]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
25 years agoThe length argument to sysctl is now size_t.
dfr [Tue, 25 Aug 1998 07:52:33 +0000 (07:52 +0000)]
The length argument to sysctl is now size_t.

25 years agoChange length arguments to sysctl to size_t.
dfr [Tue, 25 Aug 1998 07:38:19 +0000 (07:38 +0000)]
Change length arguments to sysctl to size_t.

25 years agoNew major number for SR4 streams emulation.
msmith [Mon, 24 Aug 1998 22:42:17 +0000 (22:42 +0000)]
New major number for SR4 streams emulation.

25 years agoGot two volatile sig_atomic_t and int mixed up. Spotted by Gary Palmer.
cracauer [Mon, 24 Aug 1998 19:15:48 +0000 (19:15 +0000)]
Got two volatile sig_atomic_t and int mixed up. Spotted by Gary Palmer.

25 years agoLast commit managed to get mangled somehow.
phk [Mon, 24 Aug 1998 18:23:18 +0000 (18:23 +0000)]
Last commit managed to get mangled somehow.

25 years ago- #define mask of enabled interrupts/indications in if_xlreg.h instead of
wpaul [Mon, 24 Aug 1998 17:51:38 +0000 (17:51 +0000)]
- #define mask of enabled interrupts/indications in if_xlreg.h instead of
  constructing local copy in xl_init()
- disable interrupts on entry to xl_intr(), re-enable them on exit.
- fix a few typos in some comments

25 years agoRemove the last remaining evidence of B_TAPE.
phk [Mon, 24 Aug 1998 17:47:25 +0000 (17:47 +0000)]
Remove the last remaining evidence of B_TAPE.
Reclaim 3 unused bits in b_flags

25 years agoRemove the last remaining evidence of B_TAP.
phk [Mon, 24 Aug 1998 17:47:08 +0000 (17:47 +0000)]
Remove the last remaining evidence of B_TAP.
Reclaim 3 unused bits in b_flags

25 years agoOnly claim PPS_HARDPPSONASSERT if we're compiled with PPS_SYNC
phk [Mon, 24 Aug 1998 16:31:27 +0000 (16:31 +0000)]
Only claim PPS_HARDPPSONASSERT if we're compiled with PPS_SYNC

25 years agoCheck the text segment size of the executable and the process. If
wosch [Mon, 24 Aug 1998 16:25:30 +0000 (16:25 +0000)]
Check the text segment size of the executable and the process. If
not equal, the command line arguments are wrong. E.g.:

$./gcore /bin/sh 1761

$ ./gcore /usr/tmp/chroot/bin/sh 1761
gcore: The executable /usr/tmp/chroot/bin/sh does not belong to process 1761!
Text segment size (in bytes): executable 303104, process 294912

25 years agoDo not exit on SIGINT in non-interactive shells, fixes PR 1206,
cracauer [Mon, 24 Aug 1998 10:20:37 +0000 (10:20 +0000)]
Do not exit on SIGINT in non-interactive shells, fixes PR 1206,
i.e. this makes emacs usable from system(3). Programs called from
shellscripts are now required to exit with proper signal status. That
means, they have to kill themself. Exiting with faked numerical exit
code is not sufficient.

Exit with proper signal status if script exits on signal.

Make the wait builtin interruptable, both with and without traps set.

Use volatile sig_atomic_t where (and only where) appropriate.

(Almost) fix printing of newlines on SIGINT.

Make traps setable from trap handlers. This is needed for shellscripts
that catch SIGINT for cleanup work but intend to exit on it, hance
have to kill themself from a trap handler. I.e. mkdep.

While I'm at it, make it -Wall clean. -Wall is not enabled in
Makefile, since vararg warnx() macro calls in usr.bin/printf/printf.c
are not -Wall-able.
PR: 1206
Obtained from: Basic SIGINT fix from Bruce Evans

25 years agoWhen exiting on SIGINT, exit with signal status
cracauer [Mon, 24 Aug 1998 10:17:20 +0000 (10:17 +0000)]
When exiting on SIGINT, exit with signal status

25 years agoAdd atomic.s.
dfr [Mon, 24 Aug 1998 08:48:22 +0000 (08:48 +0000)]
Add atomic.s.

25 years agoRemove a useless ifdef which made struct disklabel have a different
dfr [Mon, 24 Aug 1998 08:45:37 +0000 (08:45 +0000)]
Remove a useless ifdef which made struct disklabel have a different
size in kernel and user builds on the alpha and prevented various ioctls
from working.

25 years agoChange various syscalls to use size_t arguments instead of u_int.
dfr [Mon, 24 Aug 1998 08:39:39 +0000 (08:39 +0000)]
Change various syscalls to use size_t arguments instead of u_int.

Add some overflow checks to read/write (from bde).

Change all modifications to vm_page::flags, vm_page::busy, vm_object::flags
and vm_object::paging_in_progress to use operations which are not
interruptable.

Reviewed by: Bruce Evans <bde@zeta.org.au>

25 years agoRegnerate.
dfr [Mon, 24 Aug 1998 08:32:19 +0000 (08:32 +0000)]
Regnerate.

25 years agoFix a few syscall arguments to use size_t instead of u_int.
dfr [Mon, 24 Aug 1998 08:29:52 +0000 (08:29 +0000)]
Fix a few syscall arguments to use size_t instead of u_int.

25 years agoAdd partial KLD support for ELF. The module loading is not written yet.
dfr [Mon, 24 Aug 1998 08:25:26 +0000 (08:25 +0000)]
Add partial KLD support for ELF.  The module loading is not written yet.

25 years agoRe-implement tcp and ip fragment reassembly to not store pointers in the
dfr [Mon, 24 Aug 1998 07:47:39 +0000 (07:47 +0000)]
Re-implement tcp and ip fragment reassembly to not store pointers in the
ip header which can't work on alpha since pointers are too big.

Reviewed by: Garrett Wollman <wollman@khavrinen.lcs.mit.edu>

25 years agoDon't try to run ldconfig if it doesn't exist. Useful for elf platforms.
gpalmer [Mon, 24 Aug 1998 07:11:48 +0000 (07:11 +0000)]
Don't try to run ldconfig if it doesn't exist. Useful for elf platforms.

25 years agoUse a relative path to bsd.kern.mk if the relative mk directory exists.
bde [Mon, 24 Aug 1998 06:42:26 +0000 (06:42 +0000)]
Use a relative path to bsd.kern.mk if the relative mk directory exists.

25 years agoSupport printf format checking of FreeBSD kernel formats %[Dbrz].
bde [Mon, 24 Aug 1998 06:33:52 +0000 (06:33 +0000)]
Support printf format checking of FreeBSD kernel formats %[Dbrz].
This is enabled by the undocumented option -fformat-extensions.
This option should be named better and/or give more control over
the extensions.

Fixed a message - don't warn about the field width when it's the
precision that has the wrong type.  Didn't fix excessive checking
for the precision relative to the type - ANSI requires both to be
ints, but gcc permits the field width to be either int or unsigned
int.

25 years agoRemoved support for the gcc-2.[0-4] keywords __dead and __pure. They
bde [Mon, 24 Aug 1998 06:17:01 +0000 (06:17 +0000)]
Removed support for the gcc-2.[0-4] keywords __dead and __pure.  They
stopped being used in /usr/src almost 2 years ago.

Don't pretend to support gcc-[3-infinity].

Added __printf0like().  Feature tests for the __printf0__ feature
are problematic, so this can't be used for the err() family yet
- it's only in recent versions of FreeBSD's hacked version of gcc.

Added comments about __unused and __*like().

25 years agoShut this thing up, like the other LKMs
jkh [Mon, 24 Aug 1998 05:31:57 +0000 (05:31 +0000)]
Shut this thing up, like the other LKMs

25 years agoFixed syntax errors (__dead should have gone away years ago, and
bde [Mon, 24 Aug 1998 02:54:33 +0000 (02:54 +0000)]
Fixed syntax errors (__dead should have gone away years ago, and
__attribute__(()) is a gccism - we use __dead2).

25 years agoFixed printf format errors. Only one left in LINT on i386's.
bde [Mon, 24 Aug 1998 02:28:16 +0000 (02:28 +0000)]
Fixed printf format errors.  Only one left in LINT on i386's.

25 years agoDescribe what constitues an exceptional condition.
steve [Mon, 24 Aug 1998 01:09:34 +0000 (01:09 +0000)]
Describe what constitues an exceptional condition.

PR: 7666
Submitted by: Archie Cobbs <archie@whistle.com>

25 years agoThe host commandline option is -h and not -s.
steve [Mon, 24 Aug 1998 00:56:20 +0000 (00:56 +0000)]
The host commandline option is -h and not -s.

PR: 7703
Submitted by: Yoshishige Arai <ryo2@on.rim.or.jp>

25 years ago* Check getopt()'s return with -1, not EOF
obrien [Sun, 23 Aug 1998 22:52:09 +0000 (22:52 +0000)]
* Check getopt()'s return with -1, not EOF
* protect a few potential buffer overflows

Obtained from: NetBSD

25 years agomore files we don't need.
obrien [Sun, 23 Aug 1998 22:51:11 +0000 (22:51 +0000)]
more files we don't need.

25 years agoTwo more files FreeBSD doesn't need.
obrien [Sun, 23 Aug 1998 22:43:27 +0000 (22:43 +0000)]
Two more files FreeBSD doesn't need.

25 years agoupgrade instructions.
obrien [Sun, 23 Aug 1998 22:09:19 +0000 (22:09 +0000)]
upgrade instructions.

25 years agoThis commit was generated by cvs2svn to compensate for changes in r38494,
obrien [Sun, 23 Aug 1998 22:07:21 +0000 (22:07 +0000)]
This commit was generated by cvs2svn to compensate for changes in r38494,
which included commits to RCS files with non-trunk default branches.

25 years agoVirgin import of AMD (am-utils) v6.0a16
obrien [Sun, 23 Aug 1998 22:07:21 +0000 (22:07 +0000)]
Virgin import of AMD (am-utils) v6.0a16

25 years agoremove bdevsw arg from dsopen();
phk [Sun, 23 Aug 1998 20:16:35 +0000 (20:16 +0000)]
remove bdevsw arg from dsopen();

Forgotten by: julian
Reviewed by: bde

25 years agoUse address space numbers to reduce TLB flushes.
dfr [Sun, 23 Aug 1998 16:05:55 +0000 (16:05 +0000)]
Use address space numbers to reduce TLB flushes.

25 years agoQuick hack to allow netstat to compile once more: ifdef out MT_SOOPTS
gpalmer [Sun, 23 Aug 1998 15:23:11 +0000 (15:23 +0000)]
Quick hack to allow netstat to compile once more: ifdef out MT_SOOPTS

25 years agoDon't check minor number of dump device at all.
des [Sun, 23 Aug 1998 14:18:08 +0000 (14:18 +0000)]
Don't check minor number of dump device at all.

Discussed-with: Jörg Wunsch

25 years agoEnabled Lite2 fix for reading from dead ttys.
bde [Sun, 23 Aug 1998 11:43:29 +0000 (11:43 +0000)]
Enabled Lite2 fix for reading from dead ttys.

25 years agoFixed printf format errors.
bde [Sun, 23 Aug 1998 10:16:26 +0000 (10:16 +0000)]
Fixed printf format errors.

25 years agoFixed printf format errors. `struct eisa_device' uses a strange type
bde [Sun, 23 Aug 1998 10:12:12 +0000 (10:12 +0000)]
Fixed printf format errors.  `struct eisa_device' uses a strange type
for the unit number (like most SCSI drivers).

25 years agoFixed printf format errors.
bde [Sun, 23 Aug 1998 09:57:09 +0000 (09:57 +0000)]
Fixed printf format errors.

Fixed style bugs in the int to u_long ioctl command arg changes.  This
driver doesn't use the u_foo BSDism.

Don't smash ioctl command args back to int.

25 years agoAdded D_TTY to the cdevswitch flags for all tty drivers. This is required
bde [Sun, 23 Aug 1998 08:26:42 +0000 (08:26 +0000)]
Added D_TTY to the cdevswitch flags for all tty drivers.  This is required
for the Lite2 fix for always returning EIO in dead_read().

Cleaned up the cdevswitch initializers for all tty drivers.

Removed explicit calls to ttsetwater() from all (tty) drivers.  ttsetwater()
is now called centrally for opens, not just for parameter changes.

25 years agoRemoved cxpoll(). Just use ttpoll() like all other tty drivers. It's
bde [Sun, 23 Aug 1998 07:52:48 +0000 (07:52 +0000)]
Removed cxpoll().  Just use ttpoll() like all other tty drivers.  It's
equivalent except for the bad-minor-number case (which probably can't
happen).

25 years agoAttempt to fix my breakage of the alpha makebootarea() in rev.1.19.
bde [Sun, 23 Aug 1998 07:32:37 +0000 (07:32 +0000)]
Attempt to fix my breakage of the alpha makebootarea() in rev.1.19.
The previous attempt just converted compile time breakage to runtime
breakage.

25 years agoYow! Completely change the way socket options are handled, eliminating
wollman [Sun, 23 Aug 1998 03:07:17 +0000 (03:07 +0000)]
Yow!  Completely change the way socket options are handled, eliminating
another specialized mbuf type in the process.  Also clean up some
of the cruft surrounding IPFW, multicast routing, RSVP, and other
ill-explored corners.

25 years agoFirst cut (aka hack) at getting make release building on the alpha. Don't
gpalmer [Sun, 23 Aug 1998 00:52:18 +0000 (00:52 +0000)]
First cut (aka hack) at getting make release building on the alpha. Don't
build floppies or kernels. The i386 release process should be unaffected.

25 years agoPass me the pointy hat with the extra sequins. Just a moment, while I get
mckay [Sat, 22 Aug 1998 15:51:41 +0000 (15:51 +0000)]
Pass me the pointy hat with the extra sequins.  Just a moment, while I get
it to sit right...

The __error() hack gave out the wrong address.  It returned the address of
errno in ld.so instead of the address of errno in the main program.  Oops.

The hack is now correct, just in time to be obsoleted by elf.

25 years agoCorrect/clarify some comments.
mckay [Sat, 22 Aug 1998 15:24:09 +0000 (15:24 +0000)]
Correct/clarify some comments.

25 years agoNuke unsupported architecture files
gpalmer [Sat, 22 Aug 1998 14:34:59 +0000 (14:34 +0000)]
Nuke unsupported architecture files

25 years agoBuilding `compat' on alpha is meaningless (they are i386 libraries)
gpalmer [Sat, 22 Aug 1998 14:31:39 +0000 (14:31 +0000)]
Building `compat' on alpha is meaningless (they are i386 libraries)

25 years agoRemove bogus unused code.
dfr [Sat, 22 Aug 1998 10:32:38 +0000 (10:32 +0000)]
Remove bogus unused code.

25 years agoMinor tweaks to track a couple of i386 changes and to make it compile.
dfr [Sat, 22 Aug 1998 10:31:01 +0000 (10:31 +0000)]
Minor tweaks to track a couple of i386 changes and to make it compile.

25 years agoTry to make this compile on both alpha and i386
gpalmer [Fri, 21 Aug 1998 23:44:16 +0000 (23:44 +0000)]
Try to make this compile on both alpha and i386

25 years agoAdd BREAK_TO_DEBUGGER so that config stops whining
gpalmer [Fri, 21 Aug 1998 23:43:15 +0000 (23:43 +0000)]
Add BREAK_TO_DEBUGGER so that config stops whining

25 years agoConsolidate the two malloc()s that are done when allocating
brian [Fri, 21 Aug 1998 18:10:15 +0000 (18:10 +0000)]
Consolidate the two malloc()s that are done when allocating
an mbuf.

25 years agoMake ``show mem'' display the number of packets allocated as well as the
brian [Fri, 21 Aug 1998 18:09:57 +0000 (18:09 +0000)]
Make ``show mem'' display the number of packets allocated as well as the
number of octets.

25 years agoFix ``lp=port@machine'' syntax and mention it in printcap(5).
brian [Fri, 21 Aug 1998 18:08:46 +0000 (18:08 +0000)]
Fix ``lp=port@machine'' syntax and mention it in printcap(5).

25 years agoIncrease the number of descriptors (and, as a consequence, the number
wpaul [Fri, 21 Aug 1998 16:58:48 +0000 (16:58 +0000)]
Increase the number of descriptors (and, as a consequence, the number
of associated mbuf clusters) in the RX ring from 4 to 16. On my
really fast PI 400Mhz test machines, 4 descriptors (and associated
mbuf clusters) is enough to achieve decent performance without any
RX overruns. However, one person reported problems with the following
scenario:

- P90 system running FreeBSD with a 3c905B-TX adapter, slow IDE hard
  disk (Quantum Bigfoot?)
- PII 266 with SCSI disks running LoseNT and also with a 3c905B-TX
- Both machines connected together via crossover cable at 100Mbps
  full-duplex
- LoseNT machine writing largs amounts of data (2.5 GB work of
  files each in the neighborhood of 1 to 2 MB in size) via samba to
  the FreeBSD machine

In this case, the LoseNT machine is sending data very fast. Apparently
there weren't any problems initially because the user was writing to
one particular disk which was relatively fast, however after this disk
filled up and the user started writing to the second slower disk, RX
overruns would occur and sometimes the RX DMA engine would stall after
a 100 to 500MB had been transfered. The xl_rxeof() handler is supposed
to detect this condition and restart the upload engine; I'm not sure
why it doesn't, unless interrupts are being lost and the rx handler
isn't getting called.

This is still an improvement over the Linux driver, which uses 32
descriptors in its receive ring. :)

Problem reported by: Heiko Schaefer <hschaefer@fto.de>

25 years agoAdd "-P" display the id as a password file entry.
obrien [Fri, 21 Aug 1998 06:47:58 +0000 (06:47 +0000)]
Add "-P" display the id as a password file entry.

Submitted by: terry

25 years agoUpdate this header to use the revamped elf headers which select Elf32
jb [Fri, 21 Aug 1998 03:29:40 +0000 (03:29 +0000)]
Update this header to use the revamped elf headers which select Elf32
or Elf64 based on the inclusion of the machine dependent header.

I've left the addition of the extra fields to handle the relocation
structures with addend for a separate commit after jdp has had a chance
to review what I've done. The current change is needed to compile
csu/alpha/crt1.c

25 years agoThis commit was generated by cvs2svn to compensate for changes in r38465,
msmith [Fri, 21 Aug 1998 03:17:42 +0000 (03:17 +0000)]
This commit was generated by cvs2svn to compensate for changes in r38465,
which included commits to RCS files with non-trunk default branches.

25 years agoThis is the new unified bootstrap, sometimes known previously as the
msmith [Fri, 21 Aug 1998 03:17:42 +0000 (03:17 +0000)]
This is the new unified bootstrap, sometimes known previously as the
'three-stage' bootstrap.
There are a number of caveats with the code in its current state:
 - The i386 bootstrap only supports booting from a floppy.
 - The kernel and kld do not yet know how to deal with the extended
   information and module summary passed in.
 - PnP-based autodetection and demand loading of modules is not implemented.
 - i386 ELF kernel loading is not ready yet.
 - The i386 bootstrap is loaded via an ugly blockmap.

On the alpha, both net- and disk-booting (SRM console machines only) is
supported.  No blockmaps are used by this code.

Obtained from: Parts from the NetBSD/i386 standalone bootstrap.

25 years agoThis is the new unified bootstrap, sometimes known previously as the
msmith [Fri, 21 Aug 1998 03:17:42 +0000 (03:17 +0000)]
This is the new unified bootstrap, sometimes known previously as the
'three-stage' bootstrap.
There are a number of caveats with the code in its current state:
 - The i386 bootstrap only supports booting from a floppy.
 - The kernel and kld do not yet know how to deal with the extended
   information and module summary passed in.
 - PnP-based autodetection and demand loading of modules is not implemented.
 - i386 ELF kernel loading is not ready yet.
 - The i386 bootstrap is loaded via an ugly blockmap.

On the alpha, both net- and disk-booting (SRM console machines only) is
supported.  No blockmaps are used by this code.

Obtained from: Parts from the NetBSD/i386 standalone bootstrap.

25 years agoAdd extern declaration for `tzname[]' to make this header comply better
jkoshy [Fri, 21 Aug 1998 01:21:26 +0000 (01:21 +0000)]
Add extern declaration for `tzname[]' to make this header comply better
to POSIX90.

2-rounds-of-review-by: bde

25 years agoBuild kvm_mkdb and natd on alpha too.
jb [Thu, 20 Aug 1998 21:51:35 +0000 (21:51 +0000)]
Build kvm_mkdb and natd on alpha too.

25 years agoMake the build tools static to avoid trying to use the shared loader
jb [Thu, 20 Aug 1998 21:45:46 +0000 (21:45 +0000)]
Make the build tools static to avoid trying to use the shared loader
before it is installed.

This upsets Bruce because the host boostrap build forces tools to be
static anyway. He says I'm abusing NOTOOLS in src/Makefile by using
it to do a aout->elf transition build. One day I'll find a place to
install host tools like these to allow a true cross build.

25 years agoChange the aux field type to long instead of int so that it is large
jb [Thu, 20 Aug 1998 21:39:28 +0000 (21:39 +0000)]
Change the aux field type to long instead of int so that it is large
enough to hold a man-sized pointer.

25 years agoRemove the bootstrap hack that prevented the use of the rtld.
jb [Thu, 20 Aug 1998 21:37:22 +0000 (21:37 +0000)]
Remove the bootstrap hack that prevented the use of the rtld.

25 years agoUse %p and void * casts to print pointers.
jb [Thu, 20 Aug 1998 21:33:31 +0000 (21:33 +0000)]
Use %p and void * casts to print pointers.
Cast the hash value returned to long instead of int.

25 years agoFix the machine dependent SUBDIR I removed by mistake in the last commit.
jb [Thu, 20 Aug 1998 20:47:49 +0000 (20:47 +0000)]
Fix the machine dependent SUBDIR I removed by mistake in the last commit.

25 years agoFix small printf() bogon (forgot newline, and the message was longer that
wpaul [Thu, 20 Aug 1998 14:48:58 +0000 (14:48 +0000)]
Fix small printf() bogon (forgot newline, and the message was longer that
80 cols).

25 years agoFix my blunder in my previous commit here.
phk [Thu, 20 Aug 1998 11:22:52 +0000 (11:22 +0000)]
Fix my blunder in my previous commit here.

25 years agoAdd support for TurboChannel alphas (DEC 3000/300 and 3000/500).
dfr [Thu, 20 Aug 1998 08:27:11 +0000 (08:27 +0000)]
Add support for TurboChannel alphas (DEC 3000/300 and 3000/500).

Obtained from: NetBSD
Submitted by: Andrew Gallatin <gallatin@cs.duke.edu>

25 years agoAdd libstand.
msmith [Thu, 20 Aug 1998 08:24:07 +0000 (08:24 +0000)]
Add libstand.

25 years agoInstall libstand header <stand.h> and manpage (libstand.3)
msmith [Thu, 20 Aug 1998 08:23:12 +0000 (08:23 +0000)]
Install libstand header <stand.h> and manpage (libstand.3)

25 years agoThis is libstand; a support library for standalone executables (eg. bootstrap
msmith [Thu, 20 Aug 1998 08:19:55 +0000 (08:19 +0000)]
This is libstand; a support library for standalone executables (eg. bootstrap
modules).
Obtained from: NetBSD, with some architectural changes and many additions.

25 years agoThis commit was generated by cvs2svn to compensate for changes in r38451,
msmith [Thu, 20 Aug 1998 08:19:55 +0000 (08:19 +0000)]
This commit was generated by cvs2svn to compensate for changes in r38451,
which included commits to RCS files with non-trunk default branches.

25 years agoThis is libstand; a support library for standalone executables (eg. bootstrap
msmith [Thu, 20 Aug 1998 08:19:55 +0000 (08:19 +0000)]
This is libstand; a support library for standalone executables (eg. bootstrap
modules).
Obtained from: NetBSD, with some architectural changes and many additions.

25 years agoMoved `nx' functions to the one place where they are used (su.c).
bde [Thu, 20 Aug 1998 06:10:42 +0000 (06:10 +0000)]
Moved `nx' functions to the one place where they are used (su.c).
They shouldn't be used there either.  They should have gone away
about 3 years ago when the statically initialized devswitches went
away, but su.c unfortunately still frobs the cdevswitch in the old
way.

25 years agoA better fix for kern/7144:
msmith [Thu, 20 Aug 1998 05:49:59 +0000 (05:49 +0000)]
A better fix for kern/7144:

The check for dropping unicast packets not sent to our ethernet
address is after the bpf tap, but not conditioned on it.  All packets
received should get handed to bpf, and unicast packets not to us (mac)
should get dropped whether or not there is a bpf listener.  I believe
that the common optimization that the interface is in hw promisc mode
iff there is a bpf listener is in general wrong, but more frequently
so on wavelans.

I think Max's fix makes bpf listeners not see unicast packets sent to
others, but I'm not sure.

One can argue that checking on MOD_ENAL is wrong, but the code only
drops packets that shouldn't be received.  The correctness condition
is that it be run whenever unicast packets without our mac address can
be received.

PR: kern/7144
Submitted by: Greg Troxel <gdt@ir.bbn.com>

25 years agoFixed bogus spelling `nx*' of some `no*' devswitch functions.
bde [Thu, 20 Aug 1998 05:48:17 +0000 (05:48 +0000)]
Fixed bogus spelling `nx*' of some `no*' devswitch functions.

25 years agoFixed devfs initialization which I broke in the previous commit.
bde [Thu, 20 Aug 1998 05:21:50 +0000 (05:21 +0000)]
Fixed devfs initialization which I broke in the previous commit.

Fixed an old name and disorder in the sio dictionary.

25 years agoFixed devfs initialization which I broke in the previous commit.
bde [Thu, 20 Aug 1998 05:12:48 +0000 (05:12 +0000)]
Fixed devfs initialization which I broke in the previous commit.

25 years agoInclude opt_devfs.h which defines SLICE, to make previous commit
des [Wed, 19 Aug 1998 20:20:52 +0000 (20:20 +0000)]
Include opt_devfs.h which defines SLICE, to make previous commit
meaningful.

Pointed out by: Luoqi Chen

25 years agoMake two changes:
wpaul [Wed, 19 Aug 1998 15:07:46 +0000 (15:07 +0000)]
Make two changes:

If I'm reading the manual correctly, the 3c905B actually loses its
PCI configuration during the transition from D3(hot) back to D0, not
during the transition from D0 to D3(hot). This means it should be possible
to save the existing PCI settings, restet the power state, then restore
the PCI settings afterwards. Changed xl_attach() to attempt this first
thing before the normal PCI setup. I'm not certain this will work correctly,
but it shouldn't hurt.

If xl_init() is called while an autoneg session is in progress, the
autoneg timeout and chip state will get clobbered. Try to avoid this
by checking sc->xl_autoneg at the start of xl_init() and defer
the initialization until later if it's set. (xl_init() is always called
at the end of an autoneg session by xl_autoneg_mii().)

Problem pointed out by: Larry Baird <lab@gta.com>

25 years agoAdjust family tree to list some freebsd dates as well.
jkh [Wed, 19 Aug 1998 12:58:38 +0000 (12:58 +0000)]
Adjust family tree to list some freebsd dates as well.

25 years agoSync with sys/i386/isa/sio.c revision 1.211.
kato [Wed, 19 Aug 1998 11:48:38 +0000 (11:48 +0000)]
Sync with sys/i386/isa/sio.c revision 1.211.

25 years agoMake struct buf->b_offset reflect the real byte offset which got
sos [Wed, 19 Aug 1998 10:50:32 +0000 (10:50 +0000)]
Make struct buf->b_offset reflect the real byte offset which got
in via the uio struct. This enables device drivers to use != DEV_BSIZE
blocking on devices with wierd sector/block sizes (ie CDROM's).

25 years agoSync with sys/i386/isa/syscons.c revision 1.275.
kato [Wed, 19 Aug 1998 09:35:18 +0000 (09:35 +0000)]
Sync with sys/i386/isa/syscons.c revision 1.275.

25 years agoSync with sys/i386/i386/machdep.c revision 1.304.
kato [Wed, 19 Aug 1998 09:32:15 +0000 (09:32 +0000)]
Sync with sys/i386/i386/machdep.c revision 1.304.

25 years agoEnabled dynamically sized tty input buffers (with enough buffering
bde [Wed, 19 Aug 1998 04:17:38 +0000 (04:17 +0000)]
Enabled dynamically sized tty input buffers (with enough buffering
for 1 second's worth of input) and larger tty output buffers.  The
interrupt-level buffers are still too small for speeds above 115200
bps (only a little too small for 230400 bps if RTS flow control is
enabled).

Don't call ttsetwater() explicitly in open().  It is now called for
the TTYDISC l_open() and should be static.

Don't attempt to register the cdevsw more than once.

25 years agoA limit of 200000 for the output buffer high watermark was excessive,
bde [Wed, 19 Aug 1998 04:01:00 +0000 (04:01 +0000)]
A limit of 200000 for the output buffer high watermark was excessive,
since (hardware) ttys have too low a bandwidth to benefit significantly
from large buffers.  Use twice the old limit for the new-default case
and 8 times the old limit for the driver-specifies-watermark case.
Nothing uses these cases yet.

Removed related debugging code.

25 years agoCorrect handling of ASCII DEL (0x7F).
jkoshy [Wed, 19 Aug 1998 03:39:40 +0000 (03:39 +0000)]
Correct handling of ASCII DEL (0x7F).

I don't have access to a real VT220 to verify this against.
However, I'm committing the patch in `good faith' because
(a) getting hold of a real VT220 is going to be increasingly difficult
    the longer the PR sits around,
(b) some one was troubled enough to in a PR and
(c) the fix is minor and has no other implications.

PR: 7559
Submitted by: Christian Weisgerber <naddy@mips.rhein-neckar.de>

25 years agoCorrect glyph displayed by PCVT in ISO Latin-1 mode.
jkoshy [Wed, 19 Aug 1998 03:07:16 +0000 (03:07 +0000)]
Correct glyph displayed by PCVT in ISO Latin-1 mode.

PR: 7610
Submitted-by: Christian Weisgerber <naddy@mips.rhein-neckar.de>
25 years agoDelete dpt driver.
kato [Wed, 19 Aug 1998 02:41:51 +0000 (02:41 +0000)]
Delete dpt driver.

25 years agoAdd "pcm.4" to Makefile.
jkoshy [Wed, 19 Aug 1998 01:54:13 +0000 (01:54 +0000)]
Add "pcm.4" to Makefile.

PR: 7636
Submitted by: <Janick.Taillandier@ratp.fr>

25 years agoIncreased column widths for tty input watermark fields so that watermarks
bde [Wed, 19 Aug 1998 01:32:28 +0000 (01:32 +0000)]
Increased column widths for tty input watermark fields so that watermarks
for 11520-byte buffers for 115200 bps are displayed properly.

Fixed my recent printf format error fixes.  %p is almost unusable
in tables, since its width and format are unknown/machine-dependent.
Use %8lx and cast pointers to (u_long)(void*).  This is still quite
broken, e.g., for machines with 64-bit pointers.

25 years agoChange casts from int to u_long and formats from %x to %lx.
jb [Tue, 18 Aug 1998 13:41:38 +0000 (13:41 +0000)]
Change casts from int to u_long and formats from %x to %lx.

Didn't fix the alignment of the output fields on alpha where addresses
require 16 characters to print.

Added a dummy field to the pt_u union to help the alpha compiler align
the u_sa field in a suiable way.

25 years agoChange a variable from int to size_t.
jb [Tue, 18 Aug 1998 13:36:47 +0000 (13:36 +0000)]
Change a variable from int to size_t.