]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoEnable SCTP by default for GENERIC kernels in order to give it
Xin LI [Thu, 14 Jun 2007 17:14:27 +0000 (17:14 +0000)]
Enable SCTP by default for GENERIC kernels in order to give it
more exposure.  The current state of SCTP implementation is
considered to be ready for 32-bit platforms, but still need some
work/testing on 64-bit platforms.

Approved by: re (kensmith)
Discussed with: rrs

17 years agoExpand callout compatibility macros in the main usb bridges.
Warner Losh [Thu, 14 Jun 2007 16:23:31 +0000 (16:23 +0000)]
Expand callout compatibility macros in the main usb bridges.

17 years agoMove PWR_* from usb_port.h to usb.h
Warner Losh [Thu, 14 Jun 2007 16:12:04 +0000 (16:12 +0000)]
Move PWR_* from usb_port.h to usb.h

17 years agoFix an error message that was beside the point.
Yaroslav Tykhiy [Thu, 14 Jun 2007 15:18:44 +0000 (15:18 +0000)]
Fix an error message that was beside the point.

17 years agoMove malloc definitions to usb.h.
Warner Losh [Thu, 14 Jun 2007 15:09:21 +0000 (15:09 +0000)]
Move malloc definitions to usb.h.
Also, remove usb_malloc_type: it was unused.
Remove METHODS_NONE: it was unused.
Move include of opt_usb.h from usb_port.h to usb.h, since usb_port.h is
going away (there will be a usb_compat.h for out-of-tree drivers that want it).

17 years agoFix a typical off-by-one error that can result in an unterminated string:
Yaroslav Tykhiy [Thu, 14 Jun 2007 14:44:04 +0000 (14:44 +0000)]
Fix a typical off-by-one error that can result in an unterminated string:

strncpy(dst, src, sizeof(dst));

by substituting the safer strlcpy() for strncpy().

X-Security: none (the source string isn't user-supplied)

17 years agoUse the current user's login class for the decisions about where
Yaroslav Tykhiy [Thu, 14 Jun 2007 13:07:06 +0000 (13:07 +0000)]
Use the current user's login class for the decisions about where
the nologin(5) file is located and whether the user may bypass its
restriction.

Add some error checks.

Approved by: des
PR: bin/107612

17 years ago- Suppress compiler optimization so that orb[1] must be written first.
Hidetoshi Shimokawa [Thu, 14 Jun 2007 12:52:13 +0000 (12:52 +0000)]
- Suppress compiler optimization so that orb[1] must be written first.
  We may need an explicit memory barrier for other architectures other than i386/amd64.

MFC after: 3 days

17 years agoBuffer optimization and locking cleanup. Don't resize/malloc
Ariff Abdullah [Thu, 14 Jun 2007 11:15:51 +0000 (11:15 +0000)]
Buffer optimization and locking cleanup. Don't resize/malloc
unless it is really necessary to ease down unlock/lock sequence.

17 years agoDrain all callout handlers during driver detach appropriately.
Ariff Abdullah [Thu, 14 Jun 2007 11:13:38 +0000 (11:13 +0000)]
Drain all callout handlers during driver detach appropriately.

17 years agoDon't push too hard waiting/looping for codec detection interrupt.
Ariff Abdullah [Thu, 14 Jun 2007 11:11:06 +0000 (11:11 +0000)]
Don't push too hard waiting/looping for codec detection interrupt.
Let the interrupt do the job waking us up.

17 years ago- Add option to allow sleeping malloc(9).
Ariff Abdullah [Thu, 14 Jun 2007 11:10:21 +0000 (11:10 +0000)]
- Add option to allow sleeping malloc(9).
- Cleanup locking assertions that aren't needed anymore.

17 years agoDocument the quirks of ~/.login_conf and LOGIN_MECLASS.
Yaroslav Tykhiy [Thu, 14 Jun 2007 09:33:37 +0000 (09:33 +0000)]
Document the quirks of ~/.login_conf and LOGIN_MECLASS.

17 years agoImprove mdoc(7) markup.
Yaroslav Tykhiy [Thu, 14 Jun 2007 07:31:59 +0000 (07:31 +0000)]
Improve mdoc(7) markup.

17 years agoRequire users to provide a length information for inet_ntoa_r,
Xin LI [Thu, 14 Jun 2007 07:13:28 +0000 (07:13 +0000)]
Require users to provide a length information for inet_ntoa_r,
this is common on other platforms.

Reported by: pointyhat via kris

17 years agoUpdate some comments, mostly regarding LOGIN_MECLASS and ~/.login_conf.
Yaroslav Tykhiy [Thu, 14 Jun 2007 06:42:49 +0000 (06:42 +0000)]
Update some comments, mostly regarding LOGIN_MECLASS and ~/.login_conf.

17 years agoOops, back out previous commit since it was backwards to a wrong branch.
Bruce Evans [Thu, 14 Jun 2007 05:57:13 +0000 (05:57 +0000)]
Oops, back out previous commit since it was backwards to a wrong branch.

17 years agoMFC: 1.11: fix the threshold for (not) using the simple Taylor approximation.
Bruce Evans [Thu, 14 Jun 2007 05:51:00 +0000 (05:51 +0000)]
MFC: 1.11: fix the threshold for (not) using the simple Taylor approximation.

17 years agoAdd a missing Va option.
Christian Brueffer [Thu, 14 Jun 2007 05:32:23 +0000 (05:32 +0000)]
Add a missing Va option.

17 years agoIn the previous changeset a cast of myminor to u_int were
Xin LI [Thu, 14 Jun 2007 03:16:16 +0000 (03:16 +0000)]
In the previous changeset a cast of myminor to u_int were
removed, which will cause problems on architectures where
longs are longer than ints, for instance,
"mknod foo c 0 0xffff00ff"
would fail in such cases.

Use a proper cast instead.

Prompted by: bde

17 years agoSpell statistics more correctly in comments.
Robert Watson [Thu, 14 Jun 2007 03:02:33 +0000 (03:02 +0000)]
Spell statistics more correctly in comments.

17 years agoInclude priv.h to pick up suser(9) definitions, missed in an earlier
Robert Watson [Wed, 13 Jun 2007 22:42:43 +0000 (22:42 +0000)]
Include priv.h to pick up suser(9) definitions, missed in an earlier
commit.

Warnings spotted by: kris

17 years agoDon't clobber tf_err with the eva from a page fault as the page fault
John Baldwin [Wed, 13 Jun 2007 22:37:48 +0000 (22:37 +0000)]
Don't clobber tf_err with the eva from a page fault as the page fault
address is saved in ksi_addr already.

PR: i386/101379
Submitted by: Tijl Coosemans : tijl ulyssis org

17 years agoClose a very narrow race that might cause a trigger allocation to be
Robert Watson [Wed, 13 Jun 2007 21:17:23 +0000 (21:17 +0000)]
Close a very narrow race that might cause a trigger allocation to be
leaked if a trigger is delivered as the trigger device is closed.

Obtained from: TrustedBSD Project

17 years agoconfig_detach is just used in one place, so expand it inline.
Warner Losh [Wed, 13 Jun 2007 20:58:57 +0000 (20:58 +0000)]
config_detach is just used in one place, so expand it inline.

17 years agoImprove the ktrace locking somewhat to reduce overhead:
John Baldwin [Wed, 13 Jun 2007 20:01:42 +0000 (20:01 +0000)]
Improve the ktrace locking somewhat to reduce overhead:
- Depessimize userret() in kernels where KTRACE is enabled by doing an
  unlocked check of the per-process queue of pending events before
  acquiring any locks.  Previously ktr_userret() unconditionally acquired
  the global ktrace_sx lock on every return to userland for every thread,
  even if ktrace wasn't enabled for the thread.
- Optimize the locking in exit() to first perform an unlocked read of
  p_traceflag to see if ktrace is enabled and only acquire locks and
  teardown ktrace if the test succeeds.  Also, explicitly disable tracing
  before draining any pending events so the pending events actually get
  written out.  The unlocked read is safe because proc lock is acquired
  earlier after single-threading so p_traceflag can't change between then
  and this check (well, it can currently due to a bug in ktrace I will fix
  next, but that race existed prior to this change as well).

Reviewed by: rwatson

17 years agoConditionally acquire Giant when dropping a reference on the ktrace vnode
John Baldwin [Wed, 13 Jun 2007 19:41:47 +0000 (19:41 +0000)]
Conditionally acquire Giant when dropping a reference on the ktrace vnode
during execve() when turning off tracing due to executing a setuid binary
as non-root.  Previously this could fail to acquire Giant and fail an
assertion if the ktrace file was on a non-MPSAFE filesystem and the
executable was on an MPSAFE filesystem.

MFC after: 3 days
Reported by: kris

17 years agoCatch up with the code.
Christian Brueffer [Wed, 13 Jun 2007 19:22:29 +0000 (19:22 +0000)]
Catch up with the code.

Submitted by: peter

17 years agoFailing to set new frequency should not lead to powerd exiting.
Dmitry Morozovsky [Wed, 13 Jun 2007 19:05:11 +0000 (19:05 +0000)]
Failing to set new frequency should not lead to powerd exiting.
Change err(3) to warn(3) as three other cases.

Approved by: njl, des

17 years agoAdd the vlan tag to the bridge route table. This allows a vlan trunk to be
Andrew Thompson [Wed, 13 Jun 2007 18:58:04 +0000 (18:58 +0000)]
Add the vlan tag to the bridge route table. This allows a vlan trunk to be
bridged, previously legitimate traffic was not passed as the bridge could not
tell that it was on a different Ethernet segment.

All non-tagged traffic is treated as vlan1 as per IEEE 802.1Q-2003

17 years agoRemove IPX over IP tunneling pieces from ifconfig(8), omitted portion of
Robert Watson [Wed, 13 Jun 2007 18:07:59 +0000 (18:07 +0000)]
Remove IPX over IP tunneling pieces from ifconfig(8), omitted portion of
previous commit:

  Remove IPX over IP tunneling support, which allows IPX routing over IP
  tunnels, and was not MPSAFE.  The code can be easily restored in the
  event that someone with an IPX over IP tunnel configuration can work
  with me to test patches.

  This removes one of five remaining consumers of NET_NEEDS_GIANT.

  Approved by:    re (kensmith)

Spotted by: Artem Naluzhny <tutat nhamon dot com dot ua>

17 years agoDo not attempt to enable AHCI mode on ALi SATA controllers other
Bruce M Simpson [Wed, 13 Jun 2007 17:44:49 +0000 (17:44 +0000)]
Do not attempt to enable AHCI mode on ALi SATA controllers other
than the 5288.

It is not correctly implemented in earlier silicon, and the BIOS often
lies about AHCI capability on platforms where these chips are deployed.
With this change I am able to boot FreeBSD on the ASUS Vintage AH-1
barebones system.

Approved by: sos

17 years agoAdd myself as a commiter.
Timur I. Bakeyev [Wed, 13 Jun 2007 16:37:36 +0000 (16:37 +0000)]
Add myself as a commiter.

Approved by: shaun (mentor)

17 years ago- fix bindx to check addresses against socket's protocol family
Randall Stewart [Wed, 13 Jun 2007 14:39:41 +0000 (14:39 +0000)]
- fix bindx to check addresses against socket's protocol family

17 years agoRemove IPX over IP tunneling support, which allows IPX routing over IP
Robert Watson [Wed, 13 Jun 2007 14:01:43 +0000 (14:01 +0000)]
Remove IPX over IP tunneling support, which allows IPX routing over IP
tunnels, and was not MPSAFE.  The code can be easily restored in the
event that someone with an IPX over IP tunnel configuration can work
with me to test patches.

This removes one of five remaining consumers of NET_NEEDS_GIANT.

Approved by: re (kensmith)

17 years agoRemove two more instances of the USBDEV() macro.
Ruslan Ermilov [Wed, 13 Jun 2007 12:36:01 +0000 (12:36 +0000)]
Remove two more instances of the USBDEV() macro.

17 years agoAdd missing ng_uncallout() on node shutdown.
Alexander Motin [Wed, 13 Jun 2007 11:01:17 +0000 (11:01 +0000)]
Add missing ng_uncallout() on node shutdown.

Approved by: glebius (mentor)

17 years agoMajor update for the recently overhauled nfe(4) driver.
Christian Brueffer [Wed, 13 Jun 2007 08:44:37 +0000 (08:44 +0000)]
Major update for the recently overhauled nfe(4) driver.

Reviewed by: yongari

17 years agoUnbreak high resolution profiling a little: use dummy asms to prevent
Bruce Evans [Wed, 13 Jun 2007 06:17:48 +0000 (06:17 +0000)]
Unbreak high resolution profiling a little: use dummy asms to prevent
timing loops being optimized away.

Once apon a time, gcc promised not to optimize away timing loops, but
gcc started optimizing away the call to a null function in the timing
loop here some time between gcc-3.3.3 and gcc-3.4.6, and it started
optimizing away the timing loop itself some time between gcc-3.4.6
and gcc-4.2.

17 years agoEliminate dead code: We have not performed pageouts on the kernel object
Alan Cox [Wed, 13 Jun 2007 06:10:10 +0000 (06:10 +0000)]
Eliminate dead code: We have not performed pageouts on the kernel object
in this millenium.

17 years agoRemove more unnecessary and unwanted __OtherBSD__ ifdefs.
Warner Losh [Wed, 13 Jun 2007 06:00:10 +0000 (06:00 +0000)]
Remove more unnecessary and unwanted __OtherBSD__ ifdefs.

17 years agoRemove more __OtherBSD__ ifdefs that don't make sense, and haven't for
Warner Losh [Wed, 13 Jun 2007 05:45:48 +0000 (05:45 +0000)]
Remove more __OtherBSD__ ifdefs that don't make sense, and haven't for
some time.

17 years agoExpand USB_DETACH, USB_ATTACH and USB_MATCH inline. No functional
Warner Losh [Wed, 13 Jun 2007 05:37:59 +0000 (05:37 +0000)]
Expand USB_DETACH, USB_ATTACH and USB_MATCH inline.  No functional
change, and MD5's appear to be the same.

17 years ago- import new common code for the T304
Kip Macy [Wed, 13 Jun 2007 05:36:00 +0000 (05:36 +0000)]
- import new common code for the T304
- update to firmware version 4.1.0

- switch over to standard method for initializing cdevs (contributed by scottl@)
- break out timer_reclaim_task to be per-port
- move msix teardown into separate function
- fix bus_setup_intr for msi-x for the multi-port case so that msi-x resources
  are not corrupted on unload
- handle 10/100/1000 base-T media and auto negotiation
- bind qset to cpu even for singleq case
- white space cleanups
- remove recursive PORT_LOCK
- move mtu setting to separate function
- stop and re-init port when changing mtu
- replace all direct references to m_data with calls to mtod
- handle attach failure better by not trying to de-initialize
  taskqueues when they have not been allocated
- no longer default to jumbo frames

Sponsored by: Chelsio
MFC after: 3 days

17 years agoWARNS=6
Xin LI [Wed, 13 Jun 2007 05:30:07 +0000 (05:30 +0000)]
WARNS=6

17 years agoEnable GEOM_PART_MBR by default. On ia64 this replaces GEOM_MBR.
Marcel Moolenaar [Wed, 13 Jun 2007 05:07:42 +0000 (05:07 +0000)]
Enable GEOM_PART_MBR by default. On ia64 this replaces GEOM_MBR.

17 years agoAdd the MBR partitioning scheme to g_part. This does not yet
Marcel Moolenaar [Wed, 13 Jun 2007 04:27:36 +0000 (04:27 +0000)]
Add the MBR partitioning scheme to g_part. This does not yet
support the ability to install boot code.

17 years agoMake 'ar' write test a tad more portable.
Tim Kientzle [Wed, 13 Jun 2007 03:40:02 +0000 (03:40 +0000)]
Make 'ar' write test a tad more portable.

17 years agoRead support for the new GNU tar sparse formats added in gtar 1.15 and
Tim Kientzle [Wed, 13 Jun 2007 03:35:37 +0000 (03:35 +0000)]
Read support for the new GNU tar sparse formats added in gtar 1.15 and
gtar 1.16.

17 years agoAdd some options to libarchive_test:
Tim Kientzle [Wed, 13 Jun 2007 03:30:46 +0000 (03:30 +0000)]
Add some options to libarchive_test:
   -k: like make -k, try to keep going after errors.
   -q: quiet

17 years agoFix a broken function declaration.
Tim Kientzle [Wed, 13 Jun 2007 03:29:15 +0000 (03:29 +0000)]
Fix a broken function declaration.

17 years agosctp_process_inpcb() wants an offset parameter in size_t,
Xin LI [Wed, 13 Jun 2007 02:37:00 +0000 (02:37 +0000)]
sctp_process_inpcb() wants an offset parameter in size_t,
so define it as what it is expected.  This fixes WARNS=3
without NO_WERROR build.

17 years agoRegen.
Stephane E. Potvin [Wed, 13 Jun 2007 02:18:12 +0000 (02:18 +0000)]
Regen.

17 years agoOptions spring cleanup:
Stephane E. Potvin [Wed, 13 Jun 2007 02:08:04 +0000 (02:08 +0000)]
Options spring cleanup:
- Add and document the KVM and KVM_SUPPORT options that
are needed for the ifmcstats(3) makefile
- Garbage collect unused variables
- Add missing inclusion of bsd.own.mk where needed

Approved by: kan (mentor)
Reviewed by: ru

17 years ago- Fixed cookie handling to calc an RTO when
Randall Stewart [Wed, 13 Jun 2007 01:31:53 +0000 (01:31 +0000)]
- Fixed cookie handling to calc an RTO when
  its an INIT collision case.
- Fixed RTO calc to maintain a seperate variable to track
  if a RTO calc as been done, this allows the RTO var to be
  doubled during initial timeouts.
- Reduces the amount of stack used by process control.
- Use a constant for the peer chunk overhead.
- Name change to spell candidate correctly.

17 years agoCatch up with USB cleanups and fix the world
Maksim Yevmenkin [Wed, 13 Jun 2007 00:32:00 +0000 (00:32 +0000)]
Catch up with USB cleanups and fix the world

17 years ago - Include opt_sched.h for SCHED_STATS.
Jeff Roberson [Tue, 12 Jun 2007 23:27:31 +0000 (23:27 +0000)]
 - Include opt_sched.h for SCHED_STATS.

17 years agoFix yet another (make install) stopper with wrong sourcefilter.3 links
Andrey A. Chernov [Tue, 12 Jun 2007 22:32:52 +0000 (22:32 +0000)]
Fix yet another (make install) stopper with wrong sourcefilter.3 links

17 years agoRemove compat macros. md5 the same.
Warner Losh [Tue, 12 Jun 2007 21:35:18 +0000 (21:35 +0000)]
Remove compat macros.  md5 the same.

17 years agoRecoverdisk is now part of the base system.
Poul-Henning Kamp [Tue, 12 Jun 2007 21:33:11 +0000 (21:33 +0000)]
Recoverdisk is now part of the base system.

17 years agoAdd recoverdisk to the base system, it is far too useful to live
Poul-Henning Kamp [Tue, 12 Jun 2007 21:31:39 +0000 (21:31 +0000)]
Add recoverdisk to the base system, it is far too useful to live
in obscurity over in src/tools.

(Repocopied)

17 years agofix cassert failure by adjusting padding
Kip Macy [Tue, 12 Jun 2007 21:19:12 +0000 (21:19 +0000)]
fix cassert failure by adjusting padding

17 years agoFix typo in filename from mismerged earlier rev of this file.
Bruce M Simpson [Tue, 12 Jun 2007 21:05:31 +0000 (21:05 +0000)]
Fix typo in filename from mismerged earlier rev of this file.

17 years ago - Fix kse by moving the upcalls list back out of the zero'd section.
Jeff Roberson [Tue, 12 Jun 2007 20:22:06 +0000 (20:22 +0000)]
 - Fix kse by moving the upcalls list back out of the zero'd section.
   I had tested this with the wrong libpthread.

17 years agoUse if_capenable to allow LRO enabled drivers to bypass
Andrew Gallatin [Tue, 12 Jun 2007 19:53:44 +0000 (19:53 +0000)]
Use if_capenable to allow LRO enabled drivers to bypass
the MTU check in ether_input().

17 years ago - Garbage collect unused concurrency functions.
Jeff Roberson [Tue, 12 Jun 2007 19:50:31 +0000 (19:50 +0000)]
 - Garbage collect unused concurrency functions.

17 years ago - Garbage collect unused concurrency functions.
Jeff Roberson [Tue, 12 Jun 2007 19:49:39 +0000 (19:49 +0000)]
 - Garbage collect unused concurrency functions.
 - Remove unused kse fields from struct proc.
 - Group remaining fields and #ifdef KSE them.
 - Move some kern_kse.c only prototypes out of proc and into kern_kse.

Discussed with: Julian

17 years agoFirst pass at removing __OtherBSD__. We can't blindly remove all of
Warner Losh [Tue, 12 Jun 2007 19:40:20 +0000 (19:40 +0000)]
First pass at removing __OtherBSD__.  We can't blindly remove all of
them in bulk because there is at least one feature that's
unimplemented on FreeBSD that needs to be done and these are good
placeholders.

17 years agoUse the new IFCAP_LRO to enable/disable LRO.
Andrew Gallatin [Tue, 12 Jun 2007 19:15:16 +0000 (19:15 +0000)]
Use the new IFCAP_LRO to enable/disable LRO.

17 years agoFix a typo which crept in from an earlier version of this file.
Bruce M Simpson [Tue, 12 Jun 2007 19:08:38 +0000 (19:08 +0000)]
Fix a typo which crept in from an earlier version of this file.

17 years agoDon't pretend to support !BSD systems.
Warner Losh [Tue, 12 Jun 2007 19:01:32 +0000 (19:01 +0000)]
Don't pretend to support !BSD systems.

17 years agoRemove usb_kthread_create*. They are now unused.
Warner Losh [Tue, 12 Jun 2007 18:52:29 +0000 (18:52 +0000)]
Remove usb_kthread_create*.  They are now unused.

17 years agounifdef -D__FreeBSD__ -U__NetBSD__ -U__OpenBSD__
Warner Losh [Tue, 12 Jun 2007 18:52:07 +0000 (18:52 +0000)]
unifdef -D__FreeBSD__ -U__NetBSD__ -U__OpenBSD__
Use kthread_create directly

17 years agoAdd missing userland support files from previous commit for the new
Bruce M Simpson [Tue, 12 Jun 2007 18:03:37 +0000 (18:03 +0000)]
Add missing userland support files from previous commit for the new
multicast source filter API functions.

17 years agoSpaces to tab
Gregory Neil Shapiro [Tue, 12 Jun 2007 17:33:56 +0000 (17:33 +0000)]
Spaces to tab

17 years agoAdd a new rc.conf variable, sendmail_rebuild_aliases, which tells
Gregory Neil Shapiro [Tue, 12 Jun 2007 17:33:23 +0000 (17:33 +0000)]
Add a new rc.conf variable, sendmail_rebuild_aliases, which tells
/etc/rc.d/sendmail whether or not to run newaliases if the database
is missing or the aliases text file is newer than aliases.db.

In my opinion, the aliases file should never be automatically rebuilt.
The current text form could represent a work in progress.  Therefore,
in FreeBSD 7.0, this new option will default to "NO".  When this rc.d
change is MFC'ed, it will need to remain "YES" to maintain backward
compatibility.

PR: conf/86252
Approved by: re (kensmith)
MFC after: 3 days

17 years agoEliminate usb_thread_t.
Warner Losh [Tue, 12 Jun 2007 17:30:54 +0000 (17:30 +0000)]
Eliminate usb_thread_t.

17 years agoRemove USBDEV() macro. We do not need a macro that is defined as its
Warner Losh [Tue, 12 Jun 2007 16:52:07 +0000 (16:52 +0000)]
Remove USBDEV() macro.  We do not need a macro that is defined as its
only argument.  It was used inconsistently in the tree, so remove it.

17 years agoclalloc and clfree are useles. Remove them. Remove dead code that's
Warner Losh [Tue, 12 Jun 2007 16:41:29 +0000 (16:41 +0000)]
clalloc and clfree are useles.  Remove them.  Remove dead code that's
always been dead for years, but has been obfuscated by these macros.

17 years agoAdd regression test for SSM code.
Bruce M Simpson [Tue, 12 Jun 2007 16:29:22 +0000 (16:29 +0000)]
Add regression test for SSM code.

17 years agoImport rewrite of IPv4 socket multicast layer to support source-specific
Bruce M Simpson [Tue, 12 Jun 2007 16:24:56 +0000 (16:24 +0000)]
Import rewrite of IPv4 socket multicast layer to support source-specific
and protocol-independent host mode multicast. The code is written to
accomodate IPv6, IGMPv3 and MLDv2 with only a little additional work.

This change only pertains to FreeBSD's use as a multicast end-station and
does not concern multicast routing; for an IGMPv3/MLDv2 router
implementation, consider the XORP project.

The work is based on Wilbert de Graaf's IGMPv3 code drop for FreeBSD 4.6,
which is available at: http://www.kloosterhof.com/wilbert/igmpv3.html

Summary
 * IPv4 multicast socket processing is now moved out of ip_output.c
   into a new module, in_mcast.c.
 * The in_mcast.c module implements the IPv4 legacy any-source API in
   terms of the protocol-independent source-specific API.
 * Source filters are lazy allocated as the common case does not use them.
   They are part of per inpcb state and are covered by the inpcb lock.
 * struct ip_mreqn is now supported to allow applications to specify
   multicast joins by interface index in the legacy IPv4 any-source API.
 * In UDP, an incoming multicast datagram only requires that the source
   port matches the 4-tuple if the socket was already bound by source port.
   An unbound socket SHOULD be able to receive multicasts sent from an
   ephemeral source port.
 * The UDP socket multicast filter mode defaults to exclusive, that is,
   sources present in the per-socket list will be blocked from delivery.
 * The RFC 3678 userland functions have been added to libc: setsourcefilter,
   getsourcefilter, setipv4sourcefilter, getipv4sourcefilter.
 * Definitions for IGMPv3 are merged but not yet used.
 * struct sockaddr_storage is now referenced from <netinet/in.h>. It
   is therefore defined there if not already declared in the same way
   as for the C99 types.
 * The RFC 1724 hack (specify 0.0.0.0/8 addresses to IP_MULTICAST_IF
   which are then interpreted as interface indexes) is now deprecated.
 * A patch for the Rhyolite.com routed in the FreeBSD base system
   is available in the -net archives. This only affects individuals
   running RIPv1 or RIPv2 via point-to-point and/or unnumbered interfaces.
 * Make IPv6 detach path similar to IPv4's in code flow; functionally same.
 * Bump __FreeBSD_version to 700048; see UPDATING.

This work was financially supported by another FreeBSD committer.

Obtained from:  p4://bms_netdev
Submitted by:   Wilbert de Graaf (original work)
Reviewed by:    rwatson (locking), silence from fenner,
net@ (but with encouragement)

17 years agoExpand USB_ATTACH_{ERROR,SUCCESS}_RETURN inline and eliminate from
Warner Losh [Tue, 12 Jun 2007 15:37:19 +0000 (15:37 +0000)]
Expand USB_ATTACH_{ERROR,SUCCESS}_RETURN inline and eliminate from
usb_port.h.  They aren't needed, and are a legacy of this code's past.

17 years ago- Document support for Realtek ALC660 and Analog Devices AD1988B.
Joel Dahl [Tue, 12 Jun 2007 15:26:41 +0000 (15:26 +0000)]
- Document support for Realtek ALC660 and Analog Devices AD1988B.
- Fix minor typos.

17 years agoDelete description of non-existent options: "-4" and "-6".
Motoyuki Konno [Tue, 12 Jun 2007 13:28:55 +0000 (13:28 +0000)]
Delete description of non-existent options: "-4" and "-6".

ntpd's "-4" and "-6" options are described in the original documentation
(contrib/ntp/html/ntpd.html).  It may be original's doc bug.

PR: docs/112642
Submitted by: Seth Hieronymus<shieronymus@speakeasy.net>
Discussed with: ume
MFC after: 1 week

17 years agoDrop the argument to the OUT macro because it can't emit
Yaroslav Tykhiy [Tue, 12 Jun 2007 12:17:25 +0000 (12:17 +0000)]
Drop the argument to the OUT macro because it can't emit
anything but the pattern space anyway.  Apply style(9)
to the macro.

Tested with: md5(1)

17 years agoDon't forget to clear out the hold space for each subsequent file
Yaroslav Tykhiy [Tue, 12 Jun 2007 12:05:24 +0000 (12:05 +0000)]
Don't forget to clear out the hold space for each subsequent file
when in -i mode so that each file gets a clean context of its own.

Add a regression test for the bug.

Tested with: regression tests

17 years agoAdd cust_pkg() which iteratively tries to install packages.
Poul-Henning Kamp [Tue, 12 Jun 2007 11:35:29 +0000 (11:35 +0000)]
Add cust_pkg() which iteratively tries to install packages.

Move /usr/local/etc to /etc/local so that installed packages can be frobbed.

Use a more compatible 512MB card geometry.

17 years ago- Restructure so bindx functions are not done inline to socket option
Randall Stewart [Tue, 12 Jun 2007 11:21:00 +0000 (11:21 +0000)]
- Restructure so bindx functions are not done inline to socket option
  but are a seperate call that can be re-used if needed.
- 64 bit issues
  o re-arrange cookie so it is better 64 bit aligned
  o For wire level things we need the packed attribute.

17 years agoThere is a symbolic antonym for REPLACE as a flag to cspace()
Yaroslav Tykhiy [Tue, 12 Jun 2007 10:56:38 +0000 (10:56 +0000)]
There is a symbolic antonym for REPLACE as a flag to cspace()
and mf_fgets(): APPEND.  So use it instead of a 0 constant for
clarity.

Tested with: md5(1)

17 years agoThe maximum size of the sum of all segment lengths in a given DMA mapping
Pyun YongHyeon [Tue, 12 Jun 2007 10:51:47 +0000 (10:51 +0000)]
The maximum size of the sum of all segment lengths in a given DMA mapping
should be 65535 + link layer headers.

Pointed out by: gallatin

17 years agoCatch up with variable name changes in struct lagg_protos.
Andrew Thompson [Tue, 12 Jun 2007 10:07:57 +0000 (10:07 +0000)]
Catch up with variable name changes in struct lagg_protos.

17 years agoRemove duplicate. Was that a bug? :-)
Philip Paeps [Tue, 12 Jun 2007 09:20:31 +0000 (09:20 +0000)]
Remove duplicate.  Was that a bug? :-)

17 years ago- add myself
Chin-San Huang [Tue, 12 Jun 2007 09:05:06 +0000 (09:05 +0000)]
- add myself

Approved by: rafan (mentor)

17 years ago - Move some common code out of sched_fork_exit() and back into fork_exit().
Jeff Roberson [Tue, 12 Jun 2007 07:47:09 +0000 (07:47 +0000)]
 - Move some common code out of sched_fork_exit() and back into fork_exit().

17 years agonon-functional cleanup
Andrew Thompson [Tue, 12 Jun 2007 07:29:11 +0000 (07:29 +0000)]
non-functional cleanup
- remove dead code
- use consistent variable names
- gc unused defines
- whitespace cleanup

17 years agoSolve a complex exit race introduced with thread_lock:
Jeff Roberson [Tue, 12 Jun 2007 07:24:46 +0000 (07:24 +0000)]
Solve a complex exit race introduced with thread_lock:
 - Add a count of exiting threads, p_exitthreads, to struct proc.
 - Increment p_exithreads when we set the deadthread in thread_exit().
 - When we thread_stash() a deadthread use an atomic to drop the count.
 - Spin until the p_exithreads count reaches 0 in thread_wait().
 - Lock the last exiting thread momentarily to be certain that it has
   exited cpu_throw().
 - Restructure thread_wait().  It does not need a loop as there will only
   ever be one thread.

Tested by: moose@opera.com
Reported by: kris, moose@opera.com

17 years agoNuke duplicated __FBSDID.
Pyun YongHyeon [Tue, 12 Jun 2007 04:33:21 +0000 (04:33 +0000)]
Nuke duplicated __FBSDID.

17 years agoAdd checks for contigmalloc(9) failure.
Pyun YongHyeon [Tue, 12 Jun 2007 04:30:30 +0000 (04:30 +0000)]
Add checks for contigmalloc(9) failure.

17 years agoIncrease a maximum segment size of DMA to 4096. Previously it used
Pyun YongHyeon [Tue, 12 Jun 2007 02:35:01 +0000 (02:35 +0000)]
Increase a maximum segment size of DMA to 4096. Previously it used
MCLBYTES for the segment size but it used too many Tx descriptors in
TSO case.
While I'm here adjust maximum size of the sum of all segment lengths
in a given DMA mapping to 65535, the maximum size, in bytes, of a IP
packet.

17 years agoAdd nfe(4) to the list of drivers supported by GENERIC kernel.
Pyun YongHyeon [Tue, 12 Jun 2007 02:24:30 +0000 (02:24 +0000)]
Add nfe(4) to the list of drivers supported by GENERIC kernel.
While I'm here comment out nve(4) as nfe(4) will take over.

Approved by: re