In addition to the existing ACPI specific GPIO interface, document the new
descriptor based GPIO interface in Documentation/acpi/enumeration.txt, so
it is clear that this new interface is preferred over the ACPI specific
version.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

The ACPI GpioInt resources contain polarity field that is used to specify
whether the interrupt is active high or low. Since gpiolib supports
GPIOF_ACTIVE_LOW we can pass this information in the flags field in
acpi_find_gpio(), analogous to the DeviceTree version.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

gpiod_get_index() and gpiod_get() are now the new preferred way to request
GPIOs. Add support for finding the corresponding GPIO descriptor from ACPI
namespace.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

The new GPIO descriptor based interface is now preferred over the old
integer based one. This patch converts the ACPI GPIO helpers to use this
new interface internally. In addition to that provide compatibility
function acpi_get_gpio_by_index() that converts the returned GPIO
descriptor to an integer.

We also drop acpi_get_gpio() as it is not used anywhere outside
gpiolib-acpi and even there we use acpi_get_gpiod() instead.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

Add gpiod_get(), gpiod_get_index() and gpiod_put() functions that
provide safer management of GPIOs.

These functions put the GPIO framework in line with the conventions of
other frameworks in the kernel, and help ensure every GPIO is declared
properly and valid while it is used.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

Refactor the of_ functions of gpiolib to use the now public gpiod
interface, and export of_get_named_gpiod_flags() and
of_get_gpiod_flags() functions.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

This patch exports the gpiod_* family of API functions, a safer
alternative to the legacy GPIO interface. Differences between the gpiod
and legacy gpio APIs are:

- gpio works with integers, whereas gpiod operates on opaque handlers
  which cannot be forged or used before proper acquisition
- gpiod get/set functions are aware of the active low state of a GPIO
- gpio consumers should now include <linux/gpio/consumer.h> to access
  the new interface, whereas chips drivers will use
  <linux/gpio/driver.h>

The legacy gpio API is now built as inline functions on top of gpiod.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

Linux 3.12-rc6

Conflicts:
	drivers/gpio/gpio-lynxpoint.c

This reverts commit ee0f5b39.
The MAINTAINERS entry was not in alphabetic order.

Assign david.a.cohen@linux.intel.com as maintainer of Intel MID gpio
driver at drivers/gpio/gpio-intel-mid.c. Now do it in
alphabetic order instead.
Signed-off-by: David Cohen <david.a.cohen@linux.intel.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

Pull btrfs fix from Chris Mason:
 "Sage hit a deadlock with ceph on btrfs, and Josef tracked it down to a
  regression in our initial rc1 pull.  When doing nocow writes we were
  sometimes starting a transaction with locks held"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: release path before starting transaction in can_nocow_extent

Pull ACPI and power management fixes from Rafael Wysocki:

 - intel_pstate fix for misbehavior after system resume if sysfs
   attributes are set in a specific way before the corresponding suspend
   from Dirk Brandewie.

 - A recent intel_pstate fix has no effect if unsigned long is 32-bit,
   so fix it up to cover that case as well.

 - The s3c64xx cpufreq driver was not updated when the index field of
   struct cpufreq_frequency_table was replaced with driver_data, so
   update it now.  From Charles Keepax.

 - The Kconfig help text for ACPI_BUTTON still refers to
   /proc/acpi/event that has been dropped recently, so modify it to
   remove that reference.  From Krzysztof Mazur.

 - A Lan Tianyu's change adds a missing mutex unlock to an error code
   path in acpi_resume_power_resources().

 - Some code related to ACPI power resources, whose very purpose is
   questionable to put it lightly, turns out to cause problems to happen
   during testing on real systems, so remove it completely (we may
   revisit that in the future if there's a compelling enough reason).
   From Rafael J Wysocki and Aaron Lu.

* tag 'pm+acpi-3.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI / PM: Drop two functions that are not used any more
  ATA / ACPI: remove power dependent device handling
  cpufreq: s3c64xx: Rename index to driver_data
  ACPI / power: Drop automaitc resume of power resource dependent devices
  intel_pstate: Fix type mismatch warning
  cpufreq / intel_pstate: Fix max_perf_pct on resume
  ACPI: remove /proc/acpi/event from ACPI_BUTTON help
  ACPI / power: Release resource_lock after acpi_power_get_state() return error

The physical address of the GPIO memory address is already printed
implicitly by dev_info() as part of the device name. The virtual
address doesn't add any value. In addition, printing a resource_size_t
with %x lead to a compiler warning on some platforms.
Signed-off-by: Markus Mayer <markus.mayer@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

Pull x86 fixes from Ingo Molnar:
 "Two fixlets:

   - fix a (rare-config) build bug
   - fix a next-gen SGI/UV hw/firmware enumeration bug"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86: Update UV3 hub revision ID
  x86/microcode: Correct Kconfig dependencies

We can't be holding tree locks while we try to start a transaction, we will
deadlock.  Thanks,
Reported-by: Sage Weil <sage@inktank.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>

* acpi-fixes:
  ACPI / PM: Drop two functions that are not used any more
  ATA / ACPI: remove power dependent device handling
  ACPI / power: Drop automaitc resume of power resource dependent devices
  ACPI: remove /proc/acpi/event from ACPI_BUTTON help
  ACPI / power: Release resource_lock after acpi_power_get_state() return error

* pm-fixes:
  cpufreq: s3c64xx: Rename index to driver_data
  intel_pstate: Fix type mismatch warning
  cpufreq / intel_pstate: Fix max_perf_pct on resume

Pull CIFS fixes from Steve French:
 "Five small cifs fixes (includes fixes for: unmount hang, 2 security
  related, symlink, large file writes)"

* 'for-linus' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: ntstatus_to_dos_map[] is not terminated
  cifs: Allow LANMAN auth method for servers supporting unencapsulated authentication methods
  cifs: Fix inability to write files >2GB to SMB2/3 shares
  cifs: Avoid umount hangs with smb2 when server is unresponsive
  do not treat non-symlink reparse points as valid symlinks

Merge tag 'driver-core-3.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core

Pull driver core fix from Greg KH:
 "Here is one fix for the hotplug memory path that resolves a regression
  when removing memory that showed up in 3.12-rc1"

* tag 'driver-core-3.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  driver core: Release device_hotplug_lock when store_mem_state returns EINVAL

Pull USB fixes from Greg KH:
 "Here are some USB fixes and new device ids for 3.12-rc6

  The largest change here is a bunch of new device ids for the option
  USB serial driver for new Huawei devices.  Other than that, just some
  small bug fixes for issues that people have reported (run-time and
  build-time), nothing major"

* tag 'usb-3.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: usb_phy_gen: refine conditional declaration of usb_nop_xceiv_register
  usb: misc: usb3503: Fix compile error due to incorrect regmap depedency
  usb/chipidea: fix oops on memory allocation failure
  usb-storage: add quirk for mandatory READ_CAPACITY_16
  usb: serial: option: blacklist Olivetti Olicard200
  USB: quirks: add touchscreen that is dazzeled by remote wakeup
  Revert "usb: musb: gadget: fix otg active status flag"
  USB: quirks.c: add one device that cannot deal with suspension
  USB: serial: option: add support for Inovia SEW858 device
  USB: serial: ti_usb_3410_5052: add Abbott strip port ID to combined table as well.
  USB: support new huawei devices in option.c
  usb: musb: start musb on the udc side, too
  xhci: Fix spurious wakeups after S5 on Haswell
  xhci: fix write to USB3_PSSEN and XUSB2PRM pci config registers
  xhci: quirk for extra long delay for S4
  xhci: Don't enable/disable RWE on bus suspend/resume.

Pull serial driver fixes from Greg KH:
 "Here are two serial driver fixes for your tree.  One is a revert of a
  patch that causes a build error, the other is a fix to provide the
  correct brace placement which resolves a bug where the driver was not
  working properly"

* tag 'tty-3.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: vt8500: add missing braces
  Revert "serial: i.MX: evaluate linux,stdout-path property"

Pull char/misc driver fixes from Greg KH:
 "Here are some small iio and w1 driver fixes for 3.12-rc6.

  There is also a hyper-v fix in here, which turned out to be incorrect,
  so it was reverted.  That will probably have to wait unto 3.13-rc1 to
  get accepted as it's still being discussed"

* tag 'char-misc-3.12-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  Revert "Drivers: hv: vmbus: Fix a bug in channel rescind code"
  Drivers: hv: vmbus: Fix a bug in channel rescind code
  iio:buffer: Free active scan mask in iio_disable_all_buffers()
  iio: frequency: adf4350: add missing clk_disable_unprepare() on error in adf4350_probe()
  w1 - call request_module with w1 master mutex unlocked
  w1 - fix fops in w1_bus_notify

Pull sound fixes from Takashi Iwai:
 "All reasonably small fixes as rc6: a HD-audio mic fix, a us122l mmap
  regression fix, and kernel memory leak fix in hdsp driver.  Hopefully
  this will be the last pull request for 3.12..."

* tag 'sound-3.12' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hdsp - info leak in snd_hdsp_hwdep_ioctl()
  ALSA: us122l: Fix pcm_usb_stream mmapping regression
  ALSA: hda - Fix inverted internal mic not indicated on some machines

Pull apparmor fixes from James Morris:
 "A couple more regressions fixed"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security:
  apparmor: fix bad lock balance when introspecting policy
  apparmor: fix memleak of the profile hash

Merge tag 'iio-fixes-for-3.12c' of git://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into char-misc-linus

Jonathan writes:

Third set of IIO fixes for the 3.12 cycle.

Two little ones this time:

1) A missing clk_unprepare in adf4350.
2) A missing free of the active_scan_mask when iio_disable_all_buffers is
called during an unexpected device removal.  This leak was introduced by
the fix
a87c82e4 iio: Stop sampling when the device is removed
and hence is a regression fix.

Commit 3fa4d734 (usb: phy: rename nop_usb_xceiv => usb_phy_gen_xceiv)
changed the conditional around the declaration of usb_nop_xceiv_register
from
	#if defined(CONFIG_NOP_USB_XCEIV) ||
		(defined(CONFIG_NOP_USB_XCEIV_MODULE) && defined(MODULE))
to
	#if IS_ENABLED(CONFIG_NOP_USB_XCEIV)

While that looks the same, it is semantically different. The first expression
is true if CONFIG_NOP_USB_XCEIV is built as module and if the including
code is built as module. The second expression is true if code depending on
CONFIG_NOP_USB_XCEIV if built as module or into the kernel.

As a result, the arm:allmodconfig build fails with

arch/arm/mach-omap2/built-in.o: In function `omap3_evm_init':
arch/arm/mach-omap2/board-omap3evm.c:703: undefined reference to
	`usb_nop_xceiv_register'

Fix the problem by reverting to the old conditional.

Cc: Josh Boyer <jwboyer@redhat.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

This reverts commit 90d33f3e as it's not
the correct fix for this issue, and it causes a build warning to be
added to the kernel tree.

Cc: K. Y. Srinivasan <kys@microsoft.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Two functions defined in device_pm.c, acpi_dev_pm_add_dependent()
and acpi_dev_pm_remove_dependent(), have no callers and may be
dropped, so drop them.

Moreover, they are the only functions adding entries to and removing
entries from the power_dependent list in struct acpi_device, so drop
that list too.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Previously, we wanted SCSI devices corrsponding to ATA devices to
be runtime resumed when the power resource for those ATA device was
turned on by some other device, so we added the SCSI device to the
dependent device list of the ATA device's ACPI node.  However, this
code has no effect after commit 41863fce (ACPI / power: Drop automaitc
resume of power resource dependent devices) and the mechanism it was
supposed to implement is regarded as a bad idea now, so drop it.

[rjw: Changelog]
Signed-off-by: Aaron Lu <aaron.lu@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Whenever an IRQ is claimed or freed, call gpio_lock_as_irq() or
gpio_unlock_as_irq() on the associated GPIO, to prevent that GPIO from
being configured in a manner incompatible with an interrupt.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Reviewed-by: Javier Martinez Canillas <javier@dowhile0.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>

Merge misc fixes from Andrew Morton.

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (21 commits)
  mm: revert mremap pud_free anti-fix
  mm: fix BUG in __split_huge_page_pmd
  swap: fix set_blocksize race during swapon/swapoff
  procfs: call default get_unmapped_area on MMU-present architectures
  procfs: fix unintended truncation of returned mapped address
  writeback: fix negative bdi max pause
  percpu_refcount: export symbols
  fs: buffer: move allocation failure loop into the allocator
  mm: memcg: handle non-error OOM situations more gracefully
  tools/testing/selftests: fix uninitialized variable
  block/partitions/efi.c: treat size mismatch as a warning, not an error
  mm: hugetlb: initialize PG_reserved for tail pages of gigantic compound pages
  mm/zswap: bugfix: memory leak when re-swapon
  mm: /proc/pid/pagemap: inspect _PAGE_SOFT_DIRTY only on present pages
  mm: migration: do not lose soft dirty bit if page is in migration state
  gcov: MAINTAINERS: Add an entry for gcov
  mm/hugetlb.c: correct missing private flag clearing
  mm/vmscan.c: don't forget to free shrinker->nr_deferred
  ipc/sem.c: synchronize semop and semctl with IPC_RMID
  ipc: update locking scheme comments
  ...

Revert commit 1ecfd533 ("mm/mremap.c: call pud_free() after fail
calling pmd_alloc()").

The original code was correct: pud_alloc(), pmd_alloc(), pte_alloc_map()
ensure that the pud, pmd, pt is already allocated, and seldom do they
need to allocate; on failure, upper levels are freed if appropriate by
the subsequent do_munmap().  Whereas commit 1ecfd533 did an
unconditional pud_free() of a most-likely still-in-use pud: saved only
by the near-impossiblity of pmd_alloc() failing.
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Occasionally we hit the BUG_ON(pmd_trans_huge(*pmd)) at the end of
__split_huge_page_pmd(): seen when doing madvise(,,MADV_DONTNEED).

It's invalid: we don't always have down_write of mmap_sem there: a racing
do_huge_pmd_wp_page() might have copied-on-write to another huge page
before our split_huge_page() got the anon_vma lock.

Forget the BUG_ON, just go back and try again if this happens.
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: David Rientjes <rientjes@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Fix race between swapoff and swapon.  Swapoff used old_block_size from
swap_info outside of swapon_mutex so it could be overwritten by
concurrent swapon.

The race has visible effect only if more than one swap block device
exists with different block sizes (e.g.  /dev/sda1 with block size 4096
and /dev/sdb1 with 512).  In such case it leads to setting the blocksize
of swapped off device with wrong blocksize.

The bug can be triggered with multiple concurrent swapoff and swapon:
0. Swap for some device is on.
1. swapoff:
First the swapoff is called on this device and "struct swap_info_struct
*p" is assigned. This is done under swap_lock however this lock is
released for the call try_to_unuse().

2. swapon:
After the assignment above (and before acquiring swapon_mutex &
swap_lock by swapoff) the swapon is called on the same device.
The p->old_block_size is assigned to the value of block_size the device.
This block size should be the same as previous but sometimes it is not.
The swapon ends successfully.

3. swapoff:
Swapoff resumes, grabs the locks and mutex and continues to disable this
swap device. Now it sets the block size to value taken from swap_info
which was overwritten by swapon in 2.
Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Reported-by: Weijie Yang <weijie.yang.kh@gmail.com>
Cc: Bob Liu <bob.liu@oracle.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Shaohua Li <shli@fusionio.com>
Cc: Minchan Kim <minchan@kernel.org>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Commit c4fe2448 ("sparc: fix PCI device proc file mmap(2)") added
proc_reg_get_unmapped_area in proc_reg_file_ops and
proc_reg_file_ops_no_compat, by which now mmap always returns EIO if
get_unmapped_area method is not defined for the target procfs file,
which causes regression of mmap on /proc/vmcore.

To address this issue, like get_unmapped_area(), call default
current->mm->get_unmapped_area on MMU-present architectures if
pde->proc_fops->get_unmapped_area, i.e.  the one in actual file
operation in the procfs file, is not defined.
Reported-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Tested-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Currently, proc_reg_get_unmapped_area truncates upper 32-bit of the
mapped virtual address returned from get_unmapped_area method in
pde->proc_fops due to the variable rv of signed integer on x86_64.  This
is too small to have vitual address of unsigned long on x86_64 since on
x86_64, signed integer is of 4 bytes while unsigned long is of 8 bytes.
To fix this issue, use unsigned long instead.

Fixes a regression added in commit c4fe2448 ("sparc: fix PCI device
proc file mmap(2)").
Signed-off-by: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Tested-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Toralf runs trinity on UML/i386.  After some time it hangs and the last
message line is

	BUG: soft lockup - CPU#0 stuck for 22s! [trinity-child0:1521]

It's found that pages_dirtied becomes very large.  More than 1000000000
pages in this case:

	period = HZ * pages_dirtied / task_ratelimit;
	BUG_ON(pages_dirtied > 2000000000);
	BUG_ON(pages_dirtied > 1000000000);      <---------

UML debug printf shows that we got negative pause here:

	ick: pause : -984
	ick: pages_dirtied : 0
	ick: task_ratelimit: 0

	 pause:
	+       if (pause < 0)  {
	+               extern int printf(char *, ...);
	+               printf("ick : pause : %li\n", pause);
	+               printf("ick: pages_dirtied : %lu\n", pages_dirtied);
	+               printf("ick: task_ratelimit: %lu\n", task_ratelimit);
	+               BUG_ON(1);
	+       }
	        trace_balance_dirty_pages(bdi,

Since pause is bounded by [min_pause, max_pause] where min_pause is also
bounded by max_pause.  It's suspected and demonstrated that the
max_pause calculation goes wrong:

	ick: pause : -717
	ick: min_pause : -177
	ick: max_pause : -717
	ick: pages_dirtied : 14
	ick: task_ratelimit: 0

The problem lies in the two "long = unsigned long" assignments in
bdi_max_pause() which might go negative if the highest bit is 1, and the
min_t(long, ...) check failed to protect it falling under 0.  Fix all of
them by using "unsigned long" throughout the function.
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Reported-by: Toralf Förster <toralf.foerster@gmx.de>
Tested-by: Toralf Förster <toralf.foerster@gmx.de>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Richard Weinberger <richard@nod.at>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Export the interface to be used within modules.
Signed-off-by: Matias Bjorling <m@bjorling.me>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Buffer allocation has a very crude indefinite loop around waking the
flusher threads and performing global NOFS direct reclaim because it can
not handle allocation failures.

The most immediate problem with this is that the allocation may fail due
to a memory cgroup limit, where flushers + direct reclaim might not make
any progress towards resolving the situation at all.  Because unlike the
global case, a memory cgroup may not have any cache at all, only
anonymous pages but no swap.  This situation will lead to a reclaim
livelock with insane IO from waking the flushers and thrashing unrelated
filesystem cache in a tight loop.

Use __GFP_NOFAIL allocations for buffers for now.  This makes sure that
any looping happens in the page allocator, which knows how to
orchestrate kswapd, direct reclaim, and the flushers sensibly.  It also
allows memory cgroups to detect allocations that can't handle failure
and will allow them to ultimately bypass the limit if reclaim can not
make progress.
Reported-by: azurIt <azurit@pobox.sk>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Commit 3812c8c8 ("mm: memcg: do not trap chargers with full
callstack on OOM") assumed that only a few places that can trigger a
memcg OOM situation do not return VM_FAULT_OOM, like optional page cache
readahead.  But there are many more and it's impractical to annotate
them all.

First of all, we don't want to invoke the OOM killer when the failed
allocation is gracefully handled, so defer the actual kill to the end of
the fault handling as well.  This simplifies the code quite a bit for
added bonus.

Second, since a failed allocation might not be the abrupt end of the
fault, the memcg OOM handler needs to be re-entrant until the fault
finishes for subsequent allocation attempts.  If an allocation is
attempted after the task already OOMed, allow it to bypass the limit so
that it can quickly finish the fault and invoke the OOM killer.
Reported-by: azurIt <azurit@pobox.sk>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>