]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
12 years agoThe `end' symbol doesn't match the end of the kernel image because it's
rpaulo [Fri, 29 Jun 2012 01:55:20 +0000 (01:55 +0000)]
The `end' symbol doesn't match the end of the kernel image because it's
relative to the start address (unless the start address is 0, which is
not the case).
This is currently not a problem because all powerpc architectures are
using loader(8) which passes metadata to the kernel including the
correct `endkernel' address.  If we don't use loader(8), register 4
and 5 will have the size of the kernel ELF file, not its end address.
We fix that simply by adding `kernel_text' to `end' to compute
`endkernel'.

Discussed with: nathanw

12 years agoKick the current-state report timer when a V1 group report would
bms [Thu, 28 Jun 2012 23:48:40 +0000 (23:48 +0000)]
Kick the current-state report timer when a V1 group report would
be triggered.

Submitted by: rpaulo@
MFC after: 3 days

12 years agoFix a typo in MLD query exponent processing.
bms [Thu, 28 Jun 2012 23:45:37 +0000 (23:45 +0000)]
Fix a typo in MLD query exponent processing.

Submitted by: rpaulo@
MFC after: 3 days

12 years agoIn MLDv2 general query processing, do not enforce the strict check
bms [Thu, 28 Jun 2012 23:44:47 +0000 (23:44 +0000)]
In MLDv2 general query processing, do not enforce the strict check
on query origins.

Submitted by: Gu Yong
MFC after: 3 days

12 years agoAvoid some unnecessary PV list locking in pmap_enter().
alc [Thu, 28 Jun 2012 22:03:59 +0000 (22:03 +0000)]
Avoid some unnecessary PV list locking in pmap_enter().

12 years agoNow that the mps(4) driver is endian-safe, add it to the powerpc and
ken [Thu, 28 Jun 2012 20:48:24 +0000 (20:48 +0000)]
Now that the mps(4) driver is endian-safe, add it to the powerpc and
sparc64 GENERIC config files.

MFC after: 3 days

12 years agoFix ioctl type for compiling with clang
emaste [Thu, 28 Jun 2012 20:10:43 +0000 (20:10 +0000)]
Fix ioctl type for compiling with clang

12 years agoAdd a loader tunable, kern.cam.ctl.disable, that will disable
ken [Thu, 28 Jun 2012 19:39:30 +0000 (19:39 +0000)]
Add a loader tunable, kern.cam.ctl.disable, that will disable
loading CTL.  This may be useful in very low memory installations.

MFC after: 3 days

12 years agoFix millisecond to ticks conversion in drm_msleep().
mav [Thu, 28 Jun 2012 17:15:16 +0000 (17:15 +0000)]
Fix millisecond to ticks conversion in drm_msleep().

On systems with HZ=100 it caused Intel eDP video output initialization
(and Xorg startup) to take several minutes instead of several seconds.

Reviewed by: kib
MFC after: 3 days

12 years agoSafer fix for building with Clang.
pfg [Thu, 28 Jun 2012 16:44:29 +0000 (16:44 +0000)]
Safer fix for building with Clang.

error: format specifies type 'long long' but the argument
has type 'int64_t' (aka 'long')

Reported by: Ed Maste

12 years agoPass the src and dst address of a received packet explicitly around.
tuexen [Thu, 28 Jun 2012 16:01:08 +0000 (16:01 +0000)]
Pass the src and dst address of a received packet explicitly around.

MFC after: 3 days

12 years agoFix build with Clang.
pfg [Thu, 28 Jun 2012 15:38:14 +0000 (15:38 +0000)]
Fix build with Clang.

error: format specifies type 'long long' but the argument
has type 'int64_t' (aka 'long')

Reported by: Ed Maste

12 years agoAutomatically generate hardware notes for bwn(4).
brueffer [Thu, 28 Jun 2012 10:21:25 +0000 (10:21 +0000)]
Automatically generate hardware notes for bwn(4).

12 years agoAutomatically generate hardware notes for bxe(4).
brueffer [Thu, 28 Jun 2012 10:12:46 +0000 (10:12 +0000)]
Automatically generate hardware notes for bxe(4).

12 years agoo Restore -u <username> getopt(3) flag somehow killed in r234712.
maxim [Thu, 28 Jun 2012 08:25:19 +0000 (08:25 +0000)]
o Restore -u <username> getopt(3) flag somehow killed in r234712.

PR: bin/169490
Submitted by: amdmi3
MFC after: 2 weeks

12 years ago- Add UPDATING entry for BSD sort
gabor [Thu, 28 Jun 2012 08:22:00 +0000 (08:22 +0000)]
- Add UPDATING entry for BSD sort

12 years agoTurns out the emulators still need/want a_magic, so put it back here.
imp [Thu, 28 Jun 2012 07:37:46 +0000 (07:37 +0000)]
Turns out the emulators still need/want a_magic, so put it back here.

12 years agoRemove an old hack I noticed years ago, but never committed.
imp [Thu, 28 Jun 2012 07:33:43 +0000 (07:33 +0000)]
Remove an old hack I noticed years ago, but never committed.

12 years agoDocument some MP assumptions for sio.
imp [Thu, 28 Jun 2012 07:28:39 +0000 (07:28 +0000)]
Document some MP assumptions for sio.

Submitted by: bde (years ago)

12 years agoSimplify resource activation a bit.
imp [Thu, 28 Jun 2012 07:26:44 +0000 (07:26 +0000)]
Simplify resource activation a bit.

12 years agoAdd a sysctl to set the cdrom timeout. Data recovery operations from
imp [Thu, 28 Jun 2012 07:01:48 +0000 (07:01 +0000)]
Add a sysctl to set the cdrom timeout.  Data recovery operations from
a CD or DVD drive with a damaged disc often benefit from a shorter
timeout.  Also, when retries are set to 0, an application is expecting
errors and recovering them so do not print the error into the log.
The number of expected errors can literally be in the hundreds of
thousands which significantly slows data recovery.

Reviewed by: ken@ (but quite some time ago).

12 years agoocteon_uart_class was removed some time ago everywhere but here.
imp [Thu, 28 Jun 2012 06:49:04 +0000 (06:49 +0000)]
octeon_uart_class was removed some time ago everywhere but here.

12 years agoOptimize pmap_pv_demote_pde().
alc [Thu, 28 Jun 2012 05:42:04 +0000 (05:42 +0000)]
Optimize pmap_pv_demote_pde().

12 years agoBring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
ken [Thu, 28 Jun 2012 03:48:54 +0000 (03:48 +0000)]
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.

Their description of the changes is as follows:

1. Copyright contents has been changed in all respective .c
and .h files

2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.

3.      Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.

4.      With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls.  Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a

a. While driver is getting loaded, driver calls most of the
   commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
   CAN_SLEEP flag is used.

5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.

Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
   converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
   converts CPU format to Little Endian.

6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.

7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.

8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout.  Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.

9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;

10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond).  So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.

Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.

11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
   hw/fw support. Currently LSI released driver does not set
   this field.
b. Default of "max_lun" should not be 0 in OS, but it is
   currently set to 0 in CAM layer.
c. Export max_lun capacity to 255

12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW.  The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
   information in driver local data structure.
b. Only for WD, we have below checks after port enable
   completes, where driver clear off all information retrieved
   at #1.
if ((sc->WD_available &&
             (sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
             (sc->WD_valid_config && (sc->WD_hide_expose ==
                            MPS_WD_HIDE_IF_VOLUME)) {
  // clear off target data structure.
}
It is mainly not to attach PDs to OS.

FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS.  That is a reason
we have to clear off targ information for PDs.

Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.

Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr  5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr  5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe

Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.

13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment.  As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.

Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days

12 years agoFix a typo in a panic() call.
ken [Thu, 28 Jun 2012 03:36:13 +0000 (03:36 +0000)]
Fix a typo in a panic() call.

PR: kern/169497
Submitted by: Steven Hartland <steven.hartland@multiplay.co.uk>
MFC after: 3 days

12 years agoAdd the 'inet' keyword after the nat rule to avoid interfering with
rpaulo [Thu, 28 Jun 2012 03:30:17 +0000 (03:30 +0000)]
Add the 'inet' keyword after the nat rule to avoid interfering with
IPv6.

12 years agoUpdate entries in organization.dot for the following groups:
gjb [Thu, 28 Jun 2012 03:27:35 +0000 (03:27 +0000)]
Update entries in organization.dot for the following groups:

coresecretary, portmgr, portmgrsecretary, secteam

Sort entries by last name.

Note: core is left as-is intentionally for now.

Submitted by: bcr, rpaulo

12 years agoRemove a '\n' that crept in accidentally.
gjb [Thu, 28 Jun 2012 00:17:45 +0000 (00:17 +0000)]
Remove a '\n' that crept in accidentally.

12 years agoUpdate organization.dot to reflect current doceng@.
gjb [Thu, 28 Jun 2012 00:03:37 +0000 (00:03 +0000)]
Update organization.dot to reflect current doceng@.
While here, sort line by last name.

12 years agoAdd my entry to committers-ports.dot.
gjb [Wed, 27 Jun 2012 23:58:52 +0000 (23:58 +0000)]
Add my entry to committers-ports.dot.

12 years agoUpdate the usage with the new jail option.
thompsa [Wed, 27 Jun 2012 23:26:32 +0000 (23:26 +0000)]
Update the usage with the new jail option.

Spotted by: Jason Hellenthal
MFC after: 3 days

12 years agoFor subtractively decoding bridges, don't try to grow windows but pass
marius [Wed, 27 Jun 2012 22:17:52 +0000 (22:17 +0000)]
For subtractively decoding bridges, don't try to grow windows but pass
the request up the tree in order to be on the safe side. Growing windows
in this case would mean to switch resources to positive decoding and
it's unclear how to correctly handle this. At least with ALi/ULi M5249
PCI-PCI bridges, this also just doesn't work out of the box.

Reviewed by: jhb
MFC after: 3 days

12 years agoRegen ca(1) for r237658. This re-applies r227458, i.e., add a missing "be".
jkim [Wed, 27 Jun 2012 21:35:45 +0000 (21:35 +0000)]
Regen ca(1) for r237658.  This re-applies r227458, i.e., add a missing "be".

12 years agoAdd a test for number of CPUs configured/online.
kib [Wed, 27 Jun 2012 20:34:29 +0000 (20:34 +0000)]
Add a test for number of CPUs configured/online.

MFC after: 1 week

12 years agoOptimize the handling of SC_NPROCESSORS_CONF, by using auxv AT_NCPU
kib [Wed, 27 Jun 2012 20:32:45 +0000 (20:32 +0000)]
Optimize the handling of SC_NPROCESSORS_CONF, by using auxv AT_NCPU
value if present.

MFC after: 1 week

12 years agoEnsure that for the object which is a dependency for some filtee,
kib [Wed, 27 Jun 2012 20:24:25 +0000 (20:24 +0000)]
Ensure that for the object which is a dependency for some filtee,
relocations are performed before the object's initializer is called.
When dlopen()ing an object, relocate the whole DAG rooted in the
object instead of only relocating the object itself and list of newly
loaded dependencies.

Reversed sequence currently can occur if the same object is a
dependency for both filtee and filter, since filtees are loaded
typically during the relocation processing, when some filter
dependencies might be already loaded but not relocated yet.

Reported and tested by: swills
Reviewed by: kan
MFC after: 1 week

12 years agoPartially redo r226436, i. e., change "the the" to "the". ca(1), dgst(1),
jkim [Wed, 27 Jun 2012 19:23:29 +0000 (19:23 +0000)]
Partially redo r226436, i. e., change "the the" to "the".  ca(1), dgst(1),
and engine(3) are generated from these pod files during merge process and
we do not want to re-apply these changes over and over again.

Approved by: benl (maintainer, implicit)

12 years agoMerge OpenSSL 0.9.8x.
jkim [Wed, 27 Jun 2012 18:44:36 +0000 (18:44 +0000)]
Merge OpenSSL 0.9.8x.

Reviewed by: stas
Approved by: benl (maintainer)
MFC after: 3 days

12 years agoAdd a new line to top that provides a brief summary of the ZFS ARC memory
jhb [Wed, 27 Jun 2012 18:08:48 +0000 (18:08 +0000)]
Add a new line to top that provides a brief summary of the ZFS ARC memory
usage on hosts using ZFS.  The new line displays the total amount of RAM
used by the ARC along with the size of MFU, MRU, anonymous (in flight),
headers, and other (miscellaneous) sub-categories.  The line is not
displayed on systems that are not using ZFS.

Reviewed by: avg, fs@
MFC after: 3 days

12 years agoPoint out that /usr/src, /usr/doc, and /usr/ports may be empty or not
wblock [Wed, 27 Jun 2012 17:51:09 +0000 (17:51 +0000)]
Point out that /usr/src, /usr/doc, and /usr/ports may be empty or not
exist until populated with csup(1) or svn(1).

PR: 169487
Submitted by: Moritz Wilhelmy
MFC after: 1 day

12 years agoMFV: r237650
jkim [Wed, 27 Jun 2012 16:15:13 +0000 (16:15 +0000)]
MFV: r237650

Do not malloc(9) while holding a spin lock, to avoid panic.

Reported by: kib (and many others)
Tested by: kib (and many others)

12 years agoAdd new firmware for the g2a (6205) and g2b (623x) devices.
bschmidt [Wed, 27 Jun 2012 16:14:28 +0000 (16:14 +0000)]
Add new firmware for the g2a (6205) and g2b (623x) devices.

MFC after: 3 days

12 years agoWe need to defer passing the DELBA request to the firmware until the aggr
bschmidt [Wed, 27 Jun 2012 16:07:01 +0000 (16:07 +0000)]
We need to defer passing the DELBA request to the firmware until the aggr
queue is empty or the firmware will go nuts.

PR: kern/167806
Tested by: osa@, Brandon Gooch (earlier version),
    Bojan Petrovic (earlier version)
MFC after: 3 days

12 years agoIn g_disk_providergone(), don't continue if the softc is NULL. This may be
ken [Wed, 27 Jun 2012 16:05:09 +0000 (16:05 +0000)]
In g_disk_providergone(), don't continue if the softc is NULL.  This may be
the case if we've already gone through g_disk_destroy().

Reported by: Michael Butler <imb@protected-networks.net>
MFC after: 3 days

12 years agoFix a TX aggregation issue, if after the last compressed BA notification
bschmidt [Wed, 27 Jun 2012 15:55:34 +0000 (15:55 +0000)]
Fix a TX aggregation issue, if after the last compressed BA notification
the TX queue is empty, there won't be a TX done notification, effectly
resulting in an mbuf leak. The correct way to handle this is to free
up mbufs on both BA and TX done notifications up to the last sent seqno.

Tested by: osa@
MFC after: 3 days

12 years agoClarify that the cached file data pages included in the "Wired" count
jhb [Wed, 27 Jun 2012 12:30:56 +0000 (12:30 +0000)]
Clarify that the cached file data pages included in the "Wired" count
in top are the BIO-level cached data (i.e. "Buf"), since the previous
phrase was a bit ambiguous with the "Cache" count.

MFC after: 3 days

12 years agoThe -S option, to specify the sector size, has been usable on all types of
gavin [Wed, 27 Jun 2012 12:19:28 +0000 (12:19 +0000)]
The -S option, to specify the sector size, has been usable on all types of
memory disks since r135340.  Update the man page to reflect this.

Noticed by: avg
MFC after: 3 days

12 years agoAdd manual page for bxe(4).
trasz [Wed, 27 Jun 2012 10:07:29 +0000 (10:07 +0000)]
Add manual page for bxe(4).

Reviewed by: davidch (earlier version)
MFC after: 1 week

12 years agomdocify and update the traceroute(8) manual page.
joel [Wed, 27 Jun 2012 07:41:21 +0000 (07:41 +0000)]
mdocify and update the traceroute(8) manual page.

Reviewed by: brueffer
No objection: ru, uqs

12 years ago- Bump __FreeBSD_version: default sort is BSD sort
gabor [Wed, 27 Jun 2012 06:08:02 +0000 (06:08 +0000)]
- Bump __FreeBSD_version: default sort is BSD sort

12 years ago- Regen after BSD sort change
gabor [Wed, 27 Jun 2012 06:00:58 +0000 (06:00 +0000)]
- Regen after BSD sort change

12 years ago- Switch to BSD sort as default sort. GNU sort will still be installed as
gabor [Wed, 27 Jun 2012 05:59:01 +0000 (05:59 +0000)]
- Switch to BSD sort as default sort.  GNU sort will still be installed as
  "gnusort".  Most of the BSD sort development work was done by
  Oleg Moskalenko <oleg.moskalenko@citrix.com>.
- GNU grep can be set to default by setting WITH_GNU_GREP.  It will cause
  BSD sort to be installed as "bsdsort".

Portbuild tested by:    linimon

12 years ago- Disable threaded sort by default
gabor [Wed, 27 Jun 2012 05:50:15 +0000 (05:50 +0000)]
- Disable threaded sort by default
- Fix typo in the NLS support

12 years agoFix a subtle corner case surrounding the handling of OFDM restart along
adrian [Wed, 27 Jun 2012 05:23:33 +0000 (05:23 +0000)]
Fix a subtle corner case surrounding the handling of OFDM restart along
with AMPDU aggregate delimiters.

If there's an OFDM restart during an aggregate, the hardware ACKs
the previous frame, but communicates the RXed frame to the hardware
as having had CRC delimiter error + OFDM_RESTART phy error.
The frame however didn't have a CRC error and since the hardware ACKed
the aggregate to the sender, it thinks the frame was received.

Since I have no idea how often this occurs in the real world, add a
debug statement so trigger whenever this occurs.  I'd appreciate an
email if someone finds this particular situation is triggered.

12 years agoSome amount of style(9)
obrien [Wed, 27 Jun 2012 04:39:30 +0000 (04:39 +0000)]
Some amount of style(9)
  -- function definitions, header ordering, and $FreeBSD$.

12 years agoBring llquantize support into Dtrace.
pfg [Wed, 27 Jun 2012 04:39:30 +0000 (04:39 +0000)]
Bring llquantize support into Dtrace.

Bryan Cantrill implemented the equivalent of semi-log graph
paper for Dtrace so llquantize will use one logarithmic and
one linear scale.

Special thanks to Mark Peek for providing fix to an
assertion and to Fabian Keill for testing the port.

Illumos Revision: 13355:15b74a2a9a9d

Reference:
https://www.illumos/issues/905

Obtained from: Illumos
Tested by: Fabian Keill, mp
MFC after: 4 days

12 years agoAdd new pmap layer locks to the predefined lock order. Change the names
alc [Wed, 27 Jun 2012 03:45:25 +0000 (03:45 +0000)]
Add new pmap layer locks to the predefined lock order.  Change the names
of a few existing VM locks to follow a consistent naming scheme.

12 years agoBring over some new typedefs as part of the AR9300 HAL import.
adrian [Wed, 27 Jun 2012 03:24:27 +0000 (03:24 +0000)]
Bring over some new typedefs as part of the AR9300 HAL import.

12 years agoRemove duplicate entries.
adrian [Wed, 27 Jun 2012 03:00:29 +0000 (03:00 +0000)]
Remove duplicate entries.

12 years agoUse possessive "its", no apostrophe.
wblock [Wed, 27 Jun 2012 01:44:48 +0000 (01:44 +0000)]
Use possessive "its", no apostrophe.

MFC after: 1 day

12 years agoTry clean up some of my original text and neaten a table.
julian [Wed, 27 Jun 2012 01:13:37 +0000 (01:13 +0000)]
Try clean up some of my original text and neaten a table.

MFC after: 1 week

12 years agoAdd a -I flag which requests confirmation before action, like what is done
delphij [Wed, 27 Jun 2012 00:50:25 +0000 (00:50 +0000)]
Add a -I flag which requests confirmation before action, like what is done
in pkill(1).

MFC after: 2 weeks

12 years agoMFV: less v449.
delphij [Tue, 26 Jun 2012 23:17:33 +0000 (23:17 +0000)]
MFV: less v449.

12 years agoAdd -DWITH_INSTALL_AS_USER to the source build (naming matches the
obrien [Tue, 26 Jun 2012 23:15:55 +0000 (23:15 +0000)]
Add -DWITH_INSTALL_AS_USER to the source build (naming matches the
same functionality of The Ports Collection).
This sets BINOWN, BINGRP, etc... to match current user.  This this
allows 'install', as used in 'make install', to succeed (assuming
user has write permissions).

Submitted by: Simon J Gerraty <sjg@juniper.net>
Discussed on: freebsd-arch

12 years agoBring over the initial 802.11n bluetooth coexistence support code.
adrian [Tue, 26 Jun 2012 22:16:53 +0000 (22:16 +0000)]
Bring over the initial 802.11n bluetooth coexistence support code.

The Linux ath9k btcoex code is based off of this code.

Note this doesn't actually implement functional btcoex; there's some
driver glue and a whole lot of verification that is required.

On the other hand, I do have the AR9285+BT and AR9287+BT NICs which
this code supports..

Obtained from: Qualcomm Atheros, Linux ath9k

12 years agoAdd nand core module and module dependency information.
takawata [Tue, 26 Jun 2012 18:08:03 +0000 (18:08 +0000)]
Add nand core module and module dependency information.

Reviewed by:gber

12 years agoIntroduce RELEASE_PV_LIST_LOCK().
alc [Tue, 26 Jun 2012 16:45:18 +0000 (16:45 +0000)]
Introduce RELEASE_PV_LIST_LOCK().

12 years agoFix an issue that caused the kernel to panic inside CTL when trying
ken [Tue, 26 Jun 2012 14:51:35 +0000 (14:51 +0000)]
Fix an issue that caused the kernel to panic inside CTL when trying
to attach to target capable HBAs that implement the old immediate
notify (XPT_IMMED_NOTIFY) and notify acknowledge (XPT_NOTIFY_ACK)
CCBs.  The new API has been in place since SVN change 196008 in
2009.

The solution is two-fold:  fix CTL to handle the responses from the
HBAs, and convert the HBA drivers in question to use the new API.

These drivers have not been tested with CTL, so how well they will
interoperate with CTL is unknown.

scsi_target.c: Update the userland target example code to use the
new immediate notify API.

scsi_ctl.c: Detect when an immediate notify CCB is returned
with CAM_REQ_INVALID or CAM_PROVIDE_FAIL status,
and just free it.

Fix a duplicate assignment.

aic79xx.c,
aic79xx_osm.c: Update the aic79xx driver to use the new API.
Target mode is not enabled on for this driver, so
the changes will have no practical effect.

aic7xxx.c,
aic7xxx_osm.c: Update the aic7xxx driver to use the new API.

sbp_targ.c: Update the firewire target code to work with the
new API.

mpt_cam.c: Update the mpt(4) driver to work with the new API.
Target mode is only enabled for Fibre Channel
mpt(4) devices.

MFC after: 3 days

12 years agoMake sure the BAR TX session pause is correctly unpaused when a node
adrian [Tue, 26 Jun 2012 07:56:15 +0000 (07:56 +0000)]
Make sure the BAR TX session pause is correctly unpaused when a node
is reassociating.

PR: kern/169432

12 years agoAdd PV list locking to pmap_enter(). Its execution is no longer serialized
alc [Tue, 26 Jun 2012 06:02:43 +0000 (06:02 +0000)]
Add PV list locking to pmap_enter().  Its execution is no longer serialized
by the pvh global lock.

Add a needed atomic operation to pmap_object_init_pt().

12 years agoRemove end of line whitespace.
joel [Tue, 26 Jun 2012 05:34:31 +0000 (05:34 +0000)]
Remove end of line whitespace.

12 years agoAllow cxgbe(4) running within a VM to attach to its devices that have been
np [Tue, 26 Jun 2012 00:34:34 +0000 (00:34 +0000)]
Allow cxgbe(4) running within a VM to attach to its devices that have been
exported via PCI passthrough.

- Do not check for a specific physical function (PF) before claiming a device.
  Different PFs have different device-ids so this check is redundant anyway.

- Obtain the PF# from the WHOAMI register instead of pci_get_function().

- Setup the memory windows using the real BAR0 address, not what the VM says it
  is.

Obtained from: Chelsio Communications

12 years agoExplicitly mention that setting the change and expiry times to zero is
bjk [Mon, 25 Jun 2012 22:39:37 +0000 (22:39 +0000)]
Explicitly mention that setting the change and expiry times to zero is
equivalent to leaving the time unset. [1]

Wordsmith in the compat support section.

Use a full path to nologin(8) in the context of setting it as a user's shell,
keeping a separate cross-reference.

PR: docs/169354 [1]
Approved by: hrs (mentor)
MFC after: 3 weeks

12 years agoEnsure crunchen uses the same make binary as the rest of the build.
obrien [Mon, 25 Jun 2012 21:56:36 +0000 (21:56 +0000)]
Ensure crunchen uses the same make binary as the rest of the build.

Submitted by: Simon Gerraty <sjg@juniper.net>

12 years agoAdd more locale-specific functions to the relevant man pages and
issyl0 [Mon, 25 Jun 2012 21:51:40 +0000 (21:51 +0000)]
Add more locale-specific functions to the relevant man pages and
Makefiles:
- libc/stdtime/strftime.3
- libc/stdtime/strptime.3
- libc/stdlib/strfmon.3

Reviewed by: theraven
Approved by: gabor (mentor)

12 years agoAdd missing MLINKS for whatever -> if_whatever.
trasz [Mon, 25 Jun 2012 21:33:45 +0000 (21:33 +0000)]
Add missing MLINKS for whatever -> if_whatever.

Reviewed by: brueffer
MFC after: 1 week

12 years agoFix a LOR acquiring the if_afdata lock while holding an rtentry lock.
delphij [Mon, 25 Jun 2012 20:56:32 +0000 (20:56 +0000)]
Fix a LOR acquiring the if_afdata lock while holding an rtentry lock.
Possibly do some entra work in case we would not get into the
ifa0 != NULL paths later as we already do for the mltaddr before.

XXX We should possibly error in case in6_setscope fails.

Reference: http://lists.freebsd.org/pipermail/freebsd-net/2011-September/029829.html

Submitted by: bz
MFC after: 1 week

12 years agoUnify sctp_input() and sctp6_input().
tuexen [Mon, 25 Jun 2012 19:13:43 +0000 (19:13 +0000)]
Unify sctp_input() and sctp6_input().

MFC after: 3 days

12 years agoMFV (r237567):
delphij [Mon, 25 Jun 2012 19:01:04 +0000 (19:01 +0000)]
MFV (r237567):

Fetch both ECDSA and RSA keys by default in ssh-keyscan(1).

Approved by: des
Obtained from: OpenSSH portable
MFC after: 1 week

12 years agoHandle case when result of pmap_pte is NULL. This issue was uncovered
gonzo [Mon, 25 Jun 2012 17:50:11 +0000 (17:50 +0000)]
Handle case when result of pmap_pte is NULL. This issue was uncovered
    by r237367

12 years agoWhitespace cleanup.
tuexen [Mon, 25 Jun 2012 17:15:09 +0000 (17:15 +0000)]
Whitespace cleanup.

MFC after: 3 days

12 years agoFix clang warning when compiling iw_cxgb.
np [Mon, 25 Jun 2012 16:52:27 +0000 (16:52 +0000)]
Fix clang warning when compiling iw_cxgb.

Reported by: rene, dim

12 years agoMesh mode, potential garbage in QoS subfield.
monthadar [Mon, 25 Jun 2012 11:52:26 +0000 (11:52 +0000)]
Mesh mode, potential garbage in QoS subfield.

* qos[1] subfield is never assigned a value before this statement.
qos[1] can potentially be OR:ed with garbage. Make it an assignment instead;
* Remove brackets around if statement;

Approved by: adrian

12 years agoFix 'make depend'.
uqs [Mon, 25 Jun 2012 09:46:06 +0000 (09:46 +0000)]
Fix 'make depend'.

12 years agoAdd PV chunk and list locking to pmap_change_wiring(), pmap_protect(), and
alc [Mon, 25 Jun 2012 07:13:25 +0000 (07:13 +0000)]
Add PV chunk and list locking to pmap_change_wiring(), pmap_protect(), and
pmap_remove().  The execution of these functions is no longer serialized
by the pvh global lock.

Make some stylistic changes to the affected code for the sake of
consistency with related code elsewhere in the pmap.

12 years agoFix size of the bcopy when extracting ethernet address
kevlo [Mon, 25 Jun 2012 05:47:12 +0000 (05:47 +0000)]
Fix size of the bcopy when extracting ethernet address

Obtained from: DragonFly

12 years agoCorrect sizeof usage
kevlo [Mon, 25 Jun 2012 05:41:16 +0000 (05:41 +0000)]
Correct sizeof usage

Obtained from: DragonFly

12 years agoConsume spare fields for the providergone pointers added to the g_class and
ken [Mon, 25 Jun 2012 04:26:10 +0000 (04:26 +0000)]
Consume spare fields for the providergone pointers added to the g_class and
g_geom structures in change 237518.  The original change would have broken
the ABI.

Suggested by: ae
MFC after: 4 days

12 years agoUnbreak register tests for parallel SCSI.
mjacob [Mon, 25 Jun 2012 03:06:29 +0000 (03:06 +0000)]
Unbreak register tests for parallel SCSI.
You can't overwrite registers 7 and 8.
MFC after: 3 days

12 years agoPass the packet length explicitly around.
tuexen [Sun, 24 Jun 2012 23:12:24 +0000 (23:12 +0000)]
Pass the packet length explicitly around.

MFC after: 3 days

12 years agoRemove redundant check.
tuexen [Sun, 24 Jun 2012 22:22:44 +0000 (22:22 +0000)]
Remove redundant check.

MFC after: 3 days

12 years agoDo packet logging in a consistent way.
tuexen [Sun, 24 Jun 2012 21:25:54 +0000 (21:25 +0000)]
Do packet logging in a consistent way.

MFC after: 3 days

12 years agoClean up multi-id mode so it's driven by the f/w loaded,
mjacob [Sun, 24 Jun 2012 17:30:54 +0000 (17:30 +0000)]
Clean up multi-id mode so it's driven by the f/w loaded,
not by some hint setting.  Do more preparations for FC-Tape.
Clean up resource counting for 24XX or later chipsets so
we find out after EXEC_FIRMWARE what is actually supported.
Set target mode exchange count based upon whether or not
we are supporting simultaneous target/initiator mode. Clean
up some old (pre-24XX) xfwoption and zfwoption issues.

Sponsored by: Spectralogic
MFC after: 3 days

12 years agoAdd the sys directory we create to the list of items to clean.
andrew [Sun, 24 Jun 2012 09:19:24 +0000 (09:19 +0000)]
Add the sys directory we create to the list of items to clean.

12 years agoIn a complete lack of foresight on my part, my previous commit broke
adrian [Sun, 24 Jun 2012 08:47:19 +0000 (08:47 +0000)]
In a complete lack of foresight on my part, my previous commit broke
the assumption that ath_softc doesn't change size based on build time
configuration.

I picked up on this because suddenly radar stuff didn't work; and
although the ath_dfs code was setting sc_dodfs=1, the main ath driver
saw sc_dodfs=0.

So for now, include opt_ath.h in driver source files.  This seems like
the sane thing to do anyway.

I'll have to do a pass over the code at some later stage and turn
the radiotap TX/RX structs into malloc'ed memory, rather than in-line
inside of ath_softc.  I'd rather like to keep ath_softc the same
layout regardless of configuration parameters.

Pointy hat to:  adrian

12 years agoCreate a symlink to sys/elf32.h, sys/elf64.h, and sys/elf_common.h.
andrew [Sun, 24 Jun 2012 08:43:34 +0000 (08:43 +0000)]
Create a symlink to sys/elf32.h, sys/elf64.h, and sys/elf_common.h.

When building libelf in the bootstrap stage this would include the tree
versions of, for example, sys/_types.h. This would work as long as the
tree's version of this file was close enough to the system's version of
the file. If, however, there was a change in the tree such that the location
of a typedef was moved this would cause problems. In this case the version
of sys/_types.h in the tree no longer defines __wchar_t and expects it to
to be defined in machine/_types.h, however we pick up machine/_types.h from
the system and find it is not defined there. The solution is to restrict the
parts of sys er include from the tree to those that are needed.

This fixes the recent Tinderbox failure.

Pointy Hat to: andrew

12 years agoShuffle these initialisations to where they should be.
adrian [Sun, 24 Jun 2012 08:28:06 +0000 (08:28 +0000)]
Shuffle these initialisations to where they should be.

12 years agoChange the ath_dfs_process_phy_err() method to take an mbuf rather than
adrian [Sun, 24 Jun 2012 08:09:06 +0000 (08:09 +0000)]
Change the ath_dfs_process_phy_err() method to take an mbuf rather than
a buffer pointer.

For large radar pulses, the AR9130 and later will return a series of
FFT results for software processing.  These can overflow a single 2KB
buffer on longer pulses.  This would result in undefined buffer behaviour.

12 years agoFix mandoc "Oc breaks Op" warning.
joel [Sun, 24 Jun 2012 07:54:28 +0000 (07:54 +0000)]
Fix mandoc "Oc breaks Op" warning.

12 years agoRemove the assert added in r237286
eadler [Sun, 24 Jun 2012 07:03:22 +0000 (07:03 +0000)]
Remove the assert added in r237286
The use of assertions in libraries is not widely accepted.

Requested by: bde, scottl
Approved by: cperciva
MFC after: 3 days
X-MFC-With: r237286