]> err.no Git - linux-2.6/commitdiff
[PATCH] merge kobject_uevent and kobject_hotplug
authorKay Sievers <kay.sievers@suse.de>
Fri, 11 Nov 2005 13:43:07 +0000 (14:43 +0100)
committerGreg Kroah-Hartman <gregkh@suse.de>
Thu, 5 Jan 2006 00:18:07 +0000 (16:18 -0800)
The distinction between hotplug and uevent does not make sense these
days, netlink events are the default.

udev depends entirely on netlink uevents. Only during early boot and
in initramfs, /sbin/hotplug is needed. So merge the two functions and
provide only one interface without all the options.

The netlink layer got a nice generic interface with named slots
recently, which is probably a better facility to plug events for
subsystem specific events.
Also the new poll() interface to /proc/mounts is a nicer way to
notify about changes than sending events through the core.
The uevents should only be used for driver core related requests to
userspace now.

Signed-off-by: Kay Sievers <kay.sievers@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
drivers/scsi/ipr.c
include/linux/kobject.h
lib/kobject_uevent.c

index fa2cb3582cfa5a2c1348d974bcfa3566e5e47d73..bf44a409ba0dd72c287a3280957daf493f707f04 100644 (file)
@@ -2132,7 +2132,7 @@ restart:
        }
 
        spin_unlock_irqrestore(ioa_cfg->host->host_lock, lock_flags);
-       kobject_uevent(&ioa_cfg->host->shost_classdev.kobj, KOBJ_CHANGE, NULL);
+       kobject_hotplug(&ioa_cfg->host->shost_classdev.kobj, KOBJ_CHANGE);
        LEAVE;
 }
 
index e6926b327538dbe9f7868c52c7a4429c36b26fdc..5b08248fba72e70998a793ef10dc3209a984f195 100644 (file)
@@ -39,11 +39,11 @@ extern u64 hotplug_seqnum;
 /* the actions here must match the proper string in lib/kobject_uevent.c */
 typedef int __bitwise kobject_action_t;
 enum kobject_action {
-       KOBJ_ADD        = (__force kobject_action_t) 0x01,      /* add event, for hotplug */
-       KOBJ_REMOVE     = (__force kobject_action_t) 0x02,      /* remove event, for hotplug */
-       KOBJ_CHANGE     = (__force kobject_action_t) 0x03,      /* a sysfs attribute file has changed */
-       KOBJ_OFFLINE    = (__force kobject_action_t) 0x04,      /* offline event for hotplug devices */
-       KOBJ_ONLINE     = (__force kobject_action_t) 0x05,      /* online event for hotplug devices */
+       KOBJ_ADD        = (__force kobject_action_t) 0x01,      /* exclusive to core */
+       KOBJ_REMOVE     = (__force kobject_action_t) 0x02,      /* exclusive to core */
+       KOBJ_CHANGE     = (__force kobject_action_t) 0x03,      /* device state change */
+       KOBJ_OFFLINE    = (__force kobject_action_t) 0x04,      /* device offline */
+       KOBJ_ONLINE     = (__force kobject_action_t) 0x05,      /* device online */
 };
 
 struct kobject {
@@ -262,28 +262,13 @@ int add_hotplug_env_var(char **envp, int num_envp, int *cur_index,
                        char *buffer, int buffer_size, int *cur_len,
                        const char *format, ...)
        __attribute__((format (printf, 7, 8)));
-
-int kobject_uevent(struct kobject *kobj,
-                  enum kobject_action action,
-                  struct attribute *attr);
-int kobject_uevent_atomic(struct kobject *kobj,
-                         enum kobject_action action,
-                         struct attribute *attr);
-
 #else
 static inline void kobject_hotplug(struct kobject *kobj, enum kobject_action action) { }
+
 static inline int add_hotplug_env_var(char **envp, int num_envp, int *cur_index, 
                                      char *buffer, int buffer_size, int *cur_len, 
                                      const char *format, ...)
 { return 0; }
-int kobject_uevent(struct kobject *kobj,
-                  enum kobject_action action,
-                  struct attribute *attr)
-{ return 0; }
-int kobject_uevent_atomic(struct kobject *kobj,
-                         enum kobject_action action,
-                         struct attribute *attr)
-{ return 0; }
 #endif
 
 #endif /* __KERNEL__ */
index 845bf67d94ca1d0fe4244d5746e8ca6c657b12a8..dd061da3aba90da1cbdad7e2989145b0a7751e28 100644 (file)
@@ -29,6 +29,7 @@
 char hotplug_path[HOTPLUG_PATH_LEN] = "/sbin/hotplug";
 u64 hotplug_seqnum;
 static DEFINE_SPINLOCK(sequence_lock);
+static struct sock *uevent_sock;
 
 static char *action_to_string(enum kobject_action action)
 {
@@ -48,123 +49,6 @@ static char *action_to_string(enum kobject_action action)
        }
 }
 
-static struct sock *uevent_sock;
-
-/**
- * send_uevent - notify userspace by sending event through netlink socket
- *
- * @signal: signal name
- * @obj: object path (kobject)
- * @envp: possible hotplug environment to pass with the message
- * @gfp_mask:
- */
-static int send_uevent(const char *signal, const char *obj,
-                      char **envp, gfp_t gfp_mask)
-{
-       struct sk_buff *skb;
-       char *pos;
-       int len;
-
-       if (!uevent_sock)
-               return -EIO;
-
-       len = strlen(signal) + 1;
-       len += strlen(obj) + 1;
-
-       /* allocate buffer with the maximum possible message size */
-       skb = alloc_skb(len + BUFFER_SIZE, gfp_mask);
-       if (!skb)
-               return -ENOMEM;
-
-       pos = skb_put(skb, len);
-       sprintf(pos, "%s@%s", signal, obj);
-
-       /* copy the environment key by key to our continuous buffer */
-       if (envp) {
-               int i;
-
-               for (i = 2; envp[i]; i++) {
-                       len = strlen(envp[i]) + 1;
-                       pos = skb_put(skb, len);
-                       strcpy(pos, envp[i]);
-               }
-       }
-
-       NETLINK_CB(skb).dst_group = 1;
-       return netlink_broadcast(uevent_sock, skb, 0, 1, gfp_mask);
-}
-
-static int do_kobject_uevent(struct kobject *kobj, enum kobject_action action, 
-                            struct attribute *attr, gfp_t gfp_mask)
-{
-       char *path;
-       char *attrpath;
-       char *signal;
-       int len;
-       int rc = -ENOMEM;
-
-       path = kobject_get_path(kobj, gfp_mask);
-       if (!path)
-               return -ENOMEM;
-
-       signal = action_to_string(action);
-       if (!signal)
-               return -EINVAL;
-
-       if (attr) {
-               len = strlen(path);
-               len += strlen(attr->name) + 2;
-               attrpath = kmalloc(len, gfp_mask);
-               if (!attrpath)
-                       goto exit;
-               sprintf(attrpath, "%s/%s", path, attr->name);
-               rc = send_uevent(signal, attrpath, NULL, gfp_mask);
-               kfree(attrpath);
-       } else
-               rc = send_uevent(signal, path, NULL, gfp_mask);
-
-exit:
-       kfree(path);
-       return rc;
-}
-
-/**
- * kobject_uevent - notify userspace by sending event through netlink socket
- * 
- * @signal: signal name
- * @kobj: struct kobject that the event is happening to
- * @attr: optional struct attribute the event belongs to
- */
-int kobject_uevent(struct kobject *kobj, enum kobject_action action,
-                  struct attribute *attr)
-{
-       return do_kobject_uevent(kobj, action, attr, GFP_KERNEL);
-}
-EXPORT_SYMBOL_GPL(kobject_uevent);
-
-int kobject_uevent_atomic(struct kobject *kobj, enum kobject_action action,
-                         struct attribute *attr)
-{
-       return do_kobject_uevent(kobj, action, attr, GFP_ATOMIC);
-}
-EXPORT_SYMBOL_GPL(kobject_uevent_atomic);
-
-static int __init kobject_uevent_init(void)
-{
-       uevent_sock = netlink_kernel_create(NETLINK_KOBJECT_UEVENT, 1, NULL,
-                                           THIS_MODULE);
-
-       if (!uevent_sock) {
-               printk(KERN_ERR
-                      "kobject_uevent: unable to create netlink socket!\n");
-               return -ENODEV;
-       }
-
-       return 0;
-}
-
-postcore_initcall(kobject_uevent_init);
-
 /**
  * kobject_hotplug - notify userspace by executing /sbin/hotplug
  *
@@ -173,95 +57,84 @@ postcore_initcall(kobject_uevent_init);
  */
 void kobject_hotplug(struct kobject *kobj, enum kobject_action action)
 {
-       char *argv [3];
-       char **envp = NULL;
-       char *buffer = NULL;
-       char *seq_buff;
+       char **envp;
+       char *buffer;
        char *scratch;
+       const char *action_string;
+       const char *devpath = NULL;
+       const char *subsystem;
+       struct kobject *top_kobj;
+       struct kset *kset;
+       struct kset_hotplug_ops *hotplug_ops;
+       u64 seq;
+       char *seq_buff;
        int i = 0;
        int retval;
-       char *kobj_path = NULL;
-       const char *name = NULL;
-       char *action_string;
-       u64 seq;
-       struct kobject *top_kobj = kobj;
-       struct kset *kset;
-       static struct kset_hotplug_ops null_hotplug_ops;
-       struct kset_hotplug_ops *hotplug_ops = &null_hotplug_ops;
 
-       /* If this kobj does not belong to a kset,
-          try to find a parent that does. */
+       pr_debug("%s\n", __FUNCTION__);
+
+       action_string = action_to_string(action);
+       if (!action_string)
+               return;
+
+       /* search the kset we belong to */
+       top_kobj = kobj;
        if (!top_kobj->kset && top_kobj->parent) {
                do {
                        top_kobj = top_kobj->parent;
                } while (!top_kobj->kset && top_kobj->parent);
        }
-
-       if (top_kobj->kset)
-               kset = top_kobj->kset;
-       else
+       if (!top_kobj->kset)
                return;
 
-       if (kset->hotplug_ops)
-               hotplug_ops = kset->hotplug_ops;
+       kset = top_kobj->kset;
+       hotplug_ops = kset->hotplug_ops;
 
-       /* If the kset has a filter operation, call it.
-          Skip the event, if the filter returns zero. */
-       if (hotplug_ops->filter) {
+       /*  skip the event, if the filter returns zero. */
+       if (hotplug_ops && hotplug_ops->filter)
                if (!hotplug_ops->filter(kset, kobj))
                        return;
-       }
-
-       pr_debug ("%s\n", __FUNCTION__);
-
-       action_string = action_to_string(action);
-       if (!action_string)
-               return;
 
-       envp = kmalloc(NUM_ENVP * sizeof (char *), GFP_KERNEL);
+       /* environment index */
+       envp = kzalloc(NUM_ENVP * sizeof (char *), GFP_KERNEL);
        if (!envp)
                return;
-       memset (envp, 0x00, NUM_ENVP * sizeof (char *));
 
+       /* environment values */
        buffer = kmalloc(BUFFER_SIZE, GFP_KERNEL);
        if (!buffer)
                goto exit;
 
-       if (hotplug_ops->name)
-               name = hotplug_ops->name(kset, kobj);
-       if (name == NULL)
-               name = kobject_name(&kset->kobj);
+       /* complete object path */
+       devpath = kobject_get_path(kobj, GFP_KERNEL);
+       if (!devpath)
+               goto exit;
 
-       argv [0] = hotplug_path;
-       argv [1] = (char *)name; /* won't be changed but 'const' has to go */
-       argv [2] = NULL;
+       /* originating subsystem */
+       if (hotplug_ops && hotplug_ops->name)
+               subsystem = hotplug_ops->name(kset, kobj);
+       else
+               subsystem = kobject_name(&kset->kobj);
 
-       /* minimal command environment */
-       envp [i++] = "HOME=/";
-       envp [i++] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
+       /* event environemnt for helper process only */
+       envp[i++] = "HOME=/";
+       envp[i++] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
 
+       /* default keys */
        scratch = buffer;
-
        envp [i++] = scratch;
        scratch += sprintf(scratch, "ACTION=%s", action_string) + 1;
-
-       kobj_path = kobject_get_path(kobj, GFP_KERNEL);
-       if (!kobj_path)
-               goto exit;
-
        envp [i++] = scratch;
-       scratch += sprintf (scratch, "DEVPATH=%s", kobj_path) + 1;
-
+       scratch += sprintf (scratch, "DEVPATH=%s", devpath) + 1;
        envp [i++] = scratch;
-       scratch += sprintf(scratch, "SUBSYSTEM=%s", name) + 1;
+       scratch += sprintf(scratch, "SUBSYSTEM=%s", subsystem) + 1;
 
-       /* reserve space for the sequence,
-        * put the real one in after the hotplug call */
+       /* just reserve the space, overwrite it after kset call has returned */
        envp[i++] = seq_buff = scratch;
        scratch += strlen("SEQNUM=18446744073709551616") + 1;
 
-       if (hotplug_ops->hotplug) {
-               /* have the kset specific function add its stuff */
+       /* let the kset specific function add its stuff */
+       if (hotplug_ops && hotplug_ops->hotplug) {
                retval = hotplug_ops->hotplug (kset, kobj,
                                  &envp[i], NUM_ENVP - i, scratch,
                                  BUFFER_SIZE - (scratch - buffer));
@@ -272,27 +145,49 @@ void kobject_hotplug(struct kobject *kobj, enum kobject_action action)
                }
        }
 
+       /* we will send an event, request a new sequence number */
        spin_lock(&sequence_lock);
        seq = ++hotplug_seqnum;
        spin_unlock(&sequence_lock);
        sprintf(seq_buff, "SEQNUM=%llu", (unsigned long long)seq);
 
-       pr_debug ("%s: %s %s seq=%llu %s %s %s %s %s\n",
-                 __FUNCTION__, argv[0], argv[1], (unsigned long long)seq,
-                 envp[0], envp[1], envp[2], envp[3], envp[4]);
-
-       send_uevent(action_string, kobj_path, envp, GFP_KERNEL);
+       /* send netlink message */
+       if (uevent_sock) {
+               struct sk_buff *skb;
+               size_t len;
+
+               /* allocate message with the maximum possible size */
+               len = strlen(action_string) + strlen(devpath) + 2;
+               skb = alloc_skb(len + BUFFER_SIZE, GFP_KERNEL);
+               if (skb) {
+                       /* add header */
+                       scratch = skb_put(skb, len);
+                       sprintf(scratch, "%s@%s", action_string, devpath);
+
+                       /* copy keys to our continuous event payload buffer */
+                       for (i = 2; envp[i]; i++) {
+                               len = strlen(envp[i]) + 1;
+                               scratch = skb_put(skb, len);
+                               strcpy(scratch, envp[i]);
+                       }
+
+                       NETLINK_CB(skb).dst_group = 1;
+                       netlink_broadcast(uevent_sock, skb, 0, 1, GFP_KERNEL);
+               }
+       }
 
-       if (!hotplug_path[0])
-               goto exit;
+       /* call uevent_helper, usually only enabled during early boot */
+       if (hotplug_path[0]) {
+               char *argv [3];
 
-       retval = call_usermodehelper (argv[0], argv, envp, 0);
-       if (retval)
-               pr_debug ("%s - call_usermodehelper returned %d\n",
-                         __FUNCTION__, retval);
+               argv [0] = hotplug_path;
+               argv [1] = (char *)subsystem;
+               argv [2] = NULL;
+               call_usermodehelper (argv[0], argv, envp, 0);
+       }
 
 exit:
-       kfree(kobj_path);
+       kfree(devpath);
        kfree(buffer);
        kfree(envp);
        return;
@@ -350,4 +245,20 @@ int add_hotplug_env_var(char **envp, int num_envp, int *cur_index,
 }
 EXPORT_SYMBOL(add_hotplug_env_var);
 
+static int __init kobject_uevent_init(void)
+{
+       uevent_sock = netlink_kernel_create(NETLINK_KOBJECT_UEVENT, 1, NULL,
+                                           THIS_MODULE);
+
+       if (!uevent_sock) {
+               printk(KERN_ERR
+                      "kobject_uevent: unable to create netlink socket!\n");
+               return -ENODEV;
+       }
+
+       return 0;
+}
+
+postcore_initcall(kobject_uevent_init);
+
 #endif /* CONFIG_HOTPLUG */