Commit c21e6beb removed our queue request_fn re-enter
protection, and defaulted to always running the queues from
kblockd to be safe. This was a known potential slow down,
but should be safe.

Unfortunately this is causing big performance regressions for
some, so we need to improve this logic. Looking into the details
of the re-enter, the real issue is on requeue of requests.

Requeue of requests upon seeing a BUSY condition from the device
ends up re-running the queue, causing traces like this:

scsi_request_fn()
        scsi_dispatch_cmd()
                scsi_queue_insert()
                        __scsi_queue_insert()
                                scsi_run_queue()
					scsi_request_fn()
						...

potentially causing the issue we want to avoid. So special
case the requeue re-run of the queue, but improve it to offload
the entire run of local queue and starved queue from a single
workqueue callback. This is a lot better than potentially
kicking off a workqueue run for each device seen.

This also fixes the issue of the local device going into recursion,
since the above mentioned commit never moved that queue run out
of line.
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

Currentlly we first map the task to cgroup and then cgroup to
blkio_cgroup. There is a more direct way to get to blkio_cgroup
from task using task_subsys_state(). Use that.

The real reason for the fix is that it also avoids a race in generic
cgroup code. During remount/umount rebind_subsystems() is called and
it can do following with and rcu protection.

cgrp->subsys[i] = NULL;

That means if somebody got hold of cgroup under rcu and then it tried
to do cgroup->subsys[] to get to blkio_cgroup, it would get NULL which
is wrong. I was running into this race condition with ltp running on a
upstream derived kernel and that lead to crash.

So ideally we should also fix cgroup generic code to wait for rcu
grace period before setting pointer to NULL. Li Zefan is not very keen
on introducing synchronize_wait() as he thinks it will slow
down moun/remount/umount operations.

So for the time being atleast fix the kernel crash by taking a more
direct route to blkio_cgroup.

One tester had reported a crash while running LTP on a derived kernel
and with this fix crash is no more seen while the test has been
running for over 6 days.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Reviewed-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

__blkdev_get() doesn't rescan partitions if disk->fops->open() fails,
which leads to ghost partition devices lingering after medimum removal
is known to both the kernel and userland.  The behavior also creates a
subtle inconsistency where O_NONBLOCK open, which doesn't fail even if
there's no medium, clears the ghots partitions, which is exploited to
work around the problem from userland.

Fix it by updating __blkdev_get() to issue partition rescan after
-ENOMEDIA too.

This was reported in the following bz.

 https://bugzilla.kernel.org/show_bug.cgi?id=13029

Stable: 2.6.38
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: David Zeuthen <zeuthen@gmail.com>
Reported-by: Martin Pitt <martin.pitt@ubuntu.com>
Reported-by: Kay Sievers <kay.sievers@vrfy.org>
Tested-by: Kay Sievers <kay.sievers@vrfy.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: stable@kernel.org
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

cdrom_open() called check_disk_change() after the rest of open path
succeeded which leads to the following bizarre behavior.

* After media change, if the device opened without O_NONBLOCK,
  open_for_data() naturally fails with -ENOMEDIA and
  check_disk_change() is never called.  The media is known to be gone
  and the open failure makes it obvious to the userland but device
  invalidation never happens.

* But if the device is opened with O_NONBLOCK, all the checks are
  bypassed and cdrom_open() doesn't notice that the media is not there
  and check_disk_change() is called and invalidation happens.

There's nothing to be gained by avoiding calling check_disk_change()
on open failure.  Common cases end up calling check_disk_change()
anyway.  All we get is inconsistent behavior.

Fix it by moving check_disk_change() invocation to the top of
cdrom_open() so that it always gets called regardless of how the rest
of open proceeds.

Stable: 2.6.38
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Amit Shah <amit.shah@redhat.com>
Tested-by: Amit Shah <amit.shah@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

In-kernel disk event polling doesn't matter for legacy/fringe drivers
and may lead to infinite event loop if ->check_events() implementation
generates events on level condition instead of edge.

Now that block layer supports suppressing exporting unlisted events,
simply leaving disk->events cleared allows these drivers to keep the
internal revalidation behavior intact while avoiding weird
interactions with userland event handler.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  ide: unexport DISK_EVENT_MEDIA_CHANGE for ide-gd and ide-cd
  block: don't propagate unlisted DISK_EVENTs to userland
  elevator: check for ELEVATOR_INSERT_SORT_MERGE in !elvpriv case too

check_events() implementations in both ide-gd and ide-cd are
inadequate for in-kernel event polling.  Both generate media change
events continuously when certain conditions are met causing infinite
event loop between the driver and userland event handler.

As disk event now supports suppression of unlisted events, simply
de-listing DISK_EVENT_MEDIA_CHANGE from disk->events resolves the
problem.  Internal handling around media revalidation will behave the
same while userland will fall back to userland event polling after
detecting the device doesn't support disk events.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Jens Axboe <jaxboe@fusionio.com>
Acked-by: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

DISK_EVENT_MEDIA_CHANGE is used for both userland visible event and
internal event for revalidation of removeable devices.  Some legacy
drivers don't implement proper event detection and continuously
generate events under certain circumstances.  For example, ide-cd
generates media changed continuously if there's no media in the drive,
which can lead to infinite loop of events jumping back and forth
between the driver and userland event handler.

This patch updates disk event infrastructure such that it never
propagates events not listed in disk->events to userland.  Those
events are processed the same for internal purposes but uevent
generation is suppressed.

This also ensures that userland only gets events which are advertised
in the @events sysfs node lowering risk of confusion.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

The sort insert is the one that goes to the IO scheduler. With
the SORT_MERGE addition, we could bypass IO scheduler setup
but still ask the IO scheduler to insert the request. This would
cause an oops on switching IO schedulers through the sysfs
interface, unless the disk just happened to be idle while it
occured.
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  xfs: fix duplicate message output

Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, numa: Fix cpu nodemasks for NUMA emulation and CONFIG_DEBUG_PER_CPU_MAPS
  Revert "x86, NUMA: Fix fakenuma boot failure"

Change <sectors> from unsigned long long to sector_t.
This matches its source field.

  ERROR: "__udivdi3" [drivers/md/raid456.ko] undefined!
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  virtio: console: Enable call to hvc_remove() on console port remove
  virtio_pci: Prevent double-free of pci regions after device hot-unplug
  virtio: Decrement avail idx on buffer detach

* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  agp: fix arbitrary kernel memory writes
  agp: fix OOM and buffer overflow
  drm/radeon/kms: fix IH writeback on r6xx+ on big endian machines

* 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/keithp/linux-2.6:
  drm/i915: Initialise g4x watermarks for disabled pipes
  drm/i915: Sanitize the output registers after resume
  drm/i915/tv: Fix modeset flickering introduced in 7f58aabc
  drm/i915/tv: Only poll for TV connections
  drm/i915/tv: Remember the detected TV type

* git://git.infradead.org/iommu-2.6:
  intel_iommu: disable all VT-d PMRs when TXT launched
  intel-iommu: Fix get_domain_for_dev() error path
  intel-iommu: Unlink domain from iommu
  intel-iommu: Fix use after release during device attach

For some reason generic_setxattr() did not pass flags (XATTR_CREATE,
XATTR_REPLACE) to the filesystem specific helper. This caused that
setxattr(2) syscall just ignored these flags.

Fix the bug by passing flags correctly.
Signed-off-by: Jan Kara <jack@suse.cz>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This call was disabled as hot-unplugging one virtconsole port led to
another virtconsole port freezing.

Upon testing it again, this now works, so enable it.

In addition, a bug was found in qemu wherein removing a port of one type
caused the guest output from another port to stop working.  I doubt it
was just this bug that caused it (since disabling the hvc_remove() call
did allow other ports to continue working), but since it's all solved
now, we're fine with hot-unplugging of virtconsole ports.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

In the case where a virtio-console port is in use (opened by a program)
and a virtio-console device is removed, the port is kept around but all
the virtio-related state is assumed to be gone.

When the port is finally released (close() called), we call
device_destroy() on the port's device.  This results in the parent
device's structures to be freed as well.  This includes the PCI regions
for the virtio-console PCI device.

Once this is done, however, virtio_pci_release_dev() kicks in, as the
last ref to the virtio device is now gone, and attempts to do

     pci_iounmap(pci_dev, vp_dev->ioaddr);
     pci_release_regions(pci_dev);
     pci_disable_device(pci_dev);

which results in a double-free warning.

Move the code that releases regions, etc., to the virtio_pci_remove()
function, and all that's now left in release_dev is the final freeing of
the vp_dev.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

When detaching a buffer from a vq, the avail.idx value should be
decremented as well.

This was noticed by hot-unplugging a virtio console port and then
plugging in a new one on the same number (re-using the vqs which were
just 'disowned').  qemu reported

   'Guest moved used index from 0 to 256'

when any IO was attempted on the new port.

CC: stable@kernel.org
Reported-by: juzhang <juzhang@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

Intel VT-d Protected Memory Regions (PMRs) are supposed to be disabled,
on each VT-d engine, after DMA remapping is enabled on the engines.
This is because the behavior of having both enabled is not deterministic
and because, if TXT has been used to launch the kernel, the PMRs may be
programmed to cover memory regions that will be used for DMA.

Under some circumstances (certain quirks detected, lack of multiple
devices, etc.), the current code does not set up DMA remapping on some
VT-d engines.  In such cases it also skips disabling the PMRs.  This
causes failures when the kernel is launched with TXT (most often this
occurs on the graphics engine and results in colored vertical bars on
the display).

This patch detects when the kernel has been launched with TXT and then
disables the PMRs on all VT-d engines.  In some cases where the reason
that remapping is not being enabled is due to possible ACPI DMAR table
errors, the VT-d engine addresses may not be correct and thus not able
to be safely programmed even to disable PMRs.  Because part of the TXT
launch process is the verification of these addresses, it will always be
safe to disable PMRs if the TXT launch has succeeded and hence only
doing this in such cases.
Signed-off-by: Joseph Cihula <joseph.cihula@intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

The cpu<->node mappings under CONFIG_DEBUG_PER_CPU_MAPS=y
when NUMA emulation is enabled is currently broken because it does
not iterate through every emulated node and bind cpus that have
affinity to it.

NUMA emulation should bind each cpu to every local node to
accurately represent the true NUMA topology of the underlying
machine.

debug_cpumask_set_cpu() needs to be fixed at the same time so
that the debugging information that it emits shows the new
cpumask of the node being assigned when the cpu is being added
or removed.

It can now take responsibility of setting or clearing the cpu
itself to remove the need for duplicate code.

Also change its last parameter, "enable", to have the correct bool
type since it can only be true or false.

 -v2: Fix the return statements, by Kosaki Motohiro
Acked-and-Tested-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Andreas Herrmann <herrmann.der.user@googlemail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.00.1104201918470.12634@chino.kir.corp.google.comSigned-off-by: Ingo Molnar <mingo@elte.hu>

Andreas Herrmann reported that 7d6b4670 ("x86, NUMA: Fix fakenuma
boot failure") causes certain physical NUMA topologies (for example
AMD Magny-Cours) to move sibling cpus to a single node when in reality
they are in separate domains.

This may result in some nodes being completely void of cpus, which
doesn't accurately represent the correct topology. The system will
boot, but will have suboptimal NUMA performance.

This commit was intended as a fix for NUMA emulation, but should
not cause a regression for real NUMA machines as a side effect.

( There will be a separate fix for the numa-debug code, which
  will not affect physical topologies. )
Reported-by: Andreas Herrmann <herrmann.der.user@googlemail.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.00.1104201918110.12634@chino.kir.corp.google.comSigned-off-by: Ingo Molnar <mingo@elte.hu>

pg_start is copied from userspace on AGPIOC_BIND and AGPIOC_UNBIND ioctl
cmds of agp_ioctl() and passed to agpioc_bind_wrap().  As said in the
comment, (pg_start + mem->page_count) may wrap in case of AGPIOC_BIND,
and it is not checked at all in case of AGPIOC_UNBIND.  As a result, user
with sufficient privileges (usually "video" group) may generate either
local DoS or privilege escalation.
Signed-off-by: Vasiliy Kulikov <segoon@openwall.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

page_count is copied from userspace.  agp_allocate_memory() tries to
check whether this number is too big, but doesn't take into account the
wrap case.  Also agp_create_user_memory() doesn't check whether
alloc_size is calculated from num_agp_pages variable without overflow.
This may lead to allocation of too small buffer with following buffer
overflow.

Another problem in agp code is not addressed in the patch - kernel memory
exhaustion (AGPIOC_RESERVE and AGPIOC_ALLOCATE ioctls).  It is not checked
whether requested pid is a pid of the caller (no check in agpioc_reserve_wrap()).
Each allocation is limited to 16KB, though, there is no per-process limit.
This might lead to OOM situation, which is not even solved in case of the
caller death by OOM killer - the memory is allocated for another (faked) process.
Signed-off-by: Vasiliy Kulikov <segoon@openwall.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

* 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/staging:
  hwmon: (max34440) Add driver documentation
  hwmon: (max16064) Add driver documentation
  hwmon: (max8688) Add driver documentation
  hwmon: (pmbus) Documentation updates
  hwmon: (smm665) Fix spelling error in driver documentation
  hwmon: (pmbus) Removed unused variable from struct pmbus_data
  hwmon: Add submitting-patches checklist to documentation

* 'for-2.6.39' of git://linux-nfs.org/~bfields/linux:
  Open with O_CREAT flag set fails to open existing files on non writable directories
  nfsd4: Fix filp leak
  nfsd4: fix struct file leak on delegation

* 'fixes' of master.kernel.org:/home/rmk/linux-2.6-arm:
  ARM: 6881/1: cputype.h uses __attribute_const__ which requires including kernel.h
  ARM: Add new syscalls

* 'stable/bug-fixes-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
  xen: mask_rw_pte: do not apply the early_ioremap checks on x86_32
  xen: do not create the extra e820 region at an addr lower than 4G

* 'for-linus' of git://neil.brown.name/md:
  md: Update documentation for sync_min and sync_max entries
  md: Cleanup after raid45->raid0 takeover
  md: Fix dev_sectors on takeover from raid0 to raid4/5
  md/raid5: remove setting of ->queue_lock

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  block: Remove the extra check in queue_requests_store
  block, blk-sysfs: Fix an err return path in blk_register_queue()
  block: remove stale kerneldoc member from __blk_run_queue()
  block: get rid of QUEUE_FLAG_REENTER
  cfq-iosched: read_lock() does not always imply rcu_read_lock()
  block: kill blk_flush_plug_list() export

Commit 957935dc ("xfs: fix xfs_debug warnings" broke the logic in
__xfs_printk(). Instead of only printing one of two possible output
strings based on whether the fs has a name or not, it outputs both.
Fix it to only output one message again.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>

An open on a NFS4 share using the O_CREAT flag on an existing file for
which we have permissions to open but contained in a directory with no
write permissions will fail with EACCES.

A tcpdump shows that the client had set the open mode to UNCHECKED which
indicates that the file should be created if it doesn't exist and
encountering an existing flag is not an error. Since in this case the
file exists and can be opened by the user, the NFS server is wrong in
attempting to check create permissions on the parent directory.

The patch adds a conditional statement to check for create permissions
only if the file doesn't exist.
Signed-off-by: Sachin S. Prabhu <sprabhu@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>

The two "is_early_ioremap_ptep" checks in mask_rw_pte are only used on
x86_64, in fact early_ioremap is not used at all to setup the initial
pagetable on x86_32.
Moreover on x86_32 the two checks are wrong because the range
pgt_buf_start..pgt_buf_end initially should be mapped RW because
the pages in the range are not pagetable pages yet and haven't been
cleared yet. Afterwards considering the pgt_buf_start..pgt_buf_end is
part of the initial mapping, xen_alloc_pte is capable of turning
the ptes RO when they become pagetable pages.

Fix the issue and improve the readability of the code providing two
different implementation of mask_rw_pte for x86_32 and x86_64.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>

Do not add the extra e820 region at a physical address lower than 4G
because it breaks e820_end_of_low_ram_pfn().

It is OK for us to move the xen_extra_mem_start up and down because this
is the index of the memory that can be ballooned in/out - it is memory
not available to the kernel during bootup.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>

linux/Documentation/md.txt is missing description for sync_min and
sync_max entries.
This patch adds description for sync_min and sync_max entries.
Signed-off-by: Roman Ovchinnikov <coolthecold@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>

Problem:
After raid4->raid0 takeover operation, another takeover operation
(e.g raid0->raid10) results "kernel oops".
Root cause:
Variables 'degraded' in mddev structure is not cleared
on raid45->raid0 takeover.

This patch reset this variable.
Signed-off-by: Krzysztof Wojcik <krzysztof.wojcik@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>

A raid0 array doesn't set 'dev_sectors' as each device might
contribute a different number of sectors.
So when converting to a RAID4 or RAID5 we need to set dev_sectors
as they need the number.
We have already verified that in fact all devices do contribute
the same number of sectors, so use that number.
Signed-off-by: NeilBrown <neilb@suse.de>

We previously needed to set ->queue_lock to match the raid5
device_lock so we could safely use queue_flag_* operations (e.g. for
plugging). which test the ->queue_lock is in fact locked.

However that need has completely gone away and is unlikely to come
back to remove this now-pointless setting.
Signed-off-by: NeilBrown <neilb@suse.de>

* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm/radeon/kms: pll tweaks for r7xx
  drm/nouveau: fix allocation of notifier object
  drm/nouveau: fix notifier memory corruption bug
  drm/nouveau: fix pinning of notifier block
  drm/nouveau: populate ttm_alloced with false, when it's not
  drm/nouveau: fix nv30 pcie boards
  drm/nouveau: split ramin_lock into two locks, one hardirq safe
  drm/radeon/kms: adjust evergreen display watermark setup
  drm/radeon/kms: add connectors even if i2c fails
  drm/radeon/kms: fix bad shift in atom iio table parser