]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agoAdd missing device in tunefs entry.
remko [Sun, 13 Apr 2008 07:48:05 +0000 (07:48 +0000)]
Add missing device in tunefs entry.

PR: docs/122702
Submitted by: Yoshihiro Ota <ota@j.email.ne.jp>
MFC After: 3 days

16 years agoFreeBSD/mips port. The FreeBSD/mips port targets mips32, mips64,
imp [Sun, 13 Apr 2008 07:44:55 +0000 (07:44 +0000)]
FreeBSD/mips port.  The FreeBSD/mips port targets mips32, mips64,
mips32r2 and mips64r2 (and close relatives) processors.  There
presently is support for ADMtek ADM5120, A mips 4Kc in a malta board,
the RB533 routerboard (based on IDT RC32434) and some preliminary
support for sibtye/broadcom designs.  Other hardware support will be
forthcomcing.

This port boots multiuser under gxemul emulating the malta board and
also bootstraps on the hardware whose support is forthcoming...

Oleksandr Tymoshenko, Wojciech Koszek, Warner Losh, Olivier Houchard,
Randall Stewert and others that have contributed to the mips2 and/or
mips2-jnpr perforce branches.  Juniper contirbuted a generic mips port
late in the life cycle of the misp2 branch.  Warner Losh merged the
mips2 and Juniper code bases, and others list above have worked for
the past several months to get to multiuser.

In addition, the mips2 work owe a debt to the trail blazing efforts of
the original mips branch in perforce done by Juli Mallett.

16 years agoFreeBSD/mips port. The FreeBSD/mips port targets mips32, mips64,
imp [Sun, 13 Apr 2008 07:27:37 +0000 (07:27 +0000)]
FreeBSD/mips port.  The FreeBSD/mips port targets mips32, mips64,
mips32r2 and mips64r2 (and close relatives) processors.  There
presently is support for ADMtek ADM5120, A mips 4Kc in a malta board,
the RB533 routerboard (based on IDT RC32434) and some preliminary
support for sibtye/broadcom designs.  Other hardware support will be
forthcomcing.

This port boots multiuser under gxemul emulating the malta board and
also bootstraps on the hardware whose support is forthcoming...

Oleksandr Tymoshenko, Wojciech Koszek, Warner Losh, Olivier Houchard,
Randall Stewert and others that have contributed to the mips2 and/or
mips2-jnpr perforce branches.  Juniper contirbuted a generic mips port
late in the life cycle of the misp2 branch.  Warner Losh merged the
mips2 and Juniper code bases, and others list above have worked for
the past several months to get to multiuser.

In addition, the mips2 work owe a debt to the trail blazing efforts of
the original mips branch in perforce done by Juli Mallett.

16 years agoFreeBSD/mips port. The FreeBSD/mips port targets mips32, mips64,
imp [Sun, 13 Apr 2008 07:07:57 +0000 (07:07 +0000)]
FreeBSD/mips port.  The FreeBSD/mips port targets mips32, mips64,
mips32r2 and mips64r2 (and close relatives) processors.  There
presently is support for ADMtek ADM5120, A mips 4Kc in a malta board,
the RB533 routerboard (based on IDT RC32434) and some preliminary
support for sibtye/broadcom designs.  Other hardware support will be
forthcomcing.

This port boots multiuser under gxemul emulating the malta board and
also bootstraps on the hardware whose support is forthcoming...

Oleksandr Tymoshenko, Wojciech Koszek, Warner Losh, Olivier Houchard,
Randall Stewert and others that have contributed to the mips2 and/or
mips2-jnpr perforce branches.  Juniper contirbuted a generic mips port
late in the life cycle of the misp2 branch.  Warner Losh merged the
mips2 and Juniper code bases, and others list above have worked for
the past several months to get to multiuser.

In addition, the mips2 work owe a debt to the trail blazing efforts of
the original mips branch in perforce done by Juli Mallett.

16 years agoMerge in the mips specific configuration files and such from the
imp [Sun, 13 Apr 2008 06:25:43 +0000 (06:25 +0000)]
Merge in the mips specific configuration files and such from the
merged juniper and mips2 code base.  This represents the work of
Juniper Engineers, plus Oleksandr Tymoshenko, Wojciech Koszek, Warner
Losh, Olivier Houchard, Randall Stewert and others that have
contributed to the mips2 and/or mips2-jnpr perforce branches.

16 years agoMFi386: RTC related cleanups.
nyan [Sun, 13 Apr 2008 06:18:34 +0000 (06:18 +0000)]
MFi386: RTC related cleanups.

- Use generic RTC handling code.
- Make clock_if.m and subr_rtc.c standard.
- Nuke MD inittodr(), resettodr() functions.
- Add new "pcrtc" device driver.
- Add hints for "pcrtc" driver.

16 years agoThese files handle the radix tree for the ECMP routes.
qingli [Sun, 13 Apr 2008 06:12:13 +0000 (06:12 +0000)]
These files handle the radix tree for the ECMP routes.
The original code from KAME did not take care of address
aliases or multiple ip addresses that have the same
prefix.

Reviewed by: rwatson, gnn, sam, kmacy, julian

16 years agoThis patch provides the back end support for equal-cost multi-path
qingli [Sun, 13 Apr 2008 05:45:14 +0000 (05:45 +0000)]
This patch provides the back end support for equal-cost multi-path
(ECMP) for both IPv4 and IPv6. Previously, multipath route insertion
is disallowed. For example,

route add -net 192.103.54.0/24 10.9.44.1
route add -net 192.103.54.0/24 10.9.44.2

The second route insertion will trigger an error message of
"add net 192.103.54.0/24: gateway 10.2.5.2: route already in table"

Multiple default routes can also be inserted. Here is the netstat
output:

default 10.2.5.1 UGS 0 3074 bge0 =>
default 10.2.5.2 UGS 0 0 bge0

When multipath routes exist, the "route delete" command requires
a specific gateway to be specified or else an error message would
be displayed. For example,

route delete default

would fail and trigger the following error message:

"route: writing to routing socket: No such process"
"delete net default: not in table"

On the other hand,

route delete default 10.2.5.2

would be successful: "delete net default: gateway 10.2.5.2"

One does not have to specify a gateway if there is only a single
route for a particular destination.

I need to perform more testings on address aliases and multiple
interfaces that have the same IP prefixes. This patch as it
stands today is not yet ready for prime time. Therefore, the ECMP
code fragments are fully guarded by the RADIX_MPATH macro.
Include the "options  RADIX_MPATH" in the kernel configuration
to enable this feature.

Reviewed by: robert, sam, gnn, julian, kmacy

16 years agoUse a "rel" memory barrier for disowning the lock as it cames from an
attilio [Sun, 13 Apr 2008 01:21:56 +0000 (01:21 +0000)]
Use a "rel" memory barrier for disowning the lock as it cames from an
exclusive locking operation.

16 years agostruct lock_instance and struct lock_list_entry don't need to be in the
attilio [Sun, 13 Apr 2008 01:20:47 +0000 (01:20 +0000)]
struct lock_instance and struct lock_list_entry don't need to be in the
public namespace for WITNESS as they are only used internally so just
move them in the private namespace for the subsystem (with all related
supporting definitions).

16 years agofix printf type confusion on amd64
phk [Sat, 12 Apr 2008 21:51:54 +0000 (21:51 +0000)]
fix printf type confusion on amd64

16 years agoMove i386 to generic RTC handling code.
phk [Sat, 12 Apr 2008 20:46:06 +0000 (20:46 +0000)]
Move i386 to generic RTC handling code.

Make clock_if.m and subr_rtc.c standard on i386

Add hints for "atrtc" driver, for non-PnP, non-ACPI systems.
NB: Make sure to install GENERIC.hints into /boot/device.hints in these!

Nuke MD inittodr(), resettodr() functions.

Don't attach to PHP0B00 in the "attimer" dummy driver any more, and remove
comments that no longer apply for that reason.

Add new "atrtc" device driver, which handles IBM PC AT Real Time
Clock compatible devices using subr_rtc and clock_if.

This driver is not entirely clean: other code still fondles the
hardware to get a statclock interrupt on non-ACPI timer systems.

Wrap some overly long lines.

After it has settled in -current, this will be ported to amd64.

Technically this is MFC'able, but I fail to see a good reason.

16 years agoEmit summaries of struct c(alender)t(ime) <-> struct timespec conversions
phk [Sat, 12 Apr 2008 20:35:56 +0000 (20:35 +0000)]
Emit summaries of struct c(alender)t(ime) <-> struct timespec conversions
under bootverbose.

Struct ct is used for setting/reading real time clocks and I'm about
to Do Things to some of those, so a bit of preemptive debugging is
in order.

Remove a pointless __inline.

16 years agoRemove a trailing comma which FlexeLint whines about.
phk [Sat, 12 Apr 2008 20:26:07 +0000 (20:26 +0000)]
Remove a trailing comma which FlexeLint whines about.

16 years agoLK_NOWITNESS can be used now in lockmgr*() functions in order to skip
attilio [Sat, 12 Apr 2008 20:18:02 +0000 (20:18 +0000)]
LK_NOWITNESS can be used now in lockmgr*() functions in order to skip
witness(4) checks on a per-instance basis.

16 years ago- Re-introduce WITNESS support for lockmgr. About the old implementation
attilio [Sat, 12 Apr 2008 19:57:30 +0000 (19:57 +0000)]
- Re-introduce WITNESS support for lockmgr.  About the old implementation
  the only one difference is that lockmgr*() functions now accept
  LK_NOWITNESS flag which skips ordering for the instanced calling.
- Remove an unuseful stub in witness_checkorder() (because the above check
  doesn't allow ever happening) and allow witness_upgrade() to accept
  non-try operation too.

16 years agoDont call hw.status recursively.
sos [Sat, 12 Apr 2008 17:21:22 +0000 (17:21 +0000)]
Dont call hw.status recursively.

Spotted by: Marcel Moolenaar

16 years ago- Add codec id for Sigmatel STAC9205
ariff [Sat, 12 Apr 2008 15:07:32 +0000 (15:07 +0000)]
- Add codec id for Sigmatel STAC9205
- Fix speaker issues with Dell Vostro 1500 (GPIO0)

Tested by: John Wright <jwright.gmail.com>

- Apply ridiculous quirk on Asus A8X series (A8JC, A8M, A8xx, etc). These
  different laptop series share simmilar pci id, hardware codecs, etc.
  but works differently. A slight difference in connection type for
  widget #26 is used to differentiate it.

Tested by: eric baumbach <embaumbach.gmail.com>

- Apply GPIO0 quirk for ASUS G2K laptop
- Sort ASUS ids accordingly.

Submitted by: jkim

MFC after: 3 days

16 years agoUnbreak the build for arm and powerpc.
das [Sat, 12 Apr 2008 14:53:52 +0000 (14:53 +0000)]
Unbreak the build for arm and powerpc.

Pointy hat to yours truly.

16 years agoConnect k8temp(4) to the build.
rpaulo [Sat, 12 Apr 2008 14:20:22 +0000 (14:20 +0000)]
Connect k8temp(4) to the build.

16 years agoMFp4: k8temp(4) man page.
rpaulo [Sat, 12 Apr 2008 14:08:24 +0000 (14:08 +0000)]
MFp4: k8temp(4) man page.

16 years agoMFp4: k8temp, a driver to monitor AMD K8 CPU temperature via builtin
rpaulo [Sat, 12 Apr 2008 14:04:08 +0000 (14:04 +0000)]
MFp4: k8temp, a driver to monitor AMD K8 CPU temperature via builtin
sensors. Based on the Linux driver by the same name.

Tested by:     many (see freebsd-amd64)

16 years ago- Remove a stale comment.
attilio [Sat, 12 Apr 2008 13:56:17 +0000 (13:56 +0000)]
- Remove a stale comment.
- Add an extra assertion in order to catch malformed requested operations.

16 years agoAdd missing stubs for spinlocks cpuset and intrcnt.
attilio [Sat, 12 Apr 2008 13:51:18 +0000 (13:51 +0000)]
Add missing stubs for spinlocks cpuset and intrcnt.

Submitted by: kris

16 years agoTrylocks were implemented, remove stale comment
kris [Sat, 12 Apr 2008 12:17:49 +0000 (12:17 +0000)]
Trylocks were implemented, remove stale comment

16 years agoUpdate the list of Cx states when ACPICA notifies us. Usually, this
rpaulo [Sat, 12 Apr 2008 12:06:00 +0000 (12:06 +0000)]
Update the list of Cx states when ACPICA notifies us. Usually, this
notification is sent when the AC plug is plugged in/out.

This is required on some laptops, namely the MacBooks.

Silence on:  freebsd-acpi

16 years agoMatch the Mac Pro with 8 cores. This machine reports 35 temperature
rpaulo [Sat, 12 Apr 2008 12:04:09 +0000 (12:04 +0000)]
Match the Mac Pro with 8 cores. This machine reports 35 temperature
sensors and some fans.
As I don't own this machine, I could not test this patch.

Obtained from: applesmc Linux driver.

16 years agoCorrect an obvious typo.
pjd [Sat, 12 Apr 2008 05:49:05 +0000 (05:49 +0000)]
Correct an obvious typo.

16 years agoUpdates for changes in the way printf() handles hex floating point
das [Sat, 12 Apr 2008 03:11:56 +0000 (03:11 +0000)]
Updates for changes in the way printf() handles hex floating point
numbers.

16 years agoMake several changes to the way printf handles hex floating point (%a):
das [Sat, 12 Apr 2008 03:11:36 +0000 (03:11 +0000)]
Make several changes to the way printf handles hex floating point (%a):

1. Previously, printing the number 1.0 could produce 0x1p+0, 0x2p-1,
   0x4p-2, or 0x8p-3, depending on what happened to be convenient. This
   meant that printing a value as a double and printing the same value
   as a long double could produce different (but equivalent) results.
   The change is to always make the leading digit a 1, unless the
   number is 0. This solves the aforementioned problem and has
   several other advantages.

2. Use the FPU to do rounding. This is far simpler and more portable
   than manipulating the bits, and it fixes an obsure round-to-even
   bug. It also raises the exceptions now required by IEEE 754R.
   The drawbacks are that it is usually slightly slower, and it makes
   printf less effective as a debugging tool when the FPU is hosed
   (e.g., due to a buggy softfloat implementation).

3. On i386, twiddle the rounding precision so that (2) works properly
   for long doubles.

4. Make several simplifications that are now possible due to (2).

5. Split __hldtoa() into a separate file.

Thanks to remko for access to a sparc64 box for testing.

16 years agoFix some bugs that caused sparc64's quad precision sqrt to get
das [Sat, 12 Apr 2008 03:10:13 +0000 (03:10 +0000)]
Fix some bugs that caused sparc64's quad precision sqrt to get
the wrong answer for virtually all inputs.

Thanks to remko for access to a sparc64 box for testing.

16 years agoMake the software emulator for long doubles set the FPU exception
das [Sat, 12 Apr 2008 03:09:51 +0000 (03:09 +0000)]
Make the software emulator for long doubles set the FPU exception
flags appropriately. The next step is to make it raise a SIGFPE if
any exceptions are unmasked.

Thanks to remko for access to a sparc64 box for testing.

16 years ago- Fixed a problem with the send chain consumer index which would cause
davidch [Fri, 11 Apr 2008 23:10:40 +0000 (23:10 +0000)]
- Fixed a problem with the send chain consumer index which would cause
  TX traffic to sit in the send chain until a received packet kick
  started the interrupt handler.  This would cause extremely slow
  performance when used with NFS over UDP.
- Removed untested polling code.
- Updated copyright year in the file header.
- Removed inadvertent ^M's created by DOS text editor.

MFC after: 2 weeks

16 years ago - Pass the irq and not the vector to intr_event_create().
jeff [Fri, 11 Apr 2008 23:10:39 +0000 (23:10 +0000)]
 - Pass the irq and not the vector to intr_event_create().

Reviewed by: marcel

16 years agoHook up ZFS to the sparc64 build.
marius [Fri, 11 Apr 2008 23:04:36 +0000 (23:04 +0000)]
Hook up ZFS to the sparc64 build.

Approved by: pjd
MFC after: 2 weeks

16 years agoAdd atomic operations for ZFS/sparc64.
marius [Fri, 11 Apr 2008 22:59:33 +0000 (22:59 +0000)]
Add atomic operations for ZFS/sparc64.

Approved by: core, pjd
Obtained from: OpenSolaris (w/ adaptations)
MFC after: 2 weeks

16 years agoFix badly placed '{'
sos [Fri, 11 Apr 2008 22:56:27 +0000 (22:56 +0000)]
Fix badly placed '{'
Dont leak requests on busdma failure (not that we'd get anywhere anyhow).

Reported by: antoine@

16 years ago- Fix the path encoded in the multiple inclusion protection.
marius [Fri, 11 Apr 2008 22:53:06 +0000 (22:53 +0000)]
- Fix the path encoded in the multiple inclusion protection.
- GCC uses 32-byte function alignment for UltraSPARC CPUs.
- Remove code duplication.

Approved by: core, pjd
MFC after: 2 weeks

16 years agoAvoid printing spurious ``Header with wrong dumpdate.'' message.
mckusick [Fri, 11 Apr 2008 21:51:53 +0000 (21:51 +0000)]
Avoid printing spurious ``Header with wrong dumpdate.'' message.

16 years agoCorrectly set file group when restore is run by a user other than root.
mckusick [Fri, 11 Apr 2008 21:48:14 +0000 (21:48 +0000)]
Correctly set file group when restore is run by a user other than root.

16 years agoInstead of rolling our own jail number allocation procedure, use
delphij [Fri, 11 Apr 2008 21:31:15 +0000 (21:31 +0000)]
Instead of rolling our own jail number allocation procedure, use
alloc_unr() to do it.

Submitted by: Ed Schouten <ed 80386 nl>
PR: kern/122270
MFC after: 1 month

16 years agoUse kthread_exit() to terminate a taskqueue thread rather than kproc_exit()
jhb [Fri, 11 Apr 2008 17:35:54 +0000 (17:35 +0000)]
Use kthread_exit() to terminate a taskqueue thread rather than kproc_exit()
now that the taskqueue threads are kthreads rather than kprocs.

Reported by: kris

16 years agoDon't call acpi_disabled() because we are not part of the acpi.ko
rpaulo [Fri, 11 Apr 2008 17:16:19 +0000 (17:16 +0000)]
Don't call acpi_disabled() because we are not part of the acpi.ko
module. Instead, use resource_disabled() that doesn't depend on any
module.

Noticed by: Ian FREISLICH <ianf at clue.co.za>

16 years agoDon't mess up with CHN_F_TRIGGERED, since it should exclusively
ariff [Fri, 11 Apr 2008 15:26:25 +0000 (15:26 +0000)]
Don't mess up with CHN_F_TRIGGERED, since it should exclusively
be handled by chn_abort() and chn_start() alone. This should fix
few issues with single duplex hardware (mostly) or pre virtual record
(RELENG 6) under WINE emulation and possibly others that using
SNDCTL_DSP_SETTRIGGER.

MFC after: 3 days

16 years agoStyle nit. No functional change.
rpaulo [Fri, 11 Apr 2008 12:12:25 +0000 (12:12 +0000)]
Style nit. No functional change.

16 years agosystem_info.cpustates isn't sparse, so a bitmask of available CPU states
ru [Fri, 11 Apr 2008 11:39:26 +0000 (11:39 +0000)]
system_info.cpustates isn't sparse, so a bitmask of available CPU states
is redundant (I think it's a leftover from an older implementation).

16 years agoAllocate enough memory for pcpu_cp_time[] to stop sysctl() from
ru [Fri, 11 Apr 2008 11:34:09 +0000 (11:34 +0000)]
Allocate enough memory for pcpu_cp_time[] to stop sysctl() from
writing outside of array bounds.  This fully fixes -P display on
i386, where kern.cp_times prints zeroes for non-existing CPUs.

16 years agoFix the brokenness in the former commit, sorry for the mess.
sos [Fri, 11 Apr 2008 11:30:27 +0000 (11:30 +0000)]
Fix the brokenness in the former commit, sorry for the mess.
The problem is that the PM support is part of a much larger WIP here, but due to popular demand I decided to get some of it imported.

Also I forgot the mention:

HW sponsored by: Vitsch Electronics / VEHosting

16 years agoFix some issues that showed up during Kris' testing.
dfr [Fri, 11 Apr 2008 10:34:59 +0000 (10:34 +0000)]
Fix some issues that showed up during Kris' testing.

Reported by: kris
MFC after: 3 days

16 years ago - Use a lockmgr lock rather than a mtx to protect dirhash. This lock
jeff [Fri, 11 Apr 2008 09:48:12 +0000 (09:48 +0000)]
 - Use a lockmgr lock rather than a mtx to protect dirhash.  This lock
   may be held for the duration of the various dirhash operations which
   avoids many complex unlock/lock/revalidate sequences.
 - Permit shared locks on lookup.  To protect the ip->i_dirhash pointer we
   use the vnode interlock in the shared case.  Callers holding the
   exclusive vnode lock can run without fear of concurrent modification to
   i_dirhash.
 - Hold an exclusive dirhash lock when creating the dirhash structure for
   the first time or when re-creating a dirhash structure which has been
   recycled.

Tested by: kris, pho

16 years ago - cache dp->i_offset in the local 'i_offset' variable for use in loop
jeff [Fri, 11 Apr 2008 09:44:25 +0000 (09:44 +0000)]
 - cache dp->i_offset in the local 'i_offset' variable for use in loop
   indexes so directory lookup becomes shared lock safe.  In the modifying
   cases an exclusive lock is held here so the commit routine may
   rely on the state of i_offset.
 - Similarly handle i_diroff by fetching at the start and setting only once
   the operation is complete.  Without the exclusive lock these are only
   considered hints.
 - Assert that an exclusive lock is held when we're preparing for a commit
   routine.
 - Honor the lock type request from lookup instead of always using exclusive
   locking.

Tested by: pho, kris

16 years agoOptimize package registration/deregistration. Previously, when looking up the
pav [Fri, 11 Apr 2008 08:26:06 +0000 (08:26 +0000)]
Optimize package registration/deregistration.  Previously, when looking up the
package name for the origin of a dependency, all entries in /var/db/pkg were
traversed for each dependency of added/removed package.  Now, gather all the
origins first, then do the lookup in a single pass over /var/db/pkg.

This should provide a major speedup for packages with hundreds of dependencies.

Submitted by: rdivacky (earlier version)
MFC after: 1 month

16 years agoIdentify ICH9 USB controllers.
benno [Fri, 11 Apr 2008 05:50:53 +0000 (05:50 +0000)]
Identify ICH9 USB controllers.

I've taken a slightly different approach than is used with the ICH8 controllers
in that each controller is not identified individually (eg USB A, USB B, etc).
Instead I've given then same description to each one even though the device ID
differs.  This can easily be changed if desired, or ICH8 (and any others using
that approach) can be made to work as this does.

16 years ago - Add support for interrupt bindig to cpuset(1). Interrupts are bound
jeff [Fri, 11 Apr 2008 03:27:42 +0000 (03:27 +0000)]
 - Add support for interrupt bindig to cpuset(1).  Interrupts are bound
   by specifying the interrupt with -x <irq>.  The irq number matches
   those displayed by vmstat -i.

Sponsored by: Nokia

16 years ago - Add the interrupt vector number to intr_event_create so MI code can
jeff [Fri, 11 Apr 2008 03:26:41 +0000 (03:26 +0000)]
 - Add the interrupt vector number to intr_event_create so MI code can
   lookup hard interrupt events by number.  Ignore the irq# for soft intrs.
 - Add support to cpuset for binding hardware interrupts.  This has the
   side effect of binding any ithread associated with the hard interrupt.
   As per restrictions imposed by MD code we can only bind interrupts to
   a single cpu presently.  Interrupts can be 'unbound' by binding them
   to all cpus.

Reviewed by: jhb
Sponsored by: Nokia

16 years agoAdd a new flag, '-C' which enables a special mode that is intended for
delphij [Thu, 10 Apr 2008 23:49:23 +0000 (23:49 +0000)]
Add a new flag, '-C' which enables a special mode that is intended for
catastrophic recovery.  Currently, this mode only validates whether a
cylindergroup has good signature data, and prompts the user to decide
whether to clear it as a whole.

This mode is useful when there is data damage on a disk and you are
working on copy of the original disk, as fsck_ffs(8) tends to abnormally
exit in such case, as a last resort to recover data from the disk.

16 years agoFix a bug introduced by DEFAULTS feature. When the config file
imp [Thu, 10 Apr 2008 22:57:54 +0000 (22:57 +0000)]
Fix a bug introduced by DEFAULTS feature.  When the config file
doesn't exist, we make a directory and then say "oops, that file isn't
there" leaving the directory behind.  Add a stat for the config file
so that we detect this before making the directory.  This is
semi-lame, but less lame than having this bug.

16 years agoFix clearing of nVidia interrupts.
sos [Thu, 10 Apr 2008 20:40:25 +0000 (20:40 +0000)]
Fix clearing of nVidia interrupts.

16 years agoDon't break identity mapping set up for ACPI resume path.
takawata [Thu, 10 Apr 2008 18:38:31 +0000 (18:38 +0000)]
Don't break identity mapping set up for ACPI resume path.
With this change, BSP processor context seems to be recovered.

16 years agoFix "top -P" (`&' mistyped as `&&' and a botched logic).
ru [Thu, 10 Apr 2008 16:17:54 +0000 (16:17 +0000)]
Fix "top -P" (`&' mistyped as `&&' and a botched logic).
The bug was unnoticed on non-i386 because mp_maxid is
initialized differently, kern.cp_times doesn't print
zeroes for non-existing CPUs, so no "writing outside of
array bounds" happens.

MFC after: 3 days

16 years agoCorrect pmap_copy()'s method for extracting the physical address of a
alc [Thu, 10 Apr 2008 16:04:50 +0000 (16:04 +0000)]
Correct pmap_copy()'s method for extracting the physical address of a
2/4MB page from a PDE.  Specifically, change it to use PG_PS_FRAME,
not PG_FRAME, to extract the physical address of a 2/4MB page from a
PDE.

Change the last argument passed to pmap_pv_insert_pde() from a
vm_page_t representing the first 4KB page of a 2/4MB page to the
vm_paddr_t of the 2/4MB page.  This avoids an otherwise unnecessary
conversion from a vm_paddr_t to a vm_page_t in pmap_copy().

16 years ago- Add ASUS G2K laptop support.
jkim [Thu, 10 Apr 2008 15:17:41 +0000 (15:17 +0000)]
- Add ASUS G2K laptop support.
- Add DLED and GLED found on newer ASUS laptops.
- Turn on BLED, TLED, and WLED by default as other OSes.

Reviewed by: philip
MFC after: 3 days

16 years agoClean up makefiles and a manpage.
ru [Thu, 10 Apr 2008 14:02:00 +0000 (14:02 +0000)]
Clean up makefiles and a manpage.

OK'ed by: phk

16 years agoAdd experimental support for SATA Port Multipliers
sos [Thu, 10 Apr 2008 13:05:05 +0000 (13:05 +0000)]
Add experimental support for SATA Port Multipliers

Support is working on the Silicon Image SiI3124/3132.
Support is working on some AHCI chips but far from all.

Remember this is WIP, so test reports and (constructive) suggestions are welcome!

16 years agoIf we can't find or load the kernel NLM support, don't just go ahead and
dfr [Thu, 10 Apr 2008 12:54:53 +0000 (12:54 +0000)]
If we can't find or load the kernel NLM support, don't just go ahead and
try to use it anyway.

16 years agoFix printing of sockaddr prefixes in verbose mode.
ru [Thu, 10 Apr 2008 12:16:20 +0000 (12:16 +0000)]
Fix printing of sockaddr prefixes in verbose mode.

PR: bin/122403
Submitted by: az
MFC after: 3 days

16 years agoAdd the restrict qualifiers to the pointer arguments of the readlinkat.
kib [Thu, 10 Apr 2008 12:13:56 +0000 (12:13 +0000)]
Add the restrict qualifiers to the pointer arguments of the readlinkat.

16 years agoAdd forgotten -H, -h, and -P to usage().
ru [Thu, 10 Apr 2008 09:56:57 +0000 (09:56 +0000)]
Add forgotten -H, -h, and -P to usage().

16 years agoFix copy-n-paste typos in free text.
marcel [Thu, 10 Apr 2008 02:37:26 +0000 (02:37 +0000)]
Fix copy-n-paste typos in free text.

16 years agoDeclare _ppp_profile_cleaned, _punct, and _punct_c local in
brooks [Thu, 10 Apr 2008 01:32:49 +0000 (01:32 +0000)]
Declare _ppp_profile_cleaned, _punct, and _punct_c local in
ppp_start_profile().

Reported by: yar
MFC after: 1 week

16 years agoAdd support for MCP73 chips.
yongari [Thu, 10 Apr 2008 01:25:09 +0000 (01:25 +0000)]
Add support for MCP73 chips.

Tested by: "Daan Vreeken [PA4DAN]" ( Danovitsch AT vitsch DOT net )
MFC after: 1 week

16 years agoIt seems that RealTek 8129/8139 chip reports invalid length of
yongari [Thu, 10 Apr 2008 01:06:05 +0000 (01:06 +0000)]
It seems that RealTek 8129/8139 chip reports invalid length of
received frame under certain conditions. wpaul said the length
0xfff0 is special meaning that indicates hardware is in the
process of copying a packet into host memory. But it seems
there are other cases that hardware is busy or stuck in bad
situation even if the received frame length is not 0xfff0.
To work-around this condition, add a check that verifys that
recevied frame length is in valid range. If received length is out
of range reinitialize hardware to recover from stuck condition.

Reported by: Mike Tancsa ( mike AT sentex DOT net )
Tested by: Mike Tancsa
Obtained from: OpenBSD
MFC after: 1 week

16 years agoAdd memrchr(3).
delphij [Thu, 10 Apr 2008 00:12:44 +0000 (00:12 +0000)]
Add memrchr(3).

Obtained from: OpenBSD

16 years agoAdd support for Verizon v740 to ubsa(4).
flz [Wed, 9 Apr 2008 22:20:28 +0000 (22:20 +0000)]
Add support for Verizon v740 to ubsa(4).

PR: usb/122610
Submitted by: Randi Harper <randi@freebsdgirl.com>
MFC after: 3 days

16 years agoUpdate .Dd
emax [Wed, 9 Apr 2008 21:24:11 +0000 (21:24 +0000)]
Update .Dd

Pointed out by: Niclas Zeising <niclas-dot-zeising-at-gmail.com>
MFC after: 1 week

16 years ago- Add support for IPI_PREEMPT. [1]
marius [Wed, 9 Apr 2008 21:14:01 +0000 (21:14 +0000)]
- Add support for IPI_PREEMPT. [1]
- Add my copyright to mp_machdep.c for having implemented support for
  USIII and up and some fixes.

Obtained from: sun4v (modulo style(9) bugs) [1]

16 years agoFix spelling mistake in comment.
kan [Wed, 9 Apr 2008 20:27:53 +0000 (20:27 +0000)]
Fix spelling mistake in comment.

16 years ago- Use LK_TYPE_MASK where needed. Actually after sys/sys/lockmgr.h:1.69 it is
pjd [Wed, 9 Apr 2008 20:19:55 +0000 (20:19 +0000)]
- Use LK_TYPE_MASK where needed. Actually after sys/sys/lockmgr.h:1.69 it is
  no longer needed, but for now we still want to be consistent with other
  similar checks in the tree.
- Call ASSERT_VOP_ELOCKED() only when vget() returns 0.

Reviewed by: jeff

16 years agoDo image loading in a context known to have a root directory:
sam [Wed, 9 Apr 2008 19:07:48 +0000 (19:07 +0000)]
Do image loading in a context known to have a root directory:
o create a private task queue thread that sets up root and current
  directories (hooking mountroot event as needed); this is necessary
  because task queue threads are parented from proc0 and it does not
  have a reference to rootvnode (lost when / mounting moved to init)
o bounce image load + unload requests through the private task q so
  we can load images even when the request is made from a thread that
  does not have sufficient context (e.g. task q thread)
o add a check in the task q thread to fail requests before root is
  mounted (just in case)

Reviewed by: jhb, mlaier, luigi (glance)
MFC after: 1 month

16 years agoAdd memory barriers to the node locking operations.
mav [Wed, 9 Apr 2008 19:03:19 +0000 (19:03 +0000)]
Add memory barriers to the node locking operations.
Add some comments.

16 years agofix locking botch
sam [Wed, 9 Apr 2008 18:40:10 +0000 (18:40 +0000)]
fix locking botch

MFC after: 1 week

16 years agoRevert the previous change and let PROBE_KEYBOARD function identical to -P
jhb [Wed, 9 Apr 2008 17:59:17 +0000 (17:59 +0000)]
Revert the previous change and let PROBE_KEYBOARD function identical to -P
in boot2/gptboot.

16 years agoRemove using magic value of -1 to distinguish between linux_open()
rdivacky [Wed, 9 Apr 2008 16:42:50 +0000 (16:42 +0000)]
Remove using magic value of -1 to distinguish between linux_open()
and linux_openat(). Instead just pass AT_FDCWD into linux_common_open()
for the linux_open() case. This prevents passing -1 as a dirfd to
openat() from succeeding which is wrong.

Suggested by: rwatson, kib
Approved by: kib (mentor)

16 years agoFix a problem which stopped this from starting up on a kernel compiled
dfr [Wed, 9 Apr 2008 15:43:19 +0000 (15:43 +0000)]
Fix a problem which stopped this from starting up on a kernel compiled
without the INET6 option.

16 years agoAdd a distfile target to generate a distfile to be used by the
flz [Wed, 9 Apr 2008 15:08:31 +0000 (15:08 +0000)]
Add a distfile target to generate a distfile to be used by the
ports-mgmt/pkg_install port.

16 years agoInclude <sys/types.h> before <sys/systm.h> to get typedefs required
grehan [Wed, 9 Apr 2008 08:50:37 +0000 (08:50 +0000)]
Include <sys/types.h> before <sys/systm.h> to get typedefs required
by new atomic.h. Fixes tinderbox LINT build.

16 years agoTake the route mtu into account, if available, when sending an
bz [Wed, 9 Apr 2008 05:17:18 +0000 (05:17 +0000)]
Take the route mtu into account, if available, when sending an
ICMP unreach, frag needed.  Up to now we only looked at the
interface MTU. Make sure to only use the minimum of the two.

In case IPSEC is compiled in, loop the mtu through ip_ipsec_mtu()
to avoid any further conditional maths.

Without this, PMTU was broken in those cases when there was a
route with a lower MTU than the MTU of the outgoing interface.

PR: kern/122338
Tested by: Mark Cammidge  mark peralex.com
Reviewed by: silence on net@
MFC after: 2 weeks

16 years agoUnbreak after removal of SI_SUB_MOUNT_ROOT.
marcel [Wed, 9 Apr 2008 03:32:48 +0000 (03:32 +0000)]
Unbreak after removal of SI_SUB_MOUNT_ROOT.

16 years agoReimplement atomic_add, atomic_clear, atomic_set and atomic_subtract
marcel [Wed, 9 Apr 2008 01:00:35 +0000 (01:00 +0000)]
Reimplement atomic_add, atomic_clear, atomic_set and atomic_subtract
so that all implemented variants have proper prototypes. The 8-bit,
16-bit and 64-bit variants are not implemented.

This really fixes the current build breakages caused by type casting
and struct aliasing rules.

16 years agoSet defaults for the rfcomm_pppd_server rc script
emax [Tue, 8 Apr 2008 23:50:03 +0000 (23:50 +0000)]
Set defaults for the rfcomm_pppd_server rc script

MFC after: 1 week

16 years agoAdd rfcomm_pppd_server rc script to allow start rfcomm_pppd(8) in server
emax [Tue, 8 Apr 2008 23:34:12 +0000 (23:34 +0000)]
Add rfcomm_pppd_server rc script to allow start rfcomm_pppd(8) in server
mode at boot time. Multiple profiles can be started at the same time.
The whole idea is very similar to the ppp rc script.

Document Bluetooth knobs in rc.conf(5)

MFC after: 1 week

16 years agoo add rc.conf knobs to set the wpa_supplicant program, logging flags,
sam [Tue, 8 Apr 2008 23:12:15 +0000 (23:12 +0000)]
o add rc.conf knobs to set the wpa_supplicant program, logging flags,
  and config file
o change default logging options from -q to -s (log to syslog); this
  is currently broken for boot-time startup as syslogd is started too
  late but that'll be dealt with separately

MFC after: 2 weeks

16 years agoadd support wired interfaces
sam [Tue, 8 Apr 2008 23:00:04 +0000 (23:00 +0000)]
add support wired interfaces

MFC after: 2 weeks

16 years agoRemove ftp.hk.super.net, the DNS isn't pointing to anything at the moment.
remko [Tue, 8 Apr 2008 19:43:00 +0000 (19:43 +0000)]
Remove ftp.hk.super.net, the DNS isn't pointing to anything at the moment.

I tested this as well as the submitter and couldn't resolve this either,
since I dont want to "announce" dead mirrors, I'll remove it from the
list.

PR: 122567
Submitted by: vs
Approved by: imp (mentor, implicit for trivial changes)
MFC after: 1 week

16 years agoClean up and fix style(9) nits.
jkim [Tue, 8 Apr 2008 19:09:45 +0000 (19:09 +0000)]
Clean up and fix style(9) nits.

16 years agospell pidfile correctly so multiple wpa_supplicant processes can be run
sam [Tue, 8 Apr 2008 18:54:42 +0000 (18:54 +0000)]
spell pidfile correctly so multiple wpa_supplicant processes can be run

MFC after: 1 week

16 years ago- Add write(2) support for psm(4) in native operation level. Now arbitrary
jkim [Tue, 8 Apr 2008 17:55:26 +0000 (17:55 +0000)]
- Add write(2) support for psm(4) in native operation level.  Now arbitrary
commands can be written to /dev/psm%d and status can be read back from it.
- Reflect the change in psm(4) and bump version for ports.

MFC after: 1 week

16 years agoo add a mountroot event handler that fires when / is mounted; this information
sam [Tue, 8 Apr 2008 17:53:33 +0000 (17:53 +0000)]
o add a mountroot event handler that fires when / is mounted; this information
  was lost when root started being mounted by init
o remove SI_SUB_MOUNT_ROOT since it's no longer meaningful

MFC after: 2 weeks

16 years agochange taskqueue_start_threads to create threads instead of proc's
sam [Tue, 8 Apr 2008 17:48:02 +0000 (17:48 +0000)]
change taskqueue_start_threads to create threads instead of proc's

Reviewed by: jhb

16 years agoQuick fix for the kernel build breakage in netgraph and the
marcel [Tue, 8 Apr 2008 16:34:50 +0000 (16:34 +0000)]
Quick fix for the kernel build breakage in netgraph and the
aliasing warning in libthr. A more elaborate fix is in the
works that makes sure that all variants have proper inline
functions with proper types.