]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
15 years agoAfter r193232 rt_tables in vnet.h are no longer indirectly dependent on
bz [Mon, 8 Jun 2009 19:57:35 +0000 (19:57 +0000)]
After r193232 rt_tables in vnet.h are no longer indirectly dependent on
the ROUTETABLES kernel option thus there is no need to include opt_route.h
anymore in all consumers of vnet.h and no longer depend on it for module
builds.

Remove the hidden include in flowtable.h as well and leave the two
explicit #includes in ip_input.c and ip_output.c.

15 years agoCommit missed driver_info to driver_ivar change in usb_attach_args.
thompsa [Mon, 8 Jun 2009 19:56:23 +0000 (19:56 +0000)]
Commit missed driver_info to driver_ivar change in usb_attach_args.

Pointed out by: kib

15 years agoFix nfscl_getcl() so that it doesn't crash when it is called to
rmacklem [Mon, 8 Jun 2009 18:41:23 +0000 (18:41 +0000)]
Fix nfscl_getcl() so that it doesn't crash when it is called to
do an NFSv4 Close operation with the cred argument NULL. Also,
clarify what NULL arguments mean in the function's comment.

Approved by: kib (mentor)

15 years agoRevert my change; reintroduce __gnu89_inline.
ed [Mon, 8 Jun 2009 18:23:43 +0000 (18:23 +0000)]
Revert my change; reintroduce __gnu89_inline.

It turns out our compiler in stable/7 can't build this code anymore.
Even though my opinion is that those people should just run `make
kernel-toolchain' before building a kernel, I am willing to wait and
commit this after we've branched stable/8.

Requested by: rwatson

15 years agoChange driver_info to a ulong as it always stores a number and remove the only
thompsa [Mon, 8 Jun 2009 18:09:51 +0000 (18:09 +0000)]
Change driver_info to a ulong as it always stores a number and remove the only
diff of the usb_device_id struct to Linux.

Reviewed by: HPS

15 years agoRemove __gnu89_inline.
ed [Mon, 8 Jun 2009 17:27:25 +0000 (17:27 +0000)]
Remove __gnu89_inline.

Now that we use C99 almost everywhere, just use C99-style in the pmap
code. Since the pmap code is the only consumer of __gnu89_inline, remove
it from cdefs.h as well. Because the flag was only introduced 17 months
ago, I don't expect any problems.

Reviewed by: alc

15 years agoIntroduce an infrastructure for dismantling vnet instances.
zec [Mon, 8 Jun 2009 17:15:40 +0000 (17:15 +0000)]
Introduce an infrastructure for dismantling vnet instances.

Vnet modules and protocol domains may now register destructor
functions to clean up and release per-module state.  The destructor
mechanisms can be triggered by invoking "vimage -d", or a future
equivalent command which will be provided via the new jail framework.

While this patch introduces numerous placeholder destructor functions,
many of those are currently incomplete, thus leaking memory or (even
worse) failing to stop all running timers.  Many of such issues are
already known and will be incrementaly fixed over the next weeks in
smaller incremental commits.

Apart from introducing new fields in structs ifnet, domain, protosw
and vnet_net, which requires the kernel and modules to be rebuilt, this
change should have no impact on nooptions VIMAGE builds, since vnet
destructors can only be called in VIMAGE kernels.  Moreover,
destructor functions should be in general compiled in only in
options VIMAGE builds, except for kernel modules which can be safely
kldunloaded at run time.

Bump __FreeBSD_version to 800097.
Reviewed by: bz, julian
Approved by: rwatson, kib (re), julian (mentor)

15 years agoClean up FreeBSD-specific header file:
jkim [Mon, 8 Jun 2009 17:11:47 +0000 (17:11 +0000)]
Clean up FreeBSD-specific header file:

- Define ACPI_USE_DO_WHILE_0 to add do while(0) around return_*() macros.
- Define __cdecl only for user land.  It will never be used in kernel.
- Remove __cli() and __sti().  They were deprecated long ago.

15 years agoNow that amd64's kernel map is 512GB (SVN rev 192216), there is no reason
alc [Mon, 8 Jun 2009 16:43:40 +0000 (16:43 +0000)]
Now that amd64's kernel map is 512GB (SVN rev 192216), there is no reason
to cap its buffer map at 1GB.

MFC after: 6 weeks

15 years agoChange the printf format string to match the variable type to quiet
jhb [Mon, 8 Jun 2009 15:54:07 +0000 (15:54 +0000)]
Change the printf format string to match the variable type to quiet
warnings.

15 years agoMove zombie-reaping code out of kern_wait() and into its own function,
rwatson [Mon, 8 Jun 2009 15:26:09 +0000 (15:26 +0000)]
Move zombie-reaping code out of kern_wait() and into its own function,
proc_reap().
Reviewed by: jhb
MFC after: 3 days
Sponsored by: Google, Inc.

15 years agoCode using COMPAT_ROUTE_FLAGS option, introduced with r187094,
bz [Mon, 8 Jun 2009 15:13:20 +0000 (15:13 +0000)]
Code using COMPAT_ROUTE_FLAGS option, introduced with r187094,
was changed again in r187328, removing any use of the option from
the kernel.  The option was never in NOTES.  Garbage collect.

15 years agoInstead of packing the individual fields in the PnP structures, pack the
jhb [Mon, 8 Jun 2009 15:09:22 +0000 (15:09 +0000)]
Instead of packing the individual fields in the PnP structures, pack the
entire structures.  This trims some warnings.

Verified by: md5(1)
MFC after: 1 week

15 years agoDon't attempt to free the GPT partition list for a disk with an empty GPT.
jhb [Mon, 8 Jun 2009 15:07:35 +0000 (15:07 +0000)]
Don't attempt to free the GPT partition list for a disk with an empty GPT.

Submitted by: Yuri Pankov  yuri.pankov of gmail
MFC after: 3 days

15 years agoUpdate vimage size guard values but do not yet re-enable them.
bz [Mon, 8 Jun 2009 15:02:30 +0000 (15:02 +0000)]
Update vimage size guard values but do not yet re-enable them.
Now that rt_tables does no longer change depending on kernel option
we do no longer need any special values for LINT.

The updated values will be used to make sure that an upcoming
opt_route.h cleanup will not again show any side effects.

15 years agoAdd another PCI id for Nvidia nForce MCP67, found in several Acer laptops.
ariff [Mon, 8 Jun 2009 14:37:47 +0000 (14:37 +0000)]
Add another PCI id for Nvidia nForce MCP67, found in several Acer laptops.

15 years agoPermit the specification of bandwidth values within
luigi [Mon, 8 Jun 2009 14:32:29 +0000 (14:32 +0000)]
Permit the specification of bandwidth values within
"profile" files (bandwidth is mandatory when using a
profile, so it makes sense to have everything in one place).

Update the manpage accordingly.

Submitted by: Marta Carbone

15 years agoDo not dereference vp->v_rdev without holding any of dev_mtx or vnode
kib [Mon, 8 Jun 2009 13:34:45 +0000 (13:34 +0000)]
Do not dereference vp->v_rdev without holding any of dev_mtx or vnode
lock. Use code similar to devfs_fp_check(), but inlined to feet other
checks performed by ttyhook_register().

Reviewed by: ed

15 years agoInvalidate cache in pmap_remove_all() on ARM.
raj [Mon, 8 Jun 2009 12:15:39 +0000 (12:15 +0000)]
Invalidate cache in pmap_remove_all() on ARM.

When pages are removed from virtual address space by calling pmap_remove_all()
CPU caches were not invalidated, which led to read corruption when another
page got mapped at this same virtual address at later time (the CPU was
retrieving stale contents).

Submitted by: Piotr Ziecik
Obtained from: Semihalf

15 years agoadd a missing format in a printf
luigi [Mon, 8 Jun 2009 10:53:18 +0000 (10:53 +0000)]
add a missing format in a printf
Detected building with gcc 4.3.3

MFC after: 3 days

15 years agoFix powerpc build failure due to strict kobj signatures checking.
ariff [Mon, 8 Jun 2009 08:10:52 +0000 (08:10 +0000)]
Fix powerpc build failure due to strict kobj signatures checking.

15 years ago[patch] [locale] German locales use old %d.%m.%y date format instead of newer ISO...
edwin [Mon, 8 Jun 2009 05:33:08 +0000 (05:33 +0000)]
[patch] [locale] German locales use old %d.%m.%y date format instead of newer ISO date

From the submitter:

    DIN 5008 (German norm for text processing) defines the old date
    format (%d.%m.%Y) to be obsolete and to be used only, if unambigous.
    In international communications the new format (%Y-%m-%d) is now
    required and FreeBSD should respect this.

    References:
    - DIN 5008
    - EN 28 601
    - ISO 8601

    Thanks to Oliver Lietz for bringing this to my attention.

PR: conf/72076
Submitted by: Peter Wullinger <some-mail-drop@gmx.net>
MFC after: 1 week

15 years agoAdd notes on various SND_* options.
ariff [Mon, 8 Jun 2009 04:39:47 +0000 (04:39 +0000)]
Add notes on various SND_* options.

15 years agoDo not attempt to set source address on outgoing messages
kan [Mon, 8 Jun 2009 03:39:15 +0000 (03:39 +0000)]
Do not attempt to set source address on outgoing messages
on UDP socket if we do not have a valid IP address.

15 years agoFix grammar.
jkoshy [Mon, 8 Jun 2009 03:37:25 +0000 (03:37 +0000)]
Fix grammar.

Submitted by: richardtoohey at paradise dot net dot nz on -doc

15 years agodo not strip M_MORE_DATA on packets coming through ieee80211_start;
sam [Mon, 8 Jun 2009 03:18:45 +0000 (03:18 +0000)]
do not strip M_MORE_DATA on packets coming through ieee80211_start;
frames coming out of the ps q may have this set and removing it
causes the 802.11 header to not indicate more frames follow which can
result in the sta going to sleep and missing them

15 years agoThe change r192913 has added dependency on IP_RECVDSTADDR being
kan [Mon, 8 Jun 2009 03:15:27 +0000 (03:15 +0000)]
The change r192913 has added dependency on IP_RECVDSTADDR being
set for RPC UDP sockets. Mountd uses internal libc fuctions
directly and bypasses generic socket initialization completely,
so we need to set IP_RECVDSTADDR here to match the libc behavior.

15 years agoMake the size (-s) and start (-b) parameters of the add verb optional.
marcel [Mon, 8 Jun 2009 02:13:24 +0000 (02:13 +0000)]
Make the size (-s) and start (-b) parameters of the add verb optional.
The missing parameter(s) are automatically filled-in.

15 years agoFix build on sparc64.
ariff [Sun, 7 Jun 2009 23:38:16 +0000 (23:38 +0000)]
Fix build on sparc64.

Pointy hat: ariff@

15 years agoprefer callout_schedule
sam [Sun, 7 Jun 2009 23:16:59 +0000 (23:16 +0000)]
prefer callout_schedule

15 years agouse c99-style initializers
sam [Sun, 7 Jun 2009 23:16:10 +0000 (23:16 +0000)]
use c99-style initializers

15 years agoFix and add a workaround on an issue of EtherIP packet with reversed
hrs [Sun, 7 Jun 2009 23:00:40 +0000 (23:00 +0000)]
Fix and add a workaround on an issue of EtherIP packet with reversed
version field sent via gif(4)+if_bridge(4).  The EtherIP
implementation found on FreeBSD 6.1, 6.2, 6.3, 7.0, 7.1, and 7.2 had
an interoperability issue because it sent the incorrect EtherIP
packets and discarded the correct ones.

This change introduces the following two flags to gif(4):

 accept_rev_ethip_ver: accepts both correct EtherIP packets and ones
    with reversed version field, if enabled.  If disabled, the gif
    accepts the correct packets only.  This flag is enabled by
    default.

 send_rev_ethip_ver: sends EtherIP packets with reversed version field
    intentionally, if enabled.  If disabled, the gif sends the correct
    packets only.  This flag is disabled by default.

These flags are stored in struct gif_softc and can be set by
ifconfig(8) on per-interface basis.

Note that this is an incompatible change of EtherIP with the older
FreeBSD releases.  If you need to interoperate older FreeBSD boxes and
new versions after this commit, setting "send_rev_ethip_ver" is
needed.

Reviewed by: thompsa and rwatson
Spotted by: Shunsuke SHINOMIYA
PR: kern/125003
MFC after: 2 weeks

15 years agoDecouple the i386 native and i386 Xen APIC definitions a little further.
adrian [Sun, 7 Jun 2009 22:52:48 +0000 (22:52 +0000)]
Decouple the i386 native and i386 Xen APIC definitions a little further.

I'm experimenting locally with xen APIC emulation a bit and this
makes it easier to migrate APIC entries between being bitmapped and
not being bitmapped.

15 years agoteach ieee80211_classify about ipv6 packets
sam [Sun, 7 Jun 2009 22:06:15 +0000 (22:06 +0000)]
teach ieee80211_classify about ipv6 packets

Reviewed by: bz, rrs

15 years agoadd iv_flags_ht and ic_flags_ht
sam [Sun, 7 Jun 2009 22:03:25 +0000 (22:03 +0000)]
add iv_flags_ht and ic_flags_ht

15 years agotrack HT flags move to iv_flags_ht
sam [Sun, 7 Jun 2009 22:03:07 +0000 (22:03 +0000)]
track HT flags move to iv_flags_ht

15 years agoiv_flags_ext is full, make room by moving HT-related flags to a new
sam [Sun, 7 Jun 2009 22:00:22 +0000 (22:00 +0000)]
iv_flags_ext is full, make room by moving HT-related flags to a new
iv_flags_ht word

15 years agoTry again to add beginnings of netisr(8) man page: this time add
rwatson [Sun, 7 Jun 2009 21:32:01 +0000 (21:32 +0000)]
Try again to add beginnings of netisr(8) man page: this time add
netisr.9.

15 years agoAdd beginnings of a netisr(9) man page.
rwatson [Sun, 7 Jun 2009 21:31:06 +0000 (21:31 +0000)]
Add beginnings of a netisr(9) man page.

15 years agoAdd a temporary workaround for panics being seen on NFS servers with ZFS,
rwatson [Sun, 7 Jun 2009 20:51:31 +0000 (20:51 +0000)]
Add a temporary workaround for panics being seen on NFS servers with ZFS,
where an improperly initialized prison field could lead to a panic.  This
is not the correct solution, since it fails to address similar problems
for both AUDIT and MAC, which also rely on properly initialized
credentials, but should reduce panic reports while we work that out.

Reported by: ps, kan, others

15 years agoAdd a check to xprt_unregister() to catch the case where another
rmacklem [Sun, 7 Jun 2009 20:38:41 +0000 (20:38 +0000)]
Add a check to xprt_unregister() to catch the case where another
thread has already unregistered the structure. Also add a KASSERT()
to xprt_unregister_locked() to check that the structure hasn't already
been unregistered.

Reviewed by: jhb
Tested by: pho
Approved by: kib (mentor)

15 years agoAllow humanized numbers for LBAs, as well as partition indices for
marcel [Sun, 7 Jun 2009 20:12:14 +0000 (20:12 +0000)]
Allow humanized numbers for LBAs, as well as partition indices for
gpart(8). LBAs in particular are ugly. The ganularity is a sector,
but users expect byte granularity when specifying the size or offset
with a SI unit. Handle LBAs specially to deal with this.

15 years agoRename usb pipes to endpoints as it better represents what they are, and struct
thompsa [Sun, 7 Jun 2009 19:41:11 +0000 (19:41 +0000)]
Rename usb pipes to endpoints as it better represents what they are, and struct
usb_pipe may be used for a different purpose later on.

15 years agoEliminate unnecessary obfuscation when testing a page's valid bits.
alc [Sun, 7 Jun 2009 19:38:26 +0000 (19:38 +0000)]
Eliminate unnecessary obfuscation when testing a page's valid bits.

15 years agoBump driver revision (should have bumped it earlier).
ariff [Sun, 7 Jun 2009 19:36:25 +0000 (19:36 +0000)]
Bump driver revision (should have bumped it earlier).

15 years agoSound Mega-commit. Expect further cleanup until code freeze.
ariff [Sun, 7 Jun 2009 19:12:08 +0000 (19:12 +0000)]
Sound Mega-commit. Expect further cleanup until code freeze.

For a slightly thorough explaination, please refer to
[1] http://people.freebsd.org/~ariff/SOUND_4.TXT.html .

Summary of changes includes:

1 Volume Per-Channel (vpc).  Provides private / standalone volume control
  unique per-stream pcm channel without touching master volume / pcm.
  Applications can directly use SNDCTL_DSP_[GET|SET][PLAY|REC]VOL, or for
  backwards compatibility, SOUND_MIXER_PCM through the opened dsp device
  instead of /dev/mixer.  Special "bypass" mode is enabled through
  /dev/mixer which will automatically detect if the adjustment is made
  through /dev/mixer and forward its request to this private volume
  controller.  Changes to this volume object will not interfere with
  other channels.

  Requirements:
    - SNDCTL_DSP_[GET|SET][PLAY|REC]_VOL are newer ioctls (OSSv4) which
      require specific application modifications (preferred).
    - No modifications required for using bypass mode, so applications
      like mplayer or xmms should work out of the box.

  Kernel hints:
    - hint.pcm.%d.vpc (0 = disable vpc).

  Kernel sysctls:
    - hw.snd.vpc_mixer_bypass (default: 1).  Enable or disable /dev/mixer
      bypass mode.
    - hw.snd.vpc_autoreset (default: 1).  By default, closing/opening
      /dev/dsp will reset the volume back to 0 db gain/attenuation.
      Setting this to 0 will preserve its settings across device
      closing/opening.
    - hw.snd.vpc_reset (default: 0).  Panic/reset button to reset all
      volume settings back to 0 db.
    - hw.snd.vpc_0db (default: 45).  0 db relative to linear mixer value.

2 High quality fixed-point Bandlimited SINC sampling rate converter,
  based on Julius O'Smith's Digital Audio Resampling -
  http://ccrma.stanford.edu/~jos/resample/.  It includes a filter design
  script written in awk (the clumsiest joke I've ever written)
    - 100% 32bit fixed-point, 64bit accumulator.
    - Possibly among the fastest (if not fastest) of its kind.
    - Resampling quality is tunable, either runtime or during kernel
      compilation (FEEDER_RATE_PRESETS).
    - Quality can be further customized during kernel compilation by
      defining FEEDER_RATE_PRESETS in /etc/make.conf.

  Kernel sysctls:
    - hw.snd.feeder_rate_quality.
      0 - Zero-order Hold (ZOH).  Fastest, bad quality.
      1 - Linear Interpolation (LINEAR).  Slightly slower than ZOH,
          better quality but still does not eliminate aliasing.
      2 - (and above) - Sinc Interpolation(SINC).  Best quality.  SINC
          quality always start from 2 and above.

  Rough quality comparisons:
    - http://people.freebsd.org/~ariff/z_comparison/

3 Bit-perfect mode.  Bypasses all feeder/dsp effects.  Pure sound will be
  directly fed into the hardware.

4 Parametric (compile time) Software Equalizer (Bass/Treble mixer). Can
  be customized by defining FEEDER_EQ_PRESETS in /etc/make.conf.

5 Transparent/Adaptive Virtual Channel. Now you don't have to disable
  vchans in order to make digital format pass through.  It also makes
  vchans more dynamic by choosing a better format/rate among all the
  concurrent streams, which means that dev.pcm.X.play.vchanformat/rate
  becomes sort of optional.

6 Exclusive Stream, with special open() mode O_EXCL.  This will "mute"
  other concurrent vchan streams and only allow a single channel with
  O_EXCL set to keep producing sound.

Other Changes:
    * most feeder_* stuffs are compilable in userland. Let's not
      speculate whether we should go all out for it (save that for
      FreeBSD 16.0-RELEASE).
    * kobj signature fixups, thanks to Andriy Gapon <avg@freebsd.org>
    * pull out channel mixing logic out of vchan.c and create its own
      feeder_mixer for world justice.
    * various refactoring here and there, for good or bad.
    * activation of few more OSSv4 ioctls() (see [1] above).
    * opt_snd.h for possible compile time configuration:
      (mostly for debugging purposes, don't try these at home)
        SND_DEBUG
        SND_DIAGNOSTIC
        SND_FEEDER_MULTIFORMAT
        SND_FEEDER_FULL_MULTIFORMAT
        SND_FEEDER_RATE_HP
        SND_PCM_64
        SND_OLDSTEREO

Manual page updates are on the way.

Tested by: joel, Olivier SMEDTS <olivier at gid0 d org>, too many
           unsung / unnamed heroes.

15 years agoEliminate an unused variable from allocbuf().
alc [Sun, 7 Jun 2009 18:19:04 +0000 (18:19 +0000)]
Eliminate an unused variable from allocbuf().

Eliminate the unnecessary setting of page valid bits from a non-VMIO buffer
in vm_hold_load_pages().

15 years agoMention the range for the exit status for the exit special builtin.
jilles [Sun, 7 Jun 2009 15:04:43 +0000 (15:04 +0000)]
Mention the range for the exit status for the exit special builtin.

The exit status may exceed 255 in some cases (return); even though it seems
unwise to rely on this, it is also unwise to assume that $? is always
between 0 and 255.

This resolves bin/124748 by documenting that 'exit -1' is not valid.

PR: bin/124748
Approved by: ed (mentor)

15 years agoWelcome to a default installed /etc/ntp.conf
edwin [Sun, 7 Jun 2009 13:26:57 +0000 (13:26 +0000)]
Welcome to a default installed /etc/ntp.conf

This NTP configuration file points to the [012].pool.ntp.org servers,
which will return a list of geographical local NTP servers.
It uses the best-practice options of "iburst" and "maxpoll 9".
It gives examples on how to use the "restrict" commands, which are
unfortunately not working when you use the pool.ntp.org servers.
It sets up a fudge server so any clients syncing against this server
will always be synced even if we lose the master.

The idea of this file was briefly discussed on -net.

PR: conf/58595
Submitted by: Chris Stenton <jacs@gnome.co.uk>
MFC after: 1 week

15 years agoMake "human-readable" (-H/-h) output also "humanize" inode counts.
simon [Sun, 7 Jun 2009 09:06:21 +0000 (09:06 +0000)]
Make "human-readable" (-H/-h) output also "humanize" inode counts.
Base 10 is always used for the inode counts as I could not think of any
reason base 2 inode counts would be useful.

Minor mdoc markup fix to df(1) while here anyway.

MFC after: 3 weeks

15 years ago- Outindent long printf lines instead of splitting them in the
stas [Sun, 7 Jun 2009 08:42:26 +0000 (08:42 +0000)]
- Outindent long printf lines instead of splitting them in the
  middle of senetences. This also makes the code more consistent
  with the corresponding FFS code.
- Use 2-space sentences breaks consistently.

Suggested by: bde

15 years agoI don't know why fstab doesn't document that the fifth field specifies
imp [Sun, 7 Jun 2009 07:45:25 +0000 (07:45 +0000)]
I don't know why fstab doesn't document that the fifth field specifies
the number of days between backups.  All it says is frequency, with no
units given.  It likely should say "the interval in days between backups"
instead, but not today.

15 years agoFix compilation when compiled w/out WITNESS.
adrian [Sun, 7 Jun 2009 05:52:22 +0000 (05:52 +0000)]
Fix compilation when compiled w/out WITNESS.

Submitted by: Edwin Shao <poleris@gmail.com>

15 years agoRemove cruft at 0E11-0002. Informed upstream.
edwin [Sun, 7 Jun 2009 04:45:50 +0000 (04:45 +0000)]
Remove cruft at 0E11-0002. Informed upstream.

Submitted by: Pawel Worach <pawel.worach@gmail.com>
MFC after: 6 days.

15 years agoFix a lockorder reversal I introduced in r193436 when I moved the
rmacklem [Sun, 7 Jun 2009 01:06:56 +0000 (01:06 +0000)]
Fix a lockorder reversal I introduced in r193436 when I moved the
mtx_destroy() of the pool mutex to after SVC_RELEASE(), because
the pool mutex was still locked when soclose() was called by svc_dg_destroy().
To fix this, an mtx_unlock() was added where mtx_destroy() was before
r193436.

Reviewed by: jhb
Tested by: pho
Approved by: rwatson (mentor)

15 years agoEliminate an unneeded forward declaration. (This should have been removed
alc [Sat, 6 Jun 2009 21:23:29 +0000 (21:23 +0000)]
Eliminate an unneeded forward declaration.  (This should have been removed
in revision 1.42.)

15 years agoIf vm_pager_get_pages() returns VM_PAGER_OK, then there is no need to check
alc [Sat, 6 Jun 2009 20:13:14 +0000 (20:13 +0000)]
If vm_pager_get_pages() returns VM_PAGER_OK, then there is no need to check
the page's valid bits.  The page is guaranteed to be fully valid.  (For the
record, this is documented in vm/vm_pager.h's comments.)

15 years agoCheck that flopen() can lock against self and that children inherit the lock.
des [Sat, 6 Jun 2009 19:07:15 +0000 (19:07 +0000)]
Check that flopen() can lock against self and that children inherit the lock.

15 years agoRevert (once again, and hopefully for the last time) to flock(2) locks.
des [Sat, 6 Jun 2009 18:47:03 +0000 (18:47 +0000)]
Revert (once again, and hopefully for the last time) to flock(2) locks.
The problem with fcntl(2) locks is that they are not inherited by child
processes.  This breaks pidfile(3), where the common idiom is to open
and lock the PID file before daemonizing.

15 years agoAdd missing .Pp
des [Sat, 6 Jun 2009 17:34:49 +0000 (17:34 +0000)]
Add missing .Pp

15 years agoRemove opt_mac.h generation for various kernel modules that no longer
rwatson [Sat, 6 Jun 2009 17:01:44 +0000 (17:01 +0000)]
Remove opt_mac.h generation for various kernel modules that no longer
require it.

Submitted by: pjd

15 years agoRegen:
edwin [Sat, 6 Jun 2009 13:37:53 +0000 (13:37 +0000)]
Regen:

* Hart: rev 671 of pcidevs.txt; 22-01-2008 (D-M-Y).
* Boemler: vendors.txt (2008-06-06)

Initiated by conf/121219 by Yousif Hassan <yousif@alumni.jmu.edu>

MFC after: 1 week

15 years agopciconf(8) removes characters after an #, therefor remove them.
edwin [Sat, 6 Jun 2009 13:33:11 +0000 (13:33 +0000)]
pciconf(8) removes characters after an #, therefor remove them.
Also remove \s used to escape 's.

15 years agoWhen trying to see if we could update the share/misc/pci_vendors,
edwin [Sat, 6 Jun 2009 12:53:54 +0000 (12:53 +0000)]
When trying to see if we could update the share/misc/pci_vendors,
I found out that the input format of the Boemler list was different
than what the code expected: The last two fields were interpreted
as one. Checking the csv version of the list it showed that there
was sometimes a chipset number in the column before the card
description.

This is a rewrite to use the CSV format of the Boemler list. The
output is differently formatted: Instead of the "chip description",
it is now "description (chip)"

15 years agoAdd an interim userland utility for managing vimages / virtualized
zec [Sat, 6 Jun 2009 12:50:10 +0000 (12:50 +0000)]
Add an interim userland utility for managing vimages / virtualized
network stack infrastructure.

Requested by: julian (mentor)
Approved by: julian (mentor)

15 years agoUnbreak options VIMAGE build.
zec [Sat, 6 Jun 2009 12:43:13 +0000 (12:43 +0000)]
Unbreak options VIMAGE build.

Submitted by: julian (mentor)
Approved by: julian (mentor)

15 years agoInitial version of the sec(4) driver for the integrated security engine found
raj [Sat, 6 Jun 2009 09:37:55 +0000 (09:37 +0000)]
Initial version of the sec(4) driver for the integrated security engine found
in Freescale system-on-chip devices.

The following algorithms and schemes are currently supported:
  - 3DES, AES, DES
  - MD5, SHA1, SHA256, SHA384, SHA512

Reviewed by: philip
Obtained from: Freescale, Semihalf

15 years agoProvide 64-bit big endian bus space operations for PowerPC. They are required
raj [Sat, 6 Jun 2009 09:33:32 +0000 (09:33 +0000)]
Provide 64-bit big endian bus space operations for PowerPC. They are required
for the upcoming sec(4) driver.

Submitted by: Piotr Ziecik
Obtained from: Semihalf

15 years agoUse #ifdef APPLE_MAC instead of #ifdef MAC to conditionalize Apple-specific
rwatson [Sat, 6 Jun 2009 07:13:57 +0000 (07:13 +0000)]
Use #ifdef APPLE_MAC instead of #ifdef MAC to conditionalize Apple-specific
behavior for unicode support in UDF so as not to conflict with the MAC
Framework.

Note that Apple's XNU kernel also uses #ifdef MAC for the MAC Framework.

Suggested by: pjd
MFC after: 3 days

15 years agoEliminate trailing_slash, which was made redundant in r193028.
des [Sat, 6 Jun 2009 00:49:49 +0000 (00:49 +0000)]
Eliminate trailing_slash, which was made redundant in r193028.
Remove a couple of 4-year-old "temporary" KASSERTs.
Improve comments.

MFC after: 1 week

15 years agoDrop Giant.
des [Sat, 6 Jun 2009 00:44:13 +0000 (00:44 +0000)]
Drop Giant.

MFC after: 1 week

15 years agoo add bits for STBC and Greenfield
sam [Sat, 6 Jun 2009 00:32:15 +0000 (00:32 +0000)]
o add bits for STBC and Greenfield
o fix some comments

15 years agoOnly four out of nine arguments for ip_ipsec_output() are actually used.
pjd [Fri, 5 Jun 2009 23:53:17 +0000 (23:53 +0000)]
Only four out of nine arguments for ip_ipsec_output() are actually used.
Kill unused arguments except for 'ifp' as it might be used in the future
for detecting IPsec-capable interfaces.

15 years agoreserve node flag bits for a-msdu tx/rx
sam [Fri, 5 Jun 2009 23:37:57 +0000 (23:37 +0000)]
reserve node flag bits for a-msdu tx/rx

15 years agoreserve ioc's for Greenfield and STBC
sam [Fri, 5 Jun 2009 23:36:48 +0000 (23:36 +0000)]
reserve ioc's for Greenfield and STBC

15 years agoSimplify.
pjd [Fri, 5 Jun 2009 23:35:43 +0000 (23:35 +0000)]
Simplify.

15 years agocorrect status code returned for ht capability mismatch on assoc/reassoc
sam [Fri, 5 Jun 2009 23:21:29 +0000 (23:21 +0000)]
correct status code returned for ht capability mismatch on assoc/reassoc

15 years agoo correct/add action frame categories
sam [Fri, 5 Jun 2009 23:20:41 +0000 (23:20 +0000)]
o correct/add action frame categories
o add IEEE80211_STATUS_MISSING_HT_CAPS, added in 11n D3.0 spec

15 years agoadd tid param to ieee80211_notify_replay_failure to get the correct rsc
sam [Fri, 5 Jun 2009 23:10:30 +0000 (23:10 +0000)]
add tid param to ieee80211_notify_replay_failure to get the correct rsc

15 years agoPut intrcnt, eintrcnt, intrnames and eintrnames into the .data section.
kib [Fri, 5 Jun 2009 20:23:29 +0000 (20:23 +0000)]
Put intrcnt, eintrcnt, intrnames and eintrnames into the .data section.

Noted by: "Tseng, Kuo-Lang" <kuo-lang.tseng intel com>, bde
MFC after: 3 days

15 years agomove kernel ipfw-related sources to a separate directory,
luigi [Fri, 5 Jun 2009 19:22:47 +0000 (19:22 +0000)]
move kernel ipfw-related sources to a separate directory,
adjust conf/files and modules' Makefiles accordingly.

No code or ABI changes so this and most of previous related
changes can be easily MFC'ed

MFC after: 5 days

15 years agoImport ACPICA 20090521.
jkim [Fri, 5 Jun 2009 18:50:45 +0000 (18:50 +0000)]
Import ACPICA 20090521.

15 years agoImport ACPICA 20090521.
jkim [Fri, 5 Jun 2009 18:44:36 +0000 (18:44 +0000)]
Import ACPICA 20090521.

15 years agoImport ACPICA 20090521.
jkim [Fri, 5 Jun 2009 18:43:13 +0000 (18:43 +0000)]
Import ACPICA 20090521.

15 years agolack of power save violates 802.11 and is a bug, not a "caveat"
sam [Fri, 5 Jun 2009 17:44:43 +0000 (17:44 +0000)]
lack of power save violates 802.11 and is a bug, not a "caveat"

15 years agonote lack of power save support in ap mode
sam [Fri, 5 Jun 2009 17:26:15 +0000 (17:26 +0000)]
note lack of power save support in ap mode

Submitted by: "Paul B. Mahol" <onemda@gmail.com>

15 years agoDo not force the mtu to 2290; this was done to insure large EAPOL frames
sam [Fri, 5 Jun 2009 17:19:55 +0000 (17:19 +0000)]
Do not force the mtu to 2290; this was done to insure large EAPOL frames
could be handled w/o fragmentation but clobbers user-specified values
such as those required when the interface is bridged.

Submitted by: jim@netgate.com
Reviewed by: Jouni Malinen
MFC after: 3 days

15 years agovm_thread_swapin() needn't validate any pages. The pages are already
alc [Fri, 5 Jun 2009 17:06:20 +0000 (17:06 +0000)]
vm_thread_swapin() needn't validate any pages.  The pages are already
validated by vm_pager_get_pages().

15 years agoSimplify contigfree().
alc [Fri, 5 Jun 2009 16:55:10 +0000 (16:55 +0000)]
Simplify contigfree().

15 years agoUnlock the cache lock before returning when we run out of buffer space
marcus [Fri, 5 Jun 2009 16:44:42 +0000 (16:44 +0000)]
Unlock the cache lock before returning when we run out of buffer space
trying to fill in the full path name.

Reported by: David Naylor <naylor.b.david@gmail.com>
Approved by: kib

15 years agoSeveral ipfw options and actions use a 16-bit argument to indicate
luigi [Fri, 5 Jun 2009 16:16:07 +0000 (16:16 +0000)]
Several ipfw options and actions use a 16-bit argument to indicate
pipes, queues, tags, rule numbers and so on.
These are all different namespaces, and the only thing they have in
common is the fact they use a 16-bit slot to represent the argument.

There is some confusion in the code, mostly for historical reasons,
on how the values 0 and 65535 should be used. At the moment, 0 is
forbidden almost everywhere, while 65535 is used to represent a
'tablearg' argument, i.e. the result of the most recent table() lookup.

For now, try to use explicit constants for the min and max allowed
values, and do not overload the default rule number for that.

Also, make the MTAG_IPFW declaration only visible to the kernel.

NOTE: I think the issue needs to be revisited before 8.0 is out:
the 2^16 namespace limit for rule numbers and pipe/queue is
annoying, and we can easily bump the limit to 2^32 which gives
a lot more flexibility in partitioning the namespace.

MFC after: 5 days

15 years agoRemove clists from the kernel.
ed [Fri, 5 Jun 2009 15:31:38 +0000 (15:31 +0000)]
Remove clists from the kernel.

Clists were originally used by the TTY layer as a text buffer interface.
The advantage of clists were that it would allocate a small set of
additional buffers that could be shared between TTYs when needed. In
the modern days we can just allocate some more KBs of memory to keep the
TTYs satisfied. The global cfreelist also requires synchronisation,
which may not be useful when trying to improve scalability.

The MPSAFE TTY layer uses its own text buffers (ttyinq and ttyoutq). We
had a small amount of drivers in the tree that still uses clists, like
the old USB stack and some keyboard drivers. With the old USB stack gone
and the keyboard drivers changed to use a circular buffer, we can safely
remove clists from the kernel.

15 years agoMove buffer management into kbd and kbdmux drivers.
ed [Fri, 5 Jun 2009 15:19:05 +0000 (15:19 +0000)]
Move buffer management into kbd and kbdmux drivers.

These two drivers seem to be the last consumers of clists. clists are
quite overengineered for simple circular buffers, so I'm adding similar
buffer management routines to the kbd and kbdmux drivers. The input
buffer is now part of the softc structures, instead of having
dynamically allocated cblocks.

15 years agoMove "options MAC" from opt_mac.h to opt_global.h, as it's now in GENERIC
rwatson [Fri, 5 Jun 2009 14:55:22 +0000 (14:55 +0000)]
Move "options MAC" from opt_mac.h to opt_global.h, as it's now in GENERIC
and used in a large number of files, but also because an increasing number
of incorrect uses of MAC calls were sneaking in due to copy-and-paste of
MAC-aware code without the associated opt_mac.h include.

Discussed with: pjd

15 years agoUnifdef MAC label pointer in syncache entries -- in general, ifdef'd
rwatson [Fri, 5 Jun 2009 14:31:03 +0000 (14:31 +0000)]
Unifdef MAC label pointer in syncache entries -- in general, ifdef'd
structure contents are a bad idea in the kernel for binary
compatibility reasons, and this is a single pointer that is now included
in compiles by default anyway due to options MAC being in GENERIC.

15 years agoCorrect MAC compile problems resulting from the new RPC code copying and
rwatson [Fri, 5 Jun 2009 14:29:49 +0000 (14:29 +0000)]
Correct MAC compile problems resulting from the new RPC code copying and
pasting code from the general socket code without also bringing along
required opt_mac.h includes.

15 years agoAdd mac_framework.h include missed when MAC code was (presumably) copied
rwatson [Fri, 5 Jun 2009 14:23:24 +0000 (14:23 +0000)]
Add mac_framework.h include missed when MAC code was (presumably) copied
from another file.

15 years agoDon't check MAC in the NFS server ACL set path, right now we aren't
rwatson [Fri, 5 Jun 2009 14:15:00 +0000 (14:15 +0000)]
Don't check MAC in the NFS server ACL set path, right now we aren't
enforcing MAC for NFS clients.

15 years agoTrim old remnants of per-CPU KTR buffers.
jhb [Fri, 5 Jun 2009 14:07:00 +0000 (14:07 +0000)]
Trim old remnants of per-CPU KTR buffers.

Submitted by:  Eygene Ryabinkin

15 years agoFix spelling of MAC check for 8.x version of MAC Framework, not noticed due
rwatson [Fri, 5 Jun 2009 13:55:33 +0000 (13:55 +0000)]
Fix spelling of MAC check for 8.x version of MAC Framework, not noticed due
to a lack of an opt_mac.h include, which I won't add for now as options MAC
will soon move to opt_global.h.

Spotted by: pjd