Skip to contentSkip to conversation

change memoryX->phys_device from number to symlink [1/2] generic func

PATCH In review KAMEZAWA Hiroyuki 3 participants 5 messages Feb 17, 2006 – Feb 18, 2006

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

PATCH 1/1 change memoryX->phys_device from number to symlink [1/2] generic func

This patch is from memory hotplug tree in lhms.
This patch changes memory_device's phys_device member from just a number to symbolic link to the device. AFAIK, phys_device is not used now.

example)
$readlink /sys/devices/system/memory/memory10/phys_device ../../../../firmware/acpi/namespace/ACPI/_SB/LSB1/MEM3

This will help memory hotplug shell script.

-- Kame

struct memory_block is a struct for memory hotplug and creates sysfs entry like this.
/sys/device/system/memory/memory10/
This struct has a member phys_device and this is just a number. A user can read this member like this.
%cat /sys/devices/system/memory/memory10/state
0

AFAIK, there is no table to match this number and a real device. And this is always 0 now.

This patch changes memory_block->phys_device from a number to symbolic link to the device to which the memory block belongs to.
ex)
% readlink /sys/devices/system/memory/memory10/phys_device ../../../../firmware/acpi/namespace/ACPI/_SB/LSB1/MEM3

This is usefull for memory hotplug shell script and we don't have to maintain a table phys_device_number <-> device.

A patch making acpi-memory-hotplug use this will follow this patch.

drivers/base/memory.c +58 -27
Index: testtree/drivers/base/memory.c
===================================================================
@@ -252,31 +252,30 @@
 	return count;
 }
 
-/*
- * phys_device is a bad name for this.  What I really want
- * is a way to differentiate between memory ranges that
- * are part of physical devices that constitute
- * a complete removable unit or fru.
- * i.e. do these ranges belong to the same physical device,
- * s.t. if I offline all of these sections I can then
- * remove the physical device?
- */
-static ssize_t show_phys_device(struct sys_device *dev, char *buf)
-{
-	struct memory_block *mem =
-		container_of(dev, struct memory_block, sysdev);
-	return sprintf(buf, "%d\n", mem->phys_device);
-}
 
 static SYSDEV_ATTR(phys_index, 0444, show_mem_phys_index, NULL);
 static SYSDEV_ATTR(state, 0644, show_mem_state, store_mem_state);
-static SYSDEV_ATTR(phys_device, 0444, show_phys_device, NULL);
 
 #define mem_create_simple_file(mem, attr_name)	\
 	sysdev_create_file(&mem->sysdev, &attr_##attr_name)
 #define mem_remove_simple_file(mem, attr_name)	\
 	sysdev_remove_file(&mem->sysdev, &attr_##attr_name)
 
+static int attach_phys_device(struct memory_block *mem, struct kobject *kobj)
+{
+	int ret;
+	ret = sysfs_create_link(&mem->sysdev.kobj, kobj, "phys_device");
+	if (ret)
+		return ret;
+	mem->phys_device = kobj;
+	return 0;
+}
+
+static void remove_phys_device(struct memory_block *mem)
+{
+	sysfs_remove_link(&mem->sysdev.kobj, "phys_device");
+	mem->phys_device = NULL;
+}
 /*
  * Block size attribute stuff
  */
@@ -330,13 +329,14 @@
 #endif
 
 /*
- * Note that phys_device is optional.  It is here to allow for
- * differentiation between which *physical* devices each
+ * kobj is a kobject of physical memory device which includes the specified range
+ * It is here to allow for differentiation between which *physical* devices each
  * section belongs to...
+ * If kobj != NULL, symbolic link to device from mem_section is created.
  */
 
 static int add_memory_block(unsigned long node_id, struct mem_section *section,
-		     unsigned long state, int phys_device)
+		     unsigned long state, struct kobject *kobj)
 {
 	struct memory_block *mem = kzalloc(sizeof(*mem), GFP_KERNEL);
 	int ret = 0;
@@ -346,16 +346,16 @@
 
 	mem->phys_index = __section_nr(section);
 	mem->state = state;
+	mem->phys_device = 0;
 	init_MUTEX(&mem->state_sem);
-	mem->phys_device = phys_device;
 
 	ret = register_memory(mem, section, NULL);
 	if (!ret)
 		ret = mem_create_simple_file(mem, phys_index);
 	if (!ret)
 		ret = mem_create_simple_file(mem, state);
-	if (!ret)
-		ret = mem_create_simple_file(mem, phys_device);
+	if (!ret && kobj)
+		ret = attach_phys_device(mem, kobj);
 
 	return ret;
 }
@@ -391,27 +391,58 @@
 	return mem;
 }
 
-int remove_memory_block(unsigned long node_id, struct mem_section *section,
-		int phys_device)
+static struct memory_block *pfn_to_memory_block(unsigned long pfn)
+{
+	struct mem_section *section;
+	section = __nr_to_section(pfn_to_section_nr(pfn));
+	return find_memory_block(section);
+}
+
+
+int remove_memory_block(unsigned long node_id, struct mem_section *section)
 {
 	struct memory_block *mem;
 
 	mem = find_memory_block(section);
 	mem_remove_simple_file(mem, phys_index);
 	mem_remove_simple_file(mem, state);
-	mem_remove_simple_file(mem, phys_device);
+	if (mem->phys_device)
+		remove_phys_device(mem);
 	unregister_memory(mem, section, NULL);
 
 	return 0;
 }
 
 /*
+ * creating symbolic link from mem_section[] in specified address range
+ * to specified device. This device here is expected to be physical memory device.
+ * This symbolic link will be used to show relationship between mem_section and device.
+ */
+int attach_device_to_memsection(u64 start_addr, u64 end_addr, struct kobject *kobj)
+{
+	unsigned long pfn = start_addr >> PAGE_SHIFT;
+	unsigned long end_pfn = end_addr >> PAGE_SHIFT;
+	struct memory_block *mem;
+	int ret = 0;
+	for (; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
+		mem = pfn_to_memory_block(pfn);
+		if (mem) {
+			ret = attach_phys_device(mem, kobj);
+			if (ret)
+				break;
+		}
+	}
+	return ret;
+}
+
+
+/*
  * need an interface for the VM to add new memory regions,
  * but without onlining it.
  */
 int register_new_memory(struct mem_section *section)
 {
-	return add_memory_block(0, section, MEM_OFFLINE, 0);
+	return add_memory_block(0, section, MEM_OFFLINE, NULL);
 }
 
 int unregister_memory_section(struct mem_section *section)
@@ -419,7 +450,7 @@
 	if (!valid_section(section))
 		return -EINVAL;
 
-	return remove_memory_block(0, section, 0);
+	return remove_memory_block(0, section);
 }
 
 /*
include/linux/memory.h +7 -2
Index: testtree/include/linux/memory.h
===================================================================
@@ -31,7 +31,9 @@
 	 * initialization.
 	 */
 	struct semaphore state_sem;
-	int phys_device;		/* to which fru does this belong? */
+
+	/* information related to hardware */
+	struct kobject *phys_device;	/* a device includes this memory. */
 	void *hw;			/* optional pointer to fw/hw data */
 	int (*phys_callback)(struct memory_block *);
 	struct sys_device sysdev;
@@ -73,7 +75,10 @@
 extern int register_new_memory(struct mem_section *);
 extern int unregister_memory_section(struct mem_section *);
 extern int memory_dev_init(void);
-extern int remove_memory_block(unsigned long, struct mem_section *, int);
+extern int remove_memory_block(unsigned long, struct mem_section *);
+/* creating symbolic link between memory device and memory section */
+extern int attach_device_to_memsection(u64 start_addr, u64 end_addr, struct kobject *kobj);
+
 
 #define CONFIG_MEM_BLOCK_SIZE	(PAGES_PER_SECTION<<PAGE_SHIFT)
 
#1 lore raw
Quoted 4 lines · KAMEZAWA Hiroyuki

This patch is from memory hotplug tree in lhms.
This patch changes memory_device's phys_device member from just a number
to symbolic link to the device. AFAIK, phys_device is not used now.

We meant to use it for hardware which did not nicely export its information in sysfs. But, it seems reasonable now to say that, if you want this file to mean anything on your hardware, export your hardware in sysfs.

As you said, I don't think anybody uses this yet.

-- Dave

#2 lore raw
In reply to KAMEZAWA Hiroyuki
Quoted 9 lines · KAMEZAWA Hiroyuki

This patch is from memory hotplug tree in lhms.
This patch changes memory_device's phys_device member from just a number to symbolic link to the device. AFAIK, phys_device is not used now.

example)
$readlink /sys/devices/system/memory/memory10/phys_device ../../../../firmware/acpi/namespace/ACPI/_SB/LSB1/MEM3

This will help memory hotplug shell script.

You should bring this up on the acpi mailing list, as I know Pat Mochel has redone the whole "acpi in sysfs" thing, so this patch will break those patches (or his will break yours.)

I suggest you discuss this with him.

good luck,

greg k-h

#3 lore raw
Quoted 7 lines · Greg KH

You should bring this up on the acpi mailing list, as I know Pat Mochel has redone the whole "acpi in sysfs" thing, so this patch will break those patches (or his will break yours.)

I suggest you discuss this with him.

Thanks, I'll go acpi-list first.
--Kame

#4 lore raw
Quoted 8 lines · Kamezawa Hiroyuki
Quoted 7 lines · Greg KH

You should bring this up on the acpi mailing list, as I know Pat Mochel has redone the whole "acpi in sysfs" thing, so this patch will break those patches (or his will break yours.)

I suggest you discuss this with him.

Ah..okay. it's this ..

http://www.kernel.org/pub/linux/kernel/people/mochel/patches/acpi/driver-model/

It looks things are completely changed ;)

I'd like to look into this more before going ahead.

Thanks,
-- Kame