... making it usable with directory locked shared
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

New method: ->iterate_shared().  Same arguments as in ->iterate(),
called with the directory locked only shared.  Once all filesystems
switch, the old one will be gone.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

same as read() on regular files has, and for the same reason.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

ta-da!

The main issue is the lack of down_write_killable(), so the places
like readdir.c switched to plain inode_lock(); once killable
variants of rwsem primitives appear, that'll be dealt with.

lockdep side also might need more work
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

If we *do* run into an in-lookup match, we need to wait for it to
cease being in-lookup.  Fortunately, we do have unused space in
in-lookup dentries - d_lru is never looked at until it stops being
in-lookup.

So we can stash a pointer to wait_queue_head from stack frame of
the caller of ->lookup().  Some precautions are needed while
waiting, but it's not that hard - we do hold a reference to dentry
we are waiting for, so it can't go away.  If it's found to be
in-lookup the wait_queue_head is still alive and will remain so
at least while ->d_lock is held.  Moreover, the condition we
are waiting for becomes true at the same point where everything
on that wq gets woken up, so we can just add ourselves to the
queue once.

d_alloc_parallel() gets a pointer to wait_queue_head_t from its
caller; lookup_slow() adjusted, d_add_ci() taught to use
d_alloc_parallel() if the dentry passed to it happens to be
in-lookup one (i.e. if it's been called from the parallel lookup).

That's pretty much it - all that remains is to switch ->i_mutex
to rwsem and have lookup_slow() take it shared.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

We will need to be able to check if there is an in-lookup
dentry with matching parent/name.  Right now it's impossible,
but as soon as start locking directories shared such beasts
will appear.

Add a secondary hash for locating those.  Hash chains go through
the same space where d_alias will be once it's not in-lookup anymore.
Search is done under the same bitlock we use for modifications -
with the primary hash we can rely on d_rehash() into the wrong
chain being the worst that could happen, but here the pointers are
buggered once it's removed from the chain.  On the other hand,
the chains are not going to be long and normally we'll end up
adding to the chain anyway.  That allows us to avoid bothering with
->d_lock when doing the comparisons - everything is stable until
removed from chain.

New helper: d_alloc_parallel().  Right now it allocates, verifies
that no hashed and in-lookup matches exist and adds to in-lookup
hash.

Returns ERR_PTR() for error, hashed match (in the unlikely case it's
been found) or new dentry.  In-lookup matches trigger BUG() for
now; that will change in the next commit when we introduce waiting
for ongoing lookup to finish.  Note that in-lookup matches won't be
possible until we actually go for shared locking.

lookup_slow() switched to use of d_alloc_parallel().

Again, these commits are separated only for making it easier to
review.  All this machinery will start doing something useful only
when we go for shared locking; it's just that the combination is
too large for my taste.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

We'll need to verify that there's neither a hashed nor in-lookup
dentry with desired parent/name before adding to in-lookup set.

One possible solution would be to hold the parent's ->d_lock through
both checks, but while the in-lookup set is relatively small at any
time, dcache is not.  And holding the parent's ->d_lock through
something like __d_lookup_rcu() would suck too badly.

So we leave the parent's ->d_lock alone, which means that we watch
out for the following scenario:
	* we verify that there's no hashed match
	* existing in-lookup match gets hashed by another process
	* we verify that there's no in-lookup matches and decide
that everything's fine.

Solution: per-directory kinda-sorta seqlock, bumped around the times
we hash something that used to be in-lookup or move (and hash)
something in place of in-lookup.  Then the above would turn into
	* read the counter
	* do dcache lookup
	* if no matches found, check for in-lookup matches
	* if there had been none of those either, check if the
counter has changed; repeat if it has.

The "kinda-sorta" part is due to the fact that we don't have much spare
space in inode.  There is a spare word (shared with i_bdev/i_cdev/i_pipe),
so the counter part is not a problem, but spinlock is a different story.

We could use the parent's ->d_lock, and it would be less painful in
terms of contention, for __d_add() it would be rather inconvenient to
grab; we could do that (using lock_parent()), but...

Fortunately, we can get serialization on the counter itself, and it
might be a good idea in general; we can use cmpxchg() in a loop to
get from even to odd and smp_store_release() from odd to even.

This commit adds the counter and updating logics; the readers will be
added in the next commit.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

marked as such when (would be) parallel lookup is about to pass them
to actual ->lookup(); unmarked when
	* __d_add() is about to make it hashed, positive or not.
	* __d_move() (from d_splice_alias(), directly or via
__d_unalias()) puts a preexisting dentry in its place
	* in caller of ->lookup() if it has escaped all of the
above.  Bug (WARN_ON, actually) if it reaches the final dput()
or d_instantiate() while still marked such.

As the result, we are guaranteed that for as long as the flag is
set, dentry will
	* remain negative unhashed with positive refcount
	* never have its ->d_alias looked at
	* never have its ->d_lru looked at
	* never have its ->d_parent and ->d_name changed

Right now we have at most one such for any given parent directory.
With parallel lookups that restriction will weaken to
	* only exist when parent is locked shared
	* at most one with given (parent,name) pair (comparison of
names is according to ->d_compare())
	* only exist when there's no hashed dentry with the same
(parent,name)

Transition will take the next several commits; unfortunately, we'll
only be able to switch to rwsem at the end of this series.  The
reason for not making it a single patch is to simplify review.

New primitives: d_in_lookup() (a predicate checking if dentry is in
the in-lookup state) and d_lookup_done() (tells the system that
we are done with lookup and if it's still marked as in-lookup, it
should cease to be such).
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

will be needed as soon as lookups are not serialized by ->i_mutex
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Right now ext2_get_page() (and its analogues in a bunch of other filesystems)
relies upon the directory being locked - the way it sets and tests Checked and
Error bits would be racy without that.  Switch to a slightly different scheme,
_not_ setting Checked in case of failure.  That way the logics becomes
	if Checked => OK
	else if Error => fail
	else if !validate => fail
	else => OK
with validation setting Checked or Error on success and failure resp. and
returning which one had happened.  Equivalent to the current logics, but unlike
the current logics not sensitive to the order of set_bit, test_bit getting
reordered by CPU, etc.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

... and explain the non-obvious logics in case when lookup yields
a different dentry.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

... and have it use inode_lock()
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

grab a reference to dentry we'd got the sucker from, and return
that dentry via *wait, rather than just returning the address of
->i_mutex.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

The rest of work.xattr stuff isn't needed for this branch

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

... and do not assume they are already attached to each other
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Pull ARM fixes from Russell King:
 "A couple of small fixes, and wiring up the new syscalls which appeared
  during the merge window"

* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
  ARM: 8550/1: protect idiv patching against undefined gcc behavior
  ARM: wire up preadv2 and pwritev2 syscalls
  ARM: SMP enable of cache maintanence broadcast

Pull MMC fixes from Ulf Hansson:
 "Here are a couple of mmc fixes intended for v4.6 rc3:

  MMC host:
   - sdhci: Fix regression setting power on Trats2 board
   - sdhci-pci: Add support and PCI IDs for more Broxton host controllers"

* tag 'mmc-v4.6-rc1' of git://git.linaro.org/people/ulf.hansson/mmc:
  mmc: sdhci-pci: Add support and PCI IDs for more Broxton host controllers
  mmc: sdhci: Fix regression setting power on Trats2 board

Pull i2c fixes from Wolfram Sang:
 "Some bugfixes from I2C:

   - fix a uevent triggered boot problem by removing a useless debug
     print

   - fix sysfs-attributes of the new i2c-demux-pinctrl driver to follow
     standard kernel behaviour

   - fix a potential division-by-zero error (needed two takes)"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: jz4780: really prevent potential division by zero
  Revert "i2c: jz4780: prevent potential division by zero"
  i2c: jz4780: prevent potential division by zero
  i2c: mux: demux-pinctrl: Update docs to new sysfs-attributes
  i2c: mux: demux-pinctrl: Clean up sysfs attributes
  i2c: prevent endless uevent loop with CONFIG_I2C_DEBUG_CORE

This reverts commit 1028b55b.

It's broken: it makes ext4 return an error at an invalid point, causing
the readdir wrappers to write the the position of the last successful
directory entry into the position field, which means that the next
readdir will now return that last successful entry _again_.

You can only return fatal errors (that terminate the readdir directory
walk) from within the filesystem readdir functions, the "normal" errors
(that happen when the readdir buffer fills up, for example) happen in
the iterorator where we know the position of the actual failing entry.

I do have a very different patch that does the "signal_pending()"
handling inside the iterator function where it is allowable, but while
that one passes all the sanity checks, I screwed up something like four
times while emailing it out, so I'm not going to commit it today.

So my track record is not good enough, and the stars will have to align
better before that one gets committed.  And it would be good to get some
review too, of course, since celestial alignments are always an iffy
debugging model.

IOW, let's just revert the commit that caused the problem for now.
Reported-by: Greg Thelen <gthelen@google.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

reiserfs_xattr_[sg]et() will fail with -EOPNOTSUPP for V1 inodes anyway,
and all reiserfs instances of ->[sg]et() call it and so does ->set_acl().

Checks for name length in the instances had been bogus; they should've
been "bugger off if it's _exactly_ the prefix" (as generic would
do on its own) and not "bugger off if it's shorter than the prefix" -
that can't happen.

xattr_full_name() is needed to adjust for the fact that generic instances
will skip the prefix in the name passed to ->[gs]et(); reiserfs homegrown
analogues didn't.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

none of that stuff can ever be called for NULL or negative
dentry.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

... and neither can ever be NULL
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Pull parisc fixes from Helge Deller:
 "Since commit 0de79858 ("parisc: Use generic extable search and
  sort routines") module loading is boken on parisc, because the parisc
  module loader wasn't prepared for the new R_PARISC_PCREL32 relocations.

  In addition, due to that breakage, Mikulas Patocka noticed that
  handling exceptions from modules probably never worked on parisc.  It
  was just masked by the fact that exceptions from modules don't happen
  during normal use.

  This patch series fixes those issues and survives the tests of the
  lib/test_user_copy kernel module test.  Some patches are tagged for
  stable"

* 'parisc-4.6-3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Update comment regarding relative extable support
  parisc: Unbreak handling exceptions from kernel modules
  parisc: Fix kernel crash with reversed copy_from_user()
  parisc: Avoid function pointers for kernel exception routines
  parisc: Handle R_PARISC_PCREL32 relocations in kernel modules

Pull libnvdimm fixes from Dan Williams:
 "Three fixes, the first two are tagged for -stable:

   - The ndctl utility/library gained expanded unit tests illuminating a
     long standing bug in the libnvdimm SMART data retrieval
     implementation.

     It has been broken since its initial implementation, now fixed.

   - Another one line fix for the detection of stale info blocks.

     Without this change userspace can get into a situation where it is
     unable to reconfigure a namespace.

   - Fix the badblock initialization path in the presence of the new (in
     v4.6-rc1) section alignment workarounds.

     Without this change badblocks will be reported at the wrong offset.

  These have received a build success report from the kbuild robot and
  have appeared in -next with no reported issues"

* 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  libnvdimm, pfn: fix nvdimm_namespace_add_poison() vs section alignment
  libnvdimm, pfn: fix uuid validation
  libnvdimm: fix smart data retrieval

Pull GPIO fixes from Linus Walleij:
 "Here is a set of four GPIO fixes.  The two fixes to the core are
  serious as they are regressing minor architectures.

  Core fixes:

   - Defer GPIO device setup until after gpiolib is initialized.

     It turns out that a few very tightly integrated GPIO platform
     drivers initialize so early (befor core_initcall()) so that the
     gpiolib isn't even initialized itself.  That limits what the
     library can do, and we cannot reference uninitialized fields until
     later.

     Defer some of the initialization until right after the gpiolib is
     initialized in these (rare) cases.

   - As a consequence: do not use devm_* resources when allocating the
     states in the initial set-up of the gpiochip.

  Driver fixes:

   - In ACPI retrieveal: ignore GpioInt when looking for output GPIOs.

   - Fix legacy builds on the PXA without a backing pin controller.

   - Use correct datatype on pca953x register writes"

* tag 'gpio-v4.6-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio:
  gpio: pca953x: Use correct u16 value for register word write
  gpiolib: Defer gpio device setup until after gpiolib initialization
  gpiolib: Do not use devm functions when registering gpio chip
  gpio: pxa: fix legacy non pinctrl aware builds
  gpio / ACPI: ignore GpioInt() GPIOs when requesting GPIO_OUT_*

Pull tty fixes from Greg KH:
 "Here are two tty fixes for issues found.

  One was due to a merge error in 4.6-rc1, and the other a regression
  fix for UML consoles that broke in 4.6-rc1.

  Both have been in linux-next for a while"

* tag 'tty-4.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty: Fix merge of "tty: Refactor tty_open()"
  tty: Fix UML console breakage

Pull USB fixes from Greg KH:
 "Here are some USB fixes and new device ids for 4.6-rc3.

  Nothing major, the normal USB gadget fixes and usb-serial driver ids,
  along with some other fixes mixed in.  All except the USB serial ids
  have been tested in linux-next, the id additions should be fine as
  they are 'trivial'"

* tag 'usb-4.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (25 commits)
  USB: option: add "D-Link DWM-221 B1" device id
  USB: serial: cp210x: Adding GE Healthcare Device ID
  USB: serial: ftdi_sio: Add support for ICP DAS I-756xU devices
  usb: dwc3: keystone: drop dma_mask configuration
  usb: gadget: udc-core: remove manual dma configuration
  usb: dwc3: pci: add ID for one more Intel Broxton platform
  usb: renesas_usbhs: fix to avoid using a disabled ep in usbhsg_queue_done()
  usb: dwc2: do not override forced dr_mode in gadget setup
  usb: gadget: f_midi: unlock on error
  USB: digi_acceleport: do sanity checking for the number of ports
  USB: cypress_m8: add endpoint sanity check
  USB: mct_u232: add sanity checking in probe
  usb: fix regression in SuperSpeed endpoint descriptor parsing
  USB: usbip: fix potential out-of-bounds write
  usb: renesas_usbhs: disable TX IRQ before starting TX DMAC transfer
  usb: renesas_usbhs: avoid NULL pointer derefernce in usbhsf_pkt_handler()
  usb: gadget: f_midi: Fixed a bug when buflen was smaller than wMaxPacketSize
  usb: phy: qcom-8x16: fix regulator API abuse
  usb: ch9: Fix SSP Device Cap wFunctionalitySupport type
  usb: gadget: composite: Access SSP Dev Cap fields properly
  ...

Pull staging and IIO driver fixes from Greg KH:
 "Here are some IIO driver fixes, along with two staging driver fixes
  for 4.6-rc3.

  One staging driver patch reverts the deletion of a driver that
  happened in 4.6-rc1.  We thought that laptop.org was dead, but it's
  still alive and kicking, and has users that were mad we broke their
  hardware by deleting a driver for their machines.  So that driver is
  added back and everyone is happy again.

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'staging-4.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  Revert "Staging: olpc_dcon: Remove obsolete driver"
  staging/rdma/hfi1: select CRC32
  iio: gyro: bmg160: fix buffer read values
  iio: gyro: bmg160: fix endianness when reading axes
  iio: accel: bmc150: fix endianness when reading axes
  iio: st_magn: always define ST_MAGN_TRIGGER_SET_STATE
  iio: fix config watermark initial value
  iio: health: max30100: correct FIFO check condition
  iio: imu: Fix inv_mpu6050 dependencies
  iio: adc: Fix build error of missing devm_ioremap_resource on UM
  iio: light: apds9960: correct FIFO check condition
  iio: adc: max1363: correct reference voltage
  iio: adc: max1363: add missing adc to max1363_id

Pull SCSI fixes from James Bottomley:
 "This is a set of eight fixes.

  Two are trivial gcc-6 updates (brace additions and unused variable
  removal).  There's a couple of cxlflash regressions, a correction for
  sd being overly chatty on revalidation (causing excess log increases).
  A VPD issue which could crash USB devices because they seem very
  intolerant to VPD inquiries, an ALUA deadlock fix and a mpt3sas buffer
  overrun fix"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: Do not attach VPD to devices that don't support it
  sd: Fix excessive capacity printing on devices with blocks bigger than 512 bytes
  scsi_dh_alua: Fix a recently introduced deadlock
  scsi: Declare local symbols static
  cxlflash: Move to exponential back-off when cmd_room is not available
  cxlflash: Fix regression issue with re-ordering patch
  mpt3sas: Don't overreach ioc->reply_post[] during initialization
  aacraid: add missing curly braces

Pull MD fixes from Shaohua Li:
 "This update mainly fixes bugs:

   - fix error handling (Guoqing)
   - fix a crash when a disk is hotremoved (me)
   - fix a dead loop (Wei Fang)"

* tag 'md/4.6-rc2-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/shli/md:
  md/bitmap: clear bitmap if bitmap_create failed
  MD: add rdev reference for super write
  md: fix a trivial typo in comments
  md:raid1: fix a dead loop when read from a WriteMostly disk

Pull power management and ACPI fixes from Rafael Wysocki:
 "Fixes for some issues discovered after recent changes and for some
  that have just been found lately regardless of those changes
  (intel_pstate, intel_idle, PM core, mailbox/pcc, turbostat) plus
  support for some new CPU models (intel_idle, Intel RAPL driver,
  turbostat) and documentation updates (intel_pstate, PM core).

  Specifics:

   - intel_pstate fixes for two issues exposed by the recent switch over
     from using timers and for one issue introduced during the 4.4 cycle
     plus new comments describing data structures used by the driver
     (Rafael Wysocki, Srinivas Pandruvada).

   - intel_idle fixes related to CPU offline/online (Richard Cochran).

   - intel_idle support (new CPU IDs and state definitions mostly) for
     Skylake-X and Kabylake processors (Len Brown).

   - PCC mailbox driver fix for an out-of-bounds memory access that may
     cause the kernel to panic() (Shanker Donthineni).

   - New (missing) CPU ID for one apparently overlooked Haswell model in
     the Intel RAPL power capping driver (Srinivas Pandruvada).

   - Fix for the PM core's wakeup IRQs framework to make it work after
     wakeup settings reconfiguration from sysfs (Grygorii Strashko).

   - Runtime PM documentation update to make it describe what needs to
     be done during device removal more precisely (Krzysztof Kozlowski).

   - Stale comment removal cleanup in the cpufreq-dt driver (Viresh
     Kumar).

   - turbostat utility fixes and support for Broxton, Skylake-X and
     Kabylake processors (Len Brown)"

* tag 'pm+acpi-4.6-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (28 commits)
  PM / wakeirq: fix wakeirq setting after wakup re-configuration from sysfs
  tools/power turbostat: work around RC6 counter wrap
  tools/power turbostat: initial KBL support
  tools/power turbostat: initial SKX support
  tools/power turbostat: decode BXT TSC frequency via CPUID
  tools/power turbostat: initial BXT support
  tools/power turbostat: print IRTL MSRs
  tools/power turbostat: SGX state should print only if --debug
  intel_idle: Add KBL support
  intel_idle: Add SKX support
  intel_idle: Clean up all registered devices on exit.
  intel_idle: Propagate hot plug errors.
  intel_idle: Don't overreact to a cpuidle registration failure.
  intel_idle: Setup the timer broadcast only on successful driver load.
  intel_idle: Avoid a double free of the per-CPU data.
  intel_idle: Fix dangling registration on error path.
  intel_idle: Fix deallocation order on the driver exit path.
  intel_idle: Remove redundant initialization calls.
  intel_idle: Fix a helper function's return value.
  intel_idle: remove useless return from void function.
  ...