]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
24 years agoThis commit was generated by cvs2svn to compensate for changes in r59901,
David E. O'Brien [Tue, 2 May 2000 20:00:07 +0000 (20:00 +0000)]
This commit was generated by cvs2svn to compensate for changes in r59901,
which included commits to RCS files with non-trunk default branches.

24 years agohistory fix
David E. O'Brien [Tue, 2 May 2000 20:00:07 +0000 (20:00 +0000)]
history fix

Submitted by: ache (forwarded to Christos Zoulas where fix was agreed to)

24 years agoDon't assume *lasthook is initialized in ng_path2node().
Archie Cobbs [Tue, 2 May 2000 17:09:46 +0000 (17:09 +0000)]
Don't assume *lasthook is initialized in ng_path2node().

24 years agoRemove an unnecessary error message
Luigi Rizzo [Tue, 2 May 2000 15:39:36 +0000 (15:39 +0000)]
Remove an unnecessary error message

24 years agoDisconnect libm from the build tree. It's broken, not being
Josef Karthauser [Tue, 2 May 2000 14:16:19 +0000 (14:16 +0000)]
Disconnect libm from the build tree.  It's broken, not being
maintained, and has been replaced by msun.  The libm sources
shouldn't be removed just yet as there are parts that should be
merged into msun first.

PR: misc/17848
Discussed with: phk & bde

24 years agoUpdate my e-mail address
Nick Hibma [Tue, 2 May 2000 11:58:47 +0000 (11:58 +0000)]
Update my e-mail address

24 years agoAdd the udbp module
Nick Hibma [Tue, 2 May 2000 11:51:07 +0000 (11:51 +0000)]
Add the udbp module

24 years agoAdd missing man pages. Fix various compliance bugs, mostly having to do with
Jason Evans [Tue, 2 May 2000 06:51:40 +0000 (06:51 +0000)]
Add missing man pages.  Fix various compliance bugs, mostly having to do with
error return values.  Implement pthread_mutexattr_gettype().

PR: docs/16537, docs/17538

24 years agoAdd PnP ID for vadem based plug and play card.
Warner Losh [Tue, 2 May 2000 06:50:17 +0000 (06:50 +0000)]
Add PnP ID for vadem based plug and play card.

Submitted by: Dave Belfer-Shevett <shevett@stonekeep.com>
Fix by: Bill Paul

24 years agoAdd ENOTSUP, which is specified as an error return value for various
Jason Evans [Tue, 2 May 2000 06:02:11 +0000 (06:02 +0000)]
Add ENOTSUP, which is specified as an error return value for various
POSIX/SUSv2 functions.

24 years agoAdd the logical ID of FUJITSU Modem 33600 PNP/I2.
Seigo Tanimura [Tue, 2 May 2000 05:54:11 +0000 (05:54 +0000)]
Add the logical ID of FUJITSU Modem 33600 PNP/I2.

PR: kern/18257
Submitted by: Takanori Watanabe <takawata@shidahara1.planet.sci.kobe-u.ac.jp>

24 years agoAdd man page for pxeboot.
Doug White [Tue, 2 May 2000 05:01:38 +0000 (05:01 +0000)]
Add man page for pxeboot.

Reviewed By: asmodai

24 years agoFix a long vs. int problem that was fatal on i386s with 64bit longs (but
Tim Vanderhoek [Tue, 2 May 2000 02:36:03 +0000 (02:36 +0000)]
Fix a long vs. int problem that was fatal on i386s with 64bit longs (but
not alphas with 64bit longs).

Submitted by: bde (a while ago)

24 years agoCause modules to build with the kernel build. Modules are removed
Chuck Robey [Tue, 2 May 2000 02:26:04 +0000 (02:26 +0000)]
Cause modules to build with the kernel build.  Modules are removed
from the sys Makefile's SUBDIRs.  This is conditioned in make.conf by the
NO_MODULES variable and the existence of the modules directory.  The
actual location of the modules is not modified.  Changes in Makefiles
only, this does not affect Peter's recent changes.

Reviewed by: Peter Wemm, who warned me I would get some flack, and
he had the good idea for the NO_MODULES variable.

24 years agoAdd in include of <machine/clock.h> which defines DELAY.
Matt Jacob [Tue, 2 May 2000 01:37:02 +0000 (01:37 +0000)]
Add in include of <machine/clock.h> which defines DELAY.

24 years agoFix broken multi-link fragment reassembly algorithm.
Archie Cobbs [Tue, 2 May 2000 00:09:18 +0000 (00:09 +0000)]
Fix broken multi-link fragment reassembly algorithm.
Add hook for IPv6. Misc cleanups.

PR: kern/16335

24 years agoMacro call to ng_send_dataq() should have been to ng_send_data() instead.
Archie Cobbs [Mon, 1 May 2000 23:34:15 +0000 (23:34 +0000)]
Macro call to ng_send_dataq() should have been to ng_send_data() instead.

24 years agoField structure changed name from 'typename' -> 'type_name'.
Archie Cobbs [Mon, 1 May 2000 23:30:46 +0000 (23:30 +0000)]
Field structure changed name from 'typename' -> 'type_name'.

24 years agoUse 'type_name' structure field instead of 'typename', which is
Archie Cobbs [Mon, 1 May 2000 23:29:19 +0000 (23:29 +0000)]
Use 'type_name' structure field instead of 'typename', which is
a C++ reserved work.

Add a ng_copy_meta() function.

24 years agoInclude udbp.4 in the build
Nick Hibma [Mon, 1 May 2000 22:58:18 +0000 (22:58 +0000)]
Include udbp.4 in the build

24 years agoThe USB double bulk pipe driver (Host to host cables). Currently there
Nick Hibma [Mon, 1 May 2000 22:48:23 +0000 (22:48 +0000)]
The USB double bulk pipe driver (Host to host cables). Currently there
are two supported chips, the NetChip 1080 (only prototypes available)
and the EzLink cable. Any other cable should be supported however as they
are all very much alike (there is a difference between them wrt
performance).

It uses Netgraph.

This driver was mostly written by Doug Ambrisko and Julian Elischer and
I would like to thank Whistle for yet another contribution. And my
aplogies to them for me sitting on the driver for so long (2 months).

Also, many thanks to Reid Augustin from NetChip for providing me with a
prototype of their 1080 chip.

Be aware of the fact that this driver is very immature and has only been
tested very lightly. If someone feels like learning about Netgraph however
this is an excellent driver to start playing with.

24 years agoRemove some trailing zeroes.
Nick Hibma [Mon, 1 May 2000 22:09:15 +0000 (22:09 +0000)]
Remove some trailing zeroes.

24 years agoA temporary band-aid for ng_base. It works for some people, a better
Peter Wemm [Mon, 1 May 2000 21:26:50 +0000 (21:26 +0000)]
A temporary band-aid for ng_base.  It works for some people, a better
fix will follow.

Submitted by:   Gary Jennejohn <garyj@muc.de>

24 years agoAdd $FreeBSD$
Peter Wemm [Mon, 1 May 2000 20:32:07 +0000 (20:32 +0000)]
Add $FreeBSD$

24 years agoRegen.
Nick Hibma [Mon, 1 May 2000 20:31:01 +0000 (20:31 +0000)]
Regen.

24 years agoAdd the EzLink adapter Id
Nick Hibma [Mon, 1 May 2000 20:26:12 +0000 (20:26 +0000)]
Add the EzLink adapter Id

24 years agoRemove unused include.
Jeroen Ruigrok van der Werven [Mon, 1 May 2000 20:20:05 +0000 (20:20 +0000)]
Remove unused include.

24 years agoRemove unused include, and place sys includes at top, which enabled
Jeroen Ruigrok van der Werven [Mon, 1 May 2000 20:19:44 +0000 (20:19 +0000)]
Remove unused include, and place sys includes at top, which enabled
us to remove this include.

24 years agoRemove unused includes.
Jeroen Ruigrok van der Werven [Mon, 1 May 2000 20:01:16 +0000 (20:01 +0000)]
Remove unused includes.

24 years agoAdd missing $FreeBSD$
Peter Wemm [Mon, 1 May 2000 19:54:26 +0000 (19:54 +0000)]
Add missing $FreeBSD$

24 years agoRemove unused #include.
Jeroen Ruigrok van der Werven [Mon, 1 May 2000 19:39:36 +0000 (19:39 +0000)]
Remove unused #include.

24 years agoMove and staticize the bufchain functions so they become local to the
Poul-Henning Kamp [Mon, 1 May 2000 19:38:51 +0000 (19:38 +0000)]
Move and staticize the bufchain functions so they become local to the
only piece of code using them.  This will ease a rewrite of them.

24 years agoZap KMODDEPS line
Peter Wemm [Mon, 1 May 2000 18:58:48 +0000 (18:58 +0000)]
Zap KMODDEPS line

24 years agoRemove unused #include.
Jeroen Ruigrok van der Werven [Mon, 1 May 2000 18:52:42 +0000 (18:52 +0000)]
Remove unused #include.

24 years agoRemove unused #include.
Jeroen Ruigrok van der Werven [Mon, 1 May 2000 18:34:36 +0000 (18:34 +0000)]
Remove unused #include.

24 years agoDocument TAILQ_FOREACH_REVERSE() and CIRCLEQ_FOREACH_REVERSE() macros.
Archie Cobbs [Mon, 1 May 2000 18:22:59 +0000 (18:22 +0000)]
Document TAILQ_FOREACH_REVERSE() and CIRCLEQ_FOREACH_REVERSE() macros.

Submitted by: Jake Burkholder <jburkhol@home.com>

24 years agoAdd TAILQ_FOREACH_REVERSE() macro.
Archie Cobbs [Mon, 1 May 2000 18:17:14 +0000 (18:17 +0000)]
Add TAILQ_FOREACH_REVERSE() macro.

Submitted by: Jake Burkholder <jburkhol@home.com>

24 years agoForced commit: Rev 1.80 also changes the gensetdefs point of execution
Peter Wemm [Mon, 1 May 2000 18:08:50 +0000 (18:08 +0000)]
Forced commit: Rev 1.80 also changes the gensetdefs point of execution
so that it gets run after the .kld file generation.  If it's run before,
the linker sets are closed off and bound inside the .kld file which makes
it useless for binding into a static kernel as the linker_set's will not
be connected...

24 years agoRemove KMODDEPS support. All our existing dependencies are specified
Peter Wemm [Mon, 1 May 2000 18:06:49 +0000 (18:06 +0000)]
Remove KMODDEPS support.  All our existing dependencies are specified
via the MODULE_VERSION() and MODULE_DEPEND() macros that both the loader
and kld system know how to deal with.  The old DT_NEEDED tag is still
supported by the loader (and will remain supported for a while) - but the
kernel side presently doesn't know how to deal with DT_NEEDED.

24 years agoGC unused variable (size)
Peter Wemm [Mon, 1 May 2000 18:03:15 +0000 (18:03 +0000)]
GC unused variable (size)

24 years agoFix some warnings on the Alpha. sizeof() returns different things on
Peter Wemm [Mon, 1 May 2000 18:02:38 +0000 (18:02 +0000)]
Fix some warnings on the Alpha.  sizeof() returns different things on
x86 and Alpha. Sigh.

24 years agoWrap the failure warnings around PXE_DEBUG in the cleanup routine.
Paul Saab [Mon, 1 May 2000 17:55:05 +0000 (17:55 +0000)]
Wrap the failure warnings around PXE_DEBUG in the cleanup routine.
It does not matter if they fail, so dont print anything about it
unless we are debugging.

24 years agoUpdate loader logic to distinguish modules vs. files.
Boris Popov [Mon, 1 May 2000 17:41:25 +0000 (17:41 +0000)]
Update loader logic to distinguish modules vs. files.
Add support for module metadata. The old way of dependancy
handling will be supported for a while.

Reviewed by: peter

24 years agoAdd readdir support to the NFS filesystem in libstand.
Paul Saab [Mon, 1 May 2000 15:03:52 +0000 (15:03 +0000)]
Add readdir support to the NFS filesystem in libstand.

24 years ago. clear `.Os' macro value since this tool is not KAME only anymore
Alexey Zelkin [Mon, 1 May 2000 14:57:04 +0000 (14:57 +0000)]
. clear `.Os' macro value since this tool is not KAME only anymore
. add integration note

24 years ago. clear `.Os' macro value since this tool is not KAME only anymore
Alexey Zelkin [Mon, 1 May 2000 14:48:45 +0000 (14:48 +0000)]
. clear `.Os' macro value since this tool is not KAME only anymore
. add integration note
. avoid `.Nm' value upper case usage

24 years ago. remove some unneeded comments
Alexey Zelkin [Mon, 1 May 2000 14:42:09 +0000 (14:42 +0000)]
. remove some unneeded comments
. replace .Os value with empty value since this library is not KAME only
anymore
. add a note about IPv6 and IPsec integration to the FreeBSD

24 years ago. add a note about IPv6/IPsec integration to the FreeBSD
Alexey Zelkin [Mon, 1 May 2000 14:40:06 +0000 (14:40 +0000)]
. add a note about IPv6/IPsec integration to the FreeBSD
. replace .Po/.Pc pairs with .Pq
. remove some unneeded comments
. .Lb-ify

24 years agoAdd libipsec to the list of known libraries for .Lb macro
Alexey Zelkin [Mon, 1 May 2000 14:32:43 +0000 (14:32 +0000)]
Add libipsec to the list of known libraries for .Lb macro

24 years agoMake kame.4 manpage more FreeBSD specific. Replace some KAME kit directory and
Alexey Zelkin [Mon, 1 May 2000 14:09:23 +0000 (14:09 +0000)]
Make kame.4 manpage more FreeBSD specific. Replace some KAME kit directory and
file names with its FreeBSD equivalents.

Remove references to some debuging tools which would never appear in FreeBSD.

Use mdoc(7) macros in proper places.

Give a credit to Youshinobu Inoue for his efforts on KAME kit integration to
the FreeBSD main source tree.

24 years agoConvert to struct bio instead of struct buf.
Poul-Henning Kamp [Mon, 1 May 2000 13:36:43 +0000 (13:36 +0000)]
Convert to struct bio instead of struct buf.

24 years agoGive struct bio it's own call back mechanism.
Poul-Henning Kamp [Mon, 1 May 2000 13:36:25 +0000 (13:36 +0000)]
Give struct bio it's own call back mechanism.

24 years agoMove the MSG* and SEM* options to opt_sysvipc.h
Peter Wemm [Mon, 1 May 2000 13:33:56 +0000 (13:33 +0000)]
Move the MSG* and SEM* options to opt_sysvipc.h
Remove evil allocation macros from machdep.c (why was that there???) and
use malloc() instead.
Move paramters out of param.h and into the code itself.
Move a bunch of internal definitions from public sys/*.h headers (without
#ifdef _KERNEL even) into the code itself.

I had hoped to make some of this more dynamic, but the cost of doing
wakeups on all sleeping processes on old arrays was too frightening.
The other possibility is to initialize on the first use, and allow
dynamic sysctl changes to parameters right until that point. That would
allow /etc/rc.sysctl to change SEM* and MSG* defaults as we presently
do with SHM*, but without the nightmare of changing a running system.

24 years agoPrint options only supported by device, back out printing options separately
Andrey A. Chernov [Mon, 1 May 2000 12:14:30 +0000 (12:14 +0000)]
Print options only supported by device, back out printing options separately

Pointed-by: bde
24 years agoRegenerate (removed semconfig)
Peter Wemm [Mon, 1 May 2000 11:14:08 +0000 (11:14 +0000)]
Regenerate (removed semconfig)

24 years agoRemove the undocumented, flawed, broken-as-designed semconfig() syscall.
Peter Wemm [Mon, 1 May 2000 11:13:41 +0000 (11:13 +0000)]
Remove the undocumented, flawed, broken-as-designed semconfig() syscall.

24 years agoRemove undocumented broken-as-designed semconfig() syscall.
Peter Wemm [Mon, 1 May 2000 11:11:44 +0000 (11:11 +0000)]
Remove undocumented broken-as-designed semconfig() syscall.

24 years agoDo not attempt to free a nfs node if it is the root node. The root
Paul Saab [Mon, 1 May 2000 10:53:21 +0000 (10:53 +0000)]
Do not attempt to free a nfs node if it is the root node.  The root
node is statically allocated and is not guarded, so free will panic
in nfs_close.

24 years agoBack out t_timeout initializing, now in ttyregister
Andrey A. Chernov [Mon, 1 May 2000 10:53:12 +0000 (10:53 +0000)]
Back out t_timeout initializing, now in ttyregister

24 years agoMove t_timeout initializing to ttyregister
Andrey A. Chernov [Mon, 1 May 2000 10:51:54 +0000 (10:51 +0000)]
Move t_timeout initializing to ttyregister

Pointed-by: bde
24 years agoRemove the undocumented semconfig() system calls. These cause more trouble
Peter Wemm [Mon, 1 May 2000 10:49:41 +0000 (10:49 +0000)]
Remove the undocumented semconfig() system calls. These cause more trouble
than they are worth.

24 years ago* Move the driver_t::refs field to kobj_t to replace kobj_t::instances.
Doug Rabson [Mon, 1 May 2000 10:45:15 +0000 (10:45 +0000)]
* Move the driver_t::refs field to kobj_t to replace kobj_t::instances.
* Back out a couple of workarounds for the confusion between
  kobj_t::instances and driver_t::refs.

24 years agoFix a couple of easy cases where we make two identical devices.. one
Peter Wemm [Mon, 1 May 2000 10:43:06 +0000 (10:43 +0000)]
Fix a couple of easy cases where we make two identical devices.. one
/dev/xxx and one /dev/rxxx.  This changes them to a hard link so that
less inodes are consumed and so that the permissions are always in sync.
There are lots more of these still.

24 years agoSince ptys are allocated dynamically, there is no needs to keep their
Andrey A. Chernov [Mon, 1 May 2000 10:24:21 +0000 (10:24 +0000)]
Since ptys are allocated dynamically, there is no needs to keep their
t_timeout across close, so move t_timeout initializing to ptcopen

24 years agoInitialize t_timeout to -1 for console to set its default value once in ttyopen
Andrey A. Chernov [Mon, 1 May 2000 09:21:08 +0000 (09:21 +0000)]
Initialize t_timeout to -1 for console to set its default value once in ttyopen

24 years agoSplit out the ISA bus front end code into its own file. PCCARD attachment
Matthew N. Dodd [Mon, 1 May 2000 09:05:19 +0000 (09:05 +0000)]
Split out the ISA bus front end code into its own file.  PCCARD attachment
coming later this week.  Mitsuru IWASAKI provided a patch to -mobile which
I used to make sure I was doing the right thing but only a small part of
the actual patch was used.

24 years agoSet t_timeout to its default sysctl value only once in ttyopen
Andrey A. Chernov [Mon, 1 May 2000 09:05:03 +0000 (09:05 +0000)]
Set t_timeout to its default sysctl value only once in ttyopen
Initialize t_timeout to -1 for this reason

Pointed-by: bde
24 years agoFix for the HP burners (and possibly other broken drives to)
Søren Schmidt [Mon, 1 May 2000 07:30:28 +0000 (07:30 +0000)]
Fix for the HP burners (and possibly other broken drives to)
that fails to proberly close the disk.
The problem seems to be that the HP burners sometimes return
ready when they actually are not, the solution is to not use
immediate mode on the closing commands. This is suboptimal
for real burners, in that they now hog the ATA bus for possibly
minutes, where its really not nessesary, *sigh*.

24 years agoNow that I've got reports that this works, let's turn the debugging off.
Warner Losh [Mon, 1 May 2000 04:41:04 +0000 (04:41 +0000)]
Now that I've got reports that this works, let's turn the debugging off.
MIHIRA-san pointed out that debugging is rather big for tcpdump...

Submitted by: sanpei@sanpei.org (MIHIRA Yoshiro)

24 years ago- Merge sl_probe() and ni_probe().
Matthew N. Dodd [Mon, 1 May 2000 03:40:21 +0000 (03:40 +0000)]
- Merge sl_probe() and ni_probe().
- Add myself as 'MAINTAINER' since I've got a number
  of changes in the pipe.

24 years agoAdjust to accomodate recent changes to the rcvdata and rcvmsg
Archie Cobbs [Mon, 1 May 2000 03:31:58 +0000 (03:31 +0000)]
Adjust to accomodate recent changes to the rcvdata and rcvmsg
function prototypes.

24 years agoReserve major device number for nsmb (SMB/CIFS protocol) device.
Boris Popov [Mon, 1 May 2000 00:49:19 +0000 (00:49 +0000)]
Reserve major device number for nsmb (SMB/CIFS protocol) device.

24 years agoOnly set $MAIL_AGENT if it isn't already defined in the environment.
Nik Clayton [Sun, 30 Apr 2000 22:48:49 +0000 (22:48 +0000)]
Only set $MAIL_AGENT if it isn't already defined in the environment.
No idea why this was sent in as a docs bug. . .

PR:             docs/17014
Submitted by:   Giorgos Keramidas <keramida@ceid.upatras.gr>

24 years agoRemove reference to bsd2dos command, refer to fconv and similar.
Nik Clayton [Sun, 30 Apr 2000 22:45:13 +0000 (22:45 +0000)]
Remove reference to bsd2dos command, refer to fconv and similar.

PR:             docs/17101
Submitted by:   Udo Erdelhoff <ue@nathan.ruhr.de>

24 years agoAdd descriptions of the nm(1) keywords and their meanings.
Nik Clayton [Sun, 30 Apr 2000 22:44:05 +0000 (22:44 +0000)]
Add descriptions of the nm(1) keywords and their meanings.

PR:             docs/17269
Submitted by:   Oscar Bonilla <obonilla@fisicc-ufm.edu>

24 years agoAdd xref to cap_mkdb(1).
Nik Clayton [Sun, 30 Apr 2000 22:43:09 +0000 (22:43 +0000)]
Add xref to cap_mkdb(1).

PR:             docs/17544
Submitted by:   Christ J. Clark <cjc@cc942873-a.ewndsr1.nj.home.com>

24 years agoNote that X11 Forwarding is off by default.
Nik Clayton [Sun, 30 Apr 2000 22:41:58 +0000 (22:41 +0000)]
Note that X11 Forwarding is off by default.

PR:             docs/17566
Submitted by:   Keith Stevenson <ktstev01@louisville.edu>

24 years agoAdd another example showing how you can use pppctl to only allow
Nik Clayton [Sun, 30 Apr 2000 22:28:21 +0000 (22:28 +0000)]
Add another example showing how you can use pppctl to only allow
dial out at certain times of the day.

Approved by:    brian

24 years agoRemoved xref to mount_lfs, this died in 2.mumble.
Nik Clayton [Sun, 30 Apr 2000 22:08:13 +0000 (22:08 +0000)]
Removed xref to mount_lfs, this died in 2.mumble.

PR:             docs/18272
Submitted by:   Michael Lucas <mwlucas@blackhelicopters.org>

24 years agoPhase mismatch handling from SCRIPTS had been
Gerard Roudier [Sun, 30 Apr 2000 21:42:55 +0000 (21:42 +0000)]
Phase mismatch handling from SCRIPTS had been
broken by previous patch.

24 years agoFixes a potential buffer overflow with the pid filename.
Josef Karthauser [Sun, 30 Apr 2000 21:04:36 +0000 (21:04 +0000)]
Fixes a potential buffer overflow with the pid filename.

Submitted by: Mike Heffner <spock@techfour.net>
Submitted on: audit@freebsd.org

24 years agoFixes a potential buffer overflow with the command line arguments.
Josef Karthauser [Sun, 30 Apr 2000 20:53:54 +0000 (20:53 +0000)]
Fixes a potential buffer overflow with the command line arguments.

Submitted by:   Mike Heffner <spock@techfour.net>
Submitted on:   audit@freebsd.org

24 years agoFixes a potential buffer overflow with 'ed [MAXPATHLEN + 1 characters]'.
Josef Karthauser [Sun, 30 Apr 2000 20:46:14 +0000 (20:46 +0000)]
Fixes a potential buffer overflow with 'ed [MAXPATHLEN + 1 characters]'.

Submitted by: Mike Heffner <spock@techfour.net>
Submitted on: audit@freebsd.org

24 years agoRemove unneeded #include <vm/vm_zone.h>
Poul-Henning Kamp [Sun, 30 Apr 2000 18:52:11 +0000 (18:52 +0000)]
Remove unneeded #include <vm/vm_zone.h>

Generated by: src/tools/tools/kerninclude

24 years agoFor the M_PREPEND macro, remove an initial mbuf NULL check I added. It
Brian Feldman [Sun, 30 Apr 2000 18:42:44 +0000 (18:42 +0000)]
For the M_PREPEND macro, remove an initial mbuf NULL check I added.  It
was added accidentally, and although not terrible, it would improperly
hide the bug of calling M_PREPEND with a NULL mbuf argument.

Submitted by: jlemon (ISTR)

24 years agoChange the scheduler to actually respect the PUSER barrier. It's been
Brian Feldman [Sun, 30 Apr 2000 18:33:43 +0000 (18:33 +0000)]
Change the scheduler to actually respect the PUSER barrier.  It's been
wrong for many years that negative niceness would lower the priority
of a process below PUSER, and once below PUSER, there were conditionals
in the code that are required to test for whether a process was in
the kernel which would break.

The breakage could (and did) cause lock-ups, basically nothing else
but the least nice program being able to run in some conditions.  The
algorithm which adjusts the priority now subtracts PRIO_MIN to do
things properly, and the ESTCPULIM() algorithm was updated to use
PRIO_TOTAL (PRIO_MAX - PRIO_MIN) to calculate the estcpu.

NICE_WEIGHT is now 1 to accomodate the full range of priorities better
(a -20 process with full CPU time has the priority of a +0 process with
no CPU time).  There are now 20 queues (exactly; 80 priorities) for
use in user processes' scheduling, and PUSER has been lowered to 48
to accomplish this.

This means, to the user, that things will be scheduled more correctly
(noticeable), there is no lock-up anymore WRT a niced -20 process
never releasing the CPU time for other processes.  In this fair system,
tsleep()ed < PUSER processes now will get the proper higher priority
than priority >= PUSER user processes.

The detective work of this was done by me, along with part of the
solution.  Luoqi Chen has provided most of the solution, and really
helped me understand what was happening better, to boot :)

Submitted by:   luoqi
Concept reviewed by:    bde

24 years agoAllow "-" for working with STDIN
Andrey A. Chernov [Sun, 30 Apr 2000 18:06:04 +0000 (18:06 +0000)]
Allow "-" for working with STDIN
Allow printing of each option separately when keyword specified without a
number

24 years agoFix warn format
Andrey A. Chernov [Sun, 30 Apr 2000 17:12:49 +0000 (17:12 +0000)]
Fix warn format

Pointed-by: bde
24 years agoRemove bogus include, as per style(9).
Jeroen Ruigrok van der Werven [Sun, 30 Apr 2000 17:05:26 +0000 (17:05 +0000)]
Remove bogus include, as per style(9).
Remove unused variable.

24 years agoBack out all drainwait changes. It is enough controllable via sysctl or
Andrey A. Chernov [Sun, 30 Apr 2000 17:04:26 +0000 (17:04 +0000)]
Back out all drainwait changes. It is enough controllable via sysctl or
comcontrol, having it in stty cause too many problems with existing drivers
and tty access permissings of non-superuser.

Asked-by: bde
24 years agogfmt: set drainwait only if changed
Andrey A. Chernov [Sun, 30 Apr 2000 16:22:09 +0000 (16:22 +0000)]
gfmt: set drainwait only if changed
It allows to restore tty state without a warning for non-superuser

24 years agoAdd sysctl variable to set initial drainwait timeout on ttyopen, default to
Andrey A. Chernov [Sun, 30 Apr 2000 16:00:53 +0000 (16:00 +0000)]
Add sysctl variable to set initial drainwait timeout on ttyopen, default to
5 minutes

24 years agoAllocate space for arrays of type "char *", not "char **".
Guy Helmer [Sun, 30 Apr 2000 15:57:00 +0000 (15:57 +0000)]
Allocate space for arrays of type "char *", not "char **".
Rev 1.8 made the type consistently incorrect.

Noted by: Ben Smithurst <ben@scientia.demon.co.uk>

24 years agoFix comments, whitespace to reduce diffs between this and GENERIC.
Mark Murray [Sun, 30 Apr 2000 15:38:28 +0000 (15:38 +0000)]
Fix comments, whitespace to reduce diffs between this and GENERIC.

24 years agoFixed the type of some ivar access functions. Ivars have type uintptr_t,
Bruce Evans [Sun, 30 Apr 2000 10:01:56 +0000 (10:01 +0000)]
Fixed the type of some ivar access functions.  Ivars have type uintptr_t,
not u_long.  On i386's with 64-bit longs, returning u_longs indirectly
in (more than) the space reserved for uintptr_t's tended to corrupt the
previous frame pointer in the stack frame, so it was not easy to debug.
The type mismatches are hidden by the bogus cast in DEVMETHOD().

24 years agoUpdate the man page to reflect the recent changes to the kernel API for
Julian Elischer [Sun, 30 Apr 2000 10:01:11 +0000 (10:01 +0000)]
Update the man page to reflect the recent changes to the kernel API for
netgraph.

24 years agoInclude <sys/random.h> for rand_initialize().
Doug Rabson [Sun, 30 Apr 2000 09:41:35 +0000 (09:41 +0000)]
Include <sys/random.h> for rand_initialize().

24 years ago- Added UKBD_DFLT_KEYMAP option.
Yoshihiro Takahashi [Sun, 30 Apr 2000 09:12:48 +0000 (09:12 +0000)]
- Added UKBD_DFLT_KEYMAP option.
  Pointed out: Tomokazu HARADA <tkhara@osk4.3web.ne.jp>

- Fixed disordering.

24 years agoClean up MAXMEM routine.
Yoshihiro Takahashi [Sun, 30 Apr 2000 08:52:47 +0000 (08:52 +0000)]
Clean up MAXMEM routine.

Submitted by: "K.Magara" <magara@maizuru-ct.ac.jp>

24 years agoFixed to support JIS7 KANJI.
Yoshihiro Takahashi [Sun, 30 Apr 2000 08:40:43 +0000 (08:40 +0000)]
Fixed to support JIS7 KANJI.

Submitted by: Nobuyuki Koganemaru <kogane@koganemaru.co.jp>