dragonfly.git
14 years agoMerge from vendor branch OPENSSH:
Joerg Sonnenberger [Tue, 6 Sep 2005 10:51:17 +0000 (10:51 +0000)]
Merge from vendor branch OPENSSH:
Welcome OpenSSH 4.2.

14 years agoWelcome OpenSSH 4.2.
Joerg Sonnenberger [Tue, 6 Sep 2005 10:51:17 +0000 (10:51 +0000)]
Welcome OpenSSH 4.2.

14 years agoWelcome OpenSSH 4.2.
Joerg Sonnenberger [Tue, 6 Sep 2005 10:51:17 +0000 (10:51 +0000)]
Welcome OpenSSH 4.2.

14 years agomergemaster -> make upgrade
Sascha Wildner [Tue, 6 Sep 2005 10:11:53 +0000 (10:11 +0000)]
mergemaster -> make upgrade

14 years ago- use pci_{get, set}_powerstate() instead of reinventing the wheel in dc_acpi()
Sepherosa Ziehau [Tue, 6 Sep 2005 08:06:13 +0000 (08:06 +0000)]
- use pci_{get, set}_powerstate() instead of reinventing the wheel in dc_acpi()
- remove unused DC_PSTATE_D?

14 years agoRework and expand the algorithms in JSCAN, part 1/2. Implement a new
Matthew Dillon [Tue, 6 Sep 2005 06:42:44 +0000 (06:42 +0000)]
Rework and expand the algorithms in JSCAN, part 1/2.  Implement a new
infrastructure in the userland jscan program to create a framework for
supporting the following:

* Pipe-throughs (without storing data in a buffering file).
* File-based buffering
* Streamed replication of journaled data.
* Sequenced output files limited to a specific size (to match against
  backup media)
* Transaction id tracking to support journal restarts and resynchronization.
  This is needed to make off-site and networked backups reliable.
* Decoupled journal operations.  e.g. (once finished) we will be able to run
  a jscan in realtime to record journal data, and another one in batch to
  take that data and use it to maintain a mirror or to send it to another
  machine in a restartable and resynchronizable manner.
* Infrastructure to be able to scan a journal forwards or backwards by
  an arbitrary number of raw records and maintain a mirror in both
  the forwards (redo) and backwards (undo) direction based on that ability.
* Infrastructure that will aid in the preferred backup methodology of
  maintaining a real time mirror and UNDO/REDO journal on a 100% full
  backup partition such that one is able to free space on the partition
  simply by removing the oldest journaling file.
* Infrastructure that will make it easier to solve the overlapping
  meta-transaction problem illustrated below:

      current synchronization point
                     V
  [ ][ ][ ][ ][ ][ ][ ][ ][ ][ ][ ][ ] <= Raw records
   A      <= meta transaction A
      B        B     B B <= meta transaction B
                  C              C  C <= meta transaction C
                           D  D         <= meta transaction D
 E  E                           <= meta transaction E

  As you can see, the point at which a journal is 'broken' could
  be right smack in the middle of several meta transactions at
  once, due to the fact that meta transactions are laid out as
  their records can be generated and multiple processes could be
  modifying the filesystem at the same time.

  If we are attempting to mirror this data, only transactions A and E
  will have been completed and issued to the mirror.  If the journal
  is broken at the sychnronization point and then restarted, in
  order to pick up where we left off we have to scan existing
  journal records all the way back to the first 'B' in order to
  recover transactions B and C, then continue recording the journal
  from the synchronization point until those transactions are complete
  and can be issued to the mirror.  The difficulty here is that we
  have to carefully track exactly what needs to be synchronized.
  In recovering transaction 'B' and 'C' we do not want to reissue
  transaction E, for example, which had already been completed prior to
  the break.

  The problem exists in both the forwards and reverse direction, and
  we need to be able to go in both directions to control the 'snapshot'
  view the mirror is presenting to us.

NOTE: this is just part 1/2.  This code and the new options are
NON WORKING at the moment.

14 years agoremove nonexistent kernel tunable parameters' statments.
Sepherosa Ziehau [Tue, 6 Sep 2005 02:34:20 +0000 (02:34 +0000)]
remove nonexistent kernel tunable parameters' statments.

14 years agoDocument the dscheck(9) function and explain how it affects the slice-
Hiten Pandya [Tue, 6 Sep 2005 01:21:26 +0000 (01:21 +0000)]
Document the dscheck(9) function and explain how it affects the slice-
and driver-relative block numbers, bio->bio_blkno and bio->bio_pblkno
respectively.

The latter is modified by the said function after doing various checks
on available information, such as disklabels, etc, while the former is
provided by the VFS.

Explain how said function re-transforms the block number fields when
LABEL support is enabled in the VN driver via vnconfig(8).

14 years agoTake back last commit. It's about normal vs. context diff, not
Sascha Wildner [Tue, 6 Sep 2005 00:43:16 +0000 (00:43 +0000)]
Take back last commit. It's about normal vs. context diff, not
context vs. unified (as i was wrongly assuming). Apologies.

14 years agoRemove false statement. Rejected hunks will come out in whatever format
Sascha Wildner [Tue, 6 Sep 2005 00:20:09 +0000 (00:20 +0000)]
Remove false statement. Rejected hunks will come out in whatever format
the original patch was in. OpenBSD seems to have missed that.

14 years agoAdd gcc40 build hooks. Gcc40 isn't built per default, you will have to
Simon Schubert [Mon, 5 Sep 2005 15:36:34 +0000 (15:36 +0000)]
Add gcc40 build hooks.  Gcc40 isn't built per default, you will have to
enable it by adding WANT_GCC40=yes to make.conf.

14 years agoFreeBSD -> DragonFly
Sascha Wildner [Mon, 5 Sep 2005 04:52:13 +0000 (04:52 +0000)]
FreeBSD -> DragonFly

14 years agoRemove old wtmp conversion tool.
Sascha Wildner [Mon, 5 Sep 2005 04:50:43 +0000 (04:50 +0000)]
Remove old wtmp conversion tool.

14 years ago* Constify and initialize potentially uninitialized variable.
Sascha Wildner [Mon, 5 Sep 2005 04:22:07 +0000 (04:22 +0000)]
* Constify and initialize potentially uninitialized variable.

* Remove unnecessary casts to caddr_t.

* Remove extra newline at top of function body.

* Bump WARNS to 6.

14 years agoCleanup:
Sascha Wildner [Mon, 5 Sep 2005 04:02:43 +0000 (04:02 +0000)]
Cleanup:

* Remove unnecessary casts to void and caddr_t.

* Remove hack to suppress GCC warnings (there aren't any warnings,
  anyway).

* Remove double CVS tags.

* Remove extra newline at top of function bodies.

14 years ago* Remove parameter names in function declarations.
Sascha Wildner [Mon, 5 Sep 2005 02:43:10 +0000 (02:43 +0000)]
* Remove parameter names in function declarations.

* Remove unnecessary void and caddr_t casts.

14 years agoK&R style function declarations -> ANSI C style.
Sascha Wildner [Sun, 4 Sep 2005 23:19:12 +0000 (23:19 +0000)]
K&R style function declarations -> ANSI C style.

Submitted by: Sarunas Vancevicius <svan@redbrick.dcu.ie>

14 years agoFix loader.rc path.
Sascha Wildner [Sun, 4 Sep 2005 12:20:46 +0000 (12:20 +0000)]
Fix loader.rc path.

14 years agoAvoid panic by not using linker_* API in the early stage of
YONETANI Tomokazu [Sun, 4 Sep 2005 06:58:24 +0000 (06:58 +0000)]
Avoid panic by not using linker_* API in the early stage of
boot process, during which namei_zone is not initialized yet
but linker_* API relies on it indirectly by calling nlookup_init().

14 years ago1.101 Update code comments.
Matthew Dillon [Sun, 4 Sep 2005 05:16:59 +0000 (05:16 +0000)]
1.101   Update code comments.

1.100   Update code comments.

1.99    DEFERRED
1.98    DEFERRED

1.97    Timeout while polling is in ms not ticks.  [This should be
        computationally equivalent to the code that was there before, but
        is far simpler].

1.96    DEFERRED - Abort race.  It is unclear whether an xfer can be
                   aborted multiple times.

1.95    DEFERRED (interrupt fixes that may still be buggy)
1.94    DEFERRED (interrupt fixes that may still be buggy)
1.93    DEFERRED (interrupt fixes that may still be buggy)
1.92    DEFERRED (interrupt fixes that may still be buggy)

1.91    NOT APPLICABLE (whitespace cleanup)

1.90    WILL NOT BE APPLIED - This rev tries to setup some sort of
        memory reserve but doesn't really solve any problems.  A
        reserve is useless if it is not operated as a circular
        resource-limited pipeline.

1.89    Don't error out if the number of companion controllers is wrong, just
        complain and go ahead.

14 years agoMerge the following revs from NetBSD (in an attempt to bring new material
Matthew Dillon [Sun, 4 Sep 2005 04:46:12 +0000 (04:46 +0000)]
Merge the following revs from NetBSD (in an attempt to bring new material
in in a more piecemeal fashion):

1.105   NOT APPLICABLE TO DRAGONFLY.  This rev moved interrupt enablement
to near the end of the init.  DragonFly already goes even further
and moves interrupt enablement into the PCI module.

1.104   Constification and variable naming cleanup.

1.103   Make sure all physical addresses are filled in the qTD even when we have
        a page offset.

1.102   Make sure to honor the USBD_FORCE_SHORT_XFER flag and generate an empty
        transfer when necessary.

14 years agoRevert the very last commit to ehci.c (1.12). It was locking the system
Matthew Dillon [Sun, 4 Sep 2005 04:27:57 +0000 (04:27 +0000)]
Revert the very last commit to ehci.c (1.12).  It was locking the system
up on boot.  The other portions of the Jeroen's patch set remain intact.

14 years agoCleanup the module build and conditionalize a goto label.
Matthew Dillon [Sun, 4 Sep 2005 04:06:28 +0000 (04:06 +0000)]
Cleanup the module build and conditionalize a goto label.

14 years agotcpd.h:
Sepherosa Ziehau [Sun, 4 Sep 2005 01:53:07 +0000 (01:53 +0000)]
tcpd.h:
- properly declare functions
  Based-on: NetBSD lib/libwarp/tcpd.h rev1.2
- wrap declarations inside __{BEGIN, END}_DECLS
  Obtained-from: NetBSD lib/libwrap/tcpd.h rev1.11

scaffold.c:
- remove dummy rfc931() definition, which conflits its declaration in tcpd.h
  and is not necessary at all
  Based-on: NetBSD usr.sbin/tcpchk/scaffold.c rev1.2

14 years agoGive the kernel a native NFS mount rpc capability for mounting NFS roots by
Matthew Dillon [Sun, 4 Sep 2005 01:29:00 +0000 (01:29 +0000)]
Give the kernel a native NFS mount rpc capability for mounting NFS roots by
splitting off the mount rpc code from the BOOTP code.  The loader is no
longer required to pass the nfs root mount file handle to the kernel.

Pure tftp-based loaders with no knowledge of NFS can now pass a NFS root
mount path to the kernel without having to pass a resolved NFS file handle.
This change allows kernels booted from tftp loaders to have an NFS root
without having to specify BOOTP (which sometimes doesn't work properly when
done from both the loader and from the kernel).

14 years agoCompile up both the NFS and the TFTP version of the PXE loader and
Matthew Dillon [Sat, 3 Sep 2005 23:52:53 +0000 (23:52 +0000)]
Compile up both the NFS and the TFTP version of the PXE loader and
make both available in /boot.  The NFS version obtains all loader-level data
(including the kernel binary) from the diskless root mount.  The TFTP version
obtains all loader-level data (including the kernel binary) via TFTP.

Do not pass a bogus NFS handle to the kernel (it was previously passing an
all-zero's handle for the TFTP version).

Note that current kernels will not be able to do a NFS root mount from the
tftp version of the loader unless they have BOOTP support compiled in,
because the tftp version of the loader is not able to resolve and pass a
NFS file handle to the kernel.

14 years agoRemove old #if 0'd sections of code, add a few comments, and report a bit
Matthew Dillon [Sat, 3 Sep 2005 23:43:59 +0000 (23:43 +0000)]
Remove old #if 0'd sections of code, add a few comments, and report a bit
more information when mounting an NFS root.

14 years agoPrint out a little more information on the PXE boot configuration.
Matthew Dillon [Sat, 3 Sep 2005 22:16:24 +0000 (22:16 +0000)]
Print out a little more information on the PXE boot configuration.

14 years agoAdd a missing crit_exit(), fixing a panic. Attempt to continue with the
Matthew Dillon [Sat, 3 Sep 2005 17:56:24 +0000 (17:56 +0000)]
Add a missing crit_exit(), fixing a panic.  Attempt to continue with the
mount even if we cannot add the default routeo (since the NFS server might
be on the LAN).  Print out useful information on the console such as the
interface chosen and the mount point.

Reported-by: Scott Ullrich <geekgod@geekgod.com> (critical section panic)
14 years agoAdd strnvis, which is orthogonal to strvisx by bounding dst, not src.
Joerg Sonnenberger [Sat, 3 Sep 2005 16:25:42 +0000 (16:25 +0000)]
Add strnvis, which is orthogonal to strvisx by bounding dst, not src.

Inspired-by: OpenBSD
14 years agoSync with FreeBSD HEAD.
Joerg Sonnenberger [Sat, 3 Sep 2005 15:29:02 +0000 (15:29 +0000)]
Sync with FreeBSD HEAD.

14 years agoGC openssh-3.9
Joerg Sonnenberger [Sat, 3 Sep 2005 15:09:49 +0000 (15:09 +0000)]
GC openssh-3.9

14 years agoUpdate __DragonFly_version as well. Don't let this slip, please.
Joerg Sonnenberger [Sat, 3 Sep 2005 12:16:14 +0000 (12:16 +0000)]
Update __DragonFly_version as well. Don't let this slip, please.

14 years agoAdd a stopgap to get a proper osversion and osrelease on -Preview.
Simon Schubert [Sat, 3 Sep 2005 11:45:01 +0000 (11:45 +0000)]
Add a stopgap to get a proper osversion and osrelease on -Preview.

We should de-hack this file and put the main version number in one place.

14 years agoMake sure the vnode is unlocked across a dev_dclose() call, otherwise a
Matthew Dillon [Sat, 3 Sep 2005 08:01:32 +0000 (08:01 +0000)]
Make sure the vnode is unlocked across a dev_dclose() call, otherwise a
stuck tty can deadlock the filesystem (e.g. if you try to ls -la /dev).

VOP_CLOSE() may or may not be called with the vnode locked, so keep track
of it.

14 years ago1) merge almost identical cs_{isa, pccard}_attach() into cs_attach()
Sepherosa Ziehau [Fri, 2 Sep 2005 12:51:00 +0000 (12:51 +0000)]
1) merge almost identical cs_{isa, pccard}_attach() into cs_attach()
2) staticize csintr()
3) add cs_detach()
4) in cs_attach()
   - move ifmedia_init() to the beginning of cs_attach(), so that
     ifmedia_removeall() can be called safely if cs_attach() fails
   - add KASSERT to make sure either I/O port or I/O memory is set up before
     entering cs_attach()
   - set up intr in the last step of cs_attach()
   - remove unnecessary NULL check after malloc(), since M_WAITOK flag is
     used here
   - use newly added cs_detach() as error handling routine
   - use `#if 0' to comment out code, instead of fixing them in the comment.
     this makes the code easy to read as well as the comment

14 years agoin cs_isa_probe()
Sepherosa Ziehau [Fri, 2 Sep 2005 09:43:01 +0000 (09:43 +0000)]
in cs_isa_probe()
clean up code path by remove unnecessary `error' check and `goto'

14 years ago- printf("%s: str", if_xname) ==> if_printf/device_printf
Sepherosa Ziehau [Fri, 2 Sep 2005 08:14:36 +0000 (08:14 +0000)]
- printf("%s: str", if_xname) ==> if_printf/device_printf
- remove unused macro CS_NAME

14 years agoNow that the C language has a "void *", use it instead of caddr_t.
Jeffrey Hsu [Fri, 2 Sep 2005 07:16:58 +0000 (07:16 +0000)]
Now that the C language has a "void *", use it instead of caddr_t.

14 years agoK&R style function removal. Update functions to ANSI style. Also a few
Liam J. Foy [Thu, 1 Sep 2005 22:45:35 +0000 (22:45 +0000)]
K&R style function removal.  Update functions to ANSI style. Also a few
minor cleans.

Submitted by: Peter Avalos

14 years agoZap unused variable!
Liam J. Foy [Thu, 1 Sep 2005 22:19:26 +0000 (22:19 +0000)]
Zap unused variable!

Obtained from: FreeBSD

14 years agoUpdate to groff 1.19.1.
Jeroen Ruigrok/asmodai [Thu, 1 Sep 2005 20:47:14 +0000 (20:47 +0000)]
Update to groff 1.19.1.
This synchronises us with FreeBSD up till 1.19 (Ids adjusted accordingly).

Big help fixing PATCHES behaviour: joerg

14 years agoRemove pointless -? flag.
Sascha Wildner [Thu, 1 Sep 2005 19:08:38 +0000 (19:08 +0000)]
Remove pointless -? flag.

14 years agoDon't add files patches via *.no_obj.patch back to SRCS, make them a
Joerg Sonnenberger [Thu, 1 Sep 2005 14:17:58 +0000 (14:17 +0000)]
Don't add files patches via *.no_obj.patch back to SRCS, make them a
dependency of beforedepend instead. This avoids ordering problems with
bsd.lib.mk and bsd.prog.mk.

14 years agoRetire old perl 5.00503 optional manpath and replace with a manpath to
Jeroen Ruigrok/asmodai [Thu, 1 Sep 2005 13:13:35 +0000 (13:13 +0000)]
Retire old perl 5.00503 optional manpath and replace with a manpath to
/usr/pkg/man.

Whether we want to keep the ones to /usr/local and /usr/X11R6 remains to be
seen.

14 years agoFor Arcnet, announce link layer address in arc_ifattch(),
Sepherosa Ziehau [Thu, 1 Sep 2005 12:59:38 +0000 (12:59 +0000)]
For Arcnet, announce link layer address in arc_ifattch(),
not in each Arcnet netif drivers (currently we only have cm(4)).

Approved-by: joerg
14 years ago1) in cm_isa_attach(), setup intr in the last step
Sepherosa Ziehau [Thu, 1 Sep 2005 09:17:19 +0000 (09:17 +0000)]
1) in cm_isa_attach(), setup intr in the last step
2) in cm_attach()
   - after 1), cm_attach() is intr safe, remove crit-sect
   - initialize callout before cm_stop()
3) in cm_stop(), add missing callout_stop()
4) instead of clearing IFF_RUNNING immediately after every call of cm_stop(),
   move this clearing into cm_stop()

14 years agoRemove PC98 support (forgot these in my last commit).
Sascha Wildner [Thu, 1 Sep 2005 00:40:51 +0000 (00:40 +0000)]
Remove PC98 support (forgot these in my last commit).

14 years agoRemove PC98 support.
Sascha Wildner [Thu, 1 Sep 2005 00:18:24 +0000 (00:18 +0000)]
Remove PC98 support.

14 years agoRemove PC98 support.
Sascha Wildner [Wed, 31 Aug 2005 23:22:47 +0000 (23:22 +0000)]
Remove PC98 support.

14 years agoScrap PC98 specific snc(4) driver.
Sascha Wildner [Wed, 31 Aug 2005 22:55:28 +0000 (22:55 +0000)]
Scrap PC98 specific snc(4) driver.

OK'd by: joerg

14 years agoFix indentation with previous commit.
Jeffrey Hsu [Wed, 31 Aug 2005 22:21:23 +0000 (22:21 +0000)]
Fix indentation with previous commit.

14 years agoCheck the IP length first to avoid a memory leak later.
Jeffrey Hsu [Wed, 31 Aug 2005 22:18:46 +0000 (22:18 +0000)]
Check the IP length first to avoid a memory leak later.

14 years agoRemove pointless -? flag.
Sascha Wildner [Wed, 31 Aug 2005 18:11:05 +0000 (18:11 +0000)]
Remove pointless -? flag.

14 years agoperror() -> err()
Liam J. Foy [Wed, 31 Aug 2005 17:20:18 +0000 (17:20 +0000)]
perror() -> err()

14 years agoKill unused variables
Liam J. Foy [Wed, 31 Aug 2005 16:45:51 +0000 (16:45 +0000)]
Kill unused variables

Obtained from: FreeBSD

14 years agoRetire old sendmail.
Joerg Sonnenberger [Wed, 31 Aug 2005 16:45:44 +0000 (16:45 +0000)]
Retire old sendmail.

14 years agoUpdate zoneinfo database.
Sascha Wildner [Wed, 31 Aug 2005 16:44:29 +0000 (16:44 +0000)]
Update zoneinfo database.

Theory: 7.15 -> 7.19
africa: 7.37 -> 7.39
antarctica: 7.24 -> 7.25
asia: 7.82 -> 7.85
australasia: 7.71 -> 7.73
backward: 7.27 -> 7.29
europe: 7.92 -> 7.95
leapseconds: 7.17 -> 7.20
northamerica: 7.74 -> 7.79
southamerica: 7.59 -> 7.61
zone.tab: 1.30 -> 1.32

Note: I had already updated leapseconds to include this year's leapsecond
      so we basically just change the comments' indenting here.

14 years agoRemove the sleeper crap and use MAXPATHLEN
Liam J. Foy [Wed, 31 Aug 2005 16:27:53 +0000 (16:27 +0000)]
Remove the sleeper crap and use MAXPATHLEN

14 years agoTwo small alterations to 'make upgrade':
Chris Pressey [Wed, 31 Aug 2005 16:17:52 +0000 (16:17 +0000)]
Two small alterations to 'make upgrade':
- Create etc/rc.d directory if it does not exist, before installing
  scripts into it.
- Create etc/pam.d directory if it does not exist, and run convert.sh
  to populate it in this case.

This allows a FreeBSD 4.9 system to be binary-upgraded to DragonFly
1.3.5-DEVELOPMENT.  (A warning about missing PAM modules may still
occur during login, however.)

14 years agoDetect pkgtools fakeroot feature and use it.
Simon Schubert [Wed, 31 Aug 2005 02:38:34 +0000 (02:38 +0000)]
Detect pkgtools fakeroot feature and use it.

This enables us to use host pkgtools to add packages to the isos
instead of using the target pkgtools.  This would fail with -Release
building -Devel isos.

14 years agoThis program requires at least one option. Clean up the checking and report
Liam J. Foy [Tue, 30 Aug 2005 22:42:15 +0000 (22:42 +0000)]
This program requires at least one option. Clean up the checking and report
this fact correctly.

Obtained from: FreeBSD

14 years agoMinor typo
Liam J. Foy [Tue, 30 Aug 2005 22:02:36 +0000 (22:02 +0000)]
Minor typo

14 years agoInclude the option-arguments in the description and remove a non-existant
Liam J. Foy [Tue, 30 Aug 2005 21:25:35 +0000 (21:25 +0000)]
Include the option-arguments in the description and remove a non-existant
bug.

Obtained from: FreeBSD

14 years agoRemove undocumented historic support for treating "-" as an option
Liam J. Foy [Tue, 30 Aug 2005 21:19:13 +0000 (21:19 +0000)]
Remove undocumented historic support for treating "-" as an option
instead of a file operand; this is not compatible with POSIX.

Obtained from: FreeBSD

14 years agoForced commit to note that the review and the suggestion
YONETANI Tomokazu [Tue, 30 Aug 2005 14:22:05 +0000 (14:22 +0000)]
Forced commit to note that the review and the suggestion
for the previous commit was done by Hiten Pandya(hmp@)
and not someone named hiten@ .

14 years agoWorkaround the panic, triggered by the netgraph attempting to
YONETANI Tomokazu [Tue, 30 Aug 2005 13:25:07 +0000 (13:25 +0000)]
Workaround the panic, triggered by the netgraph attempting to
load a missing module from within the socket dispatch thread, by:

- not trying to load a netgraph module when you can't.

- preloading ng_ksocket.ko(if missing) from the module handler
  where you're guaranteed to have the process context needed by
  the linker_* functions. This makes us sure at least that the basic
  netgraph operations will work without messing with loading a kernel
  module on the fly (suggested by: hiten@).

Reviewed-by: hiten@
14 years ago1) Remove compat code
Sepherosa Ziehau [Tue, 30 Aug 2005 12:33:49 +0000 (12:33 +0000)]
1) Remove compat code
2) Do not include <machine/clock.h>
3) Remove CMSOFTCOPY option and code introduced by this option.
   softintr_{schedule, establish}() only exist in NetBSD.
4) Remove cm_softc.{sc_txcookie, sc_rxcookie}, which are only used by
   the nuked code mentioned in 3)

14 years agoin cm_isa_probe():
Sepherosa Ziehau [Tue, 30 Aug 2005 10:16:33 +0000 (10:16 +0000)]
in cm_isa_probe():
- remove unused stack variable
- clean up code path by removing unnecessary `goto'

14 years agoprintf("%s: str", ifp->if_xname) ==> if_printf(ifp, "str")
Sepherosa Ziehau [Tue, 30 Aug 2005 09:53:52 +0000 (09:53 +0000)]
printf("%s: str", ifp->if_xname) ==> if_printf(ifp, "str")

14 years agoGet rid of smp_rendezvous() and all associated support circuitry. Move
Matthew Dillon [Mon, 29 Aug 2005 21:08:06 +0000 (21:08 +0000)]
Get rid of smp_rendezvous() and all associated support circuitry.  Move
the two mechanisms still using it (User LDT and MTRR propogation) over
to the lwkt_cpusync*() API.

14 years agoUse a typedef that already conveniently exists instead of anonymous
Matthew Dillon [Mon, 29 Aug 2005 17:46:52 +0000 (17:46 +0000)]
Use a typedef that already conveniently exists instead of anonymous
TAILQ_HEAD() declarations for the LWKT queues.  No operational changes.

14 years agoAdd a space to the output for legibility.
Jeffrey Hsu [Mon, 29 Aug 2005 16:03:31 +0000 (16:03 +0000)]
Add a space to the output for legibility.

14 years agoFix the same problem described in FreeBSD-SA-05:15.tcp, which
Noritoshi Demizu [Mon, 29 Aug 2005 10:24:10 +0000 (10:24 +0000)]
Fix the same problem described in FreeBSD-SA-05:15.tcp, which
refers CAN-2005-0356 and CAN-2005-2068.  The e-mail below has more.
http://leaf.dragonflybsd.org/mailarchive/bugs/2005-08/msg00052.html

14 years agoRemove unnecessary bzero(softc).
Sepherosa Ziehau [Mon, 29 Aug 2005 10:19:52 +0000 (10:19 +0000)]
Remove unnecessary bzero(softc).
With following exceptions:
  - ed(4) if_ed_cbus.c
  - snc(4)
They are PC-98 only.

14 years agoFix a bug in NewReno which may send a huge amount of data
Noritoshi Demizu [Mon, 29 Aug 2005 10:04:01 +0000 (10:04 +0000)]
Fix a bug in NewReno which may send a huge amount of data
when a partial ACK is received.  The e-mail below has more.
http://leaf.dragonflybsd.org/mailarchive/bugs/2005-07/msg00013.html

14 years agoRemove VPLACEMARKER
Simon Schubert [Sun, 28 Aug 2005 23:35:35 +0000 (23:35 +0000)]
Remove VPLACEMARKER

14 years agoFix a deadlock in ffs_balloc(). This function was incorrectly obtaining a
Matthew Dillon [Sun, 28 Aug 2005 23:23:10 +0000 (23:23 +0000)]
Fix a deadlock in ffs_balloc().  This function was incorrectly obtaining a
locked indirect buffer followed by a locked data block buffer, where as other
procedures in the kernel generally held a locked data buffer and then
called procedures which locked the indirect buffer.  Programs like rtorrent,
which write data into the mmap'd files whos blocks had not yet been
allocated, could easily deadlock the vnode.

A typical deadlock would be: syncer calls putpages->ffs_write->balloc->getblk->
allocbuf->(blocked on VM page with indirect block and data block locked), while
at the same time some process takes a write fault which locks the VM page and
then attempts to do a BMAP, blocking on the indirect block buffer.  Deadlock.

The fix for ffs_balloc() is simply to obtain the data buffer prior to
obtaining the indirect block(s).

Reported-by: Mitja Horvat
MFC: 1 week

14 years ago- Don't write our own pid file, just use pidfile()
Liam J. Foy [Sun, 28 Aug 2005 20:55:07 +0000 (20:55 +0000)]
- Don't write our own pid file, just use pidfile()

14 years ago- Use pidfilei(3) to write a pid file in /var/run
Liam J. Foy [Sun, 28 Aug 2005 20:15:38 +0000 (20:15 +0000)]
- Use pidfilei(3) to write a pid file in /var/run

14 years agoTaken from FreeBSD-HEAD:
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 18:36:14 +0000 (18:36 +0000)]
Taken from FreeBSD-HEAD:

fr.ISO8859-1: Localisation file by Sebastien Gioria
koi8-r.tmac: Ruslan Ermilov's koi8-r tmac file

14 years agoMerge from vendor branch GROFF:
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 18:25:09 +0000 (18:25 +0000)]
Merge from vendor branch GROFF:
Remove mdoc.local, we need to make this truly local.

14 years agoRemove mdoc.local, we need to make this truly local.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 18:25:09 +0000 (18:25 +0000)]
Remove mdoc.local, we need to make this truly local.

14 years agoRemove mdoc.local, we need to make this truly local.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 18:25:09 +0000 (18:25 +0000)]
Remove mdoc.local, we need to make this truly local.

14 years agoMerge from vendor branch GROFF:
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 18:06:33 +0000 (18:06 +0000)]
Merge from vendor branch GROFF:
Update and reorder.

14 years agoUpdate and reorder.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 18:06:33 +0000 (18:06 +0000)]
Update and reorder.

14 years agoMerge from vendor branch GROFF:
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 18:03:57 +0000 (18:03 +0000)]
Merge from vendor branch GROFF:
Add groff 1.19.1, stripped down appropriately.

14 years agoAdd groff 1.19.1, stripped down appropriately.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 18:03:57 +0000 (18:03 +0000)]
Add groff 1.19.1, stripped down appropriately.

14 years agoMerge from vendor branch GROFF:
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 17:20:36 +0000 (17:20 +0000)]
Merge from vendor branch GROFF:
Add groff 1.19.1, stripped down appropriately.

14 years agoAdd groff 1.19.1, stripped down appropriately.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 17:20:36 +0000 (17:20 +0000)]
Add groff 1.19.1, stripped down appropriately.

14 years agoAdd a version identifier to enable detection of additional features.
Simon Schubert [Sun, 28 Aug 2005 17:19:16 +0000 (17:19 +0000)]
Add a version identifier to enable detection of additional features.

14 years agoAdd a hack to pkgtools to install to a different root. This differs from
Simon Schubert [Sun, 28 Aug 2005 16:56:12 +0000 (16:56 +0000)]
Add a hack to pkgtools to install to a different root.  This differs from
using PREFIX as the installed files don't get relocated, but all the tree
appears in PKG_FAKEROOT, including PKG_DBDIR.

This won't work for all packages and isn't ment to either.  Pkgtools are
evil hackery anyways and this adds even more.  This will be used in
nrelease to support addition of packages by using the host tools and not
the target tools.

14 years agoMerge from vendor branch TEXINFO:
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 15:28:56 +0000 (15:28 +0000)]
Merge from vendor branch TEXINFO:
fdl.texi is needed.

14 years agofdl.texi is needed.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 15:28:56 +0000 (15:28 +0000)]
fdl.texi is needed.

14 years agoA machine-independent spinlock implementation. It has the advantages of
Jeffrey Hsu [Sun, 28 Aug 2005 15:27:05 +0000 (15:27 +0000)]
A machine-independent spinlock implementation.  It has the advantages of
  1.  being written in C except for the most low-level atomic swap primitive,
        which is universally supported on current processor architectures
  2.  having a very small inlined memory footprint for spin_lock(),
        with the slow-path defered to a subroutine call
  3.  only requiring a bus-locked operation for lock acquisition,
         and not requiring a bus-locked operation for lock release
  4.  doing a non-bus-locked check first in the spin loop to
        reduce bus contention
  5.  doing exponential backoff in the uncommon contested case, which
        Sun has found to reduce bus contention by a factor of 5 or more

Reviewed by: Matt Dillon

14 years agoCommit a missed change.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 14:19:30 +0000 (14:19 +0000)]
Commit a missed change.

14 years agoSwitch to texinfo 4.8, which is needed for a lot of new texi files.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 12:25:05 +0000 (12:25 +0000)]
Switch to texinfo 4.8, which is needed for a lot of new texi files.

14 years agoAdd texinfo 4.8, appropriately stripped down.
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 12:23:25 +0000 (12:23 +0000)]
Add texinfo 4.8, appropriately stripped down.

14 years agoMerge from vendor branch TEXINFO:
Jeroen Ruigrok/asmodai [Sun, 28 Aug 2005 12:23:25 +0000 (12:23 +0000)]
Merge from vendor branch TEXINFO:
Add texinfo 4.8, appropriately stripped down.

14 years agoRequire HEAD users to upgrade to 1.3.5 before running installworld, due
Matthew Dillon [Sun, 28 Aug 2005 07:06:42 +0000 (07:06 +0000)]
Require HEAD users to upgrade to 1.3.5 before running installworld, due
to the dirent changes (otherwise the 'find' binary will fail half way
through the installworld).