]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
10 years agoPrint the MD5 signature information introduced in r221023 in the
Bjoern A. Zeeb [Wed, 5 Feb 2014 20:43:03 +0000 (20:43 +0000)]
Print the MD5 signature information introduced in r221023 in the
TCP statistics output.

MFC after: 3 weeks

10 years agoProperly set the alignment flags when allocating the initial range for a
John Baldwin [Wed, 5 Feb 2014 19:24:16 +0000 (19:24 +0000)]
Properly set the alignment flags when allocating the initial range for a
BAR.  This only really matters when pci_do_realloc_bars is enabled and
the initial allocation of a specific range fails.

MFC after: 1 week

10 years agoFix a typo.
John Baldwin [Wed, 5 Feb 2014 19:23:05 +0000 (19:23 +0000)]
Fix a typo.

10 years agoDescribe the use of a freebsd-boot GPT partition, brought up by Scot
Warren Block [Wed, 5 Feb 2014 18:26:30 +0000 (18:26 +0000)]
Describe the use of a freebsd-boot GPT partition, brought up by Scot
Hetzel <swhetzel@gmail.com> on the -doc mailing list.

Also modify the Author section to be clear that I wrote the man page,
not gptboot.

MFC after: 3 days

10 years agoAdd files to remove WITHOUT_NIS
Antoine Brodin [Wed, 5 Feb 2014 18:16:18 +0000 (18:16 +0000)]
Add files to remove WITHOUT_NIS

PR: 186412

10 years agoDrop the 3rd clause from all 3 clause BSD licenses where I am the sole
John Baldwin [Wed, 5 Feb 2014 18:13:27 +0000 (18:13 +0000)]
Drop the 3rd clause from all 3 clause BSD licenses where I am the sole
holder to convert them to 2 clause BSD licenses.

MFC after: 1 week

10 years agoMove a warning about LINT pins configured with a level trigger under
John Baldwin [Wed, 5 Feb 2014 18:11:46 +0000 (18:11 +0000)]
Move a warning about LINT pins configured with a level trigger under
bootverbose.

10 years ago- Update a few places to account for va_copy().
John Baldwin [Wed, 5 Feb 2014 17:41:00 +0000 (17:41 +0000)]
- Update a few places to account for va_copy().
- Create a separate 'return values' section and move some statements about
  return values to that section.
- Note that each invocation of va_start() and va_copy() must be paired with
  va_end() in the same function.

MFC after: 1 week

10 years agoSimilar to r130943 for cdboot.S, update the license on this file to a
John Baldwin [Wed, 5 Feb 2014 17:22:54 +0000 (17:22 +0000)]
Similar to r130943 for cdboot.S, update the license on this file to a
stock 2-clause BSD license.

MFC after: 1 week

10 years agoAdd the missing ')' at end of sentence. Reword it to use a more common
Luiz Otavio O Souza [Wed, 5 Feb 2014 17:08:55 +0000 (17:08 +0000)]
Add the missing ')' at end of sentence.  Reword it to use a more common
idiom.

Reviewed by: imp (on freebsd-embedded@)
Approved by: adrian (mentor, implicit)

10 years agoFix I/O freezes in some cases, caused by r257916.
Alexander Motin [Wed, 5 Feb 2014 16:22:02 +0000 (16:22 +0000)]
Fix I/O freezes in some cases, caused by r257916.

Delaying isp_reqodx update, we should be ready to update it every time
we read it.  Otherwise requests using several indexes may be requeued
ndefinitely without ever updating the variable.

MFC after: 3 days

10 years agoUse the DELTA() macro to tidy the server-side interval stats code a bit.
John Baldwin [Wed, 5 Feb 2014 14:44:59 +0000 (14:44 +0000)]
Use the DELTA() macro to tidy the server-side interval stats code a bit.

MFC after: 2 weeks

10 years agoMove Open Firmware device root on PowerPC, ARM, and MIPS systems to
Nathan Whitehorn [Wed, 5 Feb 2014 14:44:22 +0000 (14:44 +0000)]
Move Open Firmware device root on PowerPC, ARM, and MIPS systems to
a sub-node of nexus (ofwbus) rather than direct attach under nexus. This
fixes FDT on x86 and will make coexistence with ACPI on ARM systems easier.
SPARC is unchanged.

Reviewed by: imp, ian

10 years agoPartially revert r52493 and change client side interval statistics to
John Baldwin [Wed, 5 Feb 2014 14:33:22 +0000 (14:33 +0000)]
Partially revert r52493 and change client side interval statistics to
report the actual number of RPCs issued, not the theoretical number
that would be issued if all caching was disabled.

Reviewed by: rmacklem
MFC after: 2 weeks

10 years agoAdd an example of the most common use case.
Dag-Erling Smørgrav [Wed, 5 Feb 2014 09:28:02 +0000 (09:28 +0000)]
Add an example of the most common use case.

MFC after: 1 week

10 years agoUse system macro instead of own hand-rolled one.
Hans Petter Selasky [Wed, 5 Feb 2014 08:29:24 +0000 (08:29 +0000)]
Use system macro instead of own hand-rolled one.

10 years agoAdd more quirks for making builtin audio speakers work with more
Hans Petter Selasky [Wed, 5 Feb 2014 08:27:02 +0000 (08:27 +0000)]
Add more quirks for making builtin audio speakers work with more
MacBookPro's. Only tested with MacBookPro 9,2.

Obtained from: Linux
MFC after: 1 week

10 years agoCorrect setting R92C_TXAGC_MCS11_MCS08 register.
Kevin Lo [Wed, 5 Feb 2014 08:09:02 +0000 (08:09 +0000)]
Correct setting R92C_TXAGC_MCS11_MCS08 register.

10 years agoFix a regression issue. Contiguous single segment allocations above
Hans Petter Selasky [Wed, 5 Feb 2014 08:02:52 +0000 (08:02 +0000)]
Fix a regression issue. Contiguous single segment allocations above
PAGE_SIZE bytes should only use one USB page structure. Fixes a
problem with some external drivers.

MFC after: 2 days

10 years agoAdd support for FreeBSD/i386 guests under bhyve.
John Baldwin [Wed, 5 Feb 2014 04:39:03 +0000 (04:39 +0000)]
Add support for FreeBSD/i386 guests under bhyve.
- Similar to the hack for bootinfo32.c in userboot, define
  _MACHINE_ELF_WANT_32BIT in the load_elf32 file handlers in userboot.
  This allows userboot to load 32-bit kernels and modules.
- Copy the SMAP generation code out of bootinfo64.c and into its own
  file so it can be shared with bootinfo32.c to pass an SMAP to the i386
  kernel.
- Use uint32_t instead of u_long when aligning module metadata in
  bootinfo32.c in userboot, as otherwise the metadata used 64-bit
  alignment which corrupted the layout.
- Populate the basemem and extmem members of the bootinfo struct passed
  to 32-bit kernels.
- Fix the 32-bit stack in userboot to start at the top of the stack
  instead of the bottom so that there is room to grow before the
  kernel switches to its own stack.
- Push a fake return address onto the 32-bit stack in addition to the
  arguments normally passed to exec() in the loader.  This return
  address is needed to convince recover_bootinfo() in the 32-bit
  locore code that it is being invoked from a "new" boot block.
- Add a routine to libvmmapi to setup a 32-bit flat mode register state
  including a GDT and TSS that is able to start the i386 kernel and
  update bhyveload to use it when booting an i386 kernel.
- Use the guest register state to determine the CPU's current instruction
  mode (32-bit vs 64-bit) and paging mode (flat, 32-bit, PAE, or long
  mode) in the instruction emulation code.  Update the gla2gpa() routine
  used when fetching instructions to handle flat mode, 32-bit paging, and
  PAE paging in addition to long mode paging.  Don't look for a REX
  prefix when the CPU is in 32-bit mode, and use the detected mode to
  enable the existing 32-bit mode code when decoding the mod r/m byte.

Reviewed by: grehan, neel
MFC after: 1 month

10 years agoAdd support for emulating the byte move and zero extend instructions:
Tycho Nightingale [Wed, 5 Feb 2014 02:01:08 +0000 (02:01 +0000)]
Add support for emulating the byte move and zero extend instructions:
"mov r/m8, r32" and "mov r/m8, r64".

Approved by: neel (co-mentor)

10 years agolibc/net: fix a portability issue
Eitan Adler [Wed, 5 Feb 2014 02:00:31 +0000 (02:00 +0000)]
libc/net: fix a portability issue

* POSIX does not require socklen_t to be  unsigned

Submitted by: bde
MFC After: 1 week (with r261454)

10 years agoFix ! by not clearing not at the bottom of the loop.
Warner Losh [Wed, 5 Feb 2014 00:26:11 +0000 (00:26 +0000)]
Fix ! by not clearing not at the bottom of the loop.
Add a blank line

Submitted by: bde (blank line)

10 years agoAdd a license (1) and do some cleanup.
Christian Brueffer [Tue, 4 Feb 2014 22:20:17 +0000 (22:20 +0000)]
Add a license (1) and do some cleanup.

Approved by: Stefan Bethke (original author, by private mail) (1)
MFC after: 1 week

10 years agoFix installations that use kernels without CAPABILITIES support.
Pawel Jakub Dawidek [Tue, 4 Feb 2014 21:48:09 +0000 (21:48 +0000)]
Fix installations that use kernels without CAPABILITIES support.

Approved by: des

10 years agoProtect ping(8) using Capsicum and Casper. This is protection against malicious
Pawel Jakub Dawidek [Tue, 4 Feb 2014 21:43:53 +0000 (21:43 +0000)]
Protect ping(8) using Capsicum and Casper. This is protection against malicious
network packets that we parse and not against local users trying to gain root
access through ping's set-uid bit - this is handled by dropping privileges very
early in ping.

Submitted by: Mikhail <mp@lenta.ru>

10 years agoAbort when firmware isn't present in R600+ models.
Robert Millan [Tue, 4 Feb 2014 21:23:12 +0000 (21:23 +0000)]
Abort when firmware isn't present in R600+ models.

More details at:
http://anonscm.debian.org/viewvc/kernel/dists/trunk/linux/debian/patches/bugfix/all/radeon-firmware-is-required-for-drm-and-kms-on-r600-onward.patch?revision=20909&view=co

Reviewed by: dumbbell
MFC after: 1 week

10 years agoUnbreak mount_udf by passing the correct iovec length into
Christian Brueffer [Tue, 4 Feb 2014 21:15:15 +0000 (21:15 +0000)]
Unbreak mount_udf by passing the correct iovec length into
nmount().  This has been broken since r247856.

PR: bin/186193
Submitted by: Arnot Belohlavek
MFC after: 1 week

10 years agoUse right buffer to print to.
Ed Schouten [Tue, 4 Feb 2014 20:52:33 +0000 (20:52 +0000)]
Use right buffer to print to.

PR: kern/176597
Submitted by: Christoph Mallon <christoph mallon gmx de>
MFC after: 2 weeks

10 years agoActually install acpi_rapidstart.4.
Christian Brueffer [Tue, 4 Feb 2014 18:54:33 +0000 (18:54 +0000)]
Actually install acpi_rapidstart.4.

MFC after: 1 week

10 years agoImplement the '!' operator for files* files. It means 'include this
Warner Losh [Tue, 4 Feb 2014 18:28:58 +0000 (18:28 +0000)]
Implement the '!' operator for files* files. It means 'include this
only if the specified option is NOT specified.' Bump version because
old config won't be able to cope with files* files that have this
construct in them.

10 years agoBump the version of config to the latest (3 year old, so upgrade
Warner Losh [Tue, 4 Feb 2014 18:24:25 +0000 (18:24 +0000)]
Bump the version of config to the latest (3 year old, so upgrade
worries are long past). Also remove redundant MACHINE= declarations
and passing MACHINE/MACHINE_ARCH to module builds. That's now done in
common code.

10 years agoAdd a tunable "hw.mfi.mrsas_enable" to allow mfi(4) to drop priority and
Doug Ambrisko [Tue, 4 Feb 2014 17:35:41 +0000 (17:35 +0000)]
Add a tunable "hw.mfi.mrsas_enable" to allow mfi(4) to drop priority and
allow mrsas(4) from LSI to attach to newer LSI cards that are support by
mrsas(4).  If mrsas(4) is not loaded into the system at boot then mfi(4)
will always attach.  If a modified mrsas(4) is loaded in the system.  That
modification is return "-30" in it's probe since that is between
BUS_PROBE_DEFAULT and BUS_PROBE_LOW_PRIORITY.

This option is controller by a new probe flag "MFI_FLAGS_MRSAS" in mfi_ident
that denotes cards that should work with mrsas(4).  New entries that should
have this option.

This is the first step to get mrsas(4) checked into FreeBSD and to avoid
collision with people that use mrsas(4) from LSI.  Since mfi(4) takes
priority, then mrsas(4) users need to rebuild GENERIC.  Using the
.disabled="1" method doesn't work since that blocks attaching and the
probe gave it to mfi(4).

Discussed with: LSI (Kashyap Desai)

10 years agoAdd qlxgbe(4) and qlxge(4) to the hardware notes.
Christian Brueffer [Tue, 4 Feb 2014 12:34:08 +0000 (12:34 +0000)]
Add qlxgbe(4) and qlxge(4) to the hardware notes.

MFC after: 1 week

10 years agoRegenerate usb.conf
Hans Petter Selasky [Tue, 4 Feb 2014 09:11:50 +0000 (09:11 +0000)]
Regenerate usb.conf

MFC after: 2 days

10 years agoPass MACHINE and MACHINE_ARCH down into the modules
Warner Losh [Tue, 4 Feb 2014 05:35:04 +0000 (05:35 +0000)]
Pass MACHINE and MACHINE_ARCH down into the modules

10 years agoRemove trailing tabs causing false grep positives
Warner Losh [Tue, 4 Feb 2014 05:26:12 +0000 (05:26 +0000)]
Remove trailing tabs causing false grep positives

10 years agos/standard/optional/ for ohci and echi, since these files are optional
Warner Losh [Tue, 4 Feb 2014 05:21:57 +0000 (05:21 +0000)]
s/standard/optional/ for ohci and echi, since these files are optional
and not standard.

10 years agoAdd a prior version compat define.
Warner Losh [Tue, 4 Feb 2014 03:59:35 +0000 (03:59 +0000)]
Add a prior version compat define.

10 years agolibc/net: Fix some issues in inet6_opt_init() (from RFC 3542):
Eitan Adler [Tue, 4 Feb 2014 03:01:33 +0000 (03:01 +0000)]
libc/net: Fix some issues in inet6_opt_init() (from RFC 3542):

* The RFC says (in section 10.1) that only when extbuf is not NULL,
extlen shall be checked, so don't perform this check when NULL is
passed.

* socklen_t is unsigned, so checking extlen for less than zero is
not needed.

Submitted by: swildner@dragonflybsd.org
Reviewed by: Mark Martinec <Mark.Martinec+freebsd@ijs.si>
Reviewed by: hrs
Obtained by: DragonFlyBSD

10 years agoAvoid doing unnecessary nested TLB invalidations.
Neel Natu [Tue, 4 Feb 2014 02:45:08 +0000 (02:45 +0000)]
Avoid doing unnecessary nested TLB invalidations.

Prior to this change the cached value of 'pm_eptgen' was tracked per-vcpu
and per-hostcpu. In the degenerate case where 'N' vcpus were sharing
a single hostcpu this could result in 'N - 1' unnecessary TLB invalidations.
Since an 'invept' invalidates mappings for all VPIDs the first 'invept'
is sufficient.

Fix this by moving the 'eptgen[MAXCPU]' array from 'vmxctx' to 'struct vmx'.

If it is known that an 'invept' is going to be done before entering the
guest then it is safe to skip the 'invvpid'. The stat VPU_INVVPID_SAVED
counts the number of 'invvpid' invalidations that were avoided because
they were subsumed by an 'invept'.

Discussed with: grehan

10 years agoFix lock acquisition in case no request space available, missed in r260097.
Alexander Motin [Tue, 4 Feb 2014 00:00:01 +0000 (00:00 +0000)]
Fix lock acquisition in case no request space available, missed in r260097.

MFC after: 3 days

10 years agoFix a typo.
Christian Brueffer [Mon, 3 Feb 2014 22:16:46 +0000 (22:16 +0000)]
Fix a typo.

MFC after: 1 week

10 years agoConvert the loop by gotos into a for loop to improve readability. I
Warner Losh [Mon, 3 Feb 2014 19:14:36 +0000 (19:14 +0000)]
Convert the loop by gotos into a for loop to improve readability. I
did this only with the inner loop for the token parsing, and not the
outer loop which was understandable enough when the extra layers of
looping went away...

10 years agoFix a bug introduced in r261437 that failed to honor "optional
Warner Losh [Mon, 3 Feb 2014 19:10:33 +0000 (19:10 +0000)]
Fix a bug introduced in r261437 that failed to honor "optional
profiling-routine" to work, since profiling-routine is not really an
option or a device, but a special case elsewhere in the code.

10 years agoSlight cleanup to the error messaging to compress code vertically...
Warner Losh [Mon, 3 Feb 2014 18:56:41 +0000 (18:56 +0000)]
Slight cleanup to the error messaging to compress code vertically...

10 years agoOpenSSH been updated to 6.5p1.
Steven Kreuzer [Mon, 3 Feb 2014 18:44:36 +0000 (18:44 +0000)]
OpenSSH been updated to 6.5p1.
bmake has been updated to version 20140101.

Approved by: hrs (mentor)

10 years agoBetter error messages when EOF is hit in the middle of a phrase.
Warner Losh [Mon, 3 Feb 2014 18:31:51 +0000 (18:31 +0000)]
Better error messages when EOF is hit in the middle of a phrase.

10 years agoFix a logic error. Because of this inflateReset() wasn't being called and
Luiz Otavio O Souza [Mon, 3 Feb 2014 17:25:36 +0000 (17:25 +0000)]
Fix a logic error.  Because of this inflateReset() wasn't being called and
the output buffer wasn't being cleared between the inflate() calls,
producing zeroed output after the first inflate() call.

This fixes the read of mkuzip(8) images with geom_uncompress(4).

Reviewed by: ray
Approved by: adrian (mentor)

10 years agoRemove some unnecessary code. The offsets read from the first block are
Luiz Otavio O Souza [Mon, 3 Feb 2014 17:21:36 +0000 (17:21 +0000)]
Remove some unnecessary code.  The offsets read from the first block are
overwritten a few lines bellow.

Reviewed by: ray
Approved by: adrian (mentor)

10 years agoMove the check for standard keyword + optional inclusion specifier to
Warner Losh [Mon, 3 Feb 2014 16:54:53 +0000 (16:54 +0000)]
Move the check for standard keyword + optional inclusion specifier to
its proper location. Otherwise you could have 'file.c standard pci'
without an error. This construct isn't in our tree, and has no well
defined meaning.

10 years agoDon't believe we have a requirement until after we've checked all the
Warner Losh [Mon, 3 Feb 2014 16:47:10 +0000 (16:47 +0000)]
Don't believe we have a requirement until after we've checked all the
known key words. This will make error messages slightly better in
weird corner cases, but should otherwise be a nop.

10 years agoIn the 17 years since r30796, the mandatory keyword has never been used
Warner Losh [Mon, 3 Feb 2014 16:46:01 +0000 (16:46 +0000)]
In the 17 years since r30796, the mandatory keyword has never been used
in any files as far as I can tell, and is currently unused. Retire it.

10 years agoSlightly deobfuscate read_file() and likely pessimize the runtime
Warner Losh [Mon, 3 Feb 2014 15:10:44 +0000 (15:10 +0000)]
Slightly deobfuscate read_file() and likely pessimize the runtime
performance by epsilon.
(Translation: elminate bogus macros that hid 'returns' making it hard
 to read and moved a block of code inline rather than at the end of the
 fuction where it was effectively a 'gosub' kind of goto).

10 years agoFix the definition of hg_cmd.
Rui Paulo [Mon, 3 Feb 2014 08:15:09 +0000 (08:15 +0000)]
Fix the definition of hg_cmd.

MFC after: 3 days

10 years agoApply patch for CVE-2013-6393 [1] to fix heap-based buffer overflow when
Baptiste Daroussin [Mon, 3 Feb 2014 08:13:44 +0000 (08:13 +0000)]
Apply patch for CVE-2013-6393 [1] to fix heap-based buffer overflow when
parsing YAML tags.
Also apply a patch for hardenning the guards againt the issue

The only user in base in yaml is pkg(7) which uses the library a way that it is not affected

Submitted by: delphij
Obtained from: https://bugzilla.redhat.com/show_bug.cgi?id=1033990
MFC after: 3 days
Security: CVE-2013-6393

10 years agoApply patch for CVE-2013-6393 [1] to fix heap-based buffer overflow when
Baptiste Daroussin [Mon, 3 Feb 2014 08:00:45 +0000 (08:00 +0000)]
Apply patch for CVE-2013-6393 [1] to fix heap-based buffer overflow when
parsing YAML tags.
Also apply a patch for hardenning the guards againt the issue

The only user in base in yaml is pkg(7) which uses the library a way that it is not affected

Submitted by: delphij
Obtained from: https://bugzilla.redhat.com/show_bug.cgi?id=1033990
Security: CVE-2013-6393

10 years agodhclient: change the pidfile's permissions to 644
Eitan Adler [Mon, 3 Feb 2014 04:22:29 +0000 (04:22 +0000)]
dhclient: change the pidfile's permissions to 644

This change permits non-root users to determine if dhclient is running
('service dhclient status wlan0').

Discussed with: mjg, cperciva

10 years agoSort the list.
Ian Lepore [Mon, 3 Feb 2014 02:56:23 +0000 (02:56 +0000)]
Sort the list.

10 years agoAdd the imx sdhci controller.
Ian Lepore [Mon, 3 Feb 2014 02:52:07 +0000 (02:52 +0000)]
Add the imx sdhci controller.

10 years agoMake gas accept any PowerPC instruction by default. This is a local change,
Justin Hibbits [Mon, 3 Feb 2014 01:45:07 +0000 (01:45 +0000)]
Make gas accept any PowerPC instruction by default.  This is a local change,
and will not be submitted upstream.

Discussed with: nwhitehorn,rdivacky
MFC after: 1 month

10 years agoAdd driver for the ADT7460/ADT7467 fan controller found in later PowerBooks
Justin Hibbits [Mon, 3 Feb 2014 01:22:50 +0000 (01:22 +0000)]
Add driver for the ADT7460/ADT7467 fan controller found in later PowerBooks
and iBooks.  Original work by andreast.

MFC after: 1 month

10 years agoAdd missing file to Makefile.
Justin Hibbits [Mon, 3 Feb 2014 01:16:32 +0000 (01:16 +0000)]
Add missing file to Makefile.

MFC after: 1 month
X-MFC-with: 261342

10 years agoOnly use the CPU ID register if SMP is defined. Some non-MPCore armv6 cpu,
Olivier Houchard [Sun, 2 Feb 2014 23:29:51 +0000 (23:29 +0000)]
Only use the CPU ID register if SMP is defined. Some non-MPCore armv6 cpu,
such as the one found in the RPi, don't have it, and just hang when we try
to access it.

10 years agoInvalidate cachelines for bounce pages on PREREAD too, there may still be
Olivier Houchard [Sun, 2 Feb 2014 22:26:30 +0000 (22:26 +0000)]
Invalidate cachelines for bounce pages on PREREAD too, there may still be
stale entries from a previous transfer.

10 years agoAdd missing semicolon.
Ian Lepore [Sun, 2 Feb 2014 21:44:04 +0000 (21:44 +0000)]
Add missing semicolon.

10 years agoo Add prototype for tcon_bypass() used by dcu4
Ruslan Bukin [Sun, 2 Feb 2014 21:10:40 +0000 (21:10 +0000)]
o Add prototype for tcon_bypass() used by dcu4
o Add register definition

10 years agoChange the way pcpu and curthread are stored per-core:
Olivier Houchard [Sun, 2 Feb 2014 20:58:23 +0000 (20:58 +0000)]
Change the way pcpu and curthread are stored per-core:
the old way was to store pcpu in a register, and get curthread from pcpu,
which is not very atomic, and led to issues if the thread was migrated
to another core between the time we got the pcpu address and the time we
got curthread.
Instead, we now store curthread where pcpu used to be store, and we
calculate the pcpu address based on the cpu id.

10 years agoDon't call device_set_ivars() for the mmchs, it doesn't seem to be used,
Olivier Houchard [Sun, 2 Feb 2014 20:45:41 +0000 (20:45 +0000)]
Don't call device_set_ivars() for the mmchs, it doesn't seem to be used,
and it overrides the ivars set by the simplebus.

10 years agoAdd driver for Display Control Unit (DCU4).
Ruslan Bukin [Sun, 2 Feb 2014 20:25:27 +0000 (20:25 +0000)]
Add driver for Display Control Unit (DCU4).

10 years agoMake prefaulting more aggressive on hard faults. Previously, we would only
Alan Cox [Sun, 2 Feb 2014 20:21:53 +0000 (20:21 +0000)]
Make prefaulting more aggressive on hard faults.  Previously, we would only
map a fraction of the pages that were fetched by vm_pager_get_pages() from
secondary storage.  Now, we map them all in order to avoid future soft
faults.  This effect is most evident when a memory-mapped file is accessed
sequentially.  Previously, there were 6 soft faults for every hard fault.
Now, these soft faults are eliminated.

Sponsored by: EMC / Isilon Storage Division

10 years agoAdd support for Colibri VF50 Evaluation Board.
Ruslan Bukin [Sun, 2 Feb 2014 19:35:10 +0000 (19:35 +0000)]
Add support for Colibri VF50 Evaluation Board.

Colibri VF50 is a SODIMM200 Vybrid Family core module
and development board produced by Toradex AG.

Sponsored by: Machdep, Inc.

10 years agoFollow r261352 by updating all drivers which are children of simplebus
Ian Lepore [Sun, 2 Feb 2014 19:17:28 +0000 (19:17 +0000)]
Follow r261352 by updating all drivers which are children of simplebus
to check the status property in their probe routines.

Simplebus used to only instantiate its children whose status="okay"
but that was improper behavior, fixed in r261352.  Now that it doesn't
check anymore and probes all its children; the children all have to
do the check because really only the children know how to properly
interpret their status property strings.

Right now all existing drivers only understand "okay" versus something-
that's-not-okay, so they all use the new ofw_bus_status_okay() helper.

10 years agoSplit kernel configuration to chip common and board specific parts.
Ruslan Bukin [Sun, 2 Feb 2014 19:13:02 +0000 (19:13 +0000)]
Split kernel configuration to chip common and board specific parts.

10 years agoAssert input arguments to buf_send() and buf_recv().
Pawel Jakub Dawidek [Sun, 2 Feb 2014 19:06:00 +0000 (19:06 +0000)]
Assert input arguments to buf_send() and buf_recv().

Submitted by: Mariusz Zaborski <oshogbo@FreeBSD.org>

10 years agoFix sending empty nvlist.
Pawel Jakub Dawidek [Sun, 2 Feb 2014 19:03:52 +0000 (19:03 +0000)]
Fix sending empty nvlist.

Submitted by: Mariusz Zaborski <oshogbo@FreeBSD.org>

10 years agoo Expand device tree information
Ruslan Bukin [Sun, 2 Feb 2014 17:48:06 +0000 (17:48 +0000)]
o Expand device tree information
o Export iomuxc (pins) configuration to DTS
o Allow devices to assign clocks in DTS

10 years agoBe more robust with malformed interrupt config data. Instead of crashing
Ian Lepore [Sun, 2 Feb 2014 16:56:40 +0000 (16:56 +0000)]
Be more robust with malformed interrupt config data.  Instead of crashing
or going into a near-infinite loop, warn and make potentially-reasonable
assumptions.

10 years agoFix one remnant endian flaw here. The back-and-forth endian conversions are
Nathan Whitehorn [Sun, 2 Feb 2014 16:42:26 +0000 (16:42 +0000)]
Fix one remnant endian flaw here. The back-and-forth endian conversions are
confusing.

10 years agoAdd a set of helpers (ofw_bus_get_status() and ofw_bus_status_okay()) to
Nathan Whitehorn [Sun, 2 Feb 2014 16:41:54 +0000 (16:41 +0000)]
Add a set of helpers (ofw_bus_get_status() and ofw_bus_status_okay()) to
process "status" properties of OF nodes.

I've avoided adding new KOBJ methods here so that we don't have to modify
every ofw_bus in the tree. Since 100% of implementations of ofw_bus use
only ofw_bus_gen_*(), it might be worth garbage-collecting the other
methods as well.

10 years agoAdd back reference to buf(9) removed in r32223 as buf(9) was added
Bryan Drewery [Sun, 2 Feb 2014 14:13:51 +0000 (14:13 +0000)]
Add back reference to buf(9) removed in r32223 as buf(9) was added
in r42016.

Approved by: bapt (mentor)

10 years agoFix newsyslog(8) to use the size of the file instead of the blocks it
Bryan Drewery [Sun, 2 Feb 2014 14:11:34 +0000 (14:11 +0000)]
Fix newsyslog(8) to use the size of the file instead of the blocks it
takes on disk, as advertised in newsyslog.conf(5).

This fixes newsyslog(8) on ZFS with compression enabled to not have
large files compared to the expected rotation size.

  # grep remotes/messages /etc/newsyslog.conf
  /var/log/remote/messages       root:info       640  5     500  *     JC
  # ls -alh /var/log/remote/messages
  -rw-r-----  1 root  info    3.2M Jan 31 20:02 /var/log/remote/messages
  # newsyslog -vN|grep remote/messages
  /var/log/remote/messages <5J>: size (Kb): 464 [500] --> skipping
  # stat -f "st_size: %z st_blocks: %b" /var/log/remote/messages
  st_size: 3372627 st_blocks: 928
  # zfs get -H compressratio zroot/syslogs
  zroot/syslogs    compressratio   3.77x   -

With fix:

  # newsyslog -v | grep remote/messages
  /var/log/remote/messages <5J>: size (Kb): 3338 [500] --> trimming log....

Approved by: bapt (mentor)
PR: docs/150877
Reported by: Joshua Isom <jrisom@gmail.com>
MFC after: 2 weeks

10 years agoTake exclusive lock only when lle isn't NULL. We don't need write access
Andrey V. Elsukov [Sun, 2 Feb 2014 07:28:04 +0000 (07:28 +0000)]
Take exclusive lock only when lle isn't NULL. We don't need write access
to lle in most cases.

MFC after: 1 week
Sponsored by: Yandex LLC

10 years agoFix missing offset.
Nathan Whitehorn [Sun, 2 Feb 2014 05:52:34 +0000 (05:52 +0000)]
Fix missing offset.

10 years agoAllow nesting of simplebuses.
Nathan Whitehorn [Sun, 2 Feb 2014 05:41:12 +0000 (05:41 +0000)]
Allow nesting of simplebuses.

10 years agoFix typo. Sorry for breakage!
Nathan Whitehorn [Sun, 2 Feb 2014 05:21:12 +0000 (05:21 +0000)]
Fix typo. Sorry for breakage!

10 years agoUse m_defrag() instead of m_collapse() to compact a long mbuf chain
Bryan Venteicher [Sun, 2 Feb 2014 05:20:46 +0000 (05:20 +0000)]
Use m_defrag() instead of m_collapse() to compact a long mbuf chain

This should be an infrequent occurrence, so remove the per-queue
counters in favor of just global counters in the softc.

10 years agoDo not place the sglist used for Rx/Tx on the stack
Bryan Venteicher [Sun, 2 Feb 2014 05:15:36 +0000 (05:15 +0000)]
Do not place the sglist used for Rx/Tx on the stack

The sglist segment array has grown to a bit over 512 bytes (on
64-bit system) which is more than ideally should be put on the
stack. Instead allocate an appropriately sized sglist and hang
it off each Rx/Tx queue structure.

Bump the maximum number of Tx segments to 64 to make it unlikely
we'll have defragment an mbuf chain. Our previous count was
rounded up to this value since it is the next power of two, so
effective memory usage should not change.

Also only allocate the maximum number of Tx segments if TSO was
negotiated.

10 years agoUpdate all arm code that manipulates the PSR registers to use modern syntax.
Ian Lepore [Sun, 2 Feb 2014 00:48:15 +0000 (00:48 +0000)]
Update all arm code that manipulates the PSR registers to use modern syntax.

It turns out the version of gas we're using interprets the old '_all' mask
as 'fc' instead of 'fsxc'.  That is, "all" doesn't really mean "all".

This was the cause of the "wrong-endian register restore" bug that's
been causing problems with some cortex-a9 chips.  The 'endian' bit in the
spsr register would never get changed (it falls into the 'x' mask group)
and the first return-from-exception would fail if the chip had powered on
with garbage in the spsr register that included the big-endian bit.  It's
unknown why this affected only certain cortex-a9 chips.

10 years agoSwitch default Book-E scheduler to ULE, which works now, and enable
Nathan Whitehorn [Sat, 1 Feb 2014 20:56:50 +0000 (20:56 +0000)]
Switch default Book-E scheduler to ULE, which works now, and enable
CAPABILITIES stuff required to make ssh work.

Hopefully, Book-E can eventually be added to GENERIC, which would avoid
this kind of issue with bitrot. That will require figuring out how to link
Book-E and AIM kernels at the same address, however...

10 years agoULE works on Book-E since r258002, so remove statements to the contrary.
Nathan Whitehorn [Sat, 1 Feb 2014 20:46:35 +0000 (20:46 +0000)]
ULE works on Book-E since r258002, so remove statements to the contrary.

10 years agoAvoid spurious compiler warning about an uninitialized variable.
Nathan Whitehorn [Sat, 1 Feb 2014 20:06:52 +0000 (20:06 +0000)]
Avoid spurious compiler warning about an uninitialized variable.

10 years agoFix typo.
Nathan Whitehorn [Sat, 1 Feb 2014 19:25:15 +0000 (19:25 +0000)]
Fix typo.

10 years agoIn _pthread_kill(), if passed pthread is current thread, do not send
Konstantin Belousov [Sat, 1 Feb 2014 18:13:18 +0000 (18:13 +0000)]
In _pthread_kill(), if passed pthread is current thread, do not send
the signal second time, by adding the missed else before if statement.

While there, postpone initializing local curthread variable until
passed signal number is checked for validity.

Submitted by: John Wolfe <jlw@xinuos.com>
PR: threads/186309
MFC after: 1 week

10 years agoFix AT91SAM9260 to work with PA rather than VA device addresses.
Warner Losh [Sat, 1 Feb 2014 17:53:35 +0000 (17:53 +0000)]
Fix AT91SAM9260 to work with PA rather than VA device addresses.

10 years agoProvide a simpler and more standards-compliant simplebus implementation to
Nathan Whitehorn [Sat, 1 Feb 2014 17:41:54 +0000 (17:41 +0000)]
Provide a simpler and more standards-compliant simplebus implementation to
get the Routerboard 800 up and running with the vendor device tree. This
does not implement some BERI-specific features (which hopefully won't be
necessary soon), so move the old code to mips/beri, with a higher attach
priority when built, until MIPS interrupt domain support is rearranged.

10 years agoOpen Firmware interrupt specifiers can consist of arbitrary-length byte
Nathan Whitehorn [Sat, 1 Feb 2014 17:17:35 +0000 (17:17 +0000)]
Open Firmware interrupt specifiers can consist of arbitrary-length byte
strings and include arbitrary information (IRQ line/domain/sense). When the
ofw_bus_map_intr() API was introduced, it assumed that, as on most systems,
these were either 1 cell, containing an interrupt line, or 2, containing
a line number plus a sense code. It turns out a non-negligible number of
ARM systems use 3 (or even 4!) cells for interrupts, so make this more
general.

10 years agoRemove the .Ex macro that I used for testing.
Christian Brueffer [Sat, 1 Feb 2014 12:33:58 +0000 (12:33 +0000)]
Remove the .Ex macro that I used for testing.

Pointy hat: brueffer

10 years agoBring the exit status wording closer to what .Ex would produce.
Christian Brueffer [Sat, 1 Feb 2014 12:30:00 +0000 (12:30 +0000)]
Bring the exit status wording closer to what .Ex would produce.
Fixes a typo in the process.

MFC after: 1 week

10 years agoAdd definition for NetBSD 7.0, which is referenced in several manpages.
Sergey Kandaurov [Sat, 1 Feb 2014 10:36:35 +0000 (10:36 +0000)]
Add definition for NetBSD 7.0, which is referenced in several manpages.

Discussed with: uqs
MFC after: 5 days