Skip to contentSkip to conversation

Add DEVPATH env variable to hotplug helper call

PATCH RESEND In review Josef 'Jeff' Sipek 3 participants 12 messages Oct 3, 2004 – Oct 5, 2004

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 Add DEVPATH env variable to hotplug helper call

Add $DEVPATH to the environmental variables during /sbin/hotplug call.

Josef 'Jeff' Sipek.

kernel/cpu.c +4 -3
diff -Nru a/kernel/cpu.c b/kernel/cpu.c
@@ -61,13 +61,13 @@
  * cpu' with certain environment variables set.  */
 static int cpu_run_sbin_hotplug(unsigned int cpu, const char *action)
 {
-	char *argv[3], *envp[5], cpu_str[12], action_str[32];
+	char *argv[3], *envp[6], cpu_str[12], action_str[32], devpath_str[38];
 	int i;
 
 	sprintf(cpu_str, "CPU=%d", cpu);
 	sprintf(action_str, "ACTION=%s", action);
-	/* FIXME: Add DEVPATH. --RR */
-
+	sprintf(devpath_str, "DEVPATH=devices/system/cpu/cpu%d", cpu);
+	
 	i = 0;
 	argv[i++] = hotplug_path;
 	argv[i++] = "cpu";
@@ -79,6 +79,7 @@
 	envp[i++] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
 	envp[i++] = cpu_str;
 	envp[i++] = action_str;
+	envp[i++] = devpath_str;
 	envp[i] = NULL;
 
 	return call_usermodehelper(argv[0], argv, envp, 0);
#1 lore raw
Quoted 8 lines · Josef 'Jeff' Sipek

Add $DEVPATH to the environmental variables during /sbin/hotplug call.

Josef 'Jeff' Sipek.

Quoted 6 lines

diff -Nru a/kernel/cpu.c b/kernel/cpu.c
--- a/kernel/cpu.c 2004-09-24 13:08:57 -04:00
+++ b/kernel/cpu.c 2004-09-24 13:08:57 -04:00
@@ -61,13 +61,13 @@
* cpu' with certain environment variables set. */
static int cpu_run_sbin_hotplug(unsigned int cpu, const char *action)

I don't think this function should exist at all. We already have kobjects which represent the CPUs so it should be possible to find that kobject and use kobject_hotplug() on it. That gives you your $DEVPATH.

Does CPU hotplug behave correctly wrt /sys/devices/system/cpu? Given that register_cpu() is still marked __init, I assume not.

#2 lore raw
Quoted 2 lines · Andrew Morton

Does CPU hotplug behave correctly wrt /sys/devices/system/cpu? Given that register_cpu() is still marked __init, I assume not.

Currently what we have in the kernel is logical cpu hotplug, i.e once the cpu is registered via register_cpu() that cpu can only go offline and still the entry for that cpu will be present in the /sys/devices/system/cpu/cpuX/online.

So __init register_cpu() is fine untill we support unregister_cpu() which is required for physical cpu hotplug case.

I have submitted ACPI based physical cpu hotplug patches and waiting to here from ACPI mainitainer Len Brown, there I have taken care to support unregister_cpu() and register_cpu() is marked as __devinit in those patches.

Quoted 5 lines

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#3 lore raw
Quoted 9 lines · Keshavamurthy Anil S
Quoted 2 lines · Andrew Morton

Does CPU hotplug behave correctly wrt /sys/devices/system/cpu? Given that register_cpu() is still marked __init, I assume not.

Currently what we have in the kernel is logical cpu hotplug, i.e once the cpu is registered via register_cpu() that cpu can only go offline and still the entry for that cpu will be present in the /sys/devices/system/cpu/cpuX/online.

Quoted 6 lines

So __init register_cpu() is fine untill we support unregister_cpu() which is required for physical cpu hotplug case.

I have submitted ACPI based physical cpu hotplug patches and waiting to here from ACPI mainitainer Len Brown, there I have taken care to support unregister_cpu() and register_cpu() is marked as __devinit in those patches.

OK...

But still, cpu_run_sbin_hotplug() should not exist. It is duplicating (indeed, emulating) kobject_hotplug() behaviour. To the extent that it now has a hardwired sysfs path embedded in it:

	sprintf(devpath_str, "DEVPATH=devices/system/cpu/cpu%d", cpu);

which should have been obtained from kobject_get_path().

#4 lore raw
Quoted 19 lines · Andrew Morton
Quoted 9 lines · Keshavamurthy Anil S
Quoted 2 lines · Andrew Morton

Does CPU hotplug behave correctly wrt /sys/devices/system/cpu? Given that register_cpu() is still marked __init, I assume not.

Currently what we have in the kernel is logical cpu hotplug, i.e once the cpu is registered via register_cpu() that cpu can only go offline and still the entry for that cpu will be present in the /sys/devices/system/cpu/cpuX/online.

Quoted 6 lines

So __init register_cpu() is fine untill we support unregister_cpu() which is required for physical cpu hotplug case.

I have submitted ACPI based physical cpu hotplug patches and waiting to here from ACPI mainitainer Len Brown, there I have taken care to support unregister_cpu() and register_cpu() is marked as __devinit in those patches.

OK...

Quoted 7 lines

But still, cpu_run_sbin_hotplug() should not exist. It is duplicating (indeed, emulating) kobject_hotplug() behaviour. To the extent that it now has a hardwired sysfs path embedded in it:

	sprintf(devpath_str, "DEVPATH=devices/system/cpu/cpu%d", cpu);

which should have been obtained from kobject_get_path().

Yes, I agree to your point that cpu_run_sbin_hotplug() is duplication kobject_hotplug() behaviour. I will send you a patch to fix this ASAP(hopefully before end of today).

thanks,
Anil

#5 lore raw
Quoted 8 lines · Keshavamurthy Anil S
Quoted 7 lines · Andrew Morton
Quoted 1 line · Keshavamurthy Anil S

But still, cpu_run_sbin_hotplug() should not exist. It is duplicating (indeed, emulating) kobject_hotplug() behaviour. To the extent that it now has a hardwired sysfs path embedded in it:

Quoted 5 lines
Quoted 3 lines
	sprintf(devpath_str, "DEVPATH=devices/system/cpu/cpu%d", cpu);

which should have been obtained from kobject_get_path().

Yes, I agree to your point that cpu_run_sbin_hotplug() is duplication kobject_hotplug() behaviour. I will send you a patch to fix this ASAP(hopefully before end of today).

Hi Andrew,
Here is what I have come up with(please take a look at this patch).
I was successfully able to get rid of cpu_run_sbin_hotplug() function, but when I call kobject_hotplug() function, it is finding top_kobj->kset->hotplug_ops set to NULL and hence returns without calling call_usermodehelper(). Not sure if this is a bug in kobject_hotplug(), I feel kobject_hotplug() function should continue even if top_kobj->kset-hotplug_ops is NULL.

CC'ing grek-KH on this mailing list.

Attached is the patch to remove cpu_run_sbin_hotplug.

---

 linux-2.6.9-rc3-test-askeshav/drivers/base/cpu.c             |    2 
 linux-2.6.9-rc3-test-askeshav/include/linux/kobject_uevent.h |    1 
 linux-2.6.9-rc3-test-askeshav/kernel/cpu.c                   |   34 -----------
 linux-2.6.9-rc3-test-askeshav/lib/kobject_uevent.c           |    1 
 4 files changed, 4 insertions(+), 34 deletions(-)
kernel/cpu.c renamed from kernel/cpu.c~cpu_run_sbin_hotplug_v2 +0 -34
diff -puN kernel/cpu.c~cpu_run_sbin_hotplug_v2 kernel/cpu.c
@@ -57,33 +57,6 @@ static inline void check_for_tasks(int c
 	write_unlock_irq(&tasklist_lock);
 }
 
-/* Notify userspace when a cpu event occurs, by running '/sbin/hotplug
- * cpu' with certain environment variables set.  */
-static int cpu_run_sbin_hotplug(unsigned int cpu, const char *action)
-{
-	char *argv[3], *envp[5], cpu_str[12], action_str[32];
-	int i;
-
-	sprintf(cpu_str, "CPU=%d", cpu);
-	sprintf(action_str, "ACTION=%s", action);
-	/* FIXME: Add DEVPATH. --RR */
-
-	i = 0;
-	argv[i++] = hotplug_path;
-	argv[i++] = "cpu";
-	argv[i] = NULL;
-
-	i = 0;
-	/* minimal command environment */
-	envp[i++] = "HOME=/";
-	envp[i++] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
-	envp[i++] = cpu_str;
-	envp[i++] = action_str;
-	envp[i] = NULL;
-
-	return call_usermodehelper(argv[0], argv, envp, 0);
-}
-
 /* Take this CPU down. */
 static int take_cpu_down(void *unused)
 {
@@ -155,8 +128,6 @@ int cpu_down(unsigned int cpu)
 
 	check_for_tasks(cpu);
 
-	cpu_run_sbin_hotplug(cpu, "offline");
-
 out_thread:
 	err = kthread_stop(p);
 out_allowed:
@@ -165,11 +136,6 @@ out:
 	unlock_cpu_hotplug();
 	return err;
 }
-#else
-static inline int cpu_run_sbin_hotplug(unsigned int cpu, const char *action)
-{
-	return 0;
-}
 #endif /*CONFIG_HOTPLUG_CPU*/
 
 int __devinit cpu_up(unsigned int cpu)
drivers/base/cpu.c renamed from drivers/base/cpu.c~cpu_run_sbin_hotplug_v2 +2 -0
diff -puN drivers/base/cpu.c~cpu_run_sbin_hotplug_v2 drivers/base/cpu.c
@@ -32,6 +32,8 @@ static ssize_t store_online(struct sys_d
 	switch (buf[0]) {
 	case '0':
 		ret = cpu_down(cpu->sysdev.id);
+		if (!ret)
+			kobject_hotplug(&dev->kobj, KOBJ_OFFLINE);
 		break;
 	case '1':
 		ret = cpu_up(cpu->sysdev.id);
include/linux/kobject_uevent.h renamed from include/linux/kobject_uevent.h~cpu_run_sbin_hotplug_v2 +1 -0
diff -puN include/linux/kobject_uevent.h~cpu_run_sbin_hotplug_v2 include/linux/kobject_uevent.h
@@ -22,6 +22,7 @@ enum kobject_action {
 	KOBJ_CHANGE	= 0x02,	/* a sysfs attribute file has changed */
 	KOBJ_MOUNT	= 0x03,	/* mount event for block devices */
 	KOBJ_UMOUNT	= 0x04,	/* umount event for block devices */
+	KOBJ_OFFLINE	= 0x05,	/* offline event for hotplug devices */
 	KOBJ_MAX_ACTION,	/* must be last action listed */
 };
 
lib/kobject_uevent.c renamed from lib/kobject_uevent.c~cpu_run_sbin_hotplug_v2 +1 -0
diff -puN lib/kobject_uevent.c~cpu_run_sbin_hotplug_v2 lib/kobject_uevent.c
@@ -33,6 +33,7 @@ static char *actions[] = {
 	"change",	/* 0x02 */
 	"mount",	/* 0x03 */
 	"umount",	/* 0x04 */
+	"offline",	/* 0x05 */
 };
 
 static char *action_to_string(enum kobject_action action)

_

#6 lore raw
Quoted 8 lines · Keshavamurthy Anil S

Here is what I have come up with(please take a look at this patch).
I was successfully able to get rid of cpu_run_sbin_hotplug() function, but when I call kobject_hotplug() function, it is finding top_kobj->kset->hotplug_ops set to NULL and hence returns without calling call_usermodehelper(). Not sure if this is a bug in kobject_hotplug(), I feel kobject_hotplug() function should continue even if top_kobj->kset-hotplug_ops is NULL.

Yes, it doesn't seem necessary. We could give cpu_sysdev_class a valid-but-empty hotplug_ops but it seems simpler and more general to do it in kobject_hotplug().

Make kobject_hotplug() work even if the kobject's kset doesn't implement any hotplug_ops.

---

 25-akpm/lib/kobject_uevent.c |   19 ++++++++++++-------
 1 files changed, 12 insertions(+), 7 deletions(-)
lib/kobject_uevent.c renamed from lib/kobject_uevent.c~kobject_hotplug-no-hotplug_ops +12 -7
diff -puN lib/kobject_uevent.c~kobject_hotplug-no-hotplug_ops lib/kobject_uevent.c
@@ -191,6 +191,8 @@ void kobject_hotplug(struct kobject *kob
 	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 (!top_kobj->kset && top_kobj->parent) {
 		do {
@@ -198,15 +200,18 @@ void kobject_hotplug(struct kobject *kob
 		} while (!top_kobj->kset && top_kobj->parent);
 	}
 
-	if (top_kobj->kset && top_kobj->kset->hotplug_ops)
+	if (top_kobj->kset)
 		kset = top_kobj->kset;
 	else
 		return;
 
+	if (kset->hotplug_ops)
+		hotplug_ops = kset->hotplug_ops;
+
 	/* If the kset has a filter operation, call it.
 	   Skip the event, if the filter returns zero. */
-	if (kset->hotplug_ops->filter) {
-		if (!kset->hotplug_ops->filter(kset, kobj))
+	if (hotplug_ops->filter) {
+		if (!hotplug_ops->filter(kset, kobj))
 			return;
 	}
 
@@ -225,8 +230,8 @@ void kobject_hotplug(struct kobject *kob
 	if (!buffer)
 		goto exit;
 
-	if (kset->hotplug_ops->name)
-		name = kset->hotplug_ops->name(kset, kobj);
+	if (hotplug_ops->name)
+		name = hotplug_ops->name(kset, kobj);
 	if (name == NULL)
 		name = kset->kobj.name;
 
@@ -260,9 +265,9 @@ void kobject_hotplug(struct kobject *kob
 	envp [i++] = scratch;
 	scratch += sprintf(scratch, "SUBSYSTEM=%s", name) + 1;
 
-	if (kset->hotplug_ops->hotplug) {
+	if (hotplug_ops->hotplug) {
 		/* have the kset specific function add its stuff */
-		retval = kset->hotplug_ops->hotplug (kset, kobj,
+		retval = hotplug_ops->hotplug (kset, kobj,
 				  &envp[i], NUM_ENVP - i, scratch,
 				  BUFFER_SIZE - (scratch - buffer));
 		if (retval) {

_

#7 lore raw
Quoted 10 lines · Andrew Morton
Quoted 8 lines · Keshavamurthy Anil S

Here is what I have come up with(please take a look at this patch).
I was successfully able to get rid of cpu_run_sbin_hotplug() function, but when I call kobject_hotplug() function, it is finding top_kobj->kset->hotplug_ops set to NULL and hence returns without calling call_usermodehelper(). Not sure if this is a bug in kobject_hotplug(), I feel kobject_hotplug() function should continue even if top_kobj->kset-hotplug_ops is NULL.

Quoted 3 lines

Yes, it doesn't seem necessary. We could give cpu_sysdev_class a valid-but-empty hotplug_ops but it seems simpler and more general to do it in kobject_hotplug().

I tried that, but I found that parent "cpu" directory i.e /sys/devices/system/cpu itself was not getting created. Any clues?

#8 lore raw
Quoted 12 lines · Keshavamurthy Anil S
Quoted 10 lines · Andrew Morton
Quoted 8 lines · Keshavamurthy Anil S

Here is what I have come up with(please take a look at this patch).
I was successfully able to get rid of cpu_run_sbin_hotplug() function, but when I call kobject_hotplug() function, it is finding top_kobj->kset->hotplug_ops set to NULL and hence returns without calling call_usermodehelper(). Not sure if this is a bug in kobject_hotplug(), I feel kobject_hotplug() function should continue even if top_kobj->kset-hotplug_ops is NULL.

Quoted 6 lines
Quoted 3 lines

Yes, it doesn't seem necessary. We could give cpu_sysdev_class a valid-but-empty hotplug_ops but it seems simpler and more general to do it in kobject_hotplug().

I tried that, but I found that parent "cpu" directory i.e /sys/devices/system/cpu itself was not getting created. Any clues?

I don't see why the change to kobject_hotplug() would cause that directory to not be created.

With your patch and mine applied, /sys/devices/system/cpu is present and populated on my test box.

#9 lore raw
Quoted 17 lines · Andrew Morton
Quoted 16 lines · Keshavamurthy Anil S
Quoted 13 lines · Andrew Morton
Quoted 8 lines · Keshavamurthy Anil S

Here is what I have come up with(please take a look at this patch).
I was successfully able to get rid of cpu_run_sbin_hotplug() function, but when I call kobject_hotplug() function, it is finding top_kobj->kset->hotplug_ops set to NULL and hence returns without calling call_usermodehelper(). Not sure if this is a bug in kobject_hotplug(), I feel kobject_hotplug() function should continue even if top_kobj->kset-hotplug_ops is NULL.

Yes, it doesn't seem necessary. We could give cpu_sysdev_class a valid-but-empty hotplug_ops but it seems simpler and more general to do it in kobject_hotplug().

Quoted 8 lines
Quoted 2 lines

I tried that, but I found that parent "cpu" directory i.e /sys/devices/system/cpu itself was not getting created. Any clues?

I don't see why the change to kobject_hotplug() would cause that directory to not be created.

With your patch and mine applied, /sys/devices/system/cpu is present and populated on my test box.

Hi Andrew,
I am attaching the second one, just to make sure you and I have the same one.
If this is different than what you are having let me know.

By the way I am testing on IA64 box, with 2.6.9-rc3 + just bk-driver-core.patch from your 2.6.9-rc3-mm2-broken-out.tar.
I had to go for just above as I was seeing some out of memory messages on my IA64 box with complete rc3-mm2 patch.

thanks,
Anil

---

 linux-2.6.9-rc3-test-askeshav/drivers/base/cpu.c |    2 ++
 1 files changed, 2 insertions(+)
drivers/base/cpu.c renamed from drivers/base/cpu.c~test_akpm +2 -0
diff -puN drivers/base/cpu.c~test_akpm drivers/base/cpu.c
@@ -9,9 +9,11 @@
 #include <linux/topology.h>
 #include <linux/device.h>
 
+struct kset_hotplug_ops cpu_kset_hotplug_ops;
 
 struct sysdev_class cpu_sysdev_class = {
 	set_kset_name("cpu"),
+	.kset = { .hotplug_ops = &cpu_kset_hotplug_ops},
 };
 EXPORT_SYMBOL(cpu_sysdev_class);
 

_

#10 lore raw
Quoted 3 lines · Keshavamurthy Anil S

I am attaching the second one, just to make sure you and I have the same one.
If this is different than what you are having let me know.

yes, it's different from the one I'm using. I'm using the below, which I sent you an hour ago.

Quoted 4 lines

By the way I am testing on IA64 box, with 2.6.9-rc3 + just bk-driver-core.patch from your 2.6.9-rc3-mm2-broken-out.tar.
I had to go for just above as I was seeing some out of memory messages on my IA64 box with complete rc3-mm2 patch.

yes, my ia64 box goes oom during boot too. Something's broken in Tony's ia64 tree.

Make kobject_hotplug() work even if the kobject's kset doesn't implement any hotplug_ops.

---

 25-akpm/lib/kobject_uevent.c |   19 ++++++++++++-------
 1 files changed, 12 insertions(+), 7 deletions(-)
lib/kobject_uevent.c renamed from lib/kobject_uevent.c~kobject_hotplug-no-hotplug_ops +12 -7
diff -puN lib/kobject_uevent.c~kobject_hotplug-no-hotplug_ops lib/kobject_uevent.c
@@ -191,6 +191,8 @@ void kobject_hotplug(struct kobject *kob
 	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 (!top_kobj->kset && top_kobj->parent) {
 		do {
@@ -198,15 +200,18 @@ void kobject_hotplug(struct kobject *kob
 		} while (!top_kobj->kset && top_kobj->parent);
 	}
 
-	if (top_kobj->kset && top_kobj->kset->hotplug_ops)
+	if (top_kobj->kset)
 		kset = top_kobj->kset;
 	else
 		return;
 
+	if (kset->hotplug_ops)
+		hotplug_ops = kset->hotplug_ops;
+
 	/* If the kset has a filter operation, call it.
 	   Skip the event, if the filter returns zero. */
-	if (kset->hotplug_ops->filter) {
-		if (!kset->hotplug_ops->filter(kset, kobj))
+	if (hotplug_ops->filter) {
+		if (!hotplug_ops->filter(kset, kobj))
 			return;
 	}
 
@@ -225,8 +230,8 @@ void kobject_hotplug(struct kobject *kob
 	if (!buffer)
 		goto exit;
 
-	if (kset->hotplug_ops->name)
-		name = kset->hotplug_ops->name(kset, kobj);
+	if (hotplug_ops->name)
+		name = hotplug_ops->name(kset, kobj);
 	if (name == NULL)
 		name = kset->kobj.name;
 
@@ -260,9 +265,9 @@ void kobject_hotplug(struct kobject *kob
 	envp [i++] = scratch;
 	scratch += sprintf(scratch, "SUBSYSTEM=%s", name) + 1;
 
-	if (kset->hotplug_ops->hotplug) {
+	if (hotplug_ops->hotplug) {
 		/* have the kset specific function add its stuff */
-		retval = kset->hotplug_ops->hotplug (kset, kobj,
+		retval = hotplug_ops->hotplug (kset, kobj,
 				  &envp[i], NUM_ENVP - i, scratch,
 				  BUFFER_SIZE - (scratch - buffer));
 		if (retval) {

_

#11 lore raw
Quoted 1 line · Andrew Morton

Keshavamurthy Anil S <[email protected]> wrote:

Andrew,
My patch which removes cpu_run_sbin_hotplug() to use kobject_hotplug() works
fine with your patch which modifies kobject_hotplug().
I am able to get the same cpu "offline" notifications to user land.

You can safely apply both of our patches.

Thanks and regards,
Anil