]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
11 years agoTeach siglongjmp about the VFP version of the setjmp buffer.
Andrew Turner [Sun, 21 Jul 2013 12:47:33 +0000 (12:47 +0000)]
Teach siglongjmp about the VFP version of the setjmp buffer.

This fixes a bug where a call to sigsetjmp followed by siglongjmp may fail
when run on a machine with VFP enabled.

11 years ago- Fix nflag in routename().
Hiroki Sato [Sun, 21 Jul 2013 11:59:41 +0000 (11:59 +0000)]
- Fix nflag in routename().
- Display a AF_LINK address in #linkN when sdl_{nlen,alen,slen) == 0 and
  sdl_index != 0.
- Reduce unnecessary loop in pmsg_addrs().
- Remove iso_ntoa().  This is not used.

11 years agomdoc nit
Dag-Erling Smørgrav [Sun, 21 Jul 2013 07:24:25 +0000 (07:24 +0000)]
mdoc nit

11 years agoUse the correct request syntax for proxied (tunneled) HTTPS requests.
Dag-Erling Smørgrav [Sun, 21 Jul 2013 06:59:56 +0000 (06:59 +0000)]
Use the correct request syntax for proxied (tunneled) HTTPS requests.

PR: bin/180666
MFC after: 3 days

11 years agoCreate memstick images with the '.img' filename extension.
Glen Barber [Sun, 21 Jul 2013 05:23:34 +0000 (05:23 +0000)]
Create memstick images with the '.img' filename extension.

Requested by: many
Approved by: kib (mentor)
MFC after: 3 days
X-MFC-To: stable/9 only

11 years agoAdd some initial board support for the AR934x and the Qualcomm Atheros
Adrian Chadd [Sun, 21 Jul 2013 04:00:48 +0000 (04:00 +0000)]
Add some initial board support for the AR934x and the Qualcomm Atheros
DB120 development board.

The AR934x SoCs are a MIPS74k based system with increased RAM addressing
space, some scratch-pad RAM, an improved gige switch PHY and 2x2 or 3x3
on-board dual-band wifi.

This support isn't complete by any stretch; it's just enough to bring
the board up for others to tinker with.  Notably, the MIPS74k support
is broken.  However it boots enough to echo some basic probe/attach
messages, before dying somewhere in the TLB code.

Thankyou to Qualcomm Atheros for their continued support of me doing
open source work with their hardware.

Tested:

* AR9344, mips74k

11 years agoImplement some initial AR934x support routines.
Adrian Chadd [Sun, 21 Jul 2013 03:56:57 +0000 (03:56 +0000)]
Implement some initial AR934x support routines.

This code reads the PLL configuration registers and correctly programs
things so the UART and such can come up.

There's MIPS74k platform issues that need fixing; but this at least brings
things up enough to echo stuff out the serial port and allow for interactive
debugging with ddb.

Tested:

* AR71xx SoCs
* AR933x SoC
* AR9344 board (DB120)

Obtained from: Qualcomm Atheros; Linux/OpenWRT

11 years agoTeach the GPIO code about the AR934x GPIO register and pin counts.
Adrian Chadd [Sun, 21 Jul 2013 03:55:18 +0000 (03:55 +0000)]
Teach the GPIO code about the AR934x GPIO register and pin counts.

11 years agoUse the UART frequency when programming the UART clock.
Adrian Chadd [Sun, 21 Jul 2013 03:54:39 +0000 (03:54 +0000)]
Use the UART frequency when programming the UART clock.

This allows the 16550 support to work correctly on the upcoming
AR934x support.

11 years agoInitialise the watchdog and UART frequencies.
Adrian Chadd [Sun, 21 Jul 2013 03:52:52 +0000 (03:52 +0000)]
Initialise the watchdog and UART frequencies.

For all pre-AR933x chips, the frequency is just the APB frequency.
For the AR933x, the UART frequency is different but we just hacked around
it.

For the AR934x, there's a different PLL setting for these, so they have
to be broken out.

11 years agoAdd two new CPU specific definitions - the watchdog clock frequency and
Adrian Chadd [Sun, 21 Jul 2013 03:51:24 +0000 (03:51 +0000)]
Add two new CPU specific definitions - the watchdog clock frequency and
the UART clock frequency.

The AR933x and AR934x have separate PLL settings for these.

11 years agoThe NFSv4 server incorrectly assumed that the high order words of
Rick Macklem [Sat, 20 Jul 2013 22:35:32 +0000 (22:35 +0000)]
The NFSv4 server incorrectly assumed that the high order words of
the attribute bitmap argument would be non-zero. This caused an
interoperability problem for a recent patch to the Linux NFSv4 client.
The Linux folks have changed their patch to avoid this, but this
patch fixes the problem on the server.

Reported and tested by: Andre Heider (a.heider@gmail.com)
MFC after: 3 days

11 years agoFix address range specification with ifconfig(8) options such as:
Hiroki Sato [Sat, 20 Jul 2013 16:58:17 +0000 (16:58 +0000)]
Fix address range specification with ifconfig(8) options such as:

- inet 192.0.2.1-10 netmask 255.255.255.0 (inet range spec + ifconfig options)
- inet6 2001:db8:1::1-f prefixlen 60 (inet6 range spec + ifconfig options)

If prefixlen or netmask option is specified with CIDR notation at
the same time, the option is used.

Tested by: Michael Grimm
MFC after: 3 days

11 years ago- Simplify getaddr() and print_getmsg() by using RTAX_* instead of RTA_*
Hiroki Sato [Sat, 20 Jul 2013 16:46:51 +0000 (16:46 +0000)]
- Simplify getaddr() and print_getmsg() by using RTAX_* instead of RTA_*
  as the argument.
- Reduce unnecessary loop in print_getmsg().

11 years agoShow "default" for the zero-filled address consistently when nflag == 0.
Hiroki Sato [Sat, 20 Jul 2013 15:58:43 +0000 (15:58 +0000)]
Show "default" for the zero-filled address consistently when nflag == 0.

11 years agoAdd cast to (void *) to the following cases to suppress warnings by
Hiroki Sato [Sat, 20 Jul 2013 15:23:42 +0000 (15:23 +0000)]
Add cast to (void *) to the following cases to suppress warnings by
-Wcast-align.  These do not increase the alignment requirement:

- rtm = (struct rt_msghdr *)(rtm + rtm->rtm_msglen)
- struct sockaddr *sa = &sa0; sX = (struct sockaddr_X *)sa

11 years agofuse: revert birthtime support.
Pedro F. Giffuni [Sat, 20 Jul 2013 14:50:35 +0000 (14:50 +0000)]
fuse: revert birthtime support.

The creation time support breaks the data structures used in linux
fuse.  libfuse carries it's own header.

Revert the changes for now. We will try to get an agreement with the
fuse  upstream maintainers to avoid having to patch the library
headers all the time.

11 years agoUse the right name for the vm_map_find(9) flag to specify shared
Konstantin Belousov [Sat, 20 Jul 2013 13:52:40 +0000 (13:52 +0000)]
Use the right name for the vm_map_find(9) flag to specify shared
mapping.  The MAP_SHARED and MAP_INHERIT_SHARE values are identical.

Noted by: alc
MFC after: 3 days

11 years agoRegenerate
Konstantin Belousov [Sat, 20 Jul 2013 13:40:03 +0000 (13:40 +0000)]
Regenerate

11 years agoid_t is 64bit, provide the compat32 wrapper for clock_getcpuclockid2(2).
Konstantin Belousov [Sat, 20 Jul 2013 13:39:41 +0000 (13:39 +0000)]
id_t is 64bit, provide the compat32 wrapper for clock_getcpuclockid2(2).

Reported and tested by: Petr Salinger <Petr.Salinger@seznam.cz>
PR: threads/180652
Sponsored by: The FreeBSD Foundation

11 years agoAllow the code to be compiled without warnings for any combination
Michael Tuexen [Sat, 20 Jul 2013 13:14:59 +0000 (13:14 +0000)]
Allow the code to be compiled without warnings for any combination
of INET, INET6 and SCTP_DEBUG defines.
The issue was reported by Lally Singh.

MFC after: 2 weeks

11 years agoStart adding support to build bits of our code using the Thumb-2
Andrew Turner [Sat, 20 Jul 2013 09:24:48 +0000 (09:24 +0000)]
Start adding support to build bits of our code using the Thumb-2
instruction set. Thumb-2 requires an if-then instruction to implement
conditional codes.

When building for ARM mode the it-then instructions do not generate any
assembled instruction as per the ARMv7-A Architecture Reference Manual, and
are safe to use.

While this allows the atomic instructions to be built, it doesn't mean we
fully support Thumb code. It works in small tests, but is still known to
fail in a large number of places.

While here add a check for the armv6t2 architecture.

11 years agoAdd support for upcoming BCM5725 (ASIC 5762) controller. This is a
Pyun YongHyeon [Sat, 20 Jul 2013 07:58:03 +0000 (07:58 +0000)]
Add support for upcoming BCM5725 (ASIC 5762) controller.  This is a
new 1Gb server controller chip that will be going into production
soon.
BCM5725 combines MAC with triple-speed PHY, a Network Controller
Sideband Interface (NC-SI) and on-chip memory buffer in a single
device.  BCM5725 has an Application Processing Engine (APE) that is
capable of on-chip management and offloading features.  BCM5725
supports high-precision clock, time stamp registers for
receive/transmit packets and programmable trigger inputs and
watchdog timeouts. These new features are not yet supported by
bge(4).
Many thanks to Broadcom for continuing to support FreeBSD!

Submitted by: Geans Pin geanspin@Broacom (initial version)
Reviewed by: Geans Pin geanspin@Broacom
H/W donated by: Broadcom

11 years agoRecognize BCM5725C PHY.
Pyun YongHyeon [Sat, 20 Jul 2013 07:24:01 +0000 (07:24 +0000)]
Recognize BCM5725C PHY.

11 years agoSetup the PCIE Fast Training Sequence (FTS) value to prevent
Pyun YongHyeon [Sat, 20 Jul 2013 07:09:50 +0000 (07:09 +0000)]
Setup the PCIE Fast Training Sequence (FTS) value to prevent
transmit hangs for 57766 and non Ax versions of 57765.
While here, correct definition of BGE_CHIPREV_57765_AX.

11 years agoAdjust outsizes:
Pedro F. Giffuni [Sat, 20 Jul 2013 03:55:56 +0000 (03:55 +0000)]
Adjust outsizes:

Recalculate FUSE_COMPAT_ENTRY_OUT_SIZE and COMPAT_ATTR_OUT_SIZE.
These were wrong in the previous commit. They are actually unused
in FreeBSD though.

Pointed out by: Jan Beich

11 years agoAdjust outsizes:
Pedro F. Giffuni [Sat, 20 Jul 2013 03:08:50 +0000 (03:08 +0000)]
Adjust outsizes:

When birthtime was added (r253331) we missed adding the weight
of the new fields in FUSE_COMPAT_ENTRY_OUT_SIZE and
COMPAT_ATTR_OUT_SIZE. Adjust them accordingly.

Pointed out by: Jan Beich

11 years agoThere's nothing to free if the unit wasn't allocated.
Navdeep Parhar [Fri, 19 Jul 2013 21:54:48 +0000 (21:54 +0000)]
There's nothing to free if the unit wasn't allocated.

11 years agoAdd message when nvd disks are attached and detached.
Jim Harris [Fri, 19 Jul 2013 21:40:57 +0000 (21:40 +0000)]
Add message when nvd disks are attached and detached.

As part of this commit, add an nvme_strvis() function which borrows
heavily from cam_strvis().  This will allow stripping of
leading/trailing whitespace and also handle unprintable characters
in model/serial numbers.  This function goes into a new nvme_util.c
file which is used by both the driver and nvmecontrol.

Sponsored by: Intel
Reviewed by: carl
MFC after: 3 days

11 years agoAdd new Coleto Creek device support: SATA, SMBus, and Watchdog devices.
Jack F Vogel [Fri, 19 Jul 2013 21:37:40 +0000 (21:37 +0000)]
Add new Coleto Creek device support: SATA, SMBus, and Watchdog devices.

MFC after: 1 week

11 years agoFix nvme(4) and nvd(4) to support non 512-byte sector sizes.
Jim Harris [Fri, 19 Jul 2013 21:33:24 +0000 (21:33 +0000)]
Fix nvme(4) and nvd(4) to support non 512-byte sector sizes.

Recent testing with QEMU that has variable sector size support for
NVMe uncovered some of these issues.  Chatham prototype boards supported
only 512 byte sectors.

Sponsored by: Intel
Reviewed by: carl
MFC after: 3 days

11 years agoDo not call disk_create() until we have completed all initialization of our
Jim Harris [Fri, 19 Jul 2013 21:30:53 +0000 (21:30 +0000)]
Do not call disk_create() until we have completed all initialization of our
internal disk structure.

Sponsored by: Intel
Reviewed by: carl
MFC after: 3 days

11 years agoGet the code compiling without INET and INET6 being defined.
Michael Tuexen [Fri, 19 Jul 2013 21:16:59 +0000 (21:16 +0000)]
Get the code compiling without INET and INET6 being defined.
This is not possible in FreeBSD, but in the upstream code.

MFC after: 2 weeks

11 years agoBe more aggressive in using superpages in all mappings of objects:
John Baldwin [Fri, 19 Jul 2013 19:06:15 +0000 (19:06 +0000)]
Be more aggressive in using superpages in all mappings of objects:
- Add a new address space allocation method (VMFS_OPTIMAL_SPACE) for
  vm_map_find() that will try to alter the alignment of a mapping to match
  any existing superpage mappings of the object being mapped.  If no
  suitable address range is found with the necessary alignment,
  vm_map_find() will fall back to using the simple first-fit strategy
  (VMFS_ANY_SPACE).
- Change mmap() without MAP_FIXED, shmat(), and the GEM mapping ioctl to
  use VMFS_OPTIMAL_SPACE instead of VMFS_ANY_SPACE.

Reviewed by: alc (earlier version)
MFC after: 2 weeks

11 years agoSince there's no defined schema for the XML that's generated by the
Scott Long [Fri, 19 Jul 2013 06:42:15 +0000 (06:42 +0000)]
Since there's no defined schema for the XML that's generated by the
kern.geom.confxml sysctl, it's silly to warn about tags in libgeom,
especially since libgeom exists simply to build a tree out of the conf
information.

Obtained from: Netflix
MFC after: 3 days

11 years agoCorrect ELF note decoding for non-core files
Ed Maste [Fri, 19 Jul 2013 01:04:17 +0000 (01:04 +0000)]
Correct ELF note decoding for non-core files

Requested by: kib

11 years agoOverhaul error, information, and debug logging.
Scott Long [Fri, 19 Jul 2013 00:12:41 +0000 (00:12 +0000)]
Overhaul error, information, and debug logging.

Obtained from: Netflix
MFC after: 3 days

11 years agoUse _PATH_DEV (from paths.h) for the "/dev/" string, rather than
Jim Harris [Thu, 18 Jul 2013 23:25:57 +0000 (23:25 +0000)]
Use _PATH_DEV (from paths.h) for the "/dev/" string, rather than
hard-coding it.

Sponsored by: Intel
Suggested by: kib
Reviewed by: kib, carl
MFC after: 3 days

11 years agoSimplify open_dev() by returning errno values rather than just 0 or 1.
Jim Harris [Thu, 18 Jul 2013 23:12:44 +0000 (23:12 +0000)]
Simplify open_dev() by returning errno values rather than just 0 or 1.

Also remove stat() call and just rely on errno from open() call to discern
whether dev node exists or not.

Sponsored by: Intel
Reviewed by: kib, carl
MFC after: 3 days

11 years agoClose uniq(1) in the capability mode sandbox and limit descriptors using
Pawel Jakub Dawidek [Thu, 18 Jul 2013 22:11:27 +0000 (22:11 +0000)]
Close uniq(1) in the capability mode sandbox and limit descriptors using
capability rights.

11 years ago- Make localtime(3) to work in sandbox.
Pawel Jakub Dawidek [Thu, 18 Jul 2013 21:56:10 +0000 (21:56 +0000)]
- Make localtime(3) to work in sandbox.
- Move strerror(3) initialization to its own function.

11 years agoDecode PCIe ASPM capability and status.
Jung-uk Kim [Thu, 18 Jul 2013 20:59:58 +0000 (20:59 +0000)]
Decode PCIe ASPM capability and status.

11 years agoCorrect alignment in note output
Ed Maste [Thu, 18 Jul 2013 20:36:21 +0000 (20:36 +0000)]
Correct alignment in note output

11 years agoRemove accidentally cut-and-pasted colons
Ed Maste [Thu, 18 Jul 2013 19:17:58 +0000 (19:17 +0000)]
Remove accidentally cut-and-pasted colons

11 years agoSanity-check the vm exitcode, and exit the process if it's out-of-bounds
Peter Grehan [Thu, 18 Jul 2013 18:40:54 +0000 (18:40 +0000)]
Sanity-check the vm exitcode, and exit the process if it's out-of-bounds
or there is no registered handler.

Submitted by: Bela Lubkin   bela dot lubkin at tidalscale dot com

11 years agoHandle FreeBSD-specific ELF notes
Ed Maste [Thu, 18 Jul 2013 17:25:50 +0000 (17:25 +0000)]
Handle FreeBSD-specific ELF notes

Add a function to return the specific type, when the note's Name field is
'FreeBSD'.

r249558 added FreeBSD-specific ELF note types that reuse type numbers of
existing generic / Linux types.  This caused 'readelf -n' to produce
incorrect output on FreeBSD core files.

Sponsored by: DARPA, AFRL
MFC after: 3 days

11 years agoProperly handle I/O windows in bridges with the ISA enable bit set. These
John Baldwin [Thu, 18 Jul 2013 15:17:11 +0000 (15:17 +0000)]
Properly handle I/O windows in bridges with the ISA enable bit set.  These
beasts still exist unfortunately.  More details can be found in other
references, but the short version is that bridges with this bit set ignore
I/O port ranges that alias to valid ISA I/O port ranges.  In the driver
this requires not allocating these alias regions from the parent device
(so they are free to be acquired by ISA devices), and ensuring no child
devices use resources from these alias regions.
- Change the pcib_window structure to allow for an array of backing
  resources rather than a single resource and update the existing code
  to cope with this.  Some of the coping requires using the saved
  base and limit values in pcib_window instead of using rman operations
  on the backing resource.
- Add special handling for allocating and adjusting the I/O port window
  of an ISA-enabled bridge to only allocate the non-alias ranges and
  add those to the associated resource manager.
- Reject I/O port allocations for a fixed request that conflicts with an
  ISA alias range.
- Remove the "no prefected decode" verbose printf during boot.  The absence
  of a "prefetched decode" line is sufficient.
- Replace the "subtractively decoded bridge" verbose printf with a single
  printf that lists all the "special" decoding modes of a bridge: ISA,
  subtractive, and VGA.
- Add a custom bus_release_resource() method to the PCI bus driver so that
  it can properly free resources for I/O windows of PCI-PCI bridges.
  (These resources are not stored in the bridge device's resource list.)

PR: misc/179033
MFC after: 2 weeks

11 years agoRework the previous fix for the IB vs Ethernet sysctl handler to be more
John Baldwin [Thu, 18 Jul 2013 14:06:01 +0000 (14:06 +0000)]
Rework the previous fix for the IB vs Ethernet sysctl handler to be more
generic and apply to all sysfs attributes:
- Use sysctl_handle_string() instead of reimplementing it.
- Remove trailing newline from the current value before passing it to
  userland and append a newline to the new string value before passing it
  to the attribute's store function.
- Don't leak the temporary buffer if the first error check triggers.
- Revert earlier change to mlx4 port mode handler.

PR: kern/174213
Submitted by: Garrett Cooper
Reviewed by: Shakar Klein @ Mellanox
MFC after: 1 week

11 years agoInclude stdint.h when use intmax_t.
Kevin Lo [Thu, 18 Jul 2013 06:32:19 +0000 (06:32 +0000)]
Include stdint.h when use intmax_t.

Reported by: bde

11 years agoMinor mdoc fixes.
Joel Dahl [Thu, 18 Jul 2013 05:46:33 +0000 (05:46 +0000)]
Minor mdoc fixes.

11 years ago- Fix a bug in ipv6_prefix_IF. It did not work with the 64-bit prefix
Hiroki Sato [Thu, 18 Jul 2013 02:58:24 +0000 (02:58 +0000)]
- Fix a bug in ipv6_prefix_IF.  It did not work with the 64-bit prefix
  notation like 2001:db8:1:1.

- Use eui64 flag in ifconfig(8) instead of network6_getladdr()[*] for
  interface indentifier part.

Suggested by: ume [*]
MFC after: 3 days

11 years agoFix a gcc warning.
Hiroki Sato [Thu, 18 Jul 2013 02:34:25 +0000 (02:34 +0000)]
Fix a gcc warning.

Pointy hat to: hrs

11 years agoReplace PRId64 with "jd" in a printf call. Cast the corresponding value to
Kevin Lo [Thu, 18 Jul 2013 01:40:31 +0000 (01:40 +0000)]
Replace PRId64 with "jd" in a printf call. Cast the corresponding value to
intmax_t, because the original type is off_t.

Reported by: bde

11 years agoManually merge part of vendor import r238583 from Illumos.
Xin LI [Thu, 18 Jul 2013 00:22:42 +0000 (00:22 +0000)]
Manually merge part of vendor import r238583 from Illumos.

Illumos changeset: 13680:2bd022a765e2
Illumos ZFS issue:

    2671 zpool import should not fail if vdev ashift has increased

MFC after: 3 days

11 years agoMajor rework of the virtio code. Split out common parts, and modify
Peter Grehan [Wed, 17 Jul 2013 23:37:33 +0000 (23:37 +0000)]
Major rework of the virtio code. Split out common parts, and modify
the net/block devices accordingly.

Submitted by: Chris Torek   torek at torek dot net
Reviewed by: grehan

11 years agoReload font when syscons(4) is resuming without switching mode.
Jung-uk Kim [Wed, 17 Jul 2013 23:29:56 +0000 (23:29 +0000)]
Reload font when syscons(4) is resuming without switching mode.

Reported by: adrian (more than a year ago)
Prodded by: adrian (less than a month ago)

11 years agoUse pause() instead of DELAY() when polling for completion of admin
Jim Harris [Wed, 17 Jul 2013 23:26:56 +0000 (23:26 +0000)]
Use pause() instead of DELAY() when polling for completion of admin
commands during controller initialization.

DELAY() does not work here during config_intrhook context - we need to
explicitly relinquish the CPU for the admin command completion to
get processed.

Sponsored by: Intel
Reported by: Adam Brooks <adam.j.brooks@intel.com>
Reviewed by: carl
MFC after: 3 days

11 years agoDefine constants for the lengths of the serial number, model number
Jim Harris [Wed, 17 Jul 2013 23:23:38 +0000 (23:23 +0000)]
Define constants for the lengths of the serial number, model number
and firmware revision in the controller's identify structure.

Also modify consumers of these fields to ensure they only use the
specified number of bytes for their respective fields.

Sponsored by: Intel
Reviewed by: carl
MFC after: 3 days

11 years agoAlways initialize fd to 0 in open_dev().
Jim Harris [Wed, 17 Jul 2013 23:14:41 +0000 (23:14 +0000)]
Always initialize fd to 0 in open_dev().

Sponsored by: Intel
Reviewed by: carl
MFC after: 3 days

11 years agoAdd bus_dmamap_load_bio and bus_dmamap_load_ccb to bus_dma(9).
Jim Harris [Wed, 17 Jul 2013 19:41:16 +0000 (19:41 +0000)]
Add bus_dmamap_load_bio and bus_dmamap_load_ccb to bus_dma(9).

Sponsored by: Intel
Reviewed by: kib
MFC after: 3 days

11 years agoIn this GRN, Marcel Moolenaar overhauled the logic for mounting
Craig Rodrigues [Wed, 17 Jul 2013 19:32:07 +0000 (19:32 +0000)]
In this GRN, Marcel Moolenaar overhauled the logic for mounting
the root file system on bootup:

   |------------------------------------------------------------------------
   |r214006 | marcel | 2010-10-17 22:01:53 -0700 (Sun, 17 Oct 2010) | 20 lines
   |
   | Re-implement the root mount logic using a recursive approach, whereby each
   |root file system (starting with devfs and a synthesized configuration) can
   |contain directives for mounting another file system as root.
   |------------------------------------------------------------------------

This commit adds a mount.conf(8) man page which documents
the root mount logic.  mount.conf(8) also provides some examples
for the /.mount.conf file, which can be used to change the root mount behavior.

Reviewed by: marcel bjk

11 years agoUse NET_RT_DUMP.0.FIB leaf node instead of setting td_proc->p_fibnum.
Hiroki Sato [Wed, 17 Jul 2013 14:15:00 +0000 (14:15 +0000)]
Use NET_RT_DUMP.0.FIB leaf node instead of setting td_proc->p_fibnum.

11 years ago- Add support of MK_INET_SUPPORT=no.
Hiroki Sato [Wed, 17 Jul 2013 14:05:20 +0000 (14:05 +0000)]
- Add support of MK_INET_SUPPORT=no.
- Fix a bug in sodump() which prevented struct sockaddr_in6 from displaying.
- Fix a bug in in fiboptlist_csv() which could cause free() of uninitialized
  pointer.
- Style cleanups:
 . Add missing "static" keywords.
 . Use an array of struct sockaddr_storage instead of sockunion for rtmsg.
 . Use err() and errx() instead of pair of fprintf(stderr, "...") + exit(1).
 . Use nitems() macro.
 . Various style(9) fixes.

11 years agoDo not overwrite the new UTF-8 files with the old ISO one
Baptiste Daroussin [Wed, 17 Jul 2013 14:00:51 +0000 (14:00 +0000)]
Do not overwrite the new UTF-8 files with the old ISO one

11 years agoSimplify keywords.h generation.
Hiroki Sato [Wed, 17 Jul 2013 13:47:13 +0000 (13:47 +0000)]
Simplify keywords.h generation.

11 years agoRemove check forbidding requests that would result in one port being set
John Baldwin [Wed, 17 Jul 2013 13:41:54 +0000 (13:41 +0000)]
Remove check forbidding requests that would result in one port being set
to Ethernet and the subsequent port being set to IB.

Submitted by: Shakar Klein @ Mellanox
Tested by: Morgan Robertson <morganrobertson@gmail.com>
MFC after: 1 week

11 years agoFix typo.
Gleb Smirnoff [Wed, 17 Jul 2013 13:13:44 +0000 (13:13 +0000)]
Fix typo.

Submitted by: ae

11 years agoImprove example, so that it doesn't dump core when example module
Gleb Smirnoff [Wed, 17 Jul 2013 12:28:48 +0000 (12:28 +0000)]
Improve example, so that it doesn't dump core when example module
isn't loaded.

11 years agoChange Eu to the proper UTF-8 symbol this was missed in r253414
Baptiste Daroussin [Wed, 17 Jul 2013 11:51:24 +0000 (11:51 +0000)]
Change Eu to the proper UTF-8 symbol this was missed in r253414

11 years agoClear the vnode knotes before destroying vpollinfo.
Konstantin Belousov [Wed, 17 Jul 2013 10:56:21 +0000 (10:56 +0000)]
Clear the vnode knotes before destroying vpollinfo.

Reported and tested by: Patrick Lamaiziere <patfbsd@davenulle.org>
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks

11 years agoFor all european contries using the euro currency, add an UTF-8 definition with the...
Baptiste Daroussin [Wed, 17 Jul 2013 10:12:31 +0000 (10:12 +0000)]
For all european contries using the euro currency, add an UTF-8 definition with the proper Euro symbol

MFC after: 1 week

11 years ago[fix to r253380] __setenv: be explicit about placing '=' after name
Andriy Gapon [Wed, 17 Jul 2013 08:45:27 +0000 (08:45 +0000)]
[fix to r253380] __setenv: be explicit about placing '=' after name

This should a regression introduced in r253380 if malloc'ed memory
happens to have '=' at the right place.

Reported by: ache
Pointyhat to: me (avg)
MFC after: 1 day
X-MFC with: r253380

11 years agoImplement workaround for BCM5719/BCM5720 TX hang.
Pyun YongHyeon [Wed, 17 Jul 2013 04:53:53 +0000 (04:53 +0000)]
Implement workaround for BCM5719/BCM5720 TX hang.
The read DMA request logic operation is based on having sufficient
available space in the transmit data buffer (TXMBUF) before a read
DMA can be requested.  There are four read DMA channels that use
the TXMBUF, and the logic checks if the available free space in the
TXMBUF is large enough for all the data in the four Send Buffers
for which buffer descriptors have been fetched.  The Enable_Request
signal is asserted only if the free TXMBUF space is larger than the
sum of the four DMA length registers.  The power-up default value
of BGE_RDMA_LSO_CRPTEN_CTRL register bit 25 (bit 21 on BCM5720) is
zero, which selects the DMA length registers to connect to the
input of the adder block.  The DMA length registers are
asynchronously reset following BCM5719/BCM5720 power-up, and due to
the lack of synchronous deassertion of the length registers reset
signal these resisters may contain uninitialized values following
the reset deassertion.
In the case of the failure the uninitialized DMA length register
values added up to more than the TXMBUF size, which prevented the
assertion of the Enable_Request signal and any subsequent read DMA
to start.  This lockup condition is the root cause of failing to
generate any transmit traffic.

To workaround the issue, select alternate output of multiplexers
and transmit the first four Ethernet frames. This overwrites the
DMA length registers with valid values.

Reported by: Geans Pin <geanspin@broadcom.com>
Reviewed by: Geans Pin <geanspin@broadcom.com>

11 years agoSpecify a timeout for the PL block.
Navdeep Parhar [Wed, 17 Jul 2013 02:37:40 +0000 (02:37 +0000)]
Specify a timeout for the PL block.

MFC after: 3 days

11 years agoAvoid magic constant.
Pyun YongHyeon [Wed, 17 Jul 2013 01:34:25 +0000 (01:34 +0000)]
Avoid magic constant.
No functional change.

11 years ago o TxD ring requires 8 bytes alignment to work so change alignment
Pyun YongHyeon [Wed, 17 Jul 2013 01:24:23 +0000 (01:24 +0000)]
 o TxD ring requires 8 bytes alignment to work so change alignment
   constraint to 8. Previously it may have triggered watchdog
   timeouts.
 o Check whether interrupt is ours or not.
 o Enable interrupts before attemping to transmit queued packets.
   This will slightly improve TX performance.
 o No need to clear IFF_DRV_OACTIVE in a loop. AE_FLAG_TXAVAIL is
   used to know whether there are enough available TxD ring space.
 o Added missing bus_dmamap_sync(9) in ae_rx_intr() and rearranged
   code to avoid unncessary register access.
 o Make sure to clear TxD, TxS, RxD rings in driver initialization.
   Otherwise some data in these rings could be interpreted as
   'updated' which in turn will advance internally maintained
   pointers and can trigger watchdog timeouts.

PR: kern/180382

11 years agoAdd missing headers.
Kevin Lo [Wed, 17 Jul 2013 00:58:23 +0000 (00:58 +0000)]
Add missing headers.

11 years agoUse PRId64 instead of %gd to print an int64_t.
Kevin Lo [Wed, 17 Jul 2013 00:54:21 +0000 (00:54 +0000)]
Use PRId64 instead of %gd to print an int64_t.

11 years agoFix vfp:
Andrew Turner [Tue, 16 Jul 2013 23:19:05 +0000 (23:19 +0000)]
Fix vfp:
 - We should check is_d32 to see howmany registers we have
 - In vfp_restore mark vfpscr as an output register

Without the second part it appears we can return the incorrect value from
vfp_bounce if the VFP condition flags are set as it may override the
register holding the return value.

11 years agoAdd a tunable to force disable MSI use for xhci(4).
Konstantin Belousov [Tue, 16 Jul 2013 20:14:23 +0000 (20:14 +0000)]
Add a tunable to force disable MSI use for xhci(4).

Requested and tested by: delphij
Sponsored by: The FreeBSD Foundation
MFC after: 3 days

11 years agoRegen.
Andrew Turner [Tue, 16 Jul 2013 19:20:50 +0000 (19:20 +0000)]
Regen.

11 years ago2 years, 10 months, 22 days after the projects/arm_eabi branch was created
Andrew Turner [Tue, 16 Jul 2013 19:15:19 +0000 (19:15 +0000)]
2 years, 10 months, 22 days after the projects/arm_eabi branch was created
make the ARM EABI the default ABI on arm, armeb, armv6 and armv6eb.

This is intended to be the default ABI from now on with the old ABI to be
retired. Because of this all users are strongly suggested to upgrade to the
ARM EABI.

As the two ABIs are incompatible it is unlikely upgrading in place will
work. Users should perform a full backup and either use an external machine
to upgrade, or install to an alternative location on their media. They
should also reinstall all ports or packages when these are available.

The only known issues are:
 - pkg incorrectly detects the ABI. This is fixed upstream, and will a
   patch will be made to the port.
 - GDB can have issues with executables built with clang.

__FreeBSD_version has been bumped.

11 years agoFree the non-fatal "timestamp missing" debug string manually as it is
Andre Oppermann [Tue, 16 Jul 2013 16:37:08 +0000 (16:37 +0000)]
Free the non-fatal "timestamp missing" debug string manually as it is
not covered by the catch-all free for the error cases.

Found by: Coverity

11 years agoDo not throw an error if the user requests to activate the image from
Jim Harris [Tue, 16 Jul 2013 15:45:37 +0000 (15:45 +0000)]
Do not throw an error if the user requests to activate the image from
an empty firmware slot, as long as the user has specified a firmware
image to download into the empty firmware slot.

Sponsored by: Intel
Reported by: Joe Golio <joseph.golio@emc.com>
MFC after: 3 days

11 years agoWorkaround some broken BIOSes that specify edge-sensitive but active-low
John Baldwin [Tue, 16 Jul 2013 14:42:16 +0000 (14:42 +0000)]
Workaround some broken BIOSes that specify edge-sensitive but active-low
settings for ACPI-enumerated serial ports by forcing any IRQs that use
an ISA IRQ value with these settings to active-high instead of active-low.

This is known to occur with the BIOS on an Intel D2500CCE motherboard.

Tested by: Robert Ames <robertames@hotmail.com>, lev
Submitted by: Juergen Weiss weiss at uni-mainz.de (original patch)

11 years agoSweep unused nlist entries.
Gleb Smirnoff [Tue, 16 Jul 2013 12:22:36 +0000 (12:22 +0000)]
Sweep unused nlist entries.

Sponsored by: Nginx, Inc.

11 years agoNeed to define NO_MAN here.
Kevin Lo [Tue, 16 Jul 2013 09:40:59 +0000 (09:40 +0000)]
Need to define NO_MAN here.

11 years agoUse MAN= instead of NO_MAN=
Kevin Lo [Tue, 16 Jul 2013 09:38:51 +0000 (09:38 +0000)]
Use MAN= instead of NO_MAN=

For some reason it still tries to install a priv.1 when using NO_MAN,
even though there isn't one yet.

11 years agoAdd missing header needed by sleep(3).
Kevin Lo [Tue, 16 Jul 2013 09:22:50 +0000 (09:22 +0000)]
Add missing header needed by sleep(3).

11 years agolibc: name passed into __setenv is not necessarily NUL-terminated
Andriy Gapon [Tue, 16 Jul 2013 07:26:46 +0000 (07:26 +0000)]
libc: name passed into __setenv is not necessarily NUL-terminated

That's particularly true when __setenv is called from __merge_environ.

MFC after: 4 days

11 years agoAdd urtwnfw.
Rui Paulo [Tue, 16 Jul 2013 06:43:37 +0000 (06:43 +0000)]
Add urtwnfw.

11 years agoSilent warnings.
Kevin Lo [Tue, 16 Jul 2013 03:12:31 +0000 (03:12 +0000)]
Silent warnings.

11 years agoAdd -n flag for compatibility with Linux version of mount(8).
Robert Millan [Mon, 15 Jul 2013 21:57:21 +0000 (21:57 +0000)]
Add -n flag for compatibility with Linux version of mount(8).

Reviewed by: freebsd-fs, eadler, mckusick, jh, wblock

11 years agoMake some improvements to r253322 to really rescan target, not a bus.
Alexander Motin [Mon, 15 Jul 2013 18:17:31 +0000 (18:17 +0000)]
Make some improvements to r253322 to really rescan target, not a bus.
Add there and in two more places checks for NULL on xpt_alloc_ccb_nowait().

11 years agoFix an argument reversal in calls to scsi_read_element_status().
Kenneth D. Merry [Mon, 15 Jul 2013 16:38:48 +0000 (16:38 +0000)]
Fix an argument reversal in calls to scsi_read_element_status().

Reported by: Ulrich Spoerlein <uqs@FreeBSD.org>
MFC after: 3 days

11 years agoInclude sys/systm.h after sys/param.h.
Andrey V. Elsukov [Mon, 15 Jul 2013 15:40:57 +0000 (15:40 +0000)]
Include sys/systm.h after sys/param.h.

Suggested by: pluknet

11 years agoAdd IBM ServeRAID M5110 to the hardware list.
Sergey Kandaurov [Mon, 15 Jul 2013 14:28:59 +0000 (14:28 +0000)]
Add IBM ServeRAID M5110 to the hardware list.
This is a rebranded MegaRAID SAS 2208 [Thunderbolt] found e.g. on x3550 M4.

Reported by: Ilia Noskov <noskov@nic.ru>
MFC after: 3 days

11 years agoNuke mbstat. It wasn't used for mbuf statistics since FreeBSD 5.
Gleb Smirnoff [Mon, 15 Jul 2013 12:18:36 +0000 (12:18 +0000)]
Nuke mbstat. It wasn't used for mbuf statistics since FreeBSD 5.

Now that r253351 moved sendfile() stats to a separate struct, the
last field used in mbstat is m_mcfail, which is updated, but never
read or obtained from userland.

11 years agoNuke "systat -mbuf". It was broken since FreeBSD 5, and since there
Gleb Smirnoff [Mon, 15 Jul 2013 12:15:14 +0000 (12:15 +0000)]
Nuke "systat -mbuf". It was broken since FreeBSD 5, and since there
haven't been any complaints, no one used it.

11 years agoWhen stopping pfsync, remove the pfsync interface's syncpeer setting.
Dag-Erling Smørgrav [Mon, 15 Jul 2013 08:48:45 +0000 (08:48 +0000)]
When stopping pfsync, remove the pfsync interface's syncpeer setting.

MFC after: 3 days