1 /*
2  * bio-integrity.c - bio data integrity extensions
3  *
4  * Copyright (C) 2007, 2008, 2009 Oracle Corporation
5  * Written by: Martin K. Petersen <martin.petersen@oracle.com>
6  *
7  * This program is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU General Public License version
9  * 2 as published by the Free Software Foundation.
10  *
11  * This program is distributed in the hope that it will be useful, but
12  * WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License
17  * along with this program; see the file COPYING.  If not, write to
18  * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139,
19  * USA.
20  *
21  */
22 
23 #include <linux/blkdev.h>
24 #include <linux/mempool.h>
25 #include <linux/export.h>
26 #include <linux/bio.h>
27 #include <linux/workqueue.h>
28 #include <linux/slab.h>
29 #include "blk.h"
30 
31 #define BIP_INLINE_VECS	4
32 
33 static struct kmem_cache *bip_slab;
34 static struct workqueue_struct *kintegrityd_wq;
35 
blk_flush_integrity(void)36 void blk_flush_integrity(void)
37 {
38 	flush_workqueue(kintegrityd_wq);
39 }
40 
__bio_integrity_free(struct bio_set * bs,struct bio_integrity_payload * bip)41 void __bio_integrity_free(struct bio_set *bs, struct bio_integrity_payload *bip)
42 {
43 	if (bs && mempool_initialized(&bs->bio_integrity_pool)) {
44 		if (bip->bip_vec)
45 			bvec_free(&bs->bvec_integrity_pool, bip->bip_vec,
46 				  bip->bip_slab);
47 		mempool_free(bip, &bs->bio_integrity_pool);
48 	} else {
49 		kfree(bip);
50 	}
51 }
52 
53 /**
54  * bio_integrity_alloc - Allocate integrity payload and attach it to bio
55  * @bio:	bio to attach integrity metadata to
56  * @gfp_mask:	Memory allocation mask
57  * @nr_vecs:	Number of integrity metadata scatter-gather elements
58  *
59  * Description: This function prepares a bio for attaching integrity
60  * metadata.  nr_vecs specifies the maximum number of pages containing
61  * integrity metadata that can be attached.
62  */
bio_integrity_alloc(struct bio * bio,gfp_t gfp_mask,unsigned int nr_vecs)63 struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio,
64 						  gfp_t gfp_mask,
65 						  unsigned int nr_vecs)
66 {
67 	struct bio_integrity_payload *bip;
68 	struct bio_set *bs = bio->bi_pool;
69 	unsigned inline_vecs;
70 
71 	if (!bs || !mempool_initialized(&bs->bio_integrity_pool)) {
72 		bip = kmalloc(sizeof(struct bio_integrity_payload) +
73 			      sizeof(struct bio_vec) * nr_vecs, gfp_mask);
74 		inline_vecs = nr_vecs;
75 	} else {
76 		bip = mempool_alloc(&bs->bio_integrity_pool, gfp_mask);
77 		inline_vecs = BIP_INLINE_VECS;
78 	}
79 
80 	if (unlikely(!bip))
81 		return ERR_PTR(-ENOMEM);
82 
83 	memset(bip, 0, sizeof(*bip));
84 
85 	if (nr_vecs > inline_vecs) {
86 		unsigned long idx = 0;
87 
88 		bip->bip_vec = bvec_alloc(gfp_mask, nr_vecs, &idx,
89 					  &bs->bvec_integrity_pool);
90 		if (!bip->bip_vec)
91 			goto err;
92 		bip->bip_max_vcnt = bvec_nr_vecs(idx);
93 		bip->bip_slab = idx;
94 	} else {
95 		bip->bip_vec = bip->bip_inline_vecs;
96 		bip->bip_max_vcnt = inline_vecs;
97 	}
98 
99 	bip->bip_bio = bio;
100 	bio->bi_integrity = bip;
101 	bio->bi_opf |= REQ_INTEGRITY;
102 
103 	return bip;
104 err:
105 	__bio_integrity_free(bs, bip);
106 	return ERR_PTR(-ENOMEM);
107 }
108 EXPORT_SYMBOL(bio_integrity_alloc);
109 
110 /**
111  * bio_integrity_free - Free bio integrity payload
112  * @bio:	bio containing bip to be freed
113  *
114  * Description: Used to free the integrity portion of a bio. Usually
115  * called from bio_free().
116  */
bio_integrity_free(struct bio * bio)117 static void bio_integrity_free(struct bio *bio)
118 {
119 	struct bio_integrity_payload *bip = bio_integrity(bio);
120 	struct bio_set *bs = bio->bi_pool;
121 
122 	if (bip->bip_flags & BIP_BLOCK_INTEGRITY)
123 		kfree(page_address(bip->bip_vec->bv_page) +
124 		      bip->bip_vec->bv_offset);
125 
126 	__bio_integrity_free(bs, bip);
127 	bio->bi_integrity = NULL;
128 	bio->bi_opf &= ~REQ_INTEGRITY;
129 }
130 
131 /**
132  * bio_integrity_add_page - Attach integrity metadata
133  * @bio:	bio to update
134  * @page:	page containing integrity metadata
135  * @len:	number of bytes of integrity metadata in page
136  * @offset:	start offset within page
137  *
138  * Description: Attach a page containing integrity metadata to bio.
139  */
bio_integrity_add_page(struct bio * bio,struct page * page,unsigned int len,unsigned int offset)140 int bio_integrity_add_page(struct bio *bio, struct page *page,
141 			   unsigned int len, unsigned int offset)
142 {
143 	struct bio_integrity_payload *bip = bio_integrity(bio);
144 	struct bio_vec *iv;
145 
146 	if (bip->bip_vcnt >= bip->bip_max_vcnt) {
147 		printk(KERN_ERR "%s: bip_vec full\n", __func__);
148 		return 0;
149 	}
150 
151 	iv = bip->bip_vec + bip->bip_vcnt;
152 
153 	if (bip->bip_vcnt &&
154 	    bvec_gap_to_prev(bio->bi_disk->queue,
155 			     &bip->bip_vec[bip->bip_vcnt - 1], offset))
156 		return 0;
157 
158 	iv->bv_page = page;
159 	iv->bv_len = len;
160 	iv->bv_offset = offset;
161 	bip->bip_vcnt++;
162 
163 	return len;
164 }
165 EXPORT_SYMBOL(bio_integrity_add_page);
166 
167 /**
168  * bio_integrity_process - Process integrity metadata for a bio
169  * @bio:	bio to generate/verify integrity metadata for
170  * @proc_iter:  iterator to process
171  * @proc_fn:	Pointer to the relevant processing function
172  */
bio_integrity_process(struct bio * bio,struct bvec_iter * proc_iter,integrity_processing_fn * proc_fn)173 static blk_status_t bio_integrity_process(struct bio *bio,
174 		struct bvec_iter *proc_iter, integrity_processing_fn *proc_fn)
175 {
176 	struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
177 	struct blk_integrity_iter iter;
178 	struct bvec_iter bviter;
179 	struct bio_vec bv;
180 	struct bio_integrity_payload *bip = bio_integrity(bio);
181 	blk_status_t ret = BLK_STS_OK;
182 	void *prot_buf = page_address(bip->bip_vec->bv_page) +
183 		bip->bip_vec->bv_offset;
184 
185 	iter.disk_name = bio->bi_disk->disk_name;
186 	iter.interval = 1 << bi->interval_exp;
187 	iter.seed = proc_iter->bi_sector;
188 	iter.prot_buf = prot_buf;
189 
190 	__bio_for_each_segment(bv, bio, bviter, *proc_iter) {
191 		void *kaddr = kmap_atomic(bv.bv_page);
192 
193 		iter.data_buf = kaddr + bv.bv_offset;
194 		iter.data_size = bv.bv_len;
195 
196 		ret = proc_fn(&iter);
197 		if (ret) {
198 			kunmap_atomic(kaddr);
199 			return ret;
200 		}
201 
202 		kunmap_atomic(kaddr);
203 	}
204 	return ret;
205 }
206 
207 /**
208  * bio_integrity_prep - Prepare bio for integrity I/O
209  * @bio:	bio to prepare
210  *
211  * Description:  Checks if the bio already has an integrity payload attached.
212  * If it does, the payload has been generated by another kernel subsystem,
213  * and we just pass it through. Otherwise allocates integrity payload.
214  * The bio must have data direction, target device and start sector set priot
215  * to calling.  In the WRITE case, integrity metadata will be generated using
216  * the block device's integrity function.  In the READ case, the buffer
217  * will be prepared for DMA and a suitable end_io handler set up.
218  */
bio_integrity_prep(struct bio * bio)219 bool bio_integrity_prep(struct bio *bio)
220 {
221 	struct bio_integrity_payload *bip;
222 	struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
223 	struct request_queue *q = bio->bi_disk->queue;
224 	void *buf;
225 	unsigned long start, end;
226 	unsigned int len, nr_pages;
227 	unsigned int bytes, offset, i;
228 	unsigned int intervals;
229 	blk_status_t status;
230 
231 	if (!bi)
232 		return true;
233 
234 	if (bio_op(bio) != REQ_OP_READ && bio_op(bio) != REQ_OP_WRITE)
235 		return true;
236 
237 	if (!bio_sectors(bio))
238 		return true;
239 
240 	/* Already protected? */
241 	if (bio_integrity(bio))
242 		return true;
243 
244 	if (bio_data_dir(bio) == READ) {
245 		if (!bi->profile->verify_fn ||
246 		    !(bi->flags & BLK_INTEGRITY_VERIFY))
247 			return true;
248 	} else {
249 		if (!bi->profile->generate_fn ||
250 		    !(bi->flags & BLK_INTEGRITY_GENERATE))
251 			return true;
252 	}
253 	intervals = bio_integrity_intervals(bi, bio_sectors(bio));
254 
255 	/* Allocate kernel buffer for protection data */
256 	len = intervals * bi->tuple_size;
257 	buf = kmalloc(len, GFP_NOIO | q->bounce_gfp);
258 	status = BLK_STS_RESOURCE;
259 	if (unlikely(buf == NULL)) {
260 		printk(KERN_ERR "could not allocate integrity buffer\n");
261 		goto err_end_io;
262 	}
263 
264 	end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
265 	start = ((unsigned long) buf) >> PAGE_SHIFT;
266 	nr_pages = end - start;
267 
268 	/* Allocate bio integrity payload and integrity vectors */
269 	bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages);
270 	if (IS_ERR(bip)) {
271 		printk(KERN_ERR "could not allocate data integrity bioset\n");
272 		kfree(buf);
273 		status = BLK_STS_RESOURCE;
274 		goto err_end_io;
275 	}
276 
277 	bip->bip_flags |= BIP_BLOCK_INTEGRITY;
278 	bip->bip_iter.bi_size = len;
279 	bip_set_seed(bip, bio->bi_iter.bi_sector);
280 
281 	if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM)
282 		bip->bip_flags |= BIP_IP_CHECKSUM;
283 
284 	/* Map it */
285 	offset = offset_in_page(buf);
286 	for (i = 0 ; i < nr_pages ; i++) {
287 		int ret;
288 		bytes = PAGE_SIZE - offset;
289 
290 		if (len <= 0)
291 			break;
292 
293 		if (bytes > len)
294 			bytes = len;
295 
296 		ret = bio_integrity_add_page(bio, virt_to_page(buf),
297 					     bytes, offset);
298 
299 		if (ret == 0) {
300 			printk(KERN_ERR "could not attach integrity payload\n");
301 			status = BLK_STS_RESOURCE;
302 			goto err_end_io;
303 		}
304 
305 		if (ret < bytes)
306 			break;
307 
308 		buf += bytes;
309 		len -= bytes;
310 		offset = 0;
311 	}
312 
313 	/* Auto-generate integrity metadata if this is a write */
314 	if (bio_data_dir(bio) == WRITE) {
315 		bio_integrity_process(bio, &bio->bi_iter,
316 				      bi->profile->generate_fn);
317 	}
318 	return true;
319 
320 err_end_io:
321 	bio->bi_status = status;
322 	bio_endio(bio);
323 	return false;
324 
325 }
326 EXPORT_SYMBOL(bio_integrity_prep);
327 
328 /**
329  * bio_integrity_verify_fn - Integrity I/O completion worker
330  * @work:	Work struct stored in bio to be verified
331  *
332  * Description: This workqueue function is called to complete a READ
333  * request.  The function verifies the transferred integrity metadata
334  * and then calls the original bio end_io function.
335  */
bio_integrity_verify_fn(struct work_struct * work)336 static void bio_integrity_verify_fn(struct work_struct *work)
337 {
338 	struct bio_integrity_payload *bip =
339 		container_of(work, struct bio_integrity_payload, bip_work);
340 	struct bio *bio = bip->bip_bio;
341 	struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
342 	struct bvec_iter iter = bio->bi_iter;
343 
344 	/*
345 	 * At the moment verify is called bio's iterator was advanced
346 	 * during split and completion, we need to rewind iterator to
347 	 * it's original position.
348 	 */
349 	if (bio_rewind_iter(bio, &iter, iter.bi_done)) {
350 		bio->bi_status = bio_integrity_process(bio, &iter,
351 						       bi->profile->verify_fn);
352 	} else {
353 		bio->bi_status = BLK_STS_IOERR;
354 	}
355 
356 	bio_integrity_free(bio);
357 	bio_endio(bio);
358 }
359 
360 /**
361  * __bio_integrity_endio - Integrity I/O completion function
362  * @bio:	Protected bio
363  *
364  * Description: Completion for integrity I/O
365  *
366  * Normally I/O completion is done in interrupt context.  However,
367  * verifying I/O integrity is a time-consuming task which must be run
368  * in process context.	This function postpones completion
369  * accordingly.
370  */
__bio_integrity_endio(struct bio * bio)371 bool __bio_integrity_endio(struct bio *bio)
372 {
373 	struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
374 	struct bio_integrity_payload *bip = bio_integrity(bio);
375 
376 	if (bio_op(bio) == REQ_OP_READ && !bio->bi_status &&
377 	    (bip->bip_flags & BIP_BLOCK_INTEGRITY) && bi->profile->verify_fn) {
378 		INIT_WORK(&bip->bip_work, bio_integrity_verify_fn);
379 		queue_work(kintegrityd_wq, &bip->bip_work);
380 		return false;
381 	}
382 
383 	bio_integrity_free(bio);
384 	return true;
385 }
386 
387 /**
388  * bio_integrity_advance - Advance integrity vector
389  * @bio:	bio whose integrity vector to update
390  * @bytes_done:	number of data bytes that have been completed
391  *
392  * Description: This function calculates how many integrity bytes the
393  * number of completed data bytes correspond to and advances the
394  * integrity vector accordingly.
395  */
bio_integrity_advance(struct bio * bio,unsigned int bytes_done)396 void bio_integrity_advance(struct bio *bio, unsigned int bytes_done)
397 {
398 	struct bio_integrity_payload *bip = bio_integrity(bio);
399 	struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
400 	unsigned bytes = bio_integrity_bytes(bi, bytes_done >> 9);
401 
402 	bip->bip_iter.bi_sector += bio_integrity_intervals(bi, bytes_done >> 9);
403 	bvec_iter_advance(bip->bip_vec, &bip->bip_iter, bytes);
404 }
405 EXPORT_SYMBOL(bio_integrity_advance);
406 
407 /**
408  * bio_integrity_trim - Trim integrity vector
409  * @bio:	bio whose integrity vector to update
410  *
411  * Description: Used to trim the integrity vector in a cloned bio.
412  */
bio_integrity_trim(struct bio * bio)413 void bio_integrity_trim(struct bio *bio)
414 {
415 	struct bio_integrity_payload *bip = bio_integrity(bio);
416 	struct blk_integrity *bi = blk_get_integrity(bio->bi_disk);
417 
418 	bip->bip_iter.bi_size = bio_integrity_bytes(bi, bio_sectors(bio));
419 }
420 EXPORT_SYMBOL(bio_integrity_trim);
421 
422 /**
423  * bio_integrity_clone - Callback for cloning bios with integrity metadata
424  * @bio:	New bio
425  * @bio_src:	Original bio
426  * @gfp_mask:	Memory allocation mask
427  *
428  * Description:	Called to allocate a bip when cloning a bio
429  */
bio_integrity_clone(struct bio * bio,struct bio * bio_src,gfp_t gfp_mask)430 int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
431 			gfp_t gfp_mask)
432 {
433 	struct bio_integrity_payload *bip_src = bio_integrity(bio_src);
434 	struct bio_integrity_payload *bip;
435 
436 	BUG_ON(bip_src == NULL);
437 
438 	bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt);
439 	if (IS_ERR(bip))
440 		return PTR_ERR(bip);
441 
442 	memcpy(bip->bip_vec, bip_src->bip_vec,
443 	       bip_src->bip_vcnt * sizeof(struct bio_vec));
444 
445 	bip->bip_vcnt = bip_src->bip_vcnt;
446 	bip->bip_iter = bip_src->bip_iter;
447 	bip->bip_flags = bip_src->bip_flags & ~BIP_BLOCK_INTEGRITY;
448 
449 	return 0;
450 }
451 EXPORT_SYMBOL(bio_integrity_clone);
452 
bioset_integrity_create(struct bio_set * bs,int pool_size)453 int bioset_integrity_create(struct bio_set *bs, int pool_size)
454 {
455 	if (mempool_initialized(&bs->bio_integrity_pool))
456 		return 0;
457 
458 	if (mempool_init_slab_pool(&bs->bio_integrity_pool,
459 				   pool_size, bip_slab))
460 		return -1;
461 
462 	if (biovec_init_pool(&bs->bvec_integrity_pool, pool_size)) {
463 		mempool_exit(&bs->bio_integrity_pool);
464 		return -1;
465 	}
466 
467 	return 0;
468 }
469 EXPORT_SYMBOL(bioset_integrity_create);
470 
bioset_integrity_free(struct bio_set * bs)471 void bioset_integrity_free(struct bio_set *bs)
472 {
473 	mempool_exit(&bs->bio_integrity_pool);
474 	mempool_exit(&bs->bvec_integrity_pool);
475 }
476 EXPORT_SYMBOL(bioset_integrity_free);
477 
bio_integrity_init(void)478 void __init bio_integrity_init(void)
479 {
480 	/*
481 	 * kintegrityd won't block much but may burn a lot of CPU cycles.
482 	 * Make it highpri CPU intensive wq with max concurrency of 1.
483 	 */
484 	kintegrityd_wq = alloc_workqueue("kintegrityd", WQ_MEM_RECLAIM |
485 					 WQ_HIGHPRI | WQ_CPU_INTENSIVE, 1);
486 	if (!kintegrityd_wq)
487 		panic("Failed to create kintegrityd\n");
488 
489 	bip_slab = kmem_cache_create("bio_integrity_payload",
490 				     sizeof(struct bio_integrity_payload) +
491 				     sizeof(struct bio_vec) * BIP_INLINE_VECS,
492 				     0, SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
493 }
494