openwrt/staging/blogic.git
14 years ago[S390] dasd: support DIAG access for read-only devices
Stefan Weinhuber [Mon, 7 Dec 2009 11:51:48 +0000 (12:51 +0100)]
[S390] dasd: support DIAG access for read-only devices

When a DASD device is used with the DIAG discipline, the DIAG
initialization will indicate success or error with a respective
return code. So far we have interpreted a return code of 4 as error,
but it actually means that the initialization was successful, but
the device is read-only. To allow read-only devices to be used with
DIAG we need to accept a return code of 4 as success.

Re-initialization of the DIAG access is also part of the DIAG error
recovery. If we find that the access mode of a device has been
changed from writable to read-only while the device was in use,
we print an error message.

Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] Use do_exception() in pagetable walk usercopy functions.
Gerald Schaefer [Mon, 7 Dec 2009 11:51:47 +0000 (12:51 +0100)]
[S390] Use do_exception() in pagetable walk usercopy functions.

The pagetable walk usercopy functions have used a modified copy of the
do_exception() function for fault handling. This lead to inconsistencies
with recent changes to do_exception(), e.g. performance counters. This
patch changes the pagetable walk usercopy code to call do_exception()
directly, eliminating the redundancy. A new parameter is added to
do_exception() to specify the fault address.

Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] fault handler access flags check.
Martin Schwidefsky [Mon, 7 Dec 2009 11:51:46 +0000 (12:51 +0100)]
[S390] fault handler access flags check.

Simplify the check of the vma->flags in do_exception for the
different fault types.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] fault handler performance optimization.
Martin Schwidefsky [Mon, 7 Dec 2009 11:51:45 +0000 (12:51 +0100)]
[S390] fault handler performance optimization.

Slim down the do_exception function to handle only the fast path of a
fault and move the exceptional cases into a new function. That slightly
increases the performance of the fault handling.

Build fix for !CONFIG_COMPAT by
Kamalesh Babulal <kamalesh@linux.vnet.ibm.com>

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] Improve notify_page_fault implementation.
Martin Schwidefsky [Mon, 7 Dec 2009 11:51:44 +0000 (12:51 +0100)]
[S390] Improve notify_page_fault implementation.

notify_page_fault does a preempt_disable/preempt_enable for each
fault generated by a kernel access to user space. If kprobes
is not active that is unnecessary since the interrupts are not
reenabled yet. To play safe repeat the kprobe_running check after
preempt_disable().

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] Improve address space mode selection.
Martin Schwidefsky [Mon, 7 Dec 2009 11:51:43 +0000 (12:51 +0100)]
[S390] Improve address space mode selection.

Introduce user_mode to replace the two variables switch_amode and
s390_noexec. There are three valid combinations of the old values:
  1) switch_amode == 0 && s390_noexec == 0
  2) switch_amode == 1 && s390_noexec == 0
  3) switch_amode == 1 && s390_noexec == 1
They get replaced by
  1) user_mode == HOME_SPACE_MODE
  2) user_mode == PRIMARY_SPACE_MODE
  3) user_mode == SECONDARY_SPACE_MODE
The new kernel parameter user_mode=[primary,secondary,home] lets
you choose the address space mode the user space processes should
use. In addition the CONFIG_S390_SWITCH_AMODE config option
is removed.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] Improve address space check.
Martin Schwidefsky [Mon, 7 Dec 2009 11:51:42 +0000 (12:51 +0100)]
[S390] Improve address space check.

A data access in access-register mode always is a user mode access,
the code to inspect the access-registers can be removed. The second
change is to use a different test to check for no-execute fault.
The third change is to pass the translation exception identification
as parameter, in theory the trans_exc_code in the lowcore could have
been overwritten by the time the call to check_space from do_no_context
is done.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: dont unregister a busy device in ccw_device_set_offline
Sebastian Ott [Mon, 7 Dec 2009 11:51:41 +0000 (12:51 +0100)]
[S390] cio: dont unregister a busy device in ccw_device_set_offline

If we detect a busy subchannel after the driver's set_offline
callback returned in ccw_device_set_offline, the current behavior
is to unregister the device, which may lead to undesired
consequences. Change this to just quiesce the subchannel and go on
with the offline processing.

Note: This is no excuse for not fixing these drivers -
after the set_offline callback they should have no running IO!

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: improve error recovery for internal I/Os
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:40 +0000 (12:51 +0100)]
[S390] cio: improve error recovery for internal I/Os

Improve error recovery for internal I/Os by repeating each I/O
256 times per path to cope with long-running non-permanent error
conditions. Also retry each path twice to cope with link flapping,
i.e. single paths becoming unavailable in the order in which they
are tried.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: change locking in io_subchannel_remove
Sebastian Ott [Mon, 7 Dec 2009 11:51:39 +0000 (12:51 +0100)]
[S390] cio: change locking in io_subchannel_remove

IO subchannels are always unregistered in process context, so use
spin_lock_irq in the corresponding remove callback.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: quiesce subchannel in io_subchannel_remove
Sebastian Ott [Mon, 7 Dec 2009 11:51:38 +0000 (12:51 +0100)]
[S390] cio: quiesce subchannel in io_subchannel_remove

Ensure that there will be no more interrupts for an
unregistered device by using the same quiesce and disable loop
as in io_subchannel_shutdown.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: handle busy subchannel in ccw_device_move_to_sch
Sebastian Ott [Mon, 7 Dec 2009 11:51:37 +0000 (12:51 +0100)]
[S390] cio: handle busy subchannel in ccw_device_move_to_sch

Try to disable the old subchannel before we ask the driver core
to move the attached device to a new parent. This way we can use
the QUIESCE state during shutdown which prevents a possible use
after free situation in some error cases.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: handle failed disable_subchannel after device recognition
Sebastian Ott [Mon, 7 Dec 2009 11:51:36 +0000 (12:51 +0100)]
[S390] cio: handle failed disable_subchannel after device recognition

Handle a failing cio_disable_subchannel at the end of our device
recognition as if the recognition itself failed. This way
subsequent registration steps do not need to handle enabled
subchannels.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: fix quiesce state
Sebastian Ott [Mon, 7 Dec 2009 11:51:35 +0000 (12:51 +0100)]
[S390] cio: fix quiesce state

DEV_STATE_QUIESCE is used to stop all IO on a busy subchannel.
This patch fixes the following problems related to the QUIESCE
state:

* Fix a potential race condition which could occur when the
resulting state was DEV_STATE_OFFLINE.

* Add missing locking around cio_disable_subchannel,
ccw_device_cancel_halt_clear and the cdev's handler.

* Loop until we know for sure that the subchannel is disabled.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: add per device initialization status flag
Sebastian Ott [Mon, 7 Dec 2009 11:51:34 +0000 (12:51 +0100)]
[S390] cio: add per device initialization status flag

The function ccw_device_unregister has to ensure to remove
all references obtained by device_add and device_initialize.
Unfortunately it gets called for devices which are
1) uninitialized, 2) initialized but unregistered, and
3) registered devices. To distinguish 1) and 2) this patch
introduces a new flag "initialized", which is 1 as long as we
hold the initial device reference.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: remove registered flag from ccw_device_private
Sebastian Ott [Mon, 7 Dec 2009 11:51:33 +0000 (12:51 +0100)]
[S390] cio: remove registered flag from ccw_device_private

We used to maintain a "registered" flag in our ccw_device_private
structure. This patch removes the "registered" flag and converts
all users of it to device_is_registered which has the exact same
meaning.

Note: The usage the atomic operation test_and_clear_bit is replaced
by the non-atomic if (device_is_registered()) device_del(). This
will not do harm, since we serialize calls to ccw_device_unregister
with a single-threaded workqueue.

Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: make steal lock procedure more robust
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:32 +0000 (12:51 +0100)]
[S390] cio: make steal lock procedure more robust

An Unconditional Reserve + Release operation (steal lock) for a
boxed device may fail when encountering special error cases
(e.g. unit checks or path errors). Fix this by using the more
robust ccw_request infrastructure for performing the steal lock
CCW program.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: use sense-pgid operation for path verification
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:31 +0000 (12:51 +0100)]
[S390] cio: use sense-pgid operation for path verification

Set-pgid operations fail for some device types under z/VM for which
the hypervisor has already set the pgid. Also reserved devices or
changed pgids are not correctly recognized. Fix these problems by
using a combination of sense-pgid and set-pgid and by also accepting
pre-defined pgid settings.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: split PGID settings and status
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:30 +0000 (12:51 +0100)]
[S390] cio: split PGID settings and status

Split setting (driver wants feature enabled) and status (feature
setup was successful) for PGID related ccw device features so that
setup errors can be detected. Previously, incorrectly handled setup
errors could in rare cases lead to erratic I/O behavior and
permanently unusuable devices.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: remove intretry flag
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:29 +0000 (12:51 +0100)]
[S390] cio: remove intretry flag

After changing all internal I/O functions to use the newly introduced
ccw request infrastructure, retries are handled automatically after a
clear operation. Therefore remove the internal retry flag and
associated code.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: allow setting not-operational devices offline
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:28 +0000 (12:51 +0100)]
[S390] cio: allow setting not-operational devices offline

Accept a request for setting a not-operational device offline.
This way, users can remove devices from Linux which would otherwise
remain unusable until reboot.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: use ccw request infrastructure for pgid
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:27 +0000 (12:51 +0100)]
[S390] cio: use ccw request infrastructure for pgid

Use the newly introduced ccw request infrastructure to implement
pgid related operations: sense pgid, set pgid and disband pg.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: use ccw request infrastructure for sense id
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:26 +0000 (12:51 +0100)]
[S390] cio: use ccw request infrastructure for sense id

Use the newly introduced ccw request infrastructure to implement
the sense id operation.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: consistent infrastructure for internal I/O requests
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:25 +0000 (12:51 +0100)]
[S390] cio: consistent infrastructure for internal I/O requests

Reduce code duplication by introducing a central infrastructure to
perform an internal I/O operation on a CCW device.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: dont panic in non-fatal conditions
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:24 +0000 (12:51 +0100)]
[S390] cio: dont panic in non-fatal conditions

Remove the call to BUG() for situations which are unexpected
but do not cause actual problems.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: ensure proper locking during device recognition
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:23 +0000 (12:51 +0100)]
[S390] cio: ensure proper locking during device recognition

Device recognition needs to be started with the ccw device lock
held to prevent race conditions between I/O starting and interrupt
reception.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: handle error during path verification consistently
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:22 +0000 (12:51 +0100)]
[S390] cio: handle error during path verification consistently

Handle verification errors consistently through the existing
callback ccw_device_done to reduce cleanup code duplication.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: handle error during device recognition consistently
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:21 +0000 (12:51 +0100)]
[S390] cio: handle error during device recognition consistently

Remove the return code from ccw_device_recognition and handle
recognition errors through the existing callback
ccw_device_recog_done to reduce cleanup code duplication.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: inform user when online/offline processing fails
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:20 +0000 (12:51 +0100)]
[S390] cio: inform user when online/offline processing fails

Print a warning message in case a ccw device enters boxed or
not operational state during online/offline processing.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: introduce ccw device todos
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:19 +0000 (12:51 +0100)]
[S390] cio: introduce ccw device todos

Introduce a central mechanism for performing delayed ccw device work
to ensure that different types of work do not overwrite each other.
Prioritization ensures that the most important work is always
performed while less important tasks are either obsoleted or repeated
later.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: introduce subchannel todos
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:18 +0000 (12:51 +0100)]
[S390] cio: introduce subchannel todos

Ensure that current and future users of sch->work do not overwrite
each other by introducing a single mechanism for delayed subchannel
work.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: introduce parent-initiated device move
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:17 +0000 (12:51 +0100)]
[S390] cio: introduce parent-initiated device move

Change the initiative to update subchannel-ccw device associations
to the subchannel: when there is an indication that the internal
association no longer reflects the current hardware state, mark
each affected subchannel as requiring attention. Once processing
reaches a subchannel, determine the correct association for that
subchannel at that time and perform the necessary device_move
operations.

This change fixes problems with the previous approach which would
leave devices in an inconsistent state when a new hardware change
occurred while a device_move was already scheduled.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: fix repeat setting of cdev parent association
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:16 +0000 (12:51 +0100)]
[S390] cio: fix repeat setting of cdev parent association

sch_create_and_recog_new_device() associates a parent subchannel
with its ccw device child even though this is already done by
the subsequently called io_subchannel_recog(). Also make sure
io_subchannel_recog() sets the association under lock.

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] cio: fix double free in case of probe failure
Peter Oberparleiter [Mon, 7 Dec 2009 11:51:15 +0000 (12:51 +0100)]
[S390] cio: fix double free in case of probe failure

io_subchannel_probe() frees memory for sch->private which is later
freed again when io_subchannel_remove() is called. Fix this problem
by removing the cleanup in io_subchannel_probe().

Signed-off-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years ago[S390] s390: clear high-order bits of registers after sam64
Hendrik Brueckner [Mon, 7 Dec 2009 11:44:42 +0000 (12:44 +0100)]
[S390] s390: clear high-order bits of registers after sam64

When the kernel is IPLed without the CLEAR option and switches
to 64-bit, the high-order half of the registers might contain
random values.  This can cause addressing exceptions and the
kernel enters an interrupt loop.

Initialize the high-order half of the general purpose registers
with zeros after switching to 64-bit mode.

Cc: <stable@kernel.org>
Signed-off-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
14 years agoMerge branch 'x86-debug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 5 Dec 2009 23:33:27 +0000 (15:33 -0800)]
Merge branch 'x86-debug-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-debug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Limit number of per cpu TSC sync messages
  x86: dumpstack, 64-bit: Disable preemption when walking the IRQ/exception stacks
  x86: dumpstack: Clean up the x86_stack_ids[][] initalization and other details
  x86, cpu: mv display_cacheinfo -> cpu_detect_cache_sizes
  x86: Suppress stack overrun message for init_task
  x86: Fix cpu_devs[] initialization in early_cpu_init()
  x86: Remove CPU cache size output for non-Intel too
  x86: Minimise printk spew from per-vendor init code
  x86: Remove the CPU cache size printk's
  cpumask: Avoid cpumask_t in arch/x86/kernel/apic/nmi.c
  x86: Make sure we also print a Code: line for show_regs()

14 years agoMerge branch 'x86-cpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 5 Dec 2009 23:32:35 +0000 (15:32 -0800)]
Merge branch 'x86-cpu-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-cpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, msr, cpumask: Use struct cpumask rather than the deprecated cpumask_t
  x86, cpuid: Simplify the code in cpuid_open
  x86, cpuid: Remove the bkl from cpuid_open()
  x86, msr: Remove the bkl from msr_open()
  x86: AMD Geode LX optimizations
  x86, msr: Unify rdmsr_on_cpus/wrmsr_on_cpus

14 years agoMerge branch 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 23:32:18 +0000 (15:32 -0800)]
Merge branch 'x86-cleanups-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Fix a section mismatch in arch/x86/kernel/setup.c
  x86: Fixup last users of irq_chip->typename
  x86: Remove BKL from apm_32
  x86: Remove BKL from microcode
  x86: use kernel_stack_pointer() in kprobes.c
  x86: use kernel_stack_pointer() in kgdb.c
  x86: use kernel_stack_pointer() in dumpstack.c
  x86: use kernel_stack_pointer() in process_32.c

14 years agoMerge branch 'x86-asm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 5 Dec 2009 23:32:03 +0000 (15:32 -0800)]
Merge branch 'x86-asm-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-asm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  include/linux/compiler-gcc4.h: Fix build bug - gcc-4.0.2 doesn't understand __builtin_object_size
  x86/alternatives: No need for alternatives-asm.h to re-invent stuff already in asm.h
  x86/alternatives: Check replacementlen <= instrlen at build time
  x86, 64-bit: Set data segments to null after switching to 64-bit mode
  x86: Clean up the loadsegment() macro
  x86: Optimize loadsegment()
  x86: Add missing might_fault() checks to copy_{to,from}_user()
  x86-64: __copy_from_user_inatomic() adjustments
  x86: Remove unused thread_return label from switch_to()
  x86, 64-bit: Fix bstep_iret jump
  x86: Don't use the strict copy checks when branch profiling is in use
  x86, 64-bit: Move K8 B step iret fixup to fault entry asm
  x86: Generate cmpxchg build failures
  x86: Add a Kconfig option to turn the copy_from_user warnings into errors
  x86: Turn the copy_from_user check into an (optional) compile time warning
  x86: Use __builtin_memset and __builtin_memcpy for memset/memcpy
  x86: Use __builtin_object_size() to validate the buffer size for copy_from_user()

14 years agoMerge branch 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 5 Dec 2009 23:31:25 +0000 (15:31 -0800)]
Merge branch 'x86-apic-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (30 commits)
  x86, apic: Enable lapic nmi watchdog on AMD Family 11h
  x86: Remove unnecessary mdelay() from cpu_disable_common()
  x86, ioapic: Document another case when level irq is seen as an edge
  x86, ioapic: Fix the EOI register detection mechanism
  x86, io-apic: Move the effort of clearing remoteIRR explicitly before migrating the irq
  x86: SGI UV: Map low MMR ranges
  x86: apic: Print out SRAT table APIC id in hex
  x86: Re-get cfg_new in case reuse/move irq_desc
  x86: apic: Remove not needed #ifdef
  x86: io-apic: IO-APIC MMIO should not fail on resource insertion
  x86: Remove asm/apicnum.h
  x86: apic: Do not use stacked physid_mask_t
  x86, apic: Get rid of apicid_to_cpu_present assign on 64-bit
  x86, ioapic: Use snrpintf while set names for IO-APIC resourses
  x86, apic: Use PAGE_SIZE instead of numbers
  x86: Remove local_irq_enable()/local_irq_disable() in fixup_irqs()
  x86: Use EOI register in io-apic on intel platforms
  x86: Force irq complete move during cpu offline
  x86: Remove move_cleanup_count from irq_cfg
  x86, intr-remap: Avoid irq_chip mask/unmask in fixup_irqs() for intr-remapping
  ...

14 years agoMerge branch 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 23:30:49 +0000 (15:30 -0800)]
Merge branch 'sched-core-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (35 commits)
  sched, cputime: Introduce thread_group_times()
  sched, cputime: Cleanups related to task_times()
  Revert "sched, x86: Optimize branch hint in __switch_to()"
  sched: Fix isolcpus boot option
  sched: Revert 498657a478c60be092208422fefa9c7b248729c2
  sched, time: Define nsecs_to_jiffies()
  sched: Remove task_{u,s,g}time()
  sched: Introduce task_times() to replace task_{u,s}time() pair
  sched: Limit the number of scheduler debug messages
  sched.c: Call debug_show_all_locks() when dumping all tasks
  sched, x86: Optimize branch hint in __switch_to()
  sched: Optimize branch hint in context_switch()
  sched: Optimize branch hint in pick_next_task_fair()
  sched_feat_write(): Update ppos instead of file->f_pos
  sched: Sched_rt_periodic_timer vs cpu hotplug
  sched, kvm: Fix race condition involving sched_in_preempt_notifers
  sched: More generic WAKE_AFFINE vs select_idle_sibling()
  sched: Cleanup select_task_rq_fair()
  sched: Fix granularity of task_u/stime()
  sched: Fix/add missing update_rq_clock() calls
  ...

14 years agoMerge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 5 Dec 2009 23:30:21 +0000 (15:30 -0800)]
Merge branch 'perf-core-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (470 commits)
  x86: Fix comments of register/stack access functions
  perf tools: Replace %m with %a in sscanf
  hw-breakpoints: Keep track of user disabled breakpoints
  tracing/syscalls: Make syscall events print callbacks static
  tracing: Add DEFINE_EVENT(), DEFINE_SINGLE_EVENT() support to docbook
  perf: Don't free perf_mmap_data until work has been done
  perf_event: Fix compile error
  perf tools: Fix _GNU_SOURCE macro related strndup() build error
  trace_syscalls: Remove unused syscall_name_to_nr()
  trace_syscalls: Simplify syscall profile
  trace_syscalls: Remove duplicate init_enter_##sname()
  trace_syscalls: Add syscall_nr field to struct syscall_metadata
  trace_syscalls: Remove enter_id exit_id
  trace_syscalls: Set event_enter_##sname->data to its metadata
  trace_syscalls: Remove unused event_syscall_enter and event_syscall_exit
  perf_event: Initialize data.period in perf_swevent_hrtimer()
  perf probe: Simplify event naming
  perf probe: Add --list option for listing current probe events
  perf probe: Add argv_split() from lib/argv_split.c
  perf probe: Move probe event utility functions to probe-event.c
  ...

14 years agoMerge branch 'tracing-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 17:53:36 +0000 (09:53 -0800)]
Merge branch 'tracing-core-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'tracing-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (40 commits)
  tracing: Separate raw syscall from syscall tracer
  ring-buffer-benchmark: Add parameters to set produce/consumer priorities
  tracing, function tracer: Clean up strstrip() usage
  ring-buffer benchmark: Run producer/consumer threads at nice +19
  tracing: Remove the stale include/trace/power.h
  tracing: Only print objcopy version warning once from recordmcount
  tracing: Prevent build warning: 'ftrace_graph_buf' defined but not used
  ring-buffer: Move access to commit_page up into function used
  tracing: do not disable interrupts for trace_clock_local
  ring-buffer: Add multiple iterations between benchmark timestamps
  kprobes: Sanitize struct kretprobe_instance allocations
  tracing: Fix to use __always_unused attribute
  compiler: Introduce __always_unused
  tracing: Exit with error if a weak function is used in recordmcount.pl
  tracing: Move conditional into update_funcs() in recordmcount.pl
  tracing: Add regex for weak functions in recordmcount.pl
  tracing: Move mcount section search to front of loop in recordmcount.pl
  tracing: Fix objcopy revision check in recordmcount.pl
  tracing: Check absolute path of input file in recordmcount.pl
  tracing: Correct the check for number of arguments in recordmcount.pl
  ...

14 years agoMerge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 17:53:21 +0000 (09:53 -0800)]
Merge branch 'tracing-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  tracing: Fix trace_marker output
  tracing: Fix event format export
  tracing: Fix return value of tracing_stats_read()

14 years agoMerge branch 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 5 Dec 2009 17:53:08 +0000 (09:53 -0800)]
Merge branch 'irq-core-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  genirq: Fix spurious irq seqfile conversion
  genirq: switch /proc/irq/*/spurious to seq_file
  irq: Do not attempt to create subdirectories if /proc/irq/<irq> failed
  irq: Remove unused debug_poll_all_shared_irqs()
  irq: Fix docbook comments
  irq: trivial: Fix typo in comment for #endif

14 years agoMerge branch 'core-softlockup-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 17:52:46 +0000 (09:52 -0800)]
Merge branch 'core-softlockup-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-softlockup-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  softlockup: Fix hung_task_check_count sysctl

14 years agoMerge branch 'core-signal-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 17:52:33 +0000 (09:52 -0800)]
Merge branch 'core-signal-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-signal-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  signal: Print warning message when dropping signals
  signal: Fix alternate signal stack check

14 years agoMerge branch 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 5 Dec 2009 17:52:14 +0000 (09:52 -0800)]
Merge branch 'core-rcu-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (31 commits)
  rcu: Make RCU's CPU-stall detector be default
  rcu: Add expedited grace-period support for preemptible RCU
  rcu: Enable fourth level of TREE_RCU hierarchy
  rcu: Rename "quiet" functions
  rcu: Re-arrange code to reduce #ifdef pain
  rcu: Eliminate unneeded function wrapping
  rcu: Fix grace-period-stall bug on large systems with CPU hotplug
  rcu: Eliminate __rcu_pending() false positives
  rcu: Further cleanups of use of lastcomp
  rcu: Simplify association of forced quiescent states with grace periods
  rcu: Accelerate callback processing on CPUs not detecting GP end
  rcu: Mark init-time-only rcu_bootup_announce() as __init
  rcu: Simplify association of quiescent states with grace periods
  rcu: Rename dynticks_completed to completed_fqs
  rcu: Enable synchronize_sched_expedited() fastpath
  rcu: Remove inline from forward-referenced functions
  rcu: Fix note_new_gpnum() uses of ->gpnum
  rcu: Fix synchronization for rcu_process_gp_end() uses of ->completed counter
  rcu: Prepare for synchronization fixes: clean up for non-NO_HZ handling of ->completed counter
  rcu: Cleanup: balance rcu_irq_enter()/rcu_irq_exit() calls
  ...

14 years agoMerge branch 'core-printk-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 17:50:22 +0000 (09:50 -0800)]
Merge branch 'core-printk-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-printk-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  ratelimit: Make suppressed output messages more useful
  printk: Remove ratelimit.h from kernel.h
  ratelimit: Fix/allow use in atomic contexts
  ratelimit: Use per ratelimit context locking

14 years agoMerge branch 'core-locking-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 17:49:59 +0000 (09:49 -0800)]
Merge branch 'core-locking-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-locking-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  mutex: Fix missing conditions to build mutex_spin_on_owner()
  mutex: Better control mutex adaptive spinning config
  locking, task_struct: Reduce size on TRACE_IRQFLAGS and 64bit
  locking: Use __[SPIN|RW]_LOCK_UNLOCKED in [spin|rw]_lock_init()
  locking: Remove unused prototype
  locking: Reduce ifdefs in kernel/spinlock.c
  locking: Make inlining decision Kconfig based

14 years agoMerge branch 'core-ipi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 5 Dec 2009 17:49:46 +0000 (09:49 -0800)]
Merge branch 'core-ipi-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-ipi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  generic-ipi: Add smp_call_function_any()
  generic-ipi: Fix misleading smp_call_function*() description

14 years agoMerge branch 'core-iommu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 5 Dec 2009 17:49:07 +0000 (09:49 -0800)]
Merge branch 'core-iommu-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-iommu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (63 commits)
  x86, Calgary IOMMU quirk: Find nearest matching Calgary while walking up the PCI tree
  x86/amd-iommu: Remove amd_iommu_pd_table
  x86/amd-iommu: Move reset_iommu_command_buffer out of locked code
  x86/amd-iommu: Cleanup DTE flushing code
  x86/amd-iommu: Introduce iommu_flush_device() function
  x86/amd-iommu: Cleanup attach/detach_device code
  x86/amd-iommu: Keep devices per domain in a list
  x86/amd-iommu: Add device bind reference counting
  x86/amd-iommu: Use dev->arch->iommu to store iommu related information
  x86/amd-iommu: Remove support for domain sharing
  x86/amd-iommu: Rearrange dma_ops related functions
  x86/amd-iommu: Move some pte allocation functions in the right section
  x86/amd-iommu: Remove iommu parameter from dma_ops_domain_alloc
  x86/amd-iommu: Use get_device_id and check_device where appropriate
  x86/amd-iommu: Move find_protection_domain to helper functions
  x86/amd-iommu: Simplify get_device_resources()
  x86/amd-iommu: Let domain_for_device handle aliases
  x86/amd-iommu: Remove iommu specific handling from dma_ops path
  x86/amd-iommu: Remove iommu parameter from __(un)map_single
  x86/amd-iommu: Make alloc_new_range aware of multiple IOMMUs
  ...

14 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw
Linus Torvalds [Sat, 5 Dec 2009 17:47:17 +0000 (09:47 -0800)]
Merge git://git./linux/kernel/git/steve/gfs2-2.6-nmw

* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw: (31 commits)
  GFS2: Fix glock refcount issues
  writeback: remove unused nonblocking and congestion checks (gfs2)
  GFS2: drop rindex glock to refresh rindex list
  GFS2: Tag all metadata with jid
  GFS2: Locking order fix in gfs2_check_blk_state
  GFS2: Remove dirent_first() function
  GFS2: Display nobarrier option in /proc/mounts
  GFS2: add barrier/nobarrier mount options
  GFS2: remove division from new statfs code
  GFS2: Improve statfs and quota usability
  GFS2: Use dquot_send_warning()
  VFS: Export dquot_send_warning
  GFS2: Add set_xquota support
  GFS2: Add get_xquota support
  GFS2: Clean up gfs2_adjust_quota() and do_glock()
  GFS2: Remove constant argument from qd_get()
  GFS2: Remove constant argument from qdsb_get()
  GFS2: Add proper error reporting to quota sync via sysfs
  GFS2: Add get_xstate quota function
  GFS2: Remove obsolete code in quota.c
  ...

14 years agoMerge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/security...
Linus Torvalds [Sat, 5 Dec 2009 17:44:57 +0000 (09:44 -0800)]
Merge branch 'next' of git://git./linux/kernel/git/jmorris/security-testing-2.6

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/security-testing-2.6: (30 commits)
  TOMOYO: Add recursive directory matching operator support.
  remove CONFIG_SECURITY_FILE_CAPABILITIES compile option
  SELinux: print denials for buggy kernel with unknown perms
  Silence the existing API for capability version compatibility check.
  LSM: Move security_path_chmod()/security_path_chown() to after mutex_lock().
  SELinux: header generation may hit infinite loop
  selinux: Fix warnings
  security: report the module name to security_module_request
  Config option to set a default LSM
  sysctl: require CAP_SYS_RAWIO to set mmap_min_addr
  tpm: autoload tpm_tis based on system PnP IDs
  tpm_tis: TPM_STS_DATA_EXPECT workaround
  define convenient securebits masks for prctl users (v2)
  tpm: fix header for modular build
  tomoyo: improve hash bucket dispersion
  tpm add default function definitions
  LSM: imbed ima calls in the security hooks
  SELinux: add .gitignore files for dynamic classes
  security: remove root_plug
  SELinux: fix locking issue introduced with c6d3aaa4e35c71a3
  ...

14 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/brodo/pcmcia-2.6
Linus Torvalds [Sat, 5 Dec 2009 17:42:59 +0000 (09:42 -0800)]
Merge git://git./linux/kernel/git/brodo/pcmcia-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/brodo/pcmcia-2.6: (50 commits)
  pcmcia: rework the irq_req_t typedef
  pcmcia: remove deprecated handle_to_dev() macro
  pcmcia: pcmcia_request_window() doesn't need a pointer to a pointer
  pcmcia: remove unused "window_t" typedef
  pcmcia: move some window-related code to pcmcia_ioctl.c
  pcmcia: Change window_handle_t logic to unsigned long
  pcmcia: Pass struct pcmcia_socket to pcmcia_get_mem_page()
  pcmcia: Pass struct pcmcia_device to pcmcia_map_mem_page()
  pcmcia: Pass struct pcmcia_device to pcmcia_release_window()
  drivers/pcmcia: remove unnecessary kzalloc
  pcmcia: correct handling for Zoomed Video registers in topic.h
  pcmcia: fix printk formats
  pcmcia: autoload module pcmcia
  pcmcia/staging: update comedi drivers
  PCMCIA: stop duplicating pci_irq in soc_pcmcia_socket
  PCMCIA: ss: allow PCI IRQs > 255
  PCMCIA: soc_common: remove 'dev' member from soc_pcmcia_socket
  PCMCIA: soc_common: constify soc_pcmcia_socket ops member
  PCMCIA: sa1111: remove duplicated initializers
  PCMCIA: sa1111: wrap soc_pcmcia_socket to contain sa1111 specific data
  ...

14 years agoavr32: Convert BUG() to use unreachable()
David Daney [Sat, 5 Dec 2009 01:44:54 +0000 (17:44 -0800)]
avr32: Convert BUG() to use unreachable()

Use the new unreachable() macro instead of for(;;);

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Acked-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agos390: Convert BUG() to use unreachable()
David Daney [Sat, 5 Dec 2009 01:44:53 +0000 (17:44 -0800)]
s390: Convert BUG() to use unreachable()

Use the new unreachable() macro instead of for(;;);

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
CC: Heiko Carstens <heiko.carstens@de.ibm.com>
CC: linux390@de.ibm.com
CC: linux-s390@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoMIPS: Convert BUG() to use unreachable()
David Daney [Sat, 5 Dec 2009 01:44:52 +0000 (17:44 -0800)]
MIPS: Convert BUG() to use unreachable()

Use the new unreachable() macro instead of while(1);

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
CC: linux-mips@linux-mips.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agox86: Convert BUG() to use unreachable()
David Daney [Sat, 5 Dec 2009 01:44:51 +0000 (17:44 -0800)]
x86: Convert BUG() to use unreachable()

Use the new unreachable() macro instead of for(;;);.  When
allyesconfig is built with a GCC-4.5 snapshot on i686 the size of the
text segment is reduced by 3987 bytes (from 6827019 to 6823032).

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Ingo Molnar <mingo@redhat.com>
CC: x86@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoAdd support for GCC-4.5's __builtin_unreachable() to compiler.h (v2)
David Daney [Sat, 5 Dec 2009 01:44:50 +0000 (17:44 -0800)]
Add support for GCC-4.5's __builtin_unreachable() to compiler.h (v2)

Starting with version 4.5, GCC has a new built-in function
__builtin_unreachable() that can be used in places like the kernel's
BUG() where inline assembly is used to transfer control flow.  This
eliminated the need for an endless loop in these places.

The patch adds a new macro 'unreachable()' that will expand to either
__builtin_unreachable() or an endless loop depending on the compiler
version.

Change from v1: Simplify unreachable() for non-GCC 4.5 case.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoMerge branch 'perf/probes' into perf/core
Ingo Molnar [Thu, 3 Dec 2009 19:11:37 +0000 (20:11 +0100)]
Merge branch 'perf/probes' into perf/core

Merge reason: add these fixes to 'perf probe'.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoMerge branch 'perf/mce' into perf/core
Ingo Molnar [Thu, 3 Dec 2009 19:10:59 +0000 (20:10 +0100)]
Merge branch 'perf/mce' into perf/core

Merge reason: It's ready for v2.6.33.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoMerge branch 'perf/scripting' into perf/core
Ingo Molnar [Thu, 3 Dec 2009 19:10:35 +0000 (20:10 +0100)]
Merge branch 'perf/scripting' into perf/core

Merge reason: it's ready for v2.6.33.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agox86, apic: Enable lapic nmi watchdog on AMD Family 11h
Mikael Pettersson [Thu, 3 Dec 2009 14:52:44 +0000 (15:52 +0100)]
x86, apic: Enable lapic nmi watchdog on AMD Family 11h

The x86 lapic nmi watchdog does not recognize AMD Family 11h,
resulting in:

  NMI watchdog: CPU not supported

As far as I can see from available documentation (the BKDM),
family 11h looks identical to family 10h as far as the PMU
is concerned.

Extending the check to accept family 11h results in:

  Testing NMI watchdog ... OK.

I've been running with this change on a Turion X2 Ultra ZM-82
laptop for a couple of weeks now without problems.

Signed-off-by: Mikael Pettersson <mikpe@it.uu.se>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Joerg Roedel <joerg.roedel@amd.com>
Cc: <stable@kernel.org>
LKML-Reference: <19223.53436.931768.278021@pilspetsen.it.uu.se>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agoGFS2: Fix glock refcount issues
Steven Whitehouse [Fri, 27 Nov 2009 10:31:11 +0000 (10:31 +0000)]
GFS2: Fix glock refcount issues

This patch fixes some ref counting issues. Firstly by moving
the point at which we drop the ref count after a dlm lock
operation has completed we ensure that we never call
gfs2_glock_hold() on a lock with a zero ref count.

Secondly, by using atomic_dec_and_lock() in gfs2_glock_put()
we ensure that at no time will a glock with zero ref count
appear on the lru_list. That means that we can remove the
check for this in our shrinker (which was racy).

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agowriteback: remove unused nonblocking and congestion checks (gfs2)
Wu Fengguang [Wed, 18 Nov 2009 10:09:41 +0000 (18:09 +0800)]
writeback: remove unused nonblocking and congestion checks (gfs2)

No one is calling wb_writeback and write_cache_pages with
wbc.nonblocking=1 any more. And lumpy pageout will want to do
nonblocking writeback without the congestion wait.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: drop rindex glock to refresh rindex list
Benjamin Marzinski [Tue, 10 Nov 2009 18:54:56 +0000 (12:54 -0600)]
GFS2: drop rindex glock to refresh rindex list

When a gfs2 filesystem is grown, it needs to rebuild the rindex list to be able
to use the new space.  gfs2 does this when the rindex is marked not uptodate,
which happens when the rindex glock is dropped.  However, on a single node
setup, there is never any reason to drop the rindex glock, so gfs2 never
invalidates the the rindex. This patch makes gfs2 automatically drop the
rindex glock after filesystem grows, so it can refresh the rindex list.

Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Tag all metadata with jid
Steven Whitehouse [Fri, 6 Nov 2009 16:20:51 +0000 (16:20 +0000)]
GFS2: Tag all metadata with jid

There are two spare field in the header common to all GFS2
metadata. One is just the right size to fit a journal id
in it, and this patch updates the journal code so that each
time a metadata block is modified, we tag it with the journal
id of the node which is performing the modification.

The reason for this is that it should make it much easier to
debug issues which arise if we can tell which node was the
last to modify a particular metadata block.

Since the field is updated before the block is written into
the journal, each journal should only contain metadata which
is tagged with its own journal id. The one exception to this
is the journal header block, which might have a different node's
id in it, if that journal was recovered by another node in the
cluster.

Thus each journal will contain a record of which nodes recovered
it, via the journal header.

The other field in the metadata header could potentially be
used to hold information about what kind of operation was
performed, but for the time being we just zero it on each
transaction so that if we use it for that in future, we'll
know that the information (where it exists) is reliable.

I did consider using the other field to hold the journal
sequence number, however since in GFS2's journaling we write
the modified data into the journal and not the original
data, this gives no information as to what action caused the
modification, so I think we can probably come up with a better
use for those 64 bits in the future.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Locking order fix in gfs2_check_blk_state
Steven Whitehouse [Fri, 6 Nov 2009 11:10:51 +0000 (11:10 +0000)]
GFS2: Locking order fix in gfs2_check_blk_state

In some cases we already have the rindex lock when
we enter this function.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Remove dirent_first() function
Steven Whitehouse [Fri, 6 Nov 2009 11:06:37 +0000 (11:06 +0000)]
GFS2: Remove dirent_first() function

This function only had one caller left, and that caller only
called it for leaf blocks, hence one branch of the "if" was
never taken. In addition the call to get_left had already
verified the metadata type, so the function can be reduced
to a single line of code in its caller.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Display nobarrier option in /proc/mounts
Steven Whitehouse [Fri, 30 Oct 2009 10:48:53 +0000 (10:48 +0000)]
GFS2: Display nobarrier option in /proc/mounts

Since the default is barriers on, this only displays the
nobarrier option when that is active.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: add barrier/nobarrier mount options
Christoph Hellwig [Fri, 30 Oct 2009 07:03:27 +0000 (08:03 +0100)]
GFS2: add barrier/nobarrier mount options

Currently gfs2 issues barrier unconditionally.  There are various reasons
to disable them, be that just for testing or for stupid devices flushing
large battert backed caches.  Add a nobarrier option that matches xfs and
btrfs for this.  Also add a symmetric barrier option to turn it back on
at remount time.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: remove division from new statfs code
Benjamin Marzinski [Mon, 26 Oct 2009 18:29:47 +0000 (13:29 -0500)]
GFS2: remove division from new statfs code

It's not necessary to do any 64bit division for the statfs sync code, so
remove it.

Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Improve statfs and quota usability
Benjamin Marzinski [Tue, 20 Oct 2009 07:39:44 +0000 (02:39 -0500)]
GFS2: Improve statfs and quota usability

GFS2 now has three new mount options, statfs_quantum, quota_quantum and
statfs_percent.  statfs_quantum and quota_quantum simply allow you to
set the tunables of the same name.  Setting setting statfs_quantum to 0
will also turn on the statfs_slow tunable.  statfs_percent accepts an
integer between 0 and 100.  Numbers between 1 and 100 will cause GFS2 to
do any early sync when the local number of blocks free changes by at
least statfs_percent from the totoal number of blocks free.  Setting
statfs_percent to 0 disables this.

Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Use dquot_send_warning()
Steven Whitehouse [Mon, 28 Sep 2009 11:49:15 +0000 (12:49 +0100)]
GFS2: Use dquot_send_warning()

This adds support to GFS2 to send quota warnings via netlink.
Also it removes a stray \r which was left over from when the
code used to print warnings on the console.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoVFS: Export dquot_send_warning
Steven Whitehouse [Mon, 28 Sep 2009 11:35:17 +0000 (12:35 +0100)]
VFS: Export dquot_send_warning

Sending a message to userspace in a generic format to warn
of events (e.g. quota exceeded) in the quota subsystem is
a generically useful feature. This patch makes some minor
changes to the send_message function from dquot.c renaming
it quota_send_message, moving it to quota.c and exporting it
for use by filesystems which do not use the dquot code.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Add set_xquota support
Steven Whitehouse [Wed, 23 Sep 2009 12:50:49 +0000 (13:50 +0100)]
GFS2: Add set_xquota support

This patch adds the ability to set GFS2 quota limit and
warning levels via the XFS quota API.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Add get_xquota support
Steven Whitehouse [Mon, 28 Sep 2009 10:52:16 +0000 (11:52 +0100)]
GFS2: Add get_xquota support

This adds support for viewing the current GFS2 quota settings
via the XFS quota API. The setting of quotas will be addressed
in a later patch. Fields which are not supported here are left
set to zero.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Reviewed-by: Bob Peterson <rpeterso@redhat.com>
14 years agoGFS2: Clean up gfs2_adjust_quota() and do_glock()
Steven Whitehouse [Tue, 15 Sep 2009 19:42:56 +0000 (20:42 +0100)]
GFS2: Clean up gfs2_adjust_quota() and do_glock()

Both of these functions contained confusing and in one case
duplicate code. This patch adds a new check in do_glock()
so that we report -ENOENT if we are asked to sync a quota
entry which doesn't exist. Due to the previous patch this is
now reported correctly to userspace.

Also there are a few new comments, and I hope that the code
is easier to understand now.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Remove constant argument from qd_get()
Steven Whitehouse [Tue, 15 Sep 2009 15:30:38 +0000 (16:30 +0100)]
GFS2: Remove constant argument from qd_get()

This function was only ever called with the "create"
argument set to true, so we can remove it.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Remove constant argument from qdsb_get()
Steven Whitehouse [Tue, 15 Sep 2009 15:25:40 +0000 (16:25 +0100)]
GFS2: Remove constant argument from qdsb_get()

The "create" argument to qdsb_get() was only ever set to true,
so this patch removes that argument.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Add proper error reporting to quota sync via sysfs
Steven Whitehouse [Tue, 15 Sep 2009 15:20:30 +0000 (16:20 +0100)]
GFS2: Add proper error reporting to quota sync via sysfs

For some reason, the errors were not making it to userspace.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Add get_xstate quota function
Steven Whitehouse [Fri, 11 Sep 2009 14:57:27 +0000 (15:57 +0100)]
GFS2: Add get_xstate quota function

This allows querying of the quota state via the XFS quota
API.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Remove obsolete code in quota.c
Steven Whitehouse [Fri, 11 Sep 2009 14:21:56 +0000 (15:21 +0100)]
GFS2: Remove obsolete code in quota.c

There is no point in testing for GLF_DEMOTE here, we might as
well always release the glock at that point.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Hook gfs2_quota_sync into VFS via gfs2_quotactl_ops
Steven Whitehouse [Tue, 15 Sep 2009 08:59:02 +0000 (09:59 +0100)]
GFS2: Hook gfs2_quota_sync into VFS via gfs2_quotactl_ops

The plan is to add further operations to the gfs2_quotactl_ops
in future patches. The sync operation is easy, so we start with
that one.

We plan to use the XFS quota control functions because they more
closely match the GFS2 ones.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Alter arguments of gfs2_quota/statfs_sync
Steven Whitehouse [Fri, 11 Sep 2009 13:36:44 +0000 (14:36 +0100)]
GFS2: Alter arguments of gfs2_quota/statfs_sync

These two functions are altered so that gfs2_quota_sync may
in future be called directly from the VFS. The GFS2 superblock
changes to a VFS super block and there is an addition of an int
argument which is currently ignored.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoVFS: Use GFP_NOFS in posix_acl_from_xattr()
Steven Whitehouse [Tue, 29 Sep 2009 15:31:03 +0000 (16:31 +0100)]
VFS: Use GFP_NOFS in posix_acl_from_xattr()

GFS2 needs to call this from under a glock, so we need GFP_NOFS
and I suspect that other filesystems might require this too.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Add cached ACLs support
Steven Whitehouse [Tue, 29 Sep 2009 15:26:23 +0000 (16:26 +0100)]
GFS2: Add cached ACLs support

The other patches in this series have been building towards
being able to support cached ACLs like other filesystems. The
only real difference with GFS2 is that we have to invalidate
the cache when we drop a glock, but that is dealt with in earlier
patches.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Clean up ACLs
Steven Whitehouse [Fri, 2 Oct 2009 11:00:00 +0000 (12:00 +0100)]
GFS2: Clean up ACLs

To prepare for support for caching of ACLs, this cleans up the GFS2
ACL support by pushing the xattr code back into xattr.c and changing
the acl_get function into one which only returns ACLs so that we
can drop the caching function into it shortly.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Use gfs2_set_mode() instead of munge_mode()
Steven Whitehouse [Tue, 29 Sep 2009 11:40:19 +0000 (12:40 +0100)]
GFS2: Use gfs2_set_mode() instead of munge_mode()

These two functions do the same thing, so lets only use
one of them.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoGFS2: Use forget_all_cached_acls()
Steven Whitehouse [Fri, 2 Oct 2009 10:54:39 +0000 (11:54 +0100)]
GFS2: Use forget_all_cached_acls()

Invalidate all the cached ACLs when we drop the glock.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agoVFS: Add forget_all_cached_acls()
Steven Whitehouse [Tue, 29 Sep 2009 11:27:23 +0000 (12:27 +0100)]
VFS: Add forget_all_cached_acls()

This is required for cluster filesystems which want to use
cached ACLs so that they can invalidate the cache when
required.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Alexander Viro <aviro@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
14 years agoGFS2: Fix up system xattrs
Steven Whitehouse [Fri, 2 Oct 2009 10:50:54 +0000 (11:50 +0100)]
GFS2: Fix up system xattrs

This code has been shamelessly stolen from XFS at the suggestion
of Christoph Hellwig. I've not added support for cached ACLs so
far... watch for that in a later patch, although this is designed
in such a way that they should be easy to add.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
14 years agoGFS2: Fix -o meta mounts for subsequent mounts (i.e. all but the first one)
Steven Whitehouse [Mon, 28 Sep 2009 09:30:49 +0000 (10:30 +0100)]
GFS2: Fix -o meta mounts for subsequent mounts (i.e. all but the first one)

We have a long term plan to use the "-o meta" flag to GFS2 mounts to
access the alternate root which is used to store metadata for a GFS2
filesystem. This will allow us to eventually remove support for the
gfs2meta filesystem type (which is in any case just a "front end" to
the gfs2 filesystem type with the meta/master root).

Currently the "-o meta" option is only taken into account on the
initial mount of the filesystem. Subsequent mounts of the same
filesystem (i.e. on the same device) result in basically the same
as bind mounting the root of the original mount.

This patch changes that by using what is more or less a copy
of get_sb_bdev() and extending it so that it will take into
account the alternate root in all cases. The main difference
is that we have to parse the mount options a bit earlier. We can
then use them to select the appropriate root towards the end of
the function.

In addition this also fixes a bug where it was possible (but certainly
not desirable) to set different ro/rw options for the meta root
when mounted via the gfs2meta fs compared with the original mount.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Cc: Alexander Viro <aviro@redhat.com>
14 years agoGFS2: Fix potential race in glock code
Steven Whitehouse [Tue, 22 Sep 2009 09:56:16 +0000 (10:56 +0100)]
GFS2: Fix potential race in glock code

We need to be careful of the ordering between clearing the
GLF_LOCK bit and scheduling the workqueue.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
14 years agomutex: Fix missing conditions to build mutex_spin_on_owner()
Frederic Weisbecker [Wed, 2 Dec 2009 19:49:17 +0000 (20:49 +0100)]
mutex: Fix missing conditions to build mutex_spin_on_owner()

We don't need to build mutex_spin_on_owner() if we have
CONFIG_DEBUG_MUTEXES or CONFIG_HAVE_DEFAULT_NO_SPIN_MUTEXES as
it won't be used under such configs.

Use CONFIG_MUTEX_SPIN_ON_OWNER as it gathers all the necessary
checks before building it.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1259783357-8542-2-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
14 years agomutex: Better control mutex adaptive spinning config
Frederic Weisbecker [Wed, 2 Dec 2009 19:49:16 +0000 (20:49 +0100)]
mutex: Better control mutex adaptive spinning config

Introduce CONFIG_MUTEX_SPIN_ON_OWNER so that we can centralize
in a single place the conditions that determine its definition
and use.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1259783357-8542-1-git-send-regression-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
14 years agox86, Calgary IOMMU quirk: Find nearest matching Calgary while walking up the PCI...
Darrick J. Wong [Wed, 2 Dec 2009 23:05:56 +0000 (15:05 -0800)]
x86, Calgary IOMMU quirk: Find nearest matching Calgary while walking up the PCI tree

On a multi-node x3950M2 system, there's a slight oddity in the
PCI device tree for all secondary nodes:

 30:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev e1)
  \-33:00.0 PCI bridge: IBM CalIOC2 PCI-E Root Port (rev 01)
     \-34:00.0 RAID bus controller: LSI Logic / Symbios Logic MegaRAID SAS 1078 (rev 04)

...as compared to the primary node:

 00:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev e1)
  \-01:00.0 VGA compatible controller: ATI Technologies Inc ES1000 (rev 02)
 03:00.0 PCI bridge: IBM CalIOC2 PCI-E Root Port (rev 01)
  \-04:00.0 RAID bus controller: LSI Logic / Symbios Logic MegaRAID SAS 1078 (rev 04)

In both nodes, the LSI RAID controller hangs off a CalIOC2
device, but on the secondary nodes, the BIOS hides the VGA
device and substitutes the device tree ending with the disk
controller.

It would seem that Calgary devices don't necessarily appear at
the top of the PCI tree, which means that the current code to
find the Calgary IOMMU that goes with a particular device is
buggy.

Rather than walk all the way to the top of the PCI
device tree and try to match bus number with Calgary descriptor,
the code needs to examine each parent of the particular device;
if it encounters a Calgary with a matching bus number, simply
use that.

Otherwise, we BUG() when the bus number of the Calgary doesn't
match the bus number of whatever's at the top of the device tree.

Extra note: This patch appears to work correctly for the x3950
that came before the x3950 M2.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Acked-by: Muli Ben-Yehuda <muli@il.ibm.com>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Joerg Roedel <joerg.roedel@amd.com>
Cc: Yinghai Lu <yhlu.kernel@gmail.com>
Cc: Jon D. Mason <jdmason@kudzu.us>
Cc: Corinna Schultz <coschult@us.ibm.com>
Cc: <stable@kernel.org>
LKML-Reference: <20091202230556.GG10295@tux1.beaverton.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agorcu: Make RCU's CPU-stall detector be default
Paul E. McKenney [Wed, 2 Dec 2009 20:10:16 +0000 (12:10 -0800)]
rcu: Make RCU's CPU-stall detector be default

The RCU_CPU_STALL_DETECTOR costs almost nothing and has located
some bugs that might otherwise have been difficult to track
down.  Make it be default for the TREE RCU implementations.

The vmlinux size impact is limited (on 64-bit x86 defconfig):

   text    data     bss     dec     hex filename
   8440248 1260076  995588 10695912  a334e8 vmlinux.before
   8440774 1260060  995588 10696422  a336e6 vmlinux.after

+526 bytes - acceptable default cost.

For RAM starved systems, TINY_RCU does not support CPU-stall detection
and is much smaller, but then again it is a uniprocessor...

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: dipankar@in.ibm.com
Cc: mathieu.desnoyers@polymtl.ca
Cc: josh@joshtriplett.org
Cc: dvhltc@us.ibm.com
Cc: niv@us.ibm.com
Cc: peterz@infradead.org
Cc: rostedt@goodmis.org
Cc: Valdis.Kletnieks@vt.edu
Cc: dhowells@redhat.com
LKML-Reference: <12597846162906-git-send-email->
[ v2: added image size calculations to the changelog ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
14 years agorcu: Add expedited grace-period support for preemptible RCU
Paul E. McKenney [Wed, 2 Dec 2009 20:10:15 +0000 (12:10 -0800)]
rcu: Add expedited grace-period support for preemptible RCU

Implement an synchronize_rcu_expedited() for preemptible RCU
that actually is expedited.  This uses
synchronize_sched_expedited() to force all threads currently
running in a preemptible-RCU read-side critical section onto the
appropriate ->blocked_tasks[] list, then takes a snapshot of all
of these lists and waits for them to drain.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: laijs@cn.fujitsu.com
Cc: dipankar@in.ibm.com
Cc: mathieu.desnoyers@polymtl.ca
Cc: josh@joshtriplett.org
Cc: dvhltc@us.ibm.com
Cc: niv@us.ibm.com
Cc: peterz@infradead.org
Cc: rostedt@goodmis.org
Cc: Valdis.Kletnieks@vt.edu
Cc: dhowells@redhat.com
LKML-Reference: <1259784616158-git-send-email->
Signed-off-by: Ingo Molnar <mingo@elte.hu>