad numbers
on the right side with UTF-8 punctuation and figure spaces such
that numbers in different tbl(7) rows align at the decimal point.

The exact HTML output format was suggested
by <Oliver dot Corff at email dot de>;
the implementation in C is mine.

3 years agoSimplify logic in USER_edit() loop to eliminate unnecessary
krw [Thu, 9 Sep 2021 15:21:39 +0000 (15:21 +0000)]
Simplify logic in USER_edit() loop to eliminate unnecessary
goto's.

Rename CMD_SAVE to CMD_QUIT to reflect actual command (Xquit) and
help and man page verbiage.

No intentional functional change.

3 years agoWhen calling the legacy callback, ensure we catch the case where it
beck [Thu, 9 Sep 2021 15:09:43 +0000 (15:09 +0000)]
When calling the legacy callback, ensure we catch the case where it
has decided to change a succeess to a failure and change the error code.

Fixes a regression in the openssl-ruby tests which expect to test this
functionality.

ok tb@

3 years agoRework openssl-ruby-tests to run all passing tests first, then
tb [Thu, 9 Sep 2021 15:02:33 +0000 (15:02 +0000)]
Rework openssl-ruby-tests to run all passing tests first, then
run the one failing test as a separate regress test. This way,
all regressions should be caught with REGRESS_FAIL_EARLY=yes or
on bluhm's regress webpage.

This needs an up-to-date openssl-ruby-tests package and an
upcoming commit by beck in x509_verify.c to work.

ok beck bluhm

3 years agoIf the layout or data of an individual cell in a tbl(7) contains
schwarze [Thu, 9 Sep 2021 14:45:18 +0000 (14:45 +0000)]
If the layout or data of an individual cell in a tbl(7) contains
only "_", "-", or "=", requesting a horizontal line to be drawn
across the middle of the cell, print <hr/> in that cell in HTML
output.

That is arguably slightly ugly because HTML 5 regards <hr/> as
semantic markup, meaning "thematic break".  If somebody knowns
a better way to render a horizontal line across the middle of a
table cell with pure HTML and CSS, and without implying a specific
meaning, please tell me.

Missing feature reported by <Oliver dot Corff at email dot de>.

3 years agoRework how various OIDs are compared in the code.
claudio [Thu, 9 Sep 2021 14:15:49 +0000 (14:15 +0000)]
Rework how various OIDs are compared in the code.

Instead of converting the ASN1_OBJECT into a string and comparing the
strings, convert the string into an ASN1_OBJECT once and then compare
these objects with OBJ_cmp().

Makes the code a bit easier to read and removes some repetitive conversions.
With input and OK tb@

3 years agoExpand argument to run-shell again.
nicm [Thu, 9 Sep 2021 13:38:32 +0000 (13:38 +0000)]
Expand argument to run-shell again.

3 years agoNo need to initialize nuv, it is assigned to before use.
claudio [Thu, 9 Sep 2021 13:02:36 +0000 (13:02 +0000)]
No need to initialize nuv, it is assigned to before use.

3 years agoScan unit_types[] array using nitems() and eliminate the NULL
krw [Thu, 9 Sep 2021 12:36:45 +0000 (12:36 +0000)]
Scan unit_types[] array using nitems() and eliminate the NULL
sentinal entry.

No functional change.

3 years agozap trailing whitespace
tb [Thu, 9 Sep 2021 12:14:22 +0000 (12:14 +0000)]
zap trailing whitespace

3 years agozap trailing whitespace
tb [Thu, 9 Sep 2021 12:12:39 +0000 (12:12 +0000)]
zap trailing whitespace

3 years agoRelax the check for accessing variable before setting them.
mpi [Thu, 9 Sep 2021 12:09:11 +0000 (12:09 +0000)]
Relax the check for accessing variable before setting them.

First we can't assume rules are written in the order they will be executed.
Secondly filters might need to check variables before they had a chance to
be populated by the right event.

3 years agoDo not abuse the "cleandir" target to delete part of the test output.
schwarze [Thu, 9 Sep 2021 11:48:06 +0000 (11:48 +0000)]
Do not abuse the "cleandir" target to delete part of the test output.
According to /usr/share/mk/bsd.README, that target is only intended
to delete tags files, and there are no tags files here.  Instead,
make sure the "clean" target does not leave any test output behind.

Issue reported and patch OK'ed by bluhm@.

3 years agoLet ba2str() handle unitialized map just like ba2long() does.
mpi [Thu, 9 Sep 2021 11:20:40 +0000 (11:20 +0000)]
Let ba2str() handle unitialized map just like ba2long() does.

3 years agoHook test for command line arguments support.
mpi [Thu, 9 Sep 2021 10:00:04 +0000 (10:00 +0000)]
Hook test for command line arguments support.

3 years agoTest assigning multiple probes to a single action.
mpi [Thu, 9 Sep 2021 09:56:32 +0000 (09:56 +0000)]
Test assigning multiple probes to a single action.

This relies on the 'interval' provider which should always be present.

3 years agoMake it possible to associate multiple probes to a single rule.
mpi [Thu, 9 Sep 2021 09:53:11 +0000 (09:53 +0000)]
Make it possible to associate multiple probes to a single rule.

The following syntax, reducing duplication, is now allowed:

END,
interval:hz:2
{
...
}

Rule descriptors are now linked to a list of probe descriptors instead of
a single one.  Enabled kernel probes are now linked to btrace(8) probe
descriptors.

While here stop parsing filter and probe if debug is not enabled.

3 years agoHandle new NIL type in ba_name().
mpi [Thu, 9 Sep 2021 09:43:49 +0000 (09:43 +0000)]
Handle new NIL type in ba_name().

Fix filter debugging.

3 years agoPrepare for executing regress triggering kernel probes.
mpi [Thu, 9 Sep 2021 09:38:38 +0000 (09:38 +0000)]
Prepare for executing regress triggering kernel probes.

root and kern.allowdt=1 are needed for those, otherwise mark them as SKIPPED.

3 years agoDo not treat empty map has invalid.
mpi [Thu, 9 Sep 2021 07:17:31 +0000 (07:17 +0000)]
Do not treat empty map has invalid.

3 years agoUse a dynamic buffer to allow parsing scripts bigger than BUFSIZ.
mpi [Thu, 9 Sep 2021 07:03:10 +0000 (07:03 +0000)]
Use a dynamic buffer to allow parsing scripts bigger than BUFSIZ.

ok jasper@

3 years agoCheck clear(), delete(), zero() and print() after clear().
mpi [Thu, 9 Sep 2021 06:59:51 +0000 (06:59 +0000)]
Check clear(), delete(), zero() and print() after clear().

ok jasper@

3 years agoMake sure all map operators work after calling clear() on a map.
mpi [Thu, 9 Sep 2021 06:58:39 +0000 (06:58 +0000)]
Make sure all map operators work after calling clear() on a map.

Fix assertions found by jasper@ with AFL++.

ok jasper@

3 years agoFix parsing of aliases again (GitHub issue 2842), also make argument
nicm [Thu, 9 Sep 2021 06:57:48 +0000 (06:57 +0000)]
Fix parsing of aliases again (GitHub issue 2842), also make argument
parsing a bit simpler and fix the names of some client flags.

3 years ago- move CAVEATS to its correct order
jmc [Thu, 9 Sep 2021 06:17:39 +0000 (06:17 +0000)]
- move CAVEATS to its correct order
- use the term "legacy" protocol rather than "original", as the latter
made the text misleading
- uppercase SCP

ok djm

3 years agodrm: Copy drm_wait_vblank to user before returning
jsg [Thu, 9 Sep 2021 04:09:41 +0000 (04:09 +0000)]
drm: Copy drm_wait_vblank to user before returning

From Mark Yacoub
6fd6e20520ccd05a1ac3321404dd498cc28576cb in linux 5.10.y/5.10.62
fa0b1ef5f7a694f48e00804a391245f3471aa155 in mainline linux

3 years agodrm/amd/pm: change the workload type for some cards
jsg [Thu, 9 Sep 2021 04:06:55 +0000 (04:06 +0000)]
drm/amd/pm: change the workload type for some cards

From Kenneth Feng
b00ca567579a4c2f9a4cd6f9a63946f793e8b506 in linux 5.10.y/5.10.62
93c5701b00d50d192ce2247cb10d6c0b3fe25cd8 in mainline linux

3 years agoRevert "drm/amd/pm: fix workload mismatch on vega10"
jsg [Thu, 9 Sep 2021 04:04:45 +0000 (04:04 +0000)]
Revert "drm/amd/pm: fix workload mismatch on vega10"

From Kenneth Feng
3c37ec4350220a548ffc6753646913899e86b1c7 in linux 5.10.y/5.10.62
2fd31689f9e44af949f60ff4f8aca013e628ab81 in mainline linux

3 years agodrm/i915: Fix syncmap memory leak
jsg [Thu, 9 Sep 2021 04:02:53 +0000 (04:02 +0000)]
drm/i915: Fix syncmap memory leak

From Matthew Brost
257ea8a5edc04d5199db83137fbaa24e1de98e9e in linux 5.10.y/5.10.62
a63bcf08f0efb5348105bb8e0e1e8c6671077753 in mainline linux

3 years agodrm/amdgpu: Cancel delayed work when GFXOFF is disabled
jsg [Thu, 9 Sep 2021 04:01:05 +0000 (04:01 +0000)]
drm/amdgpu: Cancel delayed work when GFXOFF is disabled

From Michel Daenzer
da3067eadcc156b742657c0694beae0a7c49d157 in linux 5.10.y/5.10.62
32bc8f8373d2d6a681c96e4b25dca60d4d1c6016 in mainline linux

3 years agoDon't print references to Intel's website for downloading firmware
jcs [Thu, 9 Sep 2021 02:23:14 +0000 (02:23 +0000)]
Don't print references to Intel's website for downloading firmware
when loading fails, it will just confuse people

ok jsg

3 years agoUse the SFTP protocol by default. The original scp/rcp protocol remains
djm [Wed, 8 Sep 2021 23:31:39 +0000 (23:31 +0000)]
Use the SFTP protocol by default. The original scp/rcp protocol remains
available via the -O flag.

Note that ~user/ prefixed paths in SFTP mode require a protocol extension
that was first shipped in OpenSSH 8.7.

ok deraadt, after baking in snaps for a while without incident

3 years agoupdate rge: can do 10/100/1Gb/2.5Gb according to chris
jmc [Wed, 8 Sep 2021 20:33:42 +0000 (20:33 +0000)]
update rge: can do 10/100/1Gb/2.5Gb according to chris

3 years agoattempt to standardise the way we specify speeds in our name description (Nd)
jmc [Wed, 8 Sep 2021 20:29:21 +0000 (20:29 +0000)]
attempt to standardise the way we specify speeds in our name description (Nd)
lines;

sthen and deraadt argued for unit suffixes for speeds 1Gb+
deraadt also requested Gigabit be standardised to 1Gb

ok sthen deraadt ian benno

3 years agoPrepare to provide EC_GROUP_order_bits()
tb [Wed, 8 Sep 2021 17:29:21 +0000 (17:29 +0000)]
Prepare to provide EC_GROUP_order_bits()

ok jsing

3 years agoProvide SSL_SESSION_is_resumable and SSL_set_psk_use_session_callback stubs
tb [Wed, 8 Sep 2021 17:27:33 +0000 (17:27 +0000)]
Provide SSL_SESSION_is_resumable and SSL_set_psk_use_session_callback stubs

ok jsing

3 years agoPrepare to provide API stubs for PHA
tb [Wed, 8 Sep 2021 17:24:23 +0000 (17:24 +0000)]
Prepare to provide API stubs for PHA

ok bcook jsing

3 years agoThe number of elements being processed is known upfront. So allocate the
claudio [Wed, 8 Sep 2021 16:37:20 +0000 (16:37 +0000)]
The number of elements being processed is known upfront. So allocate the
storage needed outside of the loop. This reduces the number of recallocarray
calls.
OK tb@

3 years agoZap _THIS_PORT helper for printf("%n") tracking in ports land
jca [Wed, 8 Sep 2021 15:57:27 +0000 (15:57 +0000)]
Zap _THIS_PORT helper for printf("%n") tracking in ports land

This hack isn't very useful now that libc aborts on printf("%n") calls,
it's expected that the resulting error would lead to a build
failure, and that the coredump along with the syslog message should be
enough to find the culprit.

Hinted by naddy@ and prodded by deraadt@

3 years agoRevert a chunk committed by inadvertence in my last fix.
mpi [Wed, 8 Sep 2021 15:34:01 +0000 (15:34 +0000)]
Revert a chunk committed by inadvertence in my last fix.

3 years agoFix leak in cms_RecipientInfo_kekri_decrypt()
tb [Wed, 8 Sep 2021 14:33:02 +0000 (14:33 +0000)]
Fix leak in cms_RecipientInfo_kekri_decrypt()

Free ec->key before reassigning it.

From OpenSSL 1.1.1, 58e1e397

ok inoguchi

3 years agobtrace(8): add initial support for cli arguments
dv [Wed, 8 Sep 2021 13:29:51 +0000 (13:29 +0000)]
btrace(8): add initial support for cli arguments

This adds initial support for passing cli args to btrace(8) for use
in bt(5) scripts. Similar to bpftrace, they are referenced via $1,
$2, etc. with $# being the number of command line arguments provided.

Adds an initial regress test and a Makefile change to allow providing
arguments to regress tests in a .args file.

Currently no limit is imposed on the number of arguments, keeping
a similar approach as observed in bpftrace. References to undefined
arguments result in a new "nil" type that contextually acts as a
zero or empty string. More work can be done here to improve bpftrace
compatibility.

ok mpi@, jasper@

3 years agoBackout "Merge sysupgrade watchdog and prompt timeout code"
kn [Wed, 8 Sep 2021 13:16:53 +0000 (13:16 +0000)]
Backout "Merge sysupgrade watchdog and prompt timeout code"
(commitid 0SH0ijktpPPcSctj)

"/autoinstall[2697]: start_timeout: not found" during non-interactive
upgrade, e.g. sysupgrade(8).

Reported by Joel Carnat <joel at carnat dot net>, thanks.

3 years agoAdd a missing call to iwx_ctxt_info_free_fw_img() in an error path
stsp [Wed, 8 Sep 2021 13:06:53 +0000 (13:06 +0000)]
Add a missing call to iwx_ctxt_info_free_fw_img() in an error path
of iwx_ctxt_info_init() which should always free on error.

Also, free firmware paging DMA memory in case loading firmware has failed.
If we don't free paging on error we hit KASSERT(dram->paging == NULL)
in iwx_init_fw_sec() once we try to load firmware again.  I have hit
this while debugging firmware load failures during suspend/resume.

ok mpi@

3 years agoMake iwm(4) and iwx(4) raise IPL to splnet() while loading firmware.
stsp [Wed, 8 Sep 2021 13:06:23 +0000 (13:06 +0000)]
Make iwm(4) and iwx(4) raise IPL to splnet() while loading firmware.

ok mpi@

3 years agoPrepare to provide SSL_get_tlsext_status_type()
tb [Wed, 8 Sep 2021 12:56:14 +0000 (12:56 +0000)]
Prepare to provide SSL_get_tlsext_status_type()

Needed for nginx-lua to build with opaque SSL.

ok inoguchi jsing

3 years agoPrepare to provide SSL_set0_rbio()
tb [Wed, 8 Sep 2021 12:32:07 +0000 (12:32 +0000)]
Prepare to provide SSL_set0_rbio()

This is needed for telephony/coturn and telephony/resiprocate to compile
without opaque SSL.

ok inoguchi jsing

3 years agoPrepare to provide BN_bn2{,le}binpad() and BN_lebin2bn()
tb [Wed, 8 Sep 2021 12:19:17 +0000 (12:19 +0000)]
Prepare to provide BN_bn2{,le}binpad() and BN_lebin2bn()

As found by jsg and patrick, this is needed for newer uboot and
will also be used in upcoming elliptic curve work.

This is from OpenSSL 1.1.1l with minor style tweaks.

ok beck inoguchi

3 years agoImprove debug output when sending 802.11 action frames by showing the
stsp [Wed, 8 Sep 2021 11:40:30 +0000 (11:40 +0000)]
Improve debug output when sending 802.11 action frames by showing the
action frame subtypes we care about (i.e. those related to 11n block ack).

ok mpi@

3 years agoPrint correct RTP_PROPOSAL types with 'unwindctl status autoconf'.
tobhe [Wed, 8 Sep 2021 11:38:39 +0000 (11:38 +0000)]
Print correct RTP_PROPOSAL types with 'unwindctl status autoconf'.

Feedback and ok florian@

3 years agoLet iwm(4) and iwx(4) sleep for 1 second while loading firmware.
stsp [Wed, 8 Sep 2021 11:35:08 +0000 (11:35 +0000)]
Let iwm(4) and iwx(4) sleep for 1 second while loading firmware.

Sleeping for 1 second matches what iwn(4) does. Fixes issues where loading
firmware failed for bogus reasons. I could trigger this failure on AX200
with suspend/resume but it was not inherently specific to suspend/resume.

The previous code was looping over tsleep(9) in steps of 100msec.
This could lead to a race where the firmware's alive interrupt fired between
the endtsleep() timeout handler, which marks the sleep timeout as expired,
and sleep_finish(), which reschedules the sleeping thread. The driver would
see EWOULDBLOCK and report an error even though loading firmware did succeed.

ok mpi@

3 years agoReplace bare ; with continue;
job [Wed, 8 Sep 2021 10:49:34 +0000 (10:49 +0000)]
Replace bare ; with continue;

OK tb@

3 years agoFix indentation of comments and labels
job [Wed, 8 Sep 2021 09:49:24 +0000 (09:49 +0000)]
Fix indentation of comments and labels

OK tb@

3 years agocorrect my mistake in previous fix; spotted by halex
djm [Wed, 8 Sep 2021 03:23:44 +0000 (03:23 +0000)]
correct my mistake in previous fix; spotted by halex

3 years agodocument that SFP modules work in SFP+ cards.
benno [Tue, 7 Sep 2021 19:35:41 +0000 (19:35 +0000)]
document that SFP modules work in SFP+ cards.
change the title to show speeds of 100 and 1Gb too.
ok sthen@

3 years agoHook new tests.
mpi [Tue, 7 Sep 2021 19:31:56 +0000 (19:31 +0000)]
Hook new tests.

3 years agoCheck that clear() and zero() only work with map.
mpi [Tue, 7 Sep 2021 19:30:44 +0000 (19:30 +0000)]
Check that clear() and zero() only work with map.

3 years agoCheck that map/hist functions are called with the right argument.
mpi [Tue, 7 Sep 2021 19:29:12 +0000 (19:29 +0000)]
Check that map/hist functions are called with the right argument.

Change the parser to make clear() and zero() accept only local and
global variables as arguments.

Since the parser has no knowledge of the type of a variable abort
the execution if clear() or zero() are being called with something
other than a map or hist.

Fix assertions found by jasper@ with AFL++ (port coming soon!).

ok jasper@

3 years agoTest that syntax errors do not trigger sanity checks.
mpi [Tue, 7 Sep 2021 19:20:22 +0000 (19:20 +0000)]
Test that syntax errors do not trigger sanity checks.

3 years agoReturn early if a parsing error has been found and do not perform any
mpi [Tue, 7 Sep 2021 19:18:08 +0000 (19:18 +0000)]
Return early if a parsing error has been found and do not perform any
sanity check as they might obviously fail.

Fix an assert found by jasper@ with AFL++ (port coming soon!).

ok jasper@

3 years agoclang: add a new warning for %n format specifier usage in printf(3) family functions
semarie [Tue, 7 Sep 2021 17:39:49 +0000 (17:39 +0000)]
clang: add a new warning for %n format specifier usage in printf(3) family functions

ok deraadt@
different versions tested by jca@ naddy@ sthen@

3 years agoFix an infinite loop that could occur during some cases of horizontally
schwarze [Tue, 7 Sep 2021 17:05:30 +0000 (17:05 +0000)]
Fix an infinite loop that could occur during some cases of horizontally
overlapping horizontal spans.  One span would calculate a desired
target width and start preparations for applying it to some columns,
then the other span would overwrite the target width with a different
value and also start preparations for applying that one to some
columns, which could sometimes confuse the code doing the final
distribution to the point of not doing anything at all before
entering the next iteration.

Fix this by making sure the distribution is done step by step, doing
one step at a time rather than allowing multiple steps to conflict.
Specifically, always do the smallest useful step first.  This change
also simplifies the code.  For example, the local "colwidth" array
is no longer needed.

Note that the algorithm still differs from the one implemented in
GNU tbl(1), which appears to not even try to harmonize column widths
but seems to simply distribute the same amount to all constituent
columns, no matter whether their intrinsic width is narrow or wide.
Adopting a GNU-compatible algorithm might allow further simplifiction
in addition to yielding even more similar output, but i do not want
to implement any major changes of the algorithm at this time.

The infinite loop was reported by <Oliver dot Corff at email dot de>.

3 years agoReplace (&(x)) pattern with &x
job [Tue, 7 Sep 2021 16:50:54 +0000 (16:50 +0000)]
Replace (&(x)) pattern with &x

No functional changes.

OK tb@

3 years agoFix NULL pointer dereference introduced by previous commit.
mvs [Tue, 7 Sep 2021 16:07:46 +0000 (16:07 +0000)]
Fix NULL pointer dereference introduced by previous commit.

Reported-by: syzbot+684597dbbb9b516e76ae@syzkaller.appspotmail.com
ok mpi@

3 years agoCorrectly calculate required column widths for tables containing
schwarze [Tue, 7 Sep 2021 14:50:56 +0000 (14:50 +0000)]
Correctly calculate required column widths for tables containing
cells that horizontally span columns which contains "n" (number)
formatted cells on other rows.  This requires updating total column
widths from "n" formatted cells before starting width distribution
from the spanning cells to their constituent columns.

3 years agoFix leak of msg_cert.id_buf. ikev2_msg_cleanup() frees id_buf if we
tobhe [Tue, 7 Sep 2021 14:09:04 +0000 (14:09 +0000)]
Fix leak of msg_cert.id_buf. ikev2_msg_cleanup() frees id_buf if we
don't delete the pointer.

ok markus@

3 years agoFix leak of m if message initialization fails.
tobhe [Tue, 7 Sep 2021 14:06:23 +0000 (14:06 +0000)]
Fix leak of m if message initialization fails.

ok markus@

3 years agoRetry up to 3 times on password authentication failure
jcs [Tue, 7 Sep 2021 13:46:07 +0000 (13:46 +0000)]
Retry up to 3 times on password authentication failure

ok martijn
support from various

3 years agowe already parse the GNU tbl(7) "nospaces" option,
schwarze [Tue, 7 Sep 2021 11:47:42 +0000 (11:47 +0000)]
we already parse the GNU tbl(7) "nospaces" option,
so let it have the intended effect, too

3 years agoRegress targets for PROGS are created by default now. Do not run
bluhm [Tue, 7 Sep 2021 11:10:28 +0000 (11:10 +0000)]
Regress targets for PROGS are created by default now.  Do not run
the program rde_trie_test without a parameter.  Create separate
tests for each option.
OK claudio@

3 years agodo not crash when a tbl(7) cell uses roman font
schwarze [Tue, 7 Sep 2021 10:58:44 +0000 (10:58 +0000)]
do not crash when a tbl(7) cell uses roman font

3 years agoKNF
job [Tue, 7 Sep 2021 10:24:51 +0000 (10:24 +0000)]
KNF

OK tb@ jsing@ beck@

3 years agoThe trap receiver syntax changed.
martijn [Tue, 7 Sep 2021 10:09:28 +0000 (10:09 +0000)]
The trap receiver syntax changed.

Pointed out by and OK bluhm@

3 years agoFix the race between if_detach() and rtm_output().
mvs [Tue, 7 Sep 2021 09:56:00 +0000 (09:56 +0000)]
Fix the race between if_detach() and rtm_output().

When the dying network interface descriptor has if_get(9) obtained
reference owned by foreign thread, the if_detach() thread will sleep
just after it removed this interface from the interface index map.

The data related to this interface is still in routing table, so
if_get(9) called by concurrent rtm_output() thread will return NULL and
the following "ifp != NULL" assertion will be triggered.

So remove the "ifp != NULL" assertions from rtm_output() and try to grab
`ifp' as early as possible then hold it until we finish the work. In the
case we won the race and we have `ifp' non NULL, concurrent if_detach()
thread will wait us. In the case we lost we just return ESRCH.

The problem reported by danj@.
Diff tested by danj@.

ok mpi@