Merge remote-tracking branch 'intel-speed-select/intel-sst' into review-hans

Summary of changes:

Till the commit:
"tools/power/x86/intel-speed-select: Get punit core mapping
information"

All patches just lays foundation for adding new hardware interface
in addition to existing MSR/MMIO interface.

All patches prefixed with "Abstract" just adds individual hardware
callback for functions called from the core processing.

The mailbox processing is abstracted to a new file with callbacks.

In addition some changes to reduce duplicated code for display
of TRL level.

Patch "Introduce Punit ID concept and add related API" add further
abstraction in addition to existing package and die scope.

-----------
tools/power/x86/intel-speed-select: Get punit core mapping information
-Adds a new way to get package/die/punit information for each CPU.
-----------

Introduce TPMI interface support
- Implement all abstracted callbacks above to use TPMI IOCTL interface.
-----------

tools/power/x86/intel-speed-select: Display punit info
- With TPMI the granularity is per punit, which can be multiple in a
package/die.
----------

tools/power/x86/intel-speed-select: Remove cpu mask display for non-cpu
power domain
tools/power/x86/intel-speed-select: Display fact info for non-cpu power
domain
-There are some power domains with no CPUs, the above patches adds CPU
as invalid for these domains
----------

tools/power/x86/intel-speed-select: Use cgroup v2 isolation
- Adds new option to use Cgroup CPU isolation instead of offline of CPUs
----------

tools/power/x86/intel-speed-select: Change TRL display for Emerald Rapids
- Emerald Rapid has multiple TRL levels and moves away from TRL level
based on instruction type

tools/power/x86/intel-speed-select: Update version

Update tool and supported API version. This is the first version which
supports newer Xeon platforms with TPMI support.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: subject and changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Change TRL display for Emerald Rapids

Emerald Rapids doesn't specify TRL (Turbo Ratio Limits) based instruction
types. Instead it specifies 5 TRL levels, which can be anyone of the
instruction types.

Increase TRL levels to 5 for Emerald Rapids. Also change display to show
by level number. Show only non zero level values.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Identify Emerald Rapids

There are some differences compared to Sapphire Rapids. So, add a separate
API.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Display AMX base frequency

AMX frequency is present in non TPMI platforms also. When platform
supports, the value is non zero. So, display AMX base frequency when
non zero, irrespective of platform API version.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Use cgroup v2 isolation

On supported systems, it is possiible to isolate CPUs instead of
CPU online/offline. This is optional and can be specified using
-g option when running as daemon or in combination with -o option
for SST-PP level change.

CPU isolation doesn't isolate IRQs. So IRQs needs to be moved away
from isoolated CPUs. This can be done via IRQ sysfs or irqbalance daemon.

The IRQ balance daemon is also capable to parse thermal HFI messages to
move IRQs away from CPUS, which are supposed be isolated. But this
requires version released after July 2022.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Add missing free cpuset

During perf level change cpuset is allocated but not freed.
Add free_cpu_set() in success and failure path.

Although this is not an issue, as the program will exit after
processing of online/offline, but for completeness add the
free_cpu_set().

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Fix clos-max display with TPMI I/F

Comparing clos_config->clos_max with 255 is broken with TPMI I/F because
of different isst_get_disp_freq_multiplier() used.

Checking for clos_config->clos_max * isst_get_disp_freq_multiplier()
instead.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Add cpu id check

Some operations applies to cpu-power-domain only. Add check for cpu id
for these functions.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Avoid setting duplicate tdp level

If the new TDP level requetsted is same as the current TDP level, don't
call into driver to change level.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Remove cpu mask display for non-cpu power domain

Non CPU power domains will not have any CPUs. So don't display any CPU
count or enable mask.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: subject and changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Hide invalid TRL level

TRL levels with Zero ratio values is meaningless.
Prevent these TRL levels from being displayed.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Display fact info for non-cpu power domain

Allow displaying SST-TF info for non-cpu power domain.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Show level 0 name for new api_version

level 0 name is not shown in some case for backward compatibility reason.
No need to keep this quirk for new api_version.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Prevent cpu clos config for non-cpu power domain

Non-cpu power domain does not support cpu clos config.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Allow display non-cpu power domain info

Some power domain may not have CPUs associated, allow displaying
information for these non-cpu power domains.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Display amx_p1 and cooling_type

amx_p1 and cooling_type are newly introduced for TPMI interface.

Display amx_p1 and cooling_type info for platforms that support them.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Display punit info

Display punit info for platforms with new api_version.
For platforms with old api_version, don't display the punit info to be
backward compatible.

For example:

Intel(R) Speed Select Technology
Executing on CPU model:173[0xad]
 package-0
  die-0
    powerdomain-0
      cpu-0
        get-config-current_level:0
 package-0
  die-0
    powerdomain-3
      cpu--1
        get-config-current_level:0
 package-0
  die-0
    powerdomain-4
      cpu--1
        get-config-current_level:0

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Introduce TPMI interface support

TPMI (Topology Aware Register and PM Capsule Interface) creates a
flexible, extendable and software-PCIe-driver-enumerable MMIO interface
for PM features.

SST feature is exposed via the TPMI interface on newer Xeon platforms.

Kernel TPMI based SST driver provides a series of new IOCTLs for userspace
to use.

Introduce support for the platforms that do SST control via TPMI interface.

Compared with previous platforms, Newer Xeons also supports multi-punit in a
package/die, including cpu punit and non-cpu punit. These have already
been handled in the generic code.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Get punit core mapping information

Get punit core mapping information using format of MSR 0x54. Based
on the API version, decode is done using new format. The new format
also include a power domain ID. TPMI SST information is for each
power domain.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Introduce api_version helper

In some cases, the output format may be different with different
api_version because of different capabilities or for backward
capabilities reason.

Introduce api_version() to get the api_version of the platform running.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Support large clos_min/max

clos_min/max in TPMI interface is frequency in MHz, thus clos_min/max
needs to support larger values.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Introduce is_debug_enabled()

Platform specific code also needs to give debug output.
Introduce is_debug_enabled() for this purpose.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Allow api_version based platform callbacks

Different api_version suggests different kernel driver used and
different interface is used to communication with the hardware.

Allow setting platform specific callbacks based on api_version.

Currently, all platforms with api_version 1 uses Mbox/MMIO interfaces.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Move send_mbox_cmd to isst-core-mbox.c

After the previous cleanup, there is no user of send_mbox_cmd outside of
isst-core-mbox.c.
Thus move send_mbox_cmd to isst-core-mbox.c as internal functions.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract adjust_uncore_freq

Allow platform specific implementation to adjust the uncore frequency.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract read_pm_config

Allow platform specific implementation to get SST-CP capability and
current state.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract clos_associate

Allow platform specific implementation to set per core CLOS setting.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract clos_get_assoc_status

Allow platform specific implementation to get per core CLOS setting.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract set_clos

Allow platform specific implementation to set CLOS priority setting.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract pm_get_clos

Allow platform specific implementation to get CLOS priority setting.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract pm_qos_config

Allow platform specific implementation to set CLOS config settings.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_clos_information

Allow platform specific implementation to get CLOS config setting.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_get_trls

Allow platform specific implementation to get turbo ratio limits of each
AVX level, for a selected SST-PP level.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Enhance get_tdp_info

mbox_get_uncore_p0_p1_info/get_p1_info/get_uncore_mem_freq can be done
inside get_tdp_info().

Fold the code into get_tdp_info().

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_uncore_p0_p1_info

Allow platform specific implementation to get uncore frequency info.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_fact_info

Allow platform specific implementation to get SST-TF info.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract set_pbf_fact_status

Allow platform specific implementation to enable/disable SST-TF/BF.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Remove isst_get_pbf_info_complete

isst_get_pbf_info_complete does nothing but just free the core_mask.
Remove the function and do free core_mask directly and free core mask in
the caller.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_pbf_info

Allow platform specific implementation to get SST-BF information.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract set_tdp_level

Allow platform specific implementation to set a SST-PP level.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_trl_bucket_info

Allow platform specific implementation to get buckets info.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_get_trl

Allow platform specific implementation to get turbo ratio limit of the
selected SST-PP level, and AVX level.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_coremask_info

Allow platform specific implementation to get the core mask for a given
SST-PP level.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_tjmax_info

Allow platform specific implementation to get the Tjmax info for a
given SST-PP level.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Move code right before its caller

Some functions are defined far from its only caller.
Rearrange the code.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_pwr_info

Allow platform specific implementation to get min and max power for a
given SST-PP level.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_tdp_info

Allow platform specific implementation to get TDP information.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_ctdp_control

Allow platform specific implementation to get SST-TF/BF/CP capabilities
and status.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract get_config_levels

Allow platform specific implementation to get SST-PP level.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Abstract is_punit_valid

Allow platform specific implementation to identify a valid punit.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Introduce isst-core-mbox.c

isst-core.c should contain generic core APIs only.
Platform specific implementations/configurations should be removed from
this file.

Introduce isst-core-mbox.c and move all mbox/mmio specific functions to
this file.

Introduce struct isst_platform_ops which contains a series of callbacks
that used by the core APIs but need platform specific implementation.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Always invoke isst_fill_platform_info

isst_fill_platform_info fills platform specific information.
And it is the proper place to set platform specific callbacks, as done in
next patch.

As the platform specific callbacks are needed in all cases, including
isst_print_platform_information.

The best way to achieve both is to invoke isst_fill_platform_info
unconditionally, and make isst_print_platform_information leverage the
data already filled.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Introduce isst_get_disp_freq_multiplier

Remove hardcoded DISP_FREQ_MULTIPLIER in the code and use
isst_get_disp_freq_multiplier() instead.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Move mbox functions to isst-core.c

isst-config.c should only contain generic code.
Move mbox functions which are platform specific code to isst-core.c.

As there are some platform specific parameters set via generic
application options, introduce isst_update_platform_param to pass these
parameters to platform specific code.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Improve isst_print_extended_platform_info

The main thing done in isst_print_extended_platform_info is to get the
isst feature status by checking one of the power domains of the
platform.

This can be done using the for_each_online_power_domain_in_set()
function, which makes the code clean and easier to read.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Rename for_each_online_package_in_set

for_each_online_package_in_set is actually used to invoke callback for
each power domain.
This is not a problem when there is a single power domain within a
package/die, but it does not reflect the truth in multi-punit case.

Rename for_each_online_package_in_set to
for_each_online_power_domain_in_set.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Introduce support for multi-punit

New platforms may have more than 1 punit in a Package/Die, thus it can
have multiple power domains in a Package/Die. Package id and die id is not
sufficient to refer to a specific Power domain.

Introduce support for multi-punit per package/die.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Introduce isst_is_punit_valid()

Introduce isst_is_punit_valid() for checking a valid domain.

For current platforms, it requires a punit 0 in a valid Package/Die.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Introduce punit to isst_id

Punit id can also be retrieved from ISST_IF_GET_PHY_ID.

punit id is unique within a Package/Die, and together with Package id and
Die id, they can be used to refer to a specific SST power domain.

For current platforms, Punit id is always Zero. So no functional changes
are expected for the current platforms.

While here, prevent issuing IOCTL if the file /dev/isst_interface can't be
opened.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Follow TRL nameing for FACT info

SST-TF high priority core count and ratios and low priority core ratios
are also per TRL level.
Cleanup the code to follow the same nameing convention as TRL.

This removes hardcoded TRL level names and variables.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select: Unify TRL levels

TRL supports different levels including SSE/AVX2/AVX512.

Avoid using hardcoded level name and structure fields, so that a loop can
be used to parse each TRL level instead. This reduces several lines of
source code.

No functional changes are expected.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
[srinivas.pandruvada@linux.intel.com: changelog edits]
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

platform/x86/intel: vsec: Use intel_vsec_dev_release() to simplify init() error cleanup

On auxiliary_device_init(auxdev) failure we need to do the exact same
cleanup steps as on device.release(), so use the intel_vsec_dev_release()
callback for this.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/20230320103815.229729-1-hdegoede@redhat.com

platform/x86/intel/pmt: Add INTEL_PMT module namespace

Since the currently exported symbols in pmt_class are only used by other
Intel PMT drivers, create an INTEL_PMT module namespace for them.

Signed-off-by: David E. Box <david.e.box@linux.intel.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/20230316225736.2856521-1-david.e.box@linux.intel.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86/intel: vsec: Explicitly enable capabilities

Discovered Intel VSEC/DVSEC capabilities are enabled by default and only
get disabled by quirk. Instead, remove such quirks and only enable support
for capabilities that have been explicitly added to a new capabilities
field. While here, also reorder the device info structures alphabetically.

Signed-off-by: David E. Box <david.e.box@linux.intel.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/20230316224628.2855884-1-david.e.box@linux.intel.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/olpc: olpc-xo175-ec: Use SPI device ID data to bind device

The driver defines spi_device_id table for module autoloading, but does
not use it in id_table which causes W=1 warning:

  drivers/platform/olpc/olpc-xo175-ec.c:737:35: error: ‘olpc_xo175_ec_id_table’ defined but not used [-Werror=unused-const-variable=]

Reference the SPI device ID table, so it can be also used for device
binding.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20230316160324.78856-1-krzysztof.kozlowski@linaro.org
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: pcengines-apuv2: Drop platform:pcengines-apuv2 module-alias

The driver auto-loads based on the DMI modaliases and this platform
modalias is not necessary and broken:

1. Not necessary since the driver itself creates the platform_device,
   so it is already loaded when the modalias might be used
2. From a quick scan of the code it does not create any platform-devices
   called "platform:pcengines-apuv2"

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20230313130241.778146-1-hdegoede@redhat.com

platform/x86: sony: mark SPI related data as maybe unused

The driver can be compile tested as built-in making certain data unused:

  drivers/platform/x86/sony-laptop.c:3290:36: error: ‘sony_device_ids’ defined but not used [-Werror=unused-const-variable=]

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20230312132624.352703-3-krzysztof.kozlowski@linaro.org
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: classmate: mark SPI related data as maybe unused

The driver can be compile tested as built-in making certain data unused:

  drivers/platform/x86/classmate-laptop.c:1137:36: error: ‘cmpc_device_ids’ defined but not used [-Werror=unused-const-variable=]

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20230312132624.352703-2-krzysztof.kozlowski@linaro.org
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: think-lmi: Remove custom kobject sysfs_ops

think-lmi defines its own sysfs_ops that are identical to the standard
kobj_sysfs_ops. Use the standard definitions.

Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Link: https://lore.kernel.org/r/20230314-think-lmi-sysfs_ops-v1-1-9d4f1cf9caec@weissschuh.net
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: think-lmi: Properly interpret return value of tlmi_setting

The return value of tlmi_settings() is an errorcode, not an acpi_status.

Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Link: https://lore.kernel.org/r/20230312-think-lmi-status-v1-1-4e9f36322cc4@weissschuh.net
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: acerhdf: Remove unneeded semicolon

Fix the following warning:

./drivers/platform/x86/acerhdf.c:343:2-3: Unneeded semicolon.

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Link: https://bugzilla.openanolis.cn/show_bug.cgi?id=2271
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Link: https://lore.kernel.org/r/20230309061045.25256-1-jiapeng.chong@linux.alibaba.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: ISST: Add suspend/resume callbacks

To support S3/S4 with TPMI interface add suspend/resume callbacks.
Here HW state is stored in suspend callback and restored during
resume callback.

The hardware state which needs to be stored/restored:
- CLOS configuration
- CLOS Association
- SST-CP enable/disable status
- SST-PP perf level setting

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Suggested-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Pragya Tanwar <pragya.tanwar@intel.com>
Link: https://lore.kernel.org/r/20230308070642.1727167-9-srinivas.pandruvada@linux.intel.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: ISST: Add SST-TF support via TPMI

The support of Intel Speed Select Technology - Turbo Frequency (SST-TF)
feature enables the ability to set different “All core turbo ratio
limits” to cores based on the priority. By using this feature, some cores
can be configured to get higher turbo frequency by designating them as
high priority at the cost of lower or no turbo frequency on the low
priority cores.

One new IOCTLs are added:
ISST_IF_GET_TURBO_FREQ_INFO : Get information about turbo frequency
buckets

Once an instance is identified, read or write from correct MMIO
offset for a given field as defined in the specification.

For details on SST-TF operations using intel-speed-selet utility,
refer to:
Documentation/admin-guide/pm/intel-speed-select.rst
under the kernel documentation

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Pragya Tanwar <pragya.tanwar@intel.com>
Link: https://lore.kernel.org/r/20230308070642.1727167-8-srinivas.pandruvada@linux.intel.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: ISST: Add SST-BF support via TPMI

The Intel Speed Select Technology - Base Frequency (SST-BF) feature lets
the user control base frequency. If some critical workload threads demand
constant high guaranteed performance, then this feature can be used to
execute the thread at higher base frequency on specific sets of CPUs
(high priority CPUs) at the cost of lower base frequency (low priority
CPUs) on other CPUs.

Two new IOCTLs are added:
ISST_IF_GET_BASE_FREQ_INFO : Get frequency information for high and
low priority CPUs
ISST_IF_GET_BASE_FREQ_CPU_MASK : CPUs capable of higher frequency

Once an instance is identified, read or write from correct MMIO
offset for a given field as defined in the specification.

For details on SST-BF operations using intel-speed-selet utility,
refer to:
Documentation/admin-guide/pm/intel-speed-select.rst
under the kernel documentation

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Pragya Tanwar <pragya.tanwar@intel.com>
Link: https://lore.kernel.org/r/20230308070642.1727167-7-srinivas.pandruvada@linux.intel.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: ISST: Add SST-PP support via TPMI

This Intel Speed Select Technology - Performance Profile (SST-PP) feature
introduces a mechanism that allows multiple optimized performance profiles
per system. Each profile defines a set of CPUs that need to be online and
rest offline to sustain a guaranteed base frequency.

Five new IOCTLs are added:
ISST_IF_PERF_LEVELS : Get number of performance levels
ISST_IF_PERF_SET_LEVEL : Set to a new performance level
ISST_IF_PERF_SET_FEATURE : Activate SST-BF/SST-TF for a performance level
ISST_IF_GET_PERF_LEVEL_INFO : Get parameters for a performance level
ISST_IF_GET_PERF_LEVEL_CPU_MASK : Get CPU mask for a performance level

Once an instance is identified, read or write from correct MMIO
offset for a given field as defined in the specification.

For details on SST PP operations using intel-speed-selet utility,
refer to:
Documentation/admin-guide/pm/intel-speed-select.rst
under the kernel documentation

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Pragya Tanwar <pragya.tanwar@intel.com>
Link: https://lore.kernel.org/r/20230308070642.1727167-6-srinivas.pandruvada@linux.intel.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: ISST: Add SST-CP support via TPMI

Intel Speed Select Technology Core Power (SST-CP) is an interface that
allows users to define per core priority. This defines a mechanism to
distribute power among cores when there is a power constrained
scenario. This defines a class of service (CLOS) configuration.

Three new IOCTLs are added:
ISST_IF_CORE_POWER_STATE : Enable/Disable SST-CP
ISST_IF_CLOS_PARAM : Configure CLOS parameters
ISST_IF_CLOS_ASSOC : Associate CPUs to a CLOS

To associate CPUs to CLOS, either Linux CPU numbering or PUNIT numbering
scheme can be used, using parameter punit_cpu_map (1: for PUNIT numbering
0 for Linux CPU number).

There is no change to IOCTL to get PUNIT CPU number for a CPU.

Introduce get_instance() function, which is used by majority of IOCTLs
processing to convert a socket and power domain to
tpmi_per_power_domain_info * instance. This instance has all the MMIO
offsets stored to read a particular field.

Once an instance is identified, read or write from correct MMIO
offset for a given field as defined in the specification.

For details on SST CP operations using intel-speed-selet utility,
refer to:
Documentation/admin-guide/pm/intel-speed-select.rst
under the kernel documentation

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Pragya Tanwar <pragya.tanwar@intel.com>
Link: https://lore.kernel.org/r/20230308070642.1727167-5-srinivas.pandruvada@linux.intel.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: ISST: Parse SST MMIO and update instance

SST registers are presented to OS in multi-layer structures starting
with a SST header showing version information freezing current
definition.

For details on SST terminology refer to
Documentation/admin-guide/pm/intel-speed-select.rst
under the kernel documentation

SST TPMI details are published in the following document:
https://github.com/intel/tpmi_power_management/blob/main/SST_TPMI_public_disclosure_FINAL.docx

SST MMIO structure layout follows:
SST-HEADER
SST-CP Header
SST-CP CONTROL
SST-CP STATUS
SST-CP CONFIG0
SST-CP CONFIG1
...
...
SST-PP Header
SST-PP OFFSET_0
SST-PP OFFSET_1
SST_PP_0_INFO
SST_PP_1_INFO
SST_PP_2_INFO
SST_PP_3_INFO
SST-PP CONTROL
SST-PP STATUS

Each register bank contains information to get to next lower level
information. This information is parsed and stored in the struct
tpmi_per_power_domain_info for each domain. This information is
used to process each SST requests.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Pragya Tanwar <pragya.tanwar@intel.com>
Link: https://lore.kernel.org/r/20230308070642.1727167-4-srinivas.pandruvada@linux.intel.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: ISST: Enumerate TPMI SST and create framework

Enumerate TPMI SST driver and create basic framework to add more
features.

The basic user space interface is still same as the legacy using
/dev/isst_interface. Users of "intel-speed-select" utility should
be able to use same commands as prior gens without being aware
of new underlying hardware interface.

TPMI SST driver enumerates on device "intel_vsec.tpmi-sst". Since there
can be multiple instances and there is one common SST core, split
implementation into two parts: A common core part and an enumeration
part. The enumeration driver is loaded for each device instance and
register with the TPMI SST core driver.

On very first enumeration the TPMI SST core driver register with SST
core driver to get IOCTL callbacks. The api_version is incremented
for IOCTL ISST_IF_GET_PLATFORM_INFO, so that user space can issue
new IOCTLs.

Each TPMI package contains multiple power domains. Each power domain
has its own set of SST controls. For each domain map the MMIO memory
and update per domain struct tpmi_per_power_domain_info. This information
will be used to implement other SST interfaces.

Implement first IOCTL commands to get number of TPMI SST instances
and instance mask as some of the power domains may not have any
SST controls.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Pragya Tanwar <pragya.tanwar@intel.com>
Link: https://lore.kernel.org/r/20230308070642.1727167-3-srinivas.pandruvada@linux.intel.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: ISST: Add support for MSR 0x54

To map Linux CPU numbering scheme to hardware CPU numbering scheme
MSR 0x53 is getting used. But for new generation of CPUs, this MSR
is not valid. Since this is model specific MSR, this is possible.

A new MSR 0x54 is defined for this purpose. User space can use the
API version to distinguish format from MSR 0x53.

Intel speed select utility is updated to use the new format based
on the API version.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Reviewed-by: Zhang Rui <rui.zhang@intel.com>
Tested-by: Pragya Tanwar <pragya.tanwar@intel.com>
Link: https://lore.kernel.org/r/20230308070642.1727167-2-srinivas.pandruvada@linux.intel.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: dell-laptop: Register ctl-led for speaker-mute

Some platforms have the speaker-mute led and
current driver doesn't control it.

If the platform support the control of speaker-mute led, register it

Signed-off-by: Koba Ko <koba.ko@canonical.com>
Link: https://lore.kernel.org/r/20230308062414.1048913-1-koba.ko@canonical.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/surface: aggregator_registry: Add support for tablet-mode switch on Surface Pro 9

Add support for the POS-subsystem tablet-mode switch used on the Surface
Pro 9.

Signed-off-by: Maximilian Luz <luzmaximilian@gmail.com>
Link: https://lore.kernel.org/r/20230304194611.87770-4-luzmaximilian@gmail.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/surface: aggregator_tabletsw: Add support for Type-Cover posture source

Implement support for the Type-Cover posture source (ID 0x00), found on
the Surface Pro 9.

Signed-off-by: Maximilian Luz <luzmaximilian@gmail.com>
Link: https://lore.kernel.org/r/20230304194611.87770-3-luzmaximilian@gmail.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/surface: aggregator_tabletsw: Properly handle different posture source IDs

The device posture subsystem (POS) can provide different posture
sources. Different sources can provide different posture states and
sources can be identified by their ID.

For example, screen posture of the Surface Laptop Studio (SLS), which is
currently the only supported source, uses a source ID of 0x03. The
Surface Pro 9 uses the same subsystem for its Type-Cover, however,
provides different states for that under the ID 0x00.

To eventually support the Surface Pro 9 and potential future devices, we
need to properly disambiguate between source IDs. Therefore, add the
source ID to the state we carry and determine the tablet-mode state (as
well as state names) based on that.

Signed-off-by: Maximilian Luz <luzmaximilian@gmail.com>
Link: https://lore.kernel.org/r/20230304194611.87770-2-luzmaximilian@gmail.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: Add intel_bytcrc_pwrsrc driver

Add a new driver for the power-, wake- and reset-source functionality
of the Bay Trail (BYT) version of the Crystal Cove PMIC.

The main functionality here is detecting which power-sources (USB /
DC in / battery) are active. This is normally exposed to userspace as
a power_supply class charger device with an online sysfs attribute.

But if a charger is online or not is already exposed on BYT-CRC devices
through either an ACPI AC power_supply device, or through a native driver
for the battery charger chip (e.g. a BQ24292i).

So instead of adding duplicate info under the power_supply class this
driver exports the info through debugfs and likewise adds debugfs files
for the reset- and wake-source info / registers.

Despite this driver only exporting debugfs bits it is still useful to
have this driver because it clears the wake- and reset-source registers
after reading them. Not clearing these can have undesirable side-effects.

Specifically if the WAKESRC register contains 0x01 (wake by powerbutton)
on reboot then the firmware on some tablets turns the reboot into
a poweroff. I guess this may be necessary to make long power-presses turn
into a poweroff somehow?

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20230303221928.285477-1-hdegoede@redhat.com

Merge tag 'ib-pdx86-backlight-6.4' into review-hans

Immutable branch between pdx86 and backlight due for the v6.4 merge window

platform/x86: x86-android-tablets: Add depends on PMIC_OPREGION

Add a depends on PMIC_OPREGION to x86-android-tablets Kconfig to fix
the following build error:

ERROR: modpost: "intel_soc_pmic_exec_mipi_pmic_seq_element"
 [drivers/platform/x86/x86-android-tablets/x86-android-tablets.ko] undefined!

Fixes: 9b1d2662b8c5 ("platform/x86: x86-android-tablets: Add touchscreen support for Lenovo Yoga Tab 3 Pro YT3-X90F")
Reported-by: kernel test robot <lkp@intel.com>
Link: https://lore.kernel.org/oe-kbuild-all/202303091711.howZNrIY-lkp@intel.com/
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20230309094035.18736-1-hdegoede@redhat.com

backlight: apple_bl: Use acpi_video_get_backlight_type()

On some MacBooks both the apple_bl and the apple-gmux backlight drivers
may be able to export a /sys/class/backlight device.

To avoid having 2 backlight devices for one LCD panel until now
the apple-gmux driver has been calling apple_bl_unregister() to move
the apple_bl backlight device out of the way when it loads.

Similar problems exist on other x86 laptops and all backlight drivers
which may be used on x86 laptops have moved to using
acpi_video_get_backlight_type() to determine whether they should load
or not.

Switch apple_bl to this model too, so that it is consistent with all
the other x86 backlight drivers.

Besides code-simplification and consistency this has 2 other benefits:

1) It removes a race during boot where userspace will briefly see
   an apple_bl backlight and then have it disappear again, leading to e.g.:
   https://bbs.archlinux.org/viewtopic.php?id=269920

2) This allows user to switch between the drivers by passing
   acpi_backlight=apple_gmux or acpi_backlight=vendor on the kernel
   commandline.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-by: Daniel Thompson <daniel.thompson@linaro.org>
Acked-by: Lee Jones <lee@kernel.org>
Link: https://lore.kernel.org/r/20230307120540.389920-1-hdegoede@redhat.com

platform/x86: apple-gmux: return -EFAULT if copy fails

The copy_to/from_user() functions return the number of bytes remaining
to be copied, but we want to return -EFAULT to the user.

Fixes: ce3fef2eb235 ("platform/x86: apple-gmux: add debugfs interface")
Signed-off-by: Dan Carpenter <error27@gmail.com>
Reviewed-by: Orlando Chamberlain <orlandlch.dev@gmail.com>
Link: https://lore.kernel.org/r/0bdfa8c2-cb22-4bec-8773-584060613043@kili.mountain
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: apple-gmux: Update apple_gmux_detect documentation

Commit fc83fbc80e1a ("platform/x86: apple-gmux: refactor gmux types")
neglected to update the documentation of apple_gmux_detect's arguments.
Update the documentation to reflect that commit's changes.

include/linux/apple-gmux.h:99: warning:
Function parameter or member 'type_ret' not described in 'apple_gmux_detect'
include/linux/apple-gmux.h:99: warning:
Excess function parameter 'indexed_ret' description in 'apple_gmux_detect'

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Link: https://lore.kernel.org/lkml/20230309122822.77435e33@canb.auug.org.au/
Fixes: fc83fbc80e1a ("platform/x86: apple-gmux: refactor gmux types")
Signed-off-by: Orlando Chamberlain <orlandoch.dev@gmail.com>
Link: https://lore.kernel.org/r/20230309104353.10905-1-orlandoch.dev@gmail.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: apple-gmux: Add acpi_video_get_backlight_type() check

Make apple-gmux backlight registration honor the acpi_backlight=...
kernel commandline option which is used to select the backlight
control method on x86/ACPI devices.

Reported-and-tested-by: Aditya Garg <gargaditya08@live.com>
Link: https://lore.kernel.org/platform-driver-x86/BM1PR01MB0931B467250831916F7C55B3B8A59@BM1PR01MB0931.INDPRD01.PROD.OUTLOOK.COM/
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20230307113813.359743-1-hdegoede@redhat.com

platform/x86: apple-gmux: add debugfs interface

Allow reading and writing gmux ports from userspace.

For example:

echo 4 > /sys/kernel/debug/apple_gmux/selected_port
cat /sys/kernel/debug/apple_gmux/selected_port_data | xxd -p

Will show the gmux version information (00000005 in this case)

Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Orlando Chamberlain <orlandoch.dev@gmail.com>
Link: https://lore.kernel.org/r/20230303112842.3094-5-orlandoch.dev@gmail.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: apple-gmux: support MMIO gmux on T2 Macs

In some newer dual gpu MacBooks, the T2 Coprocessor functions as the
gmux, and the Intel side can interract with this new gmux type through
MMIO. Add support for these gmux controllers to the apple-gmux driver.

We start using the GMSP(0) acpi method on these gmux's when clearing
interrupts, as this prevents a flood of status=0 interrupts that can't
be cleared. It's unknown if this helps or hinders older gmux types, so
it isn't enabled for those.

Interestingly, the ACPI table only allocates 8 bytes for GMUX, but we
actually need 16, and as such we request 16 with request_mem_region.

Reading and writing from ports:
    16 bytes from 0xfe0b0200 are used. 0x0 to 0x4 are where data
    to read appears, and where data to write goes. Writing to 0xe
    sets the gmux port being accessed, and writing to 0xf sends commands.

    These commands are 0x40 & data_length for write, and data_length for
    read, where data_length is 1, 2 or 4. Once byte base+0xf is 0, the
    command is done.

Issues:
    As with other retina models, we can't switch DDC lines so
    switching at runtime doesn't work if the inactive gpu driver
    already disabled eDP due to it not being connected when that
    driver loaded.

    Additionally, turning on the dgpu back on on the MacBookPro16,1 does
    not work.

Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Orlando Chamberlain <orlandoch.dev@gmail.com>
Link: https://lore.kernel.org/r/20230303112842.3094-4-orlandoch.dev@gmail.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: apple-gmux: refactor gmux types

Add apple_gmux_config struct containing operations and data specific to
each mux type.

This is in preparation for adding a third, MMIO based, gmux type.

Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Orlando Chamberlain <orlandoch.dev@gmail.com>
Link: https://lore.kernel.org/r/20230303112842.3094-3-orlandoch.dev@gmail.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: apple-gmux: use first bit to check switch state

On T2 Macs with MMIO gmux, when GMUX_PORT_SWITCH_DISPLAY is read, it can
have values of 2, 3, 4, and 5. Odd values correspond to the discrete gpu,
and even values correspond to the integrated gpu. The current logic is
that only 2 corresponds to IGD, but this doesn't work for T2 Macs.
Instead, check the first bit to determine the connected gpu.

As T2 Macs with gmux only can switch the internal display, it is
untested if this change (or a similar change) would be applicable
to GMUX_PORT_SWITCH_DDC and GMUX_PORT_SWITCH_EXTERNAL.

Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Orlando Chamberlain <orlandoch.dev@gmail.com>
Link: https://lore.kernel.org/r/20230303112842.3094-2-orlandoch.dev@gmail.com
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: xo1-rfkill: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.

Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20230302144732.1903781-30-u.kleine-koenig@pengutronix.de
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: wmi: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.

Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20230302144732.1903781-29-u.kleine-koenig@pengutronix.de
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: serial-multi-instantiate: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.

Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20230302144732.1903781-28-u.kleine-koenig@pengutronix.de
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: samsung-q10: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.

Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20230302144732.1903781-27-u.kleine-koenig@pengutronix.de
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>

platform/x86: intel: vbtn: Convert to platform remove callback returning void

The .remove() callback for a platform driver returns an int which makes
many driver authors wrongly assume it's possible to do error handling by
returning an error code. However the value returned is (mostly) ignored
and this typically results in resource leaks. To improve here there is a
quest to make the remove callback return void. In the first step of this
quest all drivers are converted to .remove_new() which already returns
void.

Trivially convert this driver from always returning zero in the remove
callback to the void returning variant.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Link: https://lore.kernel.org/r/20230302144732.1903781-26-u.kleine-koenig@pengutronix.de
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>