]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
10 years agopatch: cleanup some unnecessary cruft.
Pedro F. Giffuni [Sat, 14 Jun 2014 01:58:33 +0000 (01:58 +0000)]
patch: cleanup some unnecessary cruft.

- Drop some bogus casts to size_t.
- The new_p_foo variables are not needed after r267426.

Pointed out by: bde
MFC after: 1 week

10 years agoUndo commit r267455, which was done in the wrong directory.
Edwin Groothuis [Sat, 14 Jun 2014 01:18:02 +0000 (01:18 +0000)]
Undo commit r267455, which was done in the wrong directory.

10 years agoSwitch release/Makefile knobs from WITHOUT_FOO=1 to MK_FOO=no.
Glen Barber [Sat, 14 Jun 2014 01:03:24 +0000 (01:03 +0000)]
Switch release/Makefile knobs from WITHOUT_FOO=1 to MK_FOO=no.

This allows greater granular control over the release
build, and avoids WITH_FOO=1 and WITHOUT_FOO=1 collision.

This change was verified to produce correct results
compared to an earlier build by diffing find(1) output
of the resulting bootonly/, dvd/, and release/ directories.

Tested on: head@r267449
MFC after: 1 month
X-MFC-Note: Requires commits not yet MFC'd, so likely
longer than 1 month.
Sponsored by: The FreeBSD Foundation

10 years agoAdded:
Stephen Hurd [Sat, 14 Jun 2014 00:10:10 +0000 (00:10 +0000)]
Added:

- my birthday

Approved by: db, skreuzer (mentors)

10 years agoVendor import of tzdata2014e changes
Edwin Groothuis [Fri, 13 Jun 2014 23:56:47 +0000 (23:56 +0000)]
Vendor import of tzdata2014e changes

Fix historical data for Egypt.
Better prediction for future Egypt / Morocco changes.
Update to Cocos / Cook islands.
Fix historical data for Russia.

Obtained from: ftp://ftp.iana.org/tz/releases/

10 years agoCorrect variable for loader tunable variable hw.mfi.mrsas_enable.
Xin LI [Fri, 13 Jun 2014 22:33:41 +0000 (22:33 +0000)]
Correct variable for loader tunable variable hw.mfi.mrsas_enable.

Submitted by: Sascha Wildner (DragonFly)
MFC after: 3 days

10 years ago- Don't check the dma map address for a static DMA mapping against NULL
John Baldwin [Fri, 13 Jun 2014 19:40:02 +0000 (19:40 +0000)]
- Don't check the dma map address for a static DMA mapping against NULL
  to determine if the mapping is valid.
- Don't explicitly destroy the dmap map for static DMA mappings.

Reviewed by: adrian

10 years agoDon't check maps for static DMA allocations against NULL to see if an
John Baldwin [Fri, 13 Jun 2014 19:34:34 +0000 (19:34 +0000)]
Don't check maps for static DMA allocations against NULL to see if an
allocation succeeded.  Instead, check the tag against NULL.

10 years agoAdd missing calls to bus_dmamap_unload() when freeing static DMA
John Baldwin [Fri, 13 Jun 2014 18:20:44 +0000 (18:20 +0000)]
Add missing calls to bus_dmamap_unload() when freeing static DMA
allocations.

Reviewed by: scottl

10 years agoAdd Jung-uk Kim as co-maintainer of OpenSSL per discussion on
Xin LI [Fri, 13 Jun 2014 18:07:42 +0000 (18:07 +0000)]
Add Jung-uk Kim as co-maintainer of OpenSSL per discussion on
-developers and agreed by benl@.

10 years agoDon't install GSS-API headers when the GSSAPI option has been disabled.
Tijl Coosemans [Fri, 13 Jun 2014 10:08:18 +0000 (10:08 +0000)]
Don't install GSS-API headers when the GSSAPI option has been disabled.
Some ports assume GSS-API is supported when they find the headers.

PR: 189156
Submitted by: Garrett Cooper <yanegomi@gmail.com>

10 years agoAttach the CUSE library and kernel module to the default FreeBSD
Hans Petter Selasky [Fri, 13 Jun 2014 08:53:49 +0000 (08:53 +0000)]
Attach the CUSE library and kernel module to the default FreeBSD
builds. Bump the FreeBSD version number.

10 years ago- Make invalids variable local to do_conv such that it prints the number
Tijl Coosemans [Fri, 13 Jun 2014 08:44:03 +0000 (08:44 +0000)]
- Make invalids variable local to do_conv such that it prints the number
  of invalid characters of the current file instead of an accumulated
  value.
- Make do_conv return an error when invalid characters have been found.
  Return EXIT_FAILURE from main if any file contained invalid characters.
  This matches the behaviour of GNU iconv.
- Mark usage with __dead2 attribute.
- Make the long_options array const.

10 years agoiconv_open: initialise ci_ilseq_invalid field of _citrus_iconv_shared
Tijl Coosemans [Fri, 13 Jun 2014 08:36:10 +0000 (08:36 +0000)]
iconv_open: initialise ci_ilseq_invalid field of _citrus_iconv_shared
struct after allocation with malloc.

iconvlist: reduce a memory leak by copying strings only once.

10 years agoReplace malloc+memset with calloc.
Tijl Coosemans [Fri, 13 Jun 2014 08:28:51 +0000 (08:28 +0000)]
Replace malloc+memset with calloc.

10 years agoReplace malloc+memset with calloc.
Tijl Coosemans [Fri, 13 Jun 2014 08:26:53 +0000 (08:26 +0000)]
Replace malloc+memset with calloc.

10 years agoRevert r267424 and r267425:
Glen Barber [Fri, 13 Jun 2014 00:05:06 +0000 (00:05 +0000)]
Revert r267424 and r267425:
  More exploration will be needed in order to silence the
  tinderbox mails caused by setting WITH_TESTS=1 by default.

Ref: http://tinderbox.freebsd.org/tinderbox-head-build-HEAD-amd64-amd64.full
With hat: re

10 years agoFix some leaks on LUN creation error.
Alexander Motin [Thu, 12 Jun 2014 21:50:46 +0000 (21:50 +0000)]
Fix some leaks on LUN creation error.

MFC after: 2 weeks

10 years agoAvoid zeroing during allocation.
Pedro F. Giffuni [Thu, 12 Jun 2014 19:01:57 +0000 (19:01 +0000)]
Avoid zeroing during allocation.

This change reverts a change from OpenBSD which made use of
calloc, and therefore wasted time initializing arrays that
will later be realloc'ed. Consistently use FreeBSD's
reallocf().

While here also merge the changes from OpenBSD's manpage
patch.1 Rev 1.27:
"patch was moved from user portability (UP) to base in issue 7
and is no longer optional"

MFC after: 1 week

10 years agoRegen after r267424 to replace WITH_TESTS with WITHOUT_TESTS.
Julio Merino [Thu, 12 Jun 2014 18:59:25 +0000 (18:59 +0000)]
Regen after r267424 to replace WITH_TESTS with WITHOUT_TESTS.

10 years agoEnable MK_TESTS by default.
Julio Merino [Thu, 12 Jun 2014 18:54:52 +0000 (18:54 +0000)]
Enable MK_TESTS by default.

This was never intended to be off by default but was done this way
because the initial patches broke the build.  Things seem stable now
(have been so for a while too) and "make tinderbox" is clean so let's
try again.

Announced in freebsd-current; all reported shortcomings have been
addressed.

10 years agoBuild vt(4) fonts during buildworld
Ed Maste [Thu, 12 Jun 2014 18:31:32 +0000 (18:31 +0000)]
Build vt(4) fonts during buildworld

vtfontcvt(8) is now built during buildworld, so can be used as a
bootstrap tool to create vt(4) fonts from source .hex or .bdf font
files, rather than having uuencoded binary fonts in the tree.

Sponsored by: The FreeBSD Foundation

10 years agoFix a typo in note for r266209, pmcstat(8)
Glen Barber [Thu, 12 Jun 2014 17:19:57 +0000 (17:19 +0000)]
Fix a typo in note for r266209, pmcstat(8)

Spotted by: emaste
Sponsored by: The FreeBSD Foundation

10 years agoclear the write bit... This allows my AVILA board to survive a
John-Mark Gurney [Thu, 12 Jun 2014 16:31:15 +0000 (16:31 +0000)]
clear the write bit...   This allows my AVILA board to survive a
portsnap extract, where previously it would panic..  clearly someone
who knows pmap should optimize this code per alc's comment...

Submitted by: alc
MFC after: probably

10 years agoAdd thin versions of VGAROM 8x8 and 8x16 fonts for vt(4)
Ed Maste [Thu, 12 Jun 2014 15:23:51 +0000 (15:23 +0000)]
Add thin versions of VGAROM 8x8 and 8x16 fonts for vt(4)

These are converted from syscons(4) cp437-thin-8x* fonts.

10 years agoAdd myself to the calendar
Rodrigo Osorio [Thu, 12 Jun 2014 14:20:58 +0000 (14:20 +0000)]
Add myself to the calendar

Approved by: bapt (mentor)

10 years ago- Fix out of range shifting bug in bitops.h.
Hans Petter Selasky [Thu, 12 Jun 2014 13:33:01 +0000 (13:33 +0000)]
- Fix out of range shifting bug in bitops.h.
- Make code a bit easier to read by adding parenthesis.

MFC after: 3 days
Sponsored by: Mellanox Technologies

10 years agoImplement simple direct-mapped cache for popular filesystem identifiers to
Alexander Motin [Thu, 12 Jun 2014 12:43:48 +0000 (12:43 +0000)]
Implement simple direct-mapped cache for popular filesystem identifiers to
avoid congestion on global mountlist_mtx mutex in vfs_busyfs(), while
traversing through the list of mount points.

This change significantly improves NFS server scalability, since it had
to do this translation for every request, and the global lock becomes quite
congested.

This code is more optimized for relatively small number of mount points.
On systems with hundreds of active mount points this simple cache may have
many collisions.  But the original traversal code in that case should also
behave much worse, so we are not loosing much.

Reviewed by: attilio
MFC after: 2 weeks
Sponsored by: iXsystems, Inc.

10 years agoIntroduce new "256 Bucket" zone to split requests and reduce congestion
Alexander Motin [Thu, 12 Jun 2014 11:57:07 +0000 (11:57 +0000)]
Introduce new "256 Bucket" zone to split requests and reduce congestion
on "128 Bucket" zone lock.

MFC after: 2 weeks
Sponsored by: iXsystems, Inc.

10 years agoEnable SMP for both Exynos5 models we support.
Ruslan Bukin [Thu, 12 Jun 2014 11:44:26 +0000 (11:44 +0000)]
Enable SMP for both Exynos5 models we support.

10 years agoActivate IRQ 30 (non-secure private timer IRQ) for case we are running
Ruslan Bukin [Thu, 12 Jun 2014 11:41:04 +0000 (11:41 +0000)]
Activate IRQ 30 (non-secure private timer IRQ) for case we are running
in non-secure state.

10 years agoAdd machine-dependent SMP part for Exynos5420
Ruslan Bukin [Thu, 12 Jun 2014 11:37:38 +0000 (11:37 +0000)]
Add machine-dependent SMP part for Exynos5420

10 years agoAllocating new bucket for bucket zone, never take it from the zone itself,
Alexander Motin [Thu, 12 Jun 2014 11:36:22 +0000 (11:36 +0000)]
Allocating new bucket for bucket zone, never take it from the zone itself,
since it will almost certanly fail.  Take next bigger zone instead.

This situation should not happen with original bucket zones configuration:
"32 Bucket" zone uses "64 Bucket" and vice versa.  But if "64 Bucket" zone
lock is congested, zone may grow its bucket size and start biting itself.

MFC after: 2 weeks
Sponsored by: iXsystems, Inc.

10 years agoRemove some Giant unlocks I missed in r267290.
John Baldwin [Thu, 12 Jun 2014 11:21:18 +0000 (11:21 +0000)]
Remove some Giant unlocks I missed in r267290.

10 years ago- Don't compare bus_dma map pointers for static DMA allocations against
John Baldwin [Thu, 12 Jun 2014 11:15:19 +0000 (11:15 +0000)]
- Don't compare bus_dma map pointers for static DMA allocations against
  NULL to determine if bus_dmamap_unload() or bus_dmamem_free() should be
  called.  Instead, check the associated bus and virtual addresses.
- Don't clear static DMA maps to NULL.

Reviewed by: jfv

10 years agoFix path for zlib examples, this have no runtime effect and was
Xin LI [Wed, 11 Jun 2014 20:48:40 +0000 (20:48 +0000)]
Fix path for zlib examples, this have no runtime effect and was
overlooked when I was doing zlib update.

MFC after: 2 weeks

10 years ago- Unmap static DMA buffers allocated via bus_dmemem_alloc() before
John Baldwin [Wed, 11 Jun 2014 20:46:23 +0000 (20:46 +0000)]
- Unmap static DMA buffers allocated via bus_dmemem_alloc() before
  freeing them instead of after.
- Check the bus address of a static DMA buffer to decide if the associated
  map should be unloaded.
- Don't try to destroy bus dma maps for static DMA buffers.

Reviewed by: davidcs

10 years agoAdd a zlib pkg-config file (more and more ports requires it)
Baptiste Daroussin [Wed, 11 Jun 2014 20:44:54 +0000 (20:44 +0000)]
Add a zlib pkg-config file (more and more ports requires it)

Approved by: delphij
MFC after: 1 week

10 years agoUnload maps for DMA buffers allocated via bus_dmamem_alloc() before
John Baldwin [Wed, 11 Jun 2014 20:44:02 +0000 (20:44 +0000)]
Unload maps for DMA buffers allocated via bus_dmamem_alloc() before
freeing the buffers.

Reviewed by: davidcs

10 years agoFix decoding of near CALL when address-size prefix (67h) is present.
Xin LI [Wed, 11 Jun 2014 20:38:14 +0000 (20:38 +0000)]
Fix decoding of near CALL when address-size prefix (67h) is present.

Submitted by: Wolf Ramovsky <wolf.ramovsky gmail.com>
via core (peter)
MFC after: 2 weeks

10 years agoFix decoding of LEA when address-size prefix (67h) is present.
Xin LI [Wed, 11 Jun 2014 20:31:41 +0000 (20:31 +0000)]
Fix decoding of LEA when address-size prefix (67h) is present.

This is loosly based on Xorg changeset f57bc0e by Christian
Zander.

Submitted by: Wolf Ramovsky <wolf.ramovsky gmail.com>
via core (peter)
MFC after: 2 weeks

10 years agoFix decoding of MOVSX.
Xin LI [Wed, 11 Jun 2014 20:22:14 +0000 (20:22 +0000)]
Fix decoding of MOVSX.

Submitted by: Wolf Ramovsky <wolf.ramovsky gmail.com>
via core (peter)
MFC after: 2 weeks

10 years agoUse Mt macro to properly format mailto links (patch from Franco Fichtner <franco...
Baptiste Daroussin [Wed, 11 Jun 2014 20:20:14 +0000 (20:20 +0000)]
Use Mt macro to properly format mailto links (patch from Franco Fichtner <franco@lastsummer.de> for Dragonfly)

Reviewed by: des
Approved by: des
Obtained from: Dragonfly
MFC after: 1 week

10 years agoApply vendor fixes to the High Point drivers:
Xin LI [Wed, 11 Jun 2014 18:24:51 +0000 (18:24 +0000)]
Apply vendor fixes to the High Point drivers:

 - Don't call xpt_free_path() in os_query_remove_device() and
   always return TRUE.
 - Update os_buildsgl() to support build logical SG table which
   will be used by lower RAID module.
 - Return CAM_SEL_TIMEOUTstatus for SCSIcommand failed as target
   missing.

Many thanks to HighPoint for providing this driver update.

Submitted by: Steve Chang
Reviewed by: mav
MFC after: 3 days

10 years agoAvoid leaking file pointer on error
Ed Maste [Wed, 11 Jun 2014 17:19:57 +0000 (17:19 +0000)]
Avoid leaking file pointer on error

CID: 12225061222505

10 years agoFix a few more drivers that were explicitly destroying the bus_dma
John Baldwin [Wed, 11 Jun 2014 17:03:14 +0000 (17:03 +0000)]
Fix a few more drivers that were explicitly destroying the bus_dma
map created by bus_dmamem_alloc().

10 years agoCorrect a bug in the management of the population map on big-endian
Alan Cox [Wed, 11 Jun 2014 16:11:12 +0000 (16:11 +0000)]
Correct a bug in the management of the population map on big-endian
machines.  Specifically, there was a mismatch between how the routine
allocation and deallocation operations accessed the population map
and how the aggressively optimized reservation-breaking operation
accessed it.  So, problems only occurred when reservations were broken.
This change makes the routine operations access the population map in
the same way as the reservation breaking operation.

This bug was introduced in r259999.

PR: 187080
Tested by: jmg (on an "armeb" machine)
Sponsored by: EMC / Isilon Storage Division

10 years agoFix various NIC drivers to properly cleanup static DMA resources.
John Baldwin [Wed, 11 Jun 2014 14:53:58 +0000 (14:53 +0000)]
Fix various NIC drivers to properly cleanup static DMA resources.
In particular, don't check the value of the bus_dma map against NULL
to determine if either bus_dmamem_alloc() or bus_dmamap_load() succeeded.
Instead, assume that bus_dmamap_load() succeeeded (and thus that
bus_dmamap_unload() should be called) if the bus address for a resource
is non-zero, and assume that bus_dmamem_alloc() succeeded (and thus
that bus_dmamem_free() should be called) if the virtual address for a
resource is not NULL.

In many cases these bugs could result in leaks when a driver was detached.

Reviewed by: yongari
MFC after: 2 weeks

10 years agoRemove unneeded mountlist_mtx acquisition from sync_fsync().
Alexander Motin [Wed, 11 Jun 2014 12:56:49 +0000 (12:56 +0000)]
Remove unneeded mountlist_mtx acquisition from sync_fsync().

All struct mount fields accessed by sync_fsync() are protected by MNT_MTX.

10 years agoImprove logic besides net.bpf.optimize_writers.
Alexander V. Chernikov [Wed, 11 Jun 2014 11:27:44 +0000 (11:27 +0000)]
Improve logic besides net.bpf.optimize_writers.
Direct bpf(4) consumers should now work fine with this tunable turned on.
In fact, the only case when optimized_writers can change program
behavior is direct bpf(4) consumer setting its read filter to
catch-all one.

MFC after: 2 weeks
Sponsored by: Yandex LLC

10 years agoAdd disklabel64 support
Andrey V. Elsukov [Wed, 11 Jun 2014 10:48:11 +0000 (10:48 +0000)]
Add disklabel64 support

MFC after: 2 weeks

10 years agoAdd disklabel64 support to GEOM_PART class.
Andrey V. Elsukov [Wed, 11 Jun 2014 10:42:34 +0000 (10:42 +0000)]
Add disklabel64 support to GEOM_PART class.

This partitioning scheme is used in DragonFlyBSD. It is similar to
BSD disklabel, but has the following improvements:
* metadata has own dedicated place and isn't accessible through partitions;
* all offsets are 64-bit;
* supports 16 partitions by default (has reserved place for more);
* has reserved place for backup label (but not yet implemented);
* has UUIDs for partitions and partition types;

No objections from: geom
MFC after: 2 weeks
Relnotes: yes

10 years agoAllow swapping to DragonFlyBSD's swap partition.
Andrey V. Elsukov [Wed, 11 Jun 2014 10:23:49 +0000 (10:23 +0000)]
Allow swapping to DragonFlyBSD's swap partition.

MFC after: 2 weeks

10 years agoAdd aliases for DragonFlyBSD's partition types.
Andrey V. Elsukov [Wed, 11 Jun 2014 10:19:11 +0000 (10:19 +0000)]
Add aliases for DragonFlyBSD's partition types.

MFC after: 2 weeks

10 years agoAdd DragonFlyBSD's Hammer FS types and type names.
Andrey V. Elsukov [Wed, 11 Jun 2014 10:14:40 +0000 (10:14 +0000)]
Add DragonFlyBSD's Hammer FS types and type names.

MFC after: 2 weeks

10 years agoAdd UUIDs for DragonFlyBSD's partition types.
Andrey V. Elsukov [Wed, 11 Jun 2014 10:13:08 +0000 (10:13 +0000)]
Add UUIDs for DragonFlyBSD's partition types.

MFC after: 2 weeks

10 years agoAdd a driver for Serial Peripheral Interface (SPI).
Ruslan Bukin [Wed, 11 Jun 2014 10:03:06 +0000 (10:03 +0000)]
Add a driver for Serial Peripheral Interface (SPI).

10 years agoMove root_mount_hold() functionality to separate mutex.
Alexander Motin [Wed, 11 Jun 2014 08:14:08 +0000 (08:14 +0000)]
Move root_mount_hold() functionality to separate mutex.

It has nothing to share with mutex protecting list of mounted file systems.

10 years agoUse bitwise OR instead of logical OR when constructing value for
Jim Harris [Tue, 10 Jun 2014 21:40:43 +0000 (21:40 +0000)]
Use bitwise OR instead of logical OR when constructing value for
SET_FEATURES/NUMBER_OF_QUEUES command.

Sponsored by: Intel
MFC after: 3 days

10 years agoDon't destroy bus_dma maps created by bus_dmamem_alloc(). In some cases,
John Baldwin [Tue, 10 Jun 2014 20:25:45 +0000 (20:25 +0000)]
Don't destroy bus_dma maps created by bus_dmamem_alloc().  In some cases,
don't create a map before calling bus_dmamem_alloc() (such maps were
leaked).  It is believed that the extra destroy of the map was generally
harmless since bus_dmamem_alloc() often uses special maps for which
bus_dmamap_destroy() is a no-op (e.g. on x86).

Reviewed by: scottl

10 years agoReplace enum forward declarations with complete definitions.
Tycho Nightingale [Tue, 10 Jun 2014 18:46:00 +0000 (18:46 +0000)]
Replace enum forward declarations with complete definitions.

Reviewed by: neel

10 years agovt fontcvt: move to usr.bin/vtfontcvt
Ed Maste [Tue, 10 Jun 2014 18:29:45 +0000 (18:29 +0000)]
vt fontcvt: move to usr.bin/vtfontcvt

vtfontcvt is useful for end users to convert arbitrary bitmap fonts
for use by vt(4).  It can also be used as a build tool, allowing us
to keep the source font data in the src tree rather than uuencoded
binaries.

Reviewed by: ray, wblock (D183)
Sponsored by: The FreeBSD Foundation

10 years agoPF_BLUETOOTH protocols: skip initialization of non-virtualized globals
Mikolaj Golub [Tue, 10 Jun 2014 18:21:37 +0000 (18:21 +0000)]
PF_BLUETOOTH protocols: skip initialization of non-virtualized globals
for non-default VNET instances.

This fixes panic on a vnet initialization when ng_btsocket is loaded.

MFC after: 1 week

10 years agoIn some scenarios, when generating llvm/clang .inc.h files, multiple
Dimitry Andric [Tue, 10 Jun 2014 17:54:24 +0000 (17:54 +0000)]
In some scenarios, when generating llvm/clang .inc.h files, multiple
source files could be passed to tblgen or clang-tblgen, leading to a
"Too many positional arguments specified" error message.  Fix this by
replacing the too-generic ${.ALLSRC} sources with explicit paths.

Reported by: rysto32@gmail.com, rodrigc
MFC after: 3 days

10 years agoStrip out the test suite from the installation media.
Julio Merino [Tue, 10 Jun 2014 17:05:41 +0000 (17:05 +0000)]
Strip out the test suite from the installation media.

This affects the disc1.iso, dvd1.iso and bootonly.iso files.

Obtained from: gjb

10 years agoPut the test suite in its own tests.txz distribution file.
Julio Merino [Tue, 10 Jun 2014 17:04:30 +0000 (17:04 +0000)]
Put the test suite in its own tests.txz distribution file.

Force all the contents of /usr/tests to go into a separate distribution
file so that users of binary releases can easily choose to not install it.

To make this possible, we need two fixes:
- bsd.subdir.mk needs to properly honor NO_SUBDIR in all cases so that we
  do not recurse into 'tests' subdirectories when we needn't.  Otherwise,
  we end up with some Kyuafiles in base.txz.
- etc/Makefile needs to skip installing tests in its 'distribute' target
  so that a Kyuafile doesn't leak into base.txz.

Approved by: gjb

10 years agoAdd helper functions to populate VM exit information for rendezvous and
Neel Natu [Tue, 10 Jun 2014 16:45:58 +0000 (16:45 +0000)]
Add helper functions to populate VM exit information for rendezvous and
astpending exits. This is to reduce code duplication between VT-x and
SVM implementations.

10 years agoAdd support for the SCTP_LOCAL_TRACE_BUF options.
Michael Tuexen [Tue, 10 Jun 2014 16:11:20 +0000 (16:11 +0000)]
Add support for the SCTP_LOCAL_TRACE_BUF options.
While there, fix some whitespaces.

MFC after: 1 week

10 years agochange the netmap mbuf destructor so the same code works also on FreeBSD 9.
Luigi Rizzo [Tue, 10 Jun 2014 16:06:59 +0000 (16:06 +0000)]
change the netmap mbuf destructor so the same code works also on FreeBSD 9.
For head and 10 this change has no effect, but on stable/9 it would cause
panics when using emulated netmap on top of a standard device driver.

10 years agoFix indentation level.
Glen Barber [Tue, 10 Jun 2014 15:20:41 +0000 (15:20 +0000)]
Fix indentation level.

MFC after: 3 days
X-MFC-With: r267326
Sponsored by: The FreeBSD Foundation

10 years agoAdd empty pkg-stage file to CLEANFILES if WITH_DVD=1.
Glen Barber [Tue, 10 Jun 2014 15:19:28 +0000 (15:19 +0000)]
Add empty pkg-stage file to CLEANFILES if WITH_DVD=1.

MFC after: 3 days
Sponsored by: The FreeBSD Foundation

10 years agovt fontcvt: handle failure writing output font
Ed Maste [Tue, 10 Jun 2014 12:59:56 +0000 (12:59 +0000)]
vt fontcvt: handle failure writing output font

10 years agoAvoid the USB device disconnected and controller shutdown clutter on system
Marius Strobl [Tue, 10 Jun 2014 08:20:00 +0000 (08:20 +0000)]
Avoid the USB device disconnected and controller shutdown clutter on system
shutdown by putting the former under !rebooting and turning the latter into
debug messages.

Reviewed by: hps
MFC after: 1 week
Sponsored by: Bally Wulff Games & Entertainment GmbH

10 years agoacpiconf(8): tell users not to use -k
Eitan Adler [Tue, 10 Jun 2014 08:15:41 +0000 (08:15 +0000)]
acpiconf(8): tell users not to use -k
Add language from jhb

Requested by: jhb, jkim

10 years agoCall cpp with -P to avoid printing line markings.
Rui Paulo [Tue, 10 Jun 2014 06:24:01 +0000 (06:24 +0000)]
Call cpp with -P to avoid printing line markings.

10 years agodtc: ignore lines starting with #.
Rui Paulo [Tue, 10 Jun 2014 06:16:34 +0000 (06:16 +0000)]
dtc: ignore lines starting with #.

This is necessary because we use the C pre-processor to parse #include lines
and cpp adds line markings that start with #.

10 years agodtc: don't crash if the argument is a directory.
Rui Paulo [Tue, 10 Jun 2014 06:04:25 +0000 (06:04 +0000)]
dtc: don't crash if the argument is a directory.

10 years agoThe, currently undocumented, -i option takes an argument.
Rui Paulo [Tue, 10 Jun 2014 05:58:46 +0000 (05:58 +0000)]
The, currently undocumented, -i option takes an argument.

10 years agoAlways append new bios to the tail of the queue, instead of sorting them
Bryan Venteicher [Tue, 10 Jun 2014 03:29:15 +0000 (03:29 +0000)]
Always append new bios to the tail of the queue, instead of sorting them

MFC after: 1 week

10 years agoTurn on interrupt window exiting unconditionally when an ExtINT is being
Neel Natu [Tue, 10 Jun 2014 01:38:02 +0000 (01:38 +0000)]
Turn on interrupt window exiting unconditionally when an ExtINT is being
injected into the guest. This allows the hypervisor to inject another
ExtINT or APIC vector as soon as the guest is able to process interrupts.

This change is not to address any correctness issue but to guarantee that
any pending APIC vector that was preempted by the ExtINT will be injected
as soon as possible. Prior to this change such pending interrupts could be
delayed until the next VM exit.

10 years agoCorrect unicode map for VGA ROM character 0x0e
Ed Maste [Tue, 10 Jun 2014 01:19:48 +0000 (01:19 +0000)]
Correct unicode map for VGA ROM character 0x0e

The character is a beamed pair of sixteenth notes, so should be U+266C
not U+266B (beamed eighth notes).

10 years agosiglongjmp(): Preserve floating point exception flags on i386 and amd64.
Jilles Tjoelker [Mon, 9 Jun 2014 21:35:36 +0000 (21:35 +0000)]
siglongjmp(): Preserve floating point exception flags on i386 and amd64.

Per POSIX, siglongjmp() shall be equivalent to longjmp() except that it must
match sigsetjmp() instead of setjmp() and except for the effect on the
signal mask. Therefore, it should preserve the floating point exception
flags.

This was fixed for longjmp() and _longjmp() in r180080 and r180081 for amd64
and i386 respectively.

10 years agoAdd vgarom font source
Ed Maste [Mon, 9 Jun 2014 21:06:46 +0000 (21:06 +0000)]
Add vgarom font source

These are in 'GNU Unifont' format, and are converted from syscons(4)
cp437 fonts.

10 years agoDocument r267256, OpenSSL 1.0.1h update.
Glen Barber [Mon, 9 Jun 2014 20:52:42 +0000 (20:52 +0000)]
Document r267256, OpenSSL 1.0.1h update.

Sponsored by: The FreeBSD Foundation

10 years agovt fontcvt: Speed up bold glyph map deduplication
Ed Maste [Mon, 9 Jun 2014 20:52:35 +0000 (20:52 +0000)]
vt fontcvt: Speed up bold glyph map deduplication

Perform an O(n) deduplication pass over the bold maps at the end, rather
than walking the normal map list to look for a duplicate glyph each time
a bold mapping entry is added.

Sponsored by: The FreeBSD Foundation

10 years agoAdd reserved bit checking when doing %CR8 emulation and inject #GP if required.
Neel Natu [Mon, 9 Jun 2014 20:51:08 +0000 (20:51 +0000)]
Add reserved bit checking when doing %CR8 emulation and inject #GP if required.

Pointed out by: grehan
Reviewed by: tychon

10 years agoMove the Sendmail update to keep revision numbers incremental.
Glen Barber [Mon, 9 Jun 2014 20:50:49 +0000 (20:50 +0000)]
Move the Sendmail update to keep revision numbers incremental.

Sponsored by: The FreeBSD Foundation

10 years agovt fontcvt: Hide stats by default and improve error handling
Ed Maste [Mon, 9 Jun 2014 20:49:13 +0000 (20:49 +0000)]
vt fontcvt: Hide stats by default and improve error handling

The font stats are interesting, but rather verbose.

10 years agoRe-enable -Werror for these modules. It is already enabled for the same
John Baldwin [Mon, 9 Jun 2014 20:48:38 +0000 (20:48 +0000)]
Re-enable -Werror for these modules.  It is already enabled for the same
files when built as part of a kernel.

10 years agor267169 should apply to 64-bit BARs as well.
Tycho Nightingale [Mon, 9 Jun 2014 19:55:50 +0000 (19:55 +0000)]
r267169 should apply to 64-bit BARs as well.

Reviewed by: neel

10 years agomdoc: fix AUTHORS section.
Joel Dahl [Mon, 9 Jun 2014 19:36:08 +0000 (19:36 +0000)]
mdoc: fix AUTHORS section.

10 years agoRemove blank lines.
Joel Dahl [Mon, 9 Jun 2014 19:29:10 +0000 (19:29 +0000)]
Remove blank lines.

10 years agoUse strcasecmp() instead of strcmp() when checking user-supplied encoding
John Baldwin [Mon, 9 Jun 2014 19:27:47 +0000 (19:27 +0000)]
Use strcasecmp() instead of strcmp() when checking user-supplied encoding
names so that encoding names are treated as case-insensitive.  This allows
the use of 'utf-8' instead of 'UTF-8' for example and matches the behavior
of iconv(1).

PR: 167977
Submitted by: buganini@gmail.com
MFC after: 1 week

10 years agoMake the hpt27xx(4) driver MPSAFE.
John Baldwin [Mon, 9 Jun 2014 18:15:05 +0000 (18:15 +0000)]
Make the hpt27xx(4) driver MPSAFE.
- Use the existing vbus locks instead of Giant for the CAM sim lock.
- Use callout(9) instead of timeout(9).
- Mark the interrupt handler as MPSAFE.
- Don't attempt to pass data in the softc from probe() to attach().

Reviewed by: Steve Chang <ychang@highpoint-tech.com>
Assisted by: delphij

10 years agoFixes from Fanco Ficthner on transparent mode
Luigi Rizzo [Mon, 9 Jun 2014 15:46:11 +0000 (15:46 +0000)]
Fixes from Fanco Ficthner on transparent mode

* The way rings are updated changed with the last API bump.
  Also sync ->head when moving slots in netmap_sw_to_nic().

* Remove a crashing selrecord() call.

* Unclog the logic surrounding netmap_rxsync_from_host().

* Add timestamping to RX host ring.

* Remove a couple of obsolete comments.

Submitted by: Franco Fichtner
MFC after: 3 days
Sponsored by: Packetwerk

10 years agosync the code with the one in stable/10
Luigi Rizzo [Mon, 9 Jun 2014 15:44:31 +0000 (15:44 +0000)]
sync the code with the one in stable/10
(wrap the if_t compatibilty function into a __FreeBSD_version
conditional block)

10 years agoFix race in r267221.
Alexander Motin [Mon, 9 Jun 2014 15:00:43 +0000 (15:00 +0000)]
Fix race in r267221.

MFC after: 2 weeks

10 years agoEnsure files are created during the build when using bsd.subdir.mk.
Julio Merino [Mon, 9 Jun 2014 14:36:49 +0000 (14:36 +0000)]
Ensure files are created during the build when using bsd.subdir.mk.

When FILES is defined in a Makefile that _also_ includes bsd.subdir.mk, the
build of the files (if any) was not properly triggered during the build
stage.  This was because bsd.files.mk did not define the buildfiles target
if it was already defined... and bsd.subdir.mk defined this target on its
own, thus causing a conflict.

Fix this by unconditionally defining buildfiles from bsd.files.mk; this is
safe because nothing else in the tree needs to redefine this and because the
target itself contains no commands: all it does is define dependencies.
Also ensure that bsd.files.mk is always pulled in by bsd.test.mk regardless
of what bsd.prog.mk does.

These fixes allow "make installworld" to run cleanly on a system with
read-only src and obj trees.

This is "make tinderbox" clean.

Reviewed by: imp
Obtained from: jilles

10 years agoRemoved stale comment about multi-vdev root pool config not working
Steven Hartland [Mon, 9 Jun 2014 13:04:58 +0000 (13:04 +0000)]
Removed stale comment about multi-vdev root pool config not working

MFC after: 1 week

10 years agotruncate: Detect integer overflow, fix relative sizes, add tests.
Jilles Tjoelker [Mon, 9 Jun 2014 10:39:55 +0000 (10:39 +0000)]
truncate: Detect integer overflow, fix relative sizes, add tests.

The change to expand_number (r204654) broke detection of too large sizes and
relative sizes ('+'/'-').

Also add some tests.

PR: 190735
Submitted by: Kirk Russell
MFC after: 1 week