]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
27 years agochange kvm-fbsd.c so that kernel debugging works again.
Gary Jennejohn [Sun, 27 Apr 1997 21:36:49 +0000 (21:36 +0000)]
change kvm-fbsd.c so that kernel debugging works again.

document the -k and -wcore options in the manpage.

change Makefile to use a local copy of the manpage.

27 years agoRemoved bogon from previous commit: doubly included sys/systm.h.
Alexander Langer [Sun, 27 Apr 1997 21:32:21 +0000 (21:32 +0000)]
Removed bogon from previous commit: doubly included sys/systm.h.

27 years agoRemove bogon from previous commit: doubly included sys/systm.h.
Alexander Langer [Sun, 27 Apr 1997 21:26:29 +0000 (21:26 +0000)]
Remove bogon from previous commit: doubly included sys/systm.h.

27 years ago informal discussion between Bruce Evans <bde@zeta.org.au>,
Steve Passe [Sun, 27 Apr 1997 21:18:59 +0000 (21:18 +0000)]
informal discussion between Bruce Evans <bde@zeta.org.au>,
Peter Wemm <peter@spinner.DIALix.COM>, Steve Passe <smp@csn.net>

removed all the IPI_INTS code.
made the XFAST_IPI32 code default, renaming Xfastipi32 to Xinvltlb.
cleanup of i386/isa/isa_device.h to eliminate SMP dependancies:
  made the id_irq member of struct isa_device an u_int.
  made the id_drq member of struct isa_device an int.
  removed all other '#ifdefs' concerning SMP & APIC_IO.
removed SMP/APIC_IO dependancies from if_ze.c.

27 years ago informal discussion between Bruce Evans <bde@zeta.org.au>,
Steve Passe [Sun, 27 Apr 1997 21:17:56 +0000 (21:17 +0000)]
informal discussion between Bruce Evans <bde@zeta.org.au>,
Peter Wemm <peter@spinner.DIALix.COM>, Steve Passe <smp@csn.net>

removed all the IPI_INTS code.
made the XFAST_IPI32 code default, renaming Xfastipi32 to Xinvltlb.

27 years agoTypo police.
Alexander Langer [Sun, 27 Apr 1997 20:12:34 +0000 (20:12 +0000)]
Typo police.

Added links to O'Reilly & Associates and Addison-Wesley's web sites
to accompany the book recommendations.

27 years agoCreate a special option file "opt_global.h" which is included by all
Peter Wemm [Sun, 27 Apr 1997 20:01:47 +0000 (20:01 +0000)]
Create a special option file "opt_global.h" which is included by all
source files via a 'cc -include opt_global.h ...' type arrangement.
This means we can untangle certain header files.

options.i386 has a placeholder until it has a real member so we can avoid
having to teach config about it just yet.

Reviewed by: bde

27 years agoThe long-awaited mega-massive-network-code- cleanup. Part I.
Garrett Wollman [Sun, 27 Apr 1997 20:01:29 +0000 (20:01 +0000)]
The long-awaited mega-massive-network-code- cleanup.  Part I.

This commit includes the following changes:
1) Old-style (pr_usrreq()) protocols are no longer supported, the compatibility
glue for them is deleted, and the kernel will panic on boot if any are compiled
in.

2) Certain protocol entry points are modified to take a process structure,
so they they can easily tell whether or not it is possible to sleep, and
also to access credentials.

3) SS_PRIV is no more, and with it goes the SO_PRIVSTATE setsockopt()
call.  Protocols should use the process pointer they are now passed.

4) The PF_LOCAL and PF_ROUTE families have been updated to use the new
style, as has the `raw' skeleton family.

5) PF_LOCAL sockets now obey the process's umask when creating a socket
in the filesystem.

As a result, LINT is now broken.  I'm hoping that some enterprising hacker
with a bit more time will either make the broken bits work (should be
easy for netipx) or dike them out.

27 years agoPrevent debugger attachment to init when securelevel > 0.
Alexander Langer [Sun, 27 Apr 1997 19:02:37 +0000 (19:02 +0000)]
Prevent debugger attachment to init when securelevel > 0.

Noticed by: Brian Buchanan <brian@wasteland.calbbs.com>

27 years agoDelete opt_ddb.h.
KATO Takenori [Sun, 27 Apr 1997 13:41:08 +0000 (13:41 +0000)]
Delete opt_ddb.h.

27 years agoSynchronize with sys/i386/isa/syscons.c revision 1.211.
KATO Takenori [Sun, 27 Apr 1997 13:23:29 +0000 (13:23 +0000)]
Synchronize with sys/i386/isa/syscons.c revision 1.211.

27 years agoMerge SMP code from IBM-PC tree into PC-98 tree.
KATO Takenori [Sun, 27 Apr 1997 13:22:09 +0000 (13:22 +0000)]
Merge SMP code from IBM-PC tree into PC-98 tree.

27 years agoWhoops.. We forgot to turn off the 4MB Virtual==Physical mapping at address
Peter Wemm [Sun, 27 Apr 1997 12:11:43 +0000 (12:11 +0000)]
Whoops.. We forgot to turn off the 4MB Virtual==Physical mapping at address
zero from bootstrap in the non-SMP case.

Noticed by: bde

27 years agoUndo 1.29.
KATO Takenori [Sun, 27 Apr 1997 10:49:37 +0000 (10:49 +0000)]
Undo 1.29.

27 years agoAddresses the following two problems.
Kazutaka YOKOTA [Sun, 27 Apr 1997 09:03:10 +0000 (09:03 +0000)]
Addresses the following two problems.

1) i586_bcopy() problem

There have been a number of reports that the syscons doesn't work
properly if i586_bcopy() is enabled.

The problem prevented users from installing 2.2(.1)-RELEASE.  The
symptom is that the system looks frozen during device probe or just
before the main installation menu.  The workaround was to specify the
flag 0x01 to the npx device so that i586_bcopy() is disabled.

The patch forces the syscons to call generic_bcopy() when copying
to/from the video memory, even if CPU is Pentium and i586_bcopy() is
enabled. i586_bcopy() is still called for copy operations between
non-video memory regions.

PR: kern/2277, kern/3066, kern/3107, kern/3134

2) video mode parameter table problem

The syscons reads and uses the video mode parameter table provided by
the VGA BIOS to set VGA registers when changing video mode and
modifying font data. It appears that in some VGA BIOSes the table is
not ordered as the syscons expects, and this leads to screen
corruption.

The problem prevented users from installing 2.2(.1)-RELEASE. The
symptom is the corrupt screen or strange vertical lines soon after the
kernel is loaded into memory (just after the kernel decompression).

The patch performs simplistic test and if it fails, set video_mode_ptr
to NULL so that the video mode switching won't happen.

This is an interim kludge. There should be a better way to deal with
the problem.

PR: kern/2498, conf/2775, conf/3354

Reviewed by: sos
Tested by: PR originators (not all of them, though)

27 years ago``appears'' -> ``appeared'' (closes PR#3393, Submitted-by: Josh Gilliam)
John-Mark Gurney [Sun, 27 Apr 1997 08:45:46 +0000 (08:45 +0000)]
``appears'' -> ``appeared'' (closes PR#3393, Submitted-by: Josh Gilliam)

add missing Id's
other minor clean ups

27 years agoAdd some more "member functions" to these. Still not complete.
Poul-Henning Kamp [Sun, 27 Apr 1997 08:31:17 +0000 (08:31 +0000)]
Add some more "member functions" to these.  Still not complete.

27 years agoYAMF2.2: Allow @group entries in /etc/ftpusers & /etc/ftpchroot to deny
David Nugent [Sun, 27 Apr 1997 08:29:21 +0000 (08:29 +0000)]
YAMF2.2: Allow @group entries in /etc/ftpusers & /etc/ftpchroot to deny
and allow chroot access to entire groups.

27 years agoBring in rc file changes from -current.
Jordan K. Hubbard [Sun, 27 Apr 1997 03:59:19 +0000 (03:59 +0000)]
Bring in rc file changes from -current.

27 years agoAdd ports-converters and ports-textproc targets.
Satoshi Asami [Sun, 27 Apr 1997 00:36:47 +0000 (00:36 +0000)]
Add ports-converters and ports-textproc targets.

27 years agoAdd description of ports-converters and ports-textproc.
Satoshi Asami [Sun, 27 Apr 1997 00:32:37 +0000 (00:32 +0000)]
Add description of ports-converters and ports-textproc.

27 years agoTry and make these usermode safe, Steve beat me in finding these..
Peter Wemm [Sat, 26 Apr 1997 20:04:21 +0000 (20:04 +0000)]
Try and make these usermode safe, Steve beat me in finding these..

27 years agofix & instead of && in #if statement
Peter Wemm [Sat, 26 Apr 1997 19:47:59 +0000 (19:47 +0000)]
fix & instead of && in #if statement
reorder #includes to alphabetical order

Noted by: bde

27 years agochange #if 0 to #ifdef KERNEL since this is pretty important for
Peter Wemm [Sat, 26 Apr 1997 19:26:37 +0000 (19:26 +0000)]
change #if 0 to #ifdef KERNEL since this is pretty important for
smp compiling.  It still leaves dset non-functional on smp kernels, but
we've been doing that all along with the private smp tree anyway.

27 years agoRemove the curproc printing on trap/interrupt/etc. It's outlived it's
Peter Wemm [Sat, 26 Apr 1997 19:12:03 +0000 (19:12 +0000)]
Remove the curproc printing on trap/interrupt/etc.  It's outlived it's
usefulness, and there were problems with it anyway.

Found by: bde

27 years ago#if'ed out the '#include "opt_smp.h"' line.
Steve Passe [Sat, 26 Apr 1997 19:07:36 +0000 (19:07 +0000)]
#if'ed out the '#include "opt_smp.h"' line.

The SMP source was merged into 3.0-current last nite and this broke
the make of sbin/dset.

Should make things work for non-SMP case.

People making SMP kernels will need to edit i386/isa/isa_device.h, re-enabling
'#include "opt_smp.h"'.  People making SMP worlds will have to punt for now...
I'm thinking about the real solution, but for now the goal is to NOT break
the world!

27 years agoBack out bogus code that slipped past my read of the pre-merge diff
Peter Wemm [Sat, 26 Apr 1997 18:59:42 +0000 (18:59 +0000)]
Back out bogus code that slipped past my read of the pre-merge diff
(Problems noted by Bruce)

27 years agoFix some SMP merge bugs (from Bruce) -
Peter Wemm [Sat, 26 Apr 1997 18:57:34 +0000 (18:57 +0000)]
Fix some SMP merge bugs (from Bruce) -
#include out of order
pccard_configure() called twice
munged tab (existing problem made worse)

27 years agoActivate gdbserver and gdbreplay
Paul Traina [Sat, 26 Apr 1997 17:34:05 +0000 (17:34 +0000)]
Activate gdbserver and gdbreplay

27 years agoCreate sysctl kern.fast_vfork, on for uniprocessor by default, off for
Peter Wemm [Sat, 26 Apr 1997 15:59:50 +0000 (15:59 +0000)]
Create sysctl kern.fast_vfork, on for uniprocessor by default, off for
SMP.

27 years agoDisable RFMEM in vfork for smp case.. It doesn't seem to work too well
Peter Wemm [Sat, 26 Apr 1997 14:31:36 +0000 (14:31 +0000)]
Disable RFMEM in vfork for smp case..  It doesn't seem to work too well
yet..

27 years agoDo nothing instead of adjusting un_flags when (uppervp is locked) &&
KATO Takenori [Sat, 26 Apr 1997 13:43:25 +0000 (13:43 +0000)]
Do nothing instead of adjusting un_flags when (uppervp is locked) &&
(UN_ULOCK is not set) in union_lock.  This condition may indicate
race.  DIAGNOSTIC kernel still panic here.

27 years agoDocument internal ls, how to compile it in and what it changes wrt
David Nugent [Sat, 26 Apr 1997 12:23:51 +0000 (12:23 +0000)]
Document internal ls, how to compile it in and what it changes wrt
anon ftp and chrooted users.

27 years agoAdds optional "internal ls" support for ftpd, by collecting
David Nugent [Sat, 26 Apr 1997 12:12:10 +0000 (12:12 +0000)]
Adds optional "internal ls" support for ftpd, by collecting
modules from src/bin/ls, and handling exec(_PATH_LS,..) as a
special case, very useful in an environment where many users
are given chroot access. "~/etc/{s}pwd.db" files are still
needed if uid/gid->user/group translation is desired.

To enable this it must be compiled with the make variable
FTP_INTERNAL_LS defined, either in /etc/make.conf or the
environment.

27 years agoMan the liferafts! Here comes the long awaited SMP -> -current merge!
Peter Wemm [Sat, 26 Apr 1997 11:46:25 +0000 (11:46 +0000)]
Man the liferafts!  Here comes the long awaited SMP -> -current merge!

There are various options documented in i386/conf/LINT, there is more to
come over the next few days.

The kernel should run pretty much "as before" without the options to
activate SMP mode.

There are a handful of known "loose ends" that need to be fixed, but
have been put off since the SMP kernel is in a moderately good condition
at the moment.

This commit is the result of the tinkering and testing over the last 14
months by many people.  A special thanks to Steve Passe for implementing
the APIC code!

27 years agoPrint out some more diagnostic information when we reject a message.
Justin T. Gibbs [Sat, 26 Apr 1997 05:03:18 +0000 (05:03 +0000)]
Print out some more diagnostic information when we reject a message.

When we request sense, don't allow disconnection.  This closes a window
where we might allow an overlapped tagged and non-tagged transaction.
The correct fix is to freeze the queue for the target that requests sense
which is what will happen in the new CAM framework.

27 years agoDo not clear UN_ULOCK in certain case.
KATO Takenori [Sat, 26 Apr 1997 04:09:40 +0000 (04:09 +0000)]
Do not clear UN_ULOCK in certain case.

Our vput calls vm_object_deallocate() --> vm_object_terminate().  The
vm_object_terminate() calls vn_lock(), since UN_LOCKED has been
already cleared in union_unlock().  Then, union_lock locks upper vnode
when UN_ULOCK is not set.  The upper vnode is not unlocked when
UN_KLOCK is set in union_unlock(), thus, union_lock tries to lock
locked vnode and we get panic.

27 years agoAdd new cpu type, CPU_CY486DX, which shows Cyrix 486S/DX series CPUs,
KATO Takenori [Sat, 26 Apr 1997 04:08:54 +0000 (04:08 +0000)]
Add new cpu type, CPU_CY486DX, which shows Cyrix 486S/DX series CPUs,
and initialization routine for those CPUs.

Tested by: Bob Bishop <rb@gid.co.uk>

27 years agoSynchronize with sys/i386/boot/biosboot/Makefile and
KATO Takenori [Sat, 26 Apr 1997 04:00:25 +0000 (04:00 +0000)]
Synchronize with sys/i386/boot/biosboot/Makefile and
sys/i386/boot/rawboot/Makefile revsions 1.56 and 1.7, respectively.

27 years agoWhoops -- code inspection showed I forgot to clean up getopt string
Paul Traina [Sat, 26 Apr 1997 00:03:21 +0000 (00:03 +0000)]
Whoops -- code inspection showed I forgot to clean up getopt string

27 years agoSecure mode (-s) incorrectly disabled both sending and receiving of syslog
Paul Traina [Sat, 26 Apr 1997 00:00:33 +0000 (00:00 +0000)]
Secure mode (-s) incorrectly disabled both sending and receiving of syslog
packets over UDP.  Secure boxes should still be able to send packets.

27 years agoCosmetique changes in my entry, add yet one ID
Andrey A. Chernov [Fri, 25 Apr 1997 22:58:22 +0000 (22:58 +0000)]
Cosmetique changes in my entry, add yet one ID

27 years agoUpdate ID in my pgp key
Andrey A. Chernov [Fri, 25 Apr 1997 22:54:33 +0000 (22:54 +0000)]
Update ID in my pgp key

27 years agoInitial import of gdbserver FreeBSD support
Paul Traina [Fri, 25 Apr 1997 22:25:50 +0000 (22:25 +0000)]
Initial import of gdbserver FreeBSD support

27 years agoInitial import of gdbreplay build structure
Paul Traina [Fri, 25 Apr 1997 22:25:22 +0000 (22:25 +0000)]
Initial import of gdbreplay build structure

27 years agoTypo.
Masafumi Max NAKANE [Fri, 25 Apr 1997 20:47:15 +0000 (20:47 +0000)]
Typo.
PR: 3385
Submitted by: Josh Gilliam <soil@quick.net>

27 years agoTypo.
Masafumi Max NAKANE [Fri, 25 Apr 1997 20:36:10 +0000 (20:36 +0000)]
Typo.
PR: #3178
Submitted by: Josh Gilliam <soil@quick.net>

27 years agoFixed `make depend'. The previous change reduced it to making only
Bruce Evans [Fri, 25 Apr 1997 20:20:21 +0000 (20:20 +0000)]
Fixed `make depend'.  The previous change reduced it to making only
config.h (which isn't used directly).  Added a comment about the
indirect use of config.h.

27 years agoIntegrated 1.54 -> 1.55 changes.
Masafumi Max NAKANE [Fri, 25 Apr 1997 20:09:23 +0000 (20:09 +0000)]
Integrated 1.54 -> 1.55 changes.

27 years agoMailto: was missing for one contact person's address.
Masafumi Max NAKANE [Fri, 25 Apr 1997 20:03:48 +0000 (20:03 +0000)]
Mailto: was missing for one contact person's address.
Submitted by:  Dmitrij Tejblum <dima@tejblum.dnttm.rssi.ru>
(Closing PR #3051.)

27 years agoDon't complain about no input files for `ld -f'. This fixes spurious
Bruce Evans [Fri, 25 Apr 1997 19:43:19 +0000 (19:43 +0000)]
Don't complain about no input files for `ld -f'.  This fixes spurious
(ignored :-() errors for `make depend' in /sys/i386/boot/*.  It's
natural for there to be no libraries there and inconvenient to check
for this in bsd.prog.mk.

27 years agoRemoved LDADD=-lc and associated DPADD. libc would probably break
Bruce Evans [Fri, 25 Apr 1997 19:38:01 +0000 (19:38 +0000)]
Removed LDADD=-lc and associated DPADD.  libc would probably break
things if it were actually used.

27 years agoFill the padding at the end of the runtime strings with zeros,
John Polstra [Fri, 25 Apr 1997 15:26:12 +0000 (15:26 +0000)]
Fill the padding at the end of the runtime strings with zeros,
instead of leaving random data there.  This makes the linker's
output files more deterministic -- an important property for
regression tests.

27 years agoAdd vfork(2) to SEE ALSO
Andrey A. Chernov [Fri, 25 Apr 1997 14:41:55 +0000 (14:41 +0000)]
Add vfork(2) to SEE ALSO

27 years agoAdd portsinfo, which generate list of new ports for last two weeks.
Wolfram Schneider [Fri, 25 Apr 1997 14:14:39 +0000 (14:14 +0000)]
Add portsinfo, which generate list of new ports for last two weeks.

27 years agoDon't zero b_dirtyoff and b_dirtyend on error.
Doug Rabson [Fri, 25 Apr 1997 11:14:00 +0000 (11:14 +0000)]
Don't zero b_dirtyoff and b_dirtyend on error.

Submitted by: Hidetoshi Shimokawa <simokawa@sat.t.u-tokyo.ac.jp>

27 years agoDon't try to install f2c'h now that it's installed from
Satoshi Asami [Fri, 25 Apr 1997 09:58:24 +0000 (09:58 +0000)]
Don't try to install f2c'h now that it's installed from
/usr/src/usr.bin/f2c. :)

27 years agoEliminate a small redundancy with ALL_DISTS
Jordan K. Hubbard [Fri, 25 Apr 1997 09:11:50 +0000 (09:11 +0000)]
Eliminate a small redundancy with ALL_DISTS
Submitted by: John-Mark Gurney <jmg@hydrogen.nike.efn.org>

27 years agoCatch up with the original.
Hiroyuki Hanai [Fri, 25 Apr 1997 07:24:02 +0000 (07:24 +0000)]
Catch up with the original.

27 years agocopyin the export network mask to the correct variable.
Peter Wemm [Fri, 25 Apr 1997 06:47:12 +0000 (06:47 +0000)]
copyin the export network mask to the correct variable.

Submitted by: Mike Hibler <mike@marker.cs.utah.edu>, PR#3380

27 years agocorrect vfork()/exit() to vfork()/_exit(). This is far more important
Peter Wemm [Fri, 25 Apr 1997 06:42:57 +0000 (06:42 +0000)]
correct vfork()/exit() to vfork()/_exit().  This is far more important
now that vfork() actually works in FreeBSD for the first time.

Spotted by: se

27 years agoAdd a -I switch to CFLAGS to use the f2c.h header file in the usr.bin/f2c
Jean-Marc Zucconi [Thu, 24 Apr 1997 17:10:56 +0000 (17:10 +0000)]
Add a -I switch to CFLAGS to use the f2c.h header file in the usr.bin/f2c
directory.
Reviewed by: bde

27 years agoInstall f2c.h
Jean-Marc Zucconi [Thu, 24 Apr 1997 17:04:04 +0000 (17:04 +0000)]
Install f2c.h
Reviewed by: bde

27 years agoThis file is a duplicate of src/usr.bin/f2c/f2c.h
Jean-Marc Zucconi [Thu, 24 Apr 1997 17:03:41 +0000 (17:03 +0000)]
This file is a duplicate of src/usr.bin/f2c/f2c.h
Reviewed by: bde

27 years agoThe following scenario would result in a bogus residual being reported
Justin T. Gibbs [Thu, 24 Apr 1997 16:52:18 +0000 (16:52 +0000)]
The following scenario would result in a bogus residual being reported
if SCB Paging was enabled:

    disconnect with more data to transfer
    disconnected SCB gets paged out
    target reconnects so we page SCB back in
    target completes transfer so residual is 0
    target disconnects
    SCB gets reused but not paged out since the residual is 0 (optimization)
    target reconnects so we page the SCB back in
    we report a residual because of stale residual information.

The fix for this is to set a flag that forces the SCB to be paged back
up to the host if we page in an SCB with a residual

Pointed out by: Doug Ledford <dledford@dialnet.net>

27 years agoAdd definition of PCI_SUBID_REG2, the subvendor/device ID for
Stefan Eßer [Thu, 24 Apr 1997 08:03:31 +0000 (08:03 +0000)]
Add definition of PCI_SUBID_REG2, the subvendor/device ID for
a header type 2 device.

27 years agoChange vfork to fork, too many memory-clobbering actions present in child
Andrey A. Chernov [Wed, 23 Apr 1997 23:04:58 +0000 (23:04 +0000)]
Change vfork to fork, too many memory-clobbering actions present in child

27 years agoEven better fix: use close(fileno(mfp)); to close unneded descriptor
Andrey A. Chernov [Wed, 23 Apr 1997 22:40:19 +0000 (22:40 +0000)]
Even better fix: use close(fileno(mfp)); to close unneded descriptor
and not clobber memory by fclose().

27 years agoBetter fix: return vfork() again, but remove clobbering fclose()
Andrey A. Chernov [Wed, 23 Apr 1997 22:36:51 +0000 (22:36 +0000)]
Better fix: return vfork() again, but remove clobbering fclose()
and use _exit() instead of exit()

27 years agoReplace vfork() by fork(), vfork() usage is totally bogus in this context.
Andrey A. Chernov [Wed, 23 Apr 1997 22:25:20 +0000 (22:25 +0000)]
Replace vfork() by fork(), vfork() usage is totally bogus in this context.
This program don't bother to do even minimal vfork() support as per vfork(2)
manpage required.

27 years agoRestore memory space separation (RFMEM) for vfork() after
Andrey A. Chernov [Wed, 23 Apr 1997 22:13:18 +0000 (22:13 +0000)]
Restore memory space separation (RFMEM) for vfork() after
shell imgact memory clobbering fixed

27 years agoDon't clobber user space argv0 memory on shell exec, mainly for vfork()
Andrey A. Chernov [Wed, 23 Apr 1997 22:07:05 +0000 (22:07 +0000)]
Don't clobber user space argv0 memory on shell exec, mainly for vfork()
Fix another bug: if argv[0] is NULL, garbadge args might be added for
shell script
Submitted by: Tor Egge <Tor.Egge@idi.ntnu.no> (with yet one fault detect from me)

27 years agoAdd argv0 field for imgact structure - for replacing argv0 in shell
Andrey A. Chernov [Wed, 23 Apr 1997 22:02:37 +0000 (22:02 +0000)]
Add argv0 field for imgact structure - for replacing argv0 in shell
image activator
Submitted by:  Tor Egge <Tor.Egge@idi.ntnu.no>

27 years agoAdd preliminary support for PCI config header type 2:
Stefan Eßer [Wed, 23 Apr 1997 19:43:20 +0000 (19:43 +0000)]
Add preliminary support for PCI config header type 2:
Fetch subvendor/device ID from config space register 0x40.

27 years agoFixed longstanding profiling bug. The frame pointer wasn't set up
Bruce Evans [Wed, 23 Apr 1997 19:26:04 +0000 (19:26 +0000)]
Fixed longstanding profiling bug.  The frame pointer wasn't set up
for syscalls, so one frame was lost in backtraces from syscalls.
This is handled better in the kernel by using a different mcount
entry point for profiling before the frame pointer is set up.

Expand RCSID().

Use .p2align instead of the ambiguous .align.

Added idempotency ifdef.

Removed unused macros ALTENTRY(), ALTASENTRY(), ASENTRY(), _MID_ENTRY.

Cleaned up formatting.

Reviewed by: jdp reviewed an old version
Obtained from: parts from NetBSD

27 years agoPass everything in ${LDADD} directly to the linker for `make depend'
Bruce Evans [Wed, 23 Apr 1997 18:47:42 +0000 (18:47 +0000)]
Pass everything in ${LDADD} directly to the linker for `make depend'
to prevent complaints from the compiler about nonexistent libraries.

27 years agoAdd Christoph Kukulies to donor's gallery.
Jordan K. Hubbard [Wed, 23 Apr 1997 18:36:37 +0000 (18:36 +0000)]
Add Christoph Kukulies to donor's gallery.

27 years agoDon't call getnetbyname() on a netmask. On a system with a slow net
Mike Smith [Wed, 23 Apr 1997 11:03:10 +0000 (11:03 +0000)]
Don't call getnetbyname() on a netmask.  On a system with a slow net
connection, a large set of network exports could take many minutes to
time out, giving the appearance of a total hang during boot.

27 years agoGenerate .S files for syscalls. This has many advantages:
Bruce Evans [Wed, 23 Apr 1997 10:49:54 +0000 (10:49 +0000)]
Generate .S files for syscalls.  This has many advantages:
- dependencies actually work (I need this to propagate some fixes
  in <machine/asm.h>)
- the cpp pipeline goes away, so errors can't leak out of it and
  an ANSI cpp is automatically used.
- it's simpler - standard rules get used instead of repetitive
  special rules.  (This showed bugs in the strip steps in the
  standard rules.  The wrong strip flag was also used for *.po
  here.)

Removed some ${ECHO}s and `@'s.  Normal make echoing of what is
being done is now not much more verbose than the echo messages
were, and is more useful.

27 years agoFixed missing strip step in .S.o and .S.po rules.
Bruce Evans [Wed, 23 Apr 1997 10:26:18 +0000 (10:26 +0000)]
Fixed missing strip step in .S.o and .S.po rules.

Fixed back to front -X and -x strip flags in .m.o and .m.po rules.
Fixed disordered .m.o and .m.po rules.  What is .m?

Stripping probably should be removed.  It makes problems in library
functions hard to debug...

27 years agoRemoved inconsistent declaration of malloc(). `make world' with DESTDIR
Bruce Evans [Wed, 23 Apr 1997 10:16:58 +0000 (10:16 +0000)]
Removed inconsistent declaration of malloc().  `make world' with DESTDIR
set to something other than "/" now actually works.

27 years agoAdd basic login.conf (sans authentication) support.
David Nugent [Wed, 23 Apr 1997 04:56:39 +0000 (04:56 +0000)]
Add basic login.conf (sans authentication) support.

27 years agoSpelling police.
Daniel O'Callaghan [Wed, 23 Apr 1997 04:46:38 +0000 (04:46 +0000)]
Spelling police.

27 years agoAdded login.conf support.
David Nugent [Wed, 23 Apr 1997 03:06:47 +0000 (03:06 +0000)]
Added login.conf support.

27 years agoRemove sef's email address.
Jordan K. Hubbard [Wed, 23 Apr 1997 02:54:28 +0000 (02:54 +0000)]
Remove sef's email address.

27 years agoGive up on the fast vfork() for a while.
John Dyson [Wed, 23 Apr 1997 01:59:14 +0000 (01:59 +0000)]
Give up on the fast vfork() for a while.

27 years agoCheck that the received packet length indicated by the card is at least
David Greenman [Wed, 23 Apr 1997 01:44:30 +0000 (01:44 +0000)]
Check that the received packet length indicated by the card is at least
large enough to contain the ethernet header. There appears to be a
condition where the card can return "0" in some failure cases, and this
causes bad things to happen (a panic).

27 years agoUpdate to version 1.9.
Jordan K. Hubbard [Tue, 22 Apr 1997 18:02:52 +0000 (18:02 +0000)]
Update to version 1.9.
Submitted by: Shigio Yamaguchi <shigio@wafu.netgate.net>

27 years agoRemoved wrong declarations of malloc() and realloc(). They were unused,
Bruce Evans [Tue, 22 Apr 1997 17:44:34 +0000 (17:44 +0000)]
Removed wrong declarations of malloc() and realloc().  They were unused,
and they broke when yacc started generating `#include <stdlib.h>', but
the breakage was hidden by compatibility cruft in gcc unless DESTDIR
was unusual.

27 years agoFix broken usage of nm_readdirsize and increase the socket buffers for UDP
Doug Rabson [Tue, 22 Apr 1997 17:38:01 +0000 (17:38 +0000)]
Fix broken usage of nm_readdirsize and increase the socket buffers for UDP
to prevent possible socket overflows.

2.2 candidate.

PR: kern/3304
Reviewed by: Thomas David Rivers <ponds!rivers@dg-rtp.dg.com>

27 years agoSynchronize with sys/i386/conf/Makefile.i386, sys/i386/i386/machdep.c
KATO Takenori [Tue, 22 Apr 1997 12:20:50 +0000 (12:20 +0000)]
Synchronize with sys/i386/conf/Makefile.i386, sys/i386/i386/machdep.c
and sys/i386/isa/npx.c revisions 1.94, 1.238 and 1.41, respectively.

27 years agoAdd a -2 flag to mountd to allow an admin to disable NFSv3 services.
Doug Rabson [Tue, 22 Apr 1997 10:37:27 +0000 (10:37 +0000)]
Add a -2 flag to mountd to allow an admin to disable NFSv3 services.
This is handy for testing and possibly to work around busted v3 clients.

27 years agoUpdate the host filesystem so that it works with NFSv3
Doug Rabson [Tue, 22 Apr 1997 10:24:29 +0000 (10:24 +0000)]
Update the host filesystem so that it works with NFSv3

27 years agoFIxed the cleanup. I forgot to leave stdin alone in the usual (!twoway)
Bruce Evans [Tue, 22 Apr 1997 09:44:06 +0000 (09:44 +0000)]
FIxed the cleanup.  I forgot to leave stdin alone in the usual (!twoway)
case.

27 years agoMake the necessary changes so that an ELF kernel can be built. I
John Polstra [Tue, 22 Apr 1997 06:55:47 +0000 (06:55 +0000)]
Make the necessary changes so that an ELF kernel can be built.  I
have successfully built, booted, and run a number of different ELF
kernel configurations, including GENERIC.  LINT also builds and
links cleanly, though I have not tried to boot it.

The impact on developers is virtually nil, except for two things.
All linker sets that might possibly be present in the kernel must be
listed in "sys/i386/i386/setdefs.h".  And all C symbols that are
also referenced from assembly language code must be listed in
"sys/i386/include/asnames.h".  It so happens that failure to do
these things will have no impact on the a.out kernel.  But it will
break the build of the ELF kernel.

The ELF bootloader works, but it is not ready to commit quite yet.

27 years agofix a bugon of last commit
John-Mark Gurney [Tue, 22 Apr 1997 05:52:54 +0000 (05:52 +0000)]
fix a bugon of last commit

27 years agoFixed typo in the set_keyboard_param().
Kazutaka YOKOTA [Tue, 22 Apr 1997 02:04:15 +0000 (02:04 +0000)]
Fixed typo in the set_keyboard_param().

Spotted by: Philippe Charnier@xp11.frmug.org

27 years agoIn these XXX MHz days, waiting 500ms for a process to do something is
Poul-Henning Kamp [Mon, 21 Apr 1997 20:32:11 +0000 (20:32 +0000)]
In these XXX MHz days, waiting 500ms for a process to do something is
really far too long.  Let us try 100ms instead, if you have a PP200,
maybe that's even too long.  This should speed up make -j# builds.
I wonder why SIGCHLD isn't used...

27 years agoDirty change in union_lock(). Sometimes upper vnode is locked without
KATO Takenori [Mon, 21 Apr 1997 15:32:24 +0000 (15:32 +0000)]
Dirty change in union_lock().  Sometimes upper vnode is locked without
UN_ULOCK flag.  This shows a locking violation but I couldn't find the
reason UN_ULOCK is not set or upper vnode is not unlocked.  I added
the code that detect this case and adjust un_flags.  DIAGNOSTIC kernel
doesn't adjust un_flags, but just panic here to help debug by kernel
hackers.

27 years agoLong usernames fixes
Andrey A. Chernov [Mon, 21 Apr 1997 13:53:47 +0000 (13:53 +0000)]
Long usernames fixes