Conflicts:
	fs/btrfs/extent-tree.c
	fs/btrfs/free-space-cache.c
	fs/btrfs/inode.c
	fs/btrfs/tree-log.c
Signed-off-by: Chris Mason <chris.mason@oracle.com>

Signed-off-by: Chris Mason <chris.mason@oracle.com>

Conflicts:
	fs/btrfs/inode.c
	fs/btrfs/ioctl.c
	fs/btrfs/transaction.c
Signed-off-by: Chris Mason <chris.mason@oracle.com>

Changelog V5 -> V6:
- Fix oom when the memory load is high, by storing the delayed nodes into the
  root's radix tree, and letting btrfs inodes go.

Changelog V4 -> V5:
- Fix the race on adding the delayed node to the inode, which is spotted by
  Chris Mason.
- Merge Chris Mason's incremental patch into this patch.
- Fix deadlock between readdir() and memory fault, which is reported by
  Itaru Kitayama.

Changelog V3 -> V4:
- Fix nested lock, which is reported by Itaru Kitayama, by updating space cache
  inode in time.

Changelog V2 -> V3:
- Fix the race between the delayed worker and the task which does delayed items
  balance, which is reported by Tsutomu Itoh.
- Modify the patch address David Sterba's comment.
- Fix the bug of the cpu recursion spinlock, reported by Chris Mason

Changelog V1 -> V2:
- break up the global rb-tree, use a list to manage the delayed nodes,
  which is created for every directory and file, and used to manage the
  delayed directory name index items and the delayed inode item.
- introduce a worker to deal with the delayed nodes.

Compare with Ext3/4, the performance of file creation and deletion on btrfs
is very poor. the reason is that btrfs must do a lot of b+ tree insertions,
such as inode item, directory name item, directory name index and so on.

If we can do some delayed b+ tree insertion or deletion, we can improve the
performance, so we made this patch which implemented delayed directory name
index insertion/deletion and delayed inode update.

Implementation:
- introduce a delayed root object into the filesystem, that use two lists to
  manage the delayed nodes which are created for every file/directory.
  One is used to manage all the delayed nodes that have delayed items. And the
  other is used to manage the delayed nodes which is waiting to be dealt with
  by the work thread.
- Every delayed node has two rb-tree, one is used to manage the directory name
  index which is going to be inserted into b+ tree, and the other is used to
  manage the directory name index which is going to be deleted from b+ tree.
- introduce a worker to deal with the delayed operation. This worker is used
  to deal with the works of the delayed directory name index items insertion
  and deletion and the delayed inode update.
  When the delayed items is beyond the lower limit, we create works for some
  delayed nodes and insert them into the work queue of the worker, and then
  go back.
  When the delayed items is beyond the upper bound, we create works for all
  the delayed nodes that haven't been dealt with, and insert them into the work
  queue of the worker, and then wait for that the untreated items is below some
  threshold value.
- When we want to insert a directory name index into b+ tree, we just add the
  information into the delayed inserting rb-tree.
  And then we check the number of the delayed items and do delayed items
  balance. (The balance policy is above.)
- When we want to delete a directory name index from the b+ tree, we search it
  in the inserting rb-tree at first. If we look it up, just drop it. If not,
  add the key of it into the delayed deleting rb-tree.
  Similar to the delayed inserting rb-tree, we also check the number of the
  delayed items and do delayed items balance.
  (The same to inserting manipulation)
- When we want to update the metadata of some inode, we cached the data of the
  inode into the delayed node. the worker will flush it into the b+ tree after
  dealing with the delayed insertion and deletion.
- We will move the delayed node to the tail of the list after we access the
  delayed node, By this way, we can cache more delayed items and merge more
  inode updates.
- If we want to commit transaction, we will deal with all the delayed node.
- the delayed node will be freed when we free the btrfs inode.
- Before we log the inode items, we commit all the directory name index items
  and the delayed inode update.

I did a quick test by the benchmark tool[1] and found we can improve the
performance of file creation by ~15%, and file deletion by ~20%.

Before applying this patch:
Create files:
        Total files: 50000
        Total time: 1.096108
        Average time: 0.000022
Delete files:
        Total files: 50000
        Total time: 1.510403
        Average time: 0.000030

After applying this patch:
Create files:
        Total files: 50000
        Total time: 0.932899
        Average time: 0.000019
Delete files:
        Total files: 50000
        Total time: 1.215732
        Average time: 0.000024

[1] http://marc.info/?l=linux-btrfs&m=128212635122920&q=p3

Many thanks for Kitayama-san's help!
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Reviewed-by: David Sterba <dave@jikos.cz>
Tested-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Tested-by: Itaru Kitayama <kitayama@cl.bb4u.ne.jp>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

Conflicts:
	fs/btrfs/free-space-cache.c
Signed-off-by: Chris Mason <chris.mason@oracle.com>

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2:
  configfs: Fix race between configfs_readdir() and configfs_d_iput()
  configfs: Don't try to d_delete() negative dentries.
  ocfs2/dlm: Target node death during resource migration leads to thread spin
  ocfs2: Skip mount recovery for hard-ro mounts
  ocfs2/cluster: Heartbeat mismatch message improved
  ocfs2/cluster: Increase the live threshold for global heartbeat
  ocfs2/dlm: Use negotiated o2dlm protocol version
  ocfs2: skip existing hole when removing the last extent_rec in punching-hole codes.
  ocfs2: Initialize data_ac (might be used uninitialized)

* 'devicetree/merge' of git://git.secretlab.ca/git/linux-2.6:
  drivercore: revert addition of of_match to struct device
  of: fix race when matching drivers

* 'upstream' of git://git.linux-mips.org/pub/scm/upstream-linus:
  MIPS: Kludge IP27 build for 2.6.39.
  MIPS: AR7: Fix GPIO register size for Titan variant.
  MIPS: Fix duplicate invocation of notify_die.
  MIPS: RB532: Fix iomap resource size miscalculation.

Commit b826291c, "drivercore/dt: add a match table pointer to struct
device" added an of_match pointer to struct device to cache the
of_match_table entry discovered at driver match time.  This was unsafe
because matching is not an atomic operation with probing a driver.  If
two or more drivers are attempted to be matched to a driver at the
same time, then the cached matching entry pointer could get
overwritten.

This patch reverts the of_match cache pointer and reworks all users to
call of_match_device() directly instead.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>

If two drivers are probing devices at the same time, both will write
their match table result to the dev->of_match cache at the same time.

Only write the result if the device matches.

In a thread titled "SBus devices sometimes detected, sometimes not",
Meelis reported his SBus hme was not detected about 50% of the time.
From the debug suggested by Grant it was obvious another driver matched
some devices between the call to match the hme and the hme discovery
failling.
Reported-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: Milton Miller <miltonm@bga.com>
[grant.likely: modified to only call of_match_device() once]
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  block: don't delay blk_run_queue_async
  scsi: remove performance regression due to async queue run
  blk-throttle: Use task_subsys_state() to determine a task's blkio_cgroup
  block: rescan partitions on invalidated devices on -ENOMEDIA too
  cdrom: always check_disk_change() on open
  block: unexport DISK_EVENT_MEDIA_CHANGE for legacy/fringe drivers

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

The 'size' variable contains the correct register size for both AR7
and Titan, but we never used it to ioremap the correct register size.
This problem only shows up on Titan.

[ralf@linux-mips.org: Fixed the fix.  The original patch as in patchwork
recognizes the problem correctly then fails to fix it ...]
Reported-by: Alexander Clouter <alex@digriz.org.uk>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Patchwork: https://patchwork.linux-mips.org/patch/2380/Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

Initial patch by Yury Polyanskiy <ypolyans@princeton.edu>.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Patchwork: https://patchwork.linux-mips.org/patch/2373/

This is the MIPS portion of Joe Perches <joe@perches.com>'s
https://patchwork.linux-mips.org/patch/2172/ which seems to have been
lost in time and space.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

configfs_readdir() will use the existing inode numbers of inodes in the
dcache, but it makes them up for attribute files that aren't currently
instantiated.  There is a race where a closing attribute file can be
tearing down at the same time as configfs_readdir() is trying to get its
inode number.

We want to get the inode number of open attribute files, because they
should match while instantiated.  We can't lock down the transition
where dentry->d_inode is set to NULL, so we just check for NULL there.
We can, however, ensure that an inode we find isn't iput() in
configfs_d_iput() until after we've accessed it.
Signed-off-by: Joel Becker <jlbec@evilplan.org>

When configfs is faking mkdir() on its subsystem or default group
objects, it starts by adding a negative dentry.  It then tries to
instantiate the group.  If that should fail, it must clean up after
itself.

I was using d_delete() here, but configfs_attach_group() promises to
return an empty dentry on error.  d_delete() explodes with the entry
dentry.  Let's try d_drop() instead.  The unhashing is what we want for
our dentry.
Signed-off-by: Joel Becker <jlbec@evilplan.org>

Let's check a scenario:
1. blk_delay_queue(q, SCSI_QUEUE_DELAY);
2. blk_run_queue_async();
the second one will became a noop, because q->delay_work already has
WORK_STRUCT_PENDING_BIT set, so the delayed work will still run after
SCSI_QUEUE_DELAY. But blk_run_queue_async actually hopes the delayed
work runs immediately.

Fix this by doing a cancel on potentially pending delayed work
before queuing an immediate run of the workqueue.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

* 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6:
  [media] V4L: soc-camera: regression fix: calculate .sizeimage in soc_camera.c
  [media] v4l2-subdev: fix broken subdev control enumeration
  [media] Fix cx88 remote control input
  [media] v4l: Release module if subdev registration fails

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, AMD: Fix ARAT feature setting again
  Revert "x86, AMD: Fix APIC timer erratum 400 affecting K8 Rev.A-E processors"
  x86, apic: Fix spurious error interrupts triggering on all non-boot APs
  x86, mce, AMD: Fix leaving freed data in a list
  x86: Fix UV BAU for non-consecutive nasids
  x86, UV: Fix NMI handler for UV platforms

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

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  perf evlist: Fix per thread mmap setup
  perf tools: Honour the cpu list parameter when also monitoring a thread list
  kprobes, x86: Disable irqs during optimized callback

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  cifs: fix cifsConvertToUCS() for the mapchars case
  cifs: add fallback in is_path_accessible for old servers

Provide a stub for proc_mkdir_mode() when CONFIG_PROC_FS is not
enabled, just like the stub for proc_mkdir().

Fixes this linux-next build error:

  drivers/net/wireless/airo.c:4504: error: implicit declaration of function 'proc_mkdir_mode'
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "John W. Linville" <linville@tuxdriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

os_dump_core() uses abort() to terminate UML in case of an fatal error.

glibc's abort() calls raise(SIGABRT) which makes use of tgkill().
tgkill() has no effect within UML's kernel threads because they are not
pthreads.  As fallback abort() executes an invalid instruction to
terminate the process.  Therefore UML gets killed by SIGSEGV and leaves a
ugly log entry in the host's kernel ring buffer.

To get rid of this we use our own abort routine.
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ZONE_CONGESTED should be a state of global memory reclaim.  If not, a busy
memcg sets this and give unnecessary throttoling in wait_iff_congested()
against memory recalim in other contexts.  This makes system performance
bad.

I'll think about "memcg is congested!" flag is required or not, later.
But this fix is required first.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Acked-by: Ying Han <yinghan@google.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Johannes Weiner <jweiner@redhat.com>
Cc: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Adding the necessary MODULE_DEVICE_TABLE() information allows the driver
to be automatically loaded by udev.
Signed-off-by: Axel Lin <axel.lin@gmail.com>
Cc: Shreshtha Kumar SAHU <shreshthakumar.sahu@stericsson.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Fix switch initialization to ensure that all switches have default routing
disabled.  This guarantees that no unexpected RapidIO packets arrive to
the default port set by reset and there is no default routing destination
until it is properly configured by software.

This update also unifies handling of unmapped destinations by tsi57x, IDT
Gen1 and IDT Gen2 switches.
Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Li Yang <leoli@freescale.com>
Cc: Thomas Moll <thomas.moll@sysgo.com>
Cc: <stable@kernel.org>		[2.6.37+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

As Metze pointed out, commit 84cdf74e broke mapchars option:

    Commit "cifs: fix unaligned accesses in cifsConvertToUCS"
    (84cdf74e) does multiple steps
    in just one commit (moving the function and changing it without
    testing).

    put_unaligned_le16(temp, &target[j]); is never called for any
    codepoint the goes via the 'default' switch statement. As a result
    we put just zero (or maybe uninitialized) bytes into the target
    buffer.

His proposed patch looks correct, but doesn't apply to the current head
of the tree. This patch should also fix it.

Cc: <stable@kernel.org> # .38.x: 581ade4d: cifs: clean up various nits in unicode routines (try #2)
Reported-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>

The is_path_accessible check uses a QPathInfo call, which isn't
supported by ancient win9x era servers. Fall back to an older
SMBQueryInfo call if it fails with the magic error codes.

Cc: stable@kernel.org
Reported-and-Tested-by: Sandro Bonazzola <sandro.bonazzola@gmail.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>

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

* 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  tick: Clear broadcast active bit when switching to oneshot
  rtc: mc13xxx: Don't call rtc_device_register while holding lock
  rtc: rp5c01: Initialize drvdata before registering device
  rtc: pcap: Initialize drvdata before registering device
  rtc: msm6242: Initialize drvdata before registering device
  rtc: max8998: Initialize drvdata before registering device
  rtc: max8925: Initialize drvdata before registering device
  rtc: m41t80: Initialize clientdata before registering device
  rtc: ds1286: Initialize drvdata before registering device
  rtc: ep93xx: Initialize drvdata before registering device
  rtc: davinci: Initialize drvdata before registering device
  rtc: mxc: Initialize drvdata before registering device
  clocksource: Install completely before selecting

Trying to enable the local APIC timer on early K8 revisions
uncovers a number of other issues with it, in conjunction with
the C1E enter path on AMD. Fixing those causes much more churn
and troubles than the benefit of using that timer brings so
don't enable it on K8 at all, falling back to the original
functionality the kernel had wrt to that.
Reported-and-bisected-by: Nick Bowler <nbowler@elliptictech.com>
Cc: Boris Ostrovsky <Boris.Ostrovsky@amd.com>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Greg Kroah-Hartman <greg@kroah.com>
Cc: Hans Rosenfeld <hans.rosenfeld@amd.com>
Cc: Nick Bowler <nbowler@elliptictech.com>
Cc: Joerg-Volker-Peetz <jvpeetz@web.de>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1305636919-31165-3-git-send-email-bp@amd64.orgSigned-off-by: Ingo Molnar <mingo@elte.hu>

This reverts commit e20a2d20, as it crashes
certain boxes with specific AMD CPU models.

Moving the lower endpoint of the Erratum 400 check to accomodate
earlier K8 revisions (A-E) opens a can of worms which is simply
not worth to fix properly by tweaking the errata checking
framework:

* missing IntPenging MSR on revisions < CG cause #GP:

http://marc.info/?l=linux-kernel&m=130541471818831

* makes earlier revisions use the LAPIC timer instead of the C1E
idle routine which switches to HPET, thus not waking up in
deeper C-states:

http://lkml.org/lkml/2011/4/24/20

Therefore, leave the original boundary starting with K8-revF.
Signed-off-by: Ingo Molnar <mingo@elte.hu>

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>

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  net: Change netdev_fix_features messages loglevel
  vmxnet3: Fix inconsistent LRO state after initialization
  sfc: Fix oops in register dump after mapping change
  IPVS: fix netns if reading ip_vs_* procfs entries
  bridge: fix forwarding of IPv6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc:
  Revert "mmc: fix a race between card-detect rescan and clock-gate work instances"

Fix new kernel-doc warning in mm/page_alloc.c:

  Warning(mm/page_alloc.c:2370): No description found for parameter 'nid'
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

During pci remove/rescan testing found:

  pci 0000:c0:03.0: PCI bridge to [bus c4-c9]
  pci 0000:c0:03.0:   bridge window [io  0x1000-0x0fff]
  pci 0000:c0:03.0:   bridge window [mem 0xf0000000-0xf00fffff]
  pci 0000:c0:03.0:   bridge window [mem 0xfc180000000-0xfc197ffffff 64bit pref]
  pci 0000:c0:03.0: device not available (can't reserve [io  0x1000-0x0fff])
  pci 0000:c0:03.0: Error enabling bridge (-22), continuing
  pci 0000:c0:03.0: enabling bus mastering
  pci 0000:c0:03.0: setting latency timer to 64
  pcieport 0000:c0:03.0: device not available (can't reserve [io  0x1000-0x0fff])
  pcieport: probe of 0000:c0:03.0 failed with error -22

This bug was caused by commit c8adf9a3 ("PCI: pre-allocate
additional resources to devices only after successful allocation of
essential resources.")

After that commit, pci_hotplug_io_size is changed to additional_io_size
from minium size.  So it will not go through resource_size(res) != 0
path, and will not be reset.

The root cause is: pci_bridge_check_ranges will set RESOURCE_IO flag for
pci bridge, and later if children do not need IO resource.  those bridge
resources will not need to be allocated.  but flags is still there.
that will confuse the the pci_enable_bridges later.

related code:

   static void assign_requested_resources_sorted(struct resource_list *head,
                                    struct resource_list_x *fail_head)
   {
           struct resource *res;
           struct resource_list *list;
           int idx;

           for (list = head->next; list; list = list->next) {
                   res = list->res;
                   idx = res - &list->dev->resource[0];
                   if (resource_size(res) && pci_assign_resource(list->dev, idx)) {
   ...
                           reset_resource(res);
                   }
           }
   }

At last, We have to clear the flags in pbus_size_mem/io when requested
size == 0 and !add_head.  becasue this case it will not go through
adjust_resources_sorted().

Just make size1 = size0 when !add_head. it will make flags get cleared.

At the same time when requested size == 0, add_size != 0, will still
have in head and add_list.  because we do not clear the flags for it.

After this, we will get right result:

  pci 0000:c0:03.0: PCI bridge to [bus c4-c9]
  pci 0000:c0:03.0:   bridge window [io  disabled]
  pci 0000:c0:03.0:   bridge window [mem 0xf0000000-0xf00fffff]
  pci 0000:c0:03.0:   bridge window [mem 0xfc180000000-0xfc197ffffff 64bit pref]
  pci 0000:c0:03.0: enabling bus mastering
  pci 0000:c0:03.0: setting latency timer to 64
  pcieport 0000:c0:03.0: setting latency timer to 64
  pcieport 0000:c0:03.0: irq 160 for MSI/MSI-X
  pcieport 0000:c0:03.0: Signaling PME through PCIe PME interrupt
  pci 0000:c4:00.0: Signaling PME through PCIe PME interrupt
  pcie_pme 0000:c0:03.0:pcie01: service driver pcie_pme loaded
  aer 0000:c0:03.0:pcie02: service driver aer loaded
  pciehp 0000:c0:03.0:pcie04: Hotplug Controller:

v3: more simple fix. also fix one typo in pbus_size_mem
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Reviewed-by: Ram Pai <linuxram@us.ibm.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The first cpu which switches from periodic to oneshot mode switches
also the broadcast device into oneshot mode. The broadcast device
serves as a backup for per cpu timers which stop in deeper
C-states. To avoid starvation of the cpus which might be in idle and
depend on broadcast mode it marks the other cpus as broadcast active
and sets the brodcast expiry value of those cpus to the next tick.

The oneshot mode broadcast bit for the other cpus is sticky and gets
only cleared when those cpus exit idle. If a cpu was not idle while
the bit got set in consequence the bit prevents that the broadcast
device is armed on behalf of that cpu when it enters idle for the
first time after it switched to oneshot mode.

In most cases that goes unnoticed as one of the other cpus has usually
a timer pending which keeps the broadcast device armed with a short
timeout. Now if the only cpu which has a short timer active has the
bit set then the broadcast device will not be armed on behalf of that
cpu and will fire way after the expected timer expiry. In the case of
Christians bug report it took ~145 seconds which is about half of the
wrap around time of HPET (the limit for that device) due to the fact
that all other cpus had no timers armed which expired before the 145
seconds timeframe.

The solution is simply to clear the broadcast active bit
unconditionally when a cpu switches to oneshot mode after the first
cpu switched the broadcast device over. It's not idle at that point
otherwise it would not be executing that code.

[ I fundamentally hate that broadcast crap. Why the heck thought some
  folks that when going into deep idle it's a brilliant concept to
  switch off the last device which brings the cpu back from that
  state? ]

Thanks to Christian for providing all the valuable debug information!
Reported-and-tested-by: Christian Hoffmann <email@christianhoffmann.info>
Cc: John Stultz <johnstul@us.ibm.com>
Link: http://lkml.kernel.org/r/%3Calpine.LFD.2.02.1105161105170.3078%40ionos%3E
Cc: stable@kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

Those reduced to DEBUG can possibly be triggered by unprivileged processes
and are nothing exceptional. Illegal checksum combinations can only be
caused by driver bug, so promote those messages to WARN.

Since GSO without SG will now only cause DEBUG message from
netdev_fix_features(), remove the workaround from register_netdevice().
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>