]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agothe default password policy for xdm should be pam_deny, since it is
Dag-Erling Smørgrav [Fri, 20 Feb 2004 21:59:51 +0000 (21:59 +0000)]
the default password policy for xdm should be pam_deny, since it is
incapable of holding a meaningful conversation.

20 years agoMention that securelevel 1 also blocks access to /dev/io if it
Maxime Henrion [Fri, 20 Feb 2004 21:38:23 +0000 (21:38 +0000)]
Mention that securelevel 1 also blocks access to /dev/io if it
exists (not all platforms have it).

20 years agoAdd somemore items to the list. The purpose is not to extert control over
Scott Long [Fri, 20 Feb 2004 21:27:04 +0000 (21:27 +0000)]
Add somemore items to the list.  The purpose is not to extert control over
these, but to let people know that they are maintained.

20 years agoDo not test if pDCB is not NULL, we dereference it before anyway, and it
Olivier Houchard [Fri, 20 Feb 2004 20:36:55 +0000 (20:36 +0000)]
Do not test if pDCB is not NULL, we dereference it before anyway, and it
should not happen. Add a KASSERT instead.

Reported by: Ted Unangst <tedu@coverity.com>
Spotted out by: cperciva

20 years agoadd destination address selection support for getipnodebyname(3).
Hajimu UMEMOTO [Fri, 20 Feb 2004 17:59:33 +0000 (17:59 +0000)]
add destination address selection support for getipnodebyname(3).
though getipnodebyname(3) is obsoleted api, some major applications
such as Mozilla are still using it.  so, it will help ipv4 users.

20 years agowhitespace changes only (prepare for merge from citi tree)
Jim Rees [Fri, 20 Feb 2004 16:57:36 +0000 (16:57 +0000)]
whitespace changes only (prepare for merge from citi tree)

20 years agoAdd my getaddrinfo(3) stress-tester as gaithrstress. The most obvious
Brian Feldman [Fri, 20 Feb 2004 16:54:01 +0000 (16:54 +0000)]
Add my getaddrinfo(3) stress-tester as gaithrstress.  The most obvious
regressions would be to see the program or your kernel crashing.

If you want to give it something to really test out, try a much more
reentrant version of the resolver.
<URL:http://green.homeunix.org/~green/reentrant_resolver.patch>
Any Mozilla-based browser would show you a clear difference.

20 years agomdoc(7) guru's glance.
Ruslan Ermilov [Fri, 20 Feb 2004 16:05:38 +0000 (16:05 +0000)]
mdoc(7) guru's glance.

20 years agoAdd a belated entry for the CPU_ENABLE_LONGRUN issue.
Warner Losh [Fri, 20 Feb 2004 15:25:55 +0000 (15:25 +0000)]
Add a belated entry for the CPU_ENABLE_LONGRUN issue.

20 years agoLabel @ as being AC (since it's constructed like a prosign).
Tony Finch [Fri, 20 Feb 2004 13:46:39 +0000 (13:46 +0000)]
Label @ as being AC (since it's constructed like a prosign).

Submitted by: des

20 years agoMerge from RELENG_4 1.28.2.13: Fix saving system crash dumps larger than
Tim J. Robbins [Fri, 20 Feb 2004 12:22:52 +0000 (12:22 +0000)]
Merge from RELENG_4 1.28.2.13: Fix saving system crash dumps larger than
2 GB by using fseeko() instead of fseek().

20 years agoAdd some missing punctuation symbols, in particular @ which (according to
Tony Finch [Fri, 20 Feb 2004 11:55:38 +0000 (11:55 +0000)]
Add some missing punctuation symbols, in particular @ which (according to
a recent Press Association release) was added to Morse code by the ITU in
December. @ was previously used for the SK prosign, which is now spelt &.

Obtained from:  http://en.wikipedia.org/wiki/Morse_code

20 years agoUnbreak the upgrade path from 4.9 after removal of GNU getopt and
Ruslan Ermilov [Fri, 20 Feb 2004 11:55:14 +0000 (11:55 +0000)]
Unbreak the upgrade path from 4.9 after removal of GNU getopt and
<gnuregex.h>.

20 years agoNormalize polling interval while the target is offline.
Hidetoshi Shimokawa [Fri, 20 Feb 2004 10:59:46 +0000 (10:59 +0000)]
Normalize polling interval while the target is offline.

20 years agoList EPSON Perfection 3200 as supported.
Wilko Bulte [Fri, 20 Feb 2004 09:54:32 +0000 (09:54 +0000)]
List EPSON Perfection 3200 as supported.

Reminded by: simon

20 years agoNew failure detection algorithm was backed out, so backout manual
Pawel Jakub Dawidek [Fri, 20 Feb 2004 09:52:26 +0000 (09:52 +0000)]
New failure detection algorithm was backed out, so backout manual
page change as well.

Requested by: scottl (mentor)

20 years agoBackout previous change, it breaks build and it is not needed
Pawel Jakub Dawidek [Fri, 20 Feb 2004 08:26:27 +0000 (08:26 +0000)]
Backout previous change, it breaks build and it is not needed
layering violation. As pointed out, there is much better way to do this.
Sorry guys, I need to find a better way to force reviews.

Requested by: harti, julian, scottl (mentor)
Pointy hat to: pjd

20 years agoDocument dependence of mktemp(3) on the non-reentrant arc4random(3).
Brian Feldman [Fri, 20 Feb 2004 04:08:28 +0000 (04:08 +0000)]
Document dependence of mktemp(3) on the non-reentrant arc4random(3).

20 years agoMake sure to wake up any select waiters when closing a kqueue (also, not
Brian Feldman [Fri, 20 Feb 2004 04:00:48 +0000 (04:00 +0000)]
Make sure to wake up any select waiters when closing a kqueue (also, not
crash).  I am fairly sure that only people with SMP and multi-threaded
apps using kqueue will be affected by this, so I have a stress-testing
program on my web site:
<URL:http://green.homeunix.org/~green/getaddrinfo-pthreads-stresstest.c>

20 years agoMoving forward, there's no reason to exclude libpthread on sparc64.
Marcel Moolenaar [Fri, 20 Feb 2004 03:42:34 +0000 (03:42 +0000)]
Moving forward, there's no reason to exclude libpthread on sparc64.

20 years agoChange the 'realclean' target name to be 'cleanworld', to match other
Garance A Drosehn [Fri, 20 Feb 2004 02:10:35 +0000 (02:10 +0000)]
Change the 'realclean' target name to be 'cleanworld', to match other
targets like buildworld, installworld.

Suggested by: obrien

20 years agoFix a glitch in my last commit and revert to using selwakeuppri
Mathew Kanner [Fri, 20 Feb 2004 01:24:57 +0000 (01:24 +0000)]
Fix a glitch in my last commit and revert to using selwakeuppri

Noticed by: tanimura
Noticed by: truckman

20 years agoMake uscanner recognise EPSON Perfection 3200. Tested with xsane.
Wilko Bulte [Thu, 19 Feb 2004 22:05:10 +0000 (22:05 +0000)]
Make uscanner recognise EPSON Perfection 3200. Tested with xsane.

PR: kern/63041
MFC after: 3 days

20 years agoTidy up the thread taskqueue implementation and close a lost wakeup race.
John Baldwin [Thu, 19 Feb 2004 22:03:52 +0000 (22:03 +0000)]
Tidy up the thread taskqueue implementation and close a lost wakeup race.
Instead of creating a mutex that we msleep on but don't actually lock when
doing the corresponding wakeup(), in the kthread, lock the mutex associated
with our taskqueue and msleep while the queue is empty.  Assert that the
queue is locked when the callback function is called to wake the kthread.

20 years agoAdd EPSON Perfection 3200 scanner.
Wilko Bulte [Thu, 19 Feb 2004 22:01:28 +0000 (22:01 +0000)]
Add EPSON Perfection 3200 scanner.

20 years agoAdd EPSON Perfection 3200 scanner
Wilko Bulte [Thu, 19 Feb 2004 21:58:23 +0000 (21:58 +0000)]
Add EPSON Perfection 3200 scanner

20 years agoImprovements on the 'make realclean' target. Some are style improvements,
Garance A Drosehn [Thu, 19 Feb 2004 21:09:58 +0000 (21:09 +0000)]
Improvements on the 'make realclean' target.  Some are style improvements,
but the biggest issue is that there are situatons when
${.OBJDIR} == ${.SRCDIR}, and in those situations the previous version
would happily remove all your /usr/src while it was cleaning out the objects.
Not that *you* would be happy about it...  Thanks to bde for immediately
noticing this serious possibility.

More improvements will be made to this target, but I wanted to commit this
safer version right now, before anyone lost their /usr/src due to it.

Reviewed by: bde
MFC after: 10 days

20 years agoRework jail_attach(2) so that an already jailed process cannot hop
Jacques Vidrine [Thu, 19 Feb 2004 21:03:20 +0000 (21:03 +0000)]
Rework jail_attach(2) so that an already jailed process cannot hop
to another jail.

Submitted by: rwatson

20 years agoAllow to compile with non-GCC compiler.
David E. O'Brien [Thu, 19 Feb 2004 20:02:38 +0000 (20:02 +0000)]
Allow to compile with non-GCC compiler.

20 years agos/bytes/byte/
Poul-Henning Kamp [Thu, 19 Feb 2004 19:53:07 +0000 (19:53 +0000)]
s/bytes/byte/

20 years agoDocument the acpi_toshiba driver written by Hiroyuki Aizu.
Philip Paeps [Thu, 19 Feb 2004 19:35:27 +0000 (19:35 +0000)]
Document the acpi_toshiba driver written by Hiroyuki Aizu.

Approved by: njl (mentor)

20 years agoReport the correct length for symlink entries.
Poul-Henning Kamp [Thu, 19 Feb 2004 19:09:52 +0000 (19:09 +0000)]
Report the correct length for symlink entries.

20 years agoClarify the "lp" description to mention port@host syntax.
Tim Kientzle [Thu, 19 Feb 2004 18:55:55 +0000 (18:55 +0000)]
Clarify the "lp" description to mention port@host syntax.

Suggested by: NetBSD man page

20 years agoAdd back an include to fix the build for the CPU_ELAN case.
John Baldwin [Thu, 19 Feb 2004 18:34:26 +0000 (18:34 +0000)]
Add back an include to fix the build for the CPU_ELAN case.

20 years agoUse ACPI_NEXT_RESOURCE instead of defining our own copy. The one provided
Nate Lawson [Thu, 19 Feb 2004 18:20:03 +0000 (18:20 +0000)]
Use ACPI_NEXT_RESOURCE instead of defining our own copy.  The one provided
with ACPI-CA is identical now.

20 years agoDo not remove the fixed handlers. Several systems (e.g., ASUS) only
Nate Lawson [Thu, 19 Feb 2004 18:16:34 +0000 (18:16 +0000)]
Do not remove the fixed handlers.  Several systems (e.g., ASUS) only
return events on the fixed handler even after defining a duplicate in the
AML.  While this violates the spec, hopefully we can get by with leaving
both installed.

20 years agoUpdate manual page (new failure detection algorithm was added).
Pawel Jakub Dawidek [Thu, 19 Feb 2004 17:05:32 +0000 (17:05 +0000)]
Update manual page (new failure detection algorithm was added).

Approved by: scottl (mentor)

20 years agoAdd new failure detection algorithm.
Pawel Jakub Dawidek [Thu, 19 Feb 2004 17:04:23 +0000 (17:04 +0000)]
Add new failure detection algorithm.
It works as follows:
In every 'interval' seconds defined links are checked.
If they are non-active they will not be used by to data transfer.

No response from: julian, archie
Silent on: net@
Approved by: scottl (mentor)

20 years agoExport private structure owned by ng_ether(4) module outside.
Pawel Jakub Dawidek [Thu, 19 Feb 2004 16:58:01 +0000 (16:58 +0000)]
Export private structure owned by ng_ether(4) module outside.
It'll is required by new failure detection algorithm for ng_one2many(4).

No response from: julian, archie
Silent on: net@
Approved by: scottl (mentor)

20 years agoAdd control utility for disk concatenation (GEOM_CONCAT class).
Pawel Jakub Dawidek [Thu, 19 Feb 2004 16:02:08 +0000 (16:02 +0000)]
Add control utility for disk concatenation (GEOM_CONCAT class).

Reviewed by: phk, scottl
Approved by: scottl (mentor)

20 years agoTurn non-PAM password authentication off by default when USE_PAM is
Dag-Erling Smørgrav [Thu, 19 Feb 2004 15:53:31 +0000 (15:53 +0000)]
Turn non-PAM password authentication off by default when USE_PAM is
defined.  Too many users are getting bitten by it.

20 years agoMakefile needed for building geom_concat module.
Pawel Jakub Dawidek [Thu, 19 Feb 2004 15:43:58 +0000 (15:43 +0000)]
Makefile needed for building geom_concat module.

Approved by: scottl (mentor)

20 years agoIntroduce CONCAT GEOM class for disk concatenation.
Pawel Jakub Dawidek [Thu, 19 Feb 2004 15:19:49 +0000 (15:19 +0000)]
Introduce CONCAT GEOM class for disk concatenation.
It allows manual and automatic (based on on-disk metadata) concatenation.

Reviewed by: phk, scottl
Approved by: scottl (mentor)

20 years ago- call ip6_output() instead of nd6_output() when ipsec tunnel
Hajimu UMEMOTO [Thu, 19 Feb 2004 14:57:22 +0000 (14:57 +0000)]
- call ip6_output() instead of nd6_output() when ipsec tunnel
  mode is applied, since tunneled packets are considered to be
  generated packets from a tunnel encapsulating node.
- tunnel mode may not be applied if SA mode is ANY and policy
  does not say "tunnel it".  check if we have extra IPv6 header
  on the packet after ipsec6_output_tunnel() and call ip6_output()
  only if additional IPv6 header is added.
- free the copyed packet before returning.

Obtained from: KAME

20 years agoPath correction.
Pawel Jakub Dawidek [Thu, 19 Feb 2004 14:32:44 +0000 (14:32 +0000)]
Path correction.

Approved by: scottl (mentor)

20 years agoAdded sysctl security.jail.jailed.
Pawel Jakub Dawidek [Thu, 19 Feb 2004 14:29:14 +0000 (14:29 +0000)]
Added sysctl security.jail.jailed.
It returns 1 is process is inside of jail and 0 if it is not.
Information if we are in jail or not is not a secret, there is plenty of
ways to discover it. Many people are using own hack to check this and
this will be a legal way from now on.

It will be great if our starting scripts will take advantage of this sysctl
to allow clean "boot" inside jail.

Approved by: rwatson, scottl (mentor)

20 years agoSimplify check. We are only able to check exclusive lock and if
Pawel Jakub Dawidek [Thu, 19 Feb 2004 14:19:31 +0000 (14:19 +0000)]
Simplify check. We are only able to check exclusive lock and if
2nd condition is true, first one is true for sure.

Approved by: jhb, scottl (mentor)

20 years agoFixed ucred structure leak.
Pawel Jakub Dawidek [Thu, 19 Feb 2004 14:13:21 +0000 (14:13 +0000)]
Fixed ucred structure leak.

Approved by: scottl (mentor)
PR: 54163
MFC after: 3 days

20 years agoImplement PThreads barriers and barrier attributes.
Mike Makonnen [Thu, 19 Feb 2004 13:51:52 +0000 (13:51 +0000)]
Implement PThreads barriers and barrier attributes.

20 years agoDon't wake up the thread after the signal handler
Mike Makonnen [Thu, 19 Feb 2004 13:47:12 +0000 (13:47 +0000)]
Don't wake up the thread after the signal handler
has been executed. On return from the signal handler
the call will either be restarted or EINTR will be returned,
but it will not go back to its previous state. So, it is
sufficient to simply change the state to 'running' without
actually trying to wake up the thread.

20 years agoMFi386: revision 1.466
Yoshihiro Takahashi [Thu, 19 Feb 2004 13:10:39 +0000 (13:10 +0000)]
MFi386: revision 1.466

20 years agoAdd BSD compatibility tty ioctls LINUX_TIOCSBRK and LINUX_TIOCCBRK. This
Bruce M Simpson [Thu, 19 Feb 2004 12:38:12 +0000 (12:38 +0000)]
Add BSD compatibility tty ioctls LINUX_TIOCSBRK and LINUX_TIOCCBRK. This
addition appears to allow VMware 3 Workstation to operate with nmdm(4)
as a virtual COM device.

Tested by: Guido van Rooij

20 years agoImprove formatting.
Ruslan Ermilov [Thu, 19 Feb 2004 12:02:54 +0000 (12:02 +0000)]
Improve formatting.

20 years agoWhen reparenting a process in the PT_DETACH code, only set p_sigparent
Don Lewis [Thu, 19 Feb 2004 10:39:42 +0000 (10:39 +0000)]
When reparenting a process in the PT_DETACH code, only set p_sigparent
to SIGCHLD if the new parent process is initproc.

MFC after: 2 weeks

20 years agoUse size_t or ssize_t wherever appropriate instead of casting from int *
Tim J. Robbins [Thu, 19 Feb 2004 09:56:58 +0000 (09:56 +0000)]
Use size_t or ssize_t wherever appropriate instead of casting from int *
to size_t *, which is incorrect because they may have different widths.
This caused some subtle forms of corruption, the mostly frequently
reported one being that the last character of a filename was sometimes
duplicated on amd64.

20 years agoEnforce the file size limit in VOP_WRITE() as well as VOP_TRUNCATE();
Tim J. Robbins [Thu, 19 Feb 2004 09:06:06 +0000 (09:06 +0000)]
Enforce the file size limit in VOP_WRITE() as well as VOP_TRUNCATE();
pointed out by bde.

20 years ago - Correct a long-standing race condition in vm_page_try_to_free() that
Alan Cox [Thu, 19 Feb 2004 07:43:55 +0000 (07:43 +0000)]
 - Correct a long-standing race condition in vm_page_try_to_free() that
   could result in a dirty page being unintentionally freed.
 - Simplify the dirty page check in vm_page_dontneed().

Reviewed by: tegge
MFC after: 7 days

20 years agoDon't forget to enable the ipv6 firewall once the rules are loaded.
Mike Makonnen [Thu, 19 Feb 2004 06:53:24 +0000 (06:53 +0000)]
Don't forget to enable the ipv6 firewall once the rules are loaded.

PR: misc/61501
Submitted by: Roderick van Domburg <r.s.a.vandomburg@student.utwente.nl>

20 years agoA Linux thread created using clone() should not send SIGCHLD to its
Don Lewis [Thu, 19 Feb 2004 06:43:48 +0000 (06:43 +0000)]
A Linux thread created using clone() should not send SIGCHLD to its
parent if no signal is specified in the clone() flags argument.

PR: 42457
MFC after: 2 weeks

20 years agoIn-line documentation.
Mike Makonnen [Thu, 19 Feb 2004 05:49:06 +0000 (05:49 +0000)]
In-line documentation.

Prodded by: alfred

20 years agoFix problem caused by previous commit where some users' buttons
Nate Lawson [Thu, 19 Feb 2004 05:35:20 +0000 (05:35 +0000)]
Fix problem caused by previous commit where some users' buttons
stopped returning events.  Don't disable the event when removing
the handler because it still needs to be enabled for the other
handler.  Also, remove duplicate AcpiEnableEvent calls since the
install function now does this for us.

20 years agoAdd support for 'h' and 'hh' modifiers for printf(9).
Nate Lawson [Thu, 19 Feb 2004 05:29:39 +0000 (05:29 +0000)]
Add support for 'h' and 'hh' modifiers for printf(9).

Submitted by: Bruno Ducrot <ducrot AT poupinou.org>
Reviewed by: bde

20 years agoCheckpoint the NOTES I was working on.
David E. O'Brien [Thu, 19 Feb 2004 04:39:14 +0000 (04:39 +0000)]
Checkpoint the NOTES I was working on.

20 years agoUnbreak fixit crunch: since Feb/17/2004, mount(8) requires libufs.
Makoto Matsushita [Thu, 19 Feb 2004 02:48:46 +0000 (02:48 +0000)]
Unbreak fixit crunch: since Feb/17/2004, mount(8) requires libufs.
See also src/sbin/mount/Makefile rev 1.14.

Noticed by: snapshots.jp.FreeBSD.org

20 years agoFix a long-standing bug where select on vchans doesn't work
Mathew Kanner [Thu, 19 Feb 2004 01:07:15 +0000 (01:07 +0000)]
Fix a long-standing bug where select on vchans doesn't work
(never wake up) by iterating over them when they exist.

Approved by: tanimura (mentor)

20 years agoSwitch to using the new $PIR interrupt routing code and remove the old
John Baldwin [Wed, 18 Feb 2004 22:41:53 +0000 (22:41 +0000)]
Switch to using the new $PIR interrupt routing code and remove the old
code.  The pci_cfgreg.c file now just controls reading/writing PCI config
registers.

20 years agoRework the $PIR (aka PCIBIOS) PCI interrupt routing code and split it off
John Baldwin [Wed, 18 Feb 2004 22:40:23 +0000 (22:40 +0000)]
Rework the $PIR (aka PCIBIOS) PCI interrupt routing code and split it off
into its own file:
- All of the $PIR interrupt routing is now done in a link-centric fashion.
  When a host-PCI bridge that uses the $PIR attaches, it calls pir_parse()
  to parse the table.  This scans for link devices and merges all the masks
  for each link device from the table entries.  It then looks at the intline
  register of PCI devices connected to a link to figure out if the BIOS has
  routed this link and if so to which IRQ.
- The IRQ for any given link can be overridden via a hint like so:
  'hw.pci.link.0x62.irq=10'  Any IRQ set in this matter is treated as if it
  were set that way by the BIOS.
- We only call the BIOS to route each link device once.
- When a PCI device wants to route an interrupt, we look it up in the $PIR
  to find the associated link.  If the link is routed, we simply return the
  IRQ it is using.  If it is not routed, we have to pick one.  This uses a
  different algorithm from the old code.  First off, when we try to pick
  an interrupt from a mask of possible interrupts, we try to pick the one
  that is least loaded as far as PCI devices.  We maintain this weight based
  on the number of devices attached to each link device.  When choosing an
  IRQ, we first attempt to route using any PCI only interrupts (the old
  code did this as well).  If that doesn't work, we try to use the list of
  IRQs that the BIOS has used.  This is a new step that the new code didn't
  do and avoids using IRQ 3 or 4 for every virgin interrupt routing.  If
  none of the IRQs that the BIOS used worked, then we fall back to trying
  anything.
- The fallback mask for !PC98 was fixed to include IRQ 3 and not allow IRQ
  2.
- We don't use the $PIR to route interrupts on a PCI-PCI bridge unless it
  has already been used to route on at least one Host-PCI bridge.  This
  helps to avoid mixing and matching x86 firmware PCI interrupt routing
  methods (which is a Bad Thing(tm)).

Silence on: current@

20 years agoImprove contents, ruin formatting.
Poul-Henning Kamp [Wed, 18 Feb 2004 22:10:08 +0000 (22:10 +0000)]
Improve contents, ruin formatting.

20 years agoInstall gnuregex.h as wrapper for gnu/regex.h (with warning)
Andrey A. Chernov [Wed, 18 Feb 2004 22:09:34 +0000 (22:09 +0000)]
Install gnuregex.h as wrapper for gnu/regex.h (with warning)

20 years agoUse %zu for size_t, like God intended.
Dag-Erling Smørgrav [Wed, 18 Feb 2004 21:51:02 +0000 (21:51 +0000)]
Use %zu for size_t, like God intended.

20 years agoFix printf() format bug in previous commit (size_t != int).
Dag-Erling Smørgrav [Wed, 18 Feb 2004 21:37:36 +0000 (21:37 +0000)]
Fix printf() format bug in previous commit (size_t != int).
Use %ju and an (uintmax_t) cast to print size_t values.
Use %jd and an (intmax_t) cast to print off_t values.
Use off_t variables to hold the difference between two off_t values.
Don't bother with unsigned char where char will do nicely.

20 years agoChange the disk(9) API in order to make device removal more robust.
Poul-Henning Kamp [Wed, 18 Feb 2004 21:36:53 +0000 (21:36 +0000)]
Change the disk(9) API in order to make device removal more robust.

Previously the "struct disk" were owned by the device driver and this
gave us problems when the device disappared and the users of that device
were not immediately disappearing.

Now the struct disk is allocate with a new call, disk_alloc() and owned
by geom_disk and just abandonned by the device driver when disk_create()
is called.

Unfortunately, this results in a ton of "s/\./->/" changes to device
drivers.

Since I'm doing the sweep anyway, a couple of other API improvements
have been carried out at the same time:

The Giant awareness flag has been flipped from DISKFLAG_NOGIANT to
DISKFLAG_NEEDSGIANT

A version number have been added to disk_create() so that we can detect,
report and ignore binary drivers with old ABI in the future.

Manual page update to follow shortly.

20 years agoAdd */lib/getopt* I miss somehow initially.
Andrey A. Chernov [Wed, 18 Feb 2004 18:53:13 +0000 (18:53 +0000)]
Add */lib/getopt* I miss somehow initially.

20 years agoCorrect minor typos.
Ceri Davies [Wed, 18 Feb 2004 18:52:09 +0000 (18:52 +0000)]
Correct minor typos.

20 years agoOnly call chflags() on directories once.
Ceri Davies [Wed, 18 Feb 2004 18:45:11 +0000 (18:45 +0000)]
Only call chflags() on directories once.

Approved by: ru
MFC after: 1 week

20 years agoRemove thr_getschedparam.c since it's contents have been moved into
Mike Makonnen [Wed, 18 Feb 2004 15:59:54 +0000 (15:59 +0000)]
Remove thr_getschedparam.c since it's contents have been moved into
thr_setschedparam.c

20 years agoAdd a -n option that stops ip6fw making any changes to the rules
David Malone [Wed, 18 Feb 2004 15:56:53 +0000 (15:56 +0000)]
Add a -n option that stops ip6fw making any changes to the rules
in the kernel.

Submitted by: Orla McGann <orly@redbrick.dcu.ie>
MFC after: 3 weeks

20 years agoThere are consumers of rwlocks, inluding our own libc, that depend on
Mike Makonnen [Wed, 18 Feb 2004 15:30:10 +0000 (15:30 +0000)]
There are consumers of rwlocks, inluding our own libc, that depend on
a PTHREAD_RWLOCK_INITIALIZER to do for rwlocks what
a similarly named symbol does for statically initialized mutexes.
This symbol was dropped in The Open Group Base Specifications Issue 6
and does not exist in IEEE Std 1003.1, 2003, but it should still be
supported for backwards compatibility.

Pointy hat: mtm

20 years agoo Catch up with the mutex priority protocol fixes.
Mike Makonnen [Wed, 18 Feb 2004 15:26:00 +0000 (15:26 +0000)]
o Catch up with the mutex priority protocol fixes.
o Move pthread_getschedparam() into the same file with it's
  pthread_set* counterpart. Copyright on both files is identical.

20 years agoo Style
Mike Makonnen [Wed, 18 Feb 2004 15:22:52 +0000 (15:22 +0000)]
o Style
o Instead of checking both the passed in pointer and its value
  for NULL, only check the latter. Any caller that passes in
  a NULL pointer is obviously wrong.

20 years agoo Refactor and, among other things, get rid of insane nesting levels.
Mike Makonnen [Wed, 18 Feb 2004 15:16:31 +0000 (15:16 +0000)]
o Refactor and, among other things, get rid of insane nesting levels.
o Fix mutex priority protocols. Keep separate counts of priority
  inheritance and protection mutexes to make things easier.
  This will not have much affect since this is only the
  userland side, and the rest involves kernel scheduling.

20 years agoRevamp the statistics code, and switch to a much more compact display
Dag-Erling Smørgrav [Wed, 18 Feb 2004 15:12:00 +0000 (15:12 +0000)]
Revamp the statistics code, and switch to a much more compact display
format.  The old code tried to produce the exact same output as the
pre-libfetch implementation, but I no longer see any value in this.

20 years agoMove the initialization of thread priority to a common function.
Mike Makonnen [Wed, 18 Feb 2004 15:05:56 +0000 (15:05 +0000)]
Move the initialization of thread priority to a common function.

20 years agoMove the weak references to the top of the file to conform
Mike Makonnen [Wed, 18 Feb 2004 14:56:20 +0000 (14:56 +0000)]
Move the weak references to the top of the file to conform
to the format of other similar files in libthr.

20 years agoAdd partial support for large (>4GB) files on ext2 filesystems. This
Tim J. Robbins [Wed, 18 Feb 2004 14:08:25 +0000 (14:08 +0000)]
Add partial support for large (>4GB) files on ext2 filesystems. This
support is partial in that it will refuse to create large files on
filesystems that haven't been upgraded to EXT2_DYN_REV or that don't
have the EXT2_FEATURE_RO_COMPAT_LARGE_FILE flag set in the superblock.

MFC after: 2 weeks

20 years agoo Fix a bridge example: sysctl net.inet.ip.forwarding=1 is needless,
Maxim Konovalov [Wed, 18 Feb 2004 08:30:08 +0000 (08:30 +0000)]
o Fix a bridge example: sysctl net.inet.ip.forwarding=1 is needless,
bridge(4) is working on a level below.  Suggest
sysctl net.inet.ip.check_interface=0 instead.

MFC after:      2 weeks

20 years agoAdd "-q" argument to setfmac and setfsmac to allow the patient but
Robert Watson [Wed, 18 Feb 2004 05:40:15 +0000 (05:40 +0000)]
Add "-q" argument to setfmac and setfsmac to allow the patient but
exhausted reader not to see non-fatal warnings.

20 years agoUse correct thousands separator for the Italian locale.
Colin Percival [Wed, 18 Feb 2004 05:23:09 +0000 (05:23 +0000)]
Use correct thousands separator for the Italian locale.

PR: misc/54477
Submitted by: Nicola Vitale <nivit@libero.it>
Approved by: rwatson (mentor)
Confirmed by:  #freebsd, Google

20 years agoDon't ignore errors from vfs_allocate_syncvnode.
Colin Percival [Wed, 18 Feb 2004 05:20:54 +0000 (05:20 +0000)]
Don't ignore errors from vfs_allocate_syncvnode.

PR: kern/18503
Submitted by: Anatoly Vorobey <mellon@pobox.com>
Approved by: rwatson (mentor)

20 years agoMention that kernel environment variables may be configured in
Wes Peters [Wed, 18 Feb 2004 05:20:31 +0000 (05:20 +0000)]
Mention that kernel environment variables may be configured in
/boot/loader.conf.  Make the description flow a bit better.

Complaints by: Nate Lawson <nate@root.org>

20 years agoFix misspellings in the freebsd-tips fortunes. The same PR has
Colin Percival [Wed, 18 Feb 2004 05:18:27 +0000 (05:18 +0000)]
Fix misspellings in the freebsd-tips fortunes.  The same PR has
a patch to mention portupgrade in freebsd-tips as well; I'm not
sure if that belongs here, so I'll leave it for someone else.

PR: misc/37073
Submitted by: Sean Chittenden <sean@chittenden.org>
Approved by: rwatson (mentor)

20 years agoFixed some style bugs (spaces instead of tabs in macro definitions ...).
Bruce Evans [Wed, 18 Feb 2004 02:56:14 +0000 (02:56 +0000)]
Fixed some style bugs (spaces instead of tabs in macro definitions ...).

20 years agoCheckpoint a hack to enable running i386 libc_r binaries on a 64 bit
Peter Wemm [Wed, 18 Feb 2004 00:54:17 +0000 (00:54 +0000)]
Checkpoint a hack to enable running i386 libc_r binaries on a 64 bit
kernel.  I'm not happy with it yet - refinements are to come.
This hack allows the kern.ps_strings and kern.usrstack sysctls to respond
to a 32 bit request, such as those coming from emulated i386 binaries.

20 years agoBackout MT_TAG removal (i.e. bring back MT_TAGs) for now, as dummynet is
Max Laier [Wed, 18 Feb 2004 00:04:52 +0000 (00:04 +0000)]
Backout MT_TAG removal (i.e. bring back MT_TAGs) for now, as dummynet is
not working properly with the patch in place.

Approved by: bms(mentor)

20 years agoremove dead code
Jim Rees [Tue, 17 Feb 2004 22:28:26 +0000 (22:28 +0000)]
remove dead code

Approved by: alfred

20 years agoAdd vector for memmove() (currently aliased to memcpy()) a implement
Bill Paul [Tue, 17 Feb 2004 21:50:39 +0000 (21:50 +0000)]
Add vector for memmove() (currently aliased to memcpy()) a implement
ExInterlockedAddLargeStatistic().

20 years agoMake the distinction between sysctl kernel environment variables and
Wes Peters [Tue, 17 Feb 2004 20:38:49 +0000 (20:38 +0000)]
Make the distinction between sysctl kernel environment variables and
device.hints kernel environment variables when disabling ACPI.

Some users are getting confused when they see hint.acpi.0.disabled.
They are confusing it with a sysctl MIB.

Finally, add a reference for device.hints(5) and kenv(1)
in the "SEE ALSO" section of the man page.

PR: docs/62974
Submitted by: Christian S.J. Peron <maneo@bsdpro.com>

20 years agoDont use the bio_taskqueue if we are in timeout.
Søren Schmidt [Tue, 17 Feb 2004 19:24:11 +0000 (19:24 +0000)]
Dont use the bio_taskqueue if we are in timeout.
Use taskqueue_thread rather than taskqueue_swi (maybe we should have
a taskqueue_ata).

20 years agol2ping(8) should not loose echo data if target does not responds to the ping
Maksim Yevmenkin [Tue, 17 Feb 2004 18:37:12 +0000 (18:37 +0000)]
l2ping(8) should not loose echo data if target does not responds to the ping

Reviewed by: imp (mentor), ru

20 years agoDo not place dirmask in unnamed padding. Move it to the bottom of this
Tom Rhodes [Tue, 17 Feb 2004 16:53:41 +0000 (16:53 +0000)]
Do not place dirmask in unnamed padding.  Move it to the bottom of this
list where it should have been added originally.

Prodded by: bde