]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoPreliminary support for Xircom Credit Card Adapter. Not to be
imp [Tue, 25 Oct 2005 03:49:24 +0000 (03:49 +0000)]
Preliminary support for Xircom Credit Card Adapter.  Not to be
confused with the Credit Card Adapter II and its spawn (which the xe
driver supports).  These changes get my card probing and attaching.  I
recently won one of these (and a NEC rebadged version) in an lot
auction.  The NEC didn't work, so I took it apart and found the
MB86960A chip and then modified if_fe_pccard.c to attach.  I can't
test this card further since I have no dongle for this card.

18 years agoPut pthread_condattr_init sorted order.
davidxu [Tue, 25 Oct 2005 00:09:58 +0000 (00:09 +0000)]
Put pthread_condattr_init sorted order.

18 years agoBack out most of 1.84. It was unwise to force debug kernels to always
imp [Mon, 24 Oct 2005 23:54:39 +0000 (23:54 +0000)]
Back out most of 1.84.  It was unwise to force debug kernels to always
be installed.  It should have been optional to install a non-debug
one, just like it was formerly optional to install a debug one.  In
order to do that, most of 1.84 had to go.

Instead, make installing the debug kernel the default, but create a
new option INSTALL_NODEBUG for those people that have small /
partitions and good source control habits.

This preserves the behavior of 1.84 while allowing it to be overriden
for people (like me) that do not have the time to upgrade to get a
bigger / and also don't have time for stupid makefile tricks when
upgrading their older system, but still want a kernel.debug around if
things go south.

18 years agoUse the "builtin" shell function to make sure that the requested
cperciva [Mon, 24 Oct 2005 22:32:19 +0000 (22:32 +0000)]
Use the "builtin" shell function to make sure that the requested
command is handled as a shell function.  This avoids the following
peculiar behaviour when /usr/bin is on a case-insensitive filesystem:
    # READ foo
(... long pause, depending upon the amount of swap space available ...)
    sh: Resource temporarily unavailable.

Reported by: I can't remember; someone on IRC.
MFC after: 1 week

18 years agoRename the KDB_STOP_NMI kernel option to STOP_NMI and make it apply to all
jhb [Mon, 24 Oct 2005 21:04:19 +0000 (21:04 +0000)]
Rename the KDB_STOP_NMI kernel option to STOP_NMI and make it apply to all
IPI_STOP IPIs.
- Change the i386 and amd64 MD IPI code to send an NMI if STOP_NMI is
  enabled if an attempt is made to send an IPI_STOP IPI.  If the kernel
  option is enabled, there is also a sysctl to change the behavior at
  runtime (debug.stop_cpus_with_nmi which defaults to enabled).  This
  includes removing stop_cpus_nmi() and making ipi_nmi_selected() a
  private function for i386 and amd64.
- Fix ipi_all(), ipi_all_but_self(), and ipi_self() on i386 and amd64 to
  properly handle bitmapped IPIs as well as IPI_STOP IPIs when STOP_NMI is
  enabled.
- Fix ipi_nmi_handler() to execute the restart function on the first CPU
  that is restarted making use of atomic_readandclear() rather than
  assuming that the BSP is always included in the set of restarted CPUs.
  Also, the NMI handler didn't clear the function pointer meaning that
  subsequent stop and restarts could execute the function again.
- Define a new macro HAVE_STOPPEDPCBS on i386 and amd64 to control the use
  of stoppedpcbs[] and always enable it for i386 and amd64 instead of
  being dependent on KDB_STOP_NMI.  It works fine in both the NMI and
  non-NMI cases.

18 years agoWhen restarting the BSP during cpu_reset() use a membar to ensure that
jhb [Mon, 24 Oct 2005 20:53:52 +0000 (20:53 +0000)]
When restarting the BSP during cpu_reset() use a membar to ensure that
the updated cpustop_restartfunc is seen when the BSP resumes execution.
This matches the membar already present in restart_cpus().

18 years agoUse xchg in Xcpustop to close a race and make cpustop_restartfunc truly
jhb [Mon, 24 Oct 2005 20:52:26 +0000 (20:52 +0000)]
Use xchg in Xcpustop to close a race and make cpustop_restartfunc truly
one-shot in the SMP case (before using the simple mov / cmp / mov sequence
could allow multiple CPUs to execute the restart function on resume).

18 years ago- Various small whitespace and style nits.
jhb [Mon, 24 Oct 2005 20:31:04 +0000 (20:31 +0000)]
- Various small whitespace and style nits.
- Use PCPU_GET(cpumask) in preference to 1 << PCPU_GET(cpuid) in a few
  places.

18 years agoDocument in #ifdef notnow code the actions that proc_fini would need to
jhb [Mon, 24 Oct 2005 20:15:23 +0000 (20:15 +0000)]
Document in #ifdef notnow code the actions that proc_fini would need to
take if struct procs were actually freed.

18 years agoDon't panic if a spin lock is initialized that isn't in our static order
jhb [Mon, 24 Oct 2005 20:14:24 +0000 (20:14 +0000)]
Don't panic if a spin lock is initialized that isn't in our static order
list.  Just warn about it instead.

Requested by: scottl
MFC after: 1 day

18 years agoCreate the devd thread after we have called daemon(). Otherwise, it would
njl [Mon, 24 Oct 2005 18:34:54 +0000 (18:34 +0000)]
Create the devd thread after we have called daemon().  Otherwise, it would
be killed when the parent exits.

Submitted by: Rudolf Cejka <cejkar / fit.vutbr.cz>

18 years agoInvert the check logic. No functional change, but I prefer this version.
njl [Mon, 24 Oct 2005 18:30:57 +0000 (18:30 +0000)]
Invert the check logic.  No functional change, but I prefer this version.

18 years agoDo not allow ndis_ticktask() to run after ndis_halt_nic() has been called.
wpaul [Mon, 24 Oct 2005 17:47:47 +0000 (17:47 +0000)]
Do not allow ndis_ticktask() to run after ndis_halt_nic() has been called.
(It may have been queued up in one of the workitem threads, and letting
it fire after the NIC has been halted will cause a crash in some cases.)

18 years agoRevert previous change to this file. I accidentally committed while
jhb [Mon, 24 Oct 2005 15:58:21 +0000 (15:58 +0000)]
Revert previous change to this file.  I accidentally committed while
fixing spelling in a comment.

18 years agoSpell hierarchy correctly in comments.
jhb [Mon, 24 Oct 2005 15:57:27 +0000 (15:57 +0000)]
Spell hierarchy correctly in comments.

Submitted by: Wojciech A. Koszek dunstan at freebsd dot czest dot pl

18 years ago- Rename 'traceall' to 'alltrace' so that the 'tr' shortcut for 'trace'
jhb [Mon, 24 Oct 2005 15:21:36 +0000 (15:21 +0000)]
- Rename 'traceall' to 'alltrace' so that the 'tr' shortcut for 'trace'
  still works.  Also, this is consistent with 'show pcpu' vs
  'show allpcpu'.  (And 'show allstacks' on OS X for that matter.)
- Add 'bt' as an alias for 'trace'.  We already have a 'where' alias as
  well, so this makes it easier for gdb-wired hands to work in ddb.

Ok'd by: rwatson (1)
Requested by: scottl (2)
MFC after: 1 day

18 years agoOnly set B_RAM (Read ahead mark) on an incore buffers if we can lock it.
ups [Mon, 24 Oct 2005 14:23:04 +0000 (14:23 +0000)]
Only set B_RAM (Read ahead mark) on an incore buffers if we can lock it.
This fixes a race condition caused by the unlocked write access to the
b_flags field.

MFC after: 3 days

18 years agoMoved the optimization for tiny x from __kernel_{cos,sin}[f](x) to
bde [Mon, 24 Oct 2005 14:08:36 +0000 (14:08 +0000)]
Moved the optimization for tiny x from __kernel_{cos,sin}[f](x) to
{cos_sin}[f](x) so that x doesn't need to be reclassified in the
"kernel" functions to determine if it is tiny (it still needs to be
reclassified in the cosine case for other reasons that will go away).

This optimization is quite large for exponentially distributed x, since
x is tiny for almost half of the domain, but it is a pessimization for
uniformally distributed x since it takes a little time for all cases
but rarely applies.  Arg reduction on exponentially distributed x
rarely gives a tiny x unless the reduction is null, so it is best to
only do the optimization if the initial x is tiny, which is what this
commit arranges.  The imediate result is an average optimization of
1.4% relative to the previous version in a case that doesn't favour
the optimization (double cos(x) on all float x) and a large
pessimization for the relatively unimportant cases of lgamma[f][_r](x)
on tiny, negative, exponentially distributed x.  The optimization should
be recovered for lgamma*() as part of fixing lgamma*()'s low-quality
arg reduction.

Fixed various wrong constants for the cutoff for "tiny".  For cosine,
the cutoff is when x**2/2! == {FLT or DBL}_EPSILON/2.  We round down
to an integral power of 2 (and for cos() reduce the power by another
1) because the exact cutoff doesn't matter and would take more work
to determine.  For sine, the exact cutoff is larger due to the ration
of terms being x**2/3! instead of x**2/2!, but we use the same cutoff
as for cosine.  We now use a cutoff of 2**-27 for double precision and
2**-12 for single precision.  2**-27 was used in all cases but was
misspelled 2**27 in comments.  Wrong and sloppy cutoffs just cause
missed optimizations (provided the rounding mode is to nearest --
other modes just aren't supported).

18 years agoo Grammar.
maxim [Mon, 24 Oct 2005 08:53:21 +0000 (08:53 +0000)]
o Grammar.

Submitted by: Ulrich Spoerlein
MFC after: 1 week

18 years agoNote that bugmaster would like to review commits to send-pr.
ceri [Mon, 24 Oct 2005 08:15:58 +0000 (08:15 +0000)]
Note that bugmaster would like to review commits to send-pr.

18 years agoAdd prototype for following functions, plus tab fixes.
davidxu [Mon, 24 Oct 2005 05:53:54 +0000 (05:53 +0000)]
Add prototype for following functions, plus tab fixes.
pthread_condattr_getpshared
pthread_condattr_setpshared
pthread_mutexattr_getpshared
pthread_mutexattr_setpshared

18 years agoInclude files thr_condattr_pshared.c and thr_mattr_pshare.c.
davidxu [Mon, 24 Oct 2005 05:48:32 +0000 (05:48 +0000)]
Include files thr_condattr_pshared.c and thr_mattr_pshare.c.

18 years agoExport following functions:
davidxu [Mon, 24 Oct 2005 05:37:21 +0000 (05:37 +0000)]
Export following functions:
_pthread_condattr_getpshared
_pthread_condattr_setpshared
_pthread_mutexattr_getpshared
_pthread_mutexattr_setpshared
pthread_condattr_getpshared
pthread_condattr_setpshared
pthread_mutexattr_getpshared
pthread_mutexattr_setpshared

18 years agoAdd functions pthread_mutexattr_setpshared and pthread_mutexattr_getpshared.
davidxu [Mon, 24 Oct 2005 05:35:40 +0000 (05:35 +0000)]
Add functions pthread_mutexattr_setpshared and pthread_mutexattr_getpshared.

18 years agoAdd function pthread_condattr_setpshared and pthread_condattr_getpshared.
davidxu [Mon, 24 Oct 2005 05:35:14 +0000 (05:35 +0000)]
Add function pthread_condattr_setpshared and pthread_condattr_getpshared.

18 years agoIf no AC line devices are found, go ahead and notify devd that the system
njl [Mon, 24 Oct 2005 05:34:21 +0000 (05:34 +0000)]
If no AC line devices are found, go ahead and notify devd that the system
is on AC power (i.e. not a laptop).  This allows power_profile to run once
for desktop systems as well, for instance, to set C3 or CPU frequency.

MFC after: 2 weeks

18 years agoExport following functions:
davidxu [Mon, 24 Oct 2005 05:20:04 +0000 (05:20 +0000)]
Export following functions:
_pthread_mutexattr_getpshared
_pthread_mutexattr_setpshared
pthread_condattr_getpshared
pthread_condattr_setpshared
pthread_mutexattr_getpshared
pthread_mutexattr_setpshared

18 years agoAdd functions pthread_mutexattr_setpshared and pthread_mutexattr_getpshared.
davidxu [Mon, 24 Oct 2005 05:16:41 +0000 (05:16 +0000)]
Add functions pthread_mutexattr_setpshared and pthread_mutexattr_getpshared.

18 years agoFix handling of message table messages that got broken when I
wpaul [Mon, 24 Oct 2005 05:05:09 +0000 (05:05 +0000)]
Fix handling of message table messages that got broken when I
converted NdisWriteErrorLogEntry() to use the RtlXXX unicode/ansi
conversion routines.

18 years agoWe no longer need INCLUDES+= -I$S/contrib/dev/acpica.
obrien [Mon, 24 Oct 2005 04:40:37 +0000 (04:40 +0000)]
We no longer need INCLUDES+= -I$S/contrib/dev/acpica.

18 years agoAdd a commented out version of what was done for the r20041119sysinc import.
obrien [Mon, 24 Oct 2005 04:36:14 +0000 (04:36 +0000)]
Add a commented out version of what was done for the r20041119sysinc import.

18 years agoFix conflicts of import of Intel ACPI-CA 20041119 with system includes fixups.
obrien [Mon, 24 Oct 2005 04:35:20 +0000 (04:35 +0000)]
Fix conflicts of import of Intel ACPI-CA 20041119 with system includes fixups.

18 years agoThis commit was generated by cvs2svn to compensate for changes in r151600,
obrien [Mon, 24 Oct 2005 04:31:06 +0000 (04:31 +0000)]
This commit was generated by cvs2svn to compensate for changes in r151600,
which included commits to RCS files with non-trunk default branches.

18 years agoVendor import of Intel ACPI-CA 20041119 with system includes fixups.
obrien [Mon, 24 Oct 2005 04:31:06 +0000 (04:31 +0000)]
Vendor import of Intel ACPI-CA 20041119 with system includes fixups.

18 years agoAdd vim syntax highlighting support to the send-pr(1) utility.
obrien [Mon, 24 Oct 2005 01:36:16 +0000 (01:36 +0000)]
Add vim syntax highlighting support to the send-pr(1) utility.

PR: 35333
Submitted by: Hendrik Scholz <hendrik@scholz.net>

18 years agoinclude opt_compat.h to unbreak the build
ps [Mon, 24 Oct 2005 00:00:00 +0000 (00:00 +0000)]
include opt_compat.h to unbreak the build

18 years agoAdd a 'clean' target.
obrien [Sun, 23 Oct 2005 23:58:23 +0000 (23:58 +0000)]
Add a 'clean' target.

18 years agoUnbreak for !__XSCALE__.
cognet [Sun, 23 Oct 2005 23:09:14 +0000 (23:09 +0000)]
Unbreak for !__XSCALE__.

18 years agoDon't touch last overrun if signal was already on queue.
davidxu [Sun, 23 Oct 2005 22:59:33 +0000 (22:59 +0000)]
Don't touch last overrun if signal was already on queue.

18 years agoIf we have been called from ether_ifdetach() then do not try and clear the
thompsa [Sun, 23 Oct 2005 22:30:07 +0000 (22:30 +0000)]
If we have been called from ether_ifdetach() then do not try and clear the
promisc flag from the member interface, this is a no-op anyway since the
interface is disappearing. The driver may have already released
its resources such as miibus and this is likely to panic the kernel.

Submitted and tested by: Wojciech A. Koszek
MFC after: 2 weeks

18 years agoLike acpi_throttle, set frequency to 100% in attach. Some BIOSen may set
njl [Sun, 23 Oct 2005 19:38:06 +0000 (19:38 +0000)]
Like acpi_throttle, set frequency to 100% in attach.  Some BIOSen may set
this value lower, making the system quite slow after booting.

18 years agos/RADUIS/RADIUS/
maxim [Sun, 23 Oct 2005 19:37:55 +0000 (19:37 +0000)]
s/RADUIS/RADIUS/

Submitted by: jisakiel@yahoo.es
MFC after: 1 week

18 years agoAdd prototype to be consistent.
njl [Sun, 23 Oct 2005 19:31:18 +0000 (19:31 +0000)]
Add prototype to be consistent.

18 years agoMNT_JAILDEVFS is not used anymore. Mark it as spare.
pjd [Sun, 23 Oct 2005 16:52:13 +0000 (16:52 +0000)]
MNT_JAILDEVFS is not used anymore. Mark it as spare.

OK'ed by: phk

18 years agoRestore the documentation about uid, gid or prison based rules requiring
csjp [Sun, 23 Oct 2005 16:15:02 +0000 (16:15 +0000)]
Restore the documentation about uid, gid or prison based rules requiring
that debug.mpsafenet be set to 0. It is still possible for dead locks to
occur while these filtering options are used due to the layering violation
inherent in their implementation.

Discussed: -current, rwatson, glebius

18 years agoDon't be lazy, set the "command" variable even if
yar [Sun, 23 Oct 2005 14:06:53 +0000 (14:06 +0000)]
Don't be lazy, set the "command" variable even if
/etc/defaults/rc.conf will provide foo_program, too.
By specifying "command" we explicitly say that we're
going to rely on rc.subr(8) default methods, and
rc.subr(8) will take advantage of this soon.

The majority of our rc.d scripts already set "command"
if appropriate, so fix just the non-compliant handful.

18 years agoMake p_itimers as a pointer, so file sys/proc.h does not need to include
davidxu [Sun, 23 Oct 2005 12:19:08 +0000 (12:19 +0000)]
Make p_itimers as a pointer, so file sys/proc.h does not need to include
sys/timers.h.

18 years agoDocument [:blank:].
stefanf [Sun, 23 Oct 2005 11:19:56 +0000 (11:19 +0000)]
Document [:blank:].

18 years agoregen
ps [Sun, 23 Oct 2005 10:43:39 +0000 (10:43 +0000)]
regen

18 years agoImplement for FreeBSD 3 32 binaries:
ps [Sun, 23 Oct 2005 10:43:14 +0000 (10:43 +0000)]
Implement for FreeBSD 3 32 binaries:
sigaction, sigprocmask, sigpending, sigvec, sigblock, sigsetmask,
sigsuspend, sigstack

18 years agoMake kernel structures invisible to userland.
davidxu [Sun, 23 Oct 2005 10:11:46 +0000 (10:11 +0000)]
Make kernel structures invisible to userland.

18 years agoTypo.
glebius [Sun, 23 Oct 2005 09:05:51 +0000 (09:05 +0000)]
Typo.

PR: misc/87679
Submitted by: Alan Amesbury <amesbury umn.edu>

18 years agoPreviously, nothing prevented the page that was returned by pmap_extract()
alc [Sun, 23 Oct 2005 07:41:56 +0000 (07:41 +0000)]
Previously, nothing prevented the page that was returned by pmap_extract()
from being reclaimed before it was wired.  Use pmap_extract_and_hold()
instead of pmap_extract() and retain the hold on the page until it has been
wired.

18 years agoChange _POSIX_REALTIME_SIGNALS to 200112L to indicates we support
davidxu [Sun, 23 Oct 2005 05:36:41 +0000 (05:36 +0000)]
Change _POSIX_REALTIME_SIGNALS to 200112L to indicates we support
realtime signals. Define missing _POSIX_TIMERS, it will be changed to
200112L once struct sigevent is fixed.

18 years agoRegen for POSIX timer syscalls.
davidxu [Sun, 23 Oct 2005 04:26:10 +0000 (04:26 +0000)]
Regen for POSIX timer syscalls.

18 years agoImplement POSIX timers. Current only CLOCK_REALTIME and CLOCK_MONOTONIC
davidxu [Sun, 23 Oct 2005 04:22:56 +0000 (04:22 +0000)]
Implement POSIX timers. Current only CLOCK_REALTIME and CLOCK_MONOTONIC
clock are supported. I have plan to merge XSI timer ITIMER_REAL and other
two CPU timers into the new code, current three slots are available for
the XSI timers.
The SIGEV_THREAD notification type is not supported yet because our
sigevent struct lacks of two member fields:
sigev_notify_function
sigev_notify_attributes
I have found the sigevent is used in AIO, so I won't add the two members
unless the AIO code is adjusted.

18 years ago1. Make ksiginfo_alloc and ksiginfo_free public.
davidxu [Sun, 23 Oct 2005 04:12:26 +0000 (04:12 +0000)]
1. Make ksiginfo_alloc and ksiginfo_free public.
2. Introduce flags KSI_EXT and KSI_INS. The flag KSI_EXT allows a ksiginfo
   to be managed by outside code, the KSI_INS indicates sigqueue_add should
   directly insert passed ksiginfo into queue other than copy it.

18 years agoAdd member fields for POSIX timer.
davidxu [Sun, 23 Oct 2005 03:59:52 +0000 (03:59 +0000)]
Add member fields for POSIX timer.

18 years agoBefore we export network interface data through the ifmibdata structure,
csjp [Sun, 23 Oct 2005 01:44:08 +0000 (01:44 +0000)]
Before we export network interface data through the ifmibdata structure,
OR the flags bits with the driver managed status flags. This fixes an
issue where RUNNING flags would not be reported to processes, which
conflicts with the flags information provided by ifconfig(8).

18 years agoDocument incorrect handling of multibyte characters.
tjr [Sun, 23 Oct 2005 01:37:25 +0000 (01:37 +0000)]
Document incorrect handling of multibyte characters.

PR: 87724

18 years agoWhitespace fix.
njl [Sun, 23 Oct 2005 00:55:56 +0000 (00:55 +0000)]
Whitespace fix.

18 years agoAdd a hack to get around PCI link devices that report "present" but not
njl [Sun, 23 Oct 2005 00:28:39 +0000 (00:28 +0000)]
Add a hack to get around PCI link devices that report "present" but not
"functional" (i.e., if they are disabled).  We should probe them anyway
since we may enable them later.

Tested by: thompsa
MFC after: 3 days

18 years agoHook acpi_smbat up to the build.
njl [Sun, 23 Oct 2005 00:22:02 +0000 (00:22 +0000)]
Hook acpi_smbat up to the build.

18 years agoImport ACPI smart battery support. Newer systems (Acer, mostly) do not
njl [Sun, 23 Oct 2005 00:20:13 +0000 (00:20 +0000)]
Import ACPI smart battery support.  Newer systems (Acer, mostly) do not
support the CM-battery interface.  Smart batteries can eventually be
supported without ACPI via a separate SMBus interface.  The ACPI interface
uses the embedded controller for reading/writing to the SMBus, and normal
ASL definitions for locating the battery controller (since SMBus can't be
enumerated.)  Also import definitions for the smart battery interface.

This was written by Hans Petter Selasky with minor cleanups from myself.

Submitted by: Hans Petter Selasky <hselasky / c2i.net>

18 years agoCleanups and support code for importing smart battery support.
njl [Sun, 23 Oct 2005 00:16:41 +0000 (00:16 +0000)]
Cleanups and support code for importing smart battery support.

* Use ACPI_BATT_UNKNOWN instead of constants
* Use maxunit instead of a count of devices since we may have sparse
  battery devices in the future.  Only userland should be using unit
  numbers anyway, so provide a translation function.  (Kernel use of
  batteries should be restricted to looking up a device_t and calling
  methods directly.
* Don't check acpi_BatteryIsPresent() in acpi_battery.  Leave it up to
  the hardware-specific driver (i.e. cmbat) since smart batteries seem
  to not report the "battery present" flag.
* Convert mA to mW if the battery uses those units.  CM-batteries only
  used mW so this deficiency went unnoticed.
* Clean strings reported in the battery info from any control chars.
* Only dereference the unit from ioctl_arg if the full struct is present.
  Unit wouldn't have been used later if it wasn't present but this is
  cleaner.  Translate the unit if it's not ACPI_BATTERY_ALL_UNITS.
* bzero structs before returning them to usermode for future compat.

Most of this work was submitted by Hans Petter Selasky and then majorly
reworked by myself.

Submitted by: Hans Petter Selasky <hselasky / c2i.net>

18 years agoFix typo from last commit.
glebius [Sat, 22 Oct 2005 20:47:38 +0000 (20:47 +0000)]
Fix typo from last commit.

Submitted by: simon

18 years agoVerify that access to the given address is allowed from user-space.
alc [Sat, 22 Oct 2005 20:02:59 +0000 (20:02 +0000)]
Verify that access to the given address is allowed from user-space.

Discussed with: rwatson@

18 years agoUse of the ZERO_COPY_SOCKETS options can result in an unusual state that
alc [Sat, 22 Oct 2005 18:46:38 +0000 (18:46 +0000)]
Use of the ZERO_COPY_SOCKETS options can result in an unusual state that
vm_object_backing_scan() was not written to handle.  Specifically, a wired
page within a backing object that is shadowed by a page within the shadow
object.  Handle this state by removing the wired page from the backing
object.  The wired page will be freed by socow_iodone().

Stop masking errors: If a page is being freed by vm_object_backing_scan(),
assert that it is no longer mapped rather than quietly destroying any
mappings.

Tested by: Harald Schmalzbauer

18 years agoFix obvious copy'n'paste-O in rev.1.36 While here nit style.
mlaier [Sat, 22 Oct 2005 17:50:45 +0000 (17:50 +0000)]
Fix obvious copy'n'paste-O in rev.1.36  While here nit style.

PR: bin/87783
Submitted by: Mats Palmgren
MFC after: 1 week

18 years agoFix build after in6_joingroup change. It remains unclear if DAD breaks CARP
mlaier [Sat, 22 Oct 2005 14:54:02 +0000 (14:54 +0000)]
Fix build after in6_joingroup change.  It remains unclear if DAD breaks CARP
or not.

18 years agoIn in_addprefix() compare not only route addresses, but their masks,
glebius [Sat, 22 Oct 2005 14:50:27 +0000 (14:50 +0000)]
In in_addprefix() compare not only route addresses, but their masks,
too. This fixes problem when connected prefixes overlap.

Obtained from: OpenBSD (rev. 1.40 by claudio);
[ I came to this fix myself, and then found out that
  OpenBSD had already fixed it the same way.]

18 years agoAdd bge(4) to the list of supported devices.
glebius [Sat, 22 Oct 2005 14:38:52 +0000 (14:38 +0000)]
Add bge(4) to the list of supported devices.

18 years agoIntroduce polling(4) capability for bge(4).
glebius [Sat, 22 Oct 2005 14:31:01 +0000 (14:31 +0000)]
Introduce polling(4) capability for bge(4).

Submitted by: Oleg Bulyzhin <oleg rinet.ru>

18 years agoA better, more correct explination of NO_SHARED. This could probably
trhodes [Sat, 22 Oct 2005 10:24:53 +0000 (10:24 +0000)]
A better, more correct explination of NO_SHARED.  This could probably
use a quick touch up, but at least it's correct.

Requested by: ru

18 years agoAdd missing softc
imp [Sat, 22 Oct 2005 07:43:20 +0000 (07:43 +0000)]
Add missing softc

18 years agoTurns out that we're asking for the MAC address at a bad time in the
imp [Sat, 22 Oct 2005 07:26:05 +0000 (07:26 +0000)]
Turns out that we're asking for the MAC address at a bad time in the
attach routine.  Go ahead and ask for it in the probe routine and be
just as wrong as all the other cards that ask for it there...

# this gets the RTL8019 on a SBC at work fully functional.  6.0 still treats
# the 8019 as a generic NE-2000, so these changes aren't relevant there.

18 years agoMake the multiple DPC threads an option, and create only one by default.
wpaul [Sat, 22 Oct 2005 05:15:20 +0000 (05:15 +0000)]
Make the multiple DPC threads an option, and create only one by default.
This avoids the need for sched_bind() in the default case so that you
can start up the NDIS subsystem at boot time when only CPU 0 is running.

There are potentially ways to fix it so that the DPC threads aren't
started until after the other CPUs are launched, but doing it correctly
is tricky. You need to defer the startup of the ntoskrnl subsystem
(ntoskrnl_libinit()), not just defer ndis_attach().

For now, I don't think it will make much difference having just the
single DPC thread (I started out with just one anyway). Note that this
turns the KeSetTargetProcessorDpc() routine into a no-op, since the
CPU number in struct kdpc is now ignored.

18 years agoIn the ISA case, we call ed_probe_WD80x3 before we call
imp [Sat, 22 Oct 2005 05:14:18 +0000 (05:14 +0000)]
In the ISA case, we call ed_probe_WD80x3 before we call
ed_probe_rtl80x9.  In the pci case we call ed_probe_rtl80x9 first.  In
the PCI case we were using the correct nic_offset by accident because
softc is initialized to zero.  In the isa case we were using the wrong
value by accident, since ed_probe_WD80x3 sets the offset value to
0x10.  This lead to the identification routines failing.  Fix this
problem by always initalizing the nic_offset and asic_offset before
making ed_{asic,nic}_{in,out}* calls.

18 years agofixed a compilation failure on amd64/sparc64/ia64
suz [Sat, 22 Oct 2005 05:07:16 +0000 (05:07 +0000)]
fixed a compilation failure on amd64/sparc64/ia64

Submitted by: max
MFC after: 2 month

18 years agoReplace FreeBSD 3.x syntax (controller miibus0) with 4.x syntax
imp [Sat, 22 Oct 2005 05:06:55 +0000 (05:06 +0000)]
Replace FreeBSD 3.x syntax (controller miibus0) with 4.x syntax
(device miibus) in time for 7.0 :-)

18 years agoThe controller -> device switch happened in 4.0, fix comment
imp [Sat, 22 Oct 2005 05:01:11 +0000 (05:01 +0000)]
The controller -> device switch happened in 4.0, fix comment

18 years agoSpecifically panic() in the case where pmap_insert_entry() fails to
ade [Fri, 21 Oct 2005 19:42:43 +0000 (19:42 +0000)]
Specifically panic() in the case where pmap_insert_entry() fails to
get a new pv under high system load where the available pv entries
have been exhausted before the pagedaemon has a chance to wake up
to reclaim some.

Prior to this, the NULL pointer dereference ended up causing
secondary panics with rather less than useful resulting tracebacks.

Reviewed by: alc, jhb
MFC after: 1 week

18 years agoWarn the user if the kernel driver dropped samples or ran out of event buffers
jkoshy [Fri, 21 Oct 2005 18:59:59 +0000 (18:59 +0000)]
Warn the user if the kernel driver dropped samples or ran out of event buffers
during a data collection run.

18 years agoHide a diagnostic message under if (verbose) to avoid cluttering the
wpaul [Fri, 21 Oct 2005 16:57:57 +0000 (16:57 +0000)]
Hide a diagnostic message under if (verbose) to avoid cluttering the
system log when not in verbose logging mode.

18 years agonuked non-existing commands
suz [Fri, 21 Oct 2005 16:31:39 +0000 (16:31 +0000)]
nuked non-existing commands

18 years agosync with KAME regarding NDP
suz [Fri, 21 Oct 2005 16:23:01 +0000 (16:23 +0000)]
sync with KAME regarding NDP

- introduced fine-grain-timer to manage ND-caches and IPv6 Multicast-Listeners
- supports Router-Preference <draft-ietf-ipv6-router-selection-07.txt>
- better prefix lifetime management
- more spec-comformant DAD advertisement
- updated RFC/internet-draft revisions

Obtained from: KAME
Reviewed by: ume, gnn
MFC after: 2 month

18 years agoperform NUD on an IPv6-aware point-to-point interface
suz [Fri, 21 Oct 2005 15:59:00 +0000 (15:59 +0000)]
perform NUD on an IPv6-aware point-to-point interface

Obtained from: KAME
MFC after: 1 week

18 years agosync with KAME (renamed a macro IPV6_DADOUTPUT to IPV6_UNSPECSRC)
suz [Fri, 21 Oct 2005 15:45:13 +0000 (15:45 +0000)]
sync with KAME (renamed a macro IPV6_DADOUTPUT to IPV6_UNSPECSRC)

Obtained from: KAME

18 years agoRemove redundant string length check from the previous commit.
jkim [Fri, 21 Oct 2005 15:17:18 +0000 (15:17 +0000)]
Remove redundant string length check from the previous commit.

18 years agoApply the same fix to a potential race in the ISDOTDOT code
dumbbell [Fri, 21 Oct 2005 09:15:26 +0000 (09:15 +0000)]
Apply the same fix to a potential race in the ISDOTDOT code
in reiserfs_lookup() that was used to fix an actual race in
ufs_lookup.c:1.78. This is not currently a hazard, but the
bug would be activated by marking reiserfs as MPSAFE.

Reviewed by: mux (mentor)
MFC after: 2 weeks

18 years agoEliminate spl* calls.
alc [Fri, 21 Oct 2005 05:48:38 +0000 (05:48 +0000)]
Eliminate spl* calls.

18 years agoCorrect the macro definition for KeRaiseIrql(). The official API
wpaul [Fri, 21 Oct 2005 05:23:20 +0000 (05:23 +0000)]
Correct the macro definition for KeRaiseIrql(). The official API
is KeRaiseIrql(newirql, &oldirql), not oldirql = KeRaiseIrql(newirql).
(The macro ultimately translates to KfRaiseIrql() which does use
the latter API, so this has no effect on generated code.)

Also, wait for thread termination the right way: kthread_exit()
will ultimately do a wakeup(td->td_proc). This is the event we
should wait on. Eliminate the previous synchronization machinery
for this since it was never guaranteed to work correctly.

18 years agoAdjust maxfilesize for UFS1 and old 4.4 FFS. For UFS1, increase the limit
njl [Fri, 21 Oct 2005 01:54:00 +0000 (01:54 +0000)]
Adjust maxfilesize for UFS1 and old 4.4 FFS.  For UFS1, increase the limit
to (max block - 1) * bsize.  For DEV_BSIZE, this doubles the limit from
0.5 TB to 1 TB.  For the old 4.4 FFS case, decrease the limit from 0.5 TB
to 2 GB - 1.  Older systems had a 32 bit off_t so they couldn't access the
larger files anyway.

Collaboration with: bde

18 years agoFix a longstanding buglet in bz-prefixed grep(1).
jkim [Thu, 20 Oct 2005 21:50:44 +0000 (21:50 +0000)]
Fix a longstanding buglet in bz-prefixed grep(1).

18 years agoChange format string for u_int64_t to %ju from %llu, in order to use the
rwatson [Thu, 20 Oct 2005 21:28:31 +0000 (21:28 +0000)]
Change format string for u_int64_t to %ju from %llu, in order to use the
correct format string on 64-bit systems.

Pointed out by: pjd

18 years agoRemove duplicate entry.
ru [Thu, 20 Oct 2005 20:51:30 +0000 (20:51 +0000)]
Remove duplicate entry.

18 years agoCleanup.
cognet [Thu, 20 Oct 2005 20:30:51 +0000 (20:30 +0000)]
Cleanup.

18 years agoMention the possibility of non-interactive scripts for EDITOR and add a
brooks [Thu, 20 Oct 2005 19:23:16 +0000 (19:23 +0000)]
Mention the possibility of non-interactive scripts for EDITOR and add a
BUGS section mentioning the requirement that such scripts run "sleep 1"
or equivalent to work reliably.

18 years agoFix small grammar nit.
wpaul [Thu, 20 Oct 2005 18:33:46 +0000 (18:33 +0000)]
Fix small grammar nit.

18 years agoUse sched_bind() to make sure the DPC threads are bound to the correct
wpaul [Thu, 20 Oct 2005 17:45:58 +0000 (17:45 +0000)]
Use sched_bind() to make sure the DPC threads are bound to the correct
processor, to insure DPC thread 0 runs on CPU0, DPC thread 1 runs on
CPU1, and so on.

Elevate the priority of the workitem threads, though don't use as
high a priority as the DPC threads.