Mateusz Guzik [Tue, 3 Jan 2017 21:02:30 +0000 (21:02 +0000)]
Revert r309619 "ifndef atomic_cas_* in cddl code"
It was a temporary change to ease an import of native atomic_cas primitives.
Instead, atomic_fcmpset was devised with different semantics. See r311168.
Alan Somers [Tue, 3 Jan 2017 17:35:16 +0000 (17:35 +0000)]
misc minor fixes in mpr(4)
sys/dev/mpr/mpr_sas.c
* Fix a potential null pointer dereference (CID 1305731)
* Check for overrun of the ccb_scsiio.cdb_io.cdb_bytes buffer (CID 1211934)
sys/dev/mpr/mpr_sas_lsi.c
* Nullify a dangling pointer in mprsas_get_sata_identify
* Fix a memory leak in mprsas_SSU_to_SATA_devices (CID 1211935)
Sean Bruno [Tue, 3 Jan 2017 14:52:39 +0000 (14:52 +0000)]
Restore r302384 that was dropped when r303816 updated the driver to 1.6.6.-k
Original log:
Do not initialize the adapter on MTU change when adapter status is down.
This fixes long-standing problems when changing settings of the adapter.
John Baldwin [Tue, 3 Jan 2017 01:39:05 +0000 (01:39 +0000)]
Update libsysdecode for getfsstat() 'flags' argument changing to 'mode'.
As a followup to r310638, update libsysdecode (and kdump) to decode the
'mode' argument to getfsstat(). sysdecode_getfsstat_flags() has been
renamed to sysdecode_getfsstat_mode() and now treats the argument as an
enumerated value rather than a mask of flags.
Mark Johnston [Tue, 3 Jan 2017 00:05:44 +0000 (00:05 +0000)]
Add a page queue for holding dirty anonymous unswappable pages.
On systems without a configured swap device, an attempt to launder pages
from a swap object will always fail and result in the page being
reactivated. This means that the page daemon will continuously scan pages
that can never be evicted. With this change, anonymous pages are instead
moved to PQ_UNSWAPPABLE after a failed laundering attempt when no swap
devices are configured. PQ_UNSWAPPABLE is not scanned unless a swap device
is configured, so unreferenced unswappable pages are excluded from the page
daemon's workload.
Enji Cooper [Mon, 2 Jan 2017 20:29:50 +0000 (20:29 +0000)]
Only bake krb5_config.h support in to ssh(3), etc if both MK_GSSAPI and
MK_KERBEROS_SUPPORT != no
This fixes the odd case where someone specified MK_GSSAPI=no and
MK_KERBEROS_SUPPORT=yes (which admittedly, probably doesn't make sense,
but the build system doesn't prevent this case today, and it didn't when
I filed the bug back in 2011 either).
indent(1): add option -P for loading user-provided files as profiles
Without this change, indent(1) would only look to load options from ~/.indent.pro if it's there and -npro wasn't used on the command line. This option lets the user set their own path to the file.
Enji Cooper [Mon, 2 Jan 2017 19:55:18 +0000 (19:55 +0000)]
Move the "MK_* options..." section before the "... MK_*_SUPPORT..." section
For the case that someone set WITHOUT_GSSAPI=, now WITHOUT_KERBEROS_SUPPORT
will be properly set.
This will likely fix the issue for the default case noted in the PR I filed
back in 2011. I am trying to fix the less obvious case documented in the PR
still.
Dimitry Andric [Mon, 2 Jan 2017 19:33:22 +0000 (19:33 +0000)]
Make native-xtools build correctly after clang/llvm 3.9.0 import
During the clang/llvm 3.9.0 import, the build structure for it was
completely revamped. This broke the native-xtools target.
It first attempts to build libllvmminimal, then the llvm-tblgen and
clang-tblgen executables, but these fail to link because they are linked
to the 'full' libllvm by default, as they normally are during the
'world' stage.
To make these link against libllvmminimal instead, define TOOLS_PREFIX,
similarly as during the bootstrap-tools phase. The value itself is
empty, as we don't really want to use a prefix.
Ian Lepore [Mon, 2 Jan 2017 15:19:22 +0000 (15:19 +0000)]
Update ntp.conf to use the ntpd pool feature.
Our previous ntp.conf file configured 3 servers from freebsd.pool.ntp.org
using 3 separate 'server' config lines. That is now replaced with a single
'pool' line which causes ntpd to add multiple servers from the pool.
More than just making the config smaller, the pool feature in ntpd has one
major advantage over configuring 3 separate servers from a pool: if a server
that was added using a 'pool' statement provides bad time (initially or at
some later date), ntpd automatically discards it and configures a new
different server from the pool without needing to be restarted.
These changes also add a 'tos' line to control how many pool servers get
added, a 'restrict source' line that is required to allow ntpd to add new
peers from the pool, and it deletes a 'restrict 127.127.1.0' line that does
nothing and should never have been there (127.127.1.0 is not a valid IP
address, it's a refclock identifier).
Sean Bruno [Mon, 2 Jan 2017 00:56:33 +0000 (00:56 +0000)]
2017 IFLIB updates in preparation for commits to e1000 and ixgbe.
- iflib - add checksum in place support (mmacy)
- iflib - initialize IP for TSO (going to be needed for e1000) (mmacy)
- iflib - move isc_txrx from shared context to softc context (mmacy)
- iflib - Normalize checks in TXQ drainage. (shurd)
- iflib - Fix queue capping checks (mmacy)
- iflib - Fix invalid assert, em can need 2 sentinels (mmacy)
- iflib - let the driver determine what capabilities are set and what
tx csum flags are used (mmacy)
- add INVARIANTS debugging hooks to gtaskqueue enqueue (mmacy)
- update bnxt(4) to support the changes to iflib (shurd)
Some other various, sundry updates. Slightly more verbose changelog:
Remove some uses of the GCC __nonnull() attribute.
While the checks are considered useful, the attribute does dangerous
optimizations, removing NULL checks where they can be needed. Remove the
uses of this attribute introduced in r281130: the changes were inspired on
Google's bionic where this attribute is not used anymore.
The __nonnull() attribute will be deprecrated from our headers and
replaced with the Clang _Nonnull qualifier in the future.
Enji Cooper [Sun, 1 Jan 2017 04:13:24 +0000 (04:13 +0000)]
Look for list.h in ${.CURDIR} to unbreak the build with a ports-based copy
of llvm38 on ^/stable/11 (oh, the bugs you find when you set CC,CXX,CPP
manually and it skips the bootstrap stage for the toolchain...)
Enji Cooper [Sun, 1 Jan 2017 04:01:27 +0000 (04:01 +0000)]
Make sys/vfs/lookup_cap_dotdot actually work with "kyua test"
The tests don't work when reading/writing to file descriptors in the
sandbox after entering capability mode (and wouldn't have, regardless
of the framework), so adjust the tests so they function within the
framework.
For tests that enter capability mode over the course of the test, the
following is now done:
1. Fork child process for capability mode test.
2. In child...
i. Enter capability mode.
ii. Test invariants.
iii. Exit after calling test function.
3. Collect status for child and determine whether or not it completed
successfully.
In order to test the invariants in the child process, they now use assert(3)
instead of ATF_REQUIRE*, as the atf-c-api functions right to results files
in the directories in order to determine where and how tests fail.
While in the area, fix several -Wshadow and -Wunused warnings found when I
bumped WARNS up to 6, and fix some minor style(9) issues with indentation
and type alignment.
Enji Cooper [Sat, 31 Dec 2016 22:46:17 +0000 (22:46 +0000)]
snmp_output_err_resp, snmp_output_resp: allocate `object` using calloc, not
on the stack
Some of the callers try to determine whether or not `object` is valid by
testing the value for NULL, which will never be true if it's a stack value,
so in order to be clear and correct down the call stack, use a heap
allocated object.
This also addresses a Coverity issue by initializing all of `object` via
calloc
Hiroki Sato [Sat, 31 Dec 2016 13:15:52 +0000 (13:15 +0000)]
- Use more descriptive names for variables.
- Set O_CLOEXEC to the signal pipe and /dev/klog.
- Use a single signal handler to catch both SIGHUP and SIGCHLD.
- Fix a bug which did FD_SET() the writer-end of the pipe.
Enji Cooper [Sat, 31 Dec 2016 12:14:25 +0000 (12:14 +0000)]
Set value->v.octetstring.len to a correct value on malloc success/failure
The previous code always set value->v.octetstring.len to len, regardless
of the result from the malloc call. This misleads the caller on malloc
failure. Set .len to len on success and 0 on failure.
Enji Cooper [Sat, 31 Dec 2016 10:01:25 +0000 (10:01 +0000)]
snmp_pdu_free the right object at the right time in snmptool_walk
r310892 was on the right track, but unfortunately it was resolving
the problem incorrectly and accidentally leaking memory in the
process.
- Call snmp_pdu_free on req before calling snmp_pdu_create on it
at the bottom of the outer while loop
- Call snmp_pdu_free on resp after calling snmpwalk_nextpdu_create
in the inner loop
Hiroki Sato [Sat, 31 Dec 2016 06:23:05 +0000 (06:23 +0000)]
Replace two fat signal handlers with function calls in
the main I/O multiplex loop. select() now watches
a pipe which is written by the new skinny signal
handlers and the received signals are handled inside
the loop sequencially.
This eliminates a complex signal mask to guarantee
async-signal safety.
Martin Matuska [Fri, 30 Dec 2016 22:18:22 +0000 (22:18 +0000)]
MFV r310796, r310797:
Sync libarchive with vendor.
Vendor changes (relevant to FreeBSD):
PR #771: Add NFSv4 ACL support to pax and restricted pax
NFSv4 ACL information may now be stored to and restored from tar archives.
ACL must be non-trivial and supported by the underlying filesystem, e.g.
natively by ZFS or by UFS with the NFSv4 ACL enable flag set.
Enji Cooper [Fri, 30 Dec 2016 21:33:01 +0000 (21:33 +0000)]
Unbreak ip_carp with WITHOUT_INET6 enabled by conditionalizing all IPv6
structs under the INET6 #ifdef. Similarly (even though it doesn't seem
to affect the build), conditionalize all IPv4 structs under the INET
#ifdef
This also unbreaks the LINT-NOINET6 tinderbox target on amd64; I have not
verified other MACHINE/TARGET pairs (e.g. armv6/arm).
MFC after: 2 weeks
X-MFC with: r310847
Pointyhat to: jpaetzel
Reported by: O. Hartmann <o.hartmann@walstatt.org>
Toomas Soome [Fri, 30 Dec 2016 19:06:29 +0000 (19:06 +0000)]
dosfs support in libstand is broken since r298230
Apparently the libstand dosfs optimization is a bit too optimistic
and did introduce possible memory corruption.
This patch is backing out the bad part and since this results in
dosfs reading full blocks now, we can also remove extra offset argument
from dv_strategy callback.
The analysis of the issue and the backout patch is provided by Mikhail Kupchik.
Fix two similar bugs in the populate vm_fault() code.
If pager' populate method succeeded, but other thread raced with us
and modified vm_map, we must unbusy all pages busied by the pager,
before we retry the whole fault handling. If pager instantiated more
pages than fit into the current map entry, we must unbusy the pages
which are clipped.
Also do some refactoring, clarify comments and use more clear local
variable names.
Reported and tested by: kargl, subbsd@gmail.com (previous version)
Reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Josh Paetzel [Fri, 30 Dec 2016 18:46:21 +0000 (18:46 +0000)]
Harden CARP against network loops.
If there is a loop in the network a CARP that is in MASTER state will see it's
own broadcasts, which will then cause it to assume BACKUP state. When it
assumes BACKUP it will stop sending advertisements. In that state it will no
longer see advertisements and will assume MASTER...
We can't catch all the cases where we are seeing our own CARP broadcast, but
we can catch the obvious case.