counter: ti-eqep: Convert to new counter registration

This fixes device lifetime issues where it was possible to free a live
struct device.

Fixes: f213729f6796 ("counter: new TI eQEP driver")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Acked-by: David Lechner <david@lechnology.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-23-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: stm32-lptimer-cnt: Convert to new counter registration

This fixes device lifetime issues where it was possible to free a live
struct device.

Fixes: 597f55e3f36c ("counter: stm32-lptimer: add counter device")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-22-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: stm32-timer-cnt: Convert to new counter registration

This fixes device lifetime issues where it was possible to free a live
struct device.

Fixes: ad29937e206f ("counter: Add STM32 Timer quadrature encoder")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-21-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: microchip-tcb-capture: Convert to new counter registration

This fixes device lifetime issues where it was possible to free a live
struct device.

Fixes: 106b104137fd ("counter: Add microchip TCB capture counter")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-20-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: ftm-quaddec: Convert to new counter registration

This fixes device lifetime issues where it was possible to free a live
struct device.

Fixes: a3b9a99980d9 ("counter: add FlexTimer Module Quadrature decoder counter driver")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-19-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: intel-qep: Convert to new counter registration

This fixes device lifetime issues where it was possible to free a live
struct device.

Fixes: b711f687a1c1 ("counter: Add support for Intel Quadrature Encoder Peripheral")
Tested-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-18-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: interrupt-cnt: Convert to new counter registration

This fixes device lifetime issues where it was possible to free a live
struct device.

Fixes: a55ebd47f21f ("counter: add IRQ or GPIO based counter")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-17-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: 104-quad-8: Convert to new counter registration

This fixes device lifetime issues where it was possible to free a live
struct device.

Fixes: f1d8a071d45b ("counter: 104-quad-8: Add Generic Counter interface support")
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-16-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: Update documentation for new counter registration functions

In order to replace the counter registration API also update the
documentation to the new way.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-15-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: Provide alternative counter registration functions

The current implementation gets device lifetime tracking wrong. The
problem is that allocation of struct counter_device is controlled by the
individual drivers but this structure contains a struct device that
might have to live longer than a driver is bound. As a result a command
sequence like:

{ sleep 5; echo bang; } > /dev/counter0 &
sleep 1;
echo 40000000.timer:counter > /sys/bus/platform/drivers/stm32-timer-counter/unbind

can keep a reference to the struct device and unbinding results in
freeing the memory occupied by this device resulting in an oops.

This commit provides two new functions (plus some helpers):
 - counter_alloc() to allocate a struct counter_device that is
   automatically freed once the embedded struct device is released
 - counter_add() to register such a device.

Note that this commit doesn't fix any issues, all drivers have to be
converted to these new functions to correct the lifetime problems.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-14-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: stm32-timer-cnt: Convert to counter_priv() wrapper

This is a straight forward conversion to the new counter_priv() wrapper.

Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-13-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: stm32-lptimer-cnt: Convert to counter_priv() wrapper

This is a straight forward conversion to the new counter_priv() wrapper.

Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-12-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: ti-eqep: Convert to counter_priv() wrapper

This is a straight forward conversion to the new counter_priv() wrapper.

Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Acked-by: David Lechner <david@lechnology.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-11-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: ftm-quaddec: Convert to counter_priv() wrapper

This is a straight forward conversion to the new counter_priv() wrapper.

Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-10-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: intel-qep: Convert to counter_priv() wrapper

This is a straight forward conversion to the new counter_priv() wrapper.

Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-9-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: microchip-tcb-capture: Convert to counter_priv() wrapper

This is a straight forward conversion to the new counter_priv() wrapper.

Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-8-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: interrupt-cnt: Convert to counter_priv() wrapper

This is a straight forward conversion to the new counter_priv() wrapper.

Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-7-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: 104-quad-8: Convert to counter_priv() wrapper

This is a straight forward conversion to the new counter_priv() wrapper.

Acked-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-6-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: Provide a wrapper to access device private data

For now this just wraps accessing struct counter_device::priv. However
this is about to change and converting drivers to this helper
individually makes fixing device lifetime issues result in easier to
review patches.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-5-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: microchip-tcb-capture: Drop unused platform_set_drvdata()

The driver doesn't ever use platform_get_drvdata, so drop this unused
call.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-4-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: ftm-quaddec: Drop unused platform_set_drvdata()

The driver doesn't ever use platform_get_drvdata, so drop this unused
call.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-3-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

counter: Use container_of instead of drvdata to track counter_device

The counter core uses drvdata to find a struct counter_device from a
struct device. However as the device is a member of struct counter_device,
the lookup can be done faster (and a bit type safe) using container_of.

There are no other users of drvdata, so the call to dev_set_drvdata can
go away, too.

Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Acked-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20211230150300.72196-2-u.kleine-koenig@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mei: hbm: fix client dma reply status

Don't blindly copy status value received from the firmware
into internal client status field,
It may be positive and ERR_PTR(ret) will translate it
into an invalid address and the caller will crash.

Put the error code into the client status on failure.

Fixes: 369aea845951 ("mei: implement client dma setup.")
Cc: <stable@vger.kernel.org> # v5.11+
Reported-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Tested-by: : Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Acked-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Link: https://lore.kernel.org/r/20211228082047.378115-1-tomas.winkler@intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: lattice-ecp3-config: Fix task hung when firmware load failed

When firmware load failed, kernel report task hung as follows:

INFO: task xrun:5191 blocked for more than 147 seconds.
      Tainted: G        W         5.16.0-rc5-next-20211220+ #11
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
task:xrun            state:D stack:    0 pid: 5191 ppid:   270 flags:0x00000004
Call Trace:
 __schedule+0xc12/0x4b50 kernel/sched/core.c:4986
 schedule+0xd7/0x260 kernel/sched/core.c:6369 (discriminator 1)
 schedule_timeout+0x7aa/0xa80 kernel/time/timer.c:1857
 wait_for_completion+0x181/0x290 kernel/sched/completion.c:85
 lattice_ecp3_remove+0x32/0x40 drivers/misc/lattice-ecp3-config.c:221
 spi_remove+0x72/0xb0 drivers/spi/spi.c:409

lattice_ecp3_remove() wait for signals from firmware loading, but when
load failed, firmware_load() does not send this signal. This cause
device remove hung. Fix it by sending signal even if load failed.

Fixes: 781551df57c7 ("misc: Add Lattice ECP3 FPGA configuration via SPI")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Link: https://lore.kernel.org/r/20211228125522.3122284-1-weiyongjun1@huawei.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Merge tag 'phy-for-5.17' of git://git./linux/kernel/git/phy/linux-phy into char-misc-next

Vinod writes:

phy-for-5.17

  - New support:
        - Qualcomm eDP PHY driver
- Qualcomm SM8450 UFS, USB2, USB3, PCIe0 and PCIe1 phy support
- Lan966x ethernet serdes PHY driver
- Support for uniphier NXI & Pro4 SoC
        - Qualcomm SM6350 USB2 support
- Amlogic Meson8 HDMI TX PHY driver
- Rockchip rk3568 usb2 support
- Intel Thunder Bay eMMC PHY driver
- Freescale IMX8 PCIe phy driver

  - Updates:
- Cadence Sierra driver updates for multilink configurations
        - Bcm usb2 updates for Phy reg space

* tag 'phy-for-5.17' of git://git.kernel.org/pub/scm/linux/kernel/git/phy/linux-phy: (72 commits)
  phy: cadence: Sierra: Add support for derived reference clock output
  dt-bindings: phy: cadence-sierra: Add clock ID for derived reference clock
  phy: cadence: Sierra: Add PCIe + QSGMII PHY multilink configuration
  phy: cadence: Sierra: Add support for PHY multilink configurations
  phy: cadence: Sierra: Fix to get correct parent for mux clocks
  phy: cadence: Sierra: Update single link PCIe register configuration
  phy: cadence: Sierra: Check PIPE mode PHY status to be ready for operation
  phy: cadence: Sierra: Check cmn_ready assertion during PHY power on
  phy: cadence: Sierra: Add PHY PCS common register configurations
  phy: cadence: Sierra: Rename some regmap variables to be in sync with Sierra documentation
  phy: cadence: Sierra: Add support to get SSC type from device tree
  dt-bindings: phy: cadence-sierra: Add binding to specify SSC mode
  dt-bindings: phy: cadence-torrent: Rename SSC macros to use generic names
  phy: cadence: Sierra: Prepare driver to add support for multilink configurations
  phy: cadence: Sierra: Use of_device_get_match_data() to get driver data
  phy: mediatek: Fix missing check in mtk_mipi_tx_probe
  phy: uniphier-usb3ss: fix unintended writing zeros to PHY register
  phy: phy-mtk-tphy: use new io helpers to access register
  phy: phy-mtk-xsphy: use new io helpers to access register
  phy: mediatek: add helpers to update bits of registers
  ...

Merge tag 'soundwire-5.17-rc1' of git://git./linux/kernel/git/vkoul/soundwire into char-misc-next

Vinod writes:

soundwire updates for 5.17-rc1

 - Remove redundant version number read in qcom driver

* tag 'soundwire-5.17-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/soundwire:
  soundwire: qcom: remove redundant version number read

Merge tag 'iio-for-5.17b' of https://git./linux/kernel/git/jic23/iio into char-misc-next

Jonathan writes:

2nd set of new device support etc for IIO in the 5.17 cycle.

A small additional set of things that just missed the previous
pull request and have mostly been through plenty of review before the
holiday period began (or are trivial).  I've not taken some other series
on the list to allow for more eyes after the holiday period.

New device support
* adi,admv1013
  - New driver for this wideband microwave upconverter including dt-bindings
    and some device specific ABI due to the need to describe phase calibrations
    of a differential channel on both i and q phases. Previously we could
    do differential or i/q but not both on the same channel. The driver
    ABI uses a workaround for core support which will do until we know if
    this is a common requirement for which a more generic solution is
    needed.

MAINTAINERS:
* Add Haibo Chen as a maintainer for various NXP SoC ADCs.

Minor cleanup:
* sunrise_co2
  - Make sure an uninitialized value isn't used to set *val in read_raw().
    Not a real bug, but a compiler or reviewer can't tell that based
    on what they can see locally.

* tag 'iio-for-5.17b' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio:
  iio: chemical: sunrise_co2: set val parameter only on success
  dt-bindings:iio:adc: update the maintainer of vf610-adc
  MAINTAINERS: add imx7d/imx6sx/imx6ul/imx8qxp and vf610 adc maintainer
  Documentation:ABI:testing:admv1013: add ABI docs
  dt-bindings: iio: frequency: add admv1013 doc
  iio: frequency: admv1013: add support for ADMV1013

cxl: use default_groups in kobj_type

There are currently 2 ways to create a set of sysfs files for a
kobj_type, through the default_attrs field, and the default_groups
field.  Move the cxl code to use default_groups field which has been the
preferred way since aa30f47cf666 ("kobject: Add support for default
attribute groups to kobj_type") so that we can soon get rid of the
obsolete default_attrs field.

Cc: Frederic Barrat <fbarrat@linux.ibm.com>
Cc: Andrew Donnellan <ajd@linux.ibm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20211228131350.249532-1-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

UIO: use default_groups in kobj_type

There are currently 2 ways to create a set of sysfs files for a
kobj_type, through the default_attrs field, and the default_groups
field.  Move the UIO code to use default_groups field which has been the
preferred way since aa30f47cf666 ("kobject: Add support for default
attribute groups to kobj_type") so that we can soon get rid of the
obsolete default_attrs field.

Link: https://lore.kernel.org/r/20211228131319.249324-1-gregkh@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

iio: chemical: sunrise_co2: set val parameter only on success

Clang static analysis reports this representative warning

sunrise_co2.c:410:9: warning: Assigned value is garbage or undefined
  *val = value;
       ^ ~~~~~

The ealier call to sunrise_read_word can fail without setting
value.  So defer setting val until we know the read was successful.

Fixes: c397894e24f1 ("iio: chemical: Add Senseair Sunrise 006-0-007 driver")
Signed-off-by: Tom Rix <trix@redhat.com>
Link: https://lore.kernel.org/r/20211224150833.3278236-1-trix@redhat.com
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>

phy: cadence: Sierra: Add support for derived reference clock output

Sierra has derived differential reference clock output which is sourced
after the spread spectrum generation has been added. Add support to drive
derived reference clock out of serdes. Model this derived clock as a
"clock" so that platforms using this can enable it.

Sierra Main LC VCO PLL divider 1 clock is programmed to output 100MHz
clock output.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-16-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

dt-bindings: phy: cadence-sierra: Add clock ID for derived reference clock

Add clock ID for Sierra derived reference clock.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Acked-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20211223060137.9252-15-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Add PCIe + QSGMII PHY multilink configuration

Add register sequences for PCIe + QSGMII PHY multilink configuration.
PHY configuration for multi-link operation is done in two steps.
e.g. Consider a case for a 4 lane PHY with PCIe using 2 lanes and QSGMII
other 2 lanes. Sierra PHY has 2 PLLs, viz. PLLLC and PLLLC1. So in this
case, PLLLC is used for PCIe and PLLLC1 is used for QSGMII.

PHY is configured in two steps as described below.

[1] For first step, the register values are selected as
    [TYPE_PCIE][TYPE_QSGMII][ssc].
    This will configure PHY registers associated for PCIe involving PLLLC
    registers and registers for first 2 lanes of PHY.
[2] In second step, the register values are selected as
    [TYPE_QSGMII][TYPE_PCIE][ssc].
    This will configure PHY registers associated for QSGMII involving
    PLLLC1 registers and registers for other 2 lanes of PHY.

This completes the PHY configuration for multilink operation.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-14-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Add support for PHY multilink configurations

Add support for multilink configuration of Sierra PHY. Currently,
maximum two links are supported.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-13-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Fix to get correct parent for mux clocks

Fix get_parent() callback to return the correct index of the parent for
PLL_CMNLC1 clock. Add a separate table of register values corresponding
to the parent index for PLL_CMNLC1. Update set_parent() callback
accordingly.

Fixes: 28081b72859f ("phy: cadence: Sierra: Model PLL_CMNLC and PLL_CMNLC1 as clocks (mux clocks)")
Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-12-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Update single link PCIe register configuration

Add single link PCIe register configurations for no SSC and internal
SSC. Also, add missing PMA lane registers for external SSC.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-11-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Check PIPE mode PHY status to be ready for operation

PIPE phy status is used to communicate the completion of several PHY
functions. Check if PHY is ready for operation while configured for
PIPE mode during startup.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-10-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Check cmn_ready assertion during PHY power on

Check if PMA cmn_ready is set indicating the startup process is complete.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-9-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Add PHY PCS common register configurations

Add PHY PCS common register configuration sequences for single link.
Update single link PCIe register sequence accordingly.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-8-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Rename some regmap variables to be in sync with Sierra documentation

No functional change. Rename some regmap variables as mentioned in Sierra
register description documentation.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-7-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Add support to get SSC type from device tree

Add support to get SSC type from DT.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-6-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

dt-bindings: phy: cadence-sierra: Add binding to specify SSC mode

Add binding to specify Spread Spectrum Clocking mode used.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20211223060137.9252-5-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

dt-bindings: phy: cadence-torrent: Rename SSC macros to use generic names

Rename SSC macros to use generic names instead of PHY specific names,
so that they can be used to specify SSC modes for both Torrent and
Sierra. Renaming the macros should not affect the things as these are
not being used in any DTS file yet.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Acked-by: Rob Herring <robh@kernel.org>
Link: https://lore.kernel.org/r/20211223060137.9252-4-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Prepare driver to add support for multilink configurations

Sierra driver currently supports single link configurations only. Prepare
driver to support multilink multiprotocol configurations along with
different SSC modes.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-3-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: cadence: Sierra: Use of_device_get_match_data() to get driver data

Use of_device_get_match_data() to get driver data instead of boilerplate
code.

Signed-off-by: Swapnil Jakhade <sjakhade@cadence.com>
Reviewed-by: Aswath Govindraju <a-govindraju@ti.com>
Link: https://lore.kernel.org/r/20211223060137.9252-2-sjakhade@cadence.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

phy: mediatek: Fix missing check in mtk_mipi_tx_probe

The of_device_get_match_data() function may return NULL.
Add check to prevent potential null dereference.

Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20211224082103.7658-1-linmq006@gmail.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

mei: cleanup status before client dma setup call

The upper layer may retry call to mei_cl_dma_alloc_and_map(),
in that case the client status may be non-zero after the previous call
and the wait condition will be true immediately.
Set cl->status to zero to allow waiting for an actual result
from the firmware.

Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Link: https://lore.kernel.org/r/20211223094705.204624-2-tomas.winkler@intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mei: add POWERING_DOWN into device state print

The POWERING_DOWN state string was missing from
the device states list, add it.

Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Link: https://lore.kernel.org/r/20211223094705.204624-1-tomas.winkler@intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Merge tag 'misc-habanalabs-next-2021-12-27' of https://git./linux/kernel/git/ogabbay/linux into char-misc-next

Oded writes:

This tag contains habanalabs driver changes for v5.17:

- Support reset-during-reset. In case the f/w notifies the driver
  that the f/w is going to reset the device, the driver should
  support that even if it is in the middle of doing another
  reset

- Support events from f/w that arrive during device resets.
  These events would be ignored which is bad as critical errors
  would not be reported and treated by the driver.

- Don't kill processes that hold the control device open during
  hard-reset of the device. The control device operations can't
  crash if done during hard-reset. And usually, only monitoring
  applications are using the control device, so killing them
  defies their purpose.

- Fix handling of hwmon nodes when working with legacy f/w

- Change the compute context pointer to be boolean. This pointer
  was abused by multiple code paths that wanted fast access to
  the compute context structure.

- Add uapi to fetch historical errors. This is necessary as errors
  sometimes result in hard-reset where the user application is
  being terminated.

- Optimize GAUDI's MMU cache invalidation.

- Add support for loading the latest f/w.

- Add uapi to fetch HBM replacement and pending rows information.

- Multiple bug fixes to the reset code.

- Multiple bug fixes for Multi-CS ioctl code.

- Multiple bug fixes for wait-for-interrupt ioctl code.

- Many small bug fixes and cleanups.

* tag 'misc-habanalabs-next-2021-12-27' of https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux: (70 commits)
  habanalabs: support hard-reset scheduling during soft-reset
  habanalabs: add a lock to protect multiple reset variables
  habanalabs: refactor reset information variables
  habanalabs: handle skip multi-CS if handling not done
  habanalabs: add CPU-CP packet for engine core ASID cfg
  habanalabs: replace some -ENOTTY with -EINVAL
  habanalabs: fix comments according to kernel-doc
  habanalabs: fix endianness when reading cpld version
  habanalabs: change wait_for_interrupt implementation
  habanalabs: prevent wait if CS in multi-CS list completed
  habanalabs: modify cpu boot status error print
  habanalabs: clean MMU headers definitions
  habanalabs: expose soft reset sysfs nodes for inference ASIC
  habanalabs: sysfs support for two infineon versions
  habanalabs: keep control device alive during hard reset
  habanalabs: fix hwmon handling for legacy f/w
  habanalabs: add current PI value to cpu packets
  habanalabs: remove in_debug check in device open
  habanalabs: return correct clock throttling period
  habanalabs: wait again for multi-CS if no CS completed
  ...

Merge tag 'extcon-next-for-5.17' of git://git./linux/kernel/git/chanwoo/extcon into char-misc-next

Chanwoo writes:

Update extcon next for v5.17

Detailed description for this pull request:
1. Remove duplicate code in extcon_set_state_sync() in extcon core
2. Fix non-kernel-doc comment for extcon-usb-gpio.c

* tag 'extcon-next-for-5.17' of git://git.kernel.org/pub/scm/linux/kernel/git/chanwoo/extcon:
  extcon: Deduplicate code in extcon_set_state_sync()
  extcon: usb-gpio: fix a non-kernel-doc comment

Merge tag 'icc-5.17-rc1' of git://git./linux/kernel/git/djakov/icc into char-misc-next

Georgi writes:

interconnect changes for 5.17

Here are the interconnect changes for the 5.17-rc1 merge window
consisting of new drivers, minor changes and fixes.

New drivers:
 - New driver for MSM8996 platforms
 - New driver for SC7280 EPSS L3 hardware
 - New driver for QCM2290 platforms
 - New driver for SM8450 platforms

Driver changes:
 - dt-bindings: interconnect: Combine SDM660 bindings into RPM schema
 - icc-rpm: Add support for bus power domain
 - icc-rpm: Use NOC_QOS_MODE_INVALID for qos_mode check
 - icc-rpm: Define ICC device type
 - icc-rpm: Add QNOC type QoS support
 - icc-rpm: Support child NoC device probe
 - icc-rpm: Prevent integer overflow in rate
 - icc-rpmh: Add BCMs to commit list in pre_aggregate

Signed-off-by: Georgi Djakov <djakov@kernel.org>
* tag 'icc-5.17-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/djakov/icc:
  interconnect: qcom: Add QCM2290 driver support
  dt-bindings: interconnect: Add Qualcomm QCM2290 NoC support
  interconnect: icc-rpm: Support child NoC device probe
  interconnect: icc-rpm: Add QNOC type QoS support
  interconnect: icc-rpm: Define ICC device type
  interconnect: qcom: Add SM8450 interconnect provider driver
  dt-bindings: interconnect: Add Qualcomm SM8450 DT bindings
  interconnect: qcom: rpm: Prevent integer overflow in rate
  interconnect: icc-rpm: Use NOC_QOS_MODE_INVALID for qos_mode check
  interconnect: qcom: icc-rpmh: Add BCMs to commit list in pre_aggregate
  interconnect: qcom: Add MSM8996 interconnect provider driver
  dt-bindings: interconnect: Add Qualcomm MSM8996 DT bindings
  interconnect: icc-rpm: Add support for bus power domain
  dt-bindings: interconnect: Combine SDM660 bindings into RPM schema
  interconnect: qcom: Add EPSS L3 support on SC7280
  dt-bindings: interconnect: Add EPSS L3 DT binding on SC7280

habanalabs: support hard-reset scheduling during soft-reset

As hard-reset can be requested during soft-reset, driver must allow
it or else critical events received during soft-reset will be
ignored.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add a lock to protect multiple reset variables

Atomic operations during reset are replaced by a spinlock in order
to have the ability to protect more than a single variable.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: refactor reset information variables

Unify variables related to device reset, which will help us to
add some new reset functionality in future patches.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: handle skip multi-CS if handling not done

This patch fixes issue in which we have timeout for multi-CS although
the CS in the list actually completed.

Example scenario (the two threads marked as WAIT for the thread that
handles the wait_for_multi_cs and CMPL as the thread that signal
completion for both CS and multi-CS):
1. Submit CS with sequence X
2. [WAIT]: call wait_for_multi_cs with single CS X
3. [CMPL]: CS X do invoke complete_all for both CS and multi-CS
           (multi_cs_completion_done still false)
4. [WAIT]: enter poll_fences, reinit the completion and find the CS
           as completed when asking on the fence but multi_cs_done is
   still false it returns that no CS actually completed
5. [CMPL]: set multi_cs_handling_done as true
6. [WAIT]: wait for completion but no CS to awake the wait context
           and hence wait till timeout

Solution: if CS detected as completed in poll_fences but multi_cs_done
          is still false invoke complete_all to the multi-CS completion
  and so it will not go to sleep in wait_for_completion but
  rather will have a "second chance" to wait for
  multi_cs_completion_done.

Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add CPU-CP packet for engine core ASID cfg

In some cases the driver cannot configure ASID of some engines due to
the security level of the relevant registers.
For this a new CPU-CP packet is introduced, which will allow the driver
to ask the F/W to do this configuration instead.

Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: replace some -ENOTTY with -EINVAL

-ENOTTY is returned in case of error in the ioctl arguments themselves,
such as function that doesn't exists.

In all other cases, where the error is in the arguments of the custom
data structures that we define that are passed in the various ioctls,
we need to return -EINVAL.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: fix comments according to kernel-doc

Fix missing fields, descriptions not according to kernel-doc style.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: fix endianness when reading cpld version

Current sysfs implementation does not take endianness into
consideration when dumping the cpld version.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: change wait_for_interrupt implementation

Currently the cq counters are allocated in userspace memory,
and mapped by the driver to the device address space.

A new requirement that is part of new future API related to this one,
requires that cq counters will be allocated in kernel memory.

We leverage the existing cb_create API with KERNEL_MAPPED flag set to
allocate this memory.

That way we gain two things:
1. The memory cannot be freed while in use since it's protected
by refcount in driver.

2. No need to wake up the user thread upon each interrupt from CQ,
because the kernel has direct access to the counter. Therefore,
it can make comparison with the target value in the interrupt
handler and wake up the user thread only if the counter reaches the
target value. This is instead of waking the thread up to copy counter
value from user then go sleep again if target value wasn't reached.

Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: prevent wait if CS in multi-CS list completed

By the original design we assumed that if we "miss" multi CS completion
it is of no severe consequence as we'll just call wait_for_multi_cs
again.

Sequence of events for such scenario:
1. user submit CS with sequence N
2. user calls wait for multi-CS with only CS #N in the list
3. the multi CS call starts with poll of the CSs but find that none
   completed (while CS #N did not completed yet)
4. now, multi CS #N complete but multi CS CTX was not yet created for
   the above multi-CS. so, attempt to complete multi-CS fails (as no
   multi CS CTX exist)
5. wait_for_multi_cs call now does init_wait_multi_cs_completion (and
   for this create the multi-CS CTX)
6. wait_for_multi_cs wits on completion but will not get one as CS #N
   already completed

To fix the issue we initialize the multi-CS CTX prior polling the
fences.

Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: modify cpu boot status error print

As BTL can be replaced by ROM we should modify relevant error print.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: clean MMU headers definitions

During the MMU development the MMU header files were left with unclean
definitions:

- MMU "version specific" definitions that were left in the mmu_general
  file
- unused definitions

This patch attempts, where possible, to keep definitions that can serve
multiple MMU versions (but that are not tightly bound with specific MMU
arch) in the mmu_general header file (e.g. different definitions for
number of HOPs).

Otherwise, move MMU version specific definitions (e.g. HOPs masks and
shifts) to the specific MMU version file.

Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: expose soft reset sysfs nodes for inference ASIC

As we allow soft-reset to be performed only on inference devices,
having the sysfs nodes may cause a confusion. Hence, we remove those
nodes on training ASICs.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: sysfs support for two infineon versions

Currently sysfs support dumping a single infineon version, in
future asics we will have two infineon versions.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: keep control device alive during hard reset

Need to allow user retrieve data during reset and afterwards without
the need to reopen the device.
Did it by seperating the user peocesses list into two lists:
1. fpriv_list which contains list of user processes that opened
   the device (currently only one).
2. fpriv_ctrl_list which contains list of user processes that opened
   the control device. This processes in this list shall not be
   killed during reset, only when the device is suddenly removed from
   PCI chain.

Signed-off-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: fix hwmon handling for legacy f/w

In legacy f/w that use old hwmon.h file, the values of the hwmon
enums are different than the values that are in newer kernels (5.6
and above).

Therefore, to support working with those f/w, we need to do some
fixup before registering with the hwmon subsystem and also when
calling the functions that communicate with the f/w to retrieve
sensors information.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add current PI value to cpu packets

In order to increase cpucp messaging reliability we will add
the current PI value to the descriptor sent to F/W.
F/W will wait for the PI value as an indication of a valid packet.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: remove in_debug check in device open

The driver supports only a single user anyway, so there is no point
in checking whether we are in_debug state when a user tries to open
the device, because if we are in_debug, it means a user is already
using the device.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: return correct clock throttling period

Current clock throttling period returned from driver was wrong due
to wrong time comparison.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: wait again for multi-CS if no CS completed

The original multi-CS design assumption that stream masters are used
exclusively (i.e. multi-CS with set of stream master QIDs will not get
completed by CS not from the multi-CS set) is inaccurate.

Thus multi-CS behavior is now modified not to treat such case as an
error.

Instead, if we have multi-CS completion but we detect that no CS from
the list is actually completed we will do another multi-CS wait (with
modified timeout).

Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: remove compute context pointer

It was an error to save the compute context's pointer in the device
structure, as it allowed its use without proper ref-cnt.

Change the variable to a flag that only indicates whether there is
an active compute context. Code that needs the pointer will now
be forced to use proper internal APIs to get the pointer.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add helper to get compute context

There are multiple places where the code needs to get the context's
pointer and increment its ref cnt. This is the proper way instead
of using the compute context pointer in the device structure.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: fix etr asid configuration

Pass the user's context pointer into the etr configuration function
to extract its ASID.

Using the compute_ctx pointer is an error as it is just an indication
of whether a user has opened the compute device.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: save ctx inside encaps signal

Compute context pointer in hdev shouldn't be used for fetching the
context's pointer.

If an object needs the context's pointer, it should get it while
incrementing its kref, and when the object is released, put it.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: remove redundant check on ctx_fini

The driver supports only a single context. Therefore, no need to check
if the user context that is closed is the compute context. The user
context, if exists, is always the compute context.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: free signal handle on failure

Fix a bug where in case of failure to allocate idr, the handle's
memory wasn't freed as part of the error handling code.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add missing kernel-doc comments for hl_device fields

Add missing kernel-doc comments for the "last_error" and
"stream_master_qid_arr" fields of the "hl_device" structure".

Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: pass reset flags to reset thread

The reset flags used by the reset thread are currently a mix of
hard-coded values and a specific flag which is passed from the context
that initiates the reset.
To make it easier to pass more flags in future from this context to the
reset thread, modify it to pass all the original reset flags to the
thread.

Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: enable access to info ioctl during hard reset

Because info ioctl is used to retrieve data, some of its opcodes may be
used during hard reset.
Other ioctls should be blocked while device is not operational.

Signed-off-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add SOB information to signal submission uAPI

For debug purpose, add SOB address and SOB initial counter value
before current submission to uAPI output.

Using SOB address and initial counter, user can calculate how much of
the submmision has been completed.

Signed-off-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: skip read fw errors if dynamic descriptor invalid

Reporting FW errors involves reading of the error registers.

In case we have a corrupted FW descriptor we cannot do that since the
dynamic scratchpad is potentially corrupted as well and may cause kernel
crush when attempting access to a corrupted register offset.

Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: handle events during soft-reset

Driver should handle events during soft-reset as F/W is not
going through reset and it keeps sending events towards host.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: change misleading IRQ warning during reset

Currently we dump the physical IRQ line index in host if an event
is received during reset. This ID is confusing as it means nothing
to the user.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add power information type to POWER_GET packet

In new f/w versions, it is required to explicitly indicate the power
information type when querying the F/W for power info.
When getting the current power level it should be set to power_input.

Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add more info ioctls support during reset

Some info ioctls can be served even if the device is disabled or
in reset. Hence, we enable more info ioctls during reset, as these
ioctls do not require any H/W nor F/W communication.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: fix race condition in multi CS completion

Race example scenario:
1. User have 2 threads that waits on multi CS:
   - thread_0 waits on QID 0 and uses multi CS context 0.
   - thread_1 waits on QID 1 and uses multi CS context 1.
2. thread_1 got completion and release multi CS context 1.
3. CS related to multi CS of thread_0 starts executing
   complete_multi_cs function, the first iteration of the loop
   completes the multi CS of thread_0, hence multi CS context 0
   is released.
4. thread_1 waits on QID 1 and uses multi CS context 0.
5. thread_0 waits on QID 0 and uses multi CS context 1.
6. The second iterattion of the loop (from step 3) starts, which
   means, start checking multi CS context 1:
   - multi CS contetxt is being used by thread_0 waiting on QID 0.
   - The fence of the CS (still CS from step 3) has QID map the same
     as the multi CS context 1.
   - multi CS context 1 (thread_0) gets completion on CS that triggered
     already thread_0 (with multi CS context 0) and is no longer
     being waited on.

Fixed by exiting the loop in complete_multi_cs after getting completion

Signed-off-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: move device boot warnings to the correct location

As device boot warnings clears the indication from the error mask,
they must be located together before the unknown error validation.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs/gaudi: return EPERM on non hard-reset

GAUDI supports only hard-reset. Therefore, this function should
return an error of operation not permitted.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: rename late init after reset function

The ASIC-specific soft_reset_late_init() is now called after either
soft-reset or reset-upon-device-release. Therefore, it needs a more
appropriate name.

No need to split it to two functions, as an ASIC either supports
soft-reset or reset-upon-device-release.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: fix soft reset accounting

Reset upon device release is not a soft-reset from user/system point
of view. As such, we shouldn't count that reset in the statistics we
gather and expose to the monitoring applications.

We also shouldn't print soft-reset when doing the reset upon device
release.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: Move frequency change thread to goya_late_init

Changing the frequency automatically is only done in Goya. In future
ASICs this is done inside the firmware. Therefore, move the common code
into the Goya specific files.

Main changes as part of the commit are:
    1. The thread for setting frequency is moved from device_late_init
       to goya_late_init
    2. hl_device_set_frequency is removed from hl_device_open as it is
       not relevant for other ASICs and for Goya it is taken care by
       the thread
    3. hl_device_set_frequency is renamed as goya_set_frequency

Signed-off-by: Rajaravi Krishna Katta <rkatta@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: abort reset on invalid request

Hard-reset is mutually exclusive with reset-on-device-release.
Therefore, if such a request arrives to the reset function, abort
the reset and return an error to the callee.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: fix possible deadlock in cache invl failure

Currently there is a deadlock in driver in scenarios where MMU
cache invalidation fails. The issue is basically device reset
being performed without releasing the MMU mutex.
The solution is to skip device reset as it is not necessary.
In addition we introduce a slight code refactor that prints the
invalidation error from a single location.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: skip PLL freq fetch

Getting the used PLL index with which to send the CPUPU packet relies on
the CPUCP info packet.

In case CPU queues are not enabled getting the PLL index will issue an
error and in some ASICs will also fail the driver load.

Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: align debugfs documentation to alphabetical order

Move an entry in the debugfs documentation to align with the
alphabetical order which is kept this file.

Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: prevent false heartbeat message

If a device reset has started, there is a chance that the heartbeat
function will fail because the device is disabled at the beginning
of the reset function.

In that case, we don't want the error message to appear in the log.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: add support for fetching historic errors

A new uAPI is added for debug purposes of the user-space to retrieve
errors related data from previous session (before device reset was
performed).

Inforamtion is filled when a razwi or CS timeout happens and can
contain one of the following:

1. Retrieve timestamp of last time the device was opened and razwi or
   CS timeout happened.
2. Retrieve information about last CS timeout.
3. Retrieve information about last razwi error.

This information doesn't contain user data, so no danger of data
leakage between users.

Signed-off-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: handle device TPM boot error as warning

AS TPM error indication is not fatal, driver should dump a warning
and continue booting.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

habanalabs: debugfs support for larger I2C transactions

I2C debugfs support is limited to 1 byte. We extend functionality
to more than 1 byte by using one of the pad fields as a length.
No backward compatibility issues as new F/W versions will treat 0
length as a 1 byte length transaction.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>