Linux设备模型分析之kset --转

Linux设备模型分析之kset

上一篇博客我们分析了linux设备模型中kobject的注册和使用，在这一篇文章中，我们来看一下kset的用法。首先我们看一个使用kset的例子,代码如下:[cpp] view plaincopyprint?

#include <linux/device.h>   
• #include <linux/module.h>   
  #include <linux/kernel.h>   
• #include <linux/init.h>   
  #include <linux/string.h>   
• #include <linux/sysfs.h>   
  #include <linux/stat.h>   
•    
  MODULE_AUTHOR("haoyu");  
• MODULE_LICENSE("Dual BSD/GPL");  
     
• struct my_kobject  
  {  
•     int value;  
      struct kobject kobj;  
• };  
        
• struct my_kobject my_kobj;  
     
• void kobject_release(struct kobject *kobject);  
  ssize_t kobject_attr_show(struct kobject *kobject, struct attribute *attr,char *buf);  
• ssize_t kobject_attr_store(struct kobject *kobject,struct attribute *attr,const char *buf, size_t count);  
     
• struct attribute kobject_attr1 = {  
      .name = "name",  
•     .mode = S_IRWXUGO,  
  };  
•    
  struct attribute kobject_attr2 = {  
•     .name = "value",  
      .mode = S_IRWXUGO,  
• };  
     
• static struct attribute *kobject_def_attrs[] = {  
      &kobject_attr1,  
•     &kobject_attr2,  
      NULL,  
• };  
     
• struct sysfs_ops kobject_sysfs_ops =  
  {  
•     .show = kobject_attr_show,  
      .store = kobject_attr_store,  
• };  
     
• struct kobj_type ktype =  
  {  
•     .release = kobject_release,  
      .sysfs_ops = &kobject_sysfs_ops,  
•     .default_attrs = kobject_def_attrs,  
  };  
•    
  void kobject_release(struct kobject *kobject)  
• {  
      printk("kobject release.\n");  
• }  
     
• ssize_t kobject_attr_show(struct kobject *kobject, struct attribute *attr,char *buf)  
  {  
•     int count = 0;  
      struct my_kobject *my_kobj = container_of(kobject, struct my_kobject, kobj);  
•     printk("kobject attribute show.\n");  
      if(strcmp(attr->name, "name") == 0)  
•         count = sprintf(buf, "%s\n", kobject->name);  
      else if(strcmp(attr->name, "value") == 0)  
•         count = sprintf(buf, "%d\n", my_kobj->value);  
      else  
•         printk("no this attribute.\n");  
        
•     return count;  
  }  
•    
  ssize_t kobject_attr_store(struct kobject *kobject,struct attribute *attr,const char *buf, size_t count)  
• {  
      int val;  
•     struct my_kobject *my_kobj = container_of(kobject, struct my_kobject, kobj);  
      printk("kobject attribute store.\n");  
•     if(strcmp(attr->name, "name") == 0)  
          printk("Can not change name.\n");  
•     else if(strcmp(attr->name, "value") == 0)  
      {  
•         val = buf[0] - '0';  
          if(val == 0 || val == 1)  
•             my_kobj->value = val;  
          else  
•             printk("value is '0' or '1'\n");  
      }  
•     else  
          printk("no this attribute.\n");  
•          
      return count;  
• }  
     
• int kset_fiLTEr(struct kset *kset, struct kobject *kobj)  
  {  
•     printk("UEVENT: filter. kobj %s.\n",kobj->name);  
      return 1;  
• }  
     
• const char *kset_name(struct kset *kset, struct kobject *kobj)  
  {  
•     static char buf[20];  
      printk("UEVENT: name. kobj %s.\n",kobj->name);  
•     sprintf(buf,"%s","kset_test");  
      return buf;  
• }  
     
• int kset_uevent(struct kset *kset, struct kobject *kobj,  
                  struct kobj_uevent_env *env)  
• {  
      int i = 0;  
•     printk("UEVENT: uevent. kobj %s.\n",kobj->name);  
     
•     while( i< env->envp_idx){  
          printk("%s.\n",env->envp);  
•         i++;  
      }  
•    
      return 0;  
• }  
     
• struct kset_uevent_ops uevent_ops =  
  {  
•     .filter = kset_filter,  
      .name   = kset_name,  
•     .uevent = kset_uevent,  
  };  
•    
  struct kset *kset_parent;  
• struct kset kset_child;  
     
• static int kset_test_init(void)  
  {  
•     printk("kboject test init.\n");  
      kset_parent = kset_create_and_add("kset_parent", &uevent_ops, NULL);  
•     my_kobj.kobj.kset = kset_parent;  
      kobject_init_and_add(&my_kobj.kobj,&ktype,NULL,"kobject_test");  
•    
      kobject_set_name(&kset_child.kobj,"kset_child");  
•     kset_child.kobj.kset = kset_parent;  
      kset_register(&kset_child);  
•       
      return 0;  
• }  
     
• static void kset_test_exit(void)  
  {  
•     printk("kobject test exit.\n");  
      kobject_del(&my_kobj.kobj);  
•     kset_unregister(kset_parent);  
      kset_unregister(&kset_child);  
• }  
     
• module_init(kset_test_init);  
• module_exit(kset_test_exit);  
  

#include <linux/device.h>#include <linux/module.h>#include <linux/kernel.h>#include <linux/init.h>#include <linux/string.h>#include <linux/sysfs.h>#include <linux/stat.h> MODULE_AUTHOR("haoyu");MODULE_LICENSE("Dual BSD/GPL"); struct my_kobject{    int value;    struct kobject kobj;};    struct my_kobject my_kobj; void kobject_release(struct kobject *kobject);ssize_t kobject_attr_show(struct kobject *kobject, struct attribute *attr,char *buf);ssize_t kobject_attr_store(struct kobject *kobject,struct attribute *attr,const char *buf, size_t count); struct attribute kobject_attr1 = {    .name = "name",    .mode = S_IRWXUGO,}; struct attribute kobject_attr2 = {    .name = "value",    .mode = S_IRWXUGO,}; static struct attribute *kobject_def_attrs[] = {    &kobject_attr1,    &kobject_attr2,    NULL,}; struct sysfs_ops kobject_sysfs_ops ={    .show = kobject_attr_show,    .store = kobject_attr_store,}; struct kobj_type ktype ={    .release = kobject_release,    .sysfs_ops = &kobject_sysfs_ops,    .default_attrs = kobject_def_attrs,}; void kobject_release(struct kobject *kobject){    printk("kobject release.\n");} ssize_t kobject_attr_show(struct kobject *kobject, struct attribute *attr,char *buf){    int count = 0;    struct my_kobject *my_kobj = container_of(kobject, struct my_kobject, kobj);    printk("kobject attribute show.\n");    if(strcmp(attr->name, "name") == 0)        count = sprintf(buf, "%s\n", kobject->name);    else if(strcmp(attr->name, "value") == 0)        count = sprintf(buf, "%d\n", my_kobj->value);    else        printk("no this attribute.\n");        return count;} ssize_t kobject_attr_store(struct kobject *kobject,struct attribute *attr,const char *buf, size_t count){    int val;    struct my_kobject *my_kobj = container_of(kobject, struct my_kobject, kobj);    printk("kobject attribute store.\n");    if(strcmp(attr->name, "name") == 0)        printk("Can not change name.\n");    else if(strcmp(attr->name, "value") == 0)    {        val = buf[0] - '0';        if(val == 0 || val == 1)            my_kobj->value = val;        else            printk("value is '0' or '1'\n");    }    else        printk("no this attribute.\n");            return count;} int kset_filter(struct kset *kset, struct kobject *kobj){    printk("UEVENT: filter. kobj %s.\n",kobj->name);    return 1;} const char *kset_name(struct kset *kset, struct kobject *kobj){    static char buf[20];    printk("UEVENT: name. kobj %s.\n",kobj->name);    sprintf(buf,"%s","kset_test");    return buf;} int kset_uevent(struct kset *kset, struct kobject *kobj,                struct kobj_uevent_env *env){    int i = 0;    printk("UEVENT: uevent. kobj %s.\n",kobj->name);     while( i< env->envp_idx){        printk("%s.\n",env->envp);        i++;    }     return 0;} struct kset_uevent_ops uevent_ops ={    .filter = kset_filter,    .name   = kset_name,    .uevent = kset_uevent,}; struct kset *kset_parent;struct kset kset_child; static int kset_test_init(void){    printk("kboject test init.\n");    kset_parent = kset_create_and_add("kset_parent", &uevent_ops, NULL);    my_kobj.kobj.kset = kset_parent;    kobject_init_and_add(&my_kobj.kobj,&ktype,NULL,"kobject_test");     kobject_set_name(&kset_child.kobj,"kset_child");    kset_child.kobj.kset = kset_parent;    kset_register(&kset_child);        return 0;} static void kset_test_exit(void){    printk("kobject test exit.\n");    kobject_del(&my_kobj.kobj);    kset_unregister(kset_parent);    kset_unregister(&kset_child);} module_init(kset_test_init);module_exit(kset_test_exit);
该模块执行结果如下图所示:
kset的注册使用kset_create_and_add函数，该函数定义如下：[cpp] view plaincopyprint?

820/**
• 821 * kset_create_and_add - create a struct kset dynaMICally and add it to sysfs
  822 *
• 823 * @name: the name for the kset
  824 * @uevent_ops: a struct kset_uevent_ops for the kset
• 825 * @parent_kobj: the parent kobject of this kset, if any.
  826 *
• 827 * This function creates a kset structure dynamically and registers it
  828 * with sysfs.  When you are finished with this structure, call
• 829 * kset_unregister() and the structure will be dynamically freed when it
  830 * is no longer being used.
• 831 *
  832 * If the kset was not able to be created, NULL will be returned.
• 833 */  
  834struct kset *kset_create_and_add(const char *name,  
• 835                 const struct kset_uevent_ops *uevent_ops,  
  836                 struct kobject *parent_kobj)  
• 837{  
  838    struct kset *kset;  
• 839    int error;  
  840  
• 841    kset = kset_create(name, uevent_ops, parent_kobj);  
  842    if (！kset)  
• 843        return NULL;  
  844    error = kset_register(kset);  
• 845    if (error) {  
  846        kfree(kset);  
• 847        return NULL;  
  848    }  
• 849    return kset;  
• 850}  
  

820/**821 * kset_create_and_add - create a struct kset dynamically and add it to sysfs822 *823 * @name: the name for the kset824 * @uevent_ops: a struct kset_uevent_ops for the kset825 * @parent_kobj: the parent kobject of this kset, if any.826 *827 * This function creates a kset structure dynamically and registers it828 * with sysfs.  When you are finished with this structure, call829 * kset_unregister() and the structure will be dynamically freed when it830 * is no longer being used.831 *832 * If the kset was not able to be created, NULL will be returned.833 */834struct kset *kset_create_and_add(const char *name,835                 const struct kset_uevent_ops *uevent_ops,836                 struct kobject *parent_kobj)837{838    struct kset *kset;839    int error;840841    kset = kset_create(name, uevent_ops, parent_kobj);842    if (！kset)843        return NULL;844    error = kset_register(kset);845    if (error) {846        kfree(kset);847        return NULL;848    }849    return kset;850}
该函数首先调用kset_create创建kset，然后调用kset_register将kset注册到系统中。先来看kset_create函数是如何创建kset的，该函数定义如下：[cpp] view plaincopyprint?

776/**
• 777 * kset_create - create a struct kset dynamically
  778 *
• 779 * @name: the name for the kset
  780 * @uevent_ops: a struct kset_uevent_ops for the kset
• 781 * @parent_kobj: the parent kobject of this kset, if any.
  782 *
• 783 * This function creates a kset structure dynamically.  This structure can
  784 * then be registered with the system and show up in sysfs with a call to
• 785 * kset_register().  When you are finished with this structure, if
  786 * kset_register() has been called, call kset_unregister() and the
• 787 * structure will be dynamically freed when it is no longer being used.
  788 *
• 789 * If the kset was not able to be created, NULL will be returned.
  790 */  
• 791static struct kset *kset_create(const char *name,  
  792                const struct kset_uevent_ops *uevent_ops,  
• 793                struct kobject *parent_kobj)  
  794{  
• 795    struct kset *kset;  
  796    int retval;  
• 797  
  798    kset = kzalloc(sizeof(*kset), GFP_KERNEL);  
• 799    if (！kset)  
  800        return NULL;  
• 801    retval = kobject_set_name(&kset->kobj, name);  
  802    if (retval) {  
• 803        kfree(kset);  
  804        return NULL;  
• 805    }  
  806    kset->uevent_ops = uevent_ops;  
• 807    kset->kobj.parent = parent_kobj;  
  808  
• 809    /*
  810     * The kobject of this kset will have a type of kset_ktype and belong to
• 811     * no kset itself.  That way we can properly free it when it is
  812     * finished being used.
• 813     */  
  814    kset->kobj.ktype = &kset_ktype;  
• 815    kset->kobj.kset = NULL;  
  816  
• 817    return kset;  
• 818}  
  

776/**777 * kset_create - create a struct kset dynamically778 *779 * @name: the name for the kset780 * @uevent_ops: a struct kset_uevent_ops for the kset781 * @parent_kobj: the parent kobject of this kset, if any.782 *783 * This function creates a kset structure dynamically.  This structure can784 * then be registered with the system and show up in sysfs with a call to785 * kset_register().  When you are finished with this structure, if786 * kset_register() has been called, call kset_unregister() and the787 * structure will be dynamically freed when it is no longer being used.788 *789 * If the kset was not able to be created, NULL will be returned.790 */791static struct kset *kset_create(const char *name,792                const struct kset_uevent_ops *uevent_ops,793                struct kobject *parent_kobj)794{795    struct kset *kset;796    int retval;797798    kset = kzalloc(sizeof(*kset), GFP_KERNEL);799    if (！kset)800        return NULL;801    retval = kobject_set_name(&kset->kobj, name);802    if (retval) {803        kfree(kset);804        return NULL;805    }806    kset->uevent_ops = uevent_ops;807    kset->kobj.parent = parent_kobj;808809    /*810     * The kobject of this kset will have a type of kset_ktype and belong to811     * no kset itself.  That way we can properly free it when it is812     * finished being used.813     */814    kset->kobj.ktype = &kset_ktype;815    kset->kobj.kset = NULL;816817    return kset;818}
798行，为kset分配空间。801行，设置kset->kobj的名字，代表该kset的名字。806行，设置kset->uevent_ops。807行，设置kset->kobj.parent。814行，将kset->kobj.ktype设置为&kset_ktype。kset_ktype我们在下面介绍。815行，将kset->kobj.kset设置为NULL，即该kset不属于任何kset。kset_ktype的定义如下：[cpp] view plaincopyprint?

771static struct kobj_type kset_ktype = {  
• 772    .sysfs_ops  = &kobj_sysfs_ops,  
  773    .release = kset_release,  
• 774};  
  

771static struct kobj_type kset_ktype = {772    .sysfs_ops  = &kobj_sysfs_ops,773    .release = kset_release,774};
为了分析方便，我们先来看kset_release的定义：[cpp] view plaincopyprint?

763static void kset_release(struct kobject *kobj)  
• 764{  
  765    struct kset *kset = container_of(kobj, struct kset, kobj);  
• 766    pr_debug("kobject: '%s' (%p): %s\n",  
  767         kobject_name(kobj), kobj, __func__);  
• 768    kfree(kset);  
• 769}  
  

763static void kset_release(struct kobject *kobj)764{765    struct kset *kset = container_of(kobj, struct kset, kobj);766    pr_debug("kobject: '%s' (%p): %s\n",767         kobject_name(kobj), kobj, __func__);768    kfree(kset);769}
这个函数取得kset空间指针并释放内存空间。再来看kobj_sysfs_ops的定义，注意其作用，当用户读写kset->kobj的属性文件时，就会调用kset->kobj.ktype.sysfs_ops的show和store函数，即kobj_sysfs_ops的show和store函数：[cpp] view plaincopyprint?

703const struct sysfs_ops kobj_sysfs_ops = {  
• 704    .show   = kobj_attr_show,  
  705    .store  = kobj_attr_store,  
• 706};  
  

703const struct sysfs_ops kobj_sysfs_ops = {704    .show   = kobj_attr_show,705    .store  = kobj_attr_store,706};
kobj_attr_show函数定义如下：[cpp] view plaincopyprint?

678/* default kobject attribute operations */  
• 679static ssize_t kobj_attr_show(struct kobject *kobj, struct attribute *attr,  
  680                  char *buf)  
• 681{  
  682    struct kobj_attribute *kattr;  
• 683    ssize_t ret = -EIO;  
  684  
• 685    kattr = container_of(attr, struct kobj_attribute, attr);  
  686    if (kattr->show)  
• 687        ret = kattr->show(kobj, kattr, buf);  
  688    return ret;  
• 689}  
  

678/* default kobject attribute operations */679static ssize_t kobj_attr_show(struct kobject *kobj, struct attribute *attr,680                  char *buf)681{682    struct kobj_attribute *kattr;683    ssize_t ret = -EIO;684685    kattr = container_of(attr, struct kobj_attribute, attr);686    if (kattr->show)687        ret = kattr->show(kobj, kattr, buf);688    return ret;689}
该函数首先通过container_of宏取得包含指定attribute的kobj_attribute，然后调用kobj_attribute的show函数。kobj_attr_store函数定义如下：[cpp] view plaincopyprint?

691static ssize_t kobj_attr_store(struct kobject *kobj, struct attribute *attr,  
• 692                   const char *buf, size_t count)  
  693{  
• 694    struct kobj_attribute *kattr;  
  695    ssize_t ret = -EIO;  
• 696  
  697    kattr = container_of(attr, struct kobj_attribute, attr);  
• 698    if (kattr->store)  
  699        ret = kattr->store(kobj, kattr, buf, count);  
• 700    return ret;  
• 701}  
  

691static ssize_t kobj_attr_store(struct kobject *kobj, struct attribute *attr,692                   const char *buf, size_t count)693{694    struct kobj_attribute *kattr;695    ssize_t ret = -EIO;696697    kattr = container_of(attr, struct kobj_attribute, attr);698    if (kattr->store)699        ret = kattr->store(kobj, kattr, buf, count);700    return ret;701}
该函数首先通过container_of宏取得包含指定attribute的kobj_attribute，然后调用kobj_attribute的store函数。至此,kset_create函数我们就分析完了,回到kset_create_and_add函数中,下面我们要分析的是kset_register函数,该函数定义如下：[cpp] view plaincopyprint?

708/**
• 709 * kset_register - initialize and add a kset.
  710 * @k: kset.
• 711 */  
  712int kset_register(struct kset *k)  
• 713{  
  714    int err;  
• 715  
  716    if (！k)  
• 717        return -EINVAL;  
  718  
• 719    kset_init(k);  
  720    err = kobject_add_internal(&k->kobj);  
• 721    if (err)  
  722        return err;  
• 723    kobject_uevent(&k->kobj, KOBJ_ADD);  
  724    return 0;  
• 725}  
  

708/**709 * kset_register - initialize and add a kset.710 * @k: kset.711 */712int kset_register(struct kset *k)713{714    int err;715716    if (！k)717        return -EINVAL;718719    kset_init(k);720    err = kobject_add_internal(&k->kobj);721    if (err)722        return err;723    kobject_uevent(&k->kobj, KOBJ_ADD);724    return 0;725}
719行，调用kset_init初始化kset。720行，调用kobject_add_internal将kobject注册到系统中，在/sys下建立目录结构和属性文件。该函数我们在前一篇博客<<Linux设备模型之kobject>>已经分析过，这里不再详细分析。723行，调用kobject_uevent函数发送KOBJ_ADD事件。[cpp] view plaincopyprint?

667/**
• 668 * kset_init - initialize a kset for use
  669 * @k: kset
• 670 */  
  671void kset_init(struct kset *k)  
• 672{  
  673    kobject_init_internal(&k->kobj);  
• 674    INIT_LIST_HEAD(&k->list);  
  675    SPIn_lock_init(&k->list_lock);  
• 676}  
  

667/**668 * kset_init - initialize a kset for use669 * @k: kset670 */671void kset_init(struct kset *k)672{673    kobject_init_internal(&k->kobj);674    INIT_LIST_HEAD(&k->list);675    spin_lock_init(&k->list_lock);676}
673行，调用kobject_init_internal初始化kset->kobj。这个函数我们在上一篇博客<<Linux设备模型之kobject>>中已经分析过了。kobject_uevent函数定义如下：[cpp] view plaincopyprint?

319/**
• 320 * kobject_uevent - notify userspace by sending an uevent
  321 *
• 322 * @action: action that is happening
  323 * @kobj: struct kobject that the action is happening to
• 324 *
  325 * Returns 0 if kobject_uevent() is completed with success or the
• 326 * corresponding error when it fails.
  327 */  
• 328int kobject_uevent(struct kobject *kobj, enum kobject_action action)  
  329{  
• 330    return kobject_uevent_env(kobj, action, NULL);  
• 331}  
  

319/**320 * kobject_uevent - notify userspace by sending an uevent321 *322 * @action: action that is happening323 * @kobj: struct kobject that the action is happening to324 *325 * Returns 0 if kobject_uevent() is completed with success or the326 * corresponding error when it fails.327 */328int kobject_uevent(struct kobject *kobj, enum kobject_action action)329{330    return kobject_uevent_env(kobj, action, NULL);331}
这个函数直接调用了kobject_uevent_env函数，定义如下：[cpp] view plaincopyprint?

119/**
• 120 * kobject_uevent_env - send an uevent with environmental data
  121 *
• 122 * @action: action that is happening
  123 * @kobj: struct kobject that the action is happening to
• 124 * @envp_ext: pointer to environmental data
  125 *
• 126 * Returns 0 if kobject_uevent_env() is completed with success or the
  127 * corresponding error when it fails.
• 128 */  
  129int kobject_uevent_env(struct kobject *kobj, enum kobject_action action,  
• 130               char *envp_ext[])  
  131{  
• 132    struct kobj_uevent_env *env;  
  133    const char *action_string = kobject_actions[action];  
• 134    const char *devpath = NULL;  
  135    const char *subsystem;  
• 136    struct kobject *top_kobj;  
  137    struct kset *kset;  
• 138    const struct kset_uevent_ops *uevent_ops;  
  139    u64 seq;  
• 140    int i = 0;  
  141    int retval = 0;  
• 142#ifdef CONFIG_NET  
  143    struct uevent_sock *ue_sk;  
• 144#endif  
  145  
• 146    pr_debug("kobject: '%s' (%p): %s\n",  
  147         kobject_name(kobj), kobj, __func__);  
• 148  
  149    /* search the kset we belong to */  
• 150    top_kobj = kobj;  
  151    while (！top_kobj->kset && top_kobj->parent)  
• 152        top_kobj = top_kobj->parent;  
  153  
• 154    if (！top_kobj->kset) {  
  155        pr_debug("kobject: '%s' (%p): %s: attempted to send uevent "  
• 156             "without kset！\n", kobject_name(kobj), kobj,  
  157             __func__);  
• 158        return -EINVAL;  
  159    }  
• 160  
  161    kset = top_kobj->kset;  
• 162    uevent_ops = kset->uevent_ops;  
  163  
• 164    /* skip the event, if uevent_suppress is set*/  
  165    if (kobj->uevent_suppress) {  
• 166        pr_debug("kobject: '%s' (%p): %s: uevent_suppress "  
  167                 "caused the event to drop！\n",  
• 168                 kobject_name(kobj), kobj, __func__);  
  169        return 0;  
• 170    }  
  171    /* skip the event, if the filter returns zero. */  
• 172    if (uevent_ops && uevent_ops->filter)  
  173        if (！uevent_ops->filter(kset, kobj)) {  
• 174            pr_debug("kobject: '%s' (%p): %s: filter function "  
  175                 "caused the event to drop！\n",  
• 176                 kobject_name(kobj), kobj, __func__);  
  177            return 0;  
• 178        }  
  179  
• 180    /* originating subsystem */  
  181    if (uevent_ops && uevent_ops->name)  
• 182        subsystem = uevent_ops->name(kset, kobj);  
  183    else  
• 184        subsystem = kobject_name(&kset->kobj);  
  185    if (！subsystem) {  
• 186        pr_debug("kobject: '%s' (%p): %s: unset subsystem caused the "  
  187             "event to drop！\n", kobject_name(kobj), kobj,  
• 188             __func__);  
  189        return 0;  
• 190    }  
  191  
• 192    /* environment buffer */  
  193    env = kzalloc(sizeof(struct kobj_uevent_env), GFP_KERNEL);  
• 194    if (！env)  
  195        return -ENOMEM;  
• 196  
  197    /* complete object path */  
• 198    devpath = kobject_get_path(kobj, GFP_KERNEL);  
  199    if (！devpath) {  
• 200        retval = -ENOENT;  
  201        goto exit;  
• 202    }  
  203  
• 204    /* default keys */  
  205    retval = add_uevent_var(env, "ACTION=%s", action_string);  
• 206    if (retval)  
  207        goto exit;  
• 208    retval = add_uevent_var(env, "DEVPATH=%s", devpath);  
  209    if (retval)  
• 210        goto exit;  
  211    retval = add_uevent_var(env, "SUBSYSTEM=%s", subsystem);  
• 212    if (retval)  
  213        goto exit;  
• 214  
  215    /* keys passed in fROM the caller */  
• 216    if (envp_ext) {  
  217        for (i = 0; envp_ext; i++) {  
• 218            retval = add_uevent_var(env, "%s", envp_ext);  
  219            if (retval)  
• 220                goto exit;  
  221        }  
• 222    }  
  223  
• 224    /* let the kset specific function add its stuff */  
  225    if (uevent_ops && uevent_ops->uevent) {  
• 226        retval = uevent_ops->uevent(kset, kobj, env);  
  227        if (retval) {  
• 228            pr_debug("kobject: '%s' (%p): %s: uevent() returned "  
  229                 "%d\n", kobject_name(kobj), kobj,  
• 230                 __func__, retval);  
  231            goto exit;  
• 232        }  
  233    }  
• 234  
  235    /*
• 236     * Mark "add" and "remove" events in the object to ensure proper
  237     * events to userspace during automatic cleanup. If the object did
• 238     * send an "add" event, "remove" will automatically generated by
  239     * the core, if not already done by the caller.
• 240     */  
  241    if (action == KOBJ_ADD)  
• 242        kobj->state_add_uevent_sent = 1;  
  243    else if (action == KOBJ_REMOVE)  
• 244        kobj->state_remove_uevent_sent = 1;  
  245  
• 246    /* we will send an event, so request a new sequence number */  
  247    spin_lock(&sequence_lock);  
• 248    seq = ++uevent_seqnum;  
  249    spin_unlock(&sequence_lock);  
• 250    retval = add_uevent_var(env, "SEQNUM=%llu", (unsigned long long)seq);  
  251    if (retval)  
• 252        goto exit;  
  253  
• 254#if defined(CONFIG_NET)  
  255    /* send netlink message */  
• 256    mutex_lock(&uevent_sock_mutex);  
  257    list_for_each_entry(ue_sk, &uevent_sock_list, list) {  
• 258        struct sock *uevent_sock = ue_sk->sk;  
  259        struct sk_buff *skb;  
• 260        size_t len;  
  261  
• 262        /* allocate message with the maximum possible size */  
  263        len = strlen(action_string) + strlen(devpath) + 2;  
• 264        skb = alloc_skb(len + env->buflen, GFP_KERNEL);  
  265        if (skb) {  
• 266            char *scratch;  
  267  
• 268            /* add header */  
  269            scratch = skb_put(skb, len);  
• 270            sprintf(scratch, "%s@%s", action_string, devpath);  
  271  
• 272            /* copy keys to our continuous event payload buffer */  
  273            for (i = 0; i < env->envp_idx; i++) {  
• 274                len = strlen(env->envp) + 1;  
  275                scratch = skb_put(skb, len);  
• 276                strcpy(scratch, env->envp);  
  277            }  
• 278  
  279            NETLINK_CB(skb).dst_group = 1;  
• 280            retval = netlink_broADCast_filtered(uevent_sock, skb,  
  281                                0, 1, GFP_KERNEL,  
• 282                                kobj_bcast_filter,  
  283                                kobj);  
• 284            /* ENOBUFS should be handled in userspace */  
  285            if (retval == -ENOBUFS)  
• 286                retval = 0;  
  287        } else  
• 288            retval = -ENOMEM;  
  289    }  
• 290    mutex_unlock(&uevent_sock_mutex);  
  291#endif  
• 292  
  293    /* call uevent_helper, usually only enabled during early boot */  
• 294    if (uevent_helper[0] && ！kobj_usermode_filter(kobj)) {  
  295        char *argv [3];  
• 296  
  297        argv [0] = uevent_helper;  
• 298        argv [1] = (char *)subsystem;  
  299        argv [2] = NULL;  
• 300        retval = add_uevent_var(env, "HOME=/");  
  301        if (retval)  
• 302            goto exit;  
  303        retval = add_uevent_var(env,  
• 304                    "PATH=/sbin:/bin:/usr/sbin:/usr/bin");  
  305        if (retval)  
• 306            goto exit;  
  307  
• 308        retval = call_usermodehelper(argv[0], argv,  
  309                         env->envp, UMH_WAIT_EXEC);  
• 310    }  
  311  
• 312exit:  
  313    kfree(devpath);  
• 314    kfree(env);  
  315    return retval;  
• 316}  
  

119/**120 * kobject_uevent_env - send an uevent with environmental data121 *122 * @action: action that is happening123 * @kobj: struct kobject that the action is happening to124 * @envp_ext: pointer to environmental data125 *126 * Returns 0 if kobject_uevent_env() is completed with success or the127 * corresponding error when it fails.128 */129int kobject_uevent_env(struct kobject *kobj, enum kobject_action action,130               char *envp_ext[])131{132    struct kobj_uevent_env *env;133    const char *action_string = kobject_actions[action];134    const char *devpath = NULL;135    const char *subsystem;136    struct kobject *top_kobj;137    struct kset *kset;138    const struct kset_uevent_ops *uevent_ops;139    u64 seq;140    int i = 0;141    int retval = 0;142#ifdef CONFIG_NET143    struct uevent_sock *ue_sk;144#endif145146    pr_debug("kobject: '%s' (%p): %s\n",147         kobject_name(kobj), kobj, __func__);148149    /* search the kset we belong to */150    top_kobj = kobj;151    while (！top_kobj->kset && top_kobj->parent)152        top_kobj = top_kobj->parent;153154    if (！top_kobj->kset) {155        pr_debug("kobject: '%s' (%p): %s: attempted to send uevent "156             "without kset！\n", kobject_name(kobj), kobj,157             __func__);158        return -EINVAL;159    }160161    kset = top_kobj->kset;162    uevent_ops = kset->uevent_ops;163164    /* skip the event, if uevent_suppress is set*/165    if (kobj->uevent_suppress) {166        pr_debug("kobject: '%s' (%p): %s: uevent_suppress "167                 "caused the event to drop！\n",168                 kobject_name(kobj), kobj, __func__);169        return 0;170    }171    /* skip the event, if the filter returns zero. */172    if (uevent_ops && uevent_ops->filter)173        if (！uevent_ops->filter(kset, kobj)) {174            pr_debug("kobject: '%s' (%p): %s: filter function "175                 "caused the event to drop！\n",176                 kobject_name(kobj), kobj, __func__);177            return 0;178        }179180    /* originating subsystem */181    if (uevent_ops && uevent_ops->name)182        subsystem = uevent_ops->name(kset, kobj);183    else184        subsystem = kobject_name(&kset->kobj);185    if (！subsystem) {186        pr_debug("kobject: '%s' (%p): %s: unset subsystem caused the "187             "event to drop！\n", kobject_name(kobj), kobj,188             __func__);189        return 0;190    }191192    /* environment buffer */193    env = kzalloc(sizeof(struct kobj_uevent_env), GFP_KERNEL);194    if (！env)195        return -ENOMEM;196197    /* complete object path */198    devpath = kobject_get_path(kobj, GFP_KERNEL);199    if (！devpath) {200        retval = -ENOENT;201        goto exit;202    }203204    /* default keys */205    retval = add_uevent_var(env, "ACTION=%s", action_string);206    if (retval)207        goto exit;208    retval = add_uevent_var(env, "DEVPATH=%s", devpath);209    if (retval)210        goto exit;211    retval = add_uevent_var(env, "SUBSYSTEM=%s", subsystem);212    if (retval)213        goto exit;214215    /* keys passed in from the caller */216    if (envp_ext) {217        for (i = 0; envp_ext; i++) {218            retval = add_uevent_var(env, "%s", envp_ext);219            if (retval)220                goto exit;221        }222    }223224    /* let the kset specific function add its stuff */225    if (uevent_ops && uevent_ops->uevent) {226        retval = uevent_ops->uevent(kset, kobj, env);227        if (retval) {228            pr_debug("kobject: '%s' (%p): %s: uevent() returned "229                 "%d\n", kobject_name(kobj), kobj,230                 __func__, retval);231            goto exit;232        }233    }234235    /*236     * Mark "add" and "remove" events in the object to ensure proper237     * events to userspace during automatic cleanup. If the object did238     * send an "add" event, "remove" will automatically generated by239     * the core, if not already done by the caller.240     */241    if (action == KOBJ_ADD)242        kobj->state_add_uevent_sent = 1;243    else if (action == KOBJ_REMOVE)244        kobj->state_remove_uevent_sent = 1;245246    /* we will send an event, so request a new sequence number */247    spin_lock(&sequence_lock);248    seq = ++uevent_seqnum;249    spin_unlock(&sequence_lock);250    retval = add_uevent_var(env, "SEQNUM=%llu", (unsigned long long)seq);251    if (retval)252        goto exit;253254#if defined(CONFIG_NET)255    /* send netlink message */256    mutex_lock(&uevent_sock_mutex);257    list_for_each_entry(ue_sk, &uevent_sock_list, list) {258        struct sock *uevent_sock = ue_sk->sk;259        struct sk_buff *skb;260        size_t len;261262        /* allocate message with the maximum possible size */263        len = strlen(action_string) + strlen(devpath) + 2;264        skb = alloc_skb(len + env->buflen, GFP_KERNEL);265        if (skb) {266            char *scratch;267268            /* add header */269            scratch = skb_put(skb, len);270            sprintf(scratch, "%s@%s", action_string, devpath);271272            /* copy keys to our continuous event payload buffer */273            for (i = 0; i < env->envp_idx; i++) {274                len = strlen(env->envp) + 1;275                scratch = skb_put(skb, len);276                strcpy(scratch, env->envp);277            }278279            NETLINK_CB(skb).dst_group = 1;280            retval = netlink_broadcast_filtered(uevent_sock, skb,281                                0, 1, GFP_KERNEL,282                                kobj_bcast_filter,283                                kobj);284            /* ENOBUFS should be handled in userspace */285            if (retval == -ENOBUFS)286                retval = 0;287        } else288            retval = -ENOMEM;289    }290    mutex_unlock(&uevent_sock_mutex);291#endif292293    /* call uevent_helper, usually only enabled during early boot */294    if (uevent_helper[0] && ！kobj_usermode_filter(kobj)) {295        char *argv [3];296297        argv [0] = uevent_helper;298        argv [1] = (char *)subsystem;299        argv [2] = NULL;300        retval = add_uevent_var(env, "HOME=/");301        if (retval)302            goto exit;303        retval = add_uevent_var(env,304                    "PATH=/sbin:/bin:/usr/sbin:/usr/bin");305        if (retval)306            goto exit;307308        retval = call_usermodehelper(argv[0], argv,309                         env->envp, UMH_WAIT_EXEC);310    }311312exit:313    kfree(devpath);314    kfree(env);315    return retval;316}
133由参数action取得代表事件类型的字符串。kobject_actions定义如下：[cpp] view plaincopyprint?

40/*
• 41 * The actions here must match the index to the string array
  42 * in lib/kobject_uevent.c
• 43 *
  44 * Do not add new actions here without checking with the driver-core
• 45 * maintainers. Action strings are not meant to express subsystem
  46 * or device specific properties. In most cases you want to send a
• 47 * kobject_uevent_env(kobj, KOBJ_CHANGE, env) with additional event
  48 * specific variables added to the event environment.
• 49 */  
  50enum kobject_action {  
• 51    KOBJ_ADD,  
  52    KOBJ_REMOVE,  
• 53    KOBJ_CHANGE,  
  54    KOBJ_MOVE,  
• 55    KOBJ_ONLINE,  
  56    KOBJ_OFFLINE,  
• 57    KOBJ_MAX  
  58};  
• 42/* the strings here must match the enum in include/linux/kobject.h */  
  43static const char *kobject_actions[] = {  
• 44    [KOBJ_ADD] =        "add",  
  45    [KOBJ_REMOVE] =     "remove",  
• 46    [KOBJ_CHANGE] =     "change",  
  47    [KOBJ_MOVE] =       "move",  
• 48    [KOBJ_ONLINE] =     "online",  
  49    [KOBJ_OFFLINE] =    "offline",  
• 50};  
  

40/*41 * The actions here must match the index to the string array42 * in lib/kobject_uevent.c43 *44 * Do not add new actions here without checking with the driver-core45 * maintainers. Action strings are not meant to express subsystem46 * or device specific properties. In most cases you want to send a47 * kobject_uevent_env(kobj, KOBJ_CHANGE, env) with additional event48 * specific variables added to the event environment.49 */50enum kobject_action {51    KOBJ_ADD,52    KOBJ_REMOVE,53    KOBJ_CHANGE,54    KOBJ_MOVE,55    KOBJ_ONLINE,56    KOBJ_OFFLINE,57    KOBJ_MAX58};42/* the strings here must match the enum in include/linux/kobject.h */43static const char *kobject_actions[] = {44    [KOBJ_ADD] =        "add",45    [KOBJ_REMOVE] =     "remove",46    [KOBJ_CHANGE] =     "change",47    [KOBJ_MOVE] =       "move",48    [KOBJ_ONLINE] =     "online",49    [KOBJ_OFFLINE] =    "offline",50};
149 - 162行，因为只有kset才包含处理uevent的函数，即kset.uevent_ops，所以这里查找kobj所属的kset，如果kobj->kset不存在，就沿着kobj->parent向上查找，直到找到一个不为空的kset为止。如果确实没有找到kset，则退出。165 - 170行，如果kobj->uevent_suppress的值为1，表示禁止发出uevent事件，退出。172 - 178行，如果uevent_ops不为空，并且实现了uevent_ops->filter，则执行uevent_ops->filter，如果uevent_ops->filter返回值为0，表示这个uevent被过滤了，退出。181 - 190行，如果uevent_ops不为空，并且实现了uevent_ops->name，则通过uevent_ops->name获得子系统名，如果uevent_ops为空，或者没有实现uevent_ops->name，则以kset->kobj的名字作为子系统名。193行，创建kobj_uevent_env结构体变量env，用来保存环境变量。198行，通过调用kobject_get_path函数取得kobj在/sys系统中的路径，保存在devpath变量中。205 - 213行，通过调用add_uevent_var函数将事件名，kobj路径和子系统名加入到环境变量env中。216 - 222行，如果有通过参数传递的环境变量，也调用add_uevent_var函数加入到环境变量env中。225 - 233行，如果uevent_ops不为空，并且实现了uevent_ops->uevent函数，则调用uevent_ops->uevent。241 - 244行，如果要发送的事件是KOBJ_ADD或KOBJ_REMOVE，则相应将kobj->state_add_uevent_sent或kobj->state_remove_uevent_sent置为1。247 - 252行，将发送事件序列数加1，添加到环境变量env中。254 - 291行，条件编译部分我们不分析。294行，先来看uevent_helper的定义：char uevent_helper[UEVENT_HELPER_PATH_LEN] = CONFIG_UEVENT_HELPER_PATH;#define UEVENT_HELPER_PATH_LEN      256CONFIG_UEVENT_HELPER_PATH="/sbin/hotplug"所以uevent_helper就是"/sbin/hotplug"。300 - 306行，将HOME和PATH加入环境变量env中。308 - 309行，执行用户空间的/sbin/hotplug程序，传递环境变量为env。至此, kobject_uevent_env -> kobject_uevent -> kset_register -> kset_create_and_add函数就分析完了，kset被创建和注册到系统中。