[linux-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 "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 struct memory_dev_type *dax_slowmem_type;
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;
	int i, rc, mapped = 0;
	int numa_node;

	/*
	 * 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;
	}

	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, dax_slowmem_type);

	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, 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;

		/*
		 * 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_NID_IS_MGID);

		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, dax_slowmem_type);
	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, dax_slowmem_type);
	}
}
#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;

	dax_slowmem_type = alloc_memory_type(MEMTIER_DEFAULT_DAX_ADISTANCE);
	if (IS_ERR(dax_slowmem_type)) {
		rc = PTR_ERR(dax_slowmem_type);
		goto err_dax_slowmem_type;
	}

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

	return rc;

error_dax_driver:
	destroy_memory_type(dax_slowmem_type);
err_dax_slowmem_type:
	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);
	destroy_memory_type(dax_slowmem_type);
}

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