]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoRemove remaining traces of d_boot[01]
Poul-Henning Kamp [Thu, 4 Apr 2002 20:22:15 +0000 (20:22 +0000)]
Remove remaining traces of d_boot[01]

22 years agoMinor reorg.
David E. O'Brien [Thu, 4 Apr 2002 19:36:33 +0000 (19:36 +0000)]
Minor reorg.

22 years agoMinor style tweak.
David E. O'Brien [Thu, 4 Apr 2002 19:26:13 +0000 (19:26 +0000)]
Minor style tweak.

22 years agoRemove __P
Andrey A. Chernov [Thu, 4 Apr 2002 19:18:58 +0000 (19:18 +0000)]
Remove __P

22 years agoUpgrade to something quite close, but not identical, to version 1.6 of
Dag-Erling Smørgrav [Thu, 4 Apr 2002 18:45:21 +0000 (18:45 +0000)]
Upgrade to something quite close, but not identical, to version 1.6 of
Andrew Korty's pam_ssh.  The most notable difference is that this uses
commas rather than colons to separate items in the "keyfiles" option.

Sponsored by: DARPA, NAI Labs

22 years agoDocument new categories.
Dag-Erling Smørgrav [Thu, 4 Apr 2002 18:31:36 +0000 (18:31 +0000)]
Document new categories.

Submitted by: Hiten Pandya <hiten@uk.FreeBSD.org>
MFC after: 3 days

22 years agoRemove some local cruft that snuck in yesterday.
David E. O'Brien [Thu, 4 Apr 2002 18:24:56 +0000 (18:24 +0000)]
Remove some local cruft that snuck in yesterday.

22 years agoMerge from vendor branch.
Garrett Wollman [Thu, 4 Apr 2002 18:16:19 +0000 (18:16 +0000)]
Merge from vendor branch.

22 years agoUpdated release note: ISC DHCP 3.0.1RC8.
Bruce A. Mah [Thu, 4 Apr 2002 18:11:39 +0000 (18:11 +0000)]
Updated release note:  ISC DHCP 3.0.1RC8.

22 years agoMention that options BOOTP requires options NFSCLIENT and options NFS_ROOT
Brian Somers [Thu, 4 Apr 2002 18:01:55 +0000 (18:01 +0000)]
Mention that options BOOTP requires options NFSCLIENT and options NFS_ROOT

22 years agoBack out the previous commit.
Brian Somers [Thu, 4 Apr 2002 17:58:21 +0000 (17:58 +0000)]
Back out the previous commit.

In the i386 case, options BOOTP requires options NFS_ROOT as well as
options NFSCLIENT.  With *both* the NFS options, a bootpc_init()
prototype is brought in by nfsclient/nfsdiskless.h.

In the ia64 case, it just doesn't work and my change just pushes it
further away from working.

Suggested to be wrong by: bde

22 years agoMoved signal handling and rescheduling from userret() to ast() so that
Bruce Evans [Thu, 4 Apr 2002 17:49:48 +0000 (17:49 +0000)]
Moved signal handling and rescheduling from userret() to ast() so that
they aren't in the usual path of execution for syscalls and traps.
The main complication for this is that we have to set flags to control
ast() everywhere that changes the signal mask.

Avoid locking in userret() in most of the remaining cases.

Submitted by: luoqi (first part only, long ago, reorganized by me)
Reminded by: dillon

22 years agoPlay catch up with the kernel.
Poul-Henning Kamp [Thu, 4 Apr 2002 16:40:11 +0000 (16:40 +0000)]
Play catch up with the kernel.

Sponsored by: DARPA & NAI Labs

22 years agoAdd pam_passwdqc to the build.
Dag-Erling Smørgrav [Thu, 4 Apr 2002 16:08:28 +0000 (16:08 +0000)]
Add pam_passwdqc to the build.

Sponsored by: DARPA, NAI Labs

22 years agoThis commit was generated by cvs2svn to compensate for changes in r93787,
Dag-Erling Smørgrav [Thu, 4 Apr 2002 15:50:47 +0000 (15:50 +0000)]
This commit was generated by cvs2svn to compensate for changes in r93787,
which included commits to RCS files with non-trunk default branches.

22 years agoVendor import of Solar Designer's pam_passwdqc module.
Dag-Erling Smørgrav [Thu, 4 Apr 2002 15:50:47 +0000 (15:50 +0000)]
Vendor import of Solar Designer's pam_passwdqc module.

22 years agoOptimized the check for unmasked pending signals in CURSIG() using a new
Bruce Evans [Thu, 4 Apr 2002 15:19:41 +0000 (15:19 +0000)]
Optimized the check for unmasked pending signals in CURSIG() using a new
inline function sigsetmasked() and a new macro SIGPENDING().  CURSIG()
will soon be moved out of the normal path of execution for syscalls and
traps.  Then its efficiency will be less important but the new interfaces
will be useful for checking for unmasked pending signals in more places.

Submitted by: luoqi (long ago, in a slightly different form)

Assert that sched_lock is not held in CURSIG().

22 years agoPre-declare bootpc_init() so that options BOOTP doesn't break the
Brian Somers [Thu, 4 Apr 2002 14:27:57 +0000 (14:27 +0000)]
Pre-declare bootpc_init() so that options BOOTP doesn't break the
build in ia64 and i386 due to -Werror.

22 years agoImplement collating sort
Andrey A. Chernov [Thu, 4 Apr 2002 14:10:02 +0000 (14:10 +0000)]
Implement collating sort

22 years agoCentralize EOF handling and improve access controls for bio scheduling.
Poul-Henning Kamp [Thu, 4 Apr 2002 09:58:20 +0000 (09:58 +0000)]
Centralize EOF handling and improve access controls for bio scheduling.

Sponsored by: DARPA & NAI Labs

22 years agoFixed some style bugs in axings. Whitespace before __P was not axed when
Bruce Evans [Thu, 4 Apr 2002 09:56:51 +0000 (09:56 +0000)]
Fixed some style bugs in axings.  Whitespace before __P was not axed when
__P was axed.  The ordering of several things was bogotified by axing
ifdefs.

22 years agoMove access and orphan member functions from class to geom.
Poul-Henning Kamp [Thu, 4 Apr 2002 09:54:13 +0000 (09:54 +0000)]
Move access and orphan member functions from class to geom.

Sponsored by: DARPA & NAI Labs

22 years agoFixed some English errors in previous commit.
Bruce Evans [Thu, 4 Apr 2002 09:45:11 +0000 (09:45 +0000)]
Fixed some English errors in previous commit.

Fixed some style bugs in the removal of __P(()).  Whitespace before
"__P((" was not removed.

22 years agos/classs/classes/ to fixup grammer after the previous global renaming.
Poul-Henning Kamp [Thu, 4 Apr 2002 09:41:47 +0000 (09:41 +0000)]
s/classs/classes/ to fixup grammer after the previous global renaming.

Sponsored by: DARPA & NAI Labs

22 years agoCTASSERT the size of struct dos_partition.
Poul-Henning Kamp [Thu, 4 Apr 2002 09:36:15 +0000 (09:36 +0000)]
CTASSERT the size of struct dos_partition.

22 years agoRegister major #4 for GEOM
Poul-Henning Kamp [Thu, 4 Apr 2002 09:35:08 +0000 (09:35 +0000)]
Register major #4 for GEOM

22 years agoGet tun P2P address from the local pool if RADIUS server returned
Tatsumi Hosokawa [Thu, 4 Apr 2002 08:43:00 +0000 (08:43 +0000)]
Get tun P2P address from the local pool if RADIUS server returned
255.255.255.254 as client ipaddr.

Reviewed-By: freebsd-net mailing list

22 years agoBack out last commit. (This file is installed under /etc/mail.)
Ruslan Ermilov [Thu, 4 Apr 2002 07:42:12 +0000 (07:42 +0000)]
Back out last commit.  (This file is installed under /etc/mail.)

Requested by: gshapiro

22 years agoComplete the fix in revision 1.13 -- don't even depend on termcap(5)
Ruslan Ermilov [Thu, 4 Apr 2002 07:30:46 +0000 (07:30 +0000)]
Complete the fix in revision 1.13 -- don't even depend on termcap(5)
database to be present.

PR: misc/31225

22 years agoUse a relative path to contrib/sendmail/cf.
Ruslan Ermilov [Thu, 4 Apr 2002 07:18:29 +0000 (07:18 +0000)]
Use a relative path to contrib/sendmail/cf.

22 years agoIntegrate airtools. This adds a number of parameters, which I've not
Warner Losh [Thu, 4 Apr 2002 07:04:29 +0000 (07:04 +0000)]
Integrate airtools.  This adds a number of parameters, which I've not
documented yet, the most interesting one is -l, which lists all the
access points available.

22 years ago o Kill the MD grow_stack(). Call the MI vm_map_growstack()
Alan Cox [Thu, 4 Apr 2002 06:59:18 +0000 (06:59 +0000)]
 o Kill the MD grow_stack().  Call the MI vm_map_growstack()
   in its place.
 o Eliminate the use of useracc() and grow_stack() from sendsig().

Reviewed by: peter

22 years agoUnbreak buildworld after include/Makefile,v 1.167 commit.
Ruslan Ermilov [Thu, 4 Apr 2002 06:49:46 +0000 (06:49 +0000)]
Unbreak buildworld after include/Makefile,v 1.167 commit.

22 years agoNew release notes: UCONSOLE removed, viapm(4), geom(4), ifconfig
Bruce A. Mah [Thu, 4 Apr 2002 06:43:34 +0000 (06:43 +0000)]
New release notes:  UCONSOLE removed, viapm(4), geom(4), ifconfig
link (+MFC), libusb->libusbhid (+MFC), usbhidctl(1) (+MFC).

MFCs noted:  texinfo 4.1.

22 years agoo Add architecture specific segment types.
Marcel Moolenaar [Thu, 4 Apr 2002 06:42:49 +0000 (06:42 +0000)]
o  Add architecture specific segment types.
o  Add architecture specific segment attributes.

22 years agoo Reduce the gratuitous differences with NetBSD.
Warner Losh [Thu, 4 Apr 2002 06:28:16 +0000 (06:28 +0000)]
o Reduce the gratuitous differences with NetBSD.
o OpenBSD's wiconfig tells me that a value of '2' is for sony wireless
  cards, 1 is for lucent (which we already knew) and '5' is for embedded
  lucent cards.
o Move some RID definitions to if_wavelan_ieee.h and use NetBSD names
  more often.

# more work is still needed in this area.

22 years agoUse a relative path to libstand.. /usr/src/lib/libstand may not exist
Peter Wemm [Thu, 4 Apr 2002 06:04:38 +0000 (06:04 +0000)]
Use a relative path to libstand.. /usr/src/lib/libstand may not exist
or may have the wrong header files.

22 years agoReplace (deprecated ?) FREE() macro with direct calls to free()
Luigi Rizzo [Thu, 4 Apr 2002 06:03:17 +0000 (06:03 +0000)]
Replace (deprecated ?) FREE() macro with direct calls to free()

22 years agoTry to sort out the correct way to generate async link state change
Bill Paul [Thu, 4 Apr 2002 06:01:31 +0000 (06:01 +0000)]
Try to sort out the correct way to generate async link state change
interrupts. This is a bit harder than it needs to be because there's
more than one way to generate link attentions, at least one of which
does not work on the BCM5700, but does on the 5701.

For the 5701, we can safely use the 'link changed' bit in the status
block, and we enable link change attentions in the mac event register.
For the 5700, we have to use MII interrupts, which require checking
the MAC status register rather than the status block. This requires
doing an extra register access on each interrupt which I'd prefer to
avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T
shows that we do in fact detect the link going up and down properly
on cable insertions/disconnections.

Also, avoid twiddling the autopoll enable bit in the MI mode register
when doing a PHY read. I think this coupled with the other changes
will stop the interrupt storms Paul Saab has been harassing me about.
Manually setting the link to 100baseTX full duplex seems to work ok
for me. (I'm typing over the 3c996B-T right now.)

Lastly, teach the driver how to recognize a 3c996B-SX by checking
the hardware config word in the EEPROM in order to detect the media.
We attach 5701 fiber cards correctly now, but I haven't verified that
they send/receive packets yet since I don't have a second fiber
interface at home. (I know that fiber 5700 cards work, so I'm
keeping my fingers crossed.)

22 years agoFix incorrect m_free - m_freem() usage.
Luigi Rizzo [Thu, 4 Apr 2002 06:00:21 +0000 (06:00 +0000)]
Fix incorrect m_free - m_freem() usage.

22 years agoFix a couple of incorrect m_free() vs. m_freem() usages and related issues.
Luigi Rizzo [Thu, 4 Apr 2002 05:42:09 +0000 (05:42 +0000)]
Fix a couple of incorrect m_free() vs. m_freem() usages and related issues.

Reviewed-by: brooks
22 years agoo Move some function prototypes from <string.h> to the newly rewritten
Mike Barcroft [Thu, 4 Apr 2002 05:41:57 +0000 (05:41 +0000)]
o Move some function prototypes from <string.h> to the newly rewritten
  <strings.h>, based on POSIX.1-2001's requirements.
o Add 'restrict' qualifier (spelled '__restrict') to functions in
  <string.h>, as per C99 and POSIX.1-2001.
o Properly expose new POSIX.1-2001 functions in <string.h>.

22 years agoAdd the driver for "myson" ethernet cards.
Julian Elischer [Thu, 4 Apr 2002 05:22:13 +0000 (05:22 +0000)]
Add the driver for "myson" ethernet cards.
This driver was written by Myson and is made available
by their courtesy.
The 5.x version is not fully tested (I will be testing) but
the 4.x version has been tested by many.
I will commit it soon.
Myson have their own chip design based on the DEC 214xx family but
with several differences. Myson sells this chip to several
EOMs in teh Chinese area so there may be many noname brand cards
that respond to this driver. Myson will be supplying a list
of some of these.

22 years ago o aio_process needn't fhold()/fdrop() the fp now that _aio_aqueue() and
Alan Cox [Thu, 4 Apr 2002 02:13:20 +0000 (02:13 +0000)]
 o aio_process needn't fhold()/fdrop() the fp now that _aio_aqueue() and
   aio_free_entry() do this.
 o Remove two unnecessary/unused variables from aio_process() and one field
   from aiocblist.

22 years agoMake the sed line a little bit more clear (it will get messier later).
David E. O'Brien [Thu, 4 Apr 2002 01:25:26 +0000 (01:25 +0000)]
Make the sed line a little bit more clear (it will get messier later).

22 years agoAvoid a lock order reversal by dropping the eventhandler_mutex earlier.
Alfred Perlstein [Thu, 4 Apr 2002 00:52:03 +0000 (00:52 +0000)]
Avoid a lock order reversal by dropping the eventhandler_mutex earlier.
We get enough protection from the lock on the individual lists that we
aquire later.

Noticed/Tested by: Steven G. Kargl <kargl@troutmask.apl.washington.edu>
Submitted by: Jonathan Mini <mini@haikugeek.com>

22 years agoSet NOSHARED conditionally.
David E. O'Brien [Thu, 4 Apr 2002 00:50:14 +0000 (00:50 +0000)]
Set NOSHARED conditionally.

22 years agoClean up the YACCing. I don't know why we cannot leave the .y's as .y's.
David E. O'Brien [Thu, 4 Apr 2002 00:26:20 +0000 (00:26 +0000)]
Clean up the YACCing.  I don't know why we cannot leave the .y's as .y's.
So lets see if doing so causes anyone trouble.
Also use make(1)'s assistance in using the right file.  It knows the
dependency, so lets just ask it.

22 years agoRemove duplicate objc-parse.h. While we are at it, just spell it correctly
David E. O'Brien [Thu, 4 Apr 2002 00:14:58 +0000 (00:14 +0000)]
Remove duplicate objc-parse.h.  While we are at it, just spell it correctly
as c-parse.h since that is how the consumers spell it.

22 years agoRevision 1.59 also removed the pretend sparc support.
David E. O'Brien [Thu, 4 Apr 2002 00:12:42 +0000 (00:12 +0000)]
Revision 1.59 also removed the pretend sparc support.
We wont support Sparc using GCC 2.95.

22 years agoGet rid of GCC_ARCH, and just use plain TARGET_ARCH.
David E. O'Brien [Thu, 4 Apr 2002 00:11:00 +0000 (00:11 +0000)]
Get rid of GCC_ARCH, and just use plain TARGET_ARCH.
We got rid of the MIPS le/be stuff that needed this a long time ago.

22 years agoAdd more DWIM/autoadjustment and less evil style(9) banned exit(2) codes.
Poul-Henning Kamp [Wed, 3 Apr 2002 20:48:05 +0000 (20:48 +0000)]
Add more DWIM/autoadjustment and less evil style(9) banned exit(2) codes.
Add some missing statics.

Sponsored by: DARPA & NAI Labs.

22 years agoMove the FFS parameter MAXFRAG from <sys/param.h> to <ufs/ffs/fs.h>
Poul-Henning Kamp [Wed, 3 Apr 2002 20:39:27 +0000 (20:39 +0000)]
Move the FFS parameter MAXFRAG from <sys/param.h> to <ufs/ffs/fs.h>

Sponsored by: DARPA & NAI Labs.

22 years agoIt would appear, from preliminary indications, that No Wires Necessary
Warner Losh [Wed, 3 Apr 2002 20:39:07 +0000 (20:39 +0000)]
It would appear, from preliminary indications, that No Wires Necessary
cards aren't compatible with either Lucent or Intersil firmware.

22 years agoUnifdef -DCOMPAT
Poul-Henning Kamp [Wed, 3 Apr 2002 19:53:09 +0000 (19:53 +0000)]
Unifdef -DCOMPAT

22 years agoFour fixes from NetBSD:
Warner Losh [Wed, 3 Apr 2002 19:44:17 +0000 (19:44 +0000)]
Four fixes from NetBSD:
1) Properly detect the Symbol based cards (The 3Com Airconnect and their
   ilk) and only reset them *ONCE* ever.  This appears to make them work,
   but more testing is needed.  The tests that would wedge up my machine
   completely now appear to work, but I have not real access points
   handy.
2) Report both the Station firmware and the Primary firmware on Prism
   based cards.  On Lucent based cards, only report the station firmware
   since that's all it supports.  On symbol cards, report the symbol
   specific firmware name as its station firmware.
3) Better Prism 2.5 and 3 family names.  We really need to go table
   driven for this.
4) Workaround for bugs in Intersil's firmware is only needed for at most
   0.8.2 and earlier, since 0.8.3 and later appear to work.

Obtained from: NetBSD

22 years ago- Axe a stale comment. We haven't allowed the ucred pointer passed to
John Baldwin [Wed, 3 Apr 2002 18:35:25 +0000 (18:35 +0000)]
- Axe a stale comment.  We haven't allowed the ucred pointer passed to
  securelevel_*() to be NULL for a while now.
- Use KASSERT() instead of if (foo) panic(); to optimize the
  !INVARIANTS case.

Submitted by: Martin Faxer <gmh003532@brfmasthugget.se>

22 years agoFirst round at trying to split up NOTES into MI and MD portions.
John Baldwin [Wed, 3 Apr 2002 18:09:17 +0000 (18:09 +0000)]
First round at trying to split up NOTES into MI and MD portions.
Unfortunately, this level doesn't really provide enough granularity.  We
probably need several MI NOTES type files for things that are shared by
several architectures but not by all.  For example, the PCI options could
live in a NOTES.pci.

This also updates the Makefile for i386 to generate LINT.  The only changes
in the generated LINT are the order of various options.

Suggestions for improvement welcome.

22 years agoDon't clobber headers that we didn't create.
Ruslan Ermilov [Wed, 3 Apr 2002 17:29:20 +0000 (17:29 +0000)]
Don't clobber headers that we didn't create.

Noticed by: bde
Reviewed by: bde

22 years agoRegen.
Sheldon Hearn [Wed, 3 Apr 2002 17:03:09 +0000 (17:03 +0000)]
Regen.

This restores rev 1.12 (spammed by rev 1.13), as the Cenatek Rocket
Drive is now included in at least the Hart list.

22 years agoKill only usage of the undocumnted and unuse d_boot[01] fields of
Poul-Henning Kamp [Wed, 3 Apr 2002 15:46:56 +0000 (15:46 +0000)]
Kill only usage of the undocumnted and unuse d_boot[01] fields of
struct disklabel.

Sponsored by: DARPA & NAI Labs.

22 years agoInitial deorbit burn for the undocumented and unused d_boot[01]
Poul-Henning Kamp [Wed, 3 Apr 2002 15:44:53 +0000 (15:44 +0000)]
Initial deorbit burn for the undocumented and unused d_boot[01]
fields of struct disklabel.

Sponsored by: DARPA and NAI Labs.

22 years agoBump __FreeBSD_version to reflect the new tri-state nature of the
Sheldon Hearn [Wed, 3 Apr 2002 15:03:38 +0000 (15:03 +0000)]
Bump __FreeBSD_version to reflect the new tri-state nature of the
sendmail_enable rc.conf flag.

This will help MTA port maintainers.

22 years agoAdd two forgotten vfs_unbusy() calls, in vfs_mount() and vfs_nmount().
Maxime Henrion [Wed, 3 Apr 2002 12:19:03 +0000 (12:19 +0000)]
Add two forgotten vfs_unbusy() calls, in vfs_mount() and vfs_nmount().

Reviewed by: phk

22 years agoDike out a highly insecure UCONSOLE option.
Ruslan Ermilov [Wed, 3 Apr 2002 10:56:59 +0000 (10:56 +0000)]
Dike out a highly insecure UCONSOLE option.
TIOCCONS must be able to VOP_ACCESS() /dev/console to succeed.

Obtained from: OpenBSD

22 years agoMake the kernel dump header endianness invariant by always dumping
Marcel Moolenaar [Wed, 3 Apr 2002 07:24:12 +0000 (07:24 +0000)]
Make the kernel dump header endianness invariant by always dumping
in dump byte order (=network byte order). Swap blocksize and dumptime
to avoid extraneous padding on 64-bit architectures. Use CTASSERT
instead of runtime checks to make sure the header is 512 bytes large.
Various style(9) fixes.

Reviewed by: phk, bde, mike

22 years agoFix a long standing 32bit-ism. Don't assume that the size of a chunk of
Jake Burkholder [Wed, 3 Apr 2002 06:57:52 +0000 (06:57 +0000)]
Fix a long standing 32bit-ism.  Don't assume that the size of a chunk of
memory in phys_avail will fit in 'int', use vm_size_t.  This fixes booting
on sparc64 machines with more than 2 gigs of ram.

Thanks to Jan Chrillesen for providing me with access to a 4 gig machine.

22 years agoDeal with "#define _XOPEN_SOURCE" in header files (eg: XFree86 4.2.0)
Peter Wemm [Wed, 3 Apr 2002 05:56:55 +0000 (05:56 +0000)]
Deal with "#define _XOPEN_SOURCE"  in header files (eg: XFree86 4.2.0)
which caused a syntax error here:
#if _XOPEN_SOURCE >= 600
etc.  This is a hack, but I have seen things like it around elsewhere.

Tested by: dwcjr

22 years agoFix translation
Udo Erdelhoff [Wed, 3 Apr 2002 05:50:41 +0000 (05:50 +0000)]
Fix translation

22 years agoo GC dumplo
Marcel Moolenaar [Wed, 3 Apr 2002 05:16:09 +0000 (05:16 +0000)]
o  GC dumplo
o  Replace the string lit. "ia64" with MACHINE

22 years agoUse a twiddle to show that we're busy dumping. The initial code
Marcel Moolenaar [Wed, 3 Apr 2002 05:11:59 +0000 (05:11 +0000)]
Use a twiddle to show that we're busy dumping. The initial code
emitted the total number of pages it still had to dump prior to
dumping a block of up to 16 pages. For a 128MB region this would
result in 8M number of printf()s. Barf!

The problem in general is that memory typically has one really
big region and a number of "scattered" smaller regions. Some may
even be just a few pages. The twiddle works best for now, but
it doesn't really give a good progress indication for the large
regions. Those are the cases where you definitely want good PI
to avoid having the user turn into a twiddle :-)

22 years agoActually have ia64dump.c be included in the build. Can't be bad.
Marcel Moolenaar [Wed, 3 Apr 2002 04:42:14 +0000 (04:42 +0000)]
Actually have ia64dump.c be included in the build. Can't be bad.

22 years agoRemove some 1996 GCC 2.6.3 cruft for building a shared cc_int lib.
David E. O'Brien [Wed, 3 Apr 2002 03:18:15 +0000 (03:18 +0000)]
Remove some 1996 GCC 2.6.3 cruft for building a shared cc_int lib.

22 years agoFix to handle REC_D > 127 and fold case sorting of high letters
Andrey A. Chernov [Wed, 3 Apr 2002 01:39:26 +0000 (01:39 +0000)]
Fix to handle REC_D > 127 and fold case sorting of high letters
(linear sorting still assumed, no collating support yet).

22 years agobrelse() was improperly clearing B_DELWRI in the B_DELWRI|B_INVAL case
Matthew Dillon [Wed, 3 Apr 2002 00:17:36 +0000 (00:17 +0000)]
brelse() was improperly clearing B_DELWRI in the B_DELWRI|B_INVAL case
without removing the buffer from the vnode's dirty buffer list, which
can result in a panic in NFS.  Replaced the code with a call to bundirty()
which deals with it properly.

PR: kern/36108, kern/36174
Submitted by: various people
Special mention: to Danny Schales <dan@coes.LaTech.edu> for providing a core dump that helped me track this down.
MFC after: 1 day

22 years agoRedo stuff for sparc64- primarily fix bus dma implementation. The endian
Matt Jacob [Tue, 2 Apr 2002 23:36:14 +0000 (23:36 +0000)]
Redo stuff for sparc64- primarily fix bus dma implementation. The endian
stuff was right, but the busdma stuff was massively not right.

Didn't really test on ia64 or i386- don't have the former h/w and my
FreeBSD-current disk is unwell right now. Hope that this is okay.

MFC after: 1 week

22 years agoRevert to open hashing. It makes the code simpler, and works farily well
Dag-Erling Smørgrav [Tue, 2 Apr 2002 23:26:32 +0000 (23:26 +0000)]
Revert to open hashing.  It makes the code simpler, and works farily well
even when the number of records approaches the size of the hash table.
Besides, the previous implementation (using linear probing) was broken :)

Also, use the newly introduced MTX_SYSINIT.

22 years agoRevert to vendor version, what little was left of our local patches here
Dag-Erling Smørgrav [Tue, 2 Apr 2002 23:07:31 +0000 (23:07 +0000)]
Revert to vendor version, what little was left of our local patches here
was incorrect.

Pointed out by: Markus Friedl <markus@openbsd.org>

22 years ago- Move the MI mutexes sched_lock and Giant from being declared in the
John Baldwin [Tue, 2 Apr 2002 22:19:16 +0000 (22:19 +0000)]
- Move the MI mutexes sched_lock and Giant from being declared in the
  various machdep.c's to being declared in kern_mutex.c.
- Add a new function mutex_init() used to perform early initialization
  needed for mutexes such as setting up thread0's contested lock list
  and initializing MI mutexes.  Change the various MD startup routines
  to call this function instead of duplicating all the code themselves.

Tested on: alpha, i386

22 years agoChange the FreeBSD version addendum to "FreeBSD-20020402". This shortens
Dag-Erling Smørgrav [Tue, 2 Apr 2002 21:53:54 +0000 (21:53 +0000)]
Change the FreeBSD version addendum to "FreeBSD-20020402".  This shortens
the version string to 28 characters, which is below the 40-character limit
specified in the proposed SECSH standard.  Some servers, however (like the
one built into the Foundry BigIron line of switches) will hang when
confronted with a version string longer than 24 characters, so some users
may need to shorten it further.

Sponsored by: DARPA, NAI Labs

22 years agoAdd support for the PERC3/Di found in the Dell 1650
Scott Long [Tue, 2 Apr 2002 21:49:26 +0000 (21:49 +0000)]
Add support for the PERC3/Di found in the Dell 1650

Submitted by: mp

22 years agoMake the various ssh clients understand the VersionAddendum option.
Dag-Erling Smørgrav [Tue, 2 Apr 2002 21:48:51 +0000 (21:48 +0000)]
Make the various ssh clients understand the VersionAddendum option.

Submitted by: pb

22 years agofix comment typo, s/neccisary/necessary/g
Alfred Perlstein [Tue, 2 Apr 2002 21:25:12 +0000 (21:25 +0000)]
fix comment typo, s/neccisary/necessary/g

22 years agoOops, forgot to commit the definition of the mtx_name() macro.
Dag-Erling Smørgrav [Tue, 2 Apr 2002 21:12:17 +0000 (21:12 +0000)]
Oops, forgot to commit the definition of the mtx_name() macro.

22 years agoThis commit was generated by cvs2svn to compensate for changes in r93694,
Dag-Erling Smørgrav [Tue, 2 Apr 2002 20:57:22 +0000 (20:57 +0000)]
This commit was generated by cvs2svn to compensate for changes in r93694,
which included commits to RCS files with non-trunk default branches.

22 years agoFix incorrect allocation size (perforce change 8918)
Dag-Erling Smørgrav [Tue, 2 Apr 2002 20:57:22 +0000 (20:57 +0000)]
Fix incorrect allocation size (perforce change 8918)

PR: bin/36658
Submitted by: Takanori Saneto <sanewo@ba2.so-net.ne.jp>

22 years agoSpelling police.
John Baldwin [Tue, 2 Apr 2002 20:44:30 +0000 (20:44 +0000)]
Spelling police.

22 years agoSwitch to using XFree86 version 4. We do this thru installing the package,
David E. O'Brien [Tue, 2 Apr 2002 20:42:54 +0000 (20:42 +0000)]
Switch to using XFree86 version 4.  We do this thru installing the package,
so know we have proper PKG registration and dependency information.

This is a WIP for 5.0 DP #1, so it is still rough around the edges and
does not GC the old XFree86 3.3.6 handling stuff that should be GC'ed.
Sponsored by: FreeBSD Mall, Inc.

22 years agoEnforce an implicit lock order of sleepable locks before non-sleepable
John Baldwin [Tue, 2 Apr 2002 19:27:21 +0000 (19:27 +0000)]
Enforce an implicit lock order of sleepable locks before non-sleepable
locks.

22 years agoWhen reading a UFS-style label from a floppy, don't attempt to use a
Joerg Wunsch [Tue, 2 Apr 2002 18:29:43 +0000 (18:29 +0000)]
When reading a UFS-style label from a floppy, don't attempt to use a
"raw partition" of any kind since the floppy driver doesn't support
UFS-style partitions at all.

Reported by: "Crist J. Clark" <crist.clark@attbi.com>
Reviewed by: bde
MFC after: 3 days

22 years ago- Make this compile if INVARIANTS support is not enabled.
Andrew R. Reiter [Tue, 2 Apr 2002 18:18:56 +0000 (18:18 +0000)]
- Make this compile if INVARIANTS support is not enabled.

22 years agoFix crashes that would happen when more than one 4MB page was used to
Thomas Moestl [Tue, 2 Apr 2002 17:50:13 +0000 (17:50 +0000)]
Fix crashes that would happen when more than one 4MB page was used to
hold the kernel text, data and loader metadata by not using a fixed slot
to store the TSB page(s) into. Enter fake 8k page entries into the kernel
TSB that cover the 4M kernel page(s), sot that pmap_kenter() will work
without having to treat these pages as a special case.

Problem reported by: mjacob, obrien
Problem spotted and 4M page handling proposed by:       jake

22 years ago- Add a mutex to lock the global securelevel value.
Andrew R. Reiter [Tue, 2 Apr 2002 17:43:17 +0000 (17:43 +0000)]
- Add a mutex to lock the global securelevel value.
- Make use of MTX_SYSINIT() as the means to initialize our mutex lock.

22 years agoRemove the superfluous second argument from the IOTSBSLOT() macro.
Thomas Moestl [Tue, 2 Apr 2002 17:41:06 +0000 (17:41 +0000)]
Remove the superfluous second argument from the IOTSBSLOT() macro.

22 years agoLower UPA_MEMSTART to 0x1c000000000. This is required for some larger
Thomas Moestl [Tue, 2 Apr 2002 17:38:52 +0000 (17:38 +0000)]
Lower UPA_MEMSTART to 0x1c000000000. This is required for some larger
Enterprise machines.

22 years agoSet mp_maxid so that UMA works with SMP.
Thomas Moestl [Tue, 2 Apr 2002 17:37:06 +0000 (17:37 +0000)]
Set mp_maxid so that UMA works with SMP.

Submitted by: jake

22 years agoDo not try to set up the PCI bus B error interrupt on "sabre"s, since
Thomas Moestl [Tue, 2 Apr 2002 17:27:35 +0000 (17:27 +0000)]
Do not try to set up the PCI bus B error interrupt on "sabre"s, since
it is only available on "psycho"s. The same applies to the power
management interrupt, which is not enabled by default though.

22 years agoRemove a debugging panic that was triggered when a resource that was out
Thomas Moestl [Tue, 2 Apr 2002 17:23:45 +0000 (17:23 +0000)]
Remove a debugging panic that was triggered when a resource that was out
of range was tried to be allocated; just return failure instead.

22 years ago- Add notes about SX_SYSINIT and MTX_SYSINIT in order to document the
Andrew R. Reiter [Tue, 2 Apr 2002 17:21:00 +0000 (17:21 +0000)]
- Add notes about SX_SYSINIT and MTX_SYSINIT in order to document the
  recent additions of these macros.
- Add in a MLINK to create SX_SYSINIT.9 and MTX_SYSINIT.9.

22 years agoFix leakage of p_pgrp lock.
Seigo Tanimura [Tue, 2 Apr 2002 17:12:06 +0000 (17:12 +0000)]
Fix leakage of p_pgrp lock.