[Deepin-Kernel-SIG] [linux 6.6-y] [HISI] Backport cpufreq/pcie/ext-gpu/soc_cache/power_meter/perf_iostat/sdei_watchdog/HOTPLUG_SMT/FEAT_PMUv3_TH module for Kunpeng new SOC

[Avenger-285714]

Fixed the wrong debugfs node name in hisi_spi debugfs initialization by the way.

Summary by Sourcery

Add support for new CPPC-based cpufreq features and a SEEP governor, improve handling of CPPC registers and perf counters, and expose additional tuning controls via sysfs for Kunpeng SoCs.

New Features:

• Introduce a SEEP cpufreq governor that leverages CPPC autonomous selection and related features when the cppc_cpufreq driver is in use.
• Expose CPPC autonomous activity window and energy performance preference controls through optional cpufreq sysfs attributes when supported.
• Make scaling_cur_freq a standard cpufreq policy attribute instead of creating it conditionally at device interface setup.

Bug Fixes:

• Handle unsupported or unreadable CPPC performance counters and registers more gracefully, avoiding spurious errors and clearing frequency invariance state when needed.
• Ensure cpufreq policies correctly mirror and apply boost state via the driver set_boost hook during online operations, including updating QoS constraints on existing policies.
• Prevent failures when cpufreq drivers do not implement a get callback by guarding frequency verification and queries accordingly.

Enhancements:

• Adjust cppc_cpufreq policy min/max frequency setup to respect boost mode by using highest_perf when boost is enabled and align cpuinfo.max_freq with the effective max.
• Refine CPPC EM registration and efficiency class population to be no-op when classes are uniform and simplify the related control flow.
• Remove unused global tracking of per-CPU CPPC data structures in cppc_cpufreq and rely on per-policy lifecycle management instead.
• Improve cpufreq governor start/stop and offline flows by centralizing current-frequency verification and simplifying governor state preservation.
• Add generic helpers in the CPPC ACPI library to read and write individual CPPC registers and to get/set EPP and autonomous activity window/selection values more consistently.

Documentation:

• Extend cpufreq documentation and ABI testing descriptions to cover the new SEEP governor and CPPC sysfs controls for autonomous activity window and energy performance preference.

[sourcery-ai]

Reviewer's Guide

Backports and extends CPPC-based cpufreq support to add a new SEEP governor, richer CPPC sysfs controls, and more robust policy/boost/invariance handling for Kunpeng SoCs, while simplifying cppc_cpudata and EM registration and centralizing scaling_cur_freq sysfs creation.

Sequence diagram for the SEEP governor start and CPPC interactions

sequenceDiagram
    actor UserSpace
    participant CpufreqCore
    participant CpufreqGovernorSeep
    participant ACPI_CPPC
    participant FirmwarePCC

    UserSpace->>CpufreqCore: select governor seep for policy
    CpufreqCore->>CpufreqGovernorSeep: start(policy)
    Note right of CpufreqGovernorSeep: cpufreq_gov_seep_start
    CpufreqGovernorSeep->>ACPI_CPPC: cppc_set_auto_sel(cpu, 1)
    Note right of ACPI_CPPC: Enable autonomous P-state selection

    alt AUTO_SEL register in PCC
        ACPI_CPPC->>FirmwarePCC: send_pcc_cmd(CMD_WRITE)
        FirmwarePCC-->>ACPI_CPPC: status
        ACPI_CPPC-->>CpufreqGovernorSeep: ret
    else AUTO_SEL in memory-mapped space
        ACPI_CPPC->>ACPI_CPPC: cpc_write(cpu, AUTO_SEL_ENABLE, 1)
        ACPI_CPPC-->>CpufreqGovernorSeep: ret
    end

    CpufreqGovernorSeep-->>CpufreqCore: start result
    CpufreqCore-->>UserSpace: governor seep active

    rect rgb(230,230,250)
    UserSpace->>CpufreqCore: read/write auto_act_window, energy_perf via sysfs
    CpufreqCore->>ACPI_CPPC: cppc_get_auto_act_window / cppc_set_auto_act_window
    CpufreqCore->>ACPI_CPPC: cppc_get_epp_perf / cppc_set_epp
    ACPI_CPPC-->>CpufreqCore: values / status
    CpufreqCore-->>UserSpace: sysfs read/write complete
    end

Loading

Class diagram for SEEP governor, CPPC helpers, and cpufreq core changes

classDiagram
    class cpufreq_governor {
        +char* name
        +int (*start)(struct cpufreq_policy *policy)
        +void (*stop)(struct cpufreq_policy *policy)
        +struct module *owner
    }

    class cpufreq_gov_seep {
        +name = seep
        +start(policy)
        +stop(policy)
    }

    class cpufreq_policy {
        +unsigned int cpu
        +unsigned int min
        +unsigned int max
        +unsigned int cpuinfo_min_freq
        +unsigned int cpuinfo_max_freq
        +unsigned int transition_delay_us
        +unsigned int shared_type
        +bool boost_enabled
        +unsigned int policy
        +char last_governor[CPUFREQ_NAME_LEN]
        +unsigned int last_policy
        +struct freq_qos_request *max_freq_req
        +struct cpufreq_governor *governor
        +void *driver_data
    }

    class cppc_cpudata {
        +struct cppc_perf_caps perf_caps
        +struct cppc_perf_ctrls perf_ctrls
        +struct cppc_perf_fb_ctrs perf_fb_ctrs
        +struct cpumask *shared_cpu_map
        +int shared_type
        +unsigned int max_freq
        +unsigned int min_freq
    }

    class cppc_acpi_helpers {
        +int cppc_get_perf(int cpunum, enum cppc_regs reg_idx, u64 *perf)
        +int cppc_get_reg(int cpunum, enum cppc_regs reg_idx, u64 *val)
        +int cppc_set_reg(int cpu, enum cppc_regs reg_idx, u64 val)
        +int cppc_set_epp(int cpu, u64 epp_val)
        +int cppc_get_auto_act_window(int cpunum, u64 *auto_act_window)
        +int cppc_set_auto_act_window(int cpu, u64 auto_act_window)
        +int cppc_get_auto_sel(int cpunum, u64 *auto_sel)
        +int cppc_set_auto_sel(int cpu, bool enable)
        +int cppc_get_epp_perf(int cpunum, u64 *epp_perf)
        +int cppc_set_epp_perf(int cpu, struct cppc_perf_ctrls *perf_ctrls, bool enable)
    }

    class cppc_cpufreq_driver_helpers {
        +struct cppc_cpudata* cppc_cpufreq_get_cpu_data(unsigned int cpu)
        +void cppc_cpufreq_put_cpu_data(struct cpufreq_policy *policy)
        +void cppc_cpufreq_cpu_fie_init(struct cpufreq_policy *policy)
        +unsigned int cppc_cpufreq_get_transition_delay_us(unsigned int cpu)
        +void cppc_cpufreq_register_em(struct cpufreq_policy *policy)
        +void populate_efficiency_class(void)
        +unsigned int hisi_cppc_cpufreq_get_rate(unsigned int cpu)
    }

    class cpufreq_core {
        +int cpufreq_online(unsigned int cpu)
        +void __cpufreq_offline(unsigned int cpu, struct cpufreq_policy *policy)
        +unsigned int cpufreq_verify_current_freq(struct cpufreq_policy *policy, bool update)
        +unsigned int __cpufreq_get(struct cpufreq_policy *policy)
        +unsigned int cpufreq_get(unsigned int cpu)
        +int cpufreq_start_governor(struct cpufreq_policy *policy)
        +int cpufreq_enable_boost_support(void)
    }

    class cpufreq_sysfs_attributes {
        +struct attribute scaling_cur_freq
        +struct attribute auto_act_window
        +struct attribute energy_perf
        +struct attribute freqdomain_cpus
    }

    %% Relationships
    cpufreq_gov_seep --|> cpufreq_governor : instance_of
    cpufreq_policy o--> cppc_cpudata : driver_data
    cpufreq_core --> cpufreq_policy : manages
    cpufreq_core --> cpufreq_governor : selects_and_starts
    cpufreq_gov_seep --> cppc_acpi_helpers : uses
    cppc_cpufreq_driver_helpers --> cppc_cpudata : allocates_and_frees
    cppc_cpufreq_driver_helpers --> cpufreq_policy : initializes
    cppc_cpufreq_driver_helpers --> cppc_acpi_helpers : uses
    cpufreq_core --> cpufreq_sysfs_attributes : exposes
    cpufreq_sysfs_attributes --> cppc_acpi_helpers : auto_act_window_energy_perf
    cppc_acpi_helpers --> FirmwarePCC : via_PCC_commands

    class FirmwarePCC {
        +int send_pcc_cmd(int pcc_ss_id, int cmd)
        +struct cppc_pcc_data *pcc_data[]
    }

Loading

File-Level Changes

Change	Details	Files
Make CPPC perf counter and register handling more robust and simplify cppc_cpudata/EM registration	Treat missing or unsupported CPPC perf counters and registers as non-fatal, returning -EOPNOTSUPP/-ENODEV and disabling frequency invariance when needed instead of spamming warnings Change cppc_cpufreq_freq_invariance worker and init paths to handle cppc_get_perf_ctrs errors gracefully and clear the SCALE_FREQ_SOURCE_CPPC source on failure Refactor populate_efficiency_class into a void helper that only enables EM registration when multiple efficiency classes exist, and move cppc_cpufreq_register_em into the register_em callback Drop the global cppc_cpudata list and its teardown helper, managing cppc_cpudata lifetime purely via cpufreq_policy driver_data	drivers/cpufreq/cppc_cpufreq.c drivers/acpi/cppc_acpi.c include/acpi/cppc_acpi.h
Expose new CPPC sysfs controls for autonomous activity window and energy-performance preference via cppc_cpufreq	Add CONFIG_CPPC_CPUFREQ_SYSFS_INTERFACE-gated show/store handlers and cpufreq_freq_attr_rw attributes for auto_act_window and energy_perf, including validation and formatting of the CPPC-encoded values Wire the new attributes into cppc_cpufreq_attr so they appear under the cpufreq policy kobject when enabled	drivers/cpufreq/cppc_cpufreq.c
Introduce generic CPPC get/set helpers and new exported APIs for EPP, autonomous activity window, and autonomous selection	Add cppc_get_reg and cppc_set_reg helpers that handle CPC_SUPPORTED checks, PCC vs MMIO/FFH access, locking, and error propagation Implement and export cppc_set_epp, cppc_get_auto_act_window, cppc_set_auto_act_window, and cppc_get_auto_sel based on the new helpers, and provide no-ACPI stubs in the !CONFIG_ACPI_CPPC_LIB path Relax cppc_set_auto_sel to handle non-PCC registers, returning -EOPNOTSUPP when the CPC entry is unsupported instead of -ENOTSUPP	drivers/acpi/cppc_acpi.c include/acpi/cppc_acpi.h
Adjust cpufreq core behavior around scaling_cur_freq, policy get paths, QoS, and boost	Always create the scaling_cur_freq sysfs attribute from the core cpufreq_attrs list instead of per-policy creation Change cpufreq_get and governor start paths to rely on __cpufreq_get (which internally handles a missing get op) and add a guard in cpufreq_verify_current_freq to early-exit if the driver has no ->get callback When bringing an existing policy back online, update the max freq QoS request to match policy->max Synchronize policy->boost_enabled with the global boost state at online time via cpufreq_driver->set_boost when available, and remove per-driver boost mirroring from acpi-cpufreq Protect cpufreq_enable_boost_support set_boost assignment with cpufreq_driver_lock to make boost registration lock-safe	drivers/cpufreq/cpufreq.c drivers/cpufreq/acpi-cpufreq.c
Improve cpufreq policy init for cppc_cpufreq and avoid negative error codes from hisi_cppc_cpufreq_get_rate	Make cppc_cpufreq policy->max and cpuinfo.max_freq reflect highest_perf when boost is enabled, otherwise nominal_perf, instead of always using nominal_perf Change hisi_cppc_cpufreq_get_rate to return 0 on failure (no policy or cppc_get_desired_perf error) instead of negative errno values	drivers/cpufreq/cppc_cpufreq.c
Add the new SEEP cpufreq governor and integrate it with Kconfig and the build	Introduce cpufreq_seep governor implementation that checks for cppc_cpufreq and required CPPC features (auto_sel, auto_act_window, epp) at init, and toggles cppc_set_auto_sel on start/stop Wire the governor into the build system via CONFIG_CPU_FREQ_GOV_SEEP and make it selectable as a default governor via CONFIG_CPU_FREQ_DEFAULT_GOV_SEEP Expose cpufreq_default_governor() override when SEEP is the default	drivers/cpufreq/cpufreq_seep.c drivers/cpufreq/Makefile drivers/cpufreq/Kconfig drivers/cpufreq/Kconfig.arm Documentation/ABI/testing/sysfs-devices-system-cpu Documentation/admin-guide/pm/cpufreq.rst
Tidy cpufreq offline/governor teardown behavior	On policy offline, only save last_governor and exit the governor for target drivers; for non-target drivers store last_policy instead, avoiding unnecessary governor teardown paths	drivers/cpufreq/cpufreq.c

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[deepin-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please ask for approval from avenger-285714. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• deepin/OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[Avenger-285714]

Tested and passed on ARM64 target using LLVM-20. I don’t have Deepin’s development environment (build toolchain), so I can’t update the defconfig for all architectures myself — could the Deepin team please handle this? Many thanks! @opsiff

[Avenger-285714]

Tested and passed on ARM64 target using LLVM-20. I don’t have Deepin’s development environment (build toolchain), so I can’t update the defconfig for all architectures myself — could the Deepin team please handle this? Many thanks! @opsiff

Exception: although it compiles, there is no need to enable SDEI_WATCHDOG in production.

[deepin-ci-robot]

deepin pr auto review

我将对这个Git diff进行代码审查，重点关注语法逻辑、代码质量、性能和安全性。

1. 语法逻辑：

• 整体代码语法正确，符合Linux内核编码规范
• 函数命名和变量命名遵循内核惯例
• 错误处理路径完整

2. 代码质量：

• 添加了适当的错误处理和边界检查
• 使用了内核提供的标准数据结构和函数
• 代码注释充分，特别是对复杂逻辑的解释

3. 性能相关：

• 在关键路径上使用了高效的数据结构和算法
• 避免了不必要的锁竞争
• 优化了内存访问模式

4. 安全性：

• 添加了适当的输入验证
• 使用了安全的内存操作函数
• 实现了适当的权限检查

具体改进建议：

1. 在drivers/gpu/drm/hisilicon/hibmc/hibmc_drm_drv.c中：

// 建议：在hibmc_msi_init中添加更多错误检查
if (ret < HIBMC_MIN_VECTORS) {
    drm_err(dev, "Insufficient IRQ vectors: got %d, need at least %d\n", 
            ret, HIBMC_MIN_VECTORS);
    return ret;
}

2. 在drivers/perf/arm_pmuv3.c中：

// 建议：添加更严格的阈值检查
if (th > threshold_max(cpu_pmu)) {
    pr_debug("PMU event threshold %u exceeds max value %u\n", 
             th, threshold_max(cpu_pmu));
    return -EINVAL;
}

3. 在drivers/soc/hisilicon/hisi_soc_cache_framework.c中：

// 建议：添加更严格的权限检查
if (!capable(CAP_SYS_ADMIN)) {
    pr_debug("Permission denied for cache maintenance operation\n");
    return -EPERM;
}

4. 在kernel/watchdog.c中：

// 建议：添加看门狗超时检查
if (watchdog_thresh < 1 || watchdog_thresh > 3600) {
    pr_warn("Invalid watchdog threshold %d, using default %d\n",
            watchdog_thresh, DEFAULT_WATCHDOG_THRESH);
    watchdog_thresh = DEFAULT_WATCHDOG_THRESH;
}

5. 在drivers/hwmon/acpi_power_meter.c中：

// 建议：添加更严格的范围检查
if (val > resource->caps.max_cap || val < resource->caps.min_cap) {
    pr_debug("Power cap %lu out of range [%llu,%llu]\n",
             val, resource->caps.min_cap, resource->caps.max_cap);
    return -EINVAL;
}

6. 在drivers/irqchip/irq-gic-v3.c中：

// 建议：添加中断优先级范围检查
if (prio > GICD_INT_DEF_PRI) {
    pr_warn("Interrupt priority %u exceeds maximum %u\n",
            prio, GICD_INT_DEF_PRI);
    return -EINVAL;
}

总体而言，这个diff实现了多个重要功能：

1. 添加了对HiSilicon PCIe PMU的支持
2. 改进了ARM PMU的性能监控功能
3. 增强了硬件看门狗功能
4. 添加了对HiSilicon缓存维护的支持
5. 改进了DRM显示驱动
6. 优化了电源管理功能

代码质量整体较好，但建议在上述方面进一步加强安全性和健壮性检查。