]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoThe .Nm utility
Philippe Charnier [Sun, 14 Jul 2002 14:47:15 +0000 (14:47 +0000)]
The .Nm utility

22 years agoAdd "tty" entry.
Luigi Rizzo [Sun, 14 Jul 2002 14:18:16 +0000 (14:18 +0000)]
Add "tty" entry.

22 years agoAdd entry for the "screen" program.
Luigi Rizzo [Sun, 14 Jul 2002 14:17:28 +0000 (14:17 +0000)]
Add entry for the "screen" program.

22 years agoClear local screen when ^D is typed.
Luigi Rizzo [Sun, 14 Jul 2002 14:08:43 +0000 (14:08 +0000)]
Clear local screen when ^D is typed.
Use setproctitle to remove arguments from 'ps' output.

MFC after: 3 days

22 years agoMore Russian holidays
Andrey A. Chernov [Sun, 14 Jul 2002 13:29:27 +0000 (13:29 +0000)]
More Russian holidays

PR: 39578
Submitted by: Sergey A. Osokin <osa@FreeBSD.org.ru>

22 years agoMFS: make this file compilable with gcc 2.9x as well.
Maxim Sobolev [Sun, 14 Jul 2002 13:25:51 +0000 (13:25 +0000)]
MFS: make this file compilable with gcc 2.9x as well.

22 years agoFix pl_PL collating table
Andrey A. Chernov [Sun, 14 Jul 2002 13:17:12 +0000 (13:17 +0000)]
Fix pl_PL collating table

PR: 39408
Submitted by: Michal Pasternak <doc@pasternak.w.lub.pl>

22 years agoCorrect _PATH_RCP to be /bin/rcp, not /usr/bin/rcp.
Maxim Sobolev [Sun, 14 Jul 2002 13:04:15 +0000 (13:04 +0000)]
Correct _PATH_RCP to be /bin/rcp, not /usr/bin/rcp.

22 years agoMiscellaneous fixes:
Thomas Moestl [Sun, 14 Jul 2002 12:09:48 +0000 (12:09 +0000)]
Miscellaneous fixes:
- always reinitialize the rx descriptors, even if the mbuf is kept.
  This should fix the hangs on ifconfig that were observed
- on an rx overflow, reinitialize the descriptor so that the interface
  will not hang
- correct some bus_dmamap_sync() calls
- correct some debug messages
- some minor nits

22 years agoNot needed anymore.
Luigi Rizzo [Sun, 14 Jul 2002 12:09:33 +0000 (12:09 +0000)]
Not needed anymore.

22 years agoReplace a logverbose call with log
Luigi Rizzo [Sun, 14 Jul 2002 12:07:06 +0000 (12:07 +0000)]
Replace a logverbose call with log

22 years agoMFi386: sys/i386/i386/machdep.c revision 1.522.
KATO Takenori [Sun, 14 Jul 2002 09:33:26 +0000 (09:33 +0000)]
MFi386: sys/i386/i386/machdep.c revision 1.522.

22 years agoEnable building of picobsd using CURRENT sources again.
Luigi Rizzo [Sun, 14 Jul 2002 09:07:13 +0000 (09:07 +0000)]
Enable building of picobsd using CURRENT sources again.

Following a suggestion by Ruslan, the initial creation of the
includes and libraries (and build tools) is now done by
invoking "make buildworld" (with -DPICOBSD which eventually will
limit the amount of stuff built with a 2-line change in Makefile.inc1).
The correct environment is then used for subsequent builds.

Also remove write_mfs_in_kernel.c in favour of using dd

All the above is conditional on __FreeBSD_version, as the previous
method still worked for versions earlier than 500035, and I am
unsure on how the "new" method works for earlier versions.

Finally, note that the crunch.conf files need some work because
some libraries (e.g. gmd) have gone away from the base installation.

22 years agopart of a greater patch set..
Julian Elischer [Sun, 14 Jul 2002 08:29:15 +0000 (08:29 +0000)]
part of a greater patch set..
1/ don't need to set td_state to TDS_RUNNING in fork_return.
it's already set in choosethread().
2/ Set a child process state to "normal" as opposed to "new"
when we allow it to be put on the run queue.
Allows child to receive signals from the parent if the parent
runs first and tries to immediatly signal he child.

Submitted by:  (part 2) Thomas Moestl <tmoestl@gmx.net>

22 years agoRearrange previous commit that passed the vendor id to the kernel in a way
Warner Losh [Sun, 14 Jul 2002 06:47:52 +0000 (06:47 +0000)]
Rearrange previous commit that passed the vendor id to the kernel in a way
that's binary compatible for -stable.  While binary compatibility doesn't
matter much in -current, it is critical for -stable.  This change requires
pccardd/pccardc to be recompiled.

22 years agoThinking about it I came to the conclusion that the KSE states were incorrectly
Julian Elischer [Sun, 14 Jul 2002 03:43:33 +0000 (03:43 +0000)]
Thinking about it I came to the conclusion that the KSE states were incorrectly
formulated.  The correct states should be:
IDLE:  On the idle KSE list for that KSEG
RUNQ:  Linked onto the system run queue.
THREAD: Attached to a thread and slaved to whatever state the thread is in.

This means that most places where we were adjusting kse state can go away
as it is just moving around because the thread is..
The only places we need to adjust the KSE state is in transition to and from
the idle and run queues.

Reviewed by: jhb@freebsd.org

22 years agoAdd a default case to the getopt(3) switch, remove a blank line nearby.
Juli Mallett [Sun, 14 Jul 2002 02:34:37 +0000 (02:34 +0000)]
Add a default case to the getopt(3) switch, remove a blank line nearby.

22 years agoTypo: do do -> to do.
Giorgos Keramidas [Sun, 14 Jul 2002 02:25:21 +0000 (02:25 +0000)]
Typo: do do -> to do.

Reviewed by: imp

22 years agoFunctions declared as <type> <identifier>(<nil>) should be declared as
Juli Mallett [Sun, 14 Jul 2002 02:03:23 +0000 (02:03 +0000)]
Functions declared as <type> <identifier>(<nil>) should be declared as
<type> <identifier>(<void-type>) in ANSI C.

22 years agoRemove prototype for main().
Juli Mallett [Sun, 14 Jul 2002 02:00:45 +0000 (02:00 +0000)]
Remove prototype for main().

22 years agooops, state cannot be two different values at once..
Julian Elischer [Sun, 14 Jul 2002 01:36:48 +0000 (01:36 +0000)]
oops, state cannot be two different values at once..
use || instead of &&

22 years agoTry both upa-portid and portid properties when finding the module id of a
Jake Burkholder [Sun, 14 Jul 2002 00:08:58 +0000 (00:08 +0000)]
Try both upa-portid and portid properties when finding the module id of a
secondary cpu.  Its called portid on UltraSPARCIII machines.

22 years agoRemove debug code.
Jake Burkholder [Sun, 14 Jul 2002 00:01:16 +0000 (00:01 +0000)]
Remove debug code.

22 years ago o Lock page queue accesses by vm_page_unmanage().
Alan Cox [Sat, 13 Jul 2002 23:55:30 +0000 (23:55 +0000)]
 o Lock page queue accesses by vm_page_unmanage().
 o Assert that the page queues lock is held in vm_page_unmanage().

22 years agoTypo fixo: imbedded => embedded
Juli Mallett [Sat, 13 Jul 2002 22:47:40 +0000 (22:47 +0000)]
Typo fixo: imbedded => embedded

22 years agoQuick fix for high resolution kernel profiling on i386's. Use
Bruce Evans [Sat, 13 Jul 2002 22:28:34 +0000 (22:28 +0000)]
Quick fix for high resolution kernel profiling on i386's.  Use
-finstrument-functions instead of -mprofiler-epilogue.  The former
works essentially the same as the latter but has a higher overhead
(about 22 more bytes per function for passing unused args to the
profiling functions).

Removed all traces of the IDENT Makefile variable, which had been
reduced to just a place for holding profiling's contribution to CFLAGS
(the IDENT that gives the kernel identity was renamed to KERN_IDENT).

22 years agoTwo invlpg's slipped through that were not protected from I386_CPU
Peter Wemm [Sat, 13 Jul 2002 21:55:11 +0000 (21:55 +0000)]
Two invlpg's slipped through that were not protected from I386_CPU

Pointed out by: dillon

22 years agoinvlpg() does not work too well on i386 cpus. Add token i386 support
Peter Wemm [Sat, 13 Jul 2002 21:03:08 +0000 (21:03 +0000)]
invlpg() does not work too well on i386 cpus.  Add token i386 support
back in to the pmap_zero_page* stuff.

22 years agoDo global shootdowns when switching to/from 4MB pages. I believe we can
Peter Wemm [Sat, 13 Jul 2002 20:58:56 +0000 (20:58 +0000)]
Do global shootdowns when switching to/from 4MB pages.  I believe we can
do a shootdown on a 4MB "page" though, but this should be safer for now.

Noticed by:  tegge

22 years agoBandaid for SMP. Changing APTDpde without a global shootdown is not
Peter Wemm [Sat, 13 Jul 2002 20:56:18 +0000 (20:56 +0000)]
Bandaid for SMP.  Changing APTDpde without a global shootdown is not
safe yet.  We used to do a global shootdown here anyway so another day
or so shouldn't hurt.

22 years ago o Complete the locking of page queue accesses by vm_page_unwire().
Alan Cox [Sat, 13 Jul 2002 20:55:21 +0000 (20:55 +0000)]
 o Complete the locking of page queue accesses by vm_page_unwire().
 o Assert that the page queues lock is held in vm_page_unwire().
 o Make vm_page_lock_queues() and vm_page_unlock_queues() visible
   to kernel loadable modules.

22 years ago o Lock some page queue accesses, in particular, those by vm_page_unwire().
Alan Cox [Sat, 13 Jul 2002 20:13:34 +0000 (20:13 +0000)]
 o Lock some page queue accesses, in particular, those by vm_page_unwire().

22 years ago o Lock some page queue accesses, in particular, those by vm_page_unwire().
Alan Cox [Sat, 13 Jul 2002 19:48:54 +0000 (19:48 +0000)]
 o Lock some page queue accesses, in particular, those by vm_page_unwire().

22 years agoFix a few bugs in the ERRORS section.
Garrett Wollman [Sat, 13 Jul 2002 19:38:59 +0000 (19:38 +0000)]
Fix a few bugs in the ERRORS section.

22 years agoMoved the setting of all profiling-related variables except the key one
Bruce Evans [Sat, 13 Jul 2002 19:36:14 +0000 (19:36 +0000)]
Moved the setting of all profiling-related variables except the key one
(PROFLEVEL) to kern.pre.mk so that it is easier to manage.  Bumped config
version to match.

Moved the check for cputype being configured to a less bogus place in
mkmakefile.c.

22 years agoAdd statvfs(3) to the build.
Garrett Wollman [Sat, 13 Jul 2002 19:33:20 +0000 (19:33 +0000)]
Add statvfs(3) to the build.

Tested by: Steve Kargl

22 years agoWell, it's not quite strxfrm(3) but at least it's honest.
Garrett Wollman [Sat, 13 Jul 2002 19:29:44 +0000 (19:29 +0000)]
Well, it's not quite strxfrm(3) but at least it's honest.

22 years ago o Lock some page queue accesses, in particular, those by vm_page_unwire().
Alan Cox [Sat, 13 Jul 2002 19:24:04 +0000 (19:24 +0000)]
 o Lock some page queue accesses, in particular, those by vm_page_unwire().

22 years agos/links/links1/
Bruce A. Mah [Sat, 13 Jul 2002 17:48:42 +0000 (17:48 +0000)]
s/links/links1/

This should unbreak release builds with docs but without a full ports tree.

While I'm here, fix a minor sorting glitch.

22 years agoRemove incorrect comment about now corrected manpage.
Alfred Perlstein [Sat, 13 Jul 2002 17:11:17 +0000 (17:11 +0000)]
Remove incorrect comment about now corrected manpage.

22 years agoMove COMPAT_FREEBSD4 to arch-neutral sys/conf/NOTES.
Alfred Perlstein [Sat, 13 Jul 2002 16:43:53 +0000 (16:43 +0000)]
Move COMPAT_FREEBSD4 to arch-neutral sys/conf/NOTES.
Add COMPAT_FREEBSD4 to GENERIC for arches that existed in FreeBSD 4's time,
not just i386. (alpha and pc98)

Requested by: bde

22 years agoMFCs noted: RLIMIT_VMEM, pam_opieaccess(8).
Bruce A. Mah [Sat, 13 Jul 2002 16:17:54 +0000 (16:17 +0000)]
MFCs noted:  RLIMIT_VMEM, pam_opieaccess(8).

Fix two minor grammar nits while here.

22 years agoA bunch of minor fixes:
Luigi Rizzo [Sat, 13 Jul 2002 15:57:23 +0000 (15:57 +0000)]
A bunch of minor fixes:

* accept "icmptype" as an alias for "icmptypes";
* remove an extra whitespace after "log" rules;
* print correctly the "limit" masks;
* correct a typo in parsing dummynet arguments (this caused a coredump);
* do not allow specifying both "check-state" and "limit", they are
  (and have always been) mutually exclusive;
* remove an extra print of the rule before installing it;
* make stdout buffered -- otherwise, if you log its output with syslog,
  you will see one entry for each printf(). Rather unpleasant.

22 years agoNew release notes: SA-02:29, SA-02:30.
Bruce A. Mah [Sat, 13 Jul 2002 15:49:08 +0000 (15:49 +0000)]
New release notes:  SA-02:29, SA-02:30.

22 years agoTurbochannel support / machines have been removed from -current.
Wilko Bulte [Sat, 13 Jul 2002 11:49:17 +0000 (11:49 +0000)]
Turbochannel support / machines have been removed from -current.

22 years agoWhen usb is kld, ucom and uplcom were failed to load because of
Hajimu UMEMOTO [Sat, 13 Jul 2002 10:41:35 +0000 (10:41 +0000)]
When usb is kld, ucom and uplcom were failed to load because of
usbd_abort_pipe undefined.

22 years agoAccount for space used by environment variables in a similar way to
Tim J. Robbins [Sat, 13 Jul 2002 08:08:46 +0000 (08:08 +0000)]
Account for space used by environment variables in a similar way to
xargs(1) when handling -exec ... {} + constructions.

22 years agoTypo.
Mike Heffner [Sat, 13 Jul 2002 07:21:53 +0000 (07:21 +0000)]
Typo.

22 years agoLock accesses to the page queues.
Alan Cox [Sat, 13 Jul 2002 04:37:22 +0000 (04:37 +0000)]
Lock accesses to the page queues.

22 years agoAdd prototype for cred_free_thread() when DIAGNOSTIC is defined.
Jonathan Mini [Sat, 13 Jul 2002 04:37:10 +0000 (04:37 +0000)]
Add prototype for cred_free_thread() when DIAGNOSTIC is defined.

22 years agoAdd additional cred_free_thread() calls that I had missed the first time.
Jonathan Mini [Sat, 13 Jul 2002 04:36:50 +0000 (04:36 +0000)]
Add additional cred_free_thread() calls that I had missed the first time.

Pointed out by: jhb

22 years ago o Lock accesses to the page queues.
Alan Cox [Sat, 13 Jul 2002 04:09:45 +0000 (04:09 +0000)]
 o Lock accesses to the page queues.
 o Add a comment explaining why hoisting the page queue lock outside
   of a particular loop is not possible.

22 years ago- Change chroot_refuse_vdir_fds() to require that the passed in struct
John Baldwin [Sat, 13 Jul 2002 04:07:12 +0000 (04:07 +0000)]
- Change chroot_refuse_vdir_fds() to require that the passed in struct
  filedesc is already locked rather than having chroot() unlock the
  filedesc so chroot_refuse_vdir_fds() can immediately relock it.
- Reorder chroot() a bitso that we do the namei lookup before checking
  the process's struct filedesc.  This closes at least one potential race
  and allows us to only acquire the filedsec lock once in chroot().
- Push down Giant slightly into chroot().

22 years agoUse a fixed address for KERNBASE, so it doesn't change if the size of KVA
Jake Burkholder [Sat, 13 Jul 2002 03:29:10 +0000 (03:29 +0000)]
Use a fixed address for KERNBASE, so it doesn't change if the size of KVA
is increased.  Its confusing for all the kernel addresses to change, and
doesn't serve much purpose as far as conserving address space.

22 years agoIdentify UltraSPARC-III and UltraSPARC-III+ cpus.
Jake Burkholder [Sat, 13 Jul 2002 03:23:29 +0000 (03:23 +0000)]
Identify UltraSPARC-III and UltraSPARC-III+ cpus.

22 years agoWe don't need to clear oldcred here since newcred is not NULL yet.
John Baldwin [Sat, 13 Jul 2002 03:13:15 +0000 (03:13 +0000)]
We don't need to clear oldcred here since newcred is not NULL yet.

22 years agoLock accesses to the page queues by sendfile() and friends.
Alan Cox [Sat, 13 Jul 2002 03:10:55 +0000 (03:10 +0000)]
Lock accesses to the page queues by sendfile() and friends.

22 years ago o Assert GIANT_REQUIRED on system maps in _vm_map_lock(),
Alan Cox [Fri, 12 Jul 2002 23:20:06 +0000 (23:20 +0000)]
 o Assert GIANT_REQUIRED on system maps in _vm_map_lock(),
   _vm_map_lock_read(), and _vm_map_trylock().  Submitted by: tegge
 o Remove GIANT_REQUIRED from kmem_alloc_wait() and kmem_free_wakeup().
   (This clears the way for exec_map accesses to move outside of Giant.
   The exec_map is not a system map.)
 o Remove some premature MPSAFE comments.

Reviewed by: tegge

22 years agoAvoid dereferencing a null pointer in ro_rt.
Luigi Rizzo [Fri, 12 Jul 2002 22:08:47 +0000 (22:08 +0000)]
Avoid dereferencing a null pointer in ro_rt.

This was always broken in HEAD (the offending statement was introduced
in rev. 1.123 for HEAD, while RELENG_4 included this fix (in rev.
1.99.2.12 for RELENG_4) and I inadvertently deleted it in 1.99.2.30.

So I am also restoring these two lines in RELENG_4 now.
We might need another few things from 1.99.2.30.

22 years agoRe-enable the idle page-zeroing code. Remove all IPIs from the idle
Matthew Dillon [Fri, 12 Jul 2002 20:17:06 +0000 (20:17 +0000)]
Re-enable the idle page-zeroing code.  Remove all IPIs from the idle
page-zeroing code as well as from the general page-zeroing code and use a
lazy tlb page invalidation scheme based on a callback made at the end
of mi_switch.

A number of people came up with this idea at the same time so credit
belongs to Peter, John, and Jake as well.

Two-way SMP buildworld -j 5 tests (second run, after stabilization)
    2282.76 real  2515.17 user  704.22 sys before peter's IPI commit
    2266.69 real  2467.50 user  633.77 sys after peter's commit
    2232.80 real  2468.99 user  615.89 sys after this commit

Reviewed by: peter, jhb
Approved by: peter

22 years agoalso set the KSE state for the idle KSE/thread case.
Julian Elischer [Fri, 12 Jul 2002 20:16:46 +0000 (20:16 +0000)]
also set the KSE state for the idle KSE/thread case.

22 years agoFix a type: s/your are/you are/
Tom Rhodes [Fri, 12 Jul 2002 19:56:31 +0000 (19:56 +0000)]
Fix a type: s/your are/you are/

22 years agoSet the thread state of the newly chosen to run thread to TDS_RUNNING in
John Baldwin [Fri, 12 Jul 2002 18:34:22 +0000 (18:34 +0000)]
Set the thread state of the newly chosen to run thread to TDS_RUNNING in
choosethread() in MI C code instead of doing it in in assembly in all the
various cpu_switch() functions.  This fixes problems on ia64 and sparc64.

Reviewed by: julian, peter, benno
Tested on: i386, alpha, sparc64

22 years agoNew release notes: NEWCARD default, finger(1) -g (+ MFC), whereis(1)
Bruce A. Mah [Fri, 12 Jul 2002 18:10:40 +0000 (18:10 +0000)]
New release notes:  NEWCARD default, finger(1) -g (+ MFC), whereis(1)
update.

MFCs noted:  finger(1) .pubkey.

22 years agoFix C++ exception handling. GCC was using the older BSD DBX register numbers
David E. O'Brien [Fri, 12 Jul 2002 17:44:24 +0000 (17:44 +0000)]
Fix C++ exception handling.  GCC was using the older BSD DBX register numbers
rather than the ELF ABI/SVR4 ones in its DWARF unwinding.

Submitted by: Alexander Kabaev <ak03@gte.com>

22 years agoWhen I decided to use a stock dbxout.c rather than merge the rev 1.2 change
David E. O'Brien [Fri, 12 Jul 2002 17:35:35 +0000 (17:35 +0000)]
When I decided to use a stock dbxout.c rather than merge the rev 1.2 change
(put the function stabs in traditional order on a.out, or gdb doesn't see
function local variables), I failed to remove the related knobs here.

Effectively were overrode the ELF-wide definition in elfos.h w/o providing
new infrastructure.  This is what caused GDB to fail to debug applications
compiled and linked with -stabs.  This is because GCC was unconditionally
inserts .stabs instruction for functions after the function body.  GDB was
getting confused because what it thinks is function beginning address is
actually function ending address.

Submitted by: Alexander Kabaev <ak03@gte.com>

22 years agoLock accesses to the page queues.
Alan Cox [Fri, 12 Jul 2002 17:21:22 +0000 (17:21 +0000)]
Lock accesses to the page queues.

22 years agoFix ptrace(PT_READ_*, ...) for non-little-endian architectures where
Thomas Moestl [Fri, 12 Jul 2002 16:48:05 +0000 (16:48 +0000)]
Fix ptrace(PT_READ_*, ...) for non-little-endian architectures where
sizeof(register_t) != sizeof(int).

22 years agoWhen sending cache flushing IPIs, don't try to IPI the triggering CPU
Thomas Moestl [Fri, 12 Jul 2002 16:26:49 +0000 (16:26 +0000)]
When sending cache flushing IPIs, don't try to IPI the triggering CPU
itself; this causes undefined behaviour on UltraSPARCs. In particular,
the interrupt packet data words will not necessarily be delivered
correctly, which would result in a crash.
This bug also caused the cache-flushing work to be done twice on the
triggering CPU (when it did not cause crashes).

Reviewed by: jake

22 years agoUse fgetln(3) to read lines from configuration files (ftpusers, ftphosts.)
Yaroslav Tykhiy [Fri, 12 Jul 2002 15:51:15 +0000 (15:51 +0000)]
Use fgetln(3) to read lines from configuration files (ftpusers, ftphosts.)
Thus lines of any length can be handled, unlike before.

Don't assume that each line read from the files ends with a newline.

As a side effect in inithosts(), don't use automatic buffer at all,
utilize malloc(3) when getting local host name instead.

PR: misc/21494
Reviewed by: maxim, mikeh
MFC after: 1 month

22 years agoFix ${TMAKE} to use the correct .mk files.
Ruslan Ermilov [Fri, 12 Jul 2002 15:15:35 +0000 (15:15 +0000)]
Fix ${TMAKE} to use the correct .mk files.
Removed stray -DNO_WERROR's.

22 years agoMoved the `distribute' target from bsd.obj.mk to bsd.subdir.mk,
Ruslan Ermilov [Fri, 12 Jul 2002 15:09:35 +0000 (15:09 +0000)]
Moved the `distribute' target from bsd.obj.mk to bsd.subdir.mk,
to make it call `install' in the bsd.subdir.mk-driven makefiles
too.  (share/examples/Makefile,v 1.29 changed the bsd.prog.mk
to bsd.subdir.mk and many stuff was lost during "make release".
I then merged this change in rev. 1.28.2.2 to work around the
namespace pollution (FILES) in this makefile.)

There was an added complexity here.  Both the `distribute' and
`install' targets are recursive (they propagate to SUBDIRs).
So `distribute' first calls `install' in the ${.CURDIR}, then
calls `distribute' in each SUBDIR, etc.  The problem is that
`install' (being also recursive) causes the stuff from SUBDIR
to be installed twice, first time thru `install' in ${.CURDIR}
triggered by `distribute', second time by `distribute' run in
the SUBDIR.  This problem is not new, but it became apparent
only after I moved the `distribute' target from bsd.obj.mk to
bsd.subdir.mk.  My first attempt testing the fix failed due to
this, because the whole world was distributed twice, causing
all the imaginable mess (kerberos5 stuff was installed into both
"base" and "krb5" dists, there was /sbin/init.bak, etc.)
I say the problem is not new because bsd.prog.mk and bsd.lib.mk
makefiles with SUBDIR (even without this fix) had this problem
for years.  Try e.g. running ``make distribute DISTDIR=/foo''
from usr.bin/bzip2 or from lib/libcom_err (without the fix) and
watch the output.

So the solution was to make `install' behave non-recursive when
executed by `distribute'.  My first attempt in passing SUBDIR=
to the `install' in the `distribute' body failed because of the
way how src/Makefile and src/Makefile.inc1 communicate with each
other.  SUBDIR='s assignment precedence on the "make install
SUBDIR=" command line is lowered after src/Makefile wrapper calls
"make ... -f ${.CURDIR}/Makefile.inc1 install" because SUBDIR=
is moved into environment, and Makefile.inc1's assignments now
take higher precedence.  This may be fixed someday when we merge
Makefile with Makefile.inc1.  For now, this is implemented as a
NO_SUBDIR knob.

Spotted by: Dmitry Pryanishnikov <dmitry@atlantis.dp.ua>
Prodded by: des
MFC after: 3 days

22 years agoMake dirty, rotten hack really work. As of rev. 1.16, ${BINDIR}
Ruslan Ermilov [Fri, 12 Jul 2002 14:49:25 +0000 (14:49 +0000)]
Make dirty, rotten hack really work.  As of rev. 1.16, ${BINDIR}
is only defined after <bsd.prog.mk> is included, and .if make(1)
conditionals are evaluted on the first pass.

Spotted by: Michael Bretterklieber <mbretter@jawa.at>

22 years agoConnect device.hints.5 to the build.
Ruslan Ermilov [Fri, 12 Jul 2002 14:16:46 +0000 (14:16 +0000)]
Connect device.hints.5 to the build.

22 years agoSort MAN.
Ruslan Ermilov [Fri, 12 Jul 2002 14:08:40 +0000 (14:08 +0000)]
Sort MAN.

22 years agoFixed path to /boot/device.hints.
Ruslan Ermilov [Fri, 12 Jul 2002 13:44:37 +0000 (13:44 +0000)]
Fixed path to /boot/device.hints.

22 years agoBack out the previous change, since it looks like locking udbinfo provides
Don Lewis [Fri, 12 Jul 2002 09:55:48 +0000 (09:55 +0000)]
Back out the previous change, since it looks like locking udbinfo provides
sufficient protection.

22 years agoBe specific about which reason caused vm86_addpages to panic
Peter Wemm [Fri, 12 Jul 2002 08:15:42 +0000 (08:15 +0000)]
Be specific about which reason caused vm86_addpages to panic

22 years agoLock inp while we're accessing it.
Don Lewis [Fri, 12 Jul 2002 08:05:22 +0000 (08:05 +0000)]
Lock inp while we're accessing it.

22 years agoRevive backed out pmap related changes from Feb 2002. The highlights are:
Peter Wemm [Fri, 12 Jul 2002 07:56:11 +0000 (07:56 +0000)]
Revive backed out pmap related changes from Feb 2002.  The highlights are:
- It actually works this time, honest!
- Fine grained TLB shootdowns for SMP on i386.  IPI's are very expensive,
  so try and optimize things where possible.
- Introduce ranged shootdowns that can be done as a single IPI.
- PG_G support for i386
- Specific-cpu targeted shootdowns.  For example, there is no sense in
  globally purging the TLB cache for where we are stealing a page from
  the local unshared process on the local cpu.  Use pm_active to track
  this.
- Add some instrumentation for the tlb shootdown code.
- Rip out SMP code from <machine/cpufunc.h>
- Try and fix some very bogus PG_G and PG_PS interactions that were bad
  enough to cause vm86 bios calls to break.  vm86 depended on our existing
  bugs and this was the cause of the VESA panics last time.
- Fix the silly one-line error that caused the 'panic: bad pte' last time.
- Fix a couple of other silly one-line errors that should have caused more
  pain than they did.

Some more work is needed:
- pmap_{zero,copy}_page[_idle].  These can be done without IPI's if we
  have a hook in cpu_switch.
- The IPI handlers need some cleanup.  I have a bogus %ds load that can
  be avoided.
- APTD handling is rather bogus and appears to be a large source of
  global TLB IPI shootdowns for no really good reason.

I see speedups of between 1.5% and ~4% on buildworlds in a while 1 loop.
I expect to see a bigger difference when there is significant pageout
activity or the system otherwise has memory shortages.

I have backed out a few optimizations that I had been using over the last
few days in order to be a little more conservative.  I'll revisit these
again over the next few days as the dust settles.

New option:  DISABLE_PG_G - In case I missed something.

22 years agoComplain if more than one file argument is given to unlink(1) like we did
Tim J. Robbins [Fri, 12 Jul 2002 07:20:20 +0000 (07:20 +0000)]
Complain if more than one file argument is given to unlink(1) like we did
before I made unlink use getopt().

22 years agoregen for freebsd4_sendfile(2) compat.
Alfred Perlstein [Fri, 12 Jul 2002 06:52:44 +0000 (06:52 +0000)]
regen for freebsd4_sendfile(2) compat.

22 years agoCreate a bug-for-bug FreeBSD4 compatible version of sendfile and move the
Alfred Perlstein [Fri, 12 Jul 2002 06:51:57 +0000 (06:51 +0000)]
Create a bug-for-bug FreeBSD4 compatible version of sendfile and move the
fixed sendfile over.  This is needed to preserve binary compatibility from
4.x to 5.x.

22 years agoIntroduce syscall.master option 'COMPAT4' which allows one to wrap
Alfred Perlstein [Fri, 12 Jul 2002 06:38:34 +0000 (06:38 +0000)]
Introduce syscall.master option 'COMPAT4' which allows one to wrap
syscalls for FreeBSD 4 compatibility.
Add kernel option COMPAT_FREEBSD4 to enable these syscalls.

22 years agoUnexpand a couple of 8-space indents that I added in rev 1.285.
Peter Wemm [Fri, 12 Jul 2002 04:58:51 +0000 (04:58 +0000)]
Unexpand a couple of 8-space indents that I added in rev 1.285.

22 years agoAvoid a vm_page_lookup() - that uses a spinlock protected hash. We can
Peter Wemm [Fri, 12 Jul 2002 04:38:51 +0000 (04:38 +0000)]
Avoid a vm_page_lookup() - that uses a spinlock protected hash.  We can
just use the object's memq for our nefarious purposes.

22 years ago o Lock some (unfortunately, not yet all) accesses to the page queues.
Alan Cox [Fri, 12 Jul 2002 03:17:22 +0000 (03:17 +0000)]
 o Lock some (unfortunately, not yet all) accesses to the page queues.

22 years ago o Lock accesses to the page queues.
Alan Cox [Fri, 12 Jul 2002 02:55:55 +0000 (02:55 +0000)]
 o Lock accesses to the page queues.

22 years agoFix compilation with ENABLE_VFS_IOOPT turned on and ZERO_COPY_SOCKETS
Kenneth D. Merry [Fri, 12 Jul 2002 02:23:55 +0000 (02:23 +0000)]
Fix compilation with ENABLE_VFS_IOOPT turned on and ZERO_COPY_SOCKETS
turned off.

Clean up #ifdefs, and remove a bunch of unnecessary includes.

Reviewed by: bde
Tested by: netchild

22 years agoDon't duplicate the description of the sticky bit that is already
Giorgos Keramidas [Fri, 12 Jul 2002 01:57:44 +0000 (01:57 +0000)]
Don't duplicate the description of the sticky bit that is already
part of chmod(2) and sticky(8), but refer to those manpages instead.

PR: docs/35605
Submitted by: Gary W. Swearingen <swear@blarg.net>

22 years agoCall routine to free everything obtained when filling in 'struct printer'.
Garance A Drosehn [Fri, 12 Jul 2002 01:53:36 +0000 (01:53 +0000)]
Call routine to free everything obtained when filling in 'struct printer'.

MFC after: 3 days

22 years agoFix set_qstate() so it correctly checks for any error from chmod().
Garance A Drosehn [Fri, 12 Jul 2002 01:37:06 +0000 (01:37 +0000)]
Fix set_qstate() so it correctly checks for any error from chmod().
Note that set_qstate() is only called from several 'lpc' commands.

MFC after: 3 days

22 years agoAdd a SQS_QCHANGED option to set_qstate(). This will soon be used by 'lpc'.
Garance A Drosehn [Fri, 12 Jul 2002 01:31:46 +0000 (01:31 +0000)]
Add a SQS_QCHANGED option to set_qstate().  This will soon be used by 'lpc'.

Reviewed by: freebsd-print@bostonradio.org
MFC after: 3 days

22 years agoVarious typo fixes.
Giorgos Keramidas [Fri, 12 Jul 2002 01:30:18 +0000 (01:30 +0000)]
Various typo fixes.

PR: docs/39395
Submitted by: Rich Neswold <rneswold@ameritech.net>

22 years agoAdd two variables to struct jobqueue, and change the way that getq()
Garance A Drosehn [Fri, 12 Jul 2002 01:22:57 +0000 (01:22 +0000)]
Add two variables to struct jobqueue, and change the way that getq()
calculates how much space to get for that struct, so it will get the
right amount when new variables are added.

MFC after: 3 days

22 years agoYep, revision 1.3 was 1/2assed.
David E. O'Brien [Fri, 12 Jul 2002 00:49:52 +0000 (00:49 +0000)]
Yep, revision 1.3 was 1/2assed.
This time use the right attribute for null_format_ok.

Submitted by: bde

22 years agoMove prototypes for ctl_readcf and ctl_freeinf from ctlinfo.c to ctlinfo.h,
Garance A Drosehn [Fri, 12 Jul 2002 00:33:07 +0000 (00:33 +0000)]
Move prototypes for ctl_readcf and ctl_freeinf from ctlinfo.c to ctlinfo.h,
so the routines can be called by an upcoming change for 'lpc topq/bottomq'.

MFC after: 3 days

22 years agoDefer calling SYSCTL_OUT() until after the locks have been released.
Don Lewis [Thu, 11 Jul 2002 23:18:43 +0000 (23:18 +0000)]
Defer calling SYSCTL_OUT() until after the locks have been released.

22 years agoReduce the nesting level of a code block that doesn't need to be in
Don Lewis [Thu, 11 Jul 2002 23:13:31 +0000 (23:13 +0000)]
Reduce the nesting level of a code block that doesn't need to be in
an else clause.

22 years agoA simple implementation of statvfs(3) (one step above the trivial one).
Garrett Wollman [Thu, 11 Jul 2002 22:54:11 +0000 (22:54 +0000)]
A simple implementation of statvfs(3) (one step above the trivial one).
Not yet connected to the build (awaiting documentation).