]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoAdjust code for new mailbox format.
David Xu [Mon, 18 Nov 2002 02:02:08 +0000 (02:02 +0000)]
Adjust code for new mailbox format.

Reviewed by: deischen, mini

21 years ago1. Support versioning and wall clock in kse mailbox,
David Xu [Mon, 18 Nov 2002 01:59:31 +0000 (01:59 +0000)]
1. Support versioning and wall clock in kse mailbox,
   also add rusage time in thread mailbox.
2. Minor change for thread limit code in thread_user_enter(),
   fix typo in kse_release() last I committed.

Reviewed by: deischen, mini

21 years agoMFi386 r1.369
Alan Cox [Mon, 18 Nov 2002 01:36:09 +0000 (01:36 +0000)]
MFi386 r1.369
 - Clear the PG_WRITEABLE flag in pmap_changebit() if write access is
   being removed.  Return immediately if write access is being removed and
   PG_WRITEABLE is already clear.

Note: For efficiency, pmap_changebit() should be replaced by a function
similar to sparc64's pmap_clear_write().

21 years agoFlesh out the description of the uma_zcreate callback function arguements
Alfred Perlstein [Mon, 18 Nov 2002 01:11:58 +0000 (01:11 +0000)]
Flesh out the description of the uma_zcreate callback function arguements
a bit.  As there may be changes soon we're still a bit vague unfortunatly.

21 years agoDeal with the possibility that time_t != int32_t. Otherwise ia64 thought
Peter Wemm [Sun, 17 Nov 2002 23:50:41 +0000 (23:50 +0000)]
Deal with the possibility that time_t != int32_t.  Otherwise ia64 thought
the fs_old_size was the half part of fs_old_time etc.

21 years agoOops. Some ut_time stuff slipped through the cracks. These turned out
Peter Wemm [Sun, 17 Nov 2002 23:46:45 +0000 (23:46 +0000)]
Oops.  Some ut_time stuff slipped through the cracks.  These turned out
to be non-fatal due to stack alignment roundups.

21 years agoinclude smp.h.
Julian Elischer [Sun, 17 Nov 2002 23:26:42 +0000 (23:26 +0000)]
include smp.h.
it is required by some code that was commented out until david's
last commit.

21 years agoMFi386 r1.369
Alan Cox [Sun, 17 Nov 2002 21:48:42 +0000 (21:48 +0000)]
MFi386 r1.369
 - Clear the PG_WRITEABLE flag in pmap_page_protect() if write access is
   being removed.  Return immediately if write access is being removed and
   PG_WRITEABLE is already clear.

21 years agoAdd ${DESTDIR} forgotten in the previous commit.
Hiroki Sato [Sun, 17 Nov 2002 20:38:27 +0000 (20:38 +0000)]
Add ${DESTDIR} forgotten in the previous commit.

21 years agoUpgrade the root partition mount to r/w before trying to write to it.
Luigi Rizzo [Sun, 17 Nov 2002 20:19:34 +0000 (20:19 +0000)]
Upgrade  the root partition mount to r/w before trying to write to it.

MFC after: 3 days

21 years agoAdd a new translation and activate it:
Hiroki Sato [Sun, 17 Nov 2002 19:45:00 +0000 (19:45 +0000)]
Add a new translation and activate it:

1.2   -> 1.3   ja_JP.eucJP/Makefile
(new) -> 1.3   ja_JP.eucJP/early-adopter/Makefile
(new) -> 1.5   ja_JP.eucJP/early-adopter/article.sgml

21 years agoReplace m_copy() with m_copypacket() where applicable.
Luigi Rizzo [Sun, 17 Nov 2002 18:14:04 +0000 (18:14 +0000)]
Replace m_copy() with m_copypacket() where applicable.
Replace 0 with NULL where appropriate.
Fix indentation and function headers.

21 years agoFix function headers, remove 'register' from variable declarations.
Luigi Rizzo [Sun, 17 Nov 2002 18:13:02 +0000 (18:13 +0000)]
Fix function headers, remove 'register' from variable declarations.

21 years agoFix function headers and remove 'register' variable declarations.
Luigi Rizzo [Sun, 17 Nov 2002 17:04:19 +0000 (17:04 +0000)]
Fix function headers and remove 'register' variable declarations.

21 years agoUpdate SYNOPSIS to reflect the standardized header. Add STANDARDS
Mike Barcroft [Sun, 17 Nov 2002 16:34:07 +0000 (16:34 +0000)]
Update SYNOPSIS to reflect the standardized header.  Add STANDARDS
section.

PR: 43270

21 years agoMove the ip_fragment code from ip_output() to a separate function,
Luigi Rizzo [Sun, 17 Nov 2002 16:30:44 +0000 (16:30 +0000)]
Move the ip_fragment code from ip_output() to a separate function,
so that it can be reused elsewhere (there is a number of places
where it can be useful). This also trims some 200 lines from
the body of ip_output(), which helps readability a bit.

(This change was discussed a few weeks ago on the mailing lists,
Julian agreed, silence from others. It is not a functional change,
so i expect it to be ok to commit it now but i am happy to back it
out if there are objections).

While at it, fix some function headers and replace m_copy() with
m_copypacket() where applicable.

MFC after: 1 week

21 years ago1. Hide the internals of struct fd_set in standard namespaces.
Mike Barcroft [Sun, 17 Nov 2002 16:22:18 +0000 (16:22 +0000)]
1. Hide the internals of struct fd_set in standard namespaces.
2. Avoid referencing bcopy() and bzero(), since they may not be in
   scope.

Request by:     bde (1)
Submitted by:   wollman (2)
Reviewed by: archie, bde
PR: 43270

21 years agoMinor documentation changes and indentation fix.
Luigi Rizzo [Sun, 17 Nov 2002 16:13:08 +0000 (16:13 +0000)]
Minor documentation changes and indentation fix.

Replace m_copy() with m_copypacket() where applicable.

While at it, fix some function headers and remove 'register' from
variable declarations.

21 years agoCleanup some of the comments, and reformat long lines.
Luigi Rizzo [Sun, 17 Nov 2002 16:02:17 +0000 (16:02 +0000)]
Cleanup some of the comments, and reformat long lines.

Replace m_copy() with m_copypacket() where applicable.

Replace "if (a.s_addr ...)" with "if (a.s_addr != INADDR_ANY ...)"
to make it clear what the code means.

While at it, fix some function headers and remove 'register' from
variable declarations.

MFC after: 3 days

21 years ago- Add support for ALT_BREAK_TO_DEBUGGER; this is the only reliable way to
Jake Burkholder [Sun, 17 Nov 2002 16:00:51 +0000 (16:00 +0000)]
- Add support for ALT_BREAK_TO_DEBUGGER; this is the only reliable way to
  trigger a breakpoint with this chip.
- Fiddle the right bits in the cn input and output routines to disable port
  interrupts and enable visibility of the masked interrupt status bits.
- Register a shutdown final event handler to put the chip back in the mode
  that the prom expects.

21 years agoApply some fixups in the driver_t's.
Josef Karthauser [Sun, 17 Nov 2002 14:22:37 +0000 (14:22 +0000)]
Apply some fixups in the driver_t's.

Submitted by: akiyama
MFC after: 3 days

21 years agoUse a sysctl for controlling the debugging output.
Josef Karthauser [Sun, 17 Nov 2002 14:08:48 +0000 (14:08 +0000)]
Use a sysctl for controlling the debugging output.

Submitted by: akiyama

21 years agoSave a slice name on the disk and print it at g_pc98_dumpconf().
Yoshihiro Takahashi [Sun, 17 Nov 2002 13:56:37 +0000 (13:56 +0000)]
Save a slice name on the disk and print it at g_pc98_dumpconf().

21 years ago- Ease sanity check to get cylinders.
Yoshihiro Takahashi [Sun, 17 Nov 2002 13:51:22 +0000 (13:51 +0000)]
- Ease sanity check to get cylinders.
- Get the slice name from the result of kern.geom.conftxt.

21 years agoCopy from ${EXTLOCALDIR} to ${CHROOTDIR}/usr/local if EXTLOCALDIR is defined.
Yoshihiro Takahashi [Sun, 17 Nov 2002 13:35:12 +0000 (13:35 +0000)]
Copy from ${EXTLOCALDIR} to ${CHROOTDIR}/usr/local if EXTLOCALDIR is defined.

21 years agoFix a typo in a console message.
Josef Karthauser [Sun, 17 Nov 2002 13:33:55 +0000 (13:33 +0000)]
Fix a typo in a console message.

Submitted by: akiyama
MFC after: 3 days

21 years ago1.Add sysctls to control KSE resource allocation.
David Xu [Sun, 17 Nov 2002 11:47:03 +0000 (11:47 +0000)]
1.Add sysctls to control KSE resource allocation.
  kern.threads.max_threads_per_proc
  kern.threads.max_groups_per_proc
2.Temporary disable borrower thread stash itself as
  owner thread's spare thread in thread_exit(). there
  is a race between owner thread and borrow thread:
  an owner thread may allocate a spare thread as this:
if (td->td_standin == NULL)
td->standin = thread_alloc();
  but thread_alloc() can block the thread, then a borrower
  thread would possible stash it self as owner's spare
  thread in thread_exit(), after owner is resumed, result
  is a thread leak in kernel, double check in owner can
  avoid the race, but it may be ugly and not worth to do.

21 years agoRework last exiting thread in kse_release(), wait a signal and then
David Xu [Sun, 17 Nov 2002 10:12:00 +0000 (10:12 +0000)]
Rework last exiting thread in kse_release(), wait a signal and then
schedule an upcall and call thread_exit().

21 years agoFix a typo. Add FreeBSD ID in a %% comment.
Jens Schweikhardt [Sun, 17 Nov 2002 09:53:11 +0000 (09:53 +0000)]
Fix a typo. Add FreeBSD ID in a %% comment.

MFC after: 3 days

21 years ago - Release the imgp vnode prior to freeing exec_map resources to avoid
Jeff Roberson [Sun, 17 Nov 2002 09:33:00 +0000 (09:33 +0000)]
 - Release the imgp vnode prior to freeing exec_map resources to avoid
   deadlock.

21 years agogoto break; != break;
Doug Barton [Sun, 17 Nov 2002 08:54:29 +0000 (08:54 +0000)]
goto break; != break;

I've no idea if this is the right behavior for the library, but this
at least fixes the build, and matches what seems to be alfred's intent
in the commit message for 1.19.

21 years agoDETACH_FORCE was removed recently. Remove it here, even though this
Warner Losh [Sun, 17 Nov 2002 04:52:37 +0000 (04:52 +0000)]
DETACH_FORCE was removed recently.  Remove it here, even though this
isn't supported.

21 years agoRework the sysconf(3) interaction with aio:
Alfred Perlstein [Sun, 17 Nov 2002 04:15:34 +0000 (04:15 +0000)]
Rework the sysconf(3) interaction with aio:

sysconf.c:
  Use 'break' rather than 'goto yesno' in sysconf.c so that we report a '0'
  return value from the kernel sysctl.

vfs_aio.c:
  Make aio reset its configuration parameters to -1 after unloading
  instead of 0.

posix4_mib.c:
  Initialize the aio configuration parameters to -1
  to indicate that it is not loaded.
  Add a facility (p31b_iscfg()) to determine if a posix4 facility has been
  initialized to avoid having to re-order the SYSINITs.
  Use p31b_iscfg() to determine if aio has had a chance to run yet which
  is likely if it is compiled into the kernel and avoid spamming its
  values.
  Introduce a macro P31B_VALID() instead of doing the same comparison over
  and over.

posix4.h:
  Prototype p31b_iscfg().

21 years agoMFi386: revision 1.550.
Yoshihiro Takahashi [Sun, 17 Nov 2002 02:57:06 +0000 (02:57 +0000)]
MFi386: revision 1.550.

21 years agoMerged from sys/isa/fd.c revision 1.242.
Yoshihiro Takahashi [Sun, 17 Nov 2002 02:39:55 +0000 (02:39 +0000)]
Merged from sys/isa/fd.c revision 1.242.

21 years agoMFi386 r1.369. Clear the PG_WRITEABLE flag in pmap_clear_write; return
Jake Burkholder [Sun, 17 Nov 2002 01:17:07 +0000 (01:17 +0000)]
MFi386 r1.369.  Clear the PG_WRITEABLE flag in pmap_clear_write; return
immediately if its already clear.

Suggested by: alc

21 years agoRegenerate after adding syscalls.
Daniel Eischen [Sat, 16 Nov 2002 23:48:14 +0000 (23:48 +0000)]
Regenerate after adding syscalls.

21 years ago - Don't forget the flags value when using boot pages.
Jeff Roberson [Sat, 16 Nov 2002 20:57:41 +0000 (20:57 +0000)]
 - Don't forget the flags value when using boot pages.

Reported by: grehan

21 years agoImplement the lock with a cmpxchg instruction instead of a xchg.
Marcel Moolenaar [Sat, 16 Nov 2002 17:05:06 +0000 (17:05 +0000)]
Implement the lock with a cmpxchg instruction instead of a xchg.
Both are atomic, but the cmpxchg has memory ordering hints. We
give this acquire semantics.

NOTE: The unlock in libc_r is implemented by a "normal" assign
statement. This is not correct on ia64 due to the memory ordering
characteristics of the architecture. We need release semantics
for an unlock.

21 years agoEnable selecting the type of partition menu on pc98.
Yoshihiro Takahashi [Sat, 16 Nov 2002 16:36:01 +0000 (16:36 +0000)]
Enable selecting the type of partition menu on pc98.

21 years agoAdd *context() syscalls to ia64 32-bit compatability table as requested
Daniel Eischen [Sat, 16 Nov 2002 15:15:17 +0000 (15:15 +0000)]
Add *context() syscalls to ia64 32-bit compatability table as requested
in kern/syscalls.master.

21 years agoDo not emit a message on stderr when one of the compared files
Thomas Quinot [Sat, 16 Nov 2002 14:58:39 +0000 (14:58 +0000)]
Do not emit a message on stderr when one of the compared files
is shorter than the other.

Reviewed by: roberto
MFC after: 3 days

21 years agoRemove incorrect output redirection.
Thomas Quinot [Sat, 16 Nov 2002 14:57:12 +0000 (14:57 +0000)]
Remove incorrect output redirection.

Reviewed by: roberto
Committed from: EuroBSDCon Amsterdam
MFC after: 3 days

21 years agoset watch in tcsh requires an equal sign after the variable name.
Joerg Wunsch [Sat, 16 Nov 2002 13:24:44 +0000 (13:24 +0000)]
set watch in tcsh requires an equal sign after the variable name.

MFC after: 1 day

21 years agoMfebd in a hurry
Udo Erdelhoff [Sat, 16 Nov 2002 12:14:22 +0000 (12:14 +0000)]
Mfebd in a hurry
installation/common/install.sgml: 1.15  -> 1.16
installation/sparc64/article.sgml: 1.3   -> 1.4
relnotes/common/new.sgml: 1.451 -> 1.452

21 years agoThese two patches makes it easier to compile custom versions of
Ollivier Robert [Sat, 16 Nov 2002 08:06:43 +0000 (08:06 +0000)]
These two patches makes it easier to compile custom versions of
NTP on FreeBSD:

The first one allows one to avoid installing the html files.

The second one allows one to override the CLOCKDEFS on the
make command line.

Submitted by: phk

21 years agoNow that pmap_remove_all() is exported by our pmap implementations
Alan Cox [Sat, 16 Nov 2002 07:44:25 +0000 (07:44 +0000)]
Now that pmap_remove_all() is exported by our pmap implementations
use it directly.

21 years agoDisconnect the userland get/set/swapcontext() functions from
Daniel Eischen [Sat, 16 Nov 2002 06:39:11 +0000 (06:39 +0000)]
Disconnect the userland get/set/swapcontext() functions from
libc.  I want to keep these in some version for the thread
library/ies, but don't know whether to have them repo-copied
to libc_r or renamed and kept in libc.

Change the name of an alpha macro that was changed with the
system call commit.

21 years agoExport the values for _SC_AIO_MAX and _SC_AIO_PRIO_DELTA_MAX via the p1003b
Alfred Perlstein [Sat, 16 Nov 2002 06:38:07 +0000 (06:38 +0000)]
Export the values for _SC_AIO_MAX and _SC_AIO_PRIO_DELTA_MAX via the p1003b
sysctl interface.

21 years agoRegenerate after adding system calls.
Daniel Eischen [Sat, 16 Nov 2002 06:36:56 +0000 (06:36 +0000)]
Regenerate after adding system calls.

21 years agoAdd getcontext, setcontext, and swapcontext as system calls.
Daniel Eischen [Sat, 16 Nov 2002 06:35:53 +0000 (06:35 +0000)]
Add getcontext, setcontext, and swapcontext as system calls.
Previously these were libc functions but were requested to
be made into system calls for atomicity and to coalesce what
might be two entrances into the kernel (signal mask setting
and floating point trap) into one.

A few style nits and comments from bde are also included.

Tested on alpha by: gallatin

21 years agoProvide more correct default values for sysconf(3) reporting of the AIO
Alfred Perlstein [Sat, 16 Nov 2002 06:35:20 +0000 (06:35 +0000)]
Provide more correct default values for sysconf(3) reporting of the AIO
subsystems capabilities:

_SC_AIO_LISTIO_MAX returns the default of _POSIX_AIO_LISTIO_MAX
_SC_AIO_MAX returns the default _POSIX_AIO_MAX
_SC_AIO_PRIO_DELTA_MAX returns the default of 0

Without these adjustments the values returned are -1 even when the
aio side of the kernel returns '0' for them which is incorrect.

Noticed by: Craig Rodrigues <rodrigc@attbi.com>

21 years agoCall 'p31b_setcfg(CTL_P1003_1B_AIO_LISTIO_MAX, AIO_LISTIO_MAX)'
Alfred Perlstein [Sat, 16 Nov 2002 04:22:55 +0000 (04:22 +0000)]
Call 'p31b_setcfg(CTL_P1003_1B_AIO_LISTIO_MAX, AIO_LISTIO_MAX)'
when AIO is initialized so that sysconf() gives correct results.

Reported by: Craig Rodrigues <rodrigc@attbi.com>

21 years agoRepair buglet introduced with the last import of Heimdal:
Jacques Vidrine [Sat, 16 Nov 2002 02:23:17 +0000 (02:23 +0000)]
Repair buglet introduced with the last import of Heimdal:
`krb5-config --cflags' spewed an erroneous argument.

Reported by: Gabor@Zahemszky.HU
Approved by: re (jhb)

21 years agoSort SRCS.
Marcel Moolenaar [Sat, 16 Nov 2002 01:41:33 +0000 (01:41 +0000)]
Sort SRCS.

21 years agoProperly calculate the initial number of fragments in a large filesystem.
Kirk McKusick [Fri, 15 Nov 2002 23:50:14 +0000 (23:50 +0000)]
Properly calculate the initial number of fragments in a large filesystem.

Sponsored by:   DARPA & NAI Labs.

21 years agoheaders should not really include "opt_foo.h" (in this case opt_posix.h).
Alfred Perlstein [Fri, 15 Nov 2002 22:55:06 +0000 (22:55 +0000)]
headers should not really include "opt_foo.h" (in this case opt_posix.h).
remove it from the header and add it to the files that require it.

21 years agoMassive cleanup of the ip_mroute code.
Luigi Rizzo [Fri, 15 Nov 2002 22:53:53 +0000 (22:53 +0000)]
Massive cleanup of the ip_mroute code.

No functional changes, but:

  + the mrouting module now should behave the same as the compiled-in
    version (it did not before, some of the rsvp code was not loaded
    properly);
  + netinet/ip_mroute.c is now truly optional;
  + removed some redundant/unused code;
  + changed many instances of '0' to NULL and INADDR_ANY as appropriate;
  + removed several static variables to make the code more SMP-friendly;
  + fixed some minor bugs in the mrouting code (mostly, incorrect return
    values from functions).

This commit is also a prerequisite to the addition of support for PIM,
which i would like to put in before DP2 (it does not change any of
the existing APIs, anyways).

Note, in the process we found out that some device drivers fail to
properly handle changes in IFF_ALLMULTI, leading to interesting
behaviour when a multicast router is started. This bug is not
corrected by this commit, and will be fixed with a separate commit.

Detailed changes:
--------------------
netinet/ip_mroute.c     all the above.
conf/files              make ip_mroute.c optional
net/route.c             fix mrt_ioctl hook
netinet/ip_input.c      fix ip_mforward hook, move rsvp_input() here
                        together with other rsvp code, and a couple
                        of indentation fixes.
netinet/ip_output.c     fix ip_mforward and ip_mcast_src hooks
netinet/ip_var.h        rsvp function hooks
netinet/raw_ip.c        hooks for mrouting and rsvp functions, plus
                        interface cleanup.
netinet/ip_mroute.h     remove an unused and optional field from a struct

Most of the code is from Pavlin Radoslavov and the XORP project

Reviewed by: sam
MFC after: 1 week

21 years agoA little bit of anti-foot-shooting. Use utimes(2) rather than
Peter Wemm [Fri, 15 Nov 2002 22:43:56 +0000 (22:43 +0000)]
A little bit of anti-foot-shooting.  Use utimes(2) rather than
the deprecated utime(3).  utimes(2) uses timeval, but utime(3) uses
time_t's.  If you do bad things (like I did) by mixing up include files
with libc, then install can do strange things if you mismatch the time_t
stuff.  utime() is emulated entirely within libc.

Approved by:  re (jhb)

21 years agoutmp.ut_time and lastlog.ll_time are explicitly int32_t rather than
Peter Wemm [Fri, 15 Nov 2002 22:42:00 +0000 (22:42 +0000)]
utmp.ut_time and lastlog.ll_time are explicitly int32_t rather than
time_t.  Deal with the possibility that time_t != int32_t.  This boils
down to this sort of thing:
 -   time(&ut.ut_time);
 +   ut.ut_time = time(NULL);
and similar for ctime(3) etc.  I've kept it minimal for the stuff
that may need to be portable (or 3rd party code), but used Matt's time32
stuff for cases where that isn't as much of a concern.

Approved by: re (jhb)

21 years agoDo not assume that time_t is an int.
Peter Wemm [Fri, 15 Nov 2002 22:36:57 +0000 (22:36 +0000)]
Do not assume that time_t is an int.

Approved by: re (jhb)

21 years agoTest the water. Make time_t long (64 bit) on ia64 since we do not have
Peter Wemm [Fri, 15 Nov 2002 22:35:34 +0000 (22:35 +0000)]
Test the water.  Make time_t long (64 bit) on ia64 since we do not have
to worry about ABI vs released systems yet.  This is mostly transparent
since there is no significant exposure in the syscall interface.  The
things that go wrong are mostly userland stuff - time(&intvariable).

Reviewed by: dfr, marcel
Approved by: re (jhb)

21 years agoRework the checking out of the doc, src, and ports trees a bit to make it
John Baldwin [Fri, 15 Nov 2002 22:03:25 +0000 (22:03 +0000)]
Rework the checking out of the doc, src, and ports trees a bit to make it
more manageable.
- Add some helper variables (CVS_{SRC,DOC,PORTS}ARGS) to be used when
  using CVS to checkout files.  We stick release tags in these helper
  variables if they are defined and then use only one cvs command instead
  of two cvs commands with an .ifdef to choose between them.
- rm the old src/doc/ports directories as separate commands from the
  CVS comands so that the rm commands don't need to be duplicated.
- Simplify the DOMINIMALDOCPORTS case by overriding RELEASEPORTSMODULE to
  be ${MINIMALDOCPORTS} thus removing yet another nearly-duplicate cvs
  command in an .ifdef.
- Add support for grabbing src/ and doc/ from external directories
  specified via EXTSRCDIR and EXTDOCDIR instead of from CVS.  The same
  is not done for ports/ quite yet as the DOMINIMALDOCPORTS case is a
  bit tricky.

The rerelease target scripts have not been changed to use the helper
variables yet, so there is still some room for improvement.

Submitted by: kuriyama (4)

21 years agoDocument NOCDROM and NO_FLOPPIES.
John Baldwin [Fri, 15 Nov 2002 20:55:37 +0000 (20:55 +0000)]
Document NOCDROM and NO_FLOPPIES.

21 years agoAllow a person to specify NOCDROM to skip the cdrom.1 target during a
John Baldwin [Fri, 15 Nov 2002 20:51:27 +0000 (20:51 +0000)]
Allow a person to specify NOCDROM to skip the cdrom.1 target during a
release build.

PR: 45300
Submitted by: David Yeske <dyeske@yahoo.com>

21 years agoExplicitly add lang/perl5 to the disc1 packages. Practically speaking,
Bruce A. Mah [Fri, 15 Nov 2002 19:34:31 +0000 (19:34 +0000)]
Explicitly add lang/perl5 to the disc1 packages.  Practically speaking,
this is unnecessary, because some other port is likely to depend on it
anyways.  But just to be safe...

21 years agoBackout rev. 1.263.
Ruslan Ermilov [Fri, 15 Nov 2002 18:44:09 +0000 (18:44 +0000)]
Backout rev. 1.263.

Approved by: jmallett

21 years agoBack out rev 1.150; things are more complicated than this.
Sam Leffler [Fri, 15 Nov 2002 18:42:10 +0000 (18:42 +0000)]
Back out rev 1.150; things are more complicated than this.

21 years agoif_attach should not sleep; change malloc's M_WAITOK to M_NOWAIT
Sam Leffler [Fri, 15 Nov 2002 18:35:41 +0000 (18:35 +0000)]
if_attach should not sleep; change malloc's M_WAITOK to M_NOWAIT

21 years agoUpdate sparc64 installation instructions. Because sysinstall now
Bruce A. Mah [Fri, 15 Nov 2002 17:57:51 +0000 (17:57 +0000)]
Update sparc64 installation instructions.  Because sysinstall now
works on sparc64, this document can now use (almost) the same
procedures as i386, et al.  CDROM booting instructions are cut-and-pasted
from the older sparc64/install.sgml file, which is now unlinked from
the document build.

This document is a long ways from perfect, but it's better than shipping
instructions for an installation procedure that doesn't apply anymore.

While here, add some SGML comments to help others navigate the sources.

21 years agoCatch up with sam's changes to network interfaces.
Maxime Henrion [Fri, 15 Nov 2002 14:36:33 +0000 (14:36 +0000)]
Catch up with sam's changes to network interfaces.

21 years agoFixed style(9)
Yoshihiro Takahashi [Fri, 15 Nov 2002 13:24:29 +0000 (13:24 +0000)]
Fixed style(9)

21 years agoFix to build for pc98.
Yoshihiro Takahashi [Fri, 15 Nov 2002 13:18:41 +0000 (13:18 +0000)]
Fix to build for pc98.

21 years agoFixed two typos in comments.
Ruslan Ermilov [Fri, 15 Nov 2002 08:26:36 +0000 (08:26 +0000)]
Fixed two typos in comments.

21 years agoReturn EWOULDBLOCK for last thread in kse_release().
David Xu [Fri, 15 Nov 2002 00:53:59 +0000 (00:53 +0000)]
Return EWOULDBLOCK for last thread in kse_release().

Requested by: archie

21 years agoWhen about to do an execve(), don't reset the O_NONBLOCK flag on any file
Archie Cobbs [Fri, 15 Nov 2002 00:34:28 +0000 (00:34 +0000)]
When about to do an execve(), don't reset the O_NONBLOCK flag on any file
descriptors that have the close-on-exec flag set, as that will have no
effect anyway and might screw something else up if the file descriptor
happens to be shared with another process.

PR: standards/43335
MFC after: 1 week

21 years agoupdate vlandev description; vlan code now auto-recognizes devices that
Sam Leffler [Fri, 15 Nov 2002 00:31:49 +0000 (00:31 +0000)]
update vlandev description; vlan code now auto-recognizes devices that
support h/w tagging

21 years agodocument auto-recognition of device capabilities
Sam Leffler [Fri, 15 Nov 2002 00:25:33 +0000 (00:25 +0000)]
document auto-recognition of device capabilities

21 years agoo display new interface capability bits
Sam Leffler [Fri, 15 Nov 2002 00:02:21 +0000 (00:02 +0000)]
o display new interface capability bits
o capitilize capability bit strings for consistency

Approved by: re

21 years agonetwork interface and link layer changes:
Sam Leffler [Fri, 15 Nov 2002 00:00:15 +0000 (00:00 +0000)]
network interface and link layer changes:

o on input don't strip the Ethernet header from packets
o input packet handling is now done with if_input
o track changes to ether_ifattach/ether_ifdetach API
o track changes to bpf tapping
o call ether_ioctl for default handling of ioctl's
o use constants from net/ethernet.h where possible

Reviewed by: many
Approved by: re

21 years agotrack changes to ethernet input handling to no longer strip the Ethernet header
Sam Leffler [Thu, 14 Nov 2002 23:57:09 +0000 (23:57 +0000)]
track changes to ethernet input handling to no longer strip the Ethernet header

Reviewed by: many
Approved by: re

21 years agonetwork interface driver changes:
Sam Leffler [Thu, 14 Nov 2002 23:54:55 +0000 (23:54 +0000)]
network interface driver changes:

o don't strip the Ethernet header from inbound packets; pass packets
  up the stack intact (required significant changes to some drivers)
o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN)
o track ether_ifattach/ether_ifdetach API changes
o track bpf changes (use BPF_TAP and BPF_MTAP)
o track vlan changes (ifnet capabilities, revised processing scheme, etc.)
o use if_input to pass packets "up"
o call ether_ioctl for default handling of ioctls

Reviewed by: many
Approved by: re

21 years agoo track either_ifattach/ether_ifdetach API changes
Sam Leffler [Thu, 14 Nov 2002 23:49:09 +0000 (23:49 +0000)]
o track either_ifattach/ether_ifdetach API changes
o use if_input for input packet processing
o don't strip the Ethernet header for input packets
o use BPF_* macros bpf tapping
o call ether_ioctl to handle default ioctl case
o track vlan changes

Reviewed by: many
Approved by: re

21 years agotrack changes to not strip the Ethernet header from input packets
Sam Leffler [Thu, 14 Nov 2002 23:46:04 +0000 (23:46 +0000)]
track changes to not strip the Ethernet header from input packets

Reviewed by: many
Approved by: re

21 years agotrack bpf changes
Sam Leffler [Thu, 14 Nov 2002 23:45:13 +0000 (23:45 +0000)]
track bpf changes

Reviewed by: many
Approved by: re

21 years agoo track changes to ethernet input packet handling
Sam Leffler [Thu, 14 Nov 2002 23:44:37 +0000 (23:44 +0000)]
o track changes to ethernet input packet handling
o track changes to bpf
o track changes to make ng hooks more private

Reviewed by: many
Approved by: re

21 years agoo eliminate separate callback interface for h/w tagged input packets; instead
Sam Leffler [Thu, 14 Nov 2002 23:43:16 +0000 (23:43 +0000)]
o eliminate separate callback interface for h/w tagged input packets; instead
  drivers "tag packets" with an m_tag and the input packet handling recognizes
  such packets and does the right thing
o track the number of active vlans on an interface; this lets lots of places
  only do vlan-specific processing when needed
o track changes to ether_ifdetach/ether_ifattach
o track bpf changes
o eliminate the use of M_PROTO1 for communicating to drivers about tagged
  packets
o eliminate the use of IFF_LINK0 for drivers communicating to the vlan code
  that they support h/w tagging; replaced by explicit interface capabilities
o add ifnet capabilities for h/w tagging and support of "large mtu's"
o use new interface capabilities to auto-configure use of large mtu's and h/w
  tagging
o add support for proper handling of promiscuous mode
o document driver/vlan communication conventions

Reviewed by: many
Approved by: re

21 years agoo add if_nvlans member to track the number of vlans active on an interface
Sam Leffler [Thu, 14 Nov 2002 23:36:28 +0000 (23:36 +0000)]
o add if_nvlans member to track the number of vlans active on an interface
o add if_input member for interface drivers to call through to pass packets "up"
o remove ethernet-specific function decls (moved to ethernet.h)

Reviewed by: many
Approved by: re

21 years agoo change input packet handling to eliminate the pointer to the struct
Sam Leffler [Thu, 14 Nov 2002 23:35:06 +0000 (23:35 +0000)]
o change input packet handling to eliminate the pointer to the struct
  ether_header; instead drivers are to leave the Ethernet header at the
  front of the packet
o add declarations for netgraph and vlan hooks that were removed from ethernet.h
o change various in-file calling conventions to track change in input API
o fixup bridge support to handle Ethernet header no longer being stripped
o add consistency checks to ether_input to catch problems with the change
  in the API; some of these may want to be moved to #ifdef DIAGNOSTIC at a
  later time (though they are not too expensive to leave as is)
o change ether_demux to eliminate the passing of the Ethernet header; it is
  now expected at the front of the packet a la ether_input
o add ether_sprintf compatibility shim
o change ether_ifattach API to remove "bpf supported param" and add a pointer
  to the MAC address to be installed for the LL address (this is for future
  changes to divest struct arpcom from struct ifnet)
o change ether_ifdetach API to remove "bpf support param"

Reviewed by: many
Approved by: re

21 years agogeneral cleanups mostly aimed at improving portability of drivers
Sam Leffler [Thu, 14 Nov 2002 23:28:47 +0000 (23:28 +0000)]
general cleanups mostly aimed at improving portability of drivers

o ETHER_* (ETHER_ALIGN, ETHER_MAX_FRAME, ETHER_CRC_LEN, etc.)
o M_HASFCS for drivers to indicate packets include FCS
o remove global declarations for ng_ether* and vlan_* since these
  represent a private contract between the if_ethersubr.c code and
  certain parts of the system that should not normally be abused
o add ether_* declarations that were elsewhere
o remove ETHER_BPF_* since they are no longer used with the parameter
  no longer passed to ether_ifattach and ether_ifdetach

Reviewed by: many
Approved by: re

21 years agoo add support for multiple link types per interface (e.g. 802.11 and Ethernet)
Sam Leffler [Thu, 14 Nov 2002 23:24:13 +0000 (23:24 +0000)]
o add support for multiple link types per interface (e.g. 802.11 and Ethernet)
o introduce BPF_TAP and BPF_MTAP macros to hide implementation details and
  ease code portability
o use m_getcl where appropriate

Reviewed by: many
Approved by: re
Obtained from: NetBSD (multiple link type support)

21 years agoMake sure that we don't pick up installed packages on the host when
Bruce A. Mah [Thu, 14 Nov 2002 23:17:00 +0000 (23:17 +0000)]
Make sure that we don't pick up installed packages on the host when
we print dependencies (required by rev. 1.424 of ports/Mk/bsd.port.mk).

Reviewed (and partially submitted) by: kris

21 years agoo add IF_*bps macros for netbsd compatibility
Sam Leffler [Thu, 14 Nov 2002 23:16:18 +0000 (23:16 +0000)]
o add IF_*bps macros for netbsd compatibility
o add interface capabilities for vlan use and to signal jumbo frame support

Reviewed by: many
Approved by: re

21 years agoMFbed:
Udo Erdelhoff [Thu, 14 Nov 2002 22:24:08 +0000 (22:24 +0000)]
MFbed:
early-adopter/article.sgml: add lang=de
relnotes/common/new.sgml: 1.450 -> 1.451
relnotes/ia64/article.sgml: add lang=de

21 years agoAssume that packages passed on stdin are in bzip2 format, not gzip.
Bruce A. Mah [Thu, 14 Nov 2002 21:01:17 +0000 (21:01 +0000)]
Assume that packages passed on stdin are in bzip2 format, not gzip.
(sysinstall depends on this feature for package addition.)

Comment on hard-coded bzip2 usage in the spirit of rev. 1.58.

21 years agoRemove the remaining calls to free(), they are not needed anymore now
Olivier Houchard [Thu, 14 Nov 2002 19:54:33 +0000 (19:54 +0000)]
Remove the remaining calls to free(), they are not needed anymore now
device_get_softc() is used.

21 years agoMake dynamic PAM modules depend on dynamic PAM library.
Ruslan Ermilov [Thu, 14 Nov 2002 19:24:51 +0000 (19:24 +0000)]
Make dynamic PAM modules depend on dynamic PAM library.

Requested by: des, markm

21 years agoNew release note: SA-02:43.
Bruce A. Mah [Thu, 14 Nov 2002 17:59:11 +0000 (17:59 +0000)]
New release note:  SA-02:43.

21 years agoDefine `Sudden_Underflow' when compiling for the Alpha
Robert Drehmel [Thu, 14 Nov 2002 17:06:01 +0000 (17:06 +0000)]
Define `Sudden_Underflow' when compiling for the Alpha
architecture, mainly to avoid getting a SIGFPE signal sent
when calling strtod(3) with certain input.

The SIGFPE has been sent because the code was not aware that
a Gradual Underflow is handled in software via traps on the
Alpha architecture, but is not implemented in our Alpha kernel
layer.

With `Sudden_Underflow' defined, strtod(3) should not depend
on Gradual Underflow and adjust its calculations accordingly,
which means that other, more subtle errors than the sending of
SIGFPE could be solved by this.

Discussed with: bde
PR: alpha/12623
PR: alpha/17032
PR: alpha/43567
MFC after: 7 days

21 years agoMake the msg_size, msg_bufx and msg_bufr memebers of struct msgbuf
Thomas Moestl [Thu, 14 Nov 2002 16:11:12 +0000 (16:11 +0000)]
Make the msg_size, msg_bufx and msg_bufr memebers of struct msgbuf
signed, since they describe a ring buffer and signed arithmetic is
performed on them. This avoids some evilish casts.

Since this changes all but two members of this structure, style(9)
those remaining ones, too.

Requested by: bde
Reviewed by: bde (earlier version)