xref: /openbmc/linux/drivers/crypto/atmel-sha.c (revision 7587eb18)
1 /*
2  * Cryptographic API.
3  *
4  * Support for ATMEL SHA1/SHA256 HW acceleration.
5  *
6  * Copyright (c) 2012 Eukréa Electromatique - ATMEL
7  * Author: Nicolas Royer <nicolas@eukrea.com>
8  *
9  * This program is free software; you can redistribute it and/or modify
10  * it under the terms of the GNU General Public License version 2 as published
11  * by the Free Software Foundation.
12  *
13  * Some ideas are from omap-sham.c drivers.
14  */
15 
16 
17 #include <linux/kernel.h>
18 #include <linux/module.h>
19 #include <linux/slab.h>
20 #include <linux/err.h>
21 #include <linux/clk.h>
22 #include <linux/io.h>
23 #include <linux/hw_random.h>
24 #include <linux/platform_device.h>
25 
26 #include <linux/device.h>
27 #include <linux/init.h>
28 #include <linux/errno.h>
29 #include <linux/interrupt.h>
30 #include <linux/irq.h>
31 #include <linux/scatterlist.h>
32 #include <linux/dma-mapping.h>
33 #include <linux/of_device.h>
34 #include <linux/delay.h>
35 #include <linux/crypto.h>
36 #include <linux/cryptohash.h>
37 #include <crypto/scatterwalk.h>
38 #include <crypto/algapi.h>
39 #include <crypto/sha.h>
40 #include <crypto/hash.h>
41 #include <crypto/internal/hash.h>
42 #include <linux/platform_data/crypto-atmel.h>
43 #include "atmel-sha-regs.h"
44 
45 /* SHA flags */
46 #define SHA_FLAGS_BUSY			BIT(0)
47 #define	SHA_FLAGS_FINAL			BIT(1)
48 #define SHA_FLAGS_DMA_ACTIVE	BIT(2)
49 #define SHA_FLAGS_OUTPUT_READY	BIT(3)
50 #define SHA_FLAGS_INIT			BIT(4)
51 #define SHA_FLAGS_CPU			BIT(5)
52 #define SHA_FLAGS_DMA_READY		BIT(6)
53 
54 #define SHA_FLAGS_FINUP		BIT(16)
55 #define SHA_FLAGS_SG		BIT(17)
56 #define SHA_FLAGS_ALGO_MASK	GENMASK(22, 18)
57 #define SHA_FLAGS_SHA1		BIT(18)
58 #define SHA_FLAGS_SHA224	BIT(19)
59 #define SHA_FLAGS_SHA256	BIT(20)
60 #define SHA_FLAGS_SHA384	BIT(21)
61 #define SHA_FLAGS_SHA512	BIT(22)
62 #define SHA_FLAGS_ERROR		BIT(23)
63 #define SHA_FLAGS_PAD		BIT(24)
64 #define SHA_FLAGS_RESTORE	BIT(25)
65 
66 #define SHA_OP_UPDATE	1
67 #define SHA_OP_FINAL	2
68 
69 #define SHA_BUFFER_LEN		(PAGE_SIZE / 16)
70 
71 #define ATMEL_SHA_DMA_THRESHOLD		56
72 
73 struct atmel_sha_caps {
74 	bool	has_dma;
75 	bool	has_dualbuff;
76 	bool	has_sha224;
77 	bool	has_sha_384_512;
78 	bool	has_uihv;
79 };
80 
81 struct atmel_sha_dev;
82 
83 /*
84  * .statesize = sizeof(struct atmel_sha_reqctx) must be <= PAGE_SIZE / 8 as
85  * tested by the ahash_prepare_alg() function.
86  */
87 struct atmel_sha_reqctx {
88 	struct atmel_sha_dev	*dd;
89 	unsigned long	flags;
90 	unsigned long	op;
91 
92 	u8	digest[SHA512_DIGEST_SIZE] __aligned(sizeof(u32));
93 	u64	digcnt[2];
94 	size_t	bufcnt;
95 	size_t	buflen;
96 	dma_addr_t	dma_addr;
97 
98 	/* walk state */
99 	struct scatterlist	*sg;
100 	unsigned int	offset;	/* offset in current sg */
101 	unsigned int	total;	/* total request */
102 
103 	size_t block_size;
104 
105 	u8 buffer[SHA_BUFFER_LEN + SHA512_BLOCK_SIZE] __aligned(sizeof(u32));
106 };
107 
108 struct atmel_sha_ctx {
109 	struct atmel_sha_dev	*dd;
110 
111 	unsigned long		flags;
112 };
113 
114 #define ATMEL_SHA_QUEUE_LENGTH	50
115 
116 struct atmel_sha_dma {
117 	struct dma_chan			*chan;
118 	struct dma_slave_config dma_conf;
119 };
120 
121 struct atmel_sha_dev {
122 	struct list_head	list;
123 	unsigned long		phys_base;
124 	struct device		*dev;
125 	struct clk			*iclk;
126 	int					irq;
127 	void __iomem		*io_base;
128 
129 	spinlock_t		lock;
130 	int			err;
131 	struct tasklet_struct	done_task;
132 	struct tasklet_struct	queue_task;
133 
134 	unsigned long		flags;
135 	struct crypto_queue	queue;
136 	struct ahash_request	*req;
137 
138 	struct atmel_sha_dma	dma_lch_in;
139 
140 	struct atmel_sha_caps	caps;
141 
142 	u32	hw_version;
143 };
144 
145 struct atmel_sha_drv {
146 	struct list_head	dev_list;
147 	spinlock_t		lock;
148 };
149 
150 static struct atmel_sha_drv atmel_sha = {
151 	.dev_list = LIST_HEAD_INIT(atmel_sha.dev_list),
152 	.lock = __SPIN_LOCK_UNLOCKED(atmel_sha.lock),
153 };
154 
155 static inline u32 atmel_sha_read(struct atmel_sha_dev *dd, u32 offset)
156 {
157 	return readl_relaxed(dd->io_base + offset);
158 }
159 
160 static inline void atmel_sha_write(struct atmel_sha_dev *dd,
161 					u32 offset, u32 value)
162 {
163 	writel_relaxed(value, dd->io_base + offset);
164 }
165 
166 static size_t atmel_sha_append_sg(struct atmel_sha_reqctx *ctx)
167 {
168 	size_t count;
169 
170 	while ((ctx->bufcnt < ctx->buflen) && ctx->total) {
171 		count = min(ctx->sg->length - ctx->offset, ctx->total);
172 		count = min(count, ctx->buflen - ctx->bufcnt);
173 
174 		if (count <= 0) {
175 			/*
176 			* Check if count <= 0 because the buffer is full or
177 			* because the sg length is 0. In the latest case,
178 			* check if there is another sg in the list, a 0 length
179 			* sg doesn't necessarily mean the end of the sg list.
180 			*/
181 			if ((ctx->sg->length == 0) && !sg_is_last(ctx->sg)) {
182 				ctx->sg = sg_next(ctx->sg);
183 				continue;
184 			} else {
185 				break;
186 			}
187 		}
188 
189 		scatterwalk_map_and_copy(ctx->buffer + ctx->bufcnt, ctx->sg,
190 			ctx->offset, count, 0);
191 
192 		ctx->bufcnt += count;
193 		ctx->offset += count;
194 		ctx->total -= count;
195 
196 		if (ctx->offset == ctx->sg->length) {
197 			ctx->sg = sg_next(ctx->sg);
198 			if (ctx->sg)
199 				ctx->offset = 0;
200 			else
201 				ctx->total = 0;
202 		}
203 	}
204 
205 	return 0;
206 }
207 
208 /*
209  * The purpose of this padding is to ensure that the padded message is a
210  * multiple of 512 bits (SHA1/SHA224/SHA256) or 1024 bits (SHA384/SHA512).
211  * The bit "1" is appended at the end of the message followed by
212  * "padlen-1" zero bits. Then a 64 bits block (SHA1/SHA224/SHA256) or
213  * 128 bits block (SHA384/SHA512) equals to the message length in bits
214  * is appended.
215  *
216  * For SHA1/SHA224/SHA256, padlen is calculated as followed:
217  *  - if message length < 56 bytes then padlen = 56 - message length
218  *  - else padlen = 64 + 56 - message length
219  *
220  * For SHA384/SHA512, padlen is calculated as followed:
221  *  - if message length < 112 bytes then padlen = 112 - message length
222  *  - else padlen = 128 + 112 - message length
223  */
224 static void atmel_sha_fill_padding(struct atmel_sha_reqctx *ctx, int length)
225 {
226 	unsigned int index, padlen;
227 	u64 bits[2];
228 	u64 size[2];
229 
230 	size[0] = ctx->digcnt[0];
231 	size[1] = ctx->digcnt[1];
232 
233 	size[0] += ctx->bufcnt;
234 	if (size[0] < ctx->bufcnt)
235 		size[1]++;
236 
237 	size[0] += length;
238 	if (size[0]  < length)
239 		size[1]++;
240 
241 	bits[1] = cpu_to_be64(size[0] << 3);
242 	bits[0] = cpu_to_be64(size[1] << 3 | size[0] >> 61);
243 
244 	if (ctx->flags & (SHA_FLAGS_SHA384 | SHA_FLAGS_SHA512)) {
245 		index = ctx->bufcnt & 0x7f;
246 		padlen = (index < 112) ? (112 - index) : ((128+112) - index);
247 		*(ctx->buffer + ctx->bufcnt) = 0x80;
248 		memset(ctx->buffer + ctx->bufcnt + 1, 0, padlen-1);
249 		memcpy(ctx->buffer + ctx->bufcnt + padlen, bits, 16);
250 		ctx->bufcnt += padlen + 16;
251 		ctx->flags |= SHA_FLAGS_PAD;
252 	} else {
253 		index = ctx->bufcnt & 0x3f;
254 		padlen = (index < 56) ? (56 - index) : ((64+56) - index);
255 		*(ctx->buffer + ctx->bufcnt) = 0x80;
256 		memset(ctx->buffer + ctx->bufcnt + 1, 0, padlen-1);
257 		memcpy(ctx->buffer + ctx->bufcnt + padlen, &bits[1], 8);
258 		ctx->bufcnt += padlen + 8;
259 		ctx->flags |= SHA_FLAGS_PAD;
260 	}
261 }
262 
263 static int atmel_sha_init(struct ahash_request *req)
264 {
265 	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
266 	struct atmel_sha_ctx *tctx = crypto_ahash_ctx(tfm);
267 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
268 	struct atmel_sha_dev *dd = NULL;
269 	struct atmel_sha_dev *tmp;
270 
271 	spin_lock_bh(&atmel_sha.lock);
272 	if (!tctx->dd) {
273 		list_for_each_entry(tmp, &atmel_sha.dev_list, list) {
274 			dd = tmp;
275 			break;
276 		}
277 		tctx->dd = dd;
278 	} else {
279 		dd = tctx->dd;
280 	}
281 
282 	spin_unlock_bh(&atmel_sha.lock);
283 
284 	ctx->dd = dd;
285 
286 	ctx->flags = 0;
287 
288 	dev_dbg(dd->dev, "init: digest size: %d\n",
289 		crypto_ahash_digestsize(tfm));
290 
291 	switch (crypto_ahash_digestsize(tfm)) {
292 	case SHA1_DIGEST_SIZE:
293 		ctx->flags |= SHA_FLAGS_SHA1;
294 		ctx->block_size = SHA1_BLOCK_SIZE;
295 		break;
296 	case SHA224_DIGEST_SIZE:
297 		ctx->flags |= SHA_FLAGS_SHA224;
298 		ctx->block_size = SHA224_BLOCK_SIZE;
299 		break;
300 	case SHA256_DIGEST_SIZE:
301 		ctx->flags |= SHA_FLAGS_SHA256;
302 		ctx->block_size = SHA256_BLOCK_SIZE;
303 		break;
304 	case SHA384_DIGEST_SIZE:
305 		ctx->flags |= SHA_FLAGS_SHA384;
306 		ctx->block_size = SHA384_BLOCK_SIZE;
307 		break;
308 	case SHA512_DIGEST_SIZE:
309 		ctx->flags |= SHA_FLAGS_SHA512;
310 		ctx->block_size = SHA512_BLOCK_SIZE;
311 		break;
312 	default:
313 		return -EINVAL;
314 		break;
315 	}
316 
317 	ctx->bufcnt = 0;
318 	ctx->digcnt[0] = 0;
319 	ctx->digcnt[1] = 0;
320 	ctx->buflen = SHA_BUFFER_LEN;
321 
322 	return 0;
323 }
324 
325 static void atmel_sha_write_ctrl(struct atmel_sha_dev *dd, int dma)
326 {
327 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
328 	u32 valmr = SHA_MR_MODE_AUTO;
329 	unsigned int i, hashsize = 0;
330 
331 	if (likely(dma)) {
332 		if (!dd->caps.has_dma)
333 			atmel_sha_write(dd, SHA_IER, SHA_INT_TXBUFE);
334 		valmr = SHA_MR_MODE_PDC;
335 		if (dd->caps.has_dualbuff)
336 			valmr |= SHA_MR_DUALBUFF;
337 	} else {
338 		atmel_sha_write(dd, SHA_IER, SHA_INT_DATARDY);
339 	}
340 
341 	switch (ctx->flags & SHA_FLAGS_ALGO_MASK) {
342 	case SHA_FLAGS_SHA1:
343 		valmr |= SHA_MR_ALGO_SHA1;
344 		hashsize = SHA1_DIGEST_SIZE;
345 		break;
346 
347 	case SHA_FLAGS_SHA224:
348 		valmr |= SHA_MR_ALGO_SHA224;
349 		hashsize = SHA256_DIGEST_SIZE;
350 		break;
351 
352 	case SHA_FLAGS_SHA256:
353 		valmr |= SHA_MR_ALGO_SHA256;
354 		hashsize = SHA256_DIGEST_SIZE;
355 		break;
356 
357 	case SHA_FLAGS_SHA384:
358 		valmr |= SHA_MR_ALGO_SHA384;
359 		hashsize = SHA512_DIGEST_SIZE;
360 		break;
361 
362 	case SHA_FLAGS_SHA512:
363 		valmr |= SHA_MR_ALGO_SHA512;
364 		hashsize = SHA512_DIGEST_SIZE;
365 		break;
366 
367 	default:
368 		break;
369 	}
370 
371 	/* Setting CR_FIRST only for the first iteration */
372 	if (!(ctx->digcnt[0] || ctx->digcnt[1])) {
373 		atmel_sha_write(dd, SHA_CR, SHA_CR_FIRST);
374 	} else if (dd->caps.has_uihv && (ctx->flags & SHA_FLAGS_RESTORE)) {
375 		const u32 *hash = (const u32 *)ctx->digest;
376 
377 		/*
378 		 * Restore the hardware context: update the User Initialize
379 		 * Hash Value (UIHV) with the value saved when the latest
380 		 * 'update' operation completed on this very same crypto
381 		 * request.
382 		 */
383 		ctx->flags &= ~SHA_FLAGS_RESTORE;
384 		atmel_sha_write(dd, SHA_CR, SHA_CR_WUIHV);
385 		for (i = 0; i < hashsize / sizeof(u32); ++i)
386 			atmel_sha_write(dd, SHA_REG_DIN(i), hash[i]);
387 		atmel_sha_write(dd, SHA_CR, SHA_CR_FIRST);
388 		valmr |= SHA_MR_UIHV;
389 	}
390 	/*
391 	 * WARNING: If the UIHV feature is not available, the hardware CANNOT
392 	 * process concurrent requests: the internal registers used to store
393 	 * the hash/digest are still set to the partial digest output values
394 	 * computed during the latest round.
395 	 */
396 
397 	atmel_sha_write(dd, SHA_MR, valmr);
398 }
399 
400 static int atmel_sha_xmit_cpu(struct atmel_sha_dev *dd, const u8 *buf,
401 			      size_t length, int final)
402 {
403 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
404 	int count, len32;
405 	const u32 *buffer = (const u32 *)buf;
406 
407 	dev_dbg(dd->dev, "xmit_cpu: digcnt: 0x%llx 0x%llx, length: %d, final: %d\n",
408 		ctx->digcnt[1], ctx->digcnt[0], length, final);
409 
410 	atmel_sha_write_ctrl(dd, 0);
411 
412 	/* should be non-zero before next lines to disable clocks later */
413 	ctx->digcnt[0] += length;
414 	if (ctx->digcnt[0] < length)
415 		ctx->digcnt[1]++;
416 
417 	if (final)
418 		dd->flags |= SHA_FLAGS_FINAL; /* catch last interrupt */
419 
420 	len32 = DIV_ROUND_UP(length, sizeof(u32));
421 
422 	dd->flags |= SHA_FLAGS_CPU;
423 
424 	for (count = 0; count < len32; count++)
425 		atmel_sha_write(dd, SHA_REG_DIN(count), buffer[count]);
426 
427 	return -EINPROGRESS;
428 }
429 
430 static int atmel_sha_xmit_pdc(struct atmel_sha_dev *dd, dma_addr_t dma_addr1,
431 		size_t length1, dma_addr_t dma_addr2, size_t length2, int final)
432 {
433 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
434 	int len32;
435 
436 	dev_dbg(dd->dev, "xmit_pdc: digcnt: 0x%llx 0x%llx, length: %d, final: %d\n",
437 		ctx->digcnt[1], ctx->digcnt[0], length1, final);
438 
439 	len32 = DIV_ROUND_UP(length1, sizeof(u32));
440 	atmel_sha_write(dd, SHA_PTCR, SHA_PTCR_TXTDIS);
441 	atmel_sha_write(dd, SHA_TPR, dma_addr1);
442 	atmel_sha_write(dd, SHA_TCR, len32);
443 
444 	len32 = DIV_ROUND_UP(length2, sizeof(u32));
445 	atmel_sha_write(dd, SHA_TNPR, dma_addr2);
446 	atmel_sha_write(dd, SHA_TNCR, len32);
447 
448 	atmel_sha_write_ctrl(dd, 1);
449 
450 	/* should be non-zero before next lines to disable clocks later */
451 	ctx->digcnt[0] += length1;
452 	if (ctx->digcnt[0] < length1)
453 		ctx->digcnt[1]++;
454 
455 	if (final)
456 		dd->flags |= SHA_FLAGS_FINAL; /* catch last interrupt */
457 
458 	dd->flags |=  SHA_FLAGS_DMA_ACTIVE;
459 
460 	/* Start DMA transfer */
461 	atmel_sha_write(dd, SHA_PTCR, SHA_PTCR_TXTEN);
462 
463 	return -EINPROGRESS;
464 }
465 
466 static void atmel_sha_dma_callback(void *data)
467 {
468 	struct atmel_sha_dev *dd = data;
469 
470 	/* dma_lch_in - completed - wait DATRDY */
471 	atmel_sha_write(dd, SHA_IER, SHA_INT_DATARDY);
472 }
473 
474 static int atmel_sha_xmit_dma(struct atmel_sha_dev *dd, dma_addr_t dma_addr1,
475 		size_t length1, dma_addr_t dma_addr2, size_t length2, int final)
476 {
477 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
478 	struct dma_async_tx_descriptor	*in_desc;
479 	struct scatterlist sg[2];
480 
481 	dev_dbg(dd->dev, "xmit_dma: digcnt: 0x%llx 0x%llx, length: %d, final: %d\n",
482 		ctx->digcnt[1], ctx->digcnt[0], length1, final);
483 
484 	dd->dma_lch_in.dma_conf.src_maxburst = 16;
485 	dd->dma_lch_in.dma_conf.dst_maxburst = 16;
486 
487 	dmaengine_slave_config(dd->dma_lch_in.chan, &dd->dma_lch_in.dma_conf);
488 
489 	if (length2) {
490 		sg_init_table(sg, 2);
491 		sg_dma_address(&sg[0]) = dma_addr1;
492 		sg_dma_len(&sg[0]) = length1;
493 		sg_dma_address(&sg[1]) = dma_addr2;
494 		sg_dma_len(&sg[1]) = length2;
495 		in_desc = dmaengine_prep_slave_sg(dd->dma_lch_in.chan, sg, 2,
496 			DMA_MEM_TO_DEV, DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
497 	} else {
498 		sg_init_table(sg, 1);
499 		sg_dma_address(&sg[0]) = dma_addr1;
500 		sg_dma_len(&sg[0]) = length1;
501 		in_desc = dmaengine_prep_slave_sg(dd->dma_lch_in.chan, sg, 1,
502 			DMA_MEM_TO_DEV, DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
503 	}
504 	if (!in_desc)
505 		return -EINVAL;
506 
507 	in_desc->callback = atmel_sha_dma_callback;
508 	in_desc->callback_param = dd;
509 
510 	atmel_sha_write_ctrl(dd, 1);
511 
512 	/* should be non-zero before next lines to disable clocks later */
513 	ctx->digcnt[0] += length1;
514 	if (ctx->digcnt[0] < length1)
515 		ctx->digcnt[1]++;
516 
517 	if (final)
518 		dd->flags |= SHA_FLAGS_FINAL; /* catch last interrupt */
519 
520 	dd->flags |=  SHA_FLAGS_DMA_ACTIVE;
521 
522 	/* Start DMA transfer */
523 	dmaengine_submit(in_desc);
524 	dma_async_issue_pending(dd->dma_lch_in.chan);
525 
526 	return -EINPROGRESS;
527 }
528 
529 static int atmel_sha_xmit_start(struct atmel_sha_dev *dd, dma_addr_t dma_addr1,
530 		size_t length1, dma_addr_t dma_addr2, size_t length2, int final)
531 {
532 	if (dd->caps.has_dma)
533 		return atmel_sha_xmit_dma(dd, dma_addr1, length1,
534 				dma_addr2, length2, final);
535 	else
536 		return atmel_sha_xmit_pdc(dd, dma_addr1, length1,
537 				dma_addr2, length2, final);
538 }
539 
540 static int atmel_sha_update_cpu(struct atmel_sha_dev *dd)
541 {
542 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
543 	int bufcnt;
544 
545 	atmel_sha_append_sg(ctx);
546 	atmel_sha_fill_padding(ctx, 0);
547 	bufcnt = ctx->bufcnt;
548 	ctx->bufcnt = 0;
549 
550 	return atmel_sha_xmit_cpu(dd, ctx->buffer, bufcnt, 1);
551 }
552 
553 static int atmel_sha_xmit_dma_map(struct atmel_sha_dev *dd,
554 					struct atmel_sha_reqctx *ctx,
555 					size_t length, int final)
556 {
557 	ctx->dma_addr = dma_map_single(dd->dev, ctx->buffer,
558 				ctx->buflen + ctx->block_size, DMA_TO_DEVICE);
559 	if (dma_mapping_error(dd->dev, ctx->dma_addr)) {
560 		dev_err(dd->dev, "dma %u bytes error\n", ctx->buflen +
561 				ctx->block_size);
562 		return -EINVAL;
563 	}
564 
565 	ctx->flags &= ~SHA_FLAGS_SG;
566 
567 	/* next call does not fail... so no unmap in the case of error */
568 	return atmel_sha_xmit_start(dd, ctx->dma_addr, length, 0, 0, final);
569 }
570 
571 static int atmel_sha_update_dma_slow(struct atmel_sha_dev *dd)
572 {
573 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
574 	unsigned int final;
575 	size_t count;
576 
577 	atmel_sha_append_sg(ctx);
578 
579 	final = (ctx->flags & SHA_FLAGS_FINUP) && !ctx->total;
580 
581 	dev_dbg(dd->dev, "slow: bufcnt: %u, digcnt: 0x%llx 0x%llx, final: %d\n",
582 		 ctx->bufcnt, ctx->digcnt[1], ctx->digcnt[0], final);
583 
584 	if (final)
585 		atmel_sha_fill_padding(ctx, 0);
586 
587 	if (final || (ctx->bufcnt == ctx->buflen)) {
588 		count = ctx->bufcnt;
589 		ctx->bufcnt = 0;
590 		return atmel_sha_xmit_dma_map(dd, ctx, count, final);
591 	}
592 
593 	return 0;
594 }
595 
596 static int atmel_sha_update_dma_start(struct atmel_sha_dev *dd)
597 {
598 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
599 	unsigned int length, final, tail;
600 	struct scatterlist *sg;
601 	unsigned int count;
602 
603 	if (!ctx->total)
604 		return 0;
605 
606 	if (ctx->bufcnt || ctx->offset)
607 		return atmel_sha_update_dma_slow(dd);
608 
609 	dev_dbg(dd->dev, "fast: digcnt: 0x%llx 0x%llx, bufcnt: %u, total: %u\n",
610 		ctx->digcnt[1], ctx->digcnt[0], ctx->bufcnt, ctx->total);
611 
612 	sg = ctx->sg;
613 
614 	if (!IS_ALIGNED(sg->offset, sizeof(u32)))
615 		return atmel_sha_update_dma_slow(dd);
616 
617 	if (!sg_is_last(sg) && !IS_ALIGNED(sg->length, ctx->block_size))
618 		/* size is not ctx->block_size aligned */
619 		return atmel_sha_update_dma_slow(dd);
620 
621 	length = min(ctx->total, sg->length);
622 
623 	if (sg_is_last(sg)) {
624 		if (!(ctx->flags & SHA_FLAGS_FINUP)) {
625 			/* not last sg must be ctx->block_size aligned */
626 			tail = length & (ctx->block_size - 1);
627 			length -= tail;
628 		}
629 	}
630 
631 	ctx->total -= length;
632 	ctx->offset = length; /* offset where to start slow */
633 
634 	final = (ctx->flags & SHA_FLAGS_FINUP) && !ctx->total;
635 
636 	/* Add padding */
637 	if (final) {
638 		tail = length & (ctx->block_size - 1);
639 		length -= tail;
640 		ctx->total += tail;
641 		ctx->offset = length; /* offset where to start slow */
642 
643 		sg = ctx->sg;
644 		atmel_sha_append_sg(ctx);
645 
646 		atmel_sha_fill_padding(ctx, length);
647 
648 		ctx->dma_addr = dma_map_single(dd->dev, ctx->buffer,
649 			ctx->buflen + ctx->block_size, DMA_TO_DEVICE);
650 		if (dma_mapping_error(dd->dev, ctx->dma_addr)) {
651 			dev_err(dd->dev, "dma %u bytes error\n",
652 				ctx->buflen + ctx->block_size);
653 			return -EINVAL;
654 		}
655 
656 		if (length == 0) {
657 			ctx->flags &= ~SHA_FLAGS_SG;
658 			count = ctx->bufcnt;
659 			ctx->bufcnt = 0;
660 			return atmel_sha_xmit_start(dd, ctx->dma_addr, count, 0,
661 					0, final);
662 		} else {
663 			ctx->sg = sg;
664 			if (!dma_map_sg(dd->dev, ctx->sg, 1,
665 				DMA_TO_DEVICE)) {
666 					dev_err(dd->dev, "dma_map_sg  error\n");
667 					return -EINVAL;
668 			}
669 
670 			ctx->flags |= SHA_FLAGS_SG;
671 
672 			count = ctx->bufcnt;
673 			ctx->bufcnt = 0;
674 			return atmel_sha_xmit_start(dd, sg_dma_address(ctx->sg),
675 					length, ctx->dma_addr, count, final);
676 		}
677 	}
678 
679 	if (!dma_map_sg(dd->dev, ctx->sg, 1, DMA_TO_DEVICE)) {
680 		dev_err(dd->dev, "dma_map_sg  error\n");
681 		return -EINVAL;
682 	}
683 
684 	ctx->flags |= SHA_FLAGS_SG;
685 
686 	/* next call does not fail... so no unmap in the case of error */
687 	return atmel_sha_xmit_start(dd, sg_dma_address(ctx->sg), length, 0,
688 								0, final);
689 }
690 
691 static int atmel_sha_update_dma_stop(struct atmel_sha_dev *dd)
692 {
693 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(dd->req);
694 
695 	if (ctx->flags & SHA_FLAGS_SG) {
696 		dma_unmap_sg(dd->dev, ctx->sg, 1, DMA_TO_DEVICE);
697 		if (ctx->sg->length == ctx->offset) {
698 			ctx->sg = sg_next(ctx->sg);
699 			if (ctx->sg)
700 				ctx->offset = 0;
701 		}
702 		if (ctx->flags & SHA_FLAGS_PAD) {
703 			dma_unmap_single(dd->dev, ctx->dma_addr,
704 				ctx->buflen + ctx->block_size, DMA_TO_DEVICE);
705 		}
706 	} else {
707 		dma_unmap_single(dd->dev, ctx->dma_addr, ctx->buflen +
708 						ctx->block_size, DMA_TO_DEVICE);
709 	}
710 
711 	return 0;
712 }
713 
714 static int atmel_sha_update_req(struct atmel_sha_dev *dd)
715 {
716 	struct ahash_request *req = dd->req;
717 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
718 	int err;
719 
720 	dev_dbg(dd->dev, "update_req: total: %u, digcnt: 0x%llx 0x%llx\n",
721 		ctx->total, ctx->digcnt[1], ctx->digcnt[0]);
722 
723 	if (ctx->flags & SHA_FLAGS_CPU)
724 		err = atmel_sha_update_cpu(dd);
725 	else
726 		err = atmel_sha_update_dma_start(dd);
727 
728 	/* wait for dma completion before can take more data */
729 	dev_dbg(dd->dev, "update: err: %d, digcnt: 0x%llx 0%llx\n",
730 			err, ctx->digcnt[1], ctx->digcnt[0]);
731 
732 	return err;
733 }
734 
735 static int atmel_sha_final_req(struct atmel_sha_dev *dd)
736 {
737 	struct ahash_request *req = dd->req;
738 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
739 	int err = 0;
740 	int count;
741 
742 	if (ctx->bufcnt >= ATMEL_SHA_DMA_THRESHOLD) {
743 		atmel_sha_fill_padding(ctx, 0);
744 		count = ctx->bufcnt;
745 		ctx->bufcnt = 0;
746 		err = atmel_sha_xmit_dma_map(dd, ctx, count, 1);
747 	}
748 	/* faster to handle last block with cpu */
749 	else {
750 		atmel_sha_fill_padding(ctx, 0);
751 		count = ctx->bufcnt;
752 		ctx->bufcnt = 0;
753 		err = atmel_sha_xmit_cpu(dd, ctx->buffer, count, 1);
754 	}
755 
756 	dev_dbg(dd->dev, "final_req: err: %d\n", err);
757 
758 	return err;
759 }
760 
761 static void atmel_sha_copy_hash(struct ahash_request *req)
762 {
763 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
764 	u32 *hash = (u32 *)ctx->digest;
765 	unsigned int i, hashsize;
766 
767 	switch (ctx->flags & SHA_FLAGS_ALGO_MASK) {
768 	case SHA_FLAGS_SHA1:
769 		hashsize = SHA1_DIGEST_SIZE;
770 		break;
771 
772 	case SHA_FLAGS_SHA224:
773 	case SHA_FLAGS_SHA256:
774 		hashsize = SHA256_DIGEST_SIZE;
775 		break;
776 
777 	case SHA_FLAGS_SHA384:
778 	case SHA_FLAGS_SHA512:
779 		hashsize = SHA512_DIGEST_SIZE;
780 		break;
781 
782 	default:
783 		/* Should not happen... */
784 		return;
785 	}
786 
787 	for (i = 0; i < hashsize / sizeof(u32); ++i)
788 		hash[i] = atmel_sha_read(ctx->dd, SHA_REG_DIGEST(i));
789 	ctx->flags |= SHA_FLAGS_RESTORE;
790 }
791 
792 static void atmel_sha_copy_ready_hash(struct ahash_request *req)
793 {
794 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
795 
796 	if (!req->result)
797 		return;
798 
799 	if (ctx->flags & SHA_FLAGS_SHA1)
800 		memcpy(req->result, ctx->digest, SHA1_DIGEST_SIZE);
801 	else if (ctx->flags & SHA_FLAGS_SHA224)
802 		memcpy(req->result, ctx->digest, SHA224_DIGEST_SIZE);
803 	else if (ctx->flags & SHA_FLAGS_SHA256)
804 		memcpy(req->result, ctx->digest, SHA256_DIGEST_SIZE);
805 	else if (ctx->flags & SHA_FLAGS_SHA384)
806 		memcpy(req->result, ctx->digest, SHA384_DIGEST_SIZE);
807 	else
808 		memcpy(req->result, ctx->digest, SHA512_DIGEST_SIZE);
809 }
810 
811 static int atmel_sha_finish(struct ahash_request *req)
812 {
813 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
814 	struct atmel_sha_dev *dd = ctx->dd;
815 
816 	if (ctx->digcnt[0] || ctx->digcnt[1])
817 		atmel_sha_copy_ready_hash(req);
818 
819 	dev_dbg(dd->dev, "digcnt: 0x%llx 0x%llx, bufcnt: %d\n", ctx->digcnt[1],
820 		ctx->digcnt[0], ctx->bufcnt);
821 
822 	return 0;
823 }
824 
825 static void atmel_sha_finish_req(struct ahash_request *req, int err)
826 {
827 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
828 	struct atmel_sha_dev *dd = ctx->dd;
829 
830 	if (!err) {
831 		atmel_sha_copy_hash(req);
832 		if (SHA_FLAGS_FINAL & dd->flags)
833 			err = atmel_sha_finish(req);
834 	} else {
835 		ctx->flags |= SHA_FLAGS_ERROR;
836 	}
837 
838 	/* atomic operation is not needed here */
839 	dd->flags &= ~(SHA_FLAGS_BUSY | SHA_FLAGS_FINAL | SHA_FLAGS_CPU |
840 			SHA_FLAGS_DMA_READY | SHA_FLAGS_OUTPUT_READY);
841 
842 	clk_disable(dd->iclk);
843 
844 	if (req->base.complete)
845 		req->base.complete(&req->base, err);
846 
847 	/* handle new request */
848 	tasklet_schedule(&dd->queue_task);
849 }
850 
851 static int atmel_sha_hw_init(struct atmel_sha_dev *dd)
852 {
853 	int err;
854 
855 	err = clk_enable(dd->iclk);
856 	if (err)
857 		return err;
858 
859 	if (!(SHA_FLAGS_INIT & dd->flags)) {
860 		atmel_sha_write(dd, SHA_CR, SHA_CR_SWRST);
861 		dd->flags |= SHA_FLAGS_INIT;
862 		dd->err = 0;
863 	}
864 
865 	return 0;
866 }
867 
868 static inline unsigned int atmel_sha_get_version(struct atmel_sha_dev *dd)
869 {
870 	return atmel_sha_read(dd, SHA_HW_VERSION) & 0x00000fff;
871 }
872 
873 static void atmel_sha_hw_version_init(struct atmel_sha_dev *dd)
874 {
875 	atmel_sha_hw_init(dd);
876 
877 	dd->hw_version = atmel_sha_get_version(dd);
878 
879 	dev_info(dd->dev,
880 			"version: 0x%x\n", dd->hw_version);
881 
882 	clk_disable(dd->iclk);
883 }
884 
885 static int atmel_sha_handle_queue(struct atmel_sha_dev *dd,
886 				  struct ahash_request *req)
887 {
888 	struct crypto_async_request *async_req, *backlog;
889 	struct atmel_sha_reqctx *ctx;
890 	unsigned long flags;
891 	int err = 0, ret = 0;
892 
893 	spin_lock_irqsave(&dd->lock, flags);
894 	if (req)
895 		ret = ahash_enqueue_request(&dd->queue, req);
896 
897 	if (SHA_FLAGS_BUSY & dd->flags) {
898 		spin_unlock_irqrestore(&dd->lock, flags);
899 		return ret;
900 	}
901 
902 	backlog = crypto_get_backlog(&dd->queue);
903 	async_req = crypto_dequeue_request(&dd->queue);
904 	if (async_req)
905 		dd->flags |= SHA_FLAGS_BUSY;
906 
907 	spin_unlock_irqrestore(&dd->lock, flags);
908 
909 	if (!async_req)
910 		return ret;
911 
912 	if (backlog)
913 		backlog->complete(backlog, -EINPROGRESS);
914 
915 	req = ahash_request_cast(async_req);
916 	dd->req = req;
917 	ctx = ahash_request_ctx(req);
918 
919 	dev_dbg(dd->dev, "handling new req, op: %lu, nbytes: %d\n",
920 						ctx->op, req->nbytes);
921 
922 	err = atmel_sha_hw_init(dd);
923 
924 	if (err)
925 		goto err1;
926 
927 	if (ctx->op == SHA_OP_UPDATE) {
928 		err = atmel_sha_update_req(dd);
929 		if (err != -EINPROGRESS && (ctx->flags & SHA_FLAGS_FINUP))
930 			/* no final() after finup() */
931 			err = atmel_sha_final_req(dd);
932 	} else if (ctx->op == SHA_OP_FINAL) {
933 		err = atmel_sha_final_req(dd);
934 	}
935 
936 err1:
937 	if (err != -EINPROGRESS)
938 		/* done_task will not finish it, so do it here */
939 		atmel_sha_finish_req(req, err);
940 
941 	dev_dbg(dd->dev, "exit, err: %d\n", err);
942 
943 	return ret;
944 }
945 
946 static int atmel_sha_enqueue(struct ahash_request *req, unsigned int op)
947 {
948 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
949 	struct atmel_sha_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
950 	struct atmel_sha_dev *dd = tctx->dd;
951 
952 	ctx->op = op;
953 
954 	return atmel_sha_handle_queue(dd, req);
955 }
956 
957 static int atmel_sha_update(struct ahash_request *req)
958 {
959 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
960 
961 	if (!req->nbytes)
962 		return 0;
963 
964 	ctx->total = req->nbytes;
965 	ctx->sg = req->src;
966 	ctx->offset = 0;
967 
968 	if (ctx->flags & SHA_FLAGS_FINUP) {
969 		if (ctx->bufcnt + ctx->total < ATMEL_SHA_DMA_THRESHOLD)
970 			/* faster to use CPU for short transfers */
971 			ctx->flags |= SHA_FLAGS_CPU;
972 	} else if (ctx->bufcnt + ctx->total < ctx->buflen) {
973 		atmel_sha_append_sg(ctx);
974 		return 0;
975 	}
976 	return atmel_sha_enqueue(req, SHA_OP_UPDATE);
977 }
978 
979 static int atmel_sha_final(struct ahash_request *req)
980 {
981 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
982 
983 	ctx->flags |= SHA_FLAGS_FINUP;
984 
985 	if (ctx->flags & SHA_FLAGS_ERROR)
986 		return 0; /* uncompleted hash is not needed */
987 
988 	if (ctx->flags & SHA_FLAGS_PAD)
989 		/* copy ready hash (+ finalize hmac) */
990 		return atmel_sha_finish(req);
991 
992 	return atmel_sha_enqueue(req, SHA_OP_FINAL);
993 }
994 
995 static int atmel_sha_finup(struct ahash_request *req)
996 {
997 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
998 	int err1, err2;
999 
1000 	ctx->flags |= SHA_FLAGS_FINUP;
1001 
1002 	err1 = atmel_sha_update(req);
1003 	if (err1 == -EINPROGRESS || err1 == -EBUSY)
1004 		return err1;
1005 
1006 	/*
1007 	 * final() has to be always called to cleanup resources
1008 	 * even if udpate() failed, except EINPROGRESS
1009 	 */
1010 	err2 = atmel_sha_final(req);
1011 
1012 	return err1 ?: err2;
1013 }
1014 
1015 static int atmel_sha_digest(struct ahash_request *req)
1016 {
1017 	return atmel_sha_init(req) ?: atmel_sha_finup(req);
1018 }
1019 
1020 
1021 static int atmel_sha_export(struct ahash_request *req, void *out)
1022 {
1023 	const struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
1024 
1025 	memcpy(out, ctx, sizeof(*ctx));
1026 	return 0;
1027 }
1028 
1029 static int atmel_sha_import(struct ahash_request *req, const void *in)
1030 {
1031 	struct atmel_sha_reqctx *ctx = ahash_request_ctx(req);
1032 
1033 	memcpy(ctx, in, sizeof(*ctx));
1034 	return 0;
1035 }
1036 
1037 static int atmel_sha_cra_init(struct crypto_tfm *tfm)
1038 {
1039 	crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm),
1040 				 sizeof(struct atmel_sha_reqctx));
1041 
1042 	return 0;
1043 }
1044 
1045 static struct ahash_alg sha_1_256_algs[] = {
1046 {
1047 	.init		= atmel_sha_init,
1048 	.update		= atmel_sha_update,
1049 	.final		= atmel_sha_final,
1050 	.finup		= atmel_sha_finup,
1051 	.digest		= atmel_sha_digest,
1052 	.export		= atmel_sha_export,
1053 	.import		= atmel_sha_import,
1054 	.halg = {
1055 		.digestsize	= SHA1_DIGEST_SIZE,
1056 		.statesize	= sizeof(struct atmel_sha_reqctx),
1057 		.base	= {
1058 			.cra_name		= "sha1",
1059 			.cra_driver_name	= "atmel-sha1",
1060 			.cra_priority		= 100,
1061 			.cra_flags		= CRYPTO_ALG_ASYNC,
1062 			.cra_blocksize		= SHA1_BLOCK_SIZE,
1063 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1064 			.cra_alignmask		= 0,
1065 			.cra_module		= THIS_MODULE,
1066 			.cra_init		= atmel_sha_cra_init,
1067 		}
1068 	}
1069 },
1070 {
1071 	.init		= atmel_sha_init,
1072 	.update		= atmel_sha_update,
1073 	.final		= atmel_sha_final,
1074 	.finup		= atmel_sha_finup,
1075 	.digest		= atmel_sha_digest,
1076 	.export		= atmel_sha_export,
1077 	.import		= atmel_sha_import,
1078 	.halg = {
1079 		.digestsize	= SHA256_DIGEST_SIZE,
1080 		.statesize	= sizeof(struct atmel_sha_reqctx),
1081 		.base	= {
1082 			.cra_name		= "sha256",
1083 			.cra_driver_name	= "atmel-sha256",
1084 			.cra_priority		= 100,
1085 			.cra_flags		= CRYPTO_ALG_ASYNC,
1086 			.cra_blocksize		= SHA256_BLOCK_SIZE,
1087 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1088 			.cra_alignmask		= 0,
1089 			.cra_module		= THIS_MODULE,
1090 			.cra_init		= atmel_sha_cra_init,
1091 		}
1092 	}
1093 },
1094 };
1095 
1096 static struct ahash_alg sha_224_alg = {
1097 	.init		= atmel_sha_init,
1098 	.update		= atmel_sha_update,
1099 	.final		= atmel_sha_final,
1100 	.finup		= atmel_sha_finup,
1101 	.digest		= atmel_sha_digest,
1102 	.export		= atmel_sha_export,
1103 	.import		= atmel_sha_import,
1104 	.halg = {
1105 		.digestsize	= SHA224_DIGEST_SIZE,
1106 		.statesize	= sizeof(struct atmel_sha_reqctx),
1107 		.base	= {
1108 			.cra_name		= "sha224",
1109 			.cra_driver_name	= "atmel-sha224",
1110 			.cra_priority		= 100,
1111 			.cra_flags		= CRYPTO_ALG_ASYNC,
1112 			.cra_blocksize		= SHA224_BLOCK_SIZE,
1113 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1114 			.cra_alignmask		= 0,
1115 			.cra_module		= THIS_MODULE,
1116 			.cra_init		= atmel_sha_cra_init,
1117 		}
1118 	}
1119 };
1120 
1121 static struct ahash_alg sha_384_512_algs[] = {
1122 {
1123 	.init		= atmel_sha_init,
1124 	.update		= atmel_sha_update,
1125 	.final		= atmel_sha_final,
1126 	.finup		= atmel_sha_finup,
1127 	.digest		= atmel_sha_digest,
1128 	.export		= atmel_sha_export,
1129 	.import		= atmel_sha_import,
1130 	.halg = {
1131 		.digestsize	= SHA384_DIGEST_SIZE,
1132 		.statesize	= sizeof(struct atmel_sha_reqctx),
1133 		.base	= {
1134 			.cra_name		= "sha384",
1135 			.cra_driver_name	= "atmel-sha384",
1136 			.cra_priority		= 100,
1137 			.cra_flags		= CRYPTO_ALG_ASYNC,
1138 			.cra_blocksize		= SHA384_BLOCK_SIZE,
1139 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1140 			.cra_alignmask		= 0x3,
1141 			.cra_module		= THIS_MODULE,
1142 			.cra_init		= atmel_sha_cra_init,
1143 		}
1144 	}
1145 },
1146 {
1147 	.init		= atmel_sha_init,
1148 	.update		= atmel_sha_update,
1149 	.final		= atmel_sha_final,
1150 	.finup		= atmel_sha_finup,
1151 	.digest		= atmel_sha_digest,
1152 	.export		= atmel_sha_export,
1153 	.import		= atmel_sha_import,
1154 	.halg = {
1155 		.digestsize	= SHA512_DIGEST_SIZE,
1156 		.statesize	= sizeof(struct atmel_sha_reqctx),
1157 		.base	= {
1158 			.cra_name		= "sha512",
1159 			.cra_driver_name	= "atmel-sha512",
1160 			.cra_priority		= 100,
1161 			.cra_flags		= CRYPTO_ALG_ASYNC,
1162 			.cra_blocksize		= SHA512_BLOCK_SIZE,
1163 			.cra_ctxsize		= sizeof(struct atmel_sha_ctx),
1164 			.cra_alignmask		= 0x3,
1165 			.cra_module		= THIS_MODULE,
1166 			.cra_init		= atmel_sha_cra_init,
1167 		}
1168 	}
1169 },
1170 };
1171 
1172 static void atmel_sha_queue_task(unsigned long data)
1173 {
1174 	struct atmel_sha_dev *dd = (struct atmel_sha_dev *)data;
1175 
1176 	atmel_sha_handle_queue(dd, NULL);
1177 }
1178 
1179 static void atmel_sha_done_task(unsigned long data)
1180 {
1181 	struct atmel_sha_dev *dd = (struct atmel_sha_dev *)data;
1182 	int err = 0;
1183 
1184 	if (SHA_FLAGS_CPU & dd->flags) {
1185 		if (SHA_FLAGS_OUTPUT_READY & dd->flags) {
1186 			dd->flags &= ~SHA_FLAGS_OUTPUT_READY;
1187 			goto finish;
1188 		}
1189 	} else if (SHA_FLAGS_DMA_READY & dd->flags) {
1190 		if (SHA_FLAGS_DMA_ACTIVE & dd->flags) {
1191 			dd->flags &= ~SHA_FLAGS_DMA_ACTIVE;
1192 			atmel_sha_update_dma_stop(dd);
1193 			if (dd->err) {
1194 				err = dd->err;
1195 				goto finish;
1196 			}
1197 		}
1198 		if (SHA_FLAGS_OUTPUT_READY & dd->flags) {
1199 			/* hash or semi-hash ready */
1200 			dd->flags &= ~(SHA_FLAGS_DMA_READY |
1201 						SHA_FLAGS_OUTPUT_READY);
1202 			err = atmel_sha_update_dma_start(dd);
1203 			if (err != -EINPROGRESS)
1204 				goto finish;
1205 		}
1206 	}
1207 	return;
1208 
1209 finish:
1210 	/* finish curent request */
1211 	atmel_sha_finish_req(dd->req, err);
1212 }
1213 
1214 static irqreturn_t atmel_sha_irq(int irq, void *dev_id)
1215 {
1216 	struct atmel_sha_dev *sha_dd = dev_id;
1217 	u32 reg;
1218 
1219 	reg = atmel_sha_read(sha_dd, SHA_ISR);
1220 	if (reg & atmel_sha_read(sha_dd, SHA_IMR)) {
1221 		atmel_sha_write(sha_dd, SHA_IDR, reg);
1222 		if (SHA_FLAGS_BUSY & sha_dd->flags) {
1223 			sha_dd->flags |= SHA_FLAGS_OUTPUT_READY;
1224 			if (!(SHA_FLAGS_CPU & sha_dd->flags))
1225 				sha_dd->flags |= SHA_FLAGS_DMA_READY;
1226 			tasklet_schedule(&sha_dd->done_task);
1227 		} else {
1228 			dev_warn(sha_dd->dev, "SHA interrupt when no active requests.\n");
1229 		}
1230 		return IRQ_HANDLED;
1231 	}
1232 
1233 	return IRQ_NONE;
1234 }
1235 
1236 static void atmel_sha_unregister_algs(struct atmel_sha_dev *dd)
1237 {
1238 	int i;
1239 
1240 	for (i = 0; i < ARRAY_SIZE(sha_1_256_algs); i++)
1241 		crypto_unregister_ahash(&sha_1_256_algs[i]);
1242 
1243 	if (dd->caps.has_sha224)
1244 		crypto_unregister_ahash(&sha_224_alg);
1245 
1246 	if (dd->caps.has_sha_384_512) {
1247 		for (i = 0; i < ARRAY_SIZE(sha_384_512_algs); i++)
1248 			crypto_unregister_ahash(&sha_384_512_algs[i]);
1249 	}
1250 }
1251 
1252 static int atmel_sha_register_algs(struct atmel_sha_dev *dd)
1253 {
1254 	int err, i, j;
1255 
1256 	for (i = 0; i < ARRAY_SIZE(sha_1_256_algs); i++) {
1257 		err = crypto_register_ahash(&sha_1_256_algs[i]);
1258 		if (err)
1259 			goto err_sha_1_256_algs;
1260 	}
1261 
1262 	if (dd->caps.has_sha224) {
1263 		err = crypto_register_ahash(&sha_224_alg);
1264 		if (err)
1265 			goto err_sha_224_algs;
1266 	}
1267 
1268 	if (dd->caps.has_sha_384_512) {
1269 		for (i = 0; i < ARRAY_SIZE(sha_384_512_algs); i++) {
1270 			err = crypto_register_ahash(&sha_384_512_algs[i]);
1271 			if (err)
1272 				goto err_sha_384_512_algs;
1273 		}
1274 	}
1275 
1276 	return 0;
1277 
1278 err_sha_384_512_algs:
1279 	for (j = 0; j < i; j++)
1280 		crypto_unregister_ahash(&sha_384_512_algs[j]);
1281 	crypto_unregister_ahash(&sha_224_alg);
1282 err_sha_224_algs:
1283 	i = ARRAY_SIZE(sha_1_256_algs);
1284 err_sha_1_256_algs:
1285 	for (j = 0; j < i; j++)
1286 		crypto_unregister_ahash(&sha_1_256_algs[j]);
1287 
1288 	return err;
1289 }
1290 
1291 static bool atmel_sha_filter(struct dma_chan *chan, void *slave)
1292 {
1293 	struct at_dma_slave	*sl = slave;
1294 
1295 	if (sl && sl->dma_dev == chan->device->dev) {
1296 		chan->private = sl;
1297 		return true;
1298 	} else {
1299 		return false;
1300 	}
1301 }
1302 
1303 static int atmel_sha_dma_init(struct atmel_sha_dev *dd,
1304 				struct crypto_platform_data *pdata)
1305 {
1306 	int err = -ENOMEM;
1307 	dma_cap_mask_t mask_in;
1308 
1309 	/* Try to grab DMA channel */
1310 	dma_cap_zero(mask_in);
1311 	dma_cap_set(DMA_SLAVE, mask_in);
1312 
1313 	dd->dma_lch_in.chan = dma_request_slave_channel_compat(mask_in,
1314 			atmel_sha_filter, &pdata->dma_slave->rxdata, dd->dev, "tx");
1315 	if (!dd->dma_lch_in.chan) {
1316 		dev_warn(dd->dev, "no DMA channel available\n");
1317 		return err;
1318 	}
1319 
1320 	dd->dma_lch_in.dma_conf.direction = DMA_MEM_TO_DEV;
1321 	dd->dma_lch_in.dma_conf.dst_addr = dd->phys_base +
1322 		SHA_REG_DIN(0);
1323 	dd->dma_lch_in.dma_conf.src_maxburst = 1;
1324 	dd->dma_lch_in.dma_conf.src_addr_width =
1325 		DMA_SLAVE_BUSWIDTH_4_BYTES;
1326 	dd->dma_lch_in.dma_conf.dst_maxburst = 1;
1327 	dd->dma_lch_in.dma_conf.dst_addr_width =
1328 		DMA_SLAVE_BUSWIDTH_4_BYTES;
1329 	dd->dma_lch_in.dma_conf.device_fc = false;
1330 
1331 	return 0;
1332 }
1333 
1334 static void atmel_sha_dma_cleanup(struct atmel_sha_dev *dd)
1335 {
1336 	dma_release_channel(dd->dma_lch_in.chan);
1337 }
1338 
1339 static void atmel_sha_get_cap(struct atmel_sha_dev *dd)
1340 {
1341 
1342 	dd->caps.has_dma = 0;
1343 	dd->caps.has_dualbuff = 0;
1344 	dd->caps.has_sha224 = 0;
1345 	dd->caps.has_sha_384_512 = 0;
1346 	dd->caps.has_uihv = 0;
1347 
1348 	/* keep only major version number */
1349 	switch (dd->hw_version & 0xff0) {
1350 	case 0x510:
1351 		dd->caps.has_dma = 1;
1352 		dd->caps.has_dualbuff = 1;
1353 		dd->caps.has_sha224 = 1;
1354 		dd->caps.has_sha_384_512 = 1;
1355 		dd->caps.has_uihv = 1;
1356 		break;
1357 	case 0x420:
1358 		dd->caps.has_dma = 1;
1359 		dd->caps.has_dualbuff = 1;
1360 		dd->caps.has_sha224 = 1;
1361 		dd->caps.has_sha_384_512 = 1;
1362 		dd->caps.has_uihv = 1;
1363 		break;
1364 	case 0x410:
1365 		dd->caps.has_dma = 1;
1366 		dd->caps.has_dualbuff = 1;
1367 		dd->caps.has_sha224 = 1;
1368 		dd->caps.has_sha_384_512 = 1;
1369 		break;
1370 	case 0x400:
1371 		dd->caps.has_dma = 1;
1372 		dd->caps.has_dualbuff = 1;
1373 		dd->caps.has_sha224 = 1;
1374 		break;
1375 	case 0x320:
1376 		break;
1377 	default:
1378 		dev_warn(dd->dev,
1379 				"Unmanaged sha version, set minimum capabilities\n");
1380 		break;
1381 	}
1382 }
1383 
1384 #if defined(CONFIG_OF)
1385 static const struct of_device_id atmel_sha_dt_ids[] = {
1386 	{ .compatible = "atmel,at91sam9g46-sha" },
1387 	{ /* sentinel */ }
1388 };
1389 
1390 MODULE_DEVICE_TABLE(of, atmel_sha_dt_ids);
1391 
1392 static struct crypto_platform_data *atmel_sha_of_init(struct platform_device *pdev)
1393 {
1394 	struct device_node *np = pdev->dev.of_node;
1395 	struct crypto_platform_data *pdata;
1396 
1397 	if (!np) {
1398 		dev_err(&pdev->dev, "device node not found\n");
1399 		return ERR_PTR(-EINVAL);
1400 	}
1401 
1402 	pdata = devm_kzalloc(&pdev->dev, sizeof(*pdata), GFP_KERNEL);
1403 	if (!pdata) {
1404 		dev_err(&pdev->dev, "could not allocate memory for pdata\n");
1405 		return ERR_PTR(-ENOMEM);
1406 	}
1407 
1408 	pdata->dma_slave = devm_kzalloc(&pdev->dev,
1409 					sizeof(*(pdata->dma_slave)),
1410 					GFP_KERNEL);
1411 	if (!pdata->dma_slave) {
1412 		dev_err(&pdev->dev, "could not allocate memory for dma_slave\n");
1413 		return ERR_PTR(-ENOMEM);
1414 	}
1415 
1416 	return pdata;
1417 }
1418 #else /* CONFIG_OF */
1419 static inline struct crypto_platform_data *atmel_sha_of_init(struct platform_device *dev)
1420 {
1421 	return ERR_PTR(-EINVAL);
1422 }
1423 #endif
1424 
1425 static int atmel_sha_probe(struct platform_device *pdev)
1426 {
1427 	struct atmel_sha_dev *sha_dd;
1428 	struct crypto_platform_data	*pdata;
1429 	struct device *dev = &pdev->dev;
1430 	struct resource *sha_res;
1431 	int err;
1432 
1433 	sha_dd = devm_kzalloc(&pdev->dev, sizeof(*sha_dd), GFP_KERNEL);
1434 	if (sha_dd == NULL) {
1435 		dev_err(dev, "unable to alloc data struct.\n");
1436 		err = -ENOMEM;
1437 		goto sha_dd_err;
1438 	}
1439 
1440 	sha_dd->dev = dev;
1441 
1442 	platform_set_drvdata(pdev, sha_dd);
1443 
1444 	INIT_LIST_HEAD(&sha_dd->list);
1445 	spin_lock_init(&sha_dd->lock);
1446 
1447 	tasklet_init(&sha_dd->done_task, atmel_sha_done_task,
1448 					(unsigned long)sha_dd);
1449 	tasklet_init(&sha_dd->queue_task, atmel_sha_queue_task,
1450 					(unsigned long)sha_dd);
1451 
1452 	crypto_init_queue(&sha_dd->queue, ATMEL_SHA_QUEUE_LENGTH);
1453 
1454 	sha_dd->irq = -1;
1455 
1456 	/* Get the base address */
1457 	sha_res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1458 	if (!sha_res) {
1459 		dev_err(dev, "no MEM resource info\n");
1460 		err = -ENODEV;
1461 		goto res_err;
1462 	}
1463 	sha_dd->phys_base = sha_res->start;
1464 
1465 	/* Get the IRQ */
1466 	sha_dd->irq = platform_get_irq(pdev,  0);
1467 	if (sha_dd->irq < 0) {
1468 		dev_err(dev, "no IRQ resource info\n");
1469 		err = sha_dd->irq;
1470 		goto res_err;
1471 	}
1472 
1473 	err = devm_request_irq(&pdev->dev, sha_dd->irq, atmel_sha_irq,
1474 			       IRQF_SHARED, "atmel-sha", sha_dd);
1475 	if (err) {
1476 		dev_err(dev, "unable to request sha irq.\n");
1477 		goto res_err;
1478 	}
1479 
1480 	/* Initializing the clock */
1481 	sha_dd->iclk = devm_clk_get(&pdev->dev, "sha_clk");
1482 	if (IS_ERR(sha_dd->iclk)) {
1483 		dev_err(dev, "clock initialization failed.\n");
1484 		err = PTR_ERR(sha_dd->iclk);
1485 		goto res_err;
1486 	}
1487 
1488 	sha_dd->io_base = devm_ioremap_resource(&pdev->dev, sha_res);
1489 	if (IS_ERR(sha_dd->io_base)) {
1490 		dev_err(dev, "can't ioremap\n");
1491 		err = PTR_ERR(sha_dd->io_base);
1492 		goto res_err;
1493 	}
1494 
1495 	err = clk_prepare(sha_dd->iclk);
1496 	if (err)
1497 		goto res_err;
1498 
1499 	atmel_sha_hw_version_init(sha_dd);
1500 
1501 	atmel_sha_get_cap(sha_dd);
1502 
1503 	if (sha_dd->caps.has_dma) {
1504 		pdata = pdev->dev.platform_data;
1505 		if (!pdata) {
1506 			pdata = atmel_sha_of_init(pdev);
1507 			if (IS_ERR(pdata)) {
1508 				dev_err(&pdev->dev, "platform data not available\n");
1509 				err = PTR_ERR(pdata);
1510 				goto iclk_unprepare;
1511 			}
1512 		}
1513 		if (!pdata->dma_slave) {
1514 			err = -ENXIO;
1515 			goto iclk_unprepare;
1516 		}
1517 		err = atmel_sha_dma_init(sha_dd, pdata);
1518 		if (err)
1519 			goto err_sha_dma;
1520 
1521 		dev_info(dev, "using %s for DMA transfers\n",
1522 				dma_chan_name(sha_dd->dma_lch_in.chan));
1523 	}
1524 
1525 	spin_lock(&atmel_sha.lock);
1526 	list_add_tail(&sha_dd->list, &atmel_sha.dev_list);
1527 	spin_unlock(&atmel_sha.lock);
1528 
1529 	err = atmel_sha_register_algs(sha_dd);
1530 	if (err)
1531 		goto err_algs;
1532 
1533 	dev_info(dev, "Atmel SHA1/SHA256%s%s\n",
1534 			sha_dd->caps.has_sha224 ? "/SHA224" : "",
1535 			sha_dd->caps.has_sha_384_512 ? "/SHA384/SHA512" : "");
1536 
1537 	return 0;
1538 
1539 err_algs:
1540 	spin_lock(&atmel_sha.lock);
1541 	list_del(&sha_dd->list);
1542 	spin_unlock(&atmel_sha.lock);
1543 	if (sha_dd->caps.has_dma)
1544 		atmel_sha_dma_cleanup(sha_dd);
1545 err_sha_dma:
1546 iclk_unprepare:
1547 	clk_unprepare(sha_dd->iclk);
1548 res_err:
1549 	tasklet_kill(&sha_dd->queue_task);
1550 	tasklet_kill(&sha_dd->done_task);
1551 sha_dd_err:
1552 	dev_err(dev, "initialization failed.\n");
1553 
1554 	return err;
1555 }
1556 
1557 static int atmel_sha_remove(struct platform_device *pdev)
1558 {
1559 	static struct atmel_sha_dev *sha_dd;
1560 
1561 	sha_dd = platform_get_drvdata(pdev);
1562 	if (!sha_dd)
1563 		return -ENODEV;
1564 	spin_lock(&atmel_sha.lock);
1565 	list_del(&sha_dd->list);
1566 	spin_unlock(&atmel_sha.lock);
1567 
1568 	atmel_sha_unregister_algs(sha_dd);
1569 
1570 	tasklet_kill(&sha_dd->queue_task);
1571 	tasklet_kill(&sha_dd->done_task);
1572 
1573 	if (sha_dd->caps.has_dma)
1574 		atmel_sha_dma_cleanup(sha_dd);
1575 
1576 	clk_unprepare(sha_dd->iclk);
1577 
1578 	return 0;
1579 }
1580 
1581 static struct platform_driver atmel_sha_driver = {
1582 	.probe		= atmel_sha_probe,
1583 	.remove		= atmel_sha_remove,
1584 	.driver		= {
1585 		.name	= "atmel_sha",
1586 		.of_match_table	= of_match_ptr(atmel_sha_dt_ids),
1587 	},
1588 };
1589 
1590 module_platform_driver(atmel_sha_driver);
1591 
1592 MODULE_DESCRIPTION("Atmel SHA (1/256/224/384/512) hw acceleration support.");
1593 MODULE_LICENSE("GPL v2");
1594 MODULE_AUTHOR("Nicolas Royer - Eukréa Electromatique");
1595