]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
26 years agoSync with sys/i386/isa/clock.c revision 1.121.
KATO Takenori [Thu, 28 May 1998 13:51:39 +0000 (13:51 +0000)]
Sync with sys/i386/isa/clock.c revision 1.121.

26 years agoSync with sys/i386/i386/machdep.c revision 1.297.
KATO Takenori [Thu, 28 May 1998 13:50:13 +0000 (13:50 +0000)]
Sync with sys/i386/i386/machdep.c revision 1.297.

26 years agoSync with sys/i386/boot/kzipboot/Makefile revision 1.9.
KATO Takenori [Thu, 28 May 1998 13:48:40 +0000 (13:48 +0000)]
Sync with sys/i386/boot/kzipboot/Makefile revision 1.9.

26 years agoSync with sys/i386/boot/biosboot/Makefile revision 1.64.
KATO Takenori [Thu, 28 May 1998 13:46:04 +0000 (13:46 +0000)]
Sync with sys/i386/boot/biosboot/Makefile revision 1.64.

26 years agoOnly pre-build the "default" libcrypt, since it gets to set the symlinks
Peter Wemm [Thu, 28 May 1998 13:20:35 +0000 (13:20 +0000)]
Only pre-build the "default" libcrypt, since it gets to set the symlinks
for the rest of the build.  I'm not certain, but I think this determines
which crypt() goes into /sbin/init.  This change shouldn't hurt anyway. :-)

Based on a suggestion by: bde

26 years agoTurn on the rcsid storage so that it's easier to tell the difference
Peter Wemm [Thu, 28 May 1998 12:52:03 +0000 (12:52 +0000)]
Turn on the rcsid storage so that it's easier to tell the difference
between libscrypt.a and libdescrypt.a

26 years agoUrgent fix for bootstrap target. If the old symlinks are pointing to the
Peter Wemm [Thu, 28 May 1998 12:31:42 +0000 (12:31 +0000)]
Urgent fix for bootstrap target.  If the old symlinks are pointing to the
real source tree, the bootstrap target would attempt have cpio copy the
files over themselves, unlinking them first.  I think this only happened
with make -DNOCLEAN world at the transition between a symlinked
objdir/tmp/usr/include/{sys,net,..} and real files.

26 years agoSome cleanups related to timecounters and weird ifdefs in <sys/time.h>.
Poul-Henning Kamp [Thu, 28 May 1998 09:30:28 +0000 (09:30 +0000)]
Some cleanups related to timecounters and weird ifdefs in <sys/time.h>.

Clean up (or if antipodic: down) some of the msgbuf stuff.

Use an inline function rather than a macro for timecounter delta.

Maintain process "on-cpu" time as 64 bits of microseconds to avoid
needless second rollover overhead.

Avoid calling microuptime the second time in mi_switch() if we do
not pass through _idle in cpu_switch()

This should reduce our context-switch overhead a bit, in particular
on pre-P5 and SMP systems.

WARNING:  Programs which muck about with struct proc in userland
will have to be fixed.

Reviewed, but found imperfect by:       bde

26 years agoUse AF_LINK rather than AF_UNSPEC to set an ethernet multicast address.
Julian Elischer [Thu, 28 May 1998 07:31:30 +0000 (07:31 +0000)]
Use AF_LINK rather than AF_UNSPEC to set an ethernet multicast address.
This is obviously  not a terribly used function as it's apparently been
broken forever.
It IS possible that this fix is wrong and that the KERNEL is wrong
(in which case you should fix if_ethersubr.c) either way it certainly has more hope of
working now than before. I'd take it to 2.2 except that obviously no-one cares :-)

26 years agoGrumble Grumble.. so this is how src/sys/ufs/lfs kept coming back... The
Peter Wemm [Thu, 28 May 1998 06:59:50 +0000 (06:59 +0000)]
Grumble Grumble.. so this is how src/sys/ufs/lfs kept coming back...  The
mtree in make world was following symlinks and creating directories in the
source tree and causing cvs to break (since CVS/Entries was missing).

26 years agoReviewed by: Peter Hawkins <thepish@FreeBSD.org>
Daniel O'Callaghan [Thu, 28 May 1998 00:58:29 +0000 (00:58 +0000)]
Reviewed by: Peter Hawkins <thepish@FreeBSD.org>
Add s and w flags to show duration in or with seconds.

26 years agoDon't assume stdout is a tty in interactive mode
Brian Somers [Wed, 27 May 1998 22:43:37 +0000 (22:43 +0000)]
Don't assume stdout is a tty in interactive mode
Analyzed by: dmaddox@scsn.net (Donald J. Maddox)

26 years agoCorrectly display the interrupt counts.
John Hay [Wed, 27 May 1998 21:01:37 +0000 (21:01 +0000)]
Correctly display the interrupt counts.
Reviewed with optimizations by: Tor Egge <tegge>

26 years agoReplace the partial symlink tree to src/sys/* in $OBJDIR/tmp/usr/include
Peter Wemm [Wed, 27 May 1998 18:50:01 +0000 (18:50 +0000)]
Replace the partial symlink tree to src/sys/* in $OBJDIR/tmp/usr/include
with real copies.  I'm sick of !@#&!^!@#*& mtree chowning directories in
my src/sys/* tree after it follows the symlinks.  I still believe that
mtree is broken for doing this (introduced in mtree.c rev 1.5).

26 years agoIf building and installing libdescrypt, do it first so that the symlinks
Peter Wemm [Wed, 27 May 1998 18:42:36 +0000 (18:42 +0000)]
If building and installing libdescrypt, do it first so that the symlinks
point to it rather than libscrypt.

This was how it was done prior to libscrypt being added in.  This should
stop more people getting burnt with the /usr/lib -> /usr/lib/aout
transition, and the same when the ELF libs come online.

26 years agoIf we are logging into a cvs pserver with the username of "anoncvs",
Peter Wemm [Wed, 27 May 1998 16:55:49 +0000 (16:55 +0000)]
If we are logging into a cvs pserver with the username of "anoncvs",
and we have not done an explicit 'cvs login', then use a default password
of "anoncvs".  This allows things like:
  setenv CVSROOT :pserver:anoncvs@anoncvs.freebsd.org:/cvs
  cvs checkout src  (without doing the normal 'cvs login' for pserver mode)
but this runs over the :pserver: protocol rather than the more troublesome
rsh.  Naturally, the server had better be running in -R (readonly) mode :-)

26 years agoOn the other hand, when /usr/bin/as is really a link to objformat, it's not
Peter Wemm [Wed, 27 May 1998 16:33:43 +0000 (16:33 +0000)]
On the other hand, when /usr/bin/as is really a link to objformat, it's not
helpful to stop it running /usr/libexec/aout/as for real while bootstrapping.
Only force a strict path when we really have built all the tools in
$OBJDIR/tmp/usr/libexec/*.

26 years agoMake cvs really ignore a pserver password when the passwd field of the
Peter Wemm [Wed, 27 May 1998 16:27:18 +0000 (16:27 +0000)]
Make cvs really ignore a pserver password when the passwd field of the
CVSROOT/passwd file is empty.  A 'cvs login' still seems to be required
since the cvs client doesn't seem to believe it's possible to not need
a password (yet :-).  This is intended for cheap anoncvs use.

26 years agoIt would help if it compiled. *blush*
Peter Wemm [Wed, 27 May 1998 16:00:55 +0000 (16:00 +0000)]
It would help if it compiled. *blush*

26 years agoConvert to more canonical format.. It seems that both Soren and I were
Peter Wemm [Wed, 27 May 1998 15:43:24 +0000 (15:43 +0000)]
Convert to more canonical format..  It seems that both Soren and I were
working on this in parallel.. :-(

26 years agoMove objformat after binutils, otherwise we get chicken/egg problems when
Peter Wemm [Wed, 27 May 1998 15:40:35 +0000 (15:40 +0000)]
Move objformat after binutils, otherwise we get chicken/egg problems when
building ELF.  It may be that binutils and objformat need to both be moved
earlier in the list though, but that can wait till the dust settles.

26 years agoAdd a new long flag that causes cvs to ignore the CVSROOT/passwd file.
Peter Wemm [Wed, 27 May 1998 15:26:12 +0000 (15:26 +0000)]
Add a new long flag that causes cvs to ignore the CVSROOT/passwd file.
This is mostly intended for use on freefall where we'd like to provide
a passwd file for easy anoncvs mirroring access, but don't want to open
up the pserver on freefall itself.

While here, some initial tweaks intended for allowing an empty pserver
password.  I'm not sure that this works yet.

26 years agoRemove my hacks for capturing stdout/stderr through the protocol channel
Peter Wemm [Wed, 27 May 1998 15:19:23 +0000 (15:19 +0000)]
Remove my hacks for capturing stdout/stderr through the protocol channel
while calling libdiff.  It's too ugly and not worth the recursion problems
when there is a malloc failure (which writes to stderr - now diverted via
the buf system, which calls malloc, which causes another error message etc).

We can live with the standard artificial slowdown, but reduce the time a
bit and only delay when we really need to (ie: when running as a server).
The usleep time could probably use some tuning, it basically needs to
replace the time that it used to take to fork a large process, exec gnudiff
and the time that gnudiff took before writing the initial output.

This eliminates a whole mess of other hacks I was considering that changed
use of xmalloc to alloca() etc.  It was going too fast in the wrong
direction.

26 years agoTypo police. '#ifndef lint' around CVS id. Don't use __progname in usage().
Dag-Erling Smørgrav [Wed, 27 May 1998 14:04:38 +0000 (14:04 +0000)]
Typo police. '#ifndef lint' around CVS id. Don't use __progname in usage().

Submitted by: Philippe Charnier <charnier@xp11.frmug.org>

26 years agoInstall ${DESTDIR}/boot.help during `make distribution'.
Joseph Koshy [Wed, 27 May 1998 12:13:18 +0000 (12:13 +0000)]
Install ${DESTDIR}/boot.help during `make distribution'.

PR: 6063
Submitted by: Doug <Studded@dal.net>

26 years agoCorrect manual page, based on feedback from Bruce Evans.
Joseph Koshy [Wed, 27 May 1998 11:05:59 +0000 (11:05 +0000)]
Correct manual page, based on feedback from Bruce Evans.

Reviewed by: Bruce Evans <bde@freebsd.org>

26 years agoFix some bitrot in the multicast support, the move from TAILQ macros wasn't implement...
Paul Richards [Wed, 27 May 1998 11:05:17 +0000 (11:05 +0000)]
Fix some bitrot in the multicast support, the move from TAILQ macros wasn't implemented fully.

26 years agoThis time add an Id string that'll actually get expanded :-)
Paul Richards [Wed, 27 May 1998 09:59:13 +0000 (09:59 +0000)]
This time add an Id string that'll actually get expanded :-)

26 years agoAdd an ID string (can't believe it never had one!)
Paul Richards [Wed, 27 May 1998 09:58:00 +0000 (09:58 +0000)]
Add an ID string (can't believe it never had one!)

26 years ago"make world" replaces /bin/rmail even when NO_SENDMAIL has been set to
Poul-Henning Kamp [Wed, 27 May 1998 09:25:22 +0000 (09:25 +0000)]
"make world" replaces /bin/rmail even when NO_SENDMAIL has been set to
true in /etc/make.conf.  Both qmail and smail use a different rmail, so
replacing rmail is a Bad Thing.

PR: 6762
Reviewed by: phk
Submitted by: Bill Trost <trost@cloud.rain.com>

26 years agoInstall in /usr/lib/aout
Søren Schmidt [Wed, 27 May 1998 08:06:33 +0000 (08:06 +0000)]
Install in /usr/lib/aout

26 years agoCompile & use objformat as it was intended to, I have no idea why
Søren Schmidt [Wed, 27 May 1998 07:59:37 +0000 (07:59 +0000)]
Compile & use objformat as it was intended to, I have no idea why
that failed for me before :(

26 years agoDon't use absolute path.
Jordan K. Hubbard [Wed, 27 May 1998 07:39:05 +0000 (07:39 +0000)]
Don't use absolute path.
Submitted by: jhay

26 years agoA fix to a debug test from Kirk.
Julian Elischer [Wed, 27 May 1998 03:32:23 +0000 (03:32 +0000)]
A fix to a debug test from Kirk.

26 years ago accidently omitted two file declarations.
Jonathan M. Bresler [Wed, 27 May 1998 01:26:33 +0000 (01:26 +0000)]
accidently omitted two file declarations.
so i added them.

Submitted by: Scot Elliott (scot@poptart.org)

26 years agoMake a copy of the caller's iovec array, mallocing if necessary,
John Birrell [Wed, 27 May 1998 00:44:58 +0000 (00:44 +0000)]
Make a copy of the caller's iovec array, mallocing if necessary,
and modify that if the writev() syscall does not completely write
all bytes in a single call.

26 years agoRemove error check from call to set the file descriptor to non-blocking
John Birrell [Wed, 27 May 1998 00:41:22 +0000 (00:41 +0000)]
Remove error check from call to set the file descriptor to non-blocking
instead of explicitly ignoring some errors. This allows for the case
where a device is naturally non-blocking.

26 years agoIgnore 'invalid' interrupts that occur while the interface is down.
Bill Paul [Tue, 26 May 1998 23:42:24 +0000 (23:42 +0000)]
Ignore 'invalid' interrupts that occur while the interface is down.
These are probably generated by other PCI devices sharing the TLAN's
interrupt. The programmer's guide says to simply re-enable interrupts
and return if one of these is detected.

Prompted by bug report from: Bill Fenner

26 years agoELF preparation step 2:
Søren Schmidt [Tue, 26 May 1998 20:12:56 +0000 (20:12 +0000)]
ELF preparation step 2:

Move a.out libraries to /usr/lib/aout to make space for ELF libs.
Make rtld usr /usr/lib/aout as default library path.
Make ldconfig reject /usr/lib as an a.out library path.
Fix various Makefiles for LIBDIR!=/usr/lib breakage.

This will after a make world & reboot give a system that no
longer uses /usr/lib/*, infact one could remove all the old
libraries there, they are not used anymore.

We are getting close to an ELF make world, but I'll let this
all settle for a week or two...

26 years agoGet owner/group/modes right
Søren Schmidt [Tue, 26 May 1998 12:27:37 +0000 (12:27 +0000)]
Get owner/group/modes right

26 years agoFixed logic in the test to drop ICMP echo and timestamp packets when
David Greenman [Tue, 26 May 1998 11:34:30 +0000 (11:34 +0000)]
Fixed logic in the test to drop ICMP echo and timestamp packets when
net.inet.ip.icmp.bmcastecho = 0 by removing the extra check for the
address being a multicast address. The test now relies on the link
layer flags that indicate it was received via multicast. The previous
logic was broken and replied to ICMP echo/timestamp broadcasts even
when the sysctl option disallowed them.
Reviewed by: wollman

26 years agoOops, remember CFLAGS...
Søren Schmidt [Tue, 26 May 1998 06:51:28 +0000 (06:51 +0000)]
Oops, remember CFLAGS...
Also compile directly to the target on install.

26 years agoPR: 1904
Daniel O'Callaghan [Tue, 26 May 1998 06:39:08 +0000 (06:39 +0000)]
PR: 1904
Abort if given uname is > MAXLOGNAME-1

26 years agoRemove a couple style bugs from the code snippet that
Steve Price [Tue, 26 May 1998 02:53:06 +0000 (02:53 +0000)]
Remove a couple style bugs from the code snippet that
slipped in on the previous commit.

Submitted by: bde

26 years ago code was using wrong temporary variable "i", in place of "j".
Jonathan M. Bresler [Tue, 26 May 1998 02:28:18 +0000 (02:28 +0000)]
code was using wrong temporary variable "i", in place of "j".
i contains the contents of the EP_W0_CONFIG_CTRL register.
i was being used as the array index into an array on the stack.

j is initialized to 0 as it should be.

PR: kern/6757
Reviewed by: jmb
Submitted by: Stephane E. Potvin <sepotvin@videotron.ca>

26 years ago"EOF of file..." is written to stderr and not stdout.
Steve Price [Tue, 26 May 1998 00:39:12 +0000 (00:39 +0000)]
"EOF of file..." is written to stderr and not stdout.

PR: 6756

26 years agoDon't append a '/' to the destination directory if there
Steve Price [Mon, 25 May 1998 22:44:16 +0000 (22:44 +0000)]
Don't append a '/' to the destination directory if there
is already one there.

PR: 3025

26 years agoWhen doing a blocking write, keep looping until all the bytes are
John Birrell [Mon, 25 May 1998 21:45:52 +0000 (21:45 +0000)]
When doing a blocking write, keep looping until all the bytes are
written without returning to the caller. This only occurs on pipes
where either the number of bytes written is greater than the pipe
buffer or if there is insufficient space in the pipe buffer because the
reader is reading slower than the writer is writing.

26 years agoUpdate code example to reflect current practice.
Steve Price [Mon, 25 May 1998 21:42:51 +0000 (21:42 +0000)]
Update code example to reflect current practice.

26 years agoUse select() timeouts instead of SIGALRM to schedule packet transmission.
Bill Fenner [Mon, 25 May 1998 20:16:05 +0000 (20:16 +0000)]
Use select() timeouts instead of SIGALRM to schedule packet transmission.
 Fixes bin/6649 and removes the last abusive signal handler.
Use SO_TIMESTAMP to get the kernel to timestamp packets on reception.
 Fixes bin/5658 and provides slightly better accuracy.
Explicitly zero and terminate the IP options when using -R.

PR: bin/5658
PR: bin/6649

26 years agoNorwegian dates should have a dot after day of month.
Steve Price [Mon, 25 May 1998 18:51:47 +0000 (18:51 +0000)]
Norwegian dates should have a dot after day of month.

PR: 6749
Submitted by: Hans Petter Bieker <zerium@webindex.no>

26 years agoELF preparation step 1:
Søren Schmidt [Mon, 25 May 1998 17:34:42 +0000 (17:34 +0000)]
ELF preparation step 1:

Move our old a.out utils to /usr/libexec/aout.
Enable binutils and put the utils in /usr/libexec/elf
Enable objformat, a little helper program that calls the right
utils based on /etc/objformat and $OBJFORMAT.

This will enable the ELF generating tools.
Remember that this is only step one, the system is still compiled
and run in a.out format ONLY.

Problem left to solve: The BSD manpages wins over the GNU equivalents
as the are installed last. We need to distinguish between the manpages
somehow...

26 years agoOnly compile on install.
Søren Schmidt [Mon, 25 May 1998 15:00:49 +0000 (15:00 +0000)]
Only compile on install.

26 years agoMake objformat DTRT when in the make world environment.
Søren Schmidt [Mon, 25 May 1998 14:42:22 +0000 (14:42 +0000)]
Make objformat DTRT when in the make world environment.

26 years agoBack out Steve's fix, as this was already fixed by me in revision 1.6
Dag-Erling Smørgrav [Mon, 25 May 1998 12:04:47 +0000 (12:04 +0000)]
Back out Steve's fix, as this was already fixed by me in revision 1.6

26 years agoAdd optional code to change the way that divert and ipfw work together.
Julian Elischer [Mon, 25 May 1998 10:37:48 +0000 (10:37 +0000)]
Add optional code to change the way that divert and ipfw work together.
Prior to this change, Accidental recursion protection was done by
the diverted daemon feeding back the divert port number it got
the packet on, as the port number on a sendto(). IPFW knew not to
redivert a packet to this port (again). Processing of the ruleset
started at the beginning again, skipping that divert port.

The new semantic (which is how we should have done it the first time)
is that the port number in the sendto() is the rule number AFTER which
processing should restart, and on a recvfrom(), the port number is the
rule number which caused the diversion. This is much more flexible,
and also more intuitive. If the user uses the same sockaddr received
when resending, processing resumes at the rule number following that
that caused the diversion. The user can however select to resume rule
processing at any rule. (0 is restart at the beginning)

To enable the new code use

option IPFW_DIVERT_RESTART

This should become the default as soon as people have looked at it a bit

26 years agoo Only do the /bin/cat bit when we're transferring our ctty.
Brian Somers [Mon, 25 May 1998 10:37:02 +0000 (10:37 +0000)]
o Only do the /bin/cat bit when we're transferring our ctty.

26 years agoFixed missing semicolon and added space before semicolon.
KATO Takenori [Mon, 25 May 1998 09:57:34 +0000 (09:57 +0000)]
Fixed missing semicolon and added space before semicolon.

Noticed by: Toshikazu Kaho <kaho@elam.kais.kyoto-u.ac.jp>

26 years agoManual page fix: add cross references, refer to ``dumpdev'' variable in
Joseph Koshy [Mon, 25 May 1998 08:54:40 +0000 (08:54 +0000)]
Manual page fix: add cross references, refer to ``dumpdev'' variable in
"/etc/rc.conf".

PR: 6117
Submitted by: Mark Mayo <mark@vmunix.com>

26 years agoHide the interface name in the sin_zero section of the sockaddr_in
Julian Elischer [Mon, 25 May 1998 08:44:31 +0000 (08:44 +0000)]
Hide the interface name in the sin_zero section of the sockaddr_in
passed to the user process for incoming packets. When the sockaddr_in
is passed back to the divert socket later, use thi sas the primary
interface lookup and only revert to the IP address when the name fails.
This solves a long standing bug with divert sockets:
When two interfaces had the same address (P2P for example) the interface
"assigned" to the reinjected packet was sometimes incorect.
Probably we should define a "sockaddr_div" to officially hold this
extended information in teh same manner as sockaddr_dl.

26 years agoTake the user's "IGNORE_DIVERT" argument from where the user put it
Julian Elischer [Mon, 25 May 1998 07:41:23 +0000 (07:41 +0000)]
Take the user's "IGNORE_DIVERT" argument from where the user put it
and not from the PCB which HAPPENS to contain the same number most
of the time, but not always.

26 years agoOnly allow 'who' and 'who am i' as valid usages.
Steve Price [Mon, 25 May 1998 07:19:07 +0000 (07:19 +0000)]
Only allow 'who' and 'who am i' as valid usages.

PR: 6294
Submitted by: Ruslan Ermilov <ru@ucb.crimea.ua>

26 years agoFix example in manual page.
Joseph Koshy [Mon, 25 May 1998 07:11:41 +0000 (07:11 +0000)]
Fix example in manual page.

PR: 5443
Submitted by: AMAGAI Yoshiji <amagai@nue.org>

26 years agoRevert the previous fix. As it turns out Warner Losh is
Steve Price [Mon, 25 May 1998 06:53:17 +0000 (06:53 +0000)]
Revert the previous fix.  As it turns out Warner Losh is
working a better fix.

26 years agoAllow spaces in the filename to be decoded to.
Steve Price [Mon, 25 May 1998 06:27:09 +0000 (06:27 +0000)]
Allow spaces in the filename to be decoded to.

PR: 4882
Submitted by: Slaven Rezic <eserte@cs.tu-berlin.de>

26 years agoGive correct times with old wtmp files.
Steve Price [Mon, 25 May 1998 05:21:29 +0000 (05:21 +0000)]
Give correct times with old wtmp files.

PR: 4732
Submitted by: Andrew Timonin <tim@pool1.convey.ru>

26 years agoIf no value is present for the login name set it to '-'.
Steve Price [Mon, 25 May 1998 05:07:18 +0000 (05:07 +0000)]
If no value is present for the login name set it to '-'.
Also pretty-up the display of 'ps -Ortprio'.

PR: 4947
Submitted by: Martin Kammerhofer <dada@sbox.tu-graz.ac.at>

26 years agoAllow setting of idle or realtime processing priorities per
Steve Price [Mon, 25 May 1998 03:55:23 +0000 (03:55 +0000)]
Allow setting of idle or realtime processing priorities per
login class.

PR: 6636
Submitted by: Jason Young <doogie@forbidden-donut.anet-stl.com>

26 years agoAttempt to stop another DoS attack related to ping flooding.
Steve Price [Mon, 25 May 1998 03:50:51 +0000 (03:50 +0000)]
Attempt to stop another DoS attack related to ping flooding.

PR: 6649
Submitted by: Jason Young <doogie@forbidden-donut.anet-stl.com>

26 years agoMake ftpd(8) honor its default group setting in the config files.
Steve Price [Mon, 25 May 1998 03:45:35 +0000 (03:45 +0000)]
Make ftpd(8) honor its default group setting in the config files.

PR: 6682
Submitted by: Max Euston <meuston@jmrodgers.com>

26 years agoAllow a user in group 0 to su(1) to root if their primary
Steve Price [Mon, 25 May 1998 03:34:52 +0000 (03:34 +0000)]
Allow a user in group 0 to su(1) to root if their primary
group is 0 in /etc/passwd even if they aren't listed
as a member in /etc/group.  This is more inline with
what the group manpage says.

PR: 6696
Submitted by: Max Euston <meuston@jmrodgers.com>

26 years agoAdd a graceful jumping off point with a new .error directive.
Steve Price [Mon, 25 May 1998 03:28:37 +0000 (03:28 +0000)]
Add a graceful jumping off point with a new .error directive.

PR: 6720
Submitted by: Niall Smart <njs3@doc.ic.ac.uk>

26 years agoChange shell from /nonexistent to /sbin/nologin.
Steve Price [Mon, 25 May 1998 03:19:37 +0000 (03:19 +0000)]
Change shell from /nonexistent to /sbin/nologin.

PR: 6739
Submitted by: Are Bryne <are.bryne@communique.no>

26 years agoo Don't try to transfer tty device descriptors as there's no way of
Brian Somers [Mon, 25 May 1998 02:22:38 +0000 (02:22 +0000)]
o Don't try to transfer tty device descriptors as there's no way of
  transferring session rights with them.  Instead, create two
  `/bin/cat' processes.  A new child is spawned and disassociated from
  the terminal and the parent, which continues with the rest of the ppp
  process.  Meanwhile, the parent spawns another child, and both the
  parent and child exec the `/bin/cat' commands with the appropriate
  descriptors.  This way, the session is owned by the parent, and the
  tty is held open.
o Close LCPs that have done a TLF and are now in ST_STOPPED before
  calling Down.  This prevents them from trying to come back up again
  after the peer has shut them down (it seems a bit strange that the
  rfc says that a Down in ST_STOPPED will cause a TLS etc).
o Don't try to set the physical link name pointer when we're receiving
  and renaming a datalink.  The physical hasn't been created yet, and as
  it happens, the garbage physical pointer happens to be the value of another
  physical - so we're pointing that other physical name at ourselves.
  yeuck.
o Re-arrange the order of things in main (DoLoop()).  We now handle
  signals only after the select and not before the UpdateSet.  It's
  possible that either a signal (FSM timeout) or a descriptor_Read()
  brings a link down, after which we'd better tidy up any dead direct
  and 1off descriptors before calling UpdateSet() again.
o Mention when we detect a PPP packet when we see one before the link
  is up (then start LCP as before).

26 years agoDon't blow away parts of the sourcetree on a 'make clean'. This was
Eivind Eklund [Sun, 24 May 1998 21:51:50 +0000 (21:51 +0000)]
Don't blow away parts of the sourcetree on a 'make clean'.  This was
only harmful for building from a read-only filesystem - the parts
blown away were old rm'ed directories.

26 years agoMerge updates from 2.2
Jordan K. Hubbard [Sun, 24 May 1998 20:01:33 +0000 (20:01 +0000)]
Merge updates from 2.2

26 years agoSync documentation with 2.2 updates.
Jordan K. Hubbard [Sun, 24 May 1998 20:00:27 +0000 (20:00 +0000)]
Sync documentation with 2.2 updates.

26 years agoFix Anders Email address.
Steve Price [Sun, 24 May 1998 19:48:38 +0000 (19:48 +0000)]
Fix Anders Email address.

26 years agoAdd a fuller-featured Norwegian keycap entry.
Steve Price [Sun, 24 May 1998 19:37:43 +0000 (19:37 +0000)]
Add a fuller-featured Norwegian keycap entry.

PR: 6173
Submitted by: Anders Nordby <nickerne@nome.no>

26 years agoTake IP options into account when calculating the allowable length
Bill Fenner [Sun, 24 May 1998 18:41:04 +0000 (18:41 +0000)]
Take IP options into account when calculating the allowable length
of the TCP payload.  See RFC1122 section 4.2.2.6 .  This allows
Path MTU discovery to be used along with IP options.

PR: problem discovered by Kevin Lahey <kml@nas.nasa.gov>

26 years agoReviewed by: hasty@freebsd.org
Amancio Hasty [Sun, 24 May 1998 18:29:20 +0000 (18:29 +0000)]
Reviewed by: hasty@freebsd.org
Submitted by: hasty@star-gate.com

Eliminate NTSC default setting if user choses a different video format.

26 years agoAdd a chkgrp run after backing up the passwd and group files.
Dag-Erling Smørgrav [Sun, 24 May 1998 15:57:31 +0000 (15:57 +0000)]
Add a chkgrp run after backing up the passwd and group files.

Suggested-by: Andreas Klemm
26 years agoAdd chkgrp, a small utility for checking for syntax errors in group
Dag-Erling Smørgrav [Sun, 24 May 1998 15:09:23 +0000 (15:09 +0000)]
Add chkgrp, a small utility for checking for syntax errors in group
files.

PR: bin/1387
Suggested-by: Giles Lean <giles@nemeton.com

26 years agoAdd chkgrp
Dag-Erling Smørgrav [Sun, 24 May 1998 15:07:04 +0000 (15:07 +0000)]
Add chkgrp

26 years agoThe ipt_ptr field is 1-based (see TCP/IP Illustrated, Vol. 1, pp. 91-95),
David Greenman [Sun, 24 May 1998 14:59:57 +0000 (14:59 +0000)]
The ipt_ptr field is 1-based (see TCP/IP Illustrated, Vol. 1, pp. 91-95),
so it must be adjusted (minus 1) before using it to do the length check.
I'm not sure who to give the credit to, but the bug was reported by
Jennifer Dawn Myers <jdm@enteract.com>, who also supplied a patch. It
was also fixed in OpenBSD previously by andreas.gunnarsson@emw.ericsson.se,
and of course I did the homework to verify that the fix was correct per
the specification.
PR: 6738

26 years agoConvert a couple of large allocations to use zones rather than malloc
Peter Wemm [Sun, 24 May 1998 14:41:56 +0000 (14:41 +0000)]
Convert a couple of large allocations to use zones rather than malloc
for better packing.  This means that we can choose better values for the
various hash entries without having to try and get it all to fit within
an artificial power of two limit for malloc's sake.

26 years agoHopefully fix behaviour of libreadline when prompt is longer than
Dag-Erling Smørgrav [Sun, 24 May 1998 13:06:38 +0000 (13:06 +0000)]
Hopefully fix behaviour of libreadline when prompt is longer than
screen width and contains invisible characters.

PR: gnu/6701
Submitted-by: Chet Ramey chet@po.cwru.edu

26 years agoReformat some comments to not overflow 80 chars.
Julian Elischer [Sun, 24 May 1998 04:52:31 +0000 (04:52 +0000)]
Reformat some comments to not overflow 80 chars.

26 years agoSupport a 16K first level cache for 512K 2nd level. Also, add support
John Dyson [Sun, 24 May 1998 04:25:27 +0000 (04:25 +0000)]
Support a 16K first level cache for 512K 2nd level.  Also, add support
for 1MB 2nd level cache.

26 years agoRemove stray debug message.
Steve Price [Sun, 24 May 1998 03:25:24 +0000 (03:25 +0000)]
Remove stray debug message.

PR: 6722

26 years agoLoosely document the '-s' commandline option.
Steve Price [Sun, 24 May 1998 03:11:17 +0000 (03:11 +0000)]
Loosely document the '-s' commandline option.

PR: 6713

26 years agoPrimary verison of NetBIOS over TCP/IP. Now you can connect Windows
Atsushi Murai [Sun, 24 May 1998 03:03:10 +0000 (03:03 +0000)]
Primary verison of NetBIOS over TCP/IP. Now you can connect Windows
DOMAIN as DOMAIN user through NAT function. See also RFC1002 for
futher detail of SMB structure.

Submitted by: Atsushi Murai <amurai@spec.co.jp>

26 years agoAdd 'newsyslog.conf' to the list of files to not clobber
Steve Price [Sun, 24 May 1998 02:49:54 +0000 (02:49 +0000)]
Add 'newsyslog.conf' to the list of files to not clobber
during the upgrade process.

PR: 6629

26 years agoAdd CVS option
Jordan K. Hubbard [Sun, 24 May 1998 01:29:52 +0000 (01:29 +0000)]
Add CVS option
Submitted by: Robert Baron <rvb@sicily.odyssey.cs.cmu.edu>

26 years agoRemove two CODA entries by request.
Jordan K. Hubbard [Sun, 24 May 1998 01:28:56 +0000 (01:28 +0000)]
Remove two CODA entries by request.
Submitted by: Robert Baron <rvb@sicily.odyssey.cs.cmu.edu>

26 years agoFix a mind-o in tl_setmulti(): when setting a bit in the upper
Bill Paul [Sun, 24 May 1998 00:56:49 +0000 (00:56 +0000)]
Fix a mind-o in tl_setmulti(): when setting a bit in the upper
32 bits of the 64-bit hash table, we have to use a 32-bit shift,
not 31.

Pointed out by: Bill Fenner

26 years agoShow the link mode in `show links'.
Brian Somers [Sat, 23 May 1998 22:28:19 +0000 (22:28 +0000)]
Show the link mode in `show links'.

26 years agoOops - patch failed in last commit ????
Brian Somers [Sat, 23 May 1998 22:27:53 +0000 (22:27 +0000)]
Oops - patch failed in last commit ????

26 years agoo Move our prompt descriptor list outside of the bundle.
Brian Somers [Sat, 23 May 1998 22:24:50 +0000 (22:24 +0000)]
o Move our prompt descriptor list outside of the bundle.
  It's now dealt with by the `server' object.  This simplifies
  things as we only have one list of prompt descriptors and
  the log_ routines check prompt::logactive to determine
  whether it should be used for output.
o Include the MP socket UpdateSet() result in bundle::UpdateSet().
o Don't select on the tun device unless we're in NETWORK
  phase or AUTO mode.
o Stop the idle timer when we go to DEAD phase.  We may
  have transferred a link and not had a chance to kill
  it.
o Don't fail when trying to unlink our transferred datalink
  from our descriptor lists just before the transfer.
o Add our link descriptor to the write set if we got a short
  write the last time (physical::out is set).
o Log the connection source address when a connection is closed.
o Remove descriptor::next field.  Descriptor lists are not required
  any more.

26 years agoo Make sure we adjust our min seq and process any outstanding queued
Brian Somers [Sat, 23 May 1998 17:05:28 +0000 (17:05 +0000)]
o Make sure we adjust our min seq and process any outstanding queued
  incoming fragments when a link goes down.
o Don't use the minimum sequence numbers of links that aren't open.
o Understand sequence number wrapping when determining the minimum
  sequence number.
o Add & adjust a few comments.