]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
26 years agoAdd a FreeBSD entry for alpha-elf.
jb [Mon, 9 Mar 1998 03:34:53 +0000 (03:34 +0000)]
Add a FreeBSD entry for alpha-elf.

26 years agoDon't disable the use of $ in assembler labels.. It changes libstdc++
peter [Mon, 9 Mar 1998 03:26:57 +0000 (03:26 +0000)]
Don't disable the use of $ in assembler labels..  It changes libstdc++
and libg++ so that they won't work with existing binaries (including
netscape!!).

26 years agoCorrect an ethernet framesize mismatch that caused poor
steve [Mon, 9 Mar 1998 03:07:54 +0000 (03:07 +0000)]
Correct an ethernet framesize mismatch that caused poor
device performance among other things.

PR: 4989, 5910
Submitted by: Yoshikazu Goto <gotoh@ae.anritsu.co.jp>

26 years agoRemove extraneous trailing \'s.
steve [Mon, 9 Mar 1998 03:01:47 +0000 (03:01 +0000)]
Remove extraneous trailing \'s.

PR: 5949
Submitted by: Studded@dal.net

26 years agoRemove duplicate entries by changing tcp -> udp.
steve [Mon, 9 Mar 1998 02:56:21 +0000 (02:56 +0000)]
Remove duplicate entries by changing tcp -> udp.

PR: 2882

26 years agoTypo fixes.
jkh [Mon, 9 Mar 1998 02:21:17 +0000 (02:21 +0000)]
Typo fixes.
PR: 5951
Submitted by: Doug <Studded@dal.net>

26 years agoNuke ircd entries - nobody can agree on what constitutes a reasonable
jkh [Mon, 9 Mar 1998 02:16:58 +0000 (02:16 +0000)]
Nuke ircd entries - nobody can agree on what constitutes a reasonable
range for them and I'm %&#$%@! sorry I ever added them in the first place.

26 years agoThis commit was generated by cvs2svn to compensate for changes in r34336,
jb [Mon, 9 Mar 1998 00:33:28 +0000 (00:33 +0000)]
This commit was generated by cvs2svn to compensate for changes in r34336,
which included commits to RCS files with non-trunk default branches.

26 years agoImport of alpha specific bits to configure binutils for FreeBSD/Alpha.
jb [Mon, 9 Mar 1998 00:33:28 +0000 (00:33 +0000)]
Import of alpha specific bits to configure binutils for FreeBSD/Alpha.
Other changes that affect the i386-elf work are on hold to avoid
stuffing up other work in progress.

26 years agoUpdate to build binutils (yay!), so that ar, ld, as, strip, nm, objcopy,
jb [Mon, 9 Mar 1998 00:25:30 +0000 (00:25 +0000)]
Update to build binutils (yay!), so that ar, ld, as, strip, nm, objcopy,
objdump all build out of the FreeBSD tree. At EDS we used to call this
a "significant emotional event".

Still using the lorder script from NetBSD until I can explain why the
is a difference in nm behaviour when the output looks the same.

26 years agoPreprocessor directives require a leading '#'
dufault [Mon, 9 Mar 1998 00:15:08 +0000 (00:15 +0000)]
Preprocessor directives require a leading '#'

Submitted by: ccsanady@friley585.res.iastate.edu

26 years agoFix a type mismatch between a scanf format string and the
jdp [Sun, 8 Mar 1998 23:57:00 +0000 (23:57 +0000)]
Fix a type mismatch between a scanf format string and the
corresponding argument.

26 years agoExplicitly call `caused' an `int'.
brian [Sun, 8 Mar 1998 23:39:33 +0000 (23:39 +0000)]
Explicitly call `caused' an `int'.
Sugested by: gcc-2.8

26 years agoCan build libc and libc_r on alpha now.
jb [Sun, 8 Mar 1998 23:24:05 +0000 (23:24 +0000)]
Can build libc and libc_r on alpha now.

Changed MACHINE to MACHINE_ARCH with the expectation that pc98 will
use elf the same as i386.

Nuked tahoe and vax 'cause the files they reference aren't in the
tree. If you want vax goto NetBSD. If you want tahoe... tough.

26 years agoRemove AIO_LISTIO_MAX until it is done properly.
dufault [Sun, 8 Mar 1998 22:29:56 +0000 (22:29 +0000)]
Remove AIO_LISTIO_MAX until it is done properly.

26 years agoPut sigevent and AIO_LISTIO_MAX back in aio.h so
dufault [Sun, 8 Mar 1998 22:21:12 +0000 (22:21 +0000)]
Put sigevent and AIO_LISTIO_MAX back in aio.h so
that kernels can be built.

26 years agoMy sched.h is getting walloped by Peter Dufault's. Nuke mine. Sorry.
jb [Sun, 8 Mar 1998 21:36:41 +0000 (21:36 +0000)]
My sched.h is getting walloped by Peter Dufault's. Nuke mine. Sorry.

26 years agoOur default rules tend to shove asm code through cpp, so comments
jb [Sun, 8 Mar 1998 21:09:04 +0000 (21:09 +0000)]
Our default rules tend to shove asm code through cpp, so comments
starting with hashes have a sub-optimal impact. This change adds
/* */ around the block comment in the header of each file to make
them friendly to cpp. Also added an Id keyword cause I like to
see revision numbers in source.

26 years agoLocalize it
ache [Sun, 8 Mar 1998 20:56:43 +0000 (20:56 +0000)]
Localize it

26 years agoMake this compile again.
steve [Sun, 8 Mar 1998 19:03:05 +0000 (19:03 +0000)]
Make this compile again.

PR: 5948
Submitted by: Brian Feldman

26 years agoQuell unneeded pageout daemon activity.
dyson [Sun, 8 Mar 1998 18:19:17 +0000 (18:19 +0000)]
Quell unneeded pageout daemon activity.

26 years agoRemove a very ill advised vm_page_protect. This was being called
dyson [Sun, 8 Mar 1998 18:05:59 +0000 (18:05 +0000)]
Remove a very ill advised vm_page_protect.  This was being called
for a non-managed page.  That is a big no-no.

26 years agoReviewed by: bde
dufault [Sun, 8 Mar 1998 17:25:38 +0000 (17:25 +0000)]
Reviewed by: bde

Changes to support building with _POSIX_SOURCE set to 199309L:

1. Add sys/_posix.h to handle those preprocessor defs that POSIX
says have effects when defined before including any header files;

2. Change POSIX4_VISIBLE back to _POSIX4_VISIBLE

3. Add _POSIX4_VISIBLE_HISTORICALLY for pre-existing BSD features now
defined in POSIX.  These show up when:

_POSIX_SOURCE and _POSIX_C_SOURCE are not set or
_POSIX_C_SOURCE is set >= 199309L

and vanish when:

_POSIX_SOURCE is set or _POSIX_C_SOURCE is < 199309L.

4. Explain these in man 9 posix4;

5. Include _posix.h and conditionalize on new feature test.

26 years agoUse cc -E rather than cpp when looking for __ELF__
peter [Sun, 8 Mar 1998 17:25:10 +0000 (17:25 +0000)]
Use cc -E rather than cpp when looking for __ELF__

26 years agomerge benfor branch changes onto mainline.
peter [Sun, 8 Mar 1998 16:54:00 +0000 (16:54 +0000)]
merge benfor branch changes onto mainline.

26 years agoSpell -Bshareable correctly... :-]
peter [Sun, 8 Mar 1998 16:09:57 +0000 (16:09 +0000)]
Spell -Bshareable correctly... :-]

26 years agoChange default temporary directory (TMPDIR) from /tmp to /var/tmp.
wosch [Sun, 8 Mar 1998 16:09:31 +0000 (16:09 +0000)]
Change default temporary directory (TMPDIR) from /tmp to /var/tmp.

Mosts users seems to have a larger /var/tmp partition
than /tmp - IMHO silly but who cares?

26 years agoImprove shell variable substitution.
wosch [Sun, 8 Mar 1998 15:28:37 +0000 (15:28 +0000)]
Improve shell variable substitution.

26 years agoFixed a few ancient typos, added a little missing stuff, and updated
mckay [Sun, 8 Mar 1998 15:15:33 +0000 (15:15 +0000)]
Fixed a few ancient typos, added a little missing stuff, and updated
references to abort() in light of POSIX mandated behaviour.  I'm
still not 100% happy with much of the wording, but it's better
than it was.

26 years agoConstruct the minor number for the root device taking into account the
msmith [Sun, 8 Mar 1998 15:06:55 +0000 (15:06 +0000)]
Construct the minor number for the root device taking into account the
slice number passed in by the bootblocks.  This means the kernel will
not use the compatability slice to obtain the root filesystem when
booting from a sliced disk.

Use the extraction macros from reboot.h rather than stating them in full
again.

26 years agoDefine shift, mask and extraction macros for obtaining the slice number
msmith [Sun, 8 Mar 1998 15:04:20 +0000 (15:04 +0000)]
Define shift, mask and extraction macros for obtaining the slice number
from the bootdev parameter.  This is arguably a 386-ism, although disks
partitioned in this fashion are not uncommon.

26 years agoIf we are mounting the root filesystem, and we're accessing it through
msmith [Sun, 8 Mar 1998 14:50:04 +0000 (14:50 +0000)]
If we are mounting the root filesystem, and we're accessing it through
something that might refer to the compatability slice rather than the
correct slice entry, try all the possible slice entries first.

This is a compatability hack to deal with the case where the kernel has
correctly mounted the root filesystem out of its slice, but the user
has not updated their /etc/fstab file to reflect this.  A diagnostic
is emitted if the mount succeeds, indicating that the file should be
updated.

This is a prelude to fixing the kernel to behave as alluded to above.
Reviewed by: (discussed with) julian, phk

26 years agoRewrite script as a select loop rather than as a twin reader/write process.
peter [Sun, 8 Mar 1998 14:19:18 +0000 (14:19 +0000)]
Rewrite script as a select loop rather than as a twin reader/write process.
This allows simple logging of keys sent to a session (turned on with -k).
Also allow specifying the script file flush interval.

26 years agosvr4.h defines ASM_IDENTIFY_GCC, but neglects to define
peter [Sun, 8 Mar 1998 13:26:15 +0000 (13:26 +0000)]
svr4.h defines ASM_IDENTIFY_GCC, but neglects to define
ASM_IDENTIFY_LANGUAGE.  Use the osfrose.h method, because gdb assumes
'C' by default, so there's no need to further clutter the symbol table.

26 years agoCosmetic cleanup for a.out asm generation.. Don't put the end-of-file
peter [Sun, 8 Mar 1998 12:47:45 +0000 (12:47 +0000)]
Cosmetic cleanup for a.out asm generation..  Don't put the end-of-file
markers that the stabs-in-elf system uses.

26 years agoadd `CC' as a link to `c++' for compatablity with SGI.
obrien [Sun, 8 Mar 1998 12:27:04 +0000 (12:27 +0000)]
add `CC' as a link to `c++' for compatablity with SGI.

26 years agoPut the function stabs in traditional order on a.out, or gdb doesn't see
peter [Sun, 8 Mar 1998 11:54:04 +0000 (11:54 +0000)]
Put the function stabs in traditional order on a.out, or gdb doesn't see
function local variables.

26 years agoReviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman)
julian [Sun, 8 Mar 1998 09:59:44 +0000 (09:59 +0000)]
Reviewed by: dyson@freebsd.org (john Dyson), dg@root.com (david greenman)
Submitted by: Kirk McKusick (mcKusick@mckusick.com)
Obtained from:  WHistle development tree

26 years agoMF22: update table.
jkh [Sun, 8 Mar 1998 09:29:50 +0000 (09:29 +0000)]
MF22: update table.

26 years agoNote that ed0's IRQ is now 10 to help those upgrading from 2.x.
obrien [Sun, 8 Mar 1998 09:23:20 +0000 (09:23 +0000)]
Note that ed0's IRQ is now 10 to help those upgrading from 2.x.

26 years agoReplaced by ncal
helbig [Sun, 8 Mar 1998 09:19:48 +0000 (09:19 +0000)]
Replaced by ncal

26 years agoInitialize b_resid, and also print out better diagnostics on I/O
dyson [Sun, 8 Mar 1998 08:46:18 +0000 (08:46 +0000)]
Initialize b_resid, and also print out better diagnostics on I/O
errors.  This will allow for better tracking of user error reports.

26 years agoSubmitted by: kirk McKusick
julian [Sun, 8 Mar 1998 08:38:41 +0000 (08:38 +0000)]
Submitted by: kirk McKusick

Stub file for soft updates.

26 years agoSet LIB_PATH explicitly otherwise genscripts makes (incorrect) references
peter [Sun, 8 Mar 1998 08:08:39 +0000 (08:08 +0000)]
Set LIB_PATH explicitly otherwise genscripts makes (incorrect) references
all over the place.

26 years agoAdd a quirk entry for IO DATA CDC-TX86 6 CD changer.
jkh [Sun, 8 Mar 1998 06:52:45 +0000 (06:52 +0000)]
Add a quirk entry for IO DATA CDC-TX86 6 CD changer.
PR: 5942
Submitted by: Masanao Sasai <m_sasai@mxa.meshnet.ne.jp>

26 years agoSome cruft left over from my megacommit. A page rotation optimization
dyson [Sun, 8 Mar 1998 06:27:30 +0000 (06:27 +0000)]
Some cruft left over from my megacommit.  A page rotation optimization
was a good idea, but can cause instability.  That optimization is
now removed.

26 years agoSeveral minor fixes:
dyson [Sun, 8 Mar 1998 06:25:59 +0000 (06:25 +0000)]
Several minor fixes:
1) When freeing pages, it is a good idea to protect them off.
   (This is probably gratuitious, but good form.)
2) Allow collapsing pages in the backing object that are
   PQ_CACHE.  This will improve memory utilization.
3) Correct the collapse code so that pages that were on the
   cache queue are moved to the inactive queue.  This is
   done when pages are marked dirty (so that those pages
   will be properly paged out instead of freed), so that
   cached pages will not be paradoxically marked dirty.

26 years agoFree the first page also if it is not valid.
dyson [Sun, 8 Mar 1998 06:21:33 +0000 (06:21 +0000)]
Free the first page also if it is not valid.

26 years agoAdd hooks for John Birrell's kernel thread support library.
peter [Sun, 8 Mar 1998 05:33:24 +0000 (05:33 +0000)]
Add hooks for John Birrell's kernel thread support library.

Submitted by: John Birrell <jb@cimlogic.com.au>

26 years agoFirst round of changes to support generation of assembler for the old
peter [Sun, 8 Mar 1998 05:29:49 +0000 (05:29 +0000)]
First round of changes to support generation of assembler for the old
a.out gas and the binutils gas (elf or a.out) with a single compiler.

This uses other infrastructure not yet committed, in order to support
both a.out and elf it needs to be able to get to both a.out and elf
gas, ld, libs, crt* etc.  So for now, the support is pretty much dormant.

The new freebsd.h file is based on the old freebsd-elf.h file (which has a
long lineage, right back through linux and svr4 files).  The change is
pretty dramatic from a gcc internals standpoint as it overrides a lot of
definitions in order to generate different output based on target mode.
There is potential for screw-ups, so please be on the lookout - gcc's
configuration mechanism wasn't really meant for this kind of thing.
It's believed to compile world etc just fine under both a.out and elf, can
handle global constructors and destructors, handles the differences in
a.out and elf stabs, and what sections things like exceptions go in.

The initial idea came from i386/osfrose.h which is a dual rose/elf format
target.  These two are not as diverse as a.out and elf it would seem.

The cc front-end uses external configuration to determine default object
format (still being thrashed out, so read the source if you want to see
it so far), and has a '-aout' and '-elf' override command line switch.
There are some other internal switches that can be accessed, namely -maout,
-mno-aout, -munderscores and -mnounderscores.  The underscore and local
symbol prefixing rules are controllable seperately to the output format.
(ie: it's possible to generate a.out without the _ prefixes on symbols and
also to generate elf with the _ prefixes.  This isn't quite optimal, but
does seem to work pretty well, except the linkers don't always recognise
the local symbols without their normal names)

The default format is a.out (still), nobody should see any major changes.

With both elf and a.out tools and libraries installed:

[1:26pm]/tmp-223> cc -elf -o hello hello.c
peter@beast[1:27pm]/tmp-224> file hello
hello: ELF 32-bit LSB executable, Intel 80386, version 1 (FreeBSD), dynamically linked, not stripped
[1:27pm]/tmp-225> ./hello
hello world!

[1:27pm]/tmp-226> cc -aout -o hello hello.c
[1:27pm]/tmp-227> file hello
hello: FreeBSD/i386 compact demand paged dynamically linked executable not stripped
1:27pm]/tmp-228> ./hello
hello world!

Since my co-conspirators put a lot of effort into this too, I'll add them
so they can share the blame^H^H^H^H^Hglory. :-)

Reviewed by: sos, jdp

26 years agoSync with sys/i386/i386/machdep.c revision 1.292.
kato [Sun, 8 Mar 1998 03:55:35 +0000 (03:55 +0000)]
Sync with sys/i386/i386/machdep.c revision 1.292.

26 years agoAdd sched_yield() witch is the draft 10 equivalent of pthread_yield()
jb [Sun, 8 Mar 1998 02:37:27 +0000 (02:37 +0000)]
Add sched_yield() witch is the draft 10 equivalent of pthread_yield()
from draft 4. Move some of the schedule definitions to sched.h which
is a POSIX header.

26 years agoAdd a POSIX sched header. pthread_yield() in draft 4 becomes
jb [Sun, 8 Mar 1998 02:17:26 +0000 (02:17 +0000)]
Add a POSIX sched header. pthread_yield() in draft 4 becomes
sched_yield() in the final draft (10). This header contains the
prototype. Other things in here are "future".

26 years agoAdded forgotten new file for `-Lfoo -lbar' -> `foo/bar.a' change.
bde [Sun, 8 Mar 1998 00:19:19 +0000 (00:19 +0000)]
Added forgotten new file for `-Lfoo -lbar' -> `foo/bar.a' change.

26 years agoSupport a couple of new "-s" stats.
dyson [Sat, 7 Mar 1998 23:40:23 +0000 (23:40 +0000)]
Support a couple of new "-s" stats.

26 years agoRevert previous commit. Despite the names, these files contain
jraynard [Sat, 7 Mar 1998 22:34:23 +0000 (22:34 +0000)]
Revert previous commit.  Despite the names, these files contain
troff input, not C source.

26 years agoOops, merge back 32bit fat description from -stable.
ache [Sat, 7 Mar 1998 22:00:31 +0000 (22:00 +0000)]
Oops, merge back 32bit fat description from -stable.
It was added to -stable but not to -current, strange.

26 years agoPrevent passwd locking database forever waiting for user input
ache [Sat, 7 Mar 1998 21:42:07 +0000 (21:42 +0000)]
Prevent passwd locking database forever waiting for user input
Submitted by: Antti Kaipila <anttik@iki.fi>

26 years agoThis mega-commit is meant to fix numerous interrelated problems. There
dyson [Sat, 7 Mar 1998 21:37:31 +0000 (21:37 +0000)]
This mega-commit is meant to fix numerous interrelated problems.  There
has been some bitrot and incorrect assumptions in the vfs_bio code.  These
problems have manifest themselves worse on NFS type filesystems, but can
still affect local filesystems under certain circumstances.  Most of
the problems have involved mmap consistancy, and as a side-effect broke
the vfs.ioopt code.  This code might have been committed seperately, but
almost everything is interrelated.

1) Allow (pmap_object_init_pt) prefaulting of buffer-busy pages that
are fully valid.
2) Rather than deactivating erroneously read initial (header) pages in
kern_exec, we now free them.
3) Fix the rundown of non-VMIO buffers that are in an inconsistent
(missing vp) state.
4) Fix the disassociation of pages from buffers in brelse.  The previous
code had rotted and was faulty in a couple of important circumstances.
5) Remove a gratuitious buffer wakeup in vfs_vmio_release.
6) Remove a crufty and currently unused cluster mechanism for VBLK
files in vfs_bio_awrite.  When the code is functional, I'll add back
a cleaner version.
7) The page busy count wakeups assocated with the buffer cache usage were
incorrectly cleaned up in a previous commit by me.  Revert to the
original, correct version, but with a cleaner implementation.
8) The cluster read code now tries to keep data associated with buffers
more aggressively (without breaking the heuristics) when it is presumed
that the read data (buffers) will be soon needed.
9) Change to filesystem lockmgr locks so that they use LK_NOPAUSE.  The
delay loop waiting is not useful for filesystem locks, due to the
length of the time intervals.
10) Correct and clean-up spec_getpages.
11) Implement a fully functional nfs_getpages, nfs_putpages.
12) Fix nfs_write so that modifications are coherent with the NFS data on
the server disk (at least as well as NFS seems to allow.)
13) Properly support MS_INVALIDATE on NFS.
14) Properly pass down MS_INVALIDATE to lower levels of the VM code from
vm_map_clean.
15) Better support the notion of pages being busy but valid, so that
fewer in-transit waits occur.  (use p->busy more for pageouts instead
of PG_BUSY.)  Since the page is fully valid, it is still usable for
reads.
16) It is possible (in error) for cached pages to be busy.  Make the
page allocation code handle that case correctly.  (It should probably
be a printf or panic, but I want the system to handle coding errors
robustly.  I'll probably add a printf.)
17) Correct the design and usage of vm_page_sleep.  It didn't handle
consistancy problems very well, so make the design a little less
lofty.  After vm_page_sleep, if it ever blocked, it is still important
to relookup the page (if the object generation count changed), and
verify it's status (always.)
18) In vm_pageout.c, vm_pageout_clean had rotted, so clean that up.
19) Push the page busy for writes and VM_PROT_READ into vm_pageout_flush.
20) Fix vm_pager_put_pages and it's descendents to support an int flag
instead of a boolean, so that we can pass down the invalidate bit.

26 years agoThis commit was generated by cvs2svn to compensate for changes in r34204,
jb [Sat, 7 Mar 1998 21:27:49 +0000 (21:27 +0000)]
This commit was generated by cvs2svn to compensate for changes in r34204,
which included commits to RCS files with non-trunk default branches.

26 years agoImport of the mpn asm source files for alpha from V2.0.2 which
jb [Sat, 7 Mar 1998 21:27:49 +0000 (21:27 +0000)]
Import of the mpn asm source files for alpha from V2.0.2 which
appears to match the other libgmp files in the repository.

26 years agoReenable redirection of clock interrupt to a higher priority vector.
tegge [Sat, 7 Mar 1998 20:48:16 +0000 (20:48 +0000)]
Reenable redirection of clock interrupt to a higher priority vector.
setidt() now knows about f00f_hack(), and the APs now use the same
interrupt descriptor table as the BSP.

26 years agoMake vm_fault much cleaner by removing the evil macro inlines, and
dyson [Sat, 7 Mar 1998 20:45:47 +0000 (20:45 +0000)]
Make vm_fault much cleaner by removing the evil macro inlines, and
put alot of it's context into a data structure.  This allows
significant shortening of its codepath, and will significantly
decrease it's cache footprint.

Also, add some stats to vmmeter.  Note that you'll have to
rebuild/recompile vmstat, systat, etc... Otherwise, you'll
get "very interesting" paging stats.

26 years agoEnable csu/i386-elf under the appropriate conditions.
jdp [Sat, 7 Mar 1998 20:30:18 +0000 (20:30 +0000)]
Enable csu/i386-elf under the appropriate conditions.

26 years agoThis commit was generated by cvs2svn to compensate for changes in r34198,
jdp [Sat, 7 Mar 1998 20:27:11 +0000 (20:27 +0000)]
This commit was generated by cvs2svn to compensate for changes in r34198,
which included commits to RCS files with non-trunk default branches.

26 years agoImport C startup files for ELF support.
jdp [Sat, 7 Mar 1998 20:27:11 +0000 (20:27 +0000)]
Import C startup files for ELF support.

Submitted by: John Polstra <jdp@polstra.com>

26 years agoImport C startup files for ELF support.
jdp [Sat, 7 Mar 1998 20:27:11 +0000 (20:27 +0000)]
Import C startup files for ELF support.

Submitted by: John Polstra <jdp@polstra.com>

26 years agoThe APs now reload the interrupt descriptor table pointer after
tegge [Sat, 7 Mar 1998 20:16:49 +0000 (20:16 +0000)]
The APs now reload the interrupt descriptor table pointer after
f00f_hack has run.

Use the global r_idt descriptor in f00f_hack when in SMP mode,
so the APs find the relocated interrupt descriptor table.

Submitted by: Partially from David A Adkins <adkin003@tc.umn.edu>

26 years agoAdd support for ELF.
jdp [Sat, 7 Mar 1998 19:57:05 +0000 (19:57 +0000)]
Add support for ELF.

Switch to ANSI-style function definitions.

26 years agoEnable rtld-elf when building an ELF system on the i386.
jdp [Sat, 7 Mar 1998 19:43:48 +0000 (19:43 +0000)]
Enable rtld-elf when building an ELF system on the i386.

26 years agoSome kern_lock code improvements. Add missing wakeup, and enable
dyson [Sat, 7 Mar 1998 19:25:34 +0000 (19:25 +0000)]
Some kern_lock code improvements.  Add missing wakeup, and enable
disabling some diagnostics when memory or speed is at a premium.

26 years agoImport the ELF dynamic linker. This is the ElfKit version with
jdp [Sat, 7 Mar 1998 19:24:35 +0000 (19:24 +0000)]
Import the ELF dynamic linker.  This is the ElfKit version with
quite a few enhancements and bug fixes.  There are still some known
deficiencies, but it should be adequate to get us started with ELF.

Submitted by: John Polstra <jdp@polstra.com>

26 years agoThis commit was generated by cvs2svn to compensate for changes in r34192,
jdp [Sat, 7 Mar 1998 19:24:35 +0000 (19:24 +0000)]
This commit was generated by cvs2svn to compensate for changes in r34192,
which included commits to RCS files with non-trunk default branches.

26 years agoImport the ELF dynamic linker. This is the ElfKit version with
jdp [Sat, 7 Mar 1998 19:24:35 +0000 (19:24 +0000)]
Import the ELF dynamic linker.  This is the ElfKit version with
quite a few enhancements and bug fixes.  There are still some known
deficiencies, but it should be adequate to get us started with ELF.

Submitted by: John Polstra <jdp@polstra.com>

26 years agoDelete cc(1) debug option, typo from last commit.
wosch [Sat, 7 Mar 1998 16:13:57 +0000 (16:13 +0000)]
Delete cc(1) debug option, typo from last commit.

Pointed out by: Søren Schmidt

26 years agoSync with following changes:
kato [Sat, 7 Mar 1998 15:48:29 +0000 (15:48 +0000)]
Sync with following changes:
>   Added files:
>     sys/i386/boot        Makefile.inc
>  Revision  Changes    Path
>  1.63      +3 -12     src/sys/i386/boot/biosboot/Makefile
>  1.8       +1 -17     src/sys/i386/boot/kzipboot/Makefile
>  1.16      +6 -15     src/sys/i386/boot/netboot/Makefile
>  1.10      +3 -14     src/sys/i386/boot/rawboot/Makefile

26 years agoSync with sys/i386/isa/clock.c revision 1.115.
kato [Sat, 7 Mar 1998 15:43:43 +0000 (15:43 +0000)]
Sync with sys/i386/isa/clock.c revision 1.115.

26 years agoSync with sys/i386/i386/machdep.c revision 1.291.
kato [Sat, 7 Mar 1998 15:42:54 +0000 (15:42 +0000)]
Sync with sys/i386/i386/machdep.c revision 1.291.

26 years agoSet the input and output buffer sizes and the input buffer watermarks
bde [Sat, 7 Mar 1998 15:36:29 +0000 (15:36 +0000)]
Set the input and output buffer sizes and the input buffer watermarks
dynamically depending on the line speed(s).  This should give the old
sizes and watermarks until drivers are changed.

Display the input watermarks in pstat and sicontrol.

26 years agoFixed missing simple_lock() in ffs_mountfs().
bde [Sat, 7 Mar 1998 14:59:44 +0000 (14:59 +0000)]
Fixed missing simple_lock() in ffs_mountfs().

26 years agoPass -nostdinc to mkdep for assembler and C files (it was already
bde [Sat, 7 Mar 1998 13:57:37 +0000 (13:57 +0000)]
Pass -nostdinc to mkdep for assembler and C files (it was already
passed for C++ files).

Don't pass -nostdlib to mkdep for C++ files.

Don't use temporary variables to obfuscate the main rule.

26 years agoFixed generation of extra dependencies:
bde [Sat, 7 Mar 1998 13:13:41 +0000 (13:13 +0000)]
Fixed generation of extra dependencies:
- the two `_EXTRADEPEND::' targets potentially clobbered each other for
  `make -jN'.  In practice, the output for the second target sometimes
  disappeared.
- bogus dependencies were generated for static libraries.

26 years agoMake this handle -j<large>, and avoid .depend pointing at /usr/include.
eivind [Sat, 7 Mar 1998 10:48:15 +0000 (10:48 +0000)]
Make this handle -j<large>, and avoid .depend pointing at /usr/include.

26 years agoUse `foo/bar.a' instead of `-Lfoo -lbar' for linking to static internal
bde [Sat, 7 Mar 1998 09:49:12 +0000 (09:49 +0000)]
Use `foo/bar.a' instead of `-Lfoo -lbar' for linking to static internal
libraries so that `ld -f' in can create correct dependencies for
yet-to-be-built libraries.

Get the default BINDIR correctly (by including ../Makefile.inc recursively.
Override the default it it is wrong.

Don't override defaults when the defaults are correct.

Don't give absolute paths to man page sources.

26 years agoUse `foo/bar.a' instead of `-Lfoo -lbar' for linking to static internal
bde [Sat, 7 Mar 1998 09:48:14 +0000 (09:48 +0000)]
Use `foo/bar.a' instead of `-Lfoo -lbar' for linking to static internal
libraries so that `ld -f' in can create correct dependencies for
yet-to-be-built libraries.

Get the default BINDIR correctly (by including ../Makefile.inc recursively.
Override the default it it is wrong.

Don't override defaults when the defaults are correct.

26 years agoUse `foo/bar.a' instead of `-Lfoo -lbar' for linking to static internal
bde [Sat, 7 Mar 1998 09:46:17 +0000 (09:46 +0000)]
Use `foo/bar.a' instead of `-Lfoo -lbar' for linking to static internal
libraries so that `ld -f' in can create correct dependencies for
yet-to-be-built libraries.

Get the default BINDIR correctly (by including ../Makefile.inc recursively.

26 years agoReverted part of previous change and improved formatting. Switching
bde [Sat, 7 Mar 1998 09:06:09 +0000 (09:06 +0000)]
Reverted part of previous change and improved formatting.  Switching
from `?=' to `=' for BINDIR was a mistake.  Sub-makefiles may need
to set BINDIR.

26 years agoDocumented moused flags.
jkh [Sat, 7 Mar 1998 09:04:24 +0000 (09:04 +0000)]
Documented moused flags.
Submitted by: Kazutaka YOKOTA <yokota@zodiac.mech.utsunomiya-u.ac.jp>

26 years agoAdd -I flag.
jkh [Sat, 7 Mar 1998 09:03:43 +0000 (09:03 +0000)]
Add -I flag.
Submitted by: Kazutaka YOKOTA <yokota@zodiac.mech.utsunomiya-u.ac.jp>

26 years agoNew moused flag.
jkh [Sat, 7 Mar 1998 09:02:08 +0000 (09:02 +0000)]
New moused flag.
Submitted by: Kazutaka YOKOTA <yokota@zodiac.mech.utsunomiya-u.ac.jp>

26 years agoAdd a mouse configuration menu.
jkh [Sat, 7 Mar 1998 08:59:32 +0000 (08:59 +0000)]
Add a mouse configuration menu.
Submitted by: Kazutaka YOKOTA <yokota@zodiac.mech.utsunomiya-u.ac.jp>

26 years agoUse `foo/bar.a' instead of `-Lfoo -lbar' for linking to static internal
bde [Sat, 7 Mar 1998 08:55:00 +0000 (08:55 +0000)]
Use `foo/bar.a' instead of `-Lfoo -lbar' for linking to static internal
libraries, so that `ld -f' in can create correct dependencies for
yet-to-be-built libraries.

26 years agoAdd primary fat-32
ache [Sat, 7 Mar 1998 08:45:46 +0000 (08:45 +0000)]
Add primary fat-32

26 years agoFix a #ifndef that became a #ifdef and changed the default
jb [Sat, 7 Mar 1998 05:43:59 +0000 (05:43 +0000)]
Fix a #ifndef that became a #ifdef and changed the default
behaviour of rpcgen.

Noticed by: bde

26 years agoFix resource allocation problems:
wpaul [Sat, 7 Mar 1998 05:06:10 +0000 (05:06 +0000)]
Fix resource allocation problems:

- Completely recoded the ypmatch cache code. The old code could leak
  memory: it would allow the cache to grow, but never
  shrink. The new code imposes the following limits:

o The cache is capped at a limit of 5 entries.
o Each entry expires after five seconds, at which point
  its slot is freed.
o If an insertion is to be done and all five slots
  are filled, the oldest entry is forcibly expired
  to release its slot.

  Also, the cache is implemented on a per-binding basis rather than
  having a global cache covering all bindings. This means that each
  bound domain has its own 5 slot cache.

- Changed clntudp_create() to clntudp_bufcreate() so that the
  xmit/recv message buffer sizes can be set explicitly. NIS transactions
  are rarely much larger than 1024 bytes since YPMAXRECORD is 1024.
  The defaults chosen by clntudb_create() are actually much larger
  than needed. I set the xmit buffer to a little over 1024 and the
  recv buffer to a little over 2048. This saves a few Kbytes for each
  NIS binding.

- Add my name to the copyright. I think I've made enough changes to
  this file to merit it. :)

Note: these changes should go into the 2.2.x branch, but I'm waiting
on feedback from a tester to see if the cache fixes solve the reported
memory leak problem.

26 years agoAdd ${CXX} to list of variables passed in the environment when
asami [Sat, 7 Mar 1998 04:40:19 +0000 (04:40 +0000)]
Add ${CXX} to list of variables passed in the environment when
invoking configure.

Submitted by: obrien

26 years agoDOH! Forgot a backslash.
jkh [Fri, 6 Mar 1998 23:03:49 +0000 (23:03 +0000)]
DOH!  Forgot a backslash.

26 years agoOpen temporary files with O_EXCL.
guido [Fri, 6 Mar 1998 19:00:26 +0000 (19:00 +0000)]
Open temporary files with O_EXCL.
Obtained from:bugtraq

26 years agoFixed existence test for secure sources (don't test for secure objects).
bde [Fri, 6 Mar 1998 16:05:18 +0000 (16:05 +0000)]
Fixed existence test for secure sources (don't test for secure objects).

Fixed some formatting.

26 years agoFixed `make -jN' for large N, as usual.
bde [Fri, 6 Mar 1998 15:02:45 +0000 (15:02 +0000)]
Fixed `make -jN' for large N, as usual.