2 * Persistent Memory Driver
4 * Copyright (c) 2014-2015, Intel Corporation.
5 * Copyright (c) 2015, Christoph Hellwig <hch@lst.de>.
6 * Copyright (c) 2015, Boaz Harrosh <boaz@plexistor.com>.
8 * This program is free software; you can redistribute it and/or modify it
9 * under the terms and conditions of the GNU General Public License,
10 * version 2, as published by the Free Software Foundation.
12 * This program is distributed in the hope it will be useful, but WITHOUT
13 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
14 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
18 #include <asm/cacheflush.h>
19 #include <linux/blkdev.h>
20 #include <linux/hdreg.h>
21 #include <linux/init.h>
22 #include <linux/platform_device.h>
23 #include <linux/module.h>
24 #include <linux/moduleparam.h>
25 #include <linux/slab.h>
30 struct request_queue *pmem_queue;
31 struct gendisk *pmem_disk;
33 /* One contiguous memory region per device */
34 phys_addr_t phys_addr;
39 static int pmem_major;
41 static void pmem_do_bvec(struct pmem_device *pmem, struct page *page,
42 unsigned int len, unsigned int off, int rw,
45 void *mem = kmap_atomic(page);
46 size_t pmem_off = sector << 9;
49 memcpy(mem + off, pmem->virt_addr + pmem_off, len);
50 flush_dcache_page(page);
52 flush_dcache_page(page);
53 memcpy(pmem->virt_addr + pmem_off, mem + off, len);
59 static void pmem_make_request(struct request_queue *q, struct bio *bio)
61 struct block_device *bdev = bio->bi_bdev;
62 struct pmem_device *pmem = bdev->bd_disk->private_data;
66 struct bvec_iter iter;
69 if (bio_end_sector(bio) > get_capacity(bdev->bd_disk)) {
74 BUG_ON(bio->bi_rw & REQ_DISCARD);
76 rw = bio_data_dir(bio);
77 sector = bio->bi_iter.bi_sector;
78 bio_for_each_segment(bvec, bio, iter) {
79 pmem_do_bvec(pmem, bvec.bv_page, bvec.bv_len, bvec.bv_offset,
81 sector += bvec.bv_len >> 9;
88 static int pmem_rw_page(struct block_device *bdev, sector_t sector,
89 struct page *page, int rw)
91 struct pmem_device *pmem = bdev->bd_disk->private_data;
93 pmem_do_bvec(pmem, page, PAGE_CACHE_SIZE, 0, rw, sector);
94 page_endio(page, rw & WRITE, 0);
99 static long pmem_direct_access(struct block_device *bdev, sector_t sector,
100 void **kaddr, unsigned long *pfn, long size)
102 struct pmem_device *pmem = bdev->bd_disk->private_data;
103 size_t offset = sector << 9;
108 *kaddr = pmem->virt_addr + offset;
109 *pfn = (pmem->phys_addr + offset) >> PAGE_SHIFT;
111 return pmem->size - offset;
114 static const struct block_device_operations pmem_fops = {
115 .owner = THIS_MODULE,
116 .rw_page = pmem_rw_page,
117 .direct_access = pmem_direct_access,
120 static struct pmem_device *pmem_alloc(struct device *dev,
121 struct resource *res, int id)
123 struct pmem_device *pmem;
124 struct gendisk *disk;
128 pmem = kzalloc(sizeof(*pmem), GFP_KERNEL);
132 pmem->phys_addr = res->start;
133 pmem->size = resource_size(res);
136 if (!request_mem_region(pmem->phys_addr, pmem->size, "pmem")) {
137 dev_warn(dev, "could not reserve region [0x%pa:0x%zx]\n",
138 &pmem->phys_addr, pmem->size);
143 * Map the memory as non-cachable, as we can't write back the contents
144 * of the CPU caches in case of a crash.
147 pmem->virt_addr = ioremap_nocache(pmem->phys_addr, pmem->size);
148 if (!pmem->virt_addr)
149 goto out_release_region;
151 pmem->pmem_queue = blk_alloc_queue(GFP_KERNEL);
152 if (!pmem->pmem_queue)
155 blk_queue_make_request(pmem->pmem_queue, pmem_make_request);
156 blk_queue_max_hw_sectors(pmem->pmem_queue, 1024);
157 blk_queue_bounce_limit(pmem->pmem_queue, BLK_BOUNCE_ANY);
159 disk = alloc_disk(0);
163 disk->major = pmem_major;
164 disk->first_minor = 0;
165 disk->fops = &pmem_fops;
166 disk->private_data = pmem;
167 disk->queue = pmem->pmem_queue;
168 disk->flags = GENHD_FL_EXT_DEVT;
169 sprintf(disk->disk_name, "pmem%d", id);
170 disk->driverfs_dev = dev;
171 set_capacity(disk, pmem->size >> 9);
172 pmem->pmem_disk = disk;
179 blk_cleanup_queue(pmem->pmem_queue);
181 iounmap(pmem->virt_addr);
183 release_mem_region(pmem->phys_addr, pmem->size);
190 static void pmem_free(struct pmem_device *pmem)
192 del_gendisk(pmem->pmem_disk);
193 put_disk(pmem->pmem_disk);
194 blk_cleanup_queue(pmem->pmem_queue);
195 iounmap(pmem->virt_addr);
196 release_mem_region(pmem->phys_addr, pmem->size);
200 static int nd_pmem_probe(struct device *dev)
202 struct nd_region *nd_region = to_nd_region(dev->parent);
203 struct nd_namespace_io *nsio = to_nd_namespace_io(dev);
204 struct pmem_device *pmem;
206 if (resource_size(&nsio->res) < ND_MIN_NAMESPACE_SIZE) {
207 resource_size_t size = resource_size(&nsio->res);
209 dev_dbg(dev, "%s: size: %pa, too small must be at least %#x\n",
210 __func__, &size, ND_MIN_NAMESPACE_SIZE);
214 if (nd_region_to_nstype(nd_region) == ND_DEVICE_NAMESPACE_PMEM) {
215 struct nd_namespace_pmem *nspm = to_nd_namespace_pmem(dev);
218 dev_dbg(dev, "%s: uuid not set\n", __func__);
223 pmem = pmem_alloc(dev, &nsio->res, nd_region->id);
225 return PTR_ERR(pmem);
227 dev_set_drvdata(dev, pmem);
232 static int nd_pmem_remove(struct device *dev)
234 struct pmem_device *pmem = dev_get_drvdata(dev);
240 MODULE_ALIAS("pmem");
241 MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_IO);
242 MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_PMEM);
243 static struct nd_device_driver nd_pmem_driver = {
244 .probe = nd_pmem_probe,
245 .remove = nd_pmem_remove,
249 .type = ND_DRIVER_NAMESPACE_IO | ND_DRIVER_NAMESPACE_PMEM,
252 static int __init pmem_init(void)
256 pmem_major = register_blkdev(0, "pmem");
260 error = nd_driver_register(&nd_pmem_driver);
262 unregister_blkdev(pmem_major, "pmem");
268 module_init(pmem_init);
270 static void pmem_exit(void)
272 driver_unregister(&nd_pmem_driver.drv);
273 unregister_blkdev(pmem_major, "pmem");
275 module_exit(pmem_exit);
277 MODULE_AUTHOR("Ross Zwisler <ross.zwisler@linux.intel.com>");
278 MODULE_LICENSE("GPL v2");