]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoRemove unneeded include.
Poul-Henning Kamp [Mon, 13 May 2002 11:42:27 +0000 (11:42 +0000)]
Remove unneeded include.

Sponsored by: DARPA & NAI Labs.
Submitted by: mckusick

22 years agoMark all internal libraries with INTERNALLIB.
Ruslan Ermilov [Mon, 13 May 2002 11:24:03 +0000 (11:24 +0000)]
Mark all internal libraries with INTERNALLIB.

22 years agoRemoved now unused INTERNALSTATICLIB.
Ruslan Ermilov [Mon, 13 May 2002 11:09:07 +0000 (11:09 +0000)]
Removed now unused INTERNALSTATICLIB.
INTERNALLIB now implies NOPIC and NOPROFILE.
Removed gratuitous NOMAN.

22 years agoMajor cleanup of bsd.lib.mk.
Ruslan Ermilov [Mon, 13 May 2002 10:53:24 +0000 (10:53 +0000)]
Major cleanup of bsd.lib.mk.

Get rid of the INTERNALSTATICLIB knob and just use plain INTERNALLIB.
INTERNALLIB now means to build static library only and don't install
anything.  Added a NOINSTALLLIB knob for libpam/modules.  To not
build any library at all, just do not set LIB.

22 years agoAdd ipfw hooks to ether_demux() and ether_output_frame().
Luigi Rizzo [Mon, 13 May 2002 10:37:19 +0000 (10:37 +0000)]
Add ipfw hooks to ether_demux() and ether_output_frame().
Ipfw processing of frames at layer 2 can be enabled by the sysctl variable

net.link.ether.ipfw=1

Consider this feature experimental, because right now, the firewall
is invoked in the places indicated below, and controlled by the
sysctl variables listed on the right.  As a consequence, a packet
can be filtered from 1 to 4 times depending on the path it follows,
which might make a ruleset a bit hard to follow.

I will add an ipfw option to tell if we want a given rule to apply
to ether_demux() and ether_output_frame(), but we have run out of
flags in the struct ip_fw so i need to think a bit on how to implement
this.

to upper layers
     |      |
     +----------->-----------+
     ^      V
[ip_input] [ip_output] net.inet.ip.fw.enable=1
     |      |
     ^      V
[ether_demux]      [ether_output_frame] net.link.ether.ipfw=1
     |      |
     +->- [bdg_forward]-->---+ net.link.ether.bridge_ipfw=1
     ^      V
     |      |
 to devices

22 years agostyle(9) changes before further editing that region:
Joerg Wunsch [Mon, 13 May 2002 10:28:20 +0000 (10:28 +0000)]
style(9) changes before further editing that region:

. add unnecessary parenthesis around return values
. put body of an "if" statement onto a line of its own

22 years agoRemove custom definitions (IP_FW_TCPF_SYN etc.) of TCP header flags
Luigi Rizzo [Mon, 13 May 2002 10:21:13 +0000 (10:21 +0000)]
Remove custom definitions (IP_FW_TCPF_SYN etc.) of TCP header flags
which are the same as the original ones (TH_SYN etc.)

22 years agoHandle symbolic names for common ethernet types (ip, arp etc.)
Luigi Rizzo [Mon, 13 May 2002 10:19:59 +0000 (10:19 +0000)]
Handle symbolic names for common ethernet types (ip, arp etc.)

Remove custom definitions (IP_FW_TCPF_SYN etc.) of TCP header flags
which are the same as the original ones (TH_SYN etc.)

22 years agoRemove register keyword.
Poul-Henning Kamp [Mon, 13 May 2002 09:22:31 +0000 (09:22 +0000)]
Remove register keyword.

Sponsored by: DARPA & NAI Labs.
Submitted by: mckusick

22 years agoRemoved the a.out compatibility cruft.
Ruslan Ermilov [Mon, 13 May 2002 09:01:18 +0000 (09:01 +0000)]
Removed the a.out compatibility cruft.

libgnumalloc.so.2 and libresolv.so.2 should be put under
lib/compat/compat2?/ but I don't have the 2.x releases.

22 years agoo Uncapitalize .Nd macro arguments and remove punctuation
Giorgos Keramidas [Mon, 13 May 2002 08:40:07 +0000 (08:40 +0000)]
o  Uncapitalize .Nd macro arguments and remove punctuation
o  Minor grammar fixes.
o  Sort SEE ALSO references, and add iostat(8).
o  Delete punctuation at end of AUTHORS' section only line

Reviewed by: rwatson, Hiten Pandya <hiten@uk.FreeBSD.org>

22 years agoRemove EOL space.
Giorgos Keramidas [Mon, 13 May 2002 08:37:46 +0000 (08:37 +0000)]
Remove EOL space.
This is a whitespace only change.

Reviewed by: rwatson, Hiten Pandya <hiten@uk.FreeBSD.org>

22 years agoIf we can't change the scheduling priority, warn but still execute
Tim J. Robbins [Mon, 13 May 2002 08:03:20 +0000 (08:03 +0000)]
If we can't change the scheduling priority, warn but still execute
the requested utility. This is how nice(1) traditionall behaved,
and the behaviour required by SUSv3 and POSIX.2 UPE.

Submitted by: Peter Avalos <pavalos@theshell.com> (partially)
Reviewed by: mike

22 years agoDon't use PAGE_SIZE in userland, instead use getpagesize(), this is to
Alfred Perlstein [Mon, 13 May 2002 07:58:15 +0000 (07:58 +0000)]
Don't use PAGE_SIZE in userland, instead use getpagesize(), this is to
allow running on other arches when the instructions are supported but
the page size granularity is not.

Glanced at by: peter

22 years agoBuild the fpu support routines.
Benno Rice [Mon, 13 May 2002 07:53:22 +0000 (07:53 +0000)]
Build the fpu support routines.

22 years agoFPU support.
Benno Rice [Mon, 13 May 2002 07:44:48 +0000 (07:44 +0000)]
FPU support.

Obtained from: NetBSD (portions)

22 years agoBump for GCC 3.1.
David E. O'Brien [Mon, 13 May 2002 07:14:17 +0000 (07:14 +0000)]
Bump for GCC 3.1.

22 years agoMatch the default newfs UFS block size.
David E. O'Brien [Mon, 13 May 2002 05:09:34 +0000 (05:09 +0000)]
Match the default newfs UFS block size.

22 years agoDon't call the uz free function while the zone lock is held. This can lead
Jeff Roberson [Mon, 13 May 2002 05:08:18 +0000 (05:08 +0000)]
Don't call the uz free function while the zone lock is held.  This can lead
to lock order reversals.  uma_reclaim now builds a list of freeable slabs and
then unlocks the zones to do all of the frees.

22 years agos/_ALPHA_/_MACHINE_/
Marcel Moolenaar [Mon, 13 May 2002 05:01:05 +0000 (05:01 +0000)]
s/_ALPHA_/_MACHINE_/

22 years agoRemove reference to the "Alpha Calling Standard".
Marcel Moolenaar [Mon, 13 May 2002 04:58:12 +0000 (04:58 +0000)]
Remove reference to the "Alpha Calling Standard".

22 years agoRemove the hash_free() lock order reversal. This could have happened for
Jeff Roberson [Mon, 13 May 2002 04:39:28 +0000 (04:39 +0000)]
Remove the hash_free() lock order reversal.  This could have happened for
several reasons before.  Fixing it involved restructuring the generic hash
code to require calling code to handle locking, unlocking, and freeing hashes
on error conditions.

22 years agoHandle alignment fault fixups in libc rather than in the kernel.
Jake Burkholder [Mon, 13 May 2002 04:35:08 +0000 (04:35 +0000)]
Handle alignment fault fixups in libc rather than in the kernel.

22 years agoFix IF_SEXT(val, 32). The constants need to have type long to
Jake Burkholder [Mon, 13 May 2002 04:26:38 +0000 (04:26 +0000)]
Fix IF_SEXT(val, 32).  The constants need to have type long to
handle size > 16.

22 years agoThis commit was generated by cvs2svn to compensate for changes in r96489,
David E. O'Brien [Mon, 13 May 2002 03:36:15 +0000 (03:36 +0000)]
This commit was generated by cvs2svn to compensate for changes in r96489,
which included commits to RCS files with non-trunk default branches.

22 years agoCause CVS to sync up checked out source bases with Gcc 3.1 bits after
David E. O'Brien [Mon, 13 May 2002 03:36:15 +0000 (03:36 +0000)]
Cause CVS to sync up checked out source bases with Gcc 3.1 bits after
the repo surgery to cut out the abortive Gcc 2.9[67] imports.

22 years agoRestore some of the implementation from the Bmake gcc 2.95 bits.
David E. O'Brien [Mon, 13 May 2002 03:27:03 +0000 (03:27 +0000)]
Restore some of the implementation from the Bmake gcc 2.95 bits.
In the end, I can do things more like the previous Bmake bits than was
apparent in the middle of the gcc31 WIP.

22 years agoThese were repo-copied to dump_machdep.c.
Jake Burkholder [Mon, 13 May 2002 02:52:35 +0000 (02:52 +0000)]
These were repo-copied to dump_machdep.c.

22 years ago${MACHINE_ARCH}dump.c -> dump_machdep.c.
Jake Burkholder [Mon, 13 May 2002 02:40:21 +0000 (02:40 +0000)]
${MACHINE_ARCH}dump.c -> dump_machdep.c.

22 years agoAdd another copy of the ia64 dump_machdep.c.
Jake Burkholder [Mon, 13 May 2002 02:39:19 +0000 (02:39 +0000)]
Add another copy of the ia64 dump_machdep.c.

22 years agoI was finally able to repeat the -j breakage on one of my machines. Fix it.
David E. O'Brien [Mon, 13 May 2002 01:54:26 +0000 (01:54 +0000)]
I was finally able to repeat the -j breakage on one of my machines.  Fix it.
I borrowed some ideas from Ruslan, and made the style match cc_tools/Makefile

22 years agoUFS2 preparation commit:
Poul-Henning Kamp [Sun, 12 May 2002 23:44:15 +0000 (23:44 +0000)]
UFS2 preparation commit:
Remove support for converting old FFS formats to newer.

Submitted by: mckusick
Sponspored by: DARPA & NAI Labs.

22 years agoRemove two "register" and a blank line.
Poul-Henning Kamp [Sun, 12 May 2002 22:54:48 +0000 (22:54 +0000)]
Remove two "register" and a blank line.

Submitted by: mckusick
Sponsored by: DARPA & NAI Labs.

22 years agoSupport the SUSv3 -n option and the "--" end of options marker.
Tim J. Robbins [Sun, 12 May 2002 22:49:48 +0000 (22:49 +0000)]
Support the SUSv3 -n option and the "--" end of options marker.
Replace "command" with "utility" in the manual page & source to be more
consistent with the terminology used in the standard, and to hint that
shell builtin commands won't work.

Submitted by: Peter Avalos <pavalos@theshell.com> (partially)
Approved by: mike

22 years agoRemove files not part of GCC 3.1.
David E. O'Brien [Sun, 12 May 2002 21:50:48 +0000 (21:50 +0000)]
Remove files not part of GCC 3.1.

22 years agoSigh... Yet a BBSIZE breakage.
Poul-Henning Kamp [Sun, 12 May 2002 21:43:02 +0000 (21:43 +0000)]
Sigh... Yet a BBSIZE breakage.

22 years agoSigh, more BBSIZE related breakage.
Poul-Henning Kamp [Sun, 12 May 2002 21:37:08 +0000 (21:37 +0000)]
Sigh, more BBSIZE related breakage.

Sponsored by: DARPA & NAI Labs.

22 years ago#include <sys/disklabel.h> to get BBSIZE.
Poul-Henning Kamp [Sun, 12 May 2002 20:58:18 +0000 (20:58 +0000)]
#include <sys/disklabel.h> to get BBSIZE.

22 years agoMain functional change is the implementation of matching of MAC header
Luigi Rizzo [Sun, 12 May 2002 20:52:21 +0000 (20:52 +0000)]
Main functional change is the implementation of matching of MAC header
fields as discussed in the commit to ip_fw.c:1.186

On top of this, a ton of non functional changes to clean up the code,
write functions to replace sections of code that were replicated
multiple times (e.g. the printing or matching of flags and options),
splitting long sections of inlined code into separate functions,
and the like.

I have tested the code quite a bit, but some typos (using one variable
in place of another) might have escaped.

The "embedded manpage" is a bit inconsistent, but i am leaving fixing
it for later. The current format makes no sense, it is over 40 lines
long and practically unreadable. We can either split it into sections
( ipfw -h options  , ipfw -h pipe , ipfw -h queue ...)
or remove it altogether and refer to the manpage.

22 years agoRetire the bogus uses of the disklabel field d_sbsize and begin to
Poul-Henning Kamp [Sun, 12 May 2002 20:49:41 +0000 (20:49 +0000)]
Retire the bogus uses of the disklabel field d_sbsize and begin to
initialize it to zero so we don't have to have everbody and their
aunt including FFS specific header files.

Sponsored by: DARPA & NAI Labs.

22 years agoAdd code to match MAC header fields (at the moment supported on
Luigi Rizzo [Sun, 12 May 2002 20:43:50 +0000 (20:43 +0000)]
Add code to match MAC header fields (at the moment supported on
bridged packets only, soon to come also for packets on ordinary
ether_input() and ether_output() paths. The syntax is

    ipfw add <action> MAC dst src type

where dst and src can be "any" or a MAC address optionallyfollowed
by a mask, e.g.

10:20:30:40:50
10:20:30:40:50/32
10:20:30:40:50&ff:ff:ff:f0:ff:0f

and type can be a single ethernet type, a range, or a type followed by
a mask (values are always in hexadecimal) e.g.

0800
0800-0806
0800/8
0800&03ff

Note, I am still uncertain on what is the best format for inputting
these values, having the values in hexadecimal is convenient in most
cases but can be confusing sometimes. Suggestions welcome.

Implement suggestion from PR 37778 to allow "not me" on destination
and source IP. The code in the PR was slightly wrong and interfered
with the normal handling of IP addresses. This version hopefully is
correct.

Minor cleanup of the code, in some places moving the indentation to 4
spaces because the code was becoming too deep. Eventually, in a
separate commit, I will move the whole file to 4 space indent.

22 years agoARGH! SBLOCK is not unused. Try to get this right.
Poul-Henning Kamp [Sun, 12 May 2002 20:21:40 +0000 (20:21 +0000)]
ARGH!  SBLOCK is not unused.  Try to get this right.

BBSIZE belongs in <sys/disklabel.h> (but shouldn't be a constant).

Define SBLOCK again, using the right math.

Sponsored by: DARPA & NAI Labs.

22 years agoRemove #define for BBOFF, it is assumed == 0 so many places that we might
Poul-Henning Kamp [Sun, 12 May 2002 20:00:21 +0000 (20:00 +0000)]
Remove #define for BBOFF, it is assumed == 0 so many places that we might
as well forget about it.  In fact the only thing which used it was the
SBOFF macro.

Sponsored by: DARPA & NAI Labs.

22 years agoRemove unused BBLOCK and SBLOCK #defines.
Poul-Henning Kamp [Sun, 12 May 2002 19:56:31 +0000 (19:56 +0000)]
Remove unused BBLOCK and SBLOCK #defines.

Sponsored by: DARPA & NAI Labs.

22 years agoRemove the private code for reading UFS superblocks, this does not belong
Poul-Henning Kamp [Sun, 12 May 2002 19:24:09 +0000 (19:24 +0000)]
Remove the private code for reading UFS superblocks, this does not belong
in df(1) when we have multiple filesystem types, and the complications of
handling UFS2 pushes this over the edge.

Use the .../mount/extern.h to get prototypes of the functions we
borrow from there.  Constify things to match.  (why aren't these
functions in a lib anyway ?)

Make everything static and set WARNS?=5.

The way the "df diskdevice" thing works for unmounted diskdevices
is not very general.

Sponsored by: DARPA & NAI Labs.

22 years ago o Remove GIANT_REQUIRED and an excessive number of blank lines
Alan Cox [Sun, 12 May 2002 18:42:05 +0000 (18:42 +0000)]
 o Remove GIANT_REQUIRED and an excessive number of blank lines
   from vm_map_inherit().  (minherit() need not acquire Giant
   anymore.)

22 years agoFix alpha build. The alpha has dumpsys implemented.
Marcel Moolenaar [Sun, 12 May 2002 18:27:28 +0000 (18:27 +0000)]
Fix alpha build. The alpha has dumpsys implemented.
While here, revert the condition to list the machines
for which dumpsys has not been implemented.

Reported by: wilko

22 years agoEnable KTR_TRAP by default.
Jake Burkholder [Sun, 12 May 2002 18:10:48 +0000 (18:10 +0000)]
Enable KTR_TRAP by default.

22 years agodo-while ASM_OUTPUT_ALIGN to enable its use in more places.
David E. O'Brien [Sun, 12 May 2002 17:31:12 +0000 (17:31 +0000)]
do-while ASM_OUTPUT_ALIGN to enable its use in more places.

22 years agoHide the fd autoselection messages behind the bootverbose case so they
Joerg Wunsch [Sun, 12 May 2002 17:17:46 +0000 (17:17 +0000)]
Hide the fd autoselection messages behind the bootverbose case so they
stop bothering people on their consoles.

22 years agoNOLIB is gone; put INTERNALLIB back for now.
Ruslan Ermilov [Sun, 12 May 2002 16:35:34 +0000 (16:35 +0000)]
NOLIB is gone; put INTERNALLIB back for now.

22 years agoRevert the last change. The corresponding bsd.lib.mk changes were
Ruslan Ermilov [Sun, 12 May 2002 16:21:23 +0000 (16:21 +0000)]
Revert the last change.  The corresponding bsd.lib.mk changes were
already backed out.

22 years agoAdded new bsd.incs.mk which handles installing of header files
Ruslan Ermilov [Sun, 12 May 2002 16:01:00 +0000 (16:01 +0000)]
Added new bsd.incs.mk which handles installing of header files
via INCS.  Implemented INCSLINKS (equivalent to SYMLINKS) to
handle symlinking include files.  Allow for multiple groups of
include files to be installed, with the powerful INCSGROUPS knob.
Documentation to follow.

Added standard `includes' and `incsinstall' targets, use them
in Makefile.inc1.  Headers from the following makefiles were
not installed before (during `includes' in Makefile.inc1):

kerberos5/lib/libtelnet/Makefile
lib/libbz2/Makefile
lib/libdevinfo/Makefile
lib/libform/Makefile
lib/libisc/Makefile
lib/libmenu/Makefile
lib/libmilter/Makefile
lib/libpanel/Makefile

Replaced all `beforeinstall' targets for installing includes
with the INCS stuff.

Renamed INCDIR to INCSDIR, for consistency with FILES and SCRIPTS,
and for compatibility with NetBSD.  Similarly for INCOWN, INCGRP,
and INCMODE.

Consistently use INCLUDEDIR instead of /usr/include.

gnu/lib/libstdc++/Makefile and gnu/lib/libsupc++/Makefile changes
were only lightly tested due to the missing contrib/libstdc++-v3.
I fully tested the pre-WIP_GCC31 version of this patch with the
contrib/libstdc++.295 stuff.

These changes have been tested on i386 with the -DNO_WERROR "make
world" and "make release".

22 years agoTranslated -malign-functions=4 to -falign-functions=16 for the new gcc.
Bruce Evans [Sun, 12 May 2002 15:51:38 +0000 (15:51 +0000)]
Translated -malign-functions=4 to -falign-functions=16 for the new gcc.

22 years agoSaved 176 bytes by compiling with -fno-guess-branch-probability. The
Bruce Evans [Sun, 12 May 2002 15:45:28 +0000 (15:45 +0000)]
Saved 176 bytes by compiling with -fno-guess-branch-probability.  The
default of -fguess-branch-probablility causes time optimizations (?)
like rewriting `if (foo) x++;' as
`if (!foo) goto forth; back: ; ...; forth: x++; goto back;".  This is
pessimizes space especially well on i386's because one short branch
gets converted to 2 long ones.

Removed -fno-align-foo since it is implied by -Os.  Previous commit
messages seem to have overstated the new alignment bugs in gcc.  The
only case that affects boot2 is that -fno-align-functions (or
equivalently -falign-functions=1) actually gives -falign-functions=2.
This is caused by FUNCTION_BOUNDARY being 2 (bytes) instead of 1.
The default case where the optimization level is 1 and no alignment
options are given is more broken.  All alignments are minimal, modulo
the bug in FUNCTION_BOUNDARY.  This is caused by toplev.c setting
defaults too early.

Some hacks in previous commits ar not needed now, but may as well be
kept until gcc is fixed.  The previous on in the Makefile saved 96
bytes of text due to the wrong FUNCTION_BOUNDARY and 32 bytes of data
due to unrelated bloat in the alignment of large objects.  There aren't
even any options to control alignment of data.

22 years agoIncrease the PIO timeout to approximately the value it used to have
Joerg Wunsch [Sun, 12 May 2002 15:24:21 +0000 (15:24 +0000)]
Increase the PIO timeout to approximately the value it used to have
before rev 1.229 (~ 100 ms).  According to bde, some (old) broken
hardware could require it.  In order to make timing more accurate than
what could be achieved with a loop around DELAY(1), increase loop
timing after the initial ~ 1 ms.

Also, move the declaration of FDSTS_TIMEOUT out from fdreg.h into fd.c
where it actually belongs to.

MFC after: 2 days

22 years agoAdd a man page for the devinfo command; I'd argue devinfo is the single
Robert Watson [Sun, 12 May 2002 14:23:15 +0000 (14:23 +0000)]
Add a man page for the devinfo command; I'd argue devinfo is the single
most useful new command added to FreeBSD 5.0, but maybe that's just
me :-).

Submitted by: Hiten Pandya <hiten@uk.FreeBSD.org>

22 years agoRecent zlib does not like Z_FLUSH at the end of inflate().
Hajimu UMEMOTO [Sun, 12 May 2002 14:12:30 +0000 (14:12 +0000)]
Recent zlib does not like Z_FLUSH at the end of inflate().

Reported by: quak@mydiax.ch
Obtained from: KAME
MFC after: 2 days
and approved by re

22 years agoProperly build lb1spc.asm on Sparc64.
David E. O'Brien [Sun, 12 May 2002 14:05:58 +0000 (14:05 +0000)]
Properly build lb1spc.asm on Sparc64.

22 years agoBack out last commit. I expect our bsd.*.mk gods to remove the need for
David E. O'Brien [Sun, 12 May 2002 13:54:42 +0000 (13:54 +0000)]
Back out last commit.  I expect our bsd.*.mk gods to remove the need for
defining so many extra things in addition to INTERNALLIB.  We don't like
repetitive C code and we shouldn't for make code either.

22 years agoThere is no more NOLIB.
David E. O'Brien [Sun, 12 May 2002 13:49:14 +0000 (13:49 +0000)]
There is no more NOLIB.

22 years agoBack out rev 1.118.
David E. O'Brien [Sun, 12 May 2002 13:48:53 +0000 (13:48 +0000)]
Back out rev 1.118.

22 years agoMore locking fixes.
Benno Rice [Sun, 12 May 2002 13:43:21 +0000 (13:43 +0000)]
More locking fixes.

22 years agoDocument my willingness in maintaining share/mk infrastructure.
Ruslan Ermilov [Sun, 12 May 2002 13:41:28 +0000 (13:41 +0000)]
Document my willingness in maintaining share/mk infrastructure.

22 years agoQuiet the peanut gallary and back out magical "breakage".
David E. O'Brien [Sun, 12 May 2002 13:37:59 +0000 (13:37 +0000)]
Quiet the peanut gallary and back out magical "breakage".

22 years agoI was not strict enough with my ordering of things to satisfy make(1) nieve
David E. O'Brien [Sun, 12 May 2002 12:44:04 +0000 (12:44 +0000)]
I was not strict enough with my ordering of things to satisfy make(1) nieve
symbol evaluation which causes it to be very sensitive to macro ordering.

22 years agoTidy up the cleanfiles.
David E. O'Brien [Sun, 12 May 2002 12:06:19 +0000 (12:06 +0000)]
Tidy up the cleanfiles.

22 years agoFixes for building a.out bits.
David E. O'Brien [Sun, 12 May 2002 12:01:12 +0000 (12:01 +0000)]
Fixes for building a.out bits.

Submitted by: bde

22 years agoDo as you want with GCC.
David E. O'Brien [Sun, 12 May 2002 11:45:35 +0000 (11:45 +0000)]
Do as you want with GCC.

22 years agoAdd a prototype for rad_get_vendor_attr().
Brian Somers [Sun, 12 May 2002 10:53:00 +0000 (10:53 +0000)]
Add a prototype for rad_get_vendor_attr().

Forgotten by: brian

22 years agoDon't declare krb5_mcc_ops, it's already declared in <krb5.h>
Dag-Erling Smørgrav [Sun, 12 May 2002 07:06:27 +0000 (07:06 +0000)]
Don't declare krb5_mcc_ops, it's already declared in <krb5.h>

22 years agoDo the correct locking on processes for DSI and ISI traps.
Benno Rice [Sun, 12 May 2002 06:06:24 +0000 (06:06 +0000)]
Do the correct locking on processes for DSI and ISI traps.

Copied from: sparc64

22 years agoo Rename ia64_count_aps to ia64_count_cpus and reimplement the
Marcel Moolenaar [Sun, 12 May 2002 05:54:21 +0000 (05:54 +0000)]
o  Rename ia64_count_aps to ia64_count_cpus and reimplement the
   function to return the total number of CPUs and not the highest
   CPU id.
o  Define mp_maxid based on the minimum of the actual number of
   CPUs in the system and MAXCPU.
o  In cpu_mp_add, when the CPU id of the CPU we're trying to add
   is larger than mp_maxid, don't add the CPU. Formerly this was
   based on MAXCPU. Don't count CPUs when we add them. We already
   know how many CPUs exist.
o  Replace MAXCPU with mp_maxid when used in loops that iterate
   over the id space. This avoids a couple of useless iterations.
o  In cpu_mp_unleash, use the number of CPUs to determine if we
   need to launch the CPUs.
o  Remove mp_hardware as it's not used anymore.
o  Move the IPI vector array from mp_machdep.c to sal.c. We use
   the array as a centralized place to collect vector assignments.
   Note that we still assign vectors to SMP specific IPIs in
   non-SMP configurations. Rename the array from mp_ipi_vector to
   ipi_vector.
o  Add IPI_MCA_RENDEZ and IPI_MCA_CMCV. These are used by MCA.
   Note that IPI_MCA_CMCV is not SMP specific.
o  Initialize the ipi_vector array so that we place the IPIs in
   sensible priority classes. The classes are relative to where
   the AP wake-up vector is located to guarantee that it's the
   highest priority (external) interrupt. Class assignment is
   as follows:
class IPI notes
x AP wake-up (normally x=15)
x-1 MCA rendezvous
x-2 AST, Rendezvous, stop
x-3 CMCV, test

22 years ago o Acquire and release Giant in vm_object_reference() and
Alan Cox [Sun, 12 May 2002 05:22:56 +0000 (05:22 +0000)]
 o Acquire and release Giant in vm_object_reference() and
   vm_object_deallocate(), replacing the assertion GIANT_REQUIRED.
 o Remove GIANT_REQUIRED from vm_map_protect() and vm_map_simplify_entry().
 o Acquire and release Giant around vm_map_protect()'s call to pmap_protect().

Altogether, these changes eliminate the need for mprotect() to acquire
and release Giant.

22 years agoUpdate the PCI table as per Dell's fickleness.
Scott Long [Sun, 12 May 2002 04:30:28 +0000 (04:30 +0000)]
Update the PCI table as per Dell's fickleness.

MFC after: 3 days

22 years agoFixed 5 style bugs in VCS ids.
Bruce Evans [Sun, 12 May 2002 04:02:57 +0000 (04:02 +0000)]
Fixed 5 style bugs in VCS ids.

22 years agoFinish converting the rest of the function declaration and prototypes
Mike Barcroft [Sun, 12 May 2002 03:56:34 +0000 (03:56 +0000)]
Finish converting the rest of the function declaration and prototypes
to ANSI C.  Use new source ID scheme.

22 years agoFixed printing the the strip binary's name in error messages.
Bruce Evans [Sun, 12 May 2002 03:47:23 +0000 (03:47 +0000)]
Fixed printing the the strip binary's name in error messages.

22 years agoTurn on the set-user-ID bit for k5su if ENABLE_SUID_K5SU is defined.
Jacques Vidrine [Sun, 12 May 2002 03:17:31 +0000 (03:17 +0000)]
Turn on the set-user-ID bit for k5su if ENABLE_SUID_K5SU is defined.

22 years agoCVS_RSH defaults to 'ssh' these days, and not 'rsh'.
Giorgos Keramidas [Sun, 12 May 2002 02:12:04 +0000 (02:12 +0000)]
CVS_RSH defaults to 'ssh' these days, and not 'rsh'.

PR: docs/37429
Submitted by: Martin v. Löwis <martin@v.loewis.de>
Reviewed by: peter

22 years agoRemove _PATH_CP now that it is defined in paths.h
Chris D. Faulhaber [Sun, 12 May 2002 01:52:11 +0000 (01:52 +0000)]
Remove _PATH_CP now that it is defined in paths.h

Reviewed by: des

22 years agoTo perform even basic error checking, one must have an exit code that
Brian S. Dean [Sun, 12 May 2002 01:47:50 +0000 (01:47 +0000)]
To perform even basic error checking, one must have an exit code that
indicates that not everything worked as expected.  Exit non-zero if we
timed out while transmitting or receiving a file or if the file did
not exist, etc.

MFC After: 3 days (re@ willing)

22 years agos/demon/daemon/
Dima Dorfman [Sun, 12 May 2002 00:22:38 +0000 (00:22 +0000)]
s/demon/daemon/

22 years agoRemove some duplicate types that should have been removed as part of
Mike Barcroft [Sat, 11 May 2002 23:28:51 +0000 (23:28 +0000)]
Remove some duplicate types that should have been removed as part of
the rearranging in the previous revision.

Pointy hat to: cvs update (merging), mike (for not noticing)

22 years agoDocument that <netinet/in.h> can also provide prototypes. Change
Mike Barcroft [Sat, 11 May 2002 23:22:54 +0000 (23:22 +0000)]
Document that <netinet/in.h> can also provide prototypes.  Change
wording related to standards conformance.

22 years agodemon -> daemon
Dima Dorfman [Sat, 11 May 2002 23:20:53 +0000 (23:20 +0000)]
demon -> daemon

Submitted by: zhuravlev alexander <zaa@box.ulstu.ru>

22 years agoAttempt to not crash and burn on UltraSPARC III machines; the cpuid property
Jake Burkholder [Sat, 11 May 2002 22:05:22 +0000 (22:05 +0000)]
Attempt to not crash and burn on UltraSPARC III machines; the cpuid property
is named differently.

22 years agoRestore the ability to take crashdumps on alpha. This was cut and pasted
Andrew Gallatin [Sat, 11 May 2002 21:53:46 +0000 (21:53 +0000)]
Restore the ability to take crashdumps on alpha.   This was cut and pasted
nearly in its entirety from i386, so it retains the phk/nati copyright.

Savecore likes the results, but I have no way to test it as gdb is
still broken.

22 years agoSave about 60 bytes by #define memcpy __builtin_memcpy and removing
Peter Wemm [Sat, 11 May 2002 21:49:39 +0000 (21:49 +0000)]
Save about 60 bytes by #define memcpy __builtin_memcpy and removing
the function we provided.  Restore the Keyboard: yes/no  printf for the
probe diagnostics.  We end up with 40 bytes free.

22 years agoReconnect boot2. I'm sure I'll regret this though. :-)
Peter Wemm [Sat, 11 May 2002 21:41:03 +0000 (21:41 +0000)]
Reconnect boot2.  I'm sure I'll regret this though.  :-)

22 years agoUse a crowbar and duct-tape to make boot2 fit again. This gets it down
Peter Wemm [Sat, 11 May 2002 21:39:59 +0000 (21:39 +0000)]
Use a crowbar and duct-tape to make boot2 fit again.  This gets it down
to 4 bytes free.  I removed a printf (the Keyboard yes/no) since it is of
marginal value and sed'ed the generated asm output to remove the unwanted
aligns.  There's probably a better way to gain a few extra bytes than
losing the printf.  Shortening strings is probably a better option but this
should get us over the hurdle.

22 years agoChange the disk probing so that it will actually find disks other
Jake Burkholder [Sat, 11 May 2002 21:30:46 +0000 (21:30 +0000)]
Change the disk probing so that it will actually find disks other
than the first one on a controller, and work for secondary
controllers.
Due to the prom not having nodes for each disk, but a catch-all one,
we have to iterate over each device, trying to open it to determine
whether it is actually present.
Since probing this way takese some time (and spews some spurious
warnings), it should maybe be short-circuited if we use the
device we were booted from.
Implement lazy device probing, and correct slice/partiniton
handling in the ofwd_open() code. With this, I can now actually boot
a kernel from disk, and the loader does not create unnecessary
delays.

Submitted by: tmm

22 years agoAdd a support macro to convert the 5-bit packed register field of
Jake Burkholder [Sat, 11 May 2002 21:20:05 +0000 (21:20 +0000)]
Add a support macro to convert the 5-bit packed register field of
a floating point instruction into a 6-bit register number for
double and quad arguments.
Make use of the new INSFPdq_RN macro where apporpriate; this
is required for correctly handling the "high" fp registers
(>= %f32).
Fix a number of bugs related to the handling of the high registers
which were caused by using __fpu_[gs]etreg() where __fpu_[gs]etreg64()
should be used (the former can only access the low, single-precision,
registers).

Submitted by: tmm

22 years agoAdd pointers to GCC's allowable values for -march, and restore structure
David E. O'Brien [Sat, 11 May 2002 20:18:54 +0000 (20:18 +0000)]
Add pointers to GCC's allowable values for -march, and restore structure
of rev 1.7 until someone can sit down and think thru all the GCC 3.1
related changes.

22 years agoExplicitely note ATEN UC-232A is supported. Sort list entries
Wilko Bulte [Sat, 11 May 2002 20:10:38 +0000 (20:10 +0000)]
Explicitely note ATEN UC-232A is supported. Sort list entries
while we are here.

22 years agoWith GCC 3.1, we can now treat AMD Athlon and an Athlon.
David E. O'Brien [Sat, 11 May 2002 20:05:17 +0000 (20:05 +0000)]
With GCC 3.1, we can now treat AMD Athlon and an Athlon.

Submitted by: Steven G. Kargl <kargl@troutmask.apl.washington.edu>

22 years agoFix a typo : ".Xr mount_udf 8," -> ".Xr mount_udf 8 ,".
Maxime Henrion [Sat, 11 May 2002 19:32:24 +0000 (19:32 +0000)]
Fix a typo : ".Xr mount_udf 8," -> ".Xr mount_udf 8 ,".

22 years agoSome features should have been marked as &merged;, but weren't. Fix
Bruce A. Mah [Sat, 11 May 2002 19:23:17 +0000 (19:23 +0000)]
Some features should have been marked as &merged;, but weren't.  Fix
this.

While I'm here, add a few missing words.

22 years agoAdd support for stripping non-native object files by using an
Marcel Moolenaar [Sat, 11 May 2002 19:15:15 +0000 (19:15 +0000)]
Add support for stripping non-native object files by using an
environment variable that specifies the name of the strip(1)
program to use. The envvar is "STRIPBIN". The more natural
choice would be "STRIP", but that one is taken already.

22 years agoNOPIC, NOPROFILE, NOMAN, and INTERNALSTATICLIB are redundant when using
David E. O'Brien [Sat, 11 May 2002 18:02:33 +0000 (18:02 +0000)]
NOPIC, NOPROFILE, NOMAN, and INTERNALSTATICLIB are redundant when using
INTERNALLIB now.