habanalabs: initialize variable to default value

Fix the following smatch error in unmap_device_va():
error: uninitialized symbol 'rc'.

Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Link: https://lore.kernel.org/r/20200601065648.8775-1-oded.gabbay@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Merge tag 'gnss-5.8-rc1' of https://git./linux/kernel/git/johan/gnss into char-misc-next

Johan writes:

GNSS updates for 5.8-rc1

Here are the GNSS updates for 5.8-rc1, including:

 - a fix for two broken probe errors paths in the sirf driver
 - a flexible array conversion

Both have been in linux-next with no reported issues.

* tag 'gnss-5.8-rc1' of https://git.kernel.org/pub/scm/linux/kernel/git/johan/gnss:
  gnss: replace zero-length array with flexible-array
  gnss: sirf: fix error return code in sirf_probe()

Merge tag 'extcon-next-for-5.8' of git://git./linux/kernel/git/chanwoo/extcon into char-misc-next

Chanwoo writes:

Update extcon for v5.8

Detailed description for this pull request:
1. Update extcon provider driver
- Fix an error handling code by using devm_iio_channel_get() for extcon-adc-jac.c
- Fix the usage counter imbalance of runtime PM for extcon-arizona.c.
- Add proper dt-compatible string for extcon-max14577.c.

2. Update extcon core with minor updates:
- Remove unneeded initialization of local variable in extcon_register_notifier

* tag 'extcon-next-for-5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/chanwoo/extcon:
  extcon: arizona: Fix runtime PM imbalance on error
  extcon: max14577: Add proper dt-compatible strings
  extcon: adc-jack: Fix an error handling path in 'adc_jack_probe()'
  extcon: remove redundant assignment to variable idx

extcon: arizona: Fix runtime PM imbalance on error

When arizona_request_irq() returns an error code, a
pairing runtime PM usage counter decrement is needed
to keep the counter balanced. For error paths after
this function, things are the same.

Also, remove calls to pm_runtime_disable() when
pm_runtime_enable() has not been executed.

Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>

extcon: max14577: Add proper dt-compatible strings

Add device tree compatible strings and create proper modalias structures
to let this driver load automatically if compiled as module, because
max14577 MFD driver creates MFD cells with such compatible strings.

Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>

extcon: adc-jack: Fix an error handling path in 'adc_jack_probe()'

In some error handling paths, a call to 'iio_channel_get()' is not balanced
by a corresponding call to 'iio_channel_release()'.

This can be achieved easily by using the devm_ variant of
'iio_channel_get()'.

This has the extra benefit to simplify the remove function.

Fixes: 19939860dcae ("extcon: adc_jack: adc-jack driver to support 3.5 pi or simliar devices")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>

extcon: remove redundant assignment to variable idx

The variable idx is being initialized with a value that is never read
and it is being updated later with a new value.  The initialization is
redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>

w1: omap-hdq: print dev_err if irq flags are not cleared

If irq flags are not cleared for certain operations we
print an error message.

Since this should never occur in normal operation, this
patch is an optional safety-net and debugging tool.

Signed-off-by: H. Nikolaus Schaller <hns@goldelico.com>
Link: https://lore.kernel.org/r/2de305d3046c7281a7123347899abbaa64c54fb8.1590255176.git.hns@goldelico.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

w1: omap-hdq: fix interrupt handling which did show spurious timeouts

Since

commit 27d13da8782a ("w1: omap-hdq: Simplify driver with PM runtime autosuspend")

was applied,

I did see timeouts and wrong values when reading a bq27000 connected
to hdq of the omap3. This occurred mainly after boot but remained and
only sometimes settled down after several reads.

root@letux:~# time cat /sys/class/power_supply/bq27000-battery/uevent
POWER_SUPPLY_NAME=bq27000-battery
POWER_SUPPLY_STATUS=Discharging
POWER_SUPPLY_PRESENT=1
POWER_SUPPLY_VOLTAGE_NOW=0
POWER_SUPPLY_CURRENT_NOW=0
POWER_SUPPLY_CAPACITY=0
POWER_SUPPLY_CAPACITY_LEVEL=Normal
POWER_SUPPLY_TEMP=-2731
POWER_SUPPLY_TIME_TO_EMPTY_NOW=0
POWER_SUPPLY_TIME_TO_EMPTY_AVG=0
POWER_SUPPLY_TIME_TO_FULL_NOW=0
POWER_SUPPLY_TECHNOLOGY=Li-ion
POWER_SUPPLY_CHARGE_FULL=0
POWER_SUPPLY_CHARGE_NOW=0
POWER_SUPPLY_CHARGE_FULL_DESIGN=0
POWER_SUPPLY_CYCLE_COUNT=0
POWER_SUPPLY_ENERGY_NOW=0
POWER_SUPPLY_POWER_AVG=0
POWER_SUPPLY_HEALTH=Good
POWER_SUPPLY_MANUFACTURER=Texas Instruments

real    0m15.761s
user    0m0.001s
sys     0m0.025s
root@letux:~#

Sometimes the effect did disappear after accessing
the device multiple times, speed went up and results
became correct.

All this indicates that some interrupts from the hdq
controller are lost by the driver.

Enabling debugging revealed that there were spurious tx
and rx timeouts, i.e. the driver does not always recognise
interrupts. The main problem is that rx and tx interrupts
share a single variable which was sometimes reset to
0 wiping out other interrupts. And it was overwritten
by a second interrupt, independent of whether the
previous interrupt was already processed or not.

This patch improves interrupt handling to avoid such
races and loss of interrupt flags.

The ideas are:
* only the hdq_isr() sets bits in hdq_status
* it does not reset any bits
* it does wake_up() if any interrupt is pending
* bits are only reset by the read/write/break functions
  if they were waited for
* this makes sure that no interrupts can be lost
* rx/tx/timeout bits are completely decoupled from each
  other (and not reset all after waiting for any of them)
* which bits to reset is now specified by a new parameter
  to hdq_reset_irqstatus()
* hdq_reset_irqstatus() also returns the state before
  resetting so that we can encapsulate the spinlock
* this should now handle the case that the write and read
  are both already finished quickly before the hdq_write_byte()
  ends.
* Or that two interrupts occur in succession before
  they are processed by the driver.
  Old code may have reset all status bits making the next
  hdq_read_byte() timeout.
* the spinlock now always protects changing of bits in function
  hdq_reset_irqstatus() which could become a read-write-modify
  problem if the interrupt handler tries to read-modify-write
  exactly at the same moment
* we add mutex protection also for hdq_write_byte() just to
  be safe to not to disturb a hdq_read_byte() triggered by
  some other thread/process.

This patch was tested on a GTA04 and results in no
boot problems any more. And first read after boot is now ok:

root@letux:~# time cat /sys/class/power_supply/bq27000-battery/uevent
POWER_SUPPLY_NAME=bq27000-battery
POWER_SUPPLY_STATUS=Discharging
POWER_SUPPLY_PRESENT=1
POWER_SUPPLY_VOLTAGE_NOW=3970000
POWER_SUPPLY_CURRENT_NOW=354144
POWER_SUPPLY_CAPACITY=82
POWER_SUPPLY_CAPACITY_LEVEL=Normal
POWER_SUPPLY_TEMP=266
POWER_SUPPLY_TIME_TO_EMPTY_NOW=7680
POWER_SUPPLY_TIME_TO_EMPTY_AVG=7380
POWER_SUPPLY_TECHNOLOGY=Li-ion
POWER_SUPPLY_CHARGE_FULL=934856
POWER_SUPPLY_CHARGE_NOW=763976
POWER_SUPPLY_CHARGE_FULL_DESIGN=1233792
POWER_SUPPLY_CYCLE_COUNT=82
POWER_SUPPLY_ENERGY_NOW=2852840
POWER_SUPPLY_POWER_AVG=1392840
POWER_SUPPLY_HEALTH=Good
POWER_SUPPLY_MANUFACTURER=Texas Instruments

real    0m0.233s
user    0m0.000s
sys     0m0.025s
root@letux:~#

It was also tested with dev_dbg enabled and more
printk that all activities behave correctly, especially
hdq_write_byte(), hdq_read_byte(), omap_hdq_break().

Not tested is omap_w1_triplet().

Fixes: 27d13da8782a ("w1: omap-hdq: Simplify driver with PM runtime autosuspend")
Cc: stable@vger.kernel.org # v5.6+
Signed-off-by: H. Nikolaus Schaller <hns@goldelico.com>
Link: https://lore.kernel.org/r/68fc8623ae741878beef049273696d2377526165.1590255176.git.hns@goldelico.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

w1: omap-hdq: fix return value to be -1 if there is a timeout

omap_w1_read_byte() should return -1 (or 0xff) in case of
error (e.g. missing battery).

The code accidentially overwrites the variable ret and not val,
which is returned. So it will return the initial value 0 instead
of -1.

Fixes: 27d13da8782a ("w1: omap-hdq: Simplify driver with PM runtime autosuspend")
Cc: stable@vger.kernel.org # v5.6+
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: H. Nikolaus Schaller <hns@goldelico.com>
Link: https://lore.kernel.org/r/b2c2192b461fbb9b8e9bea4ad514a49557a7210b.1590255176.git.hns@goldelico.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

w1: omap-hdq: cleanup to add missing newline for some dev_dbg

Otherwise it will corrupt the console log during debugging.

Fixes: 7b5362a603a1 ("w1: omap_hdq: Fix some error/debug handling.")
Cc: stable@vger.kernel.org
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: H. Nikolaus Schaller <hns@goldelico.com>
Link: https://lore.kernel.org/r/cd0d55749a091214106575f6e1d363c6db56622f.1590255176.git.hns@goldelico.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

/dev/mem: Revoke mappings when a driver claims the region

Close the hole of holding a mapping over kernel driver takeover event of
a given address range.

Commit 90a545e98126 ("restrict /dev/mem to idle io memory ranges")
introduced CONFIG_IO_STRICT_DEVMEM with the goal of protecting the
kernel against scenarios where a /dev/mem user tramples memory that a
kernel driver owns. However, this protection only prevents *new* read(),
write() and mmap() requests. Established mappings prior to the driver
calling request_mem_region() are left alone.

Especially with persistent memory, and the core kernel metadata that is
stored there, there are plentiful scenarios for a /dev/mem user to
violate the expectations of the driver and cause amplified damage.

Teach request_mem_region() to find and shoot down active /dev/mem
mappings that it believes it has successfully claimed for the exclusive
use of the driver. Effectively a driver call to request_mem_region()
becomes a hole-punch on the /dev/mem device.

The typical usage of unmap_mapping_range() is part of
truncate_pagecache() to punch a hole in a file, but in this case the
implementation is only doing the "first half" of a hole punch. Namely it
is just evacuating current established mappings of the "hole", and it
relies on the fact that /dev/mem establishes mappings in terms of
absolute physical address offsets. Once existing mmap users are
invalidated they can attempt to re-establish the mapping, or attempt to
continue issuing read(2) / write(2) to the invalidated extent, but they
will then be subject to the CONFIG_IO_STRICT_DEVMEM checking that can
block those subsequent accesses.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fixes: 90a545e98126 ("restrict /dev/mem to idle io memory ranges")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/159009507306.847224.8502634072429766747.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: xilinx-sdfec: convert get_user_pages() --> pin_user_pages()

This code was using get_user_pages*(), in approximately a "Case 1"
scenario (Direct IO), using the categorization from [1]. That means
that it's time to convert the get_user_pages*() + put_page() calls to
pin_user_pages*() + unpin_user_pages() calls.

There is some helpful background in [2]: basically, this is a small
part of fixing a long-standing disconnect between pinning pages, and
file systems' use of those pages.

[1] Documentation/core-api/pin_user_pages.rst

[2] "Explicit pinning of user-space pages":
    https://lwn.net/Articles/807108/

Cc: Derek Kiernan <derek.kiernan@xilinx.com>
Cc: Dragan Cvetic <dragan.cvetic@xilinx.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Michal Simek <michal.simek@xilinx.com>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Link: https://lore.kernel.org/r/20200527012628.1100649-4-jhubbard@nvidia.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: xilinx-sdfec: cleanup return value in xsdfec_table_write()

Return 0 for success, rather than the value of an incrementing
"reg" index. The reg value was never actually used, so this
simplifies the caller slightly.

Cc: Derek Kiernan <derek.kiernan@xilinx.com>
Cc: Dragan Cvetic <dragan.cvetic@xilinx.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Michal Simek <michal.simek@xilinx.com>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Link: https://lore.kernel.org/r/20200527012628.1100649-3-jhubbard@nvidia.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: xilinx-sdfec: improve get_user_pages_fast() error handling

This fixes the case of get_user_pages_fast() returning a -errno.
The result needs to be stored in a signed integer. And for safe
signed/unsigned comparisons, it's best to keep everything signed.
And get_user_pages_fast() also expects a signed value for number
of pages to pin.

Therefore, change most relevant variables, from u32 to int. Leave
"n" unsigned, for convenience in checking for overflow. And provide
a WARN_ON_ONCE() and early return, if overflow occurs.

Also, as long as we're tidying up: rename the page array from page,
to pages, in order to match the conventions used in most other call
sites.

Fixes: 20ec628e8007e ("misc: xilinx_sdfec: Add ability to configure LDPC")
Cc: Derek Kiernan <derek.kiernan@xilinx.com>
Cc: Dragan Cvetic <dragan.cvetic@xilinx.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Michal Simek <michal.simek@xilinx.com>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Link: https://lore.kernel.org/r/20200527012628.1100649-2-jhubbard@nvidia.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

nvmem: qfprom: remove incorrect write support

qfprom has different address spaces for read and write. Reads are
always done from corrected address space, where as writes are done
on raw address space.
Writing to corrected address space is invalid and ignored, so it
does not make sense to have this support in the driver which only
supports corrected address space regions at the moment.

Fixes: 4ab11996b489 ("nvmem: qfprom: Add Qualcomm QFPROM support.")
Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20200522113341.7728-1-srinivas.kandagatla@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Merge tag 'misc-habanalabs-next-2020-05-25' of git://people.freedesktop.org/~gabbayo/linux into char-misc-next

Oded writes:

This tag contains the following changes for kernel 5.8:

- Improve MMU cache invalidation code and handle case where the
  invalidation doesn't finish in a reasonable time.

- Remove the option to perform soft-reset to GAUDI. Soft-reset is where the
  driver only resets the compute and DMA engines of the ASIC. This is not
  relevant to GAUDI as we must also reset the NIC ports. And when we reset
  the NIC ports, we must also reset other stuff so we prefer to just do
  hard-reset (where we reset the entire ASIC except for PCIe).

- Fail the hard-reset procedure in case we still have user processes which
  have active file-descriptors on a device. Doing hard-reset in that case
  can result in a kernel panic because of gen_pool checks

- Don't initialize the default wait callback of dma_buf with the default
  wait function as that's the default...

* tag 'misc-habanalabs-next-2020-05-25' of git://people.freedesktop.org/~gabbayo/linux:
  habanalabs: handle MMU cache invalidation timeout
  habanalabs: don't allow hard reset with open processes
  habanalabs: GAUDI does not support soft-reset
  habanalabs: add print for soft reset due to event
  habanalabs: improve MMU cache invalidation code
  habanalabs: don't set default fence_ops->wait

habanalabs: handle MMU cache invalidation timeout

MMU cache invalidation timeout indicates that the device is unstable and
therefore unusable.
Hence in such case do hard reset and return an error to the user if was
called from ioctl.
In addition, change the print to error level and rephrase its text.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: don't allow hard reset with open processes

When the MMU is heavily used by the engines, unmapping might take a lot of
time due to a full MMU cache invalidation done as part of the unmap flow.
Hence we might not be able to kill all open processes before going to hard
reset the device, as it involves unmapping of all user memory.
In case of a failure in killing all open processes, we should stop the
hard reset flow as it might lead to a kernel crash - one thread (killing
of a process) is updating MMU structures that other thread (hard reset) is
freeing.
Stopping a hard reset flow leaves the device as nonoperational and the
user can then initiate a hard reset via sysfs to reinitialize the device.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: GAUDI does not support soft-reset

GAUDI does not support soft-reset as it leaves the NIC ports in an awkward
state, where their QMANs were reset but the NIC itself is still working.

In addition, there is not much sense in doing soft-reset when training is
done on multiple GAUDIs.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
Reviewed-by: Tomer Tayar <ttayar@habana.ai>

habanalabs: add print for soft reset due to event

Print the event name that caused the soft reset.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: improve MMU cache invalidation code

A new sequence is introduced to invalidate the MMU cache in order to avoid
timeouts.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: don't set default fence_ops->wait

It's the default.

Also so much for "we're not going to tell the graphics people how to
review their code", dma_fence is a pretty core piece of gpu driver
infrastructure. And it's very much uapi relevant, including piles of
corresponding userspace protocols and libraries for how to pass these
around.

Would be great if habanalabs would not use this (from a quick look
it's not needed at all), since open source the userspace and playing
by the usual rules isn't on the table. If that's not possible (because
it's actually using the uapi part of dma_fence to interact with gpu
drivers) then we have exactly what everyone promised we'd want to
avoid.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

misc: rtsx: Remove unnecessary rts5249_set_aspm(), rts5260_set_aspm()

rts5249_set_aspm() and rts5260_set_aspm() do nothing more than the default
rtsx_comm_set_aspm() does, so remove them and use the default.  No
functional change intended.

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20200521180545.1159896-7-helgaas@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: rtsx: Simplify rtsx_comm_set_aspm()

Simplify rtsx_comm_set_aspm() and remove the now-unused
rtsx_pci_enable_aspm().

rtsx_pci_disable_aspm() is still used by rtsx_pci_init_hw().

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20200521180545.1159896-6-helgaas@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: rtsx: Use pcie_capability_clear_and_set_word() for PCI_EXP_LNKCTL

Instead of using the driver-specific rtsx_pci_update_cfg_byte() to update
the PCIe Link Control Register, use pcie_capability_clear_and_set_word()
like the rest of the kernel does.  This makes it easier to maintain ASPM
across the PCI core and drivers.

Remove the now-unused rtsx_pci_update_cfg_byte() and ASPM_MASK_NEG
definitions.

No functional change intended.

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20200521180545.1159896-5-helgaas@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: rtsx: Use ASPM_MASK_NEG instead of hard-coded value

Use ASPM_MASK_NEG instead of hard-coded value, as other callers of
rtsx_pci_update_cfg_byte() do.  No functional change intended.

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20200521180545.1159896-4-helgaas@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: rtsx: Removed unused dev_aspm_mode

The struct rtsx_cr_option.dev_aspm_mode member is never set to anything
other than DEV_ASPM_DYNAMIC (0).  Remove it and code that tests it.  No
functional change intended.

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20200521180545.1159896-3-helgaas@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: rtsx: Remove unused pcr_ops

Remove the following unused function pointers from struct pcr_ops:

  int (*set_ltr_latency)(struct rtsx_pcr *pcr, u32 latency);
  int (*set_l1off_sub)(struct rtsx_pcr *pcr, u8 val);
  void (*full_on)(struct rtsx_pcr *pcr);
  void (*power_saving)(struct rtsx_pcr *pcr);

Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20200521180545.1159896-2-helgaas@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Handle syserr during power_up

The MHI device may be in the syserr state when we attempt to init it in
power_up().  Since we have no local state, the handling is simple -
reset the device and wait for it to transition out of the reset state.

Signed-off-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Hemant Kumar <hemantk@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-15-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Handle write lock properly in mhi_pm_m0_transition

Take write lock only to protect db_mode member of mhi channel.
This allows rest of the mhi channels to just take read lock which
fine grains the locking. It prevents channel readers to starve if
they try to enter critical section after a writer.

Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-14-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Do not process SYS_ERROR if RDDM is supported

Devices that support RDDM do not require processing SYS_ERROR as it is
deemed redundant. Avoid SYS_ERROR processing if RDDM is supported by
the device.

Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-13-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Skip handling BHI irq if MHI reg access is not allowed

Driver continues handling of BHI interrupt even if MHI register access
is not allowed. By doing so it calls the status call back and performs
early notification for the MHI client. This is not needed when MHI
register access is not allowed. Hence skip the handling in this case and
return. Also add debug log to print device state, local EE and device EE
when reg access is valid.

Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-12-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Handle disable transitions in state worker

Mission mode transition is handled by state worker thread but
power off is not. There is a possibility while mission mode
transition is in progress which calls MHI client driver probe,
power off is issued by MHI controller. This results into client
driver probe and remove running in parallel and causes use after
free situation. By queuing disable transition work when mission
mode is in progress prevents the race condition.

Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-11-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Remove the system error worker thread

Remove the system error worker thread and instead have the
execution environment worker handle that transition to serialize
processing and avoid any possible race conditions during
shutdown.

Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-10-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Ensure non-zero session or sequence ID values are used

While writing any sequence or session identifiers, it is possible that
the host could write a zero value, whereas only non-zero values should
be supported writes to those registers. Ensure that the host does not
write a non-zero value for them and also log them in debug messages. A
macro is introduced to simplify this check and the existing checks are
also converted to use this macro.

Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-9-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Improve debug logs for loading firmware

Add log messages to track boot flow errors and timeouts in SBL or AMSS
firmware loading to aid in debug.

Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-8-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Return appropriate error codes for AMSS load failure

When loading AMSS firmware using BHIe protocol, return -ETIMEDOUT if no
response is received within the timeout or return -EIO in case of a
protocol returned failure or an MHI error state.

Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-7-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Handle firmware load using state worker

Upon power up, driver queues firmware worker thread if the execution
environment is PBL. Firmware worker is blocked with a timeout until
state worker gets a chance to run and unblock firmware worker. An
endpoint power up failure can be seen if state worker gets a chance to
run after firmware worker has timed out. Remove this dependency and
handle firmware load directly using state worker thread.

Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-6-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Read transfer length from an event properly

When MHI Driver receives an EOT event, it reads xfer_len from the
event in the last TRE. The value is under control of the MHI device
and never validated by Host MHI driver. The value should never be
larger than the real size of the buffer but a malicious device can
set the value 0xFFFF as maximum. This causes driver to memory
overflow (both read or write). Fix this issue by reading minimum of
transfer length from event and the buffer length provided.

Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-5-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Add range check for channel id received in event ring

MHI data completion handler function reads channel id from event
ring element. Value is under the control of MHI devices and can be
any value between 0 and 255. In order to prevent out of bound access
add a bound check against the max channel supported by controller
and skip processing of that event ring element.

Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-4-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Cache intmod from mhi event to mhi channel

Driver is using zero initialized intmod value from mhi channel when
configuring TRE for bei field. This prevents interrupt moderation to
take effect in case it is supported by an event ring. Fix this by
copying intmod value from associated event ring to mhi channel upon
registering mhi controller.

Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-3-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bus: mhi: core: Refactor mhi queue APIs

Move all the common code to generate TRE from mhi_queue_buf,
mhi_queue_dma and mhi_queue_skb to mhi_gen_tre. This helps
to centralize the TRE generation code which makes any future
bug fixing easier to manage in these APIs.

Suggested-by: Jeffrey Hugo <jhugo@codeaurora.org>
Signed-off-by: Hemant Kumar <hemantk@codeaurora.org>
Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20200521170249.21795-2-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Merge tag 'soundwire-5.8-rc1' of git://git./linux/kernel/git/vkoul/soundwire into char-misc-next

Vinod writes:

soundwire updates for v5.8-rc1

This contains sdw_master_device patches and other updates done by Intel
folks.

Details:
  - sdw_master_device to represent the master instances.
  - sysfs properties for sdw_master_device and sdw_slave.
  - Documentation update for TDM modes.
  - some code cleanup patches and odd updates.

* tag 'soundwire-5.8-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/soundwire:
  soundwire: intel: use a single module
  soundwire: fix spelling mistake
  soundwire: fix trailing line in sysfs_slave.c
  soundwire: add Slave sysfs support
  soundwire: master: add sysfs support
  soundwire: disco: s/ch/channels/
  soundwire: master: add runtime pm support
  soundwire: bus_type: add sdw_master_device support
  soundwire: bus: add unique bus id
  soundwire: bus_type: introduce sdw_slave_type and sdw_master_type
  soundwire: bus: rename sdw_bus_master_add/delete, add arguments
  soundwire: intel: (cosmetic) remove multiple superfluous "else" statements
  soundwire: (cosmetic) remove multiple superfluous "else" statements
  soundwire: qcom: Use IRQF_ONESHOT
  soundwire: bus: reduce verbosity on enumeration
  soundwire: debugfs: clarify SDPX license with GPL-2.0-only
  soundwire: slave: don't init debugfs on device registration error
  Documentation: SoundWire: clarify TDM mode support
  soundwire: qcom: fix error handling in probe
  soundwire: intel: use asoc_rtd_to_cpu() / asoc_rtd_to_codec() macro for DAI pointer

Merge tag 'icc-5.8-rc1' of https://git.linaro.org/people/georgi.djakov/linux into char-misc-next

Georgi writes:

interconnect changes for 5.8

These are the interconnect changes for the 5.8-rc1 merge window:

Core changes:
- Convert the framework core from tristate to bool to make handling
dependencies between other core frameworks easier
- Add of_icc_get_by_index()
- Add devm_of_icc_get() helper function
- Add icc_enable() and icc_disable() helpers

New drivers:
- Platform driver for NXP i.MX8MM SoC
- Platform driver for NXP i.MX8MN SoC
- Platform driver for NXP i.MX8MQ SoC

Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
* tag 'icc-5.8-rc1' of https://git.linaro.org/people/georgi.djakov/linux:
  interconnect: Remove unused module exit code from core
  interconnect: Disallow interconnect core to be built as a module
  interconnect: Add of_icc_get_by_index() helper function
  interconnect: Add helpers for enabling/disabling a path
  interconnect: imx: Fix return value check in imx_icc_node_init_qos()
  interconnect: imx: Add platform driver for imx8mn
  interconnect: imx: Add platform driver for imx8mq
  interconnect: imx: Add platform driver for imx8mm
  interconnect: Add imx core driver
  dt-bindings: interconnect: Add bindings for imx8m noc
  interconnect: Add devm_of_icc_get() as exported API for users

w1_therm: remove redundant assignments to variable ret

The variable ret is being initialized with a value that is never read
and it is being updated later with a new value. The initialization
is redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20200519154553.873413-1-colin.king@canonical.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

w1_therm: Free the correct variable

The problem is that we change "p_args" to point to the middle of the
string so when we free it at the end of the function it's not freeing
the same pointer that we originally allocated.

Fixes: e2c94d6f5720 ("w1_therm: adding alarm sysfs entry")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/20200520120019.GA172354@mwanda
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

soundwire: intel: use a single module

It's not clear why we have two modules for the Intel controller/master
support when there is a single Kconfig. This adds complexity for no
good reason, the two parts need to work together anyways.

Signed-off-by: Rander Wang <rander.wang@intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20200519191903.6557-1-yung-chuan.liao@linux.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

soundwire: fix spelling mistake

Fix typo for paranoia spelled as paranioa

Fixes: bcac59029955 ("soundwire: add Slave sysfs support")
Signed-off-by: Vinod Koul <vkoul@kernel.org>

soundwire: fix trailing line in sysfs_slave.c

Commit bcac59029955 ("soundwire: add Slave sysfs support") added
trailing line in file sysfs_slave.c, so remove it

Fixes: bcac59029955 ("soundwire: add Slave sysfs support")
Signed-off-by: Vinod Koul <vkoul@kernel.org>

soundwire: add Slave sysfs support

Expose MIPI DisCo Slave properties in sysfs.

For Slave properties and Data Port 0, the attributes are managed with
simple devm_ support.

A Slave Device may have more than one Data Port (DPN), and each Data
Port can be sink or source. The attributes are created dynamically
using pre-canned macros, but still use devm_ with a name attribute
group to avoid creating kobjects - as requested by GregKH. In the
_show function, we use container_of() to retrieve port number and
direction required to extract the information.

Audio modes are not supported for now. Depending on the discussions
the SoundWire Device Class, we may add it later as is or follow the
new specification.

Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://lore.kernel.org/r/20200518203551.2053-4-yung-chuan.liao@linux.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

soundwire: master: add sysfs support

Add the master properties as attributes. The description is directly
derived from the MIPI DisCo specification.

Credits: this patch is based on an earlier internal contribution by
Vinod Koul, Sanyog Kale, Shreyas Nc and Hardik Shah.

Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20200518203551.2053-3-yung-chuan.liao@linux.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

soundwire: disco: s/ch/channels/

Use more meaningful member names in preparation for sysfs support.
No functionality change.

Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Link: https://lore.kernel.org/r/20200518203551.2053-2-yung-chuan.liao@linux.intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>

fpga: dfl: afu: Corrected error handling levels

Corrected error handling goto sequnece. Level put_pages should
be called when pinned pages >= 0 && pinned != npages. Level
free_pages should be called when pinned pages < 0.

Fixes: fa8dda1edef9 ("fpga: dfl: afu: add DFL_FPGA_PORT_DMA_MAP/UNMAP ioctls support")
Signed-off-by: Souptick Joarder <jrdr.linux@gmail.com>
Acked-by: Wu Hao <hao.wu@intel.com>
Reviewed-by: Xu Yilun <yilun.xu@intel.com>
Link: https://lore.kernel.org/r/1589825991-3545-1-git-send-email-jrdr.linux@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

misc: pci_endpoint_test: Add Device ID for RZ/G2E PCIe controller

Add Renesas R8A774C0 in pci_device_id table so that pci-epf-test can be
used for testing PCIe EP on RZ/G2E.

Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Reviewed-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Link: https://lore.kernel.org/r/1589493809-2602-1-git-send-email-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

genwqe: convert get_user_pages() --> pin_user_pages()

This code was using get_user_pages*(), in a "Case 2" scenario
(DMA/RDMA), using the categorization from [1]. That means that it's
time to convert the get_user_pages*() + put_page() calls to
pin_user_pages*() + unpin_user_pages() calls.

There is some helpful background in [2]: basically, this is a small
part of fixing a long-standing disconnect between pinning pages, and
file systems' use of those pages.

[1] Documentation/core-api/pin_user_pages.rst

[2] "Explicit pinning of user-space pages":
    https://lwn.net/Articles/807108/

Cc: Frank Haverkamp <haver@linux.ibm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Link: https://lore.kernel.org/r/20200518015237.1568940-1-jhubbard@nvidia.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/mic/scif: convert get_user_pages() --> pin_user_pages()

This code was using get_user_pages*(), in a "Case 2" scenario
(DMA/RDMA), using the categorization from [1]. That means that it's
time to convert the get_user_pages*() + put_page() calls to
pin_user_pages*() + unpin_user_pages() calls.

There is some helpful background in [2]: basically, this is a small
part of fixing a long-standing disconnect between pinning pages, and
file systems' use of those pages.

Note that this effectively changes the code's behavior as well: it now
ultimately calls set_page_dirty_lock(), instead of SetPageDirty(). This
is probably more accurate.

As Christoph Hellwig put it, "set_page_dirty() is only safe if we are
dealing with a file backed page where we have reference on the inode it
hangs off." [3]

[1] Documentation/core-api/pin_user_pages.rst

[2] "Explicit pinning of user-space pages":
    https://lwn.net/Articles/807108/

[3] https://lore.kernel.org/r/20190723153640.GB720@lst.de

Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Link: https://lore.kernel.org/r/20200518041307.1987328-1-jhubbard@nvidia.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: cti: Add CPU idle pm notifer to CTI devices

Adds a notify callback for CPU PM events to the CTI driver - for
CPU bound CTI devices.

Signed-off-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-24-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: cti: Add CPU Hotplug handling to CTI driver

Adds registration of CPU start and stop functions to CPU hotplug
mechanisms - for any CPU bound CTI.

Sets CTI powered flag according to state.
Will enable CTI on CPU start if there are existing enable requests.

Signed-off-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-23-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: etm4x: Fix use-after-free of per-cpu etm drvdata

etm probe could be deferred due to the dependency in the trace
path chain and may be retried. We need to clear the per-cpu
etmdrvdata entry for the etm in case of a failure to avoid
use-after-free cases as reported below:

KASAN use-after-free bug in etm4_cpu_pm_notify():

[    8.574566] coresight etm0: CPU0: ETM v4.2 initialized
[    8.581920] BUG: KASAN: use-after-free in etm4_cpu_pm_notify+0x580/0x2024
[    8.581925] Read of size 8 at addr ffffff813304f8c8 by task swapper/3/0
[    8.581927]
[    8.581934] CPU: 3 PID: 0 Comm: swapper/3 Tainted: G S      W         5.4.28 #314
[    8.587775] coresight etm1: CPU1: ETM v4.2 initialized
[    8.594195] Call trace:
[    8.594205]  dump_backtrace+0x0/0x188
[    8.594209]  show_stack+0x20/0x2c
[    8.594216]  dump_stack+0xdc/0x144
[    8.594227]  print_address_description+0x3c/0x494
[    8.594232]  __kasan_report+0x144/0x168
[    8.601598] coresight etm2: CPU2: ETM v4.2 initialized
[    8.602563]  kasan_report+0x10/0x18
[    8.602568]  check_memory_region+0x1a4/0x1b4
[    8.602572]  __kasan_check_read+0x18/0x24
[    8.602577]  etm4_cpu_pm_notify+0x580/0x2024
[    8.665945]  notifier_call_chain+0x5c/0x90
[    8.670166]  __atomic_notifier_call_chain+0x90/0xf8
[    8.675182]  cpu_pm_notify+0x40/0x6c
[    8.678858]  cpu_pm_enter+0x38/0x80
[    8.682451]  psci_enter_idle_state+0x34/0x70
[    8.686844]  cpuidle_enter_state+0xb8/0x20c
[    8.691143]  cpuidle_enter+0x38/0x4c
[    8.694820]  call_cpuidle+0x3c/0x68
[    8.698408]  do_idle+0x1a0/0x280
[    8.701729]  cpu_startup_entry+0x24/0x28
[    8.705768]  secondary_start_kernel+0x15c/0x170
[    8.710423]
[    8.711972] Allocated by task 242:
[    8.715473]  __kasan_kmalloc+0xf0/0x1ac
[    8.719426]  kasan_slab_alloc+0x14/0x1c
[    8.723375]  __kmalloc_track_caller+0x23c/0x388
[    8.728040]  devm_kmalloc+0x38/0x94
[    8.731632]  etm4_probe+0x48/0x3c8
[    8.735140]  amba_probe+0xbc/0x158
[    8.738645]  really_probe+0x144/0x408
[    8.742412]  driver_probe_device+0x70/0x140
[    8.746716]  __device_attach_driver+0x9c/0x110
[    8.751287]  bus_for_each_drv+0x90/0xd8
[    8.755236]  __device_attach+0xb4/0x164
[    8.759188]  device_initial_probe+0x20/0x2c
[    8.763490]  bus_probe_device+0x34/0x94
[    8.767436]  device_add+0x34c/0x3e0
[    8.771029]  amba_device_try_add+0x68/0x440
[    8.775332]  amba_deferred_retry_func+0x48/0xc8
[    8.779997]  process_one_work+0x344/0x648
[    8.784127]  worker_thread+0x2ac/0x47c
[    8.787987]  kthread+0x128/0x138
[    8.791313]  ret_from_fork+0x10/0x18
[    8.794993]
[    8.796532] Freed by task 242:
[    8.799684]  __kasan_slab_free+0x15c/0x22c
[    8.803897]  kasan_slab_free+0x10/0x1c
[    8.807761]  kfree+0x25c/0x4bc
[    8.810913]  release_nodes+0x240/0x2b0
[    8.814767]  devres_release_all+0x3c/0x54
[    8.818887]  really_probe+0x178/0x408
[    8.822661]  driver_probe_device+0x70/0x140
[    8.826963]  __device_attach_driver+0x9c/0x110
[    8.831539]  bus_for_each_drv+0x90/0xd8
[    8.835487]  __device_attach+0xb4/0x164
[    8.839431]  device_initial_probe+0x20/0x2c
[    8.843732]  bus_probe_device+0x34/0x94
[    8.847678]  device_add+0x34c/0x3e0
[    8.851274]  amba_device_try_add+0x68/0x440
[    8.855576]  amba_deferred_retry_func+0x48/0xc8
[    8.860240]  process_one_work+0x344/0x648
[    8.864366]  worker_thread+0x2ac/0x47c
[    8.868228]  kthread+0x128/0x138
[    8.871557]  ret_from_fork+0x10/0x18
[    8.875231]
[    8.876782] The buggy address belongs to the object at ffffff813304f800
[    8.876782]  which belongs to the cache kmalloc-1k of size 1024
[    8.889632] The buggy address is located 200 bytes inside of
[    8.889632]  1024-byte region [ffffff813304f800, ffffff813304fc00)
[    8.901761] The buggy address belongs to the page:
[    8.906695] page:ffffffff04ac1200 refcount:1 mapcount:0 mapping:ffffff8146c03800 index:0x0 compound_mapcount: 0
[    8.917047] flags: 0x4000000000010200(slab|head)
[    8.921799] raw: 4000000000010200 dead000000000100 dead000000000122 ffffff8146c03800
[    8.929753] raw: 0000000000000000 0000000000100010 00000001ffffffff 0000000000000000
[    8.937703] page dumped because: kasan: bad access detected
[    8.943433]
[    8.944974] Memory state around the buggy address:
[    8.949903]  ffffff813304f780: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[    8.957320]  ffffff813304f800: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[    8.964742] >ffffff813304f880: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[    8.972157]                                               ^
[    8.977886]  ffffff813304f900: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[    8.985298]  ffffff813304f980: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[    8.992713] ==================================================================

Fixes: f188b5e76aae ("coresight: etm4x: Save/restore state across CPU low power states")
Reported-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Tested-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-22-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Avoid casting void pointers

We don't need to cast void pointers, such as the amba_id data. Assign to
a local variable to make the code prettier and also return NULL instead
of 0 to make sparse happy.

Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Reviewed-by: Joe Perches <joe@perches.com>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Reviewed-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-21-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Include required headers in C files

We should include headers that C files use in the C files that use them
and avoid relying on implicit includes as much as possible. This helps
avoid compiler errors in the future about missing declarations when
header files change includes in the future.

Cc: Douglas Anderson <dianders@chromium.org>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-20-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Initialize arg in sparse friendly way

Sparse gets annoyed when this initializer is 0 but the first struct
member is a pointer. Just use { } to initialize instead so that sparse
is quiet.

Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-19-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Don't initialize variables unnecessarily

These variables are assigned again before they're used. Leave them
unassigned at first so that the compiler can detect problems in the
future with use before initialization.

Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-18-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Mark some functions static

These functions aren't used outside the file they're in. Mark them
static to indicate as such and silence tools like sparse.

Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
[Dropped changes in coresight-cti.c and coresight-etb10.c]
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-17-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: etm4x: Add support for Neoverse N1 ETM

Add PID for Arm Neoverse N1 ETM to the list of supported/known ETMs.

Signed-off-by: Anurag Koul <anurag.koul@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-16-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: etmv4: Update default filter and initialisation

Differing default states set on driver init / perf init and as a result
of a sysfs reset.

The ETMv4 can be programmed to trace the entire instruction address range
without the need to use address comparator filter resources.
(Described in the ETMv4.x technical reference manual)

sysfs reset was using this method, perf and default driver init were setup
with an address range comparator for the entire address range.

The perf / driver init has been altered to use the method without needing
any comparator address hardware.

Minor adjustment to the vinst_ctrl register initialisation to ensure
correct zero initialisation.

Signed-off-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-15-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: tmc: Fix TMC mode read in tmc_read_prepare_etb()

On some QCOM platforms like SC7180, SDM845 and SM8150,
reading TMC mode register without proper coresight power
management can lead to async exceptions like the one in
the call trace below in tmc_read_prepare_etb(). This can
happen if the user tries to read the TMC etf data via
device node without setting up source and the sink first.
Fix this by having a check for coresight sysfs mode
before reading TMC mode management register.

 Kernel panic - not syncing: Asynchronous SError Interrupt
 CPU: 7 PID: 2605 Comm: hexdump Tainted: G S                5.4.30 #122
 Call trace:
  dump_backtrace+0x0/0x188
  show_stack+0x20/0x2c
  dump_stack+0xdc/0x144
  panic+0x168/0x36c
  panic+0x0/0x36c
  arm64_serror_panic+0x78/0x84
  do_serror+0x130/0x138
  el1_error+0x84/0xf8
  tmc_read_prepare_etb+0x88/0xb8
  tmc_open+0x40/0xd8
  misc_open+0x120/0x158
  chrdev_open+0xb8/0x1a4
  do_dentry_open+0x268/0x3a0
  vfs_open+0x34/0x40
  path_openat+0x39c/0xdf4
  do_filp_open+0x90/0x10c
  do_sys_open+0x150/0x3e8
  __arm64_compat_sys_openat+0x28/0x34
  el0_svc_common+0xa8/0x160
  el0_svc_compat_handler+0x2c/0x38
  el0_svc_compat+0x8/0x10

Fixes: 4525412a5046 ("coresight: tmc: making prepare/unprepare functions generic")
Reported-by: Stephen Boyd <swboyd@chromium.org>
Suggested-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-14-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Fix support for sparsely populated ports

On some systems the firmware may not describe all the ports
connected to a component (e.g, for security reasons). This
could be especially problematic for "funnels" where we could
end up in modifying memory beyond the allocated space for
refcounts.

e.g, for a funnel with input ports listed 0, 3, 5, nr_inport = 3.
However the we could access refcnts[5] while checking for
references, like :

 [  526.110401] ==================================================================
 [  526.117988] BUG: KASAN: slab-out-of-bounds in funnel_enable+0x54/0x1b0
 [  526.124706] Read of size 4 at addr ffffff8135f9549c by task bash/1114
 [  526.131324]
 [  526.132886] CPU: 3 PID: 1114 Comm: bash Tainted: G S                5.4.25 #232
 [  526.140397] Hardware name: Qualcomm Technologies, Inc. SC7180 IDP (DT)
 [  526.147113] Call trace:
 [  526.149653]  dump_backtrace+0x0/0x188
 [  526.153431]  show_stack+0x20/0x2c
 [  526.156852]  dump_stack+0xdc/0x144
 [  526.160370]  print_address_description+0x3c/0x494
 [  526.165211]  __kasan_report+0x144/0x168
 [  526.169170]  kasan_report+0x10/0x18
 [  526.172769]  check_memory_region+0x1a4/0x1b4
 [  526.177164]  __kasan_check_read+0x18/0x24
 [  526.181292]  funnel_enable+0x54/0x1b0
 [  526.185072]  coresight_enable_path+0x104/0x198
 [  526.189649]  coresight_enable+0x118/0x26c

  ...

 [  526.237782] Allocated by task 280:
 [  526.241298]  __kasan_kmalloc+0xf0/0x1ac
 [  526.245249]  kasan_kmalloc+0xc/0x14
 [  526.248849]  __kmalloc+0x28c/0x3b4
 [  526.252361]  coresight_register+0x88/0x250
 [  526.256587]  funnel_probe+0x15c/0x228
 [  526.260365]  dynamic_funnel_probe+0x20/0x2c
 [  526.264679]  amba_probe+0xbc/0x158
 [  526.268193]  really_probe+0x144/0x408
 [  526.271970]  driver_probe_device+0x70/0x140

 ...

 [  526.316810]
 [  526.318364] Freed by task 0:
 [  526.321344] (stack is not available)
 [  526.325024]
 [  526.326580] The buggy address belongs to the object at ffffff8135f95480
 [  526.326580]  which belongs to the cache kmalloc-128 of size 128
 [  526.339439] The buggy address is located 28 bytes inside of
 [  526.339439]  128-byte region [ffffff8135f95480, ffffff8135f95500)
 [  526.351399] The buggy address belongs to the page:
 [  526.356342] page:ffffffff04b7e500 refcount:1 mapcount:0 mapping:ffffff814b00c380 index:0x0 compound_mapcount: 0
 [  526.366711] flags: 0x4000000000010200(slab|head)
 [  526.371475] raw: 4000000000010200 ffffffff05034008 ffffffff0501eb08 ffffff814b00c380
 [  526.379435] raw: 0000000000000000 0000000000190019 00000001ffffffff 0000000000000000
 [  526.387393] page dumped because: kasan: bad access detected
 [  526.393128]
 [  526.394681] Memory state around the buggy address:
 [  526.399619]  ffffff8135f95380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 [  526.407046]  ffffff8135f95400: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 [  526.414473] >ffffff8135f95480: 04 fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 [  526.421900]                             ^
 [  526.426029]  ffffff8135f95500: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 [  526.433456]  ffffff8135f95580: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 [  526.440883] ==================================================================

To keep the code simple, we now track the maximum number of
possible input/output connections to/from this component
@ nr_inport and nr_outport in platform_data, respectively.
Thus the output connections could be sparse and code is
adjusted to skip the unspecified connections.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Reported-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Tested-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Tested-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-13-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: etb10: Make coresight_etb_groups static

Fix the following sparse warning:

drivers/hwtracing/coresight/coresight-etb10.c:720:30: warning: symbol
'coresight_etb_groups' was not declared. Should it be static?

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-12-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: cti: Make some symbols static

Fix the following sparse warning:

drivers/hwtracing/coresight/coresight-cti.c:22:1: warning: symbol
'ect_net' was not declared. Should it be static?
drivers/hwtracing/coresight/coresight-cti.c:625:32: warning: symbol
'cti_ops_ect' was not declared. Should it be static?
drivers/hwtracing/coresight/coresight-cti.c:630:28: warning: symbol
'cti_ops' was not declared. Should it be static?

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-11-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: etm4x: Replace ETM PIDs with UCI IDs for Kryo385

Replace the AMBA ETM PIDs with UCI IDs to avoid future
conflicts when adding the CTI support for QCOM Kryo385
CPU cores.

Fixes: 17b4add0d4e0 ("coresight: etm4x: Add ETM PIDs for SDM845 and MSM8996")
Signed-off-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-10-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: etm4x: Add support for Qualcomm SC7180 SoC

Add ETM UCI IDs for Qualcomm SC7180 SoC. It has 2
big CPU cores based on Cortex-A76 and 6 LITTLE CPU
cores based on Cortex-A55.

Signed-off-by: Sai Prakash Ranjan <saiprakash.ranjan@codeaurora.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Tested-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-9-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

docs: trace: coresight-ect.rst: Fix a build warning

Sphinx wants a line after "..", as otherwise it complains with:

Documentation/trace/coresight/coresight-ect.rst:2: WARNING: Explicit markup ends without a blank line; unexpected unindent.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-8-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: docs: Add information about the topology representations

Update the CoreSight documents to describe the new connections directory
and the links between CoreSight devices in this directory.

Signed-off-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-7-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: cti: Add in sysfs links to other coresight devices

Adds in sysfs links for connections where the connected device is another
coresight device. This allows examination of the coresight topology.

Non-coresight connections remain just as a reference name.

Signed-off-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-6-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Expose device connections via sysfs

Coresight device connections are a bit complicated and is not
exposed currently to the user. One has to look at the platform
descriptions (DT bindings or ACPI bindings) to make an understanding.
Given the new naming scheme, it will be helpful to have this information
to choose the appropriate devices for tracing. This patch exposes
the device connections via links in the sysfs directories.

e.g, for a connection devA[OutputPort_X] -> devB[InputPort_Y]
is represented as two symlinks:

  /sys/bus/coresight/.../devA/out:X -> /sys/bus/coresight/.../devB
  /sys/bus/coresight/.../devB/in:Y  -> /sys/bus/coresight/.../devA

Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
[Revised to use the generic sysfs links functions & link structures.
Provides a connections sysfs group in each device to hold the links.]
Co-developed-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-5-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Add generic sysfs link creation functions

To allow the connections between coresight components to be represented
in sysfs, generic methods for creating sysfs links between two coresight
devices are added.

Signed-off-by: Mike Leach <mike.leach@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-4-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Add return value for fixup connections

Handle failures in fixing up connections for a newly registered
device. This will be useful to handle cases where we fail to expose
the links via sysfs for the connections.

Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-3-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

coresight: Pass coresight_device for coresight_release_platform_data

As we prepare to expose the links between the devices in
sysfs, pass the coresight_device instance to the
coresight_release_platform_data in order to free up the connections
when the device is removed.

No functional changes as such in this patch.

Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Mike Leach <mike.leach@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Link: https://lore.kernel.org/r/20200518180242.7916-2-mathieu.poirier@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Merge tag 'misc-habanalabs-next-2020-05-19' of git://people.freedesktop.org/~gabbayo/linux into char-misc-next

Oded writes:

This tag contains the following changes for kernel 5.8:

- GAUDI ASIC support. The tag contains code and header files needed to
  initialize the GAUDI ASIC and run workloads on it. There are changes to
  the common code that are needed for GAUDI and there is the addition of
  new ASIC-dependent code of GAUDI.

- Add new feature of signal/wait command submissions. This is relevant to
  GAUDI only and allows the user to sync between streams (queues) inside
  the device.

- Allow user to retrieve the device time alongside the host time, to allow
  a user application to synchronize device time together with host
  time during profiling.

- Change ASIC's CPU initialization by loading its boot loader code from the
  Host memory (instead of it being programmed on the on-board FLASH).

- Expose more attributes through HWMON.

- Move the ASIC event handling code to be "common code". This will be
  shared between GAUDI and future ASICs. Goya will still use its own code.

- Fix bug in command submission parsing in Goya.

- Small fixes to security configuration (open up some registers for user
  access).

- Improvements to ASIC reset code.

* tag 'misc-habanalabs-next-2020-05-19' of git://people.freedesktop.org/~gabbayo/linux: (38 commits)
  habanalabs: update patched_cb_size for Wreg32
  habanalabs: move event handling to common firmware file
  habanalabs: enable gaudi code in driver
  habanalabs: add gaudi profiler module
  habanalabs: add gaudi security module
  habanalabs: add hwmgr module for gaudi
  habanalabs: add gaudi asic-dependent code
  uapi: habanalabs: add gaudi defines
  habanalabs: add gaudi asic registers header files
  habanalabs: get card type, location from F/W
  habanalabs: support clock gating enable/disable
  habanalabs: set PM profile to auto only for goya
  habanalabs: add dedicated define for hard reset
  habanalabs: check if CoreSight is supported
  habanalabs: add signal/wait to CS IOCTL operations
  habanalabs: handle the h/w sync object
  habanalabs: define ASIC-dependent interface for signal/wait
  uapi: habanalabs: add signal/wait operations
  habanalabs: add missing MODULE_DEVICE_TABLE
  habanalabs: print all CB handles as hex numbers
  ...

habanalabs: update patched_cb_size for Wreg32

The patch_cb_size is not updated for Wreg32 in its validate function, so
updated in goya_validate_cb.

Signed-off-by: Rachel Stahl <rstahl@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: move event handling to common firmware file

Instead of writing similar event handling code for each ASIC, move the code
to the common firmware file. This code will be used for GAUDI and all
future ASICs.

In addition, add two new fields to the auto-generated events file: valid
and description. This will save the need to manually write the events
description in the source code and simplify the code.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: enable gaudi code in driver

Enable the GAUDI ASIC code in the pci probe callback of the driver so the
driver will handle GAUDI ASICs.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: add gaudi profiler module

Add the GAUDI code to initialize the ASIC's profiler. The profile receives
its initialization values from the user, same as in Goya, but the code to
initialize is in the driver because the configuration space of the
device is not directly exposed to the user.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: add gaudi security module

Add the code to initialize the security module of GAUDI. Similar to Goya,
we have two dedicated mechanisms for security: Range Registers and
Protection bits. Those mechanisms protect sensitive memory and
configuration areas inside the device.

In addition, in Gaudi we moved to a 3-level security scheme, where the F/W
runs with the highest security level (Privileged), the driver runs with a
less secured level (Secured) and the user is neither privileged nor
secured. The security module in the driver configures the Secured parts so
the user won't be able to access them. The Privileged parts are configured
by the F/W.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: add hwmgr module for gaudi

The hwmgr module is responsible for messages sent to GAUDI F/W that are
not common to all habanalabs ASICs.

In GAUDI, we provide the user a simplified mode of controlling the ASIC
clock frequency. Instead of three different clocks, we present a single
clock property that the user can configure via sysfs.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: add gaudi asic-dependent code

Add the ASIC-dependent code for GAUDI. Supply (almost) all of the function
callbacks that the driver's common code need to initialize, finalize and
submit workloads to the GAUDI ASIC.

It also contains the code to initialize the F/W of the GAUDI ASIC and to
receive events from the F/W.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

uapi: habanalabs: add gaudi defines

Add the new defines for GAUDI uapi interface. It includes the queue IDs,
the engine IDs, SRAM reserved space and Sync Manager reserved resources.

There is no new IOCTL or additional operations in existing IOCTLs.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: add gaudi asic registers header files

Add the relevant GAUDI ASIC registers header files. These files are
generated automatically from a tool maintained by the VLSI engineers.

There are more files which are not upstreamed because only very few defines
from those files are used in the driver. For those files, we copied the
relevant defines into gaudi_regs.h and gaudi_masks.h, to reduce the size of
this patch.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: get card type, location from F/W

For Gaudi the driver gets two new additional properties from the F/W:
1. The card's type - PCI or PMC
2. The card's location in the Gaudi's box (relevant only for PMC).

The card's location is also passed to the user in the HW IP info structure
as it needs this property for establishing communication between Gaudis.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: support clock gating enable/disable

In Gaudi there is a feature of clock gating certain engines.
Therefore, add this property to the device structure.

In addition, due to a limitation of this feature, the driver needs to
dynamically enable or disable this feature during run-time. Therefore, add
ASIC interface functions to enable/disable this function from the common
code.

Moreover, this feature must be turned off when the user wishes to debug the
ASIC by reading/writing registers and/or memory through the driver's
debugfs. Therefore, add an option to enable/disable clock gating via the
debugfs interface.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: set PM profile to auto only for goya

For Gaudi, the driver doesn't change the PM profile automatically due to
device-controlled PM capabilities. Therefore, set the PM profile to auto
only for Goya so the driver's code to automatically change the profile
won't run on Gaudi.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: add dedicated define for hard reset

Gaudi requires longer waiting during reset due to closing of network ports.
Add this explanation to the relevant comment in the code and add a
dedicated define for this reset timeout period, instead of multiplying
another define.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: check if CoreSight is supported

Coresight is not supported on simulator, therefore add a boolean for
checking that (currently used by un-upstreamed code).

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: add signal/wait to CS IOCTL operations

Add the following two operations to the CS IOCTL:

Signal:

The signal operation is basically a command submission, that is created by
the driver upon user request. It will be implemented using a dedicated PQE
that will increment a specific SOB. There will be a new flag:
HL_CS_FLAGS_SIGNAL. When the user set this flag in the CS IOCTL structure,
the driver will execute a dedicated code path that will prepare this
special PQE and submit it. The user only needs to provide a queue index on
which to put the signal.

Wait:

The wait operation is also a command submission that is created by the
driver upon user request. It will be implemented using a dedicated PQE that
will contain packets of "ARM a monitor" + FENCE packet. There will be a new
flag: HL_CS_FLAGS_WAIT. When the user set this flag in the CS structure,
the driver will execute a dedicated code path that will prepare this
special PQE and submit it.

The user needs to provide the following parameters:
1. queue ID
2. an array of signal_seq numbers and the number of signals to wait on
   (the length of signal_seq_arr).

The IOCTL will return the CS sequence number of the wait it put on the
queue ID.

Currently, the code supports signal_seq_nr==1. But this API definition will
allow us to put a single PQE that waits on multiple signals.

To correctly configure the monitor and fence, the driver will need to
retrieve the specified signal CS object that contains the relevant SOB and
its expected value. In case the signal CS has already been completed, there
is no point of adding a wait operation. In this case, the driver will
return to the user *without* putting anything on the PQ. The return code
should reflect to the user that the signal was completed, as we won't
return a CS sequence number for this wait.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: handle the h/w sync object

Define a structure representing the h/w sync object (SOB).

a SOB can contain up to 2^15 values. Each signal CS will increment the SOB
by 1, so after some time we will reach the maximum number the SOB can
represent. When that happens, the driver needs to move to a different SOB
for the signal operation.

A SOB can be in 1 of 4 states:

1. Working state with value < 2^15

2. We reached a value of 2^15, but the signal operations weren't completed
yet OR there are pending waits on this signal. For the next submission, the
driver will move to another SOB.

3. ALL the signal operations on the SOB have finished AND there are no more
pending waits on the SOB AND we reached a value of 2^15 (This basically
means the refcnt of the SOB is 0 - see explanation below). When that
happens, the driver can clear the SOB by simply doing WREG32 0 to it and
set the refcnt back to 1.

4. The SOB is cleared and can be used next time by the driver when it needs
to reuse an SOB.

Per SOB, the driver will maintain a single refcnt, that will be initialized
to 1. When a signal or wait operation on this SOB is submitted to the PQ,
the refcnt will be incremented. When a signal or wait operation on this SOB
completes, the refcnt will be decremented. After the submission of the
signal operation that increments the SOB to a value of 2^15, the refcnt is
also decremented.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: define ASIC-dependent interface for signal/wait

This feature requires handling h/w resources which are a bit different from
one ASIC to the other. Therefore, we need to define a set of interfaces the
ASIC code provides to the common code to signal, wait, reset sync object
and to reset and init a queue.

As this feature is not supported in Goya, provide an empty implementation
of those functions.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

uapi: habanalabs: add signal/wait operations

This is a pre-requisite to upstreaming GAUDI support.

Signal/wait operations are done by the user to perform sync between two
Primary Queues (PQs). The sync is done using the sync manager and it is
usually resolved inside the device, but sometimes it can be resolved in the
host, i.e. the user should be able to wait in the host until a signal has
been completed.

The mechanism to define signal and wait operations is done by the driver
because it needs atomicity and serialization, which is already done in the
driver when submitting work to the different queues.

To implement this feature, the driver "takes" a couple of h/w resources,
and this is reflected by the defines added to the uapi file.

The signal/wait operations are done via the existing CS IOCTL, and they use
the same data structure. There is a difference in the meaning of some of
the parameters, and for that we added unions to make the code more
readable.

Signed-off-by: Omer Shpigelman <oshpigelman@habana.ai>
Reviewed-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>

habanalabs: add missing MODULE_DEVICE_TABLE

PCI drivers should use this define to declare their PCI ID table.

Reviewed-by: Tomer Tayar <ttayar@habana.ai>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>