unix-history/.git
30 years agoFixed brokeness in the support of the 83C790/Elite Ultra (now that I
David Greenman [Sun, 10 Apr 1994 20:06:28 +0000 (20:06 +0000)]
Fixed brokeness in the support of the 83C790/Elite Ultra (now that I
finally have the f**king documentation!):

1) Changed all the numeric register offsets to symbolic ones (it should
have been this way originally).
2) If 16 bit, disable the shared memory when not using it. Apparantly
switching between 8/16bit mode makes the Ultra unhappy unless
this is done (i.e. it trashes the bus).

30 years agoCorrected a bit of confusion
Poul-Henning Kamp [Sun, 10 Apr 1994 17:58:25 +0000 (17:58 +0000)]
Corrected a bit of confusion

30 years agoFixed a little typo in the man page.
Andreas Schulz [Sun, 10 Apr 1994 12:26:35 +0000 (12:26 +0000)]
Fixed a little typo in the man page.

30 years agoModify wdprobe() to correctly recognise some 2 drive systems,
Geoff Rehmet [Sun, 10 Apr 1994 11:17:13 +0000 (11:17 +0000)]
Modify wdprobe() to correctly recognise some 2 drive systems,
on which it was failing.  Modifications follow ATA, but also
allow some weird setups which seem to contradict ATA (aaarrrrgghhh!!)

30 years agoTruncate hostnames longer than HOST_WIDTH so that
Geoff Rehmet [Sun, 10 Apr 1994 10:42:28 +0000 (10:42 +0000)]
Truncate hostnames longer than HOST_WIDTH so that
rup's output looks nicer.

30 years agoRestore badly commented out MAPLOWER/MAPUPPER
Andrey Chernov [Sun, 10 Apr 1994 01:04:24 +0000 (01:04 +0000)]
Restore badly commented out MAPLOWER/MAPUPPER

30 years agomail a+rwxt changes backed out
Andrey Chernov [Sun, 10 Apr 1994 00:56:12 +0000 (00:56 +0000)]
mail a+rwxt changes backed out

30 years agoCorrect a typo in the manpage. I didn't find anywhere the function supper().
Andreas Schulz [Sat, 9 Apr 1994 17:52:02 +0000 (17:52 +0000)]
Correct a typo in the manpage. I didn't find anywhere the function supper().

30 years agoAdd umask(022), in other case you may got /etc/passwd rw-------
Andrey Chernov [Sat, 9 Apr 1994 17:40:37 +0000 (17:40 +0000)]
Add umask(022), in other case you may got /etc/passwd rw-------
if you have umask 077

30 years agoReplace a usage of gets with fgets.
Andreas Schulz [Sat, 9 Apr 1994 17:18:50 +0000 (17:18 +0000)]
Replace a usage of gets with fgets.

30 years agoCommented out now a suspicious looking MAPLOWER 0xa3 0xa3 and a
Andreas Schulz [Sat, 9 Apr 1994 16:19:53 +0000 (16:19 +0000)]
Commented out now a suspicious looking MAPLOWER 0xa3 0xa3 and a
MAPUPPER 0xb3 0xb3. This is a no-op for me ? ache, can you please
check that , i don't know enough about russian :-).

30 years agoChanged a return(__swalk_sflush) to a return(__swalk_sflush()).
Andreas Schulz [Sat, 9 Apr 1994 15:37:44 +0000 (15:37 +0000)]
Changed a return(__swalk_sflush) to a return(__swalk_sflush()).
This seems the correct call according to the comments in the function
header.

30 years agoAdd prototype functions for flockfile and funlockfile into stdio.h.
Andreas Schulz [Sat, 9 Apr 1994 15:34:53 +0000 (15:34 +0000)]
Add prototype functions for flockfile and funlockfile into stdio.h.
They are used from some macros. This was neccessary to prevent some
compiler warnings.

30 years agoComment out a private INT_MAX/INT_MIN definition to shut up the
Andreas Schulz [Sat, 9 Apr 1994 13:03:39 +0000 (13:03 +0000)]
Comment out a private INT_MAX/INT_MIN definition to shut up the
compiler.

30 years agomake libutil be linked in dynamicly instead of staticly
Geoff Rehmet [Sat, 9 Apr 1994 11:00:30 +0000 (11:00 +0000)]
make libutil be linked in dynamicly instead of staticly

30 years agoAdded RCS id's, little tidy up.
Geoff Rehmet [Fri, 8 Apr 1994 22:23:39 +0000 (22:23 +0000)]
Added RCS id's, little tidy up.

30 years agochange to lptioctl
Geoff Rehmet [Fri, 8 Apr 1994 22:13:49 +0000 (22:13 +0000)]
change to lptioctl

30 years agolink nvi to view (this better be all now)
Geoff Rehmet [Fri, 8 Apr 1994 07:36:06 +0000 (07:36 +0000)]
link nvi to view (this better be all now)

30 years agoadded MAN1+= vi.1, and link to nvi.1
Geoff Rehmet [Fri, 8 Apr 1994 07:16:32 +0000 (07:16 +0000)]
added MAN1+= vi.1, and link to nvi.1
added rcs id.

30 years agoadded locale to Makefile for /usr/src/share
Geoff Rehmet [Fri, 8 Apr 1994 07:03:04 +0000 (07:03 +0000)]
added locale to Makefile for /usr/src/share

30 years agoDon't queue, deliver according to sendmail config.
David Greenman [Thu, 7 Apr 1994 23:42:31 +0000 (23:42 +0000)]
Don't queue, deliver according to sendmail config.

30 years agoFix arguments of CONS_GETINFO
Andrey Chernov [Thu, 7 Apr 1994 23:23:01 +0000 (23:23 +0000)]
Fix arguments of CONS_GETINFO

30 years agoMore EXCLUDE_ options added,
Andrey Chernov [Thu, 7 Apr 1994 22:47:17 +0000 (22:47 +0000)]
More EXCLUDE_ options added,
unneded irq/drq/vector removed

30 years agoRemove unneded irq/drq/vector from snd?
Andrey Chernov [Thu, 7 Apr 1994 22:18:49 +0000 (22:18 +0000)]
Remove unneded irq/drq/vector from snd?

30 years agoWanted to make sure we had an explanation of the memory test stored
Garrett Wollman [Thu, 7 Apr 1994 21:18:59 +0000 (21:18 +0000)]
Wanted to make sure we had an explanation of the memory test stored
somewhere for reference.  Anyone who wants to is welcome to re-write this
to be a ``reall'' document, rather than just a copy of DavidG's message.

30 years agoSkip idle info on uu-lines from uucpd
Andrey Chernov [Thu, 7 Apr 1994 18:38:10 +0000 (18:38 +0000)]
Skip idle info on uu-lines from uucpd

30 years agoNow wall skips uu-lines produced by uucpd.
Andrey Chernov [Thu, 7 Apr 1994 17:55:00 +0000 (17:55 +0000)]
Now wall skips uu-lines produced by uucpd.

30 years agoChange uucp%d lines to uu%d lines, because ut_line
Andrey Chernov [Thu, 7 Apr 1994 17:50:52 +0000 (17:50 +0000)]
Change uucp%d lines to uu%d lines, because ut_line
overflows with big pids, produce incorrect statistics.

30 years agoMake possibility to write utmp entry for non-tty lines,
Andrey Chernov [Thu, 7 Apr 1994 17:47:36 +0000 (17:47 +0000)]
Make possibility to write utmp entry for non-tty lines,
needed especially for uucpd connections

30 years agoo Change fake line from uucp%d to uu%d, because it produce wrong
Andrey Chernov [Thu, 7 Apr 1994 17:43:35 +0000 (17:43 +0000)]
o Change fake line from uucp%d to uu%d, because it produce wrong
  statistics with big pids (size of ut_line is only 8 chars)
o Make possibility to write utmp entries

30 years agoRemove sleep(2) waiting for DTR, now sio can do it himself
Andrey Chernov [Thu, 7 Apr 1994 17:35:52 +0000 (17:35 +0000)]
Remove sleep(2) waiting for DTR, now sio can do it himself

30 years agoChanged a macro name from _Q to _Q2 to avoid a clash with the
Andreas Schulz [Thu, 7 Apr 1994 14:34:25 +0000 (14:34 +0000)]
Changed a macro name from _Q to _Q2 to avoid a clash with the
newly introduced _Q macro in ctype.h.

30 years agoCorrected the Makefile, it has used SRCDIR our macro for that
Andreas Schulz [Thu, 7 Apr 1994 14:09:54 +0000 (14:09 +0000)]
Corrected the Makefile, it has used SRCDIR our macro for that
is CURDIR.
Corrected the syntax in the Russian.koi8-r file.

30 years agoSwap also the undef macros for tolower and toupper :-).
Andreas Schulz [Thu, 7 Apr 1994 13:48:44 +0000 (13:48 +0000)]
Swap also the undef macros for tolower and toupper :-).

30 years agofrom kimmel@varese.cs.umass.edu (Matt Kimmel):
David Greenman [Thu, 7 Apr 1994 12:10:31 +0000 (12:10 +0000)]
from kimmel@varese.cs.umass.edu (Matt Kimmel):
"el" driver for 3COM 3C501. This driver has some serious performance
problems and drops packets on the floor like hot potatos.

30 years agoPatch from Sean Vickery <seanv@cs.uq.edu.au>:
David Greenman [Thu, 7 Apr 1994 11:16:47 +0000 (11:16 +0000)]
Patch from Sean Vickery <seanv@cs.uq.edu.au>:

Description:
Strftime(3) adds garabage to result string when called with "%R" in
the format string.
Repeat-By:
$ date +%R
14:0202:02:08 PM

30 years agolocale support broke tolower and toupper -
Geoff Rehmet [Thu, 7 Apr 1994 10:59:27 +0000 (10:59 +0000)]
locale support broke tolower and toupper -
it appears they were swapped.
This caused (among others) XF86 and elm to barf.

30 years agoMake Bruce happy: silently enter ddb on a BPT or trace trap if ddb is
David Greenman [Thu, 7 Apr 1994 10:51:00 +0000 (10:51 +0000)]
Make Bruce happy: silently enter ddb on a BPT or trace trap if ddb is
configured in the kernel.

30 years agoFixed the isphonogram and isspecial macros, they used the wrong mask.
Andreas Schulz [Thu, 7 Apr 1994 10:20:14 +0000 (10:20 +0000)]
Fixed the isphonogram and isspecial macros, they used the wrong mask.

30 years agoremove trailing space after \.
Paul Richards [Thu, 7 Apr 1994 07:06:46 +0000 (07:06 +0000)]
remove trailing space after \.

30 years agoTset incorrectly assign to TERM first terminal name instead of second
Andrey Chernov [Thu, 7 Apr 1994 01:38:30 +0000 (01:38 +0000)]
Tset incorrectly assign to TERM first terminal name instead of second
from termcap entry.

30 years agoFix error with path like /a/./b (from N. Dudorov)
Andrey Chernov [Thu, 7 Apr 1994 00:30:35 +0000 (00:30 +0000)]
Fix error with path like /a/./b (from N. Dudorov)

30 years agoo Obsoleted defines removed, text cleaned
Andrey Chernov [Wed, 6 Apr 1994 23:40:36 +0000 (23:40 +0000)]
o Obsoleted defines removed, text cleaned
o Now uucpd use logwtmp from libutil instead of own hack
o Now uucpd use syslog for diagnostics
o Now uucpd don't fork, until user logged in

30 years agoUpdate lpt(4) man page to reflect recent changes to the lpt device driver.
Geoff Rehmet [Wed, 6 Apr 1994 20:21:33 +0000 (20:21 +0000)]
Update lpt(4) man page to reflect recent changes to the lpt device driver.
(also point out that lpa is now defunct.)

30 years agoAdded lptcontrol(1)
Geoff Rehmet [Wed, 6 Apr 1994 19:46:19 +0000 (19:46 +0000)]
Added lptcontrol(1)

30 years agoThis commit was generated by cvs2svn to track changes on a CVS vendor
Geoff Rehmet [Wed, 6 Apr 1994 19:38:45 +0000 (19:38 +0000)]
This commit was generated by cvs2svn to track changes on a CVS vendor
branch.

30 years agoAdded lptcontrol(1):
Geoff Rehmet [Wed, 6 Apr 1994 19:38:45 +0000 (19:38 +0000)]
Added lptcontrol(1):
used to change lpt driver between polling/interrupt driven mode

30 years agoChange default permission of /var/mail to 01777, because
Andrey Chernov [Wed, 6 Apr 1994 19:37:45 +0000 (19:37 +0000)]
Change default permission of /var/mail to 01777, because
MUA must have s-bit in other case (security hole).
This fix needed for procmail too, because procmail
uses NFS-locks in /var/mail directory

30 years agoPer Keith Bostic, wchar_t should not be defined in sys/types.h.
Garrett Wollman [Wed, 6 Apr 1994 17:42:07 +0000 (17:42 +0000)]
Per Keith Bostic, wchar_t should not be defined in sys/types.h.

30 years agoFix installation of manual pages. Each page needed locale/ prepended.
Paul Richards [Wed, 6 Apr 1994 16:48:32 +0000 (16:48 +0000)]
Fix installation of manual pages.  Each page needed locale/ prepended.

30 years agoChanges to lpt driver:
Geoff Rehmet [Wed, 6 Apr 1994 16:42:33 +0000 (16:42 +0000)]
Changes to lpt driver:
- ansi prototypes in lpt.c
- a bit of tidying in lpt.c
- ioctl in lpt.c for switching between polling and using interrupts
- added lpt.h - needed for ioctl to allow switching between polling
        and interrupt-driven modes.

30 years ago/bin/sh read builtin causes sh to dump core if read is
Geoff Rehmet [Wed, 6 Apr 1994 13:01:38 +0000 (13:01 +0000)]
/bin/sh read builtin causes sh to dump core if read is
used with either no args, or with an arg in the form "$var"
where var is not defined yet:

freefall:/home/csgr/cvswork/sh%sh
$ read
hi
Segmentation fault (core dumped)
freefall:/home/csgr/cvswork/sh%sh
$ read $foo
foobar
Segmentation fault (core dumped)

30 years agoDon't try to clean out ex since it's a directory.
Paul Richards [Wed, 6 Apr 1994 08:28:56 +0000 (08:28 +0000)]
Don't try to clean out ex since it's a directory.

30 years agoNow -ulogin passed to uucico
Andrey Chernov [Wed, 6 Apr 1994 02:50:06 +0000 (02:50 +0000)]
Now -ulogin passed to uucico

30 years agosplit -b dumps core if no count is given.
Nate Willams [Tue, 5 Apr 1994 23:05:42 +0000 (23:05 +0000)]
split -b dumps core if no count is given.

30 years agoPrevious fix didn't define rune_t like it should. Now it does.
Garrett Wollman [Tue, 5 Apr 1994 22:09:33 +0000 (22:09 +0000)]
Previous fix didn't define rune_t like it should.  Now it does.

30 years agoDon't forget to install octype.h, rune.h, and runetype.h.
Garrett Wollman [Tue, 5 Apr 1994 19:01:55 +0000 (19:01 +0000)]
Don't forget to install octype.h, rune.h, and runetype.h.

30 years agoDocument some of the latest changes from David&John.
Garrett Wollman [Tue, 5 Apr 1994 18:52:11 +0000 (18:52 +0000)]
Document some of the latest changes from David&John.

30 years agoMake stdlib.h in the non-ANSI case work.
Garrett Wollman [Tue, 5 Apr 1994 18:48:03 +0000 (18:48 +0000)]
Make stdlib.h in the non-ANSI case work.

30 years agoo Clean up my previous fix per bde suggestion
Andrey Chernov [Tue, 5 Apr 1994 10:12:41 +0000 (10:12 +0000)]
o Clean up my previous fix per bde suggestion
o Move ptcwakeup before line close
o Make all ttysleep strings unique for debug purposes

30 years agoAdd that /usr/X386/lib is also in the default path for ldconfig. Bug found
Andreas Schulz [Tue, 5 Apr 1994 08:24:19 +0000 (08:24 +0000)]
Add that /usr/X386/lib is also in the default path for ldconfig. Bug found
by Satoshi Asami ( asami@cs.berkeley.edu ). Fixed also a minor typo
problem.

30 years agoRussian locale template added
Andrey Chernov [Tue, 5 Apr 1994 07:14:48 +0000 (07:14 +0000)]
Russian locale template added

30 years agofrom John Dyson:
David Greenman [Tue, 5 Apr 1994 03:23:53 +0000 (03:23 +0000)]
from John Dyson:

1) fixed some bugs related to the bounce buffer code
2) vnode pager now supports clustered pageouts
3) experimental code for clustering all I/O via a new "cldisksort"
4) added >16MB check to Bustek driver
5) made some experimental algorithmic changes to the pageout daemon
6) fixed bugs in truncating mapped files (esp when mapped via NFS)
7) reorganized vnode pager I/O code

30 years agoFirst pass at adding locale support. This code only deals with the LC_CTYPE
Garrett Wollman [Mon, 4 Apr 1994 21:13:55 +0000 (21:13 +0000)]
First pass at adding locale support.  This code only deals with the LC_CTYPE
class of locale data, but could be extended to handle other locale
classes, as well as message catalogues and other non-locale i18n
support.

I have left the old _ctype_ array in place, and moved the ctype.h
header to octype.h, so that existing shared binaries will still be
able to find and use it as they require.

See /usr/src/share/locale for information on how to create new locale
data files (eventually this procedure will be improved).  I'd like to
have a family of locale files for various countries, languages, and
character sets, so please contribute some.

This code was originally written by Paul Borman and contributed to
4.4; I did the integration, and have somewhat tested it.  crt0.c
probably ought to call setlocale() if it doesn't already, but I'd like
for people to create some locale files and try things manually first
before I make every program do this.

30 years agoUse setuid(geteuid()) rather than setruid(geteuid()); seems tto be the
Garrett Wollman [Mon, 4 Apr 1994 19:49:12 +0000 (19:49 +0000)]
Use setuid(geteuid()) rather than setruid(geteuid()); seems tto be the
right thing to do, as this program never un-renounces.

30 years agoAdded back setruid() and setrgiod(), but make them print an obnoxious
Garrett Wollman [Mon, 4 Apr 1994 19:33:58 +0000 (19:33 +0000)]
Added back setruid() and setrgiod(), but make them print an obnoxious
warning message whenever called.  If our linker didn't have this support
broken, it would also print out a warning message at link time.  The warning
message is:

warning: this program uses setruid, which doesn't do anything
(but used to)

30 years agocorrected pointer to sound.doc
Poul-Henning Kamp [Mon, 4 Apr 1994 17:27:45 +0000 (17:27 +0000)]
corrected pointer to sound.doc

30 years agoThis addition allows compilation with EXCLUDE_AUDIO
Andrey Chernov [Sun, 3 Apr 1994 22:16:38 +0000 (22:16 +0000)]
This addition allows compilation with EXCLUDE_AUDIO
without compilation error.

30 years agoUpgraded to Delaware version 3.3p (yes, that's right, they're not increasing).
Garrett Wollman [Sun, 3 Apr 1994 20:37:34 +0000 (20:37 +0000)]
Upgraded to Delaware version 3.3p (yes, that's right, they're not increasing).
This version still doesn't talk to the kernel PLL code, but you should be
able to convince it to do so relatively easily.  Also deleted some junk
files and fixed all compilation warnings.

30 years agoxntp 3.3p from Delaware
Garrett Wollman [Sun, 3 Apr 1994 19:50:51 +0000 (19:50 +0000)]
xntp 3.3p from Delaware

30 years agoThis commit was generated by cvs2svn to track changes on a CVS vendor
Garrett Wollman [Sun, 3 Apr 1994 19:50:51 +0000 (19:50 +0000)]
This commit was generated by cvs2svn to track changes on a CVS vendor
branch.

30 years agoFix two bugs in slave ptys:
Andrey Chernov [Sun, 3 Apr 1994 19:40:01 +0000 (19:40 +0000)]
Fix two bugs in slave ptys:
1) TS_WOPEN not cleared upon exit of ptsopen(), can cause
several problems with ttymodem stuff, etc. (minor bug)
2) TS_ISOPEN not cleared upon exit of ptsclose(), can cause
serious bug with inheritance of previous tty flags on next pty open.
F.e. somethimes triggered echo in telnetd.

30 years agoDocumented MAXMEM. Added warnings about certain options being subject to
Garrett Wollman [Sun, 3 Apr 1994 18:38:48 +0000 (18:38 +0000)]
Documented MAXMEM.  Added warnings about certain options being subject to
extermination in the near future.  Added a pre-formatted `options.doc' file so
Julian Stacey can stop complaining.  Documented PANIC_REBOOT_DELAY or whatever
it's called.  Documented UPDATE_INTERVAL.  Deleted some old and bogus
option references.

30 years agoFix tty? entry in the same way like cua entry
Andrey Chernov [Sun, 3 Apr 1994 14:24:52 +0000 (14:24 +0000)]
Fix tty? entry in the same way like cua entry
Add error diagnotic about unknown device name

30 years agoAdd declaration missing from previous bde's version
Andrey Chernov [Sun, 3 Apr 1994 12:25:57 +0000 (12:25 +0000)]
Add declaration missing from previous bde's version

30 years agoCHANGES from Bruce:
Andrey Chernov [Sun, 3 Apr 1994 11:41:11 +0000 (11:41 +0000)]
CHANGES from Bruce:
---
This list of changes is in approximately chronological order (oldest first).

o Many cosmetic changes - renamed comintr1 -> siointr1, moved
  things around and fixed whitespace.
o Reduced SLIP latency (FRAME_END hack) from 20-30 ms to 16 ms
  at 115200 bps (you won't notice the average 10 ms improvement
  on slow lines).  ppp seems to use only counted transfers so
  there's no similar hack available.  It's too hard for the
  driver to know the count.
o Temporary #ifdefs for new and old interrupt handling
  (OLD_INTERRUPT_HANDLING decided by setsofttty() not being
  externally defined.
o Don't test for the IIR_NOPEND bit being set - test for the
  non-fifo part of the iir equalling it like the docs say to.
  States with other IIR_NOPEND set in combination with the
  other iir bits are undefined.  The docs may be stupid - the
  old test would not have broken when the fifo bits were
  introduced.
o Noted more problems with DTR wait.
o Rewrote console stuff.  Still some initialization and state
  preservation problems.  Same for kgdb stuff.  The driver
  doesn't do anything about the console close bug.  It needs
  to be fixed entirely in i386/cons.c.  I like chmr's version
  where the the console driver revectors the device open and
  close routines.
o Temporary (?) #ifdefs for references to tty buffers.
o Noted further things to do in (2 comments about 3 places) for
  phk's change to not touch RTS unless it is being used for flow
  control.
o Temporary #ifdefs for timestamp handling.  It needs fixing.
  The microtime() call breaks the first rule of writing fast
  interrupt handlers: NO calls to functions that might do slow
  and bad things.  microtime() enables interrupts.  This turns
  out to be only moderately harmful.  Also, I want the timestamp
  copy outside of the normal interrupt handler.
o Don't init com->tp early for the !DONT_MALLOC_TTYS case -
  both sides are NULL.
o Worry about com->tp == NULL in siopoll.  I don't see how you
  survived the (incc <= 0 || !(tp->state & TS_ISOPEN)) test.
  Perhaps early sttys or comcontrols set up the tp's for _all_
  the ports before this code is reached.

30 years agoNow cua handled like tty, i.e. name+3
Andrey Chernov [Sun, 3 Apr 1994 05:13:59 +0000 (05:13 +0000)]
Now cua handled like tty, i.e. name+3

30 years agoNow cua handled like tty, i.e. name+3
Andrey Chernov [Sun, 3 Apr 1994 05:06:29 +0000 (05:06 +0000)]
Now cua handled like tty, i.e. name+3

30 years agoFix error in making cua entries for two multiports
Andrey Chernov [Sun, 3 Apr 1994 04:39:54 +0000 (04:39 +0000)]
Fix error in making cua entries for two multiports
(> 10 devices)

30 years agoAdd RCS Id to Makefile.
Garrett Wollman [Sun, 3 Apr 1994 00:28:10 +0000 (00:28 +0000)]
Add RCS Id to Makefile.

30 years agoAdded locale files and mklocale utility from Paul Borman's `runes'
Garrett Wollman [Sun, 3 Apr 1994 00:27:38 +0000 (00:27 +0000)]
Added locale files and mklocale utility from Paul Borman's `runes'
library.  Can't be enabled until the library itself is in libc.

30 years agoBarry Morris's (bsm@freefall) changes to su.
Jordan K. Hubbard [Sun, 3 Apr 1994 00:16:36 +0000 (00:16 +0000)]
Barry Morris's (bsm@freefall) changes to su.

30 years ago Add a little extra sanity checking so that end profil requests
Jordan K. Hubbard [Sat, 2 Apr 1994 23:03:01 +0000 (23:03 +0000)]
 Add a little extra sanity checking so that end profil requests
 don't generate spurious faults.  I'm not sure how long this has been broken,
 but profiling sure won't work as it is before this patch.

30 years agoCompletely delete setr[ug]id, since they no longer do anything any more.
Garrett Wollman [Sat, 2 Apr 1994 22:25:08 +0000 (22:25 +0000)]
Completely delete setr[ug]id, since they no longer do anything any more.
Bump shared-library minor version to reflect this change.

30 years agoDocument new interrupt code.
Garrett Wollman [Sat, 2 Apr 1994 22:15:18 +0000 (22:15 +0000)]
Document new interrupt code.

30 years agoChange from Bruce:
Andrey Chernov [Sat, 2 Apr 1994 20:43:25 +0000 (20:43 +0000)]
Change from Bruce:
isa_dmarangecheck() is off by one error.
> ISARAM_END should be >= ISARAM_END. Only the first page above 16M
was mishandled.

30 years agoNewsgroups: comp.unix.bsd
Jordan K. Hubbard [Sat, 2 Apr 1994 18:22:06 +0000 (18:22 +0000)]
Newsgroups: comp.unix.bsd
From: david@wubios.wustl.edu (David J Camp)
Subject: patch for FreeBSD man page
Organization: Division of Biostatistics, WUMS, St. Louis, MO
Date: Sat, 2 Apr 1994 15:15:52 GMT

I found a minor but confusing error in one of the FreBSD man pages.
Here is a patch:  -David-

30 years agoBack out previous change.
Jordan K. Hubbard [Sat, 2 Apr 1994 17:10:19 +0000 (17:10 +0000)]
Back out previous change.

30 years agoChanged _profiling back to profiling.
David Greenman [Sat, 2 Apr 1994 08:39:20 +0000 (08:39 +0000)]
Changed _profiling back to profiling.

30 years agoOutput *_imask instead of *mask.
David Greenman [Sat, 2 Apr 1994 07:09:28 +0000 (07:09 +0000)]
Output *_imask instead of *mask.

30 years ago New interrupt code from Bruce Evans. In additional to Bruce's attached
David Greenman [Sat, 2 Apr 1994 07:01:04 +0000 (07:01 +0000)]
   New interrupt code from Bruce Evans. In additional to Bruce's attached
list of changes, I've made the following additional changes:

1) i386/include/ipl.h renamed to spl.h as the name conflicts with the
   file of the same name in i386/isa/ipl.h.
2) changed all use of *mask (i.e. netmask, biomask, ttymask, etc) to
   *_imask (net_imask, etc).
3) changed vestige of splnet use in if_is to splimp.
4) got rid of "impmask" completely (Bruce had gotten rid of netmask),
   and are now using net_imask instead.
5) dozens of minor cruft to glue in Bruce's changes.

   These require changes I made to config(8) as well, and thus it must
be rebuilt.

-DG

from Bruce Evans:

sio:
o No diff is supplied.  Remove the define of setsofttty().  I hope
  that is enough.

*.s:
o i386/isa/debug.h no longer exists.  The event counters became too
  much trouble to maintain.  All function call entry and exception
  entry counters can be recovered by using profiling kernel (the new
  profiling supports all entry points; however, it is too slow to
  leave enabled all the time; it also).  Only BDBTRAP() from debug.h
  is now used.  That is moved to exception.s.  It might be worth
  preserving SHOW_BITS() and calling it from _mcount() (if enabled).
o T_ASTFLT is now only set just before calling trap().
o All exception handlers set SWI_AST_MASK in cpl as soon as possible
  after entry and arrange for _doreti to restore it atomically with
  exiting.  It is not possible to set it atomically with entering
  the kernel, so it must be checked against the user mode bits in
  the trap frame before committing to using it.  There is no place
  to store the old value of cpl for syscalls or traps, so there are
  some complications restoring it.

Profiling stuff (mostly in *.s):
o Changes to kern/subr_mcount.c, gcc and gprof are not supplied yet.
o All interesting labels `foo' are renamed `_foo' and all
  uninteresting labels `_bar' are renamed `bar'.  A small change
  to gprof allows ignoring labels not starting with underscores.
o MCOUNT_LABEL() is to provide names for counters for times spent
  in exception handlers.
o FAKE_MCOUNT() is a version of MCOUNT() suitable for exception
  handlers.  Its arg is the pc where the exception occurred.  The
  new mcount() pretends that this was a call from that pc to a
  suitable MCOUNT_LABEL().
o MEXITCOUNT is to turn off any timer started by MCOUNT().

/usr/src/sys/i386/i386/exception.s:
o The non-BDB BPTTRAP() macros were doing a sti even when interrupts
  were disabled when the trap occurred.  The sti (fixed) sti is
  actually a no-op unless you have my changes to machdep.c that make
  the debugger trap gates interrupt gates, but fixing that would
  make the ifdefs messier.  ddb seems to be unharmed by both
  interrupts always disabled and always enabled (I had the branch in
  the fix back to front for some time :-().
o There is no known pushal bug.
o tf_err can be left as garbage for syscalls.

/usr/src/sys/i386/i386/locore.s:
o Fix and update BDE_DEBUGGER support.
o ENTRY(btext) before initialization was dangerous.
o Warm boot shot was longer than intended.

/usr/src/sys/i386/i386/machdep.c:
o DON'T APPLY ALL OF THIS DIFF.  It's what I'm using, but may require
  other changes.
  Use the following:
o Remove aston() and setsoftclock().
  Maybe use the following:
o No netisr.h.
o Spelling fix.
o Delay to read the Rebooting message.
o Fix for vm system unmapping a reduced area of memory
  after bounds_check_with_label() reduces the size of
  a physical i/o for a partition boundary.  A similar
  fix is required in kern_physio.c.
o Correct use of __CONCAT.  It never worked here for non-
  ANSI cpp's.  Is it time to drop support for non-ANSI?
o gdt_segs init.  0xffffffffUL is bogus because ssd_limit
  is not 32 bits.  The replacement may have the same
  value :-), but is more natural.
o physmem was one page too low.  Confusing variable names.
  Don't use the following:
o Better numbers of buffers.  Each 8K page requires up to
  16 buffer headers.  On my system, this results in 5576
  buffers containing [up to] 2854912 bytes of memory.
  The usual allocation of about 384 buffers only holds
  192K of disk if you use it on an fs with a block size
  of 512.
o gdt changes for bdb.
o *TGT -> *IDT changes for bdb.
o #ifdefed changes for bdb.

/usr/src/sys/i386/i386/microtime.s:
o Use the correct asm macros.  I think asm.h was copied from Mach
  just for microtime and isn't used now.  It certainly doesn't
  belong in <sys>.  Various macros are also duplicated in
  sys/i386/boot.h and libc/i386/*.h.
o Don't switch to and from the IRR; it is guaranteed to be selected
  (default after ICU init and explicitly selected in isa.c too, and
  never changed until the old microtime clobbered it).

/usr/src/sys/i386/i386/support.s:
o Non-essential changes (none related to spls or profiling).
o Removed slow loads of %gs again.  The LDT support may require
  not relying on %gs, but loading it is not the way to fix it!
  Some places (copyin ...) forgot to load it.  Loading it clobbers
  the user %gs.  trap() still loads it after certain types of
  faults so that fuword() etc can rely on it without loading it
  explicitly.  Exception handlers don't restore it.  If we want
  to preserve the user %gs, then the fastest method is to not
  touch it except for context switches.  Comparing with
  VM_MAXUSER_ADDRESS and branching takes only 2 or 4 cycles on
  a 486, while loading %gs takes 9 cycles and using it takes
  another.
o Fixed a signed branch to unsigned.

/usr/src/sys/i386/i386/swtch.s:
o Move spl0() outside of idle loop.
o Remove cli/sti from idle loop.  sw1 does a cli, and in the
  unlikely event of an interrupt occurring and whichqs becoming
  zero, sw1 will just jump back to _idle.
o There's no spl0() function in asm any more, so use splz().
o swtch() doesn't need to be superaligned, at least with the
  new mcounting.
o Fixed a signed branch to unsigned.
o Removed astoff().

/usr/src/sys/i386/i386/trap.c:
o The decentralized extern decls were inconsistent, of course.
o Fixed typo MATH_EMULTATE in comments. */
o Removed unused variables.
o Old netmask is now impmask; print it instead.  Perhaps we
  should print some of the new masks.
o BTW, trap() should not print anything for normal debugger
  traps.

/usr/src/sys/i386/include/asmacros.h:
o DON'T APPLY ALL OF THIS DIFF.  Just use some of the null macros
  as necessary.

/usr/src/sys/i386/include/cpu.h:
o CLKF_BASEPRI() changes since cpl == SWI_AST_MASK is now normal
  while the kernel is running.
o Don't use var++ to set boolean variables.  It fails after a mere
  4G times :-) and is slower than storing a constant on [3-4]86s.

/usr/src/sys/i386/include/cpufunc.h:
o DON'T APPLY ALL OF THIS DIFF.  You need mainly the include of
  <machine/ipl.h>.  Unfortunately, <machine/ipl.h> is needed by
  almost everything for the inlines.

/usr/src/sys/i386/include/ipl.h:
o New file.  Defines spl inlines and SWI macros and declares most
  variables related to hard and soft interrupt masks.

/usr/src/sys/i386/isa/icu.h:
o Moved definitions to <machine/ipl.h>

/usr/src/sys/i386/isa/icu.s:
o Software interrupts (SWIs) and delayed hardware interrupts (HWIs)
  are now handled uniformally, and dispatching them from splx() is
  more like dispatching them from _doreti.  The dispatcher is
  essentially *(handler[ffs(ipending & ~cpl)]().
o More care (not quite enough) is taken to avoid unbounded nesting
  of interrupts.
o The interface to softclock() is changed so that a trap frame is
  not required.
o Fast interrupt handlers are now handled more uniformally.
  Configuration is still too early (new handlers would require
  bits in <machine/ipl.h> and functions to vector.s).
o splnnn() and splx() are no longer here; they are inline functions
  (could be macros for other compilers).  splz() is the nontrivial
  part of the old splx().

/usr/src/sys/i386/isa/ipl.h
o New file.  Supposed to have only bus-dependent stuff.  Perhaps
  the h/w masks should be declared here.

/usr/src/sys/i386/isa/isa.c:
o DON'T APPLY ALL OF THIS DIFF.  You need only things involving
  *mask and *MASK and comments about them.  netmask is now a pure
  software mask.  It works like the softclock mask.

/usr/src/sys/i386/isa/vector.s:
o Reorganize AUTO_EOI* macros.
o Option FAST_INTR_HANDLER_USERS_ES for people who don't trust
  fastintr handlers.
o fastintr handlers need to metamorphose into ordinary interrupt
  handlers if their SWI bit has become set.  Previously, sio had
  unintended latency for handling output completions and input
  of SLIP framing characters because this was not done.

/usr/src/sys/net/netisr.h:
o The machine-dependent stuff is now imported from <machine/ipl.h>.

/usr/src/sys/sys/systm.h
o DON'T APPLY ALL OF THIS DIFF.  You need mainly the different
  splx() prototype.  The spl*() prototypes are duplicated as
  inlines in <machine/ipl.h> but they need to be duplicated here
  in case there are no inlines.  I sent systm.h and cpufunc.h
  to Garrett.  We agree that spl0 should be replaced by splnone
  and not the other way around like I've done.

/usr/src/sys/kern/kern_clock.c
o splsoftclock() now lowers cpl so the direct call to softclock()
  works as intended.
o softclock() interface changed to avoid passing the whole frame
  (some machines may need another change for profile_tick()).
o profiling renamed _profiling to avoid ANSI namespace pollution.
  (I had to improve the mcount() interface and may as well fix it.)
  The GUPROF variant doesn't actually reference profiling here,
  but the 'U' in GUPROF should mean to select the microtimer
  mcount() and not change the interface.

30 years agoRenamed executable to avoid name-clash with directory
Andrew Moore [Sat, 2 Apr 1994 06:19:12 +0000 (06:19 +0000)]
Renamed executable to avoid name-clash with directory

30 years agoReplace CAPS led with ALTGR led for ALTGR mode (soft) keyboards,
Andrey Chernov [Fri, 1 Apr 1994 18:33:12 +0000 (18:33 +0000)]
Replace CAPS led with ALTGR led for ALTGR mode (soft) keyboards,
currently affects only russian keyboard.

30 years agoChange got_status/modem_status assignment per Bruce suggestion,
Andrey Chernov [Fri, 1 Apr 1994 16:47:01 +0000 (16:47 +0000)]
Change got_status/modem_status assignment per Bruce suggestion,
because inb clears modem status port.

30 years agoLocalize previous changes (removing dma_reset)
Andrey Chernov [Fri, 1 Apr 1994 15:04:49 +0000 (15:04 +0000)]
Localize previous changes (removing dma_reset)
only for SB card, it seems that it breaks GUS at least.

30 years agoFix from Gene for bogus destinations.
Jordan K. Hubbard [Fri, 1 Apr 1994 13:11:01 +0000 (13:11 +0000)]
Fix from Gene for bogus destinations.

30 years agoFix bug with ppp.h I introduced earlier. This should do it.
Jordan K. Hubbard [Fri, 1 Apr 1994 10:13:22 +0000 (10:13 +0000)]
Fix bug with ppp.h I introduced earlier.  This should do it.

30 years agoTracker "swapping" problem finally solved,
Andrey Chernov [Thu, 31 Mar 1994 20:54:25 +0000 (20:54 +0000)]
Tracker "swapping" problem finally solved,
all (expect one) dma_reset calls commented out.

30 years agoMinor performance improvements from J.T.Conklin and Charles Hannum.
David Greenman [Thu, 31 Mar 1994 14:11:02 +0000 (14:11 +0000)]
Minor performance improvements from J.T.Conklin and Charles Hannum.

30 years agocorrected comment.
David Greenman [Thu, 31 Mar 1994 11:06:29 +0000 (11:06 +0000)]
corrected comment.