]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoImplement ARCHIVE_EXTRACT_UNLINK for regular files. This supports the
Tim Kientzle [Thu, 29 Apr 2004 05:19:42 +0000 (05:19 +0000)]
Implement ARCHIVE_EXTRACT_UNLINK for regular files.  This supports the
-U flag to bsdtar.  Essentially, this option breaks existing hard
links.  According to SUSv2, tar is supposed to overwrite existing
files on extract by default which, in particular, preserves
existing hard links.  Note that this is yet another bug in gtar; it
appears to always break existing links.  (Maybe gtar's -U is broken?)

I'm unsure about how to handle this for other file types; the current
code always unlinks first unless the NO_OVERWRITE flag is specified.
I've commented this issue liberally and will come back to it later.

20 years agoAdd a new fortune from petef.
Sean Kelly [Thu, 29 Apr 2004 04:36:36 +0000 (04:36 +0000)]
Add a new fortune from petef.

Obtained from: petef

20 years agoDon't do malloc(M_WAITOK) for sound buffers while locks are held.
Brian Feldman [Thu, 29 Apr 2004 02:51:59 +0000 (02:51 +0000)]
Don't do malloc(M_WAITOK) for sound buffers while locks are held.

20 years agoFix a memory leak in ng_get_string_token. A dynamically-allocated
John Polstra [Thu, 29 Apr 2004 01:37:11 +0000 (01:37 +0000)]
Fix a memory leak in ng_get_string_token.  A dynamically-allocated
buffer wasn't freed if the function failed.

MFC after: 3 days

20 years agoBump manual page date for last update.
Simon L. B. Nielsen [Wed, 28 Apr 2004 23:16:06 +0000 (23:16 +0000)]
Bump manual page date for last update.

Suggested by: ru

20 years agoSpelling: s/Unlinke/Unlike/
Tom Rhodes [Wed, 28 Apr 2004 22:25:47 +0000 (22:25 +0000)]
Spelling: s/Unlinke/Unlike/

PR: 66063

20 years agoFix up a little language. Remove reference to sysctl(9). Fix reference to
Sean Kelly [Wed, 28 Apr 2004 21:42:36 +0000 (21:42 +0000)]
Fix up a little language. Remove reference to sysctl(9). Fix reference to
watchdog(9) (was watchdogd(9)).

20 years agoThanks to David's patches, we can now simplify these makefiles
Ruslan Ermilov [Wed, 28 Apr 2004 21:31:21 +0000 (21:31 +0000)]
Thanks to David's patches, we can now simplify these makefiles
further, and just use PROG directly (without FILES).

Suggested by: bde

20 years agoProperly merge boot0sio.s and boot0.s into boot0.S.
Ruslan Ermilov [Wed, 28 Apr 2004 21:14:40 +0000 (21:14 +0000)]
Properly merge boot0sio.s and boot0.s into boot0.S.

20 years agoSwitch to using C99 comments in assembler preprocessed with cpp(1).
Ruslan Ermilov [Wed, 28 Apr 2004 20:55:01 +0000 (20:55 +0000)]
Switch to using C99 comments in assembler preprocessed with cpp(1).

20 years ago- Properly merge boot0sio.s and boot0.s into boot0.S.
Ruslan Ermilov [Wed, 28 Apr 2004 20:49:17 +0000 (20:49 +0000)]
- Properly merge boot0sio.s and boot0.s into boot0.S.
  boot0sio.s was repo-copied to boot0.S.

- Rename boot0ext.s to boot0ext.S, to stay consistent
  with other preprocessed asm files around here, and
  for better portability.

Repocopied by: joe

20 years agoKeep track of threads waiting in kse_release() to avoid a race
Daniel Eischen [Wed, 28 Apr 2004 20:36:53 +0000 (20:36 +0000)]
Keep track of threads waiting in kse_release() to avoid a race
condition where kse_wakeup() doesn't yet see them in (interruptible)
sleep queues.  Also add an upcall check to sleepqueue_catch_signals()
suggested by jhb.

This commit should fix recent mysql hangs.

Reviewed by: jhb, davidxu
Mysql'd by: Robin P. Blanchard <robin.blanchard at gactr uga edu>

20 years agoRefuse to extract entries with '..' in pathname.
Tim Kientzle [Wed, 28 Apr 2004 18:53:07 +0000 (18:53 +0000)]
Refuse to extract entries with '..' in pathname.

Pointed out by: David Schultz

20 years agoRemove code that fiddles with Giant in ndis_ticktask() that snuck in
Bill Paul [Wed, 28 Apr 2004 17:06:18 +0000 (17:06 +0000)]
Remove code that fiddles with Giant in ndis_ticktask() that snuck in
during previous commit.

20 years agoTemporary precaution measure until repocopy requests are
Ruslan Ermilov [Wed, 28 Apr 2004 16:16:16 +0000 (16:16 +0000)]
Temporary precaution measure until repocopy requests are
processed: don't forget to clean generated .s sources.

20 years agoo Add a missed description for m_free(9).
Maxim Konovalov [Wed, 28 Apr 2004 14:53:49 +0000 (14:53 +0000)]
o Add a missed description for m_free(9).

Optained from: NetBSD
MFC after: 1 week

20 years agoAfter talking to Bruce Evans and reading more standards specs,
Ruslan Ermilov [Wed, 28 Apr 2004 14:31:44 +0000 (14:31 +0000)]
After talking to Bruce Evans and reading more standards specs,
switch to using C99-style comments everywhere in preprocessed
assembler.  The reason is that lines starting with the regexp
'^[[:space:]]#' are treated as preprocessing directives, and
while it seems to work now with GCC, it's not necessarily has
to work.  Use C99 comments `//' for the trailing comments to
save whitespace.

20 years agoRemove the EBus stopgap of r1.248; a proper fix is in place now.
Thomas Moestl [Wed, 28 Apr 2004 13:43:11 +0000 (13:43 +0000)]
Remove the EBus stopgap of r1.248; a proper fix is in place now.

20 years agoProtect som cross-script invocations by checks to see that the target
Poul-Henning Kamp [Wed, 28 Apr 2004 13:20:15 +0000 (13:20 +0000)]
Protect som cross-script invocations by checks to see that the target
script exists.  This allows pruning of rc.d scripts without getting
too many ugly boottime error message

20 years agoInitdiskless must run before ipfw, or people will not be able to
Poul-Henning Kamp [Wed, 28 Apr 2004 13:18:22 +0000 (13:18 +0000)]
Initdiskless must run before ipfw, or people will not be able to
use the conf/* stuff for their firewall configuration.

Running ipfw before could seem to make sense in that it would allow
one to setup access to the NFS server on a "default-deny" kernel,
but it is pretty obvious to the casual observer that such a configuration
never makes it far enough to mount the NFS-root in the first place.

20 years agoFix the EBus driver to work with the new PCI code. Unlike other PCI
Thomas Moestl [Wed, 28 Apr 2004 13:06:46 +0000 (13:06 +0000)]
Fix the EBus driver to work with the new PCI code. Unlike other PCI
bridges, the EBus bridge has resource ranges it claims exclusively to
map its children into in its BARs. Hence, we need to allocate these
completely and manage them for the children, instead of just passing
allocations through to the PCI layer as we did before.

While being there, split ebus_probe(), which did also contain code
normally belonging into the attach method, into ebus_probe() and
ebus_attach(), and perform some minor cleanups.

20 years agoSheesh...
Poul-Henning Kamp [Wed, 28 Apr 2004 11:44:28 +0000 (11:44 +0000)]
Sheesh...

Turn your back on SNMP for a decade and they go and make upper
versus lowercase initial letter part of the specification.

I told them ASN.1 was a bad mistake, but would they listen ?  Ohhh no!

20 years agoMFi386: Use C (and CPP) style comments for assembler-with-cpp
Ruslan Ermilov [Wed, 28 Apr 2004 10:15:31 +0000 (10:15 +0000)]
MFi386: Use C (and CPP) style comments for assembler-with-cpp
sources, for lines that start with a comment.

20 years agoUse C (and CPP) style comments for assembler-with-cpp sources,
Ruslan Ermilov [Wed, 28 Apr 2004 10:09:21 +0000 (10:09 +0000)]
Use C (and CPP) style comments for assembler-with-cpp sources,
for lines that start with a comment.

20 years agoUse C (and CPP) style comments for assembler-with-cpp sources,
Ruslan Ermilov [Wed, 28 Apr 2004 09:57:12 +0000 (09:57 +0000)]
Use C (and CPP) style comments for assembler-with-cpp sources,
for lines that start with a comment.

20 years ago- Use C (and CPP) style comments for assembler-with-cpp sources,
Ruslan Ermilov [Wed, 28 Apr 2004 09:49:22 +0000 (09:49 +0000)]
- Use C (and CPP) style comments for assembler-with-cpp sources,
  for lines that start with a comment.
- Preserve as(1) style comments for the rest.

20 years agoReplace shell's positional parameters safely.
Ruslan Ermilov [Wed, 28 Apr 2004 09:09:27 +0000 (09:09 +0000)]
Replace shell's positional parameters safely.

20 years agoUpdate comments to reflect changes made by phk. Also no longer need
Sean Kelly [Wed, 28 Apr 2004 07:35:03 +0000 (07:35 +0000)]
Update comments to reflect changes made by phk. Also no longer need
<sys/sysctl.h>.

20 years agoRemove information about specific formats; replace with a
Tim Kientzle [Wed, 28 Apr 2004 04:43:57 +0000 (04:43 +0000)]
Remove information about specific formats; replace with a
reference to libarchive-formats(5).

20 years agoAdd statistics: track offset in compressed and uncompressed archive,
Tim Kientzle [Wed, 28 Apr 2004 04:41:27 +0000 (04:41 +0000)]
Add statistics:  track offset in compressed and uncompressed archive,
provide an interface for the client to query this information.

20 years agoCorrect spelling of "shell archive" to match /usr/share/misc/magic.
Tim Kientzle [Wed, 28 Apr 2004 04:36:57 +0000 (04:36 +0000)]
Correct spelling of "shell archive" to match /usr/share/misc/magic.

20 years agoAdd the libarchive-formats.5 manpage. This summarizes the various
Tim Kientzle [Wed, 28 Apr 2004 04:34:07 +0000 (04:34 +0000)]
Add the libarchive-formats.5 manpage.  This summarizes the various
archive formats supported by libarchive, with some information about
the relative strengths and weaknesses of each format and notes about
issues with libarchive's support for those formats.

This page should make it unnecessary to list all of the libarchive
formats in the manpage of each program that uses libarchive.
Such programs can simply refer to libarchive-formats(5).

20 years agoAdd read support for four more common cpio variants:
Tim Kientzle [Wed, 28 Apr 2004 03:44:08 +0000 (03:44 +0000)]
Add read support for four more common cpio variants:
 * little-endian old-style binary cpio archives
 * big-endian old-style binary cpio archives
 * SVR4 portable archives without CRC
 * SVR4 portable archives with CRC

Note that I don't yet verify the CRC for the last one, and I'm
not quite certain I'm correctly parsing device numbers.

20 years agoAdd LinkSys USB100M into list of supported devices and reformat list of those
Maxim Sobolev [Wed, 28 Apr 2004 02:52:13 +0000 (02:52 +0000)]
Add LinkSys USB100M into list of supported devices and reformat list of those
devices to be similar to other USB ethers.

20 years agoVarious quibbles:
David Schultz [Wed, 28 Apr 2004 01:27:36 +0000 (01:27 +0000)]
Various quibbles:
- Print a diagnostic if kdumpenv() fails.  This can occur due to MAC
  restrictions or lack of memory.  Catch all kenv(2) failures as well.
- Just of the heck of it, DTRT if the kernel environment size changes
  at the wrong time.  The old code could fail silently or fail to
  null-terminate a buffer if you got exceptionally unlucky.
- Sort and GC the #includes.

20 years agoIf the buffer supplied to kenv(KENV_DUMP, ...) isn't big enough,
David Schultz [Wed, 28 Apr 2004 01:27:33 +0000 (01:27 +0000)]
If the buffer supplied to kenv(KENV_DUMP, ...) isn't big enough,
return the number of bytes needed instead of 0.  The manpage claims
that we do this anyway.

20 years agoAdd jerk city commentary on C++.
Warner Losh [Tue, 27 Apr 2004 22:32:02 +0000 (22:32 +0000)]
Add jerk city commentary on C++.

20 years ago- The i8254 uses IRQ 0, not IRQ 8. Correct i8254_intsrc to reference the
John Baldwin [Tue, 27 Apr 2004 20:03:26 +0000 (20:03 +0000)]
- The i8254 uses IRQ 0, not IRQ 8.  Correct i8254_intsrc to reference the
  correct interrupt source.
- Cache a pointer to the i8254_intsrc's pending method to avoid several
  pointer indirections in i8254_get_timecount().

Reported by: bde

20 years agoRemoved now redundant CLEANFILES assignments.
Ruslan Ermilov [Tue, 27 Apr 2004 19:45:16 +0000 (19:45 +0000)]
Removed now redundant CLEANFILES assignments.

Not read enough of my patch by: obrien ;)

20 years agoBelatedly remove a file from the earlier apic code that is no longer used.
John Baldwin [Tue, 27 Apr 2004 19:37:17 +0000 (19:37 +0000)]
Belatedly remove a file from the earlier apic code that is no longer used.

20 years agoChange from '#' to "//" comment character to allow CPP preprocessing.
Poul-Henning Kamp [Tue, 27 Apr 2004 19:07:35 +0000 (19:07 +0000)]
Change from '#' to "//" comment character to allow CPP preprocessing.

Merge boot0.s and boot0sio.s into boot0_512.s controlled by "#ifdef SIO".

Add Makefile magic to generate boot0.s and boot0sio.s from boot0_512.s.

The compile boot0 and boot0sio have unchanged MD5 checksums.

20 years agoAdded 3ware's twa.
Vinod Kashyap [Tue, 27 Apr 2004 17:57:45 +0000 (17:57 +0000)]
Added 3ware's twa.

Reviewed by: ps

20 years agoAddress few style issues pointed out by bde
Maksim Yevmenkin [Tue, 27 Apr 2004 16:38:15 +0000 (16:38 +0000)]
Address few style issues pointed out by bde

Reviewed by: bde, ru

20 years agoMore POLA handling of CUSTOMIZE variable
Poul-Henning Kamp [Tue, 27 Apr 2004 16:04:41 +0000 (16:04 +0000)]
More POLA handling of CUSTOMIZE variable

20 years agoReduce the time spent looking for devices on channels that return
Søren Schmidt [Tue, 27 Apr 2004 15:52:08 +0000 (15:52 +0000)]
Reduce the time spent looking for devices on channels that return
what looks like garbage.

20 years agoWhen using this in non-NFSroot circumstances, `mount -t nfs` may
Poul-Henning Kamp [Tue, 27 Apr 2004 15:42:34 +0000 (15:42 +0000)]
When using this in non-NFSroot circumstances, `mount -t nfs` may
return empty.  Add a dummy element to prevent the alternate action
of the shell builtin "set"

20 years ago- Added rad_demangle() for demangling user-passwords (needed for
Ruslan Ermilov [Tue, 27 Apr 2004 15:00:29 +0000 (15:00 +0000)]
- Added rad_demangle() for demangling user-passwords (needed for
  MS-CHAPv1 MPPE-keys).
- Added rad_demangle_mppe_key() for demangling mppe-keys (needed
  for MPPE-keys).
- Added some typecasts for avoiding compiler warnings.
- Fix: better handle wrong usage of the lib (if the programmer
  has not called rad_create_request() but rad_put_*(), then a
  weird error message was returned).
- Added a new function for putting the Message-Authenticator.
- Verify the Message-Authenticator, if it was found inside a
  response packet and silently drop the packet, if the validation
  failed.
- Implicitly put the Message-Authenticator, if the EAP-Message
  attribute was added.
- Added some missing defines.

Submitted by: Michael Bretterklieber
PR: 46555

20 years agoInvoke tzset(3) within the main loop to catch changes to /etc/localtime
Guy Helmer [Tue, 27 Apr 2004 14:43:03 +0000 (14:43 +0000)]
Invoke tzset(3) within the main loop to catch changes to /etc/localtime
when running as a daemon in the hope that it will fix situations
where the CMOS clock was apparently set using the stale TZ offset.

20 years agoSync to 1.172 of usbdevs
MIHIRA Sanpei Yoshiro [Tue, 27 Apr 2004 13:56:39 +0000 (13:56 +0000)]
Sync to 1.172 of usbdevs

20 years agoAdd support DELL BC02 Bluetooth USB Adapter(TrueMobile 300)
MIHIRA Sanpei Yoshiro [Tue, 27 Apr 2004 13:55:26 +0000 (13:55 +0000)]
Add support DELL BC02 Bluetooth USB Adapter(TrueMobile 300)

PR: kern/65777
Submitted by: Patrick Hurrelmann <outi@bytephobia.de>

20 years agoRemove mlockall() and munlockall() from the list of unimplemented
Maxime Henrion [Tue, 27 Apr 2004 13:13:20 +0000 (13:13 +0000)]
Remove mlockall() and munlockall() from the list of unimplemented
syscalls in a comment, we have them since quite some time now.

Noticed by: Stefan Farfeleder <stefan@fafoe.narf.at>

20 years agoDocument that '.' stops repetition of 's..." sequences.
Poul-Henning Kamp [Tue, 27 Apr 2004 13:11:20 +0000 (13:11 +0000)]
Document that '.' stops repetition of 's..." sequences.

This is useful in order to get visual indication that something did
not succeed:

# Give it 5 seconds to run
echo 'sjjjjjA.' > /dev/led/error
random_program
echo '0' > /dev/led/error

If random_program takes more than 5 seconds or if it hangs, the LED
will be turned on.

20 years agoAdd the ability to avoid repetition of s.... sequences with a '.'
Poul-Henning Kamp [Tue, 27 Apr 2004 13:09:21 +0000 (13:09 +0000)]
Add the ability to avoid repetition of s.... sequences with a '.'

20 years agoMake sure we are all set up before creating the LED instance.
Poul-Henning Kamp [Tue, 27 Apr 2004 13:08:03 +0000 (13:08 +0000)]
Make sure we are all set up before creating the LED instance.

20 years agoUse pci_get_progif to decide if this is one of the primary/secondary
Søren Schmidt [Tue, 27 Apr 2004 12:54:59 +0000 (12:54 +0000)]
Use pci_get_progif to decide if this is one of the primary/secondary
channels. This also work when PCI native mode has been selected
(patch for /sys/dev/pci/pci.c needed for that) since pci_get_progif
uses the saved value for progif, not the one stored after we may have
changed from legacy mode to native PCI mode.

20 years agoTake boot0sio from the newly-built world, both for consistency's sake and
Thomas Quinot [Tue, 27 Apr 2004 09:49:16 +0000 (09:49 +0000)]
Take boot0sio from the newly-built world, both for consistency's sake and
to cater for host systems where it is not present.

Reviewed by: phk

20 years agoMarkup nit.
Ruslan Ermilov [Tue, 27 Apr 2004 09:18:49 +0000 (09:18 +0000)]
Markup nit.

20 years agoThis commit was generated by cvs2svn to compensate for changes in r128671,
Mike Heffner [Tue, 27 Apr 2004 00:31:36 +0000 (00:31 +0000)]
This commit was generated by cvs2svn to compensate for changes in r128671,
which included commits to RCS files with non-trunk default branches.

20 years agoImport the latest (4/26/04) CVS lukemftp. This adds the ability to
Mike Heffner [Tue, 27 Apr 2004 00:31:36 +0000 (00:31 +0000)]
Import the latest (4/26/04) CVS lukemftp. This adds the ability to
have a prefix string on the progressbar and brings in a patch by mux@
to fix the condition when connect() returns EINTR.

Requested by:      mux

20 years agoUpdate write.c to follow recent API changes to libarchive.
Tim Kientzle [Mon, 26 Apr 2004 23:39:17 +0000 (23:39 +0000)]
Update write.c to follow recent API changes to libarchive.

Also, bump up the default copy buffer from 8k to 64k
to improve performance.

20 years agoUpdate file flag handling.
Tim Kientzle [Mon, 26 Apr 2004 23:37:54 +0000 (23:37 +0000)]
Update file flag handling.

The new fflags support in archive_entry supports Linux and FreeBSD
file flags and is a bit more gracious about unrecognized flag names
than strtofflags(3).  This involves some minor API breakage.

The default tar format ("restricted pax") now enables pax extensions
when archiving files that have flags.  In particular, copying dir
heirarchies with 'bsdtar cf - -C src . | bsdtar xpf - -C dest' now
preserves file flags.  (Note the "p" on extract!)

While I'm here, fill in some additional explanation in the
archive_entry.3 manpage, fill in some missing MLINKS, mark some
overlooked internal functions 'static', and make a few minor style
fixes.

20 years agoAdd rue(4) USB ethernet driver, which for some reason has been missed.
Maxim Sobolev [Mon, 26 Apr 2004 22:52:09 +0000 (22:52 +0000)]
Add rue(4) USB ethernet driver, which for some reason has been missed.

20 years agofix the change of interface in nd6_storelladdr for multicast
Luigi Rizzo [Mon, 26 Apr 2004 20:31:46 +0000 (20:31 +0000)]
fix the change of interface in nd6_storelladdr for multicast
addresses too.

Reported by: Jun Kuriyama

20 years agoGive jail(8) the feature to allow raw sockets from within a
Bosko Milekic [Mon, 26 Apr 2004 19:46:52 +0000 (19:46 +0000)]
Give jail(8) the feature to allow raw sockets from within a
jail, which is less restrictive but allows for more flexible
jail usage (for those who are willing to make the sacrifice).
The default is off, but allowing raw sockets within jails can
now be accomplished by tuning security.jail.allow_raw_sockets
to 1.

Turning this on will allow you to use things like ping(8)
or traceroute(8) from within a jail.

The patch being committed is not identical to the patch
in the PR.  The committed version is more friendly to
APIs which pjd is working on, so it should integrate
into his work quite nicely.  This change has also been
presented and addressed on the freebsd-hackers mailing
list.

Submitted by: Christian S.J. Peron <maneo@bsdpro.com>
PR: kern/65800

20 years agoRemoves the check for the existence of the sysctl variable
Simon L. B. Nielsen [Mon, 26 Apr 2004 19:41:37 +0000 (19:41 +0000)]
Removes the check for the existence of the sysctl variable
debug.watchdog since it is not created by hardware watchdog(4) devices.
The watchdog(4) device is always compiled in the kernel, so removing the
check should not cause any problems.

Approved by: phk

20 years agomake this compile/work with CISS_DEBUG defined.
Paul Saab [Mon, 26 Apr 2004 19:28:08 +0000 (19:28 +0000)]
make this compile/work with CISS_DEBUG defined.

20 years agoMFamd64: implement the PG_UNMANAGED flag on vm_pages.
Andrew Gallatin [Mon, 26 Apr 2004 17:49:05 +0000 (17:49 +0000)]
MFamd64: implement the PG_UNMANAGED flag on vm_pages.
Alan noticed that alpha ignored PG_UNMANAGED when we were looking into
pmap_emulate_reference() panics.  Perhaps this will help.

Reviewed by: alc

20 years agoSince pmap_pinit() wants a prezeroed page, change it to ask for one. (Note
Alan Cox [Mon, 26 Apr 2004 17:22:07 +0000 (17:22 +0000)]
Since pmap_pinit() wants a prezeroed page, change it to ask for one.  (Note
that vm_page_grab(), unlike vm_page_alloc(), always returns a prezeroed
page if VM_ALLOC_ZERO is specified.  So, checking is unnecessary.)

20 years agoAlways use nd.ni_vp->v_mount as an argument for VFS_QUOTACTL(), just like
Pawel Jakub Dawidek [Mon, 26 Apr 2004 15:44:42 +0000 (15:44 +0000)]
Always use nd.ni_vp->v_mount as an argument for VFS_QUOTACTL(), just like
in RELENG_4.

Pointed out by: Alex Lyashkov <umka@sevinter.net>

20 years agoThe previous change to mount(8) to report ufs or ufs2 used
Bosko Milekic [Mon, 26 Apr 2004 15:13:46 +0000 (15:13 +0000)]
The previous change to mount(8) to report ufs or ufs2 used
libufs, which only works for Charlie root.

This change reverts the introduction of libufs and moves the
check into the kernel.  Since the f_fstypename is the same
for both ufs and ufs2, we check fs_magic for presence of
ufs2 and copy "ufs2" explicitly instead.

Submitted by: Christian S.J. Peron <maneo@bsdpro.com>

20 years agoAdd 'enableWindowing' configuration knob to the ng_pptpgre(4) netgraph node.
Archie Cobbs [Mon, 26 Apr 2004 14:26:54 +0000 (14:26 +0000)]
Add 'enableWindowing' configuration knob to the ng_pptpgre(4) netgraph node.

Submitted by: Michael Bretterklieber <mbretter@a-quadrat.at>
MFC after: 2 weeks

20 years agoFix build breakage.
Takanori Watanabe [Mon, 26 Apr 2004 08:49:11 +0000 (08:49 +0000)]
Fix build breakage.

Submitted by: Xin LI <delphij@frontfree.net>
PR: 65979

20 years agoMove the call to AcpiEnterSleepStatePrep() to before we select the BSP
Nate Lawson [Mon, 26 Apr 2004 05:25:06 +0000 (05:25 +0000)]
Move the call to AcpiEnterSleepStatePrep() to before we select the BSP
(cpuid 0) as the processor.  It mallocs some data and smp_rendezvous
calls functions with locks held.

20 years agoAdd the bad reset statistic (corresponds to tcp_input.c rev 1.235)
Mike Silbersack [Mon, 26 Apr 2004 02:57:33 +0000 (02:57 +0000)]
Add the bad reset statistic (corresponds to tcp_input.c rev 1.235)

20 years agoTighten up reset handling in order to make reset attacks as difficult as
Mike Silbersack [Mon, 26 Apr 2004 02:56:31 +0000 (02:56 +0000)]
Tighten up reset handling in order to make reset attacks as difficult as
possible while maintaining compatibility with the widest range of TCP stacks.

The algorithm is as follows:

---
For connections in the ESTABLISHED state, only resets with
sequence numbers exactly matching last_ack_sent will cause a reset,
all other segments will be silently dropped.

For connections in all other states, a reset anywhere in the window
will cause the connection to be reset.  All other segments will be
silently dropped.
---

The necessity of accepting all in-window resets was discovered
by jayanth and jlemon, both of whom have seen TCP stacks that
will respond to FIN-ACK packets with resets not meeting the
strict last_ack_sent check.

Idea by:        Darren Reed
Reviewed by:    truckman, jlemon, others(?)

20 years agoFix two typos from PR: 65694
Warner Losh [Mon, 26 Apr 2004 02:11:38 +0000 (02:11 +0000)]
Fix two typos from PR: 65694

1) In pci.c, we need to check the child device's state, not the parent
   device's state.
2) In acpi_pci.c, we have to run the power state change after the acpi
   method when the old_state is > new state, not the other way around.

Submitted by: Dmitry Remesov
PR: 65694

20 years agoUse a more compact syntax for passing the "binary" options to 'ld'.
David E. O'Brien [Sun, 25 Apr 2004 20:36:44 +0000 (20:36 +0000)]
Use a more compact syntax for passing the "binary" options to 'ld'.

20 years agoRewrite split_lines() to operate safely
Andrey A. Chernov [Sun, 25 Apr 2004 19:56:50 +0000 (19:56 +0000)]
Rewrite split_lines() to operate safely

PR:             62694
Submitted by:   moulin p <moulin.p@calyopea.com>

20 years agoSimplify the building of our i386 'binary' boot components by directly
David E. O'Brien [Sun, 25 Apr 2004 19:50:20 +0000 (19:50 +0000)]
Simplify the building of our i386 'binary' boot components by directly
producing them using 'ld' options rather than post-processing with 'objcopy'.

Idea by: Ryan Sommers <ryans@gamersimpact.com>

20 years ago- Update description of watchdogd_enable to reflect current reality.
Simon L. B. Nielsen [Sun, 25 Apr 2004 17:13:22 +0000 (17:13 +0000)]
- Update description of watchdogd_enable to reflect current reality.
- Document watchdogd_flags.

20 years agoAnother small set of changes to reduce diffs with the new arp code.
Luigi Rizzo [Sun, 25 Apr 2004 15:00:17 +0000 (15:00 +0000)]
Another small set of changes to reduce diffs with the new arp code.

20 years agomdoc(7) janitor:
Simon L. B. Nielsen [Sun, 25 Apr 2004 14:13:48 +0000 (14:13 +0000)]
mdoc(7) janitor:
- Sort SEE ALSO by section number.
- Remove redundant ','.

20 years agomdoc(7) janitor:
Simon L. B. Nielsen [Sun, 25 Apr 2004 14:09:38 +0000 (14:09 +0000)]
mdoc(7) janitor:
- Remove EOL whitespace.
- Expand contracted words.
- Correct usage of .Fl macro.
- Use complete argument to -width.
- Remove redundant ','.

20 years agoremove a stale comment on the behaviour of arpresolve
Luigi Rizzo [Sun, 25 Apr 2004 14:06:23 +0000 (14:06 +0000)]
remove a stale comment on the behaviour of arpresolve

20 years agoStart the arp timer at init time.
Luigi Rizzo [Sun, 25 Apr 2004 12:50:14 +0000 (12:50 +0000)]
Start the arp timer at init time.
It runs so rarely that it makes no sense to wait until the first request.

20 years agoMerged from sys/isa/fd.c revision 1.270.
Yoshihiro Takahashi [Sun, 25 Apr 2004 12:43:44 +0000 (12:43 +0000)]
Merged from sys/isa/fd.c revision 1.270.

20 years agoAdd support Buffalo(MELCO) USB-Key Lan Adaptor(LUA-U2-KTX)
MIHIRA Sanpei Yoshiro [Sun, 25 Apr 2004 11:39:47 +0000 (11:39 +0000)]
Add support Buffalo(MELCO) USB-Key Lan Adaptor(LUA-U2-KTX)

20 years agoSync to 1.171 of usbdevs
MIHIRA Sanpei Yoshiro [Sun, 25 Apr 2004 11:24:40 +0000 (11:24 +0000)]
Sync to 1.171 of usbdevs

20 years agoAdd support Buffalo(MELCO) USB-Key Lan Adaptor(LUA-U2-KTX)
MIHIRA Sanpei Yoshiro [Sun, 25 Apr 2004 11:21:30 +0000 (11:21 +0000)]
Add support Buffalo(MELCO) USB-Key Lan Adaptor(LUA-U2-KTX)

Submitted by: KAWAI Kenichi <kawai.kenichi@canon.co.jp>

20 years agoThis commit does two things:
Luigi Rizzo [Sun, 25 Apr 2004 09:24:52 +0000 (09:24 +0000)]
This commit does two things:

1. rt_check() cleanup:
    rt_check() is only necessary for some address families to gain access
    to the corresponding arp entry, so call it only in/near the *resolve()
    routines where it is actually used -- at the moment this is
    arpresolve(), nd6_storelladdr() (the call is embedded here),
    and atmresolve() (the call is just before atmresolve to reduce
    the number of changes).
    This change will make it a lot easier to decouple the arp table
    from the routing table.

    There is an extra call to rt_check() in if_iso88025subr.c to
    determine the routing info length. I have left it alone for
    the time being.

    The interface of arpresolve() and nd6_storelladdr() now changes slightly:
     + the 'rtentry' parameter (really a hint from the upper level layer)
       is now passed unchanged from *_output(), so it becomes the route
       to the final destination and not to the gateway.
     + the routines will return 0 if resolution is possible, non-zero
       otherwise.
     + arpresolve() returns EWOULDBLOCK in case the mbuf is being held
       waiting for an arp reply -- in this case the error code is masked
       in the caller so the upper layer protocol will not see a failure.

2. arpcom untangling
    Where possible, use 'struct ifnet' instead of 'struct arpcom' variables,
    and use the IFP2AC macro to access arpcom fields.
    This mostly affects the netatalk code.

=== Detailed changes: ===
net/if_arcsubr.c
   rt_check() cleanup, remove a useless variable

net/if_atmsubr.c
   rt_check() cleanup

net/if_ethersubr.c
   rt_check() cleanup, arpcom untangling

net/if_fddisubr.c
   rt_check() cleanup, arpcom untangling

net/if_iso88025subr.c
   rt_check() cleanup

netatalk/aarp.c
   arpcom untangling, remove a block of duplicated code

netatalk/at_extern.h
   arpcom untangling

netinet/if_ether.c
   rt_check() cleanup (change arpresolve)

netinet6/nd6.c
   rt_check() cleanup (change nd6_storelladdr)

20 years agoAdd ng_sppp(4) to the modules build process.
Roman Kurakin [Sun, 25 Apr 2004 08:56:46 +0000 (08:56 +0000)]
Add ng_sppp(4) to the modules build process.

20 years agoAdd ng_sppp.4.
Roman Kurakin [Sun, 25 Apr 2004 08:52:26 +0000 (08:52 +0000)]
Add ng_sppp.4.

20 years agoZero the physical page only if it is invalid and not prezeroed.
Alan Cox [Sun, 25 Apr 2004 07:58:59 +0000 (07:58 +0000)]
Zero the physical page only if it is invalid and not prezeroed.

20 years agoFixed breakage of the formatting operation in rev.1.266. The wrong
Bruce Evans [Sun, 25 Apr 2004 04:33:56 +0000 (04:33 +0000)]
Fixed breakage of the formatting operation in rev.1.266.  The wrong
clause of an if-else statement was removed.

Reviewed by: no response from maintainer in 12 days

20 years agoDon't mask TCD in IMR0 when we initialize the channel. Doing so makes
Marcel Moolenaar [Sun, 25 Apr 2004 04:30:40 +0000 (04:30 +0000)]
Don't mask TCD in IMR0 when we initialize the channel. Doing so makes
it impossible to check the interrupt status bit when we try to get a
character in the low level console code.

20 years agoThe paper "Hashed Timers and Hierarchical Wheels: Data Structures for the
Hiten Pandya [Sun, 25 Apr 2004 04:10:17 +0000 (04:10 +0000)]
The paper "Hashed Timers and Hierarchical Wheels: Data Structures for the
Efficient Implementation of a Timer Facility" was co-author'ed by T. Lauk,
not A. Lauk.

Adjust nearby whitespace.

20 years agoHide FLT_EVAL_METHOD and DECIMAL_DIG in pre-C99 compilation
David Schultz [Sun, 25 Apr 2004 02:36:29 +0000 (02:36 +0000)]
Hide FLT_EVAL_METHOD and DECIMAL_DIG in pre-C99 compilation
environments.

PR: 63935
Submitted by: Stefan Farfeleder <stefan@fafoe.narf.at>

20 years agoMake sure that symbols are declared in math.h iff the appropriate
David Schultz [Sun, 25 Apr 2004 02:35:42 +0000 (02:35 +0000)]
Make sure that symbols are declared in math.h iff the appropriate
namespaces are visible.  Previously, math.h failed to hide some C99-,
XSI-, and BSD-specific symbols in certain compilation environments.

The referenced PR has a nice listing of the appropriate conditions for
making symbols visible in math.h.  The only non-stylistic difference
between the patch in the PR and this commit is that I superfluously
test for __BSD_VISIBLE in a few places to be more explicit about which
symbols have historically been part of the FreeBSD environment.

PR: 65939
Submitted by: Stefan Farfeleder <stefan@fafoe.narf.at>

20 years agoRemove a stale comment referring to values.h, which has never been
David Schultz [Sun, 25 Apr 2004 02:32:46 +0000 (02:32 +0000)]
Remove a stale comment referring to values.h, which has never been
part of FreeBSD.

PR: 65939

20 years agofix one typo and remove one wrong line
Luigi Rizzo [Sun, 25 Apr 2004 01:39:00 +0000 (01:39 +0000)]
fix one typo and remove one wrong line