24 hours agoMake nfsstat(1) build with WARNS=6. master github/master
trasz [Fri, 24 Apr 2015 07:57:59 +0000 (07:57 +0000)]
Make nfsstat(1) build with WARNS=6.

Sponsored by: The FreeBSD Foundation

25 hours agoAdd Lynx-Point LP smbus controller ID.
takawata [Fri, 24 Apr 2015 07:24:31 +0000 (07:24 +0000)]
Add Lynx-Point LP smbus controller ID.

27 hours agoPlural fix.
kevlo [Fri, 24 Apr 2015 05:00:57 +0000 (05:00 +0000)]
Plural fix.

27 hours ago- Fix the length of efuse content.
kevlo [Fri, 24 Apr 2015 04:57:18 +0000 (04:57 +0000)]
- Fix the length of efuse content.
- Replace the magic numbers with something more readable.

29 hours agoFix DTrace's panic() action.
markj [Fri, 24 Apr 2015 03:19:30 +0000 (03:19 +0000)]
Fix DTrace's panic() action.

It would previously call into some unfinished Solaris compatibility code and
return without actually calling panic(9). The compatibility code is
unneeded, however, so just remove it and have dtrace_panic() call vpanic(9)

Differential Revision:
Reviewed by: avg
MFC after: 2 weeks
Sponsored by: EMC / Isilon Storage Division

29 hours agoMake vpanic() externally visible so that it can be called as part of the
markj [Fri, 24 Apr 2015 03:17:21 +0000 (03:17 +0000)]
Make vpanic() externally visible so that it can be called as part of the
DTrace panic() action.

Differential Revision:
Reviewed by: avg
MFC after: 2 weeks
Sponsored by: EMC / Isilon Storage Division

34 hours agoAdd the SOC_IMX51, SOC_IMX53, and SOC_IMX6 options. These are used to
andrew [Thu, 23 Apr 2015 22:20:43 +0000 (22:20 +0000)]
Add the SOC_IMX51, SOC_IMX53, and SOC_IMX6 options. These are used to
select which SoCs the kernel config will support.

Use these options to merge files.imx51 and files.imx53.

35 hours agoReduce the diff between files.imx51 and files.imx53
andrew [Thu, 23 Apr 2015 21:39:32 +0000 (21:39 +0000)]
Reduce the diff between files.imx51 and files.imx53

35 hours agoPull out the common parts of the ODROIDC1 and VSATV102 kernels to a new
andrew [Thu, 23 Apr 2015 21:31:06 +0000 (21:31 +0000)]
Pull out the common parts of the ODROIDC1 and VSATV102 kernels to a new
kernel config. This also removes the per-board config files from universe.

36 hours agoRemove the need for board specific std files on aml8726 by moving the
andrew [Thu, 23 Apr 2015 20:53:29 +0000 (20:53 +0000)]
Remove the need for board specific std files on aml8726 by moving the
options they set to the main kernel config.

36 hours agoMove KERNVIRTADDR to the common std.aml8726.
andrew [Thu, 23 Apr 2015 20:43:24 +0000 (20:43 +0000)]
Move KERNVIRTADDR to the common std.aml8726.

36 hours ago - Move files selected in files.smp to files.aml8726 and remove it.
andrew [Thu, 23 Apr 2015 20:39:40 +0000 (20:39 +0000)]
 - Move files selected in files.smp to files.aml8726 and remove it.
 - Remove a redundant file from files.aml8726

36 hours agoUpdate the ODROIDC1 and VSATV102 to reduce the diff to the other armv6
andrew [Thu, 23 Apr 2015 20:28:56 +0000 (20:28 +0000)]
Update the ODROIDC1 and VSATV102 to reduce the diff to the other armv6
kernel config files.

36 hours agoRevert r281889, it broke keyboard input
emaste [Thu, 23 Apr 2015 20:26:22 +0000 (20:26 +0000)]
Revert r281889, it broke keyboard input

37 hours agoNow that we have an U-Boot for RPI2, do not use the static DTB compiled
loos [Thu, 23 Apr 2015 19:50:22 +0000 (19:50 +0000)]
Now that we have an U-Boot for RPI2, do not use the static DTB compiled
into the kernel, which is used mostly on early development stages.

On RPI(2) the DTB is loaded and modified by firmware and then handed to
kernel via U-Boot and ubldr.

The RPI firmware adds (or modify) a few valuable data to the in memory
DTB, like:

 - System memory;
 - Ethernet MAC address;
 - framebuffer settings;
 - Board serial and revision;
 - clock-frequency for most of devices.

38 hours agoFix the gpio-leds node to match the LEDs on RPI2.
loos [Thu, 23 Apr 2015 18:36:36 +0000 (18:36 +0000)]
Fix the gpio-leds node to match the LEDs on RPI2.

The 'pwr' LED is connected to GPIO pin 35 (active high) and the 'act' LED
is connected to GPIO pin 47 (active high).

Remove the GPIO pin 47 from reserved pins list, previously it was used to
signaling the SD Card presence and now it is used by act LED.

41 hours agoEnable the watchdog on RPI2.
loos [Thu, 23 Apr 2015 14:58:39 +0000 (14:58 +0000)]
Enable the watchdog on RPI2.

The watchdog is used to reset the SoC on cpu_reset().

Without this, RPI2 will not reboot when asked to.

42 hours agoPlug memory leaks in kbdmux(4)
pfg [Thu, 23 Apr 2015 14:53:45 +0000 (14:53 +0000)]
Plug memory leaks in kbdmux(4)

Al kudos here for the Clang static analyzer which, unlike Coverity,
failed to flag a false positive.

Found by: clang static analyzer
CID: 1007072
CID: 1007073
CID: 1007074

MFC after: 1 week

42 hours agoRemove oldnfs support from nfsstat(1).
trasz [Thu, 23 Apr 2015 14:36:01 +0000 (14:36 +0000)]
Remove oldnfs support from nfsstat(1).

Reviewed by: rmacklem@ (earlier version)
Sponsored by: The FreeBSD Foundation

42 hours agoReassign copyright statements on several files from Advanced
jhb [Thu, 23 Apr 2015 14:22:20 +0000 (14:22 +0000)]
Reassign copyright statements on several files from Advanced
Computing Technologies LLC to Hudson River Trading LLC.

Approved by: Hudson River Trading LLC (who owns ACT LLC)
MFC after: 1 week

42 hours agoReduce diffs between the three copies of libstand's Makefile
emaste [Thu, 23 Apr 2015 14:04:33 +0000 (14:04 +0000)]
Reduce diffs between the three copies of libstand's Makefile

This should be a non-functional change.  A future change should
address the functional differences between these three and converge
on a single source.

Differential Revision:
Reviewed by: jhb
Sponsored by: The FreeBSD Foundation

43 hours agoDon't propagate SIOCSIFCAPS from a vlan(4) to its parent. This leads to
glebius [Thu, 23 Apr 2015 13:19:00 +0000 (13:19 +0000)]
Don't propagate SIOCSIFCAPS from a vlan(4) to its parent. This leads to
quite unexpected result of toggling capabilities on the neighbour vlan(4)

Reviewed by: melifaro, np
Differential Revision:
Sponsored by: Nginx, Inc.

43 hours agoA complete user message is signalled with the MSG_EOR flag, not the MSG_EOF
tuexen [Thu, 23 Apr 2015 13:10:03 +0000 (13:10 +0000)]
A complete user message is signalled with the MSG_EOR flag, not the MSG_EOF
Thanks to Valentin Nechayev for reporting the issue.

MFC after: 1 week

45 hours agoHandle incorrect ELF images specifying size for PT_GNU_STACK not being
kib [Thu, 23 Apr 2015 11:27:21 +0000 (11:27 +0000)]
Handle incorrect ELF images specifying size for PT_GNU_STACK not being
multiple of page size.

Sponsored by: The FreeBSD Foundation
MFC after: 3 days

2 days agoSimplify linux_getcwd(), removing code that was longer used.
trasz [Thu, 23 Apr 2015 08:41:50 +0000 (08:41 +0000)]
Simplify linux_getcwd(), removing code that was longer used.

Differential Revision:
Reviewed by: dchagin@, kib@
MFC after: 1 month
Sponsored by: The FreeBSD Foundation

2 days agoDisable multi process interrupts, because the current code doesn't use
hselasky [Thu, 23 Apr 2015 07:41:58 +0000 (07:41 +0000)]
Disable multi process interrupts, because the current code doesn't use
them. Else we can end up in an infinite interrupt loop in USB device

MFC after: 3 days

2 days agoMissing break in switch case.
araujo [Thu, 23 Apr 2015 02:50:06 +0000 (02:50 +0000)]
Missing break in switch case.

Differential Revision: D2342
Reviewed by: neel

2 days agoAdd another variant of BCM5708S controller to IBM HS21 workaround
yongari [Thu, 23 Apr 2015 01:39:28 +0000 (01:39 +0000)]
Add another variant of BCM5708S controller to IBM HS21 workaround

PR: 118238
MFC after: 2 weeks

2 days agoCreate the arm64/aarch64 VM disk image as MBR instead of
gjb [Thu, 23 Apr 2015 01:04:14 +0000 (01:04 +0000)]
Create the arm64/aarch64 VM disk image as MBR instead of
GPT scheme.  UEFI needs to know the unique partition GUID
with GPT, which changes each time.  Specifically, the QEMU
EFI BIOS file has this hard-coded.[1]

Since the GPT labels are now unavailable, unconditionally
label the root filesystem as 'rootfs' with newfs(8), since
it does not hurt anything anywhere else.  For the arm64 case,
'/' is mounted from /dev/ufs/rootfs; for all other VM images,
'/' is mounted from /dev/gpt/rootfs.

Unfortunately, since the /dev/gpt/swapfs label is also lost,
set NOSWAP=1 for the arm64/aarch64 images.  This is temporary,
until I figure out a scalable solution to this.  But, a certain
piece of softare was written "very fast", and ended up living
for 15 years.  We can deal with this for a week or so.

Information from: andrew, emaste [1]
Sponsored by: The FreeBSD Foundation

2 days agolibgeom: plug memory leak in EndElement.
pfg [Wed, 22 Apr 2015 22:23:56 +0000 (22:23 +0000)]
libgeom: plug memory leak in EndElement.

CID: 1016696
Found by: Clang static checker
MFC after: 1 week

2 days agoUpdate the pci_cfg_save/restore routines to operate on bridge devices
jhb [Wed, 22 Apr 2015 22:02:27 +0000 (22:02 +0000)]
Update the pci_cfg_save/restore routines to operate on bridge devices
(type 1 and type 2) as well as leaf devices (type 0).  In particular,
this allows the existing PCI bus logic to save and restore capability
registers such as MSI and PCI-express work for bridge devices rather than
requiring that code to be duplicated in bridge drivers.  It also means
that bridge drivers no longer need to save and restore basic registers
such as the PCI command register or BARs nor manage powerstates for the
bridge device.

While here, pci_setup_secbus() has been changed to initialize the 'sec'
and 'sub' fields in the 'secbus' structure instead of requiring the pcib
and pccbb drivers to do this in the NEW_PCIB + PCI_RES_BUS case.

Differential Revision:
Reviewed by: imp, jmg
MFC after: 2 weeks

2 days agoDon't explicitly manage power states for PCI-PCI bridge devices in the
jhb [Wed, 22 Apr 2015 21:56:44 +0000 (21:56 +0000)]
Don't explicitly manage power states for PCI-PCI bridge devices in the
driver's suspend and resume routines.  These have been redundant no-ops
since r214065 changed the PCI bus driver to manage power states for
all devices (including type 1/2 bridge devices) during suspend and resume.

2 days agoFix some incorrect #if conditions around older workarounds for bus
jhb [Wed, 22 Apr 2015 21:47:51 +0000 (21:47 +0000)]
Fix some incorrect #if conditions around older workarounds for bus
numbering goofs.

MFC after: 1 week

2 days agoThe minimim grant and maximum latency PCI config registers are only valid
jhb [Wed, 22 Apr 2015 21:41:59 +0000 (21:41 +0000)]
The minimim grant and maximum latency PCI config registers are only valid
for type 0 devices, not type 1 or 2 bridges.  Don't read them for bridge
devices during bus scans and return an error when attempting to read them
as ivars for bridge devices.

2 days agoCosmetic change: use PCIR_SECLAT_2 rather than PCIR_SECLAT_1.
jhb [Wed, 22 Apr 2015 21:38:21 +0000 (21:38 +0000)]
Cosmetic change: use PCIR_SECLAT_2 rather than PCIR_SECLAT_1.

2 days agoFetch the SDHCI frequency from videocore (our prefered source) and only if
loos [Wed, 22 Apr 2015 18:35:02 +0000 (18:35 +0000)]
Fetch the SDHCI frequency from videocore (our prefered source) and only if
it fails, fetch the clock-frequency from DTB.

If both methods fail, use the hardcoded default.

2 days agoMove zlib.c from net to libkern.
rodrigc [Wed, 22 Apr 2015 18:15:18 +0000 (18:15 +0000)]
Move zlib.c from net to libkern.

2 days agoAdd definition for the argument_with_type_tag attribute.
pfg [Wed, 22 Apr 2015 18:13:28 +0000 (18:13 +0000)]
Add definition for the argument_with_type_tag attribute.

This attribute originates in clang and brings support for checking types
of variadic functions' arguments for functions like fcntl() and ioctl().

Unfortunately lint(1) will complain about them: in particular as one of
the parameters is the function being tagged. For now define this attribute
in the lint-sensitive section.


2 days agoMake AIO to not allocate pbufs for unmapped I/O like r281825.
mav [Wed, 22 Apr 2015 18:11:34 +0000 (18:11 +0000)]
Make AIO to not allocate pbufs for unmapped I/O like r281825.

While there, make few more performance optimizations.

On 40-core system doing many 512-byte AIO reads from array of raw SSDs
this change removes lock congestions inside pbuf allocator and devfs,
and bottleneck on single AIO completion taskqueue thread.  It improves
peak AIO performance from ~600K to ~1.3M IOPS.

MFC after: 2 weeks

2 days agoFix the SDHCI clock frequency for RPI2, the default clock (for the newer
loos [Wed, 22 Apr 2015 17:35:58 +0000 (17:35 +0000)]
Fix the SDHCI clock frequency for RPI2, the default clock (for the newer
firmware versions) is 250MHz.

2 days agocomputematchjumps(): fix allocator sizeof operand mismatch.
pfg [Wed, 22 Apr 2015 17:09:02 +0000 (17:09 +0000)]
computematchjumps(): fix allocator sizeof operand mismatch.

Mostly cosmetical warning.

Found by: Clang static analyzer

2 days ago_pthread_cleanup_push: fix allocator sizeof operand mismatch
pfg [Wed, 22 Apr 2015 16:51:21 +0000 (16:51 +0000)]
_pthread_cleanup_push: fix allocator sizeof operand mismatch

Same fix appears to be in DragonFly's libthread_xu.

Found by: Clang Static Analyzer
MFC after: 1 week

2 days agoMove zlib.c from net to libkern.
rodrigc [Wed, 22 Apr 2015 14:38:58 +0000 (14:38 +0000)]
Move zlib.c from net to libkern.

It is not network-specific code and would
be better as part of libkern instead.
Move zlib.h and zutil.h from net/ to sys/
Update includes to use sys/zlib.h and sys/zutil.h instead of net/

Submitted by: Steve Kiernan
Obtained from: Juniper Networks, Inc.
GitHub Pull Request:
Relnotes: yes

2 days agoRemove the gptboot partition from the arm64 memory stick image,
gjb [Wed, 22 Apr 2015 13:52:29 +0000 (13:52 +0000)]
Remove the gptboot partition from the arm64 memory stick image,
which is not needed/used.

Sponsored by: The FreeBSD Foundation

2 days agoMove some common code from sys/amd64/amd64/machdep.c and
kib [Wed, 22 Apr 2015 12:32:14 +0000 (12:32 +0000)]
Move some common code from sys/amd64/amd64/machdep.c and
sys/i386/i386/machdep.c to new file sys/x86/x86/cpu_machdep.c.  Most
of the code is related to the idle handling.

Discussed with: pluknet
Sponsored by: The FreeBSD Foundation

2 days agoAdd the null implementation of stack_save and stack_save_td.
andrew [Wed, 22 Apr 2015 12:24:38 +0000 (12:24 +0000)]
Add the null implementation of stack_save and stack_save_td.

Sponsored by: The FreeBSD Foundation

2 days agoRemove the calls to panic from pmap_object_init_pt and pmap_clear_modify.
andrew [Wed, 22 Apr 2015 09:52:51 +0000 (09:52 +0000)]
Remove the calls to panic from pmap_object_init_pt and pmap_clear_modify.

Sponsored by: The FreeBSD Foundation

3 days agoSupport file verification in MAC.
rodrigc [Wed, 22 Apr 2015 01:54:25 +0000 (01:54 +0000)]
Support file verification in MAC.

* Add VCREAT flag to indicate when a new file is being created
* Add VVERIFY to indicate verification is required
* Both VCREAT and VVERIFY are only passed on the MAC method vnode_check_open
  and are removed from the accmode after
* Add O_VERIFY flag to rtld open of objects
* Add 'v' flag to __sflags to set O_VERIFY flag.

Submitted by: Steve Kiernan <>
Obtained from: Juniper Networks, Inc.
GitHub Pull Request:
Relnotes: yes

3 days agoQuiet compiler warnings in mbuf.h
rodrigc [Wed, 22 Apr 2015 01:35:29 +0000 (01:35 +0000)]
Quiet compiler warnings in mbuf.h

 * Mark unused arguments with __unused
 * Move inline before the return type (and use __inline to keep the file
   consistent in style.)

Submitted by: Steve Kiernan <>
Obtained from: Juniper Networks, Inc.

3 days agoExtend DA_Q_NO_RC16 to MXUB3* devices.
delphij [Tue, 21 Apr 2015 22:55:52 +0000 (22:55 +0000)]
Extend DA_Q_NO_RC16 to MXUB3* devices.

PR: kern/198647
MFC after: 2 weeks

3 days agoImprove carp(4) locking:
glebius [Tue, 21 Apr 2015 20:25:12 +0000 (20:25 +0000)]
Improve carp(4) locking:
- Use the carp_sx to serialize not only CARP ioctls, but also carp_attach()
  and carp_detach().
- Use cif_mtx to lock only access to those the linked list.
- These locking changes allow us to do some memory allocations with M_WAITOK
  and also properly call callout_drain() in carp_destroy().
- In carp_attach() assert that ifaddr isn't attached. We always come here
  with a pristine address from in[6]_control().

Reviewed by: oleg
Sponsored by: Nginx, Inc.

3 days agoFor igb(4), when we are doing multiqueue, we are all setup to have full 32bit
hiren [Tue, 21 Apr 2015 20:24:15 +0000 (20:24 +0000)]
For igb(4), when we are doing multiqueue, we are all setup to have full 32bit
RSS hash from the card. We do not need to hide that under "ifdef RSS" and should
expose that by default so others like lagg(4) can use that and avoid hashing the
traffic by themselves.
While here, improve comments and get rid of hidden/unimplemented RSS support
code for UDP.

Differential Revision:
Reviewed by: jfv, erj
Discussed with: adrian
Sponsored by: Limelight Networks

3 days agoRevert r281809, which did more harm than good, and apply
gjb [Tue, 21 Apr 2015 14:48:38 +0000 (14:48 +0000)]
Revert r281809, which did more harm than good, and apply
a more proper fix when attempting to locate the /boot

Sponsored by: The FreeBSD Foundation

3 days agoModify kern___getcwd() to take max pathlen limit as an additional
trasz [Tue, 21 Apr 2015 13:55:24 +0000 (13:55 +0000)]
Modify kern___getcwd() to take max pathlen limit as an additional
argument.  This will be used for the Linux emulation layer - for Linux,
PATH_MAX is 4096 and not 1024.

Differential Revision:
Reviewed by: kib@
MFC after: 1 month
Sponsored by: The FreeBSD Foundation

3 days agoFix numerous issues in iic(4) and iicbus(4):
jah [Tue, 21 Apr 2015 11:50:31 +0000 (11:50 +0000)]
Fix numerous issues in iic(4) and iicbus(4):
--Allow multiple open iic fds by storing addressing state in cdevpriv
--Fix, as much as possible, the baked-in race conditions in the iic
ioctl interface by requesting bus ownership on I2CSTART, releasing it on
I2CSTOP/I2CRSTCARD, and requiring bus ownership by the current cdevpriv
to use the I/O ioctls
--Reduce internal iic buffer size and remove 1K read/write limit by
iteratively calling iicbus_read/iicbus_write
--Eliminate dynamic allocation in I2CWRITE/I2CREAD
--Move handling of I2CRDWR to separate function and improve error handling
--Add new I2CSADDR ioctl to store address in current cdevpriv so that
I2CSTART is not needed for read(2)/write(2) to work
--Redesign iicbus_request_bus() and iicbus_release_bus():
    --iicbus_request_bus() no longer falls through if the bus is already
owned by the requesting device.  Multiple threads on the same device may
want exclusive access.  Also, iicbus_release_bus() was never
device-recursive anyway.
    --Previously, if IICBUS_CALLBACK failed in iicbus_release_bus(), but
the following iicbus_poll() call succeeded, IICBUS_CALLBACK would not be
issued again
    --Do not hold iicbus mtx during IICBUS_CALLBACK call.  There are
several drivers that may sleep in IICBUS_CALLBACK, if IIC_WAIT is passed.
    --Do not loop in iicbus_request_bus if IICBUS_CALLBACK returns
EWOULDBLOCK; instead pass that to the caller so that it can retry if so

Differential Revision:
Reviewed by: imp, jhb, loos
Approved by: kib (mentor)

3 days agoRewrite physio() to not allocate pbufs for unmapped I/O.
mav [Tue, 21 Apr 2015 10:55:53 +0000 (10:55 +0000)]
Rewrite physio() to not allocate pbufs for unmapped I/O.

pbufs is a limited resource, and their allocator is not SMP-scalable.
So instead of always allocating pbuf to immediately convert it to bio,
allocate bio just here.  If buffer needs kernel mapping, then pbuf is
still allocated, but used only as a source of KVA and storage for a list
of held pages.

On 40-core system doing many 512-byte reads from user level to array of
raw SSDs this change removes huge lock congestion inside pbuf allocator.
It improves peak performance from ~300K to ~1.2M IOPS.  On my previous
24-core system this problem also existed, but was less serious.

Reviewed by: kib
MFC after: 2 weeks

3 days agoMake IFMEDIA_DEBUG a kernel option.
glebius [Tue, 21 Apr 2015 10:35:23 +0000 (10:35 +0000)]
Make IFMEDIA_DEBUG a kernel option.

Sponsored by: Nginx, Inc.

3 days agoSerialize all of _kerberos5_bootstrap_tools to avoid build failures involving
ngie [Tue, 21 Apr 2015 10:17:25 +0000 (10:17 +0000)]
Serialize all of _kerberos5_bootstrap_tools to avoid build failures involving
make bootstrap-tools

On the plus side, this also greatly reduces complexity

MFC after: 1 week
Pointyhat to: ngie
Reported by: Willem Jan Withagen <>

3 days agoInstead of storing mii_media_table array index in ifm_data, determine
glebius [Tue, 21 Apr 2015 09:39:48 +0000 (09:39 +0000)]
Instead of storing mii_media_table array index in ifm_data, determine
it in mii_phy_setmedia() functionally.

Sponsored by: Nginx, Inc.

3 days agoAdd hint about "volmode=dev" to ctl.conf(5).
trasz [Tue, 21 Apr 2015 09:11:28 +0000 (09:11 +0000)]
Add hint about "volmode=dev" to ctl.conf(5).

Differential Revision:
Reviewed by: allanjude@, bcr@
MFC after: 1 month
Sponsored by: The FreeBSD Foundation

4 days agoThe comment on BMCR data in if_media entry is wrong. The ifm_data stores
glebius [Tue, 21 Apr 2015 08:54:14 +0000 (08:54 +0000)]
The comment on BMCR data in if_media entry is wrong. The ifm_data stores
the index array, not a value for BMCR register. In case of IFM_10_T there
could be either MII_MEDIA_10_T or MII_MEDIA_10_T_FDX, which are 1 and 2,
accordingly. Neither matches a valid BMCR value. My guessing is that this
write is harmless, since later mii_phy_setmedia() would write a proper
value there.

The code is here since the initial checkin. Note that case IFM_100_TX has
the same comment, but a proper value of BMCR_ISO is written. So, collapse
two cases into one, always writing there BMCR_ISO.

Sponsored by: Nginx, Inc.

4 days agoSince xmphy doesn't call mii_phy_setmedia(), there is no reason to set ifm_data.
glebius [Tue, 21 Apr 2015 06:59:40 +0000 (06:59 +0000)]
Since xmphy doesn't call mii_phy_setmedia(), there is no reason to set ifm_data.

Sponsored by: Nginx, Inc.

4 days agoSince brgphy doesn't call mii_phy_setmedia(), there is no reason to
glebius [Tue, 21 Apr 2015 06:46:11 +0000 (06:46 +0000)]
Since brgphy doesn't call mii_phy_setmedia(), there is no reason to
set any value to ifm_data.  If brgphy ever to call mii_phy_setmedia(),
then the value of BRGPHY_S1000 | BRGPHY_BMCR_FDX will trigger KASSERT.

While here, remove the obfuscating macro and wrap long lines.

Sponsored by: Nginx, Inc.

4 days agoDon't print uninitialized variables.
hselasky [Tue, 21 Apr 2015 06:11:47 +0000 (06:11 +0000)]
Don't print uninitialized variables.

CID: 1295242

4 days agoTweak BSS symbol handling from r281811 to not consider them unresolved
bdrewery [Tue, 21 Apr 2015 05:41:56 +0000 (05:41 +0000)]
Tweak BSS symbol handling from r281811 to not consider them unresolved

4 days agoSupport libraries linked by path.
bdrewery [Tue, 21 Apr 2015 05:10:18 +0000 (05:10 +0000)]
Support libraries linked by path.

Sponsored by: EMC / Isilon Storage Division

4 days agoMerge bmake-20150418
sjg [Tue, 21 Apr 2015 04:40:38 +0000 (04:40 +0000)]
Merge bmake-20150418

PR: 199486

4 days ago- For executables search for matching (B) global uninitialized BSS symbols from
bdrewery [Tue, 21 Apr 2015 03:29:03 +0000 (03:29 +0000)]
- For executables search for matching (B) global uninitialized BSS symbols from
  linked libraries. Only do this for BSS symbols that have a size which avoids
  __bss_start. Without this some libraries would be considered unneeded even
  though they were providing a B symbol.
- Add in the symbols from crt1.o to cover a handful of common unresolved symbols.
- Consider (C) common data symbols as provided by libraries/crt1.
- Move libkey() function to more appropriate place.

Sponsored by: EMC / Isilon Storage Division

4 days agoRevert r281805 for now as it breaks due to spaces in output
bdrewery [Tue, 21 Apr 2015 03:06:58 +0000 (03:06 +0000)]
Revert r281805 for now as it breaks due to spaces in output

4 days agoSimplify variable expansion in attempt to fix the vm-image build.
gjb [Tue, 21 Apr 2015 02:49:44 +0000 (02:49 +0000)]
Simplify variable expansion in attempt to fix the vm-image build.

Sponsored by: The FreeBSD Foundation

4 days agoPass full path to ldd(1) so it works on files in cwd.
bdrewery [Tue, 21 Apr 2015 02:02:25 +0000 (02:02 +0000)]
Pass full path to ldd(1) so it works on files in cwd.

4 days agoFix wpa/hostapd build without OpenSSL.
rpaulo [Tue, 21 Apr 2015 02:00:37 +0000 (02:00 +0000)]
Fix wpa/hostapd build without OpenSSL.

4 days agoMerge wpa_supplicant/hostapd 2.4.
rpaulo [Tue, 21 Apr 2015 01:45:11 +0000 (01:45 +0000)]
Merge wpa_supplicant/hostapd 2.4.

Major changes are: SAE, Suite B, RFC 7268, EAP-PKE, ACS, and tons of
bug fixes.

Relnotes: yes

4 days agoSupport demangling C++ symbols with -D flag.
bdrewery [Tue, 21 Apr 2015 01:37:14 +0000 (01:37 +0000)]
Support demangling C++ symbols with -D flag.

Sponsored by: EMC / Isilon Storage Division

4 days agoMerge the following from ^/projects/release-arm64 to allow
gjb [Tue, 21 Apr 2015 00:48:35 +0000 (00:48 +0000)]
Merge the following from ^/projects/release-arm64 to allow
building FreeBSD/arm64 VM images and memstick.img installation
r281786, r281788, r281792:

  Add support for building arm64/aarch64 virtual machine images.

  Copy amd64/ to arm64/ for
  aarch64 memory stick images.

  Although arm64 does not yet have USB support, the memstick
  image should be bootable with certain virtualization tools,
  such as qemu.


  Add a buildenv_setup() prototype, intended to be overridden as

  For example, the arm64/aarch64 build needs devel/aarch64-binutils,
  so buildenv_setup() in the release.conf for this architecture
  handles the installation of the port before buildworld/buildkernel.

Sponsored by: The FreeBSD Foundation

4 days agopatch(1): small include changes.
pfg [Mon, 20 Apr 2015 22:15:18 +0000 (22:15 +0000)]
patch(1): small include changes.

Mostly to match OpenBSD, no functional change.

Obtained from: OpenBSD

4 days agoBuild and install the DTB for all the supported AllWinner SoCs.
loos [Mon, 20 Apr 2015 22:12:19 +0000 (22:12 +0000)]
Build and install the DTB for all the supported AllWinner SoCs.

We will need them when we start booting using ubldr.

4 days agoFix improbable memory leak in _citrus_prop_read_str().
pfg [Mon, 20 Apr 2015 22:09:50 +0000 (22:09 +0000)]
Fix improbable memory leak in _citrus_prop_read_str().

Found by: Clang Static Analyzer

4 days agoMove the definition of struct bpf_if to bpf.c.
markj [Mon, 20 Apr 2015 22:08:11 +0000 (22:08 +0000)]
Move the definition of struct bpf_if to bpf.c.

A couple of fields are still exposed via struct bpf_if_ext so that
bpf_peers_present() can be inlined into its callers. However, this change
eliminates some type duplication in the resulting CTF container, since
otherwise ctfmerge(1) propagates the duplication through all types that
contain a struct bpf_if.

Differential Revision:
Reviewed by: melifaro, rpaulo

4 days agoscsi_parse_transportid_rdma(): fix mismatch in memoty access size.
pfg [Mon, 20 Apr 2015 21:44:55 +0000 (21:44 +0000)]
scsi_parse_transportid_rdma(): fix mismatch in memoty access size.

Independently found by Coverity and gcc49.

CID: 1230006
Reviewed by: ken
MFC after: 5 days

4 days agoMinor mdoc fixes.
joel [Mon, 20 Apr 2015 21:19:26 +0000 (21:19 +0000)]
Minor mdoc fixes.

4 days ago- Speedup significantly by not using subshells for data already fetched.
bdrewery [Mon, 20 Apr 2015 20:51:19 +0000 (20:51 +0000)]
- Speedup significantly by not using subshells for data already fetched.
  Ran against /usr/local/sbin/pkg:
   Before:  25.12 real        12.41 user        33.14 sys
   After:   0.53 real         0.49 user         0.13 sys
- Exit with 1 if any missing or unresolved symbol is detected.
- Add option '-U' to skip looking up unresolved symbols.
- Don't consider provided weak objects as unresolved (nm V).

Sponsored by: EMC / Isilon Storage Division

4 days agoFix indentation to use tabs
bdrewery [Mon, 20 Apr 2015 20:45:10 +0000 (20:45 +0000)]
Fix indentation to use tabs

4 days agophabricator related changes:
eadler [Mon, 20 Apr 2015 20:33:22 +0000 (20:33 +0000)]
phabricator related changes:
- don't lint either contrib or crypto: these are both externally written
- add additional linters for spelling (check common typos like teh ->
- chmod linter checks for executible bit on bad files
- merge-conflict checks for merge conflict tokens then may have been
  resolved incorrectly
- filename checks for back characters in filenames
- json for json syntax correctness

- remove history.immutable: it is meaningless on subversion, and causes
  workflow problems when trying to use git.  It it set to 'true' by
  default with hg

4 days agoCopy amd64/ to arm64/ for
gjb [Mon, 20 Apr 2015 20:13:02 +0000 (20:13 +0000)]
Copy amd64/ to arm64/ for
aarch64 memory stick images.

Although arm64 does not yet have USB support, the memstick
image should be bootable with certain virtualization tools,
such as qemu.

Sponsored by: The FreeBSD Foundation

4 days agodmesg: accommodate message buffer growth between the sysctl calls
vangyzen [Mon, 20 Apr 2015 20:07:39 +0000 (20:07 +0000)]
dmesg: accommodate message buffer growth between the sysctl calls

Allocate 12.5% extra space to avoid ENOMEM when the message buffer
is growing steadily.

Reported by: Steve Wahl <> (and tested)
Approved by: kib (mentor)
Obtained from: Dell Inc.
MFC after: 1 week

4 days agoAdd support for building arm64/aarch64 virtual machine images.
gjb [Mon, 20 Apr 2015 20:06:25 +0000 (20:06 +0000)]
Add support for building arm64/aarch64 virtual machine images.

Sponsored by: The FreeBSD Foundation

4 days agoAlways send log(9) messages to the message buffer.
vangyzen [Mon, 20 Apr 2015 20:03:26 +0000 (20:03 +0000)]
Always send log(9) messages to the message buffer.

It is truer to the semantics of logging for messages to *always*
go to the message buffer, where they can eventually be collected
and, in fact, be put into a log file.

This restores the behavior prior to r70239, which seems to have
changed it inadvertently.

Submitted by: Eric Badger <>
Reviewed by: jhb
Approved by: kib (mentor)
Obtained from: Dell Inc.
MFC after: 1 week

4 days agoMFH: r281668 through r281783
gjb [Mon, 20 Apr 2015 20:03:07 +0000 (20:03 +0000)]
MFH: r281668 through r281783

Sponsored by: The FreeBSD Foundation

4 days agoWhen building VM disk images, vm_copy_base() uses tar(1) to
gjb [Mon, 20 Apr 2015 19:54:54 +0000 (19:54 +0000)]
When building VM disk images, vm_copy_base() uses tar(1) to
copy the userland from one md(4)-mounted filesystem to a clean
filesystem to prevent remnants of files that were added and
removed from resulting in an unclean filesystem.  When newfs(8)
creates the first filesystem with journaled soft-updates enabled,
the /.sujournal file in the new filesystem cannot be overwritten
by the /.sujournal in the original filesystem.

To avoid this particular error case, do not enable journaled
soft-updates when creating the md(4)-backed filesystems, and
instead use tunefs(8) to enable journaled soft-updates after
the new filesystem is populated in vm_copy_base().

While here, fix a long standing bug where the build environment
/boot files were used by mkimg(1) when creating the VM disk
images by using the files in .OBJDIR.

MFC after: 3 days
Sponsored by: The FreeBSD Foundation

4 days agoDon't set bits that should be zero for SATA devices.
mav [Mon, 20 Apr 2015 19:11:27 +0000 (19:11 +0000)]
Don't set bits that should be zero for SATA devices.

Old value made Linux think that it is PATA device with SATA bridge.

MFC after: 2 weeks

4 days agocrunchide: Remove local EM_* ELF definitions provided by system ELF headers
emaste [Mon, 20 Apr 2015 18:30:06 +0000 (18:30 +0000)]
crunchide: Remove local EM_* ELF definitions provided by system ELF headers

Suggested by: imp (in review D2314)

4 days agoUpdate usage statement to align with post-r279624 functionality.
dteske [Mon, 20 Apr 2015 17:46:09 +0000 (17:46 +0000)]
Update usage statement to align with post-r279624 functionality.

MFC after: 3 days
X-MFC-to: stable/10 stable/9

4 days agovidcontrol: skip invalid video modes returned by vt(4)
emaste [Mon, 20 Apr 2015 17:43:55 +0000 (17:43 +0000)]
vidcontrol: skip invalid video modes returned by vt(4)

vt(4) has a stub CONS_MODEINFO ioctl that does not provide any data
but returns success. This needs to be fixed in the kernel, but address
it in vidcontrol(1) as well in case it's run on an older kernel.

Reviewed by: bde
Sponsored by: The FreeBSD Foundation

4 days agoAdd llvm patch corresponding to r281775.
dim [Mon, 20 Apr 2015 17:37:37 +0000 (17:37 +0000)]
Add llvm patch corresponding to r281775.

4 days agoPull in r229911 from upstream llvm trunk (by Benjamin Kramer):
dim [Mon, 20 Apr 2015 17:36:35 +0000 (17:36 +0000)]
Pull in r229911 from upstream llvm trunk (by Benjamin Kramer):

  MC: Allow multiple comma-separated expressions on the .uleb128 directive.

  For compatiblity with GNU as. Binutils documents this as
  '.uleb128 expressions'. Subtle, isn't it?

Reported by: sbruno
PR: 199554
MFC after: 3 days

4 days agoRemove unused variable.
araujo [Mon, 20 Apr 2015 17:30:13 +0000 (17:30 +0000)]
Remove unused variable.

Differential Revision: D2333
Reviewed by: royger

4 days agoAdd back ixgbe_rxeof, just remove the assignment to more.
araujo [Mon, 20 Apr 2015 17:24:39 +0000 (17:24 +0000)]
Add back ixgbe_rxeof, just remove the assignment to more.

4 days agoRemove unused variable.
araujo [Mon, 20 Apr 2015 17:21:15 +0000 (17:21 +0000)]
Remove unused variable.

Differential Revision: D2331
Reviewed by: erj

4 days agoEliminate an unused variable.
alc [Mon, 20 Apr 2015 16:48:21 +0000 (16:48 +0000)]
Eliminate an unused variable.

MFC after: 1 week