]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoPrepare for mdoc(7)NG.
ru [Tue, 16 Jan 2001 09:32:40 +0000 (09:32 +0000)]
Prepare for mdoc(7)NG.

23 years agoPrepare for mdoc(7)NG.
ru [Tue, 16 Jan 2001 09:24:36 +0000 (09:24 +0000)]
Prepare for mdoc(7)NG.

23 years agoPrepare for mdoc(7)NG.
ru [Tue, 16 Jan 2001 09:15:57 +0000 (09:15 +0000)]
Prepare for mdoc(7)NG.

23 years agoStop doing runtime checking on i386 cpus for cpu class. The cpu is
peter [Tue, 16 Jan 2001 09:10:34 +0000 (09:10 +0000)]
Stop doing runtime checking on i386 cpus for cpu class.  The cpu is
slow enough as it is, without having to constantly check that it really
is an i386 still.  It was possible to compile out the conditionals for
faster cpus by leaving out 'I386_CPU', but it was not possible to
unconditionally compile for the i386.  You got the runtime checking whether
you wanted it or not.  This makes I386_CPU mutually exclusive with the
other cpu types, and tidies things up a little in the process.

Reviewed by:  alfred, markm, phk, benno, jlemon, jhb, jake, grog, msmith,
              jasone, dcs, des (and a bunch more people who encouraged it)

23 years agoPrepare for mdoc(7)NG.
ru [Tue, 16 Jan 2001 09:08:22 +0000 (09:08 +0000)]
Prepare for mdoc(7)NG.

23 years agomdoc(7) police: use .Fa for function arguments, use .Vt for variable type.
ru [Tue, 16 Jan 2001 08:53:04 +0000 (08:53 +0000)]
mdoc(7) police: use .Fa for function arguments, use .Vt for variable type.

23 years ago- Fix a nasty bug where pkg_create doesn't gzip package if lengh of package
sobomax [Tue, 16 Jan 2001 08:27:28 +0000 (08:27 +0000)]
- Fix a nasty bug where pkg_create doesn't gzip package if lengh of package
  name is less than 5 and doesn't contain recognizeable suffix (one of .tar or
  .tgz), while gzip's it if lengh of the name greater than 4. For example
  `pkg_create [options] pkg1' will create pkg1.tar, while
  `pkg_create [options] pkg11' will create pkg11.tgz;
- use TRUE/FALSE as a values for boolean variables instead of explicit 1/0 and
  erroneous YES in one case.

MFC candidate.

23 years agoGuard against overflow of the calculated timeout value.
mjacob [Tue, 16 Jan 2001 07:15:36 +0000 (07:15 +0000)]
Guard against overflow of the calculated timeout value.

23 years agoArgh, disable the micro-ops again. I didn't test these adequately and
jhb [Tue, 16 Jan 2001 04:48:38 +0000 (04:48 +0000)]
Argh, disable the micro-ops again.  I didn't test these adequately and
managed to lock up one of my machines in world again.

Pointy-hat to: me

23 years ago- Use "+a" instead of "=&a" for several constraints. This should fix
jhb [Tue, 16 Jan 2001 03:45:54 +0000 (03:45 +0000)]
- Use "+a" instead of "=&a" for several constraints.  This should fix
  compiling errors where gcc would run out of registers.
- Add "cc" to the list of clobbers for micro-ops where we perform
  instructions that alter %eflags.
- Use xchgl instead of cmpxchgl to release a spin lock.  This could allow
  for more efficient register allocation as we no longer mandate that %eax
  be used.
- Reenable the optimized mutex micro-ops in the non-i386 case.

23 years agoFree the intrhand name when free'ing a intrhand.
jhb [Tue, 16 Jan 2001 02:17:51 +0000 (02:17 +0000)]
Free the intrhand name when free'ing a intrhand.

Submitted by: bde

23 years agoAdd some KASSERTs valid if WITNESS is defined to verify that the mbuf
bmilekic [Tue, 16 Jan 2001 01:53:13 +0000 (01:53 +0000)]
Add some KASSERTs valid if WITNESS is defined to verify that the mbuf
allocation routines are being called safely. Since we drop our relevant
mbuf mutex and acquire Giant before we call kmem_malloc(), we have
to make sure that this does not pave the way for a fatal lock order
reversal. Check that either Giant is already held (in which case it's safe
to grab it again and recurse on it) or, if Giant is not held, that no
other locks are held before we try to acquire Giant.

Similarily, add a KASSERT valid in the WITNESS case in m_reclaim() to
nail callers who end up in m_reclaim() and hold a lock.

Pointed out by: jhb

23 years agoImplement condition variables.
jasone [Tue, 16 Jan 2001 01:00:43 +0000 (01:00 +0000)]
Implement condition variables.

23 years agoUse the residual from a spacing operation to try and figure out where
mjacob [Tue, 16 Jan 2001 00:53:45 +0000 (00:53 +0000)]
Use the residual from a spacing operation to try and figure out where
we *really* are.

It should be noted that there is a degenerate case where soft tape
location will be lost (not causing a frozen state- but causing
the loss of reporting fileno/blockno)- that's where you backspace
over a filemark- you stop backspacing as soon as you cross the
filemark, but you have no idea what the record number now is because
you have no idea how many records you are into the file you just
backed into. Such is life.

While I'm at it, also pick up residuals from writing filemarks.

PR: 24222

23 years agoGet rid of those pesky "driver mistake" messages... destroy the create_d dev_t
mjacob [Tue, 16 Jan 2001 00:30:37 +0000 (00:30 +0000)]
Get rid of those pesky "driver mistake" messages... destroy the create_d dev_t
when you take down the device.

Reviewed by: audit@freebsd.org (Warner, Justin)

23 years ago- Fix atomic_load_* and atomic_store_* to generate functions for atomic.c
jhb [Tue, 16 Jan 2001 00:18:36 +0000 (00:18 +0000)]
- Fix atomic_load_* and atomic_store_* to generate functions for atomic.c
  that modules can call.
- Remove the old gcc <= 2.8 versions of the atomic ops.
- Resort the order of some things in the file so that there is only
  one #ifdef for KLD_MODULE, and so that all WANT_FUNCTIONS stuff is
  moved to the bottom of the file.
- Remove ATOMIC_ACQ_REL() and just use explicit macros instead.

23 years agoFix PR 24220 by using the periph private field0 for a CCB for not
mjacob [Mon, 15 Jan 2001 22:28:11 +0000 (22:28 +0000)]
Fix PR 24220 by using the periph private field0 for a CCB for not
only CCB type but also extra flags- one of which can be "position
updated".

In other changes: Add in a SA_QUIRK_NO_CPAGE quirk so that it's possible
to avoid using a (broken) device's implementation of he DEVICE COMPRESSION
page.

Also do a couple of printout cleanups.

As per some discussion on FreeBSD-scsi, skip doing tape flushing
if we're reading tape logical block location (MTIOCRDSPOS).

23 years agoFix typo.
grog [Mon, 15 Jan 2001 22:15:05 +0000 (22:15 +0000)]
Fix typo.

Submitted by: Jimmy Olgeni <olgeni@uli.it>
PR: 24233

Fix typos.

Submitted by: "Jeroen C. van Gelderen" <jeroen@vangelderen.org>

23 years agoClean up Makefile, and remove the last vestiges of NOBLOCKRANDOM.
markm [Mon, 15 Jan 2001 19:35:00 +0000 (19:35 +0000)]
Clean up Makefile, and remove the last vestiges of NOBLOCKRANDOM.

23 years agoAdd was_fabric_dev/fabric_dev tags to our local FC database structure
mjacob [Mon, 15 Jan 2001 18:40:37 +0000 (18:40 +0000)]
Add was_fabric_dev/fabric_dev tags to our local FC database structure
(so we can see rapidly whether something was a fabric device but is
now gone).

Add a tag which says what role this adapter should take. It can take
on the value of None, Target, Initiator or Both. None is useful for
warm failover purposes. Remove the ISP_CFG_NOINIT silliness since
a role of "None" does this.

Add a isp_lastmbxcmd tag to store the opcode for the last mailbox
command used.

23 years agoPut in offset definitions for FPM and FBM registers, plus just enough
mjacob [Mon, 15 Jan 2001 18:37:14 +0000 (18:37 +0000)]
Put in offset definitions for FPM and FBM registers, plus just enough
bits defined so we can reset them.

23 years agoSet default adapter role.
mjacob [Mon, 15 Jan 2001 18:36:39 +0000 (18:36 +0000)]
Set default adapter role.

23 years agoUse the isp_lastmbxcmd tag to report timed out mailbox commands.
mjacob [Mon, 15 Jan 2001 18:36:09 +0000 (18:36 +0000)]
Use the isp_lastmbxcmd tag to report timed out mailbox commands.

Arrrggghhhh! Very likely fix 22650 by remembering to, ahem, set
CAM_AUTOSNS_VALID when one has sense data.

23 years agoDo more cleanup of the usage of 0..125 for F-port topologies.
mjacob [Mon, 15 Jan 2001 18:34:49 +0000 (18:34 +0000)]
Do more cleanup of the usage of 0..125 for F-port topologies.

23 years agoWhen resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol
mjacob [Mon, 15 Jan 2001 18:33:08 +0000 (18:33 +0000)]
When resetting the Qlogic 2X00 units, reset the FPM (Fibre Protocol
Module) and FBM (Fibre Buffer Modules). Also remember to clear the
semaphore registers. Tell the RISC processor to not halt on FPM
parity errors.

Throw out the ISP_CFG_NOINIT silliness and instead go to the use of
adapter 'roles' to see whether one completes initialization or not
(mostly for Fibre Channel). The ultimate intent, btw, of all of this
is to have a warm standby adapter for failover reasons.  Because
we do roles now, setting of Target Capable Class 3 service parameters
in the ICB for the 2x00 cards reflects from role. Also, in isp_start,
if we're not supporting an initiator role, we bounce outgoing commands
with a Selection Timeout error. Also clean out the TOGGLE_TMODE
goop for FC- there is no toggling of target mode like there is
for parallel SCSI cards.

Do more cleanup with respect to using target ids 0..125 in F-port
topologies. Also keep track of things which *were* fabric devices
so that when you rescan the fabric you can notify the outer layers
when fabric devices go away.

Only force a LOGOUT for fabric devices if they're still logged in
(i.e., you cat their Port Database entry. Clean up the Get All Next
scanning.

Finally, use a new tag in the softc to store the opcode for the
last mailbox command used so we can report which opcode timed
out.

23 years agoThe ffs superblock includes a 128-byte region for use by temporary
iedowse [Mon, 15 Jan 2001 18:30:40 +0000 (18:30 +0000)]
The ffs superblock includes a 128-byte region for use by temporary
in-core pointers to summary information. An array in this region
(fs_csp) could overflow on filesystems with a very large number of
cylinder groups (~16000 on i386 with 8k blocks). When this happens,
other fields in the superblock get corrupted, and fsck refuses to
check the filesystem.

Solve this problem by replacing the fs_csp array in 'struct fs'
with a single pointer, and add padding to keep the length of the
128-byte region fixed. Update the kernel and userland utilities
to use just this single pointer.

With this change, the kernel no longer makes use of the superblock
fields 'fs_csshift' and 'fs_csmask'. Add a comment to newfs/mkfs.c
to indicate that these fields must be calculated for compatibility
with older kernels.

Reviewed by: mckusick

23 years agoman(7) -> mdoc(7).
ru [Mon, 15 Jan 2001 17:54:19 +0000 (17:54 +0000)]
man(7) -> mdoc(7).

23 years agoSimplify some mdoc(7) constructs.
ru [Mon, 15 Jan 2001 17:22:16 +0000 (17:22 +0000)]
Simplify some mdoc(7) constructs.

23 years agomdoc(7) police: fix the unbalanced Bk/Ek.
ru [Mon, 15 Jan 2001 16:58:14 +0000 (16:58 +0000)]
mdoc(7) police: fix the unbalanced Bk/Ek.

23 years agomdoc(7) police: fix the weird construct.
ru [Mon, 15 Jan 2001 16:49:28 +0000 (16:49 +0000)]
mdoc(7) police: fix the weird construct.

23 years agoit seems that a commit to i4b/drivers/i4b_ispppsubr.c on 2000-01-12 has
hm [Mon, 15 Jan 2001 14:58:46 +0000 (14:58 +0000)]
it seems that a commit to i4b/drivers/i4b_ispppsubr.c on 2000-01-12 has
broken the handling of uncompressed VJ packets. The attached diff should
hopefully fix that.
Submitted by: Thomas Moestl <tmoestl@gmx.net>
Reviewed by: Sergio de Souza Prallon <prallon@tmp.com.br>

23 years agoReenable support for FreeBSD 4.x and possibly the other supported BSD's
hm [Mon, 15 Jan 2001 14:54:43 +0000 (14:54 +0000)]
Reenable support for FreeBSD 4.x and possibly the other supported BSD's

23 years agoVersion bump, because of ABI incompatibility.
nsouch [Mon, 15 Jan 2001 13:19:05 +0000 (13:19 +0000)]
Version bump, because of ABI incompatibility.

Suggested by: Maxim Sobolev <sobomax@FreeBSD.org>

23 years agoAdd a couple of newlines in the output from route monitor
phk [Mon, 15 Jan 2001 12:28:48 +0000 (12:28 +0000)]
Add a couple of newlines in the output from route monitor

23 years agoAdd support for SMC91C100FD chip of MELCO LPC-TX.
toshi [Mon, 15 Jan 2001 12:01:14 +0000 (12:01 +0000)]
Add support for SMC91C100FD chip of MELCO LPC-TX.

sn1 at port 0x240-0x24f irq 10 slot 0 on pccard0
sn1:  SMC91C100FD UTP MAC address 00:a0:dc:22:26:8c

Submitted by: "KOMURO" <komujun@nifty.com>

23 years agoThe patch for the new calendar contained spaces instead of tabs.
dwmalone [Mon, 15 Jan 2001 10:18:24 +0000 (10:18 +0000)]
The patch for the new calendar contained spaces instead of tabs.
I only noticed when I went to check that it worked OK in RELENG_3.

23 years agoFix a bug arised from the ')' in the wrong place when calling printf()-like
sobomax [Mon, 15 Jan 2001 09:45:16 +0000 (09:45 +0000)]
Fix a bug arised from the ')' in the wrong place when calling printf()-like
varargs function, which lead to one of the arguments was left out. This resulted
in failure when inwoking mtree, warning message "mtree returned a non-zero
status - continuing" and probably is the reason for zillion mtree errors on
bento.

23 years agoBacktrack a bit.. Put the dangerously dedicated option under a
peter [Mon, 15 Jan 2001 08:15:19 +0000 (08:15 +0000)]
Backtrack a bit.. Put the dangerously dedicated option under a
non-advertised option (F = "FreeBSD only"), and leave the A key with
standard partitioning.  It seems people still want a runtime backdoo
to get to dangerously dedicated mode.

23 years agoMake NSWAPDEV reasonable so people do not mistakenly use unreasonable
dillon [Mon, 15 Jan 2001 06:45:01 +0000 (06:45 +0000)]
Make NSWAPDEV reasonable so people do not mistakenly use unreasonable
values when creating custom kernels from LINT.

Suggested-by: "Louis A. Mamakos" <louie@TransSys.COM>
23 years agoDon't invite trouble by waving the 'Dangerously Dedicated' (DD) option
peter [Mon, 15 Jan 2001 05:29:53 +0000 (05:29 +0000)]
Don't invite trouble by waving the 'Dangerously Dedicated' (DD) option
at people.  This has been sitting in my tree for a few months now.  I
have spoken with quite a few folks about this and the support for doing
this was pretty strong.  I dont remember names though, so I cannot share
the blame :-(.  Note that this does not *remove* DD mode, it just stops
waving it at new users.  You can still set it via config files etc, and
the bootblocks and kernel still support it.  You can still use disklabel
to make true DD disks.

23 years agoImplement an optimization for INTREN/INTRDIS that bde pointed out last
peter [Mon, 15 Jan 2001 04:18:58 +0000 (04:18 +0000)]
Implement an optimization for INTREN/INTRDIS that bde pointed out last
time I tinkered around here.  Since INTREN is called from the interrupt
critical path now, it should not be too expensive.  In this case, we
look at the bits being changed to decide which 8 bit IO port to write to
rather than unconditionally writing to both.  I could probably have gone
further and only done the write if the bits actually changed, but that
seemed overkill for the usual case in interrupt threads.

[an outb is rather expensive when it has to cross the ISA bus]

23 years agoReplace reference to replacing mkfs(8) with a paragraph actually
eivind [Mon, 15 Jan 2001 03:13:26 +0000 (03:13 +0000)]
Replace reference to replacing mkfs(8) with a paragraph actually
describing what newfs *does*.

23 years agoInstead of hard coding the major numbers for IDE and SCSI disks
joe [Sun, 14 Jan 2001 23:33:50 +0000 (23:33 +0000)]
Instead of hard coding the major numbers for IDE and SCSI disks
look in the device's cdevsw for the D_DISK flag.

23 years agoAdd a generic "queued function callin" mechanism
julian [Sun, 14 Jan 2001 23:25:54 +0000 (23:25 +0000)]
Add a generic "queued function callin" mechanism
Use it to implement (hopefully) SMP safe node/hook addition
and removal.
Fix some debug stuff.

23 years agoMake tcpwrappers use the magic in syslog.h for getting syslog facility
dwmalone [Sun, 14 Jan 2001 23:00:59 +0000 (23:00 +0000)]
Make tcpwrappers use the magic in syslog.h for getting syslog facility
and level names.

Add FreeBSD tag.

PR: 24218
Approved by: markm

23 years agoUse prober atomic operations when test&set'ing ->active.
sos [Sun, 14 Jan 2001 19:36:20 +0000 (19:36 +0000)]
Use prober atomic operations when test&set'ing ->active.

This turns some of my lockups under SMP into spontanious reboots...

23 years agoDocument the -N option in the usage message and the man page.
des [Sun, 14 Jan 2001 19:08:58 +0000 (19:08 +0000)]
Document the -N option in the usage message and the man page.

23 years agoAdd {ACD,AST,ATAPI,ATA}_DEBUG.
des [Sun, 14 Jan 2001 19:03:53 +0000 (19:03 +0000)]
Add {ACD,AST,ATAPI,ATA}_DEBUG.

Approved by: sos

23 years agoA bit of sanity-checking in bioqdisksort(): panic if we recurse.
phk [Sun, 14 Jan 2001 18:48:42 +0000 (18:48 +0000)]
A bit of sanity-checking in bioqdisksort(): panic if we recurse.

23 years agoThe size of kinfo_proc on an alpha is 904 (not 640).
mjacob [Sun, 14 Jan 2001 18:21:19 +0000 (18:21 +0000)]
The size of kinfo_proc on an alpha is 904 (not 640).

23 years agoThe Linuxulator needs sysvipc to work.
des [Sun, 14 Jan 2001 18:16:01 +0000 (18:16 +0000)]
The Linuxulator needs sysvipc to work.

23 years agoUse predictable internal names for the sysvipc modules, so we have a
des [Sun, 14 Jan 2001 18:04:30 +0000 (18:04 +0000)]
Use predictable internal names for the sysvipc modules, so we have a
chance of getting dependencies working.

23 years agoRemove NOBLOCKRANDOM as a compile-time option. Instead, provide
markm [Sun, 14 Jan 2001 17:50:15 +0000 (17:50 +0000)]
Remove NOBLOCKRANDOM as a compile-time option. Instead, provide
exactly the same functionality via a sysctl, making this feature
a run-time option.

The default is 1(ON), which means that /dev/random device will
NOT block at startup.

setting kern.random.sys.seeded to 0(OFF) will cause /dev/random
to block until the next reseed, at which stage the sysctl
will be changed back to 1(ON).

While I'm here, clean up the sysctls, and make them dynamic.
Reviewed by: des
Tested on Alpha by: obrien

23 years agofixed possibility of panic at key_acquire().
ume [Sun, 14 Jan 2001 17:25:08 +0000 (17:25 +0000)]
fixed possibility of panic at key_acquire().
key_acquire() does not require a secpolicy structure.

Obtained from: KAME

23 years agoAdd a missing argument for SYSCTL_ADD_OPAQUE.
markm [Sun, 14 Jan 2001 16:43:41 +0000 (16:43 +0000)]
Add a missing argument for SYSCTL_ADD_OPAQUE.

Fix the examples at the end which were horribly broken.

23 years agoAdd a -N option that makes sysctl(8) print out just the variable names.
des [Sun, 14 Jan 2001 16:40:06 +0000 (16:40 +0000)]
Add a -N option that makes sysctl(8) print out just the variable names.
Zsh users can add the following to their .zshrc for sysctl completion:

function listsysctls {
    case $1 in
    *.*) set -A reply $(sysctl -AN ${1%.*}) ;;
    *) set -A reply $(sysctl -AN) ;;
    esac
}
compctl -K listsysctls sysctl

While I'm here, brucify the getopt() switch.

23 years agoDescribe that the CDROM environment variable now affects which
joe [Sun, 14 Jan 2001 16:29:24 +0000 (16:29 +0000)]
Describe that the CDROM environment variable now affects which
device is used by default.

23 years agoSpell my name right in the copyright.
des [Sun, 14 Jan 2001 16:14:17 +0000 (16:14 +0000)]
Spell my name right in the copyright.

23 years agoSpecial case the error reporting when errno is ENOTDIR or ENOENT.
alfred [Sun, 14 Jan 2001 12:08:50 +0000 (12:08 +0000)]
Special case the error reporting when errno is ENOTDIR or ENOENT.

This makes "mkdir /nonexistant/foo" complain that /nonexistant
doesn't exist rather than /nonexistant/foo which doesn't make much
sense.

Submitted (in a different form) by: W.H.Scholten <whs@xs4all.nl>

23 years agoRemove references to lockinfo.plex.
grog [Sun, 14 Jan 2001 11:42:19 +0000 (11:42 +0000)]
Remove references to lockinfo.plex.

Reported by:  dougb

23 years agoArgh, remove a local customization that snuck in here.
jhb [Sun, 14 Jan 2001 10:19:42 +0000 (10:19 +0000)]
Argh, remove a local customization that snuck in here.

Noticed by: jasone

23 years agoRemove I386_CPU from GENERIC. Support for the 386 seriously pessimizes
jhb [Sun, 14 Jan 2001 10:11:10 +0000 (10:11 +0000)]
Remove I386_CPU from GENERIC.  Support for the 386 seriously pessimizes
performance on other x86 processors.  Custom kernels can still be built
that will run on the 386.

23 years agoRevert the previous revision now that atomic_store_rel_ptr() actually
jhb [Sun, 14 Jan 2001 09:56:35 +0000 (09:56 +0000)]
Revert the previous revision now that atomic_store_rel_ptr() actually
works.

23 years agoFix the atomic_load_acq() and atomic_store_rel() functions to properly
jhb [Sun, 14 Jan 2001 09:55:21 +0000 (09:55 +0000)]
Fix the atomic_load_acq() and atomic_store_rel() functions to properly
implement memory fences for the 486+.  The 386 still uses versions w/o
memory fences as all operations on the 386 are not program ordered.
The 386 versions are not MP safe.

23 years agoFix this to be more consistent with similar constructions elsewhere
dougb [Sun, 14 Jan 2001 09:29:57 +0000 (09:29 +0000)]
Fix this to be more consistent with similar constructions elsewhere

23 years agoadd new PC-Card entry, Melco/Buffalo LPC2-CLT
sanpei [Sun, 14 Jan 2001 08:40:20 +0000 (08:40 +0000)]
add new PC-Card entry, Melco/Buffalo LPC2-CLT

ed1 at port 0x300-0x31f irq 3 slot 0 on pccard0
ed1: address 00:40:26:xx:xx:xx, type NE2000 (16 bit)

Submitted by: [FreeBSD-users-jp 58182]
YANAGAWA Kazuhisa <kjana@os.xaxon.ne.jp>

23 years agoMove the process of storing entropy from /dev/random and reseeding with
dougb [Sun, 14 Jan 2001 07:18:31 +0000 (07:18 +0000)]
Move the process of storing entropy from /dev/random and reseeding with
it at boot time closer to the way we want it to be in the final version.

* Move the default directory to /var/db/entropy
* Run the entropy saving cron job every 11 minutes. This seems
  to be a better default, although still bikeshed material.
* Feed /dev/random some cheesy "entropy" from various commands
  and files before the disks are mounted. This gives /dev/random
  a better chance of running without blocking early.
* Move the reseeding with previously stored entropy to the point
  immediately after the disks are mounted.
* Make the harvesting script a little safer in regards to the
  possibility of accidentally overwriting something other
  than a regular file.

23 years agostruct rangelock: Remove the field 'plex' from the entry. Range locks
grog [Sun, 14 Jan 2001 06:34:57 +0000 (06:34 +0000)]
struct rangelock: Remove the field 'plex' from the entry.  Range locks
are accessed only via the plex, so there's never any confusion as to
the plex number.  This value was, as a result, unused.

23 years agoformat_config: If a subdisk loses its drive (due to a bug which has
grog [Sun, 14 Jan 2001 06:33:10 +0000 (06:33 +0000)]
format_config: If a subdisk loses its drive (due to a bug which has
    not yet been caught), don't save the config with a null drive
    name (which causes the drive to be renamed "plex" on the next
    start), put in the text "*invalid*" instead.

This is damage control, not a fix.

Experienced by: peter

Break some long format strings so that they fit in style(9)-sized
lines.

Remove some "outdentation".

23 years agoconfig_plex: Check that we have specified a plex organization.
grog [Sun, 14 Jan 2001 06:29:56 +0000 (06:29 +0000)]
config_plex: Check that we have specified a plex organization.

Tripped over by: "Jeroen C. van Gelderen" <jeroen@vangelderen.org>

23 years agoUse ANSI prototype and function definition for dkcksum so that this
imp [Sun, 14 Jan 2001 05:14:37 +0000 (05:14 +0000)]
Use ANSI prototype and function definition for dkcksum so that this
file can be included with C++ again.  This appears to have been broken
in 1.54 with the introduction of it as an inline function.

23 years agoFix a nasty bug whereby if the package file didn't have a version number
paul [Sun, 14 Jan 2001 02:05:02 +0000 (02:05 +0000)]
Fix a nasty bug whereby if the package file didn't have a version number
then all packages would be deinstalled!

The tightening up of version number checking also fixes a bug where
a package file such as gtk.tgz would have resulting in gtk-engines
being deinstalled.

23 years agoUse fstat to check if descriptor 0 is a socket.
brian [Sun, 14 Jan 2001 00:54:48 +0000 (00:54 +0000)]
Use fstat to check if descriptor 0 is a socket.

Suggested by: julian

23 years agoWork around the broken atomic_store_rel_ptr() on the i386 arch by just
jhb [Sun, 14 Jan 2001 00:16:17 +0000 (00:16 +0000)]
Work around the broken atomic_store_rel_ptr() on the i386 arch by just
using atomic_cmpset_rel_ptr() instead for _release_lock_quick().  When
atomic_store_rel_ptr() is functional and MP safe, then this can be
reverted.

23 years agoCorrect a typo in a product name.
dmlb [Sat, 13 Jan 2001 23:50:52 +0000 (23:50 +0000)]
Correct a typo in a product name.

Pointed Out By: ru

23 years ago- Use sched_lock to prevent the mutex name from changing out from under us
jhb [Sat, 13 Jan 2001 23:08:34 +0000 (23:08 +0000)]
- Use sched_lock to prevent the mutex name from changing out from under us
  while we are copying it to the kinfo_proc structure.
- Test against p_stat to see if we are blocked on a mutex.
- Terminate ki_mtxname with a null char rather than ki_wmesg.

23 years agoFix getsid() to use "=" instead of "==".
ben [Sat, 13 Jan 2001 22:49:59 +0000 (22:49 +0000)]
Fix getsid() to use "=" instead of "==".

Not objected to by: audit

23 years agoAdd 3 new dynamic sysctl's to control the sleep states switched to on a
jhb [Sat, 13 Jan 2001 21:28:57 +0000 (21:28 +0000)]
Add 3 new dynamic sysctl's to control the sleep states switched to on a
power button, sleep button, or lid close event.  The sysctl's use the
ACPI sleep state names S0, S1, S2, S3, S4, S4B, and S5.

Reviewed by: iwasaki

23 years agoThe prototype for cpu_switch() was changed in revision 1.111 of
chris [Sat, 13 Jan 2001 19:00:49 +0000 (19:00 +0000)]
The prototype for cpu_switch() was changed in revision 1.111 of
<sys/proc.h>

PR: 24311
Submitted by: Robert Drehmel <robert@gizmo.quizbot.org>

23 years agoo Clear up explanations of various entries (includes spelling, grammar,
chris [Sat, 13 Jan 2001 18:38:00 +0000 (18:38 +0000)]
o Clear up explanations of various entries (includes spelling, grammar,
  punctuation, and explanations that are just plain wrong)
o Add missing entries
o Remove entries for directories that do not exist

Submitted by: Rich Morin <rdm@cfcl.com> (for the most part)

23 years agofix typo; file -> files
ben [Sat, 13 Jan 2001 17:49:34 +0000 (17:49 +0000)]
fix typo; file -> files

Submitted by: Alexey Dokuchaev <danfe@inet.ssc.nsu.ru>

23 years agoOops, forgotten to add manpage.
iwasaki [Sat, 13 Jan 2001 16:18:37 +0000 (16:18 +0000)]
Oops, forgotten to add manpage.

23 years agoAdd manpage for acpiconf.
iwasaki [Sat, 13 Jan 2001 16:16:25 +0000 (16:16 +0000)]
Add manpage for acpiconf.

Submitted by: des

23 years agoMerged from sys/i386/conf/GENERIC revision from 1.286 to 1.291.
nyan [Sat, 13 Jan 2001 13:23:42 +0000 (13:23 +0000)]
Merged from sys/i386/conf/GENERIC revision from 1.286 to 1.291.

23 years agoMerged from sys/i386/conf/GENERIC.hints revision 1.6 and 1.7.
nyan [Sat, 13 Jan 2001 13:23:22 +0000 (13:23 +0000)]
Merged from sys/i386/conf/GENERIC.hints revision 1.6 and 1.7.

23 years agoCheck __i386__, not i386.
nyan [Sat, 13 Jan 2001 13:21:43 +0000 (13:21 +0000)]
Check __i386__, not i386.

23 years agoAdd Truecolor 16 and 32bits support. Note that 24bits modes are not
nsouch [Sat, 13 Jan 2001 11:30:17 +0000 (11:30 +0000)]
Add Truecolor 16 and 32bits support. Note that 24bits modes are not
supported since it's not easy to put 3 bytes accross 64Kb windows
of memory. This should not be such a problem with linear framebuffers.

There is no major interface modification except that the color type
becomes u_long instead of byte. So one just need to recompile his
application.

Approved by: Soren Schmidt <sos@freebsd.dk>

23 years ago/Really/ deprecate ConnectionsPerPeriod, ripping out the code for it
green [Sat, 13 Jan 2001 07:57:43 +0000 (07:57 +0000)]
/Really/ deprecate ConnectionsPerPeriod, ripping out the code for it
and giving a dire error to its lingering users.

23 years agoOnly install sysinstall.8 if the release/sysinstall directory exists. This
ben [Sat, 13 Jan 2001 05:50:19 +0000 (05:50 +0000)]
Only install sysinstall.8 if the release/sysinstall directory exists.  This
fixes the world for people without src-release.

PR: 24122
Submitted by: Steven G. Kargl <kargl@apl.washington.edu>

23 years agoThe -i option to ln requests user confirmation, not configuration.
dannyboy [Sat, 13 Jan 2001 01:53:23 +0000 (01:53 +0000)]
The -i option to ln requests user confirmation, not configuration.

PR: 24279
Submitted by: Roelof Osinga <roelof@nisser.com>

23 years ago/usr/libexec/cpp -> /usr/bin/cpp
ache [Sat, 13 Jan 2001 01:33:59 +0000 (01:33 +0000)]
/usr/libexec/cpp -> /usr/bin/cpp
because of new rename to /usr/libexec/cpp0

23 years agoNew release notes: Linksys Fast Ethernet cards and ed(4) driver flags,
bmah [Fri, 12 Jan 2001 23:01:15 +0000 (23:01 +0000)]
New release notes:  Linksys Fast Ethernet cards and ed(4) driver flags,
new API for hardware volume control, VESA S3 framebuffer driver, logging
of wrong-interface ARP replies sysctl, NFS client bug fixed, BurnProof(TM)
for ATAPI drives, IPFW works with ECN bits, ihfc(4), itjc(4),
<sys/selinfo.h>, pthread_* strong references, unified libgcc, SSH bug
with X11 forwarding fixed, syslogd(8) and LOG_CONSOLE, rpcgen use of
/usr/bin/cpp, rc.syscons, burncd(8) -m and -l, dmesg(a).

MFCs noted:  aac(4), OpenSSH 2.3.0.

23 years agoProperly compute the size of the final block of superblock summary information.
mckusick [Fri, 12 Jan 2001 21:56:55 +0000 (21:56 +0000)]
Properly compute the size of the final block of superblock summary information.

Submitted by: Ian Dowse <iedowse@maths.tcd.ie>

23 years agoman(7) -> mdoc(7).
ru [Fri, 12 Jan 2001 18:01:17 +0000 (18:01 +0000)]
man(7) -> mdoc(7).

23 years agoOops, the previous did not work with current mdoc(7).
ru [Fri, 12 Jan 2001 16:02:58 +0000 (16:02 +0000)]
Oops, the previous did not work with current mdoc(7).

23 years agomdoc(7) police: Ft/Vt now accept punctuation-type arguments.
ru [Fri, 12 Jan 2001 15:46:56 +0000 (15:46 +0000)]
mdoc(7) police: Ft/Vt now accept punctuation-type arguments.

23 years agoReduce diffs (mostly whitespace) to mdocNG.
ru [Fri, 12 Jan 2001 15:20:26 +0000 (15:20 +0000)]
Reduce diffs (mostly whitespace) to mdocNG.

Obtained from: mdocNG through NetBSD

23 years agoMake .Ft and .Vt macros accept punctuation characters as
ru [Fri, 12 Jan 2001 15:14:28 +0000 (15:14 +0000)]
Make .Ft and .Vt macros accept punctuation characters as
arguments, make .Vt usable outside the SYNOPSIS section.

Obtained from: mdocNG (not yet submitted for inclusion)

23 years agoMerged from sys/i386/isa/npx.c revision 1.87.
kato [Fri, 12 Jan 2001 13:42:53 +0000 (13:42 +0000)]
Merged from sys/i386/isa/npx.c revision 1.87.

23 years agoMerged from sys/i386/i386/machdep.c revisions 1.427 and 1.428.
kato [Fri, 12 Jan 2001 13:39:50 +0000 (13:39 +0000)]
Merged from sys/i386/i386/machdep.c revisions 1.427 and 1.428.