]>
Commit | Line | Data |
---|---|---|
9e853f23 RZ |
1 | /* |
2 | * Persistent Memory Driver | |
3 | * | |
9f53f9fa | 4 | * Copyright (c) 2014-2015, Intel Corporation. |
9e853f23 RZ |
5 | * Copyright (c) 2015, Christoph Hellwig <hch@lst.de>. |
6 | * Copyright (c) 2015, Boaz Harrosh <boaz@plexistor.com>. | |
7 | * | |
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. | |
11 | * | |
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 | |
15 | * more details. | |
16 | */ | |
17 | ||
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> | |
b95f5f43 | 25 | #include <linux/badblocks.h> |
9476df7d | 26 | #include <linux/memremap.h> |
32ab0a3f | 27 | #include <linux/vmalloc.h> |
71389703 | 28 | #include <linux/blk-mq.h> |
34c0fd54 | 29 | #include <linux/pfn_t.h> |
9e853f23 | 30 | #include <linux/slab.h> |
0aed55af | 31 | #include <linux/uio.h> |
c1d6e828 | 32 | #include <linux/dax.h> |
9f53f9fa | 33 | #include <linux/nd.h> |
23c47d2a | 34 | #include <linux/backing-dev.h> |
f295e53b | 35 | #include "pmem.h" |
32ab0a3f | 36 | #include "pfn.h" |
9f53f9fa | 37 | #include "nd.h" |
06e8ccda | 38 | #include "nd-core.h" |
9e853f23 | 39 | |
f284a4f2 DW |
40 | static struct device *to_dev(struct pmem_device *pmem) |
41 | { | |
42 | /* | |
43 | * nvdimm bus services need a 'dev' parameter, and we record the device | |
44 | * at init in bb.dev. | |
45 | */ | |
46 | return pmem->bb.dev; | |
47 | } | |
48 | ||
49 | static struct nd_region *to_region(struct pmem_device *pmem) | |
50 | { | |
51 | return to_nd_region(to_dev(pmem)->parent); | |
52 | } | |
9e853f23 | 53 | |
4e4cbee9 CH |
54 | static blk_status_t pmem_clear_poison(struct pmem_device *pmem, |
55 | phys_addr_t offset, unsigned int len) | |
59e64739 | 56 | { |
f284a4f2 | 57 | struct device *dev = to_dev(pmem); |
59e64739 DW |
58 | sector_t sector; |
59 | long cleared; | |
4e4cbee9 | 60 | blk_status_t rc = BLK_STS_OK; |
59e64739 DW |
61 | |
62 | sector = (offset - pmem->data_offset) / 512; | |
59e64739 | 63 | |
868f036f DW |
64 | cleared = nvdimm_clear_poison(dev, pmem->phys_addr + offset, len); |
65 | if (cleared < len) | |
4e4cbee9 | 66 | rc = BLK_STS_IOERR; |
59e64739 | 67 | if (cleared > 0 && cleared / 512) { |
868f036f | 68 | cleared /= 512; |
426824d6 | 69 | dev_dbg(dev, "%#llx clear %ld sector%s\n", |
868f036f DW |
70 | (unsigned long long) sector, cleared, |
71 | cleared > 1 ? "s" : ""); | |
0a3f27b9 | 72 | badblocks_clear(&pmem->bb, sector, cleared); |
975750a9 TK |
73 | if (pmem->bb_state) |
74 | sysfs_notify_dirent(pmem->bb_state); | |
59e64739 | 75 | } |
3115bb02 | 76 | |
f2b61257 | 77 | arch_invalidate_pmem(pmem->virt_addr + offset, len); |
868f036f DW |
78 | |
79 | return rc; | |
59e64739 DW |
80 | } |
81 | ||
bd697a80 VV |
82 | static void write_pmem(void *pmem_addr, struct page *page, |
83 | unsigned int off, unsigned int len) | |
84 | { | |
98cc093c HY |
85 | unsigned int chunk; |
86 | void *mem; | |
87 | ||
88 | while (len) { | |
89 | mem = kmap_atomic(page); | |
90 | chunk = min_t(unsigned int, len, PAGE_SIZE); | |
91 | memcpy_flushcache(pmem_addr, mem + off, chunk); | |
92 | kunmap_atomic(mem); | |
93 | len -= chunk; | |
94 | off = 0; | |
95 | page++; | |
96 | pmem_addr += PAGE_SIZE; | |
97 | } | |
bd697a80 VV |
98 | } |
99 | ||
4e4cbee9 | 100 | static blk_status_t read_pmem(struct page *page, unsigned int off, |
bd697a80 VV |
101 | void *pmem_addr, unsigned int len) |
102 | { | |
98cc093c | 103 | unsigned int chunk; |
60622d68 | 104 | unsigned long rem; |
98cc093c HY |
105 | void *mem; |
106 | ||
107 | while (len) { | |
108 | mem = kmap_atomic(page); | |
109 | chunk = min_t(unsigned int, len, PAGE_SIZE); | |
60622d68 | 110 | rem = memcpy_mcsafe(mem + off, pmem_addr, chunk); |
98cc093c | 111 | kunmap_atomic(mem); |
60622d68 | 112 | if (rem) |
98cc093c HY |
113 | return BLK_STS_IOERR; |
114 | len -= chunk; | |
115 | off = 0; | |
116 | page++; | |
117 | pmem_addr += PAGE_SIZE; | |
118 | } | |
4e4cbee9 | 119 | return BLK_STS_OK; |
bd697a80 VV |
120 | } |
121 | ||
4e4cbee9 | 122 | static blk_status_t pmem_do_bvec(struct pmem_device *pmem, struct page *page, |
c11f0c0b | 123 | unsigned int len, unsigned int off, bool is_write, |
9e853f23 RZ |
124 | sector_t sector) |
125 | { | |
4e4cbee9 | 126 | blk_status_t rc = BLK_STS_OK; |
59e64739 | 127 | bool bad_pmem = false; |
32ab0a3f | 128 | phys_addr_t pmem_off = sector * 512 + pmem->data_offset; |
7a9eb206 | 129 | void *pmem_addr = pmem->virt_addr + pmem_off; |
9e853f23 | 130 | |
59e64739 DW |
131 | if (unlikely(is_bad_pmem(&pmem->bb, sector, len))) |
132 | bad_pmem = true; | |
133 | ||
c11f0c0b | 134 | if (!is_write) { |
59e64739 | 135 | if (unlikely(bad_pmem)) |
4e4cbee9 | 136 | rc = BLK_STS_IOERR; |
b5ebc8ec | 137 | else { |
bd697a80 | 138 | rc = read_pmem(page, off, pmem_addr, len); |
b5ebc8ec DW |
139 | flush_dcache_page(page); |
140 | } | |
9e853f23 | 141 | } else { |
0a370d26 DW |
142 | /* |
143 | * Note that we write the data both before and after | |
144 | * clearing poison. The write before clear poison | |
145 | * handles situations where the latest written data is | |
146 | * preserved and the clear poison operation simply marks | |
147 | * the address range as valid without changing the data. | |
148 | * In this case application software can assume that an | |
149 | * interrupted write will either return the new good | |
150 | * data or an error. | |
151 | * | |
152 | * However, if pmem_clear_poison() leaves the data in an | |
153 | * indeterminate state we need to perform the write | |
154 | * after clear poison. | |
155 | */ | |
9e853f23 | 156 | flush_dcache_page(page); |
bd697a80 | 157 | write_pmem(pmem_addr, page, off, len); |
59e64739 | 158 | if (unlikely(bad_pmem)) { |
3115bb02 | 159 | rc = pmem_clear_poison(pmem, pmem_off, len); |
bd697a80 | 160 | write_pmem(pmem_addr, page, off, len); |
59e64739 | 161 | } |
9e853f23 RZ |
162 | } |
163 | ||
b5ebc8ec | 164 | return rc; |
9e853f23 RZ |
165 | } |
166 | ||
dece1635 | 167 | static blk_qc_t pmem_make_request(struct request_queue *q, struct bio *bio) |
9e853f23 | 168 | { |
4e4cbee9 | 169 | blk_status_t rc = 0; |
f0dc089c DW |
170 | bool do_acct; |
171 | unsigned long start; | |
9e853f23 | 172 | struct bio_vec bvec; |
9e853f23 | 173 | struct bvec_iter iter; |
bd842b8c | 174 | struct pmem_device *pmem = q->queuedata; |
7e267a8c DW |
175 | struct nd_region *nd_region = to_region(pmem); |
176 | ||
d2d6364d | 177 | if (bio->bi_opf & REQ_PREFLUSH) |
7e267a8c | 178 | nvdimm_flush(nd_region); |
9e853f23 | 179 | |
f0dc089c | 180 | do_acct = nd_iostat_start(bio, &start); |
e10624f8 DW |
181 | bio_for_each_segment(bvec, bio, iter) { |
182 | rc = pmem_do_bvec(pmem, bvec.bv_page, bvec.bv_len, | |
c11f0c0b | 183 | bvec.bv_offset, op_is_write(bio_op(bio)), |
e10624f8 DW |
184 | iter.bi_sector); |
185 | if (rc) { | |
4e4cbee9 | 186 | bio->bi_status = rc; |
e10624f8 DW |
187 | break; |
188 | } | |
189 | } | |
f0dc089c DW |
190 | if (do_acct) |
191 | nd_iostat_end(bio, start); | |
61031952 | 192 | |
1eff9d32 | 193 | if (bio->bi_opf & REQ_FUA) |
7e267a8c | 194 | nvdimm_flush(nd_region); |
61031952 | 195 | |
4246a0b6 | 196 | bio_endio(bio); |
dece1635 | 197 | return BLK_QC_T_NONE; |
9e853f23 RZ |
198 | } |
199 | ||
200 | static int pmem_rw_page(struct block_device *bdev, sector_t sector, | |
c11f0c0b | 201 | struct page *page, bool is_write) |
9e853f23 | 202 | { |
bd842b8c | 203 | struct pmem_device *pmem = bdev->bd_queue->queuedata; |
4e4cbee9 | 204 | blk_status_t rc; |
9e853f23 | 205 | |
98cc093c HY |
206 | rc = pmem_do_bvec(pmem, page, hpage_nr_pages(page) * PAGE_SIZE, |
207 | 0, is_write, sector); | |
9e853f23 | 208 | |
e10624f8 DW |
209 | /* |
210 | * The ->rw_page interface is subtle and tricky. The core | |
211 | * retries on any error, so we can only invoke page_endio() in | |
212 | * the successful completion case. Otherwise, we'll see crashes | |
213 | * caused by double completion. | |
214 | */ | |
215 | if (rc == 0) | |
c11f0c0b | 216 | page_endio(page, is_write, 0); |
e10624f8 | 217 | |
4e4cbee9 | 218 | return blk_status_to_errno(rc); |
9e853f23 RZ |
219 | } |
220 | ||
f295e53b | 221 | /* see "strong" declaration in tools/testing/nvdimm/pmem-dax.c */ |
c1d6e828 DW |
222 | __weak long __pmem_direct_access(struct pmem_device *pmem, pgoff_t pgoff, |
223 | long nr_pages, void **kaddr, pfn_t *pfn) | |
9e853f23 | 224 | { |
c1d6e828 | 225 | resource_size_t offset = PFN_PHYS(pgoff) + pmem->data_offset; |
589e75d1 | 226 | |
c1d6e828 DW |
227 | if (unlikely(is_bad_pmem(&pmem->bb, PFN_PHYS(pgoff) / 512, |
228 | PFN_PHYS(nr_pages)))) | |
0a70bd43 | 229 | return -EIO; |
e2e05394 | 230 | *kaddr = pmem->virt_addr + offset; |
34c0fd54 | 231 | *pfn = phys_to_pfn_t(pmem->phys_addr + offset, pmem->pfn_flags); |
9e853f23 | 232 | |
0a70bd43 DW |
233 | /* |
234 | * If badblocks are present, limit known good range to the | |
235 | * requested range. | |
236 | */ | |
237 | if (unlikely(pmem->bb.count)) | |
c1d6e828 DW |
238 | return nr_pages; |
239 | return PHYS_PFN(pmem->size - pmem->pfn_pad - offset); | |
9e853f23 RZ |
240 | } |
241 | ||
242 | static const struct block_device_operations pmem_fops = { | |
243 | .owner = THIS_MODULE, | |
244 | .rw_page = pmem_rw_page, | |
58138820 | 245 | .revalidate_disk = nvdimm_revalidate_disk, |
9e853f23 RZ |
246 | }; |
247 | ||
c1d6e828 DW |
248 | static long pmem_dax_direct_access(struct dax_device *dax_dev, |
249 | pgoff_t pgoff, long nr_pages, void **kaddr, pfn_t *pfn) | |
250 | { | |
251 | struct pmem_device *pmem = dax_get_private(dax_dev); | |
252 | ||
253 | return __pmem_direct_access(pmem, pgoff, nr_pages, kaddr, pfn); | |
254 | } | |
255 | ||
0aed55af DW |
256 | static size_t pmem_copy_from_iter(struct dax_device *dax_dev, pgoff_t pgoff, |
257 | void *addr, size_t bytes, struct iov_iter *i) | |
258 | { | |
259 | return copy_from_iter_flushcache(addr, bytes, i); | |
260 | } | |
261 | ||
b3a9a0c3 DW |
262 | static size_t pmem_copy_to_iter(struct dax_device *dax_dev, pgoff_t pgoff, |
263 | void *addr, size_t bytes, struct iov_iter *i) | |
264 | { | |
6dfdb2b6 | 265 | return copy_to_iter_mcsafe(addr, bytes, i); |
b3a9a0c3 DW |
266 | } |
267 | ||
c1d6e828 DW |
268 | static const struct dax_operations pmem_dax_ops = { |
269 | .direct_access = pmem_dax_direct_access, | |
0aed55af | 270 | .copy_from_iter = pmem_copy_from_iter, |
b3a9a0c3 | 271 | .copy_to_iter = pmem_copy_to_iter, |
c1d6e828 DW |
272 | }; |
273 | ||
6e0c90d6 DW |
274 | static const struct attribute_group *pmem_attribute_groups[] = { |
275 | &dax_attribute_group, | |
276 | NULL, | |
c1d6e828 DW |
277 | }; |
278 | ||
030b99e3 DW |
279 | static void pmem_release_queue(void *q) |
280 | { | |
281 | blk_cleanup_queue(q); | |
282 | } | |
283 | ||
71389703 DW |
284 | static void pmem_freeze_queue(void *q) |
285 | { | |
d3b5d352 | 286 | blk_freeze_queue_start(q); |
71389703 DW |
287 | } |
288 | ||
c1d6e828 | 289 | static void pmem_release_disk(void *__pmem) |
030b99e3 | 290 | { |
c1d6e828 DW |
291 | struct pmem_device *pmem = __pmem; |
292 | ||
293 | kill_dax(pmem->dax_dev); | |
294 | put_dax(pmem->dax_dev); | |
295 | del_gendisk(pmem->disk); | |
296 | put_disk(pmem->disk); | |
030b99e3 DW |
297 | } |
298 | ||
e7638488 DW |
299 | static void pmem_release_pgmap_ops(void *__pgmap) |
300 | { | |
301 | dev_pagemap_put_ops(); | |
302 | } | |
303 | ||
304 | static void fsdax_pagefree(struct page *page, void *data) | |
305 | { | |
306 | wake_up_var(&page->_refcount); | |
307 | } | |
308 | ||
309 | static int setup_pagemap_fsdax(struct device *dev, struct dev_pagemap *pgmap) | |
310 | { | |
311 | dev_pagemap_get_ops(); | |
312 | if (devm_add_action_or_reset(dev, pmem_release_pgmap_ops, pgmap)) | |
313 | return -ENOMEM; | |
314 | pgmap->type = MEMORY_DEVICE_FS_DAX; | |
315 | pgmap->page_free = fsdax_pagefree; | |
316 | ||
317 | return 0; | |
318 | } | |
319 | ||
200c79da DW |
320 | static int pmem_attach_disk(struct device *dev, |
321 | struct nd_namespace_common *ndns) | |
9e853f23 | 322 | { |
200c79da | 323 | struct nd_namespace_io *nsio = to_nd_namespace_io(&ndns->dev); |
f284a4f2 | 324 | struct nd_region *nd_region = to_nd_region(dev->parent); |
ce7f11a2 | 325 | int nid = dev_to_node(dev), fua; |
200c79da | 326 | struct resource *res = &nsio->res; |
e8d51348 | 327 | struct resource bb_res; |
200c79da | 328 | struct nd_pfn *nd_pfn = NULL; |
c1d6e828 | 329 | struct dax_device *dax_dev; |
200c79da | 330 | struct nd_pfn_sb *pfn_sb; |
9e853f23 | 331 | struct pmem_device *pmem; |
468ded03 | 332 | struct request_queue *q; |
6e0c90d6 | 333 | struct device *gendev; |
200c79da DW |
334 | struct gendisk *disk; |
335 | void *addr; | |
e8d51348 CH |
336 | int rc; |
337 | ||
338 | pmem = devm_kzalloc(dev, sizeof(*pmem), GFP_KERNEL); | |
339 | if (!pmem) | |
340 | return -ENOMEM; | |
200c79da DW |
341 | |
342 | /* while nsio_rw_bytes is active, parse a pfn info block if present */ | |
343 | if (is_nd_pfn(dev)) { | |
344 | nd_pfn = to_nd_pfn(dev); | |
e8d51348 CH |
345 | rc = nvdimm_setup_pfn(nd_pfn, &pmem->pgmap); |
346 | if (rc) | |
347 | return rc; | |
200c79da DW |
348 | } |
349 | ||
350 | /* we're attaching a block device, disable raw namespace access */ | |
351 | devm_nsio_disable(dev, nsio); | |
9e853f23 | 352 | |
200c79da | 353 | dev_set_drvdata(dev, pmem); |
9e853f23 RZ |
354 | pmem->phys_addr = res->start; |
355 | pmem->size = resource_size(res); | |
0b277961 DW |
356 | fua = nvdimm_has_flush(nd_region); |
357 | if (!IS_ENABLED(CONFIG_ARCH_HAS_UACCESS_FLUSHCACHE) || fua < 0) { | |
61031952 | 358 | dev_warn(dev, "unable to guarantee persistence of writes\n"); |
0b277961 DW |
359 | fua = 0; |
360 | } | |
9e853f23 | 361 | |
947df02d | 362 | if (!devm_request_mem_region(dev, res->start, resource_size(res), |
450c6633 | 363 | dev_name(&ndns->dev))) { |
947df02d | 364 | dev_warn(dev, "could not reserve region %pR\n", res); |
200c79da | 365 | return -EBUSY; |
9e853f23 RZ |
366 | } |
367 | ||
5ee0524b | 368 | q = blk_alloc_queue_node(GFP_KERNEL, dev_to_node(dev), NULL); |
468ded03 | 369 | if (!q) |
200c79da | 370 | return -ENOMEM; |
468ded03 | 371 | |
71389703 DW |
372 | if (devm_add_action_or_reset(dev, pmem_release_queue, q)) |
373 | return -ENOMEM; | |
374 | ||
34c0fd54 | 375 | pmem->pfn_flags = PFN_DEV; |
e8d51348 | 376 | pmem->pgmap.ref = &q->q_usage_counter; |
200c79da | 377 | if (is_nd_pfn(dev)) { |
e7638488 DW |
378 | if (setup_pagemap_fsdax(dev, &pmem->pgmap)) |
379 | return -ENOMEM; | |
e8d51348 | 380 | addr = devm_memremap_pages(dev, &pmem->pgmap); |
200c79da DW |
381 | pfn_sb = nd_pfn->pfn_sb; |
382 | pmem->data_offset = le64_to_cpu(pfn_sb->dataoff); | |
e8d51348 CH |
383 | pmem->pfn_pad = resource_size(res) - |
384 | resource_size(&pmem->pgmap.res); | |
200c79da | 385 | pmem->pfn_flags |= PFN_MAP; |
e8d51348 CH |
386 | memcpy(&bb_res, &pmem->pgmap.res, sizeof(bb_res)); |
387 | bb_res.start += pmem->data_offset; | |
200c79da | 388 | } else if (pmem_should_map_pages(dev)) { |
e8d51348 CH |
389 | memcpy(&pmem->pgmap.res, &nsio->res, sizeof(pmem->pgmap.res)); |
390 | pmem->pgmap.altmap_valid = false; | |
e7638488 DW |
391 | if (setup_pagemap_fsdax(dev, &pmem->pgmap)) |
392 | return -ENOMEM; | |
e8d51348 | 393 | addr = devm_memremap_pages(dev, &pmem->pgmap); |
34c0fd54 | 394 | pmem->pfn_flags |= PFN_MAP; |
e8d51348 | 395 | memcpy(&bb_res, &pmem->pgmap.res, sizeof(bb_res)); |
34c0fd54 | 396 | } else |
200c79da DW |
397 | addr = devm_memremap(dev, pmem->phys_addr, |
398 | pmem->size, ARCH_MEMREMAP_PMEM); | |
b36f4761 | 399 | |
030b99e3 | 400 | /* |
71389703 | 401 | * At release time the queue must be frozen before |
030b99e3 DW |
402 | * devm_memremap_pages is unwound |
403 | */ | |
71389703 | 404 | if (devm_add_action_or_reset(dev, pmem_freeze_queue, q)) |
200c79da | 405 | return -ENOMEM; |
8c2f7e86 | 406 | |
200c79da DW |
407 | if (IS_ERR(addr)) |
408 | return PTR_ERR(addr); | |
7a9eb206 | 409 | pmem->virt_addr = addr; |
9e853f23 | 410 | |
ce7f11a2 | 411 | blk_queue_write_cache(q, true, fua); |
5a92289f DW |
412 | blk_queue_make_request(q, pmem_make_request); |
413 | blk_queue_physical_block_size(q, PAGE_SIZE); | |
f979b13c | 414 | blk_queue_logical_block_size(q, pmem_sector_size(ndns)); |
5a92289f | 415 | blk_queue_max_hw_sectors(q, UINT_MAX); |
8b904b5b BVA |
416 | blk_queue_flag_set(QUEUE_FLAG_NONROT, q); |
417 | blk_queue_flag_set(QUEUE_FLAG_DAX, q); | |
5a92289f | 418 | q->queuedata = pmem; |
9e853f23 | 419 | |
538ea4aa | 420 | disk = alloc_disk_node(0, nid); |
030b99e3 DW |
421 | if (!disk) |
422 | return -ENOMEM; | |
c1d6e828 | 423 | pmem->disk = disk; |
9e853f23 | 424 | |
9e853f23 | 425 | disk->fops = &pmem_fops; |
5a92289f | 426 | disk->queue = q; |
9e853f23 | 427 | disk->flags = GENHD_FL_EXT_DEVT; |
23c47d2a | 428 | disk->queue->backing_dev_info->capabilities |= BDI_CAP_SYNCHRONOUS_IO; |
5212e11f | 429 | nvdimm_namespace_disk_name(ndns, disk->disk_name); |
cfe30b87 DW |
430 | set_capacity(disk, (pmem->size - pmem->pfn_pad - pmem->data_offset) |
431 | / 512); | |
b95f5f43 DW |
432 | if (devm_init_badblocks(dev, &pmem->bb)) |
433 | return -ENOMEM; | |
e8d51348 | 434 | nvdimm_badblocks_populate(nd_region, &pmem->bb, &bb_res); |
57f7f317 | 435 | disk->bb = &pmem->bb; |
f02716db | 436 | |
c1d6e828 DW |
437 | dax_dev = alloc_dax(pmem, disk->disk_name, &pmem_dax_ops); |
438 | if (!dax_dev) { | |
439 | put_disk(disk); | |
440 | return -ENOMEM; | |
441 | } | |
ce7f11a2 | 442 | dax_write_cache(dax_dev, nvdimm_has_cache(nd_region)); |
c1d6e828 DW |
443 | pmem->dax_dev = dax_dev; |
444 | ||
6e0c90d6 DW |
445 | gendev = disk_to_dev(disk); |
446 | gendev->groups = pmem_attribute_groups; | |
447 | ||
c1d6e828 DW |
448 | device_add_disk(dev, disk); |
449 | if (devm_add_action_or_reset(dev, pmem_release_disk, pmem)) | |
f02716db DW |
450 | return -ENOMEM; |
451 | ||
58138820 | 452 | revalidate_disk(disk); |
9e853f23 | 453 | |
975750a9 TK |
454 | pmem->bb_state = sysfs_get_dirent(disk_to_dev(disk)->kobj.sd, |
455 | "badblocks"); | |
6aa734a2 DW |
456 | if (!pmem->bb_state) |
457 | dev_warn(dev, "'badblocks' notification disabled\n"); | |
975750a9 | 458 | |
8c2f7e86 DW |
459 | return 0; |
460 | } | |
9e853f23 | 461 | |
9f53f9fa | 462 | static int nd_pmem_probe(struct device *dev) |
9e853f23 | 463 | { |
8c2f7e86 | 464 | struct nd_namespace_common *ndns; |
9e853f23 | 465 | |
8c2f7e86 DW |
466 | ndns = nvdimm_namespace_common_probe(dev); |
467 | if (IS_ERR(ndns)) | |
468 | return PTR_ERR(ndns); | |
bf9bccc1 | 469 | |
200c79da DW |
470 | if (devm_nsio_enable(dev, to_nd_namespace_io(&ndns->dev))) |
471 | return -ENXIO; | |
708ab62b | 472 | |
200c79da | 473 | if (is_nd_btt(dev)) |
708ab62b CH |
474 | return nvdimm_namespace_attach_btt(ndns); |
475 | ||
32ab0a3f | 476 | if (is_nd_pfn(dev)) |
200c79da | 477 | return pmem_attach_disk(dev, ndns); |
32ab0a3f | 478 | |
200c79da | 479 | /* if we find a valid info-block we'll come back as that personality */ |
c5ed9268 DW |
480 | if (nd_btt_probe(dev, ndns) == 0 || nd_pfn_probe(dev, ndns) == 0 |
481 | || nd_dax_probe(dev, ndns) == 0) | |
32ab0a3f | 482 | return -ENXIO; |
32ab0a3f | 483 | |
200c79da DW |
484 | /* ...otherwise we're just a raw pmem device */ |
485 | return pmem_attach_disk(dev, ndns); | |
9e853f23 RZ |
486 | } |
487 | ||
9f53f9fa | 488 | static int nd_pmem_remove(struct device *dev) |
9e853f23 | 489 | { |
6aa734a2 DW |
490 | struct pmem_device *pmem = dev_get_drvdata(dev); |
491 | ||
8c2f7e86 | 492 | if (is_nd_btt(dev)) |
298f2bc5 | 493 | nvdimm_namespace_detach_btt(to_nd_btt(dev)); |
6aa734a2 DW |
494 | else { |
495 | /* | |
496 | * Note, this assumes device_lock() context to not race | |
497 | * nd_pmem_notify() | |
498 | */ | |
499 | sysfs_put(pmem->bb_state); | |
500 | pmem->bb_state = NULL; | |
501 | } | |
476f848a DW |
502 | nvdimm_flush(to_nd_region(dev->parent)); |
503 | ||
9e853f23 RZ |
504 | return 0; |
505 | } | |
506 | ||
476f848a DW |
507 | static void nd_pmem_shutdown(struct device *dev) |
508 | { | |
509 | nvdimm_flush(to_nd_region(dev->parent)); | |
510 | } | |
511 | ||
71999466 DW |
512 | static void nd_pmem_notify(struct device *dev, enum nvdimm_event event) |
513 | { | |
b2518c78 | 514 | struct nd_region *nd_region; |
298f2bc5 DW |
515 | resource_size_t offset = 0, end_trunc = 0; |
516 | struct nd_namespace_common *ndns; | |
517 | struct nd_namespace_io *nsio; | |
518 | struct resource res; | |
b2518c78 | 519 | struct badblocks *bb; |
975750a9 | 520 | struct kernfs_node *bb_state; |
71999466 DW |
521 | |
522 | if (event != NVDIMM_REVALIDATE_POISON) | |
523 | return; | |
524 | ||
298f2bc5 DW |
525 | if (is_nd_btt(dev)) { |
526 | struct nd_btt *nd_btt = to_nd_btt(dev); | |
527 | ||
528 | ndns = nd_btt->ndns; | |
b2518c78 TK |
529 | nd_region = to_nd_region(ndns->dev.parent); |
530 | nsio = to_nd_namespace_io(&ndns->dev); | |
531 | bb = &nsio->bb; | |
975750a9 | 532 | bb_state = NULL; |
b2518c78 TK |
533 | } else { |
534 | struct pmem_device *pmem = dev_get_drvdata(dev); | |
a3901802 | 535 | |
b2518c78 TK |
536 | nd_region = to_region(pmem); |
537 | bb = &pmem->bb; | |
975750a9 | 538 | bb_state = pmem->bb_state; |
b2518c78 TK |
539 | |
540 | if (is_nd_pfn(dev)) { | |
541 | struct nd_pfn *nd_pfn = to_nd_pfn(dev); | |
542 | struct nd_pfn_sb *pfn_sb = nd_pfn->pfn_sb; | |
543 | ||
544 | ndns = nd_pfn->ndns; | |
545 | offset = pmem->data_offset + | |
546 | __le32_to_cpu(pfn_sb->start_pad); | |
547 | end_trunc = __le32_to_cpu(pfn_sb->end_trunc); | |
548 | } else { | |
549 | ndns = to_ndns(dev); | |
550 | } | |
551 | ||
552 | nsio = to_nd_namespace_io(&ndns->dev); | |
553 | } | |
a3901802 | 554 | |
298f2bc5 DW |
555 | res.start = nsio->res.start + offset; |
556 | res.end = nsio->res.end - end_trunc; | |
b2518c78 | 557 | nvdimm_badblocks_populate(nd_region, bb, &res); |
975750a9 TK |
558 | if (bb_state) |
559 | sysfs_notify_dirent(bb_state); | |
71999466 DW |
560 | } |
561 | ||
9f53f9fa DW |
562 | MODULE_ALIAS("pmem"); |
563 | MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_IO); | |
bf9bccc1 | 564 | MODULE_ALIAS_ND_DEVICE(ND_DEVICE_NAMESPACE_PMEM); |
9f53f9fa DW |
565 | static struct nd_device_driver nd_pmem_driver = { |
566 | .probe = nd_pmem_probe, | |
567 | .remove = nd_pmem_remove, | |
71999466 | 568 | .notify = nd_pmem_notify, |
476f848a | 569 | .shutdown = nd_pmem_shutdown, |
9f53f9fa DW |
570 | .drv = { |
571 | .name = "nd_pmem", | |
9e853f23 | 572 | }, |
bf9bccc1 | 573 | .type = ND_DRIVER_NAMESPACE_IO | ND_DRIVER_NAMESPACE_PMEM, |
9e853f23 RZ |
574 | }; |
575 | ||
03e90843 | 576 | module_nd_driver(nd_pmem_driver); |
9e853f23 RZ |
577 | |
578 | MODULE_AUTHOR("Ross Zwisler <ross.zwisler@linux.intel.com>"); | |
579 | MODULE_LICENSE("GPL v2"); |