1/*
2 * Multi buffer SHA1 algorithm Glue Code
3 *
4 * This file is provided under a dual BSD/GPLv2 license.  When using or
5 * redistributing this file, you may do so under either license.
6 *
7 * GPL LICENSE SUMMARY
8 *
9 *  Copyright(c) 2014 Intel Corporation.
10 *
11 *  This program is free software; you can redistribute it and/or modify
12 *  it under the terms of version 2 of the GNU General Public License as
13 *  published by the Free Software Foundation.
14 *
15 *  This program is distributed in the hope that it will be useful, but
16 *  WITHOUT ANY WARRANTY; without even the implied warranty of
17 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
18 *  General Public License for more details.
19 *
20 *  Contact Information:
21 *	Tim Chen <tim.c.chen@linux.intel.com>
22 *
23 *  BSD LICENSE
24 *
25 *  Copyright(c) 2014 Intel Corporation.
26 *
27 *  Redistribution and use in source and binary forms, with or without
28 *  modification, are permitted provided that the following conditions
29 *  are met:
30 *
31 *    * Redistributions of source code must retain the above copyright
32 *      notice, this list of conditions and the following disclaimer.
33 *    * Redistributions in binary form must reproduce the above copyright
34 *      notice, this list of conditions and the following disclaimer in
35 *      the documentation and/or other materials provided with the
36 *      distribution.
37 *    * Neither the name of Intel Corporation nor the names of its
38 *      contributors may be used to endorse or promote products derived
39 *      from this software without specific prior written permission.
40 *
41 *  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
42 *  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
43 *  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
44 *  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
45 *  OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
46 *  SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
47 *  LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
48 *  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
49 *  THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
50 *  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
51 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
52 */
53
54#define pr_fmt(fmt)	KBUILD_MODNAME ": " fmt
55
56#include <crypto/internal/hash.h>
57#include <linux/init.h>
58#include <linux/module.h>
59#include <linux/mm.h>
60#include <linux/cryptohash.h>
61#include <linux/types.h>
62#include <linux/list.h>
63#include <crypto/scatterwalk.h>
64#include <crypto/sha.h>
65#include <crypto/mcryptd.h>
66#include <crypto/crypto_wq.h>
67#include <asm/byteorder.h>
68#include <asm/i387.h>
69#include <asm/xcr.h>
70#include <asm/xsave.h>
71#include <linux/hardirq.h>
72#include <asm/fpu-internal.h>
73#include "sha_mb_ctx.h"
74
75#define FLUSH_INTERVAL 1000 /* in usec */
76
77static struct mcryptd_alg_state sha1_mb_alg_state;
78
79struct sha1_mb_ctx {
80	struct mcryptd_ahash *mcryptd_tfm;
81};
82
83static inline struct mcryptd_hash_request_ctx *cast_hash_to_mcryptd_ctx(struct sha1_hash_ctx *hash_ctx)
84{
85	struct shash_desc *desc;
86
87	desc = container_of((void *) hash_ctx, struct shash_desc, __ctx);
88	return container_of(desc, struct mcryptd_hash_request_ctx, desc);
89}
90
91static inline struct ahash_request *cast_mcryptd_ctx_to_req(struct mcryptd_hash_request_ctx *ctx)
92{
93	return container_of((void *) ctx, struct ahash_request, __ctx);
94}
95
96static void req_ctx_init(struct mcryptd_hash_request_ctx *rctx,
97				struct shash_desc *desc)
98{
99	rctx->flag = HASH_UPDATE;
100}
101
102static asmlinkage void (*sha1_job_mgr_init)(struct sha1_mb_mgr *state);
103static asmlinkage struct job_sha1* (*sha1_job_mgr_submit)(struct sha1_mb_mgr *state,
104							  struct job_sha1 *job);
105static asmlinkage struct job_sha1* (*sha1_job_mgr_flush)(struct sha1_mb_mgr *state);
106static asmlinkage struct job_sha1* (*sha1_job_mgr_get_comp_job)(struct sha1_mb_mgr *state);
107
108inline void sha1_init_digest(uint32_t *digest)
109{
110	static const uint32_t initial_digest[SHA1_DIGEST_LENGTH] = {SHA1_H0,
111					SHA1_H1, SHA1_H2, SHA1_H3, SHA1_H4 };
112	memcpy(digest, initial_digest, sizeof(initial_digest));
113}
114
115inline uint32_t sha1_pad(uint8_t padblock[SHA1_BLOCK_SIZE * 2],
116			 uint32_t total_len)
117{
118	uint32_t i = total_len & (SHA1_BLOCK_SIZE - 1);
119
120	memset(&padblock[i], 0, SHA1_BLOCK_SIZE);
121	padblock[i] = 0x80;
122
123	i += ((SHA1_BLOCK_SIZE - 1) &
124	      (0 - (total_len + SHA1_PADLENGTHFIELD_SIZE + 1)))
125	     + 1 + SHA1_PADLENGTHFIELD_SIZE;
126
127#if SHA1_PADLENGTHFIELD_SIZE == 16
128	*((uint64_t *) &padblock[i - 16]) = 0;
129#endif
130
131	*((uint64_t *) &padblock[i - 8]) = cpu_to_be64(total_len << 3);
132
133	/* Number of extra blocks to hash */
134	return i >> SHA1_LOG2_BLOCK_SIZE;
135}
136
137static struct sha1_hash_ctx *sha1_ctx_mgr_resubmit(struct sha1_ctx_mgr *mgr, struct sha1_hash_ctx *ctx)
138{
139	while (ctx) {
140		if (ctx->status & HASH_CTX_STS_COMPLETE) {
141			/* Clear PROCESSING bit */
142			ctx->status = HASH_CTX_STS_COMPLETE;
143			return ctx;
144		}
145
146		/*
147		 * If the extra blocks are empty, begin hashing what remains
148		 * in the user's buffer.
149		 */
150		if (ctx->partial_block_buffer_length == 0 &&
151		    ctx->incoming_buffer_length) {
152
153			const void *buffer = ctx->incoming_buffer;
154			uint32_t len = ctx->incoming_buffer_length;
155			uint32_t copy_len;
156
157			/*
158			 * Only entire blocks can be hashed.
159			 * Copy remainder to extra blocks buffer.
160			 */
161			copy_len = len & (SHA1_BLOCK_SIZE-1);
162
163			if (copy_len) {
164				len -= copy_len;
165				memcpy(ctx->partial_block_buffer,
166				       ((const char *) buffer + len),
167				       copy_len);
168				ctx->partial_block_buffer_length = copy_len;
169			}
170
171			ctx->incoming_buffer_length = 0;
172
173			/* len should be a multiple of the block size now */
174			assert((len % SHA1_BLOCK_SIZE) == 0);
175
176			/* Set len to the number of blocks to be hashed */
177			len >>= SHA1_LOG2_BLOCK_SIZE;
178
179			if (len) {
180
181				ctx->job.buffer = (uint8_t *) buffer;
182				ctx->job.len = len;
183				ctx = (struct sha1_hash_ctx *) sha1_job_mgr_submit(&mgr->mgr,
184										  &ctx->job);
185				continue;
186			}
187		}
188
189		/*
190		 * If the extra blocks are not empty, then we are
191		 * either on the last block(s) or we need more
192		 * user input before continuing.
193		 */
194		if (ctx->status & HASH_CTX_STS_LAST) {
195
196			uint8_t *buf = ctx->partial_block_buffer;
197			uint32_t n_extra_blocks = sha1_pad(buf, ctx->total_length);
198
199			ctx->status = (HASH_CTX_STS_PROCESSING |
200				       HASH_CTX_STS_COMPLETE);
201			ctx->job.buffer = buf;
202			ctx->job.len = (uint32_t) n_extra_blocks;
203			ctx = (struct sha1_hash_ctx *) sha1_job_mgr_submit(&mgr->mgr, &ctx->job);
204			continue;
205		}
206
207		ctx->status = HASH_CTX_STS_IDLE;
208		return ctx;
209	}
210
211	return NULL;
212}
213
214static struct sha1_hash_ctx *sha1_ctx_mgr_get_comp_ctx(struct sha1_ctx_mgr *mgr)
215{
216	/*
217	 * If get_comp_job returns NULL, there are no jobs complete.
218	 * If get_comp_job returns a job, verify that it is safe to return to the user.
219	 * If it is not ready, resubmit the job to finish processing.
220	 * If sha1_ctx_mgr_resubmit returned a job, it is ready to be returned.
221	 * Otherwise, all jobs currently being managed by the hash_ctx_mgr still need processing.
222	 */
223	struct sha1_hash_ctx *ctx;
224
225	ctx = (struct sha1_hash_ctx *) sha1_job_mgr_get_comp_job(&mgr->mgr);
226	return sha1_ctx_mgr_resubmit(mgr, ctx);
227}
228
229static void sha1_ctx_mgr_init(struct sha1_ctx_mgr *mgr)
230{
231	sha1_job_mgr_init(&mgr->mgr);
232}
233
234static struct sha1_hash_ctx *sha1_ctx_mgr_submit(struct sha1_ctx_mgr *mgr,
235					  struct sha1_hash_ctx *ctx,
236					  const void *buffer,
237					  uint32_t len,
238					  int flags)
239{
240	if (flags & (~HASH_ENTIRE)) {
241		/* User should not pass anything other than FIRST, UPDATE, or LAST */
242		ctx->error = HASH_CTX_ERROR_INVALID_FLAGS;
243		return ctx;
244	}
245
246	if (ctx->status & HASH_CTX_STS_PROCESSING) {
247		/* Cannot submit to a currently processing job. */
248		ctx->error = HASH_CTX_ERROR_ALREADY_PROCESSING;
249		return ctx;
250	}
251
252	if ((ctx->status & HASH_CTX_STS_COMPLETE) && !(flags & HASH_FIRST)) {
253		/* Cannot update a finished job. */
254		ctx->error = HASH_CTX_ERROR_ALREADY_COMPLETED;
255		return ctx;
256	}
257
258
259	if (flags & HASH_FIRST) {
260		/* Init digest */
261		sha1_init_digest(ctx->job.result_digest);
262
263		/* Reset byte counter */
264		ctx->total_length = 0;
265
266		/* Clear extra blocks */
267		ctx->partial_block_buffer_length = 0;
268	}
269
270	/* If we made it here, there were no errors during this call to submit */
271	ctx->error = HASH_CTX_ERROR_NONE;
272
273	/* Store buffer ptr info from user */
274	ctx->incoming_buffer = buffer;
275	ctx->incoming_buffer_length = len;
276
277	/* Store the user's request flags and mark this ctx as currently being processed. */
278	ctx->status = (flags & HASH_LAST) ?
279			(HASH_CTX_STS_PROCESSING | HASH_CTX_STS_LAST) :
280			HASH_CTX_STS_PROCESSING;
281
282	/* Advance byte counter */
283	ctx->total_length += len;
284
285	/*
286	 * If there is anything currently buffered in the extra blocks,
287	 * append to it until it contains a whole block.
288	 * Or if the user's buffer contains less than a whole block,
289	 * append as much as possible to the extra block.
290	 */
291	if ((ctx->partial_block_buffer_length) | (len < SHA1_BLOCK_SIZE)) {
292		/* Compute how many bytes to copy from user buffer into extra block */
293		uint32_t copy_len = SHA1_BLOCK_SIZE - ctx->partial_block_buffer_length;
294		if (len < copy_len)
295			copy_len = len;
296
297		if (copy_len) {
298			/* Copy and update relevant pointers and counters */
299			memcpy(&ctx->partial_block_buffer[ctx->partial_block_buffer_length],
300				buffer, copy_len);
301
302			ctx->partial_block_buffer_length += copy_len;
303			ctx->incoming_buffer = (const void *)((const char *)buffer + copy_len);
304			ctx->incoming_buffer_length = len - copy_len;
305		}
306
307		/* The extra block should never contain more than 1 block here */
308		assert(ctx->partial_block_buffer_length <= SHA1_BLOCK_SIZE);
309
310		/* If the extra block buffer contains exactly 1 block, it can be hashed. */
311		if (ctx->partial_block_buffer_length >= SHA1_BLOCK_SIZE) {
312			ctx->partial_block_buffer_length = 0;
313
314			ctx->job.buffer = ctx->partial_block_buffer;
315			ctx->job.len = 1;
316			ctx = (struct sha1_hash_ctx *) sha1_job_mgr_submit(&mgr->mgr, &ctx->job);
317		}
318	}
319
320	return sha1_ctx_mgr_resubmit(mgr, ctx);
321}
322
323static struct sha1_hash_ctx *sha1_ctx_mgr_flush(struct sha1_ctx_mgr *mgr)
324{
325	struct sha1_hash_ctx *ctx;
326
327	while (1) {
328		ctx = (struct sha1_hash_ctx *) sha1_job_mgr_flush(&mgr->mgr);
329
330		/* If flush returned 0, there are no more jobs in flight. */
331		if (!ctx)
332			return NULL;
333
334		/*
335		 * If flush returned a job, resubmit the job to finish processing.
336		 */
337		ctx = sha1_ctx_mgr_resubmit(mgr, ctx);
338
339		/*
340		 * If sha1_ctx_mgr_resubmit returned a job, it is ready to be returned.
341		 * Otherwise, all jobs currently being managed by the sha1_ctx_mgr
342		 * still need processing. Loop.
343		 */
344		if (ctx)
345			return ctx;
346	}
347}
348
349static int sha1_mb_init(struct shash_desc *desc)
350{
351	struct sha1_hash_ctx *sctx = shash_desc_ctx(desc);
352
353	hash_ctx_init(sctx);
354	sctx->job.result_digest[0] = SHA1_H0;
355	sctx->job.result_digest[1] = SHA1_H1;
356	sctx->job.result_digest[2] = SHA1_H2;
357	sctx->job.result_digest[3] = SHA1_H3;
358	sctx->job.result_digest[4] = SHA1_H4;
359	sctx->total_length = 0;
360	sctx->partial_block_buffer_length = 0;
361	sctx->status = HASH_CTX_STS_IDLE;
362
363	return 0;
364}
365
366static int sha1_mb_set_results(struct mcryptd_hash_request_ctx *rctx)
367{
368	int	i;
369	struct	sha1_hash_ctx *sctx = shash_desc_ctx(&rctx->desc);
370	__be32	*dst = (__be32 *) rctx->out;
371
372	for (i = 0; i < 5; ++i)
373		dst[i] = cpu_to_be32(sctx->job.result_digest[i]);
374
375	return 0;
376}
377
378static int sha_finish_walk(struct mcryptd_hash_request_ctx **ret_rctx,
379			struct mcryptd_alg_cstate *cstate, bool flush)
380{
381	int	flag = HASH_UPDATE;
382	int	nbytes, err = 0;
383	struct mcryptd_hash_request_ctx *rctx = *ret_rctx;
384	struct sha1_hash_ctx *sha_ctx;
385
386	/* more work ? */
387	while (!(rctx->flag & HASH_DONE)) {
388		nbytes = crypto_ahash_walk_done(&rctx->walk, 0);
389		if (nbytes < 0) {
390			err = nbytes;
391			goto out;
392		}
393		/* check if the walk is done */
394		if (crypto_ahash_walk_last(&rctx->walk)) {
395			rctx->flag |= HASH_DONE;
396			if (rctx->flag & HASH_FINAL)
397				flag |= HASH_LAST;
398
399		}
400		sha_ctx = (struct sha1_hash_ctx *) shash_desc_ctx(&rctx->desc);
401		kernel_fpu_begin();
402		sha_ctx = sha1_ctx_mgr_submit(cstate->mgr, sha_ctx, rctx->walk.data, nbytes, flag);
403		if (!sha_ctx) {
404			if (flush)
405				sha_ctx = sha1_ctx_mgr_flush(cstate->mgr);
406		}
407		kernel_fpu_end();
408		if (sha_ctx)
409			rctx = cast_hash_to_mcryptd_ctx(sha_ctx);
410		else {
411			rctx = NULL;
412			goto out;
413		}
414	}
415
416	/* copy the results */
417	if (rctx->flag & HASH_FINAL)
418		sha1_mb_set_results(rctx);
419
420out:
421	*ret_rctx = rctx;
422	return err;
423}
424
425static int sha_complete_job(struct mcryptd_hash_request_ctx *rctx,
426			    struct mcryptd_alg_cstate *cstate,
427			    int err)
428{
429	struct ahash_request *req = cast_mcryptd_ctx_to_req(rctx);
430	struct sha1_hash_ctx *sha_ctx;
431	struct mcryptd_hash_request_ctx *req_ctx;
432	int ret;
433
434	/* remove from work list */
435	spin_lock(&cstate->work_lock);
436	list_del(&rctx->waiter);
437	spin_unlock(&cstate->work_lock);
438
439	if (irqs_disabled())
440		rctx->complete(&req->base, err);
441	else {
442		local_bh_disable();
443		rctx->complete(&req->base, err);
444		local_bh_enable();
445	}
446
447	/* check to see if there are other jobs that are done */
448	sha_ctx = sha1_ctx_mgr_get_comp_ctx(cstate->mgr);
449	while (sha_ctx) {
450		req_ctx = cast_hash_to_mcryptd_ctx(sha_ctx);
451		ret = sha_finish_walk(&req_ctx, cstate, false);
452		if (req_ctx) {
453			spin_lock(&cstate->work_lock);
454			list_del(&req_ctx->waiter);
455			spin_unlock(&cstate->work_lock);
456
457			req = cast_mcryptd_ctx_to_req(req_ctx);
458			if (irqs_disabled())
459				req_ctx->complete(&req->base, ret);
460			else {
461				local_bh_disable();
462				req_ctx->complete(&req->base, ret);
463				local_bh_enable();
464			}
465		}
466		sha_ctx = sha1_ctx_mgr_get_comp_ctx(cstate->mgr);
467	}
468
469	return 0;
470}
471
472static void sha1_mb_add_list(struct mcryptd_hash_request_ctx *rctx,
473			     struct mcryptd_alg_cstate *cstate)
474{
475	unsigned long next_flush;
476	unsigned long delay = usecs_to_jiffies(FLUSH_INTERVAL);
477
478	/* initialize tag */
479	rctx->tag.arrival = jiffies;    /* tag the arrival time */
480	rctx->tag.seq_num = cstate->next_seq_num++;
481	next_flush = rctx->tag.arrival + delay;
482	rctx->tag.expire = next_flush;
483
484	spin_lock(&cstate->work_lock);
485	list_add_tail(&rctx->waiter, &cstate->work_list);
486	spin_unlock(&cstate->work_lock);
487
488	mcryptd_arm_flusher(cstate, delay);
489}
490
491static int sha1_mb_update(struct shash_desc *desc, const u8 *data,
492			  unsigned int len)
493{
494	struct mcryptd_hash_request_ctx *rctx =
495			container_of(desc, struct mcryptd_hash_request_ctx, desc);
496	struct mcryptd_alg_cstate *cstate =
497				this_cpu_ptr(sha1_mb_alg_state.alg_cstate);
498
499	struct ahash_request *req = cast_mcryptd_ctx_to_req(rctx);
500	struct sha1_hash_ctx *sha_ctx;
501	int ret = 0, nbytes;
502
503
504	/* sanity check */
505	if (rctx->tag.cpu != smp_processor_id()) {
506		pr_err("mcryptd error: cpu clash\n");
507		goto done;
508	}
509
510	/* need to init context */
511	req_ctx_init(rctx, desc);
512
513	nbytes = crypto_ahash_walk_first(req, &rctx->walk);
514
515	if (nbytes < 0) {
516		ret = nbytes;
517		goto done;
518	}
519
520	if (crypto_ahash_walk_last(&rctx->walk))
521		rctx->flag |= HASH_DONE;
522
523	/* submit */
524	sha_ctx = (struct sha1_hash_ctx *) shash_desc_ctx(desc);
525	sha1_mb_add_list(rctx, cstate);
526	kernel_fpu_begin();
527	sha_ctx = sha1_ctx_mgr_submit(cstate->mgr, sha_ctx, rctx->walk.data, nbytes, HASH_UPDATE);
528	kernel_fpu_end();
529
530	/* check if anything is returned */
531	if (!sha_ctx)
532		return -EINPROGRESS;
533
534	if (sha_ctx->error) {
535		ret = sha_ctx->error;
536		rctx = cast_hash_to_mcryptd_ctx(sha_ctx);
537		goto done;
538	}
539
540	rctx = cast_hash_to_mcryptd_ctx(sha_ctx);
541	ret = sha_finish_walk(&rctx, cstate, false);
542
543	if (!rctx)
544		return -EINPROGRESS;
545done:
546	sha_complete_job(rctx, cstate, ret);
547	return ret;
548}
549
550static int sha1_mb_finup(struct shash_desc *desc, const u8 *data,
551			     unsigned int len, u8 *out)
552{
553	struct mcryptd_hash_request_ctx *rctx =
554			container_of(desc, struct mcryptd_hash_request_ctx, desc);
555	struct mcryptd_alg_cstate *cstate =
556				this_cpu_ptr(sha1_mb_alg_state.alg_cstate);
557
558	struct ahash_request *req = cast_mcryptd_ctx_to_req(rctx);
559	struct sha1_hash_ctx *sha_ctx;
560	int ret = 0, flag = HASH_UPDATE, nbytes;
561
562	/* sanity check */
563	if (rctx->tag.cpu != smp_processor_id()) {
564		pr_err("mcryptd error: cpu clash\n");
565		goto done;
566	}
567
568	/* need to init context */
569	req_ctx_init(rctx, desc);
570
571	nbytes = crypto_ahash_walk_first(req, &rctx->walk);
572
573	if (nbytes < 0) {
574		ret = nbytes;
575		goto done;
576	}
577
578	if (crypto_ahash_walk_last(&rctx->walk)) {
579		rctx->flag |= HASH_DONE;
580		flag = HASH_LAST;
581	}
582	rctx->out = out;
583
584	/* submit */
585	rctx->flag |= HASH_FINAL;
586	sha_ctx = (struct sha1_hash_ctx *) shash_desc_ctx(desc);
587	sha1_mb_add_list(rctx, cstate);
588
589	kernel_fpu_begin();
590	sha_ctx = sha1_ctx_mgr_submit(cstate->mgr, sha_ctx, rctx->walk.data, nbytes, flag);
591	kernel_fpu_end();
592
593	/* check if anything is returned */
594	if (!sha_ctx)
595		return -EINPROGRESS;
596
597	if (sha_ctx->error) {
598		ret = sha_ctx->error;
599		goto done;
600	}
601
602	rctx = cast_hash_to_mcryptd_ctx(sha_ctx);
603	ret = sha_finish_walk(&rctx, cstate, false);
604	if (!rctx)
605		return -EINPROGRESS;
606done:
607	sha_complete_job(rctx, cstate, ret);
608	return ret;
609}
610
611static int sha1_mb_final(struct shash_desc *desc, u8 *out)
612{
613	struct mcryptd_hash_request_ctx *rctx =
614			container_of(desc, struct mcryptd_hash_request_ctx, desc);
615	struct mcryptd_alg_cstate *cstate =
616				this_cpu_ptr(sha1_mb_alg_state.alg_cstate);
617
618	struct sha1_hash_ctx *sha_ctx;
619	int ret = 0;
620	u8 data;
621
622	/* sanity check */
623	if (rctx->tag.cpu != smp_processor_id()) {
624		pr_err("mcryptd error: cpu clash\n");
625		goto done;
626	}
627
628	/* need to init context */
629	req_ctx_init(rctx, desc);
630
631	rctx->out = out;
632	rctx->flag |= HASH_DONE | HASH_FINAL;
633
634	sha_ctx = (struct sha1_hash_ctx *) shash_desc_ctx(desc);
635	/* flag HASH_FINAL and 0 data size */
636	sha1_mb_add_list(rctx, cstate);
637	kernel_fpu_begin();
638	sha_ctx = sha1_ctx_mgr_submit(cstate->mgr, sha_ctx, &data, 0, HASH_LAST);
639	kernel_fpu_end();
640
641	/* check if anything is returned */
642	if (!sha_ctx)
643		return -EINPROGRESS;
644
645	if (sha_ctx->error) {
646		ret = sha_ctx->error;
647		rctx = cast_hash_to_mcryptd_ctx(sha_ctx);
648		goto done;
649	}
650
651	rctx = cast_hash_to_mcryptd_ctx(sha_ctx);
652	ret = sha_finish_walk(&rctx, cstate, false);
653	if (!rctx)
654		return -EINPROGRESS;
655done:
656	sha_complete_job(rctx, cstate, ret);
657	return ret;
658}
659
660static int sha1_mb_export(struct shash_desc *desc, void *out)
661{
662	struct sha1_hash_ctx *sctx = shash_desc_ctx(desc);
663
664	memcpy(out, sctx, sizeof(*sctx));
665
666	return 0;
667}
668
669static int sha1_mb_import(struct shash_desc *desc, const void *in)
670{
671	struct sha1_hash_ctx *sctx = shash_desc_ctx(desc);
672
673	memcpy(sctx, in, sizeof(*sctx));
674
675	return 0;
676}
677
678
679static struct shash_alg sha1_mb_shash_alg = {
680	.digestsize	=	SHA1_DIGEST_SIZE,
681	.init		=	sha1_mb_init,
682	.update		=	sha1_mb_update,
683	.final		=	sha1_mb_final,
684	.finup		=	sha1_mb_finup,
685	.export		=	sha1_mb_export,
686	.import		=	sha1_mb_import,
687	.descsize	=	sizeof(struct sha1_hash_ctx),
688	.statesize	=	sizeof(struct sha1_hash_ctx),
689	.base		=	{
690		.cra_name	 = "__sha1-mb",
691		.cra_driver_name = "__intel_sha1-mb",
692		.cra_priority	 = 100,
693		/*
694		 * use ASYNC flag as some buffers in multi-buffer
695		 * algo may not have completed before hashing thread sleep
696		 */
697		.cra_flags	 = CRYPTO_ALG_TYPE_SHASH | CRYPTO_ALG_ASYNC |
698				   CRYPTO_ALG_INTERNAL,
699		.cra_blocksize	 = SHA1_BLOCK_SIZE,
700		.cra_module	 = THIS_MODULE,
701		.cra_list	 = LIST_HEAD_INIT(sha1_mb_shash_alg.base.cra_list),
702	}
703};
704
705static int sha1_mb_async_init(struct ahash_request *req)
706{
707	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
708	struct sha1_mb_ctx *ctx = crypto_ahash_ctx(tfm);
709	struct ahash_request *mcryptd_req = ahash_request_ctx(req);
710	struct mcryptd_ahash *mcryptd_tfm = ctx->mcryptd_tfm;
711
712	memcpy(mcryptd_req, req, sizeof(*req));
713	ahash_request_set_tfm(mcryptd_req, &mcryptd_tfm->base);
714	return crypto_ahash_init(mcryptd_req);
715}
716
717static int sha1_mb_async_update(struct ahash_request *req)
718{
719	struct ahash_request *mcryptd_req = ahash_request_ctx(req);
720
721	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
722	struct sha1_mb_ctx *ctx = crypto_ahash_ctx(tfm);
723	struct mcryptd_ahash *mcryptd_tfm = ctx->mcryptd_tfm;
724
725	memcpy(mcryptd_req, req, sizeof(*req));
726	ahash_request_set_tfm(mcryptd_req, &mcryptd_tfm->base);
727	return crypto_ahash_update(mcryptd_req);
728}
729
730static int sha1_mb_async_finup(struct ahash_request *req)
731{
732	struct ahash_request *mcryptd_req = ahash_request_ctx(req);
733
734	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
735	struct sha1_mb_ctx *ctx = crypto_ahash_ctx(tfm);
736	struct mcryptd_ahash *mcryptd_tfm = ctx->mcryptd_tfm;
737
738	memcpy(mcryptd_req, req, sizeof(*req));
739	ahash_request_set_tfm(mcryptd_req, &mcryptd_tfm->base);
740	return crypto_ahash_finup(mcryptd_req);
741}
742
743static int sha1_mb_async_final(struct ahash_request *req)
744{
745	struct ahash_request *mcryptd_req = ahash_request_ctx(req);
746
747	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
748	struct sha1_mb_ctx *ctx = crypto_ahash_ctx(tfm);
749	struct mcryptd_ahash *mcryptd_tfm = ctx->mcryptd_tfm;
750
751	memcpy(mcryptd_req, req, sizeof(*req));
752	ahash_request_set_tfm(mcryptd_req, &mcryptd_tfm->base);
753	return crypto_ahash_final(mcryptd_req);
754}
755
756static int sha1_mb_async_digest(struct ahash_request *req)
757{
758	struct crypto_ahash *tfm = crypto_ahash_reqtfm(req);
759	struct sha1_mb_ctx *ctx = crypto_ahash_ctx(tfm);
760	struct ahash_request *mcryptd_req = ahash_request_ctx(req);
761	struct mcryptd_ahash *mcryptd_tfm = ctx->mcryptd_tfm;
762
763	memcpy(mcryptd_req, req, sizeof(*req));
764	ahash_request_set_tfm(mcryptd_req, &mcryptd_tfm->base);
765	return crypto_ahash_digest(mcryptd_req);
766}
767
768static int sha1_mb_async_init_tfm(struct crypto_tfm *tfm)
769{
770	struct mcryptd_ahash *mcryptd_tfm;
771	struct sha1_mb_ctx *ctx = crypto_tfm_ctx(tfm);
772	struct mcryptd_hash_ctx *mctx;
773
774	mcryptd_tfm = mcryptd_alloc_ahash("__intel_sha1-mb",
775					  CRYPTO_ALG_INTERNAL,
776					  CRYPTO_ALG_INTERNAL);
777	if (IS_ERR(mcryptd_tfm))
778		return PTR_ERR(mcryptd_tfm);
779	mctx = crypto_ahash_ctx(&mcryptd_tfm->base);
780	mctx->alg_state = &sha1_mb_alg_state;
781	ctx->mcryptd_tfm = mcryptd_tfm;
782	crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm),
783				 sizeof(struct ahash_request) +
784				 crypto_ahash_reqsize(&mcryptd_tfm->base));
785
786	return 0;
787}
788
789static void sha1_mb_async_exit_tfm(struct crypto_tfm *tfm)
790{
791	struct sha1_mb_ctx *ctx = crypto_tfm_ctx(tfm);
792
793	mcryptd_free_ahash(ctx->mcryptd_tfm);
794}
795
796static struct ahash_alg sha1_mb_async_alg = {
797	.init           = sha1_mb_async_init,
798	.update         = sha1_mb_async_update,
799	.final          = sha1_mb_async_final,
800	.finup          = sha1_mb_async_finup,
801	.digest         = sha1_mb_async_digest,
802	.halg = {
803		.digestsize     = SHA1_DIGEST_SIZE,
804		.base = {
805			.cra_name               = "sha1",
806			.cra_driver_name        = "sha1_mb",
807			.cra_priority           = 200,
808			.cra_flags              = CRYPTO_ALG_TYPE_AHASH | CRYPTO_ALG_ASYNC,
809			.cra_blocksize          = SHA1_BLOCK_SIZE,
810			.cra_type               = &crypto_ahash_type,
811			.cra_module             = THIS_MODULE,
812			.cra_list               = LIST_HEAD_INIT(sha1_mb_async_alg.halg.base.cra_list),
813			.cra_init               = sha1_mb_async_init_tfm,
814			.cra_exit               = sha1_mb_async_exit_tfm,
815			.cra_ctxsize		= sizeof(struct sha1_mb_ctx),
816			.cra_alignmask		= 0,
817		},
818	},
819};
820
821static unsigned long sha1_mb_flusher(struct mcryptd_alg_cstate *cstate)
822{
823	struct mcryptd_hash_request_ctx *rctx;
824	unsigned long cur_time;
825	unsigned long next_flush = 0;
826	struct sha1_hash_ctx *sha_ctx;
827
828
829	cur_time = jiffies;
830
831	while (!list_empty(&cstate->work_list)) {
832		rctx = list_entry(cstate->work_list.next,
833				struct mcryptd_hash_request_ctx, waiter);
834		if (time_before(cur_time, rctx->tag.expire))
835			break;
836		kernel_fpu_begin();
837		sha_ctx = (struct sha1_hash_ctx *) sha1_ctx_mgr_flush(cstate->mgr);
838		kernel_fpu_end();
839		if (!sha_ctx) {
840			pr_err("sha1_mb error: nothing got flushed for non-empty list\n");
841			break;
842		}
843		rctx = cast_hash_to_mcryptd_ctx(sha_ctx);
844		sha_finish_walk(&rctx, cstate, true);
845		sha_complete_job(rctx, cstate, 0);
846	}
847
848	if (!list_empty(&cstate->work_list)) {
849		rctx = list_entry(cstate->work_list.next,
850				struct mcryptd_hash_request_ctx, waiter);
851		/* get the hash context and then flush time */
852		next_flush = rctx->tag.expire;
853		mcryptd_arm_flusher(cstate, get_delay(next_flush));
854	}
855	return next_flush;
856}
857
858static int __init sha1_mb_mod_init(void)
859{
860
861	int cpu;
862	int err;
863	struct mcryptd_alg_cstate *cpu_state;
864
865	/* check for dependent cpu features */
866	if (!boot_cpu_has(X86_FEATURE_AVX2) ||
867	    !boot_cpu_has(X86_FEATURE_BMI2))
868		return -ENODEV;
869
870	/* initialize multibuffer structures */
871	sha1_mb_alg_state.alg_cstate = alloc_percpu(struct mcryptd_alg_cstate);
872
873	sha1_job_mgr_init = sha1_mb_mgr_init_avx2;
874	sha1_job_mgr_submit = sha1_mb_mgr_submit_avx2;
875	sha1_job_mgr_flush = sha1_mb_mgr_flush_avx2;
876	sha1_job_mgr_get_comp_job = sha1_mb_mgr_get_comp_job_avx2;
877
878	if (!sha1_mb_alg_state.alg_cstate)
879		return -ENOMEM;
880	for_each_possible_cpu(cpu) {
881		cpu_state = per_cpu_ptr(sha1_mb_alg_state.alg_cstate, cpu);
882		cpu_state->next_flush = 0;
883		cpu_state->next_seq_num = 0;
884		cpu_state->flusher_engaged = false;
885		INIT_DELAYED_WORK(&cpu_state->flush, mcryptd_flusher);
886		cpu_state->cpu = cpu;
887		cpu_state->alg_state = &sha1_mb_alg_state;
888		cpu_state->mgr = (struct sha1_ctx_mgr *) kzalloc(sizeof(struct sha1_ctx_mgr), GFP_KERNEL);
889		if (!cpu_state->mgr)
890			goto err2;
891		sha1_ctx_mgr_init(cpu_state->mgr);
892		INIT_LIST_HEAD(&cpu_state->work_list);
893		spin_lock_init(&cpu_state->work_lock);
894	}
895	sha1_mb_alg_state.flusher = &sha1_mb_flusher;
896
897	err = crypto_register_shash(&sha1_mb_shash_alg);
898	if (err)
899		goto err2;
900	err = crypto_register_ahash(&sha1_mb_async_alg);
901	if (err)
902		goto err1;
903
904
905	return 0;
906err1:
907	crypto_unregister_shash(&sha1_mb_shash_alg);
908err2:
909	for_each_possible_cpu(cpu) {
910		cpu_state = per_cpu_ptr(sha1_mb_alg_state.alg_cstate, cpu);
911		kfree(cpu_state->mgr);
912	}
913	free_percpu(sha1_mb_alg_state.alg_cstate);
914	return -ENODEV;
915}
916
917static void __exit sha1_mb_mod_fini(void)
918{
919	int cpu;
920	struct mcryptd_alg_cstate *cpu_state;
921
922	crypto_unregister_ahash(&sha1_mb_async_alg);
923	crypto_unregister_shash(&sha1_mb_shash_alg);
924	for_each_possible_cpu(cpu) {
925		cpu_state = per_cpu_ptr(sha1_mb_alg_state.alg_cstate, cpu);
926		kfree(cpu_state->mgr);
927	}
928	free_percpu(sha1_mb_alg_state.alg_cstate);
929}
930
931module_init(sha1_mb_mod_init);
932module_exit(sha1_mb_mod_fini);
933
934MODULE_LICENSE("GPL");
935MODULE_DESCRIPTION("SHA1 Secure Hash Algorithm, multi buffer accelerated");
936
937MODULE_ALIAS_CRYPTO("sha1");
938