]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
10 years agoAdd support for some block functions that come from OS X. These are
David Chisnall [Wed, 2 Apr 2014 16:07:48 +0000 (16:07 +0000)]
Add support for some block functions that come from OS X.  These are
intended to build with any C compiler.

Reviewed by: pfg
MFC after: 3 weeks

10 years agoAdd support for event timers whose clock frequency can change while running.
Ian Lepore [Wed, 2 Apr 2014 15:56:11 +0000 (15:56 +0000)]
Add support for event timers whose clock frequency can change while running.

10 years agoMFV r258379;
Pedro F. Giffuni [Wed, 2 Apr 2014 15:32:44 +0000 (15:32 +0000)]
MFV r258379;

4248 dtrace(1M) should never create DOF with empty probes section
4249 Only probes from the first DTrace object file will be included

Illumos Revision: 4a20ab41aadcb81c53e72fc65886e964e9add59

Reference:
https://www.illumos.org/issues/4248
https://www.illumos.org/issues/4249

Obtained from: Illumos
MFC after: 1 month

10 years agoDocument more parition types.
Andrey V. Elsukov [Wed, 2 Apr 2014 11:12:48 +0000 (11:12 +0000)]
Document more parition types.

Requested by: glebius
MFC after: 1 week

10 years agoFix an issue where the locale and rune locale could become out of sync,
David Chisnall [Wed, 2 Apr 2014 11:10:46 +0000 (11:10 +0000)]
Fix an issue where the locale and rune locale could become out of sync,
causing mb* functions (and similar) to be called with the wrong data
(possibly a null pointer, causing a crash).

PR: standards/188036
MFC after: 1 week

10 years agoUse 2K buffers for IO to help achieve full device speed, rather than the
Ian Lepore [Wed, 2 Apr 2014 01:58:54 +0000 (01:58 +0000)]
Use 2K buffers for IO to help achieve full device speed, rather than the
default wMaxPacketSize (64 or 512 bytes).  This actually helps older FTDI
devices (which were USB 1/full speed) more than the new H-series high
speed, but even for the new chips it helps cut the number of interrupts
when doing very high speed (3-12mbaud).

10 years agoIf WITH_COMPRESSED_IMAGES is set, add the compressed images to
Glen Barber [Tue, 1 Apr 2014 22:54:54 +0000 (22:54 +0000)]
If WITH_COMPRESSED_IMAGES is set, add the compressed images to
the CLEANFILES list.

MFC after: 1 week
X-MFC-With: r264027,r264028,r264029
Sponsored by: The FreeBSD Foundation

10 years agoFix logic error: s/||/&&/
Glen Barber [Tue, 1 Apr 2014 22:52:27 +0000 (22:52 +0000)]
Fix logic error: s/||/&&/

Pointyhat: gjb
MFC after: 1 week
X-MFC-With: r264027,r264028
Sponsored by: The FreeBSD Foundation

10 years agoClean up trailing whitespace in release/Makefile.
Glen Barber [Tue, 1 Apr 2014 22:46:46 +0000 (22:46 +0000)]
Clean up trailing whitespace in release/Makefile.

MFC after: 1 week
X-MFC-With: r264027
Sponsored by: The FreeBSD Foundation

10 years agoAdd a new release build variable, WITH_COMPRESSED_IMAGES.
Glen Barber [Tue, 1 Apr 2014 22:41:26 +0000 (22:41 +0000)]
Add a new release build variable, WITH_COMPRESSED_IMAGES.

When set to a non-empty value, the installation medium is
compressed with gzip(1) as part of the 'install' target in
the release/ directory.

With gzip(1) compression, downloadable image are reduced in
size quite significantly.  Build test against head@263927
shows the following:

 bootonly.iso: 64% smaller
 disc1.iso: 44% smaller
 memstick.img: 47% smaller
 mini-memstick.img: 65% smaller
 dvd1.iso: untested

This option is off by default, I would eventually like to
turn it on by default, and remove the '-k' flag to gzip(1)
so only compressed images are published on FTP.

Requested by: wkoszek
MFC After: 1 week
Sponsored by: The FreeBSD Foundation

10 years agoEnable a KASSERT.
Edward Tomasz Napierala [Tue, 1 Apr 2014 22:21:56 +0000 (22:21 +0000)]
Enable a KASSERT.

Sponsored by: The FreeBSD Foundation

10 years agoGet rid of the "autoscaling", instead just set socket buffer sizes
Edward Tomasz Napierala [Tue, 1 Apr 2014 22:03:03 +0000 (22:03 +0000)]
Get rid of the "autoscaling", instead just set socket buffer sizes
in the usual way.  The only thing the old code did was making things
less predictable.

Sponsored by: The FreeBSD Foundation

10 years agoUse atomic ops instead of mutexes where appropriate.
Edward Tomasz Napierala [Tue, 1 Apr 2014 21:54:20 +0000 (21:54 +0000)]
Use atomic ops instead of mutexes where appropriate.

Submitted by: mav@
Sponsored by: The FreeBSD Foundation

10 years agoInstead of "icltx" and "iclrx", use thread names with prefix from upper
Edward Tomasz Napierala [Tue, 1 Apr 2014 21:47:22 +0000 (21:47 +0000)]
Instead of "icltx" and "iclrx", use thread names with prefix from upper
layer, so that one can see which side of the stack the threads are for.

Sponsored by: The FreeBSD Foundation

10 years agoGet rid of ICL lock; use upper-layer (initiator or target) lock instead.
Edward Tomasz Napierala [Tue, 1 Apr 2014 21:40:46 +0000 (21:40 +0000)]
Get rid of ICL lock; use upper-layer (initiator or target) lock instead.
This avoids extra locking in icl_pdu_queue(); the upper layer needs to call
it while holding its own lock anyway, to avoid sending PDUs out of order.

Sponsored by: The FreeBSD Foundation

10 years agolibnv: Don't lose big-endian flag when receiving a message.
Jilles Tjoelker [Tue, 1 Apr 2014 21:30:54 +0000 (21:30 +0000)]
libnv: Don't lose big-endian flag when receiving a message.

A bug caused the "big endian" flag to be lost when receiving a message. As a
result, the bits are interpreted as little endian and an extremely large
allocation is attempted.

This change fixes ping(8)'s communication to casperd(8) on big-endian
architectures.

Reported by: Anton Shterenlikht
Tested by: danfe

10 years agoRemove the homegrown ctl_be_block_io allocator, replacing it with UMA.
Edward Tomasz Napierala [Tue, 1 Apr 2014 21:13:05 +0000 (21:13 +0000)]
Remove the homegrown ctl_be_block_io allocator, replacing it with UMA.
There is no performance difference.

Reviewed by: mav@
Sponsored by: The FreeBSD Foundation

10 years agoFix some of the style(9) problems on ti_gpio. Note that this commit does
Luiz Otavio O Souza [Tue, 1 Apr 2014 18:46:56 +0000 (18:46 +0000)]
Fix some of the style(9) problems on ti_gpio.  Note that this commit does
not fix all the style(9) problems, but will open room for the next commits.

Remove redundant code and declarations.

No functional changes.

10 years agoUpdate the list of FTDI chips supported by uftdi(4).
Ian Lepore [Tue, 1 Apr 2014 18:41:54 +0000 (18:41 +0000)]
Update the list of FTDI chips supported by uftdi(4).

10 years agoIncrement the SSN only after processing the last fragment of an
Michael Tuexen [Tue, 1 Apr 2014 18:38:04 +0000 (18:38 +0000)]
Increment the SSN only after processing the last fragment of an
ordered user message.

MFC after: 3 days

10 years agoSilence an unnecessary warning.
George V. Neville-Neil [Tue, 1 Apr 2014 18:37:00 +0000 (18:37 +0000)]
Silence an unnecessary warning.

MFC after: 2 weeks

10 years agoAdd missing copyright date.
Ryan Stone [Tue, 1 Apr 2014 17:35:31 +0000 (17:35 +0000)]
Add missing copyright date.

MFC after: 2 months

10 years agoPrint status of ARI capability in pciconf -c
Ryan Stone [Tue, 1 Apr 2014 16:03:52 +0000 (16:03 +0000)]
Print status of ARI capability in pciconf -c

Teach pciconf how to print out the status (enabled/disabled) of the ARI
capability on PCI Root Complexes and Downstream Ports.

MFC after: 2 months
Sponsored by: Sandvine Inc.

10 years agoAdd support for PCIe ARI
Ryan Stone [Tue, 1 Apr 2014 16:02:02 +0000 (16:02 +0000)]
Add support for PCIe ARI

PCIe Alternate RID Interpretation (ARI) is an optional feature that
allows devices to have up to 256 different functions.  It is
implemented by always setting the PCI slot number to 0 and
re-purposing the 5 bits used to encode the slot number to instead
contain the function number.  Combined with the original 3 bits
allocated for the function number, this allows for 256 functions.

This is enabled by default, but it's expected to be a no-op on currently
supported hardware.  It's a prerequisite for supporting PCI SR-IOV, and
I want the ARI support to go in early to help shake out any bugs in it.
ARI can be disabled by setting the tunable hw.pci.enable_ari=0.

Reviewed by: kib
MFC after: 2 months
Sponsored by: Sandvine Inc.

10 years agoSupport serial speeds up to 12mbaud with newer FTDI chips.
Ian Lepore [Tue, 1 Apr 2014 15:56:31 +0000 (15:56 +0000)]
Support serial speeds up to 12mbaud with newer FTDI chips.

Recent FDTI chips have the ability to operate at up to 12mbps.  The newer
chips with faster clocks have the same usb vendor/product IDs as the older
chips; the bcdDevice field must be used to detect the newer versions.  This
change includes a new function to do that instead of using just the IDs from
the vendor/product table.

The code to choose the baud clock divisor is completely rewritten.  In
addition to supporting the new higher clock rates, the rewrite fixes a
longstanding bug in the old code which put the high bits of the fractional
part of the divisor into the wrong place in the wIndex field.  That bug
was mostly harmless -- it accidentally didn't affect standard baud rates
and would only show up when using relatively fast non-standard rates.

10 years agoRe-write bhyve's I/O MMU handling in terms of PCI RID.
Ryan Stone [Tue, 1 Apr 2014 15:54:03 +0000 (15:54 +0000)]
Re-write bhyve's I/O MMU handling in terms of PCI RID.

Reviewed by: neel
MFC after: 2 months
Sponsored by: Sandvine Inc.

10 years agoRe-implement the DMAR I/O MMU code in terms of PCI RIDs
Ryan Stone [Tue, 1 Apr 2014 15:48:46 +0000 (15:48 +0000)]
Re-implement the DMAR I/O MMU code in terms of PCI RIDs

Under the hood the VT-d spec is really implemented in terms of
PCI RIDs instead of bus/slot/function, even though the spec makes
pains to convert back to bus/slot/function in examples.  However
working with bus/slot/function is not correct when PCI ARI is
in use, so convert to using RIDs in most cases.  bus/slot/function
will only be used when reporting errors to a user.

Reviewed by: kib
MFC after: 2 months
Sponsored by: Sandvine Inc.

10 years agoAdd a method to get the PCI RID for a device.
Ryan Stone [Tue, 1 Apr 2014 15:47:24 +0000 (15:47 +0000)]
Add a method to get the PCI RID for a device.

Reviewed by: kib
MFC after: 2 months
Sponsored by: Sandvine Inc.

10 years agoRevert PCI RID changes.
Ryan Stone [Tue, 1 Apr 2014 15:06:03 +0000 (15:06 +0000)]
Revert PCI RID changes.

My PCI RID changes somehow got intermixed with my PCI ARI patch when I
committed it.  I may have accidentally applied a patch to a non-clean
working tree.  Revert everything while I figure out what went wrong.

Pointy hat to: rstone

10 years agoRe-write bhyve's I/O MMU handling in terms of PCI RIDs
Ryan Stone [Tue, 1 Apr 2014 14:54:43 +0000 (14:54 +0000)]
Re-write bhyve's I/O MMU handling in terms of PCI RIDs

Reviewed by: neel
Sponsored by: Sandvine Inc

10 years agoRe-implement the DMAR I/O MMU code in terms of PCI RIDs
Ryan Stone [Tue, 1 Apr 2014 14:51:45 +0000 (14:51 +0000)]
Re-implement the DMAR I/O MMU code in terms of PCI RIDs

Under the hood the VT-d spec is really implemented in terms of
PCI RIDs instead of bus/slot/function, even though the spec makes
pains to convert back to bus/slot/function in examples.  However
working with bus/slot/function is not correct when PCI ARI is
in use, so convert to using RIDs in most cases.  bus/slot/function
will only be used when reporting errors to a user.

Reviewed by: kib
Sponsored by: Sandvine Inc.

10 years agoAdd a method to get the PCI Routing ID for a device
Ryan Stone [Tue, 1 Apr 2014 14:49:25 +0000 (14:49 +0000)]
Add a method to get the PCI Routing ID for a device

Reviewed by: kib
Sponsored by: Sandvine, Inc

10 years agoAdd the svn revision number to the IPX note.
Glen Barber [Tue, 1 Apr 2014 14:48:16 +0000 (14:48 +0000)]
Add the svn revision number to the IPX note.

Sponsored by: The FreeBSD Foundation

10 years agoRename __wchar_t so it no longer conflicts with __wchar_t from clang 3.4
Tijl Coosemans [Tue, 1 Apr 2014 14:46:11 +0000 (14:46 +0000)]
Rename __wchar_t so it no longer conflicts with __wchar_t from clang 3.4
-fms-extensions.

MFC after: 2 weeks

10 years agoDon't test WITHOUT_FOO in program makefiles, test MK_FOO instead.
Warner Losh [Tue, 1 Apr 2014 14:24:32 +0000 (14:24 +0000)]
Don't test WITHOUT_FOO in program makefiles, test MK_FOO instead.

10 years agoRemove support for legacy mips*eb names. Remove tests for
Warner Losh [Tue, 1 Apr 2014 14:24:25 +0000 (14:24 +0000)]
Remove support for legacy mips*eb names. Remove tests for
TARGET_BIG_ENDIAN.

10 years agoRemove check for clang and expand the comment. Newer versions of gcc
Warner Losh [Tue, 1 Apr 2014 14:24:20 +0000 (14:24 +0000)]
Remove check for clang and expand the comment. Newer versions of gcc
generate dwarf4 by default as well, so always force dwarf2 when
generating debugging data. It is harmless on older versions of both
clang and gcc, but required on newer ones.

10 years agoThere's no need to set the default for GNUCXX based on WITHOUT_CXX
Warner Losh [Tue, 1 Apr 2014 14:24:15 +0000 (14:24 +0000)]
There's no need to set the default for GNUCXX based on WITHOUT_CXX
being defined. The system works fine without it (because GNUCXX isn't
built when WITHOUT_CXX is defined), and it is one of the few places we
test WITHOUT_FOO instead of MK_FOO in the base system. Simply
eliminate it to solve both problems. Also, minor tweak to make it
clearer that the default is always NO for GNUGCC on i386.

10 years agoEliminate FDT being treated specially. When the BSDL DTC was
Warner Losh [Tue, 1 Apr 2014 14:24:08 +0000 (14:24 +0000)]
Eliminate FDT being treated specially. When the BSDL DTC was
introduced, the meaning of WITH_FDT changed from 'do FDT in boot
loader and build dtc compiler' to be simply 'do FDT things in the boot
loader'. Now that we test for FDT in the architecture specific files,
this test here is now redunant and encourages bad behavior, so just
eliminate it and make it always yes. Those architectures whose boot
loaders don't support FDT already omit it when it is set to yes
anyway.

10 years agoTest MK_ARM_ABI rather than if WITHOUT_ARM_ABI is defined.
Warner Losh [Tue, 1 Apr 2014 14:24:03 +0000 (14:24 +0000)]
Test MK_ARM_ABI rather than if WITHOUT_ARM_ABI is defined.

10 years agoIt is possible that MK_GNUCXX is "yes" but MK_CXX is "no" so make sure
Warner Losh [Tue, 1 Apr 2014 14:23:58 +0000 (14:23 +0000)]
It is possible that MK_GNUCXX is "yes" but MK_CXX is "no" so make sure
MK_CXX isn't "no" before building these libraries.

10 years agoAdd gpiobus(4) as a link to gpio(4).
Luiz Otavio O Souza [Tue, 1 Apr 2014 14:17:38 +0000 (14:17 +0000)]
Add gpiobus(4) as a link to gpio(4).

10 years agoAdd atf m4 files from the vendor branch.
Julio Merino [Tue, 1 Apr 2014 13:47:08 +0000 (13:47 +0000)]
Add atf m4 files from the vendor branch.

These were originally deleted as "not important" but, actually we need them
in place if we want to be able to use autoconf on software that provides
atf-based tests.  (That includes being able to rebuild autotest from scratch
on the Kyua cluster machines, as the automated setup does.)

10 years ago- In the libiconv module for ISO 2022 restore the original order of the
Tijl Coosemans [Tue, 1 Apr 2014 10:36:11 +0000 (10:36 +0000)]
- In the libiconv module for ISO 2022 restore the original order of the
  fields of a private struct such that variables of this type are
  initialised correctly.  Fixes conversion from ISO 2022.
  Also do this in the BIG5 module to prevent similar errors in the future.
- In the libiconv module for EUC-TW replace 2^cs with 1<<cs.  Fixes
  conversion from EUC-TW.
- Synchronise iconv code with NetBSD.  In most cases this only updates
  the RCS id because the changes are already there or are NetBSD specific.
  + libc/iconv/citrus_csmapper.c: Add a comment.
  + libc/iconv/citrus_db_factory.c: Remove put16().
  + libc/iconv/citrus_iconv.c: Return EINVAL on error.
  + libc/iconv/citrus_mapper.c: Return EINVAL on error.
  + libc/iconv/citrus_memstream.c: Fix type of a variable.
  + libc/iconv/citrus_prop.h: Sync definition of _CITRUS_PROP_HINT_END.
  + libc/iconv/citrus_stdenc.c: Return EINVAL on error.
  + libiconv_modules/mapper_std/citrus_mapper_std.c: Plug memory leak.

Obtained from: NetBSD
MFC after: 2 weeks

10 years agoAdd new USB ID.
Hans Petter Selasky [Tue, 1 Apr 2014 10:34:39 +0000 (10:34 +0000)]
Add new USB ID.

Submitted by: Tuomo Latto <djv+fbpr@iki.fi>
PR: usb/188046
MFC after: 1 week

10 years agoEnsure make xdev respect which compiler is CC
Baptiste Daroussin [Tue, 1 Apr 2014 06:46:59 +0000 (06:46 +0000)]
Ensure make xdev respect which compiler is CC
and which compilers should be built according to standard build options

10 years agoAdd Cortex-A15 cpu id revisions.
Ruslan Bukin [Tue, 1 Apr 2014 04:56:40 +0000 (04:56 +0000)]
Add Cortex-A15 cpu id revisions.

10 years agoRelease i2c bus on detach.
Ruslan Bukin [Tue, 1 Apr 2014 04:40:08 +0000 (04:40 +0000)]
Release i2c bus on detach.

10 years agoHide CTL messages about SCSI error responses. Too many users take
Edward Tomasz Napierala [Mon, 31 Mar 2014 21:04:15 +0000 (21:04 +0000)]
Hide CTL messages about SCSI error responses.  Too many users take
them for actual target errors.  They can be enabled back by setting
kern.cam.ctl.verbose=1, or booting with bootverbose.

Sponsored by: The FreeBSD Foundation

10 years agoMake it possible to have multiple CTL worker threads. Leave the default
Edward Tomasz Napierala [Mon, 31 Mar 2014 20:49:33 +0000 (20:49 +0000)]
Make it possible to have multiple CTL worker threads.  Leave the default
of 1 for now.

Sponsored by: The FreeBSD Foundation

10 years agoTurns out name was used here when ACPI_DEBUG was defined, so refine my
Warner Losh [Mon, 31 Mar 2014 19:37:39 +0000 (19:37 +0000)]
Turns out name was used here when ACPI_DEBUG was defined, so refine my
previous patch.

10 years agoUpdate jemalloc to version 3.6.0.
Jason Evans [Mon, 31 Mar 2014 17:04:04 +0000 (17:04 +0000)]
Update jemalloc to version 3.6.0.

10 years agoddb: Minor style cleanups.
Pedro F. Giffuni [Mon, 31 Mar 2014 16:37:41 +0000 (16:37 +0000)]
ddb: Minor style cleanups.

#define should be followed by tab.

MFC after: 1 week

10 years agoRemove unused label.
Andrey V. Elsukov [Mon, 31 Mar 2014 14:40:35 +0000 (14:40 +0000)]
Remove unused label.

MFC after: 1 week

10 years agoDon't generate an ICMPv6 error message if packet was consumed by filter.
Andrey V. Elsukov [Mon, 31 Mar 2014 14:27:22 +0000 (14:27 +0000)]
Don't generate an ICMPv6 error message if packet was consumed by filter.

MFC after: 1 week
Sponsored by: Yandex LLC

10 years agoMerge r246607 by benno from projects/uefi
Ed Maste [Mon, 31 Mar 2014 14:12:27 +0000 (14:12 +0000)]
Merge r246607 by benno from projects/uefi

  Use EFI types for EFI values (silences warnings).

EFI UINTN is actually a 64-bit type on 64-bit processors.

Sponsored by: The FreeBSD Foundation

10 years agoDocument r263140, removal of IPX.
Glen Barber [Mon, 31 Mar 2014 14:04:57 +0000 (14:04 +0000)]
Document r263140, removal of IPX.

Sponsored by: The FreeBSD Foundation

10 years agoDon't copy the MF flag from original IP header to ICMP error message.
Andrey V. Elsukov [Mon, 31 Mar 2014 13:00:49 +0000 (13:00 +0000)]
Don't copy the MF flag from original IP header to ICMP error message.

PR: 188092
MFC after: 1 week
Sponsored by: Yandex LLC

10 years agoIncrease the number of TX DMA segments from 32 to 35. It turned
Pyun YongHyeon [Mon, 31 Mar 2014 01:54:59 +0000 (01:54 +0000)]
Increase the number of TX DMA segments from 32 to 35.  It turned
out 32 is not enough to support a full sized TSO packet.
While I'm here fix a long standing bug introduced in r169632 in
bce(4) where it didn't include L2 header length of TSO packet in
the maximum DMA segment size calculation.

In collaboration with: rmacklem
MFC after: 2 weeks

10 years agoRemove instances of variables that were set, but never used. gcc 4.9
Warner Losh [Sun, 30 Mar 2014 23:43:36 +0000 (23:43 +0000)]
Remove instances of variables that were set, but never used. gcc 4.9
warns about these by default.

10 years agoThere's no need to guess at the COMPILER_TYPE to pass it down. We
Warner Losh [Sun, 30 Mar 2014 23:43:30 +0000 (23:43 +0000)]
There's no need to guess at the COMPILER_TYPE to pass it down. We
guess wrong for buildkernel when CC=gcc49, say. Eliminate all the
guessing. COMPILER_TYPE propigates properly on its own, if specified,
and we guess it correctly otherwise lower in the build. Also, fix
conditionals for armv6hf when using an external compiler chain. They
were broken before, but unused. Also, prefer checking the compiler
type over CLANG_IS_CC since the latter is only supposed to be used to
determine what symlinks to install (more fixes to follow).

10 years agoMove stray targets out of the xdev section.
Warner Losh [Sun, 30 Mar 2014 22:25:01 +0000 (22:25 +0000)]
Move stray targets out of the xdev section.

10 years agoKill NO_TESTS and use MK_TESTS=no when we want to skip them.
Warner Losh [Sun, 30 Mar 2014 22:24:53 +0000 (22:24 +0000)]
Kill NO_TESTS and use MK_TESTS=no when we want to skip them.

10 years agoAllow the build system to safely set MK_FOO to avoid the ambiguity
Warner Losh [Sun, 30 Mar 2014 22:24:45 +0000 (22:24 +0000)]
Allow the build system to safely set MK_FOO to avoid the ambiguity
when both WITH_FOO and WITHOUT_FOO are set. Use this where
possible. Only disallow setting of MK_FOO on the command line. This
was preferable to inventing a new mechanism or fixing the undef bug
(bin/183762) which precludes users from turning off anything we turn
off for parts of the build with WITHOUT_FOO prior to this.

10 years agoWhen building g++, we need to build libsupc earlier to avoid a race
Warner Losh [Sun, 30 Mar 2014 22:24:37 +0000 (22:24 +0000)]
When building g++, we need to build libsupc earlier to avoid a race
with libproc. Not sure why people didn't see this before now, but I
get it often for higher (20-30) -j builds, but never for -j1 builds.

10 years agoConvert an unused banner from skeleton.c into a comment.
Baptiste Daroussin [Sun, 30 Mar 2014 21:56:13 +0000 (21:56 +0000)]
Convert an unused banner from skeleton.c into a comment.
It is added to every generated files and fix build at certain warning level with clang 3.4

Submitted by: Thomas Dickey <dickey@his.com> (byacc upstream)
Spotted by: glebius

10 years agoConvert an unused banner from skeleton.c into a comment.
Baptiste Daroussin [Sun, 30 Mar 2014 21:48:49 +0000 (21:48 +0000)]
Convert an unused banner from skeleton.c into a comment.
It is added to every generated files and fix build at certain warning level with clang 3.4

Submitted by: Thomas Dickey <dickey@his.com> (byacc upstream)
Spotted by: glebius

10 years agoadd capsicum to units(1)
Eitan Adler [Sun, 30 Mar 2014 16:04:47 +0000 (16:04 +0000)]
add capsicum to units(1)

10 years agoGive hint on how to disable the default repository.
Bryan Drewery [Sun, 30 Mar 2014 15:24:17 +0000 (15:24 +0000)]
Give hint on how to disable the default repository.

Discussed with: bapt
MFC after: instantly (preparing EN)

10 years agoAdd support for keyboard used in Samsung Chromebook (ARM machine)
Ruslan Bukin [Sun, 30 Mar 2014 15:22:36 +0000 (15:22 +0000)]
Add support for keyboard used in Samsung Chromebook (ARM machine)

Support covers device drivers for:
- Interrupt Combiner
- gpio/pad, External Interrupts Controller (pad)
- I2C Interface
- Chrome Embedded Controller
- Chrome Keyboard

Also:
- Use new gpio dev class in EHCI driver
- Expand device tree information

10 years agoAdd lowercase postfix, so SoC-common file will be ignored by tinderbox.
Ruslan Bukin [Sun, 30 Mar 2014 14:57:00 +0000 (14:57 +0000)]
Add lowercase postfix, so SoC-common file will be ignored by tinderbox.

10 years agoDirectly call kmem_alloc_contig to allocate framebuffer memory
Ruslan Bukin [Sun, 30 Mar 2014 14:42:53 +0000 (14:42 +0000)]
Directly call kmem_alloc_contig to allocate framebuffer memory
and pass VM_MEMATTR_UNCACHEABLE (no-cache, no-buffer).
This fixes screen refreshing problem when data is updated too slowly.

Discussed with: ian

10 years ago- Merge SoC-common parts
Ruslan Bukin [Sun, 30 Mar 2014 14:25:14 +0000 (14:25 +0000)]
- Merge SoC-common parts
- Enable iicbus device

10 years agounits(1): Add some missing units
Eitan Adler [Sun, 30 Mar 2014 05:06:06 +0000 (05:06 +0000)]
units(1): Add some missing units

10 years agoTeach units(1) that an erg and ergon are the same thing.
Eitan Adler [Sun, 30 Mar 2014 04:55:33 +0000 (04:55 +0000)]
Teach units(1) that an erg and ergon are the same thing.

10 years agoUse the correct variable name in the example code.
Eitan Adler [Sun, 30 Mar 2014 04:40:41 +0000 (04:40 +0000)]
Use the correct variable name in the example code.

10 years agoFix sizeof modifier for sizeof() in printf
Eitan Adler [Sun, 30 Mar 2014 00:25:31 +0000 (00:25 +0000)]
Fix sizeof modifier for sizeof() in printf

10 years agoFix build on FreeBSD 7 where:
Marcel Moolenaar [Sat, 29 Mar 2014 23:46:01 +0000 (23:46 +0000)]
Fix build on FreeBSD 7 where:
1.  DOSPTYP_FAT32 is not defined in <sys/diskmbr.h>
2.  uuid_enc_le() does not exist in libc.

10 years agoEnable all cryptocaps because net80211 can do software encryption.
Rui Paulo [Sat, 29 Mar 2014 23:25:09 +0000 (23:25 +0000)]
Enable all cryptocaps because net80211 can do software encryption.

MFC after: 1 week

10 years agoFix build on FreeBSD 8 where partition types for nandfs do not exist.
Marcel Moolenaar [Sat, 29 Mar 2014 22:10:54 +0000 (22:10 +0000)]
Fix build on FreeBSD 8 where partition types for nandfs do not exist.

10 years agoFix build on FreeBSD 9 where <sys/pc98.h> has the same defines as
Marcel Moolenaar [Sat, 29 Mar 2014 22:02:25 +0000 (22:02 +0000)]
Fix build on FreeBSD 9 where <sys/pc98.h> has the same defines as
<sys/diskmbr.h> and not the unique defines introduced later.

10 years agoHandle an edge case of address management similar to TCP.
Michael Tuexen [Sat, 29 Mar 2014 21:26:45 +0000 (21:26 +0000)]
Handle an edge case of address management similar to TCP.
This needs to be reconsidered when the address handling
will be reimplemented.
The patch is from rrs@.

MFC after: 3 days

10 years agoUse SCTP_OVER_UDP_TUNNELING_PORT more consistently.
Michael Tuexen [Sat, 29 Mar 2014 20:21:36 +0000 (20:21 +0000)]
Use SCTP_OVER_UDP_TUNNELING_PORT more consistently.

MFC after: 3 days

10 years agoAdd support to export the contents of the notification updates from the firmware.
Adrian Chadd [Sat, 29 Mar 2014 19:53:04 +0000 (19:53 +0000)]
Add support to export the contents of the notification updates from the firmware.

Right now the NIC isn't actually exporting useful data. I'm not quite sure
why this is. :(

10 years agoHook mkimg(1) to the build.
Marcel Moolenaar [Sat, 29 Mar 2014 19:04:38 +0000 (19:04 +0000)]
Hook mkimg(1) to the build.

10 years agoAdd mkimg, a utility for making disk images from raw partition contents.
Marcel Moolenaar [Sat, 29 Mar 2014 19:03:10 +0000 (19:03 +0000)]
Add mkimg, a utility for making disk images from raw partition contents.
The partitioning scheme can be one of the schemes supported by gpart.

Reviewed by: sjg
Obtained from: Juniper Networks, Inc.

10 years agoFix ah_powerMode setting.
Adrian Chadd [Sat, 29 Mar 2014 18:56:11 +0000 (18:56 +0000)]
Fix ah_powerMode setting.

Reported by: sbruno

10 years agoVFP fixes/cleanups for ARM11:
Andrew Turner [Sat, 29 Mar 2014 14:35:36 +0000 (14:35 +0000)]
VFP fixes/cleanups for ARM11:
 * Save the required VFP registers on context switch. If the exception bit
   is set we need to save and restore the FPINST register, and if the fp2v
   bit is also set we need to save and restore FPINST2.
 * Move saving and restoring the floating point control registers to C.
 * Clear the fpexc exception and fp2v flags on a floating-point exception.
 * Signal a SIGFPE if the fpexc exception flag is set on an undefined
   instruction. This is how the ARM core signals to software there is a
   floating-point exception.

10 years agoInitialise fpscr to a sane value when we create the pcb. This sets NaNs to
Andrew Turner [Sat, 29 Mar 2014 14:17:04 +0000 (14:17 +0000)]
Initialise fpscr to a sane value when we create the pcb. This sets NaNs to
be the default NaN and for denormalised numbers to be flushed to zero.

10 years agoClear the kernel grab of the FPU state on fork. The pcb_save pointer
Konstantin Belousov [Sat, 29 Mar 2014 11:56:33 +0000 (11:56 +0000)]
Clear the kernel grab of the FPU state on fork.  The pcb_save pointer
is already correctly reset to the FPU user save area, only PCB_KERNFPU
flag might leak from old thread state into the new state.

For creation of the user-mode thread, the change is nop since
corresponding syscall code does not use FPU.  On the other hand,
creation of a kernel thread forks from a thread selected arbitrary
from proc0, which might use FPU.

Reported and tested by: Chris Torek <torek@torek.net>
Sponsored by: The FreeBSD Foundation
MFC after: 1 week

10 years agoAdd more flags for the fpexc register from the ARM1176JZF-S Manual
Andrew Turner [Sat, 29 Mar 2014 10:11:19 +0000 (10:11 +0000)]
Add more flags for the fpexc register from the ARM1176JZF-S Manual

10 years agoMFV for 263899, tzdata2014b
Edwin Groothuis [Sat, 29 Mar 2014 04:15:11 +0000 (04:15 +0000)]
MFV for 263899, tzdata2014b

- Update antartica.
- Comments about historical data for Hawaii.
- Update details for Crimea on 30 March
- Move location data Simferopol.

10 years agoVendor import of tzdata 2014b
Edwin Groothuis [Sat, 29 Mar 2014 04:10:13 +0000 (04:10 +0000)]
Vendor import of tzdata 2014b

- Update antartica.
- Comments about historical data for Hawaii.
- Update details for Crimea on 30 March
- Move location data Simferopol.

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

10 years agoAdd the llvm/clang patch for r263891.
Dimitry Andric [Sat, 29 Mar 2014 00:48:50 +0000 (00:48 +0000)]
Add the llvm/clang patch for r263891.

10 years agoMake clang default to DWARF2 debug info format for FreeBSD 10.x and
Dimitry Andric [Sat, 29 Mar 2014 00:25:27 +0000 (00:25 +0000)]
Make clang default to DWARF2 debug info format for FreeBSD 10.x and
earlier.  For head, this commit does not change anything, but it is
purely meant to be MFC'd.

MFC after: 3 days

10 years agoDump default media size to 1GB to reflect a fatter environment and
Warner Losh [Fri, 28 Mar 2014 23:20:56 +0000 (23:20 +0000)]
Dump default media size to 1GB to reflect a fatter environment and
larger base media.

Submitted by: Michael Dexter

10 years agoMFV r263887:
Xin LI [Fri, 28 Mar 2014 23:12:00 +0000 (23:12 +0000)]
MFV r263887:

3993 zpool(1M) and zfs(1M) should support -p for "list" and "get"
4700 "zpool get" doesn't support -H or -o options

MFC after: 2 weeks

10 years ago3993 zpool(1M) and zfs(1M) should support -p for "list" and "get"
Xin LI [Fri, 28 Mar 2014 22:46:55 +0000 (22:46 +0000)]
3993 zpool(1M) and zfs(1M) should support -p for "list" and "get"
4700 "zpool get" doesn't support -H or -o options

illumos/illumos-gate@c58b352673e88983cd2b8a388a8c7625f35e2f18

10 years ago4573 ZFS snapshot alias
Xin LI [Fri, 28 Mar 2014 22:44:01 +0000 (22:44 +0000)]
4573 ZFS snapshot alias

illumos/illumos-gate@7dbbcd8300f0241d3ec9686f2a0a06107c1b90cb

10 years agoo Add new vd_driver method to do bitblt with mask, named vd_maskbitbltchr.
Aleksandr Rybalko [Fri, 28 Mar 2014 22:04:27 +0000 (22:04 +0000)]
o Add new vd_driver method to do bitblt with mask, named vd_maskbitbltchr.
o Move vd_bitbltchr vga's driver method to vd_maskbitbltchr.
o Implement new vd_bitbltchr method for vga driver. (It do single write for 8
pixels, have to be a bit faster).

MFC after: 7 days
Sponsored by: The FreeBSD Foundation