Patch 0fed3ac8 improved the hash mixing, but the function is slower
than necessary; there's a 7-instruction dependency chain (10 on x86)
each loop iteration.

Word-at-a-time access is a very tight loop (which is good, because
link_path_walk() is one of the hottest code paths in the entire kernel),
and the hash mixing function must not have a longer latency to avoid
slowing it down.

There do not appear to be any published fast hash functions that:
1) Operate on the input a word at a time, and
2) Don't need to know the length of the input beforehand, and
3) Have a single iterated mixing function, not needing conditional
   branches or unrolling to distinguish different loop iterations.

One of the algorithms which comes closest is Yann Collet's xxHash, but
that's two dependent multiplies per word, which is too much.

The key insights in this design are:

1) Barring expensive ops like multiplies, to diffuse one input bit
   across 64 bits of hash state takes at least log2(64) = 6 sequentially
   dependent instructions.  That is more cycles than we'd like.
2) An operation like "hash ^= hash << 13" requires a second temporary
   register anyway, and on a 2-operand machine like x86, it's three
   instructions.
3) A better use of a second register is to hold a two-word hash state.
   With careful design, no temporaries are needed at all, so it doesn't
   increase register pressure.  And this gets rid of register copying
   on 2-operand machines, so the code is smaller and faster.
4) Using two words of state weakens the requirement for one-round mixing;
   we now have two rounds of mixing before cancellation is possible.
5) A two-word hash state also allows operations on both halves to be
   done in parallel, so on a superscalar processor we get more mixing
   in fewer cycles.

I ended up using a mixing function inspired by the ChaCha and Speck
round functions.  It is 6 simple instructions and 3 cycles per iteration
(assuming multiply by 9 can be done by an "lea" instruction):

		x ^= *input++;
	y ^= x;	x = ROL(x, K1);
	x += y;	y = ROL(y, K2);
	y *= 9;

Not only is this reversible, two consecutive rounds are reversible:
if you are given the initial and final states, but not the intermediate
state, it is possible to compute both input words.  This means that at
least 3 words of input are required to create a collision.

(It also has the property, used by hash_name() to avoid a branch, that
it hashes all-zero to all-zero.)

The rotate constants K1 and K2 were found by experiment.  The search took
a sample of random initial states (I used 1023) and considered the effect
of flipping each of the 64 input bits on each of the 128 output bits two
rounds later.  Each of the 8192 pairs can be considered a biased coin, and
adding up the Shannon entropy of all of them produces a score.

The best-scoring shifts also did well in other tests (flipping bits in y,
trying 3 or 4 rounds of mixing, flipping all 64*63/2 pairs of input bits),
so the choice was made with the additional constraint that the sum of the
shifts is odd and not too close to the word size.

The final state is then folded into a 32-bit hash value by a less carefully
optimized multiply-based scheme.  This also has to be fast, as pathname
components tend to be short (the most common case is one iteration!), but
there's some room for latency, as there is a fair bit of intervening logic
before the hash value is used for anything.

(Performance verified with "bonnie++ -s 0 -n 1536:-2" on tmpfs.  I need
a better benchmark; the numbers seem to show a slight dip in performance
between 4.6.0 and this patch, but they're too noisy to quote.)

Special thanks to Bruce fields for diligent testing which uncovered a
nasty fencepost error in an earlier version of this patch.

[checkpatch.pl formatting complaints noted and respectfully disagreed with.]
Signed-off-by: NGeorge Spelvin <linux@sciencehorizons.net>
Tested-by: NJ. Bruce Fields <bfields@redhat.com>

The "simplified" prime multipliers made very bad hash functions, so get rid
of them.  This completes the work of 689de1d6.

To avoid the inefficiency which was the motivation for the "simplified"
multipliers, hash_64() on 32-bit systems is changed to use a different
algorithm.  It makes two calls to hash_32() instead.

drivers/media/usb/dvb-usb-v2/af9015.c uses the old GOLDEN_RATIO_PRIME_32
for some horrible reason, so it inherits a copy of the old definition.
Signed-off-by: NGeorge Spelvin <linux@sciencehorizons.net>
Cc: Antti Palosaari <crope@iki.fi>
Cc: Mauro Carvalho Chehab <m.chehab@samsung.com>

That's all that's ever asked for, and it makes the return
type of hash_long() consistent.

It also allows (upcoming patch) an optimized implementation
of hash_64 on 32-bit machines.

I tried adding a BUILD_BUG_ON to ensure the number of bits requested
was never more than 32 (most callers use a compile-time constant), but
adding <linux/bug.h> to <linux/hash.h> breaks the tools/perf compiler
unless tools/perf/MANIFEST is updated, and understanding that code base
well enough to update it is too much trouble.  I did the rest of an
allyesconfig build with such a check, and nothing tripped.
Signed-off-by: NGeorge Spelvin <linux@sciencehorizons.net>

Finally, the first use of previous two patches: eliminate the
separate ad-hoc string hash functions in the sunrpc code.

Now hash_str() is a wrapper around hash_string(), and hash_mem() is
likewise a wrapper around full_name_hash().

Note that sunrpc code *does* call hash_mem() with a zero length, which
is why the previous patch needed to handle that in full_name_hash().
(Thanks, Bruce, for finding that!)

This also eliminates the only caller of hash_long which asks for
more than 32 bits of output.

The comment about the quality of hashlen_string() and full_name_hash()
is jumping the gun by a few patches; they aren't very impressive now,
but will be improved greatly later in the series.
Signed-off-by: NGeorge Spelvin <linux@sciencehorizons.net>
Tested-by: NJ. Bruce Fields <bfields@redhat.com>
Acked-by: NJ. Bruce Fields <bfields@redhat.com>
Cc: Jeff Layton <jlayton@poochiereds.net>
Cc: linux-nfs@vger.kernel.org

We'd like to make more use of the highly-optimized dcache hash functions
throughout the kernel, rather than have every subsystem create its own,
and a function that hashes basic null-terminated strings is required
for that.

(The name is to emphasize that it returns both hash and length.)

It's actually useful in the dcache itself, specifically d_alloc_name().
Other uses in the next patch.

full_name_hash() is also tweaked to make it more generally useful:
1) Take a "char *" rather than "unsigned char *" argument, to
   be consistent with hash_name().
2) Handle zero-length inputs.  If we want more callers, we don't want
   to make them worry about corner cases.
Signed-off-by: NGeorge Spelvin <linux@sciencehorizons.net>

... so they can be used without the rest of <linux/dcache.h>

The hashlen_* macros will make sense next patch.
Signed-off-by: NGeorge Spelvin <linux@sciencehorizons.net>

The hash mixing between adding the next 64 bits of name
was just a bit weak.

Replaced with a still very fast but slightly more effective
mixing function.
Signed-off-by: NGeorge Spelvin <linux@horizon.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Pull x86 fix from Thomas Gleixner:
 "Just the missing compat entry for the new pread/writev2"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86: Use compat version for preadv2 and pwritev2

Pull networking fixes from David Miller:

 1) Fix mvneta/bm dependencies, from Arnd Bergmann.

 2) RX completion hw bug workaround in bnxt_en, from Michael Chan.

 3) Kernel pointer leak in nf_conntrack, from Linus.

 4) Hoplimit route attribute limits not enforced properly, from Paolo
    Abeni.

 5) qlcnic driver NULL deref fix from Dan Carpenter.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  arm64: bpf: jit JMP_JSET_{X,K}
  net/route: enforce hoplimit max value
  nf_conntrack: avoid kernel pointer value leak in slab name
  drivers: net: xgene: fix register offset
  drivers: net: xgene: fix statistics counters race condition
  drivers: net: xgene: fix ununiform latency across queues
  drivers: net: xgene: fix sharing of irqs
  drivers: net: xgene: fix IPv4 forward crash
  xen-netback: fix extra_info handling in xenvif_tx_err()
  net: mvneta: bm: fix dependencies again
  bnxt_en: Add workaround to detect bad opaque in rx completion (part 2)
  bnxt_en: Add workaround to detect bad opaque in rx completion (part 1)
  qlcnic: potential NULL dereference in qlcnic_83xx_get_minidump_template()

Original implementation commit e54bcde3 ("arm64: eBPF JIT compiler")
had the relevant code paths, but due to an oversight always fail jiting.

As a result, we had been falling back to BPF interpreter whenever a BPF
program has JMP_JSET_{X,K} instructions.

With this fix, we confirm that the corresponding tests in lib/test_bpf
continue to pass, and also jited.

...
[    2.784553] test_bpf: #30 JSET jited:1 188 192 197 PASS
[    2.791373] test_bpf: #31 tcpdump port 22 jited:1 325 677 625 PASS
[    2.808800] test_bpf: #32 tcpdump complex jited:1 323 731 991 PASS
...
[    3.190759] test_bpf: #237 JMP_JSET_K: if (0x3 & 0x2) return 1 jited:1 110 PASS
[    3.192524] test_bpf: #238 JMP_JSET_K: if (0x3 & 0xffffffff) return 1 jited:1 98 PASS
[    3.211014] test_bpf: #249 JMP_JSET_X: if (0x3 & 0x2) return 1 jited:1 120 PASS
[    3.212973] test_bpf: #250 JMP_JSET_X: if (0x3 & 0xffffffff) return 1 jited:1 89 PASS
...

Fixes: e54bcde3 ("arm64: eBPF JIT compiler")
Signed-off-by: NZi Shen Lim <zlim.lnx@gmail.com>
Acked-by: NWill Deacon <will.deacon@arm.com>
Acked-by: NYang Shi <yang.shi@linaro.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently, when creating or updating a route, no check is performed
in both ipv4 and ipv6 code to the hoplimit value.

The caller can i.e. set hoplimit to 256, and when such route will
 be used, packets will be sent with hoplimit/ttl equal to 0.

This commit adds checks for the RTAX_HOPLIMIT value, in both ipv4
ipv6 route code, substituting any value greater than 255 with 255.

This is consistent with what is currently done for ADVMSS and MTU
in the ipv4 code.
Signed-off-by: NPaolo Abeni <pabeni@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The slab name ends up being visible in the directory structure under
/sys, and even if you don't have access rights to the file you can see
the filenames.

Just use a 64-bit counter instead of the pointer to the 'net' structure
to generate a unique name.

This code will go away in 4.7 when the conntrack code moves to a single
kmemcache, but this is the backportable simple solution to avoiding
leaking kernel pointers to user space.

Fixes: 5b3501fa ("netfilter: nf_conntrack: per netns nf_conntrack_cachep")
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
Acked-by: NEric Dumazet <eric.dumazet@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pull vfs fixes from Al Viro:
 "Overlayfs fixes from Miklos, assorted fixes from me.

  Stable fodder of varying severity, all sat in -next for a while"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  ovl: ignore permissions on underlying lookup
  vfs: add lookup_hash() helper
  vfs: rename: check backing inode being equal
  vfs: add vfs_select_inode() helper
  get_rock_ridge_filename(): handle malformed NM entries
  ecryptfs: fix handling of directory opening
  atomic_open(): fix the handling of create_error
  fix the copy vs. map logics in blk_rq_map_user_iov()
  do_splice_to(): cap the size before passing to ->splice_read()

Iyappan Subramanian says:

====================
drivers: net: xgene: Bug fixes

This patch set addresses the following bug fixes that were found during testing.

  1. IPv4 forward test crash
    - drivers: net: xgene: fix IPv4 forward crash

  2. Sharing of irqs
    - drivers: net: xgene: fix sharing of irqs

  3. Ununiform latency across queues
    - drivers: net: xgene: fix ununiform latency across queues

  4. Fix statistics counters race condition
    - drivers: net: xgene: fix statistics counters race condition

  5. Correcting register offset and field lengths
    - drivers: net: xgene: fix register offset

v2: Address review comments from v1
- Defer TSO fix, and reposting all other patches from v1

v1:
- Initial version
====================
Signed-off-by: NIyappan Subramanian <isubramanian@apm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes SG_RX_DV_GATE_REG_0_ADDR register offset
and ring state field lengths.
Signed-off-by: NIyappan Subramanian <isubramanian@apm.com>
Tested-by: NToan Le <toanle@apm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes the race condition on updating the statistics
counters by moving the counters to the ring structure.
Signed-off-by: NIyappan Subramanian <isubramanian@apm.com>
Tested-by: NToan Le <toanle@apm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch addresses ununiform latency across queues by adding
more queues to match with, upto number of CPU cores.

Also, number of interrupts are increased and the channel numbers
are reordered.
Signed-off-by: NIyappan Subramanian <isubramanian@apm.com>
Tested-by: NToan Le <toanle@apm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since hardware doesn't allow sharing of interrupts,
this patch fixes the same by removing IRQF_SHARED flag.
Signed-off-by: NIyappan Subramanian <isubramanian@apm.com>
Tested-by: NToan Le <toanle@apm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes the crash observed during IPv4 forward test by
setting the drop field in the dbptr.
Signed-off-by: NIyappan Subramanian <isubramanian@apm.com>
Tested-by: NToan Le <toanle@apm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pull cgroup fixes from Tejun Heo:
 "During v4.6-rc1 cgroup namespace support was merged.  There is an
  issue where it's impossible to tell whether a given cgroup mount point
  is bind mounted or namespaced.  Serge has been working on the issue
  but it took longer than expected to resolve, so the late pull request.

  Given that it's a completely new feature and the patches don't touch
  anything else, the risk seems acceptable.  However, if this is too
  late, an alternative is plugging new cgroup ns creation for v4.6 and
  retrying for v4.7"

* 'for-4.6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  cgroup: fix compile warning
  kernfs: kernfs_sop_show_path: don't return 0 after seq_dentry call
  cgroup, kernfs: make mountinfo show properly scoped path for cgroup namespaces
  kernfs_path_from_node_locked: don't overwrite nlen

Pull workqueue fix from Tejun Heo:
 "CPU hotplug callbacks can invoke DOWN_FAILED w/o preceding
  DOWN_PREPARE which can trigger a WARN_ON() in workqueue.

  The bug has been there for a very long time.  It only triggers if CPU
  down fails at a specific point and I don't think it has adverse
  effects other than the warning messages.  The fix is very low impact"

* 'for-4.6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
  workqueue: fix rebind bound workers warning

Pull scheduler fix from Ingo Molnar:
 "This is a revert to fix an interactivity problem.

  The proper fixes for the problems that the reverted commit exposed are
  now in sched/core (consisting of 3 patches), but were too risky for
  v4.6 and will arrive in the v4.7 merge window"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  Revert "sched/fair: Fix fairness issue on migration"

Pull perf fixes from Ingo Molnar:
 "An uncharacteristically large number of bugs popped up in the last
  week:

   - various tooling fixes, two crashes and build problems
   - two Intel PT fixes
   - an KNL uncore driver fix
   - an Intel PMU driver fix"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf stat: Fallback to user only counters when perf_event_paranoid > 1
  perf evsel: Handle EACCESS + perf_event_paranoid=2 in fallback()
  perf evsel: Improve EPERM error handling in open_strerror()
  tools lib traceevent: Do not reassign parg after collapse_tree()
  perf probe: Check if dwarf_getlocations() is available
  perf dwarf: Guard !x86_64 definitions under #ifdef else clause
  perf tools: Use readdir() instead of deprecated readdir_r()
  perf thread_map: Use readdir() instead of deprecated readdir_r()
  perf script: Use readdir() instead of deprecated readdir_r()
  perf tools: Use readdir() instead of deprecated readdir_r()
  perf/core: Disable the event on a truncated AUX record
  perf/x86/intel/pt: Generate PMI in the STOP region as well
  perf/x86: Fix undefined shift on 32-bit kernels
  perf/x86/msr: Fix SMI overflow
  perf/x86/intel/uncore: Fix CHA registers configuration procedure for Knights Landing platform
  perf diff: Fix duplicated output column

Pull ARM SoC fixes from Arnd Bergmann:
 "Three more bug fixes for ARM SoCs this week:

   - The Atmel sama5d2 was registering the wrong NFC device type

   - On Atmel sam9x5, the power management controller had an incorrect
     register area size

   - On ARM64 Allwinner machine was not secting the generic irqchip
     code, causing build errors in some configurations"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: dts: at91: sam9x5: Fix the memory range assigned to the PMC
  arm64/sunxi: 4.6-rc1: Add dependency on generic irq chip
  ARM: dts: at91: sama5d2: use "atmel,sama5d3-nfc" compatible for nfc

Merge tag 'regulator-fix-v4.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator

Pull regulator fixes from Mark Brown:
 "A small collection of driver specific fixes for the regulator
  subsysetem:

   - Fix handling of probe deferral for GPIO regulators

   - Fix a typo in the module alias for DA9053

   - Fix the definition of BUCK9 in the S2MPS11 driver.  This change
     looks larger than it is because an irregularity in the hardware
     means that the macro used to define bucks 6-10 needs duplicating
     and tweaking to have a separate macro for 9

   - Fix a series of errors in the definitions of the LDOs the AXP20x
     regulators, some of which had always been present and some of which
     were introduced in the merge window"

* tag 'regulator-fix-v4.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: da9063: Correct module alias prefix to fix module autoloading
  regulator: axp20x: Fix axp22x ldo_io registration error on cold boot
  regulator: axp20x: Fix axp22x ldo_io voltage ranges
  regulator: axp20x: Fix LDO4 linear voltage range
  regulator: s2mps11: Fix invalid selector mask and voltages for buck9
  regulator: gpio: check return value of of_get_named_gpio

Pull regmap fixes from Mark Brown:
 "This is rather too late so it'd be completely understandable if you
  don't want to pull it at this point, I had thought I'd sent this
  earlier but it seems I didn't.  Everything has been in -next for some
  time now.

  The main set of fixes here are mopping up some more issues with MMIO,
  fixing handling of endianness configuration in DT (which just wasn't
  working at all) and cases where the register and value endianness are
  different.

  There is also a fix for bulk register reads on SPMI"

* tag 'regmap-fix-v4.6-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
  regmap: spmi: Fix regmap_spmi_ext_read in multi-byte case
  regmap: mmio: Explicitly say little endian is the defualt in the bus config
  regmap: mmio: Parse endianness definitions from DT
  regmap: Fix implicit inclusion of device.h
  regmap: mmio: Fix value endianness selection
  regmap: fix documentation to match code

Pull media fix from Mauro Carvalho Chehab:
 "A revert fixing a breakage that caused an OOPS on all VB2-based DVB
  drivers.

  We already have a proper fix, but it sounds safer to keep it being
  tested for a while and not hurry, to avoid the risk of another
  regression, specially since this is meant to be c/c to stable.  So,
  for now, let's just revert the broken patch"

* tag 'media/v4.6-6' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  Revert "[media] videobuf2-v4l2: Verify planes array in buffer dequeueing"

Pull drm fixes from Dave Airlie:
 "A bunch of radeon displayport mode setting fixes, and some misc i915
  fixes.

  There is one revert, the MST audio code in i915 was causing some
  oopses, so we've decided just to drop it until next kernel when we can
  fix it properly"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/amdgpu: fix DP mode validation
  drm/radeon: fix DP mode validation
  drm/i915: Bail out of pipe config compute loop on LPT
  drm/radeon: fix PLL sharing on DCE6.1 (v2)
  drm/radeon: fix DP link training issue with second 4K monitor
  Revert "drm/i915: start adding dp mst audio"
  drm/i915/bdw: Add missing delay during L3 SQC credit programming
  drm/i915/lvds: separate border enable readout from panel fitter
  drm/i915: Update CDCLK_FREQ register on BDW after changing cdclk frequency

Pull crypto fix from Herbert Xu:
 "This fixes a bug in the RSA self-test that may cause crashes on some
  architectures such as SPARC"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: testmgr - Use kmalloc memory for RSA input

Merge remote-tracking branches 'regulator/fix/axp20x', 'regulator/fix/da9063', 'regulator/fix/gpio' and 'regulator/fix/s2mps11' into regulator-linus

Merge remote-tracking branches 'regmap/fix/be', 'regmap/fix/doc' and 'regmap/fix/spmi' into regmap-linus

DP mode validation regression fix.
* 'drm-fixes-4.6' of git://people.freedesktop.org/~agd5f/linux:
  drm/amdgpu: fix DP mode validation
  drm/radeon: fix DP mode validation

Patch 562abd39 "xen-netback: support multiple extra info fragments
passed from frontend" contained a mistake which can result in an in-
correct number of responses being generated when handling errors
encountered when processing packets containing extra info fragments.
This patch fixes the problem.
Signed-off-by: NPaul Durrant <paul.durrant@citrix.com>
Reported-by: NJan Beulich <JBeulich@suse.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Acked-by: NWei Liu <wei.liu2@citrix.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Merge tag 'perf-urgent-for-mingo-20160512' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fixes from Arnaldo Carvalho de Melo:

- Fallback to usermode-only counters when perf_event_paranoid > 1, which
  is the case now (Arnaldo Carvalho de Melo)

- Do not reassign parg after collapse_tree() in libtraceevent, which
  may cause tool crashes (Steven Rostedt)

- Fix the build on Fedora Rawhide, where readdir_r() is deprecated and
  also wrt -Werror=unused-const-variable= + x86_32_regoffset_table on
  !x86_64 (Arnaldo Carvalho de Melo)

- Fix the build on Ubuntu 12.04.5, where dwarf_getlocations() isn't
  available, i.e. libdw-dev < 0.157 (Arnaldo Carvalho de Melo)
Signed-off-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: NIngo Molnar <mingo@kernel.org>

Merge fixes from Andrew Morton:
 "4 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mm: thp: calculate the mapcount correctly for THP pages during WP faults
  ksm: fix conflict between mmput and scan_get_next_rmap_item
  ocfs2: fix posix_acl_create deadlock
  ocfs2: revert using ocfs2_acl_chmod to avoid inode cluster lock hang

This will provide fully accuracy to the mapcount calculation in the
write protect faults, so page pinning will not get broken by false
positive copy-on-writes.

total_mapcount() isn't the right calculation needed in
reuse_swap_page(), so this introduces a page_trans_huge_mapcount()
that is effectively the full accurate return value for page_mapcount()
if dealing with Transparent Hugepages, however we only use the
page_trans_huge_mapcount() during COW faults where it strictly needed,
due to its higher runtime cost.

This also provide at practical zero cost the total_mapcount
information which is needed to know if we can still relocate the page
anon_vma to the local vma. If page_trans_huge_mapcount() returns 1 we
can reuse the page no matter if it's a pte or a pmd_trans_huge
triggering the fault, but we can only relocate the page anon_vma to
the local vma->anon_vma if we're sure it's only this "vma" mapping the
whole THP physical range.

Kirill A. Shutemov discovered the problem with moving the page
anon_vma to the local vma->anon_vma in a previous version of this
patch and another problem in the way page_move_anon_rmap() was called.

Andrew Morton discovered that CONFIG_SWAP=n wouldn't build in a
previous version, because reuse_swap_page must be a macro to call
page_trans_huge_mapcount from swap.h, so this uses a macro again
instead of an inline function. With this change at least it's a less
dangerous usage than it was before, because "page" is used only once
now, while with the previous code reuse_swap_page(page++) would have
called page_mapcount on page+1 and it would have increased page twice
instead of just once.

Dean Luick noticed an uninitialized variable that could result in a
rmap inefficiency for the non-THP case in a previous version.

Mike Marciniszyn said:

: Our RDMA tests are seeing an issue with memory locking that bisects to
: commit 61f5d698 ("mm: re-enable THP")
:
: The test program registers two rather large MRs (512M) and RDMA
: writes data to a passive peer using the first and RDMA reads it back
: into the second MR and compares that data.  The sizes are chosen randomly
: between 0 and 1024 bytes.
:
: The test will get through a few (<= 4 iterations) and then gets a
: compare error.
:
: Tracing indicates the kernel logical addresses associated with the individual
: pages at registration ARE correct , the data in the "RDMA read response only"
: packets ARE correct.
:
: The "corruption" occurs when the packet crosse two pages that are not physically
: contiguous.   The second page reads back as zero in the program.
:
: It looks like the user VA at the point of the compare error no longer points to
: the same physical address as was registered.
:
: This patch totally resolves the issue!

Link: http://lkml.kernel.org/r/1462547040-1737-2-git-send-email-aarcange@redhat.comSigned-off-by: NAndrea Arcangeli <aarcange@redhat.com>
Reviewed-by: N"Kirill A. Shutemov" <kirill@shutemov.name>
Reviewed-by: NDean Luick <dean.luick@intel.com>
Tested-by: NAlex Williamson <alex.williamson@redhat.com>
Tested-by: NMike Marciniszyn <mike.marciniszyn@intel.com>
Tested-by: NJosh Collier <josh.d.collier@intel.com>
Cc: Marc Haber <mh+linux-kernel@zugschlus.de>
Cc: <stable@vger.kernel.org>	[4.5]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

A concurrency issue about KSM in the function scan_get_next_rmap_item.

task A (ksmd):				|task B (the mm's task):
					|
mm = slot->mm;				|
down_read(&mm->mmap_sem);		|
					|
...					|
					|
spin_lock(&ksm_mmlist_lock);		|
					|
ksm_scan.mm_slot go to the next slot;	|
					|
spin_unlock(&ksm_mmlist_lock);		|
					|mmput() ->
					|	ksm_exit():
					|
					|spin_lock(&ksm_mmlist_lock);
					|if (mm_slot && ksm_scan.mm_slot != mm_slot) {
					|	if (!mm_slot->rmap_list) {
					|		easy_to_free = 1;
					|		...
					|
					|if (easy_to_free) {
					|	mmdrop(mm);
					|	...
					|
					|So this mm_struct may be freed in the mmput().
					|
up_read(&mm->mmap_sem);			|

As we can see above, the ksmd thread may access a mm_struct that already
been freed to the kmem_cache.  Suppose a fork will get this mm_struct from
the kmem_cache, the ksmd thread then call up_read(&mm->mmap_sem), will
cause mmap_sem.count to become -1.

As suggested by Andrea Arcangeli, unmerge_and_remove_all_rmap_items has
the same SMP race condition, so fix it too.  My prev fix in function
scan_get_next_rmap_item will introduce a different SMP race condition, so
just invert the up_read/spin_unlock order as Andrea Arcangeli said.

Link: http://lkml.kernel.org/r/1462708815-31301-1-git-send-email-zhouchengming1@huawei.comSigned-off-by: NZhou Chengming <zhouchengming1@huawei.com>
Suggested-by: NAndrea Arcangeli <aarcange@redhat.com>
Reviewed-by: NAndrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Geliang Tang <geliangtang@163.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Hanjun Guo <guohanjun@huawei.com>
Cc: Ding Tianhong <dingtianhong@huawei.com>
Cc: Li Bin <huawei.libin@huawei.com>
Cc: Zhen Lei <thunder.leizhen@huawei.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Commit 702e5bc6 ("ocfs2: use generic posix ACL infrastructure")
refactored code to use posix_acl_create.  The problem with this function
is that it is not mindful of the cluster wide inode lock making it
unsuitable for use with ocfs2 inode creation with ACLs.  For example,
when used in ocfs2_mknod, this function can cause deadlock as follows.
The parent dir inode lock is taken when calling posix_acl_create ->
get_acl -> ocfs2_iop_get_acl which takes the inode lock again.  This can
cause deadlock if there is a blocked remote lock request waiting for the
lock to be downconverted.  And same deadlock happened in ocfs2_reflink.
This fix is to revert back using ocfs2_init_acl.

Fixes: 702e5bc6 ("ocfs2: use generic posix ACL infrastructure")
Signed-off-by: NTariq Saeed <tariq.x.saeed@oracle.com>
Signed-off-by: NJunxiao Bi <junxiao.bi@oracle.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <joseph.qi@huawei.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>