]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
14 years agoMFV: Update nc to the version from OpenBSD 4.7.
Xin LI [Tue, 23 Mar 2010 23:00:35 +0000 (23:00 +0000)]
MFV: Update nc to the version from OpenBSD 4.7.

MFC after: 1 month

14 years agoImport nc from OpenBSD's OPENBSD_4_7 as of today.
Xin LI [Tue, 23 Mar 2010 22:56:19 +0000 (22:56 +0000)]
Import nc from OpenBSD's OPENBSD_4_7 as of today.

14 years agoFall back to VGA palette functions if VESA function failed and DAC is still
Jung-uk Kim [Tue, 23 Mar 2010 22:50:22 +0000 (22:50 +0000)]
Fall back to VGA palette functions if VESA function failed and DAC is still
in 6-bit mode.  Although we have to check non-VGA compatibility bit here,
it seems there are too many broken VESA BIOSes out to rely on it.

14 years agoMap entire video memory again. This is a partial backout of r203535.
Jung-uk Kim [Tue, 23 Mar 2010 22:35:52 +0000 (22:35 +0000)]
Map entire video memory again.  This is a partial backout of r203535.
Although we do not use them all directly, it seems VGA render may access
unmapped memory region and cause kernel panic.

14 years agoSeparate 24-bit pixel draw from 32-bit case. Although it is slower, we do
Jung-uk Kim [Tue, 23 Mar 2010 22:16:57 +0000 (22:16 +0000)]
Separate 24-bit pixel draw from 32-bit case.  Although it is slower, we do
not want to write a useless zero to inaccessible memory region.

14 years agoImplement /proc/filesystems.
John Baldwin [Tue, 23 Mar 2010 21:49:33 +0000 (21:49 +0000)]
Implement /proc/filesystems.

Submitted by: Fernando Apesteguia fernando.apesteguia (gmail)

14 years agoReject attempts to create a MAP_ANON mapping with a non-zero offset.
John Baldwin [Tue, 23 Mar 2010 21:08:07 +0000 (21:08 +0000)]
Reject attempts to create a MAP_ANON mapping with a non-zero offset.

PR: kern/71258
Submitted by: Alexander Best
MFC after: 2 weeks

14 years agoAdd definitions for a 4th PCI host controller. No Freescale processor
Marcel Moolenaar [Tue, 23 Mar 2010 20:12:53 +0000 (20:12 +0000)]
Add definitions for a 4th PCI host controller. No Freescale processor
has all 4 implemented, but across the processors we now support all the
combinations. For example, the MPC8533 doesn't have a PCI controller
at 0xA0000, but does at 0xB0000.

14 years agoAdd PCI ID for MCS9901's parallel port.
Xin LI [Tue, 23 Mar 2010 20:08:18 +0000 (20:08 +0000)]
Add PCI ID for MCS9901's parallel port.

PR: kern/144713
Submitted by: gcooper
MFC after: 2 weeks

14 years agoEnable power management for E500 cores. Use "doze" for now to make
Marcel Moolenaar [Tue, 23 Mar 2010 19:30:56 +0000 (19:30 +0000)]
Enable power management for E500 cores. Use "doze" for now to make
sure the caches remain coherent. For single-core configurations and
with busdma changes we could eventually switch to "nap" and force
a D-cache invalidation as part of the DMA completion. To this end,
clear PSL_WE until after we handled the decrementer or external
interrupt as it tells us whether we just woke up or not.

14 years agoCorrect cross reference.
Xin LI [Tue, 23 Mar 2010 18:00:50 +0000 (18:00 +0000)]
Correct cross reference.

MFC after: 2 weeks

14 years agoWhen receiving a management frame, pass the mbuf to bpf before calling
Rui Paulo [Tue, 23 Mar 2010 14:31:31 +0000 (14:31 +0000)]
When receiving a management frame, pass the mbuf to bpf before calling
iv_recv_mgmt(). iv_recv_mgmt() will generate management frame responses
and pass them to bpf before the management frame that triggered the
response.

PR: 144323
Submitted by: Alexander Egorenkov <egorenar at gmail.com>
MFC after: 2 weeks
Sponsored by: iXsystems, inc.

14 years agoAdd MCS to the list of media types.
Rui Paulo [Tue, 23 Mar 2010 13:15:11 +0000 (13:15 +0000)]
Add MCS to the list of media types.

Sponsored by: iXsystems, inc.

14 years agoAdd a missing LINE_BREAK() after printing the roaming parameters in
Rui Paulo [Tue, 23 Mar 2010 12:05:25 +0000 (12:05 +0000)]
Add a missing LINE_BREAK() after printing the roaming parameters in
verbose mode.

Sponsored by: iXsystems, inc.
MFC after: 2 weeks

14 years agoAdd a new field for extended HT capabilities.
Rui Paulo [Tue, 23 Mar 2010 12:03:09 +0000 (12:03 +0000)]
Add a new field for extended HT capabilities.

Submitted by: Alexander Egorenkov <egorenar at gmail.com>
MFC after: 2 weeks
Sponsored by: iXsystems, inc.

14 years agoFinish the much belated Intel XScale hwpmc(4) man page.
Rui Paulo [Tue, 23 Mar 2010 11:33:08 +0000 (11:33 +0000)]
Finish the much belated Intel XScale hwpmc(4) man page.

14 years agoAdd .snap to daily_clean_tmps_ignore; /tmp/.snap ist not supposed to
Joerg Wunsch [Tue, 23 Mar 2010 06:19:44 +0000 (06:19 +0000)]
Add .snap to daily_clean_tmps_ignore; /tmp/.snap ist not supposed to
be auto-removed (and /tmp is a filesystem of its own now by default).

MFC after: 3 days

14 years agoGet nexus(4) out of the RTC business. The interface used by nexus(4)
Nathan Whitehorn [Tue, 23 Mar 2010 03:14:44 +0000 (03:14 +0000)]
Get nexus(4) out of the RTC business. The interface used by nexus(4)
in Open Firmware was Apple-specific, and we have complete coverage of Apple
system controllers, so move RTC responsibilities into the system controller
drivers. This avoids interesting problems from manipulating these devices
through Open Firmware behind the backs of their drivers.

Obtained from: NetBSD
MFC after: 2 weeks

14 years agoFixes a bug where SACKs in the face of
Randall Stewart [Tue, 23 Mar 2010 01:36:50 +0000 (01:36 +0000)]
Fixes a bug where SACKs in the face of
mapping_array expansion would break. Basically
once we expanded the array we no longer had both
mapping arrays in sync which the sack processing code depends on.
This would mean we were randomly referring to memory that was probably
not there. This mostly just gave us bad sack results going back to the peer.
If INVARIENTS was on of course we would hit the panic routine in the sack_check
call.

We also add a print routine for the place where one would panic in
invarients so one can see what the main mapping array holds.

Reviewed by: tuexen@freebsd.org
MFC after: 2 weeks

14 years agoOpen Firmware on powerpc is generally non-reetrant, so serialize all
Nathan Whitehorn [Tue, 23 Mar 2010 01:11:10 +0000 (01:11 +0000)]
Open Firmware on powerpc is generally non-reetrant, so serialize all
OF calls with a mutex.

14 years agoDo not declare the various OFW command buffers static. It does not
Nathan Whitehorn [Tue, 23 Mar 2010 01:09:45 +0000 (01:09 +0000)]
Do not declare the various OFW command buffers static. It does not
appear to be necessary on either sparc64 or powerpc, and is a
concurrency nightmare.

Reviewed by: marius

14 years agoActually pass a pointer to the trapframe to powerpc_extr_interrupt().
Marcel Moolenaar [Tue, 23 Mar 2010 01:07:30 +0000 (01:07 +0000)]
Actually pass a pointer to the trapframe to powerpc_extr_interrupt().

14 years ago- boot-time size the ipv4 flowtable and the maximum number of flows
Kip Macy [Mon, 22 Mar 2010 23:04:12 +0000 (23:04 +0000)]
- boot-time size the ipv4 flowtable and the maximum number of flows
- increase flow cleaning frequency and decrease flow caching time
  when near the flow limit
- stop allocating new flows when within 3% of maxflows don't start
  allocating again until below 12.5%

MFC after: 7 days

14 years ago- enable alignment on amd64 only
Kip Macy [Mon, 22 Mar 2010 22:39:32 +0000 (22:39 +0000)]
- enable alignment on amd64 only
- only align pcpu caches and the volatile portion of uma_zone

14 years agoExpand $FreeBSD$.
Xin LI [Mon, 22 Mar 2010 22:12:27 +0000 (22:12 +0000)]
Expand $FreeBSD$.

14 years agoMFV of tzdata2010f:
Edwin Groothuis [Mon, 22 Mar 2010 21:27:51 +0000 (21:27 +0000)]
MFV of tzdata2010f:

The Australian Antartic Division:
- Macquarie Island will stay on UTC+11 for winter and not switch back from DST.
- Casey station reverted to its normal time of UTC+8 on 5 March 2010.
- Davis station will revert to its normal time of UTC+7 at 10 March 2010
- Mawson station stays on UTC+5.

Syria will start DST on Thursday 1 April 2010 at midnight.

Correct Samao DST start date (26 Sep vs 24 Oct)

14 years agoVendor import of tzdata2010f:
Edwin Groothuis [Mon, 22 Mar 2010 21:25:08 +0000 (21:25 +0000)]
Vendor import of tzdata2010f:

The Australian Antartic Division:
- Macquarie Island will stay on UTC+11 for winter and not switch back from DST.
- Casey station reverted to its normal time of UTC+8 on 5 March 2010.
- Davis station will revert to its normal time of UTC+7 at 10 March 2010
- Mawson station stays on UTC+5.

Syria will start DST on Thursday 1 April 2010 at midnight.

Correct Samao DST start date (26 Sep vs 24 Oct)

Obtained from: ftp://elsie.nci.nih.gov/pub/

14 years agoEnable mmap for minigzip(1).
Xin LI [Mon, 22 Mar 2010 21:19:17 +0000 (21:19 +0000)]
Enable mmap for minigzip(1).

14 years agoUpdate to zlib 1.2.4 and add versioned symbols to the
Xin LI [Mon, 22 Mar 2010 21:11:55 +0000 (21:11 +0000)]
Update to zlib 1.2.4 and add versioned symbols to the
library.

Sponsored by: iXsystems, Inc.

14 years agoSupport memory wraparound instead of high memory as VM86 mode does.
Jung-uk Kim [Mon, 22 Mar 2010 18:43:36 +0000 (18:43 +0000)]
Support memory wraparound instead of high memory as VM86 mode does.

Suggested by: delphij

14 years agoo Remove the pmap argument to pmap_invalidate_all() as it's not used
Marcel Moolenaar [Mon, 22 Mar 2010 18:24:42 +0000 (18:24 +0000)]
o   Remove the pmap argument to pmap_invalidate_all() as it's not used
    other than in a potentially dangerous KASSERT.
o   Hand-inline pmap_remove_page() as it's only called from 1 place and
    the abstraction that pmap_remove_page() provides is not enough to
    warrant the obfuscation. Eliminate the dangerous KASSERT in the
    process.
o   In pmap_remove_pte(), remove the KASSERT for pmap being the current
    one as it's not safe in the face of CPU migration.

14 years agoFix i386 PAE kernel build.
Jung-uk Kim [Mon, 22 Mar 2010 17:30:34 +0000 (17:30 +0000)]
Fix i386 PAE kernel build.

Reported by: tinderbox

14 years agoRemove unneeded type specifiers from 64-bit constants. The compiler
John Baldwin [Mon, 22 Mar 2010 15:08:26 +0000 (15:08 +0000)]
Remove unneeded type specifiers from 64-bit constants.  The compiler
infers their natural type from the constants' values.

Submitted by: bde
MFC after: 3 days

14 years agoMerge r197455 from amd64:
Ed Maste [Mon, 22 Mar 2010 11:52:53 +0000 (11:52 +0000)]
Merge r197455 from amd64:

  Add a backtrace to the "fpudna in kernel mode!" case, to help track down
  where this comes from.

  Reviewed by: bde

14 years agoForced commit to register MFC reminder.
Dag-Erling Smørgrav [Mon, 22 Mar 2010 11:32:19 +0000 (11:32 +0000)]
Forced commit to register MFC reminder.

MFC after: 2 weeks

14 years agoMerge upstream r432:
Dag-Erling Smørgrav [Mon, 22 Mar 2010 11:00:57 +0000 (11:00 +0000)]
Merge upstream r432:

pam_end() already contains a NULL check, and it is not unreasonable to
call it with a NULL pamh in a cleanup / error-handling situation.  Remove
OPENPAM_NONNULL, which may cause gcc to optimize away the NULL check.

This fixes a potential NULL pointer dereference in error-handling code
in passwd(1).

14 years agoMerge upstream r432:
Dag-Erling Smørgrav [Mon, 22 Mar 2010 10:58:16 +0000 (10:58 +0000)]
Merge upstream r432:

pam_end() already contains a NULL check, and it is not unreasonable to
call it with a NULL pamh in a cleanup / error-handling situation.  Remove
OPENPAM_NONNULL, which may cause gcc to optimize away the NULL check.

14 years agoDrop the pmap argument to pmap_invalidate_page(). It's not used other
Marcel Moolenaar [Mon, 22 Mar 2010 04:24:19 +0000 (04:24 +0000)]
Drop the pmap argument to pmap_invalidate_page(). It's not used other
than in a KASSERT. The KASSERT is broken in that it's done outside the
critical section and as such isn't protected against CPU migration.
Improve pmap_invalidate_page() as follows:
o   calculate vhpt_ofs inside the critical region for exactly the same
    reason.
o   calculate the tag outside the FOREACH loop, as it's loop-invariant.
    This is more efficient.
o   Replace the test and set with an atomic cmpset operation because we
    are changing other CPU's VHPT tables and this avoids invalidating
    after the entry got modified. Not necessarily a problem, but better
    safe than sorry.

14 years agoWith preemption, the high FP registers may get enabled by cpu_switch()
Marcel Moolenaar [Mon, 22 Mar 2010 04:01:45 +0000 (04:01 +0000)]
With preemption, the high FP registers may get enabled by cpu_switch()
before we grab the mutex. Don't assert that they must be disabled at
that point. We pretty much bypass all logic in that case anyway and
leave immediately, so there's no harm.

14 years agoFix interrupt handling by extending the critical region so that
Marcel Moolenaar [Mon, 22 Mar 2010 03:55:18 +0000 (03:55 +0000)]
Fix interrupt handling by extending the critical region so that
preemption doesn't happen until after all pending interrupt have
been services.
While here again, simplify the EOI handling by doing it after we
call the XIV-specific handlers, rather than in each of them. The
original thought was that we may want to do an EOI first and the
actual IPI handling next, but that's mostly a micro-optimization.

14 years agoDisable interrupts when calling into SAL for PCI configuration
Marcel Moolenaar [Mon, 22 Mar 2010 03:06:11 +0000 (03:06 +0000)]
Disable interrupts when calling into SAL for PCI configuration
cycles. This serves 2 purposes:
1.  It prevents preemption and CPU migration while running SAL code.
2.  It reduces the chance of stack overflows: we're supposed to enter
    SAL with at least 16KB of either memory- or register stack space,
    which we can't do without switching to a different stack.

14 years agoDefine curthread as an inline function that loads the thread pointer
Marcel Moolenaar [Mon, 22 Mar 2010 02:01:33 +0000 (02:01 +0000)]
Define curthread as an inline function that loads the thread pointer
directly from r13, the pcpu pointer. This guarantees correct behaviour
when the thread migrates to a different CPU.

14 years agoPrint MD fields in the pcpu to aid debugging.
Marcel Moolenaar [Sun, 21 Mar 2010 22:39:11 +0000 (22:39 +0000)]
Print MD fields in the pcpu to aid debugging.

14 years agoDon't include <machine/_regset.h> when _MACHINE_REGSET_H_ in defined.
Marcel Moolenaar [Sun, 21 Mar 2010 22:33:09 +0000 (22:33 +0000)]
Don't include <machine/_regset.h> when _MACHINE_REGSET_H_ in defined.
This is not for multiple inclusion purposes, because _regset.h already
handles this, but to enable inclusion of the MD header by cross-tools
on non-ia64 installations. The cross-tool can include _regset.h itself
before including MD headers that depend on it.

14 years agoReplace -b with -C and -B (as proposed by Alexander).
Edwin Groothuis [Sun, 21 Mar 2010 21:33:21 +0000 (21:33 +0000)]
Replace -b with -C and -B (as proposed by Alexander).
Add -3, -A and -B to the usage.
Update regression test for the new parameters.

14 years agoMake sure we insert and remove the PV entries related to unmanaged kernel
Olivier Houchard [Sun, 21 Mar 2010 21:03:35 +0000 (21:03 +0000)]
Make sure we insert and remove the PV entries related to unmanaged kernel
mappings into the kernel pmap, not into the pmap related to the
pmap_enter_pv()/pmap_remove_pv() call.

14 years agoLet opendir() use O_DIRECTORY.
Ed Schouten [Sun, 21 Mar 2010 20:45:06 +0000 (20:45 +0000)]
Let opendir() use O_DIRECTORY.

I am not removing the fstat() calls here, since we cannot yet assume
people will always run kernels that respect O_DIRECTORY.

14 years agoActually make O_DIRECTORY work.
Ed Schouten [Sun, 21 Mar 2010 20:43:23 +0000 (20:43 +0000)]
Actually make O_DIRECTORY work.

According to POSIX open() must return ENOTDIR when the path name does
not refer to a path name. Change vn_open() to respect this flag. This
also simplifies the Linuxolator a bit.

14 years ago- Spec tells that CCC interrupt is edge triggered. Acknowledge it as such.
Alexander Motin [Sun, 21 Mar 2010 18:18:58 +0000 (18:18 +0000)]
- Spec tells that CCC interrupt is edge triggered. Acknowledge it as such.
- Do not try to enable CCC if it is not supported.

14 years agoremove pre-vap examples
Sam Leffler [Sun, 21 Mar 2010 17:53:54 +0000 (17:53 +0000)]
remove pre-vap examples

14 years agoAdd a priority-based packet scheduler.
Luigi Rizzo [Sun, 21 Mar 2010 16:30:32 +0000 (16:30 +0000)]
Add a priority-based packet scheduler.

Sponsored by: The ONELAB2 Project
Submitted by: Riccardo Panicucci

14 years agoCorrect the type for uio_resid.
Konstantin Belousov [Sun, 21 Mar 2010 16:14:19 +0000 (16:14 +0000)]
Correct the type for uio_resid.

MFC after: 3 days

14 years agono need for ipfw_flush_tables(), we just need ipfw_destroy_tables()
Luigi Rizzo [Sun, 21 Mar 2010 15:54:07 +0000 (15:54 +0000)]
no need for ipfw_flush_tables(), we just need ipfw_destroy_tables()

14 years agorevise documentation
Luigi Rizzo [Sun, 21 Mar 2010 15:52:55 +0000 (15:52 +0000)]
revise documentation

14 years agoAdd some more codec IDs.
Alexander Motin [Sun, 21 Mar 2010 15:14:21 +0000 (15:14 +0000)]
Add some more codec IDs.

14 years agoAdd 'rotate' and 'getactive' verbs to provide some control and information
Matt Jacob [Sun, 21 Mar 2010 15:02:47 +0000 (15:02 +0000)]
Add 'rotate' and 'getactive' verbs to provide some control and information
about what the currently active path is.

Sponsored by: Panasas
MFC after: 1 month

14 years agoAvoid holding the VLAN_LOCK() over the parent interface SIOCGIFMEDIA
Ed Maste [Sun, 21 Mar 2010 15:00:33 +0000 (15:00 +0000)]
Avoid holding the VLAN_LOCK() over the parent interface SIOCGIFMEDIA
ioctl call, as it may sleep.

Reviewed by: rwatson

14 years agoAvoid aliasing which leads to incorrect results when compiling with the
Marius Strobl [Sun, 21 Mar 2010 13:18:08 +0000 (13:18 +0000)]
Avoid aliasing which leads to incorrect results when compiling with the
default strict aliasing rules.

PR: 144900
Submitted by: Peter Jeremy
MFC after: 3 days

14 years ago- The firmware of Sun Fire V1280 has a misfeature of setting %wstate to
Marius Strobl [Sun, 21 Mar 2010 13:09:54 +0000 (13:09 +0000)]
- The firmware of Sun Fire V1280 has a misfeature of setting %wstate to
  7 which corresponds to WSTATE_KMIX in OpenSolaris whenever calling into
  it which totally screws us even when restoring %wstate afterwards as
  spill/fill traps can happen while in OFW. The rather hackish OpenBSD
  approach of just setting the equivalent of WSTATE_KERNEL to 7 also is
  no option as we treat %wstate as a bit field. So in order to deal with
  this problem actually implement spill/fill handlers for %wstate 7 which
  just act as the WSTATE_KERNEL ones except of theoretically also handling
  32-bit, turn off interrupts completely so we don't even take IPIs while
  in OFW which should ensure we only take spill/fill traps at most and
  restore %wstate after calling into OFW once we have taken over the trap
  table. While at it, actually set WSTATE_{,PROM}_KMIX before calling into
  OFW just like OpenSolaris does, which should at least help testing this
  change on non-V1280.
- Remove comments referring to the %wstate usage in BSD/OS.
- Remove the no longer used RSF_ALIGN_RETRY macro.
- Correct some trap table addresses in comments.
- Ensure %wstate is set to WSTATE_KERNEL when taking over the trap table.
- Ensure PSTATE_AM is off when entering or exiting to OFW as well as that
  interrupts are also completely off when exiting to OFW as the firmware
  trap table shouldn't be used to handle our interrupts.

14 years agoEliminate a pointless TLB invalidation from pmap_bootstrap(). No mappings
Alan Cox [Sun, 21 Mar 2010 00:21:59 +0000 (00:21 +0000)]
Eliminate a pointless TLB invalidation from pmap_bootstrap().  No mappings
whatsoever are changed between the earlier load_cr3() and this invalidation.

14 years agoI am told by AMD that the machine check hardware on the instruction TLB
Alan Cox [Sun, 21 Mar 2010 00:13:11 +0000 (00:13 +0000)]
I am told by AMD that the machine check hardware on the instruction TLB
won't generate bogus exceptions.  Therefore, the implementation of the
"unofficial" workaround needn't mask L1TP errors by the instruction cache
unit.

14 years agoImprove the KVA space sizing of 186682; on machines with large dTLBs we
Marius Strobl [Sat, 20 Mar 2010 23:00:43 +0000 (23:00 +0000)]
Improve the KVA space sizing of 186682; on machines with large dTLBs we
can actually use all of the available lockable entries of the tiny dTLB
for the kernel TSB. With this change the KVA space sizing happens to be
more in line with the MI one so up to at least 24GB machines KVA doesn't
need to be limited manually. This is just another stopgap though, the
real solution is to take advantage of ASI_ATOMIC_QUAD_LDD_PHYS on CPUs
providing it so we don't need to lock the kernel TSB pages into the dTLB
in the first place.

14 years agoDo not create *.gmon files for PIE executables on i386.
Jilles Tjoelker [Sat, 20 Mar 2010 22:58:54 +0000 (22:58 +0000)]
Do not create *.gmon files for PIE executables on i386.

Scrt1_c.o was accidentally compiled with -DGCRT (profiling), like gcrt1_c.o.
This problem is i386-specific, the other architectures are OK.

If you have problems with PIE executables such as samba and cups leaving
behind gmon files, rebuild them after installing this change.

PR: ports/143924
Reviewed by: kib
MFC after: 3 days

14 years ago- While SPARC V9 allows tininess to be detected either before or after
Marius Strobl [Sat, 20 Mar 2010 22:32:13 +0000 (22:32 +0000)]
- While SPARC V9 allows tininess to be detected either before or after
  rounding (impl. dep. #55), the SPARC JPS1 responsible for SPARC64 and
  UltraSPARC processors defines that in all cases tinyness is detected
  before rounding, therefore rounding up to the smallest normalised
  number should set the underflow flag.
- If an infinite result is rounded down, the result should have an
  exponent 1 less than the value for infinity.

PR: 144900
Submitted by: Peter Jeremy
MFC after: 3 days

14 years agoDivision should take both arguments' signs into account when the
Marius Strobl [Sat, 20 Mar 2010 22:12:15 +0000 (22:12 +0000)]
Division should take both arguments' signs into account when the
the dividend is infinity or zero and the divisor is not the same.

PR: 144900
Submitted by: Peter Jeremy
MFC after: 3 days

14 years agoFPU_DEBUG requires <stdio.h>.
Marius Strobl [Sat, 20 Mar 2010 21:27:44 +0000 (21:27 +0000)]
FPU_DEBUG requires <stdio.h>.

PR: 144900
Submitted by: Peter Jeremy
MFC after: 3 days

14 years agoEnsure that __fpu_ftox() both returns the high bits and res[1] contains
Marius Strobl [Sat, 20 Mar 2010 21:15:56 +0000 (21:15 +0000)]
Ensure that __fpu_ftox() both returns the high bits and res[1] contains
the low bits also in the default case.

PR: 144900
Obtained from: OpenBSD
MFC after: 3 days

14 years ago- Remove a bogus forward declaration.
Marius Strobl [Sat, 20 Mar 2010 21:04:47 +0000 (21:04 +0000)]
- Remove a bogus forward declaration.
- Fix whitespace.

14 years ago- spread tcp timer callout load evenly across cpus if net.inet.tcp.per_cpu_timers...
Kip Macy [Sat, 20 Mar 2010 19:47:30 +0000 (19:47 +0000)]
- spread tcp timer callout load evenly across cpus if net.inet.tcp.per_cpu_timers is set to 1
- don't default to acquiring tcbinfo lock exclusively in rexmt

MFC after: 7 days

14 years agoFix command example, presumed leftovers of old markup.
Gavin Atkinson [Sat, 20 Mar 2010 16:34:51 +0000 (16:34 +0000)]
Fix command example, presumed leftovers of old markup.

MFC after: 1 week

14 years agoEscape characters unsafe for XML output in GEOM class, instance and
Jaakko Heinonen [Sat, 20 Mar 2010 16:16:13 +0000 (16:16 +0000)]
Escape characters unsafe for XML output in GEOM class, instance and
provider names.

- Characters in range 0x01-0x1f except '\t', '\n', and '\r' are replaced
  with '?'. Those characters are disallowed in XML.
- '&', '<', '>', '\'', '"' and characters in range 0x7f-0xff are
  replaced with XML numeric character reference.

If the kern.geom.confxml sysctl provides invalid XML, libgeom
geom_xml2tree() fails and utilities using it do not work. Unsafe
characters are common in msdosfs and cd9660 labels.

PR: kern/104389
Submitted by: Doug Steinwand (original version)
Reviewed by: pjd
Discussed on: freebsd-geom
MFC after: 3 weeks

14 years agoCorrect MAXPHYS and DFLTPHYS description.
Alexander Motin [Sat, 20 Mar 2010 15:30:26 +0000 (15:30 +0000)]
Correct MAXPHYS and DFLTPHYS description.

14 years agoTweak language to make one point potentially clearer for non-native spekers
Gavin Atkinson [Sat, 20 Mar 2010 14:42:16 +0000 (14:42 +0000)]
Tweak language to make one point potentially clearer for non-native spekers

PR: bin/121424
Submitted by: "Julian H. Stacey" <jhs berklix.org>

14 years agoRevisit locking in the 64-bit AIM PMAP. The PVO head for a page is
Nathan Whitehorn [Sat, 20 Mar 2010 14:35:24 +0000 (14:35 +0000)]
Revisit locking in the 64-bit AIM PMAP. The PVO head for a page is
generally protected by the VM page queue mutex. Instead of extending the
table lock to cover the PVO heads, add some asserts that the page queue
mutex is in fact held. This fixes several LORs and possible deadlocks.

This also adds an optimization to moea64_kextract() useful for
direct-mapped quantities, like UMA buffers. Being able to use this from
inside UMA removes an additional LOR.

14 years agoRemove some stale entries from MAINTAINERS:
Gavin Atkinson [Sat, 20 Mar 2010 12:39:28 +0000 (12:39 +0000)]
Remove some stale entries from MAINTAINERS:
 - aradford@amcc.com bounces.
 - gad said to drop pkill(1) and ps(1).
 - jmallett said to drop libufs.
 - joerg said to drop fdc(4) and sppp(4)

Submitted by: linimon
Approved by: gad, jmallett, joerg (via linimon)

14 years agowollman@ has OK'd dropping his maintainership of zic(8)
Gavin Atkinson [Sat, 20 Mar 2010 11:05:11 +0000 (11:05 +0000)]
wollman@ has OK'd dropping his maintainership of zic(8)

Submitted by: linimon
Approved by: wollman (via linimon)

14 years agoChop off excessive long hostnames to keep the table structure intact.
Ed Schouten [Sat, 20 Mar 2010 09:29:25 +0000 (09:29 +0000)]
Chop off excessive long hostnames to keep the table structure intact.

This makes lastlogin(8) use exactly the same printf format as last(1).

14 years agoSibyte provides a 64-bit read-only counter that counts at half the processor
Neel Natu [Sat, 20 Mar 2010 05:49:06 +0000 (05:49 +0000)]
Sibyte provides a 64-bit read-only counter that counts at half the processor
frequency. This counter can be accessed coherently from both cores.

Use this as the preferred timecounter for the SWARM kernels.

The CP0 COUNT register is unusable as the timecounter on SMP platforms because
the COUNT registers on different CPUs are not guaranteed to be in sync.

14 years agoMake sure that the registers 'v0' and 'v1' are properly sign-extended
Neel Natu [Sat, 20 Mar 2010 05:21:14 +0000 (05:21 +0000)]
Make sure that the registers 'v0' and 'v1' are properly sign-extended
when sb_load64() returns.

Some 32-bit arithmetic operations (e.g. subu) have unpredicatable results
when operating on 64-bit registers that are not properly sign-extended.

14 years agoGet rid of unused macro MIPS_MEM_RID.
Neel Natu [Sat, 20 Mar 2010 05:10:44 +0000 (05:10 +0000)]
Get rid of unused macro MIPS_MEM_RID.

Suggested by: Alexandr Rybalko (ray@dlink.ua)

14 years agoThis change enables use of physical memory that is beyond the direct
Neel Natu [Sat, 20 Mar 2010 05:07:15 +0000 (05:07 +0000)]
This change enables use of physical memory that is beyond the direct
mapped kseg0 region.

The basic idea is to use KVA from the kseg2 region for mapping page
table pages that lie beyond the direct mapped region.

The TLB miss handler can now recursively fault into the TLB invalid
handler if it dereferences a kseg2 page table page address that is not
in the TLB.

Tested by: JC (c.jayachandran@gmail.com)

14 years agoEnable MSI by default for SiI3124.
Alexander Motin [Sat, 20 Mar 2010 04:40:15 +0000 (04:40 +0000)]
Enable MSI by default for SiI3124.

14 years agoDon't check for boot_verbose in the environment. The loader does
Marcel Moolenaar [Sat, 20 Mar 2010 04:22:22 +0000 (04:22 +0000)]
Don't check for boot_verbose in the environment. The loader does
that already and sets RB_VERBOSE. The loader has always done it.

14 years agoLet unin(4) attach to U3 controllers found on G5 machines.
Nathan Whitehorn [Sat, 20 Mar 2010 03:58:00 +0000 (03:58 +0000)]
Let unin(4) attach to U3 controllers found on G5 machines.

Submitted by: Andreas Tobler

14 years agoAdd support for the Samsung S3C2xx0 family of ARM SoCs written by
Warner Losh [Sat, 20 Mar 2010 03:39:35 +0000 (03:39 +0000)]
Add support for the Samsung S3C2xx0 family of ARM SoCs written by
Andrew Turner.  The kernel supports the LN2410SBC evaluation board,
and likely others.  These parts (or similar ones) are in some open
hardware designs for phones.

Submitted by: Andrew Turner

14 years ago- Map EBDA if available and add 64KB above 1MB (high memory), just in case.
Jung-uk Kim [Fri, 19 Mar 2010 21:15:43 +0000 (21:15 +0000)]
- Map EBDA if available and add 64KB above 1MB (high memory), just in case.
- Print the initial memory map when bootverbose is set.
- Change the page fault address format from linear to %cs:%ip style.
- Move duplicate code into a newly added function.
- Add strictly aligned memory access for distant future. ;-)

14 years agoThe same code is used to import and to create pool.
Pawel Jakub Dawidek [Fri, 19 Mar 2010 20:14:27 +0000 (20:14 +0000)]
The same code is used to import and to create pool.
The order of operations is the following:
1. Try to open vdev by remembered path and guid.
2. If 1 failed, try to find vdev which guid matches and ignore the path.
3. If 2 failed this means either that the vdev we're looking for is gone
   or that pool is being created and vdev doesn't contain proper guid yet.
   To be able to handle pool creation we open vdev by path anyway.

Because of 3 it is possible that we open wrong vdev on import which can lead to
confusions.

The solution for this is to check spa_load_state. On pool creation it will be
equal to SPA_LOAD_NONE and we can open vdev only by path immediately and if it
is not equal to SPA_LOAD_NONE we first open by path+guid and when that fails,
we open by guid. We no longer open wrong vdev on import.

MFC after: 2 weeks

14 years agoSplit eventhandler_register() into an internal part and a wrapper function
Bjoern A. Zeeb [Fri, 19 Mar 2010 19:51:03 +0000 (19:51 +0000)]
Split eventhandler_register() into an internal part and a wrapper function
that provides the allocated and setup eventhandler entry.

Add a new wrapper for VIMAGE that allocates extra space to hold the
callback function and argument in addition to an extra wrapper function.
While the wrapper function goes as normal callback function the
argument points to the extra space allocated holding the original func
and arg that the wrapper function can then call.

Provide an iterator function for the virtual network stack (vnet) that
will call the callback function for each network stack.

Provide a new set of macros for VNET that in the non-VIMAGE case will
just call eventhandler_register() while in the VIMAGE case it will use
vimage_eventhandler_register() passing in the extra iterator function
but will only register once rather than per-vnet.
We need a special macro in case we are interested in the tag returned
as we must check for curvnet and can neither simply assign the
return value, nor not change it in the non-vnet0 case without that.

Sponsored by: ISPsystem
Discussed with: jhb
Reviewed by: zec (earlier version), jhb
MFC after: 1 month

14 years agoBack out revision 205307.
Xin LI [Fri, 19 Mar 2010 16:09:57 +0000 (16:09 +0000)]
Back out revision 205307.

For the record:

CPU_ENABLE_SSE enables some code that dynamically enables SSE support
but not necessarily enforce execution of SSE instructions.

14 years agoSlightly improve my previous commit.
Ed Schouten [Fri, 19 Mar 2010 15:53:02 +0000 (15:53 +0000)]
Slightly improve my previous commit.

Just comment out the atrun line instead of completely removing it. It is
not a bad idea to leave it as a reference in case someone decides to
install atrun by hand afterwards.

14 years agopmap amd64/i386: fix a typo in a comment
Andriy Gapon [Fri, 19 Mar 2010 14:48:32 +0000 (14:48 +0000)]
pmap amd64/i386: fix a typo in a comment

MFC after: 3 days

14 years agovfs_mount.9: drop cross-reference to a removed manual
Andriy Gapon [Fri, 19 Mar 2010 14:35:38 +0000 (14:35 +0000)]
vfs_mount.9: drop cross-reference to a removed manual

MFC after: 3 days

14 years agoUse the same policy for rejecting / not-reject ACPI tables with incorrect
John Baldwin [Fri, 19 Mar 2010 12:43:18 +0000 (12:43 +0000)]
Use the same policy for rejecting / not-reject ACPI tables with incorrect
checksums as the base acpi(4) driver.  This fixes a problem where the MADT
parser would reject the MADT table during early boot causing the MP Table
to be, but then the acpi(4) driver would attach and use non-SMP interrupt
routing.

Tested by: Alastair Hogge  agh of coolrhaug com
MFC after: 1 week

14 years agoDon't add the atrun-line to the crontab when MK_AT is set.
Ed Schouten [Fri, 19 Mar 2010 11:59:02 +0000 (11:59 +0000)]
Don't add the atrun-line to the crontab when MK_AT is set.

This prevents spurious calls to sendmail every 5 minutes.

MFC after: 1 week

14 years agoRegen
Konstantin Belousov [Fri, 19 Mar 2010 11:14:37 +0000 (11:14 +0000)]
Regen

14 years agoRemove empty line.
Konstantin Belousov [Fri, 19 Mar 2010 11:13:42 +0000 (11:13 +0000)]
Remove empty line.

MFC after: 2 weeks

14 years agoConvert aio syscall registration to SYSCALL_INIT_HELPER.
Konstantin Belousov [Fri, 19 Mar 2010 11:11:34 +0000 (11:11 +0000)]
Convert aio syscall registration to SYSCALL_INIT_HELPER.

Reviewed by: jhb
MFC after: 2 weeks

14 years agoImplement compat32 shims for mqueuefs.
Konstantin Belousov [Fri, 19 Mar 2010 11:10:24 +0000 (11:10 +0000)]
Implement compat32 shims for mqueuefs.

Reviewed by: jhb
MFC after: 2 weeks

14 years agoImplement compat32 shims for ksem syscalls.
Konstantin Belousov [Fri, 19 Mar 2010 11:08:43 +0000 (11:08 +0000)]
Implement compat32 shims for ksem syscalls.

Reviewed by: jhb
MFC after: 2 weeks

14 years agoMove SysV IPC freebsd32 compat shims from freebsd32_misc.c to corresponding
Konstantin Belousov [Fri, 19 Mar 2010 11:04:42 +0000 (11:04 +0000)]
Move SysV IPC freebsd32 compat shims from freebsd32_misc.c to corresponding
sysv_{msg,sem,shm}.c files.

Mark SysV IPC freebsd32 syscalls as NOSTD and add required
SYSCALL_INIT_HELPER/SYSCALL32_INIT_HELPERs to provide auto
register/unregister on module load.

This makes COMPAT_FREEBSD32 functional with SysV IPC compiled and loaded
as modules.

Reviewed by: jhb
MFC after: 2 weeks