]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
24 years agoAdd options NETGRAPH, NETGRAPH_PPPOE and NETGRAPH_SOCKET
brian [Sat, 20 Nov 1999 10:54:57 +0000 (10:54 +0000)]
Add options NETGRAPH, NETGRAPH_PPPOE and NETGRAPH_SOCKET
to the release kernel.

24 years agostruct mountlist and struct mount.mnt_list have no business being
phk [Sat, 20 Nov 1999 10:00:46 +0000 (10:00 +0000)]
struct mountlist and struct mount.mnt_list have no business being
a CIRCLEQ.  Change them to TAILQ_HEAD and TAILQ_ENTRY respectively.

This removes ugly  mp != (void*)&mountlist  comparisons.

Requested by:   phk
Submitted by:   Jake Burkholder jake@checker.org
PR:             14967

24 years agoBack out the botched attempt to update to gnu grep 2.3 (lots of history
peter [Sat, 20 Nov 1999 09:40:28 +0000 (09:40 +0000)]
Back out the botched attempt to update to gnu grep 2.3 (lots of history
was lost).  Restore original version to try and avoid breaking the build
while David O'brien does a proper set of imports and merges.

Requested by: obrien

24 years agoI forgot to remove the "NCARD == 0" hacks that kept this driver
mdodd [Sat, 20 Nov 1999 09:28:01 +0000 (09:28 +0000)]
I forgot to remove the "NCARD == 0" hacks that kept this driver
from blowing up on PCMCIA systems.  The hack isn't needed anymore.

Reminded by: Blaz Zupan <blaz@gold.amis.net>

24 years agoLatest incremental efforts on newcard:
imp [Sat, 20 Nov 1999 07:23:28 +0000 (07:23 +0000)]
Latest incremental efforts on newcard:

o Delete pcic1 from NEWCARD
o Add explicit resource hints to pcic0
o Get attach working with newbus, kinda (it does all the newbus stuff,
  but doesn't try to attach the pccard yet, too many panics).
o Disable ie0 and le0 in NEWCARD config.  There appears to be a bug in
  the isa_compat code wrt memory conflicts with newbus drivers for
  reasons unknown.
o Minor cleanups.

24 years agoKill unneeded #includes. They were left over from the short period of
imp [Sat, 20 Nov 1999 05:16:14 +0000 (05:16 +0000)]
Kill unneeded #includes.  They were left over from the short period of
time when a strcmp against the device name was required.  Sio patches
sent to peter for review.

24 years agoKill unneeded includes
imp [Sat, 20 Nov 1999 05:02:00 +0000 (05:02 +0000)]
Kill unneeded includes

Found by phk's script a while ago.

24 years agoVnode was left referenced in the case if ELF image is broken.
bp [Sat, 20 Nov 1999 03:03:14 +0000 (03:03 +0000)]
Vnode was left referenced in the case if ELF image is broken.

Reviewed by: Peter Wemm <peter@netplex.com.au>

24 years agomkstemps() has been brought into 3-STABLE, so this isn't needed here
obrien [Sat, 20 Nov 1999 00:32:05 +0000 (00:32 +0000)]
mkstemps() has been brought into 3-STABLE, so this isn't needed here
anymore as building -CURRENT sources on 3-STABLE was the reason for the
previous revision adding this.

Note that since the GCC Project moved mkstemp.c from GCC's world to
libiberty, we no longer support building -CURRENT sources on non-FreeBSD
boxes unless that box has a very simular libc mix as FreeBSD.

24 years agoRemove the `sysctl' get osversion hack. There are issues with cross-hosting
obrien [Sat, 20 Nov 1999 00:28:03 +0000 (00:28 +0000)]
Remove the `sysctl' get osversion hack.  There are issues with cross-hosting
builds, Also, `sysctl' is not a build-tool and I'd rather not make it one.
3-STABLE has been updated to deal with the main reason this came in.

24 years agoFix HISTORY - the copyright header on the file of the GCC version was
obrien [Sat, 20 Nov 1999 00:15:17 +0000 (00:15 +0000)]
Fix HISTORY - the copyright header on the file of the GCC version was
misleading.

Submitted by: Theo de Raadt <deraadt@cvs.openbsd.org>

24 years agoConditionalise unwanted chattyness.
jkh [Fri, 19 Nov 1999 23:34:01 +0000 (23:34 +0000)]
Conditionalise unwanted chattyness.

24 years agoUpdate the time delta of the first item in TimerList before
brian [Fri, 19 Nov 1999 23:21:13 +0000 (23:21 +0000)]
Update the time delta of the first item in TimerList before
inserting a new item.  Without this, it's possible to
mis-insert quite badly... but only by as much as the load of
the first item, which is almost always 1 second.

Initialise the timerservice with `restart' set if we're inserting
at the start of the list.

24 years ago Optimize two cases in the MP locking code. First, it is not necessary
dillon [Fri, 19 Nov 1999 22:47:19 +0000 (22:47 +0000)]
  Optimize two cases in the MP locking code.  First, it is not necessary
      to use a locked cmpexg when unlocking a lock that we already hold, since
      nobody else can touch the lock while we hold it.  Second, it is not
      necessary to use a locked cmpexg when locking a lock that we already
      hold, for the same reason.  These changes will allow MP locks to be used
      recursively without impacting performance.

      Modify two procedures that are called only by assembly and are already
      NOPROF entries to pass a critical argument in %edx instead of on the
      stack, removing a significant amount of code from the critical path
      as a consequence.

Reviewed by: Alfred Perlstein <bright@wintelcom.net>, Peter Wemm <peter@netplex.com.au>

24 years agoUse LIST_FOREACH to traverse the allproc list.
phk [Fri, 19 Nov 1999 21:34:50 +0000 (21:34 +0000)]
Use LIST_FOREACH to traverse the allproc list.

Submitted by:   Jake Burkholder jake@checker.org

24 years agoThe at_exit and at_fork functions currently use a 'roll your own'
phk [Fri, 19 Nov 1999 21:29:03 +0000 (21:29 +0000)]
The at_exit and at_fork functions currently use a 'roll your own'
linked list to store the callbak routines.  The patch converts the
lists to queue(3) TAILQs, making the code slightly clearer and ensuring
that callbacks are executed in FIFO order.

Man page also updated as necesary.

(discontinued use of M_TEMP malloc type while here anyway /phk)

Submitted by:   Jake Burkholder jake@checker.org
PR:             14912

24 years agoRevert previous commit now that 'acfcomp' has been removed from
archie [Fri, 19 Nov 1999 20:42:14 +0000 (20:42 +0000)]
Revert previous commit now that 'acfcomp' has been removed from
libnetgraph/debug.c to fix make world.

24 years agoAdd a field archie forgot to merge in from out sources.
julian [Fri, 19 Nov 1999 20:32:46 +0000 (20:32 +0000)]
Add a field archie forgot to merge in from out sources.

24 years agoAdd to the HISTORY.
obrien [Fri, 19 Nov 1999 17:13:31 +0000 (17:13 +0000)]
Add to the HISTORY.

24 years ago Optimize two cases in the MP locking code. First, it is not necessary
dillon [Fri, 19 Nov 1999 16:49:30 +0000 (16:49 +0000)]
Optimize two cases in the MP locking code.  First, it is not necessary
    to use a locked cmpexg when unlocking a lock that we already hold, since
    nobody else can touch the lock while we hold it.  Second, it is not
    necessary to use a locked cmpexg when locking a lock that we already
    hold, for the same reason.  These changes will allow MP locks to be used
    recursively without impacting performance.

    Modify two procedures that are called only by assembly and are already
    NOPROF entries to pass a critical argument in %edx instead of on the
    stack, removing a significant amount of code from the critical path
    as a consequence.

Reviewed by: Alfred Perlstein <bright@wintelcom.net>, Peter Wemm <peter@netplex.com.au>

24 years agoProperly attribute interrupt time on alpha. Previously, interrupt time
gallatin [Fri, 19 Nov 1999 13:38:22 +0000 (13:38 +0000)]
Properly attribute interrupt time on alpha.  Previously, interrupt time
was likely to be counted as idle time.

Note that we are counting time spent in software interrupt handlers as
interrupt time, so this invalidates the i386 meaning of intr_nesting_level.

Reviewed by: dfr, bde
Tested by: anderson@cs.duke.edu

24 years agoMake the event creation work
n_hibma [Fri, 19 Nov 1999 13:13:28 +0000 (13:13 +0000)]
Make the event creation work
- Create the /dev/usb event queue device node.
- Add usbread to the cdevsw.
- Also hide the many minor() behind a USBUNIT macro.

24 years agoAdd share/examples/netgraph.
roberto [Fri, 19 Nov 1999 12:14:10 +0000 (12:14 +0000)]
Add share/examples/netgraph.

Forgotten by: our friends at Whistle.

24 years agoThe "acfcomp" field is not in the structure anymore. There may be a better
roberto [Fri, 19 Nov 1999 10:51:16 +0000 (10:51 +0000)]
The "acfcomp" field is not in the structure anymore. There may be a better
way to do that but it fixes buildworld.

World broken by: archie :-)

24 years agofix braino.. line misplaced.
julian [Fri, 19 Nov 1999 09:19:44 +0000 (09:19 +0000)]
fix braino.. line misplaced.

24 years agoLoose ad_sleep, its of no use anymore.
sos [Fri, 19 Nov 1999 08:21:15 +0000 (08:21 +0000)]
Loose ad_sleep, its of no use anymore.

24 years agoRename es1370.c to es137x.c and merge in the patches to support
roger [Fri, 19 Nov 1999 07:29:10 +0000 (07:29 +0000)]
Rename es1370.c to es137x.c and merge in the patches to support
ES1371 and ES1373 PCI Audio Sound Chips.
The 1371 and 1373 can be found on newer CreativeLabs/Ensoniq
sound cards such as the SoundBlaster PCI 16,64 and 128.

Submitted by: Russell Cattelan <cattelan@thebarn.com>
Obtained from: Parts obtained from linux, but rewritten by Russell

24 years agoAdd an example of how to run raw IP across a sync port.
julian [Fri, 19 Nov 1999 07:04:36 +0000 (07:04 +0000)]
Add an example of how to run raw IP across a sync port.

24 years agoAdd an example of how to set up a frame relay link using netgraph and
julian [Fri, 19 Nov 1999 06:56:34 +0000 (06:56 +0000)]
Add an example of how to set up a frame relay link using netgraph and
the 'sr' sync card.

24 years agoModule name is cardbus, not pccard.
imp [Fri, 19 Nov 1999 06:31:24 +0000 (06:31 +0000)]
Module name is cardbus, not pccard.

24 years agoUse 'struct ng_xxx_private' instead of 'struct private' to help gdb
archie [Fri, 19 Nov 1999 05:50:29 +0000 (05:50 +0000)]
Use 'struct ng_xxx_private' instead of 'struct private' to help gdb
disambiguate when debugging.

24 years agoFix bug where hook pointers were not getting updated on disconnection.
archie [Fri, 19 Nov 1999 05:49:54 +0000 (05:49 +0000)]
Fix bug where hook pointers were not getting updated on disconnection.

24 years agoMove misplaced #define.
archie [Fri, 19 Nov 1999 05:49:18 +0000 (05:49 +0000)]
Move misplaced #define.

24 years agoAdd some safety using KASSERT() and splnet().
archie [Fri, 19 Nov 1999 05:45:11 +0000 (05:45 +0000)]
Add some safety using KASSERT() and splnet().

24 years agoUse KASSERT() instead of old #ifdef DIAGNOSTICS.
archie [Fri, 19 Nov 1999 05:43:33 +0000 (05:43 +0000)]
Use KASSERT() instead of old #ifdef DIAGNOSTICS.
Add more INVARIANTS-enabled sanity checking.

24 years agoprotect some more operations with splimp() under Netgraph.
julian [Fri, 19 Nov 1999 05:37:37 +0000 (05:37 +0000)]
protect some more operations with splimp() under Netgraph.

24 years agoIn revision 1.21 I changed the search order for shared libraries,
jdp [Fri, 19 Nov 1999 04:45:07 +0000 (04:45 +0000)]
In revision 1.21 I changed the search order for shared libraries,
but I forgot to make the corresponding fix to the comment.  Rectify
that.

Submitted by: Tony Finch <fanf@demon.net>

24 years agoUpdate references.
archie [Fri, 19 Nov 1999 04:41:09 +0000 (04:41 +0000)]
Update references.

24 years agoMiscellaneous edits.
archie [Fri, 19 Nov 1999 04:40:34 +0000 (04:40 +0000)]
Miscellaneous edits.

24 years agoAdd reference to ng_pppoe(8).
archie [Fri, 19 Nov 1999 04:32:28 +0000 (04:32 +0000)]
Add reference to ng_pppoe(8).

24 years agoUpdate to reflect changed functionality.
archie [Fri, 19 Nov 1999 04:30:26 +0000 (04:30 +0000)]
Update to reflect changed functionality.

24 years agoRemove the address and control field compression functionality of this
archie [Fri, 19 Nov 1999 04:27:53 +0000 (04:27 +0000)]
Remove the address and control field compression functionality of this
node type.  ACF is device independent and therefore belongs in ng_ppp.c
(which already implements it).

24 years agoRename 'struct private' to 'struct ng_xxx_private' to allow gdb
archie [Fri, 19 Nov 1999 04:25:39 +0000 (04:25 +0000)]
Rename 'struct private' to 'struct ng_xxx_private' to allow gdb
to disambiguate when debugging.

24 years agoReenable "umodem", as its build is not broken anymore.
green [Fri, 19 Nov 1999 04:04:56 +0000 (04:04 +0000)]
Reenable "umodem", as its build is not broken anymore.

24 years agoSpelling fix.
billf [Fri, 19 Nov 1999 02:16:26 +0000 (02:16 +0000)]
Spelling fix.

24 years agoMake `pkg_add -r package1 package2 package3 ...` work correctly.
cpiazza [Thu, 18 Nov 1999 23:32:43 +0000 (23:32 +0000)]
Make `pkg_add -r package1 package2 package3 ...` work correctly.

Reviewed by: billf

24 years agoStop run() from dereferencing a NULL pointer when you enter a valid command
roberto [Thu, 18 Nov 1999 23:04:38 +0000 (23:04 +0000)]
Stop run() from dereferencing a NULL pointer when you enter a valid command
then a ^D at the prompt.

cdcontrol> play 1 14
cdcontrol>
zsh: 40874 segmentation fault (core dumped)  /usr/sbin/cdcontrol -f /dev/cd1c

24 years agoCop on a bit and regenerate things correctly.
brian [Thu, 18 Nov 1999 20:45:04 +0000 (20:45 +0000)]
Cop on a bit and regenerate things correctly.

Pointed out by: bde

24 years agoCleanup of the code before we add a few thins.
n_hibma [Thu, 18 Nov 1999 18:04:17 +0000 (18:04 +0000)]
Cleanup of the code before we add a few thins.

- remove the use of NDEV. It is confusing. MAXUSBDEV should do.
- add some comments.
- add more explanation in usage()
- change the timeout value for undetected USB devices from 300 to 30
  seconds.  I don't think anyone wants to wait 5 minutes for broken
  devices to show up. The overhead CPU wise is very little.
- print 'no controllers found' as a fatal error.
- remove inclusion of malloc.h. It's unused.

24 years agoAdd usbdevs and friends.
n_hibma [Thu, 18 Nov 1999 18:02:44 +0000 (18:02 +0000)]
Add usbdevs and friends.

24 years agoDo not suggest user to change /etc/defaults/rc.conf.
phantom [Thu, 18 Nov 1999 16:09:40 +0000 (16:09 +0000)]
Do not suggest user to change /etc/defaults/rc.conf.

Requested by: sheldonh

24 years agomdoc(7)'fy
phantom [Thu, 18 Nov 1999 16:04:53 +0000 (16:04 +0000)]
mdoc(7)'fy

Mostly submitted by: "Philippe Charnier" <charnier@xp11.frmug.org>

24 years agoSet port_used and mem_used to the resource size instead of 1.
nyan [Thu, 18 Nov 1999 14:18:19 +0000 (14:18 +0000)]
Set port_used and mem_used to the resource size instead of 1.

Submitted by: chi@bd.mbn.or.jp (Chiharu Shibata)

24 years agoChange the clist reservation from 0 to buffer size.
n_hibma [Thu, 18 Nov 1999 12:50:48 +0000 (12:50 +0000)]
Change the clist reservation from 0 to buffer size.

Submitted by:  mmoeller@infolibria.com

24 years agoMissed a name change.
n_hibma [Thu, 18 Nov 1999 12:45:09 +0000 (12:45 +0000)]
Missed a name change.

24 years agoSync with sys/isa/sio.c revision up to 1.275.
nyan [Thu, 18 Nov 1999 12:22:09 +0000 (12:22 +0000)]
Sync with sys/isa/sio.c revision up to 1.275.

24 years agoSync with sys/isa/fd.c revision 1.168.
nyan [Thu, 18 Nov 1999 11:34:26 +0000 (11:34 +0000)]
Sync with sys/isa/fd.c revision 1.168.

24 years agoSync with sys/i386/conf/Makefile.i386 revision up to 1.163.
nyan [Thu, 18 Nov 1999 11:26:45 +0000 (11:26 +0000)]
Sync with sys/i386/conf/Makefile.i386 revision up to 1.163.

24 years agoArgh, don't turn the IIR test on unless it's a pccard. These tests mess
peter [Thu, 18 Nov 1999 10:29:06 +0000 (10:29 +0000)]
Argh, don't turn the IIR test on unless it's a pccard.  These tests mess
up the subsequent probes.

24 years agoMake last commit bde-compliant:
pb [Thu, 18 Nov 1999 10:20:45 +0000 (10:20 +0000)]
Make last commit bde-compliant:
- correct indentation
- change data types for consistency with the rest of ping.c
- create new variable separate from "answer" for clarity

24 years agoHack: comment out the pull-in of src/sys/../share/mk/bsd.kern.mk as
peter [Thu, 18 Nov 1999 09:54:11 +0000 (09:54 +0000)]
Hack: comment out the pull-in of src/sys/../share/mk/bsd.kern.mk as
that breaks if you try and compile a kernel before building world, as
is presently required to get past the signal changes.   I don't
particularly like doing this, but at least it will mean that a 'make world'
will activate the gcc 2.95.2-specific option in bsd.kern.mk.

24 years agoFix an unused variable warning.
peter [Thu, 18 Nov 1999 09:07:30 +0000 (09:07 +0000)]
Fix an unused variable warning.

24 years agoIf we have found pci devices via pci_cfgopen(), but don't find a
peter [Thu, 18 Nov 1999 08:58:37 +0000 (08:58 +0000)]
If we have found pci devices via pci_cfgopen(), but don't find a
host->pci bridge specifically, then add a pcib0 device on the motherboard
for the pci bus to hang off.

Requested by:  Anders Andersson <anders@sanyusan.se>
Obtained from: dfr

24 years agoremove PCATCH from a tsleep that isn't checked for (avoid infinite loop).
alfred [Thu, 18 Nov 1999 08:57:03 +0000 (08:57 +0000)]
remove PCATCH from a tsleep that isn't checked for (avoid infinite loop).

this needs to be revisited.

24 years agoFix a warning (unused static declaration without MFS_ROOT)
peter [Thu, 18 Nov 1999 08:49:40 +0000 (08:49 +0000)]
Fix a warning (unused static declaration without MFS_ROOT)

24 years agoFix a warning.
peter [Thu, 18 Nov 1999 08:47:10 +0000 (08:47 +0000)]
Fix a warning.

24 years agoFix some -Wunused warnings
peter [Thu, 18 Nov 1999 08:43:58 +0000 (08:43 +0000)]
Fix some -Wunused warnings

24 years agoFix some warnings.
peter [Thu, 18 Nov 1999 08:41:53 +0000 (08:41 +0000)]
Fix some warnings.

24 years agoFix a warning.
peter [Thu, 18 Nov 1999 08:39:02 +0000 (08:39 +0000)]
Fix a warning.

24 years agoTouch up the make_dev() usage to hopefully make it work and fix warnings.
peter [Thu, 18 Nov 1999 08:36:42 +0000 (08:36 +0000)]
Touch up the make_dev() usage to hopefully make it work and fix warnings.

24 years agoFix some warnings.
peter [Thu, 18 Nov 1999 08:32:02 +0000 (08:32 +0000)]
Fix some warnings.

24 years agoFix a warning and a potential panic if TCPDEBUG is active. (tp is
peter [Thu, 18 Nov 1999 08:28:24 +0000 (08:28 +0000)]
Fix a warning and a potential panic if TCPDEBUG is active.  (tp is
a wild pointer and used by TCPDEBUG2())

24 years agoFix a warning (unused variable RCSid)
peter [Thu, 18 Nov 1999 08:18:13 +0000 (08:18 +0000)]
Fix a warning (unused variable RCSid)

24 years ago Ensure that garbage from the kernel stack does not wind up being
dillon [Thu, 18 Nov 1999 08:14:20 +0000 (08:14 +0000)]
Ensure that garbage from the kernel stack does not wind up being
    returned to user mode in the spare fields of the stat structure.

PR: kern/14966
Reviewed by: dillon@freebsd.org
Submitted by: Kelly Yancey kbyanc@posi.net

24 years agoFollow-up on mdoc fixes in rev 1.35; this includes things that I missed
sheldonh [Thu, 18 Nov 1999 08:13:21 +0000 (08:13 +0000)]
Follow-up on mdoc fixes in rev 1.35; this includes things that I missed
in that revision as well as things I broke in that revision.  A note-
worthy instance of the latter case was the inversion of -E and -V in the
subsection on Commandline Editing.

24 years ago Only bother converting the stat structure if we intend to return it,
dillon [Thu, 18 Nov 1999 08:08:28 +0000 (08:08 +0000)]
Only bother converting the stat structure if we intend to return it,
    when no error occurs.

PR: kern/14966
Reviewed by: dillon@freebsd.org
Submitted by: Kelly Yancey kbyanc@posi.net

24 years agoZap nsio
peter [Thu, 18 Nov 1999 07:34:58 +0000 (07:34 +0000)]
Zap nsio

24 years agoSync to latest cardbusdevs file
imp [Thu, 18 Nov 1999 07:22:59 +0000 (07:22 +0000)]
Sync to latest cardbusdevs file

24 years agoMerge some typo fixes from dev/sio/sio.c (siostop -> comstop)
peter [Thu, 18 Nov 1999 07:22:41 +0000 (07:22 +0000)]
Merge some typo fixes from dev/sio/sio.c (siostop -> comstop)
Remove EXTRA_SIO/NSIOTOT and make it fully dynamic (from dev/sio/sio.c)
Make sio work for pccard here - pccardd doesn't activate interrupts
until after prove has succeeded.
Mark the initial reset of likely sio ports as broken as it depended on
config supplying a list of locations to probe, devices are now proved
standalone.
Optimize a bit of COM_NOAST4() logic.
Use bus_get_resource_start() etc rather than using isa-centric calls.
Reactivate the IIR_TXRDYBUG test, I've got a card here with it.
Try to be a bit smarter about activating interrupts (ie: don't panic
if polled)
Fix some style bugs that have crept in over time (there are still more).

24 years agoRaw import of newconfig cardbus code. This is effectively an import,
imp [Thu, 18 Nov 1999 07:21:51 +0000 (07:21 +0000)]
Raw import of newconfig cardbus code.  This is effectively an import,
so the code doesn't compile.  I added $FreeBSD$ headers.

24 years agoInitial import of cardbus bridge driver (cbb) from the latest
imp [Thu, 18 Nov 1999 07:14:54 +0000 (07:14 +0000)]
Initial import of cardbus bridge driver (cbb) from the latest
newconfig code.  This is a raw import and doesn't compile yet.

Obtained from: newconfig project

24 years agoAdd modules for cardbus and the cardbus bridge driver. Files needed
imp [Thu, 18 Nov 1999 07:11:16 +0000 (07:11 +0000)]
Add modules for cardbus and the cardbus bridge driver.  Files needed
to build these to be committed next.

Again, not added to sys/modules/Makefile

24 years agoNewcard's pccard module. It compiles, but doesn't work. This should
imp [Thu, 18 Nov 1999 07:02:18 +0000 (07:02 +0000)]
Newcard's pccard module.  It compiles, but doesn't work.  This should
make development easier and push the ability of newbus to load busses
to the test.

Not added to sys/modules/Makefile because it isn't ready to break the
nightly snapshots for alpha yet :-).

NOTE: This is only for NEWCARD.  The old pccard stuff will not build
as a loadable module.

24 years agoRemove the non-functional "swap device" userland front-end to the
peter [Thu, 18 Nov 1999 06:55:40 +0000 (06:55 +0000)]
Remove the non-functional "swap device" userland front-end to the
multiplexed underlying swap devices (/dev/drum).  The only thing it did
was to allow root to open /dev/drum, but not do anything with it.
Various utilities used to grovel around in here, but Matt has written
a much nicer (and clean) front-end to this for libkvm, and nothing uses
the old system any more.

The VM system was calling VOP_STRATEGY() on the vp of the first underlying
swap device (not the /dev/drum one, the first real device), and using
the VOP system to indirectly (and only) call swstrategy() to choose
an underlying device and enqueue it on that device.  I have changed it
to avoid diverting through the VOP system and to call the only possible
target directly, saving a little bit of time and some complexity.

In all, nothing much changes, except some scaffolding to support the
roundabout way of calling swstrategy() is gone.

Matt gave me the ok to do this some time ago, and I apologize for taking
so long to get around to it.

24 years agoHijack old, broken, obsolete, unsupported pcic module for the NEWCARD
imp [Thu, 18 Nov 1999 06:54:19 +0000 (06:54 +0000)]
Hijack old, broken, obsolete, unsupported pcic module for the NEWCARD
code.  You can now build the newcard's pcic driver as a module for all
the joy that will bring you, which currently isn't so much joy as it
is pain.

The old pccard module will never be made to work again, so I think
this is OK.  Note, it still remains disabled in sys/modules/Makefile
on purpose.

24 years agoRemove cdevsw_add() as the make_dev()'s are done in a clone-like fashion.
peter [Thu, 18 Nov 1999 06:41:30 +0000 (06:41 +0000)]
Remove cdevsw_add() as the make_dev()'s are done in a clone-like fashion.
(ie: the next device is created with make_dev as the first is opened
and so on)

24 years agoRemove cdevsw_add() - the make_dev() calls are already there.
peter [Thu, 18 Nov 1999 06:39:47 +0000 (06:39 +0000)]
Remove cdevsw_add() - the make_dev() calls are already there.
Don't '#if NSNP > 0' around the thing as it's silly and not required.
config(8) only causes it to be compiled if NSNP is > 0 by definition.

24 years agoRemove cdevsw_add() - the necessary make_dev() is already there.
peter [Thu, 18 Nov 1999 06:37:00 +0000 (06:37 +0000)]
Remove cdevsw_add() - the necessary make_dev() is already there.

24 years agoRemove cdevsw_add() - the necessary make_dev() calls appear to be there
peter [Thu, 18 Nov 1999 06:34:47 +0000 (06:34 +0000)]
Remove cdevsw_add() - the necessary make_dev() calls appear to be there
already.

24 years agoFix a warning.
peter [Thu, 18 Nov 1999 06:29:57 +0000 (06:29 +0000)]
Fix a warning.

24 years ago'const'ify a bunch of pointers in the resource_*() functions for accessing
peter [Thu, 18 Nov 1999 06:05:30 +0000 (06:05 +0000)]
'const'ify a bunch of pointers in the resource_*() functions for accessing
the config hint tables.  This fixes a few warnings elsewhere (eg: fd).

24 years agoConvert smb to use make_dev() rather than cdevsw_add(). This stops the
peter [Thu, 18 Nov 1999 05:44:56 +0000 (05:44 +0000)]
Convert smb to use make_dev() rather than cdevsw_add().  This stops the
annoying 'smb is usurping smb's cdevsw' type messages.  (Yes, there are
other ways to fix cdevsw_add(), but that is a doomed api)

24 years agoConvert iic to use make_dev() rather than cdevsw_add(). This stops the
peter [Thu, 18 Nov 1999 05:43:32 +0000 (05:43 +0000)]
Convert iic to use make_dev() rather than cdevsw_add().  This stops the
annoying 'iic is usurping iic's cdevsw' type messages.  (Yes, there are
other ways to fix cdevsw_add(), but that is a doomed api)

24 years agoTidy up a few loose ends in the fifo setup code.
peter [Thu, 18 Nov 1999 05:15:09 +0000 (05:15 +0000)]
Tidy up a few loose ends in the fifo setup code.
Don't use NFDC as an arbitary limit, it is not required and goes against
using PnP fdc devices (eg: when PNPBIOS is turned on, the motherboard
devices (sio, fdc, etc etc) are detected via PnP, not config(8).)

24 years agoUse a common MAKEDEV between the Alpha and i386 - the only differences
peter [Thu, 18 Nov 1999 04:31:31 +0000 (04:31 +0000)]
Use a common MAKEDEV between the Alpha and i386 - the only differences
of significance were due to the Alpha version going stale.  We use common
device numbering for devices.  Hopefully, this file's days are numbered.

24 years agoFor now, disable umodem module building, because it doesn't. Don't
green [Thu, 18 Nov 1999 04:27:10 +0000 (04:27 +0000)]
For now, disable umodem module building, because it doesn't.  Don't
want to break the snapshots :)

24 years agoMove Alpha conflicting entries (98,99) to 135,136. Renumber the zsc
peter [Thu, 18 Nov 1999 04:20:23 +0000 (04:20 +0000)]
Move Alpha conflicting entries (98,99) to 135,136.  Renumber the zsc
CDEV_MAJOR entry to match..  Is "ipr" in the tree?  I can't find it.

24 years agoDon't ask about SCO/IBSC2 binary support on the Alpha.
obrien [Thu, 18 Nov 1999 03:03:01 +0000 (03:03 +0000)]
Don't ask about SCO/IBSC2 binary support on the Alpha.

24 years agoFor the TCP transport, put the listening socket in non-blocking
jdp [Thu, 18 Nov 1999 03:01:06 +0000 (03:01 +0000)]
For the TCP transport, put the listening socket in non-blocking
mode.  This addresses a well-known race condition that can cause
servers to hang in accept().  The relevant case is when somebody
connects to the server and then immediately kills the connection
by sending a TCP reset.  On the server this causes select to report
a ready condition on the socket, after which the accept call blocks
because there is no longer any pending connection to accept.

In -current there is already a work-around for this in the kernel.
It was merged into -stable some time ago, but then David Greenman
reverted it because it seemed to be causing a socket leak in some
cases.  (See uipc_socket.c revision 1.51.2.3.)  Hence this userland
fix is needed in -stable, and I plan to merge it into that branch
soon because it fixes a potential DoS attack.  It may also be needed
in -current if the suspected socket leak turns out to be real.  In
any case, after thinking it over I believe the fix belongs in
userland.  An application shouldn't assume that a ready return from
select guarantees that the subsequent I/O operation cannot block.
A lot can happen between the select and the accept.

A similar fix should most likely be applied to the Unix domain
socket transport too.

Submitted by: peter
Reviewed by: jdp

24 years agofix up a whitespace screwup I made.
julian [Thu, 18 Nov 1999 01:23:06 +0000 (01:23 +0000)]
fix up a whitespace screwup I made.