]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
25 years agoPatches from DES to create three new kernel config options to control
ken [Fri, 2 Oct 1998 05:15:51 +0000 (05:15 +0000)]
Patches from DES to create three new kernel config options to control
timeouts in the SA driver (timeouts for space, rewind and erase).  Folks
can lengthen the timeouts if their hardware is especially slow, or shorten
them if they want to be notified of errors a little sooner.

Also, get rid of two OD driver options.  The od driver has been made
obsolete by the da driver.

Reviewed by: ken, gibbs
Submitted by: Dag-Erling Coidan Smørgrav <des@FreeBSD.ORG>

25 years agoDefine KLD_MODULE if building a KLD module.
msmith [Fri, 2 Oct 1998 05:08:09 +0000 (05:08 +0000)]
Define KLD_MODULE if building a KLD module.

Forgot to mention that the previous commit was
Submitted by: Doug Rabson <dfr@freebsd.org>

25 years agoChanges to support building of KLD modules. This includes the possibly
msmith [Fri, 2 Oct 1998 04:51:10 +0000 (04:51 +0000)]
Changes to support building of KLD modules.  This includes the possibly
to be removed KMODDEPS define.

25 years agoGC unused stuff.
imp [Fri, 2 Oct 1998 04:37:49 +0000 (04:37 +0000)]
GC unused stuff.

25 years agoA size field returned by the VESA BIOS is already expressed in bytes.
yokota [Fri, 2 Oct 1998 03:42:19 +0000 (03:42 +0000)]
A size field returned by the VESA BIOS is already expressed in bytes.
Submitted by: sos

25 years agoRemove a spurious, but benign statement.
gibbs [Fri, 2 Oct 1998 03:40:53 +0000 (03:40 +0000)]
Remove a spurious, but benign statement.

Correct some panic and printf strings that referenced the 'bt' driver.
I should be more careful when I Cut 'n Paste.

25 years agoadd /usr/share/perl/man
ache [Fri, 2 Oct 1998 01:06:30 +0000 (01:06 +0000)]
add /usr/share/perl/man

25 years agoAdd h2ph call to afterinstall target
ache [Fri, 2 Oct 1998 00:49:43 +0000 (00:49 +0000)]
Add h2ph call to afterinstall target
Error code ignored to allow building without perl installed

25 years agoFix 'chio params' so that it actually prints out the current picker.
ken [Thu, 1 Oct 1998 23:30:14 +0000 (23:30 +0000)]
Fix 'chio params' so that it actually prints out the current picker.
(This was broken when the volume tag changes went in.)

Submitted by: Josef Karthauser <joe@pavilion.net>

25 years agoUse `lpr' vs. `lp' to print
obrien [Thu, 1 Oct 1998 21:18:18 +0000 (21:18 +0000)]
Use `lpr' vs. `lp' to print
Add "ESC-Enter: exit" message to the emacs binding info window.

25 years agoRestore static of sc_flags.
ache [Thu, 1 Oct 1998 21:04:52 +0000 (21:04 +0000)]
Restore static of sc_flags.
Restore set_destructive_cursor prototype.

25 years agoFix a bug in which a page index was used where a byte offset was
jdp [Thu, 1 Oct 1998 20:46:41 +0000 (20:46 +0000)]
Fix a bug in which a page index was used where a byte offset was
expected.  This bug caused builds of Modula-3 to fail in mysterious
ways on SMP kernels.  More precisely, such builds failed on systems
with kern.fast_vfork equal to 0, the default and only supported
value for SMP kernels.

PR: kern/7468
Submitted by: tegge (Tor Egge)

25 years agoDon't try to save FP state if npxproc is null.
jlemon [Thu, 1 Oct 1998 20:45:28 +0000 (20:45 +0000)]
Don't try to save FP state if npxproc is null.
Submitted by: Tor Egge

25 years ago INTERNATIONAL EARTH ROTATION SERVICE (IERS)
phk [Thu, 1 Oct 1998 20:42:03 +0000 (20:42 +0000)]
             INTERNATIONAL EARTH ROTATION SERVICE (IERS)
               SERVICE INTERNATIONAL DE LA ROTATION TERRESTRE

BUREAU CENTRAL DE L'IERS
OBSERVATOIRE DE PARIS
61, Av. de l'Observatoire 75014 PARIS (France)
Tel.      : 33 (0) 1 40 51 22 26
FAX       : 33 (0) 1 40 51 22 91
Internet  : iers@obspm.fr
                                              Paris, 17 July 1998

                                              Bulletin C 16

                                              To authorities responsible for
                                              the measurement and distribution
                                              of time

                                   UTC TIME STEP
                            on the 1st of January 1999

       A positive leap second will be introduced at the end of December 1998.
       The sequence of dates of the UTC second markers will be:

                          1998 December 31,     23h 59m 59s
                          1998 December 31,     23h 59m 60s
                          1999 January   1,      0h  0m  0s

  The difference between UTC and the International Atomic Time TAI is:

  from 1997 July 1,    0h UTC, to 1999 January 1, 0h UTC  : UTC-TAI = - 31s
  from 1999 January 1, 0h UTC, until further notice       : UTC-TAI = - 32s

       Leap seconds can be introduced in UTC at the end of the months of
 December or June, depending on the evolution of UT1-TAI. Bulletin C mailed
 every six months, either to announce a time step in UTC, or to confirm that
 there will be no time step at the next possible date.

                                              Daniel GAMBIS
                                              Director
                                              Central Bureau of IERS

25 years agoBacked out "fix for PR 7575". It was bogus.
dt [Thu, 1 Oct 1998 19:40:27 +0000 (19:40 +0000)]
Backed out "fix for PR 7575". It was bogus.

25 years agoRemove lpt1 - we have userconfig if you have a weird port.
msmith [Thu, 1 Oct 1998 19:35:28 +0000 (19:35 +0000)]
Remove lpt1 - we have userconfig if you have a weird port.
Remove mse0 - the Microsoft Bus Mouse is a dinosaur.  There are probably
              more Pintos on the road than these on peoples' desks.

25 years agoPass the '-n' flag to route(8) when setting up/tearing down the default
msmith [Thu, 1 Oct 1998 19:26:02 +0000 (19:26 +0000)]
Pass the '-n' flag to route(8) when setting up/tearing down the default
route.  If your nameserver config is wrong, this will otherwise hang for
the default resolver timeout (75 seconds), leading people to think that
the system has hung.

25 years agoFix default location of *.ph file install.
markm [Thu, 1 Oct 1998 19:17:58 +0000 (19:17 +0000)]
Fix default location of *.ph file install.

Asked-for-by: ache
25 years agoWhitespace cleanup. Probable patch cut 'n paste.
markm [Thu, 1 Oct 1998 19:16:20 +0000 (19:16 +0000)]
Whitespace cleanup. Probable patch cut 'n paste.

25 years agoRemove the bi_vesa field, as vesa modeswitching is no longer performed
msmith [Thu, 1 Oct 1998 18:25:04 +0000 (18:25 +0000)]
Remove the bi_vesa field, as vesa modeswitching is no longer performed
by the bootblocks.

25 years agoPass the BIOS unit number from which the kernel/loader was obtained in
msmith [Thu, 1 Oct 1998 18:22:52 +0000 (18:22 +0000)]
Pass the BIOS unit number from which the kernel/loader was obtained in
the bootinfo struct so that it can be reused later.

25 years agoPrune unused zalloc components as recommended by Matt Dillon. Extra debugging
msmith [Thu, 1 Oct 1998 17:35:08 +0000 (17:35 +0000)]
Prune unused zalloc components as recommended by Matt Dillon.  Extra debugging
code is still enabled (it's not very expensive).

25 years agoAdd libdisk back to alpha build.
jkh [Thu, 1 Oct 1998 14:49:57 +0000 (14:49 +0000)]
Add libdisk back to alpha build.

25 years agoAdd syscons back to alpha build; works fine now.
jkh [Thu, 1 Oct 1998 14:48:58 +0000 (14:48 +0000)]
Add syscons back to alpha build; works fine now.

25 years agoFix typo.
yokota [Thu, 1 Oct 1998 11:48:38 +0000 (11:48 +0000)]
Fix typo.
PR: kern/8118
Submitted by: Sheldon Hearn

25 years ago- Express various sizes in bytes, rather than Kbytes, in the video
yokota [Thu, 1 Oct 1998 11:40:22 +0000 (11:40 +0000)]
- Express various sizes in bytes, rather than Kbytes, in the video
  mode and adapter information structures.

25 years agoYet another round of fixes for the VESA support code.
yokota [Thu, 1 Oct 1998 11:39:18 +0000 (11:39 +0000)]
Yet another round of fixes for the VESA support code.

- Express various sizes in bytes, rather than Kbytes, in the video
  mode and adapter information structures.
- Fill 0 in the linear buffer size field if the linear frame buffer
  is not available.
- Remove SW_VESA_USER ioctl. It is still experimetal and was not meant
  to be released.
- Fix missing cast operator.
- Correctly handle pointers returned by the VESA BIOS. The pointers
  may point to the area either in the BIOS ROM or in the buffer supplied
  by the caller.
- Set the destructive cursor at the right moment.

25 years agoOverride STRIP so installation doesn't try to strip the loader.
jkh [Thu, 1 Oct 1998 09:57:09 +0000 (09:57 +0000)]
Override STRIP so installation doesn't try to strip the loader.

25 years agoOops, forgot /* */ around Id string
sos [Thu, 1 Oct 1998 09:35:48 +0000 (09:35 +0000)]
Oops, forgot /* */ around Id string

25 years agoUrk!! At some point, CFLAGS got set to -O2 rather than -O (which
jkh [Thu, 1 Oct 1998 09:27:02 +0000 (09:27 +0000)]
Urk!!  At some point, CFLAGS got set to -O2 rather than -O (which
has always been our "recommended optimization flag") and this causes
problems on the Alpha in addition to being generally wrong.
Noticed by: dfr

25 years agoMisc cleanup and probe rework:
imp [Thu, 1 Oct 1998 04:53:55 +0000 (04:53 +0000)]
Misc cleanup and probe rework:

o Use the board id command to find out what kind of board
  we're talking to.  If we're talking to a board that is has
  an ID that is shared between boards supported by the aha
  driver and the bt driver, then use the bt's geometry
  register to weed out the bt cards.  Otherwise assume that we
  support this card.
o Remove bt esetup command sending to the card.  It seems to
  wedge too many cards.
o Revert to doing a soft reset after an invalid command.  This
  change didn't fix anything, so I'm backing it out.  The
  whole issue of card resetting needs to be revisisted at some
  point so that we can do it properly on all hardware.
o GC unused stuff in some places.

25 years agoMake 'make install' do something that might be useful.
peter [Wed, 30 Sep 1998 22:37:47 +0000 (22:37 +0000)]
Make 'make install' do something that might be useful.

25 years agoStop libi386.a from being installed..
peter [Wed, 30 Sep 1998 22:36:45 +0000 (22:36 +0000)]
Stop libi386.a from being installed..

25 years agoAdd /boot, while I happen to be thinking about it.
jkh [Wed, 30 Sep 1998 22:27:27 +0000 (22:27 +0000)]
Add /boot, while I happen to be thinking about it.

25 years agoOne last alpha seat belt - never ask for an MBR boot manager.
jkh [Wed, 30 Sep 1998 21:58:36 +0000 (21:58 +0000)]
One last alpha seat belt - never ask for an MBR boot manager.

25 years agoAdd another check to prevent user from hanging themselves on alpha.
jkh [Wed, 30 Sep 1998 21:55:33 +0000 (21:55 +0000)]
Add another check to prevent user from hanging themselves on alpha.

25 years agoremove keymap.h reference from incorrect location.
jkh [Wed, 30 Sep 1998 21:51:15 +0000 (21:51 +0000)]
remove keymap.h reference from incorrect location.

25 years agoAlways use dedicated disk case on alpha.
jkh [Wed, 30 Sep 1998 21:48:11 +0000 (21:48 +0000)]
Always use dedicated disk case on alpha.

25 years agoEliminate unaligned access on Alpha and also neaten up this code a little.
jkh [Wed, 30 Sep 1998 21:40:51 +0000 (21:40 +0000)]
Eliminate unaligned access on Alpha and also neaten up this code a little.
Submitted by: dfr

25 years agoSmall cleanup.
jkh [Wed, 30 Sep 1998 21:39:02 +0000 (21:39 +0000)]
Small cleanup.

25 years agoHauppauge Tech Support confirmed all Hauppauge 878 PAL/SECAM boards
sos [Wed, 30 Sep 1998 21:06:55 +0000 (21:06 +0000)]
Hauppauge Tech Support confirmed all Hauppauge 878 PAL/SECAM boards
will use PLL mode.  Added to card probe. Thanks to Ken and Fred.

Submitted by:   Roger Hardiman <roger@cs.strath.ac.uk>

25 years agoHauppauge Tech Support confirmed all Hauppauge 878 PAL/SECAM boards
sos [Wed, 30 Sep 1998 20:58:09 +0000 (20:58 +0000)]
Hauppauge Tech Support confirmed all Hauppauge 878 PAL/SECAM boards
will use PLL mode.  Added to card probe. Thanks to Ken and Fred.

Submitted by:   Roger Hardiman <roger@cs.strath.ac.uk>

25 years agoFinally conditionalize the SCSI frobbing code properly. This
jkh [Wed, 30 Sep 1998 20:54:34 +0000 (20:54 +0000)]
Finally conditionalize the SCSI frobbing code properly.  This
should return kernel config saving (under the current a.out system)
to its former functionality.

25 years agoMore rearrangement for USE_SCSI
jkh [Wed, 30 Sep 1998 20:51:01 +0000 (20:51 +0000)]
More rearrangement for USE_SCSI

25 years agoChanged tuner code to autodetect tuner i2c address.
sos [Wed, 30 Sep 1998 20:43:25 +0000 (20:43 +0000)]
Changed tuner code to autodetect tuner i2c address.
Addresses were incorrectly hardcoded.

Submitted by:   Roger Hardiman <roger@cs.strath.ac.uk>

25 years agoMore save_userconfig to a more appropriate spot.
jkh [Wed, 30 Sep 1998 20:33:56 +0000 (20:33 +0000)]
More save_userconfig to a more appropriate spot.

25 years agoMake devfs update the atime timestamp so that 'w' works when using
sos [Wed, 30 Sep 1998 20:33:46 +0000 (20:33 +0000)]
Make devfs update the atime timestamp so that 'w' works when using
options DEVFS.

25 years agoAllow NOSUIDPERL to DTRT.
markm [Wed, 30 Sep 1998 20:24:59 +0000 (20:24 +0000)]
Allow NOSUIDPERL to DTRT.

25 years agoTurn on i386-elf
peter [Wed, 30 Sep 1998 19:48:42 +0000 (19:48 +0000)]
Turn on i386-elf

25 years agoi386 ELF loader startup backend. On an ELF kernel booted with the 3-stage
peter [Wed, 30 Sep 1998 19:48:09 +0000 (19:48 +0000)]
i386 ELF loader startup backend.  On an ELF kernel booted with the 3-stage
bootblocks, the kernel shows up as the primary module:

[3:24am]~-100# kldstat
Id Refs Address  Size     Name
 1    1 0xf0100000 ff00000  /kernel
   ^^^^ oops.. :-)

Based heavily on aout_freebsd.c.  Hmm.. There's so much in common that
these could probably be combined and just check the metadata to see which
format it is.

25 years agoSave booted kernel name. Cosmetic cleanups.
peter [Wed, 30 Sep 1998 19:42:06 +0000 (19:42 +0000)]
Save booted kernel name.  Cosmetic cleanups.

25 years agoFix typos.. The vector for "int 0x12" (get base mem) is not written in
peter [Wed, 30 Sep 1998 19:41:07 +0000 (19:41 +0000)]
Fix typos..  The vector for "int 0x12" (get base mem) is not written in
hex as "0x1a". :-)
Fix a comment about the extended memory checks, that's int 0x15.

25 years agoELF loader, part 1. It works with ELF kernels generated on the i386
peter [Wed, 30 Sep 1998 19:38:26 +0000 (19:38 +0000)]
ELF loader, part 1.  It works with ELF kernels generated on the i386
so far, and should probably be able to be made to work for the alpha
without too much trouble once it's connected up and my assumptions tested.

I think (but have not tested) it will also load "old" ELF kernels that
were not linked with DYNAMIC headers.

The module glue is yet to come. (oh fun.. :-)

It does not explicitly load symbols [yet].  The _DYNAMIC data contains a
runtime symbol set that ddb can use via ddb/db_kld.c.  It'll be missing
some detail that stabs normally provides (eg: number of args to a function,
line numbers, etc).  On the other hand, those minimal symbols will always
be available even on a stripped kernel.

This is mostly stolen from load_aout.c with some ideas from
alpha/libalpha/elf_freebsd.c.

25 years agoMove variable into proper scope.
jkh [Wed, 30 Sep 1998 19:37:46 +0000 (19:37 +0000)]
Move variable into proper scope.

25 years agos/out_loadmodule/aout_loadmodule/ in a printf diagnostic.
peter [Wed, 30 Sep 1998 19:26:23 +0000 (19:26 +0000)]
s/out_loadmodule/aout_loadmodule/ in a printf diagnostic.

25 years agoUncomment prototype for elf_loadmodule
peter [Wed, 30 Sep 1998 19:25:26 +0000 (19:25 +0000)]
Uncomment prototype for elf_loadmodule

25 years agoDebug when an environment variable set, no when it is unset.
dt [Wed, 30 Sep 1998 19:17:51 +0000 (19:17 +0000)]
Debug when an environment variable set, no when it is unset.

25 years agoLimit the fingerd daemon to:
wosch [Wed, 30 Sep 1998 16:12:40 +0000 (16:12 +0000)]
Limit the fingerd daemon to:
        runs only 3 simultaneous fingerd processes and
        limit the connections-per-ip-per-minute to 10.

25 years agoPut guard shells around stuff started from $local_startup. If you type
cracauer [Wed, 30 Sep 1998 14:44:57 +0000 (14:44 +0000)]
Put guard shells around stuff started from $local_startup. If you type
SIGINT (C-c), you'll get control passed to the next script even if
the current one blocks signals. The child is not killed, though.

25 years agoDocument that we support the CS4236B.
des [Wed, 30 Sep 1998 14:09:00 +0000 (14:09 +0000)]
Document that we support the CS4236B.

25 years agoAdd PnP ID for Crystal Semiconductors CS4236B codecs (register-
des [Wed, 30 Sep 1998 14:06:25 +0000 (14:06 +0000)]
Add PnP ID for Crystal Semiconductors CS4236B codecs (register-
compatible with CS4236, so no driver changes are required)

Reviewed by: Luigi Rizzo <luigi@iet.unipi.it>

25 years agoSimplify gunzip usage even more (doh!).
jkh [Wed, 30 Sep 1998 13:36:53 +0000 (13:36 +0000)]
Simplify gunzip usage even more (doh!).

25 years agoSync with sys/i386/i386/machdep.c revision 1.311.
kato [Wed, 30 Sep 1998 13:21:44 +0000 (13:21 +0000)]
Sync with sys/i386/i386/machdep.c revision 1.311.

25 years agoFix english in prompt.
jkh [Wed, 30 Sep 1998 12:33:28 +0000 (12:33 +0000)]
Fix english in prompt.

25 years agoMake the ELF kernel build produce a dynamic executable (!). This enables
peter [Wed, 30 Sep 1998 12:14:39 +0000 (12:14 +0000)]
Make the ELF kernel build produce a dynamic executable (!).  This enables
the in-kernel linker to access the _DYNAMIC data for doing loadable elf
modules.  The alpha kernel is already done this way, I've borrowed some of
the hacks from there.

This is primarily aimed at the 3-stage boot process which is intended to
be able to do pre-loading of kernel modules.

Note that the entry point isn't 0xf0100000 any more, it'll be a little
further on - but this value is stored in the headers.  I don't think this
will be a problem, but I'm sure somebody will tell me if it is. :-)

I'm not sure if btxboot is going to like this, it doesn't do proper ELF
header checking and assumes that there are exactly two program header
entries and that they are both PT_LOAD entries - a bad assumption.

25 years agoAdd missing free()
jkh [Wed, 30 Sep 1998 12:10:16 +0000 (12:10 +0000)]
Add missing free()
Submitted by: Anatoly A. Orehovsky <tolik@mpeks.tomsk.su>

25 years agoAdd missing free().
jkh [Wed, 30 Sep 1998 12:09:12 +0000 (12:09 +0000)]
Add missing free().
Submitted by: Anatoly A. Orehovsky <tolik@mpeks.tomsk.su>

25 years agovarious bits of overdue cleanup.
jkh [Wed, 30 Sep 1998 11:49:37 +0000 (11:49 +0000)]
various bits of overdue cleanup.

25 years agoCall gunzip correctly.
jkh [Wed, 30 Sep 1998 11:44:29 +0000 (11:44 +0000)]
Call gunzip correctly.

25 years agoCorrect source file corruption in last checkin
grog [Wed, 30 Sep 1998 07:53:52 +0000 (07:53 +0000)]
Correct source file corruption in last checkin

Observed by:  jkh

25 years agoRevise test code for sigwait and add test code for sigsuspend.
jb [Wed, 30 Sep 1998 07:14:02 +0000 (07:14 +0000)]
Revise test code for sigwait and add test code for sigsuspend.

Submitted by: Daniel M. Eischen <eischen@vigrid.com>

25 years agoRevise test code for sigwait and add test code for sigsuspend.
jb [Wed, 30 Sep 1998 07:08:09 +0000 (07:08 +0000)]
Revise test code for sigwait and add test code for sigsuspend.

Submitted by: Daniel M. Eischen <eischen@vigrid.com>

25 years agoFix a comment.
jb [Wed, 30 Sep 1998 06:42:29 +0000 (06:42 +0000)]
Fix a comment.

25 years agoCosmetic cleansing. This code requires extra work to keep the garbage
jb [Wed, 30 Sep 1998 06:41:16 +0000 (06:41 +0000)]
Cosmetic cleansing. This code requires extra work to keep the garbage
collector thread running after a fork.

25 years agoMove the cleanup code that frees memory allocated for a dead thread from
jb [Wed, 30 Sep 1998 06:36:56 +0000 (06:36 +0000)]
Move the cleanup code that frees memory allocated for a dead thread from
the thread kernel into a garbage collector thread which is started when
the fisrt thread is created (other than the initial thread). This
removes the window of opportunity where a context switch will cause a
thread that has locked the malloc spinlock, to enter the thread kernel,
find there is a dead thread and try to free memory, therefore trying
to lock the malloc spinlock against itself.

The garbage collector thread acts just like any other thread, so
instead of having a spinlock to control accesses to the dead thread
list, it uses a mutex and a condition variable so that it can happily
wait to be signalled when a thread exists.

25 years agoUse snprintf instead of sprintf to avoid long source file paths from
jb [Wed, 30 Sep 1998 06:29:54 +0000 (06:29 +0000)]
Use snprintf instead of sprintf to avoid long source file paths from
launching an application into space when someone tries to debug it.

The dead thread list now has it's own link pointer, so use that when
reporting the grateful dead.

25 years agoImplementation of an additional state called SIGWAIT (with the previous
jb [Wed, 30 Sep 1998 06:27:31 +0000 (06:27 +0000)]
Implementation of an additional state called SIGWAIT (with the previous
one renamed to SIGSUSPEND) to fix sigwait().

Submitted by: Daniel M. Eischen <eischen@vigrid.com>

25 years agoNULL a pointer after it is freed to avoid trying to free it again.
jb [Wed, 30 Sep 1998 06:24:57 +0000 (06:24 +0000)]
NULL a pointer after it is freed to avoid trying to free it again.

25 years ago- Fix the debug macros.
jb [Wed, 30 Sep 1998 06:22:07 +0000 (06:22 +0000)]
-  Fix the debug macros.
-  Add support of a thread being listed in the dead thread list as well
   as the thread list.
-  Add a new thread state to make sigwait work properly. (Submitted by
   Daniel M. Eischen <eischen@vigrid.com>)
-  Add global variable for the garbage collector mutex and condition
   variable.
-  Delete a couple of prototypes that are no longer required.
-  Add a prototype for the garbage collector thread.

25 years agoDelete the XXX comments that refer to spinlock recursion. The malloc/free/
jb [Wed, 30 Sep 1998 06:13:59 +0000 (06:13 +0000)]
Delete the XXX comments that refer to spinlock recursion. The malloc/free/
realloc functions check for recursion within the malloc code itself. In
a thread-safe library, the single spinlock ensures that no two threads
go inside the protected code at the same time. The thread implementation
is responsible for ensuring that the spinlock does in fact protect malloc.
There was a window of opportunity in which this was not the case. I'll fix
that with a commit RSN.

25 years agoRemove SCSI support as the only driver in this kernel config, aic, is not
gibbs [Wed, 30 Sep 1998 03:48:23 +0000 (03:48 +0000)]
Remove SCSI support as the only driver in this kernel config, aic, is not
currently supported by CAM.

25 years agoFix typo in message.
danny [Wed, 30 Sep 1998 01:46:27 +0000 (01:46 +0000)]
Fix typo in message.

25 years agoAdd several missing ioctl handlers. One needed by Sybase, the others
jfieber [Wed, 30 Sep 1998 01:42:53 +0000 (01:42 +0000)]
Add several missing ioctl handlers.  One needed by Sybase, the others
found while looking for the one.

25 years agoDo not allow a mounted on directory to be rmdir'ed. This removal can
mckusick [Wed, 30 Sep 1998 00:53:40 +0000 (00:53 +0000)]
Do not allow a mounted on directory to be rmdir'ed. This removal can
happen when an NFS exported filesystem tries to remove a locally
mounted on directory.
PR: kern/7272
Submitted by: Andre Albsmeier <andre.albsmeier@mchp.siemens.de>

25 years agoPerform a hard reset on cards when the command fails. This should help
imp [Wed, 30 Sep 1998 00:10:44 +0000 (00:10 +0000)]
Perform a hard reset on cards when the command fails.  This should help
those people that have cards that become wedged when a bogus command is
issued that are too wedged to have a soft reset help.

25 years agoIn nfs_link(), check for a cross-device mount *before* looking
mckusick [Tue, 29 Sep 1998 23:39:37 +0000 (23:39 +0000)]
In nfs_link(), check for a cross-device mount *before* looking
in the v_data field.
Obtained from: Charles Hannum, via Frank van der Linden <frank@wins.uva.nl>

25 years agoMissing vput when cross-device link error is detected in nfs_link.
mckusick [Tue, 29 Sep 1998 23:29:48 +0000 (23:29 +0000)]
Missing vput when cross-device link error is detected in nfs_link.

25 years agoDuring truncation, have to notify the VM about the new size
mckusick [Tue, 29 Sep 1998 23:28:32 +0000 (23:28 +0000)]
During truncation, have to notify the VM about the new size
of the NFS file *before* doing the nfs_vinvalbuf operation.
Otherwise some invalid data may show up in an mmap.

25 years agoDon't require an argument for -v flag
grog [Tue, 29 Sep 1998 23:20:04 +0000 (23:20 +0000)]
Don't require an argument for -v flag
Correct checks for null special file names
Add Usage entry for -v flag
Get terminology straight in man page
Reviewed by: bde

25 years agoFrank sez: 'It fixes a problem with servers that return 0 values
mckusick [Tue, 29 Sep 1998 23:15:53 +0000 (23:15 +0000)]
Frank sez: 'It fixes a problem with servers that return 0 values
for some of the fsinfo RPC fields. It is strictly speaking not
wrong to do this, as the spec says that "it is expected that a
server will make a best effort at supporting all the attributes",
but pretty unusual. You guessed it, it's NT servers that do it.'
Obtained from: Frank van der Linden <frank@wins.uva.nl>

25 years agoDo not need (or want) to take a reference on an NFS file that
mckusick [Tue, 29 Sep 1998 23:15:25 +0000 (23:15 +0000)]
Do not need (or want) to take a reference on an NFS file that
is being deleted due to an forcible unmount. The problem is
that vgone calls vclean() which then calls calls nfs_inactive()
with VXLOCK set on the vnode. Nfs_inactive() was calling vget()
to get a reference on the vnode, which in turn hung on VXLOCK.
Nfs_inactive() now checks v_usecount to make sure that the vnode
is not coming from vclean() before it does a vget().

25 years agoThe code checks each fragment mark to see if it's valid; if the fragment
mckusick [Tue, 29 Sep 1998 22:33:05 +0000 (22:33 +0000)]
The code checks each fragment mark to see if it's valid; if the fragment
is less than NFS_MINPACKET or greater than NFS_MAXPACKET in size, it
barfs and, I think, drops the connection.

However, there's no guarantee that in a multi-fragment RPC, all the
fragments will be at least as large as NFS_MINPACKET.

In fact, with the version of "tclnfs" we have here, which supports NFS
over TCP, at least when built under SunOS 4.1.3 (i.e., with 4.1.3's
user-mode ONC RPC library), I can *repeatably* cause "tclnfs" to send a
request with more than one fragment, one of which is only 8 bytes long.
I just do a 3877-byte write to a file, at an offset of 0.

The check that "slp->ns_reclen" is greater than or equal to
NFS_MINPACKET serves no useful purpose - if the NFS server code can't
handle packets < NFS_MINPACKET bytes, it can't handle them over *any*
protocol, so the check has to be done above the RPC-over-TCP layer - and
should be removed.
Obtained from: Fix from Guy Harris, forwarded by Rick Macklem.

25 years agoIn the bootverbose case, print out error messages for all errors that will
ken [Tue, 29 Sep 1998 22:11:30 +0000 (22:11 +0000)]
In the bootverbose case, print out error messages for all errors that will
not be retried again, even if the SF_NO_PRINT flag is set.

Reviewed by: gibbs

25 years agovm86_datacall: always use workaround since temp. malloced buffer or stack
ache [Tue, 29 Sep 1998 22:06:33 +0000 (22:06 +0000)]
vm86_datacall: always use workaround since temp. malloced buffer or stack
area can be passed (and mapped to page1!) as vesa.c does. Use contigmalloc
now to get proper alignment. Bump max buffer size to PAGE_SIZE

25 years agoAdd back a few useful targets lost in the reshuffle.
jkh [Tue, 29 Sep 1998 22:03:13 +0000 (22:03 +0000)]
Add back a few useful targets lost in the reshuffle.
Reviewed by: jb

25 years agoAdded double quotes around CHMOD description to prevent garbled output.
alex [Tue, 29 Sep 1998 22:02:06 +0000 (22:02 +0000)]
Added double quotes around CHMOD description to prevent garbled output.

PR: 8094
Submitted by: Christoph Weber-Fahr <wefa@callcenter.systemhaus.net>

25 years agoMark directory buffers that have no valid data with B_INVAL
mckusick [Tue, 29 Sep 1998 22:01:10 +0000 (22:01 +0000)]
Mark directory buffers that have no valid data with B_INVAL
so that they are not put in the cache.

25 years agoWhen adding data to a buffer, we need to clear the B_NEEDCOMMIT flag
mckusick [Tue, 29 Sep 1998 21:46:54 +0000 (21:46 +0000)]
When adding data to a buffer, we need to clear the B_NEEDCOMMIT flag
which says that the data is on server but not committed.

25 years agoRestore v1.3 - page align workaround moved to vm86_datacall now
ache [Tue, 29 Sep 1998 20:38:54 +0000 (20:38 +0000)]
Restore v1.3 - page align workaround moved to vm86_datacall now

25 years agoMove workaround about page aligned data buffer directly to vm86_datacall,
ache [Tue, 29 Sep 1998 20:36:31 +0000 (20:36 +0000)]
Move workaround about page aligned data buffer directly to vm86_datacall,
it is impossible to use this func otherwise, i.e. all vesa calls are
potentially broken. Max arg size limited to 1024 for now, bump it, if needed.