]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoWe will not be responsible for damage to equipment, your ego, county wide
Wilko Bulte [Sun, 25 Dec 2005 23:29:59 +0000 (23:29 +0000)]
We will not be responsible for damage to equipment, your ego, county wide
power outages, spontaneously generated mini (or larger) black holes,
planetary disruptions, or personal injury or worse that may result from the
use of this material.
                -- taken from Samuel M. Goldwasser's
                   Sam's Strobe FAQ Notes on the Troubleshooting
                   and Repair of Electronic Flash Units and Strobe Lights

18 years agoadd a sysctl to turn debug msgs on/off when built with IFMEDIA_DEBUG
Sam Leffler [Sun, 25 Dec 2005 23:28:23 +0000 (23:28 +0000)]
add a sysctl to turn debug msgs on/off when built with IFMEDIA_DEBUG

18 years agoFix build after timeval.tv_sec changed from long to time_t.
Max Laier [Sun, 25 Dec 2005 22:57:08 +0000 (22:57 +0000)]
Fix build after timeval.tv_sec changed from long to time_t.

18 years agoAdd an example for WEP.
Wilko Bulte [Sun, 25 Dec 2005 22:09:17 +0000 (22:09 +0000)]
Add an example for WEP.

OK-ed by: sam

18 years agoKill the BUGS section, and remove a similar line noted under -T. As far
Tom Rhodes [Sun, 25 Dec 2005 10:36:30 +0000 (10:36 +0000)]
Kill the BUGS section, and remove a similar line noted under -T.  As far
as I know, Sun's NFS support works with TCP just fine.  This is even
hinted at in the PR.

PR: 71782

18 years agoWhen generating profiles (-g) warn about shared objects that were
Joseph Koshy [Sun, 25 Dec 2005 05:11:29 +0000 (05:11 +0000)]
When generating profiles (-g) warn about shared objects that were
not found.

18 years agoSuspend and resume support has been added.
Ariff Abdullah [Sun, 25 Dec 2005 00:47:25 +0000 (00:47 +0000)]
Suspend and resume support has been added.

18 years agoAdd suspend and resume support.
Ariff Abdullah [Sun, 25 Dec 2005 00:43:03 +0000 (00:43 +0000)]
Add suspend and resume support.

18 years agoAdd a64l(), l64a(), and l64a_r() XSI extentions. These functions convert
Tom Rhodes [Sat, 24 Dec 2005 22:37:59 +0000 (22:37 +0000)]
Add a64l(), l64a(), and l64a_r() XSI extentions.  These functions convert
between a 32-bit integer and a radix-64 ASCII string.  The l64a_r() function
is a NetBSD addition.

PR: 51209 (based on submission, but very different)
Reviewed by: bde, ru

18 years agoMake tv_sec a time_t on all platforms but alpha. Brings us more in line with
Tom Rhodes [Sat, 24 Dec 2005 22:22:17 +0000 (22:22 +0000)]
Make tv_sec a time_t on all platforms but alpha.  Brings us more in line with
POSIX.  This also makes the struct correct we ever implement an i386-time64
architecture.  Not that we need too.

Reviewed by: imp, brooks
Approved by: njl (acpica), des (no objects, touches procfs)
Tested with: make universe

18 years ago"A life lived in fear is a life half lived"
Wilko Bulte [Sat, 24 Dec 2005 17:55:33 +0000 (17:55 +0000)]
"A life lived in fear is a life half lived"

18 years ago- Don't print an errno string when reporting dropped samples.
Joseph Koshy [Sat, 24 Dec 2005 17:00:33 +0000 (17:00 +0000)]
- Don't print an errno string when reporting dropped samples.
- Prevent an unsuccessful exec() of a target program from triggering
  a false 'dropped samples' message.

18 years agoMake lookups for relocations from old unversioned binaries return
Alexander Kabaev [Sat, 24 Dec 2005 15:37:30 +0000 (15:37 +0000)]
Make lookups for relocations from old unversioned binaries return
oldest versioned symbol available. Do not accept hidden symbols for
all other versions.

Use "<obj->path>: <error message>" for all error messages in new
functions to make them more consistent.

18 years agoForward-declare struct trapframe to allow the aic module to compile.
Peter Grehan [Sat, 24 Dec 2005 08:27:45 +0000 (08:27 +0000)]
Forward-declare struct trapframe to allow the aic module to compile.

18 years agoAdd missing mount.h include to VFS_LOCK_GIANT() man page.
Robert Watson [Sat, 24 Dec 2005 08:25:15 +0000 (08:25 +0000)]
Add missing mount.h include to VFS_LOCK_GIANT() man page.

Submitted by: Wojciech A. Koszek <dunstan at freebsd dot czest dot pl>
MFC after: 3 days

18 years agoo NetBSD 3.0 added.
Maxim Konovalov [Sat, 24 Dec 2005 07:30:27 +0000 (07:30 +0000)]
o NetBSD 3.0 added.

18 years agoMaintain the lock on the vnode for most of exec_elfN_imgact().
Alan Cox [Sat, 24 Dec 2005 04:57:50 +0000 (04:57 +0000)]
Maintain the lock on the vnode for most of exec_elfN_imgact().
Specifically, it is required for the I/O that may be performed by
elfN_load_section().

Avoid an obscure deadlock in the a.out, elf, and gzip image
activators.  Add a comment describing why the deadlock does not occur
in the common case and how it might occur in less usual circumstances.

Eliminate an unused variable from exec_aout_imgact().

In collaboration with: tegge

18 years agoAvoid kernel panic when attaching a process which may not be stopped
David Xu [Sat, 24 Dec 2005 02:59:29 +0000 (02:59 +0000)]
Avoid kernel panic when attaching a process which may not be stopped
by debugger, e.g process is dumping core. Only access p_xthread if
P_STOPPED_TRACE is set, this means thread is ready to exchange signal
with debugger, print a warning if P_STOPPED_TRACE is not set due to
some bugs in other code, if there is.

The patch has been tested by Anish Mistry mistry.7 at osu dot edu, and
is slightly adjusted.

18 years agoBuild ed on amd64. The pci attachment works with qemu on amd64.
Warner Losh [Fri, 23 Dec 2005 21:54:56 +0000 (21:54 +0000)]
Build ed on amd64.  The pci attachment works with qemu on amd64.

I'm holding off on building on sparc64 and others because I don't know
if this driver has had all the inb/outb removed (I think it has).  Nor
do I know if there are byte ordering issues.  There are very few word
operations on an NE2000, but I've not had time to audit them all.

Suggested by: Daniel O'Connor

18 years ago - Improve the INKERNEL macro such that it can no longer give false positives.
Jeff Roberson [Fri, 23 Dec 2005 21:33:55 +0000 (21:33 +0000)]
 - Improve the INKERNEL macro such that it can no longer give false positives.
   This fixes the stack(9) functionality.

Submitted by: Antoine Brodin <antoine.brodin@laposte.net>

18 years ago - Remove and unused include.
Jeff Roberson [Fri, 23 Dec 2005 21:32:40 +0000 (21:32 +0000)]
 - Remove and unused include.

Submitted by: Antoine Brodin <antoine.brodin@laposte.net>

18 years agoRegen.
Ruslan Ermilov [Fri, 23 Dec 2005 20:06:50 +0000 (20:06 +0000)]
Regen.

18 years agoFix build.
Ruslan Ermilov [Fri, 23 Dec 2005 20:06:14 +0000 (20:06 +0000)]
Fix build.

18 years agoImplement an upper limit for packets per second sent by node.
Gleb Smirnoff [Fri, 23 Dec 2005 19:14:38 +0000 (19:14 +0000)]
Implement an upper limit for packets per second sent by node.

18 years agoTypo.
Xin LI [Fri, 23 Dec 2005 15:50:57 +0000 (15:50 +0000)]
Typo.

18 years agoBump .Dd before I get told off.
Ceri Davies [Fri, 23 Dec 2005 15:33:31 +0000 (15:33 +0000)]
Bump .Dd before I get told off.

18 years agoCommands like gmirror, graid3, ... and others which use dlopen() to load
Ceri Davies [Fri, 23 Dec 2005 15:31:37 +0000 (15:31 +0000)]
Commands like gmirror, graid3, ... and others which use dlopen() to load
classes from say, /lib/geom, cannot be statically linked completely.
Moreover, those shared objects may require other shared objects (i.e.
 for geom, libraries like -lmd, -lcrypto).

The libs_so extension to crunchgen fixes this by allowing some libraries
to be linked in dynamically.  This requires that a copy of rtld and the
shared libraries be made available to the crunched binary, and so is not
suitable for all environments.  Crunchgen configurations which do not
use the 'libs_so' keyword are unaffected and produce identical binaries
with and without this commit.

Approved by: murray (mentor, in spirit), jhb
In collaboration with: Adrian Steinmann <ast at marabu dot ch>
MFC After: 6 weeks

18 years agoRemove debugging statement that slipped into lone of the previous commits
Alexander Kabaev [Fri, 23 Dec 2005 15:30:53 +0000 (15:30 +0000)]
Remove debugging statement that slipped into lone of the previous commits
unintentionally.

18 years agoMark the return address of the call to ast() in the generic trap
Peter Grehan [Fri, 23 Dec 2005 13:05:27 +0000 (13:05 +0000)]
Mark the return address of the call to ast() in the generic trap
handling code so the stack trace unwinders don't start trying to
go into user-space.

Found by trying to create core dumps with a KTR_COMPILE/KTR_GEOM
kernel, which results in a stack_save() call in the ast() coredump
path - this created a panic, and then calling 'trace' in ddb resulted
in the black screen of death after printing out most of the backtrace.

18 years agoAdd abort2() prototype
Poul-Henning Kamp [Fri, 23 Dec 2005 12:28:10 +0000 (12:28 +0000)]
Add abort2() prototype

18 years agoAdd abort2 manual page.
Poul-Henning Kamp [Fri, 23 Dec 2005 12:27:42 +0000 (12:27 +0000)]
Add abort2 manual page.

Submitted by: "Wojciech A. Koszek" <dunstan@freebsd.czest.pl>
Edited by: phk

18 years agoFix build error.
Yoshihiro Takahashi [Fri, 23 Dec 2005 12:14:55 +0000 (12:14 +0000)]
Fix build error.

18 years agoRegenerate sysent with new abort2 system call.
Poul-Henning Kamp [Fri, 23 Dec 2005 11:58:42 +0000 (11:58 +0000)]
Regenerate sysent with new abort2 system call.

Implement abort2(const char *reason, int narg, void **args);

Submitted by: "Wojciech A. Koszek" <dunstan@freebsd.czest.pl>

18 years agoAdd missing 455-462 syscalls as unimplemented
Poul-Henning Kamp [Fri, 23 Dec 2005 11:56:39 +0000 (11:56 +0000)]
Add missing 455-462 syscalls as unimplemented

18 years agoAdd abort2() systemcall.
Poul-Henning Kamp [Fri, 23 Dec 2005 11:54:11 +0000 (11:54 +0000)]
Add abort2() systemcall.

18 years agoMake sbuf_copyin() return the number of bytes copied on success.
Poul-Henning Kamp [Fri, 23 Dec 2005 11:49:53 +0000 (11:49 +0000)]
Make sbuf_copyin() return the number of bytes copied on success.

Submitted by: "Wojciech A. Koszek" <dunstan@freebsd.czest.pl>

18 years agoCreate the taskqueue_fast handler with INTR_MPSAFE so that it doesn't run
Scott Long [Fri, 23 Dec 2005 06:18:33 +0000 (06:18 +0000)]
Create the taskqueue_fast handler with INTR_MPSAFE so that it doesn't run
with Giant.

MFC After: 3 days

18 years agoAdd the example that green@ used in his commit log for the dump to a
Warner Losh [Fri, 23 Dec 2005 02:18:01 +0000 (02:18 +0000)]
Add the example that green@ used in his commit log for the dump to a
pipe functionality.  I've done a cvs log main.c about a dozen times
now.  Maybe others will find this useful.

18 years agoUse device_printf() and if_printf() rather than printf() and axe
Pyun YongHyeon [Fri, 23 Dec 2005 02:04:41 +0000 (02:04 +0000)]
Use device_printf() and if_printf() rather than printf() and axe
bge_unit from the softc.

Requested by: marius

18 years agoMerge tzdata2005q from vendor branch. Data changes for the following
Garrett Wollman [Thu, 22 Dec 2005 23:40:21 +0000 (23:40 +0000)]
Merge tzdata2005q from vendor branch.  Data changes for the following
locales: Azerbaijan, Brazil, Cuba, Gaza, Georgia, Jordan, Kirgizstan,
Nicaragua, Uruguay.

18 years agoThis commit was generated by cvs2svn to compensate for changes in r153667,
Garrett Wollman [Thu, 22 Dec 2005 23:34:50 +0000 (23:34 +0000)]
This commit was generated by cvs2svn to compensate for changes in r153667,
which included commits to RCS files with non-trunk default branches.

18 years agoIndiana still hasn't sorted itself out, and probably won't before the
Garrett Wollman [Thu, 22 Dec 2005 23:34:50 +0000 (23:34 +0000)]
Indiana still hasn't sorted itself out, and probably won't before the
end of the year, so import what I hope to be the last database update
for 2005.

Obtained from: ftp://elsie.nci.nih.gov/pub/tzdata2005q.tar.gz

18 years agoTweak how the MD code calls the fooclock() methods some. Instead of
John Baldwin [Thu, 22 Dec 2005 22:16:09 +0000 (22:16 +0000)]
Tweak how the MD code calls the fooclock() methods some.  Instead of
passing a pointer to an opaque clockframe structure and requiring the
MD code to supply CLKF_FOO() macros to extract needed values out of the
opaque structure, just pass the needed values directly.  In practice this
means passing the pair (usermode, pc) to hardclock() and profclock() and
passing the boolean (usermode) to hardclock_cpu() and hardclock_process().
Other details:
- Axe clockframe and CLKF_FOO() macros on all architectures.  Basically,
  all the archs were taking a trapframe and converting it into a clockframe
  one way or another.  Now they can just extract the PC and usermode values
  directly out of the trapframe and pass it to fooclock().
- Renamed hardclock_process() to hardclock_cpu() as the latter is more
  accurate.
- On Alpha, we now run profclock() at hz (profhz == hz) rather than at
  the slower stathz.
- On Alpha, for the TurboLaser machines that don't have an 8254
  timecounter, call hardclock() directly.  This removes an extra
  conditional check from every clock interrupt on Alpha on the BSP.
  There is probably room for even further pruning here by changing Alpha
  to use the simplified timecounter we use on x86 with the lapic timer
  since we don't get interrupts from the 8254 on Alpha anyway.
- On x86, clkintr() shouldn't ever be called now unless using_lapic_timer
  is false, so add a KASSERT() to that affect and remove a condition
  to slightly optimize the non-lapic case.
- Change prototypeof  arm_handler_execute() so that it's first arg is a
  trapframe pointer rather than a void pointer for clarity.
- Use KCOUNT macro in profclock() to lookup the kernel profiling bucket.

Tested on: alpha, amd64, arm, i386, ia64, sparc64
Reviewed by: bde (mostly)

18 years agobge(4) is working on sparc64 now.
Christian Brueffer [Thu, 22 Dec 2005 21:46:05 +0000 (21:46 +0000)]
bge(4) is working on sparc64 now.

18 years agoAdd entry for /rescue/nextboot.sh since it was renamed to /rescue/nextboot.
John Baldwin [Thu, 22 Dec 2005 21:16:06 +0000 (21:16 +0000)]
Add entry for /rescue/nextboot.sh since it was renamed to /rescue/nextboot.

18 years agoInitialize object dagmembers list before checking version dependencies.
Alexander Kabaev [Thu, 22 Dec 2005 16:42:38 +0000 (16:42 +0000)]
Initialize object dagmembers list before checking version dependencies.

18 years agoReturn BUS_PROBE_GENERIC rather than 0 in the probe routine.
John Baldwin [Thu, 22 Dec 2005 16:25:22 +0000 (16:25 +0000)]
Return BUS_PROBE_GENERIC rather than 0 in the probe routine.

Requested by: marius

18 years agoUse the copy of the card's MAC address saved in tulip_enaddr() in the softc
John Baldwin [Thu, 22 Dec 2005 16:18:23 +0000 (16:18 +0000)]
Use the copy of the card's MAC address saved in tulip_enaddr() in the softc
if we need a valid MAC address (for probing the media for example) before
ether_ifattach() has been called since IF_LLADDR() is NULL then.

Tested by: tisco

18 years agoEnable the cs and disable the amdsmb and nfsmb on pc98.
Yoshihiro Takahashi [Thu, 22 Dec 2005 16:16:42 +0000 (16:16 +0000)]
Enable the cs and disable the amdsmb and nfsmb on pc98.

18 years ago- All bge(4) supported hardware is known to support RX/TX checksum offloading,
Gleb Smirnoff [Thu, 22 Dec 2005 15:14:42 +0000 (15:14 +0000)]
- All bge(4) supported hardware is known to support RX/TX checksum offloading,
  except for BGE_CHIPID_BCM5700_B0, which is buggy.
- All bge(4) supported hardware, has a bug that produces incorrect checksums
  on Ethernet runts. However, in case of a transmitted packet, the latter can
  be padded with zeroes, and the checksum would be correct. (Probably chip
  includes the pad data into checksum). In case of receive, we just don't
  trust checksum data in received runts.

Obtained from: NetBSD (jonathan) via Mihail Balikov

18 years agoExplicitely use a "signed char" instead of a "char", for those archs where
Olivier Houchard [Thu, 22 Dec 2005 14:23:54 +0000 (14:23 +0000)]
Explicitely use a "signed char" instead of a "char", for those archs where
char defaults to unsigned.

18 years ago1) remove useless check of loop_copy - corresponding code was removed in
Oleg Bulyzhin [Thu, 22 Dec 2005 12:16:20 +0000 (12:16 +0000)]
1) remove useless check of loop_copy - corresponding code was removed in
   rev. 1.70 five years ago.
2) convert loop_copy to "non-negative" flag

Approved by: glebius (mentor)
MFC after: 2 weeks

18 years agoFix typo
Dima Dorfman [Thu, 22 Dec 2005 10:51:14 +0000 (10:51 +0000)]
Fix typo

18 years agoCopy the description for the -r option from the traceroute(8) man
Dima Dorfman [Thu, 22 Dec 2005 10:48:42 +0000 (10:48 +0000)]
Copy the description for the -r option from the traceroute(8) man
page. I'm pretty sure that this description applies the same way to
ipv6, and at least mentioning SO_DONTROUTE is better than having no
description at all.

18 years agoIf we're operating without running external programs (-N) and
Dima Dorfman [Thu, 22 Dec 2005 10:36:58 +0000 (10:36 +0000)]
If we're operating without running external programs (-N) and
autonumbering is requested, set the unit to 0 instead of -1. This
option is just for output, and "/dev/md-1" looks disconcerting.

Submitted by: Wojciech A. Koszek

18 years agoSort the list results by the unit number. The list returned by the
Dima Dorfman [Thu, 22 Dec 2005 10:32:11 +0000 (10:32 +0000)]
Sort the list results by the unit number. The list returned by the
kernel is in the order the devices were made, which is not useful to
the user. Also, remove the "%d more" test since the kernel does not
return the complete count in md_pad[0] (maybe it should?).

Submitted by: Wojciech A. Koszek

18 years agoAdd a quirk to fix resume on some laptops.
Gleb Smirnoff [Thu, 22 Dec 2005 09:09:39 +0000 (09:09 +0000)]
Add a quirk to fix resume on some laptops.

Reported by: joe
Reported by: Huang wen hui <huang gddsn.org.cn>
Reported by: Jacques Garrigue <garrigue math.nagoya-u.ac.jp>
PR: kern/89825

18 years agoFix bge_eeprom_getbyte() to return 1 when timeout happens.
Pyun YongHyeon [Thu, 22 Dec 2005 02:03:57 +0000 (02:03 +0000)]
Fix bge_eeprom_getbyte() to return 1 when timeout happens.
Previously it always returned 0 which means success regardless of
EEPROM status.

While here, add a check whether EEPROM read is successful.

Submitted by:   jkim

18 years agoUncomment bge(4) as it's now working.
Pyun YongHyeon [Thu, 22 Dec 2005 01:46:28 +0000 (01:46 +0000)]
Uncomment bge(4) as it's now working.

18 years agoAdd bge(4) support for big-endian architectures(part 2/2).
Pyun YongHyeon [Thu, 22 Dec 2005 01:44:27 +0000 (01:44 +0000)]
Add bge(4) support for big-endian architectures(part 2/2).
 - removed unused funtion bge_handle_events().
 - removed bus_dmamap_destroy(9) calls for DMA maps created by
   bus_dmamem_alloc(9). This should fix panics seen on sparc64
   in device detach.
 - added check for parent DMA tag creation.
 - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all
   architectures.
 - added missing bus_dmamap_sync(9) in bge_txeof().
 - added missing bus_dmamap_sync(9) in bge_encap().
 - corrected memory synchronization operation on status block.
   As the driver just read status block that was DMAed by NIC it
   should use BUS_DMASYNC_POSTREAD. Likewise the driver does not
   need to write status block back, so remove unnecessary
   bus_dmamap_sync(9) calls in bge_intr().
 - corrected memory synchronization operation on RX return ring.
   The driver only read the block so remove unnecessary
   bus_dmamap_sync(9) in bge_rxeof().
 - force bus_dmamap_sync(9) for only modified descriptors. Blindly
   synching all desciptor rings would reduce performance.
 - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof().

Reviewed by: jkim(initial version)
Tested by: glebius(i386), jkim(amd64 initial version)

18 years agoAdd RFC 3378 EtherIP support. This change makes it possible to add gif
Andrew Thompson [Wed, 21 Dec 2005 21:29:45 +0000 (21:29 +0000)]
Add RFC 3378 EtherIP support. This change makes it possible to add gif
interfaces to bridges, which will then send and receive IP protocol 97 packets.
Packets are Ethernet frames with an EtherIP header prepended.

Obtained from: NetBSD
MFC after: 2 weeks

18 years agoMaintain the vnode lock throughout elfN_load_file() rather than releasing
Alan Cox [Wed, 21 Dec 2005 18:58:40 +0000 (18:58 +0000)]
Maintain the vnode lock throughout elfN_load_file() rather than releasing
it and reacquiring it in vrele().  Consequently, there is no reason to
increase the reference count on the vm object caching the file's pages.
Reviewed by: tegge

Eliminate unused parameters to elfN_load_file().

18 years agoDrivers for AMD-8111 and NVIDIA nForce2/3/4 SMBus 2.0 controllers.
Ruslan Ermilov [Wed, 21 Dec 2005 15:49:51 +0000 (15:49 +0000)]
Drivers for AMD-8111 and NVIDIA nForce2/3/4 SMBus 2.0 controllers.

18 years agoUse ETHER_ADDR_LEN rather than hardcoding 6.
John Baldwin [Wed, 21 Dec 2005 15:15:43 +0000 (15:15 +0000)]
Use ETHER_ADDR_LEN rather than hardcoding 6.

18 years agoOoops, I removed the wrong bits.
Olivier Houchard [Wed, 21 Dec 2005 15:02:31 +0000 (15:02 +0000)]
Ooops, I removed the wrong bits.
This unbreak boot from a VA which is different from the PA.

18 years agoRemove #undef DDB I shouldn't have committed.
Olivier Houchard [Wed, 21 Dec 2005 15:01:49 +0000 (15:01 +0000)]
Remove #undef DDB I shouldn't have committed.

18 years agoAdd missing MODULE_DEPEND() so that ppbus.ko and these .ko's can be
Ruslan Ermilov [Wed, 21 Dec 2005 10:54:47 +0000 (10:54 +0000)]
Add missing MODULE_DEPEND() so that ppbus.ko and these .ko's can be
loaded dynamically.

18 years agoPrevious revision was broken on SPARC, fix it by using more appropriate type.
Xin LI [Wed, 21 Dec 2005 10:12:05 +0000 (10:12 +0000)]
Previous revision was broken on SPARC, fix it by using more appropriate type.

Reported by: tindebox
Pointy hat to: delphij

18 years agosyslogd should REQUIRE newsyslog, rather than newsyslog using
Doug Barton [Wed, 21 Dec 2005 09:54:15 +0000 (09:54 +0000)]
syslogd should REQUIRE newsyslog, rather than newsyslog using
BEFORE: syslogd. This does not produce any change in the
ordering at the moment, but is cleaner style for the long term.

18 years agoSeveral users have commented (via filing PRs) that having ntp* depend
Doug Barton [Wed, 21 Dec 2005 09:48:41 +0000 (09:48 +0000)]
Several users have commented (via filing PRs) that having ntp* depend
on devfs is useful so that a hardware time device can start with its
necessary device nodes already in place. While this ordering happens
as a side effect currently in HEAD, and the PRs were generally fixed
via upgrades, etc; it's better to make it explicit.

While I'm here, ntpd should REQUIRE ntpdate, rather than ntpdate
using BEFORE: ntpd.

18 years agoAs of r1.21 all broadcast packets are reprocessed by ether_input as arriving on
Andrew Thompson [Wed, 21 Dec 2005 09:39:59 +0000 (09:39 +0000)]
As of r1.21 all broadcast packets are reprocessed by ether_input as arriving on
the bridge, this caused these packets to show up twice via bpf. Do not process
them twice with BPF_TAP.

MFC after: 3 days

18 years agoChange the default for deleting stale files from yes (which violates
Doug Barton [Wed, 21 Dec 2005 08:59:22 +0000 (08:59 +0000)]
Change the default for deleting stale files from yes (which violates
the general principle that mm should do nothing by default) to no.

PR: bin/85449
Submitted by: Kevin Oberman <oberman@es.net>

18 years agoo Fix typo: id_t -> uid_t.
Maxim Konovalov [Wed, 21 Dec 2005 08:23:04 +0000 (08:23 +0000)]
o Fix typo: id_t -> uid_t.
o Use align aware CMSG_LEN() macro for cmsg_len computation.

PR: docs/90731
Submitted by: Andrey Simonenko
Reviewed by: mdodd
MFC after: 1 week

18 years agoMFi386: revisions 1.34 and 1.36.
Yoshihiro Takahashi [Wed, 21 Dec 2005 06:10:42 +0000 (06:10 +0000)]
MFi386: revisions 1.34 and 1.36.

18 years agoMFi386: revision 1.8.
Yoshihiro Takahashi [Wed, 21 Dec 2005 06:09:47 +0000 (06:09 +0000)]
MFi386: revision 1.8.

18 years agoMFi386: revision 1.46.
Yoshihiro Takahashi [Wed, 21 Dec 2005 06:09:09 +0000 (06:09 +0000)]
MFi386: revision 1.46.

18 years agoLet _mutex_cv_lock call internal functiona mutex_lock_common.
David Xu [Wed, 21 Dec 2005 05:14:07 +0000 (05:14 +0000)]
Let _mutex_cv_lock call internal functiona mutex_lock_common.

18 years agoHide umtx API symbols as well.
David Xu [Wed, 21 Dec 2005 03:53:29 +0000 (03:53 +0000)]
Hide umtx API symbols as well.

18 years ago1. Retire macro SCLASS, instead simply use language keyword and
David Xu [Wed, 21 Dec 2005 03:14:06 +0000 (03:14 +0000)]
1. Retire macro SCLASS, instead simply use language keyword and
   put variables in thr_init.c.
2. Hide all global symbols which won't be exported.

18 years agoadd LINT build
Sam Leffler [Wed, 21 Dec 2005 02:51:41 +0000 (02:51 +0000)]
add LINT build

Discussed with: grehan
MFC after: 1 week

18 years agoFollow the mistake in libpthread, the first version name in libpthread
David Xu [Wed, 21 Dec 2005 02:34:20 +0000 (02:34 +0000)]
Follow the mistake in libpthread, the first version name in libpthread
is LIBTHREAD_1_0, but really it should be LIBPTHREAD_1_0.
Fix it so libmap.conf works again (it was broken by recent versioning
code in rtld_elf).

18 years agoUse size_t for length.
Xin LI [Wed, 21 Dec 2005 02:25:59 +0000 (02:25 +0000)]
Use size_t for length.

18 years agoFor the cases when loading bzip2-compressed kernels enabled use last
Maxim Sobolev [Wed, 21 Dec 2005 02:17:58 +0000 (02:17 +0000)]
For the cases when loading bzip2-compressed kernels enabled use last
3MB of physical memory for heap instead of range between 1MB and 4MB.
This makes this feature working with PAE and amd64 kernels, which are
loaded at 2MB. Teach i386_copyin() to avoid using range allocated by
heap in such case, so that it won't trash heap in the low memory
conditions.

This should make loading bzip2-compressed kernels/modules/mfs images
generally useable, so that re@ team is welcome to evaluate merits
of using this feature in the installation CDs.

Valuable suggestions by: jhb

18 years agoREQUIRE: syslogd and BEFORE: NETWORKING are now antithetical,
Doug Barton [Wed, 21 Dec 2005 01:19:20 +0000 (01:19 +0000)]
REQUIRE: syslogd and BEFORE: NETWORKING are now antithetical,
and including both in this file had nasty side effects on the
ordering of syslogd, as well as producing an error when running
rcorder. Remove the more bogus of the two options, which restores
proper ordering and removes the error.

There is an open question as to whether scripts with the nostart
KEYWORD should even have REQUIRE/BEFORE lines, and indeed, whether
they should be in /etc/rc.d at all, but that's for another time.

18 years agodon't assume char is signed
Sam Leffler [Wed, 21 Dec 2005 00:19:08 +0000 (00:19 +0000)]
don't assume char is signed

MFC after: 1 week

18 years agobandaid assumption that char is signed
Sam Leffler [Wed, 21 Dec 2005 00:09:11 +0000 (00:09 +0000)]
bandaid assumption that char is signed

MFC after: 1 week

18 years agoEliminate an unneeded (vm_prot_t) parameter from two functions. Eliminate
Alan Cox [Tue, 20 Dec 2005 23:42:18 +0000 (23:42 +0000)]
Eliminate an unneeded (vm_prot_t) parameter from two functions.  Eliminate
unnecessary uses of a local variable.

Reviewed by: tegge

18 years agoInclude a somewhat hackish way to make sure that we *always* test the
Doug Barton [Tue, 20 Dec 2005 23:22:47 +0000 (23:22 +0000)]
Include a somewhat hackish way to make sure that we *always* test the
new clear_tmp_X variable when start'ing.

18 years ago- Fix type in previous commit; unbreak build
Pav Lucistnik [Tue, 20 Dec 2005 23:07:38 +0000 (23:07 +0000)]
- Fix type in previous commit; unbreak build

Approved by: ssouhlal
Pointy hat to: pav

18 years agoMention PC Card support for the cs device
Warner Losh [Tue, 20 Dec 2005 23:05:50 +0000 (23:05 +0000)]
Mention PC Card support for the cs device

18 years agoMove device 'cs' into i386/pc98 specific NOTES. It is broken on ppc
Warner Losh [Tue, 20 Dec 2005 23:00:11 +0000 (23:00 +0000)]
Move device 'cs' into i386/pc98 specific NOTES.  It is broken on ppc
because it uses i386 specific calls.  Maybe it could be added to
amd64, but I'm not so sure it would work there so I've not added it
there.

18 years ago- Use PCIR_BAR() macro for the BAR for the aperture.
John Baldwin [Tue, 20 Dec 2005 22:45:24 +0000 (22:45 +0000)]
- Use PCIR_BAR() macro for the BAR for the aperture.
- Axe macros used for walking PCI capabilities list.  We now ask the PCI
  bus to find caps for us rather than doing it in the drm and agp drivers.

18 years ago- Bump FreeBSD version for the hostb(4) and vgapci(4) drivers as well as
John Baldwin [Tue, 20 Dec 2005 22:44:36 +0000 (22:44 +0000)]
- Bump FreeBSD version for the hostb(4) and vgapci(4) drivers as well as
  the addition of pci_find_extcap().
- Change the drm drivers to attach to vgapci.  This is #ifdef'd so the
  code can be shared across branches.
- Use pci_find_extcap() to look for AGP and PCIE capabilities in drm.
- GC all the drmsub stuff for i810/i830/i915.  The agp and drm devices are
  now both children of vgapci.

18 years agoAttach to the vgapci device rather than pci.
John Baldwin [Tue, 20 Dec 2005 22:42:16 +0000 (22:42 +0000)]
Attach to the vgapci device rather than pci.

18 years agoAdd a vgapci(4) stub device driver for VGA PCI devices. This device serves
John Baldwin [Tue, 20 Dec 2005 22:41:33 +0000 (22:41 +0000)]
Add a vgapci(4) stub device driver for VGA PCI devices.  This device serves
as a bus so that other drivers such as drm(4), acpi_video(4), and agp(4)
can attach to it thus allowing multiple drivers for the same device.  It
also removes the need for the drmsub hack for the i8[13]0/i915 drm and agp
drivers.

18 years agoAdd a vgapci(4) stub device driver for VGA PCI devices. This device serves
John Baldwin [Tue, 20 Dec 2005 22:35:39 +0000 (22:35 +0000)]
Add a vgapci(4) stub device driver for VGA PCI devices.  This device serves
as a bus so that other drivers such as drm(4), acpi_video(4), and agp(4)
can attach to it thus allowing multiple drivers for the same device.  It
also removes the need for the drmsub hack for the i8[13]0/i915 drm and agp
drivers.

18 years agoRemove GIANT from device random.
Paul Saab [Tue, 20 Dec 2005 21:41:52 +0000 (21:41 +0000)]
Remove GIANT from device random.

Submitted by: ups

18 years agoFirst step in bringing this page up to speed:
Christian Brueffer [Tue, 20 Dec 2005 21:40:25 +0000 (21:40 +0000)]
First step in bringing this page up to speed:

Update list of supported bridges and controllers, use markup similar to
ata.4 to display them and replace an XFree86 xref with Xorg.

MFC after: 1 week

18 years agoChange the various AGP drivers that attach to the Host-PCI bridge device to
John Baldwin [Tue, 20 Dec 2005 21:12:26 +0000 (21:12 +0000)]
Change the various AGP drivers that attach to the Host-PCI bridge device to
attach to the hostb driver instead.  This means that agp can now be loaded
at runtime (in theory at least).  Also, the drivers no longer have to
explicity call device_verbose() to cancel out any earlier calls to
device_quiet() by the hostb(4) driver (this shows a limitation in new-bus,
drivers really shouldn't be doing device_quiet() until they know they are
going to drive that device, i.e. in attach).

18 years agoMove the hostb driver out of the i386 and amd64 PCI code (where it was
John Baldwin [Tue, 20 Dec 2005 21:09:45 +0000 (21:09 +0000)]
Move the hostb driver out of the i386 and amd64 PCI code (where it was
duplicated anyways) and into a single MI driver.  Extend the driver a bit
to implement the bus and PCI kobj interfaces such that other drivers can
attach to it and transparently act as if their parent device is the PCI
bus (for the most part).