dragonfly.git
12 years agoMerge from vendor branch OPENSSH:
Peter Avalos [Sat, 8 Sep 2007 04:52:53 +0000 (04:52 +0000)]
Merge from vendor branch OPENSSH:
Import OpenSSH 4.7p1.

12 years agoImport OpenSSH 4.7p1.
Peter Avalos [Sat, 8 Sep 2007 04:52:53 +0000 (04:52 +0000)]
Import OpenSSH 4.7p1.

12 years agoYugoslavia split up into Serbia and Montenegro some time ago.
Simon Schubert [Sat, 8 Sep 2007 01:29:40 +0000 (01:29 +0000)]
Yugoslavia split up into Serbia and Montenegro some time ago.

12 years agoAllow comment lines in input files.
Simon Schubert [Sat, 8 Sep 2007 01:29:14 +0000 (01:29 +0000)]
Allow comment lines in input files.

12 years agoThe weekly script runs Saturday morning.
Sascha Wildner [Fri, 7 Sep 2007 23:24:20 +0000 (23:24 +0000)]
The weekly script runs Saturday morning.

Found-in: FreeBSD

12 years agoGeneral userfs fleshing out work. Abstract out construction and
Matthew Dillon [Fri, 7 Sep 2007 21:42:59 +0000 (21:42 +0000)]
General userfs fleshing out work.  Abstract out construction and
deconstruction functions and take advantage of the new getpages/putpages
support.

12 years agoRemove a bogus assertion. in_vm86call may have been set by some unrelated
Matthew Dillon [Fri, 7 Sep 2007 19:31:30 +0000 (19:31 +0000)]
Remove a bogus assertion.  in_vm86call may have been set by some unrelated
thread.  We have to check that we are in an emulated 8086 mode before we
can assert that the MP lock is held.

Reported-by: Rumko <rumcic@gmail.com>
12 years agoOops, forgot one .Dv
Sascha Wildner [Fri, 7 Sep 2007 08:14:57 +0000 (08:14 +0000)]
Oops, forgot one .Dv

12 years ago* Start sentences on a new line.
Sascha Wildner [Fri, 7 Sep 2007 08:01:29 +0000 (08:01 +0000)]
* Start sentences on a new line.

* Fix typo: depreciated -> deprecated

* Use .Dv for defined values.

12 years agoFix typo.
Peter Avalos [Fri, 7 Sep 2007 02:43:49 +0000 (02:43 +0000)]
Fix typo.

Reported-by: "Trevor Kendall" <trevorjkendall@gmail.com>
12 years agoUse >= instead of > to explicitly see which WARNS level we're dealing with.
Peter Avalos [Fri, 7 Sep 2007 00:09:02 +0000 (00:09 +0000)]
Use >= instead of > to explicitly see which WARNS level we're dealing with.

12 years agoUse bcd2bin().
Sascha Wildner [Thu, 6 Sep 2007 21:03:55 +0000 (21:03 +0000)]
Use bcd2bin().

12 years agoRemove redundant initialization.
Sascha Wildner [Thu, 6 Sep 2007 18:17:24 +0000 (18:17 +0000)]
Remove redundant initialization.

12 years agoUse .Dv
Sascha Wildner [Thu, 6 Sep 2007 11:29:30 +0000 (11:29 +0000)]
Use .Dv

12 years agoDocument some ATM media settings.
Sascha Wildner [Thu, 6 Sep 2007 11:11:19 +0000 (11:11 +0000)]
Document some ATM media settings.

12 years agoDescribe CCVER's default as gcc41.
Peter Avalos [Wed, 5 Sep 2007 22:11:01 +0000 (22:11 +0000)]
Describe CCVER's default as gcc41.

12 years agoAdd a build option (NO_GCC34) to prevent building gcc version 3.4.
Peter Avalos [Wed, 5 Sep 2007 21:40:10 +0000 (21:40 +0000)]
Add a build option (NO_GCC34) to prevent building gcc version 3.4.

This does not attempt to remove gcc-3.4, only to prevent it from building.

12 years agoBetter setpgid(2) documentation.
Hasso Tepper [Wed, 5 Sep 2007 08:23:43 +0000 (08:23 +0000)]
Better setpgid(2) documentation.

Obtained-from: NetBSD with modifications

12 years agoSet si_iosize_max to silence run-time warnings.
Matthew Dillon [Wed, 5 Sep 2007 05:28:32 +0000 (05:28 +0000)]
Set si_iosize_max to silence run-time warnings.

Reported-by: Vincent Stemen <vs1@crel.us>
12 years agoChange the virtual kernel's default hz to 20, because the kqueue timers we
Matthew Dillon [Tue, 4 Sep 2007 22:07:54 +0000 (22:07 +0000)]
Change the virtual kernel's default hz to 20, because the kqueue timers we
use can't really go much higher.  This fixes sleep intervals.

Reported-by: Multiple People
12 years agoUse gnuregex.h instead of a private header whose declarations are wrong.
Simon Schubert [Tue, 4 Sep 2007 17:08:38 +0000 (17:08 +0000)]
Use gnuregex.h instead of a private header whose declarations are wrong.

This private header declares the return type of re_search to be off_t,
i.e. 64 bit.  Our (old) gnu regex library however only returns ints.  As a
result, the generated opcode erroneously only looked at the sign of %edx,
which happened to contain a random value.

DragonFly-bug:  <http://bugs.dragonflybsd.org/issue801>
Reported-by: Vincent Stemen <vs1@crel.us>
12 years agoConsolidate all local CVS patches into the contrib sources.
Simon Schubert [Tue, 4 Sep 2007 16:53:57 +0000 (16:53 +0000)]
Consolidate all local CVS patches into the contrib sources.

Include file prefix ordering is a horror when patching header files,
so get over this once and for all and put the patches where they belong.

12 years agoForce an over-the-wire transaction when resolving the root of an NFS mount
Matthew Dillon [Tue, 4 Sep 2007 00:48:18 +0000 (00:48 +0000)]
Force an over-the-wire transaction when resolving the root of an NFS mount
point.  The namecache will cache the mount point for us so this should not
reduce performance.   We need to know that the mount point is good or
things like the namecache resolver could end up looping forever trying to
resolve a stale NFS mount.

Reported-by: elekktretterr@exemail.com.au (Petr)
12 years agoRemove options that don't make sense for a vkernel, and make SMP default.
Peter Avalos [Mon, 3 Sep 2007 19:24:53 +0000 (19:24 +0000)]
Remove options that don't make sense for a vkernel, and make SMP default.

12 years agoUpdate for libarchive 2.2.7.
Peter Avalos [Mon, 3 Sep 2007 18:48:12 +0000 (18:48 +0000)]
Update for libarchive 2.2.7.

12 years agoMerge from vendor branch LIBARCHIVE:
Peter Avalos [Mon, 3 Sep 2007 18:47:14 +0000 (18:47 +0000)]
Merge from vendor branch LIBARCHIVE:
Import libarchive-2.2.7.

This brings support for mtree files, gtar 1.17 sparse files, and
refines suid/sgid restore handling.

12 years agoImport libarchive-2.2.7.
Peter Avalos [Mon, 3 Sep 2007 18:47:14 +0000 (18:47 +0000)]
Import libarchive-2.2.7.

This brings support for mtree files, gtar 1.17 sparse files, and
refines suid/sgid restore handling.

12 years agoBring in FreeBSD/1.177 - fix a bug in a call made to useracc(). This
Matthew Dillon [Mon, 3 Sep 2007 17:32:32 +0000 (17:32 +0000)]
Bring in FreeBSD/1.177 - fix a bug in a call made to useracc().  This
permissions check is redundant since copyin/copyout also perform a
validation so there is no known security issue.

Reported-by: "Constantine A. Murenin" <cnst@FreeBSD.org>
Obtained-from: FreeBSD

12 years agoAdd a MNTK_ flag to the mount structure allowing a VFS to specify that
Matthew Dillon [Mon, 3 Sep 2007 17:06:24 +0000 (17:06 +0000)]
Add a MNTK_ flag to the mount structure allowing a VFS to specify that
no submounts under the VFS are to be allowed.  Adjust procfs and linprocfs
to use the feature.

Submitted-by: "Nicolas Thery" <nthery@gmail.com>
12 years agoUse .Pa
Sascha Wildner [Sun, 2 Sep 2007 19:30:48 +0000 (19:30 +0000)]
Use .Pa

12 years agoSwitch ipfw from ipfw1 to ipfw2.
Sepherosa Ziehau [Sun, 2 Sep 2007 13:27:23 +0000 (13:27 +0000)]
Switch ipfw from ipfw1 to ipfw2.

Approved-by: dillon@
Submitted-by: Gary Allan <dragonfly@gallan.plus.com> (w/ modification)
12 years agoTAILQ_FOREACH to enhance readability.
Peter Avalos [Sat, 1 Sep 2007 20:43:24 +0000 (20:43 +0000)]
TAILQ_FOREACH to enhance readability.

12 years agoSync with OpenBSD.
Peter Avalos [Sat, 1 Sep 2007 18:42:08 +0000 (18:42 +0000)]
Sync with OpenBSD.

- Save and restore not only scale, but ibase as well.

- Interchange roles: bc becomes parent, dc child. This leads to saner
tty handling.

- Avoid segv when a syntax error occurs while reading from command line.

- Various whitespace, warns cleanup, etc.

12 years agoHandle baudrate requests algorithmically with newer chips (not old SIO),
Hasso Tepper [Fri, 31 Aug 2007 13:39:35 +0000 (13:39 +0000)]
Handle baudrate requests algorithmically with newer chips (not old SIO),
allowing all sorts of strange rates one might need.

Obtained-from: OpenBSD

12 years agoSimplify the way how chip type is determined. Instead of managing insane
Hasso Tepper [Fri, 31 Aug 2007 13:25:32 +0000 (13:25 +0000)]
Simplify the way how chip type is determined. Instead of managing insane
case constructs with info what vendor/product ids combinations have which
chip, use the info device descriptors provide.

  - FT2232 dual UART chips have bNumInterfaces set to 2 (obviously).
  - Original SIO have bcdDevice set to 0x0100.

12 years agoGC even more remains of the recent old driver removal.
Sascha Wildner [Fri, 31 Aug 2007 11:26:44 +0000 (11:26 +0000)]
GC even more remains of the recent old driver removal.

12 years agoRemove leftover nv(4) remains.
Sascha Wildner [Fri, 31 Aug 2007 10:52:03 +0000 (10:52 +0000)]
Remove leftover nv(4) remains.

12 years agoSync with FreeBSD and add an entry for git.
Peter Avalos [Thu, 30 Aug 2007 22:52:39 +0000 (22:52 +0000)]
Sync with FreeBSD and add an entry for git.

12 years agoManual page update for waitpid's new WCONTINUED option and the
Peter Avalos [Thu, 30 Aug 2007 20:58:35 +0000 (20:58 +0000)]
Manual page update for waitpid's new WCONTINUED option and the
WIFCONTINUED() macro.

Obtained-from: FreeBSD

12 years agoAdd WCONTINUED option for waitpid(2).
Peter Avalos [Thu, 30 Aug 2007 20:41:00 +0000 (20:41 +0000)]
Add WCONTINUED option for waitpid(2).

P_CONTINUED is set when a stopped process receives a SIGCONT and
cleared after it has notified a parent process that has requested
notification via waitpid(2) with WCONTINUED specified in its options
operand.  The status value can be checked with the new WIFCONTINUED()
macro.

Obtained-from: FreeBSD, OpenBSD
Reviewed-by: corecode
12 years agoRemove leftover .Dx and add a reference to setcontext(3).
Sascha Wildner [Thu, 30 Aug 2007 17:47:08 +0000 (17:47 +0000)]
Remove leftover .Dx and add a reference to setcontext(3).

Submitted-by: Francis Gudin <fgudin@nerim.net>
12 years agoFix LINT: Remove cm(4) and fla(4) remains.
Sascha Wildner [Wed, 29 Aug 2007 15:42:05 +0000 (15:42 +0000)]
Fix LINT: Remove cm(4) and fla(4) remains.

12 years agoDon't use \*[Px] since it will change the font size permanently.
Sascha Wildner [Wed, 29 Aug 2007 12:02:16 +0000 (12:02 +0000)]
Don't use \*[Px] since it will change the font size permanently.
Use .Tn POSIX instead.

12 years agoSync with FreeBSD.
Peter Avalos [Tue, 28 Aug 2007 15:36:44 +0000 (15:36 +0000)]
Sync with FreeBSD.

Make it possible to have this (basename(1)) perform basename(3) on multiple
files.  The traditional behaviour, 'basename string .suffix', is preserved,
however a suffix may now also be specified via a getopt(3) option, -s, such
that if it is specified in that way, all string arguments follow.  There is
also a new flag, -a, which allows a user to say "yes, please basename(3) on
all arguments".

Add EXIT STATUS.

Cross-reference basename(3) and dirname(3).

Obtained-from:  FreeBSD

12 years agoFix a bug in vnode_pager_generic_getpages(). This function was improperly
Matthew Dillon [Tue, 28 Aug 2007 01:09:24 +0000 (01:09 +0000)]
Fix a bug in vnode_pager_generic_getpages().  This function was improperly
setting m->valid to 0 and was also improperly trying to free the page after
it had potentially become wired by the buffer cache.

Add a sysctl to UFS that allows us to force it to call vop_stdgetpages()
for debugging purposes.

12 years agoAdd vop_stdgetpages() and vop_stdputpages() and replace those filesystem
Matthew Dillon [Tue, 28 Aug 2007 01:04:33 +0000 (01:04 +0000)]
Add vop_stdgetpages() and vop_stdputpages() and replace those filesystem
getpages and putpages routines which were doing the same thing.

12 years agoMissed this file in previous commit.
Hasso Tepper [Mon, 27 Aug 2007 22:03:19 +0000 (22:03 +0000)]
Missed this file in previous commit.

12 years agoNuke nv(4), we have nfe(4) which replaces it.
Hasso Tepper [Mon, 27 Aug 2007 21:51:41 +0000 (21:51 +0000)]
Nuke nv(4), we have nfe(4) which replaces it.

12 years agoRemove fla(4) manpage during upgrade.
Hasso Tepper [Mon, 27 Aug 2007 20:51:11 +0000 (20:51 +0000)]
Remove fla(4) manpage during upgrade.

12 years agoNuke fla(4). It's known to be buggy, supports very limited set of obsolete
Hasso Tepper [Mon, 27 Aug 2007 20:40:53 +0000 (20:40 +0000)]
Nuke fla(4). It's known to be buggy, supports very limited set of obsolete
devices with limited set of firmware versions, is removed from FreeBSD more
than 3 years ago and is not supported any more more than 4 years:

http://www.daemonnews.org/mailinglists/FreeBSD/freebsd-small/msg03526.html

And it's based on blob.

12 years agoRemove remainings of the oltr(4).
Hasso Tepper [Mon, 27 Aug 2007 20:08:55 +0000 (20:08 +0000)]
Remove remainings of the oltr(4).

12 years agoNuke ARCnet, Token Ring and FDDI related headers and manpages during upgrade.
Hasso Tepper [Mon, 27 Aug 2007 17:47:15 +0000 (17:47 +0000)]
Nuke ARCnet, Token Ring and FDDI related headers and manpages during upgrade.

12 years agoPull WARNS6 into usr.bin/Makefile.inc.
Peter Avalos [Mon, 27 Aug 2007 16:51:02 +0000 (16:51 +0000)]
Pull WARNS6 into usr.bin/Makefile.inc.

Most of usr.bin is marked as WARNS6, so add exceptions in individual
Makefiles.  Also, remove any warning flags from CFLAGS in those Makefiles.

12 years agoNuke FDDI support.
Hasso Tepper [Mon, 27 Aug 2007 16:15:42 +0000 (16:15 +0000)]
Nuke FDDI support.

12 years agoNuke token ring support. This also means one blob less in DragonFly.
Hasso Tepper [Mon, 27 Aug 2007 14:56:00 +0000 (14:56 +0000)]
Nuke token ring support. This also means one blob less in DragonFly.

12 years agoNuke ARCnet support.
Hasso Tepper [Mon, 27 Aug 2007 13:15:14 +0000 (13:15 +0000)]
Nuke ARCnet support.

12 years agoCorrect off-by-one error.
Simon Schubert [Sun, 26 Aug 2007 20:40:55 +0000 (20:40 +0000)]
Correct off-by-one error.

When comparing the command names, we might wind up behind the array,
later dereferencing a dangling pointer.  Terminate the array correctly
and stop processing soon enough.  Additionally, correct the error
output to use errx(), as there is no meaningful errno available.

Differing behavior spotted by use of tinycc.

12 years agoGC unused sources.
Peter Avalos [Sun, 26 Aug 2007 20:27:36 +0000 (20:27 +0000)]
GC unused sources.

12 years agoUse the argument as a file name only if no section was specified.
Sascha Wildner [Sun, 26 Aug 2007 19:42:26 +0000 (19:42 +0000)]
Use the argument as a file name only if no section was specified.

12 years agoRe-order alphabetically.
Peter Avalos [Sun, 26 Aug 2007 18:20:49 +0000 (18:20 +0000)]
Re-order alphabetically.

12 years agoFix a misleading error message when the device specified at the mountroot
Matthew Dillon [Sun, 26 Aug 2007 16:54:51 +0000 (16:54 +0000)]
Fix a misleading error message when the device specified at the mountroot
prompt cannot be found.

Reported-by: "Simon 'corecode' Schubert" <corecode@fs.ei.tum.de>
Submitted-by: Nicolas Thery <nthery@gmail.com>
12 years agoThe function's name should begin a new line.
Peter Avalos [Sun, 26 Aug 2007 16:52:01 +0000 (16:52 +0000)]
The function's name should begin a new line.

12 years agoCheck that -w width is not above maximum. Use DWIDTH instead of hardcoded
Peter Avalos [Sun, 26 Aug 2007 16:44:32 +0000 (16:44 +0000)]
Check that -w width is not above maximum.  Use DWIDTH instead of hardcoded
constant.

Obtained-from: FreeBSD

12 years agoDocument message length limit.
Peter Avalos [Sun, 26 Aug 2007 16:41:54 +0000 (16:41 +0000)]
Document message length limit.

Obtained-from:  FreeBSD

12 years agoFix a bugaboo in the last commit. Pages are hard-busied for getpages,
Matthew Dillon [Sun, 26 Aug 2007 16:22:31 +0000 (16:22 +0000)]
Fix a bugaboo in the last commit.  Pages are hard-busied for getpages,
but only soft-busied for putpages.

Reported-by: Sascha Wildner <saw@online.de>
12 years agoSync with FreeBSD:
Peter Avalos [Sun, 26 Aug 2007 16:12:27 +0000 (16:12 +0000)]
Sync with FreeBSD:

- mktime() already does daylight savings.

- Explain the format of the at.allow and at.deny files in detail.

- mdoc improvements.

- Use MAXLOGNAME - 1 instead of UT_NAMESIZE.

- Do not pretend there is something like '/etc/utmp'.

- Add the SUSv3 -l option to at.  This is an alias for atq. Allow job ids
to be specified on the command line for which information should be
reported.

- Document the fact that at(1) is run through cron(8).  This affects the
granularity of the at queue runs.

- Organize #includes.

- Remove the -V [version number] option.

- Implement the POSIX 1003.1-2001 -r and -t options in at(1).

- Warns cleanup and diff reduction from FreeBSD.

- Remove some unsafe function calls from the signal handlers.

- Fix privileges macros.

Obtained-from: FreeBSD

12 years agoCompile in a variable default search prefix.
Simon Schubert [Sun, 26 Aug 2007 08:15:50 +0000 (08:15 +0000)]
Compile in a variable default search prefix.

Up to now, even a objdump created in the ctools phase used the standard
search path ("/") instead of using the correct prefix, unless invoked
with OBJFORMAT_PREFIX set.  Use TOOLS_PREFIX as a default search prefix to
make the ctools phase objdump work as expected.

12 years agoKmods always need their headers in place.
Simon Schubert [Sun, 26 Aug 2007 07:53:25 +0000 (07:53 +0000)]
Kmods always need their headers in place.

Now you can build a kmod without having to make depend first!

12 years agoFix a bug in linprocfs and hold lwps to prevent a race condition.
Simon Schubert [Sat, 25 Aug 2007 23:27:02 +0000 (23:27 +0000)]
Fix a bug in linprocfs and hold lwps to prevent a race condition.

Linprocfs rolls its own prototype for procfs_domem.  The function was
changed to take a struct lwp *, but neither the private prototype, nor
the procfs_domem call were changed accordingly.  As a result any access
to the linprocfs process memory file resulted in a panic.

Procfs and linprocs perform potentially blocking calls, so HOLD and
RELE the lwps so that they can't be destroyed while procfs is blocked.

12 years agoPrint the msgbuf even if using a kernel without symbols instead of aborting.
Simon Schubert [Sat, 25 Aug 2007 21:59:05 +0000 (21:59 +0000)]
Print the msgbuf even if using a kernel without symbols instead of aborting.

12 years agoAdd a README to include a cvs import hint.
Peter Avalos [Sat, 25 Aug 2007 21:09:56 +0000 (21:09 +0000)]
Add a README to include a cvs import hint.

12 years agoMerge from vendor branch TNFTP:
Peter Avalos [Sat, 25 Aug 2007 21:05:27 +0000 (21:05 +0000)]
Merge from vendor branch TNFTP:
Improve parsing of chunked transfer chunks per RFC2616:
* more stringent chunk-size parsing
* ignore optional trailing ';chunk-ext' stuff, instead of barfing
* detect EOF before final \r\n.

12 years agoImprove parsing of chunked transfer chunks per RFC2616:
Peter Avalos [Sat, 25 Aug 2007 21:05:27 +0000 (21:05 +0000)]
Improve parsing of chunked transfer chunks per RFC2616:
* more stringent chunk-size parsing
* ignore optional trailing ';chunk-ext' stuff, instead of barfing
* detect EOF before final \r\n.

12 years agoRework generating of files so that every generation tool will only run once.
Simon Schubert [Sat, 25 Aug 2007 15:29:29 +0000 (15:29 +0000)]
Rework generating of files so that every generation tool will only run once.

Before, we generated a file over and over again where it was needed by the
build process, instead of generating the files once and reusing them.
This commit changes this behavior and makes the build match more gcc's own
build way, which is quite optimized.

Additionally, use move-if-change, to avoid rebuilds if the generated files
didn't change.

12 years agoDon't hang `all' off `_EXTRADEPEND', but rather on `afterdepend'.
Simon Schubert [Sat, 25 Aug 2007 15:22:30 +0000 (15:22 +0000)]
Don't hang `all' off `_EXTRADEPEND', but rather on `afterdepend'.

The problem with using _EXTRADEPEND is that ${DEPENDFILE} depends on
_EXTRADEPEND, which means it will depend on the phony `all' source which
is always out of date.  If we make `all' a source of `afterdepend',
the .depend file won't be regernerated every time.

12 years agomdoc makeover, no content changes
Sascha Wildner [Sat, 25 Aug 2007 10:38:53 +0000 (10:38 +0000)]
mdoc makeover, no content changes

12 years agoComment out empty section.
Sascha Wildner [Sat, 25 Aug 2007 10:01:34 +0000 (10:01 +0000)]
Comment out empty section.

12 years agoOutput more detail when reporting errors. Use err() instead of warn();exit();
Simon Schubert [Sat, 25 Aug 2007 08:55:26 +0000 (08:55 +0000)]
Output more detail when reporting errors.  Use err() instead of warn();exit();

12 years agoAdd T2300 cpu support to EST.
Matthew Dillon [Fri, 24 Aug 2007 16:08:05 +0000 (16:08 +0000)]
Add T2300 cpu support to EST.

Submitted-by: Andrew Atrens <atrens@nortel.com>
Taken-from: FreeBSD and NetBSD

12 years agopart2 ppX fix for altq/pf.
Matthew Dillon [Fri, 24 Aug 2007 16:06:37 +0000 (16:06 +0000)]
part2 ppX fix for altq/pf.

Submitted-by: Andrew Atrens <atrens@nortel.com>
12 years agoAdd geode companion support to ohci probe code
Matthew Dillon [Fri, 24 Aug 2007 16:02:47 +0000 (16:02 +0000)]
Add geode companion support to ohci probe code

Submitted-by: Andrew Atrens <atrens@nortel.com>
12 years agoAdd missing if_softc assignment, allowing pppX interfaces to
Matthew Dillon [Fri, 24 Aug 2007 16:00:52 +0000 (16:00 +0000)]
Add missing if_softc assignment, allowing pppX interfaces to
work with altq/pf.

Submitted-by: Andrew Atrens <atrens@nortel.com>
12 years agoAdd missing xfer->state assignment.
Matthew Dillon [Fri, 24 Aug 2007 15:58:33 +0000 (15:58 +0000)]
Add missing xfer->state assignment.

Obtained-from: FreeBSD
Submitted-by: Andrew Atrens <atrens@nortel.com>
12 years agoDefine TARGET_PLATFORM matching TARGET_ARCH.
Simon Schubert [Thu, 23 Aug 2007 07:00:31 +0000 (07:00 +0000)]
Define TARGET_PLATFORM matching TARGET_ARCH.

This way we don't have to specify TARGET_ARCH *and* TARGET_PLATFORM for
crossbuilds.  Of course this should be generalized, etc, but as we just
have two platforms for now, this is no problem yet.

12 years agoDrop reference to non-existing variable.
Simon Schubert [Thu, 23 Aug 2007 06:57:09 +0000 (06:57 +0000)]
Drop reference to non-existing variable.

12 years agoDefine MACHINE et al like this header is supposed to do.
Simon Schubert [Thu, 23 Aug 2007 06:56:27 +0000 (06:56 +0000)]
Define MACHINE et al like this header is supposed to do.

12 years agoDon't define USED_FOR_TARGET on the command line.
Simon Schubert [Thu, 23 Aug 2007 06:52:02 +0000 (06:52 +0000)]
Don't define USED_FOR_TARGET on the command line.

This produces all kinds of side effects, ultimately leading to the csu
sources and the compiler being out of sync on compilation, thus triggering
an internal compiler error.

Details for the curious and for historical purposes:
Our gcc is configured to work with a ld which can merge ro and rw sections
to rw sections.  Nevertheless cc1 will enforce sections in one unit to be
of one type or the other, i.e. it disallows section flag changes.  Sections
are usually a compiler internal detail, so a flag change is considered a
compiler bug and thus cc1 aborts with an internal compiler error on such
occasions.

The csu libs consider themselves as part of the compiler and use section
declarations.  However if the lib and the compiler don't agree on some
settings and are out of sync, the lib will decide to create a rw .eh_frame
section (by not declaring a variable constant).  If the compiler uses the
.eh_frame section as well, like gcc41/amd64 does to store dwarf2 exception
rollback information, the above mentioned internal compiler error is
triggered.

The essential problem here was that USED_FOR_TARGET will prevent the
auto-host.h/dragonfly-native.h config.h-style headers to set the necessary
compiler configuration which is also used by the csu libs.  By moving it
to the place where the gcc people intended it to be, everything works as
desired.

12 years agoDefine USED_FOR_TARGET in tconfig.h, matching what gcc does.
Simon Schubert [Thu, 23 Aug 2007 06:22:10 +0000 (06:22 +0000)]
Define USED_FOR_TARGET in tconfig.h, matching what gcc does.

12 years agoChange gcc41 search path setup to match gcc34.
Simon Schubert [Thu, 23 Aug 2007 06:19:20 +0000 (06:19 +0000)]
Change gcc41 search path setup to match gcc34.

We need this modification to make the cross compiler use the correct
search paths.  I would prefer a solution without having to patch the
sources, but for now it is working and even reduces complexity in gcc.

12 years agoUse the i486 headers for amd64.
Simon Schubert [Thu, 23 Aug 2007 06:13:30 +0000 (06:13 +0000)]
Use the i486 headers for amd64.

12 years agoAdd usbdi(9) manpage.
Hasso Tepper [Wed, 22 Aug 2007 21:35:59 +0000 (21:35 +0000)]
Add usbdi(9) manpage.

Obtained-from: FreeBSD

12 years agoTemporary hack up a fix for msdos which avoids a deadlock between the VM
Matthew Dillon [Wed, 22 Aug 2007 16:56:52 +0000 (16:56 +0000)]
Temporary hack up a fix for msdos which avoids a deadlock between the VM
subsystem and the buffer cache subsystem.

Reported-by: "Andrew Atrens" <atrens@nortel.com>
12 years agoFix for CVE-2006-7180
Sepherosa Ziehau [Wed, 22 Aug 2007 13:41:10 +0000 (13:41 +0000)]
Fix for CVE-2006-7180

Obtained-from: MadWifi via Sam Leffler

12 years ago- Add ic_headroom through which drivers can inform 802.11 layer to reserve
Sepherosa Ziehau [Wed, 22 Aug 2007 13:24:44 +0000 (13:24 +0000)]
- Add ic_headroom through which drivers can inform 802.11 layer to reserve
  some space for device specific TX header when encapsulate 802.11 frames.
- Add another parameter to ieee80211_getmgtframe(), so it knows how many
  header space it should reserve.

Obtained-from: FreeBSD (sam@freebsd.org) w/ bug fix

12 years agoMerge from vendor branch GDB:
Simon Schubert [Wed, 22 Aug 2007 08:50:26 +0000 (08:50 +0000)]
Merge from vendor branch GDB:
Drop the stupid patch infrastructure and merge patches into contrib.

I'm not willing to spend several minutes to hours trying to figure out
how to do makefile magic to get this patch infrastructure to do what
I want, and only that what I want.

Instead, use CVS vendor branches like they are supposed to be used.

This unbreaks world and makes gdb compilable for i386 and amd64.

12 years agoDrop the stupid patch infrastructure and merge patches into contrib.
Simon Schubert [Wed, 22 Aug 2007 08:50:26 +0000 (08:50 +0000)]
Drop the stupid patch infrastructure and merge patches into contrib.

I'm not willing to spend several minutes to hours trying to figure out
how to do makefile magic to get this patch infrastructure to do what
I want, and only that what I want.

Instead, use CVS vendor branches like they are supposed to be used.

This unbreaks world and makes gdb compilable for i386 and amd64.

12 years agoDrop the stupid patch infrastructure and merge patches into contrib.
Simon Schubert [Wed, 22 Aug 2007 08:50:26 +0000 (08:50 +0000)]
Drop the stupid patch infrastructure and merge patches into contrib.

I'm not willing to spend several minutes to hours trying to figure out
how to do makefile magic to get this patch infrastructure to do what
I want, and only that what I want.

Instead, use CVS vendor branches like they are supposed to be used.

This unbreaks world and makes gdb compilable for i386 and amd64.

12 years agoDefine macros also for amd64.
Simon Schubert [Tue, 21 Aug 2007 20:04:09 +0000 (20:04 +0000)]
Define macros also for amd64.

On-behalf-of:  Noah Yan <noah.yan@gmail.com>
Submitted-by: Noah Yan <noah.yan@gmail.com>
12 years agoFix size types.
Simon Schubert [Tue, 21 Aug 2007 20:03:20 +0000 (20:03 +0000)]
Fix size types.

On-behalf-of:  Noah Yan <noah.yan@gmail.com>
Submitted-by: Noah Yan <noah.yan@gmail.com>
12 years agoDon't include the kernel target. This is handled in kgdb separately.
Simon Schubert [Tue, 21 Aug 2007 19:55:23 +0000 (19:55 +0000)]
Don't include the kernel target.  This is handled in kgdb separately.

On-behalf-of:  Noah Yan <noah.yan@gmail.com>
Submitted-by: Noah Yan <noah.yan@gmail.com>