]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoCorrect some old comments.
John Baldwin [Thu, 9 Nov 2000 21:25:58 +0000 (21:25 +0000)]
Correct some old comments.

23 years agoAdd in a few extra documents built by the latest XFree86 3.x port.
John Baldwin [Thu, 9 Nov 2000 21:22:42 +0000 (21:22 +0000)]
Add in a few extra documents built by the latest XFree86 3.x port.

23 years agoAdd more sample code having read the daemnnews article (August 2000)
Julian Elischer [Thu, 9 Nov 2000 18:55:44 +0000 (18:55 +0000)]
Add more sample code having read the daemnnews article (August 2000)

23 years agobackout my previous commit (KAME PR 296). foo != TUNNEL will
Hajimu UMEMOTO [Thu, 9 Nov 2000 17:55:17 +0000 (17:55 +0000)]
backout my previous commit (KAME PR 296).  foo != TUNNEL will
forbid "ANY" SA from being used for tnunel mode.

Reported by: Chris Cason <casonc@netplex.aussie.org>

23 years agoAdd missing delay after card reset.
Pierre Beyssac [Thu, 9 Nov 2000 17:25:49 +0000 (17:25 +0000)]
Add missing delay after card reset.

This fixes randoms lockups when probing the card at boot time, when
more than 1 similar card is found in the machine.

Reviewed by: semenu

23 years agoFix typo in END macro (END'ed enable twice)
Matt Jacob [Thu, 9 Nov 2000 17:01:21 +0000 (17:01 +0000)]
Fix typo in END macro (END'ed enable twice)
PR: 22713
Submitted by: Bernd Walter <ticso@cicely5.cicely.de>

23 years agoFound remaining -kthread option that [sort of] supported using the
David E. O'Brien [Thu, 9 Nov 2000 15:50:22 +0000 (15:50 +0000)]
Found remaining -kthread option that [sort of] supported using the
LinuxThreads port.  Dike it out as it was removed from freebsd.h on
19-July-2000 as this option depended on bits not part of the base system
and required people to install the LinuxThreads port in a manner
non-consistent with the workings of our Ports Collection.

Requested by: jasone

23 years agoFix a reference to ``Standard C'' to refer specifically to the 1990 version;
Garrett Wollman [Thu, 9 Nov 2000 15:42:05 +0000 (15:42 +0000)]
Fix a reference to ``Standard C'' to refer specifically to the 1990 version;
the requirement it describes is not in C99.

23 years agoFix i4b netgraph interface to not kernel panic at boot time
Hellmuth Michaelis [Thu, 9 Nov 2000 12:27:31 +0000 (12:27 +0000)]
Fix i4b netgraph interface to not kernel panic at boot time
Make code compile and work for FreeBSD 4.x as well as FreeBSD 5.x
Submitted by: Michael Reifenberger <root@nihil.plaut.de>

23 years agoImplement a trivial but effective interface for obtaining the kernel's
Mike Smith [Thu, 9 Nov 2000 10:21:23 +0000 (10:21 +0000)]
Implement a trivial but effective interface for obtaining the kernel's
device tree and resource manager contents.  This is the kernel side of
the upcoming libdevinfo, which will expose this information to userspace
applications in a trivial fashion.

Remove the now-obsolete DEVICE_SYSCTLS code.

23 years agoRemove the block/char device distinction. badsect only worked on bdevs,
Adrian Chadd [Thu, 9 Nov 2000 09:03:19 +0000 (09:03 +0000)]
Remove the block/char device distinction. badsect only worked on bdevs,
and then mapped /dev/foo into /dev/rfoo to get to the character device.
This isn't needed anymore.

Reviewed by: ps

23 years agoMake MINSIGSTKSZ machine dependent, and have the sigaltstack
Marcel Moolenaar [Thu, 9 Nov 2000 08:25:48 +0000 (08:25 +0000)]
Make MINSIGSTKSZ machine dependent, and have the sigaltstack
syscall compare against a variable sv_minsigstksz in struct
sysentvec as to properly take the size of the machine- and
ABI dependent struct sigframe into account.

The SVR4 and iBCS2 modules continue to have a minsigstksz of
8192 to preserve behavior. The real values (if different) are
not known at this time. Other ABI modules use the real
values.

The native MINSIGSTKSZ is now defined as follows:

Arch MINSIGSTKSZ
---- -----------
alpha     4096
i386     2048
ia64    12288

Reviewed by: mjacob
Suggested by: bde

23 years agoSync with Alpha:
Marcel Moolenaar [Thu, 9 Nov 2000 07:27:55 +0000 (07:27 +0000)]
Sync with Alpha:
Do not use sysent.c, proto.h and syscall.h in source tree;
use auto-generated versions.

23 years agoMove old aml interpreter code to amldb/aml/. This will be used for our
Mitsuru IWASAKI [Thu, 9 Nov 2000 06:24:45 +0000 (06:24 +0000)]
Move old aml interpreter code to amldb/aml/.  This will be used for our
reference and comparison with ACPICA aml interpreter's behavior.

23 years agoFarewell our code. We will switch acpica code from Intel.
Takanori Watanabe [Thu, 9 Nov 2000 05:09:52 +0000 (05:09 +0000)]
Farewell our code. We will switch acpica code from Intel.
This code has help us comprehence ACPI spec .

Contributors of this code is as follows(except for FreeBSD commiter):
Yasuo Yokoyama,
Munehiro Matsuda,
and ALL acpi-jp@jp.freebsd.org people.

Thanks.

R.I.P.

23 years agoDon't needlessly poll file descriptors when there are no
Daniel Eischen [Thu, 9 Nov 2000 05:08:26 +0000 (05:08 +0000)]
Don't needlessly poll file descriptors when there are no
file descriptors needing to be polled (Doh!).  Reported
by Dan Nelson <dnelson@emsphone.com>.

Don't install and start the scheduling timer until the
first thread is created.  This prevents the overhead of
having a periodic scheduling signal in a single threaded
program.  Reported by Dan Nelson <dnelson@emsphone.com>.

Allow builtin longjmps out of application installed
signal handlers without the need perform any post-handler
cleanup:

  o Change signal handling to save the threads interrupted
    context on the stack.  The threads current context is
    now always stored in the same place (in the pthread).
    If and when a signal handler returns, the interrupted
    context is copied back to the storage area in the pthread.

  o Before calling invoking a signal handler for a thread,
    back the thread out of any internal waiting queues
    (mutex, CV, join, etc) to which it belongs.

Rework uthread_info.c a bit to make it easier to change
the format of a thread dump.

Use an alternal signal stack for the thread library's
signal handler.  This allows us to fiddle with the main
threads stack without fear of it being in use.

Reviewed by: jasone

23 years agoChange .Os FreeBSD 5.0 to .Os FreeBSD.
Jeroen Ruigrok van der Werven [Thu, 9 Nov 2000 05:05:17 +0000 (05:05 +0000)]
Change .Os FreeBSD 5.0 to .Os FreeBSD.

PR: 22684
Submitted by: Akinori YAMADA <yamada-a@nextcom.co.jp>

23 years agoPrevent the thread-safe version of kevent from getting
Daniel Eischen [Thu, 9 Nov 2000 05:00:06 +0000 (05:00 +0000)]
Prevent the thread-safe version of kevent from getting
into an infinite loop when a timeout value is supplied
and the timeout expires.

Reported by: Dan Nelson <dnelson@emsphone.com>
Reviewed by: jasone, jlemon

23 years agoSince the 'when built' phrase is talking about the 'source file' we
Steve Price [Thu, 9 Nov 2000 02:31:02 +0000 (02:31 +0000)]
Since the 'when built' phrase is talking about the 'source file' we
should be using the word 'creates' instead.

23 years agoback out 1.66 -- the Elfxx_Brandinfo compat_3_brand field doesn't exist
Andrew Gallatin [Thu, 9 Nov 2000 01:53:05 +0000 (01:53 +0000)]
back out 1.66 -- the Elfxx_Brandinfo compat_3_brand field doesn't exist
in -current

23 years agoEliminate inconsistency where a value that contains only whitespace
Murray Stokely [Thu, 9 Nov 2000 00:28:22 +0000 (00:28 +0000)]
Eliminate inconsistency where a value that contains only whitespace
confuses the parser.

Approved by: jkh

23 years ago* Reduce diffs to RELENG_4 by specifying a full path to sysctl
Doug Barton [Thu, 9 Nov 2000 00:19:21 +0000 (00:19 +0000)]
* Reduce diffs to RELENG_4 by specifying a full path to sysctl
* Adjust a little whitespace
* Make the distrib-dirs/mtree on DESTDIR conditional on user
  actually specifying a DESTDIR. This seemed like a safe
  way to get the right directories and permissions in the
  installed tree since 'make installworld' does the same
  thing, but in practice too many people have custom hacks
  that we should leave unmolested. Still need to find a way
  to deal with 'install -d' and permissions on nonexistent
  directories in the middle of the path, but this is at
  least no worse that it was before.

PR: bin/22661

23 years agoHaving figured out a bit more of the new-bus puzzle
Julian Elischer [Wed, 8 Nov 2000 21:54:28 +0000 (21:54 +0000)]
Having figured out a bit more of the new-bus puzzle
I have added support for finding non-PNP devices to this
sample loadable ISA driver.
PCI support will come later.

If someone with a clue about newbus were to look it over it would be
really cool.

23 years agoMore paranoia against overflows
Eivind Eklund [Wed, 8 Nov 2000 21:53:05 +0000 (21:53 +0000)]
More paranoia against overflows

23 years agoRearrange the timeouts in the reset code a bit, some ATAPI devices
Søren Schmidt [Wed, 8 Nov 2000 21:25:43 +0000 (21:25 +0000)]
Rearrange the timeouts in the reset code a bit, some ATAPI devices
are picky about this.

23 years agoUpdate usage().
David E. O'Brien [Wed, 8 Nov 2000 20:41:35 +0000 (20:41 +0000)]
Update usage().

Submitted by: nectar

23 years agoHopefully solve the lost slave problem.
Søren Schmidt [Wed, 8 Nov 2000 19:31:39 +0000 (19:31 +0000)]
Hopefully solve the lost slave problem.

23 years agoClean this is up a bit for multiple MIDs... We can figure out which MID
Matt Jacob [Wed, 8 Nov 2000 18:48:21 +0000 (18:48 +0000)]
Clean this is up a bit for multiple MIDs... We can figure out which MID
for an interrupt to enable/disable from the vector (and GID too, if we
had multiple GIDs)- so, stupidly for now, search for the right mcpcia's
softc so we have the right base address for the bridge CSR to apply
IRQ bit-twiddle's to. Alas- this doesn't yet allow us to run, but it's
the right direction.

23 years agoAdd special quirk for this ancient Viper drive. This is all creeping
Matt Jacob [Wed, 8 Nov 2000 18:37:12 +0000 (18:37 +0000)]
Add special quirk for this ancient Viper drive. This is all creeping
driver bloat -I really *have* to do a quirk table that can be loaded
by the loader.

Submitted by: Hellmuth Michaelis <hm@kts.org>

23 years agoFix password clearing bug which prevented challenge/response from working.
Eivind Eklund [Wed, 8 Nov 2000 18:36:56 +0000 (18:36 +0000)]
Fix password clearing bug which prevented challenge/response from working.

Reviewed by: jdp

23 years agoFix CMSG and ALIGN macro usage.
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 16:59:25 +0000 (16:59 +0000)]
Fix CMSG and ALIGN macro usage.
Previously we had to include <machine/param.h> or <sys/param.h> bogusly
due to the fact that <sys/socket.h> CMSG macros needed the ALIGN macro,
which was defined in param.h.  However, including param.h was a disaster
for namespace pollution.
This solution, as contributed by shin a while ago, fixes it elegantly
by wrapping the definitions around some namespace pollution preventer
definitions.
This patch was long overdue.
This should allow any network programmer to use <sys/socket.h> as
before.

PR: 19971, 20530
Submitted by: Martin Kaeske <MartinKaeske@lausitz.net>
Mark Andrews <Mark.Andrews@nominum.com>
Patch submitted by: shin
Reviewed by: bde

23 years agoDocument DISABLE_PSE.
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 15:53:49 +0000 (15:53 +0000)]
Document DISABLE_PSE.

23 years agoadd a couple ESS Technology products(pci device id only).
MIHIRA Sanpei Yoshiro [Wed, 8 Nov 2000 14:43:40 +0000 (14:43 +0000)]
add a couple ESS Technology products(pci device id only).

FreeBSD src/sys/dev/sound/pcm/solo.c
NetBSD syssrc/sys/dev/pci/pcidevs
OpenBSD src/sys/dev/pci/pcidevs

23 years agoadd Texas Instruments TSB12LV22 OHCI IEEE 1394 Host Controller
MIHIRA Sanpei Yoshiro [Wed, 8 Nov 2000 14:04:54 +0000 (14:04 +0000)]
add Texas Instruments TSB12LV22 OHCI IEEE 1394 Host Controller

Obtained from: OpenBSD  src/sys/dev/pci/pcidevs

23 years agoadd Aureal Inc. AU8820/AU8830 Audio controller
MIHIRA Sanpei Yoshiro [Wed, 8 Nov 2000 13:33:23 +0000 (13:33 +0000)]
add Aureal Inc. AU8820/AU8830 Audio controller

Obtained from: OpenBSD src/sys/dev/pci/pcidevs

23 years agoDocument CLUSTERDEBUG, CPU_UPGRADE_HW_CACHE and LOCKF_DEBUG.
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 12:14:06 +0000 (12:14 +0000)]
Document CLUSTERDEBUG, CPU_UPGRADE_HW_CACHE and LOCKF_DEBUG.

23 years agoFix two typo's: teh -> the, ther -> there.
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 12:08:53 +0000 (12:08 +0000)]
Fix two typo's: teh -> the, ther -> there.

23 years agoFix some further english grammar and typo's.
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 12:00:05 +0000 (12:00 +0000)]
Fix some further english grammar and typo's.

23 years agoFix typo's: UPGRADE_CPU_HW_CACHE -> CPU_UPGRADE_HW_CACHE
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 11:58:07 +0000 (11:58 +0000)]
Fix typo's: UPGRADE_CPU_HW_CACHE -> CPU_UPGRADE_HW_CACHE

23 years ago Added PROPERTY_MAX_VALUE and PROPERTY_MAX_NAME defines to libutil.h so
Murray Stokely [Wed, 8 Nov 2000 11:57:03 +0000 (11:57 +0000)]
  Added PROPERTY_MAX_VALUE and PROPERTY_MAX_NAME defines to libutil.h so
that applications know how large of a buffer they must allocate before
calling property_find().  Also added a $FreeBSD$ tag while I'm here.

Approved by: jkh

23 years agoDocument the following options:
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 11:34:09 +0000 (11:34 +0000)]
Document the following options:
FB_DEBUG,
FB_INSTALL_CDEV,
FE_8BIT_SUPPORT,
IBCS2,
KEY,
LOUTB,
SPX_HACK

23 years agoHack to work around a probe which will lock up at least some i450GX-based
Mike Smith [Wed, 8 Nov 2000 11:25:15 +0000 (11:25 +0000)]
Hack to work around a probe which will lock up at least some i450GX-based
systems.

From the PR:

 When 'probe.slot' is PCI_SLOTMAX (== 31) and 'probe.func' is 7,
 call to 'pci_cfgread()' here and machine suddenly hangs up.
 I don't know why...  (or 450GX chipset's bug?)

PR: i386/20379
Submitted by: Masayuki FUKUI <fukui@sonic.nm.fujitsu.co.jp>

23 years agoDocument XBONEHACK option.
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 10:09:01 +0000 (10:09 +0000)]
Document XBONEHACK option.

23 years agoDocument some AHC_* options.
Jeroen Ruigrok van der Werven [Wed, 8 Nov 2000 10:01:45 +0000 (10:01 +0000)]
Document some AHC_* options.

23 years agoAdd commented-out entries for FTP_PROXY and HTTP_PROXY so those behind
Jordan K. Hubbard [Wed, 8 Nov 2000 07:30:27 +0000 (07:30 +0000)]
Add commented-out entries for FTP_PROXY and HTTP_PROXY so those behind
firewalls have examples to work from.

Requested by: chein

23 years agoAdd sn to files to install
Warner Losh [Wed, 8 Nov 2000 06:32:40 +0000 (06:32 +0000)]
Add sn to files to install

23 years agosimple sn man page, since we're missing one right now
Warner Losh [Wed, 8 Nov 2000 06:30:55 +0000 (06:30 +0000)]
simple sn man page, since we're missing one right now

23 years agorestore the freebsd local type-cast fixes that I mistakenly removed
Kenjiro Cho [Wed, 8 Nov 2000 05:45:46 +0000 (05:45 +0000)]
restore the freebsd local type-cast fixes that I mistakenly removed
in the previous commit while attempting to sync with netbsd.

Pointed out by: bde
(never thought of compiling i386's LINT with -64-bit-longs...)

23 years agoDon't build the ACPI CA debugger unless the ACPI_DEBUG option is present.
Mike Smith [Wed, 8 Nov 2000 02:57:01 +0000 (02:57 +0000)]
Don't build the ACPI CA debugger unless the ACPI_DEBUG option is present.

Only build the IA32 support on i386.  Build the IA64 support on IA64.

23 years agoRemove dependence on old acpi driver from acpi userland tools.
Mitsuru IWASAKI [Wed, 8 Nov 2000 02:37:00 +0000 (02:37 +0000)]
Remove dependence on old acpi driver from acpi userland tools.

 - acpiconf Replace include files from old acpi driver to acpica driver.
New sleep type `4b' had been added (S4BIOS) for `-s' option.
Of course this has no effect because driver doesn't
support it for now :-)

 - acpidump All needed structs in sys/dev/acpi/*.h had been merged
into local header file.  No changes on its usage.

23 years agoRemove libipsec and libradius. We're now building without support
Josef Karthauser [Wed, 8 Nov 2000 02:23:26 +0000 (02:23 +0000)]
Remove libipsec and libradius.  We're now building without support
for these by default.

23 years agoFix 'printit()' to ignore some lines it doesn't recognize (most likely
Garance A Drosehn [Wed, 8 Nov 2000 00:52:11 +0000 (00:52 +0000)]
Fix 'printit()' to ignore some lines it doesn't recognize (most likely
coming from lprNG hosts), and print a more helpful error msg for others.

23 years agoIntroduce another global (MPPE_IsServer) so that we initiate the
Brian Somers [Tue, 7 Nov 2000 23:19:11 +0000 (23:19 +0000)]
Introduce another global (MPPE_IsServer) so that we initiate the
MPPE session keys correctly.

I'm a bit dubious about this code.  It seems that the session keys
are initialised differently based on whether you're the client or
the server.  One side is the server if it issues the first challenge,
but of course you can issue a challenge from both sides.... at the
same time.  Sounds like another wonderful M$ assumption...

Ppp can now talk to itself correctly using encryption.

Problem solved by: Ustimenko Semen <semen@iclub.nsu.ru>
Hair torn out by: me

23 years agoFor High security profile, set securelevel to 1 to protect /dev/*mem
Jordan K. Hubbard [Tue, 7 Nov 2000 23:05:14 +0000 (23:05 +0000)]
For High security profile, set securelevel to 1 to protect /dev/*mem
and mounted disks, among other things.

Requested by: kirk

23 years agoRemove the superfluous call to _rtld_error() in symlook_default().
John Polstra [Tue, 7 Nov 2000 22:41:53 +0000 (22:41 +0000)]
Remove the superfluous call to _rtld_error() in symlook_default().
The function's callers generate the error message when appropriate.

This eliminates the message ``Undefined symbol "__register_frame_info"''
which was bogusly returned by dlerror() in some cases.

23 years ago(KTR debugging) Say when we're disabling a vector.
Matt Jacob [Tue, 7 Nov 2000 22:29:21 +0000 (22:29 +0000)]
(KTR debugging) Say when we're disabling a vector.

23 years agoUse size_t rather than a 16-bit data type fo the length.
David E. O'Brien [Tue, 7 Nov 2000 22:19:23 +0000 (22:19 +0000)]
Use size_t rather than a 16-bit data type fo the length.

PR: 9350
Submitted by: Danny J. Zerkel <dzerkel@columbus.rr.com>

23 years agoSync with i386's GENERIC rev 1.287 (Minor ordering changes to make more
David E. O'Brien [Tue, 7 Nov 2000 22:09:33 +0000 (22:09 +0000)]
Sync with i386's GENERIC rev 1.287 (Minor ordering changes to make more
sections strictly alphabetical.)

23 years agoSync with i386/conf/GENERIC rev 1.288 (add comments for some ISA devices).
David E. O'Brien [Tue, 7 Nov 2000 22:07:07 +0000 (22:07 +0000)]
Sync with i386/conf/GENERIC rev 1.288 (add comments for some ISA devices).

23 years agoRemove the -kthread option that matched [sort of] the LinuxThreads port.
David E. O'Brien [Tue, 7 Nov 2000 21:49:08 +0000 (21:49 +0000)]
Remove the -kthread option that matched [sort of] the LinuxThreads port.
This option depended on bits not part of the base system and required
people to install the LinuxThreads port in a manner non-consistent with
the workings of our Ports Collection.
The directions for properly linking with LinuxThreads are given by that
port at install time.

Requested by: jasone

23 years agoMerge from GENERIC 1.288
Warner Losh [Tue, 7 Nov 2000 21:26:34 +0000 (21:26 +0000)]
Merge from GENERIC 1.288

23 years agoMake the ISA nic section look like the other device sections with
Warner Losh [Tue, 7 Nov 2000 21:20:52 +0000 (21:20 +0000)]
Make the ISA nic section look like the other device sections with
comments on the same line like so:
device foo # FooInc Brand NetEther cards

Also, move the wireless NIC cards to their own section.

Add commented out wl driver in wireless section.

Remove obsolete or redundant comments about some of the wireless cards
that used to apply but don't since we've removed 'at foobus'.

There should be no functional changes in this change.

23 years agoSynchronize NEWCARD kernel to GENERIC kernel. It is still experimental,
Warner Losh [Tue, 7 Nov 2000 21:07:35 +0000 (21:07 +0000)]
Synchronize NEWCARD kernel to GENERIC kernel.  It is still experimental,
but it doesn't hurt to keep these things in sync.

23 years agoMinor ordering changes to make more sections strictly alphabetical.
Warner Losh [Tue, 7 Nov 2000 21:06:57 +0000 (21:06 +0000)]
Minor ordering changes to make more sections strictly alphabetical.

23 years agoFixup the /tmp directory permissions to 01777.
Josef Karthauser [Tue, 7 Nov 2000 20:40:21 +0000 (20:40 +0000)]
Fixup the /tmp directory permissions to 01777.

PR:     misc/15875

23 years agoSynced tx(4) driver descriptions + ``device tx'' line moved to the
Semen Ustimenko [Tue, 7 Nov 2000 20:22:19 +0000 (20:22 +0000)]
Synced tx(4) driver descriptions + ``device tx'' line moved to the
list of drivers using miibus.

PR: kern/22556

23 years agoRemove the now unused and unneeded splassert macros and prototypes.
John Baldwin [Tue, 7 Nov 2000 18:51:42 +0000 (18:51 +0000)]
Remove the now unused and unneeded splassert macros and prototypes.

23 years agoProtect against an infinite loop when prefaulting pages. This can
Alfred Perlstein [Tue, 7 Nov 2000 18:31:16 +0000 (18:31 +0000)]
Protect against an infinite loop when prefaulting pages.  This can
happen when the vm system maps past the end of an object or tries
to map a zero length object, the pmap layer misses the fact that
offsets wrap into negative numbers and we get stuck.

Found by: Joost Pol aka Nohican <nohican@marcella.niets.org>
Submitted by: tegge

23 years agoThis file has not been used for long time.
Tatsumi Hosokawa [Tue, 7 Nov 2000 14:14:38 +0000 (14:14 +0000)]
This file has not been used for long time.

23 years agoMoved if_de to mfsroot.flp (we've got 16KB free space in kern.flp)
Tatsumi Hosokawa [Tue, 7 Nov 2000 14:00:04 +0000 (14:00 +0000)]
Moved if_de to mfsroot.flp (we've got 16KB free space in kern.flp)

23 years agoLoad the kernel and root file systems from floppy:/.
Josef Karthauser [Tue, 7 Nov 2000 12:33:46 +0000 (12:33 +0000)]
Load the kernel and root file systems from floppy:/.

23 years agoFix for powering off a HP DJ950C during printing. As stated by Ian:
Nick Hibma [Tue, 7 Nov 2000 10:50:34 +0000 (10:50 +0000)]
Fix for powering off a HP DJ950C during printing. As stated by Ian:

When the printer is turned off the pipe write will cause and error,
which causes lpd to close the device and reopen it to clear the error.
After a short while the device will disappear from the bus but lpd will
have opened the ulpt0 port by then. ulpt_status will check for status
without checking the sc->dying flag and panic the kernel when the device
finally disappears from the bus.

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

23 years agonewbusify the en atm driver.
Kenjiro Cho [Tue, 7 Nov 2000 09:31:28 +0000 (09:31 +0000)]
newbusify the en atm driver.

23 years agonewbusify the en atm driver.
Kenjiro Cho [Tue, 7 Nov 2000 09:30:14 +0000 (09:30 +0000)]
newbusify the en atm driver.

also
 - sync with netbsd
 - fix a bug that miscalculates tx cell counts when the pointer size isn't 4

tested both ENI and Adaptec cards on both i386 and alpha.

23 years agoFixed the security breach I introduced in rev 1.145.
Ruslan Ermilov [Tue, 7 Nov 2000 09:20:32 +0000 (09:20 +0000)]
Fixed the security breach I introduced in rev 1.145.
Disallow getsockopt(IP_FW_ADD) if securelevel >= 3.

PR: 22600

23 years agoAdded Intel Gigabit Ethernet card to device menu.
Tatsumi Hosokawa [Tue, 7 Nov 2000 09:16:14 +0000 (09:16 +0000)]
Added Intel Gigabit Ethernet card to device menu.

23 years agoWhile we're still arguing over debugging output, at least be in a
Jordan K. Hubbard [Tue, 7 Nov 2000 08:47:11 +0000 (08:47 +0000)]
While we're still arguing over debugging output, at least be in a
self-consistent state.

Requested by: bde

23 years agoHandle being given a connect()ed udp descriptor as stdin rather
Brian Somers [Tue, 7 Nov 2000 04:29:46 +0000 (04:29 +0000)]
Handle being given a connect()ed udp descriptor as stdin rather
than assuming it's unconnected.

23 years agoPass the correct output options to the ccp output initialisation
Brian Somers [Tue, 7 Nov 2000 04:29:33 +0000 (04:29 +0000)]
Pass the correct output options to the ccp output initialisation
routine rather than passing it the first requested output option.

Ensure that options are freed correctly even if we don't reach
TLU.

23 years agoDocument the KTR_VERBOSE option.
John Baldwin [Tue, 7 Nov 2000 01:50:10 +0000 (01:50 +0000)]
Document the KTR_VERBOSE option.

23 years ago- Remove much of the inlining of the KTR tracepoints into a ktr_tracepoint()
John Baldwin [Tue, 7 Nov 2000 01:49:48 +0000 (01:49 +0000)]
- Remove much of the inlining of the KTR tracepoints into a ktr_tracepoint()
  function declared in kern_ktr.c.  The only inline checks left are the
  checks that compare KTR_COMPILE with the supplied mask and thus should
  be optimized away into either nothing or a direct call to ktr_tracepoint().
- Move several KTR-related options to opt_ktr.h now that they are only
  needed by kern_ktr.c and not by ktr.h.
- Add in the ktr_verbose functionality if KTR_EXTEND is turned on.  If the
  global variable 'ktr_verbose' is non-zero, then KTR messages will be
  dumped to the console.  This variable can be set by either kernel code
  or via the 'debug.ktr_verbose' sysctl.  It defaults to off unless the
  KTR_VERBOSE kernel option is specified in which case it defaults to on.
  This can be useful when the machine locks up spinning in a loop with
  interrupts disabled as you might be able to see what it is doing when it
  locks up.

Requested by: phk

23 years agoThe vortex driver can be compiled as a KLD now. The EISA attachment
Bill Paul [Tue, 7 Nov 2000 01:03:23 +0000 (01:03 +0000)]
The vortex driver can be compiled as a KLD now. The EISA attachment
is only built on the i386 platform for now, since it depends on symbols
which I'm pretty sure won't be present in the alpha build.

23 years agoThe vx driver no longer needs the PCI compat shims. Also should now
Bill Paul [Tue, 7 Nov 2000 00:58:35 +0000 (00:58 +0000)]
The vx driver no longer needs the PCI compat shims. Also should now
work on the alpha (at least the PCI part should).

23 years agoUpdate the vortex driver so that it no longer needs the PCI compat
Bill Paul [Tue, 7 Nov 2000 00:56:14 +0000 (00:56 +0000)]
Update the vortex driver so that it no longer needs the PCI compat
interface. In addition to using newbus, it also uses bus_space rather
than inb/outb to make it MI. The grody static softc allocation stuff
has been removed as well.

23 years agoMinor nit: missed ithd_loop -> sithd_loop in the KTR tracepoints.
John Baldwin [Tue, 7 Nov 2000 00:45:18 +0000 (00:45 +0000)]
Minor nit: missed ithd_loop -> sithd_loop in the KTR tracepoints.

23 years agoRemove an unneeded #include <machine/bus.h> that snuck in accidentally with
John Baldwin [Tue, 7 Nov 2000 00:44:24 +0000 (00:44 +0000)]
Remove an unneeded #include <machine/bus.h> that snuck in accidentally with
the MI mutexes.

Submitted indirectly by: bde

23 years agoadjust dma channels for vibra16x; recording should now work. full duplex does
Cameron Grant [Tue, 7 Nov 2000 00:38:59 +0000 (00:38 +0000)]
adjust dma channels for vibra16x; recording should now work.  full duplex does
not work on vibra16x, so is disabled.

23 years agoslight mods to isa dma and channel flushing code
Cameron Grant [Tue, 7 Nov 2000 00:32:35 +0000 (00:32 +0000)]
slight mods to isa dma and channel flushing code

23 years agoif a channel dies, rejuvenate it on close()
Cameron Grant [Tue, 7 Nov 2000 00:29:24 +0000 (00:29 +0000)]
if a channel dies, rejuvenate it on close()

23 years agoIgnore resources with a size of 0, as these are disabled (and we don't
Mike Smith [Tue, 7 Nov 2000 00:03:33 +0000 (00:03 +0000)]
Ignore resources with a size of 0, as these are disabled (and we don't
deal with them properly elsewhere).

Submitted by: Masayuki FUKUI <fukui@sonic.nm.fujitsu.co.jp>

23 years agoMFS: add ATA raid support for sysinstall
Jordan K. Hubbard [Mon, 6 Nov 2000 23:15:01 +0000 (23:15 +0000)]
MFS: add ATA raid support for sysinstall

23 years agoRemove unused PCI includes.
Mike Smith [Mon, 6 Nov 2000 22:33:49 +0000 (22:33 +0000)]
Remove unused PCI includes.

23 years agoadd AFMT_8BIT for easier determination of formats
Cameron Grant [Mon, 6 Nov 2000 22:22:52 +0000 (22:22 +0000)]
add AFMT_8BIT for easier determination of formats

23 years agoQuieten some warnings about correct usage of assignments as truth values.
Mike Smith [Mon, 6 Nov 2000 21:32:13 +0000 (21:32 +0000)]
Quieten some warnings about correct usage of assignments as truth values.

23 years agoFix async notifications for listners registered to wildcard nodes. For
Justin T. Gibbs [Mon, 6 Nov 2000 20:12:07 +0000 (20:12 +0000)]
Fix async notifications for listners registered to wildcard nodes.  For
example, a client registered to receive specific events for bus 0, target *,
lun *, was not receiving notifications.

Reviewed by: ken@FreeBSD.org

23 years agoaic7xxx.c:
Justin T. Gibbs [Mon, 6 Nov 2000 20:05:38 +0000 (20:05 +0000)]
aic7xxx.c:
When restarting the sequencer, ensure that the SCBCNT register
is 0.  A non-zero count will prevent the setting of the CCSCBDIR
bit in any future dma operations.  The only time CCSCBCNT would
be non-zero is if we happened to halt the dma during a reset,
but even that should never happen.  Better safe than sorry.

When a command completes before the target responds to an
ATN for a recovery command, we now notify the kernel so that
any recovery operation requeued in the qinfifo can be removed
safely.  In the past, we did this in ahc_done(), but ahc_done()
may be called without the card paused.  This also avoids a
recursive call to ahc_search_qinifo() which could have occurred if
ahc_search_qinififo() happened to be the routine to complete
a recovery action.

Fix 8bit math used for adjusting the qinfifo.  The index must
be wrapped properly within the 256 entry array.  We rely on the
fact that qinfifonext is a uint8_t in most cases to handle
this wrap, but we missed a few spots where the resultant
calculation was promoted to an int.

Change the way that we deal with aborting the first or second
entry from the qinfifo.  We now swap the first entry in the
qinfifo with the "next queued scb" to force the sequencer
to see an abort collision if we ever touch the qinififo while
the sequencer is mid SCB dma.

aic7xxx.reg:
Add new MKMSG_FAILED sequencer interrupt.  This displaced
the BOGUS_TAG interrupt used in some previous sequencer code
debugging.

aic7xxx.seq:
Increment our position in the qinfifo only once the dma
is complete and we have verified that the queue has not
been changed during our DMA.  This simplifies code in the
kernel.

Protect against "instruction creep" when issuing a pausing
sequencer interrupt.  On at least the 7890/91/96/97, the
sequencer will coast after issuing the interrupt for up
to two instructions.  In the past we delt with this by
using carefully placed nops.  Now we call a routine to
issue the interrupt followed by a nop and a ret.

Tell the kernel should an SCB complete with the MK_MESSAGE
flag still set.  This means the target ignored our ATN request.

Clear the channel twice as we exit the data phase.  On the
aic7890/91, the S/G preload logic may require the second
clearing to get the last S/G out of the FIFO.

aic7xxx_freebsd.c:
Don't bother searching the qinfifo for a doubly queued
recovery scb in ahc_done.  This case is handled by the
core driver now.

Free the path used to issue async callbacks after the callback
is complete.

aic7xxx_inline.h:
Split the SCB queue routine into a routine that swaps
the SCB with the "next queued SCB" and a routine that
calls the swapping routine and notifies the card of
the new SCB.  The swapping routine is now also used by
ahc_search_qinfifo.

23 years agoCosmetic change of a structure name.
Garance A Drosehn [Mon, 6 Nov 2000 19:36:38 +0000 (19:36 +0000)]
Cosmetic change of a structure name.
Turn    'struct queue { q_time, q_name }'           (loosely-speaking)
into    'struct jobqueue { job_time, job_cfname }'

Reviewed by: GAWollman

23 years agoGet rid of a minor compile-time warning.
Garance A Drosehn [Mon, 6 Nov 2000 19:19:49 +0000 (19:19 +0000)]
Get rid of a minor compile-time warning.

23 years agoThis commit was generated by cvs2svn to compensate for changes in r68396,
Jeroen Ruigrok van der Werven [Mon, 6 Nov 2000 16:10:57 +0000 (16:10 +0000)]
This commit was generated by cvs2svn to compensate for changes in r68396,
which included commits to RCS files with non-trunk default branches.

23 years agoFix AA being cleared on notify "queries"
Jeroen Ruigrok van der Werven [Mon, 6 Nov 2000 16:10:57 +0000 (16:10 +0000)]
Fix AA being cleared on notify "queries"

Submitted by: Mark Andrews <marka@nominum.com>