]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
9 years agoAdd a ROADMAP file for FreeBSD/arm release builds, which
gjb [Thu, 12 Feb 2015 00:50:54 +0000 (00:50 +0000)]
Add a ROADMAP file for FreeBSD/arm release builds, which
will make final documentation much easier when done.

Sponsored by: The FreeBSD Foundation

9 years agoFix build-arm-ports target for WANDBOARD-QUAD.
gjb [Wed, 11 Feb 2015 20:51:58 +0000 (20:51 +0000)]
Fix build-arm-ports target for WANDBOARD-QUAD.

Sponsored by: The FreeBSD Foundation

9 years agoAdd CROSS_TOOLCHAIN for arm builds.
gjb [Wed, 11 Feb 2015 20:50:25 +0000 (20:50 +0000)]
Add CROSS_TOOLCHAIN for arm builds.
Remove a BOARDS entry left in while testing.
Add build-arm-ports${BOARD} target as a prerequisite when
ARMPORTS is set.
Add build-arm-KERNCONF and build-arm-ports to CLEANFILES.

Sponsored by: The FreeBSD Foundation

9 years agoInitial restructure for FreeBSD/arm release builds.
gjb [Wed, 11 Feb 2015 20:25:21 +0000 (20:25 +0000)]
Initial restructure for FreeBSD/arm release builds.

Sponsored by: The FreeBSD Foundation

9 years agoMFH: r278202,r278205-r278590
gjb [Wed, 11 Feb 2015 18:59:46 +0000 (18:59 +0000)]
MFH: r278202,r278205-r278590

Sponsored by: The FreeBSD Foundation

9 years agoClose the file list before opening the container that holds the
marcel [Wed, 11 Feb 2015 17:56:24 +0000 (17:56 +0000)]
Close the file list before opening the container that holds the
totals, otherwise we end up emitting invalid JSON -- provided
libxo does not prevent us from doing that.

PR: 197499
Submitted by: allanjude@

9 years agoflowctl: Replace alloca() with an array.
pfg [Wed, 11 Feb 2015 17:46:35 +0000 (17:46 +0000)]
flowctl: Replace alloca() with an array.

Reviewed by: glebius

9 years agoClean up machine dependent code for DTrace on ARM.
gnn [Wed, 11 Feb 2015 17:27:37 +0000 (17:27 +0000)]
Clean up machine dependent code for DTrace on ARM.

Submitted by: markj

9 years agoFix ldscripts such that ld(1) collects the .fini_array section in the same
tijl [Wed, 11 Feb 2015 17:25:23 +0000 (17:25 +0000)]
Fix ldscripts such that ld(1) collects the .fini_array section in the same
order as the .init_array section.  Finalisation routines need to be called
in the opposite order as their corresponding initialisation routines but
rtld(1) handles that by calling the function pointers in .fini_array in
reverse order.

Reviewed by: kib
MFC after: 2 weeks

9 years agoFix formatting.
trasz [Wed, 11 Feb 2015 17:00:32 +0000 (17:00 +0000)]
Fix formatting.

MFC after: 1 month
Sponsored by: The FreeBSD Foundation

9 years agoAdd support for General Statistics and Performance log page.
mav [Wed, 11 Feb 2015 16:10:31 +0000 (16:10 +0000)]
Add support for General Statistics and Performance log page.

CTL already collects most of statistics reported there, so why not.

MFC after: 2 weeks

9 years agoMFi386:
jhb [Wed, 11 Feb 2015 15:49:14 +0000 (15:49 +0000)]
MFi386:
When building some of the boot loaders with clang, and DEBUG_FLAGS or
CFLAGS having '-g' in it, clang outputs several assembly directives that
are too new for our version of binutils.

Therefore, assemble the resulting .s files with clang instead.  A more
general solution can be implemented when a GNU as-compatible driver for
clang's integrated assembler appears.

Tested by: gjb

9 years agolibdwarf: Add aarch64 relocation support
emaste [Wed, 11 Feb 2015 14:59:35 +0000 (14:59 +0000)]
libdwarf: Add aarch64 relocation support

Reviewed by: andrew, rpaulo
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D1817

9 years agoSummary: Update CPU identification call to recent version.
gnn [Wed, 11 Feb 2015 12:08:40 +0000 (12:08 +0000)]
Summary: Update CPU identification call to recent version.

9 years agoRemove the non-EABI code from the DDB stack unwinder, we only support the
andrew [Wed, 11 Feb 2015 10:40:49 +0000 (10:40 +0000)]
Remove the non-EABI code from the DDB stack unwinder, we only support the
ARM EABI now.

9 years agoThe cpu_id macro was renamed in r278529, catch up with this new name.
andrew [Wed, 11 Feb 2015 10:37:55 +0000 (10:37 +0000)]
The cpu_id macro was renamed in r278529, catch up with this new name.

9 years agoRemove remnant from texinfo
bapt [Wed, 11 Feb 2015 08:26:36 +0000 (08:26 +0000)]
Remove remnant from texinfo

9 years agoDocument caveat with specifying WITHOUT_VI at build/install time when building
ngie [Wed, 11 Feb 2015 07:49:00 +0000 (07:49 +0000)]
Document caveat with specifying WITHOUT_VI at build/install time when building
older releases of FreeBSD

9 years agoadd an assert in case the sizeof int ever becomes bigger.. Then we will
jmg [Wed, 11 Feb 2015 07:44:53 +0000 (07:44 +0000)]
add an assert in case the sizeof int ever becomes bigger.. Then we will
have issues, at least we'll know where one of them are..

Submitted by: Erich Dollansky

9 years agoAdd preliminary support for the Ralink RT5390 and RT5392 chipsets.
kevlo [Wed, 11 Feb 2015 05:25:23 +0000 (05:25 +0000)]
Add preliminary support for the Ralink RT5390 and RT5392 chipsets.
Committed over the D-Link DWA-525 rev A2 on amd64 with WPA.

9 years agoRestore the data array in coredump(), but use a different style to
rpaulo [Wed, 11 Feb 2015 00:58:15 +0000 (00:58 +0000)]
Restore the data array in coredump(), but use a different style to
calculate the length.

Requested by: kib

9 years agofetch has supported https for a while..
jmg [Tue, 10 Feb 2015 23:18:36 +0000 (23:18 +0000)]
fetch has supported https for a while..

I would convert the url to https, but we still don't install a CA set
by default...  we even don't install the CA that authenticates
www.FreeBSD.org..

9 years agoAdd xo_finish() to w.c in case it's invoked as uptime
grembo [Tue, 10 Feb 2015 22:23:52 +0000 (22:23 +0000)]
Add xo_finish() to w.c in case it's invoked as uptime

Reviewed by: marcel
Approved by: marcel
Differential Revision: https://reviews.freebsd.org/D1821

9 years agoFix compilation with GCC in the PAE case.
marius [Tue, 10 Feb 2015 21:33:32 +0000 (21:33 +0000)]
Fix compilation with GCC in the PAE case.

MFC after: 3 days

9 years agoWhen catopen(3) returns an error, it caches the result of that error
bdrewery [Tue, 10 Feb 2015 20:45:40 +0000 (20:45 +0000)]
When catopen(3) returns an error, it caches the result of that error
from r202992. The refcount on the cache entry is not initialized, so
any attempt to clean the cache will skip over this item since it likely
has a >0 value.

This change is currently a NOP. There is work in progress to support
freeing the cache which requires this change to avoid a memory leak.

MFC after: 2 weeks
Sponsored by: EMC / Isilon Storage Division

9 years agoInitial version of DTrace on ARM32.
gnn [Tue, 10 Feb 2015 19:41:30 +0000 (19:41 +0000)]
Initial version of DTrace on ARM32.

Submitted by: Howard Su based on work by Oleksandr Tymoshenko
Reviewed by: ian, andrew, rpaulo, markj

9 years agoFix a rendering issue in the nullfs(5) manual page.
gjb [Tue, 10 Feb 2015 18:49:13 +0000 (18:49 +0000)]
Fix a rendering issue in the nullfs(5) manual page.

MFC after: 3 days
Sponsored by: The FreeBSD Foundation

9 years agoRemove a printf and an strlen() from the coredump code.
rpaulo [Tue, 10 Feb 2015 18:35:46 +0000 (18:35 +0000)]
Remove a printf and an strlen() from the coredump code.

9 years agoUpdate UPDATING after changing autofs(4) ABI.
trasz [Tue, 10 Feb 2015 18:05:58 +0000 (18:05 +0000)]
Update UPDATING after changing autofs(4) ABI.

Suggested by: bdrewery@
Sponsored by: The FreeBSD Foundation

9 years agoMountd iterating over the mount points may race with the parallel
kib [Tue, 10 Feb 2015 18:00:32 +0000 (18:00 +0000)]
Mountd iterating over the mount points may race with the parallel
unmount, which causes error from nmount(2) call when performing
MNT_DELEXPORT over the directory which ceased to be a mount point.

The race is legitimate and innocent, but results in the chatty mountd.
Silence it by providing an distinguished error code for the situation,
and ignoring the error in mountd loop.

Based on the patch by: Andreas Longwitz <longwitz@incore.de>
Prodded and tested by: bdrewery
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks

9 years agoRestore ABI compatibility, broken in r273127. Note that while this fixes
trasz [Tue, 10 Feb 2015 16:17:16 +0000 (16:17 +0000)]
Restore ABI compatibility, broken in r273127.  Note that while this fixes
ABI with 10.1, it breaks ABI for 11-CURRENT, so rebuild of automountd(8)
is neccessary.

MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation

9 years agoProperly quote EXTRA_MODULES and WITHOUT_MODULES to ensure that they
imp [Tue, 10 Feb 2015 15:15:08 +0000 (15:15 +0000)]
Properly quote EXTRA_MODULES and WITHOUT_MODULES to ensure that they
are passed down properly when there's more than one.

9 years agoResolve cache line size from CP15
zbb [Tue, 10 Feb 2015 14:11:23 +0000 (14:11 +0000)]
Resolve cache line size from CP15

Switch the cache line size during invalidations/flushes
to be read from CP15 cache type register.

Submitted by:  Wojciech Macek <wma@semihalf.com>
Reviewed by:   ian, imp
Obtained from: Semihalf

9 years agorc.d: load the network config file for netif
royger [Tue, 10 Feb 2015 13:07:08 +0000 (13:07 +0000)]
rc.d: load the network config file for netif

r272959 broke compatibility with mfsBSD that stores the default network
config file in /etc/rc.conf.d/network. In order to fix that load the network
config file from netif also.

9 years agoBump copyright, forgotten in r278502.
gjb [Tue, 10 Feb 2015 12:38:39 +0000 (12:38 +0000)]
Bump copyright, forgotten in r278502.

MFC after: 1 week
X-MFC-with: r277458, r277536, r277606, r277609,
r277836, r278118, r278119, r278206,
r278502
Sponsored by: The FreeBSD Foundation

9 years agoRevert r274918 and make a better solution. Poll the synchronisation
hselasky [Tue, 10 Feb 2015 12:08:52 +0000 (12:08 +0000)]
Revert r274918 and make a better solution. Poll the synchronisation
endpoint less frequently to make the sample rate adjustment more
accurate. This should resolve problems with the DN32-USB module for
Midas audio systems and possibly other similar products from Klark
Teknik.

MFC after: 3 days

9 years agoIn scripts/mk-vmimage.sh, prevent incorrect usage() by
gjb [Tue, 10 Feb 2015 12:04:38 +0000 (12:04 +0000)]
In scripts/mk-vmimage.sh, prevent incorrect usage() by
defaulting VMCONFIG to /dev/null, and additionally
ensuring VMCONFIG is not a character device before it
is sourced.

While here, be sure to exit if usage() is called.

This should effectively be no-op, but the usage() output
was discovered while investigating a larger issue.

MFC after: 1 week
X-MFC-with: r277458, r277536, r277606, r277609,
r277836, r278118, r278119, r278206
Sponsored by: The FreeBSD Foundation

9 years agoDo not abort already aborted tasks.
mav [Tue, 10 Feb 2015 11:37:32 +0000 (11:37 +0000)]
Do not abort already aborted tasks.

This fixes abort of new tasks with the same tags as previously aborted,
but still remaining on the queue.

MFC after: 1 week

9 years agoRemove FreeBSD/wii.
rpaulo [Tue, 10 Feb 2015 06:35:16 +0000 (06:35 +0000)]
Remove FreeBSD/wii.

This port failed to gain traction and probably only a couple Wii consoles
ran FreeBSD all the way to single user mode with an md(4). IPC
support was never implemented, so it was impossible to use any peripheral

Any further development, if any, will happen at https://github.com/rpaulo/wii.

Discussed with: nathanw (a long time ago), jhibbits

9 years agoSanitise the coredump file names sent to devd.
rpaulo [Tue, 10 Feb 2015 04:34:39 +0000 (04:34 +0000)]
Sanitise the coredump file names sent to devd.

While there, add a sysctl to turn this feature off as requested by
kib@.

9 years agoUpdate copyrights.
dteske [Tue, 10 Feb 2015 03:34:42 +0000 (03:34 +0000)]
Update copyrights.

9 years agoAdd "vt_"* aliases to bsdconfig console entries.
dteske [Tue, 10 Feb 2015 03:17:51 +0000 (03:17 +0000)]
Add "vt_"* aliases to bsdconfig console entries.
NB: "syscons_"* kept for backward compatibility

9 years agoAdd bsdconfig api function f_dialog_menutag2help()
dteske [Tue, 10 Feb 2015 03:15:54 +0000 (03:15 +0000)]
Add bsdconfig api function f_dialog_menutag2help()

9 years agoAdd bsdconfig api functions f_dialog_pause()/f_dialog_pause_no_cancel()
dteske [Tue, 10 Feb 2015 03:12:11 +0000 (03:12 +0000)]
Add bsdconfig api functions f_dialog_pause()/f_dialog_pause_no_cancel()

9 years agoEliminate sub-shells where possible for performance.
dteske [Tue, 10 Feb 2015 02:55:10 +0000 (02:55 +0000)]
Eliminate sub-shells where possible for performance.

MFC after: 7 days

9 years agoWhitespace.
dteske [Tue, 10 Feb 2015 02:53:26 +0000 (02:53 +0000)]
Whitespace.

MFC after: 3 days

9 years agoExclude 'tests' from DEBUG_DISTRIBUTIONS.
gjb [Tue, 10 Feb 2015 01:35:26 +0000 (01:35 +0000)]
Exclude 'tests' from DEBUG_DISTRIBUTIONS.

Sponsored by: The FreeBSD Foundation

9 years agocxgbe(4): allow the SET_FILTER_MODE ioctl to change the mode when it's
np [Tue, 10 Feb 2015 01:16:43 +0000 (01:16 +0000)]
cxgbe(4): allow the SET_FILTER_MODE ioctl to change the mode when it's
safe to do so.

MFC after: 1 month

9 years agoComments.
dteske [Tue, 10 Feb 2015 01:02:02 +0000 (01:02 +0000)]
Comments.

MFC after: 3 days

9 years agoAdd error reporting to interrupt CPU binding.
nwhitehorn [Tue, 10 Feb 2015 00:57:26 +0000 (00:57 +0000)]
Add error reporting to interrupt CPU binding.

9 years agoSet thread priorities on multithreaded CPUs so that threads holding a
nwhitehorn [Tue, 10 Feb 2015 00:55:42 +0000 (00:55 +0000)]
Set thread priorities on multithreaded CPUs so that threads holding a
spinlock are high-priority and threads waiting for a spinlock are set to
low priority.

9 years agoUn-revert the r278323 again - whatever Jenkins/kyua is up it, it has
jamie [Tue, 10 Feb 2015 00:48:51 +0000 (00:48 +0000)]
Un-revert the r278323 again - whatever Jenkins/kyua is up it, it has
nothing to do with this.

9 years agoNotify devd(8) when a process crashed.
rpaulo [Mon, 9 Feb 2015 23:13:50 +0000 (23:13 +0000)]
Notify devd(8) when a process crashed.

This change implements a notification (via devctl) to userland when
the kernel produces coredumps after a process has crashed.
devd can then run a specific command to produce a human readable crash
report.  The command is most usually a helper that runs gdb/lldb
commands on the file/coredump pair.  It's possible to use this
functionality for implementing automatic generation of crash reports.

devd(8) will be notified of the full path of the binary that crashed and
the full path of the coredump file.

9 years agoAdd the ability to print a vendor copyright in dmesg.
rpaulo [Mon, 9 Feb 2015 23:04:30 +0000 (23:04 +0000)]
Add the ability to print a vendor copyright in dmesg.

9 years agoFix DMA address casts. Regression issue after r278279.
hselasky [Mon, 9 Feb 2015 21:47:12 +0000 (21:47 +0000)]
Fix DMA address casts. Regression issue after r278279.

MFC after: 3 days

9 years agoUse __builtin_popcountl() instead of __builtin_popcount().
jhb [Mon, 9 Feb 2015 21:39:18 +0000 (21:39 +0000)]
Use __builtin_popcountl() instead of __builtin_popcount().

Submitted by: jkim

9 years agoAdd tests for `pw usernext'.
brd [Mon, 9 Feb 2015 21:15:52 +0000 (21:15 +0000)]
Add tests for `pw usernext'.

PR: 197120
Submitted by: Robert O'Neil <oneil.rs@gmail.com>
Approved by: will

9 years agoUse __builtin_popcnt() to implement a BIT_COUNT() operation for bitsets and
jhb [Mon, 9 Feb 2015 21:03:23 +0000 (21:03 +0000)]
Use __builtin_popcnt() to implement a BIT_COUNT() operation for bitsets and
use this to implement CPU_COUNT() to count the number of CPUs in a cpuset.

MFC after: 2 weeks

9 years agoAdd x2APIC support. Enable it by default if CPU is capable. The
kib [Mon, 9 Feb 2015 21:00:56 +0000 (21:00 +0000)]
Add x2APIC support.  Enable it by default if CPU is capable.  The
hw.x2apic_enable tunable allows disabling it from the loader prompt.

To closely repeat effects of the uncached memory ops when accessing
registers in the xAPIC mode, the x2APIC writes to MSRs are preceeded
by mfence, except for the EOI notifications.  This is probably too
strict, only ICR writes to send IPI require serialization to ensure
that other CPUs see the previous actions when IPI is delivered.  This
may be changed later.

In vmm justreturn IPI handler, call doreti_iret instead of doing iretd
inline, to handle corner conditions.

Note that the patch only switches LAPICs into x2APIC mode. It does not
enables FreeBSD to support > 255 CPUs, which requires parsing x2APIC
MADT entries and doing interrupts remapping, but is the required step
on the way.

Reviewed by: neel
Tested by: pho (real hardware), neel (on bhyve)
Discussed with: jhb, grehan
Sponsored by: The FreeBSD Foundation
MFC after: 2 months

9 years agoThis fixes a bug in the way that the LLE timers for nd6
rrs [Mon, 9 Feb 2015 19:28:11 +0000 (19:28 +0000)]
This fixes a bug in the way that the LLE timers for nd6
and arp were being used. They basically would pass in the
mutex to the callout_init. Because they used this method
to the callout system, it was possible to "stop" the callout.
When flushing the table and you stopped the running callout, the
callout_stop code would return 1 indicating that it was going
to stop the callout (that was about to run on the callout_wheel blocked
by the function calling the stop). Now when 1 was returned, it would
lower the reference count one extra time for the stopped timer, then
a few lines later delete the memory. Of course the callout_wheel was
stuck in the lock code and would then crash since it was accessing
freed memory. By using callout_init(c, 1) we always get a 0 back
and the reference counting bug does not rear its head. We do have
to make a few adjustments to the callouts themselves though to make
sure it does the proper thing if rescheduled as well as gets the lock.

Commented upon by hiren and sbruno
See Phabricator D1777 for more details.

Commented upon by hiren and sbruno
Reviewed by: adrian, jhb and bz
Sponsored by: Netflix Inc.

9 years agoDistribute interrupts across multiple CPUs in SMP configurations instead of sending...
nwhitehorn [Mon, 9 Feb 2015 19:21:54 +0000 (19:21 +0000)]
Distribute interrupts across multiple CPUs in SMP configurations instead of sending them
all to CPU 0.

9 years agoAdd new alias "bsdconfig api" (same as "bsdconfig includes")
dteske [Mon, 9 Feb 2015 19:20:59 +0000 (19:20 +0000)]
Add new alias "bsdconfig api" (same as "bsdconfig includes")
NB: My fingers like typing "api" a lot more than "includes"

MFC after: 3 days

9 years agoThis fixes two conditions that can incur when migration
rrs [Mon, 9 Feb 2015 19:19:44 +0000 (19:19 +0000)]
This fixes two conditions that can incur when migration
is being done in the callout code and harmonizes the macro
use.:
1) The callout_active() will lie. Basically if a migration
   is occuring and the callout is about to expire and the
   migration has been deferred, the callout_active will no
   longer return true until after the migration. This confuses
   and breaks callers that are doing callout_init(&c, 1); such
   as TCP.
2) The migration code had a bug in it where when migrating, if
   a two calls to callout_reset came in and they both collided with
   the callout on the wheel about to run, then the second call to
   callout_reset would corrupt the list the callout wheel uses
   putting the callout thread into a endless loop.
3) Per imp, I have fixed all the macro occurance in the code that
   were for the most part being ignored.

Phabricator D1711 and looked at by lstewart and jhb and sbruno.
Reviewed by: kostikbel, imp, adrian, hselasky
MFC after: 3 days
Sponsored by: Netflix Inc.

9 years agoAdd a few more instructions to xz/FREEBSD-upgrade.
rpaulo [Mon, 9 Feb 2015 19:19:13 +0000 (19:19 +0000)]
Add a few more instructions to xz/FREEBSD-upgrade.

9 years agoReplace the only instance of sed(1) in bsdconfig(8) with awk(1).
dteske [Mon, 9 Feb 2015 19:16:19 +0000 (19:16 +0000)]
Replace the only instance of sed(1) in bsdconfig(8) with awk(1).

MFC after: 3 days

9 years agoRemove explicit routing/sendmail rc.d inclusion in FILES
ngie [Mon, 9 Feb 2015 17:53:16 +0000 (17:53 +0000)]
Remove explicit routing/sendmail rc.d inclusion in FILES

Reported by: Guy Yur <guyyur@gmail.com>
MFC after: 14 days
X-MFC with: r278249
Sponsored by: EMC / Isilon Storage Division

9 years agoRevert SVN r278464 -- fingers were in perforce mode (premature
dteske [Mon, 9 Feb 2015 17:22:54 +0000 (17:22 +0000)]
Revert SVN r278464 -- fingers were in perforce mode (premature
commit of future stuffs not yet ready).

9 years agoReplace the only instance of sed(1) in bsdconfig(8) with awk(1).
dteske [Mon, 9 Feb 2015 17:18:10 +0000 (17:18 +0000)]
Replace the only instance of sed(1) in bsdconfig(8) with awk(1).

MFC after: 3 days

9 years agoPrefer install over mkdir to create the directory. Add test to ensure
imp [Mon, 9 Feb 2015 16:29:44 +0000 (16:29 +0000)]
Prefer install over mkdir to create the directory. Add test to ensure
the directory doesn't exist before creating it. This makes the NO_ROOT
case actually work.

Suggested by: brooks@

9 years agoMove these definitions to bsd.own.mk along side the kernel directory /
imp [Mon, 9 Feb 2015 16:21:49 +0000 (16:21 +0000)]
Move these definitions to bsd.own.mk along side the kernel directory /
user defines. Make this work with WITH_INSTALL_AS_USER as well.

Noticed by: ian@

9 years agoFix typo.
imp [Mon, 9 Feb 2015 16:10:30 +0000 (16:10 +0000)]
Fix typo.

Submitted by: jhb@

9 years agoWe need to create /boot/dtb since some use cases don't create a full
imp [Mon, 9 Feb 2015 16:04:08 +0000 (16:04 +0000)]
We need to create /boot/dtb since some use cases don't create a full
root with BSD.root.mtree, so it often times will not exist. Rather
than force the latter for an installkernel, just create the directory
with a comment about why.

Submitted by: Guy Yur

9 years agoMake sure that we define DTB{OWN,GRP,MODE} so install works.
imp [Mon, 9 Feb 2015 16:04:01 +0000 (16:04 +0000)]
Make sure that we define DTB{OWN,GRP,MODE} so install works.

Submitted by: Guy Yur

9 years agoPass MODULES_EXTRA to the modules build. While I'm here, also always
imp [Mon, 9 Feb 2015 16:03:59 +0000 (16:03 +0000)]
Pass MODULES_EXTRA to the modules build. While I'm here, also always
pass WITHOUT_MODULES down. There's no need to make this conditional.

Submitted by: Guy Yur

9 years agoEnforce that MK_foo options shall be either "yes" or "no" and nothing
imp [Mon, 9 Feb 2015 16:03:55 +0000 (16:03 +0000)]
Enforce that MK_foo options shall be either "yes" or "no" and nothing
else.

9 years agoMark invalid page table entries correctly for PMAP as well as for the
nwhitehorn [Mon, 9 Feb 2015 15:58:27 +0000 (15:58 +0000)]
Mark invalid page table entries correctly for PMAP as well as for the
hypervisor. This prevents an infinite loop where processes with evicted
pages would page fault forever when PMAP decided the evicted pages on
which the process was faulting was actually present and did not need to
be restored.

Found while building LLVM with make -j32.

Sponsored by: FreeBSD Foundation

9 years agoProperly hide a variable under #ifdef as it is only used inside the
bz [Mon, 9 Feb 2015 11:34:45 +0000 (11:34 +0000)]
Properly hide a variable under #ifdef as it is only used inside the
specific #ifdef block otherwise leaving an unused variable and breaking
other kernel builds.

9 years agoRevert r278445.
gjb [Mon, 9 Feb 2015 10:46:39 +0000 (10:46 +0000)]
Revert r278445.

I was going to use __FreeBSD_version to determine if
xz(1) should be multi-threaded by default, but doing
this will cause problems if/when the changes are merged
from head.

Sponsored by: The FreeBSD Foundation

9 years agoEnable multi-threaded xz(1) compression, after r278433.
gjb [Mon, 9 Feb 2015 10:42:27 +0000 (10:42 +0000)]
Enable multi-threaded xz(1) compression, after r278433.
Allow multi-threaded xz(1) to be turned off by specifying
NO_XZTHREADS, and allow number of threads to be overridden
by specifying XZ_THREADS=N.

MFC after: 1 week
X-MFC-needs: r278433
Sponsored by: The FreeBSD Foundation

9 years agoDefine FREEBSD_VERSION in release/Makefile, derived
gjb [Mon, 9 Feb 2015 09:56:21 +0000 (09:56 +0000)]
Define FREEBSD_VERSION in release/Makefile, derived
from sys/sys/param.h.

Sponsored by: The FreeBSD Foundation

9 years agoChase r278433: bump __FreeBSD_version for xz 5.2.0 which adds multi-thread
delphij [Mon, 9 Feb 2015 09:17:00 +0000 (09:17 +0000)]
Chase r278433: bump __FreeBSD_version for xz 5.2.0 which adds multi-thread
capability to liblzma.

9 years agoAfter r278004 was committed, Bruce Evans noted that the casts were
dim [Mon, 9 Feb 2015 07:56:50 +0000 (07:56 +0000)]
After r278004 was committed, Bruce Evans noted that the casts were
actually completely unnecessary, here:

https://lists.freebsd.org/pipermail/svn-src-all/2015-February/098478.html

Remove the casts, and just assign &xxx_io_mc_regs[0][0] directly.

Reviewed by: dumbbell
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D1748

9 years agoFix typo in PTE insertion overflow handling: use the page we're actually
nwhitehorn [Mon, 9 Feb 2015 07:08:54 +0000 (07:08 +0000)]
Fix typo in PTE insertion overflow handling: use the page we're actually
returning, not the one we just looked at.

9 years agoMerge xz 5.2.0.
rpaulo [Mon, 9 Feb 2015 06:20:34 +0000 (06:20 +0000)]
Merge xz 5.2.0.

This brings support for multi-threaded compression.  This brings close
N times faster compression where N is the number of CPU cores.
Because of this, liblzma now depends on libthr.

Soon libarchive will be modified to use the new lzma API.

Thanks to antoine@ for the exp-run.

Differential Revision:  https://reviews.freebsd.org/D1786
Reviewed by: bapt

9 years agoDo not mark shared structures as __packed, it leads to race condition
gonzo [Mon, 9 Feb 2015 02:31:27 +0000 (02:31 +0000)]
Do not mark shared structures as __packed, it leads to race condition

If structure packed as __packed clang (and probably gcc) generates
code that loads word fields (e.g. tx_pos)  byte-by-byte and if it's
modified by VideoCore in the same time as ARM loads the value result
is going to be mixed combination of bytes from previous value and
new one.

9 years agoRemove unused variables
gonzo [Mon, 9 Feb 2015 02:27:33 +0000 (02:27 +0000)]
Remove unused variables

9 years agoFix an extremely subtle concurrency bug triggered by running on 32-thread
nwhitehorn [Mon, 9 Feb 2015 02:17:21 +0000 (02:17 +0000)]
Fix an extremely subtle concurrency bug triggered by running on 32-thread
POWER8 systems. During thread switch, there was a very small window when
the stack pointer was set to the stack pointer of the outgoing thread, but
after the lock on that thread had already been released.

If, during that window, the outgoing thread were rescheduled on another CPU
and begin execution and an exception were taken on the original CPU, the
trap handler and the outgoing thread would simultaneously execute on the same
stack, causing memory corruption. Fix this by making sure to release the
old thread only after cpu_switch() is done with its stack.

MFC after: 2 weeks
Sponsored by: FreeBSD Foundation

9 years agoTechnically speaking, using one virtal processor area for all CPUs is a
nwhitehorn [Mon, 9 Feb 2015 02:13:36 +0000 (02:13 +0000)]
Technically speaking, using one virtal processor area for all CPUs is a
violation of the spec. Make duplicate entries for each CPU.

9 years agoSimplify trapcode setup by placing a copy of the generic trap handler at
nwhitehorn [Mon, 9 Feb 2015 02:12:38 +0000 (02:12 +0000)]
Simplify trapcode setup by placing a copy of the generic trap handler at
every possible trap address by default. This also makes sure the kernel
notices (and panics at) traps from newer CPUs that the kernel was not
expecting rather than executing gibberish memory.

9 years agorevert r278414
bapt [Sun, 8 Feb 2015 23:03:41 +0000 (23:03 +0000)]
revert r278414

Reported by: bz

9 years agoReapply r255891
bapt [Sun, 8 Feb 2015 22:24:18 +0000 (22:24 +0000)]
Reapply r255891

9 years agoPreset the object's color, or alignment, to maximize superpage usage.
alc [Sun, 8 Feb 2015 21:00:51 +0000 (21:00 +0000)]
Preset the object's color, or alignment, to maximize superpage usage.

MFC after: 5 days

9 years agoFix ordering of "*logout" and "*terminate"; no functional changes.
trasz [Sun, 8 Feb 2015 19:18:23 +0000 (19:18 +0000)]
Fix ordering of "*logout" and "*terminate"; no functional changes.

MFC after: 1 month
Sponsored by: The FreeBSD Foundation

9 years agoExtend ICL to add receive offload methods. For software ICL backend
trasz [Sun, 8 Feb 2015 19:15:14 +0000 (19:15 +0000)]
Extend ICL to add receive offload methods.  For software ICL backend
they are no-ops.

MFC after: 1 month
Sponsored by: The FreeBSD Foundation

9 years agoAdd some error checking on the supplied page size list. This makes sure
nwhitehorn [Sun, 8 Feb 2015 16:50:00 +0000 (16:50 +0000)]
Add some error checking on the supplied page size list. This makes sure
that we (a) get the correct large page size to provide to pmap and (b)
we can alert the user if running under incorrectly-configured PowerKVM
on POWER7 and POWER8 systems.

MFC after: 1 week

9 years agoAdd Neoway WM620 module ID.
dchagin [Sun, 8 Feb 2015 11:55:29 +0000 (11:55 +0000)]
Add Neoway WM620 module ID.

MFC after: 1 Week

9 years agoMake output of "iscsictl -v" and "ctladm islist -v" a little prettier
trasz [Sun, 8 Feb 2015 10:58:25 +0000 (10:58 +0000)]
Make output of "iscsictl -v" and "ctladm islist -v" a little prettier
by capitalizing "None".

MFC after: 1 month
Sponsored by: The FreeBSD Foundation

9 years agocxgbe(4): tidy up some of the interaction between the Upper Layer
np [Sun, 8 Feb 2015 09:28:55 +0000 (09:28 +0000)]
cxgbe(4): tidy up some of the interaction between the Upper Layer
Drivers (ULDs) and the base if_cxgbe driver.

Track the per-adapter activation of ULDs in a new "active_ulds" field.
This was done pretty arbitrarily before this change -- via TOM_INIT_DONE
in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in
adapter->offload_map for iWARP.

iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM
ULD).  The rules are:
a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then
   iWARP and/or iSCSI are enabled too.
b) When the iWARP and iSCSI modules are loaded they go looking for
   adapters with TOE enabled and enable themselves on that adapter.
c) You cannot deactivate or unload the TOM module from underneath iWARP
   or iSCSI.  Any such attempt will fail with EBUSY.

MFC after: 2 weeks

9 years agocxgbe(4): adapter_full_init is always a synchronized operation.
np [Sun, 8 Feb 2015 08:52:18 +0000 (08:52 +0000)]
cxgbe(4): adapter_full_init is always a synchronized operation.

MFC after: 1 week

9 years agocxgbe(4): a change to the synchronization rules within the the driver.
np [Sun, 8 Feb 2015 08:42:45 +0000 (08:42 +0000)]
cxgbe(4): a change to the synchronization rules within the the driver.
This is purely cosmetic because the new rules are already followed.

MFC after: 1 week