]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoAdapt sysinstall to use the new msgNoYes() function which assumes
Jordan K. Hubbard [Thu, 14 Dec 2000 02:49:02 +0000 (02:49 +0000)]
Adapt sysinstall to use the new msgNoYes() function which assumes
no as a default.  Sysinstall should be both less dangerous and less
annoying as a result of this change, though that's just my opinion
(since they're the defaults which annoy ME the least :).

23 years agoAdd a new function, dialog_noyes(), for sysinstall to be able to
Jordan K. Hubbard [Thu, 14 Dec 2000 02:35:22 +0000 (02:35 +0000)]
Add a new function, dialog_noyes(), for sysinstall to be able to
present questinos with a different default answer.  Somebody submitted
a patch to me once which did something this but I lost it (my bad) so
I'm just going to re-implement it with thanks to whomever it was who
gave me the idea.

23 years agoo Check return of strdup() for NULL, abort if so.
Robert Watson [Thu, 14 Dec 2000 01:51:39 +0000 (01:51 +0000)]
o Check return of strdup() for NULL, abort if so.

Submitted by: jedgar

23 years agoFix this based on Mike's moved pci code. We needed to delete the pci_m
Warner Losh [Thu, 14 Dec 2000 00:06:10 +0000 (00:06 +0000)]
Fix this based on Mike's moved pci code.  We needed to delete the pci_m
from this module anyway, so I just did that.

23 years agoDocument the MA_RECURSED and MA_NOTRECURSED assertion flags as well as
John Baldwin [Wed, 13 Dec 2000 23:30:43 +0000 (23:30 +0000)]
Document the MA_RECURSED and MA_NOTRECURSED assertion flags as well as
the MTX_QUIET flag.

23 years agoFix include directories for crossbuilding.
Marcel Moolenaar [Wed, 13 Dec 2000 23:15:12 +0000 (23:15 +0000)]
Fix include directories for crossbuilding.

aicasm is run on the build machine and therefore needs to be
compiled and linked against the headers and libraries (resp)
of the build machine. Since normally the default include
directories are search after any specified on the command
line, make sure we don't accidentally pick up machine
dependent headers from the kernel compile directory by
specifying /usr/include first.

This solves the (cross) build problem for ia64.

Approved by: gibbs

23 years agoFix a logical bug introduced by changing the formatting.
Thomas-Henning von Kamptz [Wed, 13 Dec 2000 22:19:05 +0000 (22:19 +0000)]
Fix a logical bug introduced by changing the formatting.

Submitted by: ru
Reviewed by: chm

23 years ago- Add a new flag MTX_QUIET that can be passed to the various mtx_*
John Baldwin [Wed, 13 Dec 2000 21:53:42 +0000 (21:53 +0000)]
- Add a new flag MTX_QUIET that can be passed to the various mtx_*
  functions.  If this flag is set, then no KTR log messages are issued.
  This is useful for blocking excessive logging, such as with the internal
  mutex used by the witness code.
- Use MTX_QUIET on all of the mtx_enter/exit operations on the internal
  mutex used by the witness code.
- If we are in a panic, don't do witness checks in witness_enter(),
  witness_exit(), and witness_try_enter(), just return.

23 years agoo Remove the COMPAT_OLDPCI option now that Mike removed it.
Warner Losh [Wed, 13 Dec 2000 21:52:44 +0000 (21:52 +0000)]
o Remove the COMPAT_OLDPCI option now that Mike removed it.
o Add pmtimer device
o Shorten scsi settle time to 1 second, since that's what I use and
  I'm tired of remerging it every time.

23 years agoReconnect linprocfs.
Dag-Erling Smørgrav [Wed, 13 Dec 2000 20:29:30 +0000 (20:29 +0000)]
Reconnect linprocfs.

23 years agoUse kinfo_proc instead of eproc (which Kirk deep-sixed earlier this week)
Dag-Erling Smørgrav [Wed, 13 Dec 2000 20:22:24 +0000 (20:22 +0000)]
Use kinfo_proc instead of eproc (which Kirk deep-sixed earlier this week)

Generate a version string that looks just like a real Linux one - almost :)

Use sbufs everywhere instead of sprintf(). Note that this is still imperfect,
as the code does not check whether the sbuf overflowed - but it'll still
work better than before, since if the sbuf overflows, the code now simply
copies out 0 bytes instead of causing a trap (or worse, corrupting kernel
structures)

23 years agoAdd dependency on linux, which is needed for proc/version.
Dag-Erling Smørgrav [Wed, 13 Dec 2000 20:19:06 +0000 (20:19 +0000)]
Add dependency on linux, which is needed for proc/version.

23 years agoFor PC164: document "ERROR: ISA table corrupt!" and workaround
Wilko Bulte [Wed, 13 Dec 2000 19:57:08 +0000 (19:57 +0000)]
For PC164: document "ERROR: ISA table corrupt!" and workaround

Thanks to: naddy@mips.inka.de

23 years agoAdd sbuf(9) man page with links to API function names.
Dag-Erling Smørgrav [Wed, 13 Dec 2000 19:53:37 +0000 (19:53 +0000)]
Add sbuf(9) man page with links to API function names.

Reviewed by: ru

23 years agoAdd subr_sbuf.c.
Dag-Erling Smørgrav [Wed, 13 Dec 2000 19:52:12 +0000 (19:52 +0000)]
Add subr_sbuf.c.

23 years agoString buffer API
Dag-Erling Smørgrav [Wed, 13 Dec 2000 19:51:07 +0000 (19:51 +0000)]
String buffer API

23 years agoFlesh out a simple framework for dependency checking rc.conf enabled
Brian S. Dean [Wed, 13 Dec 2000 19:17:54 +0000 (19:17 +0000)]
Flesh out a simple framework for dependency checking rc.conf enabled
services.  Simple YES/NO style *_enable services are supported.

Reviewed by: obrien

23 years agoIf we fail to emulate a vm86 trap in kernel mode, then we use
John Baldwin [Wed, 13 Dec 2000 18:57:15 +0000 (18:57 +0000)]
If we fail to emulate a vm86 trap in kernel mode, then we use
vm86_trap() to return to the calling program directly.  vm86_trap()
doesn't return, thus it was never returning to trap() to release
Giant.  Thus, release Giant before calling vm86_trap().

23 years agoInclude vm/vm_zone.h prior to vm/swap_pager.h.
Seigo Tanimura [Wed, 13 Dec 2000 17:36:03 +0000 (17:36 +0000)]
Include vm/vm_zone.h prior to vm/swap_pager.h.

Noticed by: Michael Harnois <mdharnois@home.com>
Submitted by: assar

23 years agoStrip the .comment and .note sections when stripping. There's no
Robert Nordier [Wed, 13 Dec 2000 17:10:38 +0000 (17:10 +0000)]
Strip the .comment and .note sections when stripping.  There's no
point in retaining this info, particularly under BTX.

23 years agoThis file was not regenerated for the latest import, thus the
Ruslan Ermilov [Wed, 13 Dec 2000 15:54:14 +0000 (15:54 +0000)]
This file was not regenerated for the latest import, thus the
reported grep version is wrong.  The correct version is 2.4d.

23 years agoI really hate it when part of a patch gets left out.
Julian Elischer [Wed, 13 Dec 2000 15:42:37 +0000 (15:42 +0000)]
I really hate it when part of a patch gets left out.
This was still sitting in my commit tree.
Luckily I always compare my before and after trees...

23 years agodocument TUNSIFINFO, TUNGIFINFO, TUNSIFPID, TUNSIFHEAD, TUNGIFHEAD
Assar Westerlund [Wed, 13 Dec 2000 13:40:42 +0000 (13:40 +0000)]
document TUNSIFINFO, TUNGIFINFO, TUNSIFPID, TUNSIFHEAD, TUNGIFHEAD

23 years agoMove `#include <sys/devicestat.h>' into #ifdef/#endif to keep
Noriaki Mitsunaga [Wed, 13 Dec 2000 13:35:46 +0000 (13:35 +0000)]
Move `#include <sys/devicestat.h>' into #ifdef/#endif to keep
comaptibility with NetBSD/pc98.

23 years ago-Change manufacture ID for REX-5535AC, REX-5535X, REX-9835X, and REX-9835Z
Noriaki Mitsunaga [Wed, 13 Dec 2000 13:29:16 +0000 (13:29 +0000)]
-Change manufacture ID for REX-5535AC, REX-5535X, REX-9835X, and REX-9835Z
 from "PCMCIA SCSI MBH1040" to "PCMCIA SCSI MBH10401" "01". They are
 based on the spc driver.
 This will fix the conflicts of entry with REX-5536AM, REX-9836A,
 and ICM PSC-2401 ("MBH10404" ones) which are based on stg driver.
 The problem was pointed out in bsd-nomads several times since PAO2 days.

-Comment out the entry for "MBH10401" ones. The spc driver is
 not supported yet.

-Add more comment about cards which has broken CIS
 (some cards which has tuple of "PCMCIA SCSI MBH10404" "01").

Reported by: takachan@running-dog.net, y-nakaga@nwsl.mesh.ad.jp,
yuki@dayo.ne.jp
Obtained from: discussion in bsd-nomads mailing list

23 years agoMdoc(7)ify.
Ruslan Ermilov [Wed, 13 Dec 2000 11:33:19 +0000 (11:33 +0000)]
Mdoc(7)ify.

23 years agoIf the URL did not specify a scheme, try to guess it from the host name.
Dag-Erling Smørgrav [Wed, 13 Dec 2000 11:26:27 +0000 (11:26 +0000)]
If the URL did not specify a scheme, try to guess it from the host name.

23 years agoAvoid a segfault (due to an unitialized pointer) when parsing URLs that have
Dag-Erling Smørgrav [Wed, 13 Dec 2000 11:21:09 +0000 (11:21 +0000)]
Avoid a segfault (due to an unitialized pointer) when parsing URLs that have
no scheme or host part.

23 years agoDo not race for the lock of an inode hash.
Seigo Tanimura [Wed, 13 Dec 2000 10:04:01 +0000 (10:04 +0000)]
Do not race for the lock of an inode hash.

Reviewed by: jhb

23 years ago- If swap metadata does not fit into the KVM, reduce the number of
Seigo Tanimura [Wed, 13 Dec 2000 10:01:00 +0000 (10:01 +0000)]
- If swap metadata does not fit into the KVM, reduce the number of
struct swblock entries by dividing the number of the entries by 2
until the swap metadata fits.

- Reject swapon(2) upon failure of swap_zone allocation.

This is just a temporary fix. Better solutions include:
(suggested by: dillon)

o reserving swap in SWAP_META_PAGES chunks, and
o swapping the swblock structures themselves.

Reviewed by: alfred, dillon

23 years agoIntroduce a new potientially cleaner interface for accessing per-cpu
Jake Burkholder [Wed, 13 Dec 2000 09:23:53 +0000 (09:23 +0000)]
Introduce a new potientially cleaner interface for accessing per-cpu
variables from i386 assembly language.  The syntax is PCPU(member)
where member is the capitalized name of the per-cpu variable, without
the gd_ prefix.  Example: movl %eax,PCPU(CURPROC).  The capitalization
is due to using the offsets generated by genassym rather than the symbols
provided by linking with globals.o.  asmacros.h is the wrong place for
this but it seemed as good a place as any for now.  The old implementation
in asnames.h has not been removed because it is still used to de-mangle
the symbols used by the C variables for the UP case.

23 years agoAdd route interrupt method.
Matt Jacob [Wed, 13 Dec 2000 09:07:16 +0000 (09:07 +0000)]
Add route interrupt method.

23 years agoLock the allproc list.
Jake Burkholder [Wed, 13 Dec 2000 09:05:45 +0000 (09:05 +0000)]
Lock the allproc list.

Approved by: DES

23 years ago#endif should not have a non-comment token after it.
David E. O'Brien [Wed, 13 Dec 2000 08:59:18 +0000 (08:59 +0000)]
#endif should not have a non-comment token after it.
GCC 2.97 (snapshot) complains about this.

23 years agoPreventing runaway kernel soft updates memory, take three.
Kirk McKusick [Wed, 13 Dec 2000 08:30:35 +0000 (08:30 +0000)]
Preventing runaway kernel soft updates memory, take three.
Previously, the syncer process was the only process in the
system that could process the soft updates background work
list. If enough other processes were adding requests to that
list, it would eventually grow without bound. Because some of
the work list requests require vnodes to be locked, it was
not generally safe to let random processes process the work
list while they already held vnodes locked. By adding a flag
to the work list queue processing function to indicate whether
the calling process could safely lock vnodes, it becomes possible
to co-opt other processes into helping out with the work list.
Now when the worklist gets too large, other processes can safely
help out by picking off those work requests that can be handled
without locking a vnode, leaving only the small number of
requests requiring a vnode lock for the syncer process. With
this change, it appears possible to keep even the nastiest
workloads under control.

Submitted by: Paul Saab <ps@yahoo-inc.com>

23 years agoSync with i386/GENERIC rev 1.294 removing "COMPAT_OLDPCI".
David E. O'Brien [Wed, 13 Dec 2000 07:34:47 +0000 (07:34 +0000)]
Sync with i386/GENERIC rev 1.294 removing "COMPAT_OLDPCI".

This fixed the broken kernel build on the Alpha.

23 years agoRemove unnecessary includes found by phk's script. I've been building
Warner Losh [Wed, 13 Dec 2000 06:28:37 +0000 (06:28 +0000)]
Remove unnecessary includes found by phk's script.  I've been building
these locally for ages.

23 years agoFix problem with ax88190 based cards trying to probe further after
Warner Losh [Wed, 13 Dec 2000 06:27:23 +0000 (06:27 +0000)]
Fix problem with ax88190 based cards trying to probe further after
matching the ax88190.

23 years agoRemove unnecessary includes found by phk's script a long time ago.
Warner Losh [Wed, 13 Dec 2000 06:25:22 +0000 (06:25 +0000)]
Remove unnecessary includes found by phk's script a long time ago.

23 years agoAdd aic and ray modules. I've been building these for a while now on
Warner Losh [Wed, 13 Dec 2000 05:54:26 +0000 (05:54 +0000)]
Add aic and ray modules.  I've been building these for a while now on
i386.

23 years agoModule for aic
Warner Losh [Wed, 13 Dec 2000 05:50:02 +0000 (05:50 +0000)]
Module for aic

Submitted by: Michael Reifenberger

23 years agoAdd module dependencies on CAM module.
Warner Losh [Wed, 13 Dec 2000 05:46:23 +0000 (05:46 +0000)]
Add module dependencies on CAM module.

Submitted by: Michael Reifenberger

23 years agoEliminate a race in MEXTFREE(). The reference counter decrement and test
Bosko Milekic [Wed, 13 Dec 2000 05:13:02 +0000 (05:13 +0000)]
Eliminate a race in MEXTFREE(). The reference counter decrement and test
was not atomic. We now make sure that we free the ext buf if the reference
count is about to reach 0 but also make sure that nobody else has done it
before us.

While I'm here, change refcnt to u_int (from long). This fixes a compiler
warning regarding use of atomic_cmpset_long on i386.

Submitted by: jasone
Reviewed by: jlemon, jake

23 years agoo Tighten restrictions on use of /proc/pid/ctl and move access checks
Robert Watson [Wed, 13 Dec 2000 04:28:24 +0000 (04:28 +0000)]
o Tighten restrictions on use of /proc/pid/ctl and move access checks
  in ctl to using centralized p_can() inter-process access control
  interface.

Reviewed by: sef

23 years agoRemove a redundant prototype.
Mike Smith [Wed, 13 Dec 2000 04:22:04 +0000 (04:22 +0000)]
Remove a redundant prototype.

23 years agoDon't try to free the now-nonexistent hdrspec field. This one snuck by
Mike Smith [Wed, 13 Dec 2000 02:45:03 +0000 (02:45 +0000)]
Don't try to free the now-nonexistent hdrspec field.  This one snuck by
me in the previous round of patches.  Oops.

23 years agoAdd isa support:
Warner Losh [Wed, 13 Dec 2000 01:47:19 +0000 (01:47 +0000)]
Add isa support:
o write isa driver routines.
o factor detach routine in sn_detach.

23 years agoUpdates to match changes elsewhere in the PCI subsystem:
Mike Smith [Wed, 13 Dec 2000 01:28:00 +0000 (01:28 +0000)]
Updates to match changes elsewhere in the PCI subsystem:

 - Remove redundant header-type-specific support in the cardbus pcibus
   clone.  The bridges don't need this anymore.
 - Use pcib_get_bus instead of the deprecated pci_get_secondarybus.
 - Implement read/write ivar support for the pccbb, and teach it how
   to report its secondary bus number.  Save the subsidiary bus number
   as well, although we don't use it yet.

23 years agoNext round of PCI subsystem updates:
Mike Smith [Wed, 13 Dec 2000 01:25:11 +0000 (01:25 +0000)]
Next round of PCI subsystem updates:

 - Break out the /dev/pci driver into a separate file.
 - Kill the COMPAT_OLDPCI support.
 - Make the EISA bridge attach a bit more like the old code; explicitly
   check for the existence of eisa0/isa0 and only attach if they don't
   already exist.  Only make one bus_generic_attach() pass over the
   bridge, once both busses are attached.  Note that the stupid Intel
   bridge's class is entirely unpredictable.
 - Add prototypes and re-layout the core PCI modules in line with
   current coding standards (not a major whitespace change, just moving
   the module data to the top of the file).
 - Remove redundant type-2 bridge support from the core PCI code; the
   PCI-CardBus code does this itself internally.  Remove the now
   entirely redundant header-class-specific support, as well as the
   secondary and subordinate bus number fields.  These are bridge
   attributes now.
 - Add support for PCI Extended Capabilities.
 - Add support for PCI Power Management.  The interface currently
   allows a driver to query and set the power state of a device.
 - Add helper functions to allow drivers to enable/disable busmastering
   and the decoding of I/O and memory ranges.
 - Use PCI_SLOTMAX and PCI_FUNCMAX rather than magic numbers in some
   places.
 - Make the PCI-PCI bridge code a little more paranoid about valid
   I/O and memory decodes.
 - Add some more PCI register definitions for the command and status
   registers.  Correct another bogus definition for type-1 bridges.

23 years agoRemove the COMPAT_OLDPCI option, it's going away.
Mike Smith [Wed, 13 Dec 2000 01:11:34 +0000 (01:11 +0000)]
Remove the COMPAT_OLDPCI option, it's going away.

Turn 'lnc' off in GENERIC for the moment, pending its update to newbus.

23 years agoRemove a couple of leftover unused variables.
Mike Smith [Wed, 13 Dec 2000 01:06:54 +0000 (01:06 +0000)]
Remove a couple of leftover unused variables.

23 years agoUse proper mutex locking when calling setrunnable from speedup_syncer().
Kirk McKusick [Wed, 13 Dec 2000 01:06:53 +0000 (01:06 +0000)]
Use proper mutex locking when calling setrunnable from speedup_syncer().

Submitted by: Tor.Egge@fast.no

23 years agoIf portmap is started as a dependancy, echo a notice during boot.
Brian S. Dean [Wed, 13 Dec 2000 00:43:10 +0000 (00:43 +0000)]
If portmap is started as a dependancy, echo a notice during boot.

Approved by: obrien

23 years agoAdd support for advertising the service we support if the
Julian Elischer [Wed, 13 Dec 2000 00:27:33 +0000 (00:27 +0000)]
Add support for advertising the service we support if the
PADI packet contains a NULL service.  This is apparently the desired
behaviour in this case, though we only allow advertising one
service. You could run multiple pppoeds to advertise multiple services.

23 years ago- Change the allproc_lock to use a macro, ALLPROC_LOCK(how), instead
Jake Burkholder [Wed, 13 Dec 2000 00:17:05 +0000 (00:17 +0000)]
- Change the allproc_lock to use a macro, ALLPROC_LOCK(how), instead
  of explicit calls to lockmgr.  Also provides macros for the flags
  pased to specify shared, exclusive or release which map to the
  lockmgr flags.  This is so that the use of lockmgr can be easily
  replaced with optimized reader-writer locks.
- Add some locking that I missed the first time.

23 years agoAnother mismatch found by Gcc:
Julian Elischer [Tue, 12 Dec 2000 23:56:16 +0000 (23:56 +0000)]
Another mismatch found by Gcc:
This is what happenss when you let the patches pile up too long without
committing them.. brain rot..

23 years agoChange initialiser to match new structure layout.
Julian Elischer [Tue, 12 Dec 2000 23:52:43 +0000 (23:52 +0000)]
Change initialiser to match new structure layout.
forgotten by: Me
Found by: GCC

23 years agoremove unused variable
Julian Elischer [Tue, 12 Dec 2000 23:50:51 +0000 (23:50 +0000)]
remove unused variable

23 years agoIt's possible for an ISA bus to be hung off an EISA bridge, so we need to
Mike Smith [Tue, 12 Dec 2000 23:33:29 +0000 (23:33 +0000)]
It's possible for an ISA bus to be hung off an EISA bridge, so we need to
reflect that here.

23 years agoCatch up to the new kinfo_proc.
John Baldwin [Tue, 12 Dec 2000 23:31:44 +0000 (23:31 +0000)]
Catch up to the new kinfo_proc.

23 years agoCatch up to the new kinfo_proc.
John Baldwin [Tue, 12 Dec 2000 23:21:24 +0000 (23:21 +0000)]
Catch up to the new kinfo_proc.

23 years agoFix bug in parse type for struct ng_one2many_config.
Archie Cobbs [Tue, 12 Dec 2000 23:12:22 +0000 (23:12 +0000)]
Fix bug in parse type for struct ng_one2many_config.

Reported by: Yian Zhu <Yian.Zhu@qobra.com>

23 years agoI always forget this file. It's netgraph, but not one of mine.....
Julian Elischer [Tue, 12 Dec 2000 23:05:19 +0000 (23:05 +0000)]
I always forget this file. It's netgraph, but not one of mine.....

23 years agoTypofix: Configration -> Configuration.
Daniel Harris [Tue, 12 Dec 2000 22:43:02 +0000 (22:43 +0000)]
Typofix: Configration -> Configuration.

Submitted by: Rich Morin <rdm@cfcl.com>

23 years ago I have no idea at all why this file was not included in the last commit.
Julian Elischer [Tue, 12 Dec 2000 22:35:36 +0000 (22:35 +0000)]
 I have no idea at all why this file was not included in the last commit.

23 years agoPoint #includes at compat/linprocfs instead of i386/linux/linprocfs.
Dag-Erling Smørgrav [Tue, 12 Dec 2000 22:06:05 +0000 (22:06 +0000)]
Point #includes at compat/linprocfs instead of i386/linux/linprocfs.

23 years agoRemove after repo-copy to sys/compat/linprocfs.
Dag-Erling Smørgrav [Tue, 12 Dec 2000 22:01:59 +0000 (22:01 +0000)]
Remove after repo-copy to sys/compat/linprocfs.

23 years agoThe linprocfs sources have moved to sys/compat/linprocfs.
Dag-Erling Smørgrav [Tue, 12 Dec 2000 22:00:05 +0000 (22:00 +0000)]
The linprocfs sources have moved to sys/compat/linprocfs.

23 years agoonly include sys/proc.h once
Matt Jacob [Tue, 12 Dec 2000 21:20:48 +0000 (21:20 +0000)]
only include sys/proc.h once

23 years agoInclude sys/proc.h so this compiles [on the Alpha].
David E. O'Brien [Tue, 12 Dec 2000 21:18:13 +0000 (21:18 +0000)]
Include sys/proc.h so this compiles [on the Alpha].

23 years agoWe reference curproc, ergo need <sys/proc.h>
Matt Jacob [Tue, 12 Dec 2000 21:14:29 +0000 (21:14 +0000)]
We reference curproc, ergo need <sys/proc.h>

23 years agoAdd a missing include of <sys/proc.h>.
John Baldwin [Tue, 12 Dec 2000 20:24:36 +0000 (20:24 +0000)]
Add a missing include of <sys/proc.h>.

23 years agocorrected spelling mistakes in comments
Thomas-Henning von Kamptz [Tue, 12 Dec 2000 20:03:17 +0000 (20:03 +0000)]
corrected spelling mistakes in comments
check a couple of mallocs
usage of errx
linebreaks of DBG_ macros,
correcting the usage of nroff macros

Submitted by: grog, charnier
Reviewed by: chm

23 years agoGrrrrr. That last commit was supposed to be to the head, not to -stable
Bill Paul [Tue, 12 Dec 2000 19:31:14 +0000 (19:31 +0000)]
Grrrrr. That last commit was supposed to be to the head, not to -stable
(even though I want the fixes in -stable anyway). I'm sure I'm going
to get flamed now for committing to -stable and -current too quickly.
*sigh*

23 years agooops that commit included a local hack... take it out..
Julian Elischer [Tue, 12 Dec 2000 18:59:09 +0000 (18:59 +0000)]
oops that commit included a local hack... take it out..

23 years agoReviewed by: Archie@freebsd.org
Julian Elischer [Tue, 12 Dec 2000 18:52:14 +0000 (18:52 +0000)]
Reviewed by: Archie@freebsd.org
This clears out my outstanding netgraph changes.
There is a netgraph change of design in the offing and this is to some
extent a superset of soem of the new functionality and some of the old
functionality that may be removed.

This code works as before, but allows some new features that I want to
work with and evaluate. It is the basis for a version of netgraph
with integral locking for SMP use.

This is running on my test machine with no new problems :-)

23 years agoAdd a magic number, for easier identification of boot0. At present,
Robert Nordier [Tue, 12 Dec 2000 17:27:35 +0000 (17:27 +0000)]
Add a magic number, for easier identification of boot0.  At present,
this is used only by libdisk.

23 years agoDo install-time configuration of the i386 boot0 boot manager. At
Robert Nordier [Tue, 12 Dec 2000 17:25:06 +0000 (17:25 +0000)]
Do install-time configuration of the i386 boot0 boot manager.  At
present, this is limited to turning on the packet option if any of
disk slices begin above cylinder 1023.  The effect of this change
should therefore be to automatically enable LBA support, as needed,
when installing FreeBSD.

Something-of-the-kind-requested-by: peter
23 years agoAdd in symbols needed in the WITNESS_ENTER and WITNESS_EXIT macros in
John Baldwin [Tue, 12 Dec 2000 16:40:20 +0000 (16:40 +0000)]
Add in symbols needed in the WITNESS_ENTER and WITNESS_EXIT macros in
i386/include/mutex.h.

23 years agoTemporarily disconnect linprocfs from the build, until I fix it so it builds
Dag-Erling Smørgrav [Tue, 12 Dec 2000 16:03:17 +0000 (16:03 +0000)]
Temporarily disconnect linprocfs from the build, until I fix it so it builds
with Kirk's big proc patch.

23 years agomdoc(7) police: Add a missing .Os.
Ruslan Ermilov [Tue, 12 Dec 2000 15:38:44 +0000 (15:38 +0000)]
mdoc(7) police: Add a missing .Os.

23 years agomdoc(7) police: .Os CMU -> .Os, split authors for better output.
Ruslan Ermilov [Tue, 12 Dec 2000 15:31:21 +0000 (15:31 +0000)]
mdoc(7) police: .Os CMU -> .Os, split authors for better output.

23 years agoRemove hard-sentence breaks, canonify .Dd (no leading zero).
Ruslan Ermilov [Tue, 12 Dec 2000 14:46:55 +0000 (14:46 +0000)]
Remove hard-sentence breaks, canonify .Dd (no leading zero).

23 years agoCatch up with the recent conversion the per-eventhandler list mutex to
Mitsuru IWASAKI [Tue, 12 Dec 2000 14:20:27 +0000 (14:20 +0000)]
Catch up with the recent conversion the per-eventhandler list mutex to
a lockmgr lock.

23 years agoDocument what Idle=* means in a short format display.
Ruslan Ermilov [Tue, 12 Dec 2000 13:37:34 +0000 (13:37 +0000)]
Document what Idle=* means in a short format display.

23 years ago - We have access to our own device_t here, so use pci_read_config
Mike Smith [Tue, 12 Dec 2000 13:20:35 +0000 (13:20 +0000)]
 - We have access to our own device_t here, so use pci_read_config
   rather than finding our parent pcib and using its PCI_READ_CONFIG
   method.

 - Fix the defines for the 32-bit I/O decode registers, and properly
   process the 16-bit versions.  Now we will correctly check that I/O
   resources behind the bridge are going to be decoded.

 - Bring the quirk for the Orion PCI:PCI bridge in here (since it
   seems to want to set the secondary/supplementary bus numbers).

 - Use PCI_SLOTMAX rather than a magic number.

23 years agoDon't try to fix up the Orion here; the interface we use is wrong for the
Mike Smith [Tue, 12 Dec 2000 13:11:42 +0000 (13:11 +0000)]
Don't try to fix up the Orion here; the interface we use is wrong for the
new code (and about to disappear too).

23 years agoStop restore from looping under certain error conditions. This
Ian Dowse [Tue, 12 Dec 2000 12:04:02 +0000 (12:04 +0000)]
Stop restore from looping under certain error conditions. This
corrects cases where restore would spew an infinite stream of
"Changing volumes on pipe input?" messages, or would loop waiting
for a response to the "set owner/mode for '.'" question.

PR: bin/14250
Reviewed by: dwmalone

23 years agomake sure we tear down the devnodes for the endpoints 1 and above
Nick Hibma [Tue, 12 Dec 2000 10:19:19 +0000 (10:19 +0000)]
make sure we tear down the devnodes for the endpoints 1 and above
when switching configuration.

23 years agomdoc(7) police: Now that .Fx macro is parsed, backout
Ruslan Ermilov [Tue, 12 Dec 2000 10:11:12 +0000 (10:11 +0000)]
mdoc(7) police: Now that .Fx macro is parsed, backout
the 1.18 -> 1.20 and fix the .Fx issue the right way.

23 years agoAlways build the ISA and EISA bridge code. This is slightly unintiuitive,
Mike Smith [Tue, 12 Dec 2000 08:23:50 +0000 (08:23 +0000)]
Always build the ISA and EISA bridge code.  This is slightly unintiuitive,
but serves to work around some uncleanliness whereby the ISA bus is not
found on Alpha systems with PCI:EISA bridges due to the lack of EISA code
for the Alpha.

23 years agoChange the proc information returned from the kernel so that it
Kirk McKusick [Tue, 12 Dec 2000 07:25:57 +0000 (07:25 +0000)]
Change the proc information returned from the kernel so that it
no longer contains kernel specific data structures, but rather
only scalar values and structures that are already part of the
kernel/user interface, specifically rusage and rtprio. It no
longer contains proc, session, pcred, ucred, procsig, vmspace,
pstats, mtx, sigiolst, klist, callout, pasleep, or mdproc. If
any of these changed in size, ps, w, fstat, gcore, systat, and
top would all stop working. The new structure has over 200 bytes
of unassigned space for future values to be added, yet is nearly
100 bytes smaller per entry than the structure that it replaced.

23 years agoDon't give up on file write errors. Just log them and continue.
Kirk McKusick [Tue, 12 Dec 2000 06:30:42 +0000 (06:30 +0000)]
Don't give up on file write errors. Just log them and continue.

23 years agoDocument /usr/include/isc.
Archie Cobbs [Tue, 12 Dec 2000 05:19:10 +0000 (05:19 +0000)]
Document /usr/include/isc.

Noticed by: ru

23 years ago- Convert the per-eventhandler list mutex to a lockmgr lock so that it can
John Baldwin [Tue, 12 Dec 2000 04:01:35 +0000 (04:01 +0000)]
- Convert the per-eventhandler list mutex to a lockmgr lock so that it can
  be safely held across an eventhandler function call.
- Fix an instance of the head of an eventhandler list being read without
  the lock being held.
- Break down and use a SYSINIT at the new SI_SUB_EVENTHANDLER to initialize
  the eventhandler global mutex and the eventhandler list of lists rather
  than using a non-MP safe initialization during the first call to
  eventhandler_register().
- Add in a KASSERT() to eventhandler_register() to ensure that we don't try
  to register an eventhandler before things have been initialized.

23 years agoFix the assembly mutex macros to call the appropriate witness functions if
John Baldwin [Tue, 12 Dec 2000 03:49:58 +0000 (03:49 +0000)]
Fix the assembly mutex macros to call the appropriate witness functions if
the witness code is compiled in.  Without this, the witness code doesn't
notice that sched_lock is released by fork_trampoline() and thus gets all
confused about spin lock order later on.

23 years agoSince _mtx_enter() and friends are static inline functions now instead of
John Baldwin [Tue, 12 Dec 2000 03:48:40 +0000 (03:48 +0000)]
Since _mtx_enter() and friends are static inline functions now instead of
macros, the mutex KTR log entries don't actually have the useful filename
and line numbers in the KTR_EXTEND case, so remove a comment claiming this
and go back to one set of KTR strings.

23 years ago - Don't return early from the PCI:EISA bridge attachment, or we will lose
Mike Smith [Tue, 12 Dec 2000 03:33:02 +0000 (03:33 +0000)]
 - Don't return early from the PCI:EISA bridge attachment, or we will lose
   the ISA bus.
 - Don't expect that a PCI:ISA bridge will have a correct class value;
   if we're checking PCI IDs, only depend on these.

This should fix the loss of ISA on machines with PCI:EISA bridges like the
AS4100.

23 years agoenable the proper cascade irq on as1000a
Andrew Gallatin [Tue, 12 Dec 2000 01:39:17 +0000 (01:39 +0000)]
enable the proper cascade irq on as1000a

tested by: wilko

23 years agofix AS1000/AS1000A support. It turns out the that iobus depends on the
Andrew Gallatin [Tue, 12 Dec 2000 01:36:26 +0000 (01:36 +0000)]
fix AS1000/AS1000A support.  It turns out the that iobus depends on the
CPU version (apecs:ev4::cia:ev5) and the irq hardware depends on the systype
previously, only ev4 AS1000s and ev5 AS1000a's would have worked.

tested by: wilko (in its -stable form)
noticed by: daniel