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> 14 #include "dax-private.h" 15 #include "bus.h" 16 17 /* Memory resource name used for add_memory_driver_managed(). */ 18 static const char *kmem_name; 19 /* Set if any memory will remain added when the driver will be unloaded. */ 20 static bool any_hotremove_failed; 21 22 static struct range dax_kmem_range(struct dev_dax *dev_dax) 23 { 24 struct range range; 25 26 /* memory-block align the hotplug range */ 27 range.start = ALIGN(dev_dax->range.start, memory_block_size_bytes()); 28 range.end = ALIGN_DOWN(dev_dax->range.end + 1, memory_block_size_bytes()) - 1; 29 return range; 30 } 31 32 int dev_dax_kmem_probe(struct device *dev) 33 { 34 struct dev_dax *dev_dax = to_dev_dax(dev); 35 struct range range = dax_kmem_range(dev_dax); 36 struct resource *res; 37 char *res_name; 38 int numa_node; 39 int rc; 40 41 /* 42 * Ensure good NUMA information for the persistent memory. 43 * Without this check, there is a risk that slow memory 44 * could be mixed in a node with faster memory, causing 45 * unavoidable performance issues. 46 */ 47 numa_node = dev_dax->target_node; 48 if (numa_node < 0) { 49 dev_warn(dev, "rejecting DAX region with invalid node: %d\n", 50 numa_node); 51 return -EINVAL; 52 } 53 54 res_name = kstrdup(dev_name(dev), GFP_KERNEL); 55 if (!res_name) 56 return -ENOMEM; 57 58 /* Region is permanently reserved if hotremove fails. */ 59 res = request_mem_region(range.start, range_len(&range), res_name); 60 if (!res) { 61 dev_warn(dev, "could not reserve region [%#llx-%#llx]\n", range.start, range.end); 62 kfree(res_name); 63 return -EBUSY; 64 } 65 66 /* 67 * Set flags appropriate for System RAM. Leave ..._BUSY clear 68 * so that add_memory() can add a child resource. Do not 69 * inherit flags from the parent since it may set new flags 70 * unknown to us that will break add_memory() below. 71 */ 72 res->flags = IORESOURCE_SYSTEM_RAM; 73 74 /* 75 * Ensure that future kexec'd kernels will not treat this as RAM 76 * automatically. 77 */ 78 rc = add_memory_driver_managed(numa_node, range.start, range_len(&range), kmem_name); 79 if (rc) { 80 release_mem_region(range.start, range_len(&range)); 81 kfree(res_name); 82 return rc; 83 } 84 85 dev_set_drvdata(dev, res_name); 86 87 return 0; 88 } 89 90 #ifdef CONFIG_MEMORY_HOTREMOVE 91 static int dev_dax_kmem_remove(struct device *dev) 92 { 93 struct dev_dax *dev_dax = to_dev_dax(dev); 94 struct range range = dax_kmem_range(dev_dax); 95 const char *res_name = dev_get_drvdata(dev); 96 int rc; 97 98 /* 99 * We have one shot for removing memory, if some memory blocks were not 100 * offline prior to calling this function remove_memory() will fail, and 101 * there is no way to hotremove this memory until reboot because device 102 * unbind will succeed even if we return failure. 103 */ 104 rc = remove_memory(dev_dax->target_node, range.start, range_len(&range)); 105 if (rc) { 106 any_hotremove_failed = true; 107 dev_err(dev, "%#llx-%#llx cannot be hotremoved until the next reboot\n", 108 range.start, range.end); 109 return rc; 110 } 111 112 /* Release and free dax resources */ 113 release_mem_region(range.start, range_len(&range)); 114 kfree(res_name); 115 116 return 0; 117 } 118 #else 119 static int dev_dax_kmem_remove(struct device *dev) 120 { 121 /* 122 * Without hotremove purposely leak the request_mem_region() for the 123 * device-dax range and return '0' to ->remove() attempts. The removal 124 * of the device from the driver always succeeds, but the region is 125 * permanently pinned as reserved by the unreleased 126 * request_mem_region(). 127 */ 128 any_hotremove_failed = true; 129 return 0; 130 } 131 #endif /* CONFIG_MEMORY_HOTREMOVE */ 132 133 static struct dax_device_driver device_dax_kmem_driver = { 134 .drv = { 135 .probe = dev_dax_kmem_probe, 136 .remove = dev_dax_kmem_remove, 137 }, 138 }; 139 140 static int __init dax_kmem_init(void) 141 { 142 int rc; 143 144 /* Resource name is permanently allocated if any hotremove fails. */ 145 kmem_name = kstrdup_const("System RAM (kmem)", GFP_KERNEL); 146 if (!kmem_name) 147 return -ENOMEM; 148 149 rc = dax_driver_register(&device_dax_kmem_driver); 150 if (rc) 151 kfree_const(kmem_name); 152 return rc; 153 } 154 155 static void __exit dax_kmem_exit(void) 156 { 157 dax_driver_unregister(&device_dax_kmem_driver); 158 if (!any_hotremove_failed) 159 kfree_const(kmem_name); 160 } 161 162 MODULE_AUTHOR("Intel Corporation"); 163 MODULE_LICENSE("GPL v2"); 164 module_init(dax_kmem_init); 165 module_exit(dax_kmem_exit); 166 MODULE_ALIAS_DAX_DEVICE(0); 167