]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
27 years ago$FreeBSD$ -> $Id$.
bde [Tue, 1 Jul 1997 16:20:56 +0000 (16:20 +0000)]
$FreeBSD$ -> $Id$.
Removed Berkeley Id.  We never did this like Berkeley (because cvs doesn't
handle symlinks) and no interesting Berkeley words are left.

27 years agoDo not use argv[0] in usage().
charnier [Tue, 1 Jul 1997 06:37:34 +0000 (06:37 +0000)]
Do not use argv[0] in usage().

27 years agoAdd usage(), rcsid. Use getopt().
charnier [Tue, 1 Jul 1997 06:34:31 +0000 (06:34 +0000)]
Add usage(), rcsid. Use getopt().
Obtained from: OpenBSD.

27 years agoChange the naming conventions for preserve'd files and also document
jkh [Tue, 1 Jul 1997 06:13:51 +0000 (06:13 +0000)]
Change the naming conventions for preserve'd files and also document
the option in pkg_create.  Now preserved files start with a . and are
named .<filename>.<pkgname>.backup so that their purpose is more clear.

Note that just using the preserve option without proper pkg_deps
is also foolish since packages being deleted in the incorrect order with
preserve on can generate some odd results.

27 years agoFix a bug (apparently very old) that can cause a TCP connection to
jdp [Tue, 1 Jul 1997 05:42:16 +0000 (05:42 +0000)]
Fix a bug (apparently very old) that can cause a TCP connection to
be dropped when it has an unusual traffic pattern.  For full details
as well as a test case that demonstrates the failure, see the
referenced PR.

Under certain circumstances involving the persist state, it is
possible for the receive side's tp->rcv_nxt to advance beyond its
tp->rcv_adv.  This causes (tp->rcv_adv - tp->rcv_nxt) to become
negative.  However, in the code affected by this fix, that difference
was interpreted as an unsigned number by max().  Since it was
negative, it was taken as a huge unsigned number.  The effect was
to cause the receiver to believe that its receive window had negative
size, thereby rejecting all received segments including ACKs.  As
the test case shows, this led to fruitless retransmissions and
eventually to a dropped connection.  Even connections using the
loopback interface could be dropped.  The fix substitutes the signed
imax() for the unsigned max() function.

PR: closes kern/3998
Reviewed by: davidg, fenner, wollman

27 years agoAllow distSetByName() to process all the relevant entries.
jkh [Tue, 1 Jul 1997 03:50:58 +0000 (03:50 +0000)]
Allow distSetByName() to process all the relevant entries.
Submitted by: msmith

27 years agoUn-inline a call to spl0(). It is not time critical, and was only inline
bde [Tue, 1 Jul 1997 01:34:30 +0000 (01:34 +0000)]
Un-inline a call to spl0().  It is not time critical, and was only inline
because there was no non-inline spl0() to call.

Don't frob intr_nesting_level in idle() or cpu_switch().  Interrupts
are mostly disabled then, so the frobbing had little effect.

27 years agoAdded ifdefs so that this compiles when neither I586_CPU nor I586_CPU
bde [Tue, 1 Jul 1997 01:18:38 +0000 (01:18 +0000)]
Added ifdefs so that this compiles when neither I586_CPU nor I586_CPU
is defined, or SMP is defined.  It is silly to configure PERFMON when
it can't work (it will be disabled at runtime), but I like to leave
the PERFMON configuration alone when I temporarily disable support for
modern CPUs to run regression tests.

Removed an unused #include.

27 years agoUn-inline a call to spl0(). It is not time critical, and was only inline
bde [Tue, 1 Jul 1997 01:02:03 +0000 (01:02 +0000)]
Un-inline a call to spl0().  It is not time critical, and was only inline
because there was no non-inline spl0() to call.

27 years agoSome staticized variables were still declared to be extern.
bde [Tue, 1 Jul 1997 00:54:37 +0000 (00:54 +0000)]
Some staticized variables were still declared to be extern.

27 years agoRemoved extra definition of constty. It is defined in subr_prf.c.
bde [Tue, 1 Jul 1997 00:52:37 +0000 (00:52 +0000)]
Removed extra definition of constty.  It is defined in subr_prf.c.

27 years agoDon't cast function pointers to (void *). This will cause warnings.
bde [Tue, 1 Jul 1997 00:45:45 +0000 (00:45 +0000)]
Don't cast function pointers to (void *).  This will cause warnings.
They should be fixed when similar warnings for the general interrupt
attach routines are fixed.

Removed unused #include.

27 years agoDon't depend on <machine/cpufunc.h> including <sys/types.h>. In fact,
bde [Tue, 1 Jul 1997 00:33:42 +0000 (00:33 +0000)]
Don't depend on <machine/cpufunc.h> including <sys/types.h>.  In fact,
don't include this kernel-only header at all, since we don't need it.

27 years agoDon't depend <machine/cpufunc.h> including <sys/types.h>.
bde [Tue, 1 Jul 1997 00:29:33 +0000 (00:29 +0000)]
Don't depend <machine/cpufunc.h> including <sys/types.h>.

27 years agoDon't depend on gcc's feature of permitting labels that aren't followed
bde [Tue, 1 Jul 1997 00:22:51 +0000 (00:22 +0000)]
Don't depend on gcc's feature of permitting labels that aren't followed
by a statement.

27 years agoEnabled some SMP options. LINT is for testing that all code compiles
bde [Tue, 1 Jul 1997 00:14:39 +0000 (00:14 +0000)]
Enabled some SMP options.  LINT is for testing that all code compiles
cleanly, so only negative options should be commented out.  Options
should have non-default values.

27 years agoDon't depend on gcc's feature of permitting returning void expressions
bde [Tue, 1 Jul 1997 00:08:34 +0000 (00:08 +0000)]
Don't depend on gcc's feature of permitting returning void expressions
in functions returning void.

27 years agoSome staticized variables were still declared to be extern.
bde [Mon, 30 Jun 1997 23:54:50 +0000 (23:54 +0000)]
Some staticized variables were still declared to be extern.

27 years agoDon't depend on gcc's feature of interpreting `int foo(c) char c; ...'
bde [Mon, 30 Jun 1997 23:49:17 +0000 (23:49 +0000)]
Don't depend on gcc's feature of interpreting `int foo(c) char c; ...'
as `int foo(char c) ...' if there is a bogus prototype `int foo(char c);'
in scope.

27 years agoRemoved temporary SMP header fix.
bde [Mon, 30 Jun 1997 23:37:54 +0000 (23:37 +0000)]
Removed temporary SMP header fix.

27 years agoRemoved -I path to gdb's readline. This didn't do much in -current
bde [Mon, 30 Jun 1997 23:10:54 +0000 (23:10 +0000)]
Removed -I path to gdb's readline.  This didn't do much in -current
because the -I path to config/libreadline had precedence, but in
2.2 it help hide the bug that the -I path to the non-contrib
libreadline was garbage.

27 years agosysconfig -> rc.conf
jkh [Mon, 30 Jun 1997 19:10:50 +0000 (19:10 +0000)]
sysconfig -> rc.conf
PR: 3978
Submitted by: Janick.Taillandier@ratp.fr

27 years agooptions.i386:
yokota [Mon, 30 Jun 1997 14:37:43 +0000 (14:37 +0000)]
options.i386:
- Added the psm options PSM_HOOKAPM and PSM_RESETAFTERSUSPEND.

LINT:
- Added the psm options PSM_HOOKAPM and PSM_RESETAFTERSUSPEND.
- Added comments on the flag 0x20 for syscons.
- Clarified descriptions on the flags (0x02, 0x04) regarding the cursor
  shape in syscons.

27 years agoFixes annoying behavior and a bug regarding the destructive
yokota [Mon, 30 Jun 1997 13:31:49 +0000 (13:31 +0000)]
Fixes annoying behavior and a bug regarding the destructive
cursor (CHAR_CURSOR)

1. Reduced the number of calls to set_destructive_cursor().  The
destructive cursor produced noticeable overhead on the system.  It was
caused by draw_cursor_image() calling set_destructive_cursor() every
so often.

set_destructive_cursor() absolutely needs to be called when

a) the character code under the cursor has changed either because
   the cursor moved or because the screen was updated or the mouse
   pointer overlapped the cursor.
b) Or a new font has been loaded,
c) or the video mode has been changed,
d) or the cursor shape has been changed,
e) or the user switched virtual consoles.

2. Turn off the configuration flag CHAR_CURSOR (destructive cursor) in
scattach() if we have a non-VGA card.  The destructive cursor works
only for VGA.

3. Removed redundant calls to set_destructive_cursor() in some places.

4. Fixed the "disappearing mouse pointer" problem. The mouse pointer
looked hidden under the destructive cursor when it overlaped the cursor.

A slightly different version of the patch was reviewd and OKed by
sos and ache.

27 years agogdb.info still needs a near-copy of hsuser.texinfo to avoid the
bde [Mon, 30 Jun 1997 12:57:58 +0000 (12:57 +0000)]
gdb.info still needs a near-copy of hsuser.texinfo to avoid the
reference to the programming manual.  Use this near-copy of the version
of hsuser.texinfo in contrib/libreadline instead of the stale near-copy
in contrib/gdb.

27 years agoAdd experimental APM support for some laptops.
yokota [Mon, 30 Jun 1997 12:52:57 +0000 (12:52 +0000)]
Add experimental APM support for some laptops.

If the configuration option PSM_HOOKAPM is defined and the APM device
is available, the psm driver will issue the ENABLE command to the
pointing device at the resume APM event if the device was open when
the system went into suspended mode. If the option
PSM_RESETAFTERSUSPEND is specified in addition to PSM_HOOKAPM, the
driver will try to reset the pointing device before sending the
ENABLE command.

Built-in PS/2-type pointing devices in some laptops (all the reports I
heard were about Toshiba models) sometimes don't work immediately
after the system is resumed. The device MAY become available after a
while. The system may exhibit the same symptom in other OS's too
(no, FreeBSD is not the only OS that is suffering :-).

I don't know the correct way of solving this yet, but it's been
reported that issuing the ENABLE command after resumption wakes up the
pointing device.

Without PSM_HOOKAPM, the psm driver behaves in the same way as before.

Problem reported in the bsd-nomads mailing list in Japan.

27 years agoUse EX_USAGE in usage().
charnier [Mon, 30 Jun 1997 11:24:18 +0000 (11:24 +0000)]
Use EX_USAGE in usage().

27 years agoRemove lines inside #ifdef 0/#endif.
charnier [Mon, 30 Jun 1997 11:10:18 +0000 (11:10 +0000)]
Remove lines inside #ifdef 0/#endif.

27 years agoAdd LC_TIME=C to date
ache [Mon, 30 Jun 1997 10:59:25 +0000 (10:59 +0000)]
Add LC_TIME=C to date

27 years agoSynchronize with sys/i386/isa/syscons.c and syscons.h revisions 1.219
kato [Mon, 30 Jun 1997 10:00:47 +0000 (10:00 +0000)]
Synchronize with sys/i386/isa/syscons.c and syscons.h revisions 1.219
and 1.30, respectively.

27 years agoSynchronize with sys/i386/i386/machdep.c revision 1.252.
kato [Mon, 30 Jun 1997 09:58:39 +0000 (09:58 +0000)]
Synchronize with sys/i386/i386/machdep.c revision 1.252.

27 years agoSynchronize with sys/i386/conf/Makefile.i386 and files.i386 revisions
kato [Mon, 30 Jun 1997 09:57:33 +0000 (09:57 +0000)]
Synchronize with sys/i386/conf/Makefile.i386 and files.i386 revisions
1.100 and 1.166, respectively.

27 years agoAdd usage().
charnier [Mon, 30 Jun 1997 06:51:20 +0000 (06:51 +0000)]
Add usage().

27 years agoDocument -k flag.
charnier [Mon, 30 Jun 1997 06:50:07 +0000 (06:50 +0000)]
Document -k flag.

27 years agoUse err(3) instead of local redefinition. Change argument name in man
charnier [Mon, 30 Jun 1997 06:48:46 +0000 (06:48 +0000)]
Use err(3) instead of local redefinition. Change argument name in man
page to sync with usage string.

27 years agoAdd usage string. Typo in man page.
charnier [Mon, 30 Jun 1997 06:45:53 +0000 (06:45 +0000)]
Add usage string. Typo in man page.

27 years agoUncompress is under .Nm not .Ar.
charnier [Mon, 30 Jun 1997 06:44:07 +0000 (06:44 +0000)]
Uncompress is under .Nm not .Ar.

27 years agoMdoc'ify man page. Use err(3). Remove whoami.
charnier [Mon, 30 Jun 1997 06:42:43 +0000 (06:42 +0000)]
Mdoc'ify man page. Use err(3). Remove whoami.

27 years agoUse err(3). Add DIAGNOSTIC section name in man page.
charnier [Mon, 30 Jun 1997 06:40:16 +0000 (06:40 +0000)]
Use err(3). Add DIAGNOSTIC section name in man page.

27 years agoAdd missing -s flag in usage string.
charnier [Mon, 30 Jun 1997 06:37:15 +0000 (06:37 +0000)]
Add missing -s flag in usage string.

27 years agoYAMF22 - deletes and preserve option fixes.
jkh [Mon, 30 Jun 1997 03:15:40 +0000 (03:15 +0000)]
YAMF22 - deletes and preserve option fixes.

27 years agoAllow specification of a umask for local socket
brian [Mon, 30 Jun 1997 03:03:38 +0000 (03:03 +0000)]
Allow specification of a umask for local socket
creation in "set server" command.

27 years agoWhoops, only attempt to back up the file if it exists. :-}
jkh [Mon, 30 Jun 1997 02:57:40 +0000 (02:57 +0000)]
Whoops, only attempt to back up the file if it exists. :-}

27 years agoDelete old comment about /bin/mail. We don't have a /bin/mail.
wosch [Sun, 29 Jun 1997 23:09:07 +0000 (23:09 +0000)]
Delete old comment about /bin/mail. We don't have a /bin/mail.

Sync with /etc/master.passwd and RFC 2142. In detail:
added:
man, news, operator, usenet (->news), xten
deleted:
ingres
added as comment:
info, marketing, sales, support, abuse, noc, security,
ftp, hostmaster, webmaster, www, bit-bucket, dev-null

27 years agoImplemented `-c command'.
bde [Sun, 29 Jun 1997 22:43:01 +0000 (22:43 +0000)]
Implemented `-c command'.

Fixed bitrot (__dead went away; EOF is now wrong for the getopt failure
value).

Moved sleep command to the end of the main loop to avoid mismatch between
main loop and the report loop.  There is an extra iteration that could be
used to calibrate the loop overhead, but was used to report wrong results.

Fixed usage message.

27 years agodev->id_flags --> flags
ache [Sun, 29 Jun 1997 22:23:32 +0000 (22:23 +0000)]
dev->id_flags --> flags
(not compiles otherwise)

27 years agoConfigure static linkage in the normal way (default to -static using
bde [Sun, 29 Jun 1997 21:40:54 +0000 (21:40 +0000)]
Configure static linkage in the normal way (default to -static using
NOSHARED?= yes).

27 years agoConfigure static linkage in the normal way (default to -static using
bde [Sun, 29 Jun 1997 21:36:33 +0000 (21:36 +0000)]
Configure static linkage in the normal way (default to -static using
NOSHARED?= yes).

Split long lines.

27 years agoPrint fs_maxfilesize.
bde [Sun, 29 Jun 1997 20:59:03 +0000 (20:59 +0000)]
Print fs_maxfilesize.

27 years agoPrepare to remove the declaration of sys_siglist from <unistd.h>.
bde [Sun, 29 Jun 1997 20:38:38 +0000 (20:38 +0000)]
Prepare to remove the declaration of sys_siglist from <unistd.h>.
It should only be declared in <signal.h> if at all.  Modload used
to depend on getting it from <unistd.h> because it only included
<sys/signal.h>.

27 years agoZap register keyword usage and convert: bcopy -> memmove, bzero -> memset,
steve [Sun, 29 Jun 1997 19:50:26 +0000 (19:50 +0000)]
Zap register keyword usage and convert: bcopy -> memmove, bzero -> memset,
index -> strchr, and rindex -> strrchr.

27 years agoMerge conflicts and make this compile -Wall clean.
steve [Sun, 29 Jun 1997 19:00:29 +0000 (19:00 +0000)]
Merge conflicts and make this compile -Wall clean.

27 years agoThis commit was generated by cvs2svn to compensate for changes in r27074,
steve [Sun, 29 Jun 1997 18:50:34 +0000 (18:50 +0000)]
This commit was generated by cvs2svn to compensate for changes in r27074,
which included commits to RCS files with non-trunk default branches.

27 years agoImport of NetBSD's rbootd version 19970629
steve [Sun, 29 Jun 1997 18:50:34 +0000 (18:50 +0000)]
Import of NetBSD's rbootd version 19970629

27 years agoRemoved most unused files (about 8M total).
bde [Sun, 29 Jun 1997 18:12:14 +0000 (18:12 +0000)]
Removed most unused files (about 8M total).

27 years agoProtect against CVSROOT not being set.
jkh [Sun, 29 Jun 1997 18:07:01 +0000 (18:07 +0000)]
Protect against CVSROOT not being set.
Submitted by: fenner

27 years agoRemoved unused files (about 8MB total).
bde [Sun, 29 Jun 1997 18:06:40 +0000 (18:06 +0000)]
Removed unused files (about 8MB total).

27 years agoUpdated generated files (makesyscalls.sh changed). Only sysproto.h
bde [Sun, 29 Jun 1997 17:47:32 +0000 (17:47 +0000)]
Updated generated files (makesyscalls.sh changed).  Only sysproto.h
really changed.

27 years agoDon't generate unused nested #include of <sys/aio.h>.
bde [Sun, 29 Jun 1997 17:39:57 +0000 (17:39 +0000)]
Don't generate unused nested #include of <sys/aio.h>.

27 years agoFixed the fix for not using -fomit-frame-pointer with -pg. The previous
bde [Sun, 29 Jun 1997 16:39:11 +0000 (16:39 +0000)]
Fixed the fix for not using -fomit-frame-pointer with -pg.  The previous
fix stopped it being used in all cases, because substitution on unset
variables does not work.

When profiling, put -malign-functions=4 in CFLAGS instead of in PROF.
This fixes the histogram counts for profiling support functions.  It
gives bogus but harmless extra alignment for genassym etc.

27 years agoA fix/work-around for ThinkPad 535.
yokota [Sun, 29 Jun 1997 15:11:40 +0000 (15:11 +0000)]
A fix/work-around for ThinkPad 535.

Add a new configuration flag, KBD_NORESET (0x20) to tell scprobe() not
to reset the keyboard.

IBM ThinkPad 535 has the `Fn' key with which the user can perform
certain functions in conjunction with other keys. For example, `Fn' +
PageUP/PageDOWN adjust speaker volume, `Fn' + Home/End change
brightness of LCD screen. It can also be used to suspend the system.

It appears that these functions are implemented at the keyboard level
or the keyboard controller level and totally independent from BIOS or
OS.  But, if the keyboard is reset (as is done in scprobe()), they
become unavailable.  (There are other laptops which have similar
functions associated with the `Fn' key. But, they aren't affected by
keyboard reset.)

ThinkPad 535 doesn't have switches or buttons to adjust brightness and
volume, or to put the system into the suspend mode. Therefore, it is
essential to preserve these `Fn' key functions in FreeBSD. The new
flag make scprobe() skip keyboard reset.

If this flag is not set, scprobe() behaves in the same say as before.

(If we only knew a way to detect ThinkPad 535, we could skip keyboard
reset automatically, but...)

27 years agoFix HUP handling while dialing.
brian [Sun, 29 Jun 1997 13:54:31 +0000 (13:54 +0000)]
Fix HUP handling while dialing.

Pointed out by: ache

27 years agoFixed clobbered vendor Id.
bde [Sun, 29 Jun 1997 11:52:51 +0000 (11:52 +0000)]
Fixed clobbered vendor Id.

27 years agoAttempt to fix clobbered vendor Id.
bde [Sun, 29 Jun 1997 11:43:31 +0000 (11:43 +0000)]
Attempt to fix clobbered vendor Id.

27 years ago``-C ${WRKDIR} -xzf'' are no longer the default args for EXTRACT_BEFORE_ARGS
obrien [Sun, 29 Jun 1997 10:51:55 +0000 (10:51 +0000)]
``-C ${WRKDIR} -xzf'' are no longer the default args for EXTRACT_BEFORE_ARGS

27 years agoAdd support for @option preserve - moves existing files out of the way
jkh [Sun, 29 Jun 1997 10:41:49 +0000 (10:41 +0000)]
Add support for @option preserve - moves existing files out of the way
before replacing them (using pkgname to make this hopefully unique).
Delete also moves them back, if they exist, resulting in a package
which can be "backed out" with reasonable safety.

27 years agoAdd magic for FIGlet fonts and controlfiles
obrien [Sun, 29 Jun 1997 10:06:56 +0000 (10:06 +0000)]
Add magic for FIGlet fonts and controlfiles

Obtained from: Figlet v2.2 release.

27 years agoAdd a blurb on how to import newer versions of patch(1) so
steve [Sun, 29 Jun 1997 07:20:16 +0000 (07:20 +0000)]
Add a blurb on how to import newer versions of patch(1) so
that others don't make the same bone-headed mistakes that
I just did.

Obtained from: Peter Wemm's version of this file in contrib_cvs

27 years ago-Wall clean.
steve [Sun, 29 Jun 1997 06:56:24 +0000 (06:56 +0000)]
-Wall clean.

27 years agoThis commit was generated by cvs2svn to compensate for changes in r27044,
steve [Sun, 29 Jun 1997 06:55:02 +0000 (06:55 +0000)]
This commit was generated by cvs2svn to compensate for changes in r27044,
which included commits to RCS files with non-trunk default branches.

27 years agoImport of GNU patch version 2.4.
steve [Sun, 29 Jun 1997 06:55:02 +0000 (06:55 +0000)]
Import of GNU patch version 2.4.

27 years agoMerge from 2.2
pst [Sun, 29 Jun 1997 06:23:14 +0000 (06:23 +0000)]
Merge from 2.2

27 years agoNOSHARED takes a yes/YES no/NO value, not "true, false, hey mon!".
pst [Sun, 29 Jun 1997 06:03:42 +0000 (06:03 +0000)]
NOSHARED takes a yes/YES no/NO value, not "true, false, hey mon!".
NOPIC is used to not generate a shared library, not NOSHARED.
Make NOSHARED advisory where appropriate.
Remove bogus NOSHARED (kbdio).

27 years ago_err() -> err().
jkh [Sun, 29 Jun 1997 00:33:17 +0000 (00:33 +0000)]
_err() -> err().

27 years agoSome minor cosmetic tweaks.
steve [Sat, 28 Jun 1997 19:27:51 +0000 (19:27 +0000)]
Some minor cosmetic tweaks.

PR: docs/3941
Partially submitted by: uenami@imasy.or.jp

27 years agoFix bogus patch (my fault)
pst [Sat, 28 Jun 1997 16:53:47 +0000 (16:53 +0000)]
Fix bogus patch (my fault)

27 years agoendif -> .endif (from BOOT_CONFIG changes).
jkh [Sat, 28 Jun 1997 16:24:09 +0000 (16:24 +0000)]
endif -> .endif (from BOOT_CONFIG changes).

27 years agoReorder things and import NOSHARED.
pst [Sat, 28 Jun 1997 08:21:10 +0000 (08:21 +0000)]
Reorder things and import NOSHARED.

27 years agoAttempt to open the device for reading before actually adding the device
pst [Sat, 28 Jun 1997 08:18:29 +0000 (08:18 +0000)]
Attempt to open the device for reading before actually adding the device
to the session list.  If the device comes back as unconfigured, just
ignore that line in /etc/ttys.  If someone HUP's init, we'll try again.

This change stops getty's from hanging on vty and sio ports that don't
exist, either due to LKM drivers not being loaded, or probes failing.
Reviewed by: bde

27 years agoMake NOSHARED = no / NOSHARED = NO do what you'd expect.
pst [Sat, 28 Jun 1997 08:14:10 +0000 (08:14 +0000)]
Make NOSHARED = no / NOSHARED = NO do what you'd expect.

27 years agoadd cvspserver (officially registered at 2401)
peter [Sat, 28 Jun 1997 04:28:07 +0000 (04:28 +0000)]
add cvspserver (officially registered at 2401)

27 years agoreplace the OpenBSD fd_set sizing code with something more efficient.
peter [Sat, 28 Jun 1997 04:19:52 +0000 (04:19 +0000)]
replace the OpenBSD fd_set sizing code with something more efficient.
Only call malloc() if the fd is too big for the compiled in fd_set size,
and don't use calloc either.  This should reduce the impact of conflicts
with private malloc implementations etc.  When using the fd_set on the
stack, only zero what is needed rather than all 1024 bits like FD_ZERO did.

27 years agoProperly make directory before moving things into it.
jkh [Sat, 28 Jun 1997 02:37:33 +0000 (02:37 +0000)]
Properly make directory before moving things into it.
Submitted by: pst

27 years agoDeal with HISADDR/MYADDR in filter rules.
brian [Sat, 28 Jun 1997 01:34:03 +0000 (01:34 +0000)]
Deal with HISADDR/MYADDR in filter rules.
Mostly submitted by: kfurge@worldnet.att.net

Allow MYADDR in add/delete commands to facilitate
dynamic additions of a loopback route to MYADDR.

27 years agoAllow command line control of ppp through both
brian [Sat, 28 Jun 1997 01:04:54 +0000 (01:04 +0000)]
Allow command line control of ppp through both
TCP and AF_LOCAL sockets.

27 years agoapic_vector.s:
fsmp [Fri, 27 Jun 1997 23:48:05 +0000 (23:48 +0000)]
apic_vector.s:
 - added Xcpustop IPI code to support stop_cpus()/restart_cpus().
   it is off by default, enable via smptests.h:TEST_CPUSTOP

intr_machdep.h:
 - moved +ICULEN to lower level.
 - added entry for Xcpustop.

27 years agoInitialize private variable other_cpus during AP boot.
fsmp [Fri, 27 Jun 1997 23:38:32 +0000 (23:38 +0000)]
Initialize private variable other_cpus during AP boot.

27 years agoAdded POST code output to various points of the startup code.
fsmp [Fri, 27 Jun 1997 23:33:17 +0000 (23:33 +0000)]
Added POST code output to various points of the startup code.

General cleanup.

New functions to stop/start CPUs via IPIs:

 - int stop_cpus( u_int map );
 - int restart_cpus( u_int map );

Turned off by default, enabled via smptests.h:TEST_CPUSTOP.
Current version has a BUG, perhaps a deadlock?

27 years agoExperimental calls to stop_cpus()/restart_cpus() within breakpoint calls.
fsmp [Fri, 27 Jun 1997 23:24:38 +0000 (23:24 +0000)]
Experimental calls to stop_cpus()/restart_cpus() within breakpoint calls.
Turned off by default in smptests.h.

27 years agoAdded other_cpus to CPU private page.
fsmp [Fri, 27 Jun 1997 23:19:43 +0000 (23:19 +0000)]
Added other_cpus to CPU private page.

This variable is a bitmap showing all CPUs present EXCEPT the CPU
owning the variable.  In other words, it is equal to the global bitmap
'all_cpus' minus its own bit.

27 years agoPreliminaries for stop_cpus()/restart_cpus().
fsmp [Fri, 27 Jun 1997 23:12:31 +0000 (23:12 +0000)]
Preliminaries for stop_cpus()/restart_cpus().
Both are turned off by default.

Added macro for displaying POST codes from kernel.

27 years agoProgram lint1 to handle NMIs.
fsmp [Fri, 27 Jun 1997 22:27:18 +0000 (22:27 +0000)]
Program lint1 to handle NMIs.

Till now NMIs would be ignored.  Now an NMI is caught by the BSP.
APs still ignore NMI, am working on code to allow a CPU to stop other CPUs
via an IPI.

27 years agoAdded fields to the LVT1/2 group.
fsmp [Fri, 27 Jun 1997 22:13:50 +0000 (22:13 +0000)]
Added fields to the LVT1/2 group.

27 years agoRemoved unused variables.
alex [Fri, 27 Jun 1997 21:51:59 +0000 (21:51 +0000)]
Removed unused variables.

27 years agoModify my copyright notice to allow the sequencer to be used with GPLed
gibbs [Fri, 27 Jun 1997 19:39:34 +0000 (19:39 +0000)]
Modify my copyright notice to allow the sequencer to be used with GPLed
software (aka Linux).

27 years agoModify my copyright notice to allow the sequencer to be used with GPLed
gibbs [Fri, 27 Jun 1997 19:38:56 +0000 (19:38 +0000)]
Modify my copyright notice to allow the sequencer to be used with GPLed
software (aka Linux).

Fix a few bugs in the sequencer assembler.

Make it easy to compiler the assembler with debugging turned on.

27 years agoKNF cleanup.
gibbs [Fri, 27 Jun 1997 19:36:27 +0000 (19:36 +0000)]
KNF cleanup.

27 years agoFix a condition where nfs_statfs() can precipitate a panic. There is
wpaul [Fri, 27 Jun 1997 19:10:46 +0000 (19:10 +0000)]
Fix a condition where nfs_statfs() can precipitate a panic. There is
code that says this:

        nfsm_request(vp, NFSPROC_FSSTAT, p, cred);
        if (v3)
                nfsm_postop_attr(vp, retattr);
        if (!error)
                nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3));

The problem here is that if error != 0, nfsm_dissect() will not be
called, which leaves sfp == NULL. But nfs_statfs() does not bail out
at this point: it continues processing until it tries to dereference
sfp, which causes a panic. I was able to generate this crash under
the following conditions:

1) Set up a machine as an NFS server and NFS client, with amd running
   (using NIS maps). /usr/local is exported, though any exported fs
   can can be used to trigger the bug.
2) Log in as normal user, with home directory mounted from a SunOS 4.1.3
   NFS server via amd (along with a few other NFS filesystems from same
   machine).
3) Su to root and type the following:
   # mount localhost:/usr/local /mnt
   # df

To fix the panic, I changed the code to read:

        if (!error) {
                nfsm_dissect(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
        } else
                goto nfsmout;

This is a bit kludgy in that nfsmout is a label defined by the nfsm_subs.h
macros, but these macros are themselves more than a little kludgy. This
stops the machine from crashing, but does not fix the overall bug: 'error'
somehow becomes 5 (EIO) when a statfs() is performed on the locally mounted
NFS filesystem. This seems to only happen the first time the filesystem
is accesed: on subsequent accesses, it seems to work fine again.

Now, I know there's no practical use in mounting a local filesystem
via NFS, but doing it shouldn't cause the system to melt down.

27 years agoRemoved '#include <machine/smptests.h>' line, no longer needed.
fsmp [Fri, 27 Jun 1997 18:29:55 +0000 (18:29 +0000)]
Removed '#include <machine/smptests.h>' line, no longer needed.

27 years agoRevive this file, it's come back from the dead in the 8.8.x dists.
peter [Fri, 27 Jun 1997 15:55:33 +0000 (15:55 +0000)]
Revive this file, it's come back from the dead in the 8.8.x dists.