linux-stable.git/drivers/platform/x86/amd, branch linux-6.4.y

platform/x86/amd/pmf: Fix a missing cleanup path

2023-09-13T07:48:34+00:00

[ Upstream commit 4dbd6e61adc7e52dd1c9165f0ccaa90806611e40 ]

On systems that support slider notifications but don't otherwise support
granular slider the SPS cleanup path doesn't run.

This means that loading/unloading/loading leads to failures because
the sysfs files don't get setup properly when reloaded.

Add the missing cleanup path.

Fixes: 33c9ab5b493a ("platform/x86/amd/pmf: Notify OS power slider update")
Signed-off-by: Mario Limonciello 
Link: https://lore.kernel.org/r/20230823185421.23959-1-mario.limonciello@amd.com
Reviewed-by: Hans de Goede 
Signed-off-by: Hans de Goede 
Signed-off-by: Sasha Levin

platform/x86/amd/pmf: Fix unsigned comparison with less than zero

2023-09-13T07:47:53+00:00

[ Upstream commit 785c00993dc4c4bb2f7b0f3a3f29c03a6f7aab2e ]

The return value from the call to amd_pmf_get_pprof_modes() is int.
However, the return value is being assigned to an unsigned char
variable 'mode', so making 'mode' an int.

silence the warning:
./drivers/platform/x86/amd/pmf/sps.c:183:5-9: WARNING: Unsigned expression compared with zero: mode < 0

Reported-by: Abaci Robot 
Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=5995
Signed-off-by: Yang Li 
Link: https://lore.kernel.org/r/20230727014315.51375-1-yang.lee@linux.alibaba.com
Reviewed-by: Hans de Goede 
Signed-off-by: Hans de Goede 
Signed-off-by: Sasha Levin

platform/x86/amd/pmf: reduce verbosity of apmf_get_system_params

2023-08-03T08:25:40+00:00

commit 839e90e75e695b3d9ee17f5a2811e7ee5aea8d4a upstream.

apmf_get_system_params() failure is not a critical event, reduce its
verbosity from dev_err to dev_dbg.

Signed-off-by: Mario Limonciello 
Signed-off-by: Shyam Sundar S K 
Link: https://lore.kernel.org/r/20230714144435.1239776-1-Shyam-sundar.S-k@amd.com
Reviewed-by: Hans de Goede 
Signed-off-by: Hans de Goede 
Signed-off-by: Greg Kroah-Hartman

platform/x86/amd/pmf: Notify OS power slider update

2023-08-03T08:25:40+00:00

commit 33c9ab5b493a0e922b06c12fed4fdcb862212cda upstream.

APMF fn8 can notify EC about the OS slider position change. Add this
capability to the PMF driver so that it can call the APMF fn8 based on
the changes in the Platform profile events.

Co-developed-by: Mario Limonciello 
Signed-off-by: Mario Limonciello 
Signed-off-by: Patil Rajesh Reddy 
Signed-off-by: Shyam Sundar S K 
Link: https://lore.kernel.org/r/20230714144435.1239776-2-Shyam-sundar.S-k@amd.com
Reviewed-by: Hans de Goede 
Signed-off-by: Hans de Goede 
Signed-off-by: Greg Kroah-Hartman

platform/x86/amd/pmf: Register notify handler only if SPS is enabled

2023-06-22T08:20:00+00:00

Power source notify handler is getting registered even when none of the
PMF feature in enabled leading to a crash.

...
[   22.592162] Call Trace:
[   22.592164]  
[   22.592164]  ? rcu_note_context_switch+0x5e0/0x660
[   22.592166]  ? __warn+0x81/0x130
[   22.592171]  ? rcu_note_context_switch+0x5e0/0x660
[   22.592172]  ? report_bug+0x171/0x1a0
[   22.592175]  ? prb_read_valid+0x1b/0x30
[   22.592177]  ? handle_bug+0x3c/0x80
[   22.592178]  ? exc_invalid_op+0x17/0x70
[   22.592179]  ? asm_exc_invalid_op+0x1a/0x20
[   22.592182]  ? rcu_note_context_switch+0x5e0/0x660
[   22.592183]  ? acpi_ut_delete_object_desc+0x86/0xb0
[   22.592186]  ? acpi_ut_update_ref_count.part.0+0x22d/0x930
[   22.592187]  __schedule+0xc0/0x1410
[   22.592189]  ? ktime_get+0x3c/0xa0
[   22.592191]  ? lapic_next_event+0x1d/0x30
[   22.592193]  ? hrtimer_start_range_ns+0x25b/0x350
[   22.592196]  schedule+0x5e/0xd0
[   22.592197]  schedule_hrtimeout_range_clock+0xbe/0x140
[   22.592199]  ? __pfx_hrtimer_wakeup+0x10/0x10
[   22.592200]  usleep_range_state+0x64/0x90
[   22.592203]  amd_pmf_send_cmd+0x106/0x2a0 [amd_pmf bddfe0fe3712aaa99acce3d5487405c5213c6616]
[   22.592207]  amd_pmf_update_slider+0x56/0x1b0 [amd_pmf bddfe0fe3712aaa99acce3d5487405c5213c6616]
[   22.592210]  amd_pmf_set_sps_power_limits+0x72/0x80 [amd_pmf bddfe0fe3712aaa99acce3d5487405c5213c6616]
[   22.592213]  amd_pmf_pwr_src_notify_call+0x49/0x90 [amd_pmf bddfe0fe3712aaa99acce3d5487405c5213c6616]
[   22.592216]  notifier_call_chain+0x5a/0xd0
[   22.592218]  atomic_notifier_call_chain+0x32/0x50
...

Fix this by moving the registration of source change notify handler only
when SPS(Static Slider) is advertised as supported.

Reported-by: Allen Zhong 
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217571
Fixes: 4c71ae414474 ("platform/x86/amd/pmf: Add support SPS PMF feature")
Tested-by: Patil Rajesh Reddy 
Reviewed-by: Mario Limonciello 
Signed-off-by: Shyam Sundar S K 
Link: https://lore.kernel.org/r/20230622060309.310001-1-Shyam-sundar.S-k@amd.com
Reviewed-by: Hans de Goede 
Signed-off-by: Hans de Goede

platform/x86/amd/pmf: Fix CnQF and auto-mode after resume

2023-05-15T12:56:13+00:00

After suspend/resume cycle there is an error message and auto-mode
or CnQF stops working.

[ 5741.447511] amd-pmf AMDI0100:00: SMU cmd failed. err: 0xff
[ 5741.447523] amd-pmf AMDI0100:00: AMD_PMF_REGISTER_RESPONSE:ff
[ 5741.447527] amd-pmf AMDI0100:00: AMD_PMF_REGISTER_ARGUMENT:7
[ 5741.447531] amd-pmf AMDI0100:00: AMD_PMF_REGISTER_MESSAGE:16
[ 5741.447540] amd-pmf AMDI0100:00: [AUTO_MODE] avg power: 0 mW mode: QUIET

This is because the DRAM address used for accessing metrics table
needs to be refreshed after a suspend resume cycle. Add a resume
callback to reset this again.

Fixes: 1a409b35c995 ("platform/x86/amd/pmf: Get performance metrics from PMFW")
Signed-off-by: Mario Limonciello 
Link: https://lore.kernel.org/r/20230513011408.958-1-mario.limonciello@amd.com
Reviewed-by: Hans de Goede 
Signed-off-by: Hans de Goede

Merge tag 'pm-6.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm

2023-04-26T01:44:10+00:00

Pull power management updates from Rafael Wysocki:
 "These update several cpufreq drivers and the cpufreq core, add sysfs
  interface for exposing the time really spent in the platform low-power
  state during suspend-to-idle, update devfreq (core and drivers) and
  the pm-graph suite of tools and clean up code.

  Specifics:

   - Fix the frequency unit in cpufreq_verify_current_freq checks()
     Sanjay Chandrashekara)

   - Make mode_state_machine in amd-pstate static (Tom Rix)

   - Make the cpufreq core require drivers with target_index() to set
     freq_table (Viresh Kumar)

   - Fix typo in the ARM_BRCMSTB_AVS_CPUFREQ Kconfig entry (Jingyu Wang)

   - Use of_property_read_bool() for boolean properties in the pmac32
     cpufreq driver (Rob Herring)

   - Make the cpufreq sysfs interface return proper error codes on
     obviously invalid input (qinyu)

   - Add guided autonomous mode support to the AMD P-state driver (Wyes
     Karny)

   - Make the Intel P-state driver enable HWP IO boost on all server
     platforms (Srinivas Pandruvada)

   - Add opp and bandwidth support to tegra194 cpufreq driver (Sumit
     Gupta)

   - Use of_property_present() for testing DT property presence (Rob
     Herring)

   - Remove MODULE_LICENSE in non-modules (Nick Alcock)

   - Add SM7225 to cpufreq-dt-platdev blocklist (Luca Weiss)

   - Optimizations and fixes for qcom-cpufreq-hw driver (Krzysztof
     Kozlowski, Konrad Dybcio, and Bjorn Andersson)

   - DT binding updates for qcom-cpufreq-hw driver (Konrad Dybcio and
     Bartosz Golaszewski)

   - Updates and fixes for mediatek driver (Jia-Wei Chang and
     AngeloGioacchino Del Regno)

   - Use of_property_present() for testing DT property presence in the
     cpuidle code (Rob Herring)

   - Drop unnecessary (void *) conversions from the PM core (Li zeming)

   - Add sysfs files to represent time spent in a platform sleep state
     during suspend-to-idle and make AMD and Intel PMC drivers use them
     Mario Limonciello)

   - Use of_property_present() for testing DT property presence (Rob
     Herring)

   - Add set_required_opps() callback to the 'struct opp_table', to make
     the code paths cleaner (Viresh Kumar)

   - Update the pm-graph siute of utilities to v5.11 with the following
     changes:
       * New script which allows users to install the latest pm-graph
         from the upstream github repo.
       * Update all the dmesg suspend/resume PM print formats to be able
         to process recent timelines using dmesg only.
       * Add ethtool output to the log for the system's ethernet device
         if ethtool exists.
       * Make the tool more robustly handle events where mangled dmesg
         or ftrace outputs do not include all the requisite data.

   - Make the sleepgraph utility recognize "CPU killed" messages (Xueqin
     Luo)

   - Remove unneeded SRCU selection in Kconfig because it's always set
     from devfreq core (Paul E. McKenney)

   - Drop of_match_ptr() macro from exynos-bus.c because this driver is
     always using the DT table for driver probe (Krzysztof Kozlowski)

   - Use the preferred of_property_present() instead of the low-level
     of_get_property() on exynos-bus.c (Rob Herring)

   - Use devm_platform_get_and_ioream_resource() in exyno-ppmu.c (Yang
     Li)"

* tag 'pm-6.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (44 commits)
  platform/x86/intel/pmc: core: Report duration of time in HW sleep state
  platform/x86/intel/pmc: core: Always capture counters on suspend
  platform/x86/amd: pmc: Report duration of time in hw sleep state
  PM: Add sysfs files to represent time spent in hardware sleep state
  cpufreq: use correct unit when verify cur freq
  cpufreq: tegra194: add OPP support and set bandwidth
  cpufreq: amd-pstate: Make varaiable mode_state_machine static
  PM: core: Remove unnecessary (void *) conversions
  cpufreq: drivers with target_index() must set freq_table
  PM / devfreq: exynos-ppmu: Use devm_platform_get_and_ioremap_resource()
  OPP: Move required opps configuration to specialized callback
  OPP: Handle all genpd cases together in _set_required_opps()
  cpufreq: qcom-cpufreq-hw: Revert adding cpufreq qos
  dt-bindings: cpufreq: cpufreq-qcom-hw: Add QCM2290
  dt-bindings: cpufreq: cpufreq-qcom-hw: Sanitize data per compatible
  dt-bindings: cpufreq: cpufreq-qcom-hw: Allow just 1 frequency domain
  cpufreq: Add SM7225 to cpufreq-dt-platdev blocklist
  cpufreq: qcom-cpufreq-hw: fix double IO unmap and resource release on exit
  cpufreq: mediatek: Raise proc and sram max voltage for MT7622/7623
  cpufreq: mediatek: raise proc/sram max voltage for MT8516
  ...

platform/x86/amd: pmc: Report duration of time in hw sleep state

2023-04-20T17:06:12+00:00

amd_pmc displays a warning when a suspend didn't get to the deepest
state and a dynamic debugging message with the duration if it did.

Rather than logging to dynamic debugging the duration spent in the
deepest state, report this to the standard kernel reporting
infrastructure so that userspace software can query after the
suspend cycle is done.

Reviewed-by: Hans de Goede 
Signed-off-by: Mario Limonciello 
Signed-off-by: Rafael J. Wysocki

platform/x86: amd: pmc: provide user message where s0ix is not supported

2023-04-17T09:33:39+00:00

Some platforms do not support hardware backed s0i3 transitions. When such
CPUs are detected, provide a warning message to the user.

Suggested-by: Sanket Goswami 
Signed-off-by: Shyam Sundar S K 
Reviewed-by: Mario Limonciello 
Link: https://lore.kernel.org/r/20230412111500.2602529-1-Shyam-sundar.S-k@amd.com
Reviewed-by: Hans de Goede 
Signed-off-by: Hans de Goede

platform/x86/amd: pmc: Fix memory leak in amd_pmc_stb_debugfs_open_v2()

2023-04-17T08:27:53+00:00

Function amd_pmc_stb_debugfs_open_v2() may be called when the STB
debug mechanism enabled.

When amd_pmc_send_cmd() fails, the 'buf' needs to be released.

Signed-off-by: Feng Jiang 
Link: https://lore.kernel.org/r/20230412093734.1126410-1-jiangfeng@kylinos.cn
Reviewed-by: Hans de Goede 
Signed-off-by: Hans de Goede