]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
29 years agoPreserve current termios speed for TIOCSET*, if it matched with
ache [Wed, 2 Aug 1995 12:53:14 +0000 (12:53 +0000)]
Preserve current termios speed for TIOCSET*, if it matched with
nearest valid. It means that gtty+stty transaction (without speed
change) not breaks non-standard speeds now.

29 years agoChange default banner fro 4.4 BSD to FreeBSD.
paul [Wed, 2 Aug 1995 12:07:31 +0000 (12:07 +0000)]
Change default banner fro 4.4 BSD to FreeBSD.
Reviewed by:
Submitted by:
Obtained from:

29 years agoBetter approximation for TIOCGETP (gtty) for non-standard speeds.
ache [Wed, 2 Aug 1995 12:03:12 +0000 (12:03 +0000)]
Better approximation for TIOCGETP (gtty) for non-standard speeds.
Old variant returns 38400 for them, now it returns nearest matched
rounded down, expect speeds in range 0 > speed < 50 rounded up
to not produce hangup.

29 years agoBack out predefined termios speeds check, it can cause troubles
ache [Wed, 2 Aug 1995 11:26:50 +0000 (11:26 +0000)]
Back out predefined termios speeds check, it can cause troubles
with interaction pty <-> serial driver with non-standard speed.
So, nothing protect us from garbadge in speed field, expect
checking for < 0 left in tty.c :-(

29 years agoUse the same DECODE_BAUD trick like in new telnetd to obtain
ache [Wed, 2 Aug 1995 11:20:05 +0000 (11:20 +0000)]
Use the same DECODE_BAUD trick like in new telnetd to obtain
termios speed.
Obtained from: Pre-Lite2 telnet

29 years agoUse the same DECODE_BAUD trick like in new telnet to set
ache [Wed, 2 Aug 1995 11:13:30 +0000 (11:13 +0000)]
Use the same DECODE_BAUD trick like in new telnet to set
termios speed
Obtained from: Pre-Lite2 telnet

29 years agoAllow any speed from 0..76800
ache [Wed, 2 Aug 1995 10:17:35 +0000 (10:17 +0000)]
Allow any speed from 0..76800
Reviewed by:
Submitted by:
Obtained from:

29 years agoLock the directory vnode before VOP_READDIR in nfsrv_readdirplus
dfr [Wed, 2 Aug 1995 10:12:47 +0000 (10:12 +0000)]
Lock the directory vnode before VOP_READDIR in nfsrv_readdirplus

29 years agoReviewed by: David Greenman
wpaul [Wed, 2 Aug 1995 09:14:23 +0000 (09:14 +0000)]
Reviewed by: David Greenman

Back out the 'help NIS rebind faster' hack. This change used a
connect()/send() pair rather than the original sendto() to allow
RPC to pass ICMP host unreachable and similar errors up to RPC
programs that use UDP. This is not a terrible thing by itself, but it can
cause trouble in environments with multi-homed hosts: if the portmapper
on the multi-homed machine sends a reply with a source address
that's different than the one associated with the connection by
connect(), the kernel will send a port unreachable message and
drop the reply. For the sake of compatibility with everybody else
on the planet, it's best to revert to the old behavior.

*long, heavy sigh*

29 years agoSubmitted by: Wolfram Schneider <w@cs.tu-berlin.de>
asami [Wed, 2 Aug 1995 07:30:37 +0000 (07:30 +0000)]
Submitted by: Wolfram Schneider <w@cs.tu-berlin.de>

The currently (1995/07/31 04:22:07) implemented if (-x "$prog") {
works only if you have '.' in your $PATH variable.

29 years agoCheck for valid speed values in pty drive
ache [Wed, 2 Aug 1995 06:55:36 +0000 (06:55 +0000)]
Check for valid speed values in pty drive
Check for negative speed values in tty drive
Back out valid speed values checking from tty drive
Suggested by: bde

29 years agoDon't set TS_ZOMBIE flag for non-open ptys. ptcclose() has always done
bde [Wed, 2 Aug 1995 02:55:47 +0000 (02:55 +0000)]
Don't set TS_ZOMBIE flag for non-open ptys.  ptcclose() has always done
too much for non-open ptys, but there is normally no problem because the
l_modem(, 0) is a no-op for closed ptys provided the line discipline is
standard and MDMBUF isn't set.

29 years agoRemove top secret version information from the banner by default.
jkh [Wed, 2 Aug 1995 02:45:46 +0000 (02:45 +0000)]
Remove top secret version information from the banner by default.
Loose lips sink ships!  Man the firewall and look alert!

29 years agoOptimize a bit valid speed search using fact that speed table sorted
ache [Tue, 1 Aug 1995 23:38:00 +0000 (23:38 +0000)]
Optimize a bit valid speed search using fact that speed table sorted
Submitted by:
Obtained from:

29 years agoCheck for valid speeds in TIOCSET* and return EINVAL for incorrect
ache [Tue, 1 Aug 1995 23:27:36 +0000 (23:27 +0000)]
Check for valid speeds in TIOCSET* and return EINVAL for incorrect
values instead of setting garbadge.

29 years agoMake strtod conforms manpage, use isspace to skip initial whitespaces
ache [Tue, 1 Aug 1995 22:20:16 +0000 (22:20 +0000)]
Make strtod conforms manpage, use isspace to skip initial whitespaces
instead of hardcoded whitespaces

29 years agoSimilar changes like in strtol, all this family is VERY broken
ache [Tue, 1 Aug 1995 22:04:57 +0000 (22:04 +0000)]
Similar changes like in strtol, all this family is VERY broken
in 8bit environment (isalpha at the end of digits)

29 years agoAdd the maybe_stripped to a CLEANFILES target to get a clean obj directory
ats [Tue, 1 Aug 1995 21:53:35 +0000 (21:53 +0000)]
Add the maybe_stripped to a CLEANFILES target to get a clean obj directory
after a clean or cleandir.

29 years agostrtol and atoi VERY broken in 8bit chars locale, i.e. if you pass something
ache [Tue, 1 Aug 1995 21:38:00 +0000 (21:38 +0000)]
strtol and atoi VERY broken in 8bit chars locale, i.e. if you pass something
like 38400<any 8bit char, isalpha> it not detect this stuff and
produce very big number instead. Fixed by operating with unsigned char
and checking for isascii. (secure/telnetd hits by it f.e.)

29 years agoRemoved my special-case hack for VOP_LINK and fixed the problem with the
dg [Tue, 1 Aug 1995 18:51:02 +0000 (18:51 +0000)]
Removed my special-case hack for VOP_LINK and fixed the problem with the
wrong vp's ops vector being used by changing the VOP_LINK's argument order.
The special-case hack doesn't go far enough and breaks the generic
bypass routine used in some non-leaf filesystems. Pointed out by Kirk
McKusick.

29 years agoAdd device support for the Matrox Meteor.
jkh [Tue, 1 Aug 1995 13:15:52 +0000 (13:15 +0000)]
Add device support for the Matrox Meteor.
Submitted by: james

29 years agoTweak the default entry a little to take advantage of new getty
jkh [Tue, 1 Aug 1995 13:13:30 +0000 (13:13 +0000)]
Tweak the default entry a little to take advantage of new getty
feature.

29 years agoA useful aid.. Add support for:
jkh [Tue, 1 Aug 1995 13:12:24 +0000 (13:12 +0000)]
A useful aid..  Add support for:

%r: current release
%m: machine architecture type (i386 for now)
%s: OS name (FreeBSD)

from uname() in banner string.

29 years agoRound this out. Now seems to transfer packages properly over ftp
jkh [Tue, 1 Aug 1995 09:49:27 +0000 (09:49 +0000)]
Round this out.  Now seems to transfer packages properly over ftp
as it was supposed to.  PKG_PATH more sorely needed than ever now..
Hmmmm.

29 years agoArgh! I hate my life! I can't seem to do anything right here today.
jkh [Tue, 1 Aug 1995 07:16:51 +0000 (07:16 +0000)]
Argh!  I hate my life!  I can't seem to do anything right here today.
This should restore functionality to the pkg_install suite.

29 years agoSync to author's Version 1.3.
jkh [Tue, 1 Aug 1995 07:07:42 +0000 (07:07 +0000)]
Sync to author's Version 1.3.
Submitted by: james

29 years agoSync to reality for the Gravis Ultrasound MAX card.
jkh [Tue, 1 Aug 1995 07:05:16 +0000 (07:05 +0000)]
Sync to reality for the Gravis Ultrasound MAX card.

29 years agoReduce passed speed to 38400, like telnet does too.
ache [Tue, 1 Aug 1995 02:23:21 +0000 (02:23 +0000)]
Reduce passed speed to 38400, like telnet does too.
Remote rlogind may not understand speeds > 38400.

29 years agoObtained from: partly from ancient patches of mine via 1.1.5
bde [Mon, 31 Jul 1995 22:50:08 +0000 (22:50 +0000)]
Obtained from: partly from ancient patches of mine via 1.1.5

Change all short variables in `struct tty' to int.  Shorts were only
right on ancient systems with ints optimized for vaxness over
efficiency.

29 years agoObtained from: partly from ancient patches of mine via 1.1.5
bde [Mon, 31 Jul 1995 22:48:46 +0000 (22:48 +0000)]
Obtained from: partly from ancient patches of mine via 1.1.5

Handle MDMBUF a little better.  Prepare to handle 4 different kinds of
output flow control.

29 years agoReserve space for Jim Lowe's impending Matrox Meteor card driver.
jkh [Mon, 31 Jul 1995 22:06:55 +0000 (22:06 +0000)]
Reserve space for Jim Lowe's impending Matrox Meteor card driver.
Submitted by: james

29 years agoObtained from: an ancient patch of mine via 1.1.5
bde [Mon, 31 Jul 1995 21:54:46 +0000 (21:54 +0000)]
Obtained from: an ancient patch of mine via 1.1.5

Call output process whether or not there is any output.  The output
process may be overloaded to handle hardware flow control and
hardware output completions.

29 years agoObtained from: an ancient patch of mine via 1.1.5
bde [Mon, 31 Jul 1995 21:43:37 +0000 (21:43 +0000)]
Obtained from: an ancient patch of mine via 1.1.5

Clear PENDIN when input is flushed so that the handling of future input
doesn't get pessimized.

29 years agoAdd Irish mirror site.
jkh [Mon, 31 Jul 1995 21:39:47 +0000 (21:39 +0000)]
Add Irish mirror site.

29 years agoEliminate the use of TS_TIMEOUT and ttstrt(). These are for handling
bde [Mon, 31 Jul 1995 21:35:17 +0000 (21:35 +0000)]
Eliminate the use of TS_TIMEOUT and ttstrt().  These are for handling
tab delays etc.  pcvt was using them to recover from a (rarely lost)
race.  Use a little more locking to avoid the race.

29 years agoUse tsleep() instead of ttysleep() to wait for carrier since a generation
bde [Mon, 31 Jul 1995 21:28:42 +0000 (21:28 +0000)]
Use tsleep() instead of ttysleep() to wait for carrier since a generation
change isn't an error.

29 years agoSleep on a better address to wait for output to drain out of the
bde [Mon, 31 Jul 1995 21:10:36 +0000 (21:10 +0000)]
Sleep on a better address to wait for output to drain out of the
hardware.  Set the sleep-on flag for the address so there is more
than a small chance that the sleep address is actually used (this
used to work by timing out).  Don't bother clearing the sleep-on
flag after a timeout here or elsewhere since leaving it set just
generates a few null calls to wakeup().

29 years agoObtained from: partly from ancient patches of mine via 1.1.5
bde [Mon, 31 Jul 1995 21:02:00 +0000 (21:02 +0000)]
Obtained from: partly from ancient patches of mine via 1.1.5

Introduce TS_CONNECTED and TS_ZOMBIE states.  TS_CONNECTED is set
while a connection is established.  It is set while (TS_CARR_ON or
CLOCAL is set) and TS_ZOMBIE is clear.  TS_ZOMBIE is set for on to
off transitions of TS_CARR_ON that occur when CLOCAL is clear and
is cleared for off to on transitions of CLOCAL.  I/o can only occur
while TS_CONNECTED is set.  TS_ZOMBIE prevents further i/o.

Split the input-event sleep address TSA_CARR_ON(tp) into TSA_CARR_ON(tp)
and TSA_HUP_OR_INPUT(tp).  The former address is now used only for
off to on carrier transitions and equivalent CLOCAL transitions.
The latter is used for all input events, all carrier transitions
and certain CLOCAL transitions.  There are some harmless extra
wakeups for rare connection- related events.  Previously there were
too many extra wakeups for non-rare input events.

Drivers now call l_modem() instead of setting TS_CARR_ON directly
to handle even the initial off to on transition of carrier.  They
should always have done this.  l_modem() now handles TS_CONNECTED
and TS_ZOMBIE as well as TS_CARR_ON.

gnu/isdn/iitty.c:
Set TS_CONNECTED for first open ourself to go with bogusly setting
CLOCAL.

i386/isa/syscons.c, i386/isa/pcvt/pcvt_drv.c:
We fake carrier, so don't also fake CLOCAL.

kern/tty.c:
Testing TS_CONNECTED instead of TS_CARR_ON fixes TIOCCONS forgetting to
test CLOCAL.  TS_ISOPEN was tested instead, but that broke when we disabled
the clearing of TS_ISOPEN for certain transitions of CLOCAL.

Testing TS_CONNECTED fixes ttyselect() returning false success for output
to devices in state !TS_CARR_ON && !CLOCAL.

Optimize the other selwakeup() call (this is not related to the other
changes).

kern/tty_pty.c:
ptcopen() can be declared in traditional C now that dev_t isn't short.

29 years agoAssorted cosmetic changes:
bde [Mon, 31 Jul 1995 19:17:19 +0000 (19:17 +0000)]
Assorted cosmetic changes:

Make more functions static.

tty.c:
Use tcflag_t (u_long) and cc_t instead of u_char and int/long.

Don't record values that are only evaluated once.

Compare ints using imin(), not min().  min() is for comparing u_ints.
Old versions of tty.c used the type-safe but multiple-evaluation-unsafe
macro MIN().  The args are apparently never negative; otherwise this
change would be non-cosmetic.

Don't repeat the loop test in ttywait().

tty.h:
Improve English in and formatting of comments.

29 years agoImprove input flow control.
bde [Mon, 31 Jul 1995 18:29:51 +0000 (18:29 +0000)]
Improve input flow control.

Use input buffer watermarks of TTYHOG-512 (high) and (high)*7/8
(low) instead of TTYHOG/2 (high) and TTYHOG/5 (low) to agree with
some drivers.  512 is magic and some things depended on TTYHOG/2
>= TTYHOG-512 to work; now they depend on the 512 magic not changing
and TTYHOG-512 being significantly larger than 0.  This should be
handled in ttsetwater().

Separate the decision about whether to do input flow control from
doing it.  ttyblock() now just starts input flow control (hardware
and/or software) and there is a new function ttyunblock() to stop
it.  The decisions are the same except for the watermark changes
and allowing for input expansion for PARMRK.

When flushing input, try harder at first to send a start character
if required, but give up if the first attempt fails.

cy.c, rc.c, sio.c:
Simplify: let ttyinput() handle input flow control if it is not
being bypassed.  Use ttyblock() to start flow control otherwise.

rc.c:
Use same input flow control test as elsewhere: test in a more
efficient order and start flow control at >= highwater instead of
at > highwater.

29 years agoThe other day someone brought me an old Apple Laserwriter II with a serial
wpaul [Mon, 31 Jul 1995 13:59:45 +0000 (13:59 +0000)]
The other day someone brought me an old Apple Laserwriter II with a serial
interface set at 57600 baud, and I found out the hard way that lpd doesn't
know about speeds greater than 38400, even though <sys/ttydev.h> also
permits 57600 and 115200 baud. Fix this by adding B57600 and B115200 to the
'bauds' table. (The Apple printer worked properly once I did this, BTW. :)

29 years agoTry to make the `syn' blocking code act a bit more sensibly - don't
gpalmer [Mon, 31 Jul 1995 13:58:35 +0000 (13:58 +0000)]
Try to make the `syn' blocking code act a bit more sensibly - don't
block `syn' packets that have `ack' set.
Reviewed by:
Submitted by:
Obtained from:

29 years agoFix panic("ifpromisc failed") when shutting down a bpf tap when the attached
peter [Mon, 31 Jul 1995 10:35:36 +0000 (10:35 +0000)]
Fix panic("ifpromisc failed") when shutting down a bpf tap when the attached
interface is no longer IFF_UP.
The test for IFF_UP in ifpromisc is only useful while enabling IFF_PROMISC
and the higher levels of the bpf code do not allow for the possibility of
failure while shutting down.  This is a trivial change.
Also, fixes PR#522.

29 years agoRemove a redundant `if' from tcp_reass().
olah [Mon, 31 Jul 1995 10:24:22 +0000 (10:24 +0000)]
Remove a redundant `if' from tcp_reass().

Correct a typo in a comment (SEND_SYN -> NEEDSYN).

Reviewed by: David Greenman

29 years agoNull terminate all strings returned by the dummy uname() routine,
mpp [Mon, 31 Jul 1995 10:10:02 +0000 (10:10 +0000)]
Null terminate all strings returned by the dummy uname() routine,
and make sure that the version string is somewhat sane.  This
closes out PR#462.

Reviewed by: Bruce Evans

29 years agoFix the sysctl string routines to return as much of the
mpp [Mon, 31 Jul 1995 10:07:31 +0000 (10:07 +0000)]
Fix the sysctl string routines to return as much of the
string as possible and return ENOMEM if the entire string cannot
be returned.  This brings the routines in line with how the man
page says they work, and how the calling routines are expecting
them to work.  This allows the dummy uname() routine in libc to
obtain the version string, since the kernel version string is
longer than that normally returned by the uname() routine.
This is 3/4 of the fix for PR# 462.

Reviewed by: Bruce Evans

29 years agoFix some incorrect comments that make reference to /dev/fd.
mpp [Mon, 31 Jul 1995 09:52:21 +0000 (09:52 +0000)]
Fix some incorrect comments that make reference to /dev/fd.

29 years agoReturn the correct number of I/O ports as part of the probe so that the
dufault [Mon, 31 Jul 1995 09:38:21 +0000 (09:38 +0000)]
Return the correct number of I/O ports as part of the probe so that the
conflict resolution works.

29 years agoFix various kernfs file system problems, which can result in
mpp [Mon, 31 Jul 1995 08:52:02 +0000 (08:52 +0000)]
Fix various kernfs file system problems, which can result in
umountable file systems, hung processes, or system panics:

- Some operations could return without decrementing the vnode
  reference count.
- Some operations could leave the vnode locked.
- Generalize the /kern/rootdev & rrootdev files so that they
  are no longer special cased in kernfs_lookup().

Note: procfs, fdescfs, and most of the other miscfs/* file systems
also suffer from the same type of problems and I will work on
fixing them one at a time.

29 years agoLong overdue, more complete, reset code. These changes implement a
gibbs [Mon, 31 Jul 1995 08:25:36 +0000 (08:25 +0000)]
Long overdue, more complete, reset code.  These changes implement a
BUS DEVICE RESET followed by BUS RESET failure recovery strategy including
the necesary renegotiation of sync/wide transfers after recovery completes.

Clean up debugging code to make it more finely selectable.  Reset code
debugging is enabled for now so I can get more feedback on how this
code behaves in real life.

29 years agoPause the sequencer when message only requests complete. This allows the
gibbs [Mon, 31 Jul 1995 08:21:59 +0000 (08:21 +0000)]
Pause the sequencer when message only requests complete.  This allows the
kernel driver to know the exact state of the sequencer after a BUS DEVICE
RESET or ABORT completes so it can properly clean up the request.

29 years agoSubmitted by: Torbjorn Granlund <tege@matematik.su.se>
asami [Mon, 31 Jul 1995 04:22:07 +0000 (04:22 +0000)]
Submitted by: Torbjorn Granlund <tege@matematik.su.se>

* /usr/bin/which doesn't work if a complete path is passed to it.  This is
  incompatible with other flavours of Unix, Here is a suggested patch:

29 years agoSync this up with ftp.c.
asami [Mon, 31 Jul 1995 04:00:01 +0000 (04:00 +0000)]
Sync this up with ftp.c.

29 years agoRemove dependency on sysinstall.
jkh [Mon, 31 Jul 1995 02:27:58 +0000 (02:27 +0000)]
Remove dependency on sysinstall.

29 years agoAdd ftp.c to makefile - that might make it work better. Sheesh!
jkh [Mon, 31 Jul 1995 01:26:58 +0000 (01:26 +0000)]
Add ftp.c to makefile - that might make it work better.  Sheesh!

29 years agoMinor corrections.
jfieber [Mon, 31 Jul 1995 01:18:46 +0000 (01:18 +0000)]
Minor corrections.
Submitted by: joerg_wunsch@uriah.heep.sax.de (Joerg Wunsch)

29 years agoIgnore trailing slashes in pathnames that "refer to a directory",
bde [Mon, 31 Jul 1995 00:35:58 +0000 (00:35 +0000)]
Ignore trailing slashes in pathnames that "refer to a directory",
as is required to be POSIXLY_CORRECT and "right".  I interpret
"referring to a directory" as being a directory or becoming a
directory.  E.g., the trailing slashes in mkdir("/nonesuch/"),
rename("/tmp", /nonesuch/") and link("/tmp", "/root_can_like_dirs/")
are ignored because the target will become a directory if the
syscall succeeds.  A trailing slash on a symlink causes the symlink
to be followed (this is a bug if the symlink doesn't point to a
directory; fix later).

29 years agoCorrect the mangled English when mentioning that the kernel config-file
dg [Sun, 30 Jul 1995 18:33:01 +0000 (18:33 +0000)]
Correct the mangled English when mentioning that the kernel config-file
"dump on" is deprecated...by removing the note completely.

29 years agoFix a bug in my disabled version of trap_pfault()...curpcb may be NULL even
dg [Sun, 30 Jul 1995 17:49:24 +0000 (17:49 +0000)]
Fix a bug in my disabled version of trap_pfault()...curpcb may be NULL even
when curproc isn't. This condition occurs at system startup and perhaps
at other times.

29 years agoRewrote shutdown_nice() to fix the init-not-yet-started panic().
dg [Sun, 30 Jul 1995 17:10:36 +0000 (17:10 +0000)]
Rewrote shutdown_nice() to fix the init-not-yet-started panic().

29 years agoTypo in log message : errros -> errors
gpalmer [Sun, 30 Jul 1995 15:18:29 +0000 (15:18 +0000)]
Typo in log message : errros -> errors

29 years agoDon't swap the queue headers to implement concatenation of the
bde [Sun, 30 Jul 1995 13:52:56 +0000 (13:52 +0000)]
Don't swap the queue headers to implement concatenation of the
queues for TIOCSETA[W].  Swapping an even number of times broke
the queue resource limits.  This would have broken CRTSCTS flow
control if the clist slush list was used up.

Don'concatenate the queues for TIOCSETA[W] if one of the queues
has a resource limit of 0.  Concatenation would cause a panic if
one of the queues is nonempty and the other is limited to length
0.  This may have caused panics in PPPDISC.

Wake up readers after all transitions of ICANON.  When ICANON is
turned off it is quite likely that characters will become available
to be read.

Reduce indentation near these changes.

29 years agoRemove a newline from the output of savecore(8), so the bogus number
joerg [Sun, 30 Jul 1995 13:10:56 +0000 (13:10 +0000)]
Remove a newline from the output of savecore(8), so the bogus number
1024 that used to remain on a line of its own after savecore completed
its job will be overwritten later in the /etc/rc process.

Reviewed by:
Submitted by: graichen@sirius.physik.fu-berlin.de (Thomas Graichen)
Obtained from:

29 years agoscsi(8) used to fall off the end of main(), returning an arbitrary
joerg [Sun, 30 Jul 1995 12:58:47 +0000 (12:58 +0000)]
scsi(8) used to fall off the end of main(), returning an arbitrary
value.

Reviewed by:
Submitted by:
Obtained from:

29 years agoFix the mknod for spigot.
jkh [Sun, 30 Jul 1995 12:56:26 +0000 (12:56 +0000)]
Fix the mknod for spigot.

29 years agoUpdate the parts on how to take a kernel crash dump.
joerg [Sun, 30 Jul 1995 12:53:39 +0000 (12:53 +0000)]
Update the parts on how to take a kernel crash dump.
Reviewed by:
Submitted by:
Obtained from:

29 years agoSplit TS_ASLEEP (sleep on output [below low water])into TS_SO_OLOWAT (sleep
bde [Sun, 30 Jul 1995 12:39:42 +0000 (12:39 +0000)]
Split TS_ASLEEP (sleep on output [below low water])into TS_SO_OLOWAT (sleep
on output below low water) and TS_SO_OCOMPLETE (sleep on output complete).
Most of the support for this has already been committed.  Drivers should
call ttwwakeup() to handle wakeups whenever output is below low water
(and some output event causes this condition to be checked) or TS_BUSY is
cleared.

tty.c:
Fix the livelock in ttywait() properly by sleeping on output complete, not
on output below low water.

Use ttwwakeup() instead of separate select and output wakeups for all
wakeups of writers.

Add wakeups of writers for output flushes and carrier/clocal transitions.

Don't go to sleep in ttycheckoutq() if ttstart() reduces the queue to below
low water.

Use the timeout built into tsleep() in ttycheckoutq().

Optimize the select wakeup in ttwwakeup().  It seems reasonable to know
too much about the internals of tp->t_wsel now that the knowledge is
localised in tty.c.

29 years agoChange the environment variable this looks for from FTP_PASSIVE to
jkh [Sun, 30 Jul 1995 09:41:21 +0000 (09:41 +0000)]
Change the environment variable this looks for from FTP_PASSIVE to
FTP_PASSIVE_MODE.  It would be really nice if we could standardise on
this name so that all tools (like ncftp) that offer passive/active
ftp selection would work seamlessly with one user environment variable
setting.

29 years agoDocument how to get pkg_add to use PASSIVE MODE ftp in such transfers.
jkh [Sun, 30 Jul 1995 09:37:31 +0000 (09:37 +0000)]
Document how to get pkg_add to use PASSIVE MODE ftp in such transfers.

29 years agoTotally eliminate the dependency on libftp (which will be removed from the
jkh [Sun, 30 Jul 1995 09:33:31 +0000 (09:33 +0000)]
Totally eliminate the dependency on libftp (which will be removed from the
FreeBSD source tree) and switch to the internal ftp routines developed
by Poul-Henning and used in sysinstall.

29 years agoo Fix PR# bin/643 - always set "where_to" properly.
jkh [Sun, 30 Jul 1995 09:11:20 +0000 (09:11 +0000)]
o Fix PR# bin/643 - always set "where_to" properly.

o Nuke dependence on libftp - it's about to die!

29 years agoRename GZIP to GZIPCMD to match previous change in bsd.doc.mk.
bde [Sun, 30 Jul 1995 08:04:50 +0000 (08:04 +0000)]
Rename GZIP to GZIPCMD to match previous change in bsd.doc.mk.

Submitted by: John Hay (jhay@mikom.csir.co.za)

29 years agoThe default for temporary directory is /var/tmp, not /tmp.
asami [Sun, 30 Jul 1995 05:25:35 +0000 (05:25 +0000)]
The default for temporary directory is /var/tmp, not /tmp.

29 years agoCorected a few typo which resulted in strange resulon results. Add URL
roberto [Sun, 30 Jul 1995 01:44:51 +0000 (01:44 +0000)]
Corected a few typo which resulted in strange resulon results. Add URL
pointing to the Handbook. Changed a few entries to mention /etc/sysconfig.
Reviewed by:
Submitted by:
Obtained from:

29 years agoYou can run pkg_info everywhere expect /var/db/pkg directory :-)
ache [Sun, 30 Jul 1995 01:44:45 +0000 (01:44 +0000)]
You can run pkg_info everywhere expect /var/db/pkg directory :-)
Running there you got any kind of strange errors from tar caused
by treating directories as tar files!
Fix it by adding new isfile(name) (check for reg. files) to simple fexists(name) calls.

29 years agoIf some files lost in /var/db/pkg/<port>/ after system crash f.e.,
ache [Sun, 30 Jul 1995 01:08:34 +0000 (01:08 +0000)]
If some files lost in /var/db/pkg/<port>/ after system crash f.e.,
pkg_manage silently dumps core, pkg_info claims about them to
stderr, which makes very difficult to find what directory cause it via
tons of pkg_info -a output. I found solution which covers both variants,
now pkg_info claims about missing files to stdout among valid output
with ERROR: prefix. It heals pkg_manage to not dump core and makes
easy to find errors in pkg_info -a output by simple /ERROR 'more' command.

29 years agoCorrected @@ into @.
roberto [Sat, 29 Jul 1995 22:39:44 +0000 (22:39 +0000)]
Corrected @@ into @.

29 years agoPut the name of the latest snapshot: 2.1.0-950726-SNAP.
roberto [Sat, 29 Jul 1995 22:35:40 +0000 (22:35 +0000)]
Put the name of the latest snapshot: 2.1.0-950726-SNAP.

29 years agoOutput statistics as unsigned in the -w section.
dg [Sat, 29 Jul 1995 22:34:15 +0000 (22:34 +0000)]
Output statistics as unsigned in the -w section.
Look at error return of kread() and stop on error.
Fix warning in kread() to not output "kvm_read:" twice.
Killed PCB cache misses stat as we no longer have it.

29 years agoRestore tabs in inetd line
ache [Sat, 29 Jul 1995 22:22:08 +0000 (22:22 +0000)]
Restore tabs in inetd line
Submitted by:
Obtained from:

29 years agoAdd "connections dropped by persist timeout" statistic.
dg [Sat, 29 Jul 1995 18:49:16 +0000 (18:49 +0000)]
Add "connections dropped by persist timeout" statistic.

29 years agoAdd connection drop capability for persist timeouts.
dg [Sat, 29 Jul 1995 18:48:44 +0000 (18:48 +0000)]
Add connection drop capability for persist timeouts.

Reviewed by: Andras Olah
Obtained from: 4.4BSD-lite2 via W. Richard Stevens

29 years agoOnly execute /usr/games/fortune if it's there.
pst [Sat, 29 Jul 1995 18:43:52 +0000 (18:43 +0000)]
Only execute /usr/games/fortune if it's there.

29 years agoAdd new ``SNAPDATE=""'' for use in snapshot building.
rgrimes [Sat, 29 Jul 1995 18:42:29 +0000 (18:42 +0000)]
Add new ``SNAPDATE=""'' for use in snapshot building.

Reviewed by: jkh

29 years agoInitialize "name" to quiet compiler.
dg [Sat, 29 Jul 1995 18:21:48 +0000 (18:21 +0000)]
Initialize "name" to quiet compiler.

29 years agoChange the ``make depend all'' to ``make depend'' then ``make all'' in
roberto [Sat, 29 Jul 1995 18:12:50 +0000 (18:12 +0000)]
Change the ``make depend all'' to ``make depend'' then ``make all'' in
the kernel configuration section.

29 years agoMajor overhaul of the FAQ. I've added several contributions
roberto [Sat, 29 Jul 1995 18:09:19 +0000 (18:09 +0000)]
Major overhaul of the FAQ. I've added several contributions
specially from Darryl Okahata. I've rewritten several URL to the proper
<url url="" name=""> tag. There is still room for improvement but it
should be closer to 2.0.5R now. I'll try to be faster for future updates...

Obtained from: Mail messages from the lists.

29 years agoDon't build xditview unless the actual include files for X11 are present.
pst [Sat, 29 Jul 1995 17:47:53 +0000 (17:47 +0000)]
Don't build xditview unless the actual include files for X11 are present.
Merely checking for the directory is -not- sufficient, since the XFree86
non-programmers distribution has that directory to hold bitmaps.

29 years agorexecd was not calling "setlogin()" when it should have. This was causing
peter [Sat, 29 Jul 1995 15:21:15 +0000 (15:21 +0000)]
rexecd was not calling "setlogin()" when it should have.  This was causing
getlogin() to return wrong answers (eg: "root").
Reviewed by: davidg
Obtained from: James Jegers, for NetBSD, slightly reworked by me.

29 years agoFix enough of the recently introduced brokenness for LINT to compile.
bde [Sat, 29 Jul 1995 14:20:54 +0000 (14:20 +0000)]
Fix enough of the recently introduced brokenness for LINT to compile.
The U6850* changes in sound_config.h are probably wrong.

Recent commits lost:
- include paths.
- copyrights.
- cvs Ids.
- infamous whitespace changes.
- other cosmetic changes.

29 years agoObtained from: partly from ancient patches by ache and me via 1.1.5
bde [Sat, 29 Jul 1995 13:40:13 +0000 (13:40 +0000)]
Obtained from: partly from ancient patches by ache and me via 1.1.5

Remove nullmodem().

It may be useful to have a null modem routine, but nullmodem()
wasn't one.  nullmodem() was identical to ttymodem() except it
didn't implement MDMBUF (carrier) flow control, didn't do any
wakeups for off to on carrier transitions, and didn't flush the
i/o queues for on to off carrier transitions (flushing has the side
effect of waking up readers and writers) although it did generate
SIGHUPs.  The wakeups must normally be done even if nullmodem() is
null in case something is sleeping waiting for a carrier transition.
In any case, the wakeups should be harmless.  They may cause bogus
results for select(), but select() is already bogus for nonstandard
line disciplines.

29 years agoAdded a new section un user ppp (iijppp).
jfieber [Sat, 29 Jul 1995 13:08:02 +0000 (13:08 +0000)]
Added a new section un user ppp (iijppp).
Submitted by: Nik Clayton <nik@blueberry.co.uk>

29 years agoThe lemac driver didn't pass transmitted packets to bpf. Now it does.
paul [Sat, 29 Jul 1995 13:00:17 +0000 (13:00 +0000)]
The lemac driver didn't pass transmitted packets to bpf. Now it does.

Reviewed by:
Submitted by:
Obtained from:

29 years agoAfter pst and ache fixed secure telnet, it was still not in the main
markm [Sat, 29 Jul 1995 12:49:25 +0000 (12:49 +0000)]
After pst and ache fixed secure telnet, it was still not in the main
makefiles. This puts it in.

PLEASE NOTE - YOU WILL NEED TO BUILD AND INSTALL THE libtelnet IN secure/
Reviewed by:
Submitted by:
Obtained from:

29 years agoEliminate sloppy common-style declarations. There should be none left for
bde [Sat, 29 Jul 1995 11:44:31 +0000 (11:44 +0000)]
Eliminate sloppy common-style declarations.  There should be none left for
the LINT configuation.

29 years agoRemoved the poorly named unused bogus common variable `total'. It was
bde [Sat, 29 Jul 1995 10:47:05 +0000 (10:47 +0000)]
Removed the poorly named unused bogus common variable `total'.  It was
obsoleted by the VM_METER sysctl a long time ago.

29 years agoDon't let IXOFF or ECHONL stop the setting of TS_CAN_BYPASS_L_RINT. IXOFF
bde [Sat, 29 Jul 1995 08:33:13 +0000 (08:33 +0000)]
Don't let IXOFF or ECHONL stop the setting of TS_CAN_BYPASS_L_RINT.  IXOFF
is handled at a low level, and ECHONL only applies if ICANON is set,
although tty.c sometimes bogusly applies it when ICANON isn't set.

29 years agoRemove /dev/ from socksys link again.
bde [Sat, 29 Jul 1995 05:25:08 +0000 (05:25 +0000)]
Remove /dev/ from socksys link again.

Move `Individual slices' case earlier so that it can be used to handle
sliced floppies.

Remove superfluous `chmod 600's.

Fix formatting of device list.

Don't group setting of umasks with setting of units.

Remove superfluous trailing semicolons.

29 years agoAlways wake up writers after clearing TS_BUSY. This will soon be
bde [Sat, 29 Jul 1995 04:05:57 +0000 (04:05 +0000)]
Always wake up writers after clearing TS_BUSY.  This will soon be
essential when I fix excessive wakeups for output-below-low-water.
In cy.c and sio.c, wake up via the driver start routine to also
eliminate duplicated code involving the clearing of TS_TTSTOP.

Always (except in code to be replaced soon) call driver start
routine directly instead of going through ttstart().

29 years agoNew sound device names.
jkh [Sat, 29 Jul 1995 00:40:40 +0000 (00:40 +0000)]
New sound device names.
Reviewed by: hasty & lowe
Submitted by: pst