[linux-2.6-block.git] / drivers / dax / kmem.c

// SPDX-License-Identifier: GPL-2.0
/* Copyright(c) 2016-2019 Intel Corporation. All rights reserved. */
#include <linux/memremap.h>
#include <linux/pagemap.h>
#include <linux/memory.h>
#include <linux/module.h>
#include <linux/device.h>
#include <linux/pfn_t.h>
#include <linux/slab.h>
#include <linux/dax.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/memory-tiers.h>
#include <linux/memory_hotplug.h>
#include "dax-private.h"
#include "bus.h"

/*
 * Default abstract distance assigned to the NUMA node onlined
 * by DAX/kmem if the low level platform driver didn't initialize
 * one for this NUMA node.
 */
#define MEMTIER_DEFAULT_DAX_ADISTANCE	(MEMTIER_ADISTANCE_DRAM * 5)

/* Memory resource name used for add_memory_driver_managed(). */
static const char *kmem_name;
/* Set if any memory will remain added when the driver will be unloaded. */
static bool any_hotremove_failed;

static int dax_kmem_range(struct dev_dax *dev_dax, int i, struct range *r)
{
	struct dev_dax_range *dax_range = &dev_dax->ranges[i];
	struct range *range = &dax_range->range;

	/* memory-block align the hotplug range */
	r->start = ALIGN(range->start, memory_block_size_bytes());
	r->end = ALIGN_DOWN(range->end + 1, memory_block_size_bytes()) - 1;
	if (r->start >= r->end) {
		r->start = range->start;
		r->end = range->end;
		return -ENOSPC;
	}
	return 0;
}

struct dax_kmem_data {
	const char *res_name;
	int mgid;
	struct resource *res[];
};

static DEFINE_MUTEX(kmem_memory_type_lock);
static LIST_HEAD(kmem_memory_types);

static struct memory_dev_type *kmem_find_alloc_memory_type(int adist)
{
	bool found = false;
	struct memory_dev_type *mtype;

	mutex_lock(&kmem_memory_type_lock);
	list_for_each_entry(mtype, &kmem_memory_types, list) {
		if (mtype->adistance == adist) {
			found = true;
			break;
		}
	}
	if (!found) {
		mtype = alloc_memory_type(adist);
		if (!IS_ERR(mtype))
			list_add(&mtype->list, &kmem_memory_types);
	}
	mutex_unlock(&kmem_memory_type_lock);

	return mtype;
}

static void kmem_put_memory_types(void)
{
	struct memory_dev_type *mtype, *mtn;

	mutex_lock(&kmem_memory_type_lock);
	list_for_each_entry_safe(mtype, mtn, &kmem_memory_types, list) {
		list_del(&mtype->list);
		put_memory_type(mtype);
	}
	mutex_unlock(&kmem_memory_type_lock);
}

static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
{
	struct device *dev = &dev_dax->dev;
	unsigned long total_len = 0;
	struct dax_kmem_data *data;
	struct memory_dev_type *mtype;
	int i, rc, mapped = 0;
	mhp_t mhp_flags;
	int numa_node;
	int adist = MEMTIER_DEFAULT_DAX_ADISTANCE;

	/*
	 * Ensure good NUMA information for the persistent memory.
	 * Without this check, there is a risk that slow memory
	 * could be mixed in a node with faster memory, causing
	 * unavoidable performance issues.
	 */
	numa_node = dev_dax->target_node;
	if (numa_node < 0) {
		dev_warn(dev, "rejecting DAX region with invalid node: %d\n",
				numa_node);
		return -EINVAL;
	}

	mt_calc_adistance(numa_node, &adist);
	mtype = kmem_find_alloc_memory_type(adist);
	if (IS_ERR(mtype))
		return PTR_ERR(mtype);

	for (i = 0; i < dev_dax->nr_range; i++) {
		struct range range;

		rc = dax_kmem_range(dev_dax, i, &range);
		if (rc) {
			dev_info(dev, "mapping%d: %#llx-%#llx too small after alignment\n",
					i, range.start, range.end);
			continue;
		}
		total_len += range_len(&range);
	}

	if (!total_len) {
		dev_warn(dev, "rejecting DAX region without any memory after alignment\n");
		return -EINVAL;
	}

	init_node_memory_type(numa_node, mtype);

	rc = -ENOMEM;
	data = kzalloc(struct_size(data, res, dev_dax->nr_range), GFP_KERNEL);
	if (!data)
		goto err_dax_kmem_data;

	data->res_name = kstrdup(dev_name(dev), GFP_KERNEL);
	if (!data->res_name)
		goto err_res_name;

	rc = memory_group_register_static(numa_node, PFN_UP(total_len));
	if (rc < 0)
		goto err_reg_mgid;
	data->mgid = rc;

	for (i = 0; i < dev_dax->nr_range; i++) {
		struct resource *res;
		struct range range;

		rc = dax_kmem_range(dev_dax, i, &range);
		if (rc)
			continue;

		/* Region is permanently reserved if hotremove fails. */
		res = request_mem_region(range.start, range_len(&range), data->res_name);
		if (!res) {
			dev_warn(dev, "mapping%d: %#llx-%#llx could not reserve region\n",
					i, range.start, range.end);
			/*
			 * Once some memory has been onlined we can't
			 * assume that it can be un-onlined safely.
			 */
			if (mapped)
				continue;
			rc = -EBUSY;
			goto err_request_mem;
		}
		data->res[i] = res;

		/*
		 * Set flags appropriate for System RAM.  Leave ..._BUSY clear
		 * so that add_memory() can add a child resource.  Do not
		 * inherit flags from the parent since it may set new flags
		 * unknown to us that will break add_memory() below.
		 */
		res->flags = IORESOURCE_SYSTEM_RAM;

		mhp_flags = MHP_NID_IS_MGID;
		if (dev_dax->memmap_on_memory)
			mhp_flags |= MHP_MEMMAP_ON_MEMORY;

		/*
		 * Ensure that future kexec'd kernels will not treat
		 * this as RAM automatically.
		 */
		rc = add_memory_driver_managed(data->mgid, range.start,
				range_len(&range), kmem_name, mhp_flags);

		if (rc) {
			dev_warn(dev, "mapping%d: %#llx-%#llx memory add failed\n",
					i, range.start, range.end);
			remove_resource(res);
			kfree(res);
			data->res[i] = NULL;
			if (mapped)
				continue;
			goto err_request_mem;
		}
		mapped++;
	}

	dev_set_drvdata(dev, data);

	return 0;

err_request_mem:
	memory_group_unregister(data->mgid);
err_reg_mgid:
	kfree(data->res_name);
err_res_name:
	kfree(data);
err_dax_kmem_data:
	clear_node_memory_type(numa_node, mtype);
	return rc;
}

#ifdef CONFIG_MEMORY_HOTREMOVE
static void dev_dax_kmem_remove(struct dev_dax *dev_dax)
{
	int i, success = 0;
	int node = dev_dax->target_node;
	struct device *dev = &dev_dax->dev;
	struct dax_kmem_data *data = dev_get_drvdata(dev);

	/*
	 * We have one shot for removing memory, if some memory blocks were not
	 * offline prior to calling this function remove_memory() will fail, and
	 * there is no way to hotremove this memory until reboot because device
	 * unbind will succeed even if we return failure.
	 */
	for (i = 0; i < dev_dax->nr_range; i++) {
		struct range range;
		int rc;

		rc = dax_kmem_range(dev_dax, i, &range);
		if (rc)
			continue;

		rc = remove_memory(range.start, range_len(&range));
		if (rc == 0) {
			remove_resource(data->res[i]);
			kfree(data->res[i]);
			data->res[i] = NULL;
			success++;
			continue;
		}
		any_hotremove_failed = true;
		dev_err(dev,
			"mapping%d: %#llx-%#llx cannot be hotremoved until the next reboot\n",
				i, range.start, range.end);
	}

	if (success >= dev_dax->nr_range) {
		memory_group_unregister(data->mgid);
		kfree(data->res_name);
		kfree(data);
		dev_set_drvdata(dev, NULL);
		/*
		 * Clear the memtype association on successful unplug.
		 * If not, we have memory blocks left which can be
		 * offlined/onlined later. We need to keep memory_dev_type
		 * for that. This implies this reference will be around
		 * till next reboot.
		 */
		clear_node_memory_type(node, NULL);
	}
}
#else
static void dev_dax_kmem_remove(struct dev_dax *dev_dax)
{
	/*
	 * Without hotremove purposely leak the request_mem_region() for the
	 * device-dax range and return '0' to ->remove() attempts. The removal
	 * of the device from the driver always succeeds, but the region is
	 * permanently pinned as reserved by the unreleased
	 * request_mem_region().
	 */
	any_hotremove_failed = true;
}
#endif /* CONFIG_MEMORY_HOTREMOVE */

static struct dax_device_driver device_dax_kmem_driver = {
	.probe = dev_dax_kmem_probe,
	.remove = dev_dax_kmem_remove,
	.type = DAXDRV_KMEM_TYPE,
};

static int __init dax_kmem_init(void)
{
	int rc;

	/* Resource name is permanently allocated if any hotremove fails. */
	kmem_name = kstrdup_const("System RAM (kmem)", GFP_KERNEL);
	if (!kmem_name)
		return -ENOMEM;

	rc = dax_driver_register(&device_dax_kmem_driver);
	if (rc)
		goto error_dax_driver;

	return rc;

error_dax_driver:
	kmem_put_memory_types();
	kfree_const(kmem_name);
	return rc;
}

static void __exit dax_kmem_exit(void)
{
	dax_driver_unregister(&device_dax_kmem_driver);
	if (!any_hotremove_failed)
		kfree_const(kmem_name);
	kmem_put_memory_types();
}

MODULE_AUTHOR("Intel Corporation");
MODULE_LICENSE("GPL v2");
module_init(dax_kmem_init);
module_exit(dax_kmem_exit);
MODULE_ALIAS_DAX_DEVICE(0);
Commit	Line	Data
c221c0b0 DH	1	// SPDX-License-Identifier: GPL-2.0
	2	/* Copyright(c) 2016-2019 Intel Corporation. All rights reserved. */
	3	#include <linux/memremap.h>
	4	#include <linux/pagemap.h>
	5	#include <linux/memory.h>
	6	#include <linux/module.h>
	7	#include <linux/device.h>
	8	#include <linux/pfn_t.h>
	9	#include <linux/slab.h>
	10	#include <linux/dax.h>
	11	#include <linux/fs.h>
	12	#include <linux/mm.h>
	13	#include <linux/mman.h>
7b88bda3	14	#include <linux/memory-tiers.h>
4eca0ef4	15	#include <linux/memory_hotplug.h>
c221c0b0 DH	16	#include "dax-private.h"
	17	#include "bus.h"
	18
7b88bda3 AK	19	/*
	20	* Default abstract distance assigned to the NUMA node onlined
	21	* by DAX/kmem if the low level platform driver didn't initialize
	22	* one for this NUMA node.
	23	*/
	24	#define MEMTIER_DEFAULT_DAX_ADISTANCE (MEMTIER_ADISTANCE_DRAM * 5)
	25
8a725e46 DH	26	/* Memory resource name used for add_memory_driver_managed(). */
	27	static const char *kmem_name;
	28	/* Set if any memory will remain added when the driver will be unloaded. */
	29	static bool any_hotremove_failed;
	30
60e93dc0	31	static int dax_kmem_range(struct dev_dax dev_dax, int i, struct range r)
59bc8d10	32	{
60e93dc0 DW	33	struct dev_dax_range *dax_range = &dev_dax->ranges[i];
60e93dc0 DW	34	struct range *range = &dax_range->range;
59bc8d10 DW	35
59bc8d10 DW	36	/* memory-block align the hotplug range */
60e93dc0 DW	37	r->start = ALIGN(range->start, memory_block_size_bytes());
	38	r->end = ALIGN_DOWN(range->end + 1, memory_block_size_bytes()) - 1;
	39	if (r->start >= r->end) {
	40	r->start = range->start;
	41	r->end = range->end;
	42	return -ENOSPC;
	43	}
	44	return 0;
59bc8d10 DW	45	}
59bc8d10 DW	46
a455aa72 DW	47	struct dax_kmem_data {
a455aa72 DW	48	const char *res_name;
eedf634a	49	int mgid;
a455aa72 DW	50	struct resource *res[];
	51	};
	52
6bc2cfdf HY	53	static DEFINE_MUTEX(kmem_memory_type_lock);
	54	static LIST_HEAD(kmem_memory_types);
	55
	56	static struct memory_dev_type *kmem_find_alloc_memory_type(int adist)
	57	{
	58	bool found = false;
	59	struct memory_dev_type *mtype;
	60
	61	mutex_lock(&kmem_memory_type_lock);
	62	list_for_each_entry(mtype, &kmem_memory_types, list) {
	63	if (mtype->adistance == adist) {
	64	found = true;
	65	break;
	66	}
	67	}
	68	if (!found) {
	69	mtype = alloc_memory_type(adist);
	70	if (!IS_ERR(mtype))
	71	list_add(&mtype->list, &kmem_memory_types);
	72	}
	73	mutex_unlock(&kmem_memory_type_lock);
	74
	75	return mtype;
	76	}
	77
	78	static void kmem_put_memory_types(void)
	79	{
	80	struct memory_dev_type mtype, mtn;
	81
	82	mutex_lock(&kmem_memory_type_lock);
	83	list_for_each_entry_safe(mtype, mtn, &kmem_memory_types, list) {
	84	list_del(&mtype->list);
	85	put_memory_type(mtype);
	86	}
	87	mutex_unlock(&kmem_memory_type_lock);
	88	}
	89
f11cf813	90	static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
c221c0b0	91	{
f11cf813	92	struct device *dev = &dev_dax->dev;
eedf634a	93	unsigned long total_len = 0;
a455aa72	94	struct dax_kmem_data *data;
6bc2cfdf	95	struct memory_dev_type *mtype;
eedf634a	96	int i, rc, mapped = 0;
4eca0ef4	97	mhp_t mhp_flags;
c221c0b0	98	int numa_node;
6bc2cfdf	99	int adist = MEMTIER_DEFAULT_DAX_ADISTANCE;
c221c0b0 DH	100
	101	/*
	102	* Ensure good NUMA information for the persistent memory.
	103	* Without this check, there is a risk that slow memory
	104	* could be mixed in a node with faster memory, causing
	105	* unavoidable performance issues.
	106	*/
	107	numa_node = dev_dax->target_node;
	108	if (numa_node < 0) {
f5516ec5 DW	109	dev_warn(dev, "rejecting DAX region with invalid node: %d\n",
f5516ec5 DW	110	numa_node);
c221c0b0 DH	111	return -EINVAL;
	112	}
	113
6bc2cfdf HY	114	mt_calc_adistance(numa_node, &adist);
	115	mtype = kmem_find_alloc_memory_type(adist);
	116	if (IS_ERR(mtype))
	117	return PTR_ERR(mtype);
	118
eedf634a DH	119	for (i = 0; i < dev_dax->nr_range; i++) {
	120	struct range range;
	121
	122	rc = dax_kmem_range(dev_dax, i, &range);
	123	if (rc) {
	124	dev_info(dev, "mapping%d: %#llx-%#llx too small after alignment\n",
	125	i, range.start, range.end);
	126	continue;
	127	}
	128	total_len += range_len(&range);
	129	}
	130
	131	if (!total_len) {
	132	dev_warn(dev, "rejecting DAX region without any memory after alignment\n");
	133	return -EINVAL;
	134	}
	135
6bc2cfdf	136	init_node_memory_type(numa_node, mtype);
7b88bda3 AK	137
7b88bda3 AK	138	rc = -ENOMEM;
7d18dd75	139	data = kzalloc(struct_size(data, res, dev_dax->nr_range), GFP_KERNEL);
a455aa72	140	if (!data)
7b88bda3	141	goto err_dax_kmem_data;
60858c00	142
a455aa72 DW	143	data->res_name = kstrdup(dev_name(dev), GFP_KERNEL);
	144	if (!data->res_name)
	145	goto err_res_name;
	146
46e66dab	147	rc = memory_group_register_static(numa_node, PFN_UP(total_len));
eedf634a DH	148	if (rc < 0)
	149	goto err_reg_mgid;
	150	data->mgid = rc;
	151
60e93dc0 DW	152	for (i = 0; i < dev_dax->nr_range; i++) {
	153	struct resource *res;
	154	struct range range;
c221c0b0	155
60e93dc0	156	rc = dax_kmem_range(dev_dax, i, &range);
eedf634a	157	if (rc)
60e93dc0	158	continue;
c221c0b0	159
60e93dc0	160	/* Region is permanently reserved if hotremove fails. */
a455aa72	161	res = request_mem_region(range.start, range_len(&range), data->res_name);
60e93dc0 DW	162	if (!res) {
	163	dev_warn(dev, "mapping%d: %#llx-%#llx could not reserve region\n",
	164	i, range.start, range.end);
	165	/*
	166	* Once some memory has been onlined we can't
	167	* assume that it can be un-onlined safely.
	168	*/
	169	if (mapped)
	170	continue;
a455aa72 DW	171	rc = -EBUSY;
a455aa72 DW	172	goto err_request_mem;
60e93dc0	173	}
a455aa72	174	data->res[i] = res;
60e93dc0 DW	175
	176	/*
	177	* Set flags appropriate for System RAM. Leave ..._BUSY clear
	178	* so that add_memory() can add a child resource. Do not
	179	* inherit flags from the parent since it may set new flags
	180	* unknown to us that will break add_memory() below.
	181	*/
	182	res->flags = IORESOURCE_SYSTEM_RAM;
	183
4eca0ef4 VV	184	mhp_flags = MHP_NID_IS_MGID;
	185	if (dev_dax->memmap_on_memory)
	186	mhp_flags \|= MHP_MEMMAP_ON_MEMORY;
	187
60e93dc0 DW	188	/*
	189	* Ensure that future kexec'd kernels will not treat
	190	* this as RAM automatically.
	191	*/
eedf634a	192	rc = add_memory_driver_managed(data->mgid, range.start,
4eca0ef4	193	range_len(&range), kmem_name, mhp_flags);
60e93dc0 DW	194
	195	if (rc) {
	196	dev_warn(dev, "mapping%d: %#llx-%#llx memory add failed\n",
	197	i, range.start, range.end);
e686c325	198	remove_resource(res);
a455aa72 DW	199	kfree(res);
a455aa72 DW	200	data->res[i] = NULL;
60e93dc0 DW	201	if (mapped)
60e93dc0 DW	202	continue;
a455aa72	203	goto err_request_mem;
60e93dc0 DW	204	}
60e93dc0 DW	205	mapped++;
31e4ca92	206	}
7e6b431a	207
a455aa72	208	dev_set_drvdata(dev, data);
c221c0b0 DH	209
c221c0b0 DH	210	return 0;
a455aa72 DW	211
a455aa72 DW	212	err_request_mem:
eedf634a DH	213	memory_group_unregister(data->mgid);
eedf634a DH	214	err_reg_mgid:
a455aa72 DW	215	kfree(data->res_name);
	216	err_res_name:
	217	kfree(data);
7b88bda3	218	err_dax_kmem_data:
6bc2cfdf	219	clear_node_memory_type(numa_node, mtype);
a455aa72	220	return rc;
c221c0b0 DH	221	}
c221c0b0 DH	222
9f960da7	223	#ifdef CONFIG_MEMORY_HOTREMOVE
0d519e0d	224	static void dev_dax_kmem_remove(struct dev_dax *dev_dax)
9f960da7	225	{
60e93dc0	226	int i, success = 0;
7b88bda3	227	int node = dev_dax->target_node;
f11cf813	228	struct device *dev = &dev_dax->dev;
a455aa72	229	struct dax_kmem_data *data = dev_get_drvdata(dev);
9f960da7 PT	230
	231	/*
	232	* We have one shot for removing memory, if some memory blocks were not
	233	* offline prior to calling this function remove_memory() will fail, and
	234	* there is no way to hotremove this memory until reboot because device
	235	* unbind will succeed even if we return failure.
	236	*/
60e93dc0 DW	237	for (i = 0; i < dev_dax->nr_range; i++) {
	238	struct range range;
	239	int rc;
	240
	241	rc = dax_kmem_range(dev_dax, i, &range);
	242	if (rc)
	243	continue;
	244
e1c158e4	245	rc = remove_memory(range.start, range_len(&range));
60e93dc0	246	if (rc == 0) {
e686c325	247	remove_resource(data->res[i]);
a455aa72 DW	248	kfree(data->res[i]);
a455aa72 DW	249	data->res[i] = NULL;
60e93dc0 DW	250	success++;
	251	continue;
	252	}
8a725e46	253	any_hotremove_failed = true;
60e93dc0 DW	254	dev_err(dev,
	255	"mapping%d: %#llx-%#llx cannot be hotremoved until the next reboot\n",
	256	i, range.start, range.end);
9f960da7 PT	257	}
9f960da7 PT	258
60e93dc0	259	if (success >= dev_dax->nr_range) {
eedf634a	260	memory_group_unregister(data->mgid);
a455aa72 DW	261	kfree(data->res_name);
a455aa72 DW	262	kfree(data);
60e93dc0	263	dev_set_drvdata(dev, NULL);
7b88bda3 AK	264	/*
	265	* Clear the memtype association on successful unplug.
	266	* If not, we have memory blocks left which can be
	267	* offlined/onlined later. We need to keep memory_dev_type
	268	* for that. This implies this reference will be around
	269	* till next reboot.
	270	*/
6bc2cfdf	271	clear_node_memory_type(node, NULL);
60e93dc0	272	}
9f960da7 PT	273	}
9f960da7 PT	274	#else
0d519e0d	275	static void dev_dax_kmem_remove(struct dev_dax *dev_dax)
c221c0b0 DH	276	{
c221c0b0 DH	277	/*
9f960da7 PT	278	* Without hotremove purposely leak the request_mem_region() for the
	279	* device-dax range and return '0' to ->remove() attempts. The removal
	280	* of the device from the driver always succeeds, but the region is
	281	* permanently pinned as reserved by the unreleased
c221c0b0 DH	282	* request_mem_region().
c221c0b0 DH	283	*/
8a725e46	284	any_hotremove_failed = true;
c221c0b0	285	}
9f960da7	286	#endif /* CONFIG_MEMORY_HOTREMOVE */
c221c0b0 DH	287
c221c0b0 DH	288	static struct dax_device_driver device_dax_kmem_driver = {
f11cf813 DW	289	.probe = dev_dax_kmem_probe,
f11cf813 DW	290	.remove = dev_dax_kmem_remove,
e9ee9fe3	291	.type = DAXDRV_KMEM_TYPE,
c221c0b0 DH	292	};
	293
	294	static int __init dax_kmem_init(void)
	295	{
8a725e46 DH	296	int rc;
	297
	298	/* Resource name is permanently allocated if any hotremove fails. */
	299	kmem_name = kstrdup_const("System RAM (kmem)", GFP_KERNEL);
	300	if (!kmem_name)
	301	return -ENOMEM;
	302
	303	rc = dax_driver_register(&device_dax_kmem_driver);
	304	if (rc)
7b88bda3 AK	305	goto error_dax_driver;
	306
	307	return rc;
	308
	309	error_dax_driver:
6bc2cfdf	310	kmem_put_memory_types();
7b88bda3	311	kfree_const(kmem_name);
8a725e46	312	return rc;
c221c0b0 DH	313	}
	314
	315	static void __exit dax_kmem_exit(void)
	316	{
	317	dax_driver_unregister(&device_dax_kmem_driver);
8a725e46 DH	318	if (!any_hotremove_failed)
8a725e46 DH	319	kfree_const(kmem_name);
6bc2cfdf	320	kmem_put_memory_types();
c221c0b0 DH	321	}
	322
	323	MODULE_AUTHOR("Intel Corporation");
	324	MODULE_LICENSE("GPL v2");
	325	module_init(dax_kmem_init);
	326	module_exit(dax_kmem_exit);
	327	MODULE_ALIAS_DAX_DEVICE(0);