projects/modsched/freebsd.git
6 years agoAdded global scheduler functions of linux and FreeBSD to the framework. p/jkrieg/kbs_fbsd
Jens Krieg [Thu, 12 Dec 2013 13:26:11 +0000 (14:26 +0100)]
Added global scheduler functions of linux and FreeBSD to the framework.
The functions has been sorted by purpose.

6 years agoOS -> FW: All OS depended functions will be placed in <platform>_sched.c
Jens Krieg [Wed, 11 Dec 2013 16:46:31 +0000 (17:46 +0100)]
OS -> FW: All OS depended functions will be placed in <platform>_sched.c
FW -> OS: All framework function will be placed in os_<domain>.c

FreeBSD compiling.

6 years agoStructure again changed. FBSD not running!
Jens Krieg [Wed, 11 Dec 2013 14:19:52 +0000 (15:19 +0100)]
Structure again changed. FBSD not running!

6 years agoRemoved debug directory and made minor changes in source code.
Anton Eisenbraun [Tue, 10 Dec 2013 12:56:50 +0000 (13:56 +0100)]
Removed debug directory and made minor changes in source code.

6 years agoMerged current BertOS progress into new directory structure.
Jens Krieg [Mon, 9 Dec 2013 13:07:13 +0000 (14:07 +0100)]
Merged current BertOS progress into new directory structure.
Framework included into FreeBSD "Make" routine.

6 years agoAdded framework files.
Anton Eisenbraun [Tue, 26 Nov 2013 16:43:00 +0000 (17:43 +0100)]
Added framework files.

6 years agoFinal changes.
Jens Krieg [Tue, 26 Nov 2013 15:50:36 +0000 (16:50 +0100)]
Final changes.

6 years agoSMP running without balancing. Idle and general tasks have old priority
Jens Krieg [Tue, 26 Nov 2013 12:20:30 +0000 (13:20 +0100)]
SMP running without balancing. Idle and general tasks have old priority
values.

6 years agoSmp boots faster.
Jens Krieg [Wed, 20 Nov 2013 16:52:15 +0000 (17:52 +0100)]
Smp boots faster.

6 years agoSingle core works. SMP thread stealing works as well but still slow.
Jens Krieg [Wed, 20 Nov 2013 14:35:52 +0000 (15:35 +0100)]
Single core works. SMP thread stealing works as well but still slow.

6 years agoGot rid of priorities.
Jens Krieg [Mon, 18 Nov 2013 15:54:50 +0000 (16:54 +0100)]
Got rid of priorities.

6 years agoFurther striped down of the kernel.
Jens Krieg [Wed, 13 Nov 2013 17:44:39 +0000 (18:44 +0100)]
Further striped down of the kernel.

6 years agoRemoved load tracking, score calculation, priority scaling. Added kernel
Anton Eisenbraun [Wed, 6 Nov 2013 17:14:22 +0000 (18:14 +0100)]
Removed load tracking, score calculation, priority scaling. Added kernel
config.

6 years agoFixed bug where idle task were placed in the wrong queue. Also added
Anton Eisenbraun [Wed, 6 Nov 2013 14:04:05 +0000 (15:04 +0100)]
Fixed bug where idle task were placed in the wrong queue. Also added
changes from previous commit, since there went something wrong in the
commit process.

6 years agoRemoved realtime runqueue and levels of still remaining. System still
Anton Eisenbraun [Tue, 5 Nov 2013 17:35:44 +0000 (18:35 +0100)]
Removed realtime runqueue and levels of still remaining. System still
runable.

6 years agoAdded originals of the patched files and excluded amd64/compile from commit.
Anton Eisenbraun [Wed, 30 Oct 2013 16:37:59 +0000 (17:37 +0100)]
Added originals of the patched files and excluded amd64/compile from commit.

6 years agoApplied patches to make compiling on gentoo fbsd possible.
Anton Eisenbraun [Wed, 30 Oct 2013 16:20:58 +0000 (17:20 +0100)]
Applied patches to make compiling on gentoo fbsd possible.

7 years agoWe think we're ready for the 9.1-RELEASE builds.
kensmith [Fri, 30 Nov 2012 19:36:20 +0000 (19:36 +0000)]
We think we're ready for the 9.1-RELEASE builds.

Approved by: re (implicit)

7 years agoMerge r243708:
kensmith [Fri, 30 Nov 2012 19:35:01 +0000 (19:35 +0000)]
Merge r243708:
Guess when we'll be ready to announce 9.1-RELEASE.

Approved by: re (implicit)

7 years agoRemove stale documents.
hrs [Fri, 30 Nov 2012 16:18:26 +0000 (16:18 +0000)]
Remove stale documents.

Approved by: re (implicitly)

7 years ago- Bump versions and revert XML migration of the release documents in
hrs [Fri, 30 Nov 2012 16:15:35 +0000 (16:15 +0000)]
- Bump versions and revert XML migration of the release documents in
  releng/9.1 branch.  The doc tree release/9.1.0 for this release still uses
  SGML toolchain[1].

- Add SVNROOT{BASE,SRC,DOC,PORTS} for subversion repository URLs and
  BRANCH{SRC,DOC,PORTS} for the branches to generate-release.sh, and
  remove -p, -r, -d options.  The revision to be built should be specified
  in the URL.

- Add {WORLD,KERNEL}_FLAGS to generate-release.sh.  These were supported in
  the old release build framework.

- Disable to use binary package for docproj port during a release build.
  This package should be built successfully.

Pointy hat to: hrs [1]
Approved by: re (implicitly)

7 years agoFix multiple Denial of Service vulnerabilities with named(8).
simon [Thu, 22 Nov 2012 22:52:15 +0000 (22:52 +0000)]
Fix multiple Denial of Service vulnerabilities with named(8).

Fix insufficient message length validation for EAP-TLS messages.

Fix Linux compatibility layer input validation error.

Security: FreeBSD-SA-12:06.bind
Security: FreeBSD-SA-12:07.hostapd
Security: FreeBSD-SA-12:08.linux
Security: CVE-2012-4244, CVE-2012-5166, CVE-2012-4445, CVE-2012-4576
Approved by: re
Approved by: security-officer

7 years agoMFC r242514:
eadler [Sat, 10 Nov 2012 06:05:04 +0000 (06:05 +0000)]
MFC r242514:
Revert the change that makes less default.

Since I've committed this I've receieved roughly an equal
amount of email thanking me for making this change
and asking me to revert it.

I've resisted making this change because
new users tend to prefer less over more
and these users are the least likely to know
how to change the PAGER on their own.

Approved by: cperciva (implicit)
Approved by: re (kib)

7 years agoReady for 9.1-RC3...
kensmith [Sun, 28 Oct 2012 16:21:46 +0000 (16:21 +0000)]
Ready for 9.1-RC3...

Approved by: re (implicit)

7 years agoMFC of r242187:
hrs [Sat, 27 Oct 2012 20:17:59 +0000 (20:17 +0000)]
MFC of r242187:

Fix an issue when ipv6_enable=YES && ipv6_gateway_enable=YES which could
prevent rtadvd(8) from working as intended.

Approved by: re (kib)

7 years agoMFC of 240586 and 240587 to unbreak release building:
hrs [Sat, 27 Oct 2012 19:40:25 +0000 (19:40 +0000)]
MFC of 240586 and 240587 to unbreak release building:

Update generate-release.sh script:

 - Use svn for ports and doc trees
 - When installing a binary textproc/docproj package,
   switch pkg_add(1) to pkg(8) [1]

Approved by: re (implicitly)

7 years agoUpdate branch tag from RELENG_9 to RELENG_9_1.
kensmith [Wed, 24 Oct 2012 02:21:39 +0000 (02:21 +0000)]
Update branch tag from RELENG_9 to RELENG_9_1.

Approved by: re (implicit)

7 years agoUpdate for being on releng/9.1 (RELENG_9_1).
kensmith [Wed, 24 Oct 2012 02:20:07 +0000 (02:20 +0000)]
Update for being on releng/9.1 (RELENG_9_1).

Approved by: re (implicit)

7 years agoMFC r241976: Add the release package directory for 9.1-RELEASE.
kensmith [Wed, 24 Oct 2012 02:09:21 +0000 (02:09 +0000)]
MFC r241976: Add the release package directory for 9.1-RELEASE.

Approved by: re (implicit)

7 years agoMFC: r241679
marius [Sun, 21 Oct 2012 12:53:33 +0000 (12:53 +0000)]
MFC: r241679

It turns out that as documented, PCF8563_R_SECOND_VL (i.e. battery low)
doesn't automatically clear when VDD rises above Vlow again and needs to be
cleared manually. However, apparently this needs all of the time registers
to be set, i.e. pcf8563_settime(), and not just PCF8563_R_SECOND in order
for PCF8563_R_SECOND_VL to stick. Thus, we just issue a warning during
pcf8563_attach() rather than failing with ENXIO in case it is set.

Approved by: re (kib)

7 years agoMFC r241753:
delphij [Sat, 20 Oct 2012 07:39:47 +0000 (07:39 +0000)]
MFC r241753:

Integrate changes from LSI vendor driver 10.80.00.005 to FreeBSD.

PR: kern/172833
Submitted by: "Charles O'Donnell" <cao bus net>
Approved by: re (kib)

7 years agoMFC r241096:
gabor [Thu, 11 Oct 2012 15:49:42 +0000 (15:49 +0000)]
MFC r241096:
  - Rename files to track the XML migration in the doc tree and make the
    release notes build again

Approved by: re (hrs)

7 years agoMFC r241414:
delphij [Wed, 10 Oct 2012 20:37:57 +0000 (20:37 +0000)]
MFC r241414:

Upgrade to 9.8.3-P4:

Prevents a lockup when queried a deliberately constructed combination
of records. [CVE-2012-5166]

For more information: https://kb.isc.org/article/AA-00801

Approved by: re (kib)

7 years agoMerge r240985 from head:
glebius [Tue, 2 Oct 2012 13:03:11 +0000 (13:03 +0000)]
Merge r240985 from head:
  Fix bug in TCP_KEEPCNT setting, which slipped in in the last round
  of reviewing of r231025.

  Unlike other options from this family TCP_KEEPCNT doesn't specify
  time interval, but a count, thus parameter supplied doesn't need
  to be multiplied by hz.

  Reported & tested by:       amdmi3

Approved by: re (kib)

7 years agoMFC r240917:
mav [Mon, 1 Oct 2012 10:52:10 +0000 (10:52 +0000)]
MFC r240917:
Reduce delays in several wait loops from 10ms to 10us, same is it is
done in Linux. This substantially increases graphics performance on Ivy
Bridge.

Submitted by: avg@
Reviewed by: kib@
Approved by: re (kib)

7 years agoMFC r240884:
mav [Thu, 27 Sep 2012 08:24:34 +0000 (08:24 +0000)]
MFC r240884:
Fix panic caused by wrong pointer dereference, left after pin sense rewrite
at r230551.

Also while there, make sense polling use reported for each node separately
instead of reporting accumulated total status.

Approved by: re (kib)

7 years agoReady for 9.1-RC2 builds.
kensmith [Wed, 26 Sep 2012 14:29:15 +0000 (14:29 +0000)]
Ready for 9.1-RC2 builds.

Approved by: re (implicit)

7 years agoMFC 240729 (dougb):
delphij [Sat, 22 Sep 2012 08:48:26 +0000 (08:48 +0000)]
MFC 240729 (dougb):

Upgrade to 9.8.3-P3:

Prevents a crash when queried for a record whose RDATA exceeds
65535 bytes.

Prevents a crash when validating caused by using "Bad cache" data
before it has been initialized.

ISC_QUEUE handling for recursive clients was updated to address
a race condition that could cause a memory leak. This rarely
occurred with UDP clients, but could be a significant problem
for a server handling a steady rate of TCP queries.

A condition has been corrected where improper handling of
zero-length RDATA could cause undesirable behavior, including
termination of the named process.

For more information: https://kb.isc.org/article/AA-00788

Approved by: re (kib)

7 years agoMFC r240079:
delphij [Thu, 20 Sep 2012 17:30:26 +0000 (17:30 +0000)]
MFC r240079:

Update arcmsr(4) to vendor version 1.20.00.25.

Many thanks to Areca for continuing to support FreeBSD.

Submitted by: Ching-Lung Huang <ching2048 areca com tw>
Approved by: re (kib)

7 years ago- Fix release notes build on releng/9.1 [1]
gjb [Wed, 19 Sep 2012 21:44:39 +0000 (21:44 +0000)]
- Fix release notes build on releng/9.1 [1]
- MFC r240508, r240516, r240519 (gabor):
  o Update releng/9.1/release/doc files post-XML conversion.
  o This commit fixes most of the 9-STABLE release build problems.
- Close colspec tags to conform to XML standards. [1]
- Convert installation article to XML stanards. [1]

[1] - These are direct commits to releng/9.1

Approved by: re (hrs)

7 years agoMFC: r240476
jkim [Wed, 19 Sep 2012 19:14:21 +0000 (19:14 +0000)]
MFC: r240476

Do not change owner, group, or mode when package database directory
and its contents are created with pkg_add(1).  It may happen when the
packing list contains @owner, @group, or @mode.

Approved by: re (kib), portmgr (bapt)

7 years agoMFC r240412:
emaste [Tue, 18 Sep 2012 13:13:28 +0000 (13:13 +0000)]
MFC r240412:

  According to a clarification at http://austingroupbugs.net/view.php?id=503
  ptsname may set errno, so avoid saving and restoring errno across the
  function.

PR: standards/171572
Approved by: re
Sponsored by: ADARA Networks

7 years agoMFC r240410:
emaste [Tue, 18 Sep 2012 13:03:00 +0000 (13:03 +0000)]
MFC r240410:

  Avoid mapping ENOENT to ENOTDIR for non-existent path components.

  The ENOTDIR mapping was introduced in r235266 for kern/128933 based on
  an interpretation of the somewhat ambiguous language in the POSIX realpath
  specification.  The interpretation is inconsistent with Solaris and Linux,
  a regression from 9.0, and does not appear to be permitted by the
  description of ENOTDIR:

       20 ENOTDIR Not a directory.  A component of the specified pathname
               existed, but it was not a directory, when a directory was
               expected.

PR: standards/171577
Approved by: re
Sponsored by: ADARA Networks

7 years agoMerge r238182 by theraven@:
glebius [Mon, 17 Sep 2012 19:25:24 +0000 (19:25 +0000)]
Merge r238182 by theraven@:
  Restore the __collate_load_error global that was accidentally removed in the
  xlocale refactoring.

Approved by: re (kib)

7 years agoMFC r240514:
jimharris [Mon, 17 Sep 2012 16:39:32 +0000 (16:39 +0000)]
MFC r240514:

isci(4): Fix SCSI/ATA translation for SCSI_WRITE_BUFFER w/ mode==0x7
(download microcode with offsets, save, and activate).

SATI translation layer was incorrectly using allocation length instead
of blocks, and was constructing the ATA command incorrectly.

Also change #define to specify that the 512 block size here is
specific for DOWNLOAD_MICROCODE, and does not relate to the device's
logical block size.

Approved by:   re (kib)

7 years agoMFC r240465:
mav [Sun, 16 Sep 2012 11:02:22 +0000 (11:02 +0000)]
MFC r240465:
Add global and per-module sysctls/tunables to enable/disable metadata taste.
That should help to handle some cases when disk has some RAID metadata that
should be ignored, especially during boot.

Approved by: re (kib)

7 years agoMFC r240286:
mav [Wed, 12 Sep 2012 10:53:08 +0000 (10:53 +0000)]
MFC r240286:
At least from A70M FCH chipsets AMD started to use their real vendor ID
(1022) in HPET. But according to report they still haven't fixed problem
with level-triggered interrupts.
Make workaround used for earlier chipsets apply to this new ID also.

PR: amd64/171355
Approved by: re (kib)

7 years agoMFC r226738:
pjd [Tue, 11 Sep 2012 13:58:03 +0000 (13:58 +0000)]
MFC r226738:

Don't look for includes in userland directories. There is no need for it
and it breaks building the kernel without userland sources.

Reported by: andre, imp
Approved by: re (kib)

7 years agoMFH (r240107, r240111): warn against using these for cryptographic purposes.
des [Tue, 11 Sep 2012 11:05:32 +0000 (11:05 +0000)]
MFH (r240107, r240111): warn against using these for cryptographic purposes.

Approved by: re (kib)

7 years agoMFC r240315:
jpaetzel [Mon, 10 Sep 2012 16:25:52 +0000 (16:25 +0000)]
MFC r240315:

Add TRIM support, enabled by default.
Fix a bug installing components from a localPath.
Allow autosizing of any partition, not just the last partition.
Adjust how ZFS is laid out to work with Boot Environments.

Submitted by: kmoore
Approved by: re (kib)
Obtained from: PC-BSD

7 years agoMFC: r239663, r239664
bapt [Mon, 10 Sep 2012 08:14:58 +0000 (08:14 +0000)]
MFC: r239663, r239664

change ALWAYS_ASSUME_YES to ASSUME_ALWAYS_YES for consistency with pkg(8)
if not on a tty prompt about the missing pkg(8) but default on 'no' except if
ASSUME_ALWAYS_YES is set

Approved by: re (kib)

7 years agoMFC r230454 (pjd):
mm [Sun, 9 Sep 2012 20:26:19 +0000 (20:26 +0000)]
MFC r230454 (pjd):
Use provided name when allocating ksid domain. It isn't really used
on FreeBSD, but should fix a panic when pool is imported from another OS
that is using this.

MFC r240162 (mm):
Make r230454 more readable and vendor-like.

PR: kern/171380
Approved by: re (kib)

7 years agoMFC r239125:
kib [Sat, 8 Sep 2012 16:47:52 +0000 (16:47 +0000)]
MFC r239125:
Do not apply errata 721 workaround when under hypervisor, since
typical hypervisor does not implement access to the required MSR,
causing #GP on boot.

PR: amd64/170388
Approved by: re (kensmith)

7 years agoMFC r232197 (by phk):
kib [Sat, 8 Sep 2012 16:45:48 +0000 (16:45 +0000)]
MFC r232197 (by phk):
Also call the low-level driver if ->c_iflag & (IXON|IXOFF|IXANY) changes.

Approved by: re (kensmith)

7 years agoMFC r238935,238960:
bz [Fri, 7 Sep 2012 09:45:08 +0000 (09:45 +0000)]
MFC r238935,238960:

  Properly apply #ifdef INET and leave a comment that we are (will) apply
  delayed IPv6 checksum processing in ip6_output.c when doing IPsec.

  In case of IPsec he have to do delayed checksum calculations before
  adding any extension header, or rather before calling into IPsec
  processing as we may send the packet and not return to IPv6 output
  processing here.

PR: kern/170116
Approved by: re (kib)

7 years agoMFC r238877-238878:
bz [Fri, 7 Sep 2012 09:22:11 +0000 (09:22 +0000)]
MFC r238877-238878:
 Fix a comment that we do not have an SA yet but need to acquire one.
 For consistency put the IPsec comment iside the #fidef section.

MFC r238934
  Improve the should-never-hit printf to ease debugging in case we'd ever hit
  it again when doing the delayed IPv6 checksum calculations.

Approved by: re (kib)

7 years agoMFS r240157 (MFC r235638,r239348):
mckay [Fri, 7 Sep 2012 00:20:46 +0000 (00:20 +0000)]
MFS r240157 (MFC r235638,r239348):
- Work around failure to compile on FreeBSD 7.x machines.
- Correct a regression introduced during the import of file(1) 5.11.

PR: 170415
Reviewed by: obrien
Approved by: re (kib)

7 years agoMFC rr240007:
tuexen [Thu, 6 Sep 2012 18:31:56 +0000 (18:31 +0000)]
MFC rr240007:

Fix a typo which results in RTT to be off by a factor of 10, if the RTT is
larger than 1 second.

Approved by: re (kib)

7 years agoMFC r239889:
jimharris [Thu, 6 Sep 2012 16:38:55 +0000 (16:38 +0000)]
MFC r239889:

Do not call sati_check_data_io() for SATI_UNMAP sequences.

This routine is intended only for commands such as INQUIRY where
the controller may fill out a smaller amount of data than allocated
by the host.

The end result of this bug was that isci(4) would report non-zero
resid for successful SCSI_UNMAP commands.

Sponsored by:   Intel
Approved by: re (kib)

7 years agoMFH (r239836): align the boot partition on a 4 kB boundary.
des [Thu, 6 Sep 2012 16:27:59 +0000 (16:27 +0000)]
MFH (r239836): align the boot partition on a 4 kB boundary.

Approved by: re (kib)

7 years agoMFS r239854 (including MFC r239601, r239602, r239621):
jamie [Wed, 29 Aug 2012 18:40:12 +0000 (18:40 +0000)]
MFS r239854 (including MFC r239601, r239602, r239621):

 Remember that I'm using length-defined strings in parameters:

  Don't include the null terminator when recomputing the parameter
  length when stripping the netmask from IP addresses.  This was
  causing later addresses in a comma-separated string to disappear.

  Use memcpy instead of strcpy.  This could just cause Bad Things.

  Add a null byte when comma-combining array parameters.

  Pre-separate IP addresses passed on the command line, so they can be
  properly parsed for interface prefixes and netmask suffixes.  This was
  already done for the old-style (fixed) command line, but missed for
  the new-style.

PR: 170832
Approved by: re (kib)

7 years agoMFC r238418: acpi_cpu: separate a notion of current deepest
avg [Wed, 29 Aug 2012 11:28:59 +0000 (11:28 +0000)]
MFC r238418: acpi_cpu: separate a notion of current deepest
allowed+available Cx level from a user-set persistent limit on the said
level

Reviewed by: jhb, gianni, sbruno
Tested by: sbruno, Vitaly Magerya <vmagerya@gmail.com>
Approved by: re(kib)

7 years agoMFC: r238933
fjoe [Tue, 28 Aug 2012 17:09:34 +0000 (17:09 +0000)]
MFC: r238933

- Change back "d_ofs" to int8_t to not pessimize padding and size of "struct puc_cfg".
- Use "puc_config_moxa" for Moxa boards that need d_ofs greater than 0x7f

Approved by: re (kib)

7 years agoMFC r239665:
jimharris [Mon, 27 Aug 2012 18:14:22 +0000 (18:14 +0000)]
MFC r239665:

  Clear freeze bit before calling xpt_release_devq.

  This ensures that any ccbs which immediately start during the call to
  xpt_release_devq see an accurate picture of the frozen_lun_mask.

Sponsored by: Intel
Approved by: re (kib)

7 years agoMFC r239655:
jimharris [Mon, 27 Aug 2012 18:12:08 +0000 (18:12 +0000)]
MFC r239655:

  Fix scsi_da's BIO_DELETE->SCSI_UNMAP translation to use correct local
  variable when determining various sizes related to SCSI UNMAP block
  descriptor lists.

Sponsored by: Intel
Approved by: re (kib)

7 years agoMFC r239545:
jimharris [Mon, 27 Aug 2012 18:10:25 +0000 (18:10 +0000)]
MFC r239545:

  Fix/add isci(4) support for SCSI UNMAP to ATA DSM translation.

  This addresses kernel panic observed when sending SCSI UNMAP
  commands to SATA disks attached to isci(4).

  1) Flesh out callback routines to allocate/free buffers needed for
     translating SCSI UNMAP data to ATA DSM data.
  2) Add controller-level pool for storing buffers previously allocated
     for UNMAP translation, to lessen chance of no buffer available
     under memory pressure.
  3) Ensure driver properly handles case where buffer pool is empty
     and contigmalloc returns NULL.

Sponsored by: Intel
Approved by: re (kib)

7 years agoMFS: r239148
imp [Sun, 26 Aug 2012 05:11:52 +0000 (05:11 +0000)]
MFS: r239148

    r239148 | imp | 2012-08-08 14:21:33 -0600 (Wed, 08 Aug 2012) | 8 lines

    Make the addition of the d_gone binary compatible.  This allows
    storage drivers compiled for 9.0 to work on 9.1 and preserves the ABI
    for disks.

    Reviewed by:    scottl, ken
    Approved by:    re@ (kensmith)
    MFS after:      releng/9.1 gets sorted

This change fixes the binary compatibility problems with additions to
the disk(9) API.  Disk drivers compiled against 9.0 will not work on
9.1-RC1, but should start working after this change or in 9.1-RC2 and
later.  Disk drivers should not be linked against 9.1-RC1 for
distribution, except specifically for testing on RC1.  This was
planned for RC1, but it was delayed due to circumstancs beyond my
control.

Approved by: re@ (kensmith)

7 years agoMFC r239649:
bschmidt [Fri, 24 Aug 2012 14:50:44 +0000 (14:50 +0000)]
MFC r239649:
Document update of 802.11s IE identifiers.

Approved by: re (kib)

7 years agoMFC r231187:
bschmidt [Fri, 24 Aug 2012 06:56:44 +0000 (06:56 +0000)]
MFC r231187:
Update the 802.11s IE numbers to represent the latest 802.11 amendment
standard.

This update breaks compatibility with older mesh setups but is necessary as
the previous IDs are used by another amendment leading to unexpected results
when trying to associate with an accesspoint using the affected IDs.

Approved by: re (kib)

7 years agoMFC r239228: Build modules along with the XENHVM kernels.
cperciva [Fri, 24 Aug 2012 05:36:37 +0000 (05:36 +0000)]
MFC r239228: Build modules along with the XENHVM kernels.

Approved by: re (kib)

7 years agoMFC r239464:
delphij [Thu, 23 Aug 2012 04:57:56 +0000 (04:57 +0000)]
MFC r239464:
>
As of r232844 we no longer need the maxpoll 9 workaround.
>
Approved by: re (kib)

7 years agoMFC r239346:
lstewart [Wed, 22 Aug 2012 01:28:16 +0000 (01:28 +0000)]
MFC r239346:

The TCP PAWS fix for kernels with fast tick rates (r231767) changed the TCP
timestamp related stack variables to reference ms directly instead of ticks. The
h_ertt(4) Khelp module relies on TCP timestamp information in order to calculate
its enhanced RTT estimates, but was not updated as part of r231767.

Consequently, h_ertt has not been calculating correct RTT estimates since
r231767 was comitted, which in turn broke all delay-based congestion control
algorithms because they rely on the h_ertt RTT estimates.

Fix the breakage by switching h_ertt to use tcp_ts_getticks() in place of all
previous uses of the ticks variable. This ensures all timestamp related
variables in h_ertt use the same units as the TCP stack and therefore results in
meaningful comparisons and RTT estimate calculations.

Reported & tested by: Naeem Khademi (naeemk at ifi uio no)
Discussed with: bz
Approved by: re (kib)

7 years agoMFC r239470: Do not call process_nodelete with NULL object pointer.
kan [Tue, 21 Aug 2012 22:42:46 +0000 (22:42 +0000)]
MFC r239470: Do not call process_nodelete with NULL object pointer.

The place where the function is called can be reached if object loading
and relocation fails too, in which case obj pointer will be NULL. Do not
call process_nodelete then, or crash will follow.

Pointy hat to: kan
Approved by: re (kib)

7 years agoMFC r238673:
mav [Tue, 21 Aug 2012 12:37:37 +0000 (12:37 +0000)]
MFC r238673:
Use 16bit PIO instead of 32bit in case of misaligned buffer.
It fixes kernel panic during CD write with cdrecord on sparc64.

Approved by: re (kib)

7 years agoMFC r232844: Remove extraneous log message
emaste [Tue, 21 Aug 2012 11:35:54 +0000 (11:35 +0000)]
MFC r232844: Remove extraneous log message

  When ntp switched between PLL and FLL mode it produced a log message
  "kernel time sync status change %04x".  This issue is reported in ntp
  bug 452[1] which claims that this behaviour is normal and the log
  message isn't necessary.  I'm not sure exactly when it was removed, but
  it's gone in the latest ntp release (4.2.6p5).

  [1] http://bugs.ntp.org/show_bug.cgi?id=452

Approved by: re

7 years agoMFC r238718: Quirk MS keyboard so that function keys work
emaste [Tue, 21 Aug 2012 11:34:40 +0000 (11:34 +0000)]
MFC r238718: Quirk MS keyboard so that function keys work

  The function keys on a Microsoft Natural Egronomic Keyboard 4000 have been
  repurposed as "Help", "Undo", "Redo" etc., and a special "F Lock" key is
  required to return them to their normal purpose.

  This change enables the UQ_KBD_BOOTPROTO quirk for the MS Natural 4000
  keyboard to get the keys working again.  More extensive changes to the USB
  keyboard infrastructure would be needed to fully support the "F Lock" mode
  and the extended keys on this keyboard.

PR:             usb/116947
Approved by: re

7 years agoMFC: r239079
marius [Tue, 21 Aug 2012 11:17:11 +0000 (11:17 +0000)]
MFC: r239079

Merge r236494 from x86:

Isolate the global TTE list lock from data and other locks to prevent false
sharing within the cache.

Approved by: re (kib)

7 years agoMFC r238631:
delphij [Tue, 21 Aug 2012 09:55:57 +0000 (09:55 +0000)]
MFC r238631:

Import an updated version of moduli(5) manual page from OpenBSD.

Approved by: re (kib)

7 years agoMFC: r239089
marius [Tue, 21 Aug 2012 09:43:03 +0000 (09:43 +0000)]
MFC: r239089

- Merge from NetBSD:
  When issuing a non-DMA command, make sure to set the "remaining length of
  command to be transferred via DMA" (sc_cmdlen) to zero up-front, otherwise
  we might get confused on command competition interrupt (no DMA active but
  still data left to transfer).
- Implement handling of MSG_IGN_WIDE_RESIDUE which some targets produce, as
  just rejecting these leads to a resend and disconnect loop.
  Reported and tested by: mjacob

Approved by: re (kib)

7 years agoMFC r239169:
delphij [Tue, 21 Aug 2012 09:05:23 +0000 (09:05 +0000)]
MFC r239169:

RFC 2289 requires all hashes be stored in little endian format before
folding to 64 bits, while SHA1 code is big endian.  Therefore, a bswap32
is required before using the value.

Without this change, the implementation does not conform to test vector
found in RFC 2289.

PR: bin/170519
Submitted by: Arthur Mesh <arthurmesh gmail com> (with changes)
Approved by: re (kib)

7 years agoMFC r239084:
delphij [Tue, 21 Aug 2012 08:31:30 +0000 (08:31 +0000)]
MFC r239084:

Add PCI IDs for various new High Point RocketRAID 43xx and 3xxx devices.

Obtained from: FreeNAS
Approved by: re (implicit, follow-up for r239460)

7 years agoMFC r239084
delphij [Mon, 20 Aug 2012 18:27:02 +0000 (18:27 +0000)]
MFC r239084

Add PCI IDs for various new High Point RocketRAID 43xx and 3xxx devices.

Obtained from: FreeNAS
Approved by: re (kib)

7 years agoMFC r238997:
delphij [Mon, 20 Aug 2012 18:10:48 +0000 (18:10 +0000)]
MFC r238997:

Correct a typo.

Reported by: Sascha Wildner <swildner dragonflybsd org>
Reviewed by: scottl
Approved by: re (kib)

7 years agoMFH r236106: avoid segfault with SSH 1 keys
des [Mon, 20 Aug 2012 17:50:20 +0000 (17:50 +0000)]
MFH r236106: avoid segfault with SSH 1 keys

Approved by: re (kib)

7 years agoMFC r239052:
tuexen [Mon, 20 Aug 2012 17:36:10 +0000 (17:36 +0000)]
MFC r239052:
Fix a refcount issue. The caller only decrements if stcb is NULL.

Approved by: re (kib)

7 years agoMFC r239041:
tuexen [Mon, 20 Aug 2012 17:06:50 +0000 (17:06 +0000)]
MFC r239041:

Fix a bug reported by Simon L. B. Nielsen:
If an SCTP endpoint receives an ASCONF with a wildcard
lookup address and incorrect verification tag, the system
crashes.
This bug was found by Shaun Colley.

Approved by: re (kib)

7 years agoMFC r239253: Pospone the DF_1_NODELETE processing until object DAG is
kan [Mon, 20 Aug 2012 16:57:27 +0000 (16:57 +0000)]
MFC r239253: Pospone the DF_1_NODELETE processing until object DAG is
fully loaded.

Trying to up the reference from the load loop risks missing
dependencies that have not been loaded yet.

Approved by: re (kib)

7 years agoMFC r239021:
jimharris [Mon, 20 Aug 2012 16:55:55 +0000 (16:55 +0000)]
MFC r239021:

In virstor_ctl_stop(), check for a valid softc before trying to update
metadata.

Approved by: re (kib)

7 years agoMFC r239233:
wblock [Mon, 20 Aug 2012 16:37:41 +0000 (16:37 +0000)]
MFC r239233:

Correct description of minfree to kilobytes rather than blocks.

PR: 125921
Submitted by: Andre Albsmeier <Andre.Albsmeier@siemens.com>
Approved by: re (implicit)

7 years agoMFC r239095: Do not add handler to event handlers list until ithread
kan [Mon, 20 Aug 2012 15:19:34 +0000 (15:19 +0000)]
MFC r239095: Do not add handler to event handlers list until ithread
is created.

In rare event when fast and ithread interrupts share the same vector
and the fast handler was registered first, we can end up trying to
schedule the ithread that is not created yet. The kernel built with
INVARIANTS then triggers an assertion.

Change the order to create the ithread first and only then add the
handler that needs it to the interrupt event handlers list.

Approved by: re (kib)
Reviewed by: jhb

7 years agoMFC r239019: Parse notes only after object structure had been allocated.
kan [Mon, 20 Aug 2012 15:05:29 +0000 (15:05 +0000)]
MFC r239019: Parse notes only after object structure had been allocated.
Approved by: re (kib)

7 years agoMFC r239216, r239217:
gjb [Mon, 20 Aug 2012 13:07:22 +0000 (13:07 +0000)]
MFC r239216, r239217:

r239216:
 - Remove a leading space that breaks rendering

r239217:
 - Fix source dataset snapshot name in Example 15.
 - Bump date.

Approved by: re (kib)

7 years agoMFC r238969:
mav [Mon, 20 Aug 2012 12:53:29 +0000 (12:53 +0000)]
MFC r238969:
Fix kernel panic on `camcontrol reset` for specific target, caused by
uninitialized cm_targ in mpssas_action_resetdev().

Approved by: re (kib)

7 years agoPrep for 9.1-RC1 builds.
kensmith [Wed, 8 Aug 2012 18:39:29 +0000 (18:39 +0000)]
Prep for 9.1-RC1 builds.

Approved by: re (implicit)

7 years agoUpdate __FreeBSD_version to reflect this being the 9.1-RELEASE release
kensmith [Mon, 6 Aug 2012 00:15:49 +0000 (00:15 +0000)]
Update __FreeBSD_version to reflect this being the 9.1-RELEASE release
branch.

Approved by: re (implicit)

7 years agoMFC r238622
emax [Thu, 2 Aug 2012 20:49:11 +0000 (20:49 +0000)]
MFC r238622

Allow to specify no source-address-selection policy

Approved by: re (kib)

7 years agoMFC r238599
emax [Thu, 2 Aug 2012 20:48:22 +0000 (20:48 +0000)]
MFC r238599

Return zero from get_addrselectpolicy() when no source-address-selection
policy is installed.

Approved by: re (kib)

7 years agoMFC r238795
ache [Thu, 2 Aug 2012 15:13:12 +0000 (15:13 +0000)]
MFC r238795

Try to avoid all files dependence on the modification time of the large
and often modified directory created symbolic links points to - it cause
unnecessary full rebuilds each time make runs when directory is changed.
So do it only if symbolic link does not exists, which usually means that
objdir is clean anyway.

Approved by:    re (kib)

7 years agoMFC 238164:
jhb [Mon, 30 Jul 2012 20:40:14 +0000 (20:40 +0000)]
MFC 238164:
Add another PS/2 keyboard PNP ID.  This ID is listed as
"Reserved by Microsoft" in the standard PNP ID table, but has been seen
in the wild on at least one laptop.

Approved by: re (kib)

7 years agoMFC r238755:
jimharris [Mon, 30 Jul 2012 15:30:42 +0000 (15:30 +0000)]
MFC r238755:

Add rmb() to tsc_read_##x to enforce serialization of rdtsc captures.

Intel Architecture Manual specifies that rdtsc instruction is not serialized,
so without this change, TSC synchronization test would periodically fail,
resulting in use of HPET timecounter instead of TSC-low.  This caused
severe performance degradation (40-50%) when running high IO/s workloads due to
HPET MMIO reads and GEOM stat collection.

Tests on Xeon E5-2600 (Sandy Bridge) 8C systems were seeing TSC synchronization
fail approximately 20% of the time.

Sponsored by: Intel
Approved by: re (kib)