pmem.c 11.4 KB
Newer Older
1 2 3
/*
 * Persistent Memory Driver
 *
4
 * Copyright (c) 2014-2015, Intel Corporation.
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
 * Copyright (c) 2015, Christoph Hellwig <hch@lst.de>.
 * Copyright (c) 2015, Boaz Harrosh <boaz@plexistor.com>.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 */

#include <asm/cacheflush.h>
#include <linux/blkdev.h>
#include <linux/hdreg.h>
#include <linux/init.h>
#include <linux/platform_device.h>
#include <linux/module.h>
#include <linux/moduleparam.h>
25
#include <linux/badblocks.h>
D
Dan Williams 已提交
26
#include <linux/memremap.h>
27
#include <linux/vmalloc.h>
D
Dan Williams 已提交
28
#include <linux/pfn_t.h>
29
#include <linux/slab.h>
30
#include <linux/pmem.h>
31
#include <linux/nd.h>
32
#include "pmem.h"
33
#include "pfn.h"
34
#include "nd.h"
35

36 37 38 39 40 41 42 43 44 45 46 47 48
static struct device *to_dev(struct pmem_device *pmem)
{
	/*
	 * nvdimm bus services need a 'dev' parameter, and we record the device
	 * at init in bb.dev.
	 */
	return pmem->bb.dev;
}

static struct nd_region *to_region(struct pmem_device *pmem)
{
	return to_nd_region(to_dev(pmem)->parent);
}
49

50 51 52
static void pmem_clear_poison(struct pmem_device *pmem, phys_addr_t offset,
		unsigned int len)
{
53
	struct device *dev = to_dev(pmem);
54
	sector_t sector, cleared;
55 56

	sector = (offset - pmem->data_offset) / 512;
57
	cleared = nvdimm_clear_poison(dev, pmem->phys_addr + offset, len) / 512;
58

59
	if (cleared) {
60
		dev_dbg(dev, "%s: %#llx clear %ld sector%s\n",
61
				__func__, (unsigned long long) sector,
62 63
				cleared, cleared > 1 ? "s" : "");
		badblocks_clear(&pmem->bb, sector, cleared);
64 65 66 67
	}
	invalidate_pmem(pmem->virt_addr + offset, len);
}

68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87
static void write_pmem(void *pmem_addr, struct page *page,
		unsigned int off, unsigned int len)
{
	void *mem = kmap_atomic(page);

	memcpy_to_pmem(pmem_addr, mem + off, len);
	kunmap_atomic(mem);
}

static int read_pmem(struct page *page, unsigned int off,
		void *pmem_addr, unsigned int len)
{
	int rc;
	void *mem = kmap_atomic(page);

	rc = memcpy_from_pmem(mem + off, pmem_addr, len);
	kunmap_atomic(mem);
	return rc;
}

88
static int pmem_do_bvec(struct pmem_device *pmem, struct page *page,
89
			unsigned int len, unsigned int off, bool is_write,
90 91
			sector_t sector)
{
92
	int rc = 0;
93
	bool bad_pmem = false;
94
	phys_addr_t pmem_off = sector * 512 + pmem->data_offset;
D
Dan Williams 已提交
95
	void *pmem_addr = pmem->virt_addr + pmem_off;
96

97 98 99
	if (unlikely(is_bad_pmem(&pmem->bb, sector, len)))
		bad_pmem = true;

100
	if (!is_write) {
101
		if (unlikely(bad_pmem))
102 103
			rc = -EIO;
		else {
104
			rc = read_pmem(page, off, pmem_addr, len);
105 106
			flush_dcache_page(page);
		}
107
	} else {
108 109 110 111 112 113 114 115 116 117 118 119 120 121
		/*
		 * Note that we write the data both before and after
		 * clearing poison.  The write before clear poison
		 * handles situations where the latest written data is
		 * preserved and the clear poison operation simply marks
		 * the address range as valid without changing the data.
		 * In this case application software can assume that an
		 * interrupted write will either return the new good
		 * data or an error.
		 *
		 * However, if pmem_clear_poison() leaves the data in an
		 * indeterminate state we need to perform the write
		 * after clear poison.
		 */
122
		flush_dcache_page(page);
123
		write_pmem(pmem_addr, page, off, len);
124 125
		if (unlikely(bad_pmem)) {
			pmem_clear_poison(pmem, pmem_off, len);
126
			write_pmem(pmem_addr, page, off, len);
127
		}
128 129
	}

130
	return rc;
131 132
}

133 134 135 136 137
/* account for REQ_FLUSH rename, replace with REQ_PREFLUSH after v4.8-rc1 */
#ifndef REQ_FLUSH
#define REQ_FLUSH REQ_PREFLUSH
#endif

138
static blk_qc_t pmem_make_request(struct request_queue *q, struct bio *bio)
139
{
140
	int rc = 0;
D
Dan Williams 已提交
141 142
	bool do_acct;
	unsigned long start;
143 144
	struct bio_vec bvec;
	struct bvec_iter iter;
145
	struct pmem_device *pmem = q->queuedata;
146 147
	struct nd_region *nd_region = to_region(pmem);

J
Jens Axboe 已提交
148
	if (bio->bi_opf & REQ_FLUSH)
149
		nvdimm_flush(nd_region);
150

D
Dan Williams 已提交
151
	do_acct = nd_iostat_start(bio, &start);
152 153
	bio_for_each_segment(bvec, bio, iter) {
		rc = pmem_do_bvec(pmem, bvec.bv_page, bvec.bv_len,
154
				bvec.bv_offset, op_is_write(bio_op(bio)),
155 156 157 158 159 160
				iter.bi_sector);
		if (rc) {
			bio->bi_error = rc;
			break;
		}
	}
D
Dan Williams 已提交
161 162
	if (do_acct)
		nd_iostat_end(bio, start);
163

J
Jens Axboe 已提交
164
	if (bio->bi_opf & REQ_FUA)
165
		nvdimm_flush(nd_region);
166

167
	bio_endio(bio);
168
	return BLK_QC_T_NONE;
169 170 171
}

static int pmem_rw_page(struct block_device *bdev, sector_t sector,
172
		       struct page *page, bool is_write)
173
{
174
	struct pmem_device *pmem = bdev->bd_queue->queuedata;
175
	int rc;
176

177
	rc = pmem_do_bvec(pmem, page, PAGE_SIZE, 0, is_write, sector);
178

179 180 181 182 183 184 185
	/*
	 * The ->rw_page interface is subtle and tricky.  The core
	 * retries on any error, so we can only invoke page_endio() in
	 * the successful completion case.  Otherwise, we'll see crashes
	 * caused by double completion.
	 */
	if (rc == 0)
186
		page_endio(page, is_write, 0);
187 188

	return rc;
189 190
}

191 192
/* see "strong" declaration in tools/testing/nvdimm/pmem-dax.c */
__weak long pmem_direct_access(struct block_device *bdev, sector_t sector,
D
Dan Williams 已提交
193
		      void **kaddr, pfn_t *pfn, long size)
194
{
195
	struct pmem_device *pmem = bdev->bd_queue->queuedata;
196
	resource_size_t offset = sector * 512 + pmem->data_offset;
197

198 199
	if (unlikely(is_bad_pmem(&pmem->bb, sector, size)))
		return -EIO;
200
	*kaddr = pmem->virt_addr + offset;
D
Dan Williams 已提交
201
	*pfn = phys_to_pfn_t(pmem->phys_addr + offset, pmem->pfn_flags);
202

203 204 205 206 207 208
	/*
	 * If badblocks are present, limit known good range to the
	 * requested range.
	 */
	if (unlikely(pmem->bb.count))
		return size;
209
	return pmem->size - pmem->pfn_pad - offset;
210 211 212 213 214 215
}

static const struct block_device_operations pmem_fops = {
	.owner =		THIS_MODULE,
	.rw_page =		pmem_rw_page,
	.direct_access =	pmem_direct_access,
216
	.revalidate_disk =	nvdimm_revalidate_disk,
217 218
};

219 220 221 222 223
static void pmem_release_queue(void *q)
{
	blk_cleanup_queue(q);
}

224
static void pmem_release_disk(void *disk)
225 226 227 228 229
{
	del_gendisk(disk);
	put_disk(disk);
}

230 231
static int pmem_attach_disk(struct device *dev,
		struct nd_namespace_common *ndns)
232
{
233
	struct nd_namespace_io *nsio = to_nd_namespace_io(&ndns->dev);
234
	struct nd_region *nd_region = to_nd_region(dev->parent);
235 236 237 238 239
	struct vmem_altmap __altmap, *altmap = NULL;
	struct resource *res = &nsio->res;
	struct nd_pfn *nd_pfn = NULL;
	int nid = dev_to_node(dev);
	struct nd_pfn_sb *pfn_sb;
240
	struct pmem_device *pmem;
241
	struct resource pfn_res;
242
	struct request_queue *q;
243 244 245 246 247 248 249 250 251 252 253 254 255
	struct gendisk *disk;
	void *addr;

	/* while nsio_rw_bytes is active, parse a pfn info block if present */
	if (is_nd_pfn(dev)) {
		nd_pfn = to_nd_pfn(dev);
		altmap = nvdimm_setup_pfn(nd_pfn, &pfn_res, &__altmap);
		if (IS_ERR(altmap))
			return PTR_ERR(altmap);
	}

	/* we're attaching a block device, disable raw namespace access */
	devm_nsio_disable(dev, nsio);
256

257
	pmem = devm_kzalloc(dev, sizeof(*pmem), GFP_KERNEL);
258
	if (!pmem)
259
		return -ENOMEM;
260

261
	dev_set_drvdata(dev, pmem);
262 263
	pmem->phys_addr = res->start;
	pmem->size = resource_size(res);
264
	if (nvdimm_has_flush(nd_region) < 0)
265
		dev_warn(dev, "unable to guarantee persistence of writes\n");
266

267
	if (!devm_request_mem_region(dev, res->start, resource_size(res),
268
				dev_name(&ndns->dev))) {
269
		dev_warn(dev, "could not reserve region %pR\n", res);
270
		return -EBUSY;
271 272
	}

273 274
	q = blk_alloc_queue_node(GFP_KERNEL, dev_to_node(dev));
	if (!q)
275
		return -ENOMEM;
276

D
Dan Williams 已提交
277
	pmem->pfn_flags = PFN_DEV;
278 279 280 281 282 283 284 285 286 287 288
	if (is_nd_pfn(dev)) {
		addr = devm_memremap_pages(dev, &pfn_res, &q->q_usage_counter,
				altmap);
		pfn_sb = nd_pfn->pfn_sb;
		pmem->data_offset = le64_to_cpu(pfn_sb->dataoff);
		pmem->pfn_pad = resource_size(res) - resource_size(&pfn_res);
		pmem->pfn_flags |= PFN_MAP;
		res = &pfn_res; /* for badblocks populate */
		res->start += pmem->data_offset;
	} else if (pmem_should_map_pages(dev)) {
		addr = devm_memremap_pages(dev, &nsio->res,
289
				&q->q_usage_counter, NULL);
D
Dan Williams 已提交
290 291
		pmem->pfn_flags |= PFN_MAP;
	} else
292 293
		addr = devm_memremap(dev, pmem->phys_addr,
				pmem->size, ARCH_MEMREMAP_PMEM);
294

295 296 297 298
	/*
	 * At release time the queue must be dead before
	 * devm_memremap_pages is unwound
	 */
299
	if (devm_add_action_or_reset(dev, pmem_release_queue, q))
300
		return -ENOMEM;
301

302 303
	if (IS_ERR(addr))
		return PTR_ERR(addr);
D
Dan Williams 已提交
304
	pmem->virt_addr = addr;
305

306
	blk_queue_write_cache(q, true, true);
307 308 309 310 311
	blk_queue_make_request(q, pmem_make_request);
	blk_queue_physical_block_size(q, PAGE_SIZE);
	blk_queue_max_hw_sectors(q, UINT_MAX);
	blk_queue_bounce_limit(q, BLK_BOUNCE_ANY);
	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q);
312
	queue_flag_set_unlocked(QUEUE_FLAG_DAX, q);
313
	q->queuedata = pmem;
314

315
	disk = alloc_disk_node(0, nid);
316 317
	if (!disk)
		return -ENOMEM;
318 319

	disk->fops		= &pmem_fops;
320
	disk->queue		= q;
321
	disk->flags		= GENHD_FL_EXT_DEVT;
V
Vishal Verma 已提交
322
	nvdimm_namespace_disk_name(ndns, disk->disk_name);
323 324
	set_capacity(disk, (pmem->size - pmem->pfn_pad - pmem->data_offset)
			/ 512);
325 326
	if (devm_init_badblocks(dev, &pmem->bb))
		return -ENOMEM;
327
	nvdimm_badblocks_populate(nd_region, &pmem->bb, res);
328
	disk->bb = &pmem->bb;
329
	device_add_disk(dev, disk);
330 331 332 333

	if (devm_add_action_or_reset(dev, pmem_release_disk, disk))
		return -ENOMEM;

334
	revalidate_disk(disk);
335

336 337
	return 0;
}
338

339
static int nd_pmem_probe(struct device *dev)
340
{
341
	struct nd_namespace_common *ndns;
342

343 344 345
	ndns = nvdimm_namespace_common_probe(dev);
	if (IS_ERR(ndns))
		return PTR_ERR(ndns);
346

347 348
	if (devm_nsio_enable(dev, to_nd_namespace_io(&ndns->dev)))
		return -ENXIO;
349

350
	if (is_nd_btt(dev))
351 352
		return nvdimm_namespace_attach_btt(ndns);

353
	if (is_nd_pfn(dev))
354
		return pmem_attach_disk(dev, ndns);
355

356
	/* if we find a valid info-block we'll come back as that personality */
357 358
	if (nd_btt_probe(dev, ndns) == 0 || nd_pfn_probe(dev, ndns) == 0
			|| nd_dax_probe(dev, ndns) == 0)
359 360
		return -ENXIO;

361 362
	/* ...otherwise we're just a raw pmem device */
	return pmem_attach_disk(dev, ndns);
363 364
}

365
static int nd_pmem_remove(struct device *dev)
366
{
367
	if (is_nd_btt(dev))
D
Dan Williams 已提交
368
		nvdimm_namespace_detach_btt(to_nd_btt(dev));
369 370
	nvdimm_flush(to_nd_region(dev->parent));

371 372 373
	return 0;
}

374 375 376 377 378
static void nd_pmem_shutdown(struct device *dev)
{
	nvdimm_flush(to_nd_region(dev->parent));
}

379 380
static void nd_pmem_notify(struct device *dev, enum nvdimm_event event)
{
D
Dan Williams 已提交
381
	struct pmem_device *pmem = dev_get_drvdata(dev);
382
	struct nd_region *nd_region = to_region(pmem);
D
Dan Williams 已提交
383 384 385 386
	resource_size_t offset = 0, end_trunc = 0;
	struct nd_namespace_common *ndns;
	struct nd_namespace_io *nsio;
	struct resource res;
387 388 389 390

	if (event != NVDIMM_REVALIDATE_POISON)
		return;

D
Dan Williams 已提交
391 392 393 394 395
	if (is_nd_btt(dev)) {
		struct nd_btt *nd_btt = to_nd_btt(dev);

		ndns = nd_btt->ndns;
	} else if (is_nd_pfn(dev)) {
396 397 398
		struct nd_pfn *nd_pfn = to_nd_pfn(dev);
		struct nd_pfn_sb *pfn_sb = nd_pfn->pfn_sb;

D
Dan Williams 已提交
399 400 401 402 403
		ndns = nd_pfn->ndns;
		offset = pmem->data_offset + __le32_to_cpu(pfn_sb->start_pad);
		end_trunc = __le32_to_cpu(pfn_sb->end_trunc);
	} else
		ndns = to_ndns(dev);
404

D
Dan Williams 已提交
405 406 407
	nsio = to_nd_namespace_io(&ndns->dev);
	res.start = nsio->res.start + offset;
	res.end = nsio->res.end - end_trunc;
408
	nvdimm_badblocks_populate(nd_region, &pmem->bb, &res);
409 410
}

411 412
MODULE_ALIAS("pmem");
MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_IO);
413
MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_PMEM);
414 415 416
static struct nd_device_driver nd_pmem_driver = {
	.probe = nd_pmem_probe,
	.remove = nd_pmem_remove,
417
	.notify = nd_pmem_notify,
418
	.shutdown = nd_pmem_shutdown,
419 420
	.drv = {
		.name = "nd_pmem",
421
	},
422
	.type = ND_DRIVER_NAMESPACE_IO | ND_DRIVER_NAMESPACE_PMEM,
423 424 425 426
};

static int __init pmem_init(void)
{
427
	return nd_driver_register(&nd_pmem_driver);
428 429 430 431 432
}
module_init(pmem_init);

static void pmem_exit(void)
{
433
	driver_unregister(&nd_pmem_driver.drv);
434 435 436 437 438
}
module_exit(pmem_exit);

MODULE_AUTHOR("Ross Zwisler <ross.zwisler@linux.intel.com>");
MODULE_LICENSE("GPL v2");