]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
10 years agoMFV of 255902, tzdata2013f
Edwin Groothuis [Tue, 5 Nov 2013 06:32:23 +0000 (06:32 +0000)]
MFV of 255902, tzdata2013f
MFV of 257651, tzdata2013h

tzdata2013f
- Jordan goes to winter time on the last Friday in October.
- Tocantins in Brazil will not go into summer time in October.
- Indonesian time zones renames.
- Lots of cleanups in with regarding to links and historical data.

tzdata2013h
- Libya didn't go back to DST.
- Fix Morocco 2038 issue.
- Brazil/Acre and Western Amazonas are chaning timezones.

10 years agoDo not coalesce if the swap object belongs to tmpfs vnode. The
Konstantin Belousov [Tue, 5 Nov 2013 06:18:50 +0000 (06:18 +0000)]
Do not coalesce if the swap object belongs to tmpfs vnode.  The
coalesce would extend the object to keep pages for the anonymous
mapping created by the process.  The pages has no relations to the
tmpfs file content which could be written into the corresponding
range, causing anonymous mapping and file content aliasing and
subsequent corruption.

Another lesser problem created by coalescing is over-accounting on the
tmpfs node destruction, since the object size is substracted from the
total count of the pages owned by the tmpfs mount.

Reported and tested by: bdrewery
Sponsored by: The FreeBSD Foundation
MFC after: 1 week

10 years agoUse suword32 and suword64 instead of copyout(9). This fixes a bug in the
Mark Johnston [Tue, 5 Nov 2013 06:13:46 +0000 (06:13 +0000)]
Use suword32 and suword64 instead of copyout(9). This fixes a bug in the
emulation of the call instruction caused by reversing the uaddr and kaddr
arguments when copying data out to userland: the suword* functions take the
uaddr as the first argument whereas copyout(9) takes the kaddr as the first
argument. This also partially undoes the fixes from r257143.

Submitted by: Prashanth Kumar <pra_udupi@yahoo.co.in> (original version)
MFC after: 1 month

10 years agoRemove local devmap code and use the essentially identical common code
Ian Lepore [Tue, 5 Nov 2013 05:18:18 +0000 (05:18 +0000)]
Remove local devmap code and use the essentially identical common code
that got moved from imx_machdep.c to arm/devmap.c.

10 years agoStyle and comment tweaks, no functional changes.
Ian Lepore [Tue, 5 Nov 2013 05:01:46 +0000 (05:01 +0000)]
Style and comment tweaks, no functional changes.

10 years agoAdd new helper routines for arm static device mapping. The new code
Ian Lepore [Tue, 5 Nov 2013 04:30:55 +0000 (04:30 +0000)]
Add new helper routines for arm static device mapping.  The new code
allocates kva space from the top down for the device mappings and builds
entries in an internal table which is automatically used later by
arm_devmap_bootstrap().  The platform code just calls the new
arm_devmap_add_entry() function as many times as it needs to (up to 32
entries allowed; most platforms use 2 or 3 at most).

There is also a new arm_devmap_lastaddr() function that returns the lowest
kva address allocated; this can be used to implement initarm_lastaddr()
which is used to initialize vm_max_kernel_address.

The new code is based on a similar concept developed for the imx family
SoCs recently.  They will soon be converted to use this new common code.

10 years agoMake PTE_DEVICE a synonym for PTE_NOCACHE on armv4, to make it easier to
Ian Lepore [Tue, 5 Nov 2013 04:06:29 +0000 (04:06 +0000)]
Make PTE_DEVICE a synonym for PTE_NOCACHE on armv4, to make it easier to
share the same code on both architectures.

10 years agoAdd myself to the calendar.
Mark Johnston [Tue, 5 Nov 2013 03:25:10 +0000 (03:25 +0000)]
Add myself to the calendar.

10 years agoModify the libproc breakpoint add/remove functions to stop the target
Mark Johnston [Tue, 5 Nov 2013 03:23:54 +0000 (03:23 +0000)]
Modify the libproc breakpoint add/remove functions to stop the target
process if it has not already been stopped, since this is required for
ptrace(2) to work.

libdtrace does not seem to stop target processes before trying to remove
their breakpoints, so we were previously failing to remove the breakpoint
on r_debug_state() in rtld. This was causing processes to die with SIGTRAP
if they called dlopen(3) after dtrace(1) had detached.

Reported by: symbolics@gmx.com
Reviewed by: rpaulo
MFC after: 1 month

10 years agoCall initarm_lastaddr() later in the init sequence, after establishing
Ian Lepore [Tue, 5 Nov 2013 02:57:34 +0000 (02:57 +0000)]
Call initarm_lastaddr() later in the init sequence, after establishing
static device mappings, rather than as the first of the initializations
that a platform can hook into.  This allows a platform to allocate KVA
from the top of the address space downwards for things like static device
mapping, and return the final "last usable address" result after that and
other early init work is done.

Because some platforms were doing work in initarm_lastaddr() that needs to
be done early, add a new initarm_early_init() routine and move the early
init code to that routine on those platforms.

Rename platform_devmap_init() to initarm_devmap_init() to match all the
other init routines called from initarm() that are designed to be
implemented by platform code.

Add a comment block that explains when these routines are called and the
type of work expected to be done in each of them.

10 years agoUse proper capitalization for FreeBSD.org
Bryan Drewery [Tue, 5 Nov 2013 02:22:04 +0000 (02:22 +0000)]
Use proper capitalization for FreeBSD.org

Approved by: bapt
MFC after: 2 days

10 years agoEnable fingerprint checking as the currently known
Bryan Drewery [Tue, 5 Nov 2013 02:20:49 +0000 (02:20 +0000)]
Enable fingerprint checking as the currently known
fingerprint has an uploaded signature on all mirrors.

Approved by: bapt
MFC after: 2 days

10 years agoremove some test code.
Luigi Rizzo [Tue, 5 Nov 2013 01:06:22 +0000 (01:06 +0000)]
remove some test code.

10 years agofix a bug when a device has 1 tx (or rx) queue and more than
Luigi Rizzo [Tue, 5 Nov 2013 00:56:07 +0000 (00:56 +0000)]
fix a bug when a device has 1 tx (or rx) queue and more than
one queue of a different type.

Submitted by: Vincenzo Maffione
MFC after: 3 days

10 years agocheck errors on return from netmap_attach()
Luigi Rizzo [Tue, 5 Nov 2013 00:50:59 +0000 (00:50 +0000)]
check errors on return from netmap_attach()

Submitted by: Giuseppe Lettieri
MFC after: 3 days

10 years agobmake does not have a -v option so remove it
Luigi Rizzo [Mon, 4 Nov 2013 23:46:20 +0000 (23:46 +0000)]
bmake does not have a -v option so remove it

10 years agoRemove the 'vdev' abstraction that was meant to sit on top of device models
Neel Natu [Mon, 4 Nov 2013 23:25:07 +0000 (23:25 +0000)]
Remove the 'vdev' abstraction that was meant to sit on top of device models
in the kernel. This abstraction was redundant because the only device emulated
inside vmm.ko is the local apic and it is always at a fixed guest physical
address.

Discussed with: grehan

10 years agoMove remaining code and data related to static device mapping into the
Ian Lepore [Mon, 4 Nov 2013 22:45:26 +0000 (22:45 +0000)]
Move remaining code and data related to static device mapping into the
new devmap.[ch] files.  Emphasize the MD nature of these things by using
the prefix arm_devmap_ on the function and type names (already a few of
these things found their way into MI code, hopefully it will be harder to
do by accident in the future).

10 years agoFor C++ programs, don't emit any explicit dependency to libstdc++.a or
Dimitry Andric [Mon, 4 Nov 2013 21:54:56 +0000 (21:54 +0000)]
For C++ programs, don't emit any explicit dependency to libstdc++.a or
libc++.a during the early build stages (bootstrap-tools, build-tools,
cross-tools), since it is not possible to know in advance which C++
library is available on the host system.

Instead, just use the bootstrap compiler's built-in default.  This
should eventually make it possible to build stable/9 on head, or on
stable/10, which ship without libstdc++ by default.

X-MFC-With: 255431
MFC after: 3 days

10 years agoQuiesce warning assigning to void * from const ctf_header_t * by explicity casting
Sean Bruno [Mon, 4 Nov 2013 21:32:07 +0000 (21:32 +0000)]
Quiesce warning assigning to void * from const ctf_header_t * by explicity casting
to void * before assignment.

Submitted as Illumos issue 4287

10 years agomdoc: remove EOL whitespace.
Joel Dahl [Mon, 4 Nov 2013 21:28:36 +0000 (21:28 +0000)]
mdoc: remove EOL whitespace.

10 years agoRemove superfluous paragraph macro and add missing "the".
Joel Dahl [Mon, 4 Nov 2013 21:27:21 +0000 (21:27 +0000)]
Remove superfluous paragraph macro and add missing "the".

10 years agocxgbe(4): Exclude MPS_RPLC_MAP_CTL (0x11114) from the register dump. Turns
Navdeep Parhar [Mon, 4 Nov 2013 21:06:21 +0000 (21:06 +0000)]
cxgbe(4): Exclude MPS_RPLC_MAP_CTL (0x11114) from the register dump.  Turns
out it's a write-only register with strange side effects on read.

Submitted by: gnn
MFC after: 3 days

10 years agoDefer loading of kernel and modules if the beastie menu is enabled. Add a
Devin Teske [Mon, 4 Nov 2013 20:28:10 +0000 (20:28 +0000)]
Defer loading of kernel and modules if the beastie menu is enabled. Add a
kernel selection menu to the beastie menu. List of kernels is taken from
`kernels' in loader.conf(5) as a space (or comma) separated list of names
to display (up to 9). If not set, default value is "kernel kernel.old".
Does not validate that kernels exist because the next enhancement will be
to allow selection of the root device.

Discussed on: -current
MFC after: 3 days

10 years agoThis files should have been included in r257648.
Ian Lepore [Mon, 4 Nov 2013 20:00:21 +0000 (20:00 +0000)]
This files should have been included in r257648.

10 years agoBegin reducing code duplication in arm pmap.c and pmap-v6.c by factoring
Ian Lepore [Mon, 4 Nov 2013 19:44:37 +0000 (19:44 +0000)]
Begin reducing code duplication in arm pmap.c and pmap-v6.c by factoring
out common code related to mapping device memory into a new devmap.c file.

Remove the growing duplication of code that used pmap_devmap_find_pa() and
then did some math with the returned results to generate a virtual address,
and likewise in reverse to get a physical address.  Now there are a pair
of functions, arm_devmap_vtop() and arm_devmap_ptov(), to do that.  The
bus_space_map() implementations are rewritten in terms of these.

10 years agospelling in comments fixup
Sean Bruno [Mon, 4 Nov 2013 19:32:35 +0000 (19:32 +0000)]
spelling in comments fixup

Submitted by: Joerg Sonnenberger <joerg@britannica.bec.de>

10 years agoUse OpenBSD's revamped description of strlcpy and strlcat.
Eitan Adler [Mon, 4 Nov 2013 19:05:31 +0000 (19:05 +0000)]
Use OpenBSD's revamped description of strlcpy and strlcat.

This explanation is supposed to be simpler and better.  In particular
"comparing it to the snprintf API provides lots of value, since it raises the
bar on understanding, so that programmers/auditors will a better job calling
all 3 of these functions."

Requested by: deraadt@cvs.openbsd.org
Obtained From: OpenBSD
Reviewed by: cperciva

10 years agoQuiesce warning around gcc_assert() for an inline macro that uses
Sean Bruno [Mon, 4 Nov 2013 18:15:45 +0000 (18:15 +0000)]
Quiesce warning around gcc_assert() for an inline macro that uses
a static variable.  This code has been moved around in gcc, but is still in
use in the latest trunk version of the compiler.

gnu/lib/libgcc/../../../contrib/gcc/unwind-dw2.c:208:36:
      warning: static variable 'dwarf_reg_size_table' is used in an inline
               function with external linkage [-Wstatic-in-inline]
gcc_assert (index < (int) sizeof(dwarf_reg_size_table));

10 years agoDocument RB_FOREACH_SAFE() and RB_FOREACH_REVERSE_SAFE().
Gleb Smirnoff [Mon, 4 Nov 2013 16:56:35 +0000 (16:56 +0000)]
Document RB_FOREACH_SAFE() and RB_FOREACH_REVERSE_SAFE().

Submitted by: Mikhail <mp lenta.ru>

10 years agoQuiesce warning about unused argument in call to rl_message() by wrapping
Sean Bruno [Mon, 4 Nov 2013 16:52:27 +0000 (16:52 +0000)]
Quiesce warning about unused argument in call to rl_message() by wrapping
this call with the same #if defined (PREFER_STDARG) directive as in display.c

Using -E to compile display.c/search.c shows that this is the code chosen by
the build when we create libreadline

10 years agoWhen building the textproc/docproj port, the ports-mgmt/pkg port
Glen Barber [Mon, 4 Nov 2013 16:28:29 +0000 (16:28 +0000)]
When building the textproc/docproj port, the ports-mgmt/pkg port
needs /var/run/ld-elf*.so.hints, which is not automatically created.

Fix reldoc build by running the ldconfig(8) startup script in the
chroot directory before starting the port build phase.

MFC after: 3 days
Sponsored by: The FreeBSD Foundation

10 years agoRemove the duplicated implementations of some bus_space functions and use
Ian Lepore [Mon, 4 Nov 2013 16:16:40 +0000 (16:16 +0000)]
Remove the duplicated implementations of some bus_space functions and use
the essentially identical generic implementations instead.  The generic
implementations differ only in the spelling of a couple variable names
and some formatting differences.

10 years agoQuiesce warning regarding %llf which has no effect.
Sean Bruno [Mon, 4 Nov 2013 16:15:43 +0000 (16:15 +0000)]
Quiesce warning regarding %llf which has no effect.

Submitted as illumos issue #4284

Reviewed by: delphij

10 years agoRework the aicasm build machinery so that it gets built along with toolchain
Ian Lepore [Mon, 4 Nov 2013 15:55:04 +0000 (15:55 +0000)]
Rework the aicasm build machinery so that it gets built along with toolchain
components instead of with the kernel and/or modules.  This ensures that it
gets built with the host compiler, not the compiler in obj/... used to build
the target components (which may be a cross-compiler outputting code for a
different architecture and using header files with types and options set up
for the wrong architecture).

Reviewed by: imp

10 years agoFix EUC ESDB and JIS CS-mapper. This also fixes an identity transformation
Hiroki Sato [Mon, 4 Nov 2013 15:49:36 +0000 (15:49 +0000)]
Fix EUC ESDB and JIS CS-mapper.  This also fixes an identity transformation
by "cat foo.euc-jp | iconv -f euc-jp -t utf-8 | iconv -f utf-8 -t euc-jp".

10 years ago- Add manual pages for capability rights (rights(4)), cap_rights_init(3)
Pawel Jakub Dawidek [Mon, 4 Nov 2013 14:10:22 +0000 (14:10 +0000)]
- Add manual pages for capability rights (rights(4)), cap_rights_init(3)
  family of functions and cap_rights_get(3) function.
- Update remaining Capsicum-related manual pages.

Reviewed by: bdrewery
MFC after: 3 days

10 years agoAdd myself to the committers-src list and to the calendar.
Julio Merino [Mon, 4 Nov 2013 11:59:44 +0000 (11:59 +0000)]
Add myself to the committers-src list and to the calendar.

Approved by: rpaulo (mentor)

10 years agoCode logic of handling PFTM_PURGE into pf_find_state().
Gleb Smirnoff [Mon, 4 Nov 2013 08:20:06 +0000 (08:20 +0000)]
Code logic of handling PFTM_PURGE into pf_find_state().

10 years agoRemove unused PFTM_UNTIL_PACKET const.
Gleb Smirnoff [Mon, 4 Nov 2013 08:15:59 +0000 (08:15 +0000)]
Remove unused PFTM_UNTIL_PACKET const.

10 years agoAdd Bluetooth/PAN (personal area networking) commands.
Adrian Chadd [Mon, 4 Nov 2013 05:52:42 +0000 (05:52 +0000)]
Add Bluetooth/PAN (personal area networking) commands.

Submitted by: Cedric Gross <cg@cgross.info>

10 years agoAdd device ids for the Centrino 2x00 devices.
Adrian Chadd [Mon, 4 Nov 2013 05:40:19 +0000 (05:40 +0000)]
Add device ids for the Centrino 2x00 devices.

Submitted by: Cedric Gross <cg@cgross.info>

10 years agoRemove trailing whitespace.
Adrian Chadd [Mon, 4 Nov 2013 04:10:36 +0000 (04:10 +0000)]
Remove trailing whitespace.

Submitted by: Cedric Gross <cg@cgross.info>

10 years agoProperly quote the included filename, now that it has a dot in it. Doh.
Ian Lepore [Mon, 4 Nov 2013 03:46:09 +0000 (03:46 +0000)]
Properly quote the included filename, now that it has a dot in it.  Doh.

10 years agoRename WANDBOARD-COMMON to WANDBOARD.common and adjust the configs that
Ian Lepore [Mon, 4 Nov 2013 03:39:23 +0000 (03:39 +0000)]
Rename WANDBOARD-COMMON to WANDBOARD.common and adjust the configs that
include it accordingly.  The build machinery for universe and tinderbox
tries to build every kernel config whose name begins and ends with [A-Z0-9]
and the common include file that has most of the options isn't buildable
by itself, so the new lowercase .common will avoid building it.

10 years agoInitialize the struct tm before handing it to strptime(3).
Mark Johnston [Mon, 4 Nov 2013 02:50:43 +0000 (02:50 +0000)]
Initialize the struct tm before handing it to strptime(3).

PR: 183550
MFC after: 2 weeks

10 years agoDuring code inspection, I spotted that there was a code path where
Rick Macklem [Sun, 3 Nov 2013 23:17:30 +0000 (23:17 +0000)]
During code inspection, I spotted that there was a code path where
CLNT_CONTROL() would be called on "client" after it was
released via CLNT_RELEASE(). It was unlikely that this
code path gets executed and I have not heard of any problem
report caused by this bug. This patch fixes the code so that
this cannot happen.

MFC after: 2 months

10 years agokqueue: Change error for kqueues rlimit from EMFILE to ENOMEM and document
Jilles Tjoelker [Sun, 3 Nov 2013 23:06:24 +0000 (23:06 +0000)]
kqueue: Change error for kqueues rlimit from EMFILE to ENOMEM and document
this error condition in the kqueue(2) manual page.

Discussed with: kib

10 years agoComments and style(9) only, no functional changes.
Ian Lepore [Sun, 3 Nov 2013 22:55:33 +0000 (22:55 +0000)]
Comments and style(9) only, no functional changes.

10 years agoAdd description of two EHCI PCI IDs.
Hans Petter Selasky [Sun, 3 Nov 2013 21:33:42 +0000 (21:33 +0000)]
Add description of two EHCI PCI IDs.

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

10 years agoThis library uses macros to define fprintf behvavior for several object types
Sean Bruno [Sun, 3 Nov 2013 21:05:44 +0000 (21:05 +0000)]
This library uses macros to define fprintf behvavior for several object types
The compiler will see the non-string literal arguments to the fprintf calls and
omit warnings for them. Quiese these warnings in contrib code:

cddl/contrib/opensolaris/lib/libnvpair/libnvpair.c:743:12: warning: format
  string is not a string literal (potentially insecure) [-Wformat-security]
  ARENDER(pctl, nvlist_array, nvl, name, val, nelem);

10 years agoUpdate libreadline with vendor patches 8-13 as fetched from:
Sean Bruno [Sun, 3 Nov 2013 20:38:51 +0000 (20:38 +0000)]
Update libreadline with vendor patches 8-13 as fetched from:
ftp://ftp.cwru.edu/pub/bash/readline-5.2-patches/

MFC after: 2 weeks

10 years agoRemove the WITH_LIBICONV_COMPAT hack that seems to do more harm than
Peter Wemm [Sun, 3 Nov 2013 19:04:57 +0000 (19:04 +0000)]
Remove the WITH_LIBICONV_COMPAT hack that seems to do more harm than
good.  This caused libc to spoof the ports libiconv namespace and
provide a colliding libiconv.so.3 to fool rtld.  This should have
been removed some time ago.

10 years agoCorrect alignment.
Pawel Jakub Dawidek [Sun, 3 Nov 2013 19:02:18 +0000 (19:02 +0000)]
Correct alignment.

10 years agoRemove utxrm(8). This command has been superseded by 'utx rm'.
Ed Schouten [Sun, 3 Nov 2013 16:43:52 +0000 (16:43 +0000)]
Remove utxrm(8). This command has been superseded by 'utx rm'.

10 years agoUnlock the lock before destroying it.
Michael Tuexen [Sun, 3 Nov 2013 14:00:17 +0000 (14:00 +0000)]
Unlock the lock before destroying it.
This issue was reported by Andrew Galante.

MFC after: 3 days

10 years agoBugfix: the attach routine needs to use the same table of fdt compat
Ian Lepore [Sat, 2 Nov 2013 22:44:35 +0000 (22:44 +0000)]
Bugfix:  the attach routine needs to use the same table of fdt compat
strings that the probe routine used.

10 years agoAdd a missing register definition.
Ian Lepore [Sat, 2 Nov 2013 21:07:39 +0000 (21:07 +0000)]
Add a missing register definition.

10 years agoArrange for uart_cpu_fdt's probe() routine to use the same table of compat
Ian Lepore [Sat, 2 Nov 2013 20:14:39 +0000 (20:14 +0000)]
Arrange for uart_cpu_fdt's probe() routine to use the same table of compat
strings as uart_bus_fdt's probe().

The bus code uses ofw_bus_search_compatible() and that's not an option in
cpu (console) code -- it runs way before the ofw routines are usable.  So
the console probe has its own loop to search the table, but now at least
there's only one table to be maintained when new devices are added.

10 years agoChanges from upstream to improve compilation when INET or INET6
Michael Tuexen [Sat, 2 Nov 2013 20:12:19 +0000 (20:12 +0000)]
Changes from upstream to improve compilation when INET or INET6
or none of them is defined.

MFC after: 3 days

10 years agocircumvent a couple of warnings:
Luigi Rizzo [Sat, 2 Nov 2013 18:03:21 +0000 (18:03 +0000)]
circumvent a couple of warnings:
- on line 2550 intentionally overriding a const qualifier
- on line 3219 intentionally converting uint64_t to a pointer

10 years agoDon't create a distinct free page pool for segregating allocations that are
Alan Cox [Sat, 2 Nov 2013 17:08:20 +0000 (17:08 +0000)]
Don't create a distinct free page pool for segregating allocations that are
accessed through the direct map unless the kernel configuration actually
includes a direct map.  Only a few configurations do, and for the rest the
unnecessary free page pool is a small pessimization.

Tested by: zbb
MFC after: 6 weeks

10 years agoRestore the entropy gathering from the m_data pointer value, not the
Adrian Chadd [Sat, 2 Nov 2013 15:13:02 +0000 (15:13 +0000)]
Restore the entropy gathering from the m_data pointer value, not the
m_data payload.

After talking with markm/bde, this is what markm actually intended.

10 years agoAdd device ID for 'Sanoxy 802.11N' usb
Alfred Perlstein [Sat, 2 Nov 2013 11:37:16 +0000 (11:37 +0000)]
Add device ID for 'Sanoxy 802.11N' usb

10 years agoFix API mismatch exposed by lagg.
Alfred Perlstein [Sat, 2 Nov 2013 10:49:47 +0000 (10:49 +0000)]
Fix API mismatch exposed by lagg.

When destroying a lagg the driver tries to restore the old mac and
fails due to API mismatch

10 years agoFix several issues with the busdma(9) KPI use in the e1000 drivers.
Konstantin Belousov [Sat, 2 Nov 2013 09:16:11 +0000 (09:16 +0000)]
Fix several issues with the busdma(9) KPI use in the e1000 drivers.
The problems do not affect bouncing busdma in a visible way, but are
critical for the dmar backend.

- The bus_dmamap_create(9) is not documented to take BUS_DMA_NOWAIT flag.
- Unload descriptor map after receive.
- Do not reset descriptor map to NULL, bus_dmamap_load(9) requires
  valid map, and also this leaks the map.

Reported and tested by: pho
Approved by: jfv
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks

10 years agoWhen zero'ing out a buffer, make sure we are using right size.
Xin LI [Sat, 2 Nov 2013 01:16:10 +0000 (01:16 +0000)]
When zero'ing out a buffer, make sure we are using right size.

Without this change, in the worst but unlikely case scenario, certain
administrative operations, including change of configuration, set or
delete key from a GEOM ELI provider, may leave potentially sensitive
information in buffer allocated from kernel memory.

We believe that it is not possible to actively exploit these issues, nor
does it impact the security of normal usage of GEOM ELI providers when
these operations are not performed after system boot.

Security: possible sensitive information disclosure
Submitted by: Clement Lecigne <clecigne google com>
MFC after: 3 days

10 years agoadd missing file from previous netmap update...
Luigi Rizzo [Sat, 2 Nov 2013 00:54:47 +0000 (00:54 +0000)]
add missing file from previous netmap update...

10 years agoTeach the netgraph code to use a const char * pointer too.
Adrian Chadd [Sat, 2 Nov 2013 00:11:38 +0000 (00:11 +0000)]
Teach the netgraph code to use a const char * pointer too.

Pointy hat to: adrian

10 years agoCreate a unique unit number for each controller and namespace cdev.
Jim Harris [Fri, 1 Nov 2013 23:30:54 +0000 (23:30 +0000)]
Create a unique unit number for each controller and namespace cdev.

Sponsored by: Intel
MFC after: 3 days

10 years agoFix this build for clang.
Adrian Chadd [Fri, 1 Nov 2013 22:38:58 +0000 (22:38 +0000)]
Fix this build for clang.

10 years agoDo not exit with error status after printing data for perftest.
Jim Harris [Fri, 1 Nov 2013 22:05:29 +0000 (22:05 +0000)]
Do not exit with error status after printing data for perftest.

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

10 years agoReimplement r257525 such that it work with the historic FreeBSD make
Brooks Davis [Fri, 1 Nov 2013 22:03:20 +0000 (22:03 +0000)]
Reimplement r257525 such that it work with the historic FreeBSD make
implementation.  This fixes the toolchain and kernel-toolchain targets
when building from older FreeBSD versions where make is fmake.

Reported by: luigi
Sponsored by: DARPA/AFRL
MFC after: 3 days

10 years agoupdate to the latest netmap snapshot.
Luigi Rizzo [Fri, 1 Nov 2013 21:21:14 +0000 (21:21 +0000)]
update to the latest netmap snapshot.
This includes the following:
- use separate memory regions for VALE ports
- locking fixes
- some simplifications in the NIC-specific routines
- performance improvements for the VALE switch
- some new features in the pkt-gen test program
- documentation updates

There are small API changes that require programs to be recompiled
(NETMAP_API has been bumped so you will detect old binaries at runtime).

In particular:
- struct netmap_slot now is 16 bytes to support an extra pointer,
  which may save one data copy when using VALE ports or VMs;
- the struct netmap_if has two extra fields;

MFC after: 3 days

10 years agoConvert the random entropy harvesting code to use a const void * pointer
Adrian Chadd [Fri, 1 Nov 2013 20:53:49 +0000 (20:53 +0000)]
Convert the random entropy harvesting code to use a const void * pointer
rather than just void *.

Then, as part of this, convert a couple of mbuf m->m_data accesses
to mtod(m, const void *).

Reviewed by: markm
Approved by: security-officer (delphij)
Sponsored by: Netflix, Inc.

10 years agoThe ability to do 8-bit implies 4-bit capability too. Rearrange the cases
Ian Lepore [Fri, 1 Nov 2013 19:29:59 +0000 (19:29 +0000)]
The ability to do 8-bit implies 4-bit capability too.  Rearrange the cases
and add a fallthrough comment to make that happen.

10 years agoTI sdhci driver improvements, mostly related to fdt data...
Ian Lepore [Fri, 1 Nov 2013 19:22:06 +0000 (19:22 +0000)]
TI sdhci driver improvements, mostly related to fdt data...

Use the published compatible strings (our own invention, "ti,mmchs" is
still accepted as well, for now).

Don't blindly turn on 8-bit bus mode, because even though the conroller
supports it, the board has to be wired appropriately as well.  Use the
published property (bus-width=<n>) and honor all the valid values (1,4,8).

The eMMC device on a Beaglebone Black is wired for 8-bit, update the dts.

The mmchs controller can inherently do both 1.8v and 3.0v on the first
device and 1.8v only on other devices, unless an external transceiver is
used.  Set the voltage automatically for the first device and honor
the published fdt property (ti,dualvolt) for other devices.

Thanks go to Ilya Bakulin for figuring out the voltage compatibility stuff.

10 years agoOpenBSD 5.4 added.
Sergey Kandaurov [Fri, 1 Nov 2013 18:55:56 +0000 (18:55 +0000)]
OpenBSD 5.4 added.

10 years agoSomehow fix LINT-NOIP.
Gleb Smirnoff [Fri, 1 Nov 2013 18:42:03 +0000 (18:42 +0000)]
Somehow fix LINT-NOIP.

10 years agosmall cleanup for building picobsd-head without CLANG.
Luigi Rizzo [Fri, 1 Nov 2013 18:29:27 +0000 (18:29 +0000)]
small cleanup for building picobsd-head without CLANG.

Note that svn 257268 gnu/usr.bin/binutils/ld/Makefile
seems to break the "toolchain" target when building HEAD
on RELENG_9, so until this is solved you may want to

    svn update -r 257267 gnu/usr.bin/binutils/ld/Makefile

before building picobsd

10 years agoAdd support for queued invalidation.
Konstantin Belousov [Fri, 1 Nov 2013 17:38:52 +0000 (17:38 +0000)]
Add support for queued invalidation.

Right now, the semaphore write is scheduled after each batch, which is
not optimal and must be tuned.

Discussed with: alc
Tested by: pho
MFC after: 1 month

10 years agoReturn BUS_PROBE_NOWILDCARD from the DMAR probe method.
Konstantin Belousov [Fri, 1 Nov 2013 17:16:44 +0000 (17:16 +0000)]
Return BUS_PROBE_NOWILDCARD from the DMAR probe method.

Confirmed by: nwhitehorn
MFC after: 1 month

10 years agosh(1),limits(1): Document kqueues (-k) rlimit.
Jilles Tjoelker [Fri, 1 Nov 2013 13:57:30 +0000 (13:57 +0000)]
sh(1),limits(1): Document kqueues (-k) rlimit.

10 years agoAdd -f support to 'pkg bootstrap' and 'pkg add' to force installation
Bryan Drewery [Fri, 1 Nov 2013 12:37:36 +0000 (12:37 +0000)]
Add -f support to 'pkg bootstrap' and 'pkg add' to force installation
of pkg(8) even if already installed.

This is useful if you somehow messup pkg(8) and need to reinstall from
remote with it already being registered in the pkg(8) /var/db/pkg database.

Also add some sanity checks to 'pkg add'.

Approved by: bapt
MFC after: 2 days

10 years agosh: Reorder union node to reduce its size on 64-bit platforms.
Jilles Tjoelker [Fri, 1 Nov 2013 11:28:56 +0000 (11:28 +0000)]
sh: Reorder union node to reduce its size on 64-bit platforms.

10 years agoRemove the dependency on procfs.
Sergey Kandaurov [Fri, 1 Nov 2013 10:42:32 +0000 (10:42 +0000)]
Remove the dependency on procfs.

Reviewed by: kib
MFC after: 1 week
X-MFC with: r257430

10 years agoMake getenv_*() functions and respectively TUNABLE_*_FETCH() macros not
Alexander Motin [Fri, 1 Nov 2013 10:32:33 +0000 (10:32 +0000)]
Make getenv_*() functions and respectively TUNABLE_*_FETCH() macros not
allocate memory and so not require sleepable environment.  getenv() has
already used on-stack temporary storage, so just use it more rationally.
getenv_string() receives buffer as argument, so don't need another one.

10 years agoin_ifadown() can be void.
Gleb Smirnoff [Fri, 1 Nov 2013 10:29:10 +0000 (10:29 +0000)]
in_ifadown() can be void.

10 years agoCleanup in_ifscrub(), which is just an entry to in_scrubprefix().
Gleb Smirnoff [Fri, 1 Nov 2013 10:18:41 +0000 (10:18 +0000)]
Cleanup in_ifscrub(), which is just an entry to in_scrubprefix().

10 years agoprison_check_ip4() can take const arguments.
Gleb Smirnoff [Fri, 1 Nov 2013 10:01:57 +0000 (10:01 +0000)]
prison_check_ip4() can take const arguments.

10 years agoPerform media change after setting IFF_DRV_RUNNING flag. Without it,
Pyun YongHyeon [Fri, 1 Nov 2013 05:03:47 +0000 (05:03 +0000)]
Perform media change after setting IFF_DRV_RUNNING flag. Without it,
driver would ignore the first link state update if controller
already established a link.

Reported by: bsam
Tested by: bsam

10 years agoKernel config for Wandboard.
Ian Lepore [Fri, 1 Nov 2013 02:04:05 +0000 (02:04 +0000)]
Kernel config for Wandboard.

10 years agoUse LOG2_ID_PAGE_SIZE again for the identity mapping in regions 6 & 7.
Marcel Moolenaar [Fri, 1 Nov 2013 01:32:01 +0000 (01:32 +0000)]
Use LOG2_ID_PAGE_SIZE again for the identity mapping in regions 6 & 7.
Make the default translation size the same as the PBVM page size to
avoid inserting overlapping translations in the TC. That generally is
very bad.

10 years agoAdd dts source for imx6 SoCs and for Wandboard boards.
Ian Lepore [Fri, 1 Nov 2013 00:45:08 +0000 (00:45 +0000)]
Add dts source for imx6 SoCs and for Wandboard boards.

The imx6.dtsi is incomplete, but has enough to support devices that work
right now.  Eventually it will need to be split into several files that
account for the differences between SoCs in the imx6 family.

The board config for Wandboard Dual is untested.  Solo and Quad work, but
one one core is turned on for the quad right now (baby steps).

10 years agoInstall include files for netpfil/pf when requested by the Makefile
Sean Bruno [Fri, 1 Nov 2013 00:32:26 +0000 (00:32 +0000)]
Install include files for netpfil/pf when requested by the Makefile

Reviewed by: glebius

10 years agoThe PAL_PTCE_INFO function returns the counts and strides of the
Marcel Moolenaar [Fri, 1 Nov 2013 00:21:38 +0000 (00:21 +0000)]
The PAL_PTCE_INFO function returns the counts and strides of the
outer and inner loop as 32-bit integers mux'd in 64-bit return
values. Change our data types for the count and stride to match
and simplify/adjust accordingly.
Note that with this change the defaults of the ptc.e parameters
have changed. Since all hardware is supposed to support the PAL
call, there should be no impact. Even ski is unaffected, because
the TC is re-initialized without considering the virtual address.
So, as long as we call ptc.e at least once, we're good. That's
what the new defaults do.
Most processor implementations need but a single ptc.e to purge
the entire TC anyway...

10 years agoAdd the Soc- / machine-dependent parts of imx6 support.
Ian Lepore [Fri, 1 Nov 2013 00:21:09 +0000 (00:21 +0000)]
Add the Soc- / machine-dependent parts of imx6 support.

10 years agoFix lock recursion, triggered by `smartctl -a /dev/adaX`.
Alexander Motin [Fri, 1 Nov 2013 00:14:15 +0000 (00:14 +0000)]
Fix lock recursion, triggered by `smartctl -a /dev/adaX`.

10 years agoConvert the if/else list of compatible devices to the table-driven
Ian Lepore [Thu, 31 Oct 2013 23:57:33 +0000 (23:57 +0000)]
Convert the if/else list of compatible devices to the table-driven
ofw_bus_search_compatible() routine.  In addition to converting existing
strings to table entries, also add compat strings for the whole imx family.

10 years agoAdd definitions for the register and data that describes the SoC type.
Ian Lepore [Thu, 31 Oct 2013 23:08:30 +0000 (23:08 +0000)]
Add definitions for the register and data that describes the SoC type.
This isn't in the chip reference manuals, it was found in u-boot and
various old mailing list threads.