]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
10 years agoChanged name of test case to a more descriptive one and moved
pho [Tue, 4 Mar 2014 10:47:35 +0000 (10:47 +0000)]
Changed name of test case to a more descriptive one and moved
comment to the "descr" property.

Suggested by:  jmmv
Sponsored by: EMC / Isilon storage division

10 years agoHandle the newer-style bluetooth message format from the (at least)
adrian [Tue, 4 Mar 2014 08:01:56 +0000 (08:01 +0000)]
Handle the newer-style bluetooth message format from the (at least)
Centrino 2230 firmware.

This fixes the general statistics block to be actually valid.

I've verified this by contrasting the output of iwnstats before and
after the change.  The general block is now correct.

Tested:

* Intel 5100 (old format stats message)
* Intel 2230 (new format stats message)

10 years agoFix multicast filtering.
hrs [Tue, 4 Mar 2014 06:55:54 +0000 (06:55 +0000)]
Fix multicast filtering.

Submitted by: Takanori Sawada
PR: arm/187179

10 years agoNaming consistency fix. The routing code defines
gnn [Tue, 4 Mar 2014 05:09:46 +0000 (05:09 +0000)]
Naming consistency fix. The routing code defines
RADIX_NODE_HEAD_LOCK as grabbing the write lock,
but RADIX_NODE_HEAD_LOCK_ASSERT as checking the read lock.

Submitted by: Vijay Singh <vijju.singh at gmail.com>
MFC after: 1 month

10 years agoWhen reading physical memory, make sure to access it using the right
marcel [Tue, 4 Mar 2014 03:19:36 +0000 (03:19 +0000)]
When reading physical memory, make sure to access it using the right
memory attributes. The same applies to the mmap(2) interface. Not
doing so results in machine checks.

We find the memory attributes in the EFI memory map, as queried by
mem_phys2virt().

10 years agoFix syntax errors and warnings emitted by the GPL dtc.
imp [Tue, 4 Mar 2014 03:19:26 +0000 (03:19 +0000)]
Fix syntax errors and warnings emitted by the GPL dtc.

10 years agoReplace use of ${.CURDIR} by ${LIBC_SRCTOP} and define ${LIBC_SRCTOP}
marcel [Tue, 4 Mar 2014 02:19:39 +0000 (02:19 +0000)]
Replace use of ${.CURDIR} by ${LIBC_SRCTOP} and define ${LIBC_SRCTOP}
if not already defined. This allows building libc from outside of
lib/libc using a reach-over makefile.

A typical use-case is to build a standard ILP32 version and a COMPAT32
version in a single iteration by building the COMPAT32 version using a
reach-over makefile.

Obtained from: Juniper Networks, Inc.

10 years agoIncrease MAXLINE to deal with longer paths.
marcel [Tue, 4 Mar 2014 01:14:37 +0000 (01:14 +0000)]
Increase MAXLINE to deal with longer paths.

Obtained from: Juniper Networks, Inc.

10 years agoDocument the fact that OpenSSH default configuration requires Capsicum
delphij [Mon, 3 Mar 2014 23:26:48 +0000 (23:26 +0000)]
Document the fact that OpenSSH default configuration requires Capsicum
capability mode support in kernel, which have been worked around later
but it's still recommended to have it enabled.

10 years agoInclude kbd(4) when vt(4) is enabled as on other platforms.
jhb [Mon, 3 Mar 2014 18:06:25 +0000 (18:06 +0000)]
Include kbd(4) when vt(4) is enabled as on other platforms.

Reported by: Anton Shterenlikht

10 years agoAdd note about dtc switch.
imp [Mon, 3 Mar 2014 16:34:42 +0000 (16:34 +0000)]
Add note about dtc switch.

10 years agoAdd a missing semicolong, noticed by the gnu dt compiler.
ian [Mon, 3 Mar 2014 16:00:11 +0000 (16:00 +0000)]
Add a missing semicolong, noticed by the gnu dt compiler.

10 years agoUse the standard __used macro instead of a bare __attribute__.
ian [Mon, 3 Mar 2014 14:37:18 +0000 (14:37 +0000)]
Use the standard __used macro instead of a bare __attribute__.

Submitted by: bde

10 years agoAdd EMAC and SRAM controller entries to FDT.
ganbold [Mon, 3 Mar 2014 11:36:39 +0000 (11:36 +0000)]
Add EMAC and SRAM controller entries to FDT.
Add EMAC device to kernel config files and
enable EMAC, SRAM drivers for build.

Approved by: stas (mentor)

10 years agoAdd EMAC 10/100 Ethernet controller driver for A10/A20.
ganbold [Mon, 3 Mar 2014 11:32:55 +0000 (11:32 +0000)]
Add EMAC 10/100 Ethernet controller driver for A10/A20.
It is available mostly in A10 devices like Hackberry, Marsboard,
Mele A1000, A2000, A100 HTPC, cubieboard1 and A20 device
like cubieboard2.
TX performance can be improved using both channels 0 and 1.
RX performance is poor and needs improvement with the assistance of
external DMA controller in case there is bulk TCP receiver.

Reviewed by: yongari@
Approved by: stas (mentor)

10 years agoAdd Static Random Access Memory controller driver for A10/A20.
ganbold [Mon, 3 Mar 2014 11:24:47 +0000 (11:24 +0000)]
Add Static Random Access Memory controller driver for A10/A20.
A10/A20's SRAM is used by devices, such as CPU, EMAC,
for extra fast memory or as cache.

Approved by: stas (mentor)

10 years agoAdd gpio and clock bits for A10/A20's EMAC ethernet controller driver, such as:
ganbold [Mon, 3 Mar 2014 11:00:52 +0000 (11:00 +0000)]
Add gpio and clock bits for A10/A20's EMAC ethernet controller driver, such as:
- EMAC gpio configuration
- EMAC clock activation

Approved by: stas (mentor)

10 years agoWorkaround a GCC warning to unbreak ARM build.
delphij [Mon, 3 Mar 2014 08:01:36 +0000 (08:01 +0000)]
Workaround a GCC warning to unbreak ARM build.

Reported by: ian

10 years agoExplicitly request review/regression testing on the
gjb [Sun, 2 Mar 2014 23:51:03 +0000 (23:51 +0000)]
Explicitly request review/regression testing on the
release/release.sh script, as it is heavily used on
a regular basis, not just during a release cycle.

Sponsored by: The FreeBSD Foundation

10 years agosh: Simplify expari().
jilles [Sun, 2 Mar 2014 22:59:34 +0000 (22:59 +0000)]
sh: Simplify expari().

Redo expari() like evalvar(). This makes the logic more understandable and
avoids possible problems if arithmetic expansion occurs if CTLESC characters
are not generated (looking backwards for CTLARI is not generally possible in
that case but the old code tried anyway).

This adds an extra argstr() recursion.

10 years agoAdd __attribute__((used)) so that the delay implementation doesn't get
ian [Sun, 2 Mar 2014 21:25:32 +0000 (21:25 +0000)]
Add __attribute__((used)) so that the delay implementation doesn't get
optimized away as unreferenced, causing linker errors when trying to
resolve the weak reference to the missing function.

10 years agoAdd a tunable to set the number of active cores, and enable SMP by default.
ian [Sun, 2 Mar 2014 19:46:03 +0000 (19:46 +0000)]
Add a tunable to set the number of active cores, and enable SMP by default.

10 years agoInclude needed header. I don't know why clang didn't complain about this
ian [Sun, 2 Mar 2014 17:19:19 +0000 (17:19 +0000)]
Include needed header.  I don't know why clang didn't complain about this
but gcc sure did.

10 years agoWhen querying a process's umask via sysctl in libprocstat(), don't
rwatson [Sun, 2 Mar 2014 13:26:08 +0000 (13:26 +0000)]
When querying a process's umask via sysctl in libprocstat(), don't
print a warning if EPERM is returned as this is an expected failure
mode rather than error -- similar to current handling of ESRCH.
This makes the output of 'procstat -as' vastly more palatable.

MFC after: 3 days
Sponsored by: DARPA, AFRL

10 years agoAdd fetch and patch to .PHONY and .NOTMAIN only if PORTNAME is defined.
ache [Sun, 2 Mar 2014 13:12:06 +0000 (13:12 +0000)]
Add fetch and patch to .PHONY and .NOTMAIN only if PORTNAME is defined.
Unconditional addition cause real fetch(1) and patch(1) programs are
builded every time.

MFC after:      1 week

10 years agoMFV r262639: ncurses 5.9 20140222 snapshot.
delphij [Sun, 2 Mar 2014 08:58:21 +0000 (08:58 +0000)]
MFV r262639: ncurses 5.9 20140222 snapshot.

Requested by: bapt

10 years agoAdd a description here.
adrian [Sun, 2 Mar 2014 07:39:37 +0000 (07:39 +0000)]
Add a description here.

10 years agoSet all of the ports into the same vlangroup; there's only one vlangroup
adrian [Sun, 2 Mar 2014 07:10:43 +0000 (07:10 +0000)]
Set all of the ports into the same vlangroup; there's only one vlangroup
(pvid=1) and we already configure them to send to other ports.

Setting pvid=portnum would mean that there were separate vlangroups
for each ports, but 'leaking' into other ports.  The result? All port
traffic flooded to all other port traffic.

Tested:

* DB120, AR9344 + AR8327 switch

10 years ago- Fix BBB's dts compilation with GNU dtc
gonzo [Sun, 2 Mar 2014 06:36:49 +0000 (06:36 +0000)]
- Fix BBB's dts compilation with GNU dtc

10 years agoAdd ATU flush support.
adrian [Sun, 2 Mar 2014 05:48:56 +0000 (05:48 +0000)]
Add ATU flush support.

The OpenWRT AR8xxx switch support flushes the ATU (address translation
unit) after each port link 'up' status change.  I've modified this to
just flush on any port transition.

Whilst here, bump the number of ports on the AR8327 to 6, rather than
the default of 5.  It's DB120 specific; I'll go and make this configurable
later.

There's some debugging code in here still; I am still debugging whether
this is or isn't working fully.

Tested:

* DB120, AR9344 + AR8327 switch

Obtained from: OpenWRT

10 years agoAdd AR8216 era ATU management/configuration register definitions.
adrian [Sun, 2 Mar 2014 05:47:05 +0000 (05:47 +0000)]
Add AR8216 era ATU management/configuration register definitions.

Obtained from: OpenWRT

10 years agoImprove upon previous commit:
marcel [Sun, 2 Mar 2014 03:34:06 +0000 (03:34 +0000)]
Improve upon previous commit:
1.  Check return of mmap(2) (*)
2.  Avoid FD leak when fstat fails.
3.  Fix style(9).

(*) Pointed out by jmg@

10 years agoufs: small formatting fixes.
pfg [Sun, 2 Mar 2014 02:52:34 +0000 (02:52 +0000)]
ufs: small formatting fixes.

Cleanup some extra space.
Use of tabs vs. spaces.
No functional change.

MFC after: 3 days
Reviewed by: mckusick

10 years agoAdd the USB EHCI flags required for the post-AR71xx devices.
adrian [Sun, 2 Mar 2014 02:49:20 +0000 (02:49 +0000)]
Add the USB EHCI flags required for the post-AR71xx devices.

Tested:

* DB120, AR9344

10 years agoAll callers of static method load_nvlist() in spa.c handles error case,
delphij [Sun, 2 Mar 2014 02:41:33 +0000 (02:41 +0000)]
All callers of static method load_nvlist() in spa.c handles error case,
so there is no reason to assert that we won't hit an error.  Instead,
just return that error to caller and have the upper layer handle it.

Obtained from: FreeNAS
Reported by: rodrigc
Reviewed by: Matthew Ahrens
MFC after: 2 weeks

10 years agoRevive support for /etc/issue, lost due revision 81963.
marcel [Sun, 2 Mar 2014 00:47:47 +0000 (00:47 +0000)]
Revive support for /etc/issue, lost due revision 81963.

Obtained from: Juniper Networks, Inc.

10 years agoUse ${MAKE} so that we always use the same version/implementation
marcel [Sun, 2 Mar 2014 00:14:57 +0000 (00:14 +0000)]
Use ${MAKE} so that we always use the same version/implementation
of make.

10 years agoWhen our linker merges .SUNW_dof sections from multiple files, it simply
markj [Sat, 1 Mar 2014 23:09:07 +0000 (23:09 +0000)]
When our linker merges .SUNW_dof sections from multiple files, it simply
concatenates the DOF tables into one section. Previously, the USDT init
code in drti.o would only look at the first table in the DOF section; with
this change, it iterates over all the tables, passing each DOF table to
the kernel.

PR: 186821
Submitted by: Fedor Indutny <fedor@indutny.com>
MFC after: 1 month

10 years agoext2fs: small formatting fixes.
pfg [Sat, 1 Mar 2014 21:22:20 +0000 (21:22 +0000)]
ext2fs: small formatting fixes.

Remove some redundant spaces.
No functional change.

MFC after: 3 days

10 years agoIf we can't find the u-boot api signature, exit with an error status that
ian [Sat, 1 Mar 2014 19:25:23 +0000 (19:25 +0000)]
If we can't find the u-boot api signature, exit with an error status that
includes "badab1"; u-boot displays the status in hex, so this may help
give some clue about why ubldr isn't working (more than -1 does anyway).

10 years agoExpose a few DTrace parameters as sysctls under kern.dtrace and add
markj [Sat, 1 Mar 2014 19:06:43 +0000 (19:06 +0000)]
Expose a few DTrace parameters as sysctls under kern.dtrace and add
descriptions for several existing sysctls.

PR: 187027
Submitted by: Fedor Indutny <fedor@indutny.com> (original version)
MFC after: 2 weeks

10 years agoPrevent fdt data loaded from a file from overwriting the kernel environment,
ian [Sat, 1 Mar 2014 19:02:00 +0000 (19:02 +0000)]
Prevent fdt data loaded from a file from overwriting the kernel environment,
by having uboot_autoload() do the fdt setup (which may load a file) rather
than waiting until we're actually in the process of launching the kernel.

As part of making this happen...
 - Define LOADER_FDT_SUPPORT on the uboot/lib compile command line when
   MK_FDT is set.
 - Make fdt_setup_fdtb() public.
 - Declare public fdt_whatever() functions in a header instead of using
   scattered extern decls in .c files.

10 years ago- Make pointer easier to control when moving slowly.
hselasky [Sat, 1 Mar 2014 18:56:50 +0000 (18:56 +0000)]
- Make pointer easier to control when moving slowly.
- Increase chance of vertical scrolling as vertical scrolling is used more often.

Submitted by: Huang Wen Hui <huanghwh@gmail.com>
MFC after: 2 weeks

10 years agoFix emulation of call and jmp instructions on i386 and for 32-bit processes
markj [Sat, 1 Mar 2014 17:55:20 +0000 (17:55 +0000)]
Fix emulation of call and jmp instructions on i386 and for 32-bit processes
on amd64.

Submitted by: Prashanth Kumar <pra_udupi@yahoo.co.in>
MFC after: 2 weeks

10 years agoApply fix for lldb not linking after the sparc backend import.
dim [Sat, 1 Mar 2014 16:07:17 +0000 (16:07 +0000)]
Apply fix for lldb not linking after the sparc backend import.

Reported by: se
Pointy hat to: dim

10 years agoSilence the gcc warning about unsigned comparison with 0 in a different
tijl [Sat, 1 Mar 2014 13:23:52 +0000 (13:23 +0000)]
Silence the gcc warning about unsigned comparison with 0 in a different
way because mnemonic_ext_size isn't a preprocessor macro.

MFC after: 3 days

10 years ago(I think!) make the AR8327 switch correctly handle traffic.
adrian [Sat, 1 Mar 2014 10:04:31 +0000 (10:04 +0000)]
(I think!) make the AR8327 switch correctly handle traffic.

This patch does four things:

* it globally disables mirroring;
* it globally sets the mirroring on each port to be disabled;
* the initial port setup now programs a portmask for the port to allow
  transmission (forwarding) to all other ports bar itself;
* the vlan setup path now programs the portmask for the port to
  allow transmission (forwarding) to all other ports bar itself.

Before this, I hard-coded the portmask to 0x3f which would mean all
ports (bar port 6, which currently isn't hooked up to anything.)
This means that traffic would be duplicated back out the port it
received it.  I bet this wasn't .. optimal.

In any case, this _seems_ to make DHCP from my macosx laptop
work through this access point.  I'll do some further testing
to ensure it's actually working correctly on all my devices.

Tested:

* DB120, AR8327 switch

10 years agossh-copy-id: add restorecon call
eadler [Sat, 1 Mar 2014 03:36:46 +0000 (03:36 +0000)]
ssh-copy-id: add restorecon call

In certain situations when creating an authorized_key file on a Linux machine
restorecon(1) may need to be called.  Therefore, attempt to run it if it exists.

MFC After: 1 week
Idea from: https://bugzilla.redhat.com/show_bug.cgi?id=739989

10 years agoFix a race between VMRUN() and vcpu_notify_event() due to 'vcpu->hostcpu'
neel [Sat, 1 Mar 2014 03:17:58 +0000 (03:17 +0000)]
Fix a race between VMRUN() and vcpu_notify_event() due to 'vcpu->hostcpu'
being updated outside of the vcpu_lock(). The race is benign and could
potentially result in a missed notification about a pending interrupt to
a vcpu. The interrupt would not be lost but rather delayed until the next
VM exit.

The vcpu's hostcpu is now updated concurrently with the vcpu state change.
When the vcpu transitions to the RUNNING state the hostcpu is set to 'curcpu'.
It is set to 'NOCPU' in all other cases.

Reviewed by: grehan

10 years agossh-copy-id: avoid sending private keys; add -v option
eadler [Sat, 1 Mar 2014 03:17:46 +0000 (03:17 +0000)]
ssh-copy-id: avoid sending private keys; add -v option

To help avoid confusion: when attempting to send a key file check to see if a
file of the same name exists with a '.pub' suffix and send that instead.  This
mimics the behavior of other ssh-copy-id scripts.

Add -v passthrough.

Reported by: dweimer <dweimer@dweimer.net>
Reported by: feld
MFC After: 1 week

10 years agoAdd the bluetooth information structs to the firmware RX statistics message.
adrian [Sat, 1 Mar 2014 02:57:22 +0000 (02:57 +0000)]
Add the bluetooth information structs to the firmware RX statistics message.

It turns out that there's a variant format of the RX statisitcs notification
from the intel firmware.  It's even more whacked - the non-BT variant has
bluetooth fields; apparently some later NICs return even _more_ bluetooth
related fields.

I'll commit the statistics structure changes here - it's a no-op for the
driver.  I'll later teach the driver code to populate a statistics structure
from the received message after reformatting things correctly.

I don't _think_ it's going to fix anything related to sensitivity programming
as the CCK/OFDM (non-11n) fields are in the same place for both formats.
But the HT structure and the general statistics aren't in the same place.

I'll go find some NIC(s) that spit out the other format and when I find one,
I'll go and update the driver to handle things correctly.

Tested:

* Intel 5100 (which returns the legacy, non-BT format)

Obtained from: Linux iwlwifi

10 years agoVendor import of ncurses 5.9 20140222 snapshot.
delphij [Sat, 1 Mar 2014 00:40:26 +0000 (00:40 +0000)]
Vendor import of ncurses 5.9 20140222 snapshot.

10 years agoBe paranoid about bit operations here.
adrian [Sat, 1 Mar 2014 00:11:45 +0000 (00:11 +0000)]
Be paranoid about bit operations here.

10 years agoCompensate previous import with the removed files.
delphij [Sat, 1 Mar 2014 00:09:36 +0000 (00:09 +0000)]
Compensate previous import with the removed files.

10 years agoRemove now dead code.
adrian [Sat, 1 Mar 2014 00:02:09 +0000 (00:02 +0000)]
Remove now dead code.

10 years agoBump __FreeBSD_version after import of ncurses 5.9.
delphij [Fri, 28 Feb 2014 23:52:12 +0000 (23:52 +0000)]
Bump __FreeBSD_version after import of ncurses 5.9.

10 years agoMFV r262617: ncurses 5.9.
delphij [Fri, 28 Feb 2014 23:48:13 +0000 (23:48 +0000)]
MFV r262617: ncurses 5.9.

10 years agoAdd a script which lists SVN log entries only for changesets
rodrigc [Fri, 28 Feb 2014 23:12:20 +0000 (23:12 +0000)]
Add a script which lists SVN log entries only for changesets
which have files which have been Added or Removed.

This can help identify changesets which should be
mentioned in Release Engineering's release notes.

10 years agoFix syntax errors (missing ;) that bsdl dtc was tolerant of, but gpl
imp [Fri, 28 Feb 2014 22:37:40 +0000 (22:37 +0000)]
Fix syntax errors (missing ;) that bsdl dtc was tolerant of, but gpl
dtc isn't.

10 years agoUse proper include path for dtc as well as cpp.
imp [Fri, 28 Feb 2014 22:37:35 +0000 (22:37 +0000)]
Use proper include path for dtc as well as cpp.

10 years agoOnly try to build the static dtb when we're building a static dtb...
imp [Fri, 28 Feb 2014 22:06:19 +0000 (22:06 +0000)]
Only try to build the static dtb when we're building a static dtb...

10 years agoCorrect VMware capitalization.
jhb [Fri, 28 Feb 2014 21:33:40 +0000 (21:33 +0000)]
Correct VMware capitalization.

Submitted by: joeld

10 years agoext2fs: use of tab vs spaces.
pfg [Fri, 28 Feb 2014 21:25:32 +0000 (21:25 +0000)]
ext2fs: use of tab vs spaces.

Consistently use a single tab after a #define as mentioned in style(9).
Use tabs instead of space for indenting.
Fix a typo: "hash_vesion".

No functional change.

MFC after: 3 days

10 years agosince -F is safe, and an update from 10-HEAD to 10-STABLE is sooo bloody
jmg [Fri, 28 Feb 2014 19:51:47 +0000 (19:51 +0000)]
since -F is safe, and an update from 10-HEAD to 10-STABLE is sooo bloody
anoying w/o it..  recommend people use -F too...

10 years agoWhitespace fix.
jhb [Fri, 28 Feb 2014 19:21:07 +0000 (19:21 +0000)]
Whitespace fix.

10 years agoRedo import of ncurses 5.9 (stripped).
delphij [Fri, 28 Feb 2014 19:18:07 +0000 (19:18 +0000)]
Redo import of ncurses 5.9 (stripped).

10 years agoUndo two previous imports which was never done in preparation of doing a
delphij [Fri, 28 Feb 2014 19:12:44 +0000 (19:12 +0000)]
Undo two previous imports which was never done in preparation of doing a
new import.

10 years agoWorkaround an apparent bug in VMWare Fusion's nested VT support where it
jhb [Fri, 28 Feb 2014 19:07:55 +0000 (19:07 +0000)]
Workaround an apparent bug in VMWare Fusion's nested VT support where it
triggers a VM exit with the exit reason of an external interrupt but
without a valid interrupt set in the exit interrupt information.

Tested by: Michael Dexter
Reviewed by: neel
MFC after: 1 week

10 years agoIntegrate device-tree upstream files into the build process:
imp [Fri, 28 Feb 2014 18:29:09 +0000 (18:29 +0000)]
Integrate device-tree upstream files into the build process:
(1) Invoke cpp to bring in files via #include (although the old
    /include/ stuff is supported still).
(2) bring in files from either vendor tree or freebsd-custom files
    when building.
(3) move all dts* files from sys/boot/fdt/dts to
    sys/boot/fdt/dts/${MACHINE} as appropriate.
(4) encode all the magic to do the build in sys/tools/fdt/make_dtb.sh
    so that the different places in the tree use the exact same logic.
(5) switch back to gpl dtc by default. the bsdl one in the tree has
    significant issues not easily addressed by those unfamiliar with
    the code.

10 years agoMerge the projects/clang-sparc64 branch back to head. This brings in
dim [Fri, 28 Feb 2014 18:06:00 +0000 (18:06 +0000)]
Merge the projects/clang-sparc64 branch back to head.  This brings in
several updates from the llvm and clang trunks to make the sparc64
backend fully functional.

Apart from one patch to sys/sparc64/include/pcpu.h which is still under
discussion, this makes it possible to let clang fully build world and
kernel for sparc64.

Any assistance with testing this on actual sparc64 hardware is greatly
appreciated, as there will unavoidably be bugs left.

Many thanks go to Roman Divacky for his upstream work on getting the
sparc64 backend into shape.

MFC after: 1 month

10 years agoMerge from head up to r262611.
dim [Fri, 28 Feb 2014 17:46:56 +0000 (17:46 +0000)]
Merge from head up to r262611.

10 years agoPull in r196874 from upstream llvm trunk:
dim [Fri, 28 Feb 2014 17:12:31 +0000 (17:12 +0000)]
Pull in r196874 from upstream llvm trunk:

  Fix a crash that occurs when PWD is invalid.

  MCJIT needs to be able to run in hostile environments, even when PWD
  is invalid. There's no need to crash MCJIT in this case.

  The obvious fix is to simply leave MCContext's CompilationDir empty
  when PWD can't be determined. This way, MCJIT clients,
  and other clients that link with LLVM don't need a valid working directory.

  If we do want to guarantee valid CompilationDir, that should be done
  only for clients of getCompilationDir(). This is as simple as checking
  for an empty string.

  The only current use of getCompilationDir is EmitGenDwarfInfo, which
  won't conceivably run with an invalid working dir. However, in the
  purely hypothetically and untestable case that this happens, the
  AT_comp_dir will be omitted from the compilation_unit DIE.

This should help fix assertions occurring with ports-mgmt/tinderbox,
when it is using jails, and sometimes invalidates clang's current
working directory.

Reported by: decke
MFC after: 2 weeks
X-MFC-With: r261991

10 years agoSymbolic bindings for the dts files...
imp [Fri, 28 Feb 2014 16:51:33 +0000 (16:51 +0000)]
Symbolic bindings for the dts files...

10 years agoAdd missing skeleton file.
imp [Fri, 28 Feb 2014 16:43:56 +0000 (16:43 +0000)]
Add missing skeleton file.

10 years agoAdd skeleton.dtsi and update to new place in tree.
imp [Fri, 28 Feb 2014 16:42:14 +0000 (16:42 +0000)]
Add skeleton.dtsi and update to new place in tree.

10 years agoAdd proper svn:keyword props
imp [Fri, 28 Feb 2014 16:35:01 +0000 (16:35 +0000)]
Add proper svn:keyword props

10 years agoThese file have a, possibly dubious, GPL claim. Until that's sorted, move to gnu.
imp [Fri, 28 Feb 2014 16:26:44 +0000 (16:26 +0000)]
These file have a, possibly dubious, GPL claim. Until that's sorted, move to gnu.

10 years agomdoc cleanup.
brueffer [Fri, 28 Feb 2014 15:55:35 +0000 (15:55 +0000)]
mdoc cleanup.

MFC after: 1 week

10 years agoAdd device and gadget bindings for fdt. These are preliminary and will
imp [Fri, 28 Feb 2014 03:00:31 +0000 (03:00 +0000)]
Add device and gadget bindings for fdt. These are preliminary and will
need work before they work, especially the gadget bindings.

10 years agotcb device for fdt
imp [Fri, 28 Feb 2014 03:00:28 +0000 (03:00 +0000)]
tcb device for fdt

10 years agoshdwc device for fdt
imp [Fri, 28 Feb 2014 03:00:25 +0000 (03:00 +0000)]
shdwc device for fdt

10 years agoAdd device node for SDRAMC device. Currently just claims device's
imp [Fri, 28 Feb 2014 02:59:51 +0000 (02:59 +0000)]
Add device node for SDRAMC device. Currently just claims device's
resources.

10 years agoAdd fdt binding for nand for Atmel parts. This does little more than
imp [Fri, 28 Feb 2014 02:38:04 +0000 (02:38 +0000)]
Add fdt binding for nand for Atmel parts. This does little more than
match the device. Pinctrl will need to be added before this will work,
in addition to migrating the current board_foo.c method of configuring
these pins to something else. Non-FDT systems won't be affected, yet.

10 years ago4478 dtrace_dof_maxsize is far too small
markj [Fri, 28 Feb 2014 02:04:41 +0000 (02:04 +0000)]
4478 dtrace_dof_maxsize is far too small

illumos/illumos-gate@d339a29bb4765c4b6883a935cf69b669cd05bca0

PR: 187027
MFC after: 1 week

10 years ago4478 dtrace_dof_maxsize is far too small
markj [Fri, 28 Feb 2014 01:53:37 +0000 (01:53 +0000)]
4478 dtrace_dof_maxsize is far too small

illumos/illumos-gate@d339a29bb4765c4b6883a935cf69b669cd05bca0

10 years agoCorrect initial import script
imp [Fri, 28 Feb 2014 01:33:03 +0000 (01:33 +0000)]
Correct initial import script

10 years agoInitial import of DTS files from Linux
imp [Fri, 28 Feb 2014 01:32:44 +0000 (01:32 +0000)]
Initial import of DTS files from Linux

10 years agoInitial import of Linux/Vendor DTS files for various embedded boards.
imp [Fri, 28 Feb 2014 01:31:35 +0000 (01:31 +0000)]
Initial import of Linux/Vendor DTS files for various embedded boards.

10 years agoAdd an armv7 implementation of cpu_sleep(). The arm11/armv6 implementation
ian [Fri, 28 Feb 2014 00:41:55 +0000 (00:41 +0000)]
Add an armv7 implementation of cpu_sleep().  The arm11/armv6 implementation
we've been using was actually just spinning due to ARM having redefined
the old 'wait for interrupt' operation via the system coprocessor as a nop
and replacing it with a WFI instruction.

10 years agoAdd some rudimentary voltage control to go with the rudimentary frequency
ian [Fri, 28 Feb 2014 00:26:57 +0000 (00:26 +0000)]
Add some rudimentary voltage control to go with the rudimentary frequency
control.  If we have to scale back the frequency due to temperature, it
will help to lower the voltage as well.

10 years agoSupply a DELAY() implementation via weak linkage, so that SoC-specific
ian [Fri, 28 Feb 2014 00:23:04 +0000 (00:23 +0000)]
Supply a DELAY() implementation via weak linkage, so that SoC-specific
code can supply a better implementation.  A SoC with variable CPU frequency
is likely to use a fixed-frequency timer for DELAY() (but still use the
mpcore private timers as eventtimers).

Also remove spaces from the eventtimer and timecounter names.

10 years agoAll our current ARM multi-core systems have all cores in one package with
ian [Fri, 28 Feb 2014 00:17:03 +0000 (00:17 +0000)]
All our current ARM multi-core systems have all cores in one package with
a shared L2 cache, reflect that in the common cpu_topo() routine.

10 years agoPull in r202422 from upstream llvm trunk (by Roman Divacky):
dim [Thu, 27 Feb 2014 23:17:00 +0000 (23:17 +0000)]
Pull in r202422 from upstream llvm trunk (by Roman Divacky):

  Lower FNEG just like FABS to fneg[ds] and fmov[ds], thus avoiding
  expensive libcall. Also, Qp_neg is not implemented on at least
  FreeBSD. This is also what gcc is doing.

10 years agoInitialize the Low Power Mode bits to keep the ARM cores running during WFI.
ian [Thu, 27 Feb 2014 22:55:33 +0000 (22:55 +0000)]
Initialize the Low Power Mode bits to keep the ARM cores running during WFI.

10 years agoMFV r262570:
delphij [Thu, 27 Feb 2014 21:50:46 +0000 (21:50 +0000)]
MFV r262570:

4626 libzfs memleak in zpool_in_use()

illumos/illumos-gate@fb13f48f1d9593453b94cd1c7277553b56f493c8

MFC after: 2 weeks

10 years agoRestore SIM freeze/release match, broken at r253549.
mav [Thu, 27 Feb 2014 21:48:34 +0000 (21:48 +0000)]
Restore SIM freeze/release match, broken at r253549.

This fixes problem with SIM left in frozen state after reinit, for example,
after firmware update.

MFC after: 1 week

10 years agoAdd tws(4) to the hardware notes.
brueffer [Thu, 27 Feb 2014 21:47:20 +0000 (21:47 +0000)]
Add tws(4) to the hardware notes.

MFC after: 3 days

10 years agoMove to dist, per layout standards
imp [Thu, 27 Feb 2014 21:46:31 +0000 (21:46 +0000)]
Move to dist, per layout standards

10 years agoGet rid of the 'chan' from softc structure and use the latter
delphij [Thu, 27 Feb 2014 21:41:52 +0000 (21:41 +0000)]
Get rid of the 'chan' from softc structure and use the latter
directly as sleep channel.

PR: kern/174684
Submitted by: jmg
MFC after: 2 weeks

10 years agoAdd missing includes and remove two unused ones.
brueffer [Thu, 27 Feb 2014 21:01:10 +0000 (21:01 +0000)]
Add missing includes and remove two unused ones.

Reviewed by: loos
MFC after: 1 week