freebsd.git
10 hours agoPOSIX states that #include <signal.h> shall make both mcontext_t and master github/master
kib [Fri, 12 Feb 2016 07:38:19 +0000 (07:38 +0000)]
POSIX states that #include <signal.h> shall make both mcontext_t and
ucontext_t available.  Our code even has XXX comment about this.

Add a bit of compliance by moving struct __ucontext definition into
sys/_ucontext.h and including it into signal.h and sys/ucontext.h.

Several machine/ucontext.h headers were changed to use namespace-safe
types (like uint64_t->__uint64_t) to not depend on sys/types.h.
struct __stack_t from sys/signal.h is made always visible in private
namespace to satisfy sys/_ucontext.h requirements.

Apparently mips _types.h pollutes global namespace with f_register_t
type definition.  This commit does not try to fix the issue.

PR: 207079
Reported and tested by: Ting-Wei Lan <lantw44@gmail.com>
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks

10 hours agoIf full width writes to the performance monitoring counters are
kib [Fri, 12 Feb 2016 07:27:24 +0000 (07:27 +0000)]
If full width writes to the performance monitoring counters are
supported, use full-width aliases MSRs for writes.  This fixes the
"[pmc,X] negative increment" assertion on the context switch when
clipped counter value is sign-extended.

Add definitions for the MSR IA32_PERF_CAPABILITIES needed to detect
the feature.

PR: 207068
Submitted by: joss.upton@yahoo.com
MFC after: 2 weeks

10 hours agoAdjust the size of PMC_DEFAULT_DEBUG_FLAGS to match the size of the
kib [Fri, 12 Feb 2016 07:20:27 +0000 (07:20 +0000)]
Adjust the size of PMC_DEFAULT_DEBUG_FLAGS to match the size of the
structure.

PR: 207068
Submitted by: joss.upton@yahoo.com
MFC after: 2 weeks

10 hours agoRemove tautological cast.
kib [Fri, 12 Feb 2016 07:19:59 +0000 (07:19 +0000)]
Remove tautological cast.

PR: 207068
Submitted by: joss.upton@yahoo.com
MFC after: 2 weeks

13 hours agoUART: Fix spurious interrupts generated by ns8250 and lpc drivers:
mmel [Fri, 12 Feb 2016 05:14:58 +0000 (05:14 +0000)]
UART: Fix spurious interrupts generated by ns8250 and lpc drivers:
 - don't enable transmitter empty interrupt before filling TX FIFO.
 - add missing uart_barrier() call in interrupt service routine

15 hours agoComments and fix small bug
dteske [Fri, 12 Feb 2016 02:53:44 +0000 (02:53 +0000)]
Comments and fix small bug

Reduce differences between jib/jng and fix a bug that would prevent
additional interfaces from being created if the first of many already
existed (counter wasn't incremented before calling only continue).

15 hours agoInterpret vnet_interface/vnet.interface as array
dteske [Fri, 12 Feb 2016 02:50:36 +0000 (02:50 +0000)]
Interpret vnet_interface/vnet.interface as array

16 hours agoAdd syntax to disable MAC allocation
dteske [Fri, 12 Feb 2016 01:41:40 +0000 (01:41 +0000)]
Add syntax to disable MAC allocation

Adding `!' before an interface name will disable MAC allocation, falling
back to driver mechanics. Alternatively adding `=' before an interface name
causes the MAC address to be cloned (for ng_bridge(4) back-end only). While
here, disable the auto-detection of wlan* since this knocks the host off;
requiring the host that defines the jail to explicitly enable this feature
by preceding the interface with `='.

17 hours agoRefactor conversion to hex
dteske [Fri, 12 Feb 2016 01:12:44 +0000 (01:12 +0000)]
Refactor conversion to hex

19 hours agoLoopback addresses are 127.0.0.0/8, not 127.0.0.1/32.
tuexen [Thu, 11 Feb 2016 22:29:39 +0000 (22:29 +0000)]
Loopback addresses are 127.0.0.0/8, not 127.0.0.1/32.

MFC after: 1 week

20 hours agoAdd support for bridging iwn(4) based wlan(4)
dteske [Thu, 11 Feb 2016 22:10:54 +0000 (22:10 +0000)]
Add support for bridging iwn(4) based wlan(4)

Documented in iwn(4), "Only one virtual interface may be configured at any
time." However, netgraph with a cloned MAC address is able to communicate
over an ng_eiface attached to an ng_bridge linked to the wlan(4) interface.
While here, introduce syntax to specify the MAC address is to be cloned if
the named interface begins with equals [=].

20 hours agoo Gather all mbuf(9) allocation functions into kern_mbuf.c, and all
glebius [Thu, 11 Feb 2016 21:32:23 +0000 (21:32 +0000)]
o Gather all mbuf(9) allocation functions into kern_mbuf.c, and all
  mbuf(9) manipulation functions into uipc_mbuf.c.  This looks like
  the initial intent, but had diffused in the last decade.

o Gather all declarations in mbuf.h in one place and sort them.

o Uninline m_clget() and m_cljget().

There are no functional changes in this patch.

The patch comes from a larger version, where all mbuf(9) allocation was
uninlined, which allowed to make mbuf(9) UMA zones private to kern_mbuf.c.
The performance impact of the total uninlining is still unclear, so we
are holding on now with larger version.

Together with: melifaro, olivier

20 hours agoCentralize MAC derivation formula
dteske [Thu, 11 Feb 2016 21:28:34 +0000 (21:28 +0000)]
Centralize MAC derivation formula

23 hours agoAdd example VIMAGE config required by vnet jails
dteske [Thu, 11 Feb 2016 18:37:02 +0000 (18:37 +0000)]
Add example VIMAGE config required by vnet jails

23 hours agoUse 4 spaces instead of a tab.
tuexen [Thu, 11 Feb 2016 18:35:46 +0000 (18:35 +0000)]
Use 4 spaces instead of a tab.

24 hours agoEnable ARM_INTRNG on all kernel configs where we use the Generic Interrupt
andrew [Thu, 11 Feb 2016 18:01:10 +0000 (18:01 +0000)]
Enable ARM_INTRNG on all kernel configs where we use the Generic Interrupt
Controller.

Sponsored by: ABT Systems Ltd

24 hours agoFix double-free error: r289419 moved all error handling in http_connect()
des [Thu, 11 Feb 2016 17:48:15 +0000 (17:48 +0000)]
Fix double-free error: r289419 moved all error handling in http_connect()
to the end of the function, but did not remove a fetch_close() call which
was made redundant by the one in the shared error-handling code.

PR: 206774
Submitted by: Christian Heckendorf <heckendorfc@gmail.com>
MFC after: 3 days

24 hours agoUse the new insecure-lan-zones option instead of listing each AS112 zone
des [Thu, 11 Feb 2016 17:37:02 +0000 (17:37 +0000)]
Use the new insecure-lan-zones option instead of listing each AS112 zone
separately.

MFC after: 3 days

24 hours agoRemove unbound-contrl-setup since we use a local control socket which
des [Thu, 11 Feb 2016 17:33:55 +0000 (17:33 +0000)]
Remove unbound-contrl-setup since we use a local control socket which
does not require keys.

MFC after: 3 days
Relnotes: yes

24 hours agonvme: avoid duplicate SET_NUM_QUEUES commands
jimharris [Thu, 11 Feb 2016 17:32:41 +0000 (17:32 +0000)]
nvme: avoid duplicate SET_NUM_QUEUES commands

nvme(4) issues a SET_NUM_QUEUES command during device
initialization to ensure enough I/O queues exists for each
of the MSI-X vectors we have allocated.  The SET_NUM_QUEUES
command is then issued again during nvme_ctrlr_start(), to
ensure that is properly set after any controller reset.

At least one NVMe drive exists which fails this second
SET_NUM_QUEUES command during device initialization.  So
change nvme_ctrlr_start() to only issue its SET_NUM_QUEUES
command when it is coming out of a reset - avoiding the
duplicate SET_NUM_QUEUES during device initialization.

Reported by: gallatin
MFC after: 3 days
Sponsored by: Intel

24 hours agoApply upstream r3619, r3620, r3621: add insecure-lan-zone option
des [Thu, 11 Feb 2016 17:25:12 +0000 (17:25 +0000)]
Apply upstream r3619, r3620, r3621: add insecure-lan-zone option

25 hours agoMerge SVN r295220 (bz) from projects/vnet/
dteske [Thu, 11 Feb 2016 17:07:19 +0000 (17:07 +0000)]
Merge SVN r295220 (bz) from projects/vnet/

Fix a panic that occurs when a vnet interface is unavailable at the time the
vnet jail referencing said interface is stopped.

Sponsored by: FIS Global, Inc.

25 hours agoInclude the correct header to get a phandle_t needed by ofw_bus_if.h. While
andrew [Thu, 11 Feb 2016 16:43:23 +0000 (16:43 +0000)]
Include the correct header to get a phandle_t needed by ofw_bus_if.h. While
here only include opt_platform.h once.

Sponsored by: ABT Systems Ltd

26 hours agoExt4: Use boolean type instead of '0' and '1'
pfg [Thu, 11 Feb 2016 15:27:14 +0000 (15:27 +0000)]
Ext4: Use boolean type instead of '0' and '1'

There are precedents of uses of bool in the kernel and
it is incorrect style to use integers as replacement for
a boolean type.

27 hours agoExplain that boot1.efi prefers ZFS over UFS
emaste [Thu, 11 Feb 2016 14:45:18 +0000 (14:45 +0000)]
Explain that boot1.efi prefers ZFS over UFS

Reported by: smh

28 hours agoo Move non-generic kernel configuration out from GENERIC.
br [Thu, 11 Feb 2016 13:28:52 +0000 (13:28 +0000)]
o Move non-generic kernel configuration out from GENERIC.
o Add kernel configuration for QEMU.

Both SPIKE and QEMU kernel configs are temporary (until
we will be able to obtain DTB from loader).

Sponsored by: DARPA, AFRL
Sponsored by: HEIF5

29 hours agoMigrate the PTE format for book-e to standardize on the 'indirect PTE' format
jhibbits [Thu, 11 Feb 2016 13:15:37 +0000 (13:15 +0000)]
Migrate the PTE format for book-e to standardize on the 'indirect PTE' format

Summary:
The revised Book-E spec, adding the specification for the MMUv2 and e6500,
includes a hardware PTE layout for indirect page tables.  In order to support
this in the future, migrate the PTE format to match the MMUv2 hardware PTE
format.

Test Plan: Boot tested on a P5020 board.  Booted to multiuser mode.

Differential Revision: https://reviews.freebsd.org/D5224

30 hours agoSupport interrupts binding in GICv3 and ITS
zbb [Thu, 11 Feb 2016 12:04:58 +0000 (12:04 +0000)]
Support interrupts binding in GICv3 and ITS

- Add MOVI command and routine for the LPI migration
- Allow to search for the ITS device descriptor using
  not only devID but also LPI number.
- Bind SPIs in the Distributor
- Don't bind its_dev to collection. Keep track of the collection
  IDs for each LPI.

Reviewed by:   wma
Obtained from: Semihalf
Sponsored by:  Cavium
Differential Revision: https://reviews.freebsd.org/D5231

30 hours agoImplement finer locking in ITS
zbb [Thu, 11 Feb 2016 12:03:11 +0000 (12:03 +0000)]
Implement finer locking in ITS

- Change locks' names to be more suitable
- Don't use blocking mutex. Lock only basic operations such
  as lists or bitmaps modifications.

Reviewed by:   wma
Obtained from: Semihalf
Sponsored by:  Cavium
Differential Revision: https://reviews.freebsd.org/D5230

30 hours agoInitially bind all interrupts to the boot CPU when using GICv3
zbb [Thu, 11 Feb 2016 12:01:33 +0000 (12:01 +0000)]
Initially bind all interrupts to the boot CPU when using GICv3

This should be done by routing all interrupts to CPU0,
different assignment will be induced by either interrupts
shuffling or bus_bind_intr().

Reviewed by:   wma
Obtained from: Semihalf
Sponsored by:  Cavium
Differential Revision: https://reviews.freebsd.org/D5229

30 hours agoCall pmc_hook() correctly in the ARM64 interrupt handler
zbb [Thu, 11 Feb 2016 11:59:32 +0000 (11:59 +0000)]
Call pmc_hook() correctly in the ARM64 interrupt handler

pmc_hook() was called only in case of the stray interrupt but should
rather be called on each interrupt. Move in to the arm_cpu_intr()
handler, out of the critical section too.

Reviewed by:   br
Obtained from: Semihalf
Sponsored by:  Cavium
Differential Revision: https://reviews.freebsd.org/D5161

30 hours agoIntroduce bus_bind_intr method for ARM64
zbb [Thu, 11 Feb 2016 11:58:27 +0000 (11:58 +0000)]
Introduce bus_bind_intr method for ARM64

It can be used to bind specific interrupt to a particular CPU.
Requires PIC support for interrupts binding.

Reviewed by:   wma
Obtained from: Semihalf
Sponsored by:  Cavium
Differential Revision: https://reviews.freebsd.org/D5122

30 hours agoFix bugs in interrupts allocation on ARM64
zbb [Thu, 11 Feb 2016 11:57:13 +0000 (11:57 +0000)]
Fix bugs in interrupts allocation on ARM64

Separate interrupt descriptors lookup from allocation. It was possible
to perform config on non-existing interrupt simply by allocating spurious
descriptor.
Must lock the interrupt descriptors table lookup to avoid mismatches.
This ought to prevent trouble while setting up new interrupt
and dispatching existing one.
Use spin mutex rather than sleep mutex. This is mainly due to lock in
arm_dispatch_intr.
This should be eventually changed to a lock-less solution without
walking through a linked list on each interrupt.

Reviewed by:   andrew, wma
Obtained from: Semihalf
Sponsored by:  Cavium
Differential Revision: https://reviews.freebsd.org/D5121

30 hours agoxen-netfront: remove useless NULL check in netif_free
royger [Thu, 11 Feb 2016 11:57:12 +0000 (11:57 +0000)]
xen-netfront: remove useless NULL check in netif_free

xn_ifp is allocated in create_netdev with if_alloc(IFT_ETHER).
According to the current arrangement it can't be NULL.

Coverity ID: 1349805
Submitted by: Wei Liu <wei.liu2@citrix.com>
Reviewed by: royger
Sponsored by: Citrix Systems R&D
Differential revision: https://reviews.freebsd.org/D5252

30 hours agoMinor clean-ups for ARM64 GICv3 and GIC drivers
zbb [Thu, 11 Feb 2016 11:55:37 +0000 (11:55 +0000)]
Minor clean-ups for ARM64 GICv3 and GIC drivers

GICv3:
- move ICC_SGI1R_EL1 definitions to armreg.h and use proper system
  register's names
GIC:
- remove unused functions

Reviewed by:   andrew
Obtained from: Semihalf
Sponsored by:  Cavium
Differential Revision: https://reviews.freebsd.org/D5119

30 hours agoxen-netfront: rearrange error paths in setup_txqs
royger [Thu, 11 Feb 2016 11:53:32 +0000 (11:53 +0000)]
xen-netfront: rearrange error paths in setup_txqs

Coverity spotted double free errors in error path. Fix that by
removing the extraneous calls.

Coverity ID: 1349798
Submitted by: Wei Liu <wei.liu2@citrix.com>
Reviewed by: royger
Sponsored by: Citrix Systems R&D
Differential revision: https://reviews.freebsd.org/D5251

30 hours agoxen-netfront: remove pointless assignment in xn_ioctl
royger [Thu, 11 Feb 2016 11:50:31 +0000 (11:50 +0000)]
xen-netfront: remove pointless assignment in xn_ioctl

The variable error is assigned to 0 before entering the switch.
Assigning error to 0 before break pointless rewrites the real error
value that should be returned.

Coverity ID: 1304974
Submitted by: Wei Liu <wei.liu2@citrix.com>
Reviewed by: royger
Sponsored by: Citrix Systems R&D
Differential revision: https://reviews.freebsd.org/D5250

30 hours agoStop defining fdt_pic_table when building for ARM_INTRNG.
andrew [Thu, 11 Feb 2016 11:49:27 +0000 (11:49 +0000)]
Stop defining fdt_pic_table when building for ARM_INTRNG.

30 hours agoStop device enumeration when we see first empty slot.
br [Thu, 11 Feb 2016 11:21:45 +0000 (11:21 +0000)]
Stop device enumeration when we see first empty slot.
This fixes operation in QEMU and saves some booting time as well.

Pointed out by: Sagar Karandikar <skarandikar@berkeley.edu>
Sponsored by: DARPA, AFRL
Sponsored by: HEIF5

31 hours agoRemove unused Zedboard config files. files.zedboard was empty other than
andrew [Thu, 11 Feb 2016 10:48:15 +0000 (10:48 +0000)]
Remove unused Zedboard config files. files.zedboard was empty other than
comments, and std.zedboard just pointed to files.zedboard and std.xilinx.

32 hours agoUse a pair of ifs when comparing the 32-bit flowid integers so that
hselasky [Thu, 11 Feb 2016 10:03:50 +0000 (10:03 +0000)]
Use a pair of ifs when comparing the 32-bit flowid integers so that
the sign bit doesn't cause an overflow. The overflow manifests itself
as a sorting index wrap around in the middle of the sorted array,
which is not a problem for the LRO code, but might be a problem for
the logic inside qsort().

Reviewed by: gnn @
Sponsored by: Mellanox Technologies
Differential Revision: https://reviews.freebsd.org/D5239

35 hours agoARM64 disassembler: support for LDR instructions
wma [Thu, 11 Feb 2016 06:50:11 +0000 (06:50 +0000)]
ARM64 disassembler: support for LDR instructions

    Implemented disassembly for a whole bunch of
    various ldr instructions.

Obtained from:         Semihalf
Sponsored by:          Cavium
Approved by:           cognet (mentor)
Reviewed by:           zbb
Differential revision: https://reviews.freebsd.org/D5217

35 hours agoMake bus_space_generic properly map/unmap memory (using pmap_mapdev and
adrian [Thu, 11 Feb 2016 06:24:34 +0000 (06:24 +0000)]
Make bus_space_generic properly map/unmap memory (using pmap_mapdev and
pmap_unmapdev respectively) so that resources are properly managed.

This is work originally done by kan@.  Stanislav picked it up as part
of his Mediatek SoC work.

Tested:

* Carambola2, AR933x SoC

Submitted by: Stanislav Galabov <sgalabov@gmail.com>
Reviewed by: kan
Differential Revision: https://reviews.freebsd.org/D5184

35 hours agoMigrate the other MIPS24K SoC cores to use the CPU_MIPS24K option.
adrian [Thu, 11 Feb 2016 06:23:02 +0000 (06:23 +0000)]
Migrate the other MIPS24K SoC cores to use the CPU_MIPS24K option.

35 hours agoMissing commit - remove MIPS fdt bus space.
adrian [Thu, 11 Feb 2016 06:19:58 +0000 (06:19 +0000)]
Missing commit - remove MIPS fdt bus space.

Differential Revision: https://reviews.freebsd.org/D5184

35 hours agoRemove bus space fdt for MIPS.
adrian [Thu, 11 Feb 2016 06:19:32 +0000 (06:19 +0000)]
Remove bus space fdt for MIPS.

This was originall done by kan@.

Submitted by: Stanislav Galabov <sgalabov@gmail.com>
Reviewed by: kan
Differential Revision: https://reviews.freebsd.org/D5184

36 hours agoConvert MIPS nexus and mips_pic to BUS_PASS
adrian [Thu, 11 Feb 2016 06:15:43 +0000 (06:15 +0000)]
Convert MIPS nexus and mips_pic to BUS_PASS

Submitted by: Stanislav Galabov <sgalabov@gmail.com>
Reviewed by: kan
Differential Revision: https://reviews.freebsd.org/D5196

36 hours agoTeach the MIPS ticker to attach itself properly when using INTRNG.
adrian [Thu, 11 Feb 2016 06:13:53 +0000 (06:13 +0000)]
Teach the MIPS ticker to attach itself properly when using INTRNG.

Submitted by: Stanislav Galabov <sgalabov@gmail.com>
Reviewed by: kan
Differential Revision: https://reviews.freebsd.org/D5183

36 hours agoBegin the MIPS_INTRNG support.
adrian [Thu, 11 Feb 2016 06:09:27 +0000 (06:09 +0000)]
Begin the MIPS_INTRNG support.

This is a prelude to intr-ng support for MIPS boards that need it -
notably the CI20 port from kan@ that's upcoming, but also work that
Stanislav is doing for the Mediatek platforms.

This is the initial platform dependent bits in include/intr.h, some
#defines for the nexus code for the intrng initialisation/runtime
bits, some changed naming (which I'll fix later to be the same, much
like what I did for ARM intr-ng) in exception.S, and the first cut
at a PIC.

Stanislav and I refactored out the common code for intrng support,
so the mips intrng definitions are quite small (sys/mips/include/intr.h.)

This is all work done by kan@, which stanislav has been cherry picking
into common code for his mediatek chipset work.

Tested:

* Carambola2 - no regressions (not intr-ng though!)

Submitted by: Stanislav Galabov <sgalabov@gmail.com>
Reviewed by: kan (original author)
Differential Revision: https://reviews.freebsd.org/D5182

40 hours agoUpdate uefi.8 for ZFS and multi device boot support
emaste [Thu, 11 Feb 2016 02:02:01 +0000 (02:02 +0000)]
Update uefi.8 for ZFS and multi device boot support

As of r294068 boot1.efi can load loader.efi from ZFS.

As of r295320 boot1.efi prefers to load loader.efi from the same device
it was loaded from.

MFC after: 3 days
Sponsored by: The FreeBSD Foundation

40 hours agoDocument boot1.efi's handling of /boot.config, added in r294768
emaste [Thu, 11 Feb 2016 01:30:43 +0000 (01:30 +0000)]
Document boot1.efi's handling of /boot.config, added in r294768

MFC after: 3 days
Sponsored by: The FreeBSD Foundation

40 hours agoUpdate leapsecond file in non-chroot environments.
cy [Thu, 11 Feb 2016 01:16:57 +0000 (01:16 +0000)]
Update leapsecond file in non-chroot environments.

PR: 207095
Submitted by: madpilot
MFC after: 3 days

41 hours agoExt4: fix handling of files with sparse blocks before extent's index.
pfg [Thu, 11 Feb 2016 00:34:11 +0000 (00:34 +0000)]
Ext4: fix handling of files with sparse blocks before extent's index.

This is ongoing work from Damjan Jovanovic to improve ext4 read support
with sparse files:

Keep track of the first and last block in each extent as it descends down
the extent tree, thus being able to work out that some blocks are sparse
earlier. This solves an issue on r293680.

In ext4_bmapext() start supporting the runb parameter, which appears to be
the number of adjacent blocks prior to the block being converted in the
same way that runp is the number of blocks after, speding up random access
to mmaped files.

PR: 206652

41 hours agoAdd WITH_GDB src.conf(5) description
emaste [Thu, 11 Feb 2016 00:30:51 +0000 (00:30 +0000)]
Add WITH_GDB src.conf(5) description

42 hours agoAdd missing src.conf(5) descriptions for tool chain components
emaste [Thu, 11 Feb 2016 00:14:00 +0000 (00:14 +0000)]
Add missing src.conf(5) descriptions for tool chain components

Sponsored by: The FreeBSD Foundation

42 hours agoUpdate list of binutils controlled by WITHOUT_BINUTILS_BOOTSTRAP
emaste [Wed, 10 Feb 2016 23:57:09 +0000 (23:57 +0000)]
Update list of binutils controlled by WITHOUT_BINUTILS_BOOTSTRAP

Tools removed from the list are now provided by the ELF Tool Chain
project and are not controlled by the WITHOUT_BINUTILS_BOOTSTRAP knob.

Sponsored by: The FreeBSD Foundation

44 hours agoRemove useless checks for NULL before calling free(9), in the kernel
kib [Wed, 10 Feb 2016 21:35:00 +0000 (21:35 +0000)]
Remove useless checks for NULL before calling free(9), in the kernel
elf linkers.

Found by: Related PVS-Studio diagnostic
Sponsored by: The FreeBSD Foundation
MFC after: 1 week

44 hours agoFinish r173600. There is no need to test a condition if both cases
kib [Wed, 10 Feb 2016 21:16:37 +0000 (21:16 +0000)]
Finish r173600.  There is no need to test a condition if both cases
result in the same value.

Found by: PVS-Studio
Sponsored by: The FreeBSD Foundation
MFC after: 1 week

45 hours agontb_hw(4): Print correct PAT name for non-WC/WB types mapped at load
cem [Wed, 10 Feb 2016 20:49:22 +0000 (20:49 +0000)]
ntb_hw(4): Print correct PAT name for non-WC/WB types mapped at load

Sponsored by: EMC / Isilon Storage Division

45 hours agontb_hw(4): Allow any x86 PAT caching flags for MW defaults
cem [Wed, 10 Feb 2016 20:28:28 +0000 (20:28 +0000)]
ntb_hw(4): Allow any x86 PAT caching flags for MW defaults

Replace the hw.ntb.enable_writecombine tunable with
hw.ntb.default_mw_pat.  It can be set with several specific numerical
values to select a caching type.  Any bogus value is treated as
Uncacheable (UC).

The ntb_mw_set_wc() KPI has removed the restriction that the selected
mode must be one of UC, WC, or WB.

Sponsored by: EMC / Isilon Storage Division

47 hours agoAdd a kern.icl.drivers sysctl, to retrieve the list of registered
trasz [Wed, 10 Feb 2016 19:01:26 +0000 (19:01 +0000)]
Add a kern.icl.drivers sysctl, to retrieve the list of registered
ICL drivers.

MFC after: 1 month
Sponsored by: The FreeBSD Foundation

47 hours agoGarbage collect unused arguments of m_init().
glebius [Wed, 10 Feb 2016 18:54:18 +0000 (18:54 +0000)]
Garbage collect unused arguments of m_init().

47 hours agoGarbage collect m_getclr().
glebius [Wed, 10 Feb 2016 18:36:51 +0000 (18:36 +0000)]
Garbage collect m_getclr().

2 days agoSimilar to r295116, add an additional 1 second sleep after calling ggatel
ngie [Wed, 10 Feb 2016 17:57:24 +0000 (17:57 +0000)]
Similar to r295116, add an additional 1 second sleep after calling ggatel
before calling dd to defeat a race when writing out to the geom_gate(4)
device

MFC after: 1 month
Reported by: Jenkins
Sponsored by: EMC / Isilon Storage Division

2 days agoRemove stray semicolons from the iSCSI code.
trasz [Wed, 10 Feb 2016 17:55:39 +0000 (17:55 +0000)]
Remove stray semicolons from the iSCSI code.

MFC after: 1 month
Sponsored by: The FreeBSD Foundation

2 days agoExplicitly include <sys/systm.h>. Previously we were pulling it in due to
andrew [Wed, 10 Feb 2016 16:49:20 +0000 (16:49 +0000)]
Explicitly include <sys/systm.h>. Previously we were pulling it in due to
header pollution.

2 days agoFix a logic inversion when checking for ARM_INTRNG
andrew [Wed, 10 Feb 2016 16:44:20 +0000 (16:44 +0000)]
Fix a logic inversion when checking for ARM_INTRNG

Pointy hat to: andrew

2 days agoOnly use fdt_pic_table when not using ARM_INTRNG
andrew [Wed, 10 Feb 2016 16:39:02 +0000 (16:39 +0000)]
Only use fdt_pic_table when not using ARM_INTRNG

2 days agoAdd new rc.conf parameter "jail_reverse_stop"
feld [Wed, 10 Feb 2016 16:13:59 +0000 (16:13 +0000)]
Add new rc.conf parameter "jail_reverse_stop"

When a user defines "jail_list" in rc.conf the jails are started in the
order defined. Currently the jails are not are stopped in reverse order
which may break dependencies between jails/services and prevent a clean
shutdown. The new parameter "jail_reverse_stop" will shutdown jails in
"jail_list" in reverse order when set to "YES".

Please note that this does not affect manual invocation of the jail rc
script. If a user runs the command

  # service jail stop jail1 jail2 jail3

the jails will be stopped in exactly the order specified regardless of
jail_reverse_stop being defined in rc.conf.

PR: 196152
Approved by: jamie
MFC after: 1 week
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D5233

2 days agoUse OF_getencprop to read a property, it already handles the endian
andrew [Wed, 10 Feb 2016 15:56:52 +0000 (15:56 +0000)]
Use OF_getencprop to read a property, it already handles the endian
conversion so we don't need to pass the data through fdt32_to_cpu.

2 days agoRemove man page references to rndassociates.com, which has been taken over
jamie [Wed, 10 Feb 2016 14:48:49 +0000 (14:48 +0000)]
Remove man page references to rndassociates.com, which has been taken over
by a domain squatter.

2 days agosfxge: implement SIOCGI2C to read information from phy modules
arybchik [Wed, 10 Feb 2016 12:14:56 +0000 (12:14 +0000)]
sfxge: implement SIOCGI2C to read information from phy modules

The IOCTL is used by 'ifconfig -v' to show SFP+/QSFP+ information
including inventory information and dianostics (temperature, light
levels, voltage etc).

Reviewed by:    gnn,melifaro
Sponsored by:   Solarflare Communications, Inc.
MFC after:      2 days
Differential Revision: https://reviews.freebsd.org/D5240

2 days agoRevert an arm64 change that sneaked in with r295464.
andrew [Wed, 10 Feb 2016 10:28:33 +0000 (10:28 +0000)]
Revert an arm64 change that sneaked in with r295464.

2 days agoRemove O_SYNC from the options passed to dbmopen().
se [Wed, 10 Feb 2016 09:40:45 +0000 (09:40 +0000)]
Remove O_SYNC from the options passed to dbmopen().

The services db is created as a temporary file that is moved over the
existing file after completion. Thus there is no need to immediately
flush all created db records to the temporary file.

This speeds up creation of the services db by a factor of 500 on my
ZFS based /var/db filesytem (from 110 seconds to 0.235 seconds).

MFC after: 1 week

2 days agoUpdate of the Allwinner drivers to:
andrew [Wed, 10 Feb 2016 09:19:29 +0000 (09:19 +0000)]
Update of the Allwinner drivers to:
 * Use the Linux compat string
 * Use EARLY_DRIVER_MODULE to attach at the right time
 * Add a generic A10 kernel config file
 * A20 now use generic_timer
 * Add two new dts files for Olimex boards
 * Update our custom DTS file for A10 and A20 to use the same compatible
   property names as the vendor ones.

Submitted by: Emmanuel Vadot <manu@bidouilliste.com>
Differential Revision: https://reviews.freebsd.org/D4792

2 days agoCorrect PCI device description.
hselasky [Wed, 10 Feb 2016 08:03:10 +0000 (08:03 +0000)]
Correct PCI device description.

Submitted by: Dmitry Luhtionov <dmitryluhtionov@gmail.com>

2 days agoAdd `stats' command to get ng_bridge(4) stats
dteske [Wed, 10 Feb 2016 04:56:38 +0000 (04:56 +0000)]
Add `stats' command to get ng_bridge(4) stats

2 days agoBreak out the shared bits of the arm intrng definitions into sys/intr.h;
adrian [Wed, 10 Feb 2016 04:43:08 +0000 (04:43 +0000)]
Break out the shared bits of the arm intrng definitions into sys/intr.h;
leave the machine dependent bits in sys/arm/.

This is in preparation for MIPS INTRNG work.

Submitted by: Stanislav Galabov <sgalabov@gmail.com>

2 days agoAdd vnet jail configuration examples
dteske [Wed, 10 Feb 2016 04:04:31 +0000 (04:04 +0000)]
Add vnet jail configuration examples

2 days agoFix make universe when running with non POSIX/C locales using a locale sensitive
bapt [Wed, 10 Feb 2016 00:26:01 +0000 (00:26 +0000)]
Fix make universe when running with non POSIX/C locales using a locale sensitive
pattern

Reported by: many
Submitted by: jilles

2 days agoZero pf rule counters so daily reports make sense
lidl [Tue, 9 Feb 2016 21:00:38 +0000 (21:00 +0000)]
Zero pf rule counters so daily reports make sense

Zero pf rule counters so that each daily report lists an absolute
number of rejected packets, not the total since the last time the
machine rebooted (or the counters were manually cleared).

PR: 206467
Submitted by: Rick Adams
Approved by: rpaulo (mentor)
Differential Revision: https://reviews.freebsd.org/D5172

2 days agoInclude sys/_task.h into uma_int.h, so that taskqueue.h isn't a
glebius [Tue, 9 Feb 2016 20:22:35 +0000 (20:22 +0000)]
Include sys/_task.h into uma_int.h, so that taskqueue.h isn't a
requirement for uma_int.h.

Suggested by: jhb

2 days agoRename variable to reflect the condition.
kib [Tue, 9 Feb 2016 18:35:37 +0000 (18:35 +0000)]
Rename variable to reflect the condition.

Suggested by: jhb
Sponsored by: The FreeBSD Foundation

2 days agoRemove unneeded whitespace, and fix a license for a new file. Extracted
andrew [Tue, 9 Feb 2016 18:19:24 +0000 (18:19 +0000)]
Remove unneeded whitespace, and fix a license for a new file. Extracted
from a larger patch.

Submitted by: Emmanuel Vadot <manu@bidouilliste.com>
X-Differential Revision: https://reviews.freebsd.org/D4792

3 days agoBring generated interfaces up always
dteske [Tue, 9 Feb 2016 18:11:18 +0000 (18:11 +0000)]
Bring generated interfaces up always

3 days agoChange MAC algo to support mixing w/jng interfaces
dteske [Tue, 9 Feb 2016 18:10:36 +0000 (18:10 +0000)]
Change MAC algo to support mixing w/jng interfaces

3 days agoFix bpf unhiding example /etc/devfs.rules
dteske [Tue, 9 Feb 2016 18:08:40 +0000 (18:08 +0000)]
Fix bpf unhiding example /etc/devfs.rules

3 days agoTo support userspace audio daemons like Virtual OSS, /dev/sndstat is
hselasky [Tue, 9 Feb 2016 17:09:14 +0000 (17:09 +0000)]
To support userspace audio daemons like Virtual OSS, /dev/sndstat is
made writeable by the root user. Userspace audio daemons can add or
update an entry in /dev/sndstat by doing a single system write call to
any /dev/sndstat file descriptor handle. When the audio daemon closes the
file handle or is killed the entry disappears.

While at it, cleanup the sound status code a bit:
- keep the device list sorted to avoid sorting the list every time a
  /dev/sndstat read request is made.
- factor out locking into a pair of locking macros.
- use the sound status lock to protect all per file handle states,
  when generating the output for /dev/sndstat and when removing or
  adding sound status devices. This way sndstat_acquire() and
  sndstat_release() become superfluous and can be removed.

Reviewed by: mav @
Differential Revision: https://reviews.freebsd.org/D5191

3 days agoUpdate our copy of the Linux dts files to be in sync with Linux 4.5-rc1. We
andrew [Tue, 9 Feb 2016 16:42:32 +0000 (16:42 +0000)]
Update our copy of the Linux dts files to be in sync with Linux 4.5-rc1. We
now uave a full copy of the arm device tree sources to help when adding
support for newer boards.

Sponsored by: ABT Systems Ltd

3 days agoRename P_KTHREAD struct proc p_flag to P_KPROC.
kib [Tue, 9 Feb 2016 16:30:16 +0000 (16:30 +0000)]
Rename P_KTHREAD struct proc p_flag to P_KPROC.

I left as is an apparent bug in ntoskrnl_var.h:AT_PASSIVE_LEVEL()
definition.

Suggested by: jhb
Sponsored by: The FreeBSD Foundation

3 days agoUse designated initialisers for the db disassembler interface to help with
andrew [Tue, 9 Feb 2016 10:18:22 +0000 (10:18 +0000)]
Use designated initialisers for the db disassembler interface to help with
finding the struct when searching for one of its members.

3 days agoCorrect the year, it's 2016.
andrew [Tue, 9 Feb 2016 10:06:06 +0000 (10:06 +0000)]
Correct the year, it's 2016.

3 days agoAdd an ARMv7 config file to build nanobsd images for the qemu virt
andrew [Tue, 9 Feb 2016 09:57:50 +0000 (09:57 +0000)]
Add an ARMv7 config file to build nanobsd images for the qemu virt
platform.

Sponsored by: ABT Systems Ltd

3 days agoDon't set the offset when loading the kernel on the arm loader.efi. The
andrew [Tue, 9 Feb 2016 09:39:30 +0000 (09:39 +0000)]
Don't set the offset when loading the kernel on the arm loader.efi. The
copyin and copyout code handle virtual addresses such that they will take
a virtual address and convert it into a valid physical address. It may
also mean we fail to boot as the elf files load address could be 0.

Sponsored by: ABT Systems Ltd

3 days agoImprove comment to reflect recent changes.
trasz [Tue, 9 Feb 2016 09:38:05 +0000 (09:38 +0000)]
Improve comment to reflect recent changes.

MFC after: 1 month
Sponsored by: The FreeBSD Foundation

3 days agoUse a proper type for return value of postfix_to_mult(); the way it's
trasz [Tue, 9 Feb 2016 09:37:04 +0000 (09:37 +0000)]
Use a proper type for return value of postfix_to_mult(); the way it's
being used wouldn't work if the value returned didn't fit in intmax_t.

Submitted by: bde@
MFC after: 1 month
Sponsored by: The FreeBSD Foundation

3 days agoUse proper functions for parsing the numeric values.
trasz [Tue, 9 Feb 2016 09:32:10 +0000 (09:32 +0000)]
Use proper functions for parsing the numeric values.

Submitted by: bde@
MFC after: 1 month
Sponsored by: The FreeBSD Foundation

3 days agoIgnore invalid page descriptors in ARM64 pmap_mincore
wma [Tue, 9 Feb 2016 06:26:27 +0000 (06:26 +0000)]
Ignore invalid page descriptors in ARM64 pmap_mincore

    Prevent the function from null-pointer-dereference when unexisting
    mapping is being processed.

Obtained from:         Semihalf
Sponsored by:          Cavium
Approved by:           cognet (mentor)
Reviewed by:           zbb, cognet
Differential revision: https://reviews.freebsd.org/D5228

3 days agoTeach ofw_bus_parse_xref_list_alloc to be able to return the length of the parsed...
adrian [Tue, 9 Feb 2016 03:35:40 +0000 (03:35 +0000)]
Teach ofw_bus_parse_xref_list_alloc to be able to return the length of the parsed list.

Currently, there is no easy way to know in advance how many entries a list parsed by
ofw_bus_parse_xref_list_alloc() in sys/dev/ofw/ofw_bus_subr.c has.

This patch:

* teaches the existing function about handling idx == -1 and returning how big
  the set is; then renames it as _internal;
* create a new function that asserts idx != -1, so the old API is maintained;
* add a new function that returns just the list length.

Submitted by: Stanislav Galabov <sgalabov@gmail.com>
Differential Revision: https://reviews.freebsd.org/D5043

3 days agoAdjust MAC algo to support interface branching
dteske [Tue, 9 Feb 2016 01:41:03 +0000 (01:41 +0000)]
Adjust MAC algo to support interface branching