[net-next-2.6.git] / drivers / base / memory.c

/*
 * drivers/base/memory.c - basic Memory class support
 *
 * Written by Matt Tolentino <matthew.e.tolentino@intel.com>
 *            Dave Hansen <haveblue@us.ibm.com>
 *
 * This file provides the necessary infrastructure to represent
 * a SPARSEMEM-memory-model system's physical memory in /sysfs.
 * All arch-independent code that assumes MEMORY_HOTPLUG requires
 * SPARSEMEM should be contained here, or in mm/memory_hotplug.c.
 */

#include <linux/sysdev.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/topology.h>
#include <linux/capability.h>
#include <linux/device.h>
#include <linux/memory.h>
#include <linux/kobject.h>
#include <linux/memory_hotplug.h>
#include <linux/mm.h>
#include <linux/mutex.h>
#include <linux/stat.h>

#include <asm/atomic.h>
#include <asm/uaccess.h>

#define MEMORY_CLASS_NAME	"memory"

static struct sysdev_class memory_sysdev_class = {
	.name = MEMORY_CLASS_NAME,
};

static const char *memory_uevent_name(struct kset *kset, struct kobject *kobj)
{
	return MEMORY_CLASS_NAME;
}

static int memory_uevent(struct kset *kset, struct kobject *obj, struct kobj_uevent_env *env)
{
	int retval = 0;

	return retval;
}

static struct kset_uevent_ops memory_uevent_ops = {
	.name		= memory_uevent_name,
	.uevent		= memory_uevent,
};

static BLOCKING_NOTIFIER_HEAD(memory_chain);

int register_memory_notifier(struct notifier_block *nb)
{
        return blocking_notifier_chain_register(&memory_chain, nb);
}
EXPORT_SYMBOL(register_memory_notifier);

void unregister_memory_notifier(struct notifier_block *nb)
{
        blocking_notifier_chain_unregister(&memory_chain, nb);
}
EXPORT_SYMBOL(unregister_memory_notifier);

static ATOMIC_NOTIFIER_HEAD(memory_isolate_chain);

int register_memory_isolate_notifier(struct notifier_block *nb)
{
	return atomic_notifier_chain_register(&memory_isolate_chain, nb);
}
EXPORT_SYMBOL(register_memory_isolate_notifier);

void unregister_memory_isolate_notifier(struct notifier_block *nb)
{
	atomic_notifier_chain_unregister(&memory_isolate_chain, nb);
}
EXPORT_SYMBOL(unregister_memory_isolate_notifier);

/*
 * register_memory - Setup a sysfs device for a memory block
 */
static
int register_memory(struct memory_block *memory, struct mem_section *section)
{
	int error;

	memory->sysdev.cls = &memory_sysdev_class;
	memory->sysdev.id = __section_nr(section);

	error = sysdev_register(&memory->sysdev);
	return error;
}

static void
unregister_memory(struct memory_block *memory, struct mem_section *section)
{
	BUG_ON(memory->sysdev.cls != &memory_sysdev_class);
	BUG_ON(memory->sysdev.id != __section_nr(section));

	/* drop the ref. we got in remove_memory_block() */
	kobject_put(&memory->sysdev.kobj);
	sysdev_unregister(&memory->sysdev);
}

/*
 * use this as the physical section index that this memsection
 * uses.
 */

static ssize_t show_mem_phys_index(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
{
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);
	return sprintf(buf, "%08lx\n", mem->phys_index);
}

/*
 * Show whether the section of memory is likely to be hot-removable
 */
static ssize_t show_mem_removable(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
{
	unsigned long start_pfn;
	int ret;
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);

	start_pfn = section_nr_to_pfn(mem->phys_index);
	ret = is_mem_section_removable(start_pfn, PAGES_PER_SECTION);
	return sprintf(buf, "%d\n", ret);
}

/*
 * online, offline, going offline, etc.
 */
static ssize_t show_mem_state(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
{
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);
	ssize_t len = 0;

	/*
	 * We can probably put these states in a nice little array
	 * so that they're not open-coded
	 */
	switch (mem->state) {
		case MEM_ONLINE:
			len = sprintf(buf, "online\n");
			break;
		case MEM_OFFLINE:
			len = sprintf(buf, "offline\n");
			break;
		case MEM_GOING_OFFLINE:
			len = sprintf(buf, "going-offline\n");
			break;
		default:
			len = sprintf(buf, "ERROR-UNKNOWN-%ld\n",
					mem->state);
			WARN_ON(1);
			break;
	}

	return len;
}

int memory_notify(unsigned long val, void *v)
{
	return blocking_notifier_call_chain(&memory_chain, val, v);
}

int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
static int
memory_block_action(struct memory_block *mem, unsigned long action)
{
	int i;
	unsigned long psection;
	unsigned long start_pfn, start_paddr;
	struct page *first_page;
	int ret;
	int old_state = mem->state;

	psection = mem->phys_index;
	first_page = pfn_to_page(psection << PFN_SECTION_SHIFT);

	/*
	 * The probe routines leave the pages reserved, just
	 * as the bootmem code does.  Make sure they're still
	 * that way.
	 */
	if (action == MEM_ONLINE) {
		for (i = 0; i < PAGES_PER_SECTION; i++) {
			if (PageReserved(first_page+i))
				continue;

			printk(KERN_WARNING "section number %ld page number %d "
				"not reserved, was it already online? \n",
				psection, i);
			return -EBUSY;
		}
	}

	switch (action) {
		case MEM_ONLINE:
			start_pfn = page_to_pfn(first_page);
			ret = online_pages(start_pfn, PAGES_PER_SECTION);
			break;
		case MEM_OFFLINE:
			mem->state = MEM_GOING_OFFLINE;
			start_paddr = page_to_pfn(first_page) << PAGE_SHIFT;
			ret = remove_memory(start_paddr,
					    PAGES_PER_SECTION << PAGE_SHIFT);
			if (ret) {
				mem->state = old_state;
				break;
			}
			break;
		default:
			WARN(1, KERN_WARNING "%s(%p, %ld) unknown action: %ld\n",
					__func__, mem, action, action);
			ret = -EINVAL;
	}

	return ret;
}

static int memory_block_change_state(struct memory_block *mem,
		unsigned long to_state, unsigned long from_state_req)
{
	int ret = 0;
	mutex_lock(&mem->state_mutex);

	if (mem->state != from_state_req) {
		ret = -EINVAL;
		goto out;
	}

	ret = memory_block_action(mem, to_state);
	if (!ret)
		mem->state = to_state;

out:
	mutex_unlock(&mem->state_mutex);
	return ret;
}

static ssize_t
store_mem_state(struct sys_device *dev,
		struct sysdev_attribute *attr, const char *buf, size_t count)
{
	struct memory_block *mem;
	unsigned int phys_section_nr;
	int ret = -EINVAL;

	mem = container_of(dev, struct memory_block, sysdev);
	phys_section_nr = mem->phys_index;

	if (!present_section_nr(phys_section_nr))
		goto out;

	if (!strncmp(buf, "online", min((int)count, 6)))
		ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
	else if(!strncmp(buf, "offline", min((int)count, 7)))
		ret = memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
out:
	if (ret)
		return ret;
	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,
				struct sysdev_attribute *attr, 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);
static SYSDEV_ATTR(removable, 0444, show_mem_removable, 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)

/*
 * Block size attribute stuff
 */
static ssize_t
print_block_size(struct sysdev_class *class,
		 struct sysdev_class_attribute *class_attr,
		 char *buf)
{
	return sprintf(buf, "%#lx\n", (unsigned long)PAGES_PER_SECTION * PAGE_SIZE);
}

static SYSDEV_CLASS_ATTR(block_size_bytes, 0444, print_block_size, NULL);

static int block_size_init(void)
{
	return sysfs_create_file(&memory_sysdev_class.kset.kobj,
				&attr_block_size_bytes.attr);
}

/*
 * Some architectures will have custom drivers to do this, and
 * will not need to do it from userspace.  The fake hot-add code
 * as well as ppc64 will do all of their discovery in userspace
 * and will require this interface.
 */
#ifdef CONFIG_ARCH_MEMORY_PROBE
static ssize_t
memory_probe_store(struct sysdev_class *class,
		   struct sysdev_class_attribute *class_attr,
		   const char *buf, size_t count)
{
	u64 phys_addr;
	int nid;
	int ret;

	phys_addr = simple_strtoull(buf, NULL, 0);

	nid = memory_add_physaddr_to_nid(phys_addr);
	ret = add_memory(nid, phys_addr, PAGES_PER_SECTION << PAGE_SHIFT);

	if (ret)
		count = ret;

	return count;
}
static SYSDEV_CLASS_ATTR(probe, S_IWUSR, NULL, memory_probe_store);

static int memory_probe_init(void)
{
	return sysfs_create_file(&memory_sysdev_class.kset.kobj,
				&attr_probe.attr);
}
#else
static inline int memory_probe_init(void)
{
	return 0;
}
#endif

#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
static ssize_t
store_soft_offline_page(struct sysdev_class *class,
			struct sysdev_class_attribute *class_attr,
			const char *buf, size_t count)
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	if (strict_strtoull(buf, 0, &pfn) < 0)
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
	if (!pfn_valid(pfn))
		return -ENXIO;
	ret = soft_offline_page(pfn_to_page(pfn), 0);
	return ret == 0 ? count : ret;
}

/* Forcibly offline a page, including killing processes. */
static ssize_t
store_hard_offline_page(struct sysdev_class *class,
			struct sysdev_class_attribute *class_attr,
			const char *buf, size_t count)
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	if (strict_strtoull(buf, 0, &pfn) < 0)
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
	ret = __memory_failure(pfn, 0, 0);
	return ret ? ret : count;
}

static SYSDEV_CLASS_ATTR(soft_offline_page, 0644, NULL, store_soft_offline_page);
static SYSDEV_CLASS_ATTR(hard_offline_page, 0644, NULL, store_hard_offline_page);

static __init int memory_fail_init(void)
{
	int err;

	err = sysfs_create_file(&memory_sysdev_class.kset.kobj,
				&attr_soft_offline_page.attr);
	if (!err)
		err = sysfs_create_file(&memory_sysdev_class.kset.kobj,
				&attr_hard_offline_page.attr);
	return err;
}
#else
static inline int memory_fail_init(void)
{
	return 0;
}
#endif

/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */

static int add_memory_block(int nid, struct mem_section *section,
			unsigned long state, int phys_device,
			enum mem_add_context context)
{
	struct memory_block *mem = kzalloc(sizeof(*mem), GFP_KERNEL);
	int ret = 0;

	if (!mem)
		return -ENOMEM;

	mem->phys_index = __section_nr(section);
	mem->state = state;
	mutex_init(&mem->state_mutex);
	mem->phys_device = phys_device;

	ret = register_memory(mem, section);
	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)
		ret = mem_create_simple_file(mem, removable);
	if (!ret) {
		if (context == HOTPLUG)
			ret = register_mem_sect_under_node(mem, nid);
	}

	return ret;
}

/*
 * For now, we have a linear search to go find the appropriate
 * memory_block corresponding to a particular phys_index. If
 * this gets to be a real problem, we can always use a radix
 * tree or something here.
 *
 * This could be made generic for all sysdev classes.
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	struct kobject *kobj;
	struct sys_device *sysdev;
	struct memory_block *mem;
	char name[sizeof(MEMORY_CLASS_NAME) + 9 + 1];

	/*
	 * This only works because we know that section == sysdev->id
	 * slightly redundant with sysdev_register()
	 */
	sprintf(&name[0], "%s%d", MEMORY_CLASS_NAME, __section_nr(section));

	kobj = kset_find_obj(&memory_sysdev_class.kset, name);
	if (!kobj)
		return NULL;

	sysdev = container_of(kobj, struct sys_device, kobj);
	mem = container_of(sysdev, struct memory_block, sysdev);

	return mem;
}

int remove_memory_block(unsigned long node_id, struct mem_section *section,
		int phys_device)
{
	struct memory_block *mem;

	mem = find_memory_block(section);
	unregister_mem_sect_under_nodes(mem);
	mem_remove_simple_file(mem, phys_index);
	mem_remove_simple_file(mem, state);
	mem_remove_simple_file(mem, phys_device);
	mem_remove_simple_file(mem, removable);
	unregister_memory(mem, section);

	return 0;
}

/*
 * need an interface for the VM to add new memory regions,
 * but without onlining it.
 */
int register_new_memory(int nid, struct mem_section *section)
{
	return add_memory_block(nid, section, MEM_OFFLINE, 0, HOTPLUG);
}

int unregister_memory_section(struct mem_section *section)
{
	if (!present_section(section))
		return -EINVAL;

	return remove_memory_block(0, section, 0);
}

/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	unsigned int i;
	int ret;
	int err;

	memory_sysdev_class.kset.uevent_ops = &memory_uevent_ops;
	ret = sysdev_class_register(&memory_sysdev_class);
	if (ret)
		goto out;

	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
	for (i = 0; i < NR_MEM_SECTIONS; i++) {
		if (!present_section_nr(i))
			continue;
		err = add_memory_block(0, __nr_to_section(i), MEM_ONLINE,
					0, BOOT);
		if (!ret)
			ret = err;
	}

	err = memory_probe_init();
	if (!ret)
		ret = err;
	err = memory_fail_init();
	if (!ret)
		ret = err;
	err = block_size_init();
	if (!ret)
		ret = err;
out:
	if (ret)
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
	return ret;
}
Commit	Line	Data
3947be19 DH	1	/*
	2	* drivers/base/memory.c - basic Memory class support
	3	*
	4	* Written by Matt Tolentino <matthew.e.tolentino@intel.com>
	5	* Dave Hansen <haveblue@us.ibm.com>
	6	*
	7	* This file provides the necessary infrastructure to represent
	8	* a SPARSEMEM-memory-model system's physical memory in /sysfs.
	9	* All arch-independent code that assumes MEMORY_HOTPLUG requires
	10	* SPARSEMEM should be contained here, or in mm/memory_hotplug.c.
	11	*/
	12
	13	#include <linux/sysdev.h>
	14	#include <linux/module.h>
	15	#include <linux/init.h>
3947be19	16	#include <linux/topology.h>
c59ede7b	17	#include <linux/capability.h>
3947be19 DH	18	#include <linux/device.h>
	19	#include <linux/memory.h>
	20	#include <linux/kobject.h>
	21	#include <linux/memory_hotplug.h>
	22	#include <linux/mm.h>
da19cbcf	23	#include <linux/mutex.h>
9f1b16a5 SL	24	#include <linux/stat.h>
9f1b16a5 SL	25
3947be19 DH	26	#include <asm/atomic.h>
	27	#include <asm/uaccess.h>
	28
	29	#define MEMORY_CLASS_NAME "memory"
	30
	31	static struct sysdev_class memory_sysdev_class = {
af5ca3f4	32	.name = MEMORY_CLASS_NAME,
3947be19	33	};
3947be19	34
312c004d	35	static const char memory_uevent_name(struct kset kset, struct kobject *kobj)
3947be19 DH	36	{
	37	return MEMORY_CLASS_NAME;
	38	}
	39
9ec0fd4e	40	static int memory_uevent(struct kset kset, struct kobject obj, struct kobj_uevent_env *env)
3947be19 DH	41	{
	42	int retval = 0;
	43
	44	return retval;
	45	}
	46
312c004d KS	47	static struct kset_uevent_ops memory_uevent_ops = {
	48	.name = memory_uevent_name,
	49	.uevent = memory_uevent,
3947be19 DH	50	};
3947be19 DH	51
e041c683	52	static BLOCKING_NOTIFIER_HEAD(memory_chain);
3947be19	53
98a38ebd	54	int register_memory_notifier(struct notifier_block *nb)
3947be19	55	{
e041c683	56	return blocking_notifier_chain_register(&memory_chain, nb);
3947be19	57	}
3c82c30c	58	EXPORT_SYMBOL(register_memory_notifier);
3947be19	59
98a38ebd	60	void unregister_memory_notifier(struct notifier_block *nb)
3947be19	61	{
e041c683	62	blocking_notifier_chain_unregister(&memory_chain, nb);
3947be19	63	}
3c82c30c	64	EXPORT_SYMBOL(unregister_memory_notifier);
3947be19	65
925cc71e RJ	66	static ATOMIC_NOTIFIER_HEAD(memory_isolate_chain);
	67
	68	int register_memory_isolate_notifier(struct notifier_block *nb)
	69	{
	70	return atomic_notifier_chain_register(&memory_isolate_chain, nb);
	71	}
	72	EXPORT_SYMBOL(register_memory_isolate_notifier);
	73
	74	void unregister_memory_isolate_notifier(struct notifier_block *nb)
	75	{
	76	atomic_notifier_chain_unregister(&memory_isolate_chain, nb);
	77	}
	78	EXPORT_SYMBOL(unregister_memory_isolate_notifier);
	79
3947be19 DH	80	/*
	81	* register_memory - Setup a sysfs device for a memory block
	82	*/
00a41db5 BP	83	static
00a41db5 BP	84	int register_memory(struct memory_block memory, struct mem_section section)
3947be19 DH	85	{
	86	int error;
	87
	88	memory->sysdev.cls = &memory_sysdev_class;
	89	memory->sysdev.id = __section_nr(section);
	90
	91	error = sysdev_register(&memory->sysdev);
3947be19 DH	92	return error;
	93	}
	94
	95	static void
00a41db5	96	unregister_memory(struct memory_block memory, struct mem_section section)
3947be19 DH	97	{
	98	BUG_ON(memory->sysdev.cls != &memory_sysdev_class);
	99	BUG_ON(memory->sysdev.id != __section_nr(section));
	100
00a41db5 BP	101	/* drop the ref. we got in remove_memory_block() */
00a41db5 BP	102	kobject_put(&memory->sysdev.kobj);
3947be19	103	sysdev_unregister(&memory->sysdev);
3947be19 DH	104	}
	105
	106	/*
	107	* use this as the physical section index that this memsection
	108	* uses.
	109	*/
	110
4a0b2b4d AK	111	static ssize_t show_mem_phys_index(struct sys_device *dev,
4a0b2b4d AK	112	struct sysdev_attribute attr, char buf)
3947be19 DH	113	{
	114	struct memory_block *mem =
	115	container_of(dev, struct memory_block, sysdev);
	116	return sprintf(buf, "%08lx\n", mem->phys_index);
	117	}
	118
5c755e9f BP	119	/*
	120	* Show whether the section of memory is likely to be hot-removable
	121	*/
1f07be1c SR	122	static ssize_t show_mem_removable(struct sys_device *dev,
1f07be1c SR	123	struct sysdev_attribute attr, char buf)
5c755e9f BP	124	{
	125	unsigned long start_pfn;
	126	int ret;
	127	struct memory_block *mem =
	128	container_of(dev, struct memory_block, sysdev);
	129
	130	start_pfn = section_nr_to_pfn(mem->phys_index);
	131	ret = is_mem_section_removable(start_pfn, PAGES_PER_SECTION);
	132	return sprintf(buf, "%d\n", ret);
	133	}
	134
3947be19 DH	135	/*
	136	* online, offline, going offline, etc.
	137	*/
4a0b2b4d AK	138	static ssize_t show_mem_state(struct sys_device *dev,
4a0b2b4d AK	139	struct sysdev_attribute attr, char buf)
3947be19 DH	140	{
	141	struct memory_block *mem =
	142	container_of(dev, struct memory_block, sysdev);
	143	ssize_t len = 0;
	144
	145	/*
	146	* We can probably put these states in a nice little array
	147	* so that they're not open-coded
	148	*/
	149	switch (mem->state) {
	150	case MEM_ONLINE:
	151	len = sprintf(buf, "online\n");
	152	break;
	153	case MEM_OFFLINE:
	154	len = sprintf(buf, "offline\n");
	155	break;
	156	case MEM_GOING_OFFLINE:
	157	len = sprintf(buf, "going-offline\n");
	158	break;
	159	default:
	160	len = sprintf(buf, "ERROR-UNKNOWN-%ld\n",
	161	mem->state);
	162	WARN_ON(1);
	163	break;
	164	}
	165
	166	return len;
	167	}
	168
7b78d335	169	int memory_notify(unsigned long val, void *v)
3947be19	170	{
e041c683	171	return blocking_notifier_call_chain(&memory_chain, val, v);
3947be19 DH	172	}
3947be19 DH	173
925cc71e RJ	174	int memory_isolate_notify(unsigned long val, void *v)
	175	{
	176	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
	177	}
	178
3947be19 DH	179	/*
	180	* MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
	181	* OK to have direct references to sparsemem variables in here.
	182	*/
	183	static int
	184	memory_block_action(struct memory_block *mem, unsigned long action)
	185	{
	186	int i;
	187	unsigned long psection;
	188	unsigned long start_pfn, start_paddr;
	189	struct page *first_page;
	190	int ret;
	191	int old_state = mem->state;
	192
	193	psection = mem->phys_index;
	194	first_page = pfn_to_page(psection << PFN_SECTION_SHIFT);
	195
	196	/*
	197	* The probe routines leave the pages reserved, just
	198	* as the bootmem code does. Make sure they're still
	199	* that way.
	200	*/
	201	if (action == MEM_ONLINE) {
	202	for (i = 0; i < PAGES_PER_SECTION; i++) {
	203	if (PageReserved(first_page+i))
	204	continue;
	205
	206	printk(KERN_WARNING "section number %ld page number %d "
	207	"not reserved, was it already online? \n",
	208	psection, i);
	209	return -EBUSY;
	210	}
	211	}
	212
	213	switch (action) {
	214	case MEM_ONLINE:
	215	start_pfn = page_to_pfn(first_page);
	216	ret = online_pages(start_pfn, PAGES_PER_SECTION);
	217	break;
	218	case MEM_OFFLINE:
	219	mem->state = MEM_GOING_OFFLINE;
3947be19 DH	220	start_paddr = page_to_pfn(first_page) << PAGE_SHIFT;
	221	ret = remove_memory(start_paddr,
	222	PAGES_PER_SECTION << PAGE_SHIFT);
	223	if (ret) {
	224	mem->state = old_state;
	225	break;
	226	}
3947be19 DH	227	break;
3947be19 DH	228	default:
f810a5cf	229	WARN(1, KERN_WARNING "%s(%p, %ld) unknown action: %ld\n",
2b3a302a	230	__func__, mem, action, action);
3947be19 DH	231	ret = -EINVAL;
3947be19 DH	232	}
3947be19 DH	233
	234	return ret;
	235	}
	236
	237	static int memory_block_change_state(struct memory_block *mem,
	238	unsigned long to_state, unsigned long from_state_req)
	239	{
	240	int ret = 0;
da19cbcf	241	mutex_lock(&mem->state_mutex);
3947be19 DH	242
	243	if (mem->state != from_state_req) {
	244	ret = -EINVAL;
	245	goto out;
	246	}
	247
	248	ret = memory_block_action(mem, to_state);
	249	if (!ret)
	250	mem->state = to_state;
	251
	252	out:
da19cbcf	253	mutex_unlock(&mem->state_mutex);
3947be19 DH	254	return ret;
	255	}
	256
	257	static ssize_t
4a0b2b4d AK	258	store_mem_state(struct sys_device *dev,
4a0b2b4d AK	259	struct sysdev_attribute attr, const char buf, size_t count)
3947be19 DH	260	{
	261	struct memory_block *mem;
	262	unsigned int phys_section_nr;
	263	int ret = -EINVAL;
	264
	265	mem = container_of(dev, struct memory_block, sysdev);
	266	phys_section_nr = mem->phys_index;
	267
540557b9	268	if (!present_section_nr(phys_section_nr))
3947be19 DH	269	goto out;
	270
	271	if (!strncmp(buf, "online", min((int)count, 6)))
	272	ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
	273	else if(!strncmp(buf, "offline", min((int)count, 7)))
	274	ret = memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
	275	out:
	276	if (ret)
	277	return ret;
	278	return count;
	279	}
	280
	281	/*
	282	* phys_device is a bad name for this. What I really want
	283	* is a way to differentiate between memory ranges that
	284	* are part of physical devices that constitute
	285	* a complete removable unit or fru.
	286	* i.e. do these ranges belong to the same physical device,
	287	* s.t. if I offline all of these sections I can then
	288	* remove the physical device?
	289	*/
4a0b2b4d AK	290	static ssize_t show_phys_device(struct sys_device *dev,
4a0b2b4d AK	291	struct sysdev_attribute attr, char buf)
3947be19 DH	292	{
	293	struct memory_block *mem =
	294	container_of(dev, struct memory_block, sysdev);
	295	return sprintf(buf, "%d\n", mem->phys_device);
	296	}
	297
	298	static SYSDEV_ATTR(phys_index, 0444, show_mem_phys_index, NULL);
	299	static SYSDEV_ATTR(state, 0644, show_mem_state, store_mem_state);
	300	static SYSDEV_ATTR(phys_device, 0444, show_phys_device, NULL);
5c755e9f	301	static SYSDEV_ATTR(removable, 0444, show_mem_removable, NULL);
3947be19 DH	302
	303	#define mem_create_simple_file(mem, attr_name) \
	304	sysdev_create_file(&mem->sysdev, &attr_##attr_name)
	305	#define mem_remove_simple_file(mem, attr_name) \
	306	sysdev_remove_file(&mem->sysdev, &attr_##attr_name)
	307
	308	/*
	309	* Block size attribute stuff
	310	*/
	311	static ssize_t
8ff410da WF	312	print_block_size(struct sysdev_class *class,
	313	struct sysdev_class_attribute *class_attr,
	314	char *buf)
3947be19	315	{
ba168fc3	316	return sprintf(buf, "%#lx\n", (unsigned long)PAGES_PER_SECTION * PAGE_SIZE);
3947be19 DH	317	}
3947be19 DH	318
8ff410da	319	static SYSDEV_CLASS_ATTR(block_size_bytes, 0444, print_block_size, NULL);
3947be19 DH	320
	321	static int block_size_init(void)
	322	{
28ec24e2	323	return sysfs_create_file(&memory_sysdev_class.kset.kobj,
8ff410da	324	&attr_block_size_bytes.attr);
3947be19 DH	325	}
	326
	327	/*
	328	* Some architectures will have custom drivers to do this, and
	329	* will not need to do it from userspace. The fake hot-add code
	330	* as well as ppc64 will do all of their discovery in userspace
	331	* and will require this interface.
	332	*/
	333	#ifdef CONFIG_ARCH_MEMORY_PROBE
	334	static ssize_t
8ff410da WF	335	memory_probe_store(struct sysdev_class *class,
	336	struct sysdev_class_attribute *class_attr,
	337	const char *buf, size_t count)
3947be19 DH	338	{
3947be19 DH	339	u64 phys_addr;
bc02af93	340	int nid;
3947be19 DH	341	int ret;
	342
	343	phys_addr = simple_strtoull(buf, NULL, 0);
	344
bc02af93 YG	345	nid = memory_add_physaddr_to_nid(phys_addr);
bc02af93 YG	346	ret = add_memory(nid, phys_addr, PAGES_PER_SECTION << PAGE_SHIFT);
3947be19 DH	347
	348	if (ret)
	349	count = ret;
	350
	351	return count;
	352	}
8ff410da	353	static SYSDEV_CLASS_ATTR(probe, S_IWUSR, NULL, memory_probe_store);
3947be19 DH	354
	355	static int memory_probe_init(void)
	356	{
28ec24e2	357	return sysfs_create_file(&memory_sysdev_class.kset.kobj,
8ff410da	358	&attr_probe.attr);
3947be19 DH	359	}
3947be19 DH	360	#else
28ec24e2 AM	361	static inline int memory_probe_init(void)
	362	{
	363	return 0;
	364	}
3947be19 DH	365	#endif
3947be19 DH	366
facb6011 AK	367	#ifdef CONFIG_MEMORY_FAILURE
	368	/*
	369	* Support for offlining pages of memory
	370	*/
	371
	372	/* Soft offline a page */
	373	static ssize_t
8ff410da WF	374	store_soft_offline_page(struct sysdev_class *class,
	375	struct sysdev_class_attribute *class_attr,
	376	const char *buf, size_t count)
facb6011 AK	377	{
	378	int ret;
	379	u64 pfn;
	380	if (!capable(CAP_SYS_ADMIN))
	381	return -EPERM;
	382	if (strict_strtoull(buf, 0, &pfn) < 0)
	383	return -EINVAL;
	384	pfn >>= PAGE_SHIFT;
	385	if (!pfn_valid(pfn))
	386	return -ENXIO;
	387	ret = soft_offline_page(pfn_to_page(pfn), 0);
	388	return ret == 0 ? count : ret;
	389	}
	390
	391	/* Forcibly offline a page, including killing processes. */
	392	static ssize_t
8ff410da WF	393	store_hard_offline_page(struct sysdev_class *class,
	394	struct sysdev_class_attribute *class_attr,
	395	const char *buf, size_t count)
facb6011 AK	396	{
	397	int ret;
	398	u64 pfn;
	399	if (!capable(CAP_SYS_ADMIN))
	400	return -EPERM;
	401	if (strict_strtoull(buf, 0, &pfn) < 0)
	402	return -EINVAL;
	403	pfn >>= PAGE_SHIFT;
	404	ret = __memory_failure(pfn, 0, 0);
	405	return ret ? ret : count;
	406	}
	407
8ff410da WF	408	static SYSDEV_CLASS_ATTR(soft_offline_page, 0644, NULL, store_soft_offline_page);
8ff410da WF	409	static SYSDEV_CLASS_ATTR(hard_offline_page, 0644, NULL, store_hard_offline_page);
facb6011 AK	410
	411	static __init int memory_fail_init(void)
	412	{
	413	int err;
	414
	415	err = sysfs_create_file(&memory_sysdev_class.kset.kobj,
8ff410da	416	&attr_soft_offline_page.attr);
facb6011 AK	417	if (!err)
facb6011 AK	418	err = sysfs_create_file(&memory_sysdev_class.kset.kobj,
8ff410da	419	&attr_hard_offline_page.attr);
facb6011 AK	420	return err;
	421	}
	422	#else
	423	static inline int memory_fail_init(void)
	424	{
	425	return 0;
	426	}
	427	#endif
	428
3947be19 DH	429	/*
	430	* Note that phys_device is optional. It is here to allow for
	431	* differentiation between which physical devices each
	432	* section belongs to...
	433	*/
	434
c04fc586 GH	435	static int add_memory_block(int nid, struct mem_section *section,
	436	unsigned long state, int phys_device,
	437	enum mem_add_context context)
3947be19	438	{
0b0acbec	439	struct memory_block mem = kzalloc(sizeof(mem), GFP_KERNEL);
3947be19 DH	440	int ret = 0;
	441
	442	if (!mem)
	443	return -ENOMEM;
	444
3947be19 DH	445	mem->phys_index = __section_nr(section);
3947be19 DH	446	mem->state = state;
da19cbcf	447	mutex_init(&mem->state_mutex);
3947be19 DH	448	mem->phys_device = phys_device;
3947be19 DH	449
00a41db5	450	ret = register_memory(mem, section);
3947be19 DH	451	if (!ret)
	452	ret = mem_create_simple_file(mem, phys_index);
	453	if (!ret)
	454	ret = mem_create_simple_file(mem, state);
	455	if (!ret)
	456	ret = mem_create_simple_file(mem, phys_device);
5c755e9f BP	457	if (!ret)
5c755e9f BP	458	ret = mem_create_simple_file(mem, removable);
c04fc586 GH	459	if (!ret) {
	460	if (context == HOTPLUG)
	461	ret = register_mem_sect_under_node(mem, nid);
	462	}
3947be19 DH	463
	464	return ret;
	465	}
	466
	467	/*
	468	* For now, we have a linear search to go find the appropriate
	469	* memory_block corresponding to a particular phys_index. If
	470	* this gets to be a real problem, we can always use a radix
	471	* tree or something here.
	472	*
	473	* This could be made generic for all sysdev classes.
	474	*/
c04fc586	475	struct memory_block find_memory_block(struct mem_section section)
3947be19 DH	476	{
	477	struct kobject *kobj;
	478	struct sys_device *sysdev;
	479	struct memory_block *mem;
	480	char name[sizeof(MEMORY_CLASS_NAME) + 9 + 1];
	481
	482	/*
	483	* This only works because we know that section == sysdev->id
	484	* slightly redundant with sysdev_register()
	485	*/
	486	sprintf(&name[0], "%s%d", MEMORY_CLASS_NAME, __section_nr(section));
	487
	488	kobj = kset_find_obj(&memory_sysdev_class.kset, name);
	489	if (!kobj)
	490	return NULL;
	491
	492	sysdev = container_of(kobj, struct sys_device, kobj);
	493	mem = container_of(sysdev, struct memory_block, sysdev);
	494
	495	return mem;
	496	}
	497
	498	int remove_memory_block(unsigned long node_id, struct mem_section *section,
	499	int phys_device)
	500	{
	501	struct memory_block *mem;
	502
	503	mem = find_memory_block(section);
c04fc586	504	unregister_mem_sect_under_nodes(mem);
3947be19 DH	505	mem_remove_simple_file(mem, phys_index);
	506	mem_remove_simple_file(mem, state);
	507	mem_remove_simple_file(mem, phys_device);
5c755e9f	508	mem_remove_simple_file(mem, removable);
00a41db5	509	unregister_memory(mem, section);
3947be19 DH	510
	511	return 0;
	512	}
	513
	514	/*
	515	* need an interface for the VM to add new memory regions,
	516	* but without onlining it.
	517	*/
c04fc586	518	int register_new_memory(int nid, struct mem_section *section)
3947be19	519	{
c04fc586	520	return add_memory_block(nid, section, MEM_OFFLINE, 0, HOTPLUG);
3947be19 DH	521	}
	522
	523	int unregister_memory_section(struct mem_section *section)
	524	{
540557b9	525	if (!present_section(section))
3947be19 DH	526	return -EINVAL;
	527
	528	return remove_memory_block(0, section, 0);
	529	}
	530
	531	/*
	532	* Initialize the sysfs support for memory devices...
	533	*/
	534	int __init memory_dev_init(void)
	535	{
	536	unsigned int i;
	537	int ret;
28ec24e2	538	int err;
3947be19	539
312c004d	540	memory_sysdev_class.kset.uevent_ops = &memory_uevent_ops;
3947be19	541	ret = sysdev_class_register(&memory_sysdev_class);
28ec24e2 AM	542	if (ret)
28ec24e2 AM	543	goto out;
3947be19 DH	544
	545	/*
	546	* Create entries for memory sections that were found
	547	* during boot and have been initialized
	548	*/
	549	for (i = 0; i < NR_MEM_SECTIONS; i++) {
540557b9	550	if (!present_section_nr(i))
3947be19	551	continue;
c04fc586 GH	552	err = add_memory_block(0, __nr_to_section(i), MEM_ONLINE,
c04fc586 GH	553	0, BOOT);
28ec24e2 AM	554	if (!ret)
28ec24e2 AM	555	ret = err;
3947be19 DH	556	}
3947be19 DH	557
28ec24e2	558	err = memory_probe_init();
facb6011 AK	559	if (!ret)
	560	ret = err;
	561	err = memory_fail_init();
28ec24e2 AM	562	if (!ret)
	563	ret = err;
	564	err = block_size_init();
	565	if (!ret)
	566	ret = err;
	567	out:
	568	if (ret)
2b3a302a	569	printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
3947be19 DH	570	return ret;
3947be19 DH	571	}