xref: /rk3399_ARM-atf/drivers/nxp/ddr/phy-gen2/phy.c (revision fa0105693c85eacf6eda22eca63f220d304f7768)
1 /*
2  * Copyright 2021-2022 NXP
3  * SPDX-License-Identifier: BSD-3-Clause
4  *
5  */
6 
7 #include <errno.h>
8 #include <stdint.h>
9 #include <stdio.h>
10 #include <stdlib.h>
11 #include <string.h>
12 
13 #include <common/debug.h>
14 #include "csr.h"
15 #include <ddr.h>
16 #include "ddr4fw.h"
17 #include <drivers/delay_timer.h>
18 #ifdef NXP_WARM_BOOT
19 #include <fspi_api.h>
20 #endif
21 #include "input.h"
22 #include <lib/mmio.h>
23 #include <lib/utils.h>
24 #include <lib/xlat_tables/xlat_tables_v2.h>
25 #ifdef DDR_PHY_DEBUG
26 #include "messages.h"
27 #endif
28 #ifdef NXP_WARM_BOOT
29 #include "phy.h"
30 #endif
31 #include "pie.h"
32 
33 #define TIMEOUTDEFAULT 500
34 #define MAP_PHY_ADDR(pstate, n, instance, offset, c) \
35 		((((pstate * n) + instance + c) << 12) + offset)
36 
37 static uint32_t map_phy_addr_space(uint32_t addr)
38 {
39 	/* 23 bit addressing */
40 	uint32_t pstate =     (addr & U(0x700000)) >> 20U; /* bit 22:20 */
41 	uint32_t block_type = (addr & U(0x0f0000)) >> 16U; /* bit 19:16 */
42 	uint32_t instance =   (addr & U(0x00f000)) >> 12U; /* bit 15:12 */
43 	uint32_t offset =     (addr & U(0x000fff));        /* bit 11:0 */
44 
45 	switch (block_type) {
46 	case 0x0: /* 0x0 : ANIB */
47 		return MAP_PHY_ADDR(pstate, 12, instance, offset, 0);
48 	case 0x1: /* 0x1 : DBYTE */
49 		return MAP_PHY_ADDR(pstate, 10, instance, offset, 0x30);
50 	case 0x2: /* 0x2 : MASTER */
51 		return MAP_PHY_ADDR(pstate, 1, 0, offset, 0x58);
52 	case 0x4: /* 0x4 : ACSM */
53 		return MAP_PHY_ADDR(pstate, 1, 0, offset, 0x5c);
54 	case 0x5: /* 0x5 : μCTL Memory */
55 		return MAP_PHY_ADDR(pstate, 0, instance, offset, 0x60);
56 	case 0x7: /* 0x7 : PPGC */
57 		return MAP_PHY_ADDR(pstate, 0, 0, offset, 0x68);
58 	case 0x9: /* 0x9 : INITENG */
59 		return MAP_PHY_ADDR(pstate, 1, 0, offset, 0x69);
60 	case 0xc: /* 0xC : DRTUB */
61 		return MAP_PHY_ADDR(pstate, 0, 0, offset, 0x6d);
62 	case 0xd: /* 0xD : APB Only */
63 		return MAP_PHY_ADDR(pstate, 0, 0, offset, 0x6e);
64 	default:
65 		printf("ERR: Invalid block_type = 0x%x\n", block_type);
66 		return 0;
67 	}
68 }
69 
70 static inline uint16_t *phy_io_addr(void *phy, uint32_t addr)
71 {
72 	return phy + (map_phy_addr_space(addr) << 2);
73 }
74 
75 static inline void phy_io_write16(uint16_t *phy, uint32_t addr, uint16_t data)
76 {
77 	mmio_write_16((uintptr_t)phy_io_addr(phy, addr), data);
78 #ifdef DEBUG_PHY_IO
79 	printf("0x%06x,0x%x\n", addr, data);
80 #endif
81 }
82 
83 static inline uint16_t phy_io_read16(uint16_t *phy, uint32_t addr)
84 {
85 	uint16_t reg = mmio_read_16((uintptr_t) phy_io_addr(phy, addr));
86 
87 #ifdef DEBUG_PHY_IO
88 	printf("R: 0x%06x,0x%x\n", addr, reg);
89 #endif
90 
91 	return reg;
92 }
93 
94 #ifdef NXP_APPLY_MAX_CDD
95 
96 #define CDD_VAL_READ_ADDR (0x054012)
97 #define CDD_DATA_LEN    (60)
98 
99 static void read_phy_reg(uint16_t *phy, uint32_t addr,
100 		uint16_t *buf, uint32_t len)
101 {
102 	uint32_t i = 0U;
103 
104 	for (i = 0U; i < len/2; i++) {
105 		buf[i] = phy_io_read16(phy, (addr + i));
106 	}
107 }
108 
109 static uint32_t findrank(uint32_t cs_in_use)
110 {
111 	uint32_t val = 0U;
112 
113 	switch (cs_in_use) {
114 	case U(0xf):
115 		val = 4U;
116 		break;
117 	case U(0x3):
118 		val = 2U;
119 		break;
120 	case U(0x1):
121 		val = 1U;
122 		break;
123 	default:
124 		printf("Error - Invalid cs_in_use value\n");
125 	}
126 	return val;
127 }
128 
129 static uint8_t findmax(uint8_t *buf, uint32_t len)
130 {
131 	uint8_t max = 0U;
132 	uint32_t i = 0U;
133 
134 	for (i = 0U; i < len; i++) {
135 		if (buf[i] > max) {
136 			max = buf[i];
137 		}
138 	}
139 
140 	return max;
141 }
142 
143 static void get_cdd_val(uint16_t **phy_ptr, uint32_t rank, uint32_t freq,
144 		uint32_t *tcfg0, uint32_t *tcfg4)
145 {
146 	uint8_t cdd[CDD_DATA_LEN+4] = {0U};
147 	uint32_t i, val = 0U;
148 	uint16_t *phy;
149 	uint8_t buf[16] = {U(0x0)};
150 	uint8_t trr = 0U, tww = 0U, trw = 0U, twr = 0U;
151 	uint8_t rrmax = 0U, wwmax = 0U, rwmax = 0U, wrmax = 0U;
152 	uint8_t tmp = U(0x0);
153 	uint8_t *c =  NULL;
154 
155 	for (i = 0U; i < NUM_OF_DDRC; i++) {
156 
157 		phy = phy_ptr[i];
158 		if (phy == NULL) {
159 			continue;
160 		}
161 
162 		phy_io_write16(phy, t_apbonly |
163 				csr_micro_cont_mux_sel_addr, U(0x0));
164 
165 		read_phy_reg(phy, CDD_VAL_READ_ADDR,
166 				(uint16_t *)&cdd, CDD_DATA_LEN);
167 
168 		phy_io_write16(phy, t_apbonly |
169 				csr_micro_cont_mux_sel_addr, U(0x1));
170 
171 	/* CDD values and address
172 	 *
173 	 *   0x054012    0x24    cdd[0]  CDD[X][X]
174 	 *   0x054012    0x25    cdd[1]  RR[3][2]
175 	 *   0x054013    0x26    cdd[2]  RR[3][1]
176 	 *   0x054013    0x27    cdd[3]  RR[3][0]
177 	 *   0x054014    0x28    cdd[4]  RR[2][3]
178 	 *   0x054014    0x29    cdd[5]  RR[2][1]
179 	 *   0x054015    0x2a    cdd[6]  RR[2][0]
180 	 *   0x054015    0x2b    cdd[7]  RR[1][3]
181 	 *   0x054016    0x2c    cdd[8]  RR[1][2]
182 	 *   0x054016    0x2d    cdd[9]  RR[1][0]
183 	 *   0x054017    0x2e    cdd[10] RR[0][3]
184 	 *   0x054017    0x2f    cdd[11] RR[0][2]
185 	 *   0x054018    0x30    cdd[12] RR[0][1]
186 
187 	 *   0x054018    0x31    cdd[13] WW[3][2]
188 	 *   0x054019    0x32    cdd[14] WW[3][1]
189 	 *   0x054019    0x33    cdd[15] WW[3][0]
190 	 *   0x05401a    0x34    cdd[16] WW[2][3]
191 	 *   0x05401a    0x35    cdd[17] WW[2][1]
192 	 *   0x05401b    0x36    cdd[18] WW[2][0]
193 	 *   0x05401b    0x37    cdd[19] WW[1][3]
194 	 *   0x05401c    0x38    cdd[20] WW[1][2]
195 	 *   0x05401c    0x39    cdd[21] WW[1][0]
196 	 *   0x05401d    0x3a    cdd[22] WW[0][3]
197 	 *   0x05401d    0x3b    cdd[23] WW[0][2]
198 	 *   0x05401e    0x3c    cdd[24] WW[0][1]
199 
200 	 *   0x05401e    0x3d    cdd[25] RW[3][3]
201 	 *   0x05401f    0x3e    cdd[26] RW[3][2]
202 	 *   0x05401f    0x3f    cdd[27] RW[3][1]
203 	 *   0x054020    0x40    cdd[28] RW[3][0]
204 	 *   0x054020    0x41    cdd[29] RW[2][3]
205 	 *   0x054021    0x42    cdd[30] RW[2][2]
206 	 *   0x054021    0x43    cdd[31] RW[2][1]
207 	 *   0x054022    0x44    cdd[32] RW[2][0]
208 	 *   0x054022    0x45    cdd[33] RW[1][3]
209 	 *   0x054023    0x46    cdd[34] RW[1][2]
210 	 *   0x054023    0x47    cdd[35] RW[1][1]
211 	 *   0x054024    0x48    cdd[36] RW[1][0]
212 	 *   0x054024    0x49    cdd[37] RW[0][3]
213 	 *   0x054025    0x4a    cdd[38] RW[0][2]
214 	 *   0x054025    0x4b    cdd[39] RW[0][1]
215 	 *   0x054026    0x4c    cdd[40] RW[0][0]
216 
217 	 *   0x054026    0x4d    cdd[41] WR[3][3]
218 	 *   0x054027    0x4e    cdd[42] WR[3][2]
219 	 *   0x054027    0x4f    cdd[43] WR[3][1]
220 	 *   0x054028    0x50    cdd[44] WR[3][0]
221 	 *   0x054028    0x51    cdd[45] WR[2][3]
222 	 *   0x054029    0x52    cdd[46] WR[2][2]
223 	 *   0x054029    0x53    cdd[47] WR[2][1]
224 	 *   0x05402a    0x54    cdd[48] WR[2][0]
225 	 *   0x05402a    0x55    cdd[49] WR[1][3]
226 	 *   0x05402b    0x56    cdd[50] WR[1][2]
227 	 *   0x05402b    0x57    cdd[51] WR[1][1]
228 	 *   0x05402c    0x58    cdd[52] WR[1][0]
229 	 *   0x05402c    0x59    cdd[53] WR[0][3]
230 	 *   0x05402d    0x5a    cdd[54] WR[0][2]
231 	 *   0x05402d    0x5b    cdd[55] WR[0][1]
232 	 *   0x05402e    0x5c    cdd[56] WR[0][0]
233 	 *   0x05402e    0x5d    cdd[57] CDD[Y][Y]
234 	 */
235 
236 		switch (rank) {
237 		case 1U:
238 			tmp = rwmax;
239 			rwmax = cdd[40];
240 			if (tmp > rwmax) {
241 				rwmax = tmp;
242 			}
243 
244 			break;
245 
246 		case 2U:
247 			buf[0] = cdd[12];
248 			buf[1] = cdd[9];
249 			tmp = rrmax;
250 			rrmax = findmax(buf, 2U);
251 			if (tmp > rrmax) {
252 				rrmax = tmp;
253 			}
254 
255 			buf[0] = cdd[24];
256 			buf[1] = cdd[21];
257 			tmp = wwmax;
258 			wwmax = findmax(buf, 2U);
259 			if (tmp > wwmax) {
260 				wwmax = tmp;
261 			}
262 
263 			buf[0] = cdd[40];
264 			buf[1] = cdd[39];
265 			buf[2] = cdd[36];
266 			buf[3] = cdd[35];
267 			tmp = rwmax;
268 			rwmax = findmax(buf, 4U);
269 			if (tmp > rwmax) {
270 				rwmax = tmp;
271 			}
272 
273 			wrmax = wwmax;
274 
275 			break;
276 
277 		case 4U:
278 			tmp = rrmax;
279 			c = &cdd[1];
280 			rrmax = findmax(c, 12U);
281 			if (tmp > rrmax) {
282 				rrmax = tmp;
283 			}
284 
285 			tmp = wwmax;
286 			c = &cdd[13];
287 			wwmax = findmax(c, 12U);
288 			if (tmp > wwmax) {
289 				wwmax = tmp;
290 			}
291 
292 			tmp = rwmax;
293 			c = &cdd[25];
294 			rwmax = findmax(c, 16U);
295 			if (tmp > rwmax) {
296 				rwmax = tmp;
297 			}
298 
299 			wrmax = wwmax;
300 
301 			break;
302 
303 		}
304 	}
305 
306 	rrmax += 3U;
307 	wwmax += 4U;
308 
309 	if (wwmax > 7U) {
310 		wwmax = 7U;
311 	}
312 
313 	if (rrmax > 7U) {
314 		rrmax = 7U;
315 	}
316 
317 	if (wrmax > U(0xf)) {
318 		wrmax = 0U;
319 	}
320 
321 	if (rwmax > U(0x7)) {
322 		rwmax = U(0x7);
323 	}
324 
325 	val = *tcfg0;
326 	tww = (val >> 24U) & U(0x3);
327 	trr = (val >> 26U) & U(0x3);
328 	twr = (val >> 28U) & U(0x3);
329 	trw = (val >> 30U) & U(0x3);
330 
331 	val = *tcfg4;
332 	tww = tww | (((val >> 8U) & U(0x1)) << 2U);
333 	trr = trr | (((val >> 10U) & U(0x1)) << 2U);
334 	twr = twr | (((val >> 12U) & U(0x1)) << 2U);
335 	trw = trw | (((val >> 14U) & U(0x3)) << 2U);
336 
337 	if (trr > rrmax) {
338 		rrmax = trr;
339 	}
340 
341 	if (tww > wwmax) {
342 		wwmax = tww;
343 	}
344 
345 	if (trw > rwmax) {
346 		rwmax = trw;
347 	}
348 
349 	if (twr > wrmax) {
350 		wrmax = twr;
351 	}
352 
353 	debug("CDD rrmax %x wwmax %x rwmax %x wrmax %x\n",
354 			rrmax, wwmax, rwmax, wrmax);
355 
356 	val = ((wwmax & U(0x3)) << 24U)
357 		| ((rrmax & U(0x3)) << 26U)
358 		| ((wrmax & U(0x3)) << 28U)
359 		| ((rwmax & U(0x3)) << 30U);
360 
361 	*tcfg0 = (*tcfg0 & U(0x00FFFFFF)) | (val);
362 
363 	val = (((wwmax >> 2U) & U(0x1)) << 8U)
364 		| (((rrmax >> 2U) & U(0x1)) << 10U)
365 		| (((wrmax >> 2U) & U(0x1)) << 12U)
366 		| (((rwmax >> 2U) & U(0x3)) << 14U);
367 
368 	*tcfg4 = (*tcfg4 & U(0xffff00ff)) | val;
369 }
370 #endif
371 
372 #ifdef NXP_WARM_BOOT
373 int save_phy_training_values(uint16_t **phy_ptr, uint32_t address_to_store,
374 		uint32_t num_of_phy, int train2d)
375 {
376 	uint16_t *phy = NULL, value = 0x0;
377 	uint32_t size = 1U, num_of_regs = 1U, phy_store = 0U;
378 	int i = 0, j = 0, ret = -EINVAL;
379 
380 	ret = xspi_sector_erase(address_to_store, PHY_ERASE_SIZE);
381 	if (ret != 0) {
382 		return -EINVAL;
383 	}
384 
385 	for (j = 0; j < num_of_phy; j++) {
386 		/* Save training values of all PHYs */
387 		phy = phy_ptr[j];
388 		size = sizeof(training_1D_values);
389 		num_of_regs = ARRAY_SIZE(training_1D_values);
390 
391 		/* Enable access to the internal CSRs */
392 		phy_io_write16(phy, t_apbonly |
393 				csr_micro_cont_mux_sel_addr, 0x0);
394 		/* Enable clocks in case they were disabled. */
395 		phy_io_write16(phy, t_drtub |
396 				csr_ucclk_hclk_enables_addr, 0x3);
397 		if (train2d != 0) {
398 		/* Address to store training values is
399 		 * to be appended for next PHY
400 		 */
401 			phy_store = address_to_store + (j *
402 					(sizeof(training_1D_values) +
403 					 sizeof(training_2D_values)));
404 		} else {
405 			phy_store = address_to_store + (j *
406 					(sizeof(training_1D_values)));
407 		}
408 		debug("Saving 1D Training reg val at: %d\n", phy_store);
409 		for (i = 0; i < num_of_regs; i++) {
410 			value = phy_io_read16(phy, training_1D_values[i].addr);
411 #ifdef DEBUG_WARM_RESET
412 			debug("%d. Reg: %x, value: %x PHY: %p\n", i,
413 					training_1D_values[i].addr, value,
414 					phy_io_addr(phy,
415 						training_1D_values[i].addr));
416 #endif
417 			training_1D_values[i].data = value;
418 		}
419 		/* Storing 1D training values on flash */
420 		ret = xspi_write(phy_store, (void *)training_1D_values, size);
421 		if (train2d != 0) {
422 			phy_store = phy_store+size;
423 			size = sizeof(training_2D_values);
424 			num_of_regs = ARRAY_SIZE(training_2D_values);
425 			debug("Saving 2D Training reg val at:%d\n", phy_store);
426 			for (i = 0; i < num_of_regs; i++) {
427 				value = phy_io_read16(phy,
428 						training_2D_values[i].addr);
429 				training_2D_values[i].data = value;
430 #ifdef DEBUG_WARM_RESET
431 				debug("%d.2D addr:0x%x,val:0x%x,PHY:0x%p\n",
432 						i, training_2D_values[i].addr,
433 						value, phy_io_addr(phy,
434 						training_2D_values[i].addr));
435 #endif
436 			}
437 			/* Storing 2D training values on flash */
438 			ret = xspi_write(phy_store, training_2D_values,
439 					size);
440 		}
441 		/* Disable clocks in case they were disabled. */
442 		phy_io_write16(phy, t_drtub |
443 				csr_ucclk_hclk_enables_addr, 0x0);
444 		/* Disable access to the internal CSRs */
445 		phy_io_write16(phy, t_apbonly |
446 				csr_micro_cont_mux_sel_addr, 0x1);
447 	}
448 	if (ret != 0) {
449 		return -EINVAL;
450 	}
451 
452 	return 0;
453 }
454 
455 int restore_phy_training_values(uint16_t **phy_ptr, uint32_t address_to_restore,
456 		uint32_t num_of_phy, int train2d)
457 {
458 	uint16_t *phy = NULL;
459 	uint32_t size = 1U, num_of_regs = 1U, phy_store = 0U;
460 	int i = 0, j = 0, ret = -EINVAL;
461 
462 	debug("Restoring Training register values\n");
463 	for (j = 0; j < num_of_phy; j++) {
464 		phy = phy_ptr[j];
465 		size = sizeof(training_1D_values);
466 		num_of_regs = ARRAY_SIZE(training_1D_values);
467 		if (train2d != 0) {
468 		/* The address to restore training values is
469 		 * to be appended for next PHY
470 		 */
471 			phy_store = address_to_restore + (j *
472 					(sizeof(training_1D_values) +
473 					 sizeof(training_2D_values)));
474 		} else {
475 			phy_store = address_to_restore + (j *
476 					(sizeof(training_1D_values)));
477 		}
478 		/* Enable access to the internal CSRs */
479 		phy_io_write16(phy, t_apbonly |
480 				csr_micro_cont_mux_sel_addr, 0x0);
481 		/* Enable clocks in case they were disabled. */
482 		phy_io_write16(phy, t_drtub |
483 				csr_ucclk_hclk_enables_addr, 0x3);
484 
485 		/* Reading 1D training values from flash*/
486 		ret = xspi_read(phy_store, (uint32_t *)training_1D_values,
487 				size);
488 		debug("Restoring 1D Training reg val at:%08x\n", phy_store);
489 		for (i = 0; i < num_of_regs; i++) {
490 			phy_io_write16(phy, training_1D_values[i].addr,
491 					training_1D_values[i].data);
492 #ifdef DEBUG_WARM_RESET
493 			debug("%d. Reg: %x, value: %x PHY: %p\n", i,
494 					training_1D_values[i].addr,
495 					training_1D_values[i].data,
496 					phy_io_addr(phy,
497 						training_1D_values[i].addr));
498 #endif
499 		}
500 		if (train2d != 0) {
501 			phy_store = phy_store + size;
502 			size = sizeof(training_2D_values);
503 			num_of_regs = ARRAY_SIZE(training_2D_values);
504 			/* Reading 2D training values from flash */
505 			ret = xspi_read(phy_store,
506 					(uint32_t *)training_2D_values,	size);
507 			debug("Restoring 2D Training reg val at:%08x\n",
508 					phy_store);
509 			for (i = 0; i < num_of_regs; i++) {
510 				phy_io_write16(phy, training_2D_values[i].addr,
511 						training_2D_values[i].data);
512 #ifdef DEBUG_WARM_RESET
513 				debug("%d. Reg: %x, value: %x PHY: %p\n", i,
514 						training_2D_values[i].addr,
515 						training_2D_values[i].data,
516 						phy_io_addr(phy,
517 						training_1D_values[i].addr));
518 #endif
519 			}
520 		}
521 		/* Disable clocks in case they were disabled. */
522 		phy_io_write16(phy, t_drtub |
523 				csr_ucclk_hclk_enables_addr, 0x0);
524 		/* Disable access to the internal CSRs */
525 		phy_io_write16(phy, t_apbonly |
526 				csr_micro_cont_mux_sel_addr, 0x1);
527 	}
528 	if (ret != 0) {
529 		return -EINVAL;
530 	}
531 	return 0;
532 }
533 #endif
534 
535 static void load_pieimage(uint16_t *phy,
536 			  enum dimm_types dimm_type)
537 {
538 	int i;
539 	int size;
540 	const struct pie *image = NULL;
541 
542 	switch (dimm_type) {
543 	case UDIMM:
544 	case SODIMM:
545 	case NODIMM:
546 		image = pie_udimm;
547 		size = ARRAY_SIZE(pie_udimm);
548 		break;
549 	case RDIMM:
550 		image = pie_rdimm;
551 		size = ARRAY_SIZE(pie_rdimm);
552 		break;
553 	case LRDIMM:
554 		image = pie_lrdimm;
555 		size = ARRAY_SIZE(pie_lrdimm);
556 		break;
557 	default:
558 		printf("Unsupported DIMM type\n");
559 		break;
560 	}
561 
562 	if (image != NULL) {
563 		for (i = 0; i < size; i++)
564 			phy_io_write16(phy, image[i].addr, image[i].data);
565 	}
566 }
567 
568 static void prog_acsm_playback(uint16_t *phy,
569 			       const struct input *input, const void *msg)
570 {
571 	int vec;
572 	const struct ddr4r1d *msg_blk;
573 	uint16_t acsmplayback[2][3];
574 	uint32_t f0rc0a;
575 	uint32_t f0rc3x;
576 	uint32_t f0rc5x;
577 
578 	if (input->basic.dimm_type != RDIMM) {
579 		return;
580 	}
581 
582 	msg_blk = msg;
583 	f0rc0a = (msg_blk->f0rc0a_d0 & U(0xf)) | U(0xa0);
584 	f0rc3x = (msg_blk->f0rc3x_d0 & U(0xff)) | U(0x300);
585 	f0rc5x = (input->adv.phy_gen2_umctl_f0rc5x & U(0xff)) | U(0x500);
586 
587 	acsmplayback[0][0] = U(0x3ff) & f0rc0a;
588 	acsmplayback[1][0] = (U(0x1c00) & f0rc0a) >> 10U;
589 	acsmplayback[0][1] = U(0x3ff) & f0rc3x;
590 	acsmplayback[1][1] = (U(0x1c00) & f0rc3x) >> 10U;
591 	acsmplayback[0][2] = U(0x3ff) & f0rc5x;
592 	acsmplayback[1][2] = (U(0x1c00) & f0rc5x) >> 10U;
593 	for (vec = 0; vec < 3; vec++) {
594 		phy_io_write16(phy, t_acsm | (csr_acsm_playback0x0_addr +
595 			       (vec << 1)), acsmplayback[0][vec]);
596 		phy_io_write16(phy, t_acsm | (csr_acsm_playback1x0_addr +
597 			       (vec << 1)), acsmplayback[1][vec]);
598 	}
599 }
600 
601 static void prog_acsm_ctr(uint16_t *phy,
602 			  const struct input *input)
603 {
604 	if (input->basic.dimm_type != RDIMM) {
605 		return;
606 	}
607 
608 	phy_io_write16(phy, t_acsm | csr_acsm_ctrl13_addr,
609 		       0xf << csr_acsm_cke_enb_lsb);
610 
611 	phy_io_write16(phy, t_acsm | csr_acsm_ctrl0_addr,
612 		       csr_acsm_par_mode_mask | csr_acsm_2t_mode_mask);
613 }
614 
615 static void prog_cal_rate_run(uint16_t *phy,
616 			  const struct input *input)
617 {
618 	int cal_rate;
619 	int cal_interval;
620 	int cal_once;
621 	uint32_t addr;
622 
623 	cal_interval = input->adv.cal_interval;
624 	cal_once = input->adv.cal_once;
625 	cal_rate = 0x1 << csr_cal_run_lsb		|
626 			cal_once << csr_cal_once_lsb	|
627 			cal_interval << csr_cal_interval_lsb;
628 	addr = t_master | csr_cal_rate_addr;
629 	phy_io_write16(phy, addr, cal_rate);
630 }
631 
632 static void prog_seq0bdly0(uint16_t *phy,
633 		    const struct input *input)
634 {
635 	int ps_count[4];
636 	int frq;
637 	uint32_t addr;
638 	int lower_freq_opt = 0;
639 
640 	__unused const soc_info_t *soc_info;
641 
642 	frq = input->basic.frequency >> 1;
643 	ps_count[0] = frq >> 3; /* 0.5 * frq / 4*/
644 	if (input->basic.frequency < 400) {
645 		lower_freq_opt = (input->basic.dimm_type == RDIMM) ? 7 : 3;
646 	} else if (input->basic.frequency < 533) {
647 		lower_freq_opt = (input->basic.dimm_type == RDIMM) ? 14 : 11;
648 	}
649 
650 	/* 1.0 * frq / 4 - lower_freq */
651 	ps_count[1] = (frq >> 2) - lower_freq_opt;
652 	ps_count[2] = (frq << 1) +  (frq >> 1); /* 10.0 * frq / 4 */
653 
654 #ifdef DDR_PLL_FIX
655 	soc_info = get_soc_info();
656 	if (soc_info->svr_reg.bf.maj_ver == 1) {
657 		ps_count[0] = 0x520; /* seq0bdly0 */
658 		ps_count[1] = 0xa41; /* seq0bdly1 */
659 		ps_count[2] = 0x668a; /* seq0bdly2 */
660 	}
661 #endif
662 	if (frq > 266) {
663 		ps_count[3] = 44;
664 	} else if (frq > 200) {
665 		ps_count[3] = 33;
666 	} else {
667 		ps_count[3] = 16;
668 	}
669 
670 	addr = t_master | csr_seq0bdly0_addr;
671 	phy_io_write16(phy, addr, ps_count[0]);
672 
673 	debug("seq0bdly0 = 0x%x\n", phy_io_read16(phy, addr));
674 
675 	addr = t_master | csr_seq0bdly1_addr;
676 	phy_io_write16(phy, addr, ps_count[1]);
677 
678 	debug("seq0bdly1 = 0x%x\n", phy_io_read16(phy, addr));
679 
680 	addr = t_master | csr_seq0bdly2_addr;
681 	phy_io_write16(phy, addr, ps_count[2]);
682 
683 	debug("seq0bdly2 = 0x%x\n", phy_io_read16(phy, addr));
684 
685 	addr = t_master | csr_seq0bdly3_addr;
686 	phy_io_write16(phy, addr, ps_count[3]);
687 
688 	debug("seq0bdly3 = 0x%x\n", phy_io_read16(phy, addr));
689 }
690 
691 /* Only RDIMM requires msg_blk */
692 static void i_load_pie(uint16_t **phy_ptr,
693 		       const struct input *input,
694 		       const void *msg)
695 {
696 	int i;
697 	uint16_t *phy;
698 
699 	for (i = 0; i < NUM_OF_DDRC; i++) {
700 		phy = phy_ptr[i];
701 		if (phy == NULL) {
702 			continue;
703 		}
704 
705 		phy_io_write16(phy,
706 			       t_apbonly | csr_micro_cont_mux_sel_addr,
707 			       0U);
708 
709 		load_pieimage(phy, input->basic.dimm_type);
710 
711 		prog_seq0bdly0(phy, input);
712 		phy_io_write16(phy, t_initeng | csr_seq0bdisable_flag0_addr,
713 			       U(0x0000));
714 		phy_io_write16(phy, t_initeng | csr_seq0bdisable_flag1_addr,
715 			       U(0x0173));
716 		phy_io_write16(phy, t_initeng | csr_seq0bdisable_flag2_addr,
717 			       U(0x0060));
718 		phy_io_write16(phy, t_initeng | csr_seq0bdisable_flag3_addr,
719 			       U(0x6110));
720 		phy_io_write16(phy, t_initeng | csr_seq0bdisable_flag4_addr,
721 			       U(0x2152));
722 		phy_io_write16(phy, t_initeng | csr_seq0bdisable_flag5_addr,
723 			       U(0xdfbd));
724 		phy_io_write16(phy, t_initeng | csr_seq0bdisable_flag6_addr,
725 			       input->basic.dimm_type == RDIMM &&
726 			       input->adv.phy_gen2_umctl_opt == 1U ?
727 			       U(0x6000) : U(0xffff));
728 		phy_io_write16(phy, t_initeng | csr_seq0bdisable_flag7_addr,
729 			       U(0x6152));
730 		prog_acsm_playback(phy, input, msg);		/* rdimm */
731 		prog_acsm_ctr(phy, input);			/* rdimm */
732 
733 		phy_io_write16(phy, t_master | csr_cal_zap_addr, U(0x1));
734 		prog_cal_rate_run(phy, input);
735 
736 		phy_io_write16(phy, t_drtub | csr_ucclk_hclk_enables_addr,
737 			       input->basic.dimm_type == RDIMM ? U(0x2) : 0U);
738 
739 		phy_io_write16(phy, t_apbonly | csr_micro_cont_mux_sel_addr, 1U);
740 	}
741 }
742 
743 static void phy_gen2_init_input(struct input *input)
744 {
745 	int i;
746 
747 	input->adv.dram_byte_swap		= 0;
748 	input->adv.ext_cal_res_val		= 0;
749 	input->adv.tx_slew_rise_dq		= 0xf;
750 	input->adv.tx_slew_fall_dq		= 0xf;
751 	input->adv.tx_slew_rise_ac		= 0xf;
752 	input->adv.tx_slew_fall_ac		= 0xf;
753 	input->adv.mem_alert_en			= 0;
754 	input->adv.mem_alert_puimp		= 5;
755 	input->adv.mem_alert_vref_level		= 0x29;
756 	input->adv.mem_alert_sync_bypass	= 0;
757 	input->adv.cal_interval			= 0x9;
758 	input->adv.cal_once			= 0;
759 	input->adv.dis_dyn_adr_tri		= 0;
760 	input->adv.is2ttiming			= 0;
761 	input->adv.d4rx_preamble_length		= 0;
762 	input->adv.d4tx_preamble_length		= 0;
763 
764 	for (i = 0; i < 7; i++) {
765 		debug("mr[%d] = 0x%x\n", i, input->mr[i]);
766 	}
767 
768 	debug("input->cs_d0 = 0x%x\n", input->cs_d0);
769 	debug("input->cs_d1 = 0x%x\n", input->cs_d1);
770 	debug("input->mirror = 0x%x\n", input->mirror);
771 	debug("PHY ODT impedance = %d ohm\n", input->adv.odtimpedance);
772 	debug("PHY DQ driver impedance = %d ohm\n", input->adv.tx_impedance);
773 	debug("PHY Addr driver impedance = %d ohm\n", input->adv.atx_impedance);
774 
775 	for (i = 0; i < 4; i++) {
776 		debug("odt[%d] = 0x%x\n", i, input->odt[i]);
777 	}
778 
779 	if (input->basic.dimm_type == RDIMM) {
780 		for (i = 0; i < 16; i++) {
781 			debug("input->rcw[%d] = 0x%x\n", i, input->rcw[i]);
782 		}
783 		debug("input->rcw3x = 0x%x\n", input->rcw3x);
784 	}
785 }
786 
787 /*
788  * All protocols share the same base structure of message block.
789  * RDIMM and LRDIMM have more entries defined than UDIMM.
790  * Create message blocks for 1D and 2D training.
791  * Update len with message block size.
792  */
793 static int phy_gen2_msg_init(void *msg_1d,
794 			     void *msg_2d,
795 			     const struct input *input)
796 {
797 	struct ddr4u1d *msg_blk = msg_1d;
798 	struct ddr4u2d *msg_blk_2d = msg_2d;
799 	struct ddr4r1d *msg_blk_r;
800 	struct ddr4lr1d *msg_blk_lr;
801 
802 	switch (input->basic.dimm_type) {
803 	case UDIMM:
804 	case SODIMM:
805 	case NODIMM:
806 		msg_blk->dram_type	= U(0x2);
807 		break;
808 	case RDIMM:
809 		msg_blk->dram_type	= U(0x4);
810 		break;
811 	case LRDIMM:
812 		msg_blk->dram_type	= U(0x5);
813 		break;
814 	default:
815 		ERROR("Unsupported DIMM type\n");
816 		return -EINVAL;
817 	}
818 	msg_blk->pstate			= 0U;
819 
820 	/*Enable quickRd2D, a substage of read deskew, to 1D training.*/
821 	msg_blk->reserved00             = U(0x20);
822 
823 	/*Enable High-Effort WrDQ1D.*/
824 	msg_blk->reserved00             |= U(0x40);
825 
826 	/* Enable 1D extra effort training.*/
827 	msg_blk->reserved1c[3]		= U(0x3);
828 
829 	if (input->basic.dimm_type == LRDIMM) {
830 		msg_blk->sequence_ctrl	= U(0x3f1f);
831 	} else {
832 		msg_blk->sequence_ctrl	= U(0x031f);
833 	}
834 	msg_blk->phy_config_override	= 0U;
835 #ifdef DDR_PHY_DEBUG
836 	msg_blk->hdt_ctrl		= U(0x5);
837 #else
838 	msg_blk->hdt_ctrl		= U(0xc9);
839 #endif
840 	msg_blk->msg_misc		= U(0x0);
841 	msg_blk->dfimrlmargin		= U(0x1);
842 	msg_blk->phy_vref		= input->vref ? input->vref : U(0x61);
843 	msg_blk->cs_present		= input->cs_d0 | input->cs_d1;
844 	msg_blk->cs_present_d0		= input->cs_d0;
845 	msg_blk->cs_present_d1		= input->cs_d1;
846 	if (input->mirror != 0) {
847 		msg_blk->addr_mirror	= U(0x0a);	/* odd CS are mirrored */
848 	}
849 	msg_blk->share2dvref_result	= 1U;
850 
851 	msg_blk->acsm_odt_ctrl0		= input->odt[0];
852 	msg_blk->acsm_odt_ctrl1		= input->odt[1];
853 	msg_blk->acsm_odt_ctrl2		= input->odt[2];
854 	msg_blk->acsm_odt_ctrl3		= input->odt[3];
855 	msg_blk->enabled_dqs = (input->basic.num_active_dbyte_dfi0 +
856 				input->basic.num_active_dbyte_dfi1) * 8;
857 	msg_blk->x16present		= input->basic.dram_data_width == 0x10 ?
858 					  msg_blk->cs_present : 0;
859 	msg_blk->d4misc			= U(0x1);
860 	msg_blk->cs_setup_gddec		= U(0x1);
861 	msg_blk->rtt_nom_wr_park0	= 0U;
862 	msg_blk->rtt_nom_wr_park1	= 0U;
863 	msg_blk->rtt_nom_wr_park2	= 0U;
864 	msg_blk->rtt_nom_wr_park3	= 0U;
865 	msg_blk->rtt_nom_wr_park4	= 0U;
866 	msg_blk->rtt_nom_wr_park5	= 0U;
867 	msg_blk->rtt_nom_wr_park6	= 0U;
868 	msg_blk->rtt_nom_wr_park7	= 0U;
869 	msg_blk->mr0			= input->mr[0];
870 	msg_blk->mr1			= input->mr[1];
871 	msg_blk->mr2			= input->mr[2];
872 	msg_blk->mr3			= input->mr[3];
873 	msg_blk->mr4			= input->mr[4];
874 	msg_blk->mr5			= input->mr[5];
875 	msg_blk->mr6			= input->mr[6];
876 	if ((msg_blk->mr4 & U(0x1c0)) != 0U) {
877 		ERROR("Setting DRAM CAL mode is not supported\n");
878 	}
879 
880 	msg_blk->alt_cas_l		= 0U;
881 	msg_blk->alt_wcas_l		= 0U;
882 
883 	msg_blk->dramfreq		= input->basic.frequency * 2U;
884 	msg_blk->pll_bypass_en		= input->basic.pll_bypass;
885 	msg_blk->dfi_freq_ratio		= input->basic.dfi_freq_ratio == 0U ? 1U :
886 					  input->basic.dfi_freq_ratio == 1U ? 2U :
887 					  4U;
888 	msg_blk->bpznres_val		= input->adv.ext_cal_res_val;
889 	msg_blk->disabled_dbyte		= 0U;
890 
891 	debug("msg_blk->dram_type = 0x%x\n", msg_blk->dram_type);
892 	debug("msg_blk->sequence_ctrl = 0x%x\n", msg_blk->sequence_ctrl);
893 	debug("msg_blk->phy_cfg = 0x%x\n", msg_blk->phy_cfg);
894 	debug("msg_blk->x16present = 0x%x\n", msg_blk->x16present);
895 	debug("msg_blk->dramfreq = 0x%x\n", msg_blk->dramfreq);
896 	debug("msg_blk->pll_bypass_en = 0x%x\n", msg_blk->pll_bypass_en);
897 	debug("msg_blk->dfi_freq_ratio = 0x%x\n", msg_blk->dfi_freq_ratio);
898 	debug("msg_blk->phy_odt_impedance = 0x%x\n",
899 						msg_blk->phy_odt_impedance);
900 	debug("msg_blk->phy_drv_impedance = 0x%x\n",
901 						msg_blk->phy_drv_impedance);
902 	debug("msg_blk->bpznres_val = 0x%x\n", msg_blk->bpznres_val);
903 	debug("msg_blk->enabled_dqs = 0x%x\n", msg_blk->enabled_dqs);
904 	debug("msg_blk->acsm_odt_ctrl0 = 0x%x\n", msg_blk->acsm_odt_ctrl0);
905 	debug("msg_blk->acsm_odt_ctrl1 = 0x%x\n", msg_blk->acsm_odt_ctrl1);
906 	debug("msg_blk->acsm_odt_ctrl2 = 0x%x\n", msg_blk->acsm_odt_ctrl2);
907 	debug("msg_blk->acsm_odt_ctrl3 = 0x%x\n", msg_blk->acsm_odt_ctrl3);
908 
909 	/* RDIMM only */
910 	if (input->basic.dimm_type == RDIMM ||
911 	    input->basic.dimm_type == LRDIMM) {
912 		msg_blk_r = (struct ddr4r1d *)msg_blk;
913 		if (msg_blk_r->cs_present_d0 != 0U) {
914 			msg_blk_r->f0rc00_d0 = input->rcw[0];
915 			msg_blk_r->f0rc01_d0 = input->rcw[1];
916 			msg_blk_r->f0rc02_d0 = input->rcw[2];
917 			msg_blk_r->f0rc03_d0 = input->rcw[3];
918 			msg_blk_r->f0rc04_d0 = input->rcw[4];
919 			msg_blk_r->f0rc05_d0 = input->rcw[5];
920 			msg_blk_r->f0rc06_d0 = input->rcw[6];
921 			msg_blk_r->f0rc07_d0 = input->rcw[7];
922 			msg_blk_r->f0rc08_d0 = input->rcw[8];
923 			msg_blk_r->f0rc09_d0 = input->rcw[9];
924 			msg_blk_r->f0rc0a_d0 = input->rcw[10];
925 			msg_blk_r->f0rc0b_d0 = input->rcw[11];
926 			msg_blk_r->f0rc0c_d0 = input->rcw[12];
927 			msg_blk_r->f0rc0d_d0 = input->rcw[13];
928 			msg_blk_r->f0rc0e_d0 = input->rcw[14];
929 			msg_blk_r->f0rc0f_d0 = input->rcw[15];
930 			msg_blk_r->f0rc3x_d0 = input->rcw3x;
931 		}
932 		if (msg_blk_r->cs_present_d1 != 0) {
933 			msg_blk_r->f0rc00_d1 = input->rcw[0];
934 			msg_blk_r->f0rc01_d1 = input->rcw[1];
935 			msg_blk_r->f0rc02_d1 = input->rcw[2];
936 			msg_blk_r->f0rc03_d1 = input->rcw[3];
937 			msg_blk_r->f0rc04_d1 = input->rcw[4];
938 			msg_blk_r->f0rc05_d1 = input->rcw[5];
939 			msg_blk_r->f0rc06_d1 = input->rcw[6];
940 			msg_blk_r->f0rc07_d1 = input->rcw[7];
941 			msg_blk_r->f0rc08_d1 = input->rcw[8];
942 			msg_blk_r->f0rc09_d1 = input->rcw[9];
943 			msg_blk_r->f0rc0a_d1 = input->rcw[10];
944 			msg_blk_r->f0rc0b_d1 = input->rcw[11];
945 			msg_blk_r->f0rc0c_d1 = input->rcw[12];
946 			msg_blk_r->f0rc0d_d1 = input->rcw[13];
947 			msg_blk_r->f0rc0e_d1 = input->rcw[14];
948 			msg_blk_r->f0rc0f_d1 = input->rcw[15];
949 			msg_blk_r->f0rc3x_d1 = input->rcw3x;
950 		}
951 		if (input->basic.dimm_type == LRDIMM) {
952 			msg_blk_lr = (struct ddr4lr1d *)msg_blk;
953 			msg_blk_lr->bc0a_d0 = msg_blk_lr->f0rc0a_d0;
954 			msg_blk_lr->bc0a_d1 = msg_blk_lr->f0rc0a_d1;
955 			msg_blk_lr->f0bc6x_d0 = msg_blk_lr->f0rc3x_d0;
956 			msg_blk_lr->f0bc6x_d1 = msg_blk_lr->f0rc3x_d1;
957 		}
958 	}
959 
960 	/* below is different for 1D and 2D message block */
961 	if (input->basic.train2d != 0) {
962 		memcpy(msg_blk_2d, msg_blk, sizeof(struct ddr4u1d));
963 		/*High-Effort WrDQ1D is applicable to 2D traning also*/
964 		msg_blk_2d->reserved00          |= U(0x40);
965 		msg_blk_2d->sequence_ctrl	= U(0x0061);
966 		msg_blk_2d->rx2d_train_opt	= 0U;
967 		msg_blk_2d->tx2d_train_opt	= 0U;
968 		msg_blk_2d->share2dvref_result	= 1U;
969 		msg_blk_2d->delay_weight2d	= U(0x20);
970 		msg_blk_2d->voltage_weight2d	= U(0x80);
971 		debug("rx2d_train_opt %d, tx2d_train_opt %d\n",
972 				msg_blk_2d->rx2d_train_opt,
973 				msg_blk_2d->tx2d_train_opt);
974 	}
975 
976 	msg_blk->phy_cfg = (((msg_blk->mr3 & U(0x8)) != 0U) ||
977 				((msg_blk_2d->mr3 & 0x8) != 0U)) ? 0U
978 				: input->adv.is2ttiming;
979 
980 	return 0;
981 }
982 
983 static void prog_tx_pre_drv_mode(uint16_t *phy,
984 				 const struct input *input)
985 {
986 	int lane, byte, b_addr, c_addr, p_addr;
987 	int tx_slew_rate, tx_pre_p, tx_pre_n;
988 	int tx_pre_drv_mode = 0x2;
989 	uint32_t addr;
990 
991 	/* Program TxPreDrvMode with 0x2 */
992 	/* FIXME: TxPreDrvMode depends on DramType? */
993 	tx_pre_p = input->adv.tx_slew_rise_dq;
994 	tx_pre_n = input->adv.tx_slew_fall_dq;
995 	tx_slew_rate = tx_pre_drv_mode << csr_tx_pre_drv_mode_lsb	|
996 		     tx_pre_p << csr_tx_pre_p_lsb			|
997 		     tx_pre_n << csr_tx_pre_n_lsb;
998 	p_addr = 0;
999 	for (byte = 0; byte < input->basic.num_dbyte; byte++) {
1000 		c_addr = byte << 12;
1001 		for (lane = 0; lane <= 1; lane++) {
1002 			b_addr = lane << 8;
1003 			addr = p_addr | t_dbyte | c_addr | b_addr |
1004 					csr_tx_slew_rate_addr;
1005 			phy_io_write16(phy, addr, tx_slew_rate);
1006 		}
1007 	}
1008 }
1009 
1010 static void prog_atx_pre_drv_mode(uint16_t *phy,
1011 				  const struct input *input)
1012 {
1013 	int anib, c_addr;
1014 	int atx_slew_rate, atx_pre_p, atx_pre_n, atx_pre_drv_mode,
1015 		ck_anib_inst[2];
1016 	uint32_t addr;
1017 
1018 	atx_pre_n = input->adv.tx_slew_fall_ac;
1019 	atx_pre_p = input->adv.tx_slew_rise_ac;
1020 
1021 	if (input->basic.num_anib == 8) {
1022 		ck_anib_inst[0] = 1;
1023 		ck_anib_inst[1] = 1;
1024 	} else if (input->basic.num_anib == 10 || input->basic.num_anib == 12 ||
1025 	    input->basic.num_anib == 13) {
1026 		ck_anib_inst[0] = 4;
1027 		ck_anib_inst[1] = 5;
1028 	} else {
1029 		ERROR("Invalid number of aNIBs: %d\n", input->basic.num_anib);
1030 		return;
1031 	}
1032 
1033 	for (anib = 0; anib < input->basic.num_anib; anib++) {
1034 		c_addr = anib << 12;
1035 		if (anib == ck_anib_inst[0] || anib == ck_anib_inst[1]) {
1036 			atx_pre_drv_mode = 0;
1037 		} else {
1038 			atx_pre_drv_mode = 3;
1039 		}
1040 		atx_slew_rate = atx_pre_drv_mode << csr_atx_pre_drv_mode_lsb |
1041 				atx_pre_n << csr_atx_pre_n_lsb		     |
1042 				atx_pre_p << csr_atx_pre_p_lsb;
1043 		addr = t_anib | c_addr | csr_atx_slew_rate_addr;
1044 		phy_io_write16(phy, addr, atx_slew_rate);
1045 	}
1046 }
1047 
1048 static void prog_enable_cs_multicast(uint16_t *phy,
1049 				     const struct input *input)
1050 {
1051 	uint32_t addr = t_master | csr_enable_cs_multicast_addr;
1052 
1053 	if (input->basic.dimm_type != RDIMM &&
1054 	    input->basic.dimm_type != LRDIMM) {
1055 		return;
1056 	}
1057 
1058 	phy_io_write16(phy, addr, input->adv.cast_cs_to_cid);
1059 }
1060 
1061 static void prog_dfi_rd_data_cs_dest_map(uint16_t *phy,
1062 					 unsigned int ip_rev,
1063 					 const struct input *input,
1064 					 const struct ddr4lr1d *msg)
1065 {
1066 	const struct ddr4lr1d *msg_blk;
1067 	uint16_t dfi_xxdestm0 = 0U;
1068 	uint16_t dfi_xxdestm1 = 0U;
1069 	uint16_t dfi_xxdestm2 = 0U;
1070 	uint16_t dfi_xxdestm3 = 0U;
1071 	uint16_t dfi_rd_data_cs_dest_map;
1072 	uint16_t dfi_wr_data_cs_dest_map;
1073 	__unused const soc_info_t *soc_info;
1074 
1075 #ifdef ERRATA_DDR_A011396
1076 	/* Only apply to DDRC 5.05.00 */
1077 	soc_info = get_soc_info();
1078 	if ((soc_info->svr_reg.bf.maj_ver == 1U) && (ip_rev == U(0x50500))) {
1079 		phy_io_write16(phy,
1080 				t_master | csr_dfi_rd_data_cs_dest_map_addr,
1081 				0U);
1082 		return;
1083 	}
1084 #endif
1085 
1086 	msg_blk = msg;
1087 
1088 	switch (input->basic.dimm_type) {
1089 	case UDIMM:
1090 	case SODIMM:
1091 	case NODIMM:
1092 		if ((msg_blk->msg_misc & U(0x40)) != 0U) {
1093 			dfi_rd_data_cs_dest_map = U(0xa0);
1094 			dfi_wr_data_cs_dest_map = U(0xa0);
1095 
1096 			phy_io_write16(phy,
1097 				t_master | csr_dfi_rd_data_cs_dest_map_addr,
1098 				dfi_rd_data_cs_dest_map);
1099 			phy_io_write16(phy,
1100 				t_master | csr_dfi_wr_data_cs_dest_map_addr,
1101 				dfi_wr_data_cs_dest_map);
1102 		}
1103 		break;
1104 	case LRDIMM:
1105 		if (msg->cs_present_d1 != 0U) {
1106 			dfi_xxdestm2 = 1U;
1107 			dfi_xxdestm3 = 1U;
1108 		}
1109 
1110 		dfi_rd_data_cs_dest_map =
1111 			dfi_xxdestm0 << csr_dfi_rd_destm0_lsb	|
1112 			dfi_xxdestm1 << csr_dfi_rd_destm1_lsb	|
1113 			dfi_xxdestm2 << csr_dfi_rd_destm2_lsb	|
1114 			dfi_xxdestm3 << csr_dfi_rd_destm3_lsb;
1115 		dfi_wr_data_cs_dest_map =
1116 			dfi_xxdestm0 << csr_dfi_wr_destm0_lsb	|
1117 			dfi_xxdestm1 << csr_dfi_wr_destm1_lsb	|
1118 			dfi_xxdestm2 << csr_dfi_wr_destm2_lsb	|
1119 			dfi_xxdestm3 << csr_dfi_wr_destm3_lsb;
1120 		phy_io_write16(phy, t_master | csr_dfi_rd_data_cs_dest_map_addr,
1121 				dfi_rd_data_cs_dest_map);
1122 		phy_io_write16(phy, t_master | csr_dfi_wr_data_cs_dest_map_addr,
1123 				dfi_wr_data_cs_dest_map);
1124 
1125 		break;
1126 	default:
1127 		break;
1128 	}
1129 }
1130 
1131 static void prog_pll_ctrl(uint16_t *phy,
1132 			   const struct input *input)
1133 {
1134 	uint32_t addr;
1135 	int pll_ctrl1 = 0x21; /* 000100001b */
1136 	int pll_ctrl4 = 0x17f; /* 101111111b */
1137 	int pll_test_mode = 0x24; /* 00100100b */
1138 
1139 	addr = t_master | csr_pll_ctrl1_addr;
1140 	phy_io_write16(phy, addr, pll_ctrl1);
1141 
1142 	debug("pll_ctrl1 = 0x%x\n", phy_io_read16(phy, addr));
1143 
1144 	addr = t_master | csr_pll_test_mode_addr;
1145 	phy_io_write16(phy, addr, pll_test_mode);
1146 
1147 	debug("pll_test_mode = 0x%x\n", phy_io_read16(phy, addr));
1148 
1149 	addr = t_master | csr_pll_ctrl4_addr;
1150 	phy_io_write16(phy, addr, pll_ctrl4);
1151 
1152 	debug("pll_ctrl4 = 0x%x\n", phy_io_read16(phy, addr));
1153 }
1154 
1155 static void prog_pll_ctrl2(uint16_t *phy,
1156 			   const struct input *input)
1157 {
1158 	int pll_ctrl2;
1159 	uint32_t addr = t_master | csr_pll_ctrl2_addr;
1160 
1161 	if (input->basic.frequency / 2 < 235) {
1162 		pll_ctrl2 = 0x7;
1163 	} else if (input->basic.frequency / 2 < 313) {
1164 		pll_ctrl2 = 0x6;
1165 	} else if (input->basic.frequency / 2 < 469) {
1166 		pll_ctrl2 = 0xb;
1167 	} else if (input->basic.frequency / 2 < 625) {
1168 		pll_ctrl2 = 0xa;
1169 	} else if (input->basic.frequency / 2 < 938) {
1170 		pll_ctrl2 = 0x19;
1171 	} else if (input->basic.frequency / 2 < 1067) {
1172 		pll_ctrl2 = 0x18;
1173 	} else {
1174 		pll_ctrl2 = 0x19;
1175 	}
1176 
1177 	phy_io_write16(phy, addr, pll_ctrl2);
1178 
1179 	debug("pll_ctrl2 = 0x%x\n", phy_io_read16(phy, addr));
1180 }
1181 
1182 static void prog_dll_lck_param(uint16_t *phy, const struct input *input)
1183 {
1184 	uint32_t addr = t_master | csr_dll_lockparam_addr;
1185 
1186 	phy_io_write16(phy, addr, U(0x212));
1187 	debug("dll_lck_param = 0x%x\n", phy_io_read16(phy, addr));
1188 }
1189 
1190 static void prog_dll_gain_ctl(uint16_t *phy, const struct input *input)
1191 {
1192 	uint32_t addr = t_master | csr_dll_gain_ctl_addr;
1193 
1194 	phy_io_write16(phy, addr, U(0x61));
1195 	debug("dll_gain_ctl = 0x%x\n", phy_io_read16(phy, addr));
1196 }
1197 
1198 static void prog_pll_pwr_dn(uint16_t *phy,
1199 			   const struct input *input)
1200 {
1201 	uint32_t addr;
1202 
1203 	addr = t_master | csr_pll_pwr_dn_addr;
1204 	phy_io_write16(phy, addr, 0U);
1205 
1206 	debug("pll_pwrdn = 0x%x\n", phy_io_read16(phy, addr));
1207 }
1208 
1209 static void prog_ard_ptr_init_val(uint16_t *phy,
1210 				  const struct input *input)
1211 {
1212 	int ard_ptr_init_val;
1213 	uint32_t addr = t_master | csr_ard_ptr_init_val_addr;
1214 
1215 	if (input->basic.frequency >= 933) {
1216 		ard_ptr_init_val = 0x2;
1217 	} else {
1218 		ard_ptr_init_val = 0x1;
1219 	}
1220 
1221 	phy_io_write16(phy, addr, ard_ptr_init_val);
1222 }
1223 
1224 static void prog_dqs_preamble_control(uint16_t *phy,
1225 				      const struct input *input)
1226 {
1227 	int data;
1228 	uint32_t addr = t_master | csr_dqs_preamble_control_addr;
1229 	const int wdqsextension = 0;
1230 	const int lp4sttc_pre_bridge_rx_en = 0;
1231 	const int lp4postamble_ext = 0;
1232 	const int lp4tgl_two_tck_tx_dqs_pre = 0;
1233 	const int position_dfe_init = 2;
1234 	const int dll_rx_preamble_mode = 1;
1235 	int two_tck_tx_dqs_pre = input->adv.d4tx_preamble_length;
1236 	int two_tck_rx_dqs_pre = input->adv.d4rx_preamble_length;
1237 
1238 	data = wdqsextension << csr_wdqsextension_lsb			|
1239 	       lp4sttc_pre_bridge_rx_en << csr_lp4sttc_pre_bridge_rx_en_lsb |
1240 	       lp4postamble_ext << csr_lp4postamble_ext_lsb		|
1241 	       lp4tgl_two_tck_tx_dqs_pre << csr_lp4tgl_two_tck_tx_dqs_pre_lsb |
1242 	       position_dfe_init << csr_position_dfe_init_lsb		|
1243 	       two_tck_tx_dqs_pre << csr_two_tck_tx_dqs_pre_lsb		|
1244 	       two_tck_rx_dqs_pre << csr_two_tck_rx_dqs_pre_lsb;
1245 	phy_io_write16(phy, addr, data);
1246 
1247 	data = dll_rx_preamble_mode << csr_dll_rx_preamble_mode_lsb;
1248 	addr = t_master | csr_dbyte_dll_mode_cntrl_addr;
1249 	phy_io_write16(phy, addr, data);
1250 }
1251 
1252 static void prog_proc_odt_time_ctl(uint16_t *phy,
1253 				   const struct input *input)
1254 {
1255 	int proc_odt_time_ctl;
1256 	uint32_t addr = t_master | csr_proc_odt_time_ctl_addr;
1257 
1258 	if (input->adv.wdqsext != 0) {
1259 		proc_odt_time_ctl = 0x3;
1260 	} else if (input->basic.frequency <= 933) {
1261 		proc_odt_time_ctl = 0xa;
1262 	} else if (input->basic.frequency <= 1200) {
1263 		if (input->adv.d4rx_preamble_length == 1) {
1264 			proc_odt_time_ctl = 0x2;
1265 		} else {
1266 			proc_odt_time_ctl = 0x6;
1267 		}
1268 	} else {
1269 		if (input->adv.d4rx_preamble_length == 1) {
1270 			proc_odt_time_ctl = 0x3;
1271 		} else {
1272 			proc_odt_time_ctl = 0x7;
1273 		}
1274 	}
1275 	phy_io_write16(phy, addr, proc_odt_time_ctl);
1276 }
1277 
1278 static const struct impedance_mapping map[] = {
1279 	{	29,	0x3f	},
1280 	{	31,	0x3e	},
1281 	{	33,	0x3b	},
1282 	{	36,	0x3a	},
1283 	{	39,	0x39	},
1284 	{	42,	0x38	},
1285 	{	46,	0x1b	},
1286 	{	51,	0x1a	},
1287 	{	57,	0x19	},
1288 	{	64,	0x18	},
1289 	{	74,	0x0b	},
1290 	{	88,	0x0a	},
1291 	{	108,	0x09	},
1292 	{	140,	0x08	},
1293 	{	200,	0x03	},
1294 	{	360,	0x02	},
1295 	{	481,	0x01	},
1296 	{}
1297 };
1298 
1299 static int map_impedance(int strength)
1300 {
1301 	const struct impedance_mapping *tbl = map;
1302 	int val = 0;
1303 
1304 	if (strength == 0) {
1305 		return 0;
1306 	}
1307 
1308 	while (tbl->ohm != 0U) {
1309 		if (strength < tbl->ohm) {
1310 			val = tbl->code;
1311 			break;
1312 		}
1313 		tbl++;
1314 	}
1315 
1316 	return val;
1317 }
1318 
1319 static int map_odtstren_p(int strength, int hard_macro_ver)
1320 {
1321 	int val = -1;
1322 
1323 	if (hard_macro_ver == 4) {
1324 		if (strength == 0) {
1325 			val = 0;
1326 		} else if (strength == 120) {
1327 			val = 0x8;
1328 		} else if (strength == 60) {
1329 			val = 0x18;
1330 		} else if (strength == 40) {
1331 			val = 0x38;
1332 		} else {
1333 			printf("error: unsupported ODTStrenP %d\n", strength);
1334 		}
1335 	} else {
1336 		val = map_impedance(strength);
1337 	}
1338 
1339 	return val;
1340 }
1341 
1342 static void prog_tx_odt_drv_stren(uint16_t *phy,
1343 				  const struct input *input)
1344 {
1345 	int lane, byte, b_addr, c_addr;
1346 	int tx_odt_drv_stren;
1347 	int odtstren_p, odtstren_n;
1348 	uint32_t addr;
1349 
1350 	odtstren_p = map_odtstren_p(input->adv.odtimpedance,
1351 				input->basic.hard_macro_ver);
1352 	if (odtstren_p < 0) {
1353 		return;
1354 	}
1355 
1356 	odtstren_n = 0;	/* always high-z */
1357 	tx_odt_drv_stren = odtstren_n << csr_odtstren_n_lsb | odtstren_p;
1358 	for (byte = 0; byte < input->basic.num_dbyte; byte++) {
1359 		c_addr = byte << 12;
1360 		for (lane = 0; lane <= 1; lane++) {
1361 			b_addr = lane << 8;
1362 			addr = t_dbyte | c_addr | b_addr |
1363 				csr_tx_odt_drv_stren_addr;
1364 			phy_io_write16(phy, addr, tx_odt_drv_stren);
1365 		}
1366 	}
1367 }
1368 
1369 static int map_drvstren_fsdq_p(int strength, int hard_macro_ver)
1370 {
1371 	int val = -1;
1372 
1373 	if (hard_macro_ver == 4) {
1374 		if (strength == 0) {
1375 			val = 0x07;
1376 		} else if (strength == 120) {
1377 			val = 0x0F;
1378 		} else if (strength == 60) {
1379 			val = 0x1F;
1380 		} else if (strength == 40) {
1381 			val = 0x3F;
1382 		} else {
1383 			printf("error: unsupported drv_stren_fSDq_p %d\n",
1384 			       strength);
1385 		}
1386 	} else {
1387 		val = map_impedance(strength);
1388 	}
1389 
1390 	return val;
1391 }
1392 
1393 static int map_drvstren_fsdq_n(int strength, int hard_macro_ver)
1394 {
1395 	int val = -1;
1396 
1397 	if (hard_macro_ver == 4) {
1398 		if (strength == 0) {
1399 			val = 0x00;
1400 		} else if (strength == 120) {
1401 			val = 0x08;
1402 		} else if (strength == 60) {
1403 			val = 0x18;
1404 		} else if (strength == 40) {
1405 			val = 0x38;
1406 		} else {
1407 			printf("error: unsupported drvStrenFSDqN %d\n",
1408 			       strength);
1409 		}
1410 	} else {
1411 		val = map_impedance(strength);
1412 	}
1413 
1414 	return val;
1415 }
1416 
1417 static void prog_tx_impedance_ctrl1(uint16_t *phy,
1418 				    const struct input *input)
1419 {
1420 	int lane, byte, b_addr, c_addr;
1421 	int tx_impedance_ctrl1;
1422 	int drv_stren_fsdq_p, drv_stren_fsdq_n;
1423 	uint32_t addr;
1424 
1425 	drv_stren_fsdq_p = map_drvstren_fsdq_p(input->adv.tx_impedance,
1426 					input->basic.hard_macro_ver);
1427 	drv_stren_fsdq_n = map_drvstren_fsdq_n(input->adv.tx_impedance,
1428 					input->basic.hard_macro_ver);
1429 	tx_impedance_ctrl1 = drv_stren_fsdq_n << csr_drv_stren_fsdq_n_lsb |
1430 			   drv_stren_fsdq_p << csr_drv_stren_fsdq_p_lsb;
1431 
1432 	for (byte = 0; byte < input->basic.num_dbyte; byte++) {
1433 		c_addr = byte << 12;
1434 		for (lane = 0; lane <= 1; lane++) {
1435 			b_addr = lane << 8;
1436 			addr = t_dbyte | c_addr | b_addr |
1437 				csr_tx_impedance_ctrl1_addr;
1438 			phy_io_write16(phy, addr, tx_impedance_ctrl1);
1439 		}
1440 	}
1441 }
1442 
1443 static int map_adrv_stren_p(int strength, int hard_macro_ver)
1444 {
1445 	int val = -1;
1446 
1447 	if (hard_macro_ver == 4) {
1448 		if (strength == 120) {
1449 			val = 0x1c;
1450 		} else if (strength == 60) {
1451 			val = 0x1d;
1452 		} else if (strength == 40) {
1453 			val = 0x1f;
1454 		} else {
1455 			printf("error: unsupported aDrv_stren_p %d\n",
1456 			       strength);
1457 		}
1458 	} else {
1459 		if (strength == 120) {
1460 			val = 0x00;
1461 		} else if (strength == 60) {
1462 			val = 0x01;
1463 		} else if (strength == 40) {
1464 			val = 0x03;
1465 		} else if (strength == 30) {
1466 			val = 0x07;
1467 		} else if (strength == 24) {
1468 			val = 0x0f;
1469 		} else if (strength == 20) {
1470 			val = 0x1f;
1471 		} else {
1472 			printf("error: unsupported aDrv_stren_p %d\n",
1473 			       strength);
1474 		}
1475 	}
1476 
1477 	return val;
1478 }
1479 
1480 static int map_adrv_stren_n(int strength, int hard_macro_ver)
1481 {
1482 	int val = -1;
1483 
1484 	if (hard_macro_ver == 4) {
1485 		if (strength == 120) {
1486 			val = 0x00;
1487 		} else if (strength == 60) {
1488 			val = 0x01;
1489 		} else if (strength == 40) {
1490 			val = 0x03;
1491 		} else {
1492 			printf("Error: unsupported ADrvStrenP %d\n", strength);
1493 		}
1494 	} else {
1495 		if (strength == 120) {
1496 			val = 0x00;
1497 		} else if (strength == 60) {
1498 			val = 0x01;
1499 		} else if (strength == 40) {
1500 			val = 0x03;
1501 		} else if (strength == 30) {
1502 			val = 0x07;
1503 		} else if (strength == 24) {
1504 			val = 0x0f;
1505 		} else if (strength == 20) {
1506 			val = 0x1f;
1507 		} else {
1508 			printf("Error: unsupported ADrvStrenP %d\n", strength);
1509 		}
1510 	}
1511 
1512 	return val;
1513 }
1514 
1515 static void prog_atx_impedance(uint16_t *phy,
1516 			       const struct input *input)
1517 {
1518 	int anib, c_addr;
1519 	int atx_impedance;
1520 	int adrv_stren_p;
1521 	int adrv_stren_n;
1522 	uint32_t addr;
1523 
1524 	if (input->basic.hard_macro_ver == 4 &&
1525 	    input->adv.atx_impedance == 20) {
1526 		printf("Error:ATxImpedance has to be 40 for HardMacroVer 4\n");
1527 		return;
1528 	}
1529 
1530 	adrv_stren_p = map_adrv_stren_p(input->adv.atx_impedance,
1531 					input->basic.hard_macro_ver);
1532 	adrv_stren_n = map_adrv_stren_n(input->adv.atx_impedance,
1533 					input->basic.hard_macro_ver);
1534 	atx_impedance = adrv_stren_n << csr_adrv_stren_n_lsb		|
1535 		       adrv_stren_p << csr_adrv_stren_p_lsb;
1536 	for (anib = 0; anib < input->basic.num_anib; anib++) {
1537 		c_addr = anib << 12;
1538 		addr = t_anib | c_addr | csr_atx_impedance_addr;
1539 		phy_io_write16(phy, addr, atx_impedance);
1540 	}
1541 }
1542 
1543 static void prog_dfi_mode(uint16_t *phy,
1544 			  const struct input *input)
1545 {
1546 	int dfi_mode;
1547 	uint32_t addr;
1548 
1549 	if (input->basic.dfi1exists == 1) {
1550 		dfi_mode = 0x5;	/* DFI1 exists but disabled */
1551 	} else {
1552 		dfi_mode = 0x1;	/* DFI1 does not physically exists */
1553 	}
1554 	addr = t_master | csr_dfi_mode_addr;
1555 	phy_io_write16(phy, addr, dfi_mode);
1556 }
1557 
1558 static void prog_acx4_anib_dis(uint16_t *phy, const struct input *input)
1559 {
1560 	uint32_t addr;
1561 
1562 	addr = t_master | csr_acx4_anib_dis_addr;
1563 	phy_io_write16(phy, addr, 0x0);
1564 	debug("%s 0x%x\n", __func__, phy_io_read16(phy, addr));
1565 }
1566 
1567 static void prog_dfi_camode(uint16_t *phy,
1568 			    const struct input *input)
1569 {
1570 	int dfi_camode = 2;
1571 	uint32_t addr = t_master | csr_dfi_camode_addr;
1572 
1573 	phy_io_write16(phy, addr, dfi_camode);
1574 }
1575 
1576 static void prog_cal_drv_str0(uint16_t *phy,
1577 			      const struct input *input)
1578 {
1579 	int cal_drv_str0;
1580 	int cal_drv_str_pd50;
1581 	int cal_drv_str_pu50;
1582 	uint32_t addr;
1583 
1584 	cal_drv_str_pu50 = input->adv.ext_cal_res_val;
1585 	cal_drv_str_pd50 = cal_drv_str_pu50;
1586 	cal_drv_str0 = cal_drv_str_pu50 << csr_cal_drv_str_pu50_lsb |
1587 			cal_drv_str_pd50;
1588 	addr = t_master | csr_cal_drv_str0_addr;
1589 	phy_io_write16(phy, addr, cal_drv_str0);
1590 }
1591 
1592 static void prog_cal_uclk_info(uint16_t *phy,
1593 			       const struct input *input)
1594 {
1595 	int cal_uclk_ticks_per1u_s;
1596 	uint32_t addr;
1597 
1598 	cal_uclk_ticks_per1u_s = input->basic.frequency >> 1;
1599 	if (cal_uclk_ticks_per1u_s < 24) {
1600 		cal_uclk_ticks_per1u_s = 24;
1601 	}
1602 
1603 	addr = t_master | csr_cal_uclk_info_addr;
1604 	phy_io_write16(phy, addr, cal_uclk_ticks_per1u_s);
1605 }
1606 
1607 static void prog_cal_rate(uint16_t *phy,
1608 			  const struct input *input)
1609 {
1610 	int cal_rate;
1611 	int cal_interval;
1612 	int cal_once;
1613 	uint32_t addr;
1614 
1615 	cal_interval = input->adv.cal_interval;
1616 	cal_once = input->adv.cal_once;
1617 	cal_rate = cal_once << csr_cal_once_lsb		|
1618 		  cal_interval << csr_cal_interval_lsb;
1619 	addr = t_master | csr_cal_rate_addr;
1620 	phy_io_write16(phy, addr, cal_rate);
1621 }
1622 
1623 static void prog_vref_in_global(uint16_t *phy,
1624 				const struct input *input,
1625 				const struct ddr4u1d *msg)
1626 {
1627 	int vref_in_global;
1628 	int global_vref_in_dac = 0;
1629 	int global_vref_in_sel = 0;
1630 	uint32_t addr;
1631 
1632 	/*
1633 	 * phy_vref_prcnt = msg->phy_vref / 128.0
1634 	 *  global_vref_in_dac = (phy_vref_prcnt - 0.345) / 0.005;
1635 	 */
1636 	global_vref_in_dac = (msg->phy_vref * 1000 - 345 * 128 + 320) /
1637 			     (5 * 128);
1638 
1639 	vref_in_global = global_vref_in_dac << csr_global_vref_in_dac_lsb |
1640 		       global_vref_in_sel;
1641 	addr = t_master | csr_vref_in_global_addr;
1642 	phy_io_write16(phy, addr, vref_in_global);
1643 }
1644 
1645 static void prog_dq_dqs_rcv_cntrl(uint16_t *phy,
1646 				  const struct input *input)
1647 {
1648 	int lane, byte, b_addr, c_addr;
1649 	int dq_dqs_rcv_cntrl;
1650 	int gain_curr_adj_defval = 0xb;
1651 	int major_mode_dbyte = 3;
1652 	int dfe_ctrl_defval = 0;
1653 	int ext_vref_range_defval = 0;
1654 	int sel_analog_vref = 1;
1655 	uint32_t addr;
1656 
1657 #ifdef ERRATA_DDR_A050958
1658 	gain_curr_adj_defval = 0x1f;
1659 #endif
1660 
1661 	dq_dqs_rcv_cntrl = gain_curr_adj_defval << csr_gain_curr_adj_lsb |
1662 			major_mode_dbyte << csr_major_mode_dbyte_lsb	|
1663 			dfe_ctrl_defval << csr_dfe_ctrl_lsb		|
1664 			ext_vref_range_defval << csr_ext_vref_range_lsb	|
1665 			sel_analog_vref << csr_sel_analog_vref_lsb;
1666 	for (byte = 0; byte < input->basic.num_dbyte; byte++) {
1667 		c_addr = byte << 12;
1668 		for (lane = 0; lane <= 1; lane++) {
1669 			b_addr = lane << 8;
1670 			addr = t_dbyte | c_addr | b_addr |
1671 					csr_dq_dqs_rcv_cntrl_addr;
1672 			phy_io_write16(phy, addr, dq_dqs_rcv_cntrl);
1673 		}
1674 	}
1675 }
1676 
1677 static void prog_mem_alert_control(uint16_t *phy,
1678 				   const struct input *input)
1679 {
1680 	int mem_alert_control;
1681 	int mem_alert_control2;
1682 	int malertpu_en;
1683 	int malertrx_en;
1684 	int malertvref_level;
1685 	int malertpu_stren;
1686 	int malertsync_bypass;
1687 	int malertdisable_val_defval = 1;
1688 	uint32_t addr;
1689 
1690 	if (input->basic.dram_type == DDR4 && input->adv.mem_alert_en == 1) {
1691 		malertpu_en = 1;
1692 		malertrx_en = 1;
1693 		malertpu_stren = input->adv.mem_alert_puimp;
1694 		malertvref_level = input->adv.mem_alert_vref_level;
1695 		malertsync_bypass = input->adv.mem_alert_sync_bypass;
1696 		mem_alert_control = malertdisable_val_defval << 14	|
1697 				  malertrx_en << 13		|
1698 				  malertpu_en << 12		|
1699 				  malertpu_stren << 8		|
1700 				  malertvref_level;
1701 		mem_alert_control2 = malertsync_bypass <<
1702 					csr_malertsync_bypass_lsb;
1703 		addr = t_master | csr_mem_alert_control_addr;
1704 		phy_io_write16(phy, addr, mem_alert_control);
1705 		addr = t_master | csr_mem_alert_control2_addr;
1706 		phy_io_write16(phy, addr, mem_alert_control2);
1707 	}
1708 }
1709 
1710 static void prog_dfi_freq_ratio(uint16_t *phy,
1711 				const struct input *input)
1712 {
1713 	int dfi_freq_ratio;
1714 	uint32_t addr = t_master | csr_dfi_freq_ratio_addr;
1715 
1716 	dfi_freq_ratio = input->basic.dfi_freq_ratio;
1717 	phy_io_write16(phy, addr, dfi_freq_ratio);
1718 }
1719 
1720 static void prog_tristate_mode_ca(uint16_t *phy,
1721 				  const struct input *input)
1722 {
1723 	int tristate_mode_ca;
1724 	int dis_dyn_adr_tri;
1725 	int ddr2tmode;
1726 	int ck_dis_val_def = 1;
1727 	uint32_t addr = t_master | csr_tristate_mode_ca_addr;
1728 
1729 	dis_dyn_adr_tri = input->adv.dis_dyn_adr_tri;
1730 	ddr2tmode = input->adv.is2ttiming;
1731 	tristate_mode_ca = ck_dis_val_def << csr_ck_dis_val_lsb	|
1732 			 ddr2tmode << csr_ddr2tmode_lsb		|
1733 			 dis_dyn_adr_tri << csr_dis_dyn_adr_tri_lsb;
1734 	phy_io_write16(phy, addr, tristate_mode_ca);
1735 }
1736 
1737 static void prog_dfi_xlat(uint16_t *phy,
1738 			  const struct input *input)
1739 {
1740 	uint16_t loop_vector;
1741 	int dfifreqxlat_dat;
1742 	int pllbypass_dat;
1743 	uint32_t addr;
1744 
1745 	/* fIXME: Shall unused P1, P2, P3 be bypassed? */
1746 	pllbypass_dat = input->basic.pll_bypass; /* only [0] is used */
1747 	for (loop_vector = 0; loop_vector < 8; loop_vector++) {
1748 		if (loop_vector == 0) {
1749 			dfifreqxlat_dat = pllbypass_dat + 0x5555;
1750 		} else if (loop_vector == 7) {
1751 			dfifreqxlat_dat = 0xf000;
1752 		} else {
1753 			dfifreqxlat_dat = 0x5555;
1754 		}
1755 		addr = t_master | (csr_dfi_freq_xlat0_addr + loop_vector);
1756 		phy_io_write16(phy, addr, dfifreqxlat_dat);
1757 	}
1758 }
1759 
1760 static void prog_dbyte_misc_mode(uint16_t *phy,
1761 				 const struct input *input,
1762 				 const struct ddr4u1d *msg)
1763 {
1764 	int dbyte_misc_mode;
1765 	int dq_dqs_rcv_cntrl1;
1766 	int dq_dqs_rcv_cntrl1_1;
1767 	int byte, c_addr;
1768 	uint32_t addr;
1769 
1770 	dbyte_misc_mode = 0x1 << csr_dbyte_disable_lsb;
1771 	dq_dqs_rcv_cntrl1 = 0x1ff << csr_power_down_rcvr_lsb		|
1772 			 0x1 << csr_power_down_rcvr_dqs_lsb	|
1773 			 0x1 << csr_rx_pad_standby_en_lsb;
1774 	dq_dqs_rcv_cntrl1_1 = (0x100 << csr_power_down_rcvr_lsb |
1775 			csr_rx_pad_standby_en_mask);
1776 	for (byte = 0; byte < input->basic.num_dbyte; byte++) {
1777 		c_addr = byte << 12;
1778 		if (byte <= input->basic.num_active_dbyte_dfi0 - 1) {
1779 			/* disable RDBI lane if not used. */
1780 			if ((input->basic.dram_data_width != 4) &&
1781 				(((msg->mr5 >> 12) & 0x1) == 0)) {
1782 				addr = t_dbyte
1783 					| c_addr
1784 					| csr_dq_dqs_rcv_cntrl1_addr;
1785 				phy_io_write16(phy, addr, dq_dqs_rcv_cntrl1_1);
1786 			}
1787 		} else {
1788 			addr = t_dbyte | c_addr | csr_dbyte_misc_mode_addr;
1789 			phy_io_write16(phy, addr, dbyte_misc_mode);
1790 			addr = t_dbyte | c_addr | csr_dq_dqs_rcv_cntrl1_addr;
1791 			phy_io_write16(phy, addr, dq_dqs_rcv_cntrl1);
1792 		}
1793 	}
1794 }
1795 
1796 static void prog_master_x4config(uint16_t *phy,
1797 				 const struct input *input)
1798 {
1799 	int master_x4config;
1800 	int x4tg;
1801 	uint32_t addr = t_master | csr_master_x4config_addr;
1802 
1803 	x4tg = input->basic.dram_data_width == 4 ? 0xf : 0;
1804 	master_x4config = x4tg << csr_x4tg_lsb;
1805 	phy_io_write16(phy, addr, master_x4config);
1806 }
1807 
1808 static void prog_dmipin_present(uint16_t *phy,
1809 				const struct input *input,
1810 				const struct ddr4u1d *msg)
1811 {
1812 	int dmipin_present;
1813 	uint32_t addr = t_master | csr_dmipin_present_addr;
1814 
1815 	dmipin_present = (msg->mr5 >> 12) & 0x1;
1816 	phy_io_write16(phy, addr, dmipin_present);
1817 }
1818 
1819 static void prog_dfi_phyupd(uint16_t *phy,
1820 			  const struct input *input)
1821 {
1822 	int dfiphyupd_dat;
1823 	uint32_t addr;
1824 
1825 	addr = t_master | (csr_dfiphyupd_addr);
1826 	dfiphyupd_dat = phy_io_read16(phy, addr) &
1827 				~csr_dfiphyupd_threshold_mask;
1828 
1829 	phy_io_write16(phy, addr, dfiphyupd_dat);
1830 }
1831 
1832 static void prog_cal_misc2(uint16_t *phy,
1833 			  const struct input *input)
1834 {
1835 	int cal_misc2_dat, cal_drv_pdth_data, cal_offsets_dat;
1836 	uint32_t addr;
1837 
1838 	addr = t_master | (csr_cal_misc2_addr);
1839 	cal_misc2_dat = phy_io_read16(phy, addr) |
1840 			(1 << csr_cal_misc2_err_dis);
1841 
1842 	phy_io_write16(phy, addr, cal_misc2_dat);
1843 
1844 	addr = t_master | (csr_cal_offsets_addr);
1845 
1846 	cal_drv_pdth_data = 0x9 << 6;
1847 	cal_offsets_dat = (phy_io_read16(phy, addr) & ~csr_cal_drv_pdth_mask)
1848 			| cal_drv_pdth_data;
1849 
1850 	phy_io_write16(phy, addr, cal_offsets_dat);
1851 }
1852 
1853 static int c_init_phy_config(uint16_t **phy_ptr,
1854 			     unsigned int ip_rev,
1855 			     const struct input *input,
1856 			     const void *msg)
1857 {
1858 	int i;
1859 	uint16_t *phy;
1860 	__unused const soc_info_t *soc_info;
1861 
1862 	for (i = 0; i < NUM_OF_DDRC; i++) {
1863 		phy = phy_ptr[i];
1864 		if (phy == NULL) {
1865 			continue;
1866 		}
1867 
1868 		debug("Initialize PHY %d config\n", i);
1869 		prog_dfi_phyupd(phy, input);
1870 		prog_cal_misc2(phy, input);
1871 		prog_tx_pre_drv_mode(phy, input);
1872 		prog_atx_pre_drv_mode(phy, input);
1873 		prog_enable_cs_multicast(phy, input);	/* rdimm and lrdimm */
1874 		prog_dfi_rd_data_cs_dest_map(phy, ip_rev, input, msg);
1875 		prog_pll_ctrl2(phy, input);
1876 #ifdef DDR_PLL_FIX
1877 		soc_info = get_soc_info();
1878 		debug("SOC_SI_REV = %x\n", soc_info->svr_reg.bf.maj_ver);
1879 		if (soc_info->svr_reg.bf.maj_ver == 1) {
1880 			prog_pll_pwr_dn(phy, input);
1881 
1882 			/*Enable FFE aka TxEqualizationMode for rev1 SI*/
1883 			phy_io_write16(phy, 0x010048, 0x1);
1884 		}
1885 #endif
1886 		prog_ard_ptr_init_val(phy, input);
1887 		prog_dqs_preamble_control(phy, input);
1888 		prog_dll_lck_param(phy, input);
1889 		prog_dll_gain_ctl(phy, input);
1890 		prog_proc_odt_time_ctl(phy, input);
1891 		prog_tx_odt_drv_stren(phy, input);
1892 		prog_tx_impedance_ctrl1(phy, input);
1893 		prog_atx_impedance(phy, input);
1894 		prog_dfi_mode(phy, input);
1895 		prog_dfi_camode(phy, input);
1896 		prog_cal_drv_str0(phy, input);
1897 		prog_cal_uclk_info(phy, input);
1898 		prog_cal_rate(phy, input);
1899 		prog_vref_in_global(phy, input, msg);
1900 		prog_dq_dqs_rcv_cntrl(phy, input);
1901 		prog_mem_alert_control(phy, input);
1902 		prog_dfi_freq_ratio(phy, input);
1903 		prog_tristate_mode_ca(phy, input);
1904 		prog_dfi_xlat(phy, input);
1905 		prog_dbyte_misc_mode(phy, input, msg);
1906 		prog_master_x4config(phy, input);
1907 		prog_dmipin_present(phy, input, msg);
1908 		prog_acx4_anib_dis(phy, input);
1909 	}
1910 
1911 	return 0;
1912 }
1913 
1914 static uint32_t get_mail(uint16_t *phy, int stream)
1915 {
1916 	int timeout;
1917 	uint32_t mail = 0U;
1918 
1919 	timeout = TIMEOUTDEFAULT;
1920 	while (((--timeout) != 0) &&
1921 	       ((phy_io_read16(phy, t_apbonly | csr_uct_shadow_regs)
1922 		& uct_write_prot_shadow_mask) != 0)) {
1923 		mdelay(10);
1924 	}
1925 	if (timeout == 0) {
1926 		ERROR("Timeout getting mail from PHY\n");
1927 		return 0xFFFF;
1928 	}
1929 
1930 	mail = phy_io_read16(phy, t_apbonly |
1931 			     csr_uct_write_only_shadow);
1932 	if (stream != 0) {
1933 		mail |= phy_io_read16(phy, t_apbonly |
1934 				      csr_uct_dat_write_only_shadow) << 16;
1935 	}
1936 
1937 	/* Ack */
1938 	phy_io_write16(phy, t_apbonly | csr_dct_write_prot, 0);
1939 
1940 	timeout = TIMEOUTDEFAULT;
1941 	while (((--timeout) != 0) &&
1942 	       ((phy_io_read16(phy, t_apbonly | csr_uct_shadow_regs)
1943 		 & uct_write_prot_shadow_mask) == 0)) {
1944 		mdelay(1);
1945 	}
1946 	if (timeout == 0) {
1947 		ERROR("Timeout ack PHY mail\n");
1948 	}
1949 
1950 	/* completed */
1951 	phy_io_write16(phy, t_apbonly | csr_dct_write_prot, 1U);
1952 
1953 	return mail;
1954 }
1955 
1956 #ifdef DDR_PHY_DEBUG
1957 static const char *lookup_msg(uint32_t index, int train2d)
1958 {
1959 	int i;
1960 	int size;
1961 	const struct phy_msg *messages;
1962 	const char *ptr = NULL;
1963 
1964 	if (train2d != 0) {
1965 		messages = messages_2d;
1966 		size = ARRAY_SIZE(messages_2d);
1967 	} else {
1968 		messages = messages_1d;
1969 		size = ARRAY_SIZE(messages_1d);
1970 	}
1971 	for (i = 0; i < size; i++) {
1972 		if (messages[i].index == index) {
1973 			ptr = messages[i].msg;
1974 			break;
1975 		}
1976 	}
1977 
1978 	return ptr;
1979 }
1980 #endif
1981 
1982 #define MAX_ARGS 32
1983 static void decode_stream_message(uint16_t *phy, int train2d)
1984 {
1985 	uint32_t index __unused;
1986 
1987 	__unused const char *format;
1988 	__unused uint32_t args[MAX_ARGS];
1989 	__unused int i;
1990 
1991 #ifdef DDR_PHY_DEBUG
1992 	index = get_mail(phy, 1);
1993 	if ((index & 0xffff) > MAX_ARGS) {	/* up to MAX_ARGS args so far */
1994 		printf("Program error in %s\n", __func__);
1995 	}
1996 	for (i = 0; i < (index & 0xffff) && i < MAX_ARGS; i++) {
1997 		args[i] = get_mail(phy, 1);
1998 	}
1999 
2000 	format = lookup_msg(index, train2d);
2001 	if (format != NULL) {
2002 		printf("0x%08x: ", index);
2003 		printf(format, args[0], args[1], args[2], args[3], args[4],
2004 		       args[5], args[6], args[7], args[8], args[9], args[10],
2005 		       args[11], args[12], args[13], args[14], args[15],
2006 		       args[16], args[17], args[18], args[19], args[20],
2007 		       args[21], args[22], args[23], args[24], args[25],
2008 		       args[26], args[27], args[28], args[29], args[30],
2009 		       args[31]);
2010 	}
2011 #endif
2012 }
2013 
2014 static int wait_fw_done(uint16_t *phy, int train2d)
2015 {
2016 	uint32_t mail = 0U;
2017 
2018 	while (mail == U(0x0)) {
2019 		mail = get_mail(phy, 0);
2020 		switch (mail) {
2021 		case U(0x7):
2022 			debug("%s Training completed\n", train2d ? "2D" : "1D");
2023 			break;
2024 		case U(0xff):
2025 			debug("%s Training failure\n", train2d ? "2D" : "1D");
2026 			break;
2027 		case U(0x0):
2028 			debug("End of initialization\n");
2029 			mail = 0U;
2030 			break;
2031 		case U(0x1):
2032 			debug("End of fine write leveling\n");
2033 			mail = 0U;
2034 			break;
2035 		case U(0x2):
2036 			debug("End of read enable training\n");
2037 			mail = 0U;
2038 			break;
2039 		case U(0x3):
2040 			debug("End of read delay center optimization\n");
2041 			mail = 0U;
2042 			break;
2043 		case U(0x4):
2044 			debug("End of write delay center optimization\n");
2045 			mail = 0U;
2046 			break;
2047 		case U(0x5):
2048 			debug("End of 2D read delay/voltage center optimztn\n");
2049 			mail = 0U;
2050 			break;
2051 		case U(0x6):
2052 			debug("End of 2D write delay/voltage center optmztn\n");
2053 			mail = 0U;
2054 			break;
2055 		case U(0x8):
2056 			decode_stream_message(phy, train2d);
2057 			mail = 0U;
2058 			break;
2059 		case U(0x9):
2060 			debug("End of max read latency training\n");
2061 			mail = 0U;
2062 			break;
2063 		case U(0xa):
2064 			debug("End of read dq deskew training\n");
2065 			mail = 0U;
2066 			break;
2067 		case U(0xc):
2068 			debug("End of LRDIMM Specific training, including:\n");
2069 			debug("/tDWL, MREP, MRD and MWD\n");
2070 			mail = 0U;
2071 			break;
2072 		case U(0xd):
2073 			debug("End of CA training\n");
2074 			mail = 0U;
2075 			break;
2076 		case U(0xfd):
2077 			debug("End of MPR read delay center optimization\n");
2078 			mail = 0U;
2079 			break;
2080 		case U(0xfe):
2081 			debug("End of Write leveling coarse delay\n");
2082 			mail = 0U;
2083 			break;
2084 		case U(0xffff):
2085 			debug("Timed out\n");
2086 			break;
2087 		default:
2088 			mail = 0U;
2089 			break;
2090 		}
2091 	}
2092 
2093 	if (mail == U(0x7)) {
2094 		return 0;
2095 	} else if (mail == U(0xff)) {
2096 		return -EIO;
2097 	} else if (mail == U(0xffff)) {
2098 		return -ETIMEDOUT;
2099 	}
2100 
2101 	debug("PHY_GEN2 FW: Unxpected mail = 0x%x\n", mail);
2102 
2103 	return -EINVAL;
2104 }
2105 
2106 static int g_exec_fw(uint16_t **phy_ptr, int train2d, struct input *input)
2107 {
2108 	int ret = -EINVAL;
2109 	int i;
2110 	uint16_t *phy;
2111 
2112 	for (i = 0; i < NUM_OF_DDRC; i++) {
2113 		phy = phy_ptr[i];
2114 		if (phy == NULL) {
2115 			continue;
2116 		}
2117 		debug("Applying PLL optimal settings\n");
2118 		prog_pll_ctrl2(phy, input);
2119 		prog_pll_ctrl(phy, input);
2120 		phy_io_write16(phy,
2121 			       t_apbonly | csr_micro_cont_mux_sel_addr,
2122 			       0x1);
2123 		phy_io_write16(phy,
2124 			       t_apbonly | csr_micro_reset_addr,
2125 			       csr_reset_to_micro_mask |
2126 			       csr_stall_to_micro_mask);
2127 		phy_io_write16(phy,
2128 			       t_apbonly | csr_micro_reset_addr,
2129 			       csr_stall_to_micro_mask);
2130 		phy_io_write16(phy,
2131 			       t_apbonly | csr_micro_reset_addr,
2132 			       0);
2133 
2134 		ret = wait_fw_done(phy, train2d);
2135 		if (ret == -ETIMEDOUT) {
2136 			ERROR("Wait timed out: Firmware execution on PHY %d\n",
2137 			      i);
2138 		}
2139 	}
2140 	return ret;
2141 }
2142 
2143 static inline int send_fw(uint16_t *phy,
2144 			   uint32_t dst,
2145 			   uint16_t *img,
2146 			   uint32_t size)
2147 {
2148 	uint32_t i;
2149 
2150 	if ((size % 2U) != 0U) {
2151 		ERROR("Wrong image size 0x%x\n", size);
2152 		return -EINVAL;
2153 	}
2154 
2155 	for (i = 0U; i < size / 2; i++) {
2156 		phy_io_write16(phy, dst + i, *(img + i));
2157 	}
2158 
2159 	return 0;
2160 }
2161 
2162 static int load_fw(uint16_t **phy_ptr,
2163 		   struct input *input,
2164 		   int train2d,
2165 		   void *msg,
2166 		   size_t len,
2167 		   uintptr_t phy_gen2_fw_img_buf,
2168 		   int (*img_loadr)(unsigned int, uintptr_t *, uint32_t *),
2169 		   uint32_t warm_boot_flag)
2170 {
2171 	uint32_t imem_id, dmem_id;
2172 	uintptr_t image_buf;
2173 	uint32_t size;
2174 	int ret;
2175 	int i;
2176 	uint16_t *phy;
2177 
2178 	switch (input->basic.dimm_type) {
2179 	case UDIMM:
2180 	case SODIMM:
2181 	case NODIMM:
2182 		imem_id = train2d ? DDR_IMEM_UDIMM_2D_IMAGE_ID :
2183 			  DDR_IMEM_UDIMM_1D_IMAGE_ID;
2184 		dmem_id = train2d ? DDR_DMEM_UDIMM_2D_IMAGE_ID :
2185 			  DDR_DMEM_UDIMM_1D_IMAGE_ID;
2186 		break;
2187 	case RDIMM:
2188 		imem_id = train2d ? DDR_IMEM_RDIMM_2D_IMAGE_ID :
2189 			  DDR_IMEM_RDIMM_1D_IMAGE_ID;
2190 		dmem_id = train2d ? DDR_DMEM_RDIMM_2D_IMAGE_ID :
2191 			  DDR_DMEM_RDIMM_1D_IMAGE_ID;
2192 		break;
2193 	default:
2194 		ERROR("Unsupported DIMM type\n");
2195 		return -EINVAL;
2196 	}
2197 
2198 	size = PHY_GEN2_MAX_IMAGE_SIZE;
2199 	image_buf = (uintptr_t)phy_gen2_fw_img_buf;
2200 	ret = img_loadr(imem_id, &image_buf, &size);
2201 	if (ret != 0) {
2202 		ERROR("Failed to load %d firmware.\n", imem_id);
2203 		return ret;
2204 	}
2205 	debug("Loaded Imaged id %d of size %x at address %lx\n",
2206 						imem_id, size, image_buf);
2207 
2208 	for (i = 0; i < NUM_OF_DDRC; i++) {
2209 		phy = phy_ptr[i];
2210 		if (phy == NULL) {
2211 			continue;
2212 		}
2213 
2214 		if (warm_boot_flag != DDR_WARM_BOOT) {
2215 			if (train2d == 0) {
2216 				phy_io_write16(phy, t_master |
2217 						csr_mem_reset_l_addr,
2218 						csr_protect_mem_reset_mask);
2219 			}
2220 		}
2221 		/* Enable access to the internal CSRs */
2222 		phy_io_write16(phy, t_apbonly | csr_micro_cont_mux_sel_addr, 0);
2223 
2224 		ret = send_fw(phy, PHY_GEN2_IMEM_ADDR,
2225 			      (uint16_t *)image_buf, size);
2226 		if (ret != 0) {
2227 			return ret;
2228 		}
2229 	}
2230 
2231 	size = PHY_GEN2_MAX_IMAGE_SIZE;
2232 	image_buf = (uintptr_t)phy_gen2_fw_img_buf;
2233 	ret = img_loadr(dmem_id, &image_buf, &size);
2234 	if (ret != 0) {
2235 		ERROR("Failed to load %d firmware.\n", dmem_id);
2236 		return ret;
2237 	}
2238 	debug("Loaded Imaged id %d of size %x at address %lx\n",
2239 						dmem_id, size, image_buf);
2240 	image_buf += len;
2241 	size -= len;
2242 
2243 	for (i = 0; i < NUM_OF_DDRC; i++) {
2244 		phy = phy_ptr[i];
2245 		if (phy == NULL) {
2246 			continue;
2247 		}
2248 
2249 		ret = send_fw(phy, PHY_GEN2_DMEM_ADDR, msg, len);
2250 		if (ret != 0) {
2251 			return ret;
2252 		}
2253 
2254 		ret = send_fw(phy, PHY_GEN2_DMEM_ADDR + len / 2,
2255 			      (uint16_t *)image_buf, size);
2256 		if (ret != 0) {
2257 			return ret;
2258 		}
2259 	}
2260 
2261 	return ret;
2262 }
2263 
2264 static void parse_odt(const unsigned int val,
2265 		       const int read,
2266 		       const int i,
2267 		       const unsigned int cs_d0,
2268 		       const unsigned int cs_d1,
2269 		       unsigned int *odt)
2270 {
2271 	int shift = read ? 4 : 0;
2272 	int j;
2273 
2274 	if (i < 0 || i > 3) {
2275 		printf("Error: invalid chip-select value\n");
2276 	}
2277 	switch (val) {
2278 	case DDR_ODT_CS:
2279 		odt[i] |= (1 << i) << shift;
2280 		break;
2281 	case DDR_ODT_ALL_OTHER_CS:
2282 		for (j = 0; j < DDRC_NUM_CS; j++) {
2283 			if (i == j) {
2284 				continue;
2285 			}
2286 			if (((cs_d0 | cs_d1) & (1 << j)) == 0) {
2287 				continue;
2288 			}
2289 			odt[j] |= (1 << i) << shift;
2290 		}
2291 		break;
2292 	case DDR_ODT_CS_AND_OTHER_DIMM:
2293 		odt[i] |= (1 << i) << 4;
2294 		/* fallthrough */
2295 	case DDR_ODT_OTHER_DIMM:
2296 		for (j = 0; j < DDRC_NUM_CS; j++) {
2297 			if ((((cs_d0 & (1 << i)) != 0) &&
2298 						((cs_d1 & (1 << j)) != 0)) ||
2299 			    (((cs_d1 & (1 << i)) != 0) &&
2300 						((cs_d0 & (1 << j)) != 0))) {
2301 				odt[j] |= (1 << i) << shift;
2302 			}
2303 		}
2304 		break;
2305 	case DDR_ODT_ALL:
2306 		for (j = 0; j < DDRC_NUM_CS; j++) {
2307 			if (((cs_d0 | cs_d1) & (1 << j)) == 0) {
2308 				continue;
2309 			}
2310 			odt[j] |= (1 << i) << shift;
2311 		}
2312 		break;
2313 	case DDR_ODT_SAME_DIMM:
2314 		for (j = 0; j < DDRC_NUM_CS; j++) {
2315 			if ((((cs_d0 & (1 << i)) != 0) &&
2316 						((cs_d0 & (1 << j)) != 0)) ||
2317 			    (((cs_d1 & (1 << i)) != 0) &&
2318 						((cs_d1 & (1 << j)) != 0))) {
2319 				odt[j] |= (1 << i) << shift;
2320 			}
2321 		}
2322 		break;
2323 	case DDR_ODT_OTHER_CS_ONSAMEDIMM:
2324 		for (j = 0; j < DDRC_NUM_CS; j++) {
2325 			if (i == j) {
2326 				continue;
2327 			}
2328 			if ((((cs_d0 & (1 << i)) != 0) &&
2329 						((cs_d0 & (1 << j)) != 0)) ||
2330 			    (((cs_d1 & (1 << i)) != 0) &&
2331 						((cs_d1 & (1 << j)) != 0))) {
2332 				odt[j] |= (1 << i) << shift;
2333 			}
2334 		}
2335 		break;
2336 	case DDR_ODT_NEVER:
2337 		break;
2338 	default:
2339 		break;
2340 	}
2341 }
2342 
2343 #ifdef DEBUG_DDR_INPUT_CONFIG
2344 char *dram_types_str[] = {
2345 		"DDR4",
2346 		"DDR3",
2347 		"LDDDR4",
2348 		"LPDDR3",
2349 		"LPDDR2",
2350 		"DDR5"
2351 };
2352 
2353 char *dimm_types_str[] = {
2354 		"UDIMM",
2355 		"SODIMM",
2356 		"RDIMM",
2357 		"LRDIMM",
2358 		"NODIMM",
2359 };
2360 
2361 
2362 static void print_jason_format(struct input *input,
2363 			       struct ddr4u1d *msg_1d,
2364 			       struct ddr4u2d *msg_2d)
2365 {
2366 
2367 	printf("\n{");
2368 	printf("\n    \"dram_type\": \"%s\",", dram_types_str[input->basic.dram_type]);
2369 	printf("\n    \"dimm_type\": \"%s\",", dimm_types_str[input->basic.dimm_type]);
2370 	printf("\n    \"hard_macro_ver\": \"%d\",", input->basic.hard_macro_ver);
2371 	printf("\n    \"num_dbyte\": \"0x%04x\",", (unsigned int)input->basic.num_dbyte);
2372 	printf("\n    \"num_active_dbyte_dfi0\": \"0x%04x\",", (unsigned int)input->basic.num_active_dbyte_dfi0);
2373 	printf("\n    \"num_anib\": \"0x%04x\",", (unsigned int)input->basic.num_anib);
2374 	printf("\n    \"num_rank_dfi0\": \"0x%04x\",", (unsigned int)input->basic.num_rank_dfi0);
2375 	printf("\n    \"num_pstates\": \"0x%04x\",", (unsigned int)input->basic.num_pstates);
2376 	printf("\n    \"frequency\": \"%d\",", input->basic.frequency);
2377 	printf("\n    \"pll_bypass\": \"0x%04x\",", (unsigned int)input->basic.dfi_freq_ratio);
2378 	printf("\n    \"dfi_freq_ratio\": \"0x%04x\",", (unsigned int)input->basic.dfi_freq_ratio);
2379 	printf("\n    \"dfi1_exists\":  \"0x%04x\",", (unsigned int)input->basic.dfi1exists);
2380 	printf("\n    \"dram_data_width\": \"0x%04x\",", (unsigned int)input->basic.dram_data_width);
2381 	printf("\n    \"dram_byte_swap\": \"0x%04x\",", (unsigned int)input->adv.dram_byte_swap);
2382 	printf("\n    \"ext_cal_res_val\": \"0x%04x\",", (unsigned int)input->adv.ext_cal_res_val);
2383 	printf("\n    \"tx_slew_rise_dq\": \"0x%04x\",", (unsigned int)input->adv.tx_slew_rise_dq);
2384 	printf("\n    \"tx_slew_fall_dq\": \"0x%04x\",", (unsigned int)input->adv.tx_slew_fall_dq);
2385 	printf("\n    \"tx_slew_rise_ac\": \"0x%04x\",", (unsigned int)input->adv.tx_slew_rise_ac);
2386 	printf("\n    \"tx_slew_fall_ac\": \"0x%04x\",", (unsigned int)input->adv.tx_slew_fall_ac);
2387 	printf("\n    \"odt_impedance\": \"%d\",", input->adv.odtimpedance);
2388 	printf("\n    \"tx_impedance\": \"%d\",", input->adv.tx_impedance);
2389 	printf("\n    \"atx_impedance\": \"%d\",", input->adv.atx_impedance);
2390 	printf("\n    \"mem_alert_en\": \"0x%04x\",", (unsigned int)input->adv.mem_alert_en);
2391 	printf("\n    \"mem_alert_pu_imp\": \"0x%04x\",", (unsigned int)input->adv.mem_alert_puimp);
2392 	printf("\n    \"mem_alert_vref_level\": \"0x%04x\",", (unsigned int)input->adv.mem_alert_vref_level);
2393 	printf("\n    \"mem_alert_sync_bypass\": \"0x%04x\",", (unsigned int)input->adv.mem_alert_sync_bypass);
2394 	printf("\n    \"cal_interval\": \"0x%04x\",", (unsigned int)input->adv.cal_interval);
2395 	printf("\n    \"cal_once\": \"0x%04x\",", (unsigned int)input->adv.cal_once);
2396 	printf("\n    \"dis_dyn_adr_tri\": \"0x%04x\",", (unsigned int)input->adv.dis_dyn_adr_tri);
2397 	printf("\n    \"is2t_timing\": \"0x%04x\",", (unsigned int)input->adv.is2ttiming);
2398 	printf("\n    \"d4rx_preabmle_length\": \"0x%04x\",", (unsigned int)input->adv.d4rx_preamble_length);
2399 	printf("\n    \"d4tx_preamble_length\": \"0x%04x\",", (unsigned int)input->adv.d4tx_preamble_length);
2400 	printf("\n    \"msg_misc\": \"0x%02x\",", (unsigned int)msg_1d->msg_misc);
2401 	printf("\n    \"reserved00\": \"0x%01x\",", (unsigned int)msg_1d->reserved00);
2402 	printf("\n    \"hdt_ctrl\": \"0x%02x\",", (unsigned int)msg_1d->hdt_ctrl);
2403 	printf("\n    \"cs_present\": \"0x%02x\",", (unsigned int)msg_1d->cs_present);
2404 	printf("\n    \"phy_vref\": \"0x%02x\",", (unsigned int)msg_1d->phy_vref);
2405 	printf("\n    \"dfi_mrl_margin\": \"0x%02x\",", (unsigned int)msg_1d->dfimrlmargin);
2406 	printf("\n    \"addr_mirror\": \"0x%02x\",", (unsigned int)msg_1d->addr_mirror);
2407 	printf("\n    \"wr_odt_pat_rank0\": \"0x%02x\",", (unsigned int)(msg_1d->acsm_odt_ctrl0 & 0x0f));
2408 	printf("\n    \"wr_odt_pat_rank1\": \"0x%02x\",", (unsigned int)(msg_1d->acsm_odt_ctrl1 & 0x0f));
2409 	printf("\n    \"wr_odt_pat_rank2\": \"0x%02x\",", (unsigned int)(msg_1d->acsm_odt_ctrl2 & 0x0f));
2410 	printf("\n    \"wr_odt_pat_rank3\": \"0x%02x\",", (unsigned int)(msg_1d->acsm_odt_ctrl3 & 0x0f));
2411 	printf("\n    \"rd_odt_pat_rank0\": \"0x%02x\",", (unsigned int)(msg_1d->acsm_odt_ctrl0 & 0xf0));
2412 	printf("\n    \"rd_odt_pat_rank1\": \"0x%02x\",", (unsigned int)(msg_1d->acsm_odt_ctrl1 & 0xf0));
2413 	printf("\n    \"rd_odt_pat_rank2\": \"0x%02x\",", (unsigned int)(msg_1d->acsm_odt_ctrl2 & 0xf0));
2414 	printf("\n    \"rd_odt_pat_rank3\": \"0x%02x\",", (unsigned int)(msg_1d->acsm_odt_ctrl3 & 0xf0));
2415 	printf("\n    \"d4_misc\": \"0x%01x\",", (unsigned int)msg_1d->d4misc);
2416 	printf("\n    \"share_2d_vref_results\": \"0x%01x\",", (unsigned int)msg_1d->share2dvref_result);
2417 	printf("\n    \"sequence_ctrl\": \"0x%04x\",", (unsigned int)msg_1d->sequence_ctrl);
2418 	printf("\n    \"mr0\": \"0x%04x\",", (unsigned int)msg_1d->mr0);
2419 	printf("\n    \"mr1\": \"0x%04x\",", (unsigned int)msg_1d->mr1);
2420 	printf("\n    \"mr2\": \"0x%04x\",", (unsigned int)msg_1d->mr2);
2421 	printf("\n    \"mr3\": \"0x%04x\",", (unsigned int)msg_1d->mr3);
2422 	printf("\n    \"mr4\": \"0x%04x\",", (unsigned int)msg_1d->mr4);
2423 	printf("\n    \"mr5\": \"0x%04x\",", (unsigned int)msg_1d->mr5);
2424 	printf("\n    \"mr6\": \"0x%04x\",", (unsigned int)msg_1d->mr6);
2425 	printf("\n    \"alt_cal_l\": \"0x%04x\",", (unsigned int)msg_1d->alt_cas_l);
2426 	printf("\n    \"alt_wcal_l\": \"0x%04x\",", (unsigned int)msg_1d->alt_wcas_l);
2427 	printf("\n    \"sequence_ctrl_2d\": \"0x%04x\",", (unsigned int)msg_2d->sequence_ctrl);
2428 	printf("\n    \"rtt_nom_wr_park0\": \"0x%01x\",", (unsigned int)msg_1d->rtt_nom_wr_park0);
2429 	printf("\n    \"rtt_nom_wr_park1\": \"0x%01x\",", (unsigned int)msg_1d->rtt_nom_wr_park1);
2430 	printf("\n    \"rtt_nom_wr_park2\": \"0x%01x\",", (unsigned int)msg_1d->rtt_nom_wr_park2);
2431 	printf("\n    \"rtt_nom_wr_park3\": \"0x%01x\",", (unsigned int)msg_1d->rtt_nom_wr_park3);
2432 	printf("\n    \"rtt_nom_wr_park4\": \"0x%01x\",", (unsigned int)msg_1d->rtt_nom_wr_park4);
2433 	printf("\n    \"rtt_nom_wr_park5\": \"0x%01x\",", (unsigned int)msg_1d->rtt_nom_wr_park5);
2434 	printf("\n    \"rtt_nom_wr_park6\": \"0x%01x\",", (unsigned int)msg_1d->rtt_nom_wr_park6);
2435 	printf("\n    \"rtt_nom_wr_park7\": \"0x%01x\"", (unsigned int)msg_1d->rtt_nom_wr_park7);
2436 	printf("\n}");
2437 	printf("\n");
2438 }
2439 #endif
2440 
2441 int compute_ddr_phy(struct ddr_info *priv)
2442 {
2443 	const unsigned long clk = priv->clk;
2444 	const struct memctl_opt *popts = &priv->opt;
2445 	const struct ddr_conf *conf = &priv->conf;
2446 	const struct dimm_params *dimm_param = &priv->dimm;
2447 	struct ddr_cfg_regs *regs = &priv->ddr_reg;
2448 	int ret;
2449 	static struct input input;
2450 	static struct ddr4u1d msg_1d;
2451 	static struct ddr4u2d msg_2d;
2452 	unsigned int i;
2453 	unsigned int odt_rd, odt_wr;
2454 	__unused const soc_info_t *soc_info;
2455 #ifdef NXP_APPLY_MAX_CDD
2456 	unsigned int tcfg0, tcfg4, rank;
2457 #endif
2458 
2459 	if (dimm_param == NULL) {
2460 		ERROR("Empty DIMM parameters.\n");
2461 		return -EINVAL;
2462 	}
2463 
2464 	zeromem(&input, sizeof(input));
2465 	zeromem(&msg_1d, sizeof(msg_1d));
2466 	zeromem(&msg_2d, sizeof(msg_2d));
2467 
2468 	input.basic.dram_type = DDR4;
2469 	/* FIXME: Add condition for LRDIMM */
2470 	input.basic.dimm_type = (dimm_param->rdimm != 0) ? RDIMM : UDIMM;
2471 	input.basic.num_dbyte = dimm_param->primary_sdram_width / 8 +
2472 				 dimm_param->ec_sdram_width / 8;
2473 	input.basic.num_active_dbyte_dfi0 = input.basic.num_dbyte;
2474 	input.basic.num_rank_dfi0 = dimm_param->n_ranks;
2475 	input.basic.dram_data_width = dimm_param->device_width;
2476 	input.basic.hard_macro_ver	= 0xa;
2477 	input.basic.num_pstates	= 1;
2478 	input.basic.dfi_freq_ratio	= 1;
2479 	input.basic.num_anib		= 0xc;
2480 	input.basic.train2d		= popts->skip2d ? 0 : 1;
2481 	input.basic.frequency = (int) (clk / 2000000ul);
2482 	debug("frequency = %dMHz\n", input.basic.frequency);
2483 	input.cs_d0 = conf->cs_on_dimm[0];
2484 #if DDRC_NUM_DIMM > 1
2485 	input.cs_d1 = conf->cs_on_dimm[1];
2486 #endif
2487 	input.mirror = dimm_param->mirrored_dimm;
2488 	input.mr[0] = regs->sdram_mode[0] & U(0xffff);
2489 	input.mr[1] = regs->sdram_mode[0] >> 16U;
2490 	input.mr[2] = regs->sdram_mode[1] >> 16U;
2491 	input.mr[3] = regs->sdram_mode[1] & U(0xffff);
2492 	input.mr[4] = regs->sdram_mode[8] >> 16U;
2493 	input.mr[5] = regs->sdram_mode[8] & U(0xffff);
2494 	input.mr[6] = regs->sdram_mode[9] >> 16U;
2495 	input.vref = popts->vref_phy;
2496 	debug("Vref_phy = %d percent\n", (input.vref * 100U) >> 7U);
2497 	for (i = 0U; i < DDRC_NUM_CS; i++) {
2498 		if ((regs->cs[i].config & SDRAM_CS_CONFIG_EN) == 0U) {
2499 			continue;
2500 		}
2501 		odt_rd = (regs->cs[i].config >> 20U) & U(0x7);
2502 		odt_wr = (regs->cs[i].config >> 16U) & U(0x7);
2503 		parse_odt(odt_rd, true, i, input.cs_d0, input.cs_d1,
2504 			   input.odt);
2505 		parse_odt(odt_wr, false, i, input.cs_d0, input.cs_d1,
2506 			   input.odt);
2507 	}
2508 
2509 	/* Do not set sdram_cfg[RD_EN] or sdram_cfg2[RCW_EN] for RDIMM */
2510 	if (dimm_param->rdimm != 0U) {
2511 		regs->sdram_cfg[0] &= ~(1 << 28U);
2512 		regs->sdram_cfg[1] &= ~(1 << 2U);
2513 		input.rcw[0] = (regs->sdram_rcw[0] >> 28U) & U(0xf);
2514 		input.rcw[1] = (regs->sdram_rcw[0] >> 24U) & U(0xf);
2515 		input.rcw[2] = (regs->sdram_rcw[0] >> 20U) & U(0xf);
2516 		input.rcw[3] = (regs->sdram_rcw[0] >> 16U) & U(0xf);
2517 		input.rcw[4] = (regs->sdram_rcw[0] >> 12U) & U(0xf);
2518 		input.rcw[5] = (regs->sdram_rcw[0] >> 8U) & U(0xf);
2519 		input.rcw[6] = (regs->sdram_rcw[0] >> 4U) & U(0xf);
2520 		input.rcw[7] = (regs->sdram_rcw[0] >> 0U) & U(0xf);
2521 		input.rcw[8] = (regs->sdram_rcw[1] >> 28U) & U(0xf);
2522 		input.rcw[9] = (regs->sdram_rcw[1] >> 24U) & U(0xf);
2523 		input.rcw[10] = (regs->sdram_rcw[1] >> 20U) & U(0xf);
2524 		input.rcw[11] = (regs->sdram_rcw[1] >> 16U) & U(0xf);
2525 		input.rcw[12] = (regs->sdram_rcw[1] >> 12U) & U(0xf);
2526 		input.rcw[13] = (regs->sdram_rcw[1] >> 8U) & U(0xf);
2527 		input.rcw[14] = (regs->sdram_rcw[1] >> 4U) & U(0xf);
2528 		input.rcw[15] = (regs->sdram_rcw[1] >> 0U) & U(0xf);
2529 		input.rcw3x = (regs->sdram_rcw[2] >> 8U) & U(0xff);
2530 	}
2531 
2532 	input.adv.odtimpedance = popts->odt ? popts->odt : 60;
2533 	input.adv.tx_impedance = popts->phy_tx_impedance ?
2534 					popts->phy_tx_impedance : 28;
2535 	input.adv.atx_impedance = popts->phy_atx_impedance ?
2536 					popts->phy_atx_impedance : 30;
2537 
2538 	debug("Initializing input adv data structure\n");
2539 	phy_gen2_init_input(&input);
2540 
2541 	debug("Initializing message block\n");
2542 	ret = phy_gen2_msg_init(&msg_1d, &msg_2d, &input);
2543 	if (ret != 0) {
2544 		ERROR("Init msg failed (error code %d)\n", ret);
2545 		return ret;
2546 	}
2547 
2548 	ret = c_init_phy_config(priv->phy, priv->ip_rev, &input, &msg_1d);
2549 	if (ret != 0) {
2550 		ERROR("Init PHY failed (error code %d)\n", ret);
2551 		return ret;
2552 	}
2553 #ifdef NXP_WARM_BOOT
2554 	debug("Warm boot flag value %0x\n", priv->warm_boot_flag);
2555 	if (priv->warm_boot_flag == DDR_WARM_BOOT) {
2556 		debug("Restoring the Phy training data\n");
2557 		// Restore the training data
2558 		ret = restore_phy_training_values(priv->phy,
2559 						  PHY_TRAINING_REGS_ON_FLASH,
2560 						  priv->num_ctlrs,
2561 						  input.basic.train2d);
2562 		if (ret != 0) {
2563 			ERROR("Restoring of training data failed %d\n", ret);
2564 			return ret;
2565 		}
2566 	} else {
2567 #endif
2568 		/* Mapping IMG buffer firstly */
2569 		ret = mmap_add_dynamic_region(priv->phy_gen2_fw_img_buf,
2570 			priv->phy_gen2_fw_img_buf,
2571 			PHY_GEN2_MAX_IMAGE_SIZE,
2572 			MT_MEMORY | MT_RW | MT_SECURE);
2573 		if (ret != 0) {
2574 			ERROR("Failed to add dynamic memory region.\n");
2575 			return ret;
2576 		}
2577 
2578 		debug("Load 1D firmware\n");
2579 		ret = load_fw(priv->phy, &input, 0, &msg_1d,
2580 			      sizeof(struct ddr4u1d), priv->phy_gen2_fw_img_buf,
2581 					priv->img_loadr, priv->warm_boot_flag);
2582 		if (ret != 0) {
2583 			ERROR("Loading firmware failed (error code %d)\n", ret);
2584 			return ret;
2585 		}
2586 
2587 		debug("Execute firmware\n");
2588 		ret = g_exec_fw(priv->phy, 0, &input);
2589 		if (ret != 0) {
2590 			ERROR("Execution FW failed (error code %d)\n", ret);
2591 		}
2592 
2593 #ifdef NXP_APPLY_MAX_CDD
2594 		soc_info = get_soc_info();
2595 		if (soc_info->svr_reg.bf.maj_ver == 2) {
2596 			tcfg0 = regs->timing_cfg[0];
2597 			tcfg4 = regs->timing_cfg[4];
2598 			rank = findrank(conf->cs_in_use);
2599 			get_cdd_val(priv->phy, rank, input.basic.frequency,
2600 					&tcfg0, &tcfg4);
2601 			regs->timing_cfg[0] = tcfg0;
2602 			regs->timing_cfg[4] = tcfg4;
2603 		}
2604 #endif
2605 
2606 		if ((ret == 0) && (input.basic.train2d != 0)) {
2607 			/* 2D training starts here */
2608 			debug("Load 2D firmware\n");
2609 			ret = load_fw(priv->phy, &input, 1, &msg_2d,
2610 				      sizeof(struct ddr4u2d),
2611 				      priv->phy_gen2_fw_img_buf,
2612 				      priv->img_loadr,
2613 				      priv->warm_boot_flag);
2614 			if (ret != 0) {
2615 				ERROR("Loading fw failed (err code %d)\n", ret);
2616 			} else {
2617 				debug("Execute 2D firmware\n");
2618 				ret = g_exec_fw(priv->phy, 1, &input);
2619 				if (ret != 0) {
2620 					ERROR("Execution FW failed (err %d)\n",
2621 					       ret);
2622 				}
2623 			}
2624 		}
2625 #ifdef NXP_WARM_BOOT
2626 		if (priv->warm_boot_flag != DDR_WRM_BOOT_NT_SUPPORTED &&
2627 		    ret == 0) {
2628 			debug("save the phy training data\n");
2629 			//Save training data TBD
2630 			ret = save_phy_training_values(priv->phy,
2631 						PHY_TRAINING_REGS_ON_FLASH,
2632 						priv->num_ctlrs,
2633 						input.basic.train2d);
2634 			if (ret != 0) {
2635 				ERROR("Saving training data failed.");
2636 				ERROR("Warm boot will fail. Error=%d.\n", ret);
2637 			}
2638 		}
2639 	} /* else */
2640 #endif
2641 
2642 	if (ret == 0) {
2643 		debug("Load PIE\n");
2644 		i_load_pie(priv->phy, &input, &msg_1d);
2645 
2646 		NOTICE("DDR4 %s with %d-rank %d-bit bus (x%d)\n",
2647 		       input.basic.dimm_type == RDIMM ? "RDIMM" :
2648 		       input.basic.dimm_type == LRDIMM ? "LRDIMM" :
2649 		       "UDIMM",
2650 		       dimm_param->n_ranks,
2651 		       dimm_param->primary_sdram_width,
2652 		       dimm_param->device_width);
2653 	}
2654 #ifdef DEBUG_DDR_INPUT_CONFIG
2655 	print_jason_format(&input, &msg_1d, &msg_2d);
2656 #endif
2657 
2658 	return ret;
2659 }
2660