]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agoMFC: Add kvm_getcptime(3).
jhb [Tue, 26 Aug 2008 18:00:06 +0000 (18:00 +0000)]
MFC: Add kvm_getcptime(3).

16 years agoMFC: Add kvm_getmaxcpu() and kvm_getpcpu().
jhb [Tue, 26 Aug 2008 17:57:38 +0000 (17:57 +0000)]
MFC: Add kvm_getmaxcpu() and kvm_getpcpu().

16 years agoMFC: Export 'struct pcpu' to userland w/o requiring _KERNEL.
jhb [Tue, 26 Aug 2008 17:48:10 +0000 (17:48 +0000)]
MFC: Export 'struct pcpu' to userland w/o requiring _KERNEL.

16 years agoMFC: Build a /usr/bin/ldd32 binary on amd64 and use it for 32-bit objects
jhb [Tue, 26 Aug 2008 17:36:38 +0000 (17:36 +0000)]
MFC: Build a /usr/bin/ldd32 binary on amd64 and use it for 32-bit objects
passed to ldd.  This enables tracing of 32-bit shared libraries.

16 years agoAnnounce cron(8) -m switch.
marck [Tue, 26 Aug 2008 10:20:07 +0000 (10:20 +0000)]
Announce cron(8) -m switch.

Approved by: blackend

16 years agoMFC r180943
mav [Mon, 25 Aug 2008 22:26:46 +0000 (22:26 +0000)]
MFC r180943

Don't use memcpy() to copy several bytes.
Store IDs is host order. It is not so important to bloat code for it.
Combine m_adj() and M_PREPEND() into single M_PREPEND().

16 years agoMFC r180850, r180932
mav [Mon, 25 Aug 2008 22:12:38 +0000 (22:12 +0000)]
MFC r180850, r180932

According to in_pcb.h protocol binding information has double locking.
It allows access it while list travercing holding only global pcbinfo lock.

16 years agoMFC r180305
mav [Mon, 25 Aug 2008 22:09:19 +0000 (22:09 +0000)]
MFC r180305

Improve approximation of style(9) in raw socket code.

16 years agoMFC r180851
mav [Mon, 25 Aug 2008 21:59:00 +0000 (21:59 +0000)]
MFC r180851

According to in_pcb.h protocol binding information has double locking.
It allows access it while list travercing holding only global pcbinfo lock.

16 years agoMFC r180828
mav [Mon, 25 Aug 2008 21:46:45 +0000 (21:46 +0000)]
MFC r180828

Add hash table lookup for a fully connected raw sockets.

This gives significant performance improvements when many raw sockets used.
Benchmarks of mpd handeling 1000 simultaneous PPTP connections show up to 50%
performance boost. With higher number of connections benefit becomes even
bigger. PopTop snd others should also get some benefits.

MFC r180833, r180874

According to in_pcb.h protocol binding information has double locking.
It allows access it while list travercing holding only global pcbinfo lock.
This relaxed locking noticably increses receive socket lookup performance.

16 years agoMFC r180305
mav [Mon, 25 Aug 2008 21:42:41 +0000 (21:42 +0000)]
MFC r180305

Improve approximation of style(9) in raw socket code.

16 years agoMFC r180836
mav [Mon, 25 Aug 2008 21:19:36 +0000 (21:19 +0000)]
MFC r180836

Increase UDBHASHSIZE from 16 to 128 items.
Previous value was chosen 10 years ago and not very effective now.
This change gives several percents speedup on 1000 L2TP mpd links.

16 years agoMFC r176819:
rodrigc [Mon, 25 Aug 2008 20:09:27 +0000 (20:09 +0000)]
MFC r176819:
Remove hacks to filter out MNT_ROOTFS, since r173252 of vfs_mount.c
filters out MNT_ROOTFS if it is passed from userspace -> kernel.

16 years agoMFC: Sync. BPF JIT compiler with HEAD.
jkim [Mon, 25 Aug 2008 19:06:02 +0000 (19:06 +0000)]
MFC: Sync. BPF JIT compiler with HEAD.

- r179967, r179968, r179977, r179978:
Rehash and clean up BPF JIT compiler macros to emit opcodes closer
to GNU as(1) generated codes and micro-optimize.

- r181644, r181645:
Replace number of stack usages with registers.  Note amd64 version
does not use stack any more because it has more registers.

- r181697:
Remove unnecessary jump instructions of zero offsets.

- r181700, r181848, r181853:
Correctly check unsignedness of all registers used for load instructions
with (in)direct offsets and sync. with sys/net/bpf_filter.c

- r181648, r181649, 181846:
Update copyrights, fix style(9) nits, remove unused macros, and make
JIT compilers compilable on user land.

16 years agoMFC: r182018
marius [Mon, 25 Aug 2008 18:36:19 +0000 (18:36 +0000)]
MFC: r182018

Avoid misaligned access of struct frame.

16 years agoMFC r181964
gallatin [Mon, 25 Aug 2008 17:58:15 +0000 (17:58 +0000)]
MFC r181964

Adapt mxge shims to detect phys contig jumbo frames
in RELENG_7

16 years agoWe're less than a week from code freeze for the 7.1-REL release cycle.
kensmith [Mon, 25 Aug 2008 17:33:43 +0000 (17:33 +0000)]
We're less than a week from code freeze for the 7.1-REL release cycle.
Start calling RELENG_7 "7.1-PRERELEASE" now to warn people that we're
nearing code freeze so things might get a bit bumpy.

16 years agoMFH: make obrien happy
des [Mon, 25 Aug 2008 16:37:58 +0000 (16:37 +0000)]
MFH: make obrien happy

16 years agoMFC r181965: Increase cryptotest tool initialization vector (IV) size.
raj [Mon, 25 Aug 2008 12:11:07 +0000 (12:11 +0000)]
MFC r181965: Increase cryptotest tool initialization vector (IV) size.

16 years agoMFC r181984: Don't clean before testing, since clean also removes
kientzle [Mon, 25 Aug 2008 02:18:12 +0000 (02:18 +0000)]
MFC r181984: Don't clean before testing, since clean also removes
results from previous tests.

16 years agoMFC r181979, r181981, r181985: Test -q option and make it actually work.
kientzle [Mon, 25 Aug 2008 02:14:52 +0000 (02:14 +0000)]
MFC r181979, r181981, r181985: Test -q option and make it actually work.

16 years agoMFC r181821, r181845: "./foo" now matches "foo"
kientzle [Mon, 25 Aug 2008 02:06:56 +0000 (02:06 +0000)]
MFC r181821, r181845: "./foo" now matches "foo"

16 years agoMFC r181750: various minor test updates
kientzle [Mon, 25 Aug 2008 02:01:47 +0000 (02:01 +0000)]
MFC r181750: various minor test updates

16 years agoMFC r181542,r181555: Minor test suite fixup so that the test suite
kientzle [Mon, 25 Aug 2008 01:55:40 +0000 (01:55 +0000)]
MFC r181542,r181555: Minor test suite fixup so that the test suite
works against the libarchive 1.x API.  Not needed by 7-STABLE, but
it's nice to minimize the differences between 7-STABLE and CURRENT.

16 years agoMFC: r181701, r182020
marius [Sun, 24 Aug 2008 18:28:13 +0000 (18:28 +0000)]
MFC: r181701, r182020

cosmetic changes and style fixes

16 years agoMFC: r180484
rodrigc [Sat, 23 Aug 2008 22:40:29 +0000 (22:40 +0000)]
MFC: r180484

Delete "update" from the mount options after we have set
MNT_UPDATE in fsflags.

Requested by: yar

16 years agoMFC r181843:
kaiw [Sat, 23 Aug 2008 17:40:35 +0000 (17:40 +0000)]
MFC r181843:

Fix a typo: According to the hid spec, Global item #3
is Physical Minimum.

16 years agoMFC r181841:
kaiw [Sat, 23 Aug 2008 17:38:24 +0000 (17:38 +0000)]
MFC r181841:

In the hid parser, if a INPUT/OUTPUT/FEATURE item is skipped, its
corresponding USAGE should be skipped as well.

Tested by:        Grzegorz Blach
PR:               usb/125941

16 years agoMFC r181844:
kaiw [Sat, 23 Aug 2008 17:35:15 +0000 (17:35 +0000)]
MFC r181844:

sc->sc_ibuf should be malloc'ed after quirks applied, as
sc->sc_isize might have changed.

16 years agoMFC r181839:
kaiw [Sat, 23 Aug 2008 17:32:43 +0000 (17:32 +0000)]
MFC r181839:

Re-add Microsoft Intellimouse 2.0 TWHEEL quirk.

Tested by:      Merritt Draney, Brian Cox
PR:             kern/123224
PR:             kern/123510

16 years agoMFH colldef and timedef for nb_NO, nn_NO.
des [Sat, 23 Aug 2008 13:25:56 +0000 (13:25 +0000)]
MFH colldef and timedef for nb_NO, nn_NO.

PR: conf/51920

16 years agoMFC r180753 by luoqi@:
mlaier [Sat, 23 Aug 2008 03:40:49 +0000 (03:40 +0000)]
MFC r180753 by luoqi@:
  Only enable MSI mappings for devices that use MSI.  Fixes interrupt loss
  on some nForce based boards.

PR:             kern/118842

16 years agoMFC: Back FreeBSD(98)'s copyright to 2003.
nyan [Fri, 22 Aug 2008 12:40:59 +0000 (12:40 +0000)]
MFC: Back FreeBSD(98)'s copyright to 2003.

16 years agoMFC: sync with HEAD.
luigi [Fri, 22 Aug 2008 10:18:18 +0000 (10:18 +0000)]
MFC: sync with HEAD.

BTW, the change in rev.1.42 in head is totally unrelated to the
commit message: it only adds .svn to the exclude list when copying
files using tar.

16 years agoMFC r180758:
kib [Fri, 22 Aug 2008 09:54:56 +0000 (09:54 +0000)]
MFC r180758:
Prevent the bdflush() activity when several buffers are locked at once
by setting the TDP_INBDFUSH for the problematic code blocks.

16 years agoMFC r181868:
kib [Fri, 22 Aug 2008 09:47:01 +0000 (09:47 +0000)]
MFC r181868:
In brelse, put the B_NEEDSGIANT buffer on the QUEUE_DIRTY_GIANT queue,
instead of QUEUE_DIRTY.

16 years agoMFC r181753:
philip [Fri, 22 Aug 2008 08:09:13 +0000 (08:09 +0000)]
MFC r181753:

  Introduce a new loader tunable "hw.ata.ata_dma_check_80pin", defaulting
  to 1.  This can be used to disable the 80pin cable check on systems which
  forget to set the bit -- such as certain laptops and Soekris boards.

16 years agoMFC r179894, r181224
thompsa [Fri, 22 Aug 2008 03:55:37 +0000 (03:55 +0000)]
MFC r179894, r181224

 Add support for the optional key in the GRE header.

 PR: kern/114714
 Submitted by: Cristian KLEIN

16 years agoMFC r178214
thompsa [Fri, 22 Aug 2008 03:15:50 +0000 (03:15 +0000)]
MFC r178214

 If the .inf file did not have a Default entry for the registry key then write
 out a blank value and close the brackets on the ndis_regvals array.

16 years agoMFC r174150, r174240, r179009, r179720
thompsa [Fri, 22 Aug 2008 03:08:31 +0000 (03:08 +0000)]
MFC r174150, r174240, r179009, r179720

Sync sys/compat/ndis to HEAD.

 - Correct the calculation for the number of 100ns intervals since January 1,
   1601. The 1601 - 1970 period was in seconds rather than 100ns units.
 - Remove duplication by having NdisGetCurrentSystemTime call ntoskrnl_time.
 - Implement functions required by some ndis drivers.
 - Fix a panic when it occurred during initializing the ndis driver because it
   try to read network address through ifnet structure which is NULL until the
   ndis driver's initialization is finished.
 - fix a page fault that it occurred during ifp is NULL.  This bug happens when
   NDIS driver's initialization is failed and NDIS driver's trying to call
   NdisWriteErrorLogEntry().

16 years agoMFC 181653:
kmacy [Fri, 22 Aug 2008 01:30:39 +0000 (01:30 +0000)]
MFC 181653:
Fix runt TSO packet issue.

Obtained from: Chelsio Inc.

16 years agoMFC 181652:
kmacy [Fri, 22 Aug 2008 01:28:57 +0000 (01:28 +0000)]
MFC 181652:

Add LRO and MAC statistics to exported sysctls.

Obtained from: Chelsio Inc.

16 years agoMFC 181616:
kmacy [Fri, 22 Aug 2008 01:23:39 +0000 (01:23 +0000)]
MFC 181616:
Remove cxgb private lro implementation and switch to using system implementation.

Obtained from: Chelsio Inc.

16 years agoMFC r181823:
kib [Thu, 21 Aug 2008 09:58:18 +0000 (09:58 +0000)]
MFC r181823:
Remove the check for mode and conditional swapgs in doreti_iret_fault,
we already have right gs base in the MSR.

16 years agoMFC r180754
weongyo [Thu, 21 Aug 2008 05:40:38 +0000 (05:40 +0000)]
MFC r180754

when NDIS framework try to query/set informations NDIS drivers can
return NDIS_STATUS_PENDING.  In this case, it's waiting for 5 secs to
get the response from drivers now.  However, some NDIS drivers can send
the response before NDIS framework gets ready to receive it so we might
always be blocked for 5 secs in current implementation.  NDIS framework
should reset the event before calling NDIS driver's callback not after.

16 years agoMFC r180730
weongyo [Thu, 21 Aug 2008 05:34:37 +0000 (05:34 +0000)]
MFC r180730

remove reference for unexisting ndisapi(9)

16 years agoMFC r179855
weongyo [Thu, 21 Aug 2008 05:30:26 +0000 (05:30 +0000)]
MFC r179855

handle .INF files for PCMCIA correctly that specify multiple entries in
their [Manufacturer] sections and prevent a case that NDIS_PCI_DEV_TABLE
definition was always emitted that it's only emitted once if a .INF file
is for PCI.

16 years agoMFC r179423
weongyo [Thu, 21 Aug 2008 05:13:55 +0000 (05:13 +0000)]
MFC r179423

Fix a panic that a priority value which is passed to cv_broadcastpri(9)
can be < 0.  We don't ignore a `increment' argument but at least we keep
a priority value of NDIS threads over PRI_MIN_KERN.

Reviewed by: thompsa

16 years agoMFC r181557:
yongari [Thu, 21 Aug 2008 03:45:50 +0000 (03:45 +0000)]
MFC r181557:
  Use device_set_desc() instead of device_set_desc_copy() as we don't
  manipulate the verbose description of a device.

16 years agoMFC r180954:
yongari [Thu, 21 Aug 2008 03:34:14 +0000 (03:34 +0000)]
MFC r180954:
  style(9) - space after keywords, don't indent case.

16 years agoMFC r180952, r181556:
yongari [Thu, 21 Aug 2008 03:25:44 +0000 (03:25 +0000)]
MFC r180952, r181556:
  Nuke referencing softc in device probe.

16 years agoMFC r180950:
yongari [Thu, 21 Aug 2008 02:03:01 +0000 (02:03 +0000)]
MFC r180950:
  s/printf/device_printf/g
  Don't hard code function name in device_printf() and use __func__.
  While I'm here nuke bfe_unit in softc as it's not needed anymore.

16 years agoBump _FreeBSD_version for physically contiguous jumbo frame support
kmacy [Wed, 20 Aug 2008 22:42:20 +0000 (22:42 +0000)]
Bump _FreeBSD_version for physically contiguous jumbo frame support

Requested by: Andrew Gallatin (gallatin@)

16 years agoMFC r179374
ghelmer [Wed, 20 Aug 2008 19:57:58 +0000 (19:57 +0000)]
MFC r179374

Similar to changes previously made to src/usr.bin/uniq/uniq.c,
fix truncation of lines at LINE_MAX characters by dynamically
extending line buffers.

16 years agoMFC: Attach the cpufreq child devices with specific orders to enforce
jhb [Wed, 20 Aug 2008 18:33:09 +0000 (18:33 +0000)]
MFC: Attach the cpufreq child devices with specific orders to enforce
relative priority of some of the drivers that manage the same state (e.g.
ichss0 vs est0).

16 years agoMFC: Decode "exotic" instructions such as pause as well as "cmov*" on i386.
jhb [Wed, 20 Aug 2008 17:44:02 +0000 (17:44 +0000)]
MFC: Decode "exotic" instructions such as pause as well as "cmov*" on i386.

16 years agoMFC r181468-181469,181594:
philip [Wed, 20 Aug 2008 14:16:07 +0000 (14:16 +0000)]
MFC r181468-181469,181594:

  Add the glxsb(4) manual page.

16 years agoMFC r181467, r181593:
philip [Wed, 20 Aug 2008 11:33:13 +0000 (11:33 +0000)]
MFC r181467, r181593:

  Add glxsb(4) driver for the Security Block in AMD Geode LX processors (as
  found in Soekris hardware, for instance).  The hardware supports
  acceleration of AES-128-CBC accessible through crypto(4) and supplies
  entropy to random(4).

16 years agoMFC r176265:
yongari [Wed, 20 Aug 2008 05:23:31 +0000 (05:23 +0000)]
MFC r176265:
  Prefer NULL over integer 0 for pointer type.

16 years agoMFC r176264:
yongari [Wed, 20 Aug 2008 05:19:41 +0000 (05:19 +0000)]
MFC r176264:
  Nuke local jumbo allocator and switch to use of UMA backed page
  allocator for jumbo frame.
   o Removed unneeded jlist lock which was used to manage jumbo
     buffers.
   o Don't reinitialize hardware if MTU was not changed.
   o Added additional check for minimal MTU size.
   o Added a new tunable hw.skc.jumbo_disable to disable jumbo frame
     support for the driver. The tunable could be set for systems that
     do not need to use jumbo frames and it would save
     (9K * number of Rx descriptors) bytes kernel memory.
   o Jumbo buffer allocation failure is no longer critical error for
     the operation of sk(4). If sk(4) encounter the allocation failure
     it just disables jumbo frame support and continues to work without
     user intervention.

  With these changes jumbo frame performance of sk(4) was slightly
  increased and users should not encounter jumbo buffer allocation
  failure. Previously sk(4) tried to allocate physically contiguous
  memory, 3388KB for 256 Rx descriptors. Sometimes that amount of
  contiguous memory region could not be available for running systems
  which in turn resulted in failure of loading the driver.

16 years agoMFC r176859:
yongari [Wed, 20 Aug 2008 05:12:30 +0000 (05:12 +0000)]
MFC r176859:
  Nuke local jumbo allocator and switch to use of UMA backed page
  allocator for jumbo frame. Also remove unneeded jlist lock which
  is no longer required to protect jumbo buffers.
  With these changes jumbo frame performance of nfe(4) was slightly
  increased and users should not encounter jumbo buffer allocation
  failure anymore.

16 years agoMFC r181655:
yongari [Wed, 20 Aug 2008 04:55:40 +0000 (04:55 +0000)]
MFC r181655:
  Fix VLAN hardware tag insertion/stripping on big-endian
  architectures.

  Reported by:  naddy
  Tested on:    sparc64

16 years agoMFC r181619:
yongari [Wed, 20 Aug 2008 04:46:46 +0000 (04:46 +0000)]
MFC r181619:
  Restore link state handling which was broken in rev 1.69.
  Also report current link state while auto-negotiation is in
  progress.
  With this change link loss should be reported within a second
  and drivers that rely on link state should work.

16 years agoMFC r181617-r181618:
yongari [Wed, 20 Aug 2008 04:19:28 +0000 (04:19 +0000)]
MFC r181617-r181618:
 r181617: Remove whitespace at the end of line.
 r181618: Remove 'cr' at the end of line.

16 years agoMFC teh ability to use tablearg with a skipto.
julian [Tue, 19 Aug 2008 19:58:42 +0000 (19:58 +0000)]
MFC teh ability to use tablearg with a skipto.
This was already supported in userland.

16 years agoUpdate locking table, based on changes in r173444:
emaste [Tue, 19 Aug 2008 19:31:58 +0000 (19:31 +0000)]
Update locking table, based on changes in r173444:
 - Have SPIN / want SPIN adds the "Recursion is defined per lock" footnote.
 - Have sx / want Slp & have sx / want rw changed from no to ok.

16 years agoMFC r181731:
kensmith [Tue, 19 Aug 2008 13:40:54 +0000 (13:40 +0000)]
MFC r181731:

> Catch up on removal of /usr/src/compat.

16 years agoMFC r179175:
kib [Tue, 19 Aug 2008 12:36:40 +0000 (12:36 +0000)]
MFC r179175:
Implement the per-open file data for the cdev.
The td_fpop member of the struct thread is appended to the end, and cleared
in the thread allocator to keep struct thread KBI-compatible on RELENG_7.

MFC r181635:
Remove unnecessary locking around pointer fetch.

16 years agoMFC: r179773, r181862
kevlo [Tue, 19 Aug 2008 01:51:37 +0000 (01:51 +0000)]
MFC: r179773, r181862

- Add the D-Link DWA-110
Tested by: Jonathan Lee <spamtrap at tczyhatczsche dot eu>

- Add the D-Link DWA-111

16 years agoMFC 174247:
kmacy [Tue, 19 Aug 2008 01:36:45 +0000 (01:36 +0000)]
MFC 174247:

Introduce an UMA backend page allocator for the jumbo frame zones that
allocates physically contiguous memory.

MFC after: 3 months
Requested and reviewed by: Kip Macy
Tested by: Andrew Gallatin and Pyun YongHyeon

Approved by: alc

16 years agoAvoid possible namespace collision between function scope sin and locally sub-express...
kmacy [Tue, 19 Aug 2008 01:25:08 +0000 (01:25 +0000)]
Avoid possible namespace collision between function scope sin and locally sub-expression scoped sin
by moving sin declaration to point of use.

16 years agoIn anticipation of MFCing physically contiguous jumbo frames MFC r177921:
kmacy [Tue, 19 Aug 2008 01:16:12 +0000 (01:16 +0000)]
In anticipation of MFCing physically contiguous jumbo frames MFC r177921:
 Reintroduce UMA_SLAB_KMAP; however, change its spelling to
 UMA_SLAB_KERNEL for consistency with its sibling UMA_SLAB_KMEM.
 (UMA_SLAB_KMAP met its original demise in revision 1.30 of
 vm/uma_core.c.)  UMA_SLAB_KERNEL is now required by the jumbo frame
 allocators.  Without it, UMA cannot correctly return pages from the
 jumbo frame zones to the VM system because it resets the pages' object
 field to NULL instead of the kernel object.  In more detail, the jumbo
 frame zones are created with the option UMA_ZONE_REFCNT.  This causes
 UMA to overwrite the pages' object field with the address of the slab.
 However, when UMA wants to release these pages, it doesn't know how to
 restore the object field, so it sets it to NULL.  This change teaches
 UMA how to reset the object field to the kernel object.

Crashes reported by: kris
Fix tested by: kris

Approved by: alc

16 years agoMFC: r181782, cvs 1.235 udp_usrreq.c, 1.102 ip6_input.c, 1.44 ip6_var.h
bz [Mon, 18 Aug 2008 23:00:41 +0000 (23:00 +0000)]
MFC: r181782, cvs 1.235 udp_usrreq.c, 1.102 ip6_input.c, 1.44 ip6_var.h

  Fix a regression introduced in r179289 splitting up ip6_savecontrol()
  into v4-only vs. v6-only inp_flags processing.
  When ip6_savecontrol_v4() is called from ip6_savecontrol() we
  were not passing back the **mp thus the information will be missing
  in userland.
  Istead of going with a *** as suggested in the PR we are returning
  **mp now and passing in the v4only flag as a pointer argument.

PR: kern/126349

16 years agoMFC: Allow comment to be placed at the end of a configuration line.
sobomax [Mon, 18 Aug 2008 20:35:02 +0000 (20:35 +0000)]
MFC: Allow comment to be placed at the end of a configuration line.

16 years agoMFC: Fix 6-year old cut&paste error. The # could be escaped with '\', not
sobomax [Mon, 18 Aug 2008 20:31:50 +0000 (20:31 +0000)]
MFC: Fix 6-year old cut&paste error. The # could be escaped with '\', not
with '\\'.

16 years agoMerge r178285, r178318, r178319, r178320, r178321, r178322, r178325,
rwatson [Mon, 18 Aug 2008 08:11:48 +0000 (08:11 +0000)]
Merge r178285, r178318, r178319, r178320, r178321, r178322, r178325,
r178376, r178377, r178378, r178419, r179412, r179414, r180127, r180338,
r180343, r180344, r180346, r180348, r180368, r180422, r180429, r180536,
r180558, r180589, r181364, r181365 from head to stable/7:

Introduce and use rwlocks throughout the inpcbinfo and inpcb infrastructure,
and protocols that depend on that infrastructure, including UDP, TCP, and
IP raw sockets.  Significant parts of this work were reviewed by Bjoern Zeeb,
and tested by Paul Saab, Kris Keneway, and George Neville-Neil, whose
contributions to this work are greatly appreciated.

Tested by:      ps, kris, gnn, Mike Tancsa <mike at sentex dot net>
Reviewed by:    bz, des

16 years agoMFC r179467
sam [Mon, 18 Aug 2008 01:20:39 +0000 (01:20 +0000)]
MFC r179467

5416 and similar chips grew another region in the pci clock domain
where register accesses do not pass through the byte-lane hardware;
extend the register op macros to deal with this

16 years agoMFC r179640: correct code returned in AsssocResp
sam [Mon, 18 Aug 2008 01:16:04 +0000 (01:16 +0000)]
MFC r179640: correct code returned in AsssocResp

16 years agoMFC r180837: don't deauth a station because it sends a ps-poll w/ a bogus aid
sam [Mon, 18 Aug 2008 01:05:46 +0000 (01:05 +0000)]
MFC r180837: don't deauth a station because it sends a ps-poll w/ a bogus aid

16 years agoMFC r181083
sam [Mon, 18 Aug 2008 00:59:07 +0000 (00:59 +0000)]
MFC r181083

unset TARGET_CPUTYPE and TARGET_BIG_ENDIAN when building the kernel
to silence compiler complaints when cross-building

16 years agoMFC r181454: recognize WPS ie's
sam [Mon, 18 Aug 2008 00:51:19 +0000 (00:51 +0000)]
MFC r181454: recognize WPS ie's

16 years agoMFC r181347: add WPS oui
sam [Mon, 18 Aug 2008 00:16:57 +0000 (00:16 +0000)]
MFC r181347: add WPS oui

16 years agoMFC support to return all received information elements in scan results
sam [Sun, 17 Aug 2008 23:00:47 +0000 (23:00 +0000)]
MFC support to return all received information elements in scan results
and station information

16 years agoSomewhat belatedly, update for SA-08:06.bind and EN-08:02.tcp.
bmah [Sun, 17 Aug 2008 12:58:32 +0000 (12:58 +0000)]
Somewhat belatedly, update for SA-08:06.bind and EN-08:02.tcp.

16 years agoMFC allocator improvements and fixes:
jasone [Sat, 16 Aug 2008 20:14:21 +0000 (20:14 +0000)]
MFC allocator improvements and fixes:

* Enhance the chunk map to support run coalescing, and substantially reduce
  the number of red-black tree operations performed.

* Remove unused code.

* Fix arena_run_reg_dalloc() to use the entire precomputed division table.

* Improve lock preemption performance for hyperthreaded CPUs.

16 years agoMFC r173217,176789,176792-176793,176839,176842
yar [Sat, 16 Aug 2008 19:17:36 +0000 (19:17 +0000)]
MFC r173217,176789,176792-176793,176839,176842

Don't forget to set MAKEFLAGS in the childs' environment
from the .MAKEFLAGS global variable even if it's empty or
unset.  This means setting MAKEFLAGS to just an empty string
in the latter case.

The non-POSIX environment variable MAKE was superseded
by MAKEFLAGS ages ago, so don't mention it in comments.

Update the manpage accordingly and make the whole description of
MAKEFLAGS and .MAKEFLAGS more consistent as this change indeed
brings more consistency into the reliable behaviour of make(1).

Give more details in the manpage on the following topics:
o How global, command-line, and environment variables relate.
o What peculiarities the -f option has WRT ${MAKEFLAGS}.

(.Dd in the manpage is already beyond the date of those changes
in HEAD.)

16 years agoMFC: Update man page for addition of BIOCSETFNR.
dwmalone [Sat, 16 Aug 2008 11:48:44 +0000 (11:48 +0000)]
MFC: Update man page for addition of BIOCSETFNR.

16 years agoMFC:
dwmalone [Sat, 16 Aug 2008 11:48:10 +0000 (11:48 +0000)]
MFC:
Add a new ioctl for changing the read filter (BIOCSETFNR). This is
just like BIOCSETF but it doesn't drop all the packets buffered on
the discriptor and reset the statistics.

Also, when setting the write filter, don't drop packets waiting to
be read or reset the statistics.

16 years agoMFC: Use an explicit path to pw(8) so /usr/sbin does not have to be in
jhb [Fri, 15 Aug 2008 20:20:23 +0000 (20:20 +0000)]
MFC: Use an explicit path to pw(8) so /usr/sbin does not have to be in
the user's path.

16 years agoMFC: If the kernel fails to allocate resources for the initial value of a
jhb [Fri, 15 Aug 2008 20:16:21 +0000 (20:16 +0000)]
MFC: If the kernel fails to allocate resources for the initial value of a
PCI BAR, then zero the BAR and clear the resource list entry so that a
new resource will be allocated by the device driver.

16 years agoMFC: Fix some of the hangs reported with the real mode BTX by disabling
jhb [Fri, 15 Aug 2008 19:51:23 +0000 (19:51 +0000)]
MFC: Fix some of the hangs reported with the real mode BTX by disabling
interrupts for software interrupts.

16 years agoMFC: Initialize the vm86 structure to a known-good state, specifically
jhb [Fri, 15 Aug 2008 19:31:12 +0000 (19:31 +0000)]
MFC: Initialize the vm86 structure to a known-good state, specifically
eflags.

16 years agoMFC: Close two different races with concurrent opens of pty master devices.
jhb [Fri, 15 Aug 2008 19:25:40 +0000 (19:25 +0000)]
MFC: Close two different races with concurrent opens of pty master devices.

16 years agoMFC: Close a race with concurrent open's of a pts master device which
jhb [Fri, 15 Aug 2008 19:22:33 +0000 (19:22 +0000)]
MFC: Close a race with concurrent open's of a pts master device which
could result in leaked tty structures.

16 years agoMFC r180495 to stable/7:
antoine [Fri, 15 Aug 2008 16:42:07 +0000 (16:42 +0000)]
MFC r180495 to stable/7:
  Staticize M_STACK.

  Approved by:    rwatson (mentor)
  MFC after:      1 month

16 years agoMFC: r177373
marius [Fri, 15 Aug 2008 11:37:29 +0000 (11:37 +0000)]
MFC: r177373

Oops. Use atomic_add_long() for atomic_fetchadd_long() (not atomic_add_int())
for sparc64 and sun4v.

16 years agoMFC r177535 by peter:
kib [Fri, 15 Aug 2008 09:08:35 +0000 (09:08 +0000)]
MFC r177535 by peter:
(Micro)optimization of cpu_switch for amd64.
Because r173004 (pure kernel threads) is not MFCed to stable/7, I use
P_KTHREAD flag from curthread->td_proc->p_flag instead of TDP_KTHREAD
flag from curthread->td_pflags to test whether the fs/gs bases need to
be restored [1].

No objections for MFC from: peter
Suggested by: jhb [1]

MFC r180992:
Bring back the save/restore of the %ds, %es, %fs and %gs registers for
the 32bit images on amd64.

The end result is almost complete syncronization of the
amd64/amd64/cpu_switch.S between head and stable/7, besides TDP_KTHREAD
difference.

16 years agoMFC of r180208, r181413, r181418, r181421, r181424
edwin [Fri, 15 Aug 2008 01:12:28 +0000 (01:12 +0000)]
MFC of r180208, r181413, r181418, r181421, r181424

  Set magic fbsd:nokeywords property

  Remove Theory, which isn't part of the zoneinfo module but came out
  of /head/usr.sbin/zic (and isn't installed from there neither).

  MFC of tzdata2008c

  - Africa/Morocco will have DST in 2008.
  - Asia/Choibalsan should be GMT+08:00.
  - Asia/Pakistan will have DST in 2008.

  MFC of tzdata2008d

  - Mauritius will have a DST experiment between 2008-11-01 and 2009-03-31.
  - Add/Fix historical data for C-Eur, the SovietZone, Germany,
    Bahamas, San Luis.
  - Add information about West Para (America/Santarem)
  - America/Eirunepe and America/Rio_Branco go to UTC-4

  MFC on tzdata2008e:

  - Fix dates in 2008/2009 for Africa/Mauritius.
  - Leap second notification for the end of 2008.

16 years agoMFC: Add support for cpusets to the 4BSD scheduler.
jhb [Thu, 14 Aug 2008 22:04:52 +0000 (22:04 +0000)]
MFC: Add support for cpusets to the 4BSD scheduler.

Committed from: my iPhone

16 years agoMFC: Various and sundry whitespace fixes.
jhb [Thu, 14 Aug 2008 19:27:47 +0000 (19:27 +0000)]
MFC: Various and sundry whitespace fixes.