Pull rdma fixes from Doug Ledford:
 "This is the second pull request for the rdma subsystem.  Most of the
  patches are small and obvious.  I took two patches in that are larger
  than I wanted this late in the cycle.

  The first is the hfi1 patch that implements a work queue to test the
  QSFP read state.  I originally rejected the first patch for this
  (which would have place up to 20 seconds worth of udelays in their
  probe routine).  They then rewrote it the way I wanted (use delayed
  work tasks to wait asynchronously up to 20 seconds for the QSFP to
  come alive), so I can't really complain about the size of getting what
  I asked for :-/.

  The second is large because it switches the rcu locking in the debugfs
  code.  Since a locking change like this is done all at once, the size
  it what it is.  It resolves a litany of debug messages from the
  kernel, so I pulled it in for -rc.

  The rest are all typical -rc worthy patches I think.

  There will still be a third -rc pull request from the rdma subsystem
  this release.  I hope to have that one ready to go by the end of this
  week or early next.

  Summary:

   - a smattering of small fixes across the core, ipoib, i40iw, isert,
     cxgb4, and mlx4

   - a slightly larger group of fixes to each of mlx5 and hfi1"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma:
  IB/hfi1: Rework debugfs to use SRCU
  IB/hfi1: Make n_krcvqs be an unsigned long integer
  IB/hfi1: Add QSFP sanity pre-check
  IB/hfi1: Fix AHG KDETH Intr shift
  IB/hfi1: Fix SGE length for misaligned PIO copy
  IB/mlx5: Don't return errors from poll_cq
  IB/mlx5: Use TIR number based on selector
  IB/mlx5: Simplify code by removing return variable
  IB/mlx5: Return EINVAL when caller specifies too many SGEs
  IB/mlx4: Don't return errors from poll_cq
  Revert "IB/mlx4: Return EAGAIN for any error in mlx4_ib_poll_one"
  IB/ipoib: Fix memory corruption in ipoib cm mode connect flow
  IB/core: Fix use after free in send_leave function
  IB/cxgb4: Make _free_qp static to silence build warning
  IB/isert: Properly release resources on DEVICE_REMOVAL
  IB/hfi1: Fix the size parameter to find_first_bit
  IB/mlx5: Fix the size parameter to find_first_bit
  IB/hfi1: Clean up type used and casting
  i40iw: Receive notification events correctly
  i40iw: Update hw_iwarp_state

Pull mailbox fixes from Jassi Brar:
 "Misc fixes for BCM mailbox driver

   - Fix build warnings by making static functions used within the file.
   - Check for potential NULL before dereferencing
   - Fix link error by defining HAS_DMA dependency"

* 'mailbox-devel' of git://git.linaro.org/landing-teams/working/fujitsu/integration:
  fix:mailbox:bcm-pdc-mailbox:mark symbols static where possible
  mailbox: bcm-pdc: potential NULL dereference in pdc_shutdown()
  mailbox: Add HAS_DMA Kconfig dependency to BCM_PDC_MBOX

Pull SCSI fixes from James Bottomley:
 "This is really three fixes, but the SES one comes in a bundle of three
  (making the replacement API available properly, using it and removing
  the non-working one).  The SES problem causes an oops on hpsa devices
  because they attach virtual disks to the host which aren't SAS
  attached (the replacement API ignores them).

  The other two fixes are fairly minor: the sense key one means we
  actually resolve a newly added sense key and the RDAC device
  blacklisting is needed to prevent us annoying the universal XPORT lun
  of various RDAC arrays"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: sas: remove is_sas_attached()
  scsi: ses: use scsi_is_sas_rphy instead of is_sas_attached
  scsi: sas: provide stub implementation for scsi_is_sas_rphy
  scsi: blacklist all RDAC devices for BLIST_NO_ULD_ATTACH
  scsi: fix upper bounds check of sense key in scsi_sense_key_string()

Pull regmap fixes from Mark Brown:
 "Several fixes here, the main one being the change from Lars-Peter
  which I'd been letting soak in -next since the merge window in case it
  uncovered further issues as it's a minimal fix rather than a change
  addressing the root cause of the problems (which would've been too
  invasive for -rc):

   - The biggest change is a fix from Lars-Peter to ensure that we don't
     create overlapping rbtree nodes which in turn avoids returning
     corrupt cache values to users, fixing some issues that were exposed
     by some recent optimisations with certain access patterns but had
     been present for a long time.

   - A fix from Elaine Zhang to stop us updating the cache if we get an
     I/O error when writing to the hardware.

   - A fix fromm Maarten ter Huurne to avoid uninitialized defaults in
     cases where we have non-readable registers but are initializing the
     cache by reading from the device"

* tag 'regmap-fix-v4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
  regmap: drop cache if the bus transfer error
  regmap: rbtree: Avoid overlapping nodes
  regmap: cache: Fix num_reg_defaults computation from reg_defaults_raw

Pull spi fixes from Mark Brown:
 "As well as the usual driver fixes there's a couple of non-trivial core
  fixes in here:

   - Fixes for issues reported by Julia Lawall in the changes that were
     sent last time to fix interaction between the bus lock and the
     locking done for the SPI thread.  I'd let this one cook for a while
     to make sure nothing else came up in testing.

   - A fix from Sien Wu for arithmetic overflows when calculating the
     timeout for larger transfers (espcially common with slow buses with
     flashes on them)"

* tag 'spi-fix-v4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: Prevent unexpected SPI time out due to arithmetic overflow
  spi: pxa2xx-pci: fix ACPI-based enumeration of SPI devices
  MAINTAINERS: add myself as Samsung SPI maintainer
  spi: Drop io_mutex in error paths
  spi: sh-msiof: Avoid invalid clock generator parameters
  spi: img-spfi: Remove spi_master_put in img_spfi_remove()
  spi: mediatek: remove spi_master_put in mtk_spi_remove()
  spi: qup: Remove spi_master_put in spi_qup_remove()

Merge tag 'regulator-fix-v4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator

Pull regulator fixes from Mark Brown:
 "Two things here, one an e-mail update for Krzysztof Kozlowski and the
  other a couple of fixes for issues with incorrectly described voltages
  in a couple of the Qualcomm regulator drivers that were breaking MMC
  on some platforms"

* tag 'regulator-fix-v4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: Change Krzysztof Kozlowski's email to kernel.org
  regulator: qcom_smd: Fix voltage ranges for pma8084 ftsmps and pldo
  regulator: qcom_smd: Fix voltage ranges for pm8x41

Pull pin control fixes from Linus Walleij:
 "Nothing special at all, just three SoC-specific driver fixes:

   - Fix routing problems in pistachio (Imagination) and sunxi
     (AllWinner)

   - Fix an interrupt problem in the Cherryview (Intel)"

* tag 'pinctrl-v4.8-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: sunxi: fix uart1 CTS/RTS pins at PG on A23/A33
  pinctrl: cherryview: Do not mask all interrupts in probe
  pinctrl: pistachio: fix mfio pll_lock pinmux

Merge remote-tracking branches 'spi/fix/lock', 'spi/fix/maintainers', 'spi/fix/put', 'spi/fix/pxa2xx', 'spi/fix/sh-msiof' and 'spi/fix/timeout' into spi-linus

Merge remote-tracking branches 'regulator/fix/email' and 'regulator/fix/qcom-smd' into regulator-linus

Pull crypto fixes from Herbert Xu:
 "This fixes a regression in the cryptd code that breaks certain
  accelerated AED algorithms as well as an older regression in the
  caam driver that breaks IPsec"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: caam - fix IV loading for authenc (giv)decryption
  crypto: cryptd - Use correct tfm object for AEAD tracking

Pull kbuild fix from Michal Marek:
 "Fix for 'make deb-pkg'.  The bug got introduced in v4.8-rc1"

* 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:
  builddeb: Skip gcc-plugins when not configured

Pull x86 fix from Thomas Gleixner:
 "A single fix for an AMD erratum so machines without a BIOS fix work"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/AMD: Apply erratum 665 on machines without a BIOS fix

Pull timer fixes from Thomas Gleixner:
 "Two fixlet from the timers departement:

   - A fix for scheduler stalls in the tick idle code affecting
     NOHZ_FULL kernels

   - A trivial compile fix"

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tick/nohz: Fix softlockup on scheduler stalls in kvm guest
  clocksource/drivers/atmel-pit: Fix compilation error

Pull device mapper fixes from Mike Snitzer:

 - a stable fix in both DM crypt and DM log-writes for too large bios
   (as generated by bcache)

 - two other stable fixes for DM log-writes

 - a stable fix for a DM crypt bug that could result in freeing pointers
   from uninitialized memory in the tfm allocation error path

 - a DM bufio cleanup to discontinue using create_singlethread_workqueue()

* tag 'dm-4.8-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm bufio: remove use of deprecated create_singlethread_workqueue()
  dm crypt: fix free of bad values after tfm allocation failure
  dm crypt: fix error with too large bios
  dm log writes: fix check of kthread_run() return value
  dm log writes: fix bug with too large bios
  dm log writes: move IO accounting earlier to fix error path

Pull btrfs fixes from Chris Mason:
 "I'm still prepping a set of fixes for btrfs fsync, just nailing down a
  hard to trigger memory corruption.  For now, these are tested and ready."

* 'for-linus-4.8' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  btrfs: fix one bug that process may endlessly wait for ticket in wait_reserve_ticket()
  Btrfs: fix endless loop in balancing block groups
  Btrfs: kill invalid ASSERT() in process_all_refs()

Pull arm64 fixes from Catalin Marinas:
 "arm64 and arm/perf fixes:

   - arm64 fix: debug exception unmasking on the CPU resume path

   - ARM PMU fixes: memory leak on error path and NULL pointer
     dereference"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: kernel: Fix unmasked debug exceptions when restoring mdscr_el1
  drivers/perf: arm_pmu: Fix NULL pointer dereference during probe
  drivers/perf: arm_pmu: Fix leak in error path

Pull char/misc driver fixes from Greg KH:
 "Here are a number of small driver fixes for 4.8-rc5.

  The largest thing here is deleting an obsolete driver,
  drivers/misc/bh1780gli.c, as the functionality of it was replaced by
  an iio driver a while ago.

  The other fixes are things that have been reported, or reverts of
  broken stuff (the binder change).  All of these changes have been in
  linux-next for a while with no reported issues"

* tag 'char-misc-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  thunderbolt: Don't declare Falcon Ridge unsupported
  thunderbolt: Add support for INTEL_FALCON_RIDGE_2C controller.
  thunderbolt: Fix resume quirk for Falcon Ridge 4C.
  lkdtm: Mark lkdtm_rodata_do_nothing() notrace
  mei: me: disable driver on SPT SPS firmware
  Revert "android: binder: fix dangling pointer comparison"
  drivers/iio/light/Kconfig: SENSORS_BH1780 cleanup
  android: binder: fix dangling pointer comparison
  misc: delete bh1780 driver

Pull driver core fixes from Greg KH:
 "Here are three small fixes for 4.8-rc5.

  One for sysfs, one for kernfs, and one documentation fix, all for
  reported issues.  All of these have been in linux-next for a while"

* tag 'driver-core-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  sysfs: correctly handle read offset on PREALLOC attrs
  documentation: drivers/core/of: fix name of of_node symlink
  kernfs: don't depend on d_find_any_alias() when generating notifications

Pull staging/IIO driver fixes from Greg KH:
 "Here are a number of small fixes for staging and IIO drivers that
  resolve reported problems.

  Full details are in the shortlog.  All of these have been in
  linux-next with no reported issues"

* tag 'staging-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging: (35 commits)
  arm: dts: rockchip: add reset node for the exist saradc SoCs
  arm64: dts: rockchip: add reset saradc node for rk3368 SoCs
  iio: adc: rockchip_saradc: reset saradc controller before programming it
  iio: accel: kxsd9: Fix raw read return
  iio: adc: ti_am335x_adc: Increase timeout value waiting for ADC sample
  iio: adc: ti_am335x_adc: Protect FIFO1 from concurrent access
  include/linux: fix excess fence.h kernel-doc notation
  staging: wilc1000: correctly check if associatedsta has not been found
  staging: wilc1000: NULL dereference on error
  staging: wilc1000: txq_event: Fix coding error
  MAINTAINERS: Add file patterns for ion device tree bindings
  MAINTAINERS: Update maintainer entry for wilc1000
  iio: chemical: atlas-ph-sensor: fix typo in val assignment
  iio: fix sched WARNING "do not call blocking ops when !TASK_RUNNING"
  staging: comedi: ni_mio_common: fix AO inttrig backwards compatibility
  staging: comedi: dt2811: fix a precedence bug
  staging: comedi: adv_pci1760: Do not return EINVAL for CMDF_ROUND_DOWN.
  staging: comedi: ni_mio_common: fix wrong insn_write handler
  staging: comedi: comedi_test: fix timer race conditions
  staging: comedi: daqboard2000: bug fix board type matching code
  ...

Pull serial driver fixes from Greg KH:
 "Here are some small serial driver fixes for 4.8-rc5.  One fixes an
  oft-reported build issue with the fintek driver, another reverts a
  patch that was causing problems, one fixes a crash, and some new
  device ids were added.

  All of these have been in linux-next for a while"

* tag 'tty-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: 8250: added acces i/o products quad and octal serial cards
  serial: 8250_mid: fix divide error bug if baud rate is 0
  Revert "tty/serial/8250: use mctrl_gpio helpers"
  8250/fintek: rename IRQ_MODE macro

Pull USB/PHY fixes from Greg KH:
 "Here are some USB and PHY driver fixes for 4.8-rc5

  Nothing major, lots of little fixes for reported bugs, and a build fix
  for a missing .h file that the phy drivers needed.  All of these have
  been in linux-next for a while with no reported issues"

* tag 'usb-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (24 commits)
  usb: musb: Fix locking errors for host only mode
  usb: dwc3: gadget: always decrement by 1
  usb: dwc3: debug: fix ep name on trace output
  usb: gadget: udc: core: don't starve DMA resources
  USB: serial: option: add WeTelecom 0x6802 and 0x6803 products
  USB: avoid left shift by -1
  USB: fix typo in wMaxPacketSize validation
  usb: gadget: Add the gserial port checking in gs_start_tx()
  usb: dwc3: gadget: don't rely on jiffies while holding spinlock
  usb: gadget: fsl_qe_udc: signedness bug in qe_get_frame()
  usb: gadget: function: f_rndis: socket buffer may be NULL
  usb: gadget: function: f_eem: socket buffer may be NULL
  usb: renesas_usbhs: gadget: fix return value check in usbhs_mod_gadget_probe()
  usb: dwc2: Add reset control to dwc2
  usb: dwc3: core: allow device to runtime_suspend several times
  usb: dwc3: pci: runtime_resume child device
  USB: serial: option: add WeTelecom WM-D200
  usb: chipidea: udc: don't touch DP when controller is in host mode
  USB: serial: mos7840: fix non-atomic allocation in write path
  USB: serial: mos7720: fix non-atomic allocation in write path
  ...

In commit 8ead9dd5 ("devpts: more pty driver interface cleanups") I
made devpts_get_priv() just return the dentry->fs_data directly.  And
because I thought it wouldn't happen, I added a warning if you ever saw
a pts node that wasn't on devpts.

And no, that warning never triggered under any actual real use, but you
can trigger it by creating nonsensical pts nodes by hand.

So just revert the warning, and make devpts_get_priv() return NULL for
that case like it used to.
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: stable@vger.kernel.org # 4.6+
Cc: Eric W Biederman" <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

When reading SPI flash as MTD device, the transfer length is
directly passed to the spi driver. If the requested data size
exceeds 512KB, it will cause the time out calculation to
overflow since transfer length is 32-bit unsigned integer.
This issue is resolved by using 64-bit unsigned integer
to perform the arithmetic.
Signed-off-by: Sien Wu <sien.wu@ni.com>
Acked-by: Brad Keryan <brad.keryan@ni.com>
Acked-by: Gratian Crisan <gratian.crisan@ni.com>
Acked-by: Brad Mouring <brad.mouring@ni.com>

Natinst-ReviewBoard-ID 150232
Signed-off-by: Mark Brown <broonie@kernel.org>

Pull block fixes from Jens Axboe:
 "A collection of fixes for the nvme over fabrics code"

* 'for-linus' of git://git.kernel.dk/linux-block:
  nvme-rdma: Get rid of redundant defines
  nvme-rdma: Get rid of duplicate variable
  nvme: fabrics drivers don't need the nvme-pci driver
  nvme-fabrics: get a reference when reusing a nvme_host structure
  nvme-fabrics: change NQN UUID to big-endian format
  nvme-loop: set sqsize to 0-based value, per spec
  nvme-rdma: fix sqsize/hsqsize per spec
  fabrics: define admin sqsize min default, per spec
  nvmet-rdma: +1 to *queue_size from hsqsize/hrqsize
  nvmet-rdma: Fix use after free
  nvme-rdma: initialize ret to zero to avoid returning garbage

Pull TPM bugfix from James Morris.

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security:
  tpm: invalid self test error message

The driver emits invalid self test error message even though the init
succeeds.
Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Fixes: cae8b441 ("tpm: Factor out common startup code")
Reviewed-by: James Morris <james.l.morris@oracle.com>
Signed-off-by: James Morris <james.l.morris@oracle.com>

Pull ACPI fixes ffrom Rafael Wysocki:
 "Two stable-candidate fixes for the ACPI early device probing code
  added during the 4.4 cycle, one fixing a typo in a stub macro used
  when CONFIG_ACPI is unset and one that prevents sleeping functions
  from being called under a spinlock (Lorenzo Pieralisi)"

* tag 'acpi-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI / drivers: replace acpi_probe_lock spinlock with mutex
  ACPI / drivers: fix typo in ACPI_DECLARE_PROBE_ENTRY macro

Pull power management fixes from Rafael Wysocki:
 "This includes a stable-candidate cpufreq-dt driver problem fix and
  annotations of tracepoints in the runtime PM framework.

  Specifics:

   - Fix the definition of the cpufreq-dt driver's machines table
     introduced during the 4.7 cycle that should be NULL-terminated, but
     the termination entry is missing from it (Wei Yongjun).

   - Annotate tracepoints in the runtime PM framework's core so as to
     allow the functions containing them to be called from the idle code
     path without causing RCU to complain about illegal usage (Paul
     McKenney)"

* tag 'pm-4.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM / runtime: Add _rcuidle suffix to allow rpm_idle() use from idle
  PM / runtime: Add _rcuidle suffix to allow rpm_resume() to be called from idle
  cpufreq: dt: Add terminate entry for of_device_id tables

* pm-cpufreq-fixes:
  cpufreq: dt: Add terminate entry for of_device_id tables

* pm-core-fixes:
  PM / runtime: Add _rcuidle suffix to allow rpm_idle() use from idle
  PM / runtime: Add _rcuidle suffix to allow rpm_resume() to be called from idle

Commit e647b532 ("ACPI: Add early device probing infrastructure")
introduced code that allows inserting driver specific
struct acpi_probe_entry probe entries into ACPI linker sections
(one per-subsystem, eg irqchip, clocksource) that are then walked
to retrieve the data and function hooks required to probe the
respective kernel components.

Probing for all entries in a section is triggered through
the __acpi_probe_device_table() function, that in turn, according
to the table ID a given probe entry reports parses the table
with the function retrieved from the respective section structures
(ie struct acpi_probe_entry). Owing to the current ACPI table
parsing implementation, the __acpi_probe_device_table() function
has to share global variables with the acpi_match_madt() function, so
in order to guarantee mutual exclusion locking is required
between the two functions.

Current kernel code implements the locking through the acpi_probe_lock
spinlock; this has the side effect of requiring all code called
within the lock (ie struct acpi_probe_entry.probe_{table/subtbl} hooks)
not to sleep.

However, kernel subsystems that make use of the early probing
infrastructure are relying on kernel APIs that may sleep (eg
irq_domain_alloc_fwnode(), among others) in the function calls
pointed at by struct acpi_probe_entry.{probe_table/subtbl} entries
(eg gic_v2_acpi_init()), which is a bug.

Since __acpi_probe_device_table() is called from context
that is allowed to sleep the acpi_probe_lock spinlock can be replaced
with a mutex; this fixes the issue whilst still guaranteeing
mutual exclusion.
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Fixes: e647b532 (ACPI: Add early device probing infrastructure)
Cc: 4.4+ <stable@vger.kernel.org> # 4.4+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

When the ACPI_DECLARE_PROBE_ENTRY macro was added in
commit e647b532 ("ACPI: Add early device probing infrastructure"),
a stub macro adding an unused entry was added for the !CONFIG_ACPI
Kconfig option case to make sure kernel code making use of the
macro did not require to be guarded within CONFIG_ACPI in order to
be compiled.

The stub macro was never used since all kernel code that defines
ACPI_DECLARE_PROBE_ENTRY entries is currently guarded within
CONFIG_ACPI; it contains a typo that should be nonetheless fixed.

Fix the typo in the stub (ie !CONFIG_ACPI) ACPI_DECLARE_PROBE_ENTRY()
macro so that it can actually be used if needed.
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Fixes: e647b532 (ACPI: Add early device probing infrastructure)
Cc: 4.4+ <stable@vger.kernel.org> # 4.4+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

AMD F12h machines have an erratum which can cause DIV/IDIV to behave
unpredictably. The workaround is to set MSRC001_1029[31] but sometimes
there is no BIOS update containing that workaround so let's do it
ourselves unconditionally. It is simple enough.

[ Borislav: Wrote commit message. ]
Signed-off-by: Emanuel Czirai <icanrealizeum@gmail.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Yaowu Xu <yaowu@google.com>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/20160902053550.18097-1-bp@alien8.deSigned-off-by: Thomas Gleixner <tglx@linutronix.de>

The debugfs RCU trips many debug kernel warnings because of potential
sleeps with an RCU read lock held. This includes both user copy calls
and slab allocations throughout the file.

This patch switches the RCU to use SRCU for file remove/access
race protection.

In one case, the SRCU is implicit in the use of the raw debugfs file
object and just works.

In the seq_file case, a wrapper around seq_read() and seq_lseek() is
used to enforce the SRCU using the debugfs supplied functions
debugfs_use_file_start() and debugfs_use_file_stop().

The sychronize_rcu() is deleted since the SRCU prevents the remove
access race.

The RCU locking is kept for qp_stats since the QP hash list is
protected using the non-sleepable RCU.
Reviewed-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

The global variable n_krcvqs stores the sum of the number of kernel
receive queues of VLs 0-7 which the user can pass to the driver through
the module parameter array krcvqs which is of type unsigned integer. If
the user passes large value(s) into krcvqs parameter array, it can cause
an arithmetic overflow while calculating n_krcvqs which is also of type
unsigned int. The overflow results in an incorrect value of n_krcvqs
which can lead to kernel crash while loading the driver.

Fix by changing the data type of n_krcvqs to unsigned long. This patch
also changes the data type of other variables that get their values from
n_krcvqs.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

Sometimes a QSFP device does not respond in the expected time
after a power-on.  Add a read pre-check/retry when starting
the link on driver load.
Reviewed-by: Easwar Hariharan <easwar.hariharan@intel.com>
Signed-off-by: Dean Luick <dean.luick@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

In the set_txreq_header_ahg(), The KDETH Intr bit is obtained from the
header in the user sdma request using a KDETH_GET shift and mask macro.
This value is then futher right shifted by 16 causing us to lose the
value i.e it is shifted to zero, leading to the following
smatch warning:
drivers/infiniband/hw/hfi1/user_sdma.c:1482 set_txreq_header_ahg()
warn: mask and shift to zero

The Intr bit should be left shifted into its correct position in the
KDETH header before the AHG update.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Mitko Haralanov <mitko.haralanov@intel.com>
Reviewed-by: Harish Chegondi <harish.chegondi@intel.com>
Signed-off-by: Jubin John <jubin.john@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

When trying to align the source pointer and there's a byte carry
in an SGE copy, bytes are borrowed from the next quad-word X to
complete the required quad-word copy. Then, the SGE length is
reduced by the number of borrowed bytes. After this, if the
remaining number of bytes from quad-word X (extra bytes) is
greater than the new SGE length, the number of extra bytes needs
to be updated to the new SGE length. Otherwise, when the
SGE length gets updated again after the extra bytes are read to
create the new byte carry, it goes negative, which then becomes
a very large number as the SGE length is an unsigned integer.
This causes SGE buffer to be over-read.
Reviewed-by: Dean Luick <dean.luick@intel.com>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>