]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
26 years agoELF versions of __warn_references() and __weak_reference(). Note that this
Peter Wemm [Thu, 18 Jun 1998 18:35:04 +0000 (18:35 +0000)]
ELF versions of __warn_references() and __weak_reference().  Note that this
doesn't work with libc/net/res_stubs.c since gas wants the weak reference
to take place in the same context as the real definition, presumably so
that it can get hold of the symbol typing etc.
However, this doesn't matter for libc/elf since we don't have binaries
that use the old symbol names.

26 years agoConverted isa_strayintr() to take a `void *' arg.
Bruce Evans [Thu, 18 Jun 1998 16:08:46 +0000 (16:08 +0000)]
Converted isa_strayintr() to take a `void *' arg.

26 years agoChanged the type of an isa/general interrupt handler to take a
Bruce Evans [Thu, 18 Jun 1998 15:32:09 +0000 (15:32 +0000)]
Changed the type of an isa/general interrupt handler to take a
`void *' arg.  Fixed or hid most of the resulting type mismatches.
Handlers can now be updated locally (except for reworking their
global declarations in isa_device.h).

26 years agoCreate a save TMPDIR using `mktemp -d'.
Wolfram Schneider [Thu, 18 Jun 1998 09:26:22 +0000 (09:26 +0000)]
Create a save TMPDIR using `mktemp -d'.

26 years agopreserve errno across the readlink() calls. There is no value in trashing
Peter Wemm [Thu, 18 Jun 1998 09:13:16 +0000 (09:13 +0000)]
preserve errno across the readlink() calls.  There is no value in trashing
errno during a successful malloc() call.

26 years agoMention the ``allow users'' command when describing
Brian Somers [Thu, 18 Jun 1998 01:24:29 +0000 (01:24 +0000)]
Mention the ``allow users'' command when describing
how to set things up for incoming connections.

26 years agoDon't allow a SIGCHLD to wake up a thread if the process has the default
John Birrell [Wed, 17 Jun 1998 22:29:12 +0000 (22:29 +0000)]
Don't allow a SIGCHLD to wake up a thread if the process has the default
signal handler installed for SIGCHLD. The ACE MT_SOCK_Test was hanging
as the result of being interrupted when it didn't expect to be.

26 years agoRenambed sbintr to sb_intr to avoid a conflict with the (now
Bruce Evans [Wed, 17 Jun 1998 16:59:30 +0000 (16:59 +0000)]
Renambed sbintr to sb_intr to avoid a conflict with the (now
unconditionally declared although it can't be configured together
with snd) extern function sbintr.

26 years agoDon't depend on <sys/conf.h> declaring isa device structs or isa
Bruce Evans [Wed, 17 Jun 1998 16:54:22 +0000 (16:54 +0000)]
Don't depend on <sys/conf.h> declaring isa device structs or isa
interrupt handlers.  pcm is not in LINT, so it is hard to test.

26 years agoDelete redundant declaration of lptintr.
KATO Takenori [Wed, 17 Jun 1998 16:33:14 +0000 (16:33 +0000)]
Delete redundant declaration of lptintr.

26 years agoDeclare bsintr when PC98 is defined.
KATO Takenori [Wed, 17 Jun 1998 16:32:23 +0000 (16:32 +0000)]
Declare bsintr when PC98 is defined.

26 years agoSync with sys/i386/conf/Makefile.i386 revision 1.114.
KATO Takenori [Wed, 17 Jun 1998 16:30:53 +0000 (16:30 +0000)]
Sync with sys/i386/conf/Makefile.i386 revision 1.114.

26 years agoSync with sys/i386/i386/machdep.c revision 1.300.
KATO Takenori [Wed, 17 Jun 1998 16:30:16 +0000 (16:30 +0000)]
Sync with sys/i386/i386/machdep.c revision 1.300.

26 years agoSync with sys/i386/isa/if_ed.c revision 1.142.
KATO Takenori [Wed, 17 Jun 1998 16:29:27 +0000 (16:29 +0000)]
Sync with sys/i386/isa/if_ed.c revision 1.142.

26 years agoThe isa device configuration hook was especially useless here.
Bruce Evans [Wed, 17 Jun 1998 15:55:22 +0000 (15:55 +0000)]
The isa device configuration hook was especially useless here.

26 years agoDon't generate declarations for isa device structs in "ioconf.h".
Bruce Evans [Wed, 17 Jun 1998 15:16:53 +0000 (15:16 +0000)]
Don't generate declarations for isa device structs in "ioconf.h".

Don't generate declarations for isa interrupt handlers at all.
Isa interrupt handlers are now declared in <i386/isa/isa_device.h>
but should be converted take a `void *' arg and staticized as
soon as possible.

Updated CONFIGVERS.  New configs are very incompatible with
previous versions.

26 years agoDon't declare isa device structs or isa interrupt handlers in <sys/conf>,
Bruce Evans [Wed, 17 Jun 1998 14:58:04 +0000 (14:58 +0000)]
Don't declare isa device structs or isa interrupt handlers in <sys/conf>,
and don't depend on them being declared there.  This will cause lots of
warnings for a few minutes until config is updated.  Interrupt handlers
should never have been configured by config, and the machine generated
declarations get in the way of changing the arg type from int to void *.

26 years agoAdded used include of "ioconf.h" - don't depend on pollution in
Bruce Evans [Wed, 17 Jun 1998 14:13:15 +0000 (14:13 +0000)]
Added used include of "ioconf.h" - don't depend on pollution in
<sys/conf.h>.  I'm fixing isa interrupt handler configuration and
this is just a quick fix to keep SCSI configuration unharmed.

26 years agoInclude <i386/isa/pnp.h> after including its prerequisite
Bruce Evans [Wed, 17 Jun 1998 14:01:00 +0000 (14:01 +0000)]
Include <i386/isa/pnp.h> after including its prerequisite
<i386/isa/isa_device.h> - don't depend on pollution in <sys/conf.h>
including the prerequisite earlier.

26 years agoAdded used include of <i386/isa/isa_device.h> - don't depend on
Bruce Evans [Wed, 17 Jun 1998 13:54:59 +0000 (13:54 +0000)]
Added used include of <i386/isa/isa_device.h> - don't depend on
pollution in <sys/conf.h>

26 years agoRemove reference to non-existent htable(8).
Joseph Koshy [Wed, 17 Jun 1998 13:13:41 +0000 (13:13 +0000)]
Remove reference to non-existent htable(8).

PR: 6940

26 years agoRemove compile time dependency on ARG_MAX.
Joseph Koshy [Wed, 17 Jun 1998 12:58:43 +0000 (12:58 +0000)]
Remove compile time dependency on ARG_MAX.

This fix only removes the dependency on compile time constants.  The code
has other (old) problems that need to be addressed.

PR: 1791
Reviewed-by: bde, tegge
26 years agoFixed a misdeclaration. This unhides type mismatches which will be
Bruce Evans [Wed, 17 Jun 1998 12:14:55 +0000 (12:14 +0000)]
Fixed a misdeclaration.  This unhides type mismatches which will be
fixed soon.

26 years agoRemove compile time dependency on ARG_MAX.
Joseph Koshy [Wed, 17 Jun 1998 11:33:10 +0000 (11:33 +0000)]
Remove compile time dependency on ARG_MAX.

PR: 1791 (partial)
Reviewed by: Bruce Evans <bde@freebsd.org>, Tor Egge <tegge@freebsd.org>

26 years agoAdded yacc to the bootstrap tools. It is needed very early to
Bruce Evans [Wed, 17 Jun 1998 09:34:42 +0000 (09:34 +0000)]
Added yacc to the bootstrap tools.  It is needed very early to
build at least compile_et and lex, and although almost any version
of yacc could work, the version in -stable doesn't actually work
with -current makefiles because it doesn't support -o.

Submitted by: Ian Holland <ianh@tortuga.com.au>

26 years agoSync with sys/i386/isa/sio.c revision 1.207.
KATO Takenori [Wed, 17 Jun 1998 09:27:15 +0000 (09:27 +0000)]
Sync with sys/i386/isa/sio.c revision 1.207.

26 years agoSync with sys/i386/isa/conf/files.i386 revision 1.198.
KATO Takenori [Wed, 17 Jun 1998 09:26:23 +0000 (09:26 +0000)]
Sync with sys/i386/isa/conf/files.i386 revision 1.198.

26 years agoAdd mention of Intel EtherExpress 16 cards in the supported list.
Joseph Koshy [Wed, 17 Jun 1998 08:33:10 +0000 (08:33 +0000)]
Add mention of Intel EtherExpress 16 cards in the supported list.

Noticed by: Bob Van Valzah <Bob@WhiteBarn.com> on the -doc list.

26 years agoIf a thread is waiting on a child process to complete, the SIGCHLD
John Birrell [Wed, 17 Jun 1998 03:53:16 +0000 (03:53 +0000)]
If a thread is waiting on a child process to complete, the SIGCHLD
signal can arrive before the thread is woken from it's wait4. In this
case, don't return an EINTR, just set the thread state to running and
the wait4 wrapper will loop and get the exit status of the process.

26 years agoo Allow ``set mrru'' or ``set mrru 0'' to disable
Brian Somers [Tue, 16 Jun 1998 23:23:57 +0000 (23:23 +0000)]
o Allow ``set mrru'' or ``set mrru 0'' to disable
  multi-link mode.
o Fix a typo in the ``set mrru'' description.

26 years agoChange some log levels. ALERTs are only logged when
Brian Somers [Tue, 16 Jun 1998 19:40:42 +0000 (19:40 +0000)]
Change some log levels.  ALERTs are only logged when
something that can't happen happens or when everyone
needs to know.  ERRORs are only logged when something
unexpected happens.

26 years agoDon't bring the modem offline or hangup when ``down lcp''
Brian Somers [Tue, 16 Jun 1998 19:40:28 +0000 (19:40 +0000)]
Don't bring the modem offline or hangup when ``down lcp''
is done.  Instead, behave like ``close lcp''.

26 years agoZero mbuf::pnext when Dequeue()ing.
Brian Somers [Tue, 16 Jun 1998 19:40:20 +0000 (19:40 +0000)]
Zero mbuf::pnext when Dequeue()ing.

26 years agoBacked out rev.1.183, which had nothing to do with its log message.
Bruce Evans [Tue, 16 Jun 1998 14:55:27 +0000 (14:55 +0000)]
Backed out rev.1.183, which had nothing to do with its log message.
It was to support a half-baked optimization of certain long long
divisions in gcc-2.8 and/or egcs.  We now avoid these divisions.

26 years agoUse copyout() instead of bcopy() to copy the image to user space.
Bruce Evans [Tue, 16 Jun 1998 14:36:40 +0000 (14:36 +0000)]
Use copyout() instead of bcopy() to copy the image to user space.
bcopy() caused panics under heavy paging (not quite as suspected -
the kernel stack seemed to get corrupted).

Fixed long lines.

Reviewed by: phk

26 years agoDon't log "unexpected" events on never-opened devices. Events left
Bruce Evans [Tue, 16 Jun 1998 11:05:59 +0000 (11:05 +0000)]
Don't log "unexpected" events on never-opened devices.  Events left
over from the probe are now expected for incompatible UARTs that
deliver IRQs as a strobe (low) instead of a level (high).

Discard events on going-away devices too.  Endless loops may have
been possible when an active pccard was removed.

26 years agoOnly set the datalink::stayonline flag if we
Brian Somers [Tue, 16 Jun 1998 07:15:16 +0000 (07:15 +0000)]
Only set the datalink::stayonline flag if we
close/down with CLOSE_LCP when our state is
DATALINK_READY or higher.

26 years agoo Don't dump core with ~m in term mode.
Brian Somers [Tue, 16 Jun 1998 07:15:11 +0000 (07:15 +0000)]
o Don't dump core with ~m in term mode.
o Always put a '\r' before a '\n' at the end of a line
  in prompt_vPrintf() in term mode, and make prompt_Printf()
  use prompt_vPrintf().
o Fix ~? message.

26 years agoo De-staticise things that don't need to be static.
Brian Somers [Mon, 15 Jun 1998 19:06:58 +0000 (19:06 +0000)]
o De-staticise things that don't need to be static.
o Bring the static ``ttystate'' into struct prompt so that
  the tilde context is per prompt and not global.
o Comment the remaining static variables so that it's
  clear why they're static.
o Add some XXX comments suggesting that our interface list
  and our hostname should be re-generated after a signal
  (say SIGUSR1) so that a machine with PCCARDs has a chance.

26 years agoFix a rather nasty use of `static'. This caused a SEGV
Brian Somers [Mon, 15 Jun 1998 19:06:25 +0000 (19:06 +0000)]
Fix a rather nasty use of `static'.  This caused a SEGV
when running ``link * load label'' as we ended up recursing
back into command_Interpret after nuking our command arg list.

26 years agoGive ``load'' optional context. It's now possible to
Brian Somers [Mon, 15 Jun 1998 19:05:51 +0000 (19:05 +0000)]
Give ``load'' optional context.  It's now possible to
``link 1,2,3 load label'' for people that want to set
up their links in a more mpd-like manner.

26 years agoMake `close lcp' just close the LCP layer and not hangup. This is
Brian Somers [Mon, 15 Jun 1998 19:05:27 +0000 (19:05 +0000)]
Make `close lcp' just close the LCP layer and not hangup.  This is
useful for slirp users that wish to get their shell back after the
ppp session.  `close' with no args still hangs up as expected.
Required by: jmz

26 years agoUpdated entry for Xircom CreditCard Modem
Nate Williams [Mon, 15 Jun 1998 17:51:03 +0000 (17:51 +0000)]
Updated entry for Xircom CreditCard Modem

Submitted by: "Mike Del" <repenting@hotmail.com>

26 years agoOops, the previous commit missed one line of code.
Joerg Wunsch [Mon, 15 Jun 1998 15:43:13 +0000 (15:43 +0000)]
Oops, the previous commit missed one line of code.

26 years agoWatch out for null hostnames in netgroup entries, to avoid dumping core.
Joerg Wunsch [Mon, 15 Jun 1998 15:41:41 +0000 (15:41 +0000)]
Watch out for null hostnames in netgroup entries, to avoid dumping core.
This happens if someone tries to export to a netgroup like:

mygroup (,,mynisdomain)

26 years agoSync with sys/i386/isa/syscons.c revision 1.263.
KATO Takenori [Mon, 15 Jun 1998 13:06:33 +0000 (13:06 +0000)]
Sync with sys/i386/isa/syscons.c revision 1.263.

26 years agoCorrect use of .Nm. Use .Bl/.El for enumerating options. Use .An. Correct
Philippe Charnier [Mon, 15 Jun 1998 07:12:20 +0000 (07:12 +0000)]
Correct use of .Nm. Use .Bl/.El for enumerating options. Use .An. Correct
formatting of rcsid. Remove unused #includes. Do not use memory after
freeing it.

26 years agoCorrect use of .Nm. Add rcsid. Remove unused #includes. Use err(3).
Philippe Charnier [Mon, 15 Jun 1998 07:07:21 +0000 (07:07 +0000)]
Correct use of .Nm. Add rcsid. Remove unused #includes. Use err(3).

26 years agoCapitalize at the start of sentence. Add rcsid. Remove unused #includes.
Philippe Charnier [Mon, 15 Jun 1998 07:03:47 +0000 (07:03 +0000)]
Capitalize at the start of sentence. Add rcsid. Remove unused #includes.
Use `dumpon' instead of argv[0].

26 years agoAdd rcsid. Remove unused #includes.
Philippe Charnier [Mon, 15 Jun 1998 07:00:01 +0000 (07:00 +0000)]
Add rcsid. Remove unused #includes.

26 years agoCorrect use of .Nm. Add rcsid. Remove unused #includes.
Philippe Charnier [Mon, 15 Jun 1998 06:58:13 +0000 (06:58 +0000)]
Correct use of .Nm. Add rcsid. Remove unused #includes.

26 years agoOops, uoff_t can't be used here yet.
Bruce Evans [Mon, 15 Jun 1998 04:50:55 +0000 (04:50 +0000)]
Oops, uoff_t can't be used here yet.

26 years agofix another typo
Julian Elischer [Mon, 15 Jun 1998 00:35:47 +0000 (00:35 +0000)]
fix another typo

26 years agoOops
Julian Elischer [Sun, 14 Jun 1998 23:53:43 +0000 (23:53 +0000)]
Oops
left a "break;" out of the last patch
it complains for every loopback packet..

26 years agoMake it compatible with long usernames
Andrey A. Chernov [Sun, 14 Jun 1998 22:56:31 +0000 (22:56 +0000)]
Make it compatible with long usernames

26 years agoTry narrow down the culprit sending undefined packet types through the loopback
Julian Elischer [Sun, 14 Jun 1998 20:58:17 +0000 (20:58 +0000)]
Try narrow down the culprit sending undefined packet types through the loopback

26 years agoReviewed by: Amancio
Amancio Hasty [Sun, 14 Jun 1998 20:05:27 +0000 (20:05 +0000)]
Reviewed by: Amancio
Submitted by: Randall Hopper <rhh@ct.picker.com>
    The patch supports using the X10 Mouse Remote in both stand-alone and
pass-through configurations, so you can plug your mouse and remote into the
same serial port, use the mouse for X, and use the remote for other apps
like Fxtv. For instance, we can now control fxtv via the remote control
just like a TV : change channels, mute, increase volume, zoom video,
freeze frame 8)

     The mouse events are channeled through the syscons/sysmouse I/F like
normal, and the remote buttons are "syphoned off" to a UNIX-domain stream
socket (defined as _PATH_MOUSEREMOTE in <machine/mouse.h>) for a
remote-aware app to grab and use.

For further info on the X10 Mouse Remote see:
http://www.x10.com/products/x10_mk19a.htm

26 years agoSlight change to directory cleanup
Julian Elischer [Sun, 14 Jun 1998 19:31:28 +0000 (19:31 +0000)]
Slight change to directory cleanup
Makes soft updates a bit cleaner. Eliminates some warnings about
'corrupted directories' from fsck.

26 years agoHorrible hack to allow newfs to work again in SLICE kernels.
Julian Elischer [Sun, 14 Jun 1998 19:00:12 +0000 (19:00 +0000)]
Horrible hack to allow newfs to work again in SLICE kernels.
Real fix in the making but requires editing 60+ files. (every damn driver)

26 years agoOops, wrong file again.
Doug Rabson [Sun, 14 Jun 1998 17:35:55 +0000 (17:35 +0000)]
Oops, wrong file again.

26 years agoAdd support for the Storage Dimensions SDC3222F.
Justin T. Gibbs [Sun, 14 Jun 1998 16:41:57 +0000 (16:41 +0000)]
Add support for the Storage Dimensions SDC3222F.
PR: 6261
Submitted by: "Matthew N. Dodd" <winter@jurai.net>

26 years agoCleanup natd startup test.
Steve Price [Sun, 14 Jun 1998 16:31:03 +0000 (16:31 +0000)]
Cleanup natd startup test.

PR: 6946
Submitted by: Jacques Vidrine <n@nectar.com>

26 years ago'They They' -> 'They'
Steve Price [Sun, 14 Jun 1998 16:06:00 +0000 (16:06 +0000)]
'They They' -> 'They'

PR: 6912
Submitted by: Stefan Eggers <seggers@semyam.dinoco.de>

26 years agoComplete the switch to using gcc-2.4 features to declare fpos_t ,
Bruce Evans [Sun, 14 Jun 1998 16:04:20 +0000 (16:04 +0000)]
Complete the switch to using gcc-2.4 features to declare fpos_t ,
as threatened in the 4-5 year old comment.  Most of the change is
in <machine/ansi.h>; this commit just removes cruft for the strict
ANSI case.  64-bit types couldn't actually be defined using
__attribute__(()) in gcc-2.4, since attribute names in the
implementation namespace only became available in gcc-2.7.  I've
probably just broken gcc-2.6 support by using the gcc-2.7 names
unconditionally.

26 years agosprintf -> snprintf to avoid potential buffer overflow.
Steve Price [Sun, 14 Jun 1998 16:03:40 +0000 (16:03 +0000)]
sprintf -> snprintf to avoid potential buffer overflow.

PR: 6907
Submitted by: Archie Cobbs <archie@whistle.com>

26 years agoAvoid an egcs pessimization for 64-bit signed division on i386's.
Bruce Evans [Sun, 14 Jun 1998 15:52:00 +0000 (15:52 +0000)]
Avoid an egcs pessimization for 64-bit signed division on i386's.
Pre-2.8 versions of gcc generate a call to __divdi3() for all 64-bit
signed divisions, but egcs optimizes them to a shift and fixup when
the divisor is a constant power of 2.  Unfortunately, it generates
a call to __cmpdi2() for the fixup, although all except possibly
ancient versions of gcc and egcs do ordinary 64-bit comparisons
inline.

26 years agoAdded uoff_t.
Bruce Evans [Sun, 14 Jun 1998 15:26:03 +0000 (15:26 +0000)]
Added uoff_t.

26 years agoRemoved LOCORE ifdef. This header isn't used for locore stuff.
Bruce Evans [Sun, 14 Jun 1998 15:21:27 +0000 (15:21 +0000)]
Removed LOCORE ifdef.  This header isn't used for locore stuff.

Use uoff_t instead of `unsigned long long' for unsigned offsets.

Fixed some style bugs.

26 years agoDefine 64-bit types in terms of __int64_t and __uint64_t.
Bruce Evans [Sun, 14 Jun 1998 14:08:21 +0000 (14:08 +0000)]
Define 64-bit types in terms of __int64_t and __uint64_t.

Define an unsigned version of off_t (uoff_t) if KERNEL is defined.
This is useful for micro-optimizing expressions involving off_t's
where you know that the values are non-negative but the compiler
doesn't know it.

Fixed a gratuitous ANSIism.

26 years agoAdded typedefs for __int64_t and __uint64_t. These will be used to avoid
Bruce Evans [Sun, 14 Jun 1998 14:00:47 +0000 (14:00 +0000)]
Added typedefs for __int64_t and __uint64_t.  These will be used to avoid
explicit use of `[unsigned] long long' and `[u]quad_t'.

Fixed disorder of and lack of comments for the types added in the previous
two commits.  Improved some comments.  The comments should be elswhere
(not duplicated in every MD file).

26 years agoDeclare zs_cnattach.
Doug Rabson [Sun, 14 Jun 1998 13:58:57 +0000 (13:58 +0000)]
Declare zs_cnattach.

26 years ago[Add missing files from previous commit]
Doug Rabson [Sun, 14 Jun 1998 13:53:12 +0000 (13:53 +0000)]
[Add missing files from previous commit]

Major changes to the generic device framework for FreeBSD/alpha:

* Eliminate bus_t and make it possible for all devices to have
  attached children.

* Support dynamically extendable interfaces for drivers to replace
  both the function pointers in driver_t and bus_ops_t (which has been
  removed entirely.  Two system defined interfaces have been defined,
  'device' which is mandatory for all devices and 'bus' which is
  recommended for all devices which support attached children.

* In addition, the alpha port defines two simple interfaces 'clock'
  for attaching various real time clocks to the system and 'mcclock'
  for the many different variations of mc146818 clocks which can be
  attached to different alpha platforms.  This eliminates two more
  function pointer tables in favour of the generic method dispatch
  system provided by the device framework.

Future device interfaces may include:

* cdev and bdev interfaces for devfs to use in replacement for specfs
  and the fixed interfaces bdevsw and cdevsw.

* scsi interface to replace struct scsi_adapter (not sure how this
  works in CAM but I imagine there is something similar there).

* various tailored interfaces for different bus types such as pci,
  isa, pccard etc.

26 years agoMajor changes to the generic device framework for FreeBSD/alpha:
Doug Rabson [Sun, 14 Jun 1998 13:46:10 +0000 (13:46 +0000)]
Major changes to the generic device framework for FreeBSD/alpha:

* Eliminate bus_t and make it possible for all devices to have
  attached children.

* Support dynamically extendable interfaces for drivers to replace
  both the function pointers in driver_t and bus_ops_t (which has been
  removed entirely.  Two system defined interfaces have been defined,
  'device' which is mandatory for all devices and 'bus' which is
  recommended for all devices which support attached children.

* In addition, the alpha port defines two simple interfaces 'clock'
  for attaching various real time clocks to the system and 'mcclock'
  for the many different variations of mc146818 clocks which can be
  attached to different alpha platforms.  This eliminates two more
  function pointer tables in favour of the generic method dispatch
  system provided by the device framework.

Future device interfaces may include:

* cdev and bdev interfaces for devfs to use in replacement for specfs
  and the fixed interfaces bdevsw and cdevsw.

* scsi interface to replace struct scsi_adapter (not sure how this
  works in CAM but I imagine there is something similar there).

* various tailored interfaces for different bus types such as pci,
  isa, pccard etc.

26 years agoDon't define away __attribute__() for the non-gcc and old-gcc cases.
Bruce Evans [Sun, 14 Jun 1998 13:40:01 +0000 (13:40 +0000)]
Don't define away __attribute__() for the non-gcc and old-gcc cases.
Using __attribute__() in non-ifdefed code is just wrong if it changes
the semantics, and there is no way to ignore __attribute__() only in
the benign cases.

Don't use __attribute__ in non-ifdefed code here.  It was a benign
case, but we already have a macro (__unused) for this case.

Expanded comment about gotchas for __CONCAT() in the non-ANSI case.

26 years agoElf64_Shdr.sh_size should be Elf64_Size not Elf64_Word.
Doug Rabson [Sun, 14 Jun 1998 13:24:09 +0000 (13:24 +0000)]
Elf64_Shdr.sh_size should be Elf64_Size not Elf64_Word.

26 years agoAvoid a 64-bit division in procfs_readdir(). Fixed related overflows.
Bruce Evans [Sun, 14 Jun 1998 12:53:39 +0000 (12:53 +0000)]
Avoid a 64-bit division in procfs_readdir().  Fixed related overflows.
Check args using the same expression as in fdesc and kernfs.  The check
was actually already correct, modulo overflow.  It could be tightened
up to either allow huge (aligned) offsets, treating them as EOF, or
disallow all offsets beyond EOF.

Didn't fix invalid address calculation &foo[i] where i may be out of
bounds.

Didn't fix shooting of foot using a private unportable dirent struct.

26 years agoAvoid a 64-bit division in kernfs_readdir(). Fixed related overflows
Bruce Evans [Sun, 14 Jun 1998 12:34:42 +0000 (12:34 +0000)]
Avoid a 64-bit division in kernfs_readdir().  Fixed related overflows
and arg checking.

26 years agoDon't compile in the use of poll() when building libc_r. This isn't
Peter Wemm [Sun, 14 Jun 1998 11:25:46 +0000 (11:25 +0000)]
Don't compile in the use of poll() when building libc_r.  This isn't
so much a "fix", rather a bandaid to buy time to fix it properly
within the thread engine.

26 years agoAvoid 64-bit divisions and modulos. Fixed related overflows for weird
Bruce Evans [Sun, 14 Jun 1998 10:52:52 +0000 (10:52 +0000)]
Avoid 64-bit divisions and modulos.  Fixed related overflows for weird
args.

This driver should not be used, since it calls uiomove() with interrupts
disabled.

26 years agoIf a short write, only loop if no error.
John Birrell [Sun, 14 Jun 1998 09:36:14 +0000 (09:36 +0000)]
If a short write, only loop if no error.

26 years agoAvoid a 64-bit division in fdesc_readdir(). Fixed related overflows
Bruce Evans [Sun, 14 Jun 1998 08:46:41 +0000 (08:46 +0000)]
Avoid a 64-bit division in fdesc_readdir().  Fixed related overflows
and missing arg checking.

Panic instead of returning bogus error codes or forgetting to check
all cases if fdesc_readdir() gets called for a non-directory.  This
can't happen.

26 years agoDocument EINVAL return value.
Bruce Evans [Sun, 14 Jun 1998 08:22:05 +0000 (08:22 +0000)]
Document EINVAL return value.

26 years agoDeal with PROTO_IGMP and PROTO_IPIP packets correctly.
Brian Somers [Sun, 14 Jun 1998 01:21:24 +0000 (01:21 +0000)]
Deal with PROTO_IGMP and PROTO_IPIP packets correctly.

Submitted by: Mark Tinguely <tinguely@plains.NoDak.edu>
  This change will allow a PPP host enabled with the "-alias" option to
  run mrouted. This does not intend to forward the IGMP nor tunneled packets
  to another host on the far side if the tun0 interface.

26 years agoo Pass our negotiated number of VJ slots into
Brian Somers [Sun, 14 Jun 1998 00:56:13 +0000 (00:56 +0000)]
o Pass our negotiated number of VJ slots into
  sl_uncompress_tcp() and drop packets with
  slot numbers that are out of range.
o Drop packets that want to use a slot that still
  has an IP header length of 0 (ie, the requested
  slot number is bogus again).

Without this code, if the other side mis-behaves (and
sends us garbage slot numbers), we happily ``adjust''
a memset(..., '\0', ...) TCP/IP header and promptly
cr*p all over the stack before returning.... quickly
followed by a SIGBUS.

Dodgy ISP used by, and help locating the problem from: jmz
Problem also seen by: Mourad de Riche <omnibus@image.dk>

There's still a link lockup after this happens, but my
bets are on the other side (who has already started sending
rubbish) being to blame.

26 years agoAdd a macro tweak.
Steve Price [Sat, 13 Jun 1998 19:36:22 +0000 (19:36 +0000)]
Add a macro tweak.

PR: 6932
Submitted by: Nick Sayer <nsayer@quack.kfu.com>

26 years agoMention all the environment variables that can be used
Steve Price [Sat, 13 Jun 1998 19:28:51 +0000 (19:28 +0000)]
Mention all the environment variables that can be used
to specify the device.

PR: 6927
Submitted by: Stefan Eggers <seggers@semyam.dinoco.de>

26 years agoAdd doc/ja/FAQ and handbook, so that releases can build again.
John Hay [Sat, 13 Jun 1998 19:26:38 +0000 (19:26 +0000)]
Add doc/ja/FAQ and handbook, so that releases can build again.

26 years agoCorrect the description of the last four fields.
Steve Price [Sat, 13 Jun 1998 19:14:14 +0000 (19:14 +0000)]
Correct the description of the last four fields.

PR: 6926
Submitted by: Josh Gilliam <josh@quick.net>

26 years agomiserable -> miserably
Steve Price [Sat, 13 Jun 1998 19:06:49 +0000 (19:06 +0000)]
miserable -> miserably

PR: 6922
Submitted by: Sue Blake <sue@welearn.com.au>

26 years agoMinor tweaks in the AUTHORS section.
Steve Price [Sat, 13 Jun 1998 18:55:55 +0000 (18:55 +0000)]
Minor tweaks in the AUTHORS section.

PR: 6921
Submitted by: Josh Gilliam <josh@quick.net>

26 years agoAdd a symlink from /dev/vga to /dev/ttyv0 when using DEVFS
Steve Price [Sat, 13 Jun 1998 18:53:22 +0000 (18:53 +0000)]
Add a symlink from /dev/vga to /dev/ttyv0 when using DEVFS
for programs like doscmd.

PR: 6920
Submitted by: Luoqi Chen <luoqi@chen.ml.org>

26 years agoAdd a swissfrench keyboard mapping.
Steve Price [Sat, 13 Jun 1998 18:39:22 +0000 (18:39 +0000)]
Add a swissfrench keyboard mapping.

PR: 6916
Submitted by: tamone@eig.unige.ch

26 years agoMerge changes from vendor branch;
Peter Wemm [Sat, 13 Jun 1998 17:20:03 +0000 (17:20 +0000)]
Merge changes from vendor branch;
- connector selection values (should fix aui/bnc),
- non-shifting version of crc calculation using a table,
- interrupt mask adjustments,
- add some brackets where a #ifdef could break an if(),
- don't reset the card unless it's up.

26 years agoMake -V expand it's arguments. This means that instead of 'make -V OBJS'
Peter Wemm [Sat, 13 Jun 1998 11:55:57 +0000 (11:55 +0000)]
Make -V expand it's arguments.  This means that instead of 'make -V OBJS'
printing something useless (to a shell) like: ${SRCS:N*.h:R:S/$/.o/g}
it will instead print the actual ${OBJS} value.

26 years agoIntroduce std_pps_ioctl() to automagically DTRT.
Poul-Henning Kamp [Sat, 13 Jun 1998 09:30:26 +0000 (09:30 +0000)]
Introduce std_pps_ioctl() to automagically DTRT.

Add scaling capability to timex.offset, ntpd-4.0.73 will support this.

26 years agoAllow a protocol to specify that it does NOT want to be looped back
Julian Elischer [Sat, 13 Jun 1998 02:27:10 +0000 (02:27 +0000)]
Allow a protocol to specify that it does NOT want to be looped back
even if it looks like it should (backwards compatibility with
old broken code) should get rid of some annoying messags.

26 years agojoin(1) is used by lorder, add it to build-tools. (otherwise, things get
Peter Wemm [Sat, 13 Jun 1998 02:03:36 +0000 (02:03 +0000)]
join(1) is used by lorder, add it to build-tools.  (otherwise, things get
ugly when an a.out /usr/bin/join sees the elf LD_LIBRARY_PATH from a
'make world' and it's a.out ld.so tries to load the elf libc.so.)

26 years agoUpdate to draft-mogul-pps-api-02.txt as submitted to IETF
Poul-Henning Kamp [Fri, 12 Jun 1998 23:15:53 +0000 (23:15 +0000)]
Update to draft-mogul-pps-api-02.txt as submitted to IETF

26 years agoNote which version of Kirk's sources this corresponds to.
Julian Elischer [Fri, 12 Jun 1998 21:21:26 +0000 (21:21 +0000)]
Note which version of Kirk's sources this corresponds to.

26 years agoFix the case when renaming to a file that you've just created and deleted,
Julian Elischer [Fri, 12 Jun 1998 20:48:30 +0000 (20:48 +0000)]
Fix the case when renaming to a file that you've just created and deleted,
that had an inode that has not yet been written to disk, when the inode of the
new file is also not yet written to disk, and your old directory entry is not
yet on disk but you need to remove it and the new name exists in memory
but has been deleted but the transaction to write the deleted name to disk
exists and has not yet been cancelled by the request to delete the non
existant name.  I don't know how kirk could have missed such a glaring
problem for so long. :-) Especially since the inconsitency survived on
the disk for a whole 4 second on average before being fixed by other code.
This was not a crashing bug but just led to filesystem inconsitencies
if you crashed.

Submitted by: Kirk McKusick (mckusick@mckusick.com)