forked from ~ljy/RK356X_SDK_RELEASE

blame | history | patch | commit | commitdiff
hc
2024-12-19 9370bb92b2d16684ee45cf24e879c93c509162da 
 kernel/crypto/async_tx/async_raid6_recov.c
....@@ -1,24 +1,10 @@
1
+// SPDX-License-Identifier: GPL-2.0-or-later
12 /*
23  * Asynchronous RAID-6 recovery calculations ASYNC_TX API.
34  * Copyright(c) 2009 Intel Corporation
45  *
56  * based on raid6recov.c:
67  *   Copyright 2002 H. Peter Anvin
7
- *
8
- * This program is free software; you can redistribute it and/or modify it
9
- * under the terms of the GNU General Public License as published by the Free
10
- * Software Foundation; either version 2 of the License, or (at your option)
11
- * any later version.
12
- *
13
- * This program is distributed in the hope that it will be useful, but WITHOUT
14
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
15
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
16
- * more details.
17
- *
18
- * You should have received a copy of the GNU General Public License along with
19
- * this program; if not, write to the Free Software Foundation, Inc., 51
20
- * Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
21
- *
228  */
239 #include <linux/kernel.h>
2410 #include <linux/interrupt.h>
....@@ -29,8 +15,9 @@
2915 #include <linux/dmaengine.h>
3016 
3117 static struct dma_async_tx_descriptor *
32
-async_sum_product(struct page *dest, struct page **srcs, unsigned char *coef,
33
-		  size_t len, struct async_submit_ctl *submit)
18
+async_sum_product(struct page *dest, unsigned int d_off,
19
+		struct page **srcs, unsigned int *src_offs, unsigned char *coef,
20
+		size_t len, struct async_submit_ctl *submit)
3421 {
3522 	struct dma_chan *chan = async_tx_find_channel(submit, DMA_PQ,
3623 						      &dest, 1, srcs, 2, len);
....@@ -51,11 +38,14 @@
5138 
5239 		if (submit->flags & ASYNC_TX_FENCE)
5340 			dma_flags |= DMA_PREP_FENCE;
54
-		unmap->addr[0] = dma_map_page(dev, srcs[0], 0, len, DMA_TO_DEVICE);
55
-		unmap->addr[1] = dma_map_page(dev, srcs[1], 0, len, DMA_TO_DEVICE);
41
+		unmap->addr[0] = dma_map_page(dev, srcs[0], src_offs[0],
42
+						len, DMA_TO_DEVICE);
43
+		unmap->addr[1] = dma_map_page(dev, srcs[1], src_offs[1],
44
+						len, DMA_TO_DEVICE);
5645 		unmap->to_cnt = 2;
5746 
58
-		unmap->addr[2] = dma_map_page(dev, dest, 0, len, DMA_BIDIRECTIONAL);
47
+		unmap->addr[2] = dma_map_page(dev, dest, d_off,
48
+						len, DMA_BIDIRECTIONAL);
5949 		unmap->bidi_cnt = 1;
6050 		/* engine only looks at Q, but expects it to follow P */
6151 		pq[1] = unmap->addr[2];
....@@ -80,9 +70,9 @@
8070 	async_tx_quiesce(&submit->depend_tx);
8171 	amul = raid6_gfmul[coef[0]];
8272 	bmul = raid6_gfmul[coef[1]];
83
-	a = page_address(srcs[0]);
84
-	b = page_address(srcs[1]);
85
-	c = page_address(dest);
73
+	a = page_address(srcs[0]) + src_offs[0];
74
+	b = page_address(srcs[1]) + src_offs[1];
75
+	c = page_address(dest) + d_off;
8676 
8777 	while (len--) {
8878 		ax    = amul[*a++];
....@@ -94,8 +84,9 @@
9484 }
9585 
9686 static struct dma_async_tx_descriptor *
97
-async_mult(struct page *dest, struct page *src, u8 coef, size_t len,
98
-	   struct async_submit_ctl *submit)
87
+async_mult(struct page *dest, unsigned int d_off, struct page *src,
88
+		unsigned int s_off, u8 coef, size_t len,
89
+		struct async_submit_ctl *submit)
9990 {
10091 	struct dma_chan *chan = async_tx_find_channel(submit, DMA_PQ,
10192 						      &dest, 1, &src, 1, len);
....@@ -115,9 +106,11 @@
115106 
116107 		if (submit->flags & ASYNC_TX_FENCE)
117108 			dma_flags |= DMA_PREP_FENCE;
118
-		unmap->addr[0] = dma_map_page(dev, src, 0, len, DMA_TO_DEVICE);
109
+		unmap->addr[0] = dma_map_page(dev, src, s_off,
110
+						len, DMA_TO_DEVICE);
119111 		unmap->to_cnt++;
120
-		unmap->addr[1] = dma_map_page(dev, dest, 0, len, DMA_BIDIRECTIONAL);
112
+		unmap->addr[1] = dma_map_page(dev, dest, d_off,
113
+						len, DMA_BIDIRECTIONAL);
121114 		dma_dest[1] = unmap->addr[1];
122115 		unmap->bidi_cnt++;
123116 		unmap->len = len;
....@@ -147,8 +140,8 @@
147140 	 */
148141 	async_tx_quiesce(&submit->depend_tx);
149142 	qmul  = raid6_gfmul[coef];
150
-	d = page_address(dest);
151
-	s = page_address(src);
143
+	d = page_address(dest) + d_off;
144
+	s = page_address(src) + s_off;
152145 
153146 	while (len--)
154147 		*d++ = qmul[*s++];
....@@ -158,11 +151,14 @@
158151 
159152 static struct dma_async_tx_descriptor *
160153 __2data_recov_4(int disks, size_t bytes, int faila, int failb,
161
-		struct page **blocks, struct async_submit_ctl *submit)
154
+		struct page **blocks, unsigned int *offs,
155
+		struct async_submit_ctl *submit)
162156 {
163157 	struct dma_async_tx_descriptor *tx = NULL;
164158 	struct page *p, *q, *a, *b;
159
+	unsigned int p_off, q_off, a_off, b_off;
165160 	struct page *srcs[2];
161
+	unsigned int src_offs[2];
166162 	unsigned char coef[2];
167163 	enum async_tx_flags flags = submit->flags;
168164 	dma_async_tx_callback cb_fn = submit->cb_fn;
....@@ -170,26 +166,34 @@
170166 	void *scribble = submit->scribble;
171167 
172168 	p = blocks[disks-2];
169
+	p_off = offs[disks-2];
173170 	q = blocks[disks-1];
171
+	q_off = offs[disks-1];
174172 
175173 	a = blocks[faila];
174
+	a_off = offs[faila];
176175 	b = blocks[failb];
176
+	b_off = offs[failb];
177177 
178178 	/* in the 4 disk case P + Pxy == P and Q + Qxy == Q */
179179 	/* Dx = A*(P+Pxy) + B*(Q+Qxy) */
180180 	srcs[0] = p;
181
+	src_offs[0] = p_off;
181182 	srcs[1] = q;
183
+	src_offs[1] = q_off;
182184 	coef[0] = raid6_gfexi[failb-faila];
183185 	coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]];
184186 	init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble);
185
-	tx = async_sum_product(b, srcs, coef, bytes, submit);
187
+	tx = async_sum_product(b, b_off, srcs, src_offs, coef, bytes, submit);
186188 
187189 	/* Dy = P+Pxy+Dx */
188190 	srcs[0] = p;
191
+	src_offs[0] = p_off;
189192 	srcs[1] = b;
193
+	src_offs[1] = b_off;
190194 	init_async_submit(submit, flags | ASYNC_TX_XOR_ZERO_DST, tx, cb_fn,
191195 			  cb_param, scribble);
192
-	tx = async_xor(a, srcs, 0, 2, bytes, submit);
196
+	tx = async_xor_offs(a, a_off, srcs, src_offs, 2, bytes, submit);
193197 
194198 	return tx;
195199 
....@@ -197,11 +201,14 @@
197201 
198202 static struct dma_async_tx_descriptor *
199203 __2data_recov_5(int disks, size_t bytes, int faila, int failb,
200
-		struct page **blocks, struct async_submit_ctl *submit)
204
+		struct page **blocks, unsigned int *offs,
205
+		struct async_submit_ctl *submit)
201206 {
202207 	struct dma_async_tx_descriptor *tx = NULL;
203208 	struct page *p, *q, *g, *dp, *dq;
209
+	unsigned int p_off, q_off, g_off, dp_off, dq_off;
204210 	struct page *srcs[2];
211
+	unsigned int src_offs[2];
205212 	unsigned char coef[2];
206213 	enum async_tx_flags flags = submit->flags;
207214 	dma_async_tx_callback cb_fn = submit->cb_fn;
....@@ -222,60 +229,77 @@
222229 	BUG_ON(good_srcs > 1);
223230 
224231 	p = blocks[disks-2];
232
+	p_off = offs[disks-2];
225233 	q = blocks[disks-1];
234
+	q_off = offs[disks-1];
226235 	g = blocks[good];
236
+	g_off = offs[good];
227237 
228238 	/* Compute syndrome with zero for the missing data pages
229239 	 * Use the dead data pages as temporary storage for delta p and
230240 	 * delta q
231241 	 */
232242 	dp = blocks[faila];
243
+	dp_off = offs[faila];
233244 	dq = blocks[failb];
245
+	dq_off = offs[failb];
234246 
235247 	init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble);
236
-	tx = async_memcpy(dp, g, 0, 0, bytes, submit);
248
+	tx = async_memcpy(dp, g, dp_off, g_off, bytes, submit);
237249 	init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble);
238
-	tx = async_mult(dq, g, raid6_gfexp[good], bytes, submit);
250
+	tx = async_mult(dq, dq_off, g, g_off,
251
+			raid6_gfexp[good], bytes, submit);
239252 
240253 	/* compute P + Pxy */
241254 	srcs[0] = dp;
255
+	src_offs[0] = dp_off;
242256 	srcs[1] = p;
257
+	src_offs[1] = p_off;
243258 	init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
244259 			  NULL, NULL, scribble);
245
-	tx = async_xor(dp, srcs, 0, 2, bytes, submit);
260
+	tx = async_xor_offs(dp, dp_off, srcs, src_offs, 2, bytes, submit);
246261 
247262 	/* compute Q + Qxy */
248263 	srcs[0] = dq;
264
+	src_offs[0] = dq_off;
249265 	srcs[1] = q;
266
+	src_offs[1] = q_off;
250267 	init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
251268 			  NULL, NULL, scribble);
252
-	tx = async_xor(dq, srcs, 0, 2, bytes, submit);
269
+	tx = async_xor_offs(dq, dq_off, srcs, src_offs, 2, bytes, submit);
253270 
254271 	/* Dx = A*(P+Pxy) + B*(Q+Qxy) */
255272 	srcs[0] = dp;
273
+	src_offs[0] = dp_off;
256274 	srcs[1] = dq;
275
+	src_offs[1] = dq_off;
257276 	coef[0] = raid6_gfexi[failb-faila];
258277 	coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]];
259278 	init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble);
260
-	tx = async_sum_product(dq, srcs, coef, bytes, submit);
279
+	tx = async_sum_product(dq, dq_off, srcs, src_offs, coef, bytes, submit);
261280 
262281 	/* Dy = P+Pxy+Dx */
263282 	srcs[0] = dp;
283
+	src_offs[0] = dp_off;
264284 	srcs[1] = dq;
285
+	src_offs[1] = dq_off;
265286 	init_async_submit(submit, flags | ASYNC_TX_XOR_DROP_DST, tx, cb_fn,
266287 			  cb_param, scribble);
267
-	tx = async_xor(dp, srcs, 0, 2, bytes, submit);
288
+	tx = async_xor_offs(dp, dp_off, srcs, src_offs, 2, bytes, submit);
268289 
269290 	return tx;
270291 }
271292 
272293 static struct dma_async_tx_descriptor *
273294 __2data_recov_n(int disks, size_t bytes, int faila, int failb,
274
-	      struct page **blocks, struct async_submit_ctl *submit)
295
+	      struct page **blocks, unsigned int *offs,
296
+		  struct async_submit_ctl *submit)
275297 {
276298 	struct dma_async_tx_descriptor *tx = NULL;
277299 	struct page *p, *q, *dp, *dq;
300
+	unsigned int p_off, q_off, dp_off, dq_off;
278301 	struct page *srcs[2];
302
+	unsigned int src_offs[2];
279303 	unsigned char coef[2];
280304 	enum async_tx_flags flags = submit->flags;
281305 	dma_async_tx_callback cb_fn = submit->cb_fn;
....@@ -283,56 +307,74 @@
283307 	void *scribble = submit->scribble;
284308 
285309 	p = blocks[disks-2];
310
+	p_off = offs[disks-2];
286311 	q = blocks[disks-1];
312
+	q_off = offs[disks-1];
287313 
288314 	/* Compute syndrome with zero for the missing data pages
289315 	 * Use the dead data pages as temporary storage for
290316 	 * delta p and delta q
291317 	 */
292318 	dp = blocks[faila];
319
+	dp_off = offs[faila];
293320 	blocks[faila] = NULL;
294321 	blocks[disks-2] = dp;
322
+	offs[disks-2] = dp_off;
295323 	dq = blocks[failb];
324
+	dq_off = offs[failb];
296325 	blocks[failb] = NULL;
297326 	blocks[disks-1] = dq;
327
+	offs[disks-1] = dq_off;
298328 
299329 	init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble);
300
-	tx = async_gen_syndrome(blocks, 0, disks, bytes, submit);
330
+	tx = async_gen_syndrome(blocks, offs, disks, bytes, submit);
301331 
302332 	/* Restore pointer table */
303333 	blocks[faila]   = dp;
334
+	offs[faila] = dp_off;
304335 	blocks[failb]   = dq;
336
+	offs[failb] = dq_off;
305337 	blocks[disks-2] = p;
338
+	offs[disks-2] = p_off;
306339 	blocks[disks-1] = q;
340
+	offs[disks-1] = q_off;
307341 
308342 	/* compute P + Pxy */
309343 	srcs[0] = dp;
344
+	src_offs[0] = dp_off;
310345 	srcs[1] = p;
346
+	src_offs[1] = p_off;
311347 	init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
312348 			  NULL, NULL, scribble);
313
-	tx = async_xor(dp, srcs, 0, 2, bytes, submit);
349
+	tx = async_xor_offs(dp, dp_off, srcs, src_offs, 2, bytes, submit);
314350 
315351 	/* compute Q + Qxy */
316352 	srcs[0] = dq;
353
+	src_offs[0] = dq_off;
317354 	srcs[1] = q;
355
+	src_offs[1] = q_off;
318356 	init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
319357 			  NULL, NULL, scribble);
320
-	tx = async_xor(dq, srcs, 0, 2, bytes, submit);
358
+	tx = async_xor_offs(dq, dq_off, srcs, src_offs, 2, bytes, submit);
321359 
322360 	/* Dx = A*(P+Pxy) + B*(Q+Qxy) */
323361 	srcs[0] = dp;
362
+	src_offs[0] = dp_off;
324363 	srcs[1] = dq;
364
+	src_offs[1] = dq_off;
325365 	coef[0] = raid6_gfexi[failb-faila];
326366 	coef[1] = raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]];
327367 	init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble);
328
-	tx = async_sum_product(dq, srcs, coef, bytes, submit);
368
+	tx = async_sum_product(dq, dq_off, srcs, src_offs, coef, bytes, submit);
329369 
330370 	/* Dy = P+Pxy+Dx */
331371 	srcs[0] = dp;
372
+	src_offs[0] = dp_off;
332373 	srcs[1] = dq;
374
+	src_offs[1] = dq_off;
333375 	init_async_submit(submit, flags | ASYNC_TX_XOR_DROP_DST, tx, cb_fn,
334376 			  cb_param, scribble);
335
-	tx = async_xor(dp, srcs, 0, 2, bytes, submit);
377
+	tx = async_xor_offs(dp, dp_off, srcs, src_offs, 2, bytes, submit);
336378 
337379 	return tx;
338380 }
....@@ -344,11 +386,13 @@
344386  * @faila: first failed drive index
345387  * @failb: second failed drive index
346388  * @blocks: array of source pointers where the last two entries are p and q
389
+ * @offs: array of offset for pages in blocks
347390  * @submit: submission/completion modifiers
348391  */
349392 struct dma_async_tx_descriptor *
350393 async_raid6_2data_recov(int disks, size_t bytes, int faila, int failb,
351
-			struct page **blocks, struct async_submit_ctl *submit)
394
+			struct page **blocks, unsigned int *offs,
395
+			struct async_submit_ctl *submit)
352396 {
353397 	void *scribble = submit->scribble;
354398 	int non_zero_srcs, i;
....@@ -372,7 +416,7 @@
372416 			if (blocks[i] == NULL)
373417 				ptrs[i] = (void *) raid6_empty_zero_page;
374418 			else
375
-				ptrs[i] = page_address(blocks[i]);
419
+				ptrs[i] = page_address(blocks[i]) + offs[i];
376420 
377421 		raid6_2data_recov(disks, bytes, faila, failb, ptrs);
378422 
....@@ -397,16 +441,19 @@
397441 		 * explicitly handle the special case of a 4 disk array with
398442 		 * both data disks missing.
399443 		 */
400
-		return __2data_recov_4(disks, bytes, faila, failb, blocks, submit);
444
+		return __2data_recov_4(disks, bytes, faila, failb,
445
+				blocks, offs, submit);
401446 	case 3:
402447 		/* dma devices do not uniformly understand a single
403448 		 * source pq operation (in contrast to the synchronous
404449 		 * case), so explicitly handle the special case of a 5 disk
405450 		 * array with 2 of 3 data disks missing.
406451 		 */
407
-		return __2data_recov_5(disks, bytes, faila, failb, blocks, submit);
452
+		return __2data_recov_5(disks, bytes, faila, failb,
453
+				blocks, offs, submit);
408454 	default:
409
-		return __2data_recov_n(disks, bytes, faila, failb, blocks, submit);
455
+		return __2data_recov_n(disks, bytes, faila, failb,
456
+				blocks, offs, submit);
410457 	}
411458 }
412459 EXPORT_SYMBOL_GPL(async_raid6_2data_recov);
....@@ -417,14 +464,17 @@
417464  * @bytes: block size
418465  * @faila: failed drive index
419466  * @blocks: array of source pointers where the last two entries are p and q
467
+ * @offs: array of offset for pages in blocks
420468  * @submit: submission/completion modifiers
421469  */
422470 struct dma_async_tx_descriptor *
423471 async_raid6_datap_recov(int disks, size_t bytes, int faila,
424
-			struct page **blocks, struct async_submit_ctl *submit)
472
+			struct page **blocks, unsigned int *offs,
473
+			struct async_submit_ctl *submit)
425474 {
426475 	struct dma_async_tx_descriptor *tx = NULL;
427476 	struct page *p, *q, *dq;
477
+	unsigned int p_off, q_off, dq_off;
428478 	u8 coef;
429479 	enum async_tx_flags flags = submit->flags;
430480 	dma_async_tx_callback cb_fn = submit->cb_fn;
....@@ -432,6 +482,7 @@
432482 	void *scribble = submit->scribble;
433483 	int good_srcs, good, i;
434484 	struct page *srcs[2];
485
+	unsigned int src_offs[2];
435486 
436487 	pr_debug("%s: disks: %d len: %zu\n", __func__, disks, bytes);
437488 
....@@ -448,7 +499,7 @@
448499 			if (blocks[i] == NULL)
449500 				ptrs[i] = (void*)raid6_empty_zero_page;
450501 			else
451
-				ptrs[i] = page_address(blocks[i]);
502
+				ptrs[i] = page_address(blocks[i]) + offs[i];
452503 
453504 		raid6_datap_recov(disks, bytes, faila, ptrs);
454505 
....@@ -472,55 +523,67 @@
472523 	BUG_ON(good_srcs == 0);
473524 
474525 	p = blocks[disks-2];
526
+	p_off = offs[disks-2];
475527 	q = blocks[disks-1];
528
+	q_off = offs[disks-1];
476529 
477530 	/* Compute syndrome with zero for the missing data page
478531 	 * Use the dead data page as temporary storage for delta q
479532 	 */
480533 	dq = blocks[faila];
534
+	dq_off = offs[faila];
481535 	blocks[faila] = NULL;
482536 	blocks[disks-1] = dq;
537
+	offs[disks-1] = dq_off;
483538 
484539 	/* in the 4-disk case we only need to perform a single source
485540 	 * multiplication with the one good data block.
486541 	 */
487542 	if (good_srcs == 1) {
488543 		struct page *g = blocks[good];
544
+		unsigned int g_off = offs[good];
489545 
490546 		init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL,
491547 				  scribble);
492
-		tx = async_memcpy(p, g, 0, 0, bytes, submit);
548
+		tx = async_memcpy(p, g, p_off, g_off, bytes, submit);
493549 
494550 		init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL,
495551 				  scribble);
496
-		tx = async_mult(dq, g, raid6_gfexp[good], bytes, submit);
552
+		tx = async_mult(dq, dq_off, g, g_off,
553
+				raid6_gfexp[good], bytes, submit);
497554 	} else {
498555 		init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL,
499556 				  scribble);
500
-		tx = async_gen_syndrome(blocks, 0, disks, bytes, submit);
557
+		tx = async_gen_syndrome(blocks, offs, disks, bytes, submit);
501558 	}
502559 
503560 	/* Restore pointer table */
504561 	blocks[faila]   = dq;
562
+	offs[faila] = dq_off;
505563 	blocks[disks-1] = q;
564
+	offs[disks-1] = q_off;
506565 
507566 	/* calculate g^{-faila} */
508567 	coef = raid6_gfinv[raid6_gfexp[faila]];
509568 
510569 	srcs[0] = dq;
570
+	src_offs[0] = dq_off;
511571 	srcs[1] = q;
572
+	src_offs[1] = q_off;
512573 	init_async_submit(submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
513574 			  NULL, NULL, scribble);
514
-	tx = async_xor(dq, srcs, 0, 2, bytes, submit);
575
+	tx = async_xor_offs(dq, dq_off, srcs, src_offs, 2, bytes, submit);
515576 
516577 	init_async_submit(submit, ASYNC_TX_FENCE, tx, NULL, NULL, scribble);
517
-	tx = async_mult(dq, dq, coef, bytes, submit);
578
+	tx = async_mult(dq, dq_off, dq, dq_off, coef, bytes, submit);
518579 
519580 	srcs[0] = p;
581
+	src_offs[0] = p_off;
520582 	srcs[1] = dq;
583
+	src_offs[1] = dq_off;
521584 	init_async_submit(submit, flags | ASYNC_TX_XOR_DROP_DST, tx, cb_fn,
522585 			  cb_param, scribble);
523
-	tx = async_xor(p, srcs, 0, 2, bytes, submit);
586
+	tx = async_xor_offs(p, p_off, srcs, src_offs, 2, bytes, submit);
524587 
525588 	return tx;
526589 }