]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
12 years agonetmap-related changes:
luigi [Thu, 19 Jan 2012 09:36:19 +0000 (09:36 +0000)]
netmap-related changes:
1. correct the initialization of RDT when there is an ixgbe_init()
   while a netmap client is active. This code was previously
   in ixgbe_initialize_receive_units() but RDT is overwritten
   shortly afterwards in ixgbe_init_locked()

2. add code (not active yet) to disable CRCSTRIP while in netmap mode.
   From all evidence i could gather, it seems that when the 82599 has to
   write a data block that is not a full cache line, it first reads
   the line (64 bytes) and then writes back the updated version.
   This hurts reception of min-sized frames, which are only 60 bytes
   if the CRC is stripped: i could never get above 11Mpps
   (received from one queue) with CRCSTRIP enabled, whyle 64+4-byte
   packets reach 14.2 Mpps (the theoretical maximum).
   Leaving the CRC in gets us 14.88Mpps for 60+4 byte frames,
   (and penalizes 64+4). The min-size case is important not just because
   it looks good in benchmarks, but also because this is the size
   of pure acks.
   Note we cannot leave CRCSTRIP on by default because it is
   incompatible with some other features (LRO etc.)

12 years agoAdd missing MLINKS to INDEX.5
kevlo [Thu, 19 Jan 2012 02:49:21 +0000 (02:49 +0000)]
Add missing MLINKS to INDEX.5

12 years agoAdd support for Sony Ericsson GC89 EDGE/Wirelles LAN PC Card
eadler [Thu, 19 Jan 2012 02:47:11 +0000 (02:47 +0000)]
Add support for Sony Ericsson GC89 EDGE/Wirelles LAN PC Card

PR: kern/131933
Submitted by: Alex Keda <admin@lissyara.su>
Approved by: jhb
Approved by: cperciva (mentor, blanket for pre-mentorship already-approved commits)
MFC after: 1 week

12 years agoTwo 192/24/8 playback streams overflow single mandatory output line (SDO)
mav [Thu, 19 Jan 2012 01:55:48 +0000 (01:55 +0000)]
Two 192/24/8 playback streams overflow single mandatory output line (SDO)
of HDA bus. Handle that from two directions:
 - Add support for "striping" (using several SDO lines), if supported.
 - Account HDA bus utilization and return error on new stream allocation
attempt if remaining bandwidth is unsifficient.

Most of HDA controllers have one SDO line with 46Mbps output bandwidth.
NVIDIA GF210 has 2 lines - 92Mbps. NVIDIA GF520 has 4 lines - 184Mbps!

MFC after: 2 months
Sponsored by: iXsystems, Inc.

12 years agoWrap changes from svn r212126 inside LOADER_NFS_SUPPORT such that
sbruno [Thu, 19 Jan 2012 01:25:50 +0000 (01:25 +0000)]
Wrap changes from svn r212126 inside LOADER_NFS_SUPPORT such that
using LOADER_TFTP_SUPPORT excludes this code.  Fixes compilation of pxeldr
with -DLOADER_TFTP_SUPPORT

Applicable to stable/9 and stable/8 now.

This appears to not be needed on stable/7 as r212126 has not been MFC'd.

Obtained from: Yahoo! Inc.
MFC after: 2 weeks

12 years agoUpdate .Dd date on this man page.
sbruno [Wed, 18 Jan 2012 22:52:02 +0000 (22:52 +0000)]
Update .Dd date on this man page.

MFC with r230313 and r230316

Caught by: bz
Obtained from: Yahoo! Inc.

12 years agoUpdate man page to show that ciss(4) now support the P420
sbruno [Wed, 18 Jan 2012 21:27:49 +0000 (21:27 +0000)]
Update man page to show that ciss(4) now support the P420

MFC with r230313

Caught by:  brueffer
Obtained from: Yahoo! Inc
MFC after: 2 weeks

12 years agoAdd support for HP P420 to ciss(4)
sbruno [Wed, 18 Jan 2012 19:35:16 +0000 (19:35 +0000)]
Add support for HP P420 to ciss(4)

Tested on upcoming Gen 8 releases of hardware from HP.

MFC to all supported releases.

Obtained from: Yahoo! Inc.
MFC after: 2 weeks

12 years agoImprove HDMI/DisplayPort audio support in snd_hda(4):
mav [Wed, 18 Jan 2012 19:12:33 +0000 (19:12 +0000)]
Improve HDMI/DisplayPort audio support in snd_hda(4):

 - Enable and handle unsolicited responses from digital display pins,
reporting connection and EDID-Like Data (ELD) validity status changes.
 - Fetch ELD data, describing connected digital display device audio
capabilities. These data not really used at the moment (user is not
denied to use audio formats not supported by the device), only printed to
verbose logs. But they are useful for debugging. The fact that ELD was
received tells that HDMI link was established and video driver enabled
HDMI audio passthrough. Some old chips may not return ELD, so lack of it
is not necessary a problem.
 - Add some more points to CODEC configuration sequence:
   - For converter widgets, supporting more then two channels (HDMI/DP
     converter widgets support 8), set number of channels to handle.
   - For digital display pins (HDMI/DP) fill audio infoframe, reporting
     connected device about number of channels and speakers allocation.
   - For digital display pins (HDMI/DP) set mapping between channels seen
     by software and channels transferred via HDMI/DisplayPort.
 - Allow more audio formats, not used for analog connections because of
stereo pairs orientation, but easily applicable to HDMI/DisplayPort: 2.1,
3.0, 3.1, 4.1, 5.0, 6.0, 6.1, 7.0. That list may be filtered later using
info from ELD.
 - Disable MSI interrupts for NVIDIA HDA controllers before GT520.

At this point I can successfully play audio over HDMI from NVIDIA GT210
and GT520 cards with nvidia-driver-290.10 driver to Marantz SR4001
receiver in 2.0, 2.1, 3.0, 4.0, 4.1, 5.0 and 5.1 PCM formats at 44, 48,
88 and 96KHz at 16 and 24 bits, same as do AC3/DTS passthrough.
6.0, 6.1, 7.0 and 7.1 PCM formats are not working for me, but I think
it is because of receiver age.

MFC after: 2 months
Sponsored by: iXsystems, Inc.

12 years agoRev 228065 (change bsd.own.mk -> bsd.init.mk) broke pam_unix.so by causing
peter [Wed, 18 Jan 2012 18:26:56 +0000 (18:26 +0000)]
Rev 228065 (change bsd.own.mk -> bsd.init.mk) broke pam_unix.so by causing
the LDADD/DPADD to lose the -lpam, and causing openpam_dynamic() to fail
due to "openpam_get_options" being undefined.

This would cause obscure console log messages like:
  openpam_dynamic(): No error: 0
  openpam_load_module(): no pam_unix.so found
and other helpful messages which are no help in diagnosing the problem.

Fortunately this change was not mfc'ed to 9.x, it isn't broken there.

12 years agoWarn if trying to install over an existing partition, which usually fails
nwhitehorn [Wed, 18 Jan 2012 15:59:23 +0000 (15:59 +0000)]
Warn if trying to install over an existing partition, which usually fails
anyway due to libarchive not being able to overwrite schg flags.

PR: bin/164278
MFC after: 4 days

12 years agoUnbreak several mips kernel configs after r230150 and r230152 to make
bz [Wed, 18 Jan 2012 15:25:34 +0000 (15:25 +0000)]
Unbreak several mips kernel configs after r230150 and r230152 to make
a universe complete successfully again.

12 years agoFix two issues related to the use of SIGINFO in fetch(1) to display
des [Wed, 18 Jan 2012 15:13:21 +0000 (15:13 +0000)]
Fix two issues related to the use of SIGINFO in fetch(1) to display
progress information.  The first is that fetch_read() (used in the HTTP
code but not the FTP code) can enter an infinite loop if it has previously
been interrupted by a signal.  The second is that when it is interrupted,
fetch_read() will discard any data it may have read up to that point.
Luckily, both bugs are extremely timing-sensitive and therefore difficult
to trigger.

PR: bin/153240
Submitted by: Mark <markjdb@gmail.com>
MFC after: 3 weeks

12 years agoComment cosmetics: end more sentences with full stops.
schweikh [Wed, 18 Jan 2012 14:41:26 +0000 (14:41 +0000)]
Comment cosmetics: end more sentences with full stops.

12 years agoSubject: NULLFS: properly destroy node hash
rea [Wed, 18 Jan 2012 11:23:46 +0000 (11:23 +0000)]
Subject: NULLFS: properly destroy node hash

Use hashdestroy() instead of naive free().

Approved by: kib
MFC after: 2 weeks

12 years agoConnect the bktr module(s) to the amd64 build.
schweikh [Wed, 18 Jan 2012 08:56:26 +0000 (08:56 +0000)]
Connect the bktr module(s) to the amd64 build.

MFC after: 7 days

12 years agoFix #ifdef VERBOSE for UTC case
emaste [Wed, 18 Jan 2012 04:37:35 +0000 (04:37 +0000)]
Fix #ifdef VERBOSE for UTC case

Update verbose output to match what actually happens when selecting
the UTC option, and when the -n option is used.

Patch updated slightly for new libdialog.

PR: bin/164042
Submitted by: Devin Teske <dteske vicor com>
MFC after: 2 weeks

12 years agoDon't write /var/db/zoneinfo when zone is not actually changed.
emaste [Wed, 18 Jan 2012 04:12:32 +0000 (04:12 +0000)]
Don't write /var/db/zoneinfo when zone is not actually changed.

If the specified zone file does not exist or the -n flag is specified,
do not update /var/db/zoneinfo.

PR: bin/164039
Submitted by: Devin Teske <dteske vicor com>
MFC after: 1 week

12 years agoSpecify the correct section (4 instead of 9) in the h_ertt man page's title and
lstewart [Wed, 18 Jan 2012 03:03:21 +0000 (03:03 +0000)]
Specify the correct section (4 instead of 9) in the h_ertt man page's title and
bump the document date.

MFC after: 3 days

12 years agoAdd missing line continuation \. It did not cause any issue because
emaste [Wed, 18 Jan 2012 02:44:22 +0000 (02:44 +0000)]
Add missing line continuation \.  It did not cause any issue because
the same path is already being included in ../Makefile.inc.

PR: 164192
Submitted by: Devin Teske <dteske vicor com>
MFC after: 2 weeks

12 years agoNOOBJ is long time dead
kevlo [Wed, 18 Jan 2012 02:29:11 +0000 (02:29 +0000)]
NOOBJ is long time dead

Reviewed by: kib

12 years agovfs_object_create() function is obsolete. Use vnode_create_vobject()
kevlo [Wed, 18 Jan 2012 02:27:09 +0000 (02:27 +0000)]
vfs_object_create() function is obsolete. Use vnode_create_vobject()
to create the backing object.

Reviewed by: kib

12 years agoAllow growfs to be built with GCC 4.7 and -Werror.
ed [Tue, 17 Jan 2012 22:19:35 +0000 (22:19 +0000)]
Allow growfs to be built with GCC 4.7 and -Werror.

The dp1 variable is only used when FSIRAND is defined. Just place the
variable behind #ifdefs entirely.

12 years agoDocument dev.bge.%d.msi_disable tunable.
yongari [Tue, 17 Jan 2012 22:17:11 +0000 (22:17 +0000)]
Document dev.bge.%d.msi_disable tunable.

12 years agoRemove unused variables.
ed [Tue, 17 Jan 2012 22:17:10 +0000 (22:17 +0000)]
Remove unused variables.

Simply annotate the function parameters with __unused, instead of adding
the ve-variables. This makes the code build with GCC 4.7 and -Werror.

12 years agoIntroduce a tunable that disables use of MSI.
yongari [Tue, 17 Jan 2012 22:15:33 +0000 (22:15 +0000)]
Introduce a tunable that disables use of MSI.
Non-zero value will use INTx.

12 years agoRemove unneeded dtv variable.
ed [Tue, 17 Jan 2012 21:55:20 +0000 (21:55 +0000)]
Remove unneeded dtv variable.

It is only assigned and not used at all. The object files stay identical
when the variables are removed.

Approved by: kib

12 years agoRevert changes in r228790. It prevented the ypserv daemon from working with
hrs [Tue, 17 Jan 2012 20:39:33 +0000 (20:39 +0000)]
Revert changes in r228790.  It prevented the ypserv daemon from working with
with multiple socktypes.

12 years agoOnly use the static assertion when __generic is available.
ed [Tue, 17 Jan 2012 20:22:10 +0000 (20:22 +0000)]
Only use the static assertion when __generic is available.

Reported by: tijl

12 years agoDon't expose __generic() when not using C++.
ed [Tue, 17 Jan 2012 20:21:31 +0000 (20:21 +0000)]
Don't expose __generic() when not using C++.

According to the GCC documentation, the constructs used to implement
<tgmath.h> are only available in C mode. They only cause breakage when
used used with g++.

Reported by: tijl

12 years agoFree allocated jumbo buffers when controller is stopped.
yongari [Tue, 17 Jan 2012 19:36:53 +0000 (19:36 +0000)]
Free allocated jumbo buffers when controller is stopped.

12 years agoUse a RX DMA tag to free loaded RX DMA maps.
yongari [Tue, 17 Jan 2012 19:31:03 +0000 (19:31 +0000)]
Use a RX DMA tag to free loaded RX DMA maps.
Previously it used a TX DMA tag.

12 years agoTidy-up.
adrian [Tue, 17 Jan 2012 18:20:34 +0000 (18:20 +0000)]
Tidy-up.

12 years agoThe newhook method can be called in ISR context at
glebius [Tue, 17 Jan 2012 18:10:25 +0000 (18:10 +0000)]
The newhook method can be called in ISR context at
certain circumstances, so better use M_NOWAIT in it.

12 years agoAdd definitions for the FPU extended state header, legacy extended
kib [Tue, 17 Jan 2012 17:07:13 +0000 (17:07 +0000)]
Add definitions for the FPU extended state header, legacy extended
state and AVX state.

MFC after: 1 week

12 years agoModernize the fpusave structures definitions by using uint*_t types.
kib [Tue, 17 Jan 2012 16:53:41 +0000 (16:53 +0000)]
Modernize the fpusave structures definitions by using uint*_t types.

MFC after: 1 week

12 years agoExplain why it is safe to unlock the vnode.
alc [Tue, 17 Jan 2012 16:20:50 +0000 (16:20 +0000)]
Explain why it is safe to unlock the vnode.

Requested by: kib

12 years agoFix clang atomic to use for atomic_is_lock_free().
theraven [Tue, 17 Jan 2012 15:20:41 +0000 (15:20 +0000)]
Fix clang atomic to use for atomic_is_lock_free().

Reviewed by: ed
Approved by: dim (mentor)

12 years agoAllocate our mbuf with m_get2().
glebius [Tue, 17 Jan 2012 12:14:26 +0000 (12:14 +0000)]
Allocate our mbuf with m_get2().

12 years agoProvide a function m_get2() that allocates a minimal mbuf that
glebius [Tue, 17 Jan 2012 12:13:36 +0000 (12:13 +0000)]
Provide a function m_get2() that allocates a minimal mbuf that
would fit specified size. Returned mbuf may be a single mbuf,
an mbuf with a cluster from packet zone, or an mbuf with jumbo
cluster of sufficient size.

12 years agoImplement xsetbv(), xsave() and xrstor() providing C access to the
kib [Tue, 17 Jan 2012 07:30:36 +0000 (07:30 +0000)]
Implement xsetbv(), xsave() and xrstor() providing C access to the
similarly named CPU instructions.

Since our in-tree binutils gas is not aware of the instructions, and
I have to use the byte-sequence to encode them, hardcode the r/m operand
as (%rdi). This way, first argument of the pseudo-function is already
placed into proper register.

MFC after: 1 week

12 years agoAdd definitions related to XCR0.
kib [Tue, 17 Jan 2012 07:23:43 +0000 (07:23 +0000)]
Add definitions related to XCR0.

MFC after: 1 week

12 years agoAdd macro IS_BSP() to check whether the current CPU is BSP.
kib [Tue, 17 Jan 2012 07:21:23 +0000 (07:21 +0000)]
Add macro IS_BSP() to check whether the current CPU is BSP.

MFC after: 1 week

12 years agoFix the "lock &zrl->zr_mtx already initialized" assertion by initializing
pluknet [Tue, 17 Jan 2012 06:23:25 +0000 (06:23 +0000)]
Fix the "lock &zrl->zr_mtx already initialized" assertion by initializing
the allocated memory before calling mtx_init(9) on mtx pointing to it.
Otherwize, random contents of uninitialized memory might occasionally
trigger the assertion.

Reported by: Pavel Polyakov <bsd kobyla org>
Reviewed by: pjd
MFC after: 1 week

12 years agoReturn EOPNOTSUPP since we only support update mounts for NFS export.
kevlo [Tue, 17 Jan 2012 01:25:53 +0000 (01:25 +0000)]
Return EOPNOTSUPP since we only support update mounts for NFS export.

Spotted by: trociny

12 years agoThere are several bugs/hangs when trying to take a snapshot on a UFS/FFS
mckusick [Tue, 17 Jan 2012 01:14:56 +0000 (01:14 +0000)]
There are several bugs/hangs when trying to take a snapshot on a UFS/FFS
filesystem running with journaled soft updates. Until these problems
have been tracked down, return ENOTSUPP when an attempt is made to
take a snapshot on a filesystem running with journaled soft updates.

MFC after: 2 weeks

12 years agoMake sure all intermediate variables holding mount flags (mnt_flag)
mckusick [Tue, 17 Jan 2012 01:08:01 +0000 (01:08 +0000)]
Make sure all intermediate variables holding mount flags (mnt_flag)
and that all internal kernel calls passing mount flags are declared
as uint64_t so that flags in the top 32-bits are not lost.

MFC after: 2 weeks

12 years agoRevert r212360 now that PowerPC can handle large sparse arguments to
nwhitehorn [Tue, 17 Jan 2012 00:31:09 +0000 (00:31 +0000)]
Revert r212360 now that PowerPC can handle large sparse arguments to
pmap_remove() (changed in r228412).

MFC after: 2 weeks

12 years agoImprove abstraction. Eliminate direct access by elf*_load_section()
alc [Tue, 17 Jan 2012 00:27:32 +0000 (00:27 +0000)]
Improve abstraction.  Eliminate direct access by elf*_load_section()
to an OBJT_VNODE-specific field of the vm object.  The same
information can be just as easily obtained from the struct vattr that
is in struct image_params if the latter is passed to
elf*_load_section().  Moreover, by replacing the vmspace and vm
object parameters to elf*_load_section() with a struct image_params
parameter, we actually reduce the size of the object code.

In collaboration with: kib

12 years ago- Add ID for the BeagleBone FTDI serial over usb port.
stas [Mon, 16 Jan 2012 23:14:23 +0000 (23:14 +0000)]
- Add ID for the BeagleBone FTDI serial over usb port.

MFC after: 3 days

12 years agoAdd support for more USB devices.
hselasky [Mon, 16 Jan 2012 22:26:25 +0000 (22:26 +0000)]
Add support for more USB devices.

Submitted by: pav @
MFC after: 1 week

12 years agoRemove unnecessary includes from these libutil man pages.
ghelmer [Mon, 16 Jan 2012 21:25:41 +0000 (21:25 +0000)]
Remove unnecessary includes from these libutil man pages.

Requested by bde (as was the previous commit).

12 years agoFix more disorder in prototypes and constants.
ghelmer [Mon, 16 Jan 2012 21:19:23 +0000 (21:19 +0000)]
Fix more disorder in prototypes and constants.
Fix header comments for each section of constants.
Fix whitespace in #define lines.
Fix unnecessary parenthesis in constants.

12 years agoBe pedantic and change // comment to C-style one.
pluknet [Mon, 16 Jan 2012 20:42:56 +0000 (20:42 +0000)]
Be pedantic and change // comment to C-style one.

Noticed by: Bruce Evans

12 years agoGenerate a warning if the kernel's arc4random() is seeded with bogus entropy.
das [Mon, 16 Jan 2012 20:18:10 +0000 (20:18 +0000)]
Generate a warning if the kernel's arc4random() is seeded with bogus entropy.

12 years agoFix the value of float_t to match what is implied by FLT_EVAL_METHOD.
das [Mon, 16 Jan 2012 20:17:51 +0000 (20:17 +0000)]
Fix the value of float_t to match what is implied by FLT_EVAL_METHOD.

12 years agoChange the definition of FLT_EVAL_METHOD from 1 to 0. A value of 1 implies
das [Mon, 16 Jan 2012 20:17:29 +0000 (20:17 +0000)]
Change the definition of FLT_EVAL_METHOD from 1 to 0.  A value of 1 implies
that the compiler promotes floats to double precision in computations, but
inspection of the output of a cross-compiler indicates that this isn't the
case on powerpc.

12 years agoChange checkpath() to not exit on error. This is a prerequisite for
jh [Mon, 16 Jan 2012 19:34:21 +0000 (19:34 +0000)]
Change checkpath() to not exit on error. This is a prerequisite for
fixing the mount(8) "failok" option.

PR: 163668
Reviewed by: Garrett Cooper, delphij (previous version)

12 years agoUse the signal fence builtin in stdatomic.h when using the clang atomic
theraven [Mon, 16 Jan 2012 18:19:53 +0000 (18:19 +0000)]
Use the signal fence builtin in stdatomic.h when using the clang atomic
builtins, rather than the __asm hack.  Somehow I missed the existence of this
builtin originally and only noticed that it was there when I went to implement
it...

Note: Trunk clang now has support for (most of) the C[++]11 atomics stuff.
Please test!

Approved by: brooks (mentor)

12 years agoAdd a bit of verbosity to the comment.
ivoras [Mon, 16 Jan 2012 15:47:42 +0000 (15:47 +0000)]
Add a bit of verbosity to the comment.

12 years agoFix a style bug
kevlo [Mon, 16 Jan 2012 14:54:48 +0000 (14:54 +0000)]
Fix a style bug

Spotted by: avg

12 years agoMake panic strings in m_gettype(), m_getzone(), m_cljset()
glebius [Mon, 16 Jan 2012 13:23:19 +0000 (13:23 +0000)]
Make panic strings in m_gettype(), m_getzone(), m_cljset()
consistent.

12 years agoAdd missing static.
glebius [Mon, 16 Jan 2012 12:33:55 +0000 (12:33 +0000)]
Add missing static.

12 years agoRemove some disabled NOTYET code. Probability of enabling it is low,
glebius [Mon, 16 Jan 2012 12:31:33 +0000 (12:31 +0000)]
Remove some disabled NOTYET code. Probability of enabling it is low,
if anyone wants, he/she can take it from svn.

12 years agosh: Fix execution of multiple statements in a trap when evalskip is set
dumbbell [Mon, 16 Jan 2012 11:07:46 +0000 (11:07 +0000)]
sh: Fix execution of multiple statements in a trap when evalskip is set

Before this fix, only the first statement of the trap was executed if
evalskip was set. This is for example the case when:
    o  "-e" is set for this shell
    o  a trap is set on EXIT
    o  a function returns 1 and causes the script to abort

Reviewed by: jilles
MFC after: 2 weeks

12 years agosh: Test EXIT trap with multiple statements in it
dumbbell [Mon, 16 Jan 2012 10:59:44 +0000 (10:59 +0000)]
sh: Test EXIT trap with multiple statements in it

Reviewed by: jilles
MFC after: 2 weeks

12 years agom_getzone() should return only cluster zones.
glebius [Mon, 16 Jan 2012 10:58:14 +0000 (10:58 +0000)]
m_getzone() should return only cluster zones.

12 years agoExport ttyname instead of ttyunit via the sysctl interface.
hselasky [Mon, 16 Jan 2012 10:42:43 +0000 (10:42 +0000)]
Export ttyname instead of ttyunit via the sysctl interface.

Submitted by: Mykhaylo Yehorov
PR: usb/164090
MFC after: 1 week

12 years agoAdd nfs export support to tmpfs(5)
kevlo [Mon, 16 Jan 2012 10:25:22 +0000 (10:25 +0000)]
Add nfs export support to tmpfs(5)

Reviewed by: kib

12 years agoDrop support for SIOCSIFADDR, SIOCSIFNETMASK, SIOCSIFBRDADDR, SIOCSIFDSTADDR
glebius [Mon, 16 Jan 2012 09:53:24 +0000 (09:53 +0000)]
Drop support for SIOCSIFADDR, SIOCSIFNETMASK, SIOCSIFBRDADDR, SIOCSIFDSTADDR
ioctl commands.

PR: 163524
Reviewed by: net

12 years agoExport information about USB serial port unit and port numbers
hselasky [Mon, 16 Jan 2012 08:13:25 +0000 (08:13 +0000)]
Export information about USB serial port unit and port numbers
directly via the sysctl interface.

Submitted by: Mykhaylo Yehorov
PR: usb/164090
MFC after: 1 week

12 years agoInsert read memory barriers.
davidxu [Mon, 16 Jan 2012 06:15:14 +0000 (06:15 +0000)]
Insert read memory barriers.

12 years agoMatch surrounding style.
delphij [Mon, 16 Jan 2012 06:00:44 +0000 (06:00 +0000)]
Match surrounding style.

Noticed by: avg

12 years agoRemove a confused comment and fix some minor bugs.
das [Mon, 16 Jan 2012 05:23:27 +0000 (05:23 +0000)]
Remove a confused comment and fix some minor bugs.

12 years agoFix the definition of FLT_EVAL_METHOD and some minor bugs.
das [Mon, 16 Jan 2012 05:23:13 +0000 (05:23 +0000)]
Fix the definition of FLT_EVAL_METHOD and some minor bugs.

12 years agoStop overloading opt_global.h.
adrian [Mon, 16 Jan 2012 05:07:32 +0000 (05:07 +0000)]
Stop overloading opt_global.h.

12 years agoEliminate branch and insert an explicit reader memory barrier to ensure
davidxu [Mon, 16 Jan 2012 04:39:10 +0000 (04:39 +0000)]
Eliminate branch and insert an explicit reader memory barrier to ensure
that waiter bit is set before reading semaphore count.

12 years agoComputations on NaNs are supposed to return one of the input NaNs unchanged.
das [Mon, 16 Jan 2012 04:09:45 +0000 (04:09 +0000)]
Computations on NaNs are supposed to return one of the input NaNs unchanged.
Fix a few places in the sparc64 floating-point emulator where this wasn't
being handled properly.

Submitted by: bde

12 years agoAdd an implementation of fenv.h intended for platforms that lack an FPU and
das [Mon, 16 Jan 2012 04:09:17 +0000 (04:09 +0000)]
Add an implementation of fenv.h intended for platforms that lack an FPU and
use softfloat.

Thanks to Ian Lepore for testing and debugging this patch.  The fenv
regression tests pass (at least for Ian's arm chip) with this change.

12 years agoImplement FLT_ROUNDS for arm. Some (all?) arm FPUs lack support for
das [Mon, 16 Jan 2012 04:08:29 +0000 (04:08 +0000)]
Implement FLT_ROUNDS for arm.  Some (all?) arm FPUs lack support  for
dynamic rounding modes, but FPUless chips that use softfloat can support it
because everything is emulated anyway.  (We presently have incomplete
support for hardware FPUs.)

Submitted by: Ian Lepore

12 years agoRemove functions from softfloat's Symbol.map that don't need to be exported.
das [Mon, 16 Jan 2012 04:06:56 +0000 (04:06 +0000)]
Remove functions from softfloat's Symbol.map that don't need to be exported.
Also use the proper number of underscores for internal names.  (Changing
the names should be fine, since apparently the symbols are currently
unused.)

12 years agoConvert softfloat to use the standard exception flag and rounding macros
das [Mon, 16 Jan 2012 04:05:53 +0000 (04:05 +0000)]
Convert softfloat to use the standard exception flag and rounding macros
in fenv.h instead of the non-standard and incomplete ones in ieeefp.h.

Thanks to Ian Lepore for testing this patch.

12 years agoCorrect some bugs that resulted from arm/_fpmath.h being blindly copied
das [Mon, 16 Jan 2012 04:04:35 +0000 (04:04 +0000)]
Correct some bugs that resulted from arm/_fpmath.h being blindly copied
from the x86 version, which has a completely different long double
format.

Submitted by: Maks Verver

12 years agos/amd64/i386/ in comment.
cperciva [Mon, 16 Jan 2012 02:42:41 +0000 (02:42 +0000)]
s/amd64/i386/ in comment.

12 years agoCopy XENHVM config file from amd64, now that i386+XENHVM works.
cperciva [Mon, 16 Jan 2012 02:42:16 +0000 (02:42 +0000)]
Copy XENHVM config file from amd64, now that i386+XENHVM works.

12 years agoMake XENHVM work on i386. The __ffs() function counts bits starting from
cperciva [Mon, 16 Jan 2012 02:38:45 +0000 (02:38 +0000)]
Make XENHVM work on i386.  The __ffs() function counts bits starting from
zero, unlike ffs(3), which starts counting from 1.

12 years agoAdd check to avoid assertion panic on duplicate stop.
mav [Mon, 16 Jan 2012 00:26:52 +0000 (00:26 +0000)]
Add check to avoid assertion panic on duplicate stop.

Reported by: Yuri Pankov <yuri.pankov@gmail.com>
MFC after: 2 months

12 years agoWhen tmpfs_write() resets an extended file to its original size after an
alc [Mon, 16 Jan 2012 00:26:49 +0000 (00:26 +0000)]
When tmpfs_write() resets an extended file to its original size after an
error, we want tmpfs_reg_resize() to ignore I/O errors and unconditionally
update the file's size.

Reviewed by: kib
MFC after: 3 weeks

12 years agoBeagleBone uses an FTDI chip with
kientzle [Sun, 15 Jan 2012 23:00:33 +0000 (23:00 +0000)]
BeagleBone uses an FTDI chip with
an altered Product ID.

12 years agosh: Fix some bugs with exit status from case containing ;&.
jilles [Sun, 15 Jan 2012 21:39:38 +0000 (21:39 +0000)]
sh: Fix some bugs with exit status from case containing ;&.

Also, rework evalcase() to not evaluate any tree. Instead, return the
NCLISTFALLTHRU node and handle it in evaltree().

Fixed bugs:

* If a ;& list with non-zero exit status is followed by an empty ;; or final
  list, the exit status of the case command should be equal to the exit
  status of the ;& list, not 0.

* An empty ;& case should not reset $?.

12 years agodadump: don't leak the periph lock on i/o error
avg [Sun, 15 Jan 2012 20:43:39 +0000 (20:43 +0000)]
dadump: don't leak the periph lock on i/o error

Reported by: az
MFC after: 1 week

12 years agoxlocale.h: consistently use __va_list
avg [Sun, 15 Jan 2012 20:37:39 +0000 (20:37 +0000)]
xlocale.h: consistently use __va_list

Plain 'va_list' in this header seems to cause troubles with non-base GCC
which creates and uses "tortured" versions of some sysem header files
including stdio.h (installed in a private 'include-fixed' directory).

Reviewed by: theraven
X-MFC with: r227753

12 years agoMake man page wording more clear:
eadler [Sun, 15 Jan 2012 20:14:52 +0000 (20:14 +0000)]
Make man page wording more clear:

PR: docs/164078
Submitted by: Taras <ds@ukrhub.net>
Approved by: bcr
MFC after: 3 days

12 years agosh: Fix two bugs with case and exit status:
jilles [Sun, 15 Jan 2012 20:04:05 +0000 (20:04 +0000)]
sh: Fix two bugs with case and exit status:

* If no pattern is matched, POSIX says the exit status shall be 0 (even if
  there are command substitutions).
* If a pattern is matched and there are no command substitutions, the first
  command should see the $? from before the case command, not always 0.

12 years agoFix the situation where net80211 is built with IEEE80211_SUPPORT_TDMA but a module
adrian [Sun, 15 Jan 2012 19:45:23 +0000 (19:45 +0000)]
Fix the situation where net80211 is built with IEEE80211_SUPPORT_TDMA but a module
is used.

Although the module _builds_, it fails to load because of a missing symbol from
ieee80211_tdma.c.

Specifics:

* Always build ieee80211_tdma.c in the module;
* only compile in the code if IEEE80211_SUPPORT_TDMA is defined.

12 years agoBuild some more things (random, bridge/gif/gre, gpio, USB) as modules as well
adrian [Sun, 15 Jan 2012 19:43:56 +0000 (19:43 +0000)]
Build some more things (random, bridge/gif/gre, gpio, USB) as modules as well
so some embedded platform builds can use these instead of a fully monolithic
kernel.

12 years agoBegin breaking out the AR71xx specific functional drivers into modules.
adrian [Sun, 15 Jan 2012 19:42:55 +0000 (19:42 +0000)]
Begin breaking out the AR71xx specific functional drivers into modules.

The USB code as it stands includes the bus glue along _with_ the controller
code. So the ohci/ehci modules actually build the USB controller code and
the PCI bus glue.

It'd be nice to ship separate modules for the PCI glue and the USB
controller (so for example if there were a USB controller hanging off
the internal SoC bus as well as an external PCI device) it could be done.

This is primarily done to save a few bytes here and there on embedded
systems with limited flash space for kernels - a very limited (sub-1MB)
space may be available for the kernel and may only support gzip encoding.
The rootfs can be LZMA compressed.

12 years agoAllow building the GPIO bus and associated bits as modules.
adrian [Sun, 15 Jan 2012 19:40:59 +0000 (19:40 +0000)]
Allow building the GPIO bus and associated bits as modules.

This is primarily done to save a few bytes here and there on embedded
systems with limited flash space for kernels - a very limited (sub-1MB)
space may be available for the kernel and may only support gzip encoding.
The rootfs can be LZMA compressed.

12 years agoAdd the new option introduced in the previous commit.
adrian [Sun, 15 Jan 2012 19:30:32 +0000 (19:30 +0000)]
Add the new option introduced in the previous commit.

12 years agoSome of the atheros based embedded devices use one or more PCI NICs
adrian [Sun, 15 Jan 2012 19:29:33 +0000 (19:29 +0000)]
Some of the atheros based embedded devices use one or more PCI NICs
on-board, glued to the AR71xx CPU.  These may forgo separate WMAC EEPROMs
(which store configuration and calibration data) and instead store
it in the main board SPI flash.

Normally the NIC reads the EEPROM attached to it to setup various PCI
configuration registers.  If this isn't done, the device will probe as
something different (eg 0x168c:abcd, or 0x168c:ff??.)  Other setup registers
are also written to which may control important functions.

This introduces a new compile option, AR71XX_ATH_EEPROM, which enables the
use of this particular code.  The ART offset in the SPI flash can be
specified as a hint against the relevant slot/device number, for example:

hint.pcib.0.bus.0.17.0.ath_fixup_addr=0x1fff1000
hint.pcib.0.bus.0.18.0.ath_fixup_addr=0x1fff5000

TODO:

* Think of a better name;
* Make the PCIe version of this fixup code also use this option;
* Maybe also check slot 19;
* This has to happen _before_ the SPI flash is set from memory-mapped
  to SPI-IO - so document that somewhere.