1 /*
2  * caam descriptor construction helper functions
3  *
4  * Copyright 2008-2012 Freescale Semiconductor, Inc.
5  */
6 
7 #include "desc.h"
8 #include "regs.h"
9 
10 #define IMMEDIATE (1 << 23)
11 #define CAAM_CMD_SZ sizeof(u32)
12 #define CAAM_PTR_SZ sizeof(dma_addr_t)
13 #define CAAM_DESC_BYTES_MAX (CAAM_CMD_SZ * MAX_CAAM_DESCSIZE)
14 #define DESC_JOB_IO_LEN (CAAM_CMD_SZ * 5 + CAAM_PTR_SZ * 3)
15 
16 #ifdef DEBUG
17 #define PRINT_POS do { printk(KERN_DEBUG "%02d: %s\n", desc_len(desc),\
18 			      &__func__[sizeof("append")]); } while (0)
19 #else
20 #define PRINT_POS
21 #endif
22 
23 #define SET_OK_NO_PROP_ERRORS (IMMEDIATE | LDST_CLASS_DECO | \
24 			       LDST_SRCDST_WORD_DECOCTRL | \
25 			       (LDOFF_CHG_SHARE_OK_NO_PROP << \
26 				LDST_OFFSET_SHIFT))
27 #define DISABLE_AUTO_INFO_FIFO (IMMEDIATE | LDST_CLASS_DECO | \
28 				LDST_SRCDST_WORD_DECOCTRL | \
29 				(LDOFF_DISABLE_AUTO_NFIFO << LDST_OFFSET_SHIFT))
30 #define ENABLE_AUTO_INFO_FIFO (IMMEDIATE | LDST_CLASS_DECO | \
31 			       LDST_SRCDST_WORD_DECOCTRL | \
32 			       (LDOFF_ENABLE_AUTO_NFIFO << LDST_OFFSET_SHIFT))
33 
34 extern bool caam_little_end;
35 
36 static inline int desc_len(u32 * const desc)
37 {
38 	return caam32_to_cpu(*desc) & HDR_DESCLEN_MASK;
39 }
40 
41 static inline int desc_bytes(void * const desc)
42 {
43 	return desc_len(desc) * CAAM_CMD_SZ;
44 }
45 
46 static inline u32 *desc_end(u32 * const desc)
47 {
48 	return desc + desc_len(desc);
49 }
50 
51 static inline void *sh_desc_pdb(u32 * const desc)
52 {
53 	return desc + 1;
54 }
55 
56 static inline void init_desc(u32 * const desc, u32 options)
57 {
58 	*desc = cpu_to_caam32((options | HDR_ONE) + 1);
59 }
60 
61 static inline void init_sh_desc(u32 * const desc, u32 options)
62 {
63 	PRINT_POS;
64 	init_desc(desc, CMD_SHARED_DESC_HDR | options);
65 }
66 
67 static inline void init_sh_desc_pdb(u32 * const desc, u32 options,
68 				    size_t pdb_bytes)
69 {
70 	u32 pdb_len = (pdb_bytes + CAAM_CMD_SZ - 1) / CAAM_CMD_SZ;
71 
72 	init_sh_desc(desc, (((pdb_len + 1) << HDR_START_IDX_SHIFT) + pdb_len) |
73 		     options);
74 }
75 
76 static inline void init_job_desc(u32 * const desc, u32 options)
77 {
78 	init_desc(desc, CMD_DESC_HDR | options);
79 }
80 
81 static inline void init_job_desc_pdb(u32 * const desc, u32 options,
82 				     size_t pdb_bytes)
83 {
84 	u32 pdb_len = (pdb_bytes + CAAM_CMD_SZ - 1) / CAAM_CMD_SZ;
85 
86 	init_job_desc(desc, (((pdb_len + 1) << HDR_START_IDX_SHIFT)) | options);
87 }
88 
89 static inline void append_ptr(u32 * const desc, dma_addr_t ptr)
90 {
91 	dma_addr_t *offset = (dma_addr_t *)desc_end(desc);
92 
93 	*offset = cpu_to_caam_dma(ptr);
94 
95 	(*desc) = cpu_to_caam32(caam32_to_cpu(*desc) +
96 				CAAM_PTR_SZ / CAAM_CMD_SZ);
97 }
98 
99 static inline void init_job_desc_shared(u32 * const desc, dma_addr_t ptr,
100 					int len, u32 options)
101 {
102 	PRINT_POS;
103 	init_job_desc(desc, HDR_SHARED | options |
104 		      (len << HDR_START_IDX_SHIFT));
105 	append_ptr(desc, ptr);
106 }
107 
108 static inline void append_data(u32 * const desc, void *data, int len)
109 {
110 	u32 *offset = desc_end(desc);
111 
112 	if (len) /* avoid sparse warning: memcpy with byte count of 0 */
113 		memcpy(offset, data, len);
114 
115 	(*desc) = cpu_to_caam32(caam32_to_cpu(*desc) +
116 				(len + CAAM_CMD_SZ - 1) / CAAM_CMD_SZ);
117 }
118 
119 static inline void append_cmd(u32 * const desc, u32 command)
120 {
121 	u32 *cmd = desc_end(desc);
122 
123 	*cmd = cpu_to_caam32(command);
124 
125 	(*desc) = cpu_to_caam32(caam32_to_cpu(*desc) + 1);
126 }
127 
128 #define append_u32 append_cmd
129 
130 static inline void append_u64(u32 * const desc, u64 data)
131 {
132 	u32 *offset = desc_end(desc);
133 
134 	/* Only 32-bit alignment is guaranteed in descriptor buffer */
135 	if (caam_little_end) {
136 		*offset = cpu_to_caam32(lower_32_bits(data));
137 		*(++offset) = cpu_to_caam32(upper_32_bits(data));
138 	} else {
139 		*offset = cpu_to_caam32(upper_32_bits(data));
140 		*(++offset) = cpu_to_caam32(lower_32_bits(data));
141 	}
142 
143 	(*desc) = cpu_to_caam32(caam32_to_cpu(*desc) + 2);
144 }
145 
146 /* Write command without affecting header, and return pointer to next word */
147 static inline u32 *write_cmd(u32 * const desc, u32 command)
148 {
149 	*desc = cpu_to_caam32(command);
150 
151 	return desc + 1;
152 }
153 
154 static inline void append_cmd_ptr(u32 * const desc, dma_addr_t ptr, int len,
155 				  u32 command)
156 {
157 	append_cmd(desc, command | len);
158 	append_ptr(desc, ptr);
159 }
160 
161 /* Write length after pointer, rather than inside command */
162 static inline void append_cmd_ptr_extlen(u32 * const desc, dma_addr_t ptr,
163 					 unsigned int len, u32 command)
164 {
165 	append_cmd(desc, command);
166 	if (!(command & (SQIN_RTO | SQIN_PRE)))
167 		append_ptr(desc, ptr);
168 	append_cmd(desc, len);
169 }
170 
171 static inline void append_cmd_data(u32 * const desc, void *data, int len,
172 				   u32 command)
173 {
174 	append_cmd(desc, command | IMMEDIATE | len);
175 	append_data(desc, data, len);
176 }
177 
178 #define APPEND_CMD_RET(cmd, op) \
179 static inline u32 *append_##cmd(u32 * const desc, u32 options) \
180 { \
181 	u32 *cmd = desc_end(desc); \
182 	PRINT_POS; \
183 	append_cmd(desc, CMD_##op | options); \
184 	return cmd; \
185 }
186 APPEND_CMD_RET(jump, JUMP)
187 APPEND_CMD_RET(move, MOVE)
188 
189 static inline void set_jump_tgt_here(u32 * const desc, u32 *jump_cmd)
190 {
191 	*jump_cmd = cpu_to_caam32(caam32_to_cpu(*jump_cmd) |
192 				  (desc_len(desc) - (jump_cmd - desc)));
193 }
194 
195 static inline void set_move_tgt_here(u32 * const desc, u32 *move_cmd)
196 {
197 	u32 val = caam32_to_cpu(*move_cmd);
198 
199 	val &= ~MOVE_OFFSET_MASK;
200 	val |= (desc_len(desc) << (MOVE_OFFSET_SHIFT + 2)) & MOVE_OFFSET_MASK;
201 	*move_cmd = cpu_to_caam32(val);
202 }
203 
204 #define APPEND_CMD(cmd, op) \
205 static inline void append_##cmd(u32 * const desc, u32 options) \
206 { \
207 	PRINT_POS; \
208 	append_cmd(desc, CMD_##op | options); \
209 }
210 APPEND_CMD(operation, OPERATION)
211 
212 #define APPEND_CMD_LEN(cmd, op) \
213 static inline void append_##cmd(u32 * const desc, unsigned int len, \
214 				u32 options) \
215 { \
216 	PRINT_POS; \
217 	append_cmd(desc, CMD_##op | len | options); \
218 }
219 
220 APPEND_CMD_LEN(seq_load, SEQ_LOAD)
221 APPEND_CMD_LEN(seq_store, SEQ_STORE)
222 APPEND_CMD_LEN(seq_fifo_load, SEQ_FIFO_LOAD)
223 APPEND_CMD_LEN(seq_fifo_store, SEQ_FIFO_STORE)
224 
225 #define APPEND_CMD_PTR(cmd, op) \
226 static inline void append_##cmd(u32 * const desc, dma_addr_t ptr, \
227 				unsigned int len, u32 options) \
228 { \
229 	PRINT_POS; \
230 	append_cmd_ptr(desc, ptr, len, CMD_##op | options); \
231 }
232 APPEND_CMD_PTR(key, KEY)
233 APPEND_CMD_PTR(load, LOAD)
234 APPEND_CMD_PTR(fifo_load, FIFO_LOAD)
235 APPEND_CMD_PTR(fifo_store, FIFO_STORE)
236 
237 static inline void append_store(u32 * const desc, dma_addr_t ptr,
238 				unsigned int len, u32 options)
239 {
240 	u32 cmd_src;
241 
242 	cmd_src = options & LDST_SRCDST_MASK;
243 
244 	append_cmd(desc, CMD_STORE | options | len);
245 
246 	/* The following options do not require pointer */
247 	if (!(cmd_src == LDST_SRCDST_WORD_DESCBUF_SHARED ||
248 	      cmd_src == LDST_SRCDST_WORD_DESCBUF_JOB    ||
249 	      cmd_src == LDST_SRCDST_WORD_DESCBUF_JOB_WE ||
250 	      cmd_src == LDST_SRCDST_WORD_DESCBUF_SHARED_WE))
251 		append_ptr(desc, ptr);
252 }
253 
254 #define APPEND_SEQ_PTR_INTLEN(cmd, op) \
255 static inline void append_seq_##cmd##_ptr_intlen(u32 * const desc, \
256 						 dma_addr_t ptr, \
257 						 unsigned int len, \
258 						 u32 options) \
259 { \
260 	PRINT_POS; \
261 	if (options & (SQIN_RTO | SQIN_PRE)) \
262 		append_cmd(desc, CMD_SEQ_##op##_PTR | len | options); \
263 	else \
264 		append_cmd_ptr(desc, ptr, len, CMD_SEQ_##op##_PTR | options); \
265 }
266 APPEND_SEQ_PTR_INTLEN(in, IN)
267 APPEND_SEQ_PTR_INTLEN(out, OUT)
268 
269 #define APPEND_CMD_PTR_TO_IMM(cmd, op) \
270 static inline void append_##cmd##_as_imm(u32 * const desc, void *data, \
271 					 unsigned int len, u32 options) \
272 { \
273 	PRINT_POS; \
274 	append_cmd_data(desc, data, len, CMD_##op | options); \
275 }
276 APPEND_CMD_PTR_TO_IMM(load, LOAD);
277 APPEND_CMD_PTR_TO_IMM(fifo_load, FIFO_LOAD);
278 
279 #define APPEND_CMD_PTR_EXTLEN(cmd, op) \
280 static inline void append_##cmd##_extlen(u32 * const desc, dma_addr_t ptr, \
281 					 unsigned int len, u32 options) \
282 { \
283 	PRINT_POS; \
284 	append_cmd_ptr_extlen(desc, ptr, len, CMD_##op | SQIN_EXT | options); \
285 }
286 APPEND_CMD_PTR_EXTLEN(seq_in_ptr, SEQ_IN_PTR)
287 APPEND_CMD_PTR_EXTLEN(seq_out_ptr, SEQ_OUT_PTR)
288 
289 /*
290  * Determine whether to store length internally or externally depending on
291  * the size of its type
292  */
293 #define APPEND_CMD_PTR_LEN(cmd, op, type) \
294 static inline void append_##cmd(u32 * const desc, dma_addr_t ptr, \
295 				type len, u32 options) \
296 { \
297 	PRINT_POS; \
298 	if (sizeof(type) > sizeof(u16)) \
299 		append_##cmd##_extlen(desc, ptr, len, options); \
300 	else \
301 		append_##cmd##_intlen(desc, ptr, len, options); \
302 }
303 APPEND_CMD_PTR_LEN(seq_in_ptr, SEQ_IN_PTR, u32)
304 APPEND_CMD_PTR_LEN(seq_out_ptr, SEQ_OUT_PTR, u32)
305 
306 /*
307  * 2nd variant for commands whose specified immediate length differs
308  * from length of immediate data provided, e.g., split keys
309  */
310 #define APPEND_CMD_PTR_TO_IMM2(cmd, op) \
311 static inline void append_##cmd##_as_imm(u32 * const desc, void *data, \
312 					 unsigned int data_len, \
313 					 unsigned int len, u32 options) \
314 { \
315 	PRINT_POS; \
316 	append_cmd(desc, CMD_##op | IMMEDIATE | len | options); \
317 	append_data(desc, data, data_len); \
318 }
319 APPEND_CMD_PTR_TO_IMM2(key, KEY);
320 
321 #define APPEND_CMD_RAW_IMM(cmd, op, type) \
322 static inline void append_##cmd##_imm_##type(u32 * const desc, type immediate, \
323 					     u32 options) \
324 { \
325 	PRINT_POS; \
326 	append_cmd(desc, CMD_##op | IMMEDIATE | options | sizeof(type)); \
327 	append_cmd(desc, immediate); \
328 }
329 APPEND_CMD_RAW_IMM(load, LOAD, u32);
330 
331 /*
332  * ee - endianness
333  * size - size of immediate type in bytes
334  */
335 #define APPEND_CMD_RAW_IMM2(cmd, op, ee, size) \
336 static inline void append_##cmd##_imm_##ee##size(u32 *desc, \
337 						   u##size immediate, \
338 						   u32 options) \
339 { \
340 	__##ee##size data = cpu_to_##ee##size(immediate); \
341 	PRINT_POS; \
342 	append_cmd(desc, CMD_##op | IMMEDIATE | options | sizeof(data)); \
343 	append_data(desc, &data, sizeof(data)); \
344 }
345 
346 APPEND_CMD_RAW_IMM2(load, LOAD, be, 32);
347 
348 /*
349  * Append math command. Only the last part of destination and source need to
350  * be specified
351  */
352 #define APPEND_MATH(op, desc, dest, src_0, src_1, len) \
353 append_cmd(desc, CMD_MATH | MATH_FUN_##op | MATH_DEST_##dest | \
354 	MATH_SRC0_##src_0 | MATH_SRC1_##src_1 | (u32)len);
355 
356 #define append_math_add(desc, dest, src0, src1, len) \
357 	APPEND_MATH(ADD, desc, dest, src0, src1, len)
358 #define append_math_sub(desc, dest, src0, src1, len) \
359 	APPEND_MATH(SUB, desc, dest, src0, src1, len)
360 #define append_math_add_c(desc, dest, src0, src1, len) \
361 	APPEND_MATH(ADDC, desc, dest, src0, src1, len)
362 #define append_math_sub_b(desc, dest, src0, src1, len) \
363 	APPEND_MATH(SUBB, desc, dest, src0, src1, len)
364 #define append_math_and(desc, dest, src0, src1, len) \
365 	APPEND_MATH(AND, desc, dest, src0, src1, len)
366 #define append_math_or(desc, dest, src0, src1, len) \
367 	APPEND_MATH(OR, desc, dest, src0, src1, len)
368 #define append_math_xor(desc, dest, src0, src1, len) \
369 	APPEND_MATH(XOR, desc, dest, src0, src1, len)
370 #define append_math_lshift(desc, dest, src0, src1, len) \
371 	APPEND_MATH(LSHIFT, desc, dest, src0, src1, len)
372 #define append_math_rshift(desc, dest, src0, src1, len) \
373 	APPEND_MATH(RSHIFT, desc, dest, src0, src1, len)
374 #define append_math_ldshift(desc, dest, src0, src1, len) \
375 	APPEND_MATH(SHLD, desc, dest, src0, src1, len)
376 
377 /* Exactly one source is IMM. Data is passed in as u32 value */
378 #define APPEND_MATH_IMM_u32(op, desc, dest, src_0, src_1, data) \
379 do { \
380 	APPEND_MATH(op, desc, dest, src_0, src_1, CAAM_CMD_SZ); \
381 	append_cmd(desc, data); \
382 } while (0)
383 
384 #define append_math_add_imm_u32(desc, dest, src0, src1, data) \
385 	APPEND_MATH_IMM_u32(ADD, desc, dest, src0, src1, data)
386 #define append_math_sub_imm_u32(desc, dest, src0, src1, data) \
387 	APPEND_MATH_IMM_u32(SUB, desc, dest, src0, src1, data)
388 #define append_math_add_c_imm_u32(desc, dest, src0, src1, data) \
389 	APPEND_MATH_IMM_u32(ADDC, desc, dest, src0, src1, data)
390 #define append_math_sub_b_imm_u32(desc, dest, src0, src1, data) \
391 	APPEND_MATH_IMM_u32(SUBB, desc, dest, src0, src1, data)
392 #define append_math_and_imm_u32(desc, dest, src0, src1, data) \
393 	APPEND_MATH_IMM_u32(AND, desc, dest, src0, src1, data)
394 #define append_math_or_imm_u32(desc, dest, src0, src1, data) \
395 	APPEND_MATH_IMM_u32(OR, desc, dest, src0, src1, data)
396 #define append_math_xor_imm_u32(desc, dest, src0, src1, data) \
397 	APPEND_MATH_IMM_u32(XOR, desc, dest, src0, src1, data)
398 #define append_math_lshift_imm_u32(desc, dest, src0, src1, data) \
399 	APPEND_MATH_IMM_u32(LSHIFT, desc, dest, src0, src1, data)
400 #define append_math_rshift_imm_u32(desc, dest, src0, src1, data) \
401 	APPEND_MATH_IMM_u32(RSHIFT, desc, dest, src0, src1, data)
402 
403 /* Exactly one source is IMM. Data is passed in as u64 value */
404 #define APPEND_MATH_IMM_u64(op, desc, dest, src_0, src_1, data) \
405 do { \
406 	u32 upper = (data >> 16) >> 16; \
407 	APPEND_MATH(op, desc, dest, src_0, src_1, CAAM_CMD_SZ * 2 | \
408 		    (upper ? 0 : MATH_IFB)); \
409 	if (upper) \
410 		append_u64(desc, data); \
411 	else \
412 		append_u32(desc, lower_32_bits(data)); \
413 } while (0)
414 
415 #define append_math_add_imm_u64(desc, dest, src0, src1, data) \
416 	APPEND_MATH_IMM_u64(ADD, desc, dest, src0, src1, data)
417 #define append_math_sub_imm_u64(desc, dest, src0, src1, data) \
418 	APPEND_MATH_IMM_u64(SUB, desc, dest, src0, src1, data)
419 #define append_math_add_c_imm_u64(desc, dest, src0, src1, data) \
420 	APPEND_MATH_IMM_u64(ADDC, desc, dest, src0, src1, data)
421 #define append_math_sub_b_imm_u64(desc, dest, src0, src1, data) \
422 	APPEND_MATH_IMM_u64(SUBB, desc, dest, src0, src1, data)
423 #define append_math_and_imm_u64(desc, dest, src0, src1, data) \
424 	APPEND_MATH_IMM_u64(AND, desc, dest, src0, src1, data)
425 #define append_math_or_imm_u64(desc, dest, src0, src1, data) \
426 	APPEND_MATH_IMM_u64(OR, desc, dest, src0, src1, data)
427 #define append_math_xor_imm_u64(desc, dest, src0, src1, data) \
428 	APPEND_MATH_IMM_u64(XOR, desc, dest, src0, src1, data)
429 #define append_math_lshift_imm_u64(desc, dest, src0, src1, data) \
430 	APPEND_MATH_IMM_u64(LSHIFT, desc, dest, src0, src1, data)
431 #define append_math_rshift_imm_u64(desc, dest, src0, src1, data) \
432 	APPEND_MATH_IMM_u64(RSHIFT, desc, dest, src0, src1, data)
433 
434 /**
435  * struct alginfo - Container for algorithm details
436  * @algtype: algorithm selector; for valid values, see documentation of the
437  *           functions where it is used.
438  * @keylen: length of the provided algorithm key, in bytes
439  * @keylen_pad: padded length of the provided algorithm key, in bytes
440  * @key: address where algorithm key resides; virtual address if key_inline
441  *       is true, dma (bus) address if key_inline is false.
442  * @key_inline: true - key can be inlined in the descriptor; false - key is
443  *              referenced by the descriptor
444  */
445 struct alginfo {
446 	u32 algtype;
447 	unsigned int keylen;
448 	unsigned int keylen_pad;
449 	union {
450 		dma_addr_t key_dma;
451 		void *key_virt;
452 	};
453 	bool key_inline;
454 };
455 
456 /**
457  * desc_inline_query() - Provide indications on which data items can be inlined
458  *                       and which shall be referenced in a shared descriptor.
459  * @sd_base_len: Shared descriptor base length - bytes consumed by the commands,
460  *               excluding the data items to be inlined (or corresponding
461  *               pointer if an item is not inlined). Each cnstr_* function that
462  *               generates descriptors should have a define mentioning
463  *               corresponding length.
464  * @jd_len: Maximum length of the job descriptor(s) that will be used
465  *          together with the shared descriptor.
466  * @data_len: Array of lengths of the data items trying to be inlined
467  * @inl_mask: 32bit mask with bit x = 1 if data item x can be inlined, 0
468  *            otherwise.
469  * @count: Number of data items (size of @data_len array); must be <= 32
470  *
471  * Return: 0 if data can be inlined / referenced, negative value if not. If 0,
472  *         check @inl_mask for details.
473  */
474 static inline int desc_inline_query(unsigned int sd_base_len,
475 				    unsigned int jd_len, unsigned int *data_len,
476 				    u32 *inl_mask, unsigned int count)
477 {
478 	int rem_bytes = (int)(CAAM_DESC_BYTES_MAX - sd_base_len - jd_len);
479 	unsigned int i;
480 
481 	*inl_mask = 0;
482 	for (i = 0; (i < count) && (rem_bytes > 0); i++) {
483 		if (rem_bytes - (int)(data_len[i] +
484 			(count - i - 1) * CAAM_PTR_SZ) >= 0) {
485 			rem_bytes -= data_len[i];
486 			*inl_mask |= (1 << i);
487 		} else {
488 			rem_bytes -= CAAM_PTR_SZ;
489 		}
490 	}
491 
492 	return (rem_bytes >= 0) ? 0 : -1;
493 }
494