xref: /linux/drivers/crypto/omap-sham.c (revision 6fdcba32711044c35c0e1b094cbd8f3f0b4472c9)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Cryptographic API.
4  *
5  * Support for OMAP SHA1/MD5 HW acceleration.
6  *
7  * Copyright (c) 2010 Nokia Corporation
8  * Author: Dmitry Kasatkin <dmitry.kasatkin@nokia.com>
9  * Copyright (c) 2011 Texas Instruments Incorporated
10  *
11  * Some ideas are from old omap-sha1-md5.c driver.
12  */
13 
14 #define pr_fmt(fmt) "%s: " fmt, __func__
15 
16 #include <linux/err.h>
17 #include <linux/device.h>
18 #include <linux/module.h>
19 #include <linux/init.h>
20 #include <linux/errno.h>
21 #include <linux/interrupt.h>
22 #include <linux/kernel.h>
23 #include <linux/irq.h>
24 #include <linux/io.h>
25 #include <linux/platform_device.h>
26 #include <linux/scatterlist.h>
27 #include <linux/dma-mapping.h>
28 #include <linux/dmaengine.h>
29 #include <linux/pm_runtime.h>
30 #include <linux/of.h>
31 #include <linux/of_device.h>
32 #include <linux/of_address.h>
33 #include <linux/of_irq.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/hmac.h>
42 #include <crypto/internal/hash.h>
43 
44 #define MD5_DIGEST_SIZE			16
45 
46 #define SHA_REG_IDIGEST(dd, x)		((dd)->pdata->idigest_ofs + ((x)*0x04))
47 #define SHA_REG_DIN(dd, x)		((dd)->pdata->din_ofs + ((x) * 0x04))
48 #define SHA_REG_DIGCNT(dd)		((dd)->pdata->digcnt_ofs)
49 
50 #define SHA_REG_ODIGEST(dd, x)		((dd)->pdata->odigest_ofs + (x * 0x04))
51 
52 #define SHA_REG_CTRL			0x18
53 #define SHA_REG_CTRL_LENGTH		(0xFFFFFFFF << 5)
54 #define SHA_REG_CTRL_CLOSE_HASH		(1 << 4)
55 #define SHA_REG_CTRL_ALGO_CONST		(1 << 3)
56 #define SHA_REG_CTRL_ALGO		(1 << 2)
57 #define SHA_REG_CTRL_INPUT_READY	(1 << 1)
58 #define SHA_REG_CTRL_OUTPUT_READY	(1 << 0)
59 
60 #define SHA_REG_REV(dd)			((dd)->pdata->rev_ofs)
61 
62 #define SHA_REG_MASK(dd)		((dd)->pdata->mask_ofs)
63 #define SHA_REG_MASK_DMA_EN		(1 << 3)
64 #define SHA_REG_MASK_IT_EN		(1 << 2)
65 #define SHA_REG_MASK_SOFTRESET		(1 << 1)
66 #define SHA_REG_AUTOIDLE		(1 << 0)
67 
68 #define SHA_REG_SYSSTATUS(dd)		((dd)->pdata->sysstatus_ofs)
69 #define SHA_REG_SYSSTATUS_RESETDONE	(1 << 0)
70 
71 #define SHA_REG_MODE(dd)		((dd)->pdata->mode_ofs)
72 #define SHA_REG_MODE_HMAC_OUTER_HASH	(1 << 7)
73 #define SHA_REG_MODE_HMAC_KEY_PROC	(1 << 5)
74 #define SHA_REG_MODE_CLOSE_HASH		(1 << 4)
75 #define SHA_REG_MODE_ALGO_CONSTANT	(1 << 3)
76 
77 #define SHA_REG_MODE_ALGO_MASK		(7 << 0)
78 #define SHA_REG_MODE_ALGO_MD5_128	(0 << 1)
79 #define SHA_REG_MODE_ALGO_SHA1_160	(1 << 1)
80 #define SHA_REG_MODE_ALGO_SHA2_224	(2 << 1)
81 #define SHA_REG_MODE_ALGO_SHA2_256	(3 << 1)
82 #define SHA_REG_MODE_ALGO_SHA2_384	(1 << 0)
83 #define SHA_REG_MODE_ALGO_SHA2_512	(3 << 0)
84 
85 #define SHA_REG_LENGTH(dd)		((dd)->pdata->length_ofs)
86 
87 #define SHA_REG_IRQSTATUS		0x118
88 #define SHA_REG_IRQSTATUS_CTX_RDY	(1 << 3)
89 #define SHA_REG_IRQSTATUS_PARTHASH_RDY (1 << 2)
90 #define SHA_REG_IRQSTATUS_INPUT_RDY	(1 << 1)
91 #define SHA_REG_IRQSTATUS_OUTPUT_RDY	(1 << 0)
92 
93 #define SHA_REG_IRQENA			0x11C
94 #define SHA_REG_IRQENA_CTX_RDY		(1 << 3)
95 #define SHA_REG_IRQENA_PARTHASH_RDY	(1 << 2)
96 #define SHA_REG_IRQENA_INPUT_RDY	(1 << 1)
97 #define SHA_REG_IRQENA_OUTPUT_RDY	(1 << 0)
98 
99 #define DEFAULT_TIMEOUT_INTERVAL	HZ
100 
101 #define DEFAULT_AUTOSUSPEND_DELAY	1000
102 
103 /* mostly device flags */
104 #define FLAGS_BUSY		0
105 #define FLAGS_FINAL		1
106 #define FLAGS_DMA_ACTIVE	2
107 #define FLAGS_OUTPUT_READY	3
108 #define FLAGS_INIT		4
109 #define FLAGS_CPU		5
110 #define FLAGS_DMA_READY		6
111 #define FLAGS_AUTO_XOR		7
112 #define FLAGS_BE32_SHA1		8
113 #define FLAGS_SGS_COPIED	9
114 #define FLAGS_SGS_ALLOCED	10
115 /* context flags */
116 #define FLAGS_FINUP		16
117 
118 #define FLAGS_MODE_SHIFT	18
119 #define FLAGS_MODE_MASK		(SHA_REG_MODE_ALGO_MASK	<< FLAGS_MODE_SHIFT)
120 #define FLAGS_MODE_MD5		(SHA_REG_MODE_ALGO_MD5_128 << FLAGS_MODE_SHIFT)
121 #define FLAGS_MODE_SHA1		(SHA_REG_MODE_ALGO_SHA1_160 << FLAGS_MODE_SHIFT)
122 #define FLAGS_MODE_SHA224	(SHA_REG_MODE_ALGO_SHA2_224 << FLAGS_MODE_SHIFT)
123 #define FLAGS_MODE_SHA256	(SHA_REG_MODE_ALGO_SHA2_256 << FLAGS_MODE_SHIFT)
124 #define FLAGS_MODE_SHA384	(SHA_REG_MODE_ALGO_SHA2_384 << FLAGS_MODE_SHIFT)
125 #define FLAGS_MODE_SHA512	(SHA_REG_MODE_ALGO_SHA2_512 << FLAGS_MODE_SHIFT)
126 
127 #define FLAGS_HMAC		21
128 #define FLAGS_ERROR		22
129 
130 #define OP_UPDATE		1
131 #define OP_FINAL		2
132 
133 #define OMAP_ALIGN_MASK		(sizeof(u32)-1)
134 #define OMAP_ALIGNED		__attribute__((aligned(sizeof(u32))))
135 
136 #define BUFLEN			SHA512_BLOCK_SIZE
137 #define OMAP_SHA_DMA_THRESHOLD	256
138 
139 struct omap_sham_dev;
140 
141 struct omap_sham_reqctx {
142 	struct omap_sham_dev	*dd;
143 	unsigned long		flags;
144 	unsigned long		op;
145 
146 	u8			digest[SHA512_DIGEST_SIZE] OMAP_ALIGNED;
147 	size_t			digcnt;
148 	size_t			bufcnt;
149 	size_t			buflen;
150 
151 	/* walk state */
152 	struct scatterlist	*sg;
153 	struct scatterlist	sgl[2];
154 	int			offset;	/* offset in current sg */
155 	int			sg_len;
156 	unsigned int		total;	/* total request */
157 
158 	u8			buffer[0] OMAP_ALIGNED;
159 };
160 
161 struct omap_sham_hmac_ctx {
162 	struct crypto_shash	*shash;
163 	u8			ipad[SHA512_BLOCK_SIZE] OMAP_ALIGNED;
164 	u8			opad[SHA512_BLOCK_SIZE] OMAP_ALIGNED;
165 };
166 
167 struct omap_sham_ctx {
168 	struct omap_sham_dev	*dd;
169 
170 	unsigned long		flags;
171 
172 	/* fallback stuff */
173 	struct crypto_shash	*fallback;
174 
175 	struct omap_sham_hmac_ctx base[0];
176 };
177 
178 #define OMAP_SHAM_QUEUE_LENGTH	10
179 
180 struct omap_sham_algs_info {
181 	struct ahash_alg	*algs_list;
182 	unsigned int		size;
183 	unsigned int		registered;
184 };
185 
186 struct omap_sham_pdata {
187 	struct omap_sham_algs_info	*algs_info;
188 	unsigned int	algs_info_size;
189 	unsigned long	flags;
190 	int		digest_size;
191 
192 	void		(*copy_hash)(struct ahash_request *req, int out);
193 	void		(*write_ctrl)(struct omap_sham_dev *dd, size_t length,
194 				      int final, int dma);
195 	void		(*trigger)(struct omap_sham_dev *dd, size_t length);
196 	int		(*poll_irq)(struct omap_sham_dev *dd);
197 	irqreturn_t	(*intr_hdlr)(int irq, void *dev_id);
198 
199 	u32		odigest_ofs;
200 	u32		idigest_ofs;
201 	u32		din_ofs;
202 	u32		digcnt_ofs;
203 	u32		rev_ofs;
204 	u32		mask_ofs;
205 	u32		sysstatus_ofs;
206 	u32		mode_ofs;
207 	u32		length_ofs;
208 
209 	u32		major_mask;
210 	u32		major_shift;
211 	u32		minor_mask;
212 	u32		minor_shift;
213 };
214 
215 struct omap_sham_dev {
216 	struct list_head	list;
217 	unsigned long		phys_base;
218 	struct device		*dev;
219 	void __iomem		*io_base;
220 	int			irq;
221 	spinlock_t		lock;
222 	int			err;
223 	struct dma_chan		*dma_lch;
224 	struct tasklet_struct	done_task;
225 	u8			polling_mode;
226 	u8			xmit_buf[BUFLEN] OMAP_ALIGNED;
227 
228 	unsigned long		flags;
229 	int			fallback_sz;
230 	struct crypto_queue	queue;
231 	struct ahash_request	*req;
232 
233 	const struct omap_sham_pdata	*pdata;
234 };
235 
236 struct omap_sham_drv {
237 	struct list_head	dev_list;
238 	spinlock_t		lock;
239 	unsigned long		flags;
240 };
241 
242 static struct omap_sham_drv sham = {
243 	.dev_list = LIST_HEAD_INIT(sham.dev_list),
244 	.lock = __SPIN_LOCK_UNLOCKED(sham.lock),
245 };
246 
247 static inline u32 omap_sham_read(struct omap_sham_dev *dd, u32 offset)
248 {
249 	return __raw_readl(dd->io_base + offset);
250 }
251 
252 static inline void omap_sham_write(struct omap_sham_dev *dd,
253 					u32 offset, u32 value)
254 {
255 	__raw_writel(value, dd->io_base + offset);
256 }
257 
258 static inline void omap_sham_write_mask(struct omap_sham_dev *dd, u32 address,
259 					u32 value, u32 mask)
260 {
261 	u32 val;
262 
263 	val = omap_sham_read(dd, address);
264 	val &= ~mask;
265 	val |= value;
266 	omap_sham_write(dd, address, val);
267 }
268 
269 static inline int omap_sham_wait(struct omap_sham_dev *dd, u32 offset, u32 bit)
270 {
271 	unsigned long timeout = jiffies + DEFAULT_TIMEOUT_INTERVAL;
272 
273 	while (!(omap_sham_read(dd, offset) & bit)) {
274 		if (time_is_before_jiffies(timeout))
275 			return -ETIMEDOUT;
276 	}
277 
278 	return 0;
279 }
280 
281 static void omap_sham_copy_hash_omap2(struct ahash_request *req, int out)
282 {
283 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
284 	struct omap_sham_dev *dd = ctx->dd;
285 	u32 *hash = (u32 *)ctx->digest;
286 	int i;
287 
288 	for (i = 0; i < dd->pdata->digest_size / sizeof(u32); i++) {
289 		if (out)
290 			hash[i] = omap_sham_read(dd, SHA_REG_IDIGEST(dd, i));
291 		else
292 			omap_sham_write(dd, SHA_REG_IDIGEST(dd, i), hash[i]);
293 	}
294 }
295 
296 static void omap_sham_copy_hash_omap4(struct ahash_request *req, int out)
297 {
298 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
299 	struct omap_sham_dev *dd = ctx->dd;
300 	int i;
301 
302 	if (ctx->flags & BIT(FLAGS_HMAC)) {
303 		struct crypto_ahash *tfm = crypto_ahash_reqtfm(dd->req);
304 		struct omap_sham_ctx *tctx = crypto_ahash_ctx(tfm);
305 		struct omap_sham_hmac_ctx *bctx = tctx->base;
306 		u32 *opad = (u32 *)bctx->opad;
307 
308 		for (i = 0; i < dd->pdata->digest_size / sizeof(u32); i++) {
309 			if (out)
310 				opad[i] = omap_sham_read(dd,
311 						SHA_REG_ODIGEST(dd, i));
312 			else
313 				omap_sham_write(dd, SHA_REG_ODIGEST(dd, i),
314 						opad[i]);
315 		}
316 	}
317 
318 	omap_sham_copy_hash_omap2(req, out);
319 }
320 
321 static void omap_sham_copy_ready_hash(struct ahash_request *req)
322 {
323 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
324 	u32 *in = (u32 *)ctx->digest;
325 	u32 *hash = (u32 *)req->result;
326 	int i, d, big_endian = 0;
327 
328 	if (!hash)
329 		return;
330 
331 	switch (ctx->flags & FLAGS_MODE_MASK) {
332 	case FLAGS_MODE_MD5:
333 		d = MD5_DIGEST_SIZE / sizeof(u32);
334 		break;
335 	case FLAGS_MODE_SHA1:
336 		/* OMAP2 SHA1 is big endian */
337 		if (test_bit(FLAGS_BE32_SHA1, &ctx->dd->flags))
338 			big_endian = 1;
339 		d = SHA1_DIGEST_SIZE / sizeof(u32);
340 		break;
341 	case FLAGS_MODE_SHA224:
342 		d = SHA224_DIGEST_SIZE / sizeof(u32);
343 		break;
344 	case FLAGS_MODE_SHA256:
345 		d = SHA256_DIGEST_SIZE / sizeof(u32);
346 		break;
347 	case FLAGS_MODE_SHA384:
348 		d = SHA384_DIGEST_SIZE / sizeof(u32);
349 		break;
350 	case FLAGS_MODE_SHA512:
351 		d = SHA512_DIGEST_SIZE / sizeof(u32);
352 		break;
353 	default:
354 		d = 0;
355 	}
356 
357 	if (big_endian)
358 		for (i = 0; i < d; i++)
359 			hash[i] = be32_to_cpu(in[i]);
360 	else
361 		for (i = 0; i < d; i++)
362 			hash[i] = le32_to_cpu(in[i]);
363 }
364 
365 static int omap_sham_hw_init(struct omap_sham_dev *dd)
366 {
367 	int err;
368 
369 	err = pm_runtime_get_sync(dd->dev);
370 	if (err < 0) {
371 		dev_err(dd->dev, "failed to get sync: %d\n", err);
372 		return err;
373 	}
374 
375 	if (!test_bit(FLAGS_INIT, &dd->flags)) {
376 		set_bit(FLAGS_INIT, &dd->flags);
377 		dd->err = 0;
378 	}
379 
380 	return 0;
381 }
382 
383 static void omap_sham_write_ctrl_omap2(struct omap_sham_dev *dd, size_t length,
384 				 int final, int dma)
385 {
386 	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
387 	u32 val = length << 5, mask;
388 
389 	if (likely(ctx->digcnt))
390 		omap_sham_write(dd, SHA_REG_DIGCNT(dd), ctx->digcnt);
391 
392 	omap_sham_write_mask(dd, SHA_REG_MASK(dd),
393 		SHA_REG_MASK_IT_EN | (dma ? SHA_REG_MASK_DMA_EN : 0),
394 		SHA_REG_MASK_IT_EN | SHA_REG_MASK_DMA_EN);
395 	/*
396 	 * Setting ALGO_CONST only for the first iteration
397 	 * and CLOSE_HASH only for the last one.
398 	 */
399 	if ((ctx->flags & FLAGS_MODE_MASK) == FLAGS_MODE_SHA1)
400 		val |= SHA_REG_CTRL_ALGO;
401 	if (!ctx->digcnt)
402 		val |= SHA_REG_CTRL_ALGO_CONST;
403 	if (final)
404 		val |= SHA_REG_CTRL_CLOSE_HASH;
405 
406 	mask = SHA_REG_CTRL_ALGO_CONST | SHA_REG_CTRL_CLOSE_HASH |
407 			SHA_REG_CTRL_ALGO | SHA_REG_CTRL_LENGTH;
408 
409 	omap_sham_write_mask(dd, SHA_REG_CTRL, val, mask);
410 }
411 
412 static void omap_sham_trigger_omap2(struct omap_sham_dev *dd, size_t length)
413 {
414 }
415 
416 static int omap_sham_poll_irq_omap2(struct omap_sham_dev *dd)
417 {
418 	return omap_sham_wait(dd, SHA_REG_CTRL, SHA_REG_CTRL_INPUT_READY);
419 }
420 
421 static int get_block_size(struct omap_sham_reqctx *ctx)
422 {
423 	int d;
424 
425 	switch (ctx->flags & FLAGS_MODE_MASK) {
426 	case FLAGS_MODE_MD5:
427 	case FLAGS_MODE_SHA1:
428 		d = SHA1_BLOCK_SIZE;
429 		break;
430 	case FLAGS_MODE_SHA224:
431 	case FLAGS_MODE_SHA256:
432 		d = SHA256_BLOCK_SIZE;
433 		break;
434 	case FLAGS_MODE_SHA384:
435 	case FLAGS_MODE_SHA512:
436 		d = SHA512_BLOCK_SIZE;
437 		break;
438 	default:
439 		d = 0;
440 	}
441 
442 	return d;
443 }
444 
445 static void omap_sham_write_n(struct omap_sham_dev *dd, u32 offset,
446 				    u32 *value, int count)
447 {
448 	for (; count--; value++, offset += 4)
449 		omap_sham_write(dd, offset, *value);
450 }
451 
452 static void omap_sham_write_ctrl_omap4(struct omap_sham_dev *dd, size_t length,
453 				 int final, int dma)
454 {
455 	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
456 	u32 val, mask;
457 
458 	/*
459 	 * Setting ALGO_CONST only for the first iteration and
460 	 * CLOSE_HASH only for the last one. Note that flags mode bits
461 	 * correspond to algorithm encoding in mode register.
462 	 */
463 	val = (ctx->flags & FLAGS_MODE_MASK) >> (FLAGS_MODE_SHIFT);
464 	if (!ctx->digcnt) {
465 		struct crypto_ahash *tfm = crypto_ahash_reqtfm(dd->req);
466 		struct omap_sham_ctx *tctx = crypto_ahash_ctx(tfm);
467 		struct omap_sham_hmac_ctx *bctx = tctx->base;
468 		int bs, nr_dr;
469 
470 		val |= SHA_REG_MODE_ALGO_CONSTANT;
471 
472 		if (ctx->flags & BIT(FLAGS_HMAC)) {
473 			bs = get_block_size(ctx);
474 			nr_dr = bs / (2 * sizeof(u32));
475 			val |= SHA_REG_MODE_HMAC_KEY_PROC;
476 			omap_sham_write_n(dd, SHA_REG_ODIGEST(dd, 0),
477 					  (u32 *)bctx->ipad, nr_dr);
478 			omap_sham_write_n(dd, SHA_REG_IDIGEST(dd, 0),
479 					  (u32 *)bctx->ipad + nr_dr, nr_dr);
480 			ctx->digcnt += bs;
481 		}
482 	}
483 
484 	if (final) {
485 		val |= SHA_REG_MODE_CLOSE_HASH;
486 
487 		if (ctx->flags & BIT(FLAGS_HMAC))
488 			val |= SHA_REG_MODE_HMAC_OUTER_HASH;
489 	}
490 
491 	mask = SHA_REG_MODE_ALGO_CONSTANT | SHA_REG_MODE_CLOSE_HASH |
492 	       SHA_REG_MODE_ALGO_MASK | SHA_REG_MODE_HMAC_OUTER_HASH |
493 	       SHA_REG_MODE_HMAC_KEY_PROC;
494 
495 	dev_dbg(dd->dev, "ctrl: %08x, flags: %08lx\n", val, ctx->flags);
496 	omap_sham_write_mask(dd, SHA_REG_MODE(dd), val, mask);
497 	omap_sham_write(dd, SHA_REG_IRQENA, SHA_REG_IRQENA_OUTPUT_RDY);
498 	omap_sham_write_mask(dd, SHA_REG_MASK(dd),
499 			     SHA_REG_MASK_IT_EN |
500 				     (dma ? SHA_REG_MASK_DMA_EN : 0),
501 			     SHA_REG_MASK_IT_EN | SHA_REG_MASK_DMA_EN);
502 }
503 
504 static void omap_sham_trigger_omap4(struct omap_sham_dev *dd, size_t length)
505 {
506 	omap_sham_write(dd, SHA_REG_LENGTH(dd), length);
507 }
508 
509 static int omap_sham_poll_irq_omap4(struct omap_sham_dev *dd)
510 {
511 	return omap_sham_wait(dd, SHA_REG_IRQSTATUS,
512 			      SHA_REG_IRQSTATUS_INPUT_RDY);
513 }
514 
515 static int omap_sham_xmit_cpu(struct omap_sham_dev *dd, size_t length,
516 			      int final)
517 {
518 	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
519 	int count, len32, bs32, offset = 0;
520 	const u32 *buffer;
521 	int mlen;
522 	struct sg_mapping_iter mi;
523 
524 	dev_dbg(dd->dev, "xmit_cpu: digcnt: %d, length: %d, final: %d\n",
525 						ctx->digcnt, length, final);
526 
527 	dd->pdata->write_ctrl(dd, length, final, 0);
528 	dd->pdata->trigger(dd, length);
529 
530 	/* should be non-zero before next lines to disable clocks later */
531 	ctx->digcnt += length;
532 	ctx->total -= length;
533 
534 	if (final)
535 		set_bit(FLAGS_FINAL, &dd->flags); /* catch last interrupt */
536 
537 	set_bit(FLAGS_CPU, &dd->flags);
538 
539 	len32 = DIV_ROUND_UP(length, sizeof(u32));
540 	bs32 = get_block_size(ctx) / sizeof(u32);
541 
542 	sg_miter_start(&mi, ctx->sg, ctx->sg_len,
543 		       SG_MITER_FROM_SG | SG_MITER_ATOMIC);
544 
545 	mlen = 0;
546 
547 	while (len32) {
548 		if (dd->pdata->poll_irq(dd))
549 			return -ETIMEDOUT;
550 
551 		for (count = 0; count < min(len32, bs32); count++, offset++) {
552 			if (!mlen) {
553 				sg_miter_next(&mi);
554 				mlen = mi.length;
555 				if (!mlen) {
556 					pr_err("sg miter failure.\n");
557 					return -EINVAL;
558 				}
559 				offset = 0;
560 				buffer = mi.addr;
561 			}
562 			omap_sham_write(dd, SHA_REG_DIN(dd, count),
563 					buffer[offset]);
564 			mlen -= 4;
565 		}
566 		len32 -= min(len32, bs32);
567 	}
568 
569 	sg_miter_stop(&mi);
570 
571 	return -EINPROGRESS;
572 }
573 
574 static void omap_sham_dma_callback(void *param)
575 {
576 	struct omap_sham_dev *dd = param;
577 
578 	set_bit(FLAGS_DMA_READY, &dd->flags);
579 	tasklet_schedule(&dd->done_task);
580 }
581 
582 static int omap_sham_xmit_dma(struct omap_sham_dev *dd, size_t length,
583 			      int final)
584 {
585 	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
586 	struct dma_async_tx_descriptor *tx;
587 	struct dma_slave_config cfg;
588 	int ret;
589 
590 	dev_dbg(dd->dev, "xmit_dma: digcnt: %d, length: %d, final: %d\n",
591 						ctx->digcnt, length, final);
592 
593 	if (!dma_map_sg(dd->dev, ctx->sg, ctx->sg_len, DMA_TO_DEVICE)) {
594 		dev_err(dd->dev, "dma_map_sg error\n");
595 		return -EINVAL;
596 	}
597 
598 	memset(&cfg, 0, sizeof(cfg));
599 
600 	cfg.dst_addr = dd->phys_base + SHA_REG_DIN(dd, 0);
601 	cfg.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
602 	cfg.dst_maxburst = get_block_size(ctx) / DMA_SLAVE_BUSWIDTH_4_BYTES;
603 
604 	ret = dmaengine_slave_config(dd->dma_lch, &cfg);
605 	if (ret) {
606 		pr_err("omap-sham: can't configure dmaengine slave: %d\n", ret);
607 		return ret;
608 	}
609 
610 	tx = dmaengine_prep_slave_sg(dd->dma_lch, ctx->sg, ctx->sg_len,
611 				     DMA_MEM_TO_DEV,
612 				     DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
613 
614 	if (!tx) {
615 		dev_err(dd->dev, "prep_slave_sg failed\n");
616 		return -EINVAL;
617 	}
618 
619 	tx->callback = omap_sham_dma_callback;
620 	tx->callback_param = dd;
621 
622 	dd->pdata->write_ctrl(dd, length, final, 1);
623 
624 	ctx->digcnt += length;
625 	ctx->total -= length;
626 
627 	if (final)
628 		set_bit(FLAGS_FINAL, &dd->flags); /* catch last interrupt */
629 
630 	set_bit(FLAGS_DMA_ACTIVE, &dd->flags);
631 
632 	dmaengine_submit(tx);
633 	dma_async_issue_pending(dd->dma_lch);
634 
635 	dd->pdata->trigger(dd, length);
636 
637 	return -EINPROGRESS;
638 }
639 
640 static int omap_sham_copy_sg_lists(struct omap_sham_reqctx *ctx,
641 				   struct scatterlist *sg, int bs, int new_len)
642 {
643 	int n = sg_nents(sg);
644 	struct scatterlist *tmp;
645 	int offset = ctx->offset;
646 
647 	if (ctx->bufcnt)
648 		n++;
649 
650 	ctx->sg = kmalloc_array(n, sizeof(*sg), GFP_KERNEL);
651 	if (!ctx->sg)
652 		return -ENOMEM;
653 
654 	sg_init_table(ctx->sg, n);
655 
656 	tmp = ctx->sg;
657 
658 	ctx->sg_len = 0;
659 
660 	if (ctx->bufcnt) {
661 		sg_set_buf(tmp, ctx->dd->xmit_buf, ctx->bufcnt);
662 		tmp = sg_next(tmp);
663 		ctx->sg_len++;
664 	}
665 
666 	while (sg && new_len) {
667 		int len = sg->length - offset;
668 
669 		if (offset) {
670 			offset -= sg->length;
671 			if (offset < 0)
672 				offset = 0;
673 		}
674 
675 		if (new_len < len)
676 			len = new_len;
677 
678 		if (len > 0) {
679 			new_len -= len;
680 			sg_set_page(tmp, sg_page(sg), len, sg->offset);
681 			if (new_len <= 0)
682 				sg_mark_end(tmp);
683 			tmp = sg_next(tmp);
684 			ctx->sg_len++;
685 		}
686 
687 		sg = sg_next(sg);
688 	}
689 
690 	set_bit(FLAGS_SGS_ALLOCED, &ctx->dd->flags);
691 
692 	ctx->bufcnt = 0;
693 
694 	return 0;
695 }
696 
697 static int omap_sham_copy_sgs(struct omap_sham_reqctx *ctx,
698 			      struct scatterlist *sg, int bs, int new_len)
699 {
700 	int pages;
701 	void *buf;
702 	int len;
703 
704 	len = new_len + ctx->bufcnt;
705 
706 	pages = get_order(ctx->total);
707 
708 	buf = (void *)__get_free_pages(GFP_ATOMIC, pages);
709 	if (!buf) {
710 		pr_err("Couldn't allocate pages for unaligned cases.\n");
711 		return -ENOMEM;
712 	}
713 
714 	if (ctx->bufcnt)
715 		memcpy(buf, ctx->dd->xmit_buf, ctx->bufcnt);
716 
717 	scatterwalk_map_and_copy(buf + ctx->bufcnt, sg, ctx->offset,
718 				 ctx->total - ctx->bufcnt, 0);
719 	sg_init_table(ctx->sgl, 1);
720 	sg_set_buf(ctx->sgl, buf, len);
721 	ctx->sg = ctx->sgl;
722 	set_bit(FLAGS_SGS_COPIED, &ctx->dd->flags);
723 	ctx->sg_len = 1;
724 	ctx->bufcnt = 0;
725 	ctx->offset = 0;
726 
727 	return 0;
728 }
729 
730 static int omap_sham_align_sgs(struct scatterlist *sg,
731 			       int nbytes, int bs, bool final,
732 			       struct omap_sham_reqctx *rctx)
733 {
734 	int n = 0;
735 	bool aligned = true;
736 	bool list_ok = true;
737 	struct scatterlist *sg_tmp = sg;
738 	int new_len;
739 	int offset = rctx->offset;
740 
741 	if (!sg || !sg->length || !nbytes)
742 		return 0;
743 
744 	new_len = nbytes;
745 
746 	if (offset)
747 		list_ok = false;
748 
749 	if (final)
750 		new_len = DIV_ROUND_UP(new_len, bs) * bs;
751 	else
752 		new_len = (new_len - 1) / bs * bs;
753 
754 	if (nbytes != new_len)
755 		list_ok = false;
756 
757 	while (nbytes > 0 && sg_tmp) {
758 		n++;
759 
760 #ifdef CONFIG_ZONE_DMA
761 		if (page_zonenum(sg_page(sg_tmp)) != ZONE_DMA) {
762 			aligned = false;
763 			break;
764 		}
765 #endif
766 
767 		if (offset < sg_tmp->length) {
768 			if (!IS_ALIGNED(offset + sg_tmp->offset, 4)) {
769 				aligned = false;
770 				break;
771 			}
772 
773 			if (!IS_ALIGNED(sg_tmp->length - offset, bs)) {
774 				aligned = false;
775 				break;
776 			}
777 		}
778 
779 		if (offset) {
780 			offset -= sg_tmp->length;
781 			if (offset < 0) {
782 				nbytes += offset;
783 				offset = 0;
784 			}
785 		} else {
786 			nbytes -= sg_tmp->length;
787 		}
788 
789 		sg_tmp = sg_next(sg_tmp);
790 
791 		if (nbytes < 0) {
792 			list_ok = false;
793 			break;
794 		}
795 	}
796 
797 	if (!aligned)
798 		return omap_sham_copy_sgs(rctx, sg, bs, new_len);
799 	else if (!list_ok)
800 		return omap_sham_copy_sg_lists(rctx, sg, bs, new_len);
801 
802 	rctx->sg_len = n;
803 	rctx->sg = sg;
804 
805 	return 0;
806 }
807 
808 static int omap_sham_prepare_request(struct ahash_request *req, bool update)
809 {
810 	struct omap_sham_reqctx *rctx = ahash_request_ctx(req);
811 	int bs;
812 	int ret;
813 	int nbytes;
814 	bool final = rctx->flags & BIT(FLAGS_FINUP);
815 	int xmit_len, hash_later;
816 
817 	bs = get_block_size(rctx);
818 
819 	if (update)
820 		nbytes = req->nbytes;
821 	else
822 		nbytes = 0;
823 
824 	rctx->total = nbytes + rctx->bufcnt;
825 
826 	if (!rctx->total)
827 		return 0;
828 
829 	if (nbytes && (!IS_ALIGNED(rctx->bufcnt, bs))) {
830 		int len = bs - rctx->bufcnt % bs;
831 
832 		if (len > nbytes)
833 			len = nbytes;
834 		scatterwalk_map_and_copy(rctx->buffer + rctx->bufcnt, req->src,
835 					 0, len, 0);
836 		rctx->bufcnt += len;
837 		nbytes -= len;
838 		rctx->offset = len;
839 	}
840 
841 	if (rctx->bufcnt)
842 		memcpy(rctx->dd->xmit_buf, rctx->buffer, rctx->bufcnt);
843 
844 	ret = omap_sham_align_sgs(req->src, nbytes, bs, final, rctx);
845 	if (ret)
846 		return ret;
847 
848 	xmit_len = rctx->total;
849 
850 	if (!IS_ALIGNED(xmit_len, bs)) {
851 		if (final)
852 			xmit_len = DIV_ROUND_UP(xmit_len, bs) * bs;
853 		else
854 			xmit_len = xmit_len / bs * bs;
855 	} else if (!final) {
856 		xmit_len -= bs;
857 	}
858 
859 	hash_later = rctx->total - xmit_len;
860 	if (hash_later < 0)
861 		hash_later = 0;
862 
863 	if (rctx->bufcnt && nbytes) {
864 		/* have data from previous operation and current */
865 		sg_init_table(rctx->sgl, 2);
866 		sg_set_buf(rctx->sgl, rctx->dd->xmit_buf, rctx->bufcnt);
867 
868 		sg_chain(rctx->sgl, 2, req->src);
869 
870 		rctx->sg = rctx->sgl;
871 
872 		rctx->sg_len++;
873 	} else if (rctx->bufcnt) {
874 		/* have buffered data only */
875 		sg_init_table(rctx->sgl, 1);
876 		sg_set_buf(rctx->sgl, rctx->dd->xmit_buf, xmit_len);
877 
878 		rctx->sg = rctx->sgl;
879 
880 		rctx->sg_len = 1;
881 	}
882 
883 	if (hash_later) {
884 		int offset = 0;
885 
886 		if (hash_later > req->nbytes) {
887 			memcpy(rctx->buffer, rctx->buffer + xmit_len,
888 			       hash_later - req->nbytes);
889 			offset = hash_later - req->nbytes;
890 		}
891 
892 		if (req->nbytes) {
893 			scatterwalk_map_and_copy(rctx->buffer + offset,
894 						 req->src,
895 						 offset + req->nbytes -
896 						 hash_later, hash_later, 0);
897 		}
898 
899 		rctx->bufcnt = hash_later;
900 	} else {
901 		rctx->bufcnt = 0;
902 	}
903 
904 	if (!final)
905 		rctx->total = xmit_len;
906 
907 	return 0;
908 }
909 
910 static int omap_sham_update_dma_stop(struct omap_sham_dev *dd)
911 {
912 	struct omap_sham_reqctx *ctx = ahash_request_ctx(dd->req);
913 
914 	dma_unmap_sg(dd->dev, ctx->sg, ctx->sg_len, DMA_TO_DEVICE);
915 
916 	clear_bit(FLAGS_DMA_ACTIVE, &dd->flags);
917 
918 	return 0;
919 }
920 
921 static int omap_sham_init(struct ahash_request *req)
922 {
923 	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
924 	struct omap_sham_ctx *tctx = crypto_ahash_ctx(tfm);
925 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
926 	struct omap_sham_dev *dd = NULL, *tmp;
927 	int bs = 0;
928 
929 	spin_lock_bh(&sham.lock);
930 	if (!tctx->dd) {
931 		list_for_each_entry(tmp, &sham.dev_list, list) {
932 			dd = tmp;
933 			break;
934 		}
935 		tctx->dd = dd;
936 	} else {
937 		dd = tctx->dd;
938 	}
939 	spin_unlock_bh(&sham.lock);
940 
941 	ctx->dd = dd;
942 
943 	ctx->flags = 0;
944 
945 	dev_dbg(dd->dev, "init: digest size: %d\n",
946 		crypto_ahash_digestsize(tfm));
947 
948 	switch (crypto_ahash_digestsize(tfm)) {
949 	case MD5_DIGEST_SIZE:
950 		ctx->flags |= FLAGS_MODE_MD5;
951 		bs = SHA1_BLOCK_SIZE;
952 		break;
953 	case SHA1_DIGEST_SIZE:
954 		ctx->flags |= FLAGS_MODE_SHA1;
955 		bs = SHA1_BLOCK_SIZE;
956 		break;
957 	case SHA224_DIGEST_SIZE:
958 		ctx->flags |= FLAGS_MODE_SHA224;
959 		bs = SHA224_BLOCK_SIZE;
960 		break;
961 	case SHA256_DIGEST_SIZE:
962 		ctx->flags |= FLAGS_MODE_SHA256;
963 		bs = SHA256_BLOCK_SIZE;
964 		break;
965 	case SHA384_DIGEST_SIZE:
966 		ctx->flags |= FLAGS_MODE_SHA384;
967 		bs = SHA384_BLOCK_SIZE;
968 		break;
969 	case SHA512_DIGEST_SIZE:
970 		ctx->flags |= FLAGS_MODE_SHA512;
971 		bs = SHA512_BLOCK_SIZE;
972 		break;
973 	}
974 
975 	ctx->bufcnt = 0;
976 	ctx->digcnt = 0;
977 	ctx->total = 0;
978 	ctx->offset = 0;
979 	ctx->buflen = BUFLEN;
980 
981 	if (tctx->flags & BIT(FLAGS_HMAC)) {
982 		if (!test_bit(FLAGS_AUTO_XOR, &dd->flags)) {
983 			struct omap_sham_hmac_ctx *bctx = tctx->base;
984 
985 			memcpy(ctx->buffer, bctx->ipad, bs);
986 			ctx->bufcnt = bs;
987 		}
988 
989 		ctx->flags |= BIT(FLAGS_HMAC);
990 	}
991 
992 	return 0;
993 
994 }
995 
996 static int omap_sham_update_req(struct omap_sham_dev *dd)
997 {
998 	struct ahash_request *req = dd->req;
999 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1000 	int err;
1001 	bool final = ctx->flags & BIT(FLAGS_FINUP);
1002 
1003 	dev_dbg(dd->dev, "update_req: total: %u, digcnt: %d, finup: %d\n",
1004 		 ctx->total, ctx->digcnt, (ctx->flags & BIT(FLAGS_FINUP)) != 0);
1005 
1006 	if (ctx->total < get_block_size(ctx) ||
1007 	    ctx->total < dd->fallback_sz)
1008 		ctx->flags |= BIT(FLAGS_CPU);
1009 
1010 	if (ctx->flags & BIT(FLAGS_CPU))
1011 		err = omap_sham_xmit_cpu(dd, ctx->total, final);
1012 	else
1013 		err = omap_sham_xmit_dma(dd, ctx->total, final);
1014 
1015 	/* wait for dma completion before can take more data */
1016 	dev_dbg(dd->dev, "update: err: %d, digcnt: %d\n", err, ctx->digcnt);
1017 
1018 	return err;
1019 }
1020 
1021 static int omap_sham_final_req(struct omap_sham_dev *dd)
1022 {
1023 	struct ahash_request *req = dd->req;
1024 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1025 	int err = 0, use_dma = 1;
1026 
1027 	if ((ctx->total <= get_block_size(ctx)) || dd->polling_mode)
1028 		/*
1029 		 * faster to handle last block with cpu or
1030 		 * use cpu when dma is not present.
1031 		 */
1032 		use_dma = 0;
1033 
1034 	if (use_dma)
1035 		err = omap_sham_xmit_dma(dd, ctx->total, 1);
1036 	else
1037 		err = omap_sham_xmit_cpu(dd, ctx->total, 1);
1038 
1039 	ctx->bufcnt = 0;
1040 
1041 	dev_dbg(dd->dev, "final_req: err: %d\n", err);
1042 
1043 	return err;
1044 }
1045 
1046 static int omap_sham_finish_hmac(struct ahash_request *req)
1047 {
1048 	struct omap_sham_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
1049 	struct omap_sham_hmac_ctx *bctx = tctx->base;
1050 	int bs = crypto_shash_blocksize(bctx->shash);
1051 	int ds = crypto_shash_digestsize(bctx->shash);
1052 	SHASH_DESC_ON_STACK(shash, bctx->shash);
1053 
1054 	shash->tfm = bctx->shash;
1055 
1056 	return crypto_shash_init(shash) ?:
1057 	       crypto_shash_update(shash, bctx->opad, bs) ?:
1058 	       crypto_shash_finup(shash, req->result, ds, req->result);
1059 }
1060 
1061 static int omap_sham_finish(struct ahash_request *req)
1062 {
1063 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1064 	struct omap_sham_dev *dd = ctx->dd;
1065 	int err = 0;
1066 
1067 	if (ctx->digcnt) {
1068 		omap_sham_copy_ready_hash(req);
1069 		if ((ctx->flags & BIT(FLAGS_HMAC)) &&
1070 				!test_bit(FLAGS_AUTO_XOR, &dd->flags))
1071 			err = omap_sham_finish_hmac(req);
1072 	}
1073 
1074 	dev_dbg(dd->dev, "digcnt: %d, bufcnt: %d\n", ctx->digcnt, ctx->bufcnt);
1075 
1076 	return err;
1077 }
1078 
1079 static void omap_sham_finish_req(struct ahash_request *req, int err)
1080 {
1081 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1082 	struct omap_sham_dev *dd = ctx->dd;
1083 
1084 	if (test_bit(FLAGS_SGS_COPIED, &dd->flags))
1085 		free_pages((unsigned long)sg_virt(ctx->sg),
1086 			   get_order(ctx->sg->length + ctx->bufcnt));
1087 
1088 	if (test_bit(FLAGS_SGS_ALLOCED, &dd->flags))
1089 		kfree(ctx->sg);
1090 
1091 	ctx->sg = NULL;
1092 
1093 	dd->flags &= ~(BIT(FLAGS_SGS_ALLOCED) | BIT(FLAGS_SGS_COPIED));
1094 
1095 	if (!err) {
1096 		dd->pdata->copy_hash(req, 1);
1097 		if (test_bit(FLAGS_FINAL, &dd->flags))
1098 			err = omap_sham_finish(req);
1099 	} else {
1100 		ctx->flags |= BIT(FLAGS_ERROR);
1101 	}
1102 
1103 	/* atomic operation is not needed here */
1104 	dd->flags &= ~(BIT(FLAGS_BUSY) | BIT(FLAGS_FINAL) | BIT(FLAGS_CPU) |
1105 			BIT(FLAGS_DMA_READY) | BIT(FLAGS_OUTPUT_READY));
1106 
1107 	pm_runtime_mark_last_busy(dd->dev);
1108 	pm_runtime_put_autosuspend(dd->dev);
1109 
1110 	if (req->base.complete)
1111 		req->base.complete(&req->base, err);
1112 }
1113 
1114 static int omap_sham_handle_queue(struct omap_sham_dev *dd,
1115 				  struct ahash_request *req)
1116 {
1117 	struct crypto_async_request *async_req, *backlog;
1118 	struct omap_sham_reqctx *ctx;
1119 	unsigned long flags;
1120 	int err = 0, ret = 0;
1121 
1122 retry:
1123 	spin_lock_irqsave(&dd->lock, flags);
1124 	if (req)
1125 		ret = ahash_enqueue_request(&dd->queue, req);
1126 	if (test_bit(FLAGS_BUSY, &dd->flags)) {
1127 		spin_unlock_irqrestore(&dd->lock, flags);
1128 		return ret;
1129 	}
1130 	backlog = crypto_get_backlog(&dd->queue);
1131 	async_req = crypto_dequeue_request(&dd->queue);
1132 	if (async_req)
1133 		set_bit(FLAGS_BUSY, &dd->flags);
1134 	spin_unlock_irqrestore(&dd->lock, flags);
1135 
1136 	if (!async_req)
1137 		return ret;
1138 
1139 	if (backlog)
1140 		backlog->complete(backlog, -EINPROGRESS);
1141 
1142 	req = ahash_request_cast(async_req);
1143 	dd->req = req;
1144 	ctx = ahash_request_ctx(req);
1145 
1146 	err = omap_sham_prepare_request(req, ctx->op == OP_UPDATE);
1147 	if (err || !ctx->total)
1148 		goto err1;
1149 
1150 	dev_dbg(dd->dev, "handling new req, op: %lu, nbytes: %d\n",
1151 						ctx->op, req->nbytes);
1152 
1153 	err = omap_sham_hw_init(dd);
1154 	if (err)
1155 		goto err1;
1156 
1157 	if (ctx->digcnt)
1158 		/* request has changed - restore hash */
1159 		dd->pdata->copy_hash(req, 0);
1160 
1161 	if (ctx->op == OP_UPDATE) {
1162 		err = omap_sham_update_req(dd);
1163 		if (err != -EINPROGRESS && (ctx->flags & BIT(FLAGS_FINUP)))
1164 			/* no final() after finup() */
1165 			err = omap_sham_final_req(dd);
1166 	} else if (ctx->op == OP_FINAL) {
1167 		err = omap_sham_final_req(dd);
1168 	}
1169 err1:
1170 	dev_dbg(dd->dev, "exit, err: %d\n", err);
1171 
1172 	if (err != -EINPROGRESS) {
1173 		/* done_task will not finish it, so do it here */
1174 		omap_sham_finish_req(req, err);
1175 		req = NULL;
1176 
1177 		/*
1178 		 * Execute next request immediately if there is anything
1179 		 * in queue.
1180 		 */
1181 		goto retry;
1182 	}
1183 
1184 	return ret;
1185 }
1186 
1187 static int omap_sham_enqueue(struct ahash_request *req, unsigned int op)
1188 {
1189 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1190 	struct omap_sham_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
1191 	struct omap_sham_dev *dd = tctx->dd;
1192 
1193 	ctx->op = op;
1194 
1195 	return omap_sham_handle_queue(dd, req);
1196 }
1197 
1198 static int omap_sham_update(struct ahash_request *req)
1199 {
1200 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1201 	struct omap_sham_dev *dd = ctx->dd;
1202 
1203 	if (!req->nbytes)
1204 		return 0;
1205 
1206 	if (ctx->bufcnt + req->nbytes <= ctx->buflen) {
1207 		scatterwalk_map_and_copy(ctx->buffer + ctx->bufcnt, req->src,
1208 					 0, req->nbytes, 0);
1209 		ctx->bufcnt += req->nbytes;
1210 		return 0;
1211 	}
1212 
1213 	if (dd->polling_mode)
1214 		ctx->flags |= BIT(FLAGS_CPU);
1215 
1216 	return omap_sham_enqueue(req, OP_UPDATE);
1217 }
1218 
1219 static int omap_sham_shash_digest(struct crypto_shash *tfm, u32 flags,
1220 				  const u8 *data, unsigned int len, u8 *out)
1221 {
1222 	SHASH_DESC_ON_STACK(shash, tfm);
1223 
1224 	shash->tfm = tfm;
1225 
1226 	return crypto_shash_digest(shash, data, len, out);
1227 }
1228 
1229 static int omap_sham_final_shash(struct ahash_request *req)
1230 {
1231 	struct omap_sham_ctx *tctx = crypto_tfm_ctx(req->base.tfm);
1232 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1233 	int offset = 0;
1234 
1235 	/*
1236 	 * If we are running HMAC on limited hardware support, skip
1237 	 * the ipad in the beginning of the buffer if we are going for
1238 	 * software fallback algorithm.
1239 	 */
1240 	if (test_bit(FLAGS_HMAC, &ctx->flags) &&
1241 	    !test_bit(FLAGS_AUTO_XOR, &ctx->dd->flags))
1242 		offset = get_block_size(ctx);
1243 
1244 	return omap_sham_shash_digest(tctx->fallback, req->base.flags,
1245 				      ctx->buffer + offset,
1246 				      ctx->bufcnt - offset, req->result);
1247 }
1248 
1249 static int omap_sham_final(struct ahash_request *req)
1250 {
1251 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1252 
1253 	ctx->flags |= BIT(FLAGS_FINUP);
1254 
1255 	if (ctx->flags & BIT(FLAGS_ERROR))
1256 		return 0; /* uncompleted hash is not needed */
1257 
1258 	/*
1259 	 * OMAP HW accel works only with buffers >= 9.
1260 	 * HMAC is always >= 9 because ipad == block size.
1261 	 * If buffersize is less than fallback_sz, we use fallback
1262 	 * SW encoding, as using DMA + HW in this case doesn't provide
1263 	 * any benefit.
1264 	 */
1265 	if (!ctx->digcnt && ctx->bufcnt < ctx->dd->fallback_sz)
1266 		return omap_sham_final_shash(req);
1267 	else if (ctx->bufcnt)
1268 		return omap_sham_enqueue(req, OP_FINAL);
1269 
1270 	/* copy ready hash (+ finalize hmac) */
1271 	return omap_sham_finish(req);
1272 }
1273 
1274 static int omap_sham_finup(struct ahash_request *req)
1275 {
1276 	struct omap_sham_reqctx *ctx = ahash_request_ctx(req);
1277 	int err1, err2;
1278 
1279 	ctx->flags |= BIT(FLAGS_FINUP);
1280 
1281 	err1 = omap_sham_update(req);
1282 	if (err1 == -EINPROGRESS || err1 == -EBUSY)
1283 		return err1;
1284 	/*
1285 	 * final() has to be always called to cleanup resources
1286 	 * even if udpate() failed, except EINPROGRESS
1287 	 */
1288 	err2 = omap_sham_final(req);
1289 
1290 	return err1 ?: err2;
1291 }
1292 
1293 static int omap_sham_digest(struct ahash_request *req)
1294 {
1295 	return omap_sham_init(req) ?: omap_sham_finup(req);
1296 }
1297 
1298 static int omap_sham_setkey(struct crypto_ahash *tfm, const u8 *key,
1299 		      unsigned int keylen)
1300 {
1301 	struct omap_sham_ctx *tctx = crypto_ahash_ctx(tfm);
1302 	struct omap_sham_hmac_ctx *bctx = tctx->base;
1303 	int bs = crypto_shash_blocksize(bctx->shash);
1304 	int ds = crypto_shash_digestsize(bctx->shash);
1305 	struct omap_sham_dev *dd = NULL, *tmp;
1306 	int err, i;
1307 
1308 	spin_lock_bh(&sham.lock);
1309 	if (!tctx->dd) {
1310 		list_for_each_entry(tmp, &sham.dev_list, list) {
1311 			dd = tmp;
1312 			break;
1313 		}
1314 		tctx->dd = dd;
1315 	} else {
1316 		dd = tctx->dd;
1317 	}
1318 	spin_unlock_bh(&sham.lock);
1319 
1320 	err = crypto_shash_setkey(tctx->fallback, key, keylen);
1321 	if (err)
1322 		return err;
1323 
1324 	if (keylen > bs) {
1325 		err = omap_sham_shash_digest(bctx->shash,
1326 				crypto_shash_get_flags(bctx->shash),
1327 				key, keylen, bctx->ipad);
1328 		if (err)
1329 			return err;
1330 		keylen = ds;
1331 	} else {
1332 		memcpy(bctx->ipad, key, keylen);
1333 	}
1334 
1335 	memset(bctx->ipad + keylen, 0, bs - keylen);
1336 
1337 	if (!test_bit(FLAGS_AUTO_XOR, &dd->flags)) {
1338 		memcpy(bctx->opad, bctx->ipad, bs);
1339 
1340 		for (i = 0; i < bs; i++) {
1341 			bctx->ipad[i] ^= HMAC_IPAD_VALUE;
1342 			bctx->opad[i] ^= HMAC_OPAD_VALUE;
1343 		}
1344 	}
1345 
1346 	return err;
1347 }
1348 
1349 static int omap_sham_cra_init_alg(struct crypto_tfm *tfm, const char *alg_base)
1350 {
1351 	struct omap_sham_ctx *tctx = crypto_tfm_ctx(tfm);
1352 	const char *alg_name = crypto_tfm_alg_name(tfm);
1353 
1354 	/* Allocate a fallback and abort if it failed. */
1355 	tctx->fallback = crypto_alloc_shash(alg_name, 0,
1356 					    CRYPTO_ALG_NEED_FALLBACK);
1357 	if (IS_ERR(tctx->fallback)) {
1358 		pr_err("omap-sham: fallback driver '%s' "
1359 				"could not be loaded.\n", alg_name);
1360 		return PTR_ERR(tctx->fallback);
1361 	}
1362 
1363 	crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm),
1364 				 sizeof(struct omap_sham_reqctx) + BUFLEN);
1365 
1366 	if (alg_base) {
1367 		struct omap_sham_hmac_ctx *bctx = tctx->base;
1368 		tctx->flags |= BIT(FLAGS_HMAC);
1369 		bctx->shash = crypto_alloc_shash(alg_base, 0,
1370 						CRYPTO_ALG_NEED_FALLBACK);
1371 		if (IS_ERR(bctx->shash)) {
1372 			pr_err("omap-sham: base driver '%s' "
1373 					"could not be loaded.\n", alg_base);
1374 			crypto_free_shash(tctx->fallback);
1375 			return PTR_ERR(bctx->shash);
1376 		}
1377 
1378 	}
1379 
1380 	return 0;
1381 }
1382 
1383 static int omap_sham_cra_init(struct crypto_tfm *tfm)
1384 {
1385 	return omap_sham_cra_init_alg(tfm, NULL);
1386 }
1387 
1388 static int omap_sham_cra_sha1_init(struct crypto_tfm *tfm)
1389 {
1390 	return omap_sham_cra_init_alg(tfm, "sha1");
1391 }
1392 
1393 static int omap_sham_cra_sha224_init(struct crypto_tfm *tfm)
1394 {
1395 	return omap_sham_cra_init_alg(tfm, "sha224");
1396 }
1397 
1398 static int omap_sham_cra_sha256_init(struct crypto_tfm *tfm)
1399 {
1400 	return omap_sham_cra_init_alg(tfm, "sha256");
1401 }
1402 
1403 static int omap_sham_cra_md5_init(struct crypto_tfm *tfm)
1404 {
1405 	return omap_sham_cra_init_alg(tfm, "md5");
1406 }
1407 
1408 static int omap_sham_cra_sha384_init(struct crypto_tfm *tfm)
1409 {
1410 	return omap_sham_cra_init_alg(tfm, "sha384");
1411 }
1412 
1413 static int omap_sham_cra_sha512_init(struct crypto_tfm *tfm)
1414 {
1415 	return omap_sham_cra_init_alg(tfm, "sha512");
1416 }
1417 
1418 static void omap_sham_cra_exit(struct crypto_tfm *tfm)
1419 {
1420 	struct omap_sham_ctx *tctx = crypto_tfm_ctx(tfm);
1421 
1422 	crypto_free_shash(tctx->fallback);
1423 	tctx->fallback = NULL;
1424 
1425 	if (tctx->flags & BIT(FLAGS_HMAC)) {
1426 		struct omap_sham_hmac_ctx *bctx = tctx->base;
1427 		crypto_free_shash(bctx->shash);
1428 	}
1429 }
1430 
1431 static int omap_sham_export(struct ahash_request *req, void *out)
1432 {
1433 	struct omap_sham_reqctx *rctx = ahash_request_ctx(req);
1434 
1435 	memcpy(out, rctx, sizeof(*rctx) + rctx->bufcnt);
1436 
1437 	return 0;
1438 }
1439 
1440 static int omap_sham_import(struct ahash_request *req, const void *in)
1441 {
1442 	struct omap_sham_reqctx *rctx = ahash_request_ctx(req);
1443 	const struct omap_sham_reqctx *ctx_in = in;
1444 
1445 	memcpy(rctx, in, sizeof(*rctx) + ctx_in->bufcnt);
1446 
1447 	return 0;
1448 }
1449 
1450 static struct ahash_alg algs_sha1_md5[] = {
1451 {
1452 	.init		= omap_sham_init,
1453 	.update		= omap_sham_update,
1454 	.final		= omap_sham_final,
1455 	.finup		= omap_sham_finup,
1456 	.digest		= omap_sham_digest,
1457 	.halg.digestsize	= SHA1_DIGEST_SIZE,
1458 	.halg.base	= {
1459 		.cra_name		= "sha1",
1460 		.cra_driver_name	= "omap-sha1",
1461 		.cra_priority		= 400,
1462 		.cra_flags		= CRYPTO_ALG_KERN_DRIVER_ONLY |
1463 						CRYPTO_ALG_ASYNC |
1464 						CRYPTO_ALG_NEED_FALLBACK,
1465 		.cra_blocksize		= SHA1_BLOCK_SIZE,
1466 		.cra_ctxsize		= sizeof(struct omap_sham_ctx),
1467 		.cra_alignmask		= OMAP_ALIGN_MASK,
1468 		.cra_module		= THIS_MODULE,
1469 		.cra_init		= omap_sham_cra_init,
1470 		.cra_exit		= omap_sham_cra_exit,
1471 	}
1472 },
1473 {
1474 	.init		= omap_sham_init,
1475 	.update		= omap_sham_update,
1476 	.final		= omap_sham_final,
1477 	.finup		= omap_sham_finup,
1478 	.digest		= omap_sham_digest,
1479 	.halg.digestsize	= MD5_DIGEST_SIZE,
1480 	.halg.base	= {
1481 		.cra_name		= "md5",
1482 		.cra_driver_name	= "omap-md5",
1483 		.cra_priority		= 400,
1484 		.cra_flags		= CRYPTO_ALG_KERN_DRIVER_ONLY |
1485 						CRYPTO_ALG_ASYNC |
1486 						CRYPTO_ALG_NEED_FALLBACK,
1487 		.cra_blocksize		= SHA1_BLOCK_SIZE,
1488 		.cra_ctxsize		= sizeof(struct omap_sham_ctx),
1489 		.cra_alignmask		= OMAP_ALIGN_MASK,
1490 		.cra_module		= THIS_MODULE,
1491 		.cra_init		= omap_sham_cra_init,
1492 		.cra_exit		= omap_sham_cra_exit,
1493 	}
1494 },
1495 {
1496 	.init		= omap_sham_init,
1497 	.update		= omap_sham_update,
1498 	.final		= omap_sham_final,
1499 	.finup		= omap_sham_finup,
1500 	.digest		= omap_sham_digest,
1501 	.setkey		= omap_sham_setkey,
1502 	.halg.digestsize	= SHA1_DIGEST_SIZE,
1503 	.halg.base	= {
1504 		.cra_name		= "hmac(sha1)",
1505 		.cra_driver_name	= "omap-hmac-sha1",
1506 		.cra_priority		= 400,
1507 		.cra_flags		= CRYPTO_ALG_KERN_DRIVER_ONLY |
1508 						CRYPTO_ALG_ASYNC |
1509 						CRYPTO_ALG_NEED_FALLBACK,
1510 		.cra_blocksize		= SHA1_BLOCK_SIZE,
1511 		.cra_ctxsize		= sizeof(struct omap_sham_ctx) +
1512 					sizeof(struct omap_sham_hmac_ctx),
1513 		.cra_alignmask		= OMAP_ALIGN_MASK,
1514 		.cra_module		= THIS_MODULE,
1515 		.cra_init		= omap_sham_cra_sha1_init,
1516 		.cra_exit		= omap_sham_cra_exit,
1517 	}
1518 },
1519 {
1520 	.init		= omap_sham_init,
1521 	.update		= omap_sham_update,
1522 	.final		= omap_sham_final,
1523 	.finup		= omap_sham_finup,
1524 	.digest		= omap_sham_digest,
1525 	.setkey		= omap_sham_setkey,
1526 	.halg.digestsize	= MD5_DIGEST_SIZE,
1527 	.halg.base	= {
1528 		.cra_name		= "hmac(md5)",
1529 		.cra_driver_name	= "omap-hmac-md5",
1530 		.cra_priority		= 400,
1531 		.cra_flags		= CRYPTO_ALG_KERN_DRIVER_ONLY |
1532 						CRYPTO_ALG_ASYNC |
1533 						CRYPTO_ALG_NEED_FALLBACK,
1534 		.cra_blocksize		= SHA1_BLOCK_SIZE,
1535 		.cra_ctxsize		= sizeof(struct omap_sham_ctx) +
1536 					sizeof(struct omap_sham_hmac_ctx),
1537 		.cra_alignmask		= OMAP_ALIGN_MASK,
1538 		.cra_module		= THIS_MODULE,
1539 		.cra_init		= omap_sham_cra_md5_init,
1540 		.cra_exit		= omap_sham_cra_exit,
1541 	}
1542 }
1543 };
1544 
1545 /* OMAP4 has some algs in addition to what OMAP2 has */
1546 static struct ahash_alg algs_sha224_sha256[] = {
1547 {
1548 	.init		= omap_sham_init,
1549 	.update		= omap_sham_update,
1550 	.final		= omap_sham_final,
1551 	.finup		= omap_sham_finup,
1552 	.digest		= omap_sham_digest,
1553 	.halg.digestsize	= SHA224_DIGEST_SIZE,
1554 	.halg.base	= {
1555 		.cra_name		= "sha224",
1556 		.cra_driver_name	= "omap-sha224",
1557 		.cra_priority		= 400,
1558 		.cra_flags		= CRYPTO_ALG_ASYNC |
1559 						CRYPTO_ALG_NEED_FALLBACK,
1560 		.cra_blocksize		= SHA224_BLOCK_SIZE,
1561 		.cra_ctxsize		= sizeof(struct omap_sham_ctx),
1562 		.cra_alignmask		= OMAP_ALIGN_MASK,
1563 		.cra_module		= THIS_MODULE,
1564 		.cra_init		= omap_sham_cra_init,
1565 		.cra_exit		= omap_sham_cra_exit,
1566 	}
1567 },
1568 {
1569 	.init		= omap_sham_init,
1570 	.update		= omap_sham_update,
1571 	.final		= omap_sham_final,
1572 	.finup		= omap_sham_finup,
1573 	.digest		= omap_sham_digest,
1574 	.halg.digestsize	= SHA256_DIGEST_SIZE,
1575 	.halg.base	= {
1576 		.cra_name		= "sha256",
1577 		.cra_driver_name	= "omap-sha256",
1578 		.cra_priority		= 400,
1579 		.cra_flags		= CRYPTO_ALG_ASYNC |
1580 						CRYPTO_ALG_NEED_FALLBACK,
1581 		.cra_blocksize		= SHA256_BLOCK_SIZE,
1582 		.cra_ctxsize		= sizeof(struct omap_sham_ctx),
1583 		.cra_alignmask		= OMAP_ALIGN_MASK,
1584 		.cra_module		= THIS_MODULE,
1585 		.cra_init		= omap_sham_cra_init,
1586 		.cra_exit		= omap_sham_cra_exit,
1587 	}
1588 },
1589 {
1590 	.init		= omap_sham_init,
1591 	.update		= omap_sham_update,
1592 	.final		= omap_sham_final,
1593 	.finup		= omap_sham_finup,
1594 	.digest		= omap_sham_digest,
1595 	.setkey		= omap_sham_setkey,
1596 	.halg.digestsize	= SHA224_DIGEST_SIZE,
1597 	.halg.base	= {
1598 		.cra_name		= "hmac(sha224)",
1599 		.cra_driver_name	= "omap-hmac-sha224",
1600 		.cra_priority		= 400,
1601 		.cra_flags		= CRYPTO_ALG_ASYNC |
1602 						CRYPTO_ALG_NEED_FALLBACK,
1603 		.cra_blocksize		= SHA224_BLOCK_SIZE,
1604 		.cra_ctxsize		= sizeof(struct omap_sham_ctx) +
1605 					sizeof(struct omap_sham_hmac_ctx),
1606 		.cra_alignmask		= OMAP_ALIGN_MASK,
1607 		.cra_module		= THIS_MODULE,
1608 		.cra_init		= omap_sham_cra_sha224_init,
1609 		.cra_exit		= omap_sham_cra_exit,
1610 	}
1611 },
1612 {
1613 	.init		= omap_sham_init,
1614 	.update		= omap_sham_update,
1615 	.final		= omap_sham_final,
1616 	.finup		= omap_sham_finup,
1617 	.digest		= omap_sham_digest,
1618 	.setkey		= omap_sham_setkey,
1619 	.halg.digestsize	= SHA256_DIGEST_SIZE,
1620 	.halg.base	= {
1621 		.cra_name		= "hmac(sha256)",
1622 		.cra_driver_name	= "omap-hmac-sha256",
1623 		.cra_priority		= 400,
1624 		.cra_flags		= CRYPTO_ALG_ASYNC |
1625 						CRYPTO_ALG_NEED_FALLBACK,
1626 		.cra_blocksize		= SHA256_BLOCK_SIZE,
1627 		.cra_ctxsize		= sizeof(struct omap_sham_ctx) +
1628 					sizeof(struct omap_sham_hmac_ctx),
1629 		.cra_alignmask		= OMAP_ALIGN_MASK,
1630 		.cra_module		= THIS_MODULE,
1631 		.cra_init		= omap_sham_cra_sha256_init,
1632 		.cra_exit		= omap_sham_cra_exit,
1633 	}
1634 },
1635 };
1636 
1637 static struct ahash_alg algs_sha384_sha512[] = {
1638 {
1639 	.init		= omap_sham_init,
1640 	.update		= omap_sham_update,
1641 	.final		= omap_sham_final,
1642 	.finup		= omap_sham_finup,
1643 	.digest		= omap_sham_digest,
1644 	.halg.digestsize	= SHA384_DIGEST_SIZE,
1645 	.halg.base	= {
1646 		.cra_name		= "sha384",
1647 		.cra_driver_name	= "omap-sha384",
1648 		.cra_priority		= 400,
1649 		.cra_flags		= CRYPTO_ALG_ASYNC |
1650 						CRYPTO_ALG_NEED_FALLBACK,
1651 		.cra_blocksize		= SHA384_BLOCK_SIZE,
1652 		.cra_ctxsize		= sizeof(struct omap_sham_ctx),
1653 		.cra_alignmask		= OMAP_ALIGN_MASK,
1654 		.cra_module		= THIS_MODULE,
1655 		.cra_init		= omap_sham_cra_init,
1656 		.cra_exit		= omap_sham_cra_exit,
1657 	}
1658 },
1659 {
1660 	.init		= omap_sham_init,
1661 	.update		= omap_sham_update,
1662 	.final		= omap_sham_final,
1663 	.finup		= omap_sham_finup,
1664 	.digest		= omap_sham_digest,
1665 	.halg.digestsize	= SHA512_DIGEST_SIZE,
1666 	.halg.base	= {
1667 		.cra_name		= "sha512",
1668 		.cra_driver_name	= "omap-sha512",
1669 		.cra_priority		= 400,
1670 		.cra_flags		= CRYPTO_ALG_ASYNC |
1671 						CRYPTO_ALG_NEED_FALLBACK,
1672 		.cra_blocksize		= SHA512_BLOCK_SIZE,
1673 		.cra_ctxsize		= sizeof(struct omap_sham_ctx),
1674 		.cra_alignmask		= OMAP_ALIGN_MASK,
1675 		.cra_module		= THIS_MODULE,
1676 		.cra_init		= omap_sham_cra_init,
1677 		.cra_exit		= omap_sham_cra_exit,
1678 	}
1679 },
1680 {
1681 	.init		= omap_sham_init,
1682 	.update		= omap_sham_update,
1683 	.final		= omap_sham_final,
1684 	.finup		= omap_sham_finup,
1685 	.digest		= omap_sham_digest,
1686 	.setkey		= omap_sham_setkey,
1687 	.halg.digestsize	= SHA384_DIGEST_SIZE,
1688 	.halg.base	= {
1689 		.cra_name		= "hmac(sha384)",
1690 		.cra_driver_name	= "omap-hmac-sha384",
1691 		.cra_priority		= 400,
1692 		.cra_flags		= CRYPTO_ALG_ASYNC |
1693 						CRYPTO_ALG_NEED_FALLBACK,
1694 		.cra_blocksize		= SHA384_BLOCK_SIZE,
1695 		.cra_ctxsize		= sizeof(struct omap_sham_ctx) +
1696 					sizeof(struct omap_sham_hmac_ctx),
1697 		.cra_alignmask		= OMAP_ALIGN_MASK,
1698 		.cra_module		= THIS_MODULE,
1699 		.cra_init		= omap_sham_cra_sha384_init,
1700 		.cra_exit		= omap_sham_cra_exit,
1701 	}
1702 },
1703 {
1704 	.init		= omap_sham_init,
1705 	.update		= omap_sham_update,
1706 	.final		= omap_sham_final,
1707 	.finup		= omap_sham_finup,
1708 	.digest		= omap_sham_digest,
1709 	.setkey		= omap_sham_setkey,
1710 	.halg.digestsize	= SHA512_DIGEST_SIZE,
1711 	.halg.base	= {
1712 		.cra_name		= "hmac(sha512)",
1713 		.cra_driver_name	= "omap-hmac-sha512",
1714 		.cra_priority		= 400,
1715 		.cra_flags		= CRYPTO_ALG_ASYNC |
1716 						CRYPTO_ALG_NEED_FALLBACK,
1717 		.cra_blocksize		= SHA512_BLOCK_SIZE,
1718 		.cra_ctxsize		= sizeof(struct omap_sham_ctx) +
1719 					sizeof(struct omap_sham_hmac_ctx),
1720 		.cra_alignmask		= OMAP_ALIGN_MASK,
1721 		.cra_module		= THIS_MODULE,
1722 		.cra_init		= omap_sham_cra_sha512_init,
1723 		.cra_exit		= omap_sham_cra_exit,
1724 	}
1725 },
1726 };
1727 
1728 static void omap_sham_done_task(unsigned long data)
1729 {
1730 	struct omap_sham_dev *dd = (struct omap_sham_dev *)data;
1731 	int err = 0;
1732 
1733 	if (!test_bit(FLAGS_BUSY, &dd->flags)) {
1734 		omap_sham_handle_queue(dd, NULL);
1735 		return;
1736 	}
1737 
1738 	if (test_bit(FLAGS_CPU, &dd->flags)) {
1739 		if (test_and_clear_bit(FLAGS_OUTPUT_READY, &dd->flags))
1740 			goto finish;
1741 	} else if (test_bit(FLAGS_DMA_READY, &dd->flags)) {
1742 		if (test_and_clear_bit(FLAGS_DMA_ACTIVE, &dd->flags)) {
1743 			omap_sham_update_dma_stop(dd);
1744 			if (dd->err) {
1745 				err = dd->err;
1746 				goto finish;
1747 			}
1748 		}
1749 		if (test_and_clear_bit(FLAGS_OUTPUT_READY, &dd->flags)) {
1750 			/* hash or semi-hash ready */
1751 			clear_bit(FLAGS_DMA_READY, &dd->flags);
1752 			goto finish;
1753 		}
1754 	}
1755 
1756 	return;
1757 
1758 finish:
1759 	dev_dbg(dd->dev, "update done: err: %d\n", err);
1760 	/* finish curent request */
1761 	omap_sham_finish_req(dd->req, err);
1762 
1763 	/* If we are not busy, process next req */
1764 	if (!test_bit(FLAGS_BUSY, &dd->flags))
1765 		omap_sham_handle_queue(dd, NULL);
1766 }
1767 
1768 static irqreturn_t omap_sham_irq_common(struct omap_sham_dev *dd)
1769 {
1770 	if (!test_bit(FLAGS_BUSY, &dd->flags)) {
1771 		dev_warn(dd->dev, "Interrupt when no active requests.\n");
1772 	} else {
1773 		set_bit(FLAGS_OUTPUT_READY, &dd->flags);
1774 		tasklet_schedule(&dd->done_task);
1775 	}
1776 
1777 	return IRQ_HANDLED;
1778 }
1779 
1780 static irqreturn_t omap_sham_irq_omap2(int irq, void *dev_id)
1781 {
1782 	struct omap_sham_dev *dd = dev_id;
1783 
1784 	if (unlikely(test_bit(FLAGS_FINAL, &dd->flags)))
1785 		/* final -> allow device to go to power-saving mode */
1786 		omap_sham_write_mask(dd, SHA_REG_CTRL, 0, SHA_REG_CTRL_LENGTH);
1787 
1788 	omap_sham_write_mask(dd, SHA_REG_CTRL, SHA_REG_CTRL_OUTPUT_READY,
1789 				 SHA_REG_CTRL_OUTPUT_READY);
1790 	omap_sham_read(dd, SHA_REG_CTRL);
1791 
1792 	return omap_sham_irq_common(dd);
1793 }
1794 
1795 static irqreturn_t omap_sham_irq_omap4(int irq, void *dev_id)
1796 {
1797 	struct omap_sham_dev *dd = dev_id;
1798 
1799 	omap_sham_write_mask(dd, SHA_REG_MASK(dd), 0, SHA_REG_MASK_IT_EN);
1800 
1801 	return omap_sham_irq_common(dd);
1802 }
1803 
1804 static struct omap_sham_algs_info omap_sham_algs_info_omap2[] = {
1805 	{
1806 		.algs_list	= algs_sha1_md5,
1807 		.size		= ARRAY_SIZE(algs_sha1_md5),
1808 	},
1809 };
1810 
1811 static const struct omap_sham_pdata omap_sham_pdata_omap2 = {
1812 	.algs_info	= omap_sham_algs_info_omap2,
1813 	.algs_info_size	= ARRAY_SIZE(omap_sham_algs_info_omap2),
1814 	.flags		= BIT(FLAGS_BE32_SHA1),
1815 	.digest_size	= SHA1_DIGEST_SIZE,
1816 	.copy_hash	= omap_sham_copy_hash_omap2,
1817 	.write_ctrl	= omap_sham_write_ctrl_omap2,
1818 	.trigger	= omap_sham_trigger_omap2,
1819 	.poll_irq	= omap_sham_poll_irq_omap2,
1820 	.intr_hdlr	= omap_sham_irq_omap2,
1821 	.idigest_ofs	= 0x00,
1822 	.din_ofs	= 0x1c,
1823 	.digcnt_ofs	= 0x14,
1824 	.rev_ofs	= 0x5c,
1825 	.mask_ofs	= 0x60,
1826 	.sysstatus_ofs	= 0x64,
1827 	.major_mask	= 0xf0,
1828 	.major_shift	= 4,
1829 	.minor_mask	= 0x0f,
1830 	.minor_shift	= 0,
1831 };
1832 
1833 #ifdef CONFIG_OF
1834 static struct omap_sham_algs_info omap_sham_algs_info_omap4[] = {
1835 	{
1836 		.algs_list	= algs_sha1_md5,
1837 		.size		= ARRAY_SIZE(algs_sha1_md5),
1838 	},
1839 	{
1840 		.algs_list	= algs_sha224_sha256,
1841 		.size		= ARRAY_SIZE(algs_sha224_sha256),
1842 	},
1843 };
1844 
1845 static const struct omap_sham_pdata omap_sham_pdata_omap4 = {
1846 	.algs_info	= omap_sham_algs_info_omap4,
1847 	.algs_info_size	= ARRAY_SIZE(omap_sham_algs_info_omap4),
1848 	.flags		= BIT(FLAGS_AUTO_XOR),
1849 	.digest_size	= SHA256_DIGEST_SIZE,
1850 	.copy_hash	= omap_sham_copy_hash_omap4,
1851 	.write_ctrl	= omap_sham_write_ctrl_omap4,
1852 	.trigger	= omap_sham_trigger_omap4,
1853 	.poll_irq	= omap_sham_poll_irq_omap4,
1854 	.intr_hdlr	= omap_sham_irq_omap4,
1855 	.idigest_ofs	= 0x020,
1856 	.odigest_ofs	= 0x0,
1857 	.din_ofs	= 0x080,
1858 	.digcnt_ofs	= 0x040,
1859 	.rev_ofs	= 0x100,
1860 	.mask_ofs	= 0x110,
1861 	.sysstatus_ofs	= 0x114,
1862 	.mode_ofs	= 0x44,
1863 	.length_ofs	= 0x48,
1864 	.major_mask	= 0x0700,
1865 	.major_shift	= 8,
1866 	.minor_mask	= 0x003f,
1867 	.minor_shift	= 0,
1868 };
1869 
1870 static struct omap_sham_algs_info omap_sham_algs_info_omap5[] = {
1871 	{
1872 		.algs_list	= algs_sha1_md5,
1873 		.size		= ARRAY_SIZE(algs_sha1_md5),
1874 	},
1875 	{
1876 		.algs_list	= algs_sha224_sha256,
1877 		.size		= ARRAY_SIZE(algs_sha224_sha256),
1878 	},
1879 	{
1880 		.algs_list	= algs_sha384_sha512,
1881 		.size		= ARRAY_SIZE(algs_sha384_sha512),
1882 	},
1883 };
1884 
1885 static const struct omap_sham_pdata omap_sham_pdata_omap5 = {
1886 	.algs_info	= omap_sham_algs_info_omap5,
1887 	.algs_info_size	= ARRAY_SIZE(omap_sham_algs_info_omap5),
1888 	.flags		= BIT(FLAGS_AUTO_XOR),
1889 	.digest_size	= SHA512_DIGEST_SIZE,
1890 	.copy_hash	= omap_sham_copy_hash_omap4,
1891 	.write_ctrl	= omap_sham_write_ctrl_omap4,
1892 	.trigger	= omap_sham_trigger_omap4,
1893 	.poll_irq	= omap_sham_poll_irq_omap4,
1894 	.intr_hdlr	= omap_sham_irq_omap4,
1895 	.idigest_ofs	= 0x240,
1896 	.odigest_ofs	= 0x200,
1897 	.din_ofs	= 0x080,
1898 	.digcnt_ofs	= 0x280,
1899 	.rev_ofs	= 0x100,
1900 	.mask_ofs	= 0x110,
1901 	.sysstatus_ofs	= 0x114,
1902 	.mode_ofs	= 0x284,
1903 	.length_ofs	= 0x288,
1904 	.major_mask	= 0x0700,
1905 	.major_shift	= 8,
1906 	.minor_mask	= 0x003f,
1907 	.minor_shift	= 0,
1908 };
1909 
1910 static const struct of_device_id omap_sham_of_match[] = {
1911 	{
1912 		.compatible	= "ti,omap2-sham",
1913 		.data		= &omap_sham_pdata_omap2,
1914 	},
1915 	{
1916 		.compatible	= "ti,omap3-sham",
1917 		.data		= &omap_sham_pdata_omap2,
1918 	},
1919 	{
1920 		.compatible	= "ti,omap4-sham",
1921 		.data		= &omap_sham_pdata_omap4,
1922 	},
1923 	{
1924 		.compatible	= "ti,omap5-sham",
1925 		.data		= &omap_sham_pdata_omap5,
1926 	},
1927 	{},
1928 };
1929 MODULE_DEVICE_TABLE(of, omap_sham_of_match);
1930 
1931 static int omap_sham_get_res_of(struct omap_sham_dev *dd,
1932 		struct device *dev, struct resource *res)
1933 {
1934 	struct device_node *node = dev->of_node;
1935 	int err = 0;
1936 
1937 	dd->pdata = of_device_get_match_data(dev);
1938 	if (!dd->pdata) {
1939 		dev_err(dev, "no compatible OF match\n");
1940 		err = -EINVAL;
1941 		goto err;
1942 	}
1943 
1944 	err = of_address_to_resource(node, 0, res);
1945 	if (err < 0) {
1946 		dev_err(dev, "can't translate OF node address\n");
1947 		err = -EINVAL;
1948 		goto err;
1949 	}
1950 
1951 	dd->irq = irq_of_parse_and_map(node, 0);
1952 	if (!dd->irq) {
1953 		dev_err(dev, "can't translate OF irq value\n");
1954 		err = -EINVAL;
1955 		goto err;
1956 	}
1957 
1958 err:
1959 	return err;
1960 }
1961 #else
1962 static const struct of_device_id omap_sham_of_match[] = {
1963 	{},
1964 };
1965 
1966 static int omap_sham_get_res_of(struct omap_sham_dev *dd,
1967 		struct device *dev, struct resource *res)
1968 {
1969 	return -EINVAL;
1970 }
1971 #endif
1972 
1973 static int omap_sham_get_res_pdev(struct omap_sham_dev *dd,
1974 		struct platform_device *pdev, struct resource *res)
1975 {
1976 	struct device *dev = &pdev->dev;
1977 	struct resource *r;
1978 	int err = 0;
1979 
1980 	/* Get the base address */
1981 	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
1982 	if (!r) {
1983 		dev_err(dev, "no MEM resource info\n");
1984 		err = -ENODEV;
1985 		goto err;
1986 	}
1987 	memcpy(res, r, sizeof(*res));
1988 
1989 	/* Get the IRQ */
1990 	dd->irq = platform_get_irq(pdev, 0);
1991 	if (dd->irq < 0) {
1992 		err = dd->irq;
1993 		goto err;
1994 	}
1995 
1996 	/* Only OMAP2/3 can be non-DT */
1997 	dd->pdata = &omap_sham_pdata_omap2;
1998 
1999 err:
2000 	return err;
2001 }
2002 
2003 static ssize_t fallback_show(struct device *dev, struct device_attribute *attr,
2004 			     char *buf)
2005 {
2006 	struct omap_sham_dev *dd = dev_get_drvdata(dev);
2007 
2008 	return sprintf(buf, "%d\n", dd->fallback_sz);
2009 }
2010 
2011 static ssize_t fallback_store(struct device *dev, struct device_attribute *attr,
2012 			      const char *buf, size_t size)
2013 {
2014 	struct omap_sham_dev *dd = dev_get_drvdata(dev);
2015 	ssize_t status;
2016 	long value;
2017 
2018 	status = kstrtol(buf, 0, &value);
2019 	if (status)
2020 		return status;
2021 
2022 	/* HW accelerator only works with buffers > 9 */
2023 	if (value < 9) {
2024 		dev_err(dev, "minimum fallback size 9\n");
2025 		return -EINVAL;
2026 	}
2027 
2028 	dd->fallback_sz = value;
2029 
2030 	return size;
2031 }
2032 
2033 static ssize_t queue_len_show(struct device *dev, struct device_attribute *attr,
2034 			      char *buf)
2035 {
2036 	struct omap_sham_dev *dd = dev_get_drvdata(dev);
2037 
2038 	return sprintf(buf, "%d\n", dd->queue.max_qlen);
2039 }
2040 
2041 static ssize_t queue_len_store(struct device *dev,
2042 			       struct device_attribute *attr, const char *buf,
2043 			       size_t size)
2044 {
2045 	struct omap_sham_dev *dd = dev_get_drvdata(dev);
2046 	ssize_t status;
2047 	long value;
2048 	unsigned long flags;
2049 
2050 	status = kstrtol(buf, 0, &value);
2051 	if (status)
2052 		return status;
2053 
2054 	if (value < 1)
2055 		return -EINVAL;
2056 
2057 	/*
2058 	 * Changing the queue size in fly is safe, if size becomes smaller
2059 	 * than current size, it will just not accept new entries until
2060 	 * it has shrank enough.
2061 	 */
2062 	spin_lock_irqsave(&dd->lock, flags);
2063 	dd->queue.max_qlen = value;
2064 	spin_unlock_irqrestore(&dd->lock, flags);
2065 
2066 	return size;
2067 }
2068 
2069 static DEVICE_ATTR_RW(queue_len);
2070 static DEVICE_ATTR_RW(fallback);
2071 
2072 static struct attribute *omap_sham_attrs[] = {
2073 	&dev_attr_queue_len.attr,
2074 	&dev_attr_fallback.attr,
2075 	NULL,
2076 };
2077 
2078 static struct attribute_group omap_sham_attr_group = {
2079 	.attrs = omap_sham_attrs,
2080 };
2081 
2082 static int omap_sham_probe(struct platform_device *pdev)
2083 {
2084 	struct omap_sham_dev *dd;
2085 	struct device *dev = &pdev->dev;
2086 	struct resource res;
2087 	dma_cap_mask_t mask;
2088 	int err, i, j;
2089 	u32 rev;
2090 
2091 	dd = devm_kzalloc(dev, sizeof(struct omap_sham_dev), GFP_KERNEL);
2092 	if (dd == NULL) {
2093 		dev_err(dev, "unable to alloc data struct.\n");
2094 		err = -ENOMEM;
2095 		goto data_err;
2096 	}
2097 	dd->dev = dev;
2098 	platform_set_drvdata(pdev, dd);
2099 
2100 	INIT_LIST_HEAD(&dd->list);
2101 	spin_lock_init(&dd->lock);
2102 	tasklet_init(&dd->done_task, omap_sham_done_task, (unsigned long)dd);
2103 	crypto_init_queue(&dd->queue, OMAP_SHAM_QUEUE_LENGTH);
2104 
2105 	err = (dev->of_node) ? omap_sham_get_res_of(dd, dev, &res) :
2106 			       omap_sham_get_res_pdev(dd, pdev, &res);
2107 	if (err)
2108 		goto data_err;
2109 
2110 	dd->io_base = devm_ioremap_resource(dev, &res);
2111 	if (IS_ERR(dd->io_base)) {
2112 		err = PTR_ERR(dd->io_base);
2113 		goto data_err;
2114 	}
2115 	dd->phys_base = res.start;
2116 
2117 	err = devm_request_irq(dev, dd->irq, dd->pdata->intr_hdlr,
2118 			       IRQF_TRIGGER_NONE, dev_name(dev), dd);
2119 	if (err) {
2120 		dev_err(dev, "unable to request irq %d, err = %d\n",
2121 			dd->irq, err);
2122 		goto data_err;
2123 	}
2124 
2125 	dma_cap_zero(mask);
2126 	dma_cap_set(DMA_SLAVE, mask);
2127 
2128 	dd->dma_lch = dma_request_chan(dev, "rx");
2129 	if (IS_ERR(dd->dma_lch)) {
2130 		err = PTR_ERR(dd->dma_lch);
2131 		if (err == -EPROBE_DEFER)
2132 			goto data_err;
2133 
2134 		dd->polling_mode = 1;
2135 		dev_dbg(dev, "using polling mode instead of dma\n");
2136 	}
2137 
2138 	dd->flags |= dd->pdata->flags;
2139 
2140 	pm_runtime_use_autosuspend(dev);
2141 	pm_runtime_set_autosuspend_delay(dev, DEFAULT_AUTOSUSPEND_DELAY);
2142 
2143 	dd->fallback_sz = OMAP_SHA_DMA_THRESHOLD;
2144 
2145 	pm_runtime_enable(dev);
2146 	pm_runtime_irq_safe(dev);
2147 
2148 	err = pm_runtime_get_sync(dev);
2149 	if (err < 0) {
2150 		dev_err(dev, "failed to get sync: %d\n", err);
2151 		goto err_pm;
2152 	}
2153 
2154 	rev = omap_sham_read(dd, SHA_REG_REV(dd));
2155 	pm_runtime_put_sync(&pdev->dev);
2156 
2157 	dev_info(dev, "hw accel on OMAP rev %u.%u\n",
2158 		(rev & dd->pdata->major_mask) >> dd->pdata->major_shift,
2159 		(rev & dd->pdata->minor_mask) >> dd->pdata->minor_shift);
2160 
2161 	spin_lock(&sham.lock);
2162 	list_add_tail(&dd->list, &sham.dev_list);
2163 	spin_unlock(&sham.lock);
2164 
2165 	for (i = 0; i < dd->pdata->algs_info_size; i++) {
2166 		for (j = 0; j < dd->pdata->algs_info[i].size; j++) {
2167 			struct ahash_alg *alg;
2168 
2169 			alg = &dd->pdata->algs_info[i].algs_list[j];
2170 			alg->export = omap_sham_export;
2171 			alg->import = omap_sham_import;
2172 			alg->halg.statesize = sizeof(struct omap_sham_reqctx) +
2173 					      BUFLEN;
2174 			err = crypto_register_ahash(alg);
2175 			if (err)
2176 				goto err_algs;
2177 
2178 			dd->pdata->algs_info[i].registered++;
2179 		}
2180 	}
2181 
2182 	err = sysfs_create_group(&dev->kobj, &omap_sham_attr_group);
2183 	if (err) {
2184 		dev_err(dev, "could not create sysfs device attrs\n");
2185 		goto err_algs;
2186 	}
2187 
2188 	return 0;
2189 
2190 err_algs:
2191 	for (i = dd->pdata->algs_info_size - 1; i >= 0; i--)
2192 		for (j = dd->pdata->algs_info[i].registered - 1; j >= 0; j--)
2193 			crypto_unregister_ahash(
2194 					&dd->pdata->algs_info[i].algs_list[j]);
2195 err_pm:
2196 	pm_runtime_disable(dev);
2197 	if (!dd->polling_mode)
2198 		dma_release_channel(dd->dma_lch);
2199 data_err:
2200 	dev_err(dev, "initialization failed.\n");
2201 
2202 	return err;
2203 }
2204 
2205 static int omap_sham_remove(struct platform_device *pdev)
2206 {
2207 	struct omap_sham_dev *dd;
2208 	int i, j;
2209 
2210 	dd = platform_get_drvdata(pdev);
2211 	if (!dd)
2212 		return -ENODEV;
2213 	spin_lock(&sham.lock);
2214 	list_del(&dd->list);
2215 	spin_unlock(&sham.lock);
2216 	for (i = dd->pdata->algs_info_size - 1; i >= 0; i--)
2217 		for (j = dd->pdata->algs_info[i].registered - 1; j >= 0; j--)
2218 			crypto_unregister_ahash(
2219 					&dd->pdata->algs_info[i].algs_list[j]);
2220 	tasklet_kill(&dd->done_task);
2221 	pm_runtime_disable(&pdev->dev);
2222 
2223 	if (!dd->polling_mode)
2224 		dma_release_channel(dd->dma_lch);
2225 
2226 	return 0;
2227 }
2228 
2229 #ifdef CONFIG_PM_SLEEP
2230 static int omap_sham_suspend(struct device *dev)
2231 {
2232 	pm_runtime_put_sync(dev);
2233 	return 0;
2234 }
2235 
2236 static int omap_sham_resume(struct device *dev)
2237 {
2238 	int err = pm_runtime_get_sync(dev);
2239 	if (err < 0) {
2240 		dev_err(dev, "failed to get sync: %d\n", err);
2241 		return err;
2242 	}
2243 	return 0;
2244 }
2245 #endif
2246 
2247 static SIMPLE_DEV_PM_OPS(omap_sham_pm_ops, omap_sham_suspend, omap_sham_resume);
2248 
2249 static struct platform_driver omap_sham_driver = {
2250 	.probe	= omap_sham_probe,
2251 	.remove	= omap_sham_remove,
2252 	.driver	= {
2253 		.name	= "omap-sham",
2254 		.pm	= &omap_sham_pm_ops,
2255 		.of_match_table	= omap_sham_of_match,
2256 	},
2257 };
2258 
2259 module_platform_driver(omap_sham_driver);
2260 
2261 MODULE_DESCRIPTION("OMAP SHA1/MD5 hw acceleration support.");
2262 MODULE_LICENSE("GPL v2");
2263 MODULE_AUTHOR("Dmitry Kasatkin");
2264 MODULE_ALIAS("platform:omap-sham");
2265