linux PMBus总线及设备驱动分析

PMBus协议规范介绍

PMBus是一套对电源进行配置、控制和监控的通讯协议标准。其最新版本为1.3，该规范还在不断演进中，比如新标准中新增的zone PMBus、AVSBus等特性。在其官网上有详细的规范文档，本节不尝试翻译规范文档，重点记录作者在了解PMBus过程中的疑问和解答。

PMBus与I2C、SMBus的区别？

PMBus在SMBus（System Management Bus）基础上增加了一套电源配置、控制和监控规范。SMBus最初是为电池智能管理而开发的一套标准，其基于I2C协议，并针对I2C协议的弱健壮性做了如下改进：

• 支持SMBALERT#中断；
• 支持错包检测（PEC）；
• 支持包超时；
• 支持START/STOP保护；
• 支持Host Notify Protocol协议；

PMBus监控哪些参数？告警分为几级？不同告警级别有什么样的应对措施？

PMBus支持电压、电流、功率、温度和风扇等参数的上下限监控，支持warning和fault 2级告警级别（如上图所示）。

• warning告警：表示监控参数异常，系统需引起关注，但可以继续运行，系统无需任何响应措施；
• fault告警：比warning告警级别高，系统会根据异常对设备的危害情况，进行设备控制电路重启(restart)或输出切断(shutdown)等处理；

告警产生时如何上报给主机？

告警上报一般有如下几种方式：

• 主机轮询PMBus设备；
• PMBus设备通过SMBALERT#中断通知主机；
• Host Notify Protocol(PMBus设备临时切换成总线主机(bus master),并发送一组特定协议通知系统主机)。

什么情况下告警会取消或清除？重启是否会清除告警？

任何warning或fault告警一旦上报，只有通过如下几种方式可以取消清除：

• PMBus设备接收到CLEAR_FAULTS命令；
• PMBus设备RESET引脚生效；
• PMBus设备通过CONTROL引脚或OPERATION命令关闭并重新打开；
• 断电；
• 如果异常一直存在，那么即使进行告警清除操作，告警会马上重新上报。

linux PMBus驱动设计分析

PMBus设备驱动位于linux/drivers/hwmon/pmbus，文件组织划分为3个部分：

• pmbus_core.c：PMBus通用设备驱动，总线规范实现，sysfs外部接口呈现；
• pmbus.c：标准PMBus设备注册；
• [device].c：非标准PMBus设备注册。

比较有意思的是PMBus的通用设备驱动框架设计部分，其设计方案主要要解决如下2个问题：

1. 支持PMBus设备厂商的自定义功能集。PMBus规范定义一套功能集，其中有些是基本功能，有些是可选功能；
2. 支持PMBus设备厂商的自定义寄存器。

pmbus驱动框架的数据模型如下，其核心对象为i2c_client，即i2c设备对象，i2c_client继承于linux设备驱动模型device对象。pmbus设备信息通过设备驱动模型抽象接口driver_data访问，由pmbus_data对象实现。pmbus_data对象又关联如下2个主要对象：

• pmbus_driver_info：PMBus设备支持的功能集描述及相关接口，由pmbus设备实现。
• pmbus_sensor：PMBus设备支持的监控传感器对象链表，由voltage/current/power/temp/fan实现。

pmbus设备功能集识别有2种实现方式：

• 由pmbus_driver_info对象的identify接口完成，其工作原理是通过读取功能寄存器，如果读取成功，则说明设备支持此功能，否则不支持；
• 直接静态初始化pmbus_driver_info。

设备自定义寄存器通过虚拟寄存器（Virtual registers）统一到pmbus驱动框架中。pmbus通用设备驱动只看到标准寄存器和虚拟寄存器。虚拟寄存器到设备自定义寄存器的映射过程通过设备注册的4个接口：read_byte_data/read_word_data/write_word_data/write_byte来完成。　

应用示例

1. 编写pmbus设备的smbus总线设备驱动并注册。如下i2c-10[1-3]为epld实现的4个i2c总线设备

/ # ls /sys/bus/i2c/devices/ -l
total 0
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 i2c-100 -> ../../../devices/i2c-100
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 i2c-101 -> ../../../devices/i2c-101
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 i2c-102 -> ../../../devices/i2c-102
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 i2c-103 -> ../../../devices/i2c-103
 
/ # cats '/sys/bus/i2c/devices/i2c-10*/name'
/sys/bus/i2c/devices/i2c-100/name: WORK_EPLD1
/sys/bus/i2c/devices/i2c-101/name: WORK_EPLD2.0
/sys/bus/i2c/devices/i2c-102/name: WORK_EPLD2.1
/sys/bus/i2c/devices/i2c-103/name: WORK_NSE
/ #

　　

2. 编写pmbus设备驱动并注册。如下为注册方法，hwmon[0-9]为注册的pmbus设备

echo tps53667 0x60 > /sys/bus/i2c/devices/i2c-100/new_device
echo tps53667 0x62 > /sys/bus/i2c/devices/i2c-100/new_device
echo tps53667 0x1060 > /sys/bus/i2c/devices/i2c-100/new_device
echo tps53667 0x1062 > /sys/bus/i2c/devices/i2c-100/new_device
 
echo tps53667 0x60 > /sys/bus/i2c/devices/i2c-101/new_device
echo tps53667 0x60 > /sys/bus/i2c/devices/i2c-102/new_device
 
echo tps53667 0x70 > /sys/bus/i2c/devices/i2c-103/new_device
echo tps53667 0x1071 > /sys/bus/i2c/devices/i2c-103/new_device
echo tps53667 0x2072 > /sys/bus/i2c/devices/i2c-103/new_device
echo tps53667 0x3073 > /sys/bus/i2c/devices/i2c-103/new_device
 
/ # ls /sys/class/hwmon/ -l
total 0
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon0 -> ../../devices/i2c-100/100-0060/hwmon/hwmon0
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon1 -> ../../devices/i2c-100/100-0062/hwmon/hwmon1
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon2 -> ../../devices/i2c-100/100-1060/hwmon/hwmon2
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon3 -> ../../devices/i2c-100/100-1062/hwmon/hwmon3
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon4 -> ../../devices/i2c-101/101-0060/hwmon/hwmon4
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon5 -> ../../devices/i2c-102/102-0060/hwmon/hwmon5
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon6 -> ../../devices/i2c-103/103-0070/hwmon/hwmon6
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon7 -> ../../devices/i2c-103/103-1071/hwmon/hwmon7
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon8 -> ../../devices/i2c-103/103-2072/hwmon/hwmon8
lrwxrwxrwx    1 root     root             0 Jan  1 00:00 hwmon9 -> ../../devices/i2c-103/103-3073/hwmon/hwmon9

　　

3. 查看pmbus设备监控数据，所有字段解释详见kernel文档。

/sys/class/hwmon # cats 'hwmon2/device/*'
# 电流监控数据  单位：mA 毫安  /1000
hwmon2/device/curr1_crit:       255000
hwmon2/device/curr1_crit_alarm: 0
hwmon2/device/curr1_input:      2679
hwmon2/device/curr1_label:      iin
hwmon2/device/curr1_max:        25000
hwmon2/device/curr1_max_alarm:  0
hwmon2/device/curr2_crit:       122000
hwmon2/device/curr2_crit_alarm: 0
hwmon2/device/curr2_input:      22968
hwmon2/device/curr2_label:      iout1
hwmon2/device/curr2_max:        98000
hwmon2/device/curr2_max_alarm:  0
hwmon2/device/driver:   cat: read error: Is a directory
hwmon2/device/hwmon:    cat: read error: Is a directory
# 电压监控数据  单位：mV 毫伏 1/1000
hwmon2/device/in1_crit: 17000
hwmon2/device/in1_crit_alarm:   0
hwmon2/device/in1_input:        11906
hwmon2/device/in1_label:        vin
hwmon2/device/in2_alarm:        0
hwmon2/device/in2_input:        631
hwmon2/device/in2_label:        vout1
hwmon2/device/modalias: i2c:tps53667
hwmon2/device/name:     tps53667
# 功率监控数据  单位：uW 微伏 1/1000000
hwmon2/device/power1_input:     31625000
hwmon2/device/power1_label:     pin
hwmon2/device/power2_input:     23343750
hwmon2/device/power2_label:     pout1
hwmon2/device/subsystem:        cat: read error: Is a directory
# 温度监控数据  单位：m℃ 毫摄氏度 1/1000
hwmon2/device/temp1_crit:       125000
hwmon2/device/temp1_crit_alarm: 0
hwmon2/device/temp1_input:      35500
hwmon2/device/temp1_max:        95000
hwmon2/device/temp1_max_alarm:  0
hwmon2/device/uevent:   DRIVER=tps53667
MODALIAS=i2c:tps53667

　　

4. 构造一个UV fault alarm，如下。可见异常恢复后，告警依然保持，不会清除；重启也不是清除告警；手动清除后，告警清除。

/sys/devices/i2c-103/103-3073 # cats 'in*'
in1_crit:       17000
in1_crit_alarm: 0
in1_input:      11968
in1_label:      vin
in2_alarm:      0
in2_input:      0
in2_label:      vout1
/sys/devices/i2c-103/103-3073 # echo 10000 > in1_crit　　# 将in voltage的UV fault阈值设成10V
/sys/devices/i2c-103/103-3073 # cats 'in*'
in1_crit:       10000
in1_crit_alarm: 1　　# 告警触发
in1_input:      11953
in1_label:      vin
in2_alarm:      0
in2_input:      0
in2_label:      vout1
/sys/devices/i2c-103/103-3073 # echo 17000 > in1_crit　　# 将in voltage的UV fault阈值恢复成17V
/sys/devices/i2c-103/103-3073 # cats 'in*'
in1_crit:       17000
in1_crit_alarm: 1　　# 异常后再恢复正常，告警依然保持
in1_input:      11968
in1_label:      vin
in2_alarm:      0
in2_input:      0
in2_label:      vout1
/sys/devices/i2c-103/103-3073 # reboot
。。。（启动过程省略）
/sys/devices/i2c-103/103-3073 # cats 'in*'
in1_crit:       17000
in1_crit_alarm: 1　　# 重启不会清除告警
in1_input:      11968
in1_label:      vin
in2_alarm:      0
in2_input:      0
in2_label:      vout1
/sys/devices/i2c-103/103-3073 # echo 0 > clear_fault 　　# 手动清除告警
/sys/devices/i2c-103/103-3073 # cats 'in*'
in1_crit:       17000
in1_crit_alarm: 0　　# 告警清除
in1_input:      11968
in1_label:      vin
in2_alarm:      0
in2_input:      631
in2_label:      vout1

　　

附主要数据结构：

struct pmbus_data {
    struct device *dev;
    struct device *hwmon_dev;
 
    u32 flags;        /* from platform data */
 
    int exponent;        /* linear mode: exponent for output voltages */
 
    const struct pmbus_driver_info *info;
 
    int max_attributes;
    int num_attributes;
    struct attribute_group group;
 
    struct pmbus_sensor *sensors;
 
    struct mutex update_lock;
    bool valid;
    unsigned long last_updated;    /* in jiffies */
 
    /*
     * A single status register covers multiple attributes,
     * so we keep them all together.
     */
    u8 status[PB_NUM_STATUS_REG];
    u8 status_register;
 
    u8 currpage;
};
 
struct pmbus_driver_info {
    int pages;        /* Total number of pages */
    enum pmbus_data_format format[PSC_NUM_CLASSES];
    /*
     * Support one set of coefficients for each sensor type
     * Used for chips providing data in direct mode.
     */
    int m[PSC_NUM_CLASSES];    /* mantissa for direct data format */
    int b[PSC_NUM_CLASSES];    /* offset */
    int R[PSC_NUM_CLASSES];    /* exponent */
 
    u32 func[PMBUS_PAGES];    /* Functionality, per page */
    /*
     * The following functions map manufacturing specific register values
     * to PMBus standard register values. Specify only if mapping is
     * necessary.
     * Functions return the register value (read) or zero (write) if
     * successful. A return value of -ENODATA indicates that there is no
     * manufacturer specific register, but that a standard PMBus register
     * may exist. Any other negative return value indicates that the
     * register does not exist, and that no attempt should be made to read
     * the standard register.
     */
    int (*read_byte_data)(struct i2c_client *client, int page, int reg);
    int (*read_word_data)(struct i2c_client *client, int page, int reg);
    int (*write_word_data)(struct i2c_client *client, int page, int reg,
                   u16 word);
    int (*write_byte)(struct i2c_client *client, int page, u8 value);
    /*
     * The identify function determines supported PMBus functionality.
     * This function is only necessary if a chip driver supports multiple
     * chips, and the chip functionality is not pre-determined.
     */
    int (*identify)(struct i2c_client *client,
            struct pmbus_driver_info *info);
};
 
struct pmbus_sensor {
    struct pmbus_sensor *next;
    char name[PMBUS_NAME_SIZE];    /* sysfs sensor name */
    struct device_attribute attribute;
    u8 page;        /* page number */
    u16 reg;        /* register */
    enum pmbus_sensor_classes class;    /* sensor class */
    bool update;        /* runtime sensor update needed */
    int data;        /* Sensor data.
                   Negative if there was a read error */
};
 
/*
 * Virtual registers.
 * Useful to support attributes which are not supported by standard PMBus
 * registers but exist as manufacturer specific registers on individual chips.
 * Must be mapped to real registers in device specific code.
 *
 * Semantics:
 * Virtual registers are all word size.
 * READ registers are read-only; writes are either ignored or return an error.
 * RESET registers are read/write. Reading reset registers returns zero
 * (used for detection), writing any value causes the associated history to be
 * reset.
 * Virtual registers have to be handled in device specific driver code. Chip
 * driver code returns non-negative register values if a virtual register is
 * supported, or a negative error code if not. The chip driver may return
 * -ENODATA or any other error code in this case, though an error code other
 * than -ENODATA is handled more efficiently and thus preferred. Either case,
 * the calling PMBus core code will abort if the chip driver returns an error
 * code when reading or writing virtual registers.
 */
#define PMBUS_VIRT_BASE            0x100
#define PMBUS_VIRT_READ_TEMP_AVG    (PMBUS_VIRT_BASE + 0)
#define PMBUS_VIRT_READ_TEMP_MIN    (PMBUS_VIRT_BASE + 1)
#define PMBUS_VIRT_READ_TEMP_MAX    (PMBUS_VIRT_BASE + 2)
#define PMBUS_VIRT_RESET_TEMP_HISTORY    (PMBUS_VIRT_BASE + 3)
#define PMBUS_VIRT_READ_VIN_AVG        (PMBUS_VIRT_BASE + 4)
#define PMBUS_VIRT_READ_VIN_MIN        (PMBUS_VIRT_BASE + 5)
#define PMBUS_VIRT_READ_VIN_MAX        (PMBUS_VIRT_BASE + 6)
#define PMBUS_VIRT_RESET_VIN_HISTORY    (PMBUS_VIRT_BASE + 7)
#define PMBUS_VIRT_READ_IIN_AVG        (PMBUS_VIRT_BASE + 8)
#define PMBUS_VIRT_READ_IIN_MIN        (PMBUS_VIRT_BASE + 9)
#define PMBUS_VIRT_READ_IIN_MAX        (PMBUS_VIRT_BASE + 10)
#define PMBUS_VIRT_RESET_IIN_HISTORY    (PMBUS_VIRT_BASE + 11)
#define PMBUS_VIRT_READ_PIN_AVG        (PMBUS_VIRT_BASE + 12)
#define PMBUS_VIRT_READ_PIN_MAX        (PMBUS_VIRT_BASE + 13)
#define PMBUS_VIRT_RESET_PIN_HISTORY    (PMBUS_VIRT_BASE + 14)
#define PMBUS_VIRT_READ_POUT_AVG    (PMBUS_VIRT_BASE + 15)
#define PMBUS_VIRT_READ_POUT_MAX    (PMBUS_VIRT_BASE + 16)
#define PMBUS_VIRT_RESET_POUT_HISTORY    (PMBUS_VIRT_BASE + 17)
#define PMBUS_VIRT_READ_VOUT_AVG    (PMBUS_VIRT_BASE + 18)
#define PMBUS_VIRT_READ_VOUT_MIN    (PMBUS_VIRT_BASE + 19)
#define PMBUS_VIRT_READ_VOUT_MAX    (PMBUS_VIRT_BASE + 20)
#define PMBUS_VIRT_RESET_VOUT_HISTORY    (PMBUS_VIRT_BASE + 21)
#define PMBUS_VIRT_READ_IOUT_AVG    (PMBUS_VIRT_BASE + 22)
#define PMBUS_VIRT_READ_IOUT_MIN    (PMBUS_VIRT_BASE + 23)
#define PMBUS_VIRT_READ_IOUT_MAX    (PMBUS_VIRT_BASE + 24)
#define PMBUS_VIRT_RESET_IOUT_HISTORY    (PMBUS_VIRT_BASE + 25)
#define PMBUS_VIRT_READ_TEMP2_AVG    (PMBUS_VIRT_BASE + 26)
#define PMBUS_VIRT_READ_TEMP2_MIN    (PMBUS_VIRT_BASE + 27)
#define PMBUS_VIRT_READ_TEMP2_MAX    (PMBUS_VIRT_BASE + 28)
#define PMBUS_VIRT_RESET_TEMP2_HISTORY    (PMBUS_VIRT_BASE + 29)
 
#define PMBUS_VIRT_READ_VMON        (PMBUS_VIRT_BASE + 30)
#define PMBUS_VIRT_VMON_UV_WARN_LIMIT    (PMBUS_VIRT_BASE + 31)
#define PMBUS_VIRT_VMON_OV_WARN_LIMIT    (PMBUS_VIRT_BASE + 32)
#define PMBUS_VIRT_VMON_UV_FAULT_LIMIT    (PMBUS_VIRT_BASE + 33)
#define PMBUS_VIRT_VMON_OV_FAULT_LIMIT    (PMBUS_VIRT_BASE + 34)
#define PMBUS_VIRT_STATUS_VMON        (PMBUS_VIRT_BASE + 35)