]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoThe previous clause 3 commit was also
Mark Murray [Fri, 9 Apr 2004 20:34:51 +0000 (20:34 +0000)]
The previous clause 3 commit was also

Approved by: scottl
Thanks to: scottl

20 years agoRemove advertising clause from University of California Regent's license,
Mark Murray [Fri, 9 Apr 2004 19:58:40 +0000 (19:58 +0000)]
Remove advertising clause from University of California Regent's license,
per letter dated July 22, 1999.

Approved by: core, imp

20 years agoAdd MODULE_DEPEND entries so some of these drivers can eventually be
Nate Lawson [Fri, 9 Apr 2004 18:14:32 +0000 (18:14 +0000)]
Add MODULE_DEPEND entries so some of these drivers can eventually be
loaded separately from ACPI (i.e., embedded use).

20 years agoBand-aid fix to extract MAC address from some CEM2/CEM28 cards with broken
Scott Mitchell [Fri, 9 Apr 2004 17:34:54 +0000 (17:34 +0000)]
Band-aid fix to extract MAC address from some CEM2/CEM28 cards with broken
CIS.  Really needs a better interface to the CIS in pccard driver.

Reviewed by: imp (mentor)

20 years agoBackout previous commit. It seems this comment applied to something
Maxime Henrion [Fri, 9 Apr 2004 17:30:26 +0000 (17:30 +0000)]
Backout previous commit.  It seems this comment applied to something
else than I thought, and thus really needs to go away.

Noticed by: ru

20 years agoFix probe routine to use card IDs from pccarddevs for NEWCARD and OLDCARD.
Scott Mitchell [Fri, 9 Apr 2004 17:27:36 +0000 (17:27 +0000)]
Fix probe routine to use card IDs from pccarddevs for NEWCARD and OLDCARD.
Should now correctly probe and attach all supported cards in either mode.

Reviewed by: imp (mentor)

20 years agoo Fix an incorrect parsing of 0.0.0.0/0 expression.
Maxim Konovalov [Fri, 9 Apr 2004 17:26:01 +0000 (17:26 +0000)]
o Fix an incorrect parsing of 0.0.0.0/0 expression.

PR: kern/64778
MFC after: 6 weeks

20 years agoAdd Xircom XEM5600 card (appears to be a renamed REM56).
Scott Mitchell [Fri, 9 Apr 2004 17:15:18 +0000 (17:15 +0000)]
Add Xircom XEM5600 card (appears to be a renamed REM56).
Observe that Xircom CEM28 and CEM33 are known to work in Ethernet mode.

Reviewed by: imp (mentor)

20 years agoSync to pccarddevs 1.83
Scott Mitchell [Fri, 9 Apr 2004 17:10:12 +0000 (17:10 +0000)]
Sync to pccarddevs 1.83

Reviewed by: imp (mentor)

20 years agoAdd Xircom XEM5600 and known versions of CE2, CEM33 and CEM56.
Scott Mitchell [Fri, 9 Apr 2004 17:08:12 +0000 (17:08 +0000)]
Add Xircom XEM5600 and known versions of CE2, CEM33 and CEM56.
Xircom had an unfortunate habit of re-using PCMCIA IDs for quite different
cards - the xe driver knows about this and uses the first byte of 'extra'
PCMCIA ID info to identify cards with ambiguous IDs.

Reviewed by: imp (mentor)

20 years agoI hate noticing bugs after committing. :-(
Mark Murray [Fri, 9 Apr 2004 17:00:03 +0000 (17:00 +0000)]
I hate noticing bugs after committing. :-(
ALWAYS set up the CPU base identity string. THEN optionally
add features.

20 years agoo SIOCGIFCONF->getifaddrs(2) conversion.
Maxim Konovalov [Fri, 9 Apr 2004 16:59:05 +0000 (16:59 +0000)]
o SIOCGIFCONF->getifaddrs(2) conversion.

PR: bin/9379
Obtained from: NetBSD

20 years agoDocument the recent upgrade to the entropy device WRT hardware
Mark Murray [Fri, 9 Apr 2004 16:02:54 +0000 (16:02 +0000)]
Document the recent upgrade to the entropy device WRT hardware
generators.

20 years agoTake into account hardware-supplied entropy. If the entropy source
Mark Murray [Fri, 9 Apr 2004 15:56:16 +0000 (15:56 +0000)]
Take into account hardware-supplied entropy. If the entropy source
is hardware, the Yarrow initialisations don't need to be done.

20 years agoReorganise the entropy device so that high-yield entropy sources
Mark Murray [Fri, 9 Apr 2004 15:47:10 +0000 (15:47 +0000)]
Reorganise the entropy device so that high-yield entropy sources
can more easily be used INSTEAD OF the hard-working Yarrow.
The only hardware source used at this point is the one inside
the VIA C3 Nehemiah (Stepping 3 and above) CPU. More sources will
be added in due course. Contributions welcome!

20 years agoOmnibus PCI commit:
Warner Losh [Fri, 9 Apr 2004 15:44:34 +0000 (15:44 +0000)]
Omnibus PCI commit:
o Save and restore bars for suspend/resume as well as for D3->D0
  transitions.
o preallocate resources that the PCI devices use to avoid resource
  conflicts
o lazy allocation of resources not allocated by the BIOS.
o set unattached drivers to state D3.  Set power state to D0
  before probe/attach.  Right now there's two special cases
  for this (display and memory devices) that need work in other
  areas of the tree.

Please report any bugs to me.

20 years agoPlug minor memory leak of module_t structures when unloading a file
Peter Edwards [Fri, 9 Apr 2004 15:27:38 +0000 (15:27 +0000)]
Plug minor memory leak of module_t structures when unloading a file
from the kernel.

Reviewed By: Doug Rabson (dfr@)

20 years agoAdd extra output to show when VIA C3 Nehemiah CPUs have hardware
Mark Murray [Fri, 9 Apr 2004 15:01:44 +0000 (15:01 +0000)]
Add extra output to show when VIA C3 Nehemiah CPUs have hardware
Random Number Generator (RNG) and/or Advanced Cryptography Engine
(ACE).

20 years agoSpell "switches" a more conventional way.
Olivier Houchard [Fri, 9 Apr 2004 14:31:29 +0000 (14:31 +0000)]
Spell "switches" a more conventional way.

20 years ago- Change several errx() calls that should have been err() calls.
Maxime Henrion [Fri, 9 Apr 2004 14:27:28 +0000 (14:27 +0000)]
- Change several errx() calls that should have been err() calls.
- Handle empty ARP tables properly.
- Remove register keyword.
- arp(8) is WARNS?=4 clean, so mark it as such to avoid regressions.

20 years agoReintroduce a comment that was wrongly removed in revision 1.34.
Maxime Henrion [Fri, 9 Apr 2004 14:19:12 +0000 (14:19 +0000)]
Reintroduce a comment that was wrongly removed in revision 1.34.

20 years agoCompare pointers with NULL rather than using pointers are booleans in
Robert Watson [Fri, 9 Apr 2004 13:23:51 +0000 (13:23 +0000)]
Compare pointers with NULL rather than using pointers are booleans in
if/for statements.  Assign pointers to NULL rather than typecast 0.
Compare pointers with NULL rather than 0.

20 years agoFixed missing forward declaration of struct rusage in the _KERNEL case.
Bruce Evans [Fri, 9 Apr 2004 13:14:32 +0000 (13:14 +0000)]
Fixed missing forward declaration of struct rusage in the _KERNEL case.

Removed namespace-polluting forward declaration of struct rusage in the
!_KERNEL && !_BSD_VISIBLE case.

Fixed some nearby style bugs.

20 years agoTerminate execl()'s argument list with a null pointer instead of a
Tim J. Robbins [Fri, 9 Apr 2004 11:32:32 +0000 (11:32 +0000)]
Terminate execl()'s argument list with a null pointer instead of a
null pointer constant. (The latter may be an integer constant, which
is not correct here.)

Submitted by: Stefan Farfeleder

20 years agoImprove robustness of multibyte character handling (-m option), and
Tim J. Robbins [Fri, 9 Apr 2004 11:17:29 +0000 (11:17 +0000)]
Improve robustness of multibyte character handling (-m option), and
simplify the read buffering now that we can feed partial multibyte
characters to mbrtowc().

20 years agoStyle fixes and add gen to CLEANFILES.
Nate Lawson [Fri, 9 Apr 2004 07:13:27 +0000 (07:13 +0000)]
Style fixes and add gen to CLEANFILES.

Submitted by: bde

20 years agoReplace more ad-hoc versions of acpi_GetReference(). Since the type of
Nate Lawson [Fri, 9 Apr 2004 06:55:50 +0000 (06:55 +0000)]
Replace more ad-hoc versions of acpi_GetReference().  Since the type of
Reference objects changed from ACPI_TYPE_ANY to ACPI_TYPE_LOCAL_REFERENCE
in Oct. 2002, this may help systems where switching the cooler on failed.
We support both types for now until this sorts out.

20 years agoInclude the prototype for acpi_GetReference.
Nate Lawson [Fri, 9 Apr 2004 06:53:50 +0000 (06:53 +0000)]
Include the prototype for acpi_GetReference.

20 years agoAdd support for packages as the first element of _PRW. This may allow
Nate Lawson [Fri, 9 Apr 2004 06:40:03 +0000 (06:40 +0000)]
Add support for packages as the first element of _PRW.  This may allow
some machines to enable wake events for more devices although I haven't
seen a system yet that uses this form.  Also, introduce acpi_GetReference()
which retrieves an object reference from various types.

20 years agoOoops, removed this acknowledgement bogusly.
Warner Losh [Fri, 9 Apr 2004 05:12:47 +0000 (05:12 +0000)]
Ooops, removed this acknowledgement bogusly.

Eagle Eyes: bde

20 years agoForced commit to add note to previous commit message.
Robert Watson [Fri, 9 Apr 2004 03:17:21 +0000 (03:17 +0000)]
Forced commit to add note to previous commit message.

Also tested by: "Michael L. Squires" <mikes@siralan.org>

20 years agoLock down the netatalk AARP code, which is responsible for appletalk
Robert Watson [Fri, 9 Apr 2004 01:40:12 +0000 (01:40 +0000)]
Lock down the netatalk AARP code, which is responsible for appletalk
address discovery and caching (similar to inet ARP).  Use a single
global mutex, aarptab_mtx, to protect the table.  Remove spl/spx.

Tested by: Bob Bishop <rb@gid.co.uk>

20 years agoAfter forking and initializing the library to single-threaded
Daniel Eischen [Thu, 8 Apr 2004 23:16:21 +0000 (23:16 +0000)]
After forking and initializing the library to single-threaded
mode (where the forked thread is the one and only thread and
is marked as system scope), set the system scope flag before
initializing the signal mask.  This prevents trying to use
internal locks that haven't yet been initialized.

Reported by: Dan Nelson <dnelson at allantgroup.com>
Reviewed by: davidxu

20 years agoAdd a LIBRARY section.
Simon L. B. Nielsen [Thu, 8 Apr 2004 23:04:54 +0000 (23:04 +0000)]
Add a LIBRARY section.

Submitted by: Craig Rodrigues <rodrigc@crodrigues.org>
PR: docs/65208

20 years agoAllow arguments to appear in any order.
Tim Kientzle [Thu, 8 Apr 2004 19:37:01 +0000 (19:37 +0000)]
Allow arguments to appear in any order.

While I'm here, add a couple of extra sanity-checks to
the argument parsing (reject -j -z, for instance) and
update the docs a bit.

Requested by: most everyone ;-)

20 years agoThe demise of vm_pager_map_page() in revision 1.93 of vm/vm_pager.c permits
Alan Cox [Thu, 8 Apr 2004 19:08:49 +0000 (19:08 +0000)]
The demise of vm_pager_map_page() in revision 1.93 of vm/vm_pager.c permits
the reduction of the pager map's size by 8M bytes.  In other words, eight
megabytes of largely wasted KVA are returned to the kernel map for use
elsewhere.

20 years agoReadd -Werror flag to compilation pf.c after the compiler warnings have been
Max Laier [Thu, 8 Apr 2004 18:17:13 +0000 (18:17 +0000)]
Readd -Werror flag to compilation pf.c after the compiler warnings have been
shut up.

Tested by: LINT(i386)
Approved by: bms(mentor)

20 years agoUnify on version 1 to be similar to the rest of the tree. After 5-stable
Nate Lawson [Thu, 8 Apr 2004 16:45:12 +0000 (16:45 +0000)]
Unify on version 1 to be similar to the rest of the tree.  After 5-stable
branches, increment version on any API change visible to other modules.

20 years agoIn the examples, make the bitmap_name an absolute path name.
Jens Schweikhardt [Thu, 8 Apr 2004 16:31:17 +0000 (16:31 +0000)]
In the examples, make the bitmap_name an absolute path name.
This avoids hair pulling when someone expects the splash screen
to just work when the bitmap file is in /boot.

20 years agoAdd Netgear WG311T (PCI version of the WG511T).
Stefan Eßer [Thu, 8 Apr 2004 16:09:40 +0000 (16:09 +0000)]
Add Netgear WG311T (PCI version of the WG511T).
I just received one and have verified that it works just fine.

20 years agoFix typo.
Bruce A. Mah [Thu, 8 Apr 2004 15:47:30 +0000 (15:47 +0000)]
Fix typo.

Submitted by: schweikh

20 years agoUpdate manual pages for change to C99 mbrtowc() semantics.
Tim J. Robbins [Thu, 8 Apr 2004 09:59:02 +0000 (09:59 +0000)]
Update manual pages for change to C99 mbrtowc() semantics.

20 years agoFix a regression in my change which sends headers along with data; a
Mike Silbersack [Thu, 8 Apr 2004 07:14:34 +0000 (07:14 +0000)]
Fix a regression in my change which sends headers along with data; a
side effect of that change caused headers to not be sent if a 0 byte
file was passed to sendfile.  This change fixes that behavior, allowing
sendfile to send out the headers even with a 0 byte file again.

Noticed by: Dirk Engling

20 years agoDo not assume that the initial thread (i.e. the thread with the ID
Marcel Moolenaar [Thu, 8 Apr 2004 06:37:00 +0000 (06:37 +0000)]
Do not assume that the initial thread (i.e. the thread with the ID
equal to the process ID) is still present when we dump a core. It
already may have been destroyed. In that case we would end up
dereferencing a NULL pointer, so specifically test for that as well.

Reported & tested by: Dan Nelson <dnelson@allantgroup.com>

20 years agoDocument the newly-implemented -T option.
Tim Kientzle [Thu, 8 Apr 2004 06:21:03 +0000 (06:21 +0000)]
Document the newly-implemented -T option.

20 years agoAdd support for -T option on create.
Tim Kientzle [Thu, 8 Apr 2004 06:12:01 +0000 (06:12 +0000)]
Add support for -T option on create.
(Required by pkg_create.)

20 years agoWhen invoking tar, make sure the mode option (-x) is first.
Tim Kientzle [Thu, 8 Apr 2004 06:06:42 +0000 (06:06 +0000)]
When invoking tar, make sure the mode option (-x) is first.

20 years agoAdd whitespace before comment blocks. (reported by njl)
Colin Percival [Thu, 8 Apr 2004 02:03:49 +0000 (02:03 +0000)]
Add whitespace before comment blocks. (reported by njl)
Remove spurious whitespace, add indent protection, fix punctuation,
remove initialization of static variables to zero, put wakeup_ctr
and wakeup_needed in the correct order. (reported by bde)

This doesn't fix all the style bugs I introduced, but the remaining
style bugs make it easier for me to understand what I did here.

20 years agoAdding Vinod Kashyap (vkashyap) as the maintainer for twa.
Vinod Kashyap [Thu, 8 Apr 2004 01:43:34 +0000 (01:43 +0000)]
Adding Vinod Kashyap (vkashyap) as the maintainer for twa.

20 years agoThis manual page will not first appear in 4.10 as RELENG_4 has a different
Tom Rhodes [Wed, 7 Apr 2004 22:16:04 +0000 (22:16 +0000)]
This manual page will not first appear in 4.10 as RELENG_4 has a different
version of the adduser utility.

Noticed by: simon

20 years agoBack out last bad commit (again!)
Warner Losh [Wed, 7 Apr 2004 21:56:20 +0000 (21:56 +0000)]
Back out last bad commit (again!)

20 years agoRemove advertising clause from University of California Regent's
Warner Losh [Wed, 7 Apr 2004 20:46:16 +0000 (20:46 +0000)]
Remove advertising clause from University of California Regent's
license, per letter dated July 22, 1999 and email from Peter Wemm,
Alan Cox and Robert Watson.

Approved by: core, peter, alc, rwatson

20 years agoImplement an ACPI-aware pci_set_powerstate() method for PCI busses that
John Baldwin [Wed, 7 Apr 2004 19:42:21 +0000 (19:42 +0000)]
Implement an ACPI-aware pci_set_powerstate() method for PCI busses that
are enumerated in the ACPI device tree.  In addition to the normal PCI
powerstate functionality, the ACPI _PSx methods are executed and ACPI
PowerResources are switched on and off via the acpi_pwr_switch_consumer()
function.

Glanced at by: imp, njl

20 years agoDon't turn off the regular SIGINFO status information. The use of
Ian Dowse [Wed, 7 Apr 2004 18:48:11 +0000 (18:48 +0000)]
Don't turn off the regular SIGINFO status information. The use of
the NOKERNINFO flag only marginally de-clutters the output and
has a number of unwanted side effects:
  o The kernel info might be what you want to see
  o ^T is left non-functional if ping is killed non-cleanly
  o "ping -q foo &" gets suspended on tty output

Encouraged by: bde

20 years agoLinux still uses 32-bit off_t by default. Ask for 64-bit off_t on
Tim Kientzle [Wed, 7 Apr 2004 17:47:06 +0000 (17:47 +0000)]
Linux still uses 32-bit off_t by default.  Ask for 64-bit off_t on
Linux.

20 years agoCorrect signature in alternative stub implementation of setup_acls()
Tim Kientzle [Wed, 7 Apr 2004 17:43:48 +0000 (17:43 +0000)]
Correct signature in alternative stub implementation of setup_acls()

20 years agoIn ndis_convert_res(), initialize the head of our temporary list
Bill Paul [Wed, 7 Apr 2004 17:02:55 +0000 (17:02 +0000)]
In ndis_convert_res(), initialize the head of our temporary list
before calling BUS_GET_RESOURCE_LIST(). Previously, the list head would
only be initialized if BUS_GET_RESOURCE_LIST() succeeded; it needs to
be initialized unconditionally so that the list cleanup code won't
trip over potential stack garbage.

20 years agoAdd new ID for Intel 82562ET (ICH5/ICH5R) Pro/100 VE Ethernet.
Warner Losh [Wed, 7 Apr 2004 15:47:14 +0000 (15:47 +0000)]
Add new ID for Intel 82562ET (ICH5/ICH5R) Pro/100 VE Ethernet.

Submitted by: Stefan Bethke
PR: 61320

20 years agoAdd -acl and a small blurb about it to the options list.
Tom Rhodes [Wed, 7 Apr 2004 14:48:30 +0000 (14:48 +0000)]
Add -acl and a small blurb about it to the options list.
Update date.

20 years agoRecord where half the bits in this file came from (from ufs_readwrite.c).
Bruce Evans [Wed, 7 Apr 2004 11:21:18 +0000 (11:21 +0000)]
Record where half the bits in this file came from (from ufs_readwrite.c).
Damage to history from moving bits was especially large since a repo copy
is not feasible for partial files.

20 years agoReset internal conversion state after triggering conversion errors;
Tim J. Robbins [Wed, 7 Apr 2004 11:02:51 +0000 (11:02 +0000)]
Reset internal conversion state after triggering conversion errors;
the state is undefined according to the standards, and our undefined
behaviour has changed.

20 years agoAllow partial multibyte characters to accumulate in conversion state
Tim J. Robbins [Wed, 7 Apr 2004 10:48:19 +0000 (10:48 +0000)]
Allow partial multibyte characters to accumulate in conversion state
objects passed to mbrtowc(), mbsrtowcs(), and mbrlen(), as required
by C99.

20 years agoFixed a bug in previous revision: compute the payload checksum before
Ruslan Ermilov [Wed, 7 Apr 2004 10:01:39 +0000 (10:01 +0000)]
Fixed a bug in previous revision: compute the payload checksum before
we convert ip_len into a network byte order; in_delayed_cksum() still
expects it in host byte order.

The symtom was the ``in_cksum_skip: out of data by %d'' complaints
from the kernel.

To add to the previous commit log.  These fixes make tcpdump(1) happy
by not complaining about UDP/TCP checksum being bad for looped back
IP multicast when multicast router is deactivated.

Reported by: Vsevolod Lobko

20 years agoPrepare to handle trivial state-dependent encodings. Full support for
Tim J. Robbins [Wed, 7 Apr 2004 09:55:05 +0000 (09:55 +0000)]
Prepare to handle trivial state-dependent encodings. Full support for
state-dependent encodings with locking shifts will come later if there
is demand for it.

20 years agoBegin conversions for sgetrune() and sputrune() in the initial
Tim J. Robbins [Wed, 7 Apr 2004 09:49:10 +0000 (09:49 +0000)]
Begin conversions for sgetrune() and sputrune() in the initial
conversion state.

20 years agoPrepare to handle state-dependent encodings. This mainly involves not
Tim J. Robbins [Wed, 7 Apr 2004 09:47:56 +0000 (09:47 +0000)]
Prepare to handle state-dependent encodings. This mainly involves not
taking shortcuts when it comes to storing and passing around conversion
states.

20 years agoBegin in the initial shift state in mbstowcs() and wcstombs().
Tim J. Robbins [Wed, 7 Apr 2004 08:33:23 +0000 (08:33 +0000)]
Begin in the initial shift state in mbstowcs() and wcstombs().
(This change is non-functional since nothing uses states yet.)

20 years agoFix filt_timer* races: Finish initializing a knote before we pass it to
Colin Percival [Wed, 7 Apr 2004 05:59:57 +0000 (05:59 +0000)]
Fix filt_timer* races: Finish initializing a knote before we pass it to
a callout, and use the new callout_drain API to make sure that a callout
has finished before we deallocate memory it is using.

PR: kern/64121
Discussed with: gallatin

20 years agoThe bs driver was replaced with the ct(pc98) driver. takahashi-san
Warner Losh [Wed, 7 Apr 2004 05:44:51 +0000 (05:44 +0000)]
The bs driver was replaced with the ct(pc98) driver.  takahashi-san
(nyan) says this driver is now obsolete and can be removed.

20 years agoLast change was a bogus
Warner Losh [Wed, 7 Apr 2004 05:30:54 +0000 (05:30 +0000)]
Last change was a bogus

20 years agoFix path to rc.sendmail
Gregory Neil Shapiro [Wed, 7 Apr 2004 05:29:03 +0000 (05:29 +0000)]
Fix path to rc.sendmail

20 years agoRemove advertising clause from University of California Regent's
Warner Losh [Wed, 7 Apr 2004 05:00:01 +0000 (05:00 +0000)]
Remove advertising clause from University of California Regent's
license, per letter dated July 22, 1999 and email from Peter Wemm,
Alan Cox and Robert Watson.

Approved by: core, peter, alc, rwatson

20 years agoRemove advertising clause from University of California Regent's license,
Warner Losh [Wed, 7 Apr 2004 04:19:52 +0000 (04:19 +0000)]
Remove advertising clause from University of California Regent's license,
per letter dated July 22, 1999.

Approved by: core

20 years agoRemove advertising clause from University of California Regent's
Warner Losh [Wed, 7 Apr 2004 03:47:21 +0000 (03:47 +0000)]
Remove advertising clause from University of California Regent's
license, per letter dated July 22, 1999 and irc message from Robert
Watson saying that clause 3 can be removed from those files with an
NAI copyright that also have only a University of California
copyrights.

Approved by: core, rwatson

20 years agoUpdate to include both the L1 and L2 TLB stats, as well as the seperate
Peter Wemm [Wed, 7 Apr 2004 00:44:15 +0000 (00:44 +0000)]
Update to include both the L1 and L2 TLB stats, as well as the seperate
2M/4M page TLB vs 4K page TLB stats.  This also applies to the i386
platform, as does the cpu features fixes.

20 years agoMFi386: move rss() from db_interface.c to cpufunc.h
Peter Wemm [Wed, 7 Apr 2004 00:41:05 +0000 (00:41 +0000)]
MFi386: move rss() from db_interface.c to cpufunc.h

20 years agoUnbreak FAST_IPSEC build on 64 bit archs with INVARIANTS.
Pawel Jakub Dawidek [Wed, 7 Apr 2004 00:19:02 +0000 (00:19 +0000)]
Unbreak FAST_IPSEC build on 64 bit archs with INVARIANTS.

Approved by: sam

20 years agoFix some issues with ACL handling:
Tim Kientzle [Tue, 6 Apr 2004 23:16:50 +0000 (23:16 +0000)]
Fix some issues with ACL handling:
  * ACL storage is no longer erased before a group of entries are added.
  * ACL text creation no longer tries to skip over non-existent text.
  * UTF8 encoder no longer blows up on invalid wide characters.
  * Fixed ACL state management for default ACLs.
Also, publicize function for obtaining text-format ACL in various
formats.  The interface is now extensible through a "flags" argument
that allows you to select a variant format.

20 years agoStyle fixes, as suggested by Jens Schweikhardt <schweikh@schweikhardt.net>
Wes Peters [Tue, 6 Apr 2004 23:15:48 +0000 (23:15 +0000)]
Style fixes, as suggested by Jens Schweikhardt <schweikh@schweikhardt.net>

20 years agoIntroduce a callout_drain() function. This acts in the same manner as
Colin Percival [Tue, 6 Apr 2004 23:08:49 +0000 (23:08 +0000)]
Introduce a callout_drain() function.  This acts in the same manner as
callout_stop(), except that if the callout being stopped is currently
in progress, it blocks attempts to reset the callout and waits until the
callout is completed before it returns.

This makes it possible to clean up callout-using code safely, e.g.,
without potentially freeing memory which is still being used by a callout.

Reviewed by: mux, gallatin, rwatson, jhb

20 years agoArchive both access and default ACLs.
Tim Kientzle [Tue, 6 Apr 2004 23:08:31 +0000 (23:08 +0000)]
Archive both access and default ACLs.

20 years agoBetter checks to make sure that we get good alignment. This code is a
Warner Losh [Tue, 6 Apr 2004 22:50:50 +0000 (22:50 +0000)]
Better checks to make sure that we get good alignment.  This code is a
bit of a bandaide until I get better pci bus code committed to head
from my p4 tree.

20 years agoFix mis-merge from p4 by adding line getting sc.
Warner Losh [Tue, 6 Apr 2004 22:41:14 +0000 (22:41 +0000)]
Fix mis-merge from p4 by adding line getting sc.
Attempt to deal with larger memory allocation better.

20 years agoGive date(1) a much better SYNOPSIS section which is useful as more than
Brian Feldman [Tue, 6 Apr 2004 22:03:38 +0000 (22:03 +0000)]
Give date(1) a much better SYNOPSIS section which is useful as more than
decoration.  Further improvements are welcome, but at least this
is a separate of the various modes of operation date has, as well as
sectioning off the two deprecated options for settimeofday(tz) that
don't even apply to actual operation of date as such, anyway.

20 years agoFix some syntax errors in examples. These were discovered when trying
Mark Murray [Tue, 6 Apr 2004 20:16:10 +0000 (20:16 +0000)]
Fix some syntax errors in examples. These were discovered when trying
to follow the examples concerned.

20 years agoRemove advertising clause from University of California Regent's license,
Warner Losh [Tue, 6 Apr 2004 20:15:37 +0000 (20:15 +0000)]
Remove advertising clause from University of California Regent's license,
per letter dated July 22, 1999.

Approved by: core

20 years agoMFP4: Power up with OE disabled. Similar patches went into NetBSD a
Warner Losh [Tue, 6 Apr 2004 20:13:29 +0000 (20:13 +0000)]
MFP4: Power up with OE disabled.  Similar patches went into NetBSD a
while ago, and it does seem to help at least one card I have and has
been in my p4 tree for many months.

20 years agoRemove clause 3 from the UCB licenses.
Mark Murray [Tue, 6 Apr 2004 20:06:54 +0000 (20:06 +0000)]
Remove clause 3 from the UCB licenses.

OK'ed by: imp, core

20 years ago- Overdue bump of the document date.
Ruslan Ermilov [Tue, 6 Apr 2004 19:49:21 +0000 (19:49 +0000)]
- Overdue bump of the document date.
- Fixed markup.
- Minor wordsmithing.

20 years agoUse the correct flag for mbuf allocations (M_DONTWAIT, not M_NOWAIT).
Ian Dowse [Tue, 6 Apr 2004 19:32:00 +0000 (19:32 +0000)]
Use the correct flag for mbuf allocations (M_DONTWAIT, not M_NOWAIT).

20 years agoFix a paste-o from the buf_prewrite() cleanup commit and check for the
John Baldwin [Tue, 6 Apr 2004 19:20:24 +0000 (19:20 +0000)]
Fix a paste-o from the buf_prewrite() cleanup commit and check for the
MNTK_SUSPEND flag on the correct vnode pointer in softdep_disk_prewrite().

Reviewed by: phk
Tested by: kensmith

20 years agoAssociate a simple count of waiters with each condition variable. The
John Baldwin [Tue, 6 Apr 2004 19:17:46 +0000 (19:17 +0000)]
Associate a simple count of waiters with each condition variable.  The
count is protected by the mutex that protects the condition, so the count
does not require any extra locking or atomic operations.  It serves as an
optimization to avoid calling into the sleepqueue code at all if there are
no waiters.

Note that the count can get temporarily out of sync when threads sleeping
on a condition variable time out or are aborted.  However, it doesn't hurt
to call the sleepqueue code for either a signal or a broadcast when there
are no waiters, and the count is never out of sync in the opposite
direction unless we have more than INT_MAX sleeping threads.

20 years agoAdd a new kernel option MUTEX_WAKE_ALL that changes the mutex unlock code
John Baldwin [Tue, 6 Apr 2004 19:12:24 +0000 (19:12 +0000)]
Add a new kernel option MUTEX_WAKE_ALL that changes the mutex unlock code
to awaken all waiters when a contested mutex is released instead of just
the highest priority waiter.  If the various threads are awakened in
sequence then each thread may acquire and release the lock in question
without contention resulting in fewer expensive unlock and lock
operations.  This old behavior of waking just the highest priority is
still used if this option is specified.  Making the algorithm conditional
on a kernel option will allows us to benchmark both cases later and
determine which one should be used by default.

Requested by: tanimura-san

20 years agoSort function prototypes.
John Baldwin [Tue, 6 Apr 2004 19:08:40 +0000 (19:08 +0000)]
Sort function prototypes.

20 years agoRename turnstile_wakeup() to turnstile_broadcast() to make the naming
John Baldwin [Tue, 6 Apr 2004 19:07:21 +0000 (19:07 +0000)]
Rename turnstile_wakeup() to turnstile_broadcast() to make the naming
more consistent with other APIs. sleepq and cv's use signal/broadcast, and
msleep uses wakeup_one/wakeup.  Prior to this turnstiles were using a
signal/wakeup mixture.

20 years agoEnable the memory arbiter before turning off the PXE restart. This
Paul Saab [Tue, 6 Apr 2004 18:28:15 +0000 (18:28 +0000)]
Enable the memory arbiter before turning off the PXE restart.  This
prevents NMI's from happening when resetting the chip on some
hardware I have seen.

Mis-behaving box made available by: John Cagle <john.cagle@hp.com>

20 years agoSplit ramdisk processing in two so it can actually work. Ownerships
Wes Peters [Tue, 6 Apr 2004 18:26:43 +0000 (18:26 +0000)]
Split ramdisk processing in two so it can actually work.  Ownerships
and permissions specified per rc.conf(5) now apply both to the md
device and to the mountpoint directory, after the mount has completed.
This has to be done in two steps, because chown is not available
until after /usr has been mounted, but the mdconfig and newfs steps
have to complete before fstab processing.

20 years agoConnect bsdtar to the build.
Tim Kientzle [Tue, 6 Apr 2004 17:04:38 +0000 (17:04 +0000)]
Connect bsdtar to the build.

20 years agoMake pf* modules respect NOINET6 from make.conf(5) in order to build them
Max Laier [Tue, 6 Apr 2004 15:12:50 +0000 (15:12 +0000)]
Make pf* modules respect NOINET6 from make.conf(5) in order to build them
for INET6-less kernel.

Requested by: many
Approved by: bms(mentor)

20 years agoMFi386: Enable the cy driver.
Yoshihiro Takahashi [Tue, 6 Apr 2004 14:19:45 +0000 (14:19 +0000)]
MFi386: Enable the cy driver.