1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  Copyright (C) 2017 Chelsio Communications.  All rights reserved.
4  */
5 
6 #include <linux/sort.h>
7 #include <linux/string.h>
8 
9 #include "t4_regs.h"
10 #include "cxgb4.h"
11 #include "cxgb4_cudbg.h"
12 #include "cudbg_if.h"
13 #include "cudbg_lib_common.h"
14 #include "cudbg_entity.h"
15 #include "cudbg_lib.h"
16 #include "cudbg_zlib.h"
17 
18 static int cudbg_do_compression(struct cudbg_init *pdbg_init,
19 				struct cudbg_buffer *pin_buff,
20 				struct cudbg_buffer *dbg_buff)
21 {
22 	struct cudbg_buffer temp_in_buff = { 0 };
23 	int bytes_left, bytes_read, bytes;
24 	u32 offset = dbg_buff->offset;
25 	int rc;
26 
27 	temp_in_buff.offset = pin_buff->offset;
28 	temp_in_buff.data = pin_buff->data;
29 	temp_in_buff.size = pin_buff->size;
30 
31 	bytes_left = pin_buff->size;
32 	bytes_read = 0;
33 	while (bytes_left > 0) {
34 		/* Do compression in smaller chunks */
35 		bytes = min_t(unsigned long, bytes_left,
36 			      (unsigned long)CUDBG_CHUNK_SIZE);
37 		temp_in_buff.data = (char *)pin_buff->data + bytes_read;
38 		temp_in_buff.size = bytes;
39 		rc = cudbg_compress_buff(pdbg_init, &temp_in_buff, dbg_buff);
40 		if (rc)
41 			return rc;
42 		bytes_left -= bytes;
43 		bytes_read += bytes;
44 	}
45 
46 	pin_buff->size = dbg_buff->offset - offset;
47 	return 0;
48 }
49 
50 static int cudbg_write_and_release_buff(struct cudbg_init *pdbg_init,
51 					struct cudbg_buffer *pin_buff,
52 					struct cudbg_buffer *dbg_buff)
53 {
54 	int rc = 0;
55 
56 	if (pdbg_init->compress_type == CUDBG_COMPRESSION_NONE) {
57 		cudbg_update_buff(pin_buff, dbg_buff);
58 	} else {
59 		rc = cudbg_do_compression(pdbg_init, pin_buff, dbg_buff);
60 		if (rc)
61 			goto out;
62 	}
63 
64 out:
65 	cudbg_put_buff(pdbg_init, pin_buff);
66 	return rc;
67 }
68 
69 static int is_fw_attached(struct cudbg_init *pdbg_init)
70 {
71 	struct adapter *padap = pdbg_init->adap;
72 
73 	if (!(padap->flags & CXGB4_FW_OK) || padap->use_bd)
74 		return 0;
75 
76 	return 1;
77 }
78 
79 /* This function will add additional padding bytes into debug_buffer to make it
80  * 4 byte aligned.
81  */
82 void cudbg_align_debug_buffer(struct cudbg_buffer *dbg_buff,
83 			      struct cudbg_entity_hdr *entity_hdr)
84 {
85 	u8 zero_buf[4] = {0};
86 	u8 padding, remain;
87 
88 	remain = (dbg_buff->offset - entity_hdr->start_offset) % 4;
89 	padding = 4 - remain;
90 	if (remain) {
91 		memcpy(((u8 *)dbg_buff->data) + dbg_buff->offset, &zero_buf,
92 		       padding);
93 		dbg_buff->offset += padding;
94 		entity_hdr->num_pad = padding;
95 	}
96 	entity_hdr->size = dbg_buff->offset - entity_hdr->start_offset;
97 }
98 
99 struct cudbg_entity_hdr *cudbg_get_entity_hdr(void *outbuf, int i)
100 {
101 	struct cudbg_hdr *cudbg_hdr = (struct cudbg_hdr *)outbuf;
102 
103 	return (struct cudbg_entity_hdr *)
104 	       ((char *)outbuf + cudbg_hdr->hdr_len +
105 		(sizeof(struct cudbg_entity_hdr) * (i - 1)));
106 }
107 
108 static int cudbg_read_vpd_reg(struct adapter *padap, u32 addr, u32 len,
109 			      void *dest)
110 {
111 	int vaddr, rc;
112 
113 	vaddr = t4_eeprom_ptov(addr, padap->pf, EEPROMPFSIZE);
114 	if (vaddr < 0)
115 		return vaddr;
116 
117 	rc = pci_read_vpd(padap->pdev, vaddr, len, dest);
118 	if (rc < 0)
119 		return rc;
120 
121 	return 0;
122 }
123 
124 static int cudbg_mem_desc_cmp(const void *a, const void *b)
125 {
126 	return ((const struct cudbg_mem_desc *)a)->base -
127 	       ((const struct cudbg_mem_desc *)b)->base;
128 }
129 
130 int cudbg_fill_meminfo(struct adapter *padap,
131 		       struct cudbg_meminfo *meminfo_buff)
132 {
133 	struct cudbg_mem_desc *md;
134 	u32 lo, hi, used, alloc;
135 	int n, i;
136 
137 	memset(meminfo_buff->avail, 0,
138 	       ARRAY_SIZE(meminfo_buff->avail) *
139 	       sizeof(struct cudbg_mem_desc));
140 	memset(meminfo_buff->mem, 0,
141 	       (ARRAY_SIZE(cudbg_region) + 3) * sizeof(struct cudbg_mem_desc));
142 	md  = meminfo_buff->mem;
143 
144 	for (i = 0; i < ARRAY_SIZE(meminfo_buff->mem); i++) {
145 		meminfo_buff->mem[i].limit = 0;
146 		meminfo_buff->mem[i].idx = i;
147 	}
148 
149 	/* Find and sort the populated memory ranges */
150 	i = 0;
151 	lo = t4_read_reg(padap, MA_TARGET_MEM_ENABLE_A);
152 	if (lo & EDRAM0_ENABLE_F) {
153 		hi = t4_read_reg(padap, MA_EDRAM0_BAR_A);
154 		meminfo_buff->avail[i].base =
155 			cudbg_mbytes_to_bytes(EDRAM0_BASE_G(hi));
156 		meminfo_buff->avail[i].limit =
157 			meminfo_buff->avail[i].base +
158 			cudbg_mbytes_to_bytes(EDRAM0_SIZE_G(hi));
159 		meminfo_buff->avail[i].idx = 0;
160 		i++;
161 	}
162 
163 	if (lo & EDRAM1_ENABLE_F) {
164 		hi =  t4_read_reg(padap, MA_EDRAM1_BAR_A);
165 		meminfo_buff->avail[i].base =
166 			cudbg_mbytes_to_bytes(EDRAM1_BASE_G(hi));
167 		meminfo_buff->avail[i].limit =
168 			meminfo_buff->avail[i].base +
169 			cudbg_mbytes_to_bytes(EDRAM1_SIZE_G(hi));
170 		meminfo_buff->avail[i].idx = 1;
171 		i++;
172 	}
173 
174 	if (is_t5(padap->params.chip)) {
175 		if (lo & EXT_MEM0_ENABLE_F) {
176 			hi = t4_read_reg(padap, MA_EXT_MEMORY0_BAR_A);
177 			meminfo_buff->avail[i].base =
178 				cudbg_mbytes_to_bytes(EXT_MEM_BASE_G(hi));
179 			meminfo_buff->avail[i].limit =
180 				meminfo_buff->avail[i].base +
181 				cudbg_mbytes_to_bytes(EXT_MEM_SIZE_G(hi));
182 			meminfo_buff->avail[i].idx = 3;
183 			i++;
184 		}
185 
186 		if (lo & EXT_MEM1_ENABLE_F) {
187 			hi = t4_read_reg(padap, MA_EXT_MEMORY1_BAR_A);
188 			meminfo_buff->avail[i].base =
189 				cudbg_mbytes_to_bytes(EXT_MEM1_BASE_G(hi));
190 			meminfo_buff->avail[i].limit =
191 				meminfo_buff->avail[i].base +
192 				cudbg_mbytes_to_bytes(EXT_MEM1_SIZE_G(hi));
193 			meminfo_buff->avail[i].idx = 4;
194 			i++;
195 		}
196 	} else {
197 		if (lo & EXT_MEM_ENABLE_F) {
198 			hi = t4_read_reg(padap, MA_EXT_MEMORY_BAR_A);
199 			meminfo_buff->avail[i].base =
200 				cudbg_mbytes_to_bytes(EXT_MEM_BASE_G(hi));
201 			meminfo_buff->avail[i].limit =
202 				meminfo_buff->avail[i].base +
203 				cudbg_mbytes_to_bytes(EXT_MEM_SIZE_G(hi));
204 			meminfo_buff->avail[i].idx = 2;
205 			i++;
206 		}
207 
208 		if (lo & HMA_MUX_F) {
209 			hi = t4_read_reg(padap, MA_EXT_MEMORY1_BAR_A);
210 			meminfo_buff->avail[i].base =
211 				cudbg_mbytes_to_bytes(EXT_MEM1_BASE_G(hi));
212 			meminfo_buff->avail[i].limit =
213 				meminfo_buff->avail[i].base +
214 				cudbg_mbytes_to_bytes(EXT_MEM1_SIZE_G(hi));
215 			meminfo_buff->avail[i].idx = 5;
216 			i++;
217 		}
218 	}
219 
220 	if (!i) /* no memory available */
221 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
222 
223 	meminfo_buff->avail_c = i;
224 	sort(meminfo_buff->avail, i, sizeof(struct cudbg_mem_desc),
225 	     cudbg_mem_desc_cmp, NULL);
226 	(md++)->base = t4_read_reg(padap, SGE_DBQ_CTXT_BADDR_A);
227 	(md++)->base = t4_read_reg(padap, SGE_IMSG_CTXT_BADDR_A);
228 	(md++)->base = t4_read_reg(padap, SGE_FLM_CACHE_BADDR_A);
229 	(md++)->base = t4_read_reg(padap, TP_CMM_TCB_BASE_A);
230 	(md++)->base = t4_read_reg(padap, TP_CMM_MM_BASE_A);
231 	(md++)->base = t4_read_reg(padap, TP_CMM_TIMER_BASE_A);
232 	(md++)->base = t4_read_reg(padap, TP_CMM_MM_RX_FLST_BASE_A);
233 	(md++)->base = t4_read_reg(padap, TP_CMM_MM_TX_FLST_BASE_A);
234 	(md++)->base = t4_read_reg(padap, TP_CMM_MM_PS_FLST_BASE_A);
235 
236 	/* the next few have explicit upper bounds */
237 	md->base = t4_read_reg(padap, TP_PMM_TX_BASE_A);
238 	md->limit = md->base - 1 +
239 		    t4_read_reg(padap, TP_PMM_TX_PAGE_SIZE_A) *
240 		    PMTXMAXPAGE_G(t4_read_reg(padap, TP_PMM_TX_MAX_PAGE_A));
241 	md++;
242 
243 	md->base = t4_read_reg(padap, TP_PMM_RX_BASE_A);
244 	md->limit = md->base - 1 +
245 		    t4_read_reg(padap, TP_PMM_RX_PAGE_SIZE_A) *
246 		    PMRXMAXPAGE_G(t4_read_reg(padap, TP_PMM_RX_MAX_PAGE_A));
247 	md++;
248 
249 	if (t4_read_reg(padap, LE_DB_CONFIG_A) & HASHEN_F) {
250 		if (CHELSIO_CHIP_VERSION(padap->params.chip) <= CHELSIO_T5) {
251 			hi = t4_read_reg(padap, LE_DB_TID_HASHBASE_A) / 4;
252 			md->base = t4_read_reg(padap, LE_DB_HASH_TID_BASE_A);
253 		} else {
254 			hi = t4_read_reg(padap, LE_DB_HASH_TID_BASE_A);
255 			md->base = t4_read_reg(padap,
256 					       LE_DB_HASH_TBL_BASE_ADDR_A);
257 		}
258 		md->limit = 0;
259 	} else {
260 		md->base = 0;
261 		md->idx = ARRAY_SIZE(cudbg_region);  /* hide it */
262 	}
263 	md++;
264 
265 #define ulp_region(reg) do { \
266 	md->base = t4_read_reg(padap, ULP_ ## reg ## _LLIMIT_A);\
267 	(md++)->limit = t4_read_reg(padap, ULP_ ## reg ## _ULIMIT_A);\
268 } while (0)
269 
270 	ulp_region(RX_ISCSI);
271 	ulp_region(RX_TDDP);
272 	ulp_region(TX_TPT);
273 	ulp_region(RX_STAG);
274 	ulp_region(RX_RQ);
275 	ulp_region(RX_RQUDP);
276 	ulp_region(RX_PBL);
277 	ulp_region(TX_PBL);
278 #undef ulp_region
279 	md->base = 0;
280 	md->idx = ARRAY_SIZE(cudbg_region);
281 	if (!is_t4(padap->params.chip)) {
282 		u32 fifo_size = t4_read_reg(padap, SGE_DBVFIFO_SIZE_A);
283 		u32 sge_ctrl = t4_read_reg(padap, SGE_CONTROL2_A);
284 		u32 size = 0;
285 
286 		if (is_t5(padap->params.chip)) {
287 			if (sge_ctrl & VFIFO_ENABLE_F)
288 				size = DBVFIFO_SIZE_G(fifo_size);
289 		} else {
290 			size = T6_DBVFIFO_SIZE_G(fifo_size);
291 		}
292 
293 		if (size) {
294 			md->base = BASEADDR_G(t4_read_reg(padap,
295 							  SGE_DBVFIFO_BADDR_A));
296 			md->limit = md->base + (size << 2) - 1;
297 		}
298 	}
299 
300 	md++;
301 
302 	md->base = t4_read_reg(padap, ULP_RX_CTX_BASE_A);
303 	md->limit = 0;
304 	md++;
305 	md->base = t4_read_reg(padap, ULP_TX_ERR_TABLE_BASE_A);
306 	md->limit = 0;
307 	md++;
308 
309 	md->base = padap->vres.ocq.start;
310 	if (padap->vres.ocq.size)
311 		md->limit = md->base + padap->vres.ocq.size - 1;
312 	else
313 		md->idx = ARRAY_SIZE(cudbg_region);  /* hide it */
314 	md++;
315 
316 	/* add any address-space holes, there can be up to 3 */
317 	for (n = 0; n < i - 1; n++)
318 		if (meminfo_buff->avail[n].limit <
319 		    meminfo_buff->avail[n + 1].base)
320 			(md++)->base = meminfo_buff->avail[n].limit;
321 
322 	if (meminfo_buff->avail[n].limit)
323 		(md++)->base = meminfo_buff->avail[n].limit;
324 
325 	n = md - meminfo_buff->mem;
326 	meminfo_buff->mem_c = n;
327 
328 	sort(meminfo_buff->mem, n, sizeof(struct cudbg_mem_desc),
329 	     cudbg_mem_desc_cmp, NULL);
330 
331 	lo = t4_read_reg(padap, CIM_SDRAM_BASE_ADDR_A);
332 	hi = t4_read_reg(padap, CIM_SDRAM_ADDR_SIZE_A) + lo - 1;
333 	meminfo_buff->up_ram_lo = lo;
334 	meminfo_buff->up_ram_hi = hi;
335 
336 	lo = t4_read_reg(padap, CIM_EXTMEM2_BASE_ADDR_A);
337 	hi = t4_read_reg(padap, CIM_EXTMEM2_ADDR_SIZE_A) + lo - 1;
338 	meminfo_buff->up_extmem2_lo = lo;
339 	meminfo_buff->up_extmem2_hi = hi;
340 
341 	lo = t4_read_reg(padap, TP_PMM_RX_MAX_PAGE_A);
342 	for (i = 0, meminfo_buff->free_rx_cnt = 0; i < 2; i++)
343 		meminfo_buff->free_rx_cnt +=
344 			FREERXPAGECOUNT_G(t4_read_reg(padap,
345 						      TP_FLM_FREE_RX_CNT_A));
346 
347 	meminfo_buff->rx_pages_data[0] =  PMRXMAXPAGE_G(lo);
348 	meminfo_buff->rx_pages_data[1] =
349 		t4_read_reg(padap, TP_PMM_RX_PAGE_SIZE_A) >> 10;
350 	meminfo_buff->rx_pages_data[2] = (lo & PMRXNUMCHN_F) ? 2 : 1;
351 
352 	lo = t4_read_reg(padap, TP_PMM_TX_MAX_PAGE_A);
353 	hi = t4_read_reg(padap, TP_PMM_TX_PAGE_SIZE_A);
354 	for (i = 0, meminfo_buff->free_tx_cnt = 0; i < 4; i++)
355 		meminfo_buff->free_tx_cnt +=
356 			FREETXPAGECOUNT_G(t4_read_reg(padap,
357 						      TP_FLM_FREE_TX_CNT_A));
358 
359 	meminfo_buff->tx_pages_data[0] = PMTXMAXPAGE_G(lo);
360 	meminfo_buff->tx_pages_data[1] =
361 		hi >= (1 << 20) ? (hi >> 20) : (hi >> 10);
362 	meminfo_buff->tx_pages_data[2] =
363 		hi >= (1 << 20) ? 'M' : 'K';
364 	meminfo_buff->tx_pages_data[3] = 1 << PMTXNUMCHN_G(lo);
365 
366 	meminfo_buff->p_structs = t4_read_reg(padap, TP_CMM_MM_MAX_PSTRUCT_A);
367 	meminfo_buff->p_structs_free_cnt =
368 		FREEPSTRUCTCOUNT_G(t4_read_reg(padap, TP_FLM_FREE_PS_CNT_A));
369 
370 	for (i = 0; i < 4; i++) {
371 		if (CHELSIO_CHIP_VERSION(padap->params.chip) > CHELSIO_T5)
372 			lo = t4_read_reg(padap,
373 					 MPS_RX_MAC_BG_PG_CNT0_A + i * 4);
374 		else
375 			lo = t4_read_reg(padap, MPS_RX_PG_RSV0_A + i * 4);
376 		if (is_t5(padap->params.chip)) {
377 			used = T5_USED_G(lo);
378 			alloc = T5_ALLOC_G(lo);
379 		} else {
380 			used = USED_G(lo);
381 			alloc = ALLOC_G(lo);
382 		}
383 		meminfo_buff->port_used[i] = used;
384 		meminfo_buff->port_alloc[i] = alloc;
385 	}
386 
387 	for (i = 0; i < padap->params.arch.nchan; i++) {
388 		if (CHELSIO_CHIP_VERSION(padap->params.chip) > CHELSIO_T5)
389 			lo = t4_read_reg(padap,
390 					 MPS_RX_LPBK_BG_PG_CNT0_A + i * 4);
391 		else
392 			lo = t4_read_reg(padap, MPS_RX_PG_RSV4_A + i * 4);
393 		if (is_t5(padap->params.chip)) {
394 			used = T5_USED_G(lo);
395 			alloc = T5_ALLOC_G(lo);
396 		} else {
397 			used = USED_G(lo);
398 			alloc = ALLOC_G(lo);
399 		}
400 		meminfo_buff->loopback_used[i] = used;
401 		meminfo_buff->loopback_alloc[i] = alloc;
402 	}
403 
404 	return 0;
405 }
406 
407 int cudbg_collect_reg_dump(struct cudbg_init *pdbg_init,
408 			   struct cudbg_buffer *dbg_buff,
409 			   struct cudbg_error *cudbg_err)
410 {
411 	struct adapter *padap = pdbg_init->adap;
412 	struct cudbg_buffer temp_buff = { 0 };
413 	u32 buf_size = 0;
414 	int rc = 0;
415 
416 	if (is_t4(padap->params.chip))
417 		buf_size = T4_REGMAP_SIZE;
418 	else if (is_t5(padap->params.chip) || is_t6(padap->params.chip))
419 		buf_size = T5_REGMAP_SIZE;
420 
421 	rc = cudbg_get_buff(pdbg_init, dbg_buff, buf_size, &temp_buff);
422 	if (rc)
423 		return rc;
424 	t4_get_regs(padap, (void *)temp_buff.data, temp_buff.size);
425 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
426 }
427 
428 int cudbg_collect_fw_devlog(struct cudbg_init *pdbg_init,
429 			    struct cudbg_buffer *dbg_buff,
430 			    struct cudbg_error *cudbg_err)
431 {
432 	struct adapter *padap = pdbg_init->adap;
433 	struct cudbg_buffer temp_buff = { 0 };
434 	struct devlog_params *dparams;
435 	int rc = 0;
436 
437 	rc = t4_init_devlog_params(padap);
438 	if (rc < 0) {
439 		cudbg_err->sys_err = rc;
440 		return rc;
441 	}
442 
443 	dparams = &padap->params.devlog;
444 	rc = cudbg_get_buff(pdbg_init, dbg_buff, dparams->size, &temp_buff);
445 	if (rc)
446 		return rc;
447 
448 	/* Collect FW devlog */
449 	if (dparams->start != 0) {
450 		spin_lock(&padap->win0_lock);
451 		rc = t4_memory_rw(padap, padap->params.drv_memwin,
452 				  dparams->memtype, dparams->start,
453 				  dparams->size,
454 				  (__be32 *)(char *)temp_buff.data,
455 				  1);
456 		spin_unlock(&padap->win0_lock);
457 		if (rc) {
458 			cudbg_err->sys_err = rc;
459 			cudbg_put_buff(pdbg_init, &temp_buff);
460 			return rc;
461 		}
462 	}
463 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
464 }
465 
466 int cudbg_collect_cim_la(struct cudbg_init *pdbg_init,
467 			 struct cudbg_buffer *dbg_buff,
468 			 struct cudbg_error *cudbg_err)
469 {
470 	struct adapter *padap = pdbg_init->adap;
471 	struct cudbg_buffer temp_buff = { 0 };
472 	int size, rc;
473 	u32 cfg = 0;
474 
475 	if (is_t6(padap->params.chip)) {
476 		size = padap->params.cim_la_size / 10 + 1;
477 		size *= 10 * sizeof(u32);
478 	} else {
479 		size = padap->params.cim_la_size / 8;
480 		size *= 8 * sizeof(u32);
481 	}
482 
483 	size += sizeof(cfg);
484 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
485 	if (rc)
486 		return rc;
487 
488 	rc = t4_cim_read(padap, UP_UP_DBG_LA_CFG_A, 1, &cfg);
489 	if (rc) {
490 		cudbg_err->sys_err = rc;
491 		cudbg_put_buff(pdbg_init, &temp_buff);
492 		return rc;
493 	}
494 
495 	memcpy((char *)temp_buff.data, &cfg, sizeof(cfg));
496 	rc = t4_cim_read_la(padap,
497 			    (u32 *)((char *)temp_buff.data + sizeof(cfg)),
498 			    NULL);
499 	if (rc < 0) {
500 		cudbg_err->sys_err = rc;
501 		cudbg_put_buff(pdbg_init, &temp_buff);
502 		return rc;
503 	}
504 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
505 }
506 
507 int cudbg_collect_cim_ma_la(struct cudbg_init *pdbg_init,
508 			    struct cudbg_buffer *dbg_buff,
509 			    struct cudbg_error *cudbg_err)
510 {
511 	struct adapter *padap = pdbg_init->adap;
512 	struct cudbg_buffer temp_buff = { 0 };
513 	int size, rc;
514 
515 	size = 2 * CIM_MALA_SIZE * 5 * sizeof(u32);
516 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
517 	if (rc)
518 		return rc;
519 
520 	t4_cim_read_ma_la(padap,
521 			  (u32 *)temp_buff.data,
522 			  (u32 *)((char *)temp_buff.data +
523 				  5 * CIM_MALA_SIZE));
524 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
525 }
526 
527 int cudbg_collect_cim_qcfg(struct cudbg_init *pdbg_init,
528 			   struct cudbg_buffer *dbg_buff,
529 			   struct cudbg_error *cudbg_err)
530 {
531 	struct adapter *padap = pdbg_init->adap;
532 	struct cudbg_buffer temp_buff = { 0 };
533 	struct cudbg_cim_qcfg *cim_qcfg_data;
534 	int rc;
535 
536 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_cim_qcfg),
537 			    &temp_buff);
538 	if (rc)
539 		return rc;
540 
541 	cim_qcfg_data = (struct cudbg_cim_qcfg *)temp_buff.data;
542 	cim_qcfg_data->chip = padap->params.chip;
543 	rc = t4_cim_read(padap, UP_IBQ_0_RDADDR_A,
544 			 ARRAY_SIZE(cim_qcfg_data->stat), cim_qcfg_data->stat);
545 	if (rc) {
546 		cudbg_err->sys_err = rc;
547 		cudbg_put_buff(pdbg_init, &temp_buff);
548 		return rc;
549 	}
550 
551 	rc = t4_cim_read(padap, UP_OBQ_0_REALADDR_A,
552 			 ARRAY_SIZE(cim_qcfg_data->obq_wr),
553 			 cim_qcfg_data->obq_wr);
554 	if (rc) {
555 		cudbg_err->sys_err = rc;
556 		cudbg_put_buff(pdbg_init, &temp_buff);
557 		return rc;
558 	}
559 
560 	t4_read_cimq_cfg(padap, cim_qcfg_data->base, cim_qcfg_data->size,
561 			 cim_qcfg_data->thres);
562 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
563 }
564 
565 static int cudbg_read_cim_ibq(struct cudbg_init *pdbg_init,
566 			      struct cudbg_buffer *dbg_buff,
567 			      struct cudbg_error *cudbg_err, int qid)
568 {
569 	struct adapter *padap = pdbg_init->adap;
570 	struct cudbg_buffer temp_buff = { 0 };
571 	int no_of_read_words, rc = 0;
572 	u32 qsize;
573 
574 	/* collect CIM IBQ */
575 	qsize = CIM_IBQ_SIZE * 4 * sizeof(u32);
576 	rc = cudbg_get_buff(pdbg_init, dbg_buff, qsize, &temp_buff);
577 	if (rc)
578 		return rc;
579 
580 	/* t4_read_cim_ibq will return no. of read words or error */
581 	no_of_read_words = t4_read_cim_ibq(padap, qid,
582 					   (u32 *)temp_buff.data, qsize);
583 	/* no_of_read_words is less than or equal to 0 means error */
584 	if (no_of_read_words <= 0) {
585 		if (!no_of_read_words)
586 			rc = CUDBG_SYSTEM_ERROR;
587 		else
588 			rc = no_of_read_words;
589 		cudbg_err->sys_err = rc;
590 		cudbg_put_buff(pdbg_init, &temp_buff);
591 		return rc;
592 	}
593 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
594 }
595 
596 int cudbg_collect_cim_ibq_tp0(struct cudbg_init *pdbg_init,
597 			      struct cudbg_buffer *dbg_buff,
598 			      struct cudbg_error *cudbg_err)
599 {
600 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 0);
601 }
602 
603 int cudbg_collect_cim_ibq_tp1(struct cudbg_init *pdbg_init,
604 			      struct cudbg_buffer *dbg_buff,
605 			      struct cudbg_error *cudbg_err)
606 {
607 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 1);
608 }
609 
610 int cudbg_collect_cim_ibq_ulp(struct cudbg_init *pdbg_init,
611 			      struct cudbg_buffer *dbg_buff,
612 			      struct cudbg_error *cudbg_err)
613 {
614 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 2);
615 }
616 
617 int cudbg_collect_cim_ibq_sge0(struct cudbg_init *pdbg_init,
618 			       struct cudbg_buffer *dbg_buff,
619 			       struct cudbg_error *cudbg_err)
620 {
621 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 3);
622 }
623 
624 int cudbg_collect_cim_ibq_sge1(struct cudbg_init *pdbg_init,
625 			       struct cudbg_buffer *dbg_buff,
626 			       struct cudbg_error *cudbg_err)
627 {
628 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 4);
629 }
630 
631 int cudbg_collect_cim_ibq_ncsi(struct cudbg_init *pdbg_init,
632 			       struct cudbg_buffer *dbg_buff,
633 			       struct cudbg_error *cudbg_err)
634 {
635 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 5);
636 }
637 
638 u32 cudbg_cim_obq_size(struct adapter *padap, int qid)
639 {
640 	u32 value;
641 
642 	t4_write_reg(padap, CIM_QUEUE_CONFIG_REF_A, OBQSELECT_F |
643 		     QUENUMSELECT_V(qid));
644 	value = t4_read_reg(padap, CIM_QUEUE_CONFIG_CTRL_A);
645 	value = CIMQSIZE_G(value) * 64; /* size in number of words */
646 	return value * sizeof(u32);
647 }
648 
649 static int cudbg_read_cim_obq(struct cudbg_init *pdbg_init,
650 			      struct cudbg_buffer *dbg_buff,
651 			      struct cudbg_error *cudbg_err, int qid)
652 {
653 	struct adapter *padap = pdbg_init->adap;
654 	struct cudbg_buffer temp_buff = { 0 };
655 	int no_of_read_words, rc = 0;
656 	u32 qsize;
657 
658 	/* collect CIM OBQ */
659 	qsize =  cudbg_cim_obq_size(padap, qid);
660 	rc = cudbg_get_buff(pdbg_init, dbg_buff, qsize, &temp_buff);
661 	if (rc)
662 		return rc;
663 
664 	/* t4_read_cim_obq will return no. of read words or error */
665 	no_of_read_words = t4_read_cim_obq(padap, qid,
666 					   (u32 *)temp_buff.data, qsize);
667 	/* no_of_read_words is less than or equal to 0 means error */
668 	if (no_of_read_words <= 0) {
669 		if (!no_of_read_words)
670 			rc = CUDBG_SYSTEM_ERROR;
671 		else
672 			rc = no_of_read_words;
673 		cudbg_err->sys_err = rc;
674 		cudbg_put_buff(pdbg_init, &temp_buff);
675 		return rc;
676 	}
677 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
678 }
679 
680 int cudbg_collect_cim_obq_ulp0(struct cudbg_init *pdbg_init,
681 			       struct cudbg_buffer *dbg_buff,
682 			       struct cudbg_error *cudbg_err)
683 {
684 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 0);
685 }
686 
687 int cudbg_collect_cim_obq_ulp1(struct cudbg_init *pdbg_init,
688 			       struct cudbg_buffer *dbg_buff,
689 			       struct cudbg_error *cudbg_err)
690 {
691 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 1);
692 }
693 
694 int cudbg_collect_cim_obq_ulp2(struct cudbg_init *pdbg_init,
695 			       struct cudbg_buffer *dbg_buff,
696 			       struct cudbg_error *cudbg_err)
697 {
698 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 2);
699 }
700 
701 int cudbg_collect_cim_obq_ulp3(struct cudbg_init *pdbg_init,
702 			       struct cudbg_buffer *dbg_buff,
703 			       struct cudbg_error *cudbg_err)
704 {
705 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 3);
706 }
707 
708 int cudbg_collect_cim_obq_sge(struct cudbg_init *pdbg_init,
709 			      struct cudbg_buffer *dbg_buff,
710 			      struct cudbg_error *cudbg_err)
711 {
712 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 4);
713 }
714 
715 int cudbg_collect_cim_obq_ncsi(struct cudbg_init *pdbg_init,
716 			       struct cudbg_buffer *dbg_buff,
717 			       struct cudbg_error *cudbg_err)
718 {
719 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 5);
720 }
721 
722 int cudbg_collect_obq_sge_rx_q0(struct cudbg_init *pdbg_init,
723 				struct cudbg_buffer *dbg_buff,
724 				struct cudbg_error *cudbg_err)
725 {
726 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 6);
727 }
728 
729 int cudbg_collect_obq_sge_rx_q1(struct cudbg_init *pdbg_init,
730 				struct cudbg_buffer *dbg_buff,
731 				struct cudbg_error *cudbg_err)
732 {
733 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 7);
734 }
735 
736 static int cudbg_meminfo_get_mem_index(struct adapter *padap,
737 				       struct cudbg_meminfo *mem_info,
738 				       u8 mem_type, u8 *idx)
739 {
740 	u8 i, flag;
741 
742 	switch (mem_type) {
743 	case MEM_EDC0:
744 		flag = EDC0_FLAG;
745 		break;
746 	case MEM_EDC1:
747 		flag = EDC1_FLAG;
748 		break;
749 	case MEM_MC0:
750 		/* Some T5 cards have both MC0 and MC1. */
751 		flag = is_t5(padap->params.chip) ? MC0_FLAG : MC_FLAG;
752 		break;
753 	case MEM_MC1:
754 		flag = MC1_FLAG;
755 		break;
756 	case MEM_HMA:
757 		flag = HMA_FLAG;
758 		break;
759 	default:
760 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
761 	}
762 
763 	for (i = 0; i < mem_info->avail_c; i++) {
764 		if (mem_info->avail[i].idx == flag) {
765 			*idx = i;
766 			return 0;
767 		}
768 	}
769 
770 	return CUDBG_STATUS_ENTITY_NOT_FOUND;
771 }
772 
773 /* Fetch the @region_name's start and end from @meminfo. */
774 static int cudbg_get_mem_region(struct adapter *padap,
775 				struct cudbg_meminfo *meminfo,
776 				u8 mem_type, const char *region_name,
777 				struct cudbg_mem_desc *mem_desc)
778 {
779 	u8 mc, found = 0;
780 	u32 idx = 0;
781 	int rc, i;
782 
783 	rc = cudbg_meminfo_get_mem_index(padap, meminfo, mem_type, &mc);
784 	if (rc)
785 		return rc;
786 
787 	i = match_string(cudbg_region, ARRAY_SIZE(cudbg_region), region_name);
788 	if (i < 0)
789 		return -EINVAL;
790 
791 	idx = i;
792 	for (i = 0; i < meminfo->mem_c; i++) {
793 		if (meminfo->mem[i].idx >= ARRAY_SIZE(cudbg_region))
794 			continue; /* Skip holes */
795 
796 		if (!(meminfo->mem[i].limit))
797 			meminfo->mem[i].limit =
798 				i < meminfo->mem_c - 1 ?
799 				meminfo->mem[i + 1].base - 1 : ~0;
800 
801 		if (meminfo->mem[i].idx == idx) {
802 			/* Check if the region exists in @mem_type memory */
803 			if (meminfo->mem[i].base < meminfo->avail[mc].base &&
804 			    meminfo->mem[i].limit < meminfo->avail[mc].base)
805 				return -EINVAL;
806 
807 			if (meminfo->mem[i].base > meminfo->avail[mc].limit)
808 				return -EINVAL;
809 
810 			memcpy(mem_desc, &meminfo->mem[i],
811 			       sizeof(struct cudbg_mem_desc));
812 			found = 1;
813 			break;
814 		}
815 	}
816 	if (!found)
817 		return -EINVAL;
818 
819 	return 0;
820 }
821 
822 /* Fetch and update the start and end of the requested memory region w.r.t 0
823  * in the corresponding EDC/MC/HMA.
824  */
825 static int cudbg_get_mem_relative(struct adapter *padap,
826 				  struct cudbg_meminfo *meminfo,
827 				  u8 mem_type, u32 *out_base, u32 *out_end)
828 {
829 	u8 mc_idx;
830 	int rc;
831 
832 	rc = cudbg_meminfo_get_mem_index(padap, meminfo, mem_type, &mc_idx);
833 	if (rc)
834 		return rc;
835 
836 	if (*out_base < meminfo->avail[mc_idx].base)
837 		*out_base = 0;
838 	else
839 		*out_base -= meminfo->avail[mc_idx].base;
840 
841 	if (*out_end > meminfo->avail[mc_idx].limit)
842 		*out_end = meminfo->avail[mc_idx].limit;
843 	else
844 		*out_end -= meminfo->avail[mc_idx].base;
845 
846 	return 0;
847 }
848 
849 /* Get TX and RX Payload region */
850 static int cudbg_get_payload_range(struct adapter *padap, u8 mem_type,
851 				   const char *region_name,
852 				   struct cudbg_region_info *payload)
853 {
854 	struct cudbg_mem_desc mem_desc = { 0 };
855 	struct cudbg_meminfo meminfo;
856 	int rc;
857 
858 	rc = cudbg_fill_meminfo(padap, &meminfo);
859 	if (rc)
860 		return rc;
861 
862 	rc = cudbg_get_mem_region(padap, &meminfo, mem_type, region_name,
863 				  &mem_desc);
864 	if (rc) {
865 		payload->exist = false;
866 		return 0;
867 	}
868 
869 	payload->exist = true;
870 	payload->start = mem_desc.base;
871 	payload->end = mem_desc.limit;
872 
873 	return cudbg_get_mem_relative(padap, &meminfo, mem_type,
874 				      &payload->start, &payload->end);
875 }
876 
877 static int cudbg_memory_read(struct cudbg_init *pdbg_init, int win,
878 			     int mtype, u32 addr, u32 len, void *hbuf)
879 {
880 	u32 win_pf, memoffset, mem_aperture, mem_base;
881 	struct adapter *adap = pdbg_init->adap;
882 	u32 pos, offset, resid;
883 	u32 *res_buf;
884 	u64 *buf;
885 	int ret;
886 
887 	/* Argument sanity checks ...
888 	 */
889 	if (addr & 0x3 || (uintptr_t)hbuf & 0x3)
890 		return -EINVAL;
891 
892 	buf = (u64 *)hbuf;
893 
894 	/* Try to do 64-bit reads.  Residual will be handled later. */
895 	resid = len & 0x7;
896 	len -= resid;
897 
898 	ret = t4_memory_rw_init(adap, win, mtype, &memoffset, &mem_base,
899 				&mem_aperture);
900 	if (ret)
901 		return ret;
902 
903 	addr = addr + memoffset;
904 	win_pf = is_t4(adap->params.chip) ? 0 : PFNUM_V(adap->pf);
905 
906 	pos = addr & ~(mem_aperture - 1);
907 	offset = addr - pos;
908 
909 	/* Set up initial PCI-E Memory Window to cover the start of our
910 	 * transfer.
911 	 */
912 	t4_memory_update_win(adap, win, pos | win_pf);
913 
914 	/* Transfer data from the adapter */
915 	while (len > 0) {
916 		*buf++ = le64_to_cpu((__force __le64)
917 				     t4_read_reg64(adap, mem_base + offset));
918 		offset += sizeof(u64);
919 		len -= sizeof(u64);
920 
921 		/* If we've reached the end of our current window aperture,
922 		 * move the PCI-E Memory Window on to the next.
923 		 */
924 		if (offset == mem_aperture) {
925 			pos += mem_aperture;
926 			offset = 0;
927 			t4_memory_update_win(adap, win, pos | win_pf);
928 		}
929 	}
930 
931 	res_buf = (u32 *)buf;
932 	/* Read residual in 32-bit multiples */
933 	while (resid > sizeof(u32)) {
934 		*res_buf++ = le32_to_cpu((__force __le32)
935 					 t4_read_reg(adap, mem_base + offset));
936 		offset += sizeof(u32);
937 		resid -= sizeof(u32);
938 
939 		/* If we've reached the end of our current window aperture,
940 		 * move the PCI-E Memory Window on to the next.
941 		 */
942 		if (offset == mem_aperture) {
943 			pos += mem_aperture;
944 			offset = 0;
945 			t4_memory_update_win(adap, win, pos | win_pf);
946 		}
947 	}
948 
949 	/* Transfer residual < 32-bits */
950 	if (resid)
951 		t4_memory_rw_residual(adap, resid, mem_base + offset,
952 				      (u8 *)res_buf, T4_MEMORY_READ);
953 
954 	return 0;
955 }
956 
957 #define CUDBG_YIELD_ITERATION 256
958 
959 static int cudbg_read_fw_mem(struct cudbg_init *pdbg_init,
960 			     struct cudbg_buffer *dbg_buff, u8 mem_type,
961 			     unsigned long tot_len,
962 			     struct cudbg_error *cudbg_err)
963 {
964 	static const char * const region_name[] = { "Tx payload:",
965 						    "Rx payload:" };
966 	unsigned long bytes, bytes_left, bytes_read = 0;
967 	struct adapter *padap = pdbg_init->adap;
968 	struct cudbg_buffer temp_buff = { 0 };
969 	struct cudbg_region_info payload[2];
970 	u32 yield_count = 0;
971 	int rc = 0;
972 	u8 i;
973 
974 	/* Get TX/RX Payload region range if they exist */
975 	memset(payload, 0, sizeof(payload));
976 	for (i = 0; i < ARRAY_SIZE(region_name); i++) {
977 		rc = cudbg_get_payload_range(padap, mem_type, region_name[i],
978 					     &payload[i]);
979 		if (rc)
980 			return rc;
981 
982 		if (payload[i].exist) {
983 			/* Align start and end to avoid wrap around */
984 			payload[i].start = roundup(payload[i].start,
985 						   CUDBG_CHUNK_SIZE);
986 			payload[i].end = rounddown(payload[i].end,
987 						   CUDBG_CHUNK_SIZE);
988 		}
989 	}
990 
991 	bytes_left = tot_len;
992 	while (bytes_left > 0) {
993 		/* As MC size is huge and read through PIO access, this
994 		 * loop will hold cpu for a longer time. OS may think that
995 		 * the process is hanged and will generate CPU stall traces.
996 		 * So yield the cpu regularly.
997 		 */
998 		yield_count++;
999 		if (!(yield_count % CUDBG_YIELD_ITERATION))
1000 			schedule();
1001 
1002 		bytes = min_t(unsigned long, bytes_left,
1003 			      (unsigned long)CUDBG_CHUNK_SIZE);
1004 		rc = cudbg_get_buff(pdbg_init, dbg_buff, bytes, &temp_buff);
1005 		if (rc)
1006 			return rc;
1007 
1008 		for (i = 0; i < ARRAY_SIZE(payload); i++)
1009 			if (payload[i].exist &&
1010 			    bytes_read >= payload[i].start &&
1011 			    bytes_read + bytes <= payload[i].end)
1012 				/* TX and RX Payload regions can't overlap */
1013 				goto skip_read;
1014 
1015 		spin_lock(&padap->win0_lock);
1016 		rc = cudbg_memory_read(pdbg_init, MEMWIN_NIC, mem_type,
1017 				       bytes_read, bytes, temp_buff.data);
1018 		spin_unlock(&padap->win0_lock);
1019 		if (rc) {
1020 			cudbg_err->sys_err = rc;
1021 			cudbg_put_buff(pdbg_init, &temp_buff);
1022 			return rc;
1023 		}
1024 
1025 skip_read:
1026 		bytes_left -= bytes;
1027 		bytes_read += bytes;
1028 		rc = cudbg_write_and_release_buff(pdbg_init, &temp_buff,
1029 						  dbg_buff);
1030 		if (rc) {
1031 			cudbg_put_buff(pdbg_init, &temp_buff);
1032 			return rc;
1033 		}
1034 	}
1035 	return rc;
1036 }
1037 
1038 static void cudbg_t4_fwcache(struct cudbg_init *pdbg_init,
1039 			     struct cudbg_error *cudbg_err)
1040 {
1041 	struct adapter *padap = pdbg_init->adap;
1042 	int rc;
1043 
1044 	if (is_fw_attached(pdbg_init)) {
1045 		/* Flush uP dcache before reading edcX/mcX  */
1046 		rc = t4_fwcache(padap, FW_PARAM_DEV_FWCACHE_FLUSH);
1047 		if (rc)
1048 			cudbg_err->sys_warn = rc;
1049 	}
1050 }
1051 
1052 static int cudbg_mem_region_size(struct cudbg_init *pdbg_init,
1053 				 struct cudbg_error *cudbg_err,
1054 				 u8 mem_type, unsigned long *region_size)
1055 {
1056 	struct adapter *padap = pdbg_init->adap;
1057 	struct cudbg_meminfo mem_info;
1058 	u8 mc_idx;
1059 	int rc;
1060 
1061 	memset(&mem_info, 0, sizeof(struct cudbg_meminfo));
1062 	rc = cudbg_fill_meminfo(padap, &mem_info);
1063 	if (rc) {
1064 		cudbg_err->sys_err = rc;
1065 		return rc;
1066 	}
1067 
1068 	cudbg_t4_fwcache(pdbg_init, cudbg_err);
1069 	rc = cudbg_meminfo_get_mem_index(padap, &mem_info, mem_type, &mc_idx);
1070 	if (rc) {
1071 		cudbg_err->sys_err = rc;
1072 		return rc;
1073 	}
1074 
1075 	if (region_size)
1076 		*region_size = mem_info.avail[mc_idx].limit -
1077 			       mem_info.avail[mc_idx].base;
1078 
1079 	return 0;
1080 }
1081 
1082 static int cudbg_collect_mem_region(struct cudbg_init *pdbg_init,
1083 				    struct cudbg_buffer *dbg_buff,
1084 				    struct cudbg_error *cudbg_err,
1085 				    u8 mem_type)
1086 {
1087 	unsigned long size = 0;
1088 	int rc;
1089 
1090 	rc = cudbg_mem_region_size(pdbg_init, cudbg_err, mem_type, &size);
1091 	if (rc)
1092 		return rc;
1093 
1094 	return cudbg_read_fw_mem(pdbg_init, dbg_buff, mem_type, size,
1095 				 cudbg_err);
1096 }
1097 
1098 int cudbg_collect_edc0_meminfo(struct cudbg_init *pdbg_init,
1099 			       struct cudbg_buffer *dbg_buff,
1100 			       struct cudbg_error *cudbg_err)
1101 {
1102 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1103 					MEM_EDC0);
1104 }
1105 
1106 int cudbg_collect_edc1_meminfo(struct cudbg_init *pdbg_init,
1107 			       struct cudbg_buffer *dbg_buff,
1108 			       struct cudbg_error *cudbg_err)
1109 {
1110 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1111 					MEM_EDC1);
1112 }
1113 
1114 int cudbg_collect_mc0_meminfo(struct cudbg_init *pdbg_init,
1115 			      struct cudbg_buffer *dbg_buff,
1116 			      struct cudbg_error *cudbg_err)
1117 {
1118 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1119 					MEM_MC0);
1120 }
1121 
1122 int cudbg_collect_mc1_meminfo(struct cudbg_init *pdbg_init,
1123 			      struct cudbg_buffer *dbg_buff,
1124 			      struct cudbg_error *cudbg_err)
1125 {
1126 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1127 					MEM_MC1);
1128 }
1129 
1130 int cudbg_collect_hma_meminfo(struct cudbg_init *pdbg_init,
1131 			      struct cudbg_buffer *dbg_buff,
1132 			      struct cudbg_error *cudbg_err)
1133 {
1134 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1135 					MEM_HMA);
1136 }
1137 
1138 int cudbg_collect_rss(struct cudbg_init *pdbg_init,
1139 		      struct cudbg_buffer *dbg_buff,
1140 		      struct cudbg_error *cudbg_err)
1141 {
1142 	struct adapter *padap = pdbg_init->adap;
1143 	struct cudbg_buffer temp_buff = { 0 };
1144 	int rc, nentries;
1145 
1146 	nentries = t4_chip_rss_size(padap);
1147 	rc = cudbg_get_buff(pdbg_init, dbg_buff, nentries * sizeof(u16),
1148 			    &temp_buff);
1149 	if (rc)
1150 		return rc;
1151 
1152 	rc = t4_read_rss(padap, (u16 *)temp_buff.data);
1153 	if (rc) {
1154 		cudbg_err->sys_err = rc;
1155 		cudbg_put_buff(pdbg_init, &temp_buff);
1156 		return rc;
1157 	}
1158 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1159 }
1160 
1161 int cudbg_collect_rss_vf_config(struct cudbg_init *pdbg_init,
1162 				struct cudbg_buffer *dbg_buff,
1163 				struct cudbg_error *cudbg_err)
1164 {
1165 	struct adapter *padap = pdbg_init->adap;
1166 	struct cudbg_buffer temp_buff = { 0 };
1167 	struct cudbg_rss_vf_conf *vfconf;
1168 	int vf, rc, vf_count;
1169 
1170 	vf_count = padap->params.arch.vfcount;
1171 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
1172 			    vf_count * sizeof(struct cudbg_rss_vf_conf),
1173 			    &temp_buff);
1174 	if (rc)
1175 		return rc;
1176 
1177 	vfconf = (struct cudbg_rss_vf_conf *)temp_buff.data;
1178 	for (vf = 0; vf < vf_count; vf++)
1179 		t4_read_rss_vf_config(padap, vf, &vfconf[vf].rss_vf_vfl,
1180 				      &vfconf[vf].rss_vf_vfh, true);
1181 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1182 }
1183 
1184 int cudbg_collect_path_mtu(struct cudbg_init *pdbg_init,
1185 			   struct cudbg_buffer *dbg_buff,
1186 			   struct cudbg_error *cudbg_err)
1187 {
1188 	struct adapter *padap = pdbg_init->adap;
1189 	struct cudbg_buffer temp_buff = { 0 };
1190 	int rc;
1191 
1192 	rc = cudbg_get_buff(pdbg_init, dbg_buff, NMTUS * sizeof(u16),
1193 			    &temp_buff);
1194 	if (rc)
1195 		return rc;
1196 
1197 	t4_read_mtu_tbl(padap, (u16 *)temp_buff.data, NULL);
1198 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1199 }
1200 
1201 int cudbg_collect_pm_stats(struct cudbg_init *pdbg_init,
1202 			   struct cudbg_buffer *dbg_buff,
1203 			   struct cudbg_error *cudbg_err)
1204 {
1205 	struct adapter *padap = pdbg_init->adap;
1206 	struct cudbg_buffer temp_buff = { 0 };
1207 	struct cudbg_pm_stats *pm_stats_buff;
1208 	int rc;
1209 
1210 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_pm_stats),
1211 			    &temp_buff);
1212 	if (rc)
1213 		return rc;
1214 
1215 	pm_stats_buff = (struct cudbg_pm_stats *)temp_buff.data;
1216 	t4_pmtx_get_stats(padap, pm_stats_buff->tx_cnt, pm_stats_buff->tx_cyc);
1217 	t4_pmrx_get_stats(padap, pm_stats_buff->rx_cnt, pm_stats_buff->rx_cyc);
1218 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1219 }
1220 
1221 int cudbg_collect_hw_sched(struct cudbg_init *pdbg_init,
1222 			   struct cudbg_buffer *dbg_buff,
1223 			   struct cudbg_error *cudbg_err)
1224 {
1225 	struct adapter *padap = pdbg_init->adap;
1226 	struct cudbg_buffer temp_buff = { 0 };
1227 	struct cudbg_hw_sched *hw_sched_buff;
1228 	int i, rc = 0;
1229 
1230 	if (!padap->params.vpd.cclk)
1231 		return CUDBG_STATUS_CCLK_NOT_DEFINED;
1232 
1233 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_hw_sched),
1234 			    &temp_buff);
1235 
1236 	if (rc)
1237 		return rc;
1238 
1239 	hw_sched_buff = (struct cudbg_hw_sched *)temp_buff.data;
1240 	hw_sched_buff->map = t4_read_reg(padap, TP_TX_MOD_QUEUE_REQ_MAP_A);
1241 	hw_sched_buff->mode = TIMERMODE_G(t4_read_reg(padap, TP_MOD_CONFIG_A));
1242 	t4_read_pace_tbl(padap, hw_sched_buff->pace_tab);
1243 	for (i = 0; i < NTX_SCHED; ++i)
1244 		t4_get_tx_sched(padap, i, &hw_sched_buff->kbps[i],
1245 				&hw_sched_buff->ipg[i], true);
1246 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1247 }
1248 
1249 int cudbg_collect_tp_indirect(struct cudbg_init *pdbg_init,
1250 			      struct cudbg_buffer *dbg_buff,
1251 			      struct cudbg_error *cudbg_err)
1252 {
1253 	struct adapter *padap = pdbg_init->adap;
1254 	struct cudbg_buffer temp_buff = { 0 };
1255 	struct ireg_buf *ch_tp_pio;
1256 	int i, rc, n = 0;
1257 	u32 size;
1258 
1259 	if (is_t5(padap->params.chip))
1260 		n = sizeof(t5_tp_pio_array) +
1261 		    sizeof(t5_tp_tm_pio_array) +
1262 		    sizeof(t5_tp_mib_index_array);
1263 	else
1264 		n = sizeof(t6_tp_pio_array) +
1265 		    sizeof(t6_tp_tm_pio_array) +
1266 		    sizeof(t6_tp_mib_index_array);
1267 
1268 	n = n / (IREG_NUM_ELEM * sizeof(u32));
1269 	size = sizeof(struct ireg_buf) * n;
1270 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1271 	if (rc)
1272 		return rc;
1273 
1274 	ch_tp_pio = (struct ireg_buf *)temp_buff.data;
1275 
1276 	/* TP_PIO */
1277 	if (is_t5(padap->params.chip))
1278 		n = sizeof(t5_tp_pio_array) / (IREG_NUM_ELEM * sizeof(u32));
1279 	else if (is_t6(padap->params.chip))
1280 		n = sizeof(t6_tp_pio_array) / (IREG_NUM_ELEM * sizeof(u32));
1281 
1282 	for (i = 0; i < n; i++) {
1283 		struct ireg_field *tp_pio = &ch_tp_pio->tp_pio;
1284 		u32 *buff = ch_tp_pio->outbuf;
1285 
1286 		if (is_t5(padap->params.chip)) {
1287 			tp_pio->ireg_addr = t5_tp_pio_array[i][0];
1288 			tp_pio->ireg_data = t5_tp_pio_array[i][1];
1289 			tp_pio->ireg_local_offset = t5_tp_pio_array[i][2];
1290 			tp_pio->ireg_offset_range = t5_tp_pio_array[i][3];
1291 		} else if (is_t6(padap->params.chip)) {
1292 			tp_pio->ireg_addr = t6_tp_pio_array[i][0];
1293 			tp_pio->ireg_data = t6_tp_pio_array[i][1];
1294 			tp_pio->ireg_local_offset = t6_tp_pio_array[i][2];
1295 			tp_pio->ireg_offset_range = t6_tp_pio_array[i][3];
1296 		}
1297 		t4_tp_pio_read(padap, buff, tp_pio->ireg_offset_range,
1298 			       tp_pio->ireg_local_offset, true);
1299 		ch_tp_pio++;
1300 	}
1301 
1302 	/* TP_TM_PIO */
1303 	if (is_t5(padap->params.chip))
1304 		n = sizeof(t5_tp_tm_pio_array) / (IREG_NUM_ELEM * sizeof(u32));
1305 	else if (is_t6(padap->params.chip))
1306 		n = sizeof(t6_tp_tm_pio_array) / (IREG_NUM_ELEM * sizeof(u32));
1307 
1308 	for (i = 0; i < n; i++) {
1309 		struct ireg_field *tp_pio = &ch_tp_pio->tp_pio;
1310 		u32 *buff = ch_tp_pio->outbuf;
1311 
1312 		if (is_t5(padap->params.chip)) {
1313 			tp_pio->ireg_addr = t5_tp_tm_pio_array[i][0];
1314 			tp_pio->ireg_data = t5_tp_tm_pio_array[i][1];
1315 			tp_pio->ireg_local_offset = t5_tp_tm_pio_array[i][2];
1316 			tp_pio->ireg_offset_range = t5_tp_tm_pio_array[i][3];
1317 		} else if (is_t6(padap->params.chip)) {
1318 			tp_pio->ireg_addr = t6_tp_tm_pio_array[i][0];
1319 			tp_pio->ireg_data = t6_tp_tm_pio_array[i][1];
1320 			tp_pio->ireg_local_offset = t6_tp_tm_pio_array[i][2];
1321 			tp_pio->ireg_offset_range = t6_tp_tm_pio_array[i][3];
1322 		}
1323 		t4_tp_tm_pio_read(padap, buff, tp_pio->ireg_offset_range,
1324 				  tp_pio->ireg_local_offset, true);
1325 		ch_tp_pio++;
1326 	}
1327 
1328 	/* TP_MIB_INDEX */
1329 	if (is_t5(padap->params.chip))
1330 		n = sizeof(t5_tp_mib_index_array) /
1331 		    (IREG_NUM_ELEM * sizeof(u32));
1332 	else if (is_t6(padap->params.chip))
1333 		n = sizeof(t6_tp_mib_index_array) /
1334 		    (IREG_NUM_ELEM * sizeof(u32));
1335 
1336 	for (i = 0; i < n ; i++) {
1337 		struct ireg_field *tp_pio = &ch_tp_pio->tp_pio;
1338 		u32 *buff = ch_tp_pio->outbuf;
1339 
1340 		if (is_t5(padap->params.chip)) {
1341 			tp_pio->ireg_addr = t5_tp_mib_index_array[i][0];
1342 			tp_pio->ireg_data = t5_tp_mib_index_array[i][1];
1343 			tp_pio->ireg_local_offset =
1344 				t5_tp_mib_index_array[i][2];
1345 			tp_pio->ireg_offset_range =
1346 				t5_tp_mib_index_array[i][3];
1347 		} else if (is_t6(padap->params.chip)) {
1348 			tp_pio->ireg_addr = t6_tp_mib_index_array[i][0];
1349 			tp_pio->ireg_data = t6_tp_mib_index_array[i][1];
1350 			tp_pio->ireg_local_offset =
1351 				t6_tp_mib_index_array[i][2];
1352 			tp_pio->ireg_offset_range =
1353 				t6_tp_mib_index_array[i][3];
1354 		}
1355 		t4_tp_mib_read(padap, buff, tp_pio->ireg_offset_range,
1356 			       tp_pio->ireg_local_offset, true);
1357 		ch_tp_pio++;
1358 	}
1359 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1360 }
1361 
1362 static void cudbg_read_sge_qbase_indirect_reg(struct adapter *padap,
1363 					      struct sge_qbase_reg_field *qbase,
1364 					      u32 func, bool is_pf)
1365 {
1366 	u32 *buff, i;
1367 
1368 	if (is_pf) {
1369 		buff = qbase->pf_data_value[func];
1370 	} else {
1371 		buff = qbase->vf_data_value[func];
1372 		/* In SGE_QBASE_INDEX,
1373 		 * Entries 0->7 are PF0->7, Entries 8->263 are VFID0->256.
1374 		 */
1375 		func += 8;
1376 	}
1377 
1378 	t4_write_reg(padap, qbase->reg_addr, func);
1379 	for (i = 0; i < SGE_QBASE_DATA_REG_NUM; i++, buff++)
1380 		*buff = t4_read_reg(padap, qbase->reg_data[i]);
1381 }
1382 
1383 int cudbg_collect_sge_indirect(struct cudbg_init *pdbg_init,
1384 			       struct cudbg_buffer *dbg_buff,
1385 			       struct cudbg_error *cudbg_err)
1386 {
1387 	struct adapter *padap = pdbg_init->adap;
1388 	struct cudbg_buffer temp_buff = { 0 };
1389 	struct sge_qbase_reg_field *sge_qbase;
1390 	struct ireg_buf *ch_sge_dbg;
1391 	int i, rc;
1392 
1393 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
1394 			    sizeof(*ch_sge_dbg) * 2 + sizeof(*sge_qbase),
1395 			    &temp_buff);
1396 	if (rc)
1397 		return rc;
1398 
1399 	ch_sge_dbg = (struct ireg_buf *)temp_buff.data;
1400 	for (i = 0; i < 2; i++) {
1401 		struct ireg_field *sge_pio = &ch_sge_dbg->tp_pio;
1402 		u32 *buff = ch_sge_dbg->outbuf;
1403 
1404 		sge_pio->ireg_addr = t5_sge_dbg_index_array[i][0];
1405 		sge_pio->ireg_data = t5_sge_dbg_index_array[i][1];
1406 		sge_pio->ireg_local_offset = t5_sge_dbg_index_array[i][2];
1407 		sge_pio->ireg_offset_range = t5_sge_dbg_index_array[i][3];
1408 		t4_read_indirect(padap,
1409 				 sge_pio->ireg_addr,
1410 				 sge_pio->ireg_data,
1411 				 buff,
1412 				 sge_pio->ireg_offset_range,
1413 				 sge_pio->ireg_local_offset);
1414 		ch_sge_dbg++;
1415 	}
1416 
1417 	if (CHELSIO_CHIP_VERSION(padap->params.chip) > CHELSIO_T5) {
1418 		sge_qbase = (struct sge_qbase_reg_field *)ch_sge_dbg;
1419 		/* 1 addr reg SGE_QBASE_INDEX and 4 data reg
1420 		 * SGE_QBASE_MAP[0-3]
1421 		 */
1422 		sge_qbase->reg_addr = t6_sge_qbase_index_array[0];
1423 		for (i = 0; i < SGE_QBASE_DATA_REG_NUM; i++)
1424 			sge_qbase->reg_data[i] =
1425 				t6_sge_qbase_index_array[i + 1];
1426 
1427 		for (i = 0; i <= PCIE_FW_MASTER_M; i++)
1428 			cudbg_read_sge_qbase_indirect_reg(padap, sge_qbase,
1429 							  i, true);
1430 
1431 		for (i = 0; i < padap->params.arch.vfcount; i++)
1432 			cudbg_read_sge_qbase_indirect_reg(padap, sge_qbase,
1433 							  i, false);
1434 
1435 		sge_qbase->vfcount = padap->params.arch.vfcount;
1436 	}
1437 
1438 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1439 }
1440 
1441 int cudbg_collect_ulprx_la(struct cudbg_init *pdbg_init,
1442 			   struct cudbg_buffer *dbg_buff,
1443 			   struct cudbg_error *cudbg_err)
1444 {
1445 	struct adapter *padap = pdbg_init->adap;
1446 	struct cudbg_buffer temp_buff = { 0 };
1447 	struct cudbg_ulprx_la *ulprx_la_buff;
1448 	int rc;
1449 
1450 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_ulprx_la),
1451 			    &temp_buff);
1452 	if (rc)
1453 		return rc;
1454 
1455 	ulprx_la_buff = (struct cudbg_ulprx_la *)temp_buff.data;
1456 	t4_ulprx_read_la(padap, (u32 *)ulprx_la_buff->data);
1457 	ulprx_la_buff->size = ULPRX_LA_SIZE;
1458 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1459 }
1460 
1461 int cudbg_collect_tp_la(struct cudbg_init *pdbg_init,
1462 			struct cudbg_buffer *dbg_buff,
1463 			struct cudbg_error *cudbg_err)
1464 {
1465 	struct adapter *padap = pdbg_init->adap;
1466 	struct cudbg_buffer temp_buff = { 0 };
1467 	struct cudbg_tp_la *tp_la_buff;
1468 	int size, rc;
1469 
1470 	size = sizeof(struct cudbg_tp_la) + TPLA_SIZE *  sizeof(u64);
1471 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1472 	if (rc)
1473 		return rc;
1474 
1475 	tp_la_buff = (struct cudbg_tp_la *)temp_buff.data;
1476 	tp_la_buff->mode = DBGLAMODE_G(t4_read_reg(padap, TP_DBG_LA_CONFIG_A));
1477 	t4_tp_read_la(padap, (u64 *)tp_la_buff->data, NULL);
1478 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1479 }
1480 
1481 int cudbg_collect_meminfo(struct cudbg_init *pdbg_init,
1482 			  struct cudbg_buffer *dbg_buff,
1483 			  struct cudbg_error *cudbg_err)
1484 {
1485 	struct adapter *padap = pdbg_init->adap;
1486 	struct cudbg_buffer temp_buff = { 0 };
1487 	struct cudbg_meminfo *meminfo_buff;
1488 	struct cudbg_ver_hdr *ver_hdr;
1489 	int rc;
1490 
1491 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
1492 			    sizeof(struct cudbg_ver_hdr) +
1493 			    sizeof(struct cudbg_meminfo),
1494 			    &temp_buff);
1495 	if (rc)
1496 		return rc;
1497 
1498 	ver_hdr = (struct cudbg_ver_hdr *)temp_buff.data;
1499 	ver_hdr->signature = CUDBG_ENTITY_SIGNATURE;
1500 	ver_hdr->revision = CUDBG_MEMINFO_REV;
1501 	ver_hdr->size = sizeof(struct cudbg_meminfo);
1502 
1503 	meminfo_buff = (struct cudbg_meminfo *)(temp_buff.data +
1504 						sizeof(*ver_hdr));
1505 	rc = cudbg_fill_meminfo(padap, meminfo_buff);
1506 	if (rc) {
1507 		cudbg_err->sys_err = rc;
1508 		cudbg_put_buff(pdbg_init, &temp_buff);
1509 		return rc;
1510 	}
1511 
1512 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1513 }
1514 
1515 int cudbg_collect_cim_pif_la(struct cudbg_init *pdbg_init,
1516 			     struct cudbg_buffer *dbg_buff,
1517 			     struct cudbg_error *cudbg_err)
1518 {
1519 	struct cudbg_cim_pif_la *cim_pif_la_buff;
1520 	struct adapter *padap = pdbg_init->adap;
1521 	struct cudbg_buffer temp_buff = { 0 };
1522 	int size, rc;
1523 
1524 	size = sizeof(struct cudbg_cim_pif_la) +
1525 	       2 * CIM_PIFLA_SIZE * 6 * sizeof(u32);
1526 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1527 	if (rc)
1528 		return rc;
1529 
1530 	cim_pif_la_buff = (struct cudbg_cim_pif_la *)temp_buff.data;
1531 	cim_pif_la_buff->size = CIM_PIFLA_SIZE;
1532 	t4_cim_read_pif_la(padap, (u32 *)cim_pif_la_buff->data,
1533 			   (u32 *)cim_pif_la_buff->data + 6 * CIM_PIFLA_SIZE,
1534 			   NULL, NULL);
1535 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1536 }
1537 
1538 int cudbg_collect_clk_info(struct cudbg_init *pdbg_init,
1539 			   struct cudbg_buffer *dbg_buff,
1540 			   struct cudbg_error *cudbg_err)
1541 {
1542 	struct adapter *padap = pdbg_init->adap;
1543 	struct cudbg_buffer temp_buff = { 0 };
1544 	struct cudbg_clk_info *clk_info_buff;
1545 	u64 tp_tick_us;
1546 	int rc;
1547 
1548 	if (!padap->params.vpd.cclk)
1549 		return CUDBG_STATUS_CCLK_NOT_DEFINED;
1550 
1551 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_clk_info),
1552 			    &temp_buff);
1553 	if (rc)
1554 		return rc;
1555 
1556 	clk_info_buff = (struct cudbg_clk_info *)temp_buff.data;
1557 	clk_info_buff->cclk_ps = 1000000000 / padap->params.vpd.cclk; /* psec */
1558 	clk_info_buff->res = t4_read_reg(padap, TP_TIMER_RESOLUTION_A);
1559 	clk_info_buff->tre = TIMERRESOLUTION_G(clk_info_buff->res);
1560 	clk_info_buff->dack_re = DELAYEDACKRESOLUTION_G(clk_info_buff->res);
1561 	tp_tick_us = (clk_info_buff->cclk_ps << clk_info_buff->tre) / 1000000;
1562 
1563 	clk_info_buff->dack_timer =
1564 		(clk_info_buff->cclk_ps << clk_info_buff->dack_re) / 1000000 *
1565 		t4_read_reg(padap, TP_DACK_TIMER_A);
1566 	clk_info_buff->retransmit_min =
1567 		tp_tick_us * t4_read_reg(padap, TP_RXT_MIN_A);
1568 	clk_info_buff->retransmit_max =
1569 		tp_tick_us * t4_read_reg(padap, TP_RXT_MAX_A);
1570 	clk_info_buff->persist_timer_min =
1571 		tp_tick_us * t4_read_reg(padap, TP_PERS_MIN_A);
1572 	clk_info_buff->persist_timer_max =
1573 		tp_tick_us * t4_read_reg(padap, TP_PERS_MAX_A);
1574 	clk_info_buff->keepalive_idle_timer =
1575 		tp_tick_us * t4_read_reg(padap, TP_KEEP_IDLE_A);
1576 	clk_info_buff->keepalive_interval =
1577 		tp_tick_us * t4_read_reg(padap, TP_KEEP_INTVL_A);
1578 	clk_info_buff->initial_srtt =
1579 		tp_tick_us * INITSRTT_G(t4_read_reg(padap, TP_INIT_SRTT_A));
1580 	clk_info_buff->finwait2_timer =
1581 		tp_tick_us * t4_read_reg(padap, TP_FINWAIT2_TIMER_A);
1582 
1583 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1584 }
1585 
1586 int cudbg_collect_pcie_indirect(struct cudbg_init *pdbg_init,
1587 				struct cudbg_buffer *dbg_buff,
1588 				struct cudbg_error *cudbg_err)
1589 {
1590 	struct adapter *padap = pdbg_init->adap;
1591 	struct cudbg_buffer temp_buff = { 0 };
1592 	struct ireg_buf *ch_pcie;
1593 	int i, rc, n;
1594 	u32 size;
1595 
1596 	n = sizeof(t5_pcie_pdbg_array) / (IREG_NUM_ELEM * sizeof(u32));
1597 	size = sizeof(struct ireg_buf) * n * 2;
1598 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1599 	if (rc)
1600 		return rc;
1601 
1602 	ch_pcie = (struct ireg_buf *)temp_buff.data;
1603 	/* PCIE_PDBG */
1604 	for (i = 0; i < n; i++) {
1605 		struct ireg_field *pcie_pio = &ch_pcie->tp_pio;
1606 		u32 *buff = ch_pcie->outbuf;
1607 
1608 		pcie_pio->ireg_addr = t5_pcie_pdbg_array[i][0];
1609 		pcie_pio->ireg_data = t5_pcie_pdbg_array[i][1];
1610 		pcie_pio->ireg_local_offset = t5_pcie_pdbg_array[i][2];
1611 		pcie_pio->ireg_offset_range = t5_pcie_pdbg_array[i][3];
1612 		t4_read_indirect(padap,
1613 				 pcie_pio->ireg_addr,
1614 				 pcie_pio->ireg_data,
1615 				 buff,
1616 				 pcie_pio->ireg_offset_range,
1617 				 pcie_pio->ireg_local_offset);
1618 		ch_pcie++;
1619 	}
1620 
1621 	/* PCIE_CDBG */
1622 	n = sizeof(t5_pcie_cdbg_array) / (IREG_NUM_ELEM * sizeof(u32));
1623 	for (i = 0; i < n; i++) {
1624 		struct ireg_field *pcie_pio = &ch_pcie->tp_pio;
1625 		u32 *buff = ch_pcie->outbuf;
1626 
1627 		pcie_pio->ireg_addr = t5_pcie_cdbg_array[i][0];
1628 		pcie_pio->ireg_data = t5_pcie_cdbg_array[i][1];
1629 		pcie_pio->ireg_local_offset = t5_pcie_cdbg_array[i][2];
1630 		pcie_pio->ireg_offset_range = t5_pcie_cdbg_array[i][3];
1631 		t4_read_indirect(padap,
1632 				 pcie_pio->ireg_addr,
1633 				 pcie_pio->ireg_data,
1634 				 buff,
1635 				 pcie_pio->ireg_offset_range,
1636 				 pcie_pio->ireg_local_offset);
1637 		ch_pcie++;
1638 	}
1639 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1640 }
1641 
1642 int cudbg_collect_pm_indirect(struct cudbg_init *pdbg_init,
1643 			      struct cudbg_buffer *dbg_buff,
1644 			      struct cudbg_error *cudbg_err)
1645 {
1646 	struct adapter *padap = pdbg_init->adap;
1647 	struct cudbg_buffer temp_buff = { 0 };
1648 	struct ireg_buf *ch_pm;
1649 	int i, rc, n;
1650 	u32 size;
1651 
1652 	n = sizeof(t5_pm_rx_array) / (IREG_NUM_ELEM * sizeof(u32));
1653 	size = sizeof(struct ireg_buf) * n * 2;
1654 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1655 	if (rc)
1656 		return rc;
1657 
1658 	ch_pm = (struct ireg_buf *)temp_buff.data;
1659 	/* PM_RX */
1660 	for (i = 0; i < n; i++) {
1661 		struct ireg_field *pm_pio = &ch_pm->tp_pio;
1662 		u32 *buff = ch_pm->outbuf;
1663 
1664 		pm_pio->ireg_addr = t5_pm_rx_array[i][0];
1665 		pm_pio->ireg_data = t5_pm_rx_array[i][1];
1666 		pm_pio->ireg_local_offset = t5_pm_rx_array[i][2];
1667 		pm_pio->ireg_offset_range = t5_pm_rx_array[i][3];
1668 		t4_read_indirect(padap,
1669 				 pm_pio->ireg_addr,
1670 				 pm_pio->ireg_data,
1671 				 buff,
1672 				 pm_pio->ireg_offset_range,
1673 				 pm_pio->ireg_local_offset);
1674 		ch_pm++;
1675 	}
1676 
1677 	/* PM_TX */
1678 	n = sizeof(t5_pm_tx_array) / (IREG_NUM_ELEM * sizeof(u32));
1679 	for (i = 0; i < n; i++) {
1680 		struct ireg_field *pm_pio = &ch_pm->tp_pio;
1681 		u32 *buff = ch_pm->outbuf;
1682 
1683 		pm_pio->ireg_addr = t5_pm_tx_array[i][0];
1684 		pm_pio->ireg_data = t5_pm_tx_array[i][1];
1685 		pm_pio->ireg_local_offset = t5_pm_tx_array[i][2];
1686 		pm_pio->ireg_offset_range = t5_pm_tx_array[i][3];
1687 		t4_read_indirect(padap,
1688 				 pm_pio->ireg_addr,
1689 				 pm_pio->ireg_data,
1690 				 buff,
1691 				 pm_pio->ireg_offset_range,
1692 				 pm_pio->ireg_local_offset);
1693 		ch_pm++;
1694 	}
1695 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1696 }
1697 
1698 int cudbg_collect_tid(struct cudbg_init *pdbg_init,
1699 		      struct cudbg_buffer *dbg_buff,
1700 		      struct cudbg_error *cudbg_err)
1701 {
1702 	struct adapter *padap = pdbg_init->adap;
1703 	struct cudbg_tid_info_region_rev1 *tid1;
1704 	struct cudbg_buffer temp_buff = { 0 };
1705 	struct cudbg_tid_info_region *tid;
1706 	u32 para[2], val[2];
1707 	int rc;
1708 
1709 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
1710 			    sizeof(struct cudbg_tid_info_region_rev1),
1711 			    &temp_buff);
1712 	if (rc)
1713 		return rc;
1714 
1715 	tid1 = (struct cudbg_tid_info_region_rev1 *)temp_buff.data;
1716 	tid = &tid1->tid;
1717 	tid1->ver_hdr.signature = CUDBG_ENTITY_SIGNATURE;
1718 	tid1->ver_hdr.revision = CUDBG_TID_INFO_REV;
1719 	tid1->ver_hdr.size = sizeof(struct cudbg_tid_info_region_rev1) -
1720 			     sizeof(struct cudbg_ver_hdr);
1721 
1722 	/* If firmware is not attached/alive, use backdoor register
1723 	 * access to collect dump.
1724 	 */
1725 	if (!is_fw_attached(pdbg_init))
1726 		goto fill_tid;
1727 
1728 #define FW_PARAM_PFVF_A(param) \
1729 	(FW_PARAMS_MNEM_V(FW_PARAMS_MNEM_PFVF) | \
1730 	 FW_PARAMS_PARAM_X_V(FW_PARAMS_PARAM_PFVF_##param) | \
1731 	 FW_PARAMS_PARAM_Y_V(0) | \
1732 	 FW_PARAMS_PARAM_Z_V(0))
1733 
1734 	para[0] = FW_PARAM_PFVF_A(ETHOFLD_START);
1735 	para[1] = FW_PARAM_PFVF_A(ETHOFLD_END);
1736 	rc = t4_query_params(padap, padap->mbox, padap->pf, 0, 2, para, val);
1737 	if (rc <  0) {
1738 		cudbg_err->sys_err = rc;
1739 		cudbg_put_buff(pdbg_init, &temp_buff);
1740 		return rc;
1741 	}
1742 	tid->uotid_base = val[0];
1743 	tid->nuotids = val[1] - val[0] + 1;
1744 
1745 	if (is_t5(padap->params.chip)) {
1746 		tid->sb = t4_read_reg(padap, LE_DB_SERVER_INDEX_A) / 4;
1747 	} else if (is_t6(padap->params.chip)) {
1748 		tid1->tid_start =
1749 			t4_read_reg(padap, LE_DB_ACTIVE_TABLE_START_INDEX_A);
1750 		tid->sb = t4_read_reg(padap, LE_DB_SRVR_START_INDEX_A);
1751 
1752 		para[0] = FW_PARAM_PFVF_A(HPFILTER_START);
1753 		para[1] = FW_PARAM_PFVF_A(HPFILTER_END);
1754 		rc = t4_query_params(padap, padap->mbox, padap->pf, 0, 2,
1755 				     para, val);
1756 		if (rc < 0) {
1757 			cudbg_err->sys_err = rc;
1758 			cudbg_put_buff(pdbg_init, &temp_buff);
1759 			return rc;
1760 		}
1761 		tid->hpftid_base = val[0];
1762 		tid->nhpftids = val[1] - val[0] + 1;
1763 	}
1764 
1765 #undef FW_PARAM_PFVF_A
1766 
1767 fill_tid:
1768 	tid->ntids = padap->tids.ntids;
1769 	tid->nstids = padap->tids.nstids;
1770 	tid->stid_base = padap->tids.stid_base;
1771 	tid->hash_base = padap->tids.hash_base;
1772 
1773 	tid->natids = padap->tids.natids;
1774 	tid->nftids = padap->tids.nftids;
1775 	tid->ftid_base = padap->tids.ftid_base;
1776 	tid->aftid_base = padap->tids.aftid_base;
1777 	tid->aftid_end = padap->tids.aftid_end;
1778 
1779 	tid->sftid_base = padap->tids.sftid_base;
1780 	tid->nsftids = padap->tids.nsftids;
1781 
1782 	tid->flags = padap->flags;
1783 	tid->le_db_conf = t4_read_reg(padap, LE_DB_CONFIG_A);
1784 	tid->ip_users = t4_read_reg(padap, LE_DB_ACT_CNT_IPV4_A);
1785 	tid->ipv6_users = t4_read_reg(padap, LE_DB_ACT_CNT_IPV6_A);
1786 
1787 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1788 }
1789 
1790 int cudbg_collect_pcie_config(struct cudbg_init *pdbg_init,
1791 			      struct cudbg_buffer *dbg_buff,
1792 			      struct cudbg_error *cudbg_err)
1793 {
1794 	struct adapter *padap = pdbg_init->adap;
1795 	struct cudbg_buffer temp_buff = { 0 };
1796 	u32 size, *value, j;
1797 	int i, rc, n;
1798 
1799 	size = sizeof(u32) * CUDBG_NUM_PCIE_CONFIG_REGS;
1800 	n = sizeof(t5_pcie_config_array) / (2 * sizeof(u32));
1801 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1802 	if (rc)
1803 		return rc;
1804 
1805 	value = (u32 *)temp_buff.data;
1806 	for (i = 0; i < n; i++) {
1807 		for (j = t5_pcie_config_array[i][0];
1808 		     j <= t5_pcie_config_array[i][1]; j += 4) {
1809 			t4_hw_pci_read_cfg4(padap, j, value);
1810 			value++;
1811 		}
1812 	}
1813 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1814 }
1815 
1816 static int cudbg_sge_ctxt_check_valid(u32 *buf, int type)
1817 {
1818 	int index, bit, bit_pos = 0;
1819 
1820 	switch (type) {
1821 	case CTXT_EGRESS:
1822 		bit_pos = 176;
1823 		break;
1824 	case CTXT_INGRESS:
1825 		bit_pos = 141;
1826 		break;
1827 	case CTXT_FLM:
1828 		bit_pos = 89;
1829 		break;
1830 	}
1831 	index = bit_pos / 32;
1832 	bit =  bit_pos % 32;
1833 	return buf[index] & (1U << bit);
1834 }
1835 
1836 static int cudbg_get_ctxt_region_info(struct adapter *padap,
1837 				      struct cudbg_region_info *ctx_info,
1838 				      u8 *mem_type)
1839 {
1840 	struct cudbg_mem_desc mem_desc;
1841 	struct cudbg_meminfo meminfo;
1842 	u32 i, j, value, found;
1843 	u8 flq;
1844 	int rc;
1845 
1846 	rc = cudbg_fill_meminfo(padap, &meminfo);
1847 	if (rc)
1848 		return rc;
1849 
1850 	/* Get EGRESS and INGRESS context region size */
1851 	for (i = CTXT_EGRESS; i <= CTXT_INGRESS; i++) {
1852 		found = 0;
1853 		memset(&mem_desc, 0, sizeof(struct cudbg_mem_desc));
1854 		for (j = 0; j < ARRAY_SIZE(meminfo.avail); j++) {
1855 			rc = cudbg_get_mem_region(padap, &meminfo, j,
1856 						  cudbg_region[i],
1857 						  &mem_desc);
1858 			if (!rc) {
1859 				found = 1;
1860 				rc = cudbg_get_mem_relative(padap, &meminfo, j,
1861 							    &mem_desc.base,
1862 							    &mem_desc.limit);
1863 				if (rc) {
1864 					ctx_info[i].exist = false;
1865 					break;
1866 				}
1867 				ctx_info[i].exist = true;
1868 				ctx_info[i].start = mem_desc.base;
1869 				ctx_info[i].end = mem_desc.limit;
1870 				mem_type[i] = j;
1871 				break;
1872 			}
1873 		}
1874 		if (!found)
1875 			ctx_info[i].exist = false;
1876 	}
1877 
1878 	/* Get FLM and CNM max qid. */
1879 	value = t4_read_reg(padap, SGE_FLM_CFG_A);
1880 
1881 	/* Get number of data freelist queues */
1882 	flq = HDRSTARTFLQ_G(value);
1883 	ctx_info[CTXT_FLM].exist = true;
1884 	ctx_info[CTXT_FLM].end = (CUDBG_MAX_FL_QIDS >> flq) * SGE_CTXT_SIZE;
1885 
1886 	/* The number of CONM contexts are same as number of freelist
1887 	 * queues.
1888 	 */
1889 	ctx_info[CTXT_CNM].exist = true;
1890 	ctx_info[CTXT_CNM].end = ctx_info[CTXT_FLM].end;
1891 
1892 	return 0;
1893 }
1894 
1895 int cudbg_dump_context_size(struct adapter *padap)
1896 {
1897 	struct cudbg_region_info region_info[CTXT_CNM + 1] = { {0} };
1898 	u8 mem_type[CTXT_INGRESS + 1] = { 0 };
1899 	u32 i, size = 0;
1900 	int rc;
1901 
1902 	/* Get max valid qid for each type of queue */
1903 	rc = cudbg_get_ctxt_region_info(padap, region_info, mem_type);
1904 	if (rc)
1905 		return rc;
1906 
1907 	for (i = 0; i < CTXT_CNM; i++) {
1908 		if (!region_info[i].exist) {
1909 			if (i == CTXT_EGRESS || i == CTXT_INGRESS)
1910 				size += CUDBG_LOWMEM_MAX_CTXT_QIDS *
1911 					SGE_CTXT_SIZE;
1912 			continue;
1913 		}
1914 
1915 		size += (region_info[i].end - region_info[i].start + 1) /
1916 			SGE_CTXT_SIZE;
1917 	}
1918 	return size * sizeof(struct cudbg_ch_cntxt);
1919 }
1920 
1921 static void cudbg_read_sge_ctxt(struct cudbg_init *pdbg_init, u32 cid,
1922 				enum ctxt_type ctype, u32 *data)
1923 {
1924 	struct adapter *padap = pdbg_init->adap;
1925 	int rc = -1;
1926 
1927 	/* Under heavy traffic, the SGE Queue contexts registers will be
1928 	 * frequently accessed by firmware.
1929 	 *
1930 	 * To avoid conflicts with firmware, always ask firmware to fetch
1931 	 * the SGE Queue contexts via mailbox. On failure, fallback to
1932 	 * accessing hardware registers directly.
1933 	 */
1934 	if (is_fw_attached(pdbg_init))
1935 		rc = t4_sge_ctxt_rd(padap, padap->mbox, cid, ctype, data);
1936 	if (rc)
1937 		t4_sge_ctxt_rd_bd(padap, cid, ctype, data);
1938 }
1939 
1940 static void cudbg_get_sge_ctxt_fw(struct cudbg_init *pdbg_init, u32 max_qid,
1941 				  u8 ctxt_type,
1942 				  struct cudbg_ch_cntxt **out_buff)
1943 {
1944 	struct cudbg_ch_cntxt *buff = *out_buff;
1945 	int rc;
1946 	u32 j;
1947 
1948 	for (j = 0; j < max_qid; j++) {
1949 		cudbg_read_sge_ctxt(pdbg_init, j, ctxt_type, buff->data);
1950 		rc = cudbg_sge_ctxt_check_valid(buff->data, ctxt_type);
1951 		if (!rc)
1952 			continue;
1953 
1954 		buff->cntxt_type = ctxt_type;
1955 		buff->cntxt_id = j;
1956 		buff++;
1957 		if (ctxt_type == CTXT_FLM) {
1958 			cudbg_read_sge_ctxt(pdbg_init, j, CTXT_CNM, buff->data);
1959 			buff->cntxt_type = CTXT_CNM;
1960 			buff->cntxt_id = j;
1961 			buff++;
1962 		}
1963 	}
1964 
1965 	*out_buff = buff;
1966 }
1967 
1968 int cudbg_collect_dump_context(struct cudbg_init *pdbg_init,
1969 			       struct cudbg_buffer *dbg_buff,
1970 			       struct cudbg_error *cudbg_err)
1971 {
1972 	struct cudbg_region_info region_info[CTXT_CNM + 1] = { {0} };
1973 	struct adapter *padap = pdbg_init->adap;
1974 	u32 j, size, max_ctx_size, max_ctx_qid;
1975 	u8 mem_type[CTXT_INGRESS + 1] = { 0 };
1976 	struct cudbg_buffer temp_buff = { 0 };
1977 	struct cudbg_ch_cntxt *buff;
1978 	u64 *dst_off, *src_off;
1979 	u8 *ctx_buf;
1980 	u8 i, k;
1981 	int rc;
1982 
1983 	/* Get max valid qid for each type of queue */
1984 	rc = cudbg_get_ctxt_region_info(padap, region_info, mem_type);
1985 	if (rc)
1986 		return rc;
1987 
1988 	rc = cudbg_dump_context_size(padap);
1989 	if (rc <= 0)
1990 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
1991 
1992 	size = rc;
1993 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1994 	if (rc)
1995 		return rc;
1996 
1997 	/* Get buffer with enough space to read the biggest context
1998 	 * region in memory.
1999 	 */
2000 	max_ctx_size = max(region_info[CTXT_EGRESS].end -
2001 			   region_info[CTXT_EGRESS].start + 1,
2002 			   region_info[CTXT_INGRESS].end -
2003 			   region_info[CTXT_INGRESS].start + 1);
2004 
2005 	ctx_buf = kvzalloc(max_ctx_size, GFP_KERNEL);
2006 	if (!ctx_buf) {
2007 		cudbg_put_buff(pdbg_init, &temp_buff);
2008 		return -ENOMEM;
2009 	}
2010 
2011 	buff = (struct cudbg_ch_cntxt *)temp_buff.data;
2012 
2013 	/* Collect EGRESS and INGRESS context data.
2014 	 * In case of failures, fallback to collecting via FW or
2015 	 * backdoor access.
2016 	 */
2017 	for (i = CTXT_EGRESS; i <= CTXT_INGRESS; i++) {
2018 		if (!region_info[i].exist) {
2019 			max_ctx_qid = CUDBG_LOWMEM_MAX_CTXT_QIDS;
2020 			cudbg_get_sge_ctxt_fw(pdbg_init, max_ctx_qid, i,
2021 					      &buff);
2022 			continue;
2023 		}
2024 
2025 		max_ctx_size = region_info[i].end - region_info[i].start + 1;
2026 		max_ctx_qid = max_ctx_size / SGE_CTXT_SIZE;
2027 
2028 		/* If firmware is not attached/alive, use backdoor register
2029 		 * access to collect dump.
2030 		 */
2031 		if (is_fw_attached(pdbg_init)) {
2032 			t4_sge_ctxt_flush(padap, padap->mbox, i);
2033 
2034 			rc = t4_memory_rw(padap, MEMWIN_NIC, mem_type[i],
2035 					  region_info[i].start, max_ctx_size,
2036 					  (__be32 *)ctx_buf, 1);
2037 		}
2038 
2039 		if (rc || !is_fw_attached(pdbg_init)) {
2040 			max_ctx_qid = CUDBG_LOWMEM_MAX_CTXT_QIDS;
2041 			cudbg_get_sge_ctxt_fw(pdbg_init, max_ctx_qid, i,
2042 					      &buff);
2043 			continue;
2044 		}
2045 
2046 		for (j = 0; j < max_ctx_qid; j++) {
2047 			src_off = (u64 *)(ctx_buf + j * SGE_CTXT_SIZE);
2048 			dst_off = (u64 *)buff->data;
2049 
2050 			/* The data is stored in 64-bit cpu order.  Convert it
2051 			 * to big endian before parsing.
2052 			 */
2053 			for (k = 0; k < SGE_CTXT_SIZE / sizeof(u64); k++)
2054 				dst_off[k] = cpu_to_be64(src_off[k]);
2055 
2056 			rc = cudbg_sge_ctxt_check_valid(buff->data, i);
2057 			if (!rc)
2058 				continue;
2059 
2060 			buff->cntxt_type = i;
2061 			buff->cntxt_id = j;
2062 			buff++;
2063 		}
2064 	}
2065 
2066 	kvfree(ctx_buf);
2067 
2068 	/* Collect FREELIST and CONGESTION MANAGER contexts */
2069 	max_ctx_size = region_info[CTXT_FLM].end -
2070 		       region_info[CTXT_FLM].start + 1;
2071 	max_ctx_qid = max_ctx_size / SGE_CTXT_SIZE;
2072 	/* Since FLM and CONM are 1-to-1 mapped, the below function
2073 	 * will fetch both FLM and CONM contexts.
2074 	 */
2075 	cudbg_get_sge_ctxt_fw(pdbg_init, max_ctx_qid, CTXT_FLM, &buff);
2076 
2077 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2078 }
2079 
2080 static inline void cudbg_tcamxy2valmask(u64 x, u64 y, u8 *addr, u64 *mask)
2081 {
2082 	*mask = x | y;
2083 	y = (__force u64)cpu_to_be64(y);
2084 	memcpy(addr, (char *)&y + 2, ETH_ALEN);
2085 }
2086 
2087 static void cudbg_mps_rpl_backdoor(struct adapter *padap,
2088 				   struct fw_ldst_mps_rplc *mps_rplc)
2089 {
2090 	if (is_t5(padap->params.chip)) {
2091 		mps_rplc->rplc255_224 = htonl(t4_read_reg(padap,
2092 							  MPS_VF_RPLCT_MAP3_A));
2093 		mps_rplc->rplc223_192 = htonl(t4_read_reg(padap,
2094 							  MPS_VF_RPLCT_MAP2_A));
2095 		mps_rplc->rplc191_160 = htonl(t4_read_reg(padap,
2096 							  MPS_VF_RPLCT_MAP1_A));
2097 		mps_rplc->rplc159_128 = htonl(t4_read_reg(padap,
2098 							  MPS_VF_RPLCT_MAP0_A));
2099 	} else {
2100 		mps_rplc->rplc255_224 = htonl(t4_read_reg(padap,
2101 							  MPS_VF_RPLCT_MAP7_A));
2102 		mps_rplc->rplc223_192 = htonl(t4_read_reg(padap,
2103 							  MPS_VF_RPLCT_MAP6_A));
2104 		mps_rplc->rplc191_160 = htonl(t4_read_reg(padap,
2105 							  MPS_VF_RPLCT_MAP5_A));
2106 		mps_rplc->rplc159_128 = htonl(t4_read_reg(padap,
2107 							  MPS_VF_RPLCT_MAP4_A));
2108 	}
2109 	mps_rplc->rplc127_96 = htonl(t4_read_reg(padap, MPS_VF_RPLCT_MAP3_A));
2110 	mps_rplc->rplc95_64 = htonl(t4_read_reg(padap, MPS_VF_RPLCT_MAP2_A));
2111 	mps_rplc->rplc63_32 = htonl(t4_read_reg(padap, MPS_VF_RPLCT_MAP1_A));
2112 	mps_rplc->rplc31_0 = htonl(t4_read_reg(padap, MPS_VF_RPLCT_MAP0_A));
2113 }
2114 
2115 static int cudbg_collect_tcam_index(struct cudbg_init *pdbg_init,
2116 				    struct cudbg_mps_tcam *tcam, u32 idx)
2117 {
2118 	struct adapter *padap = pdbg_init->adap;
2119 	u64 tcamy, tcamx, val;
2120 	u32 ctl, data2;
2121 	int rc = 0;
2122 
2123 	if (CHELSIO_CHIP_VERSION(padap->params.chip) >= CHELSIO_T6) {
2124 		/* CtlReqID   - 1: use Host Driver Requester ID
2125 		 * CtlCmdType - 0: Read, 1: Write
2126 		 * CtlTcamSel - 0: TCAM0, 1: TCAM1
2127 		 * CtlXYBitSel- 0: Y bit, 1: X bit
2128 		 */
2129 
2130 		/* Read tcamy */
2131 		ctl = CTLREQID_V(1) | CTLCMDTYPE_V(0) | CTLXYBITSEL_V(0);
2132 		if (idx < 256)
2133 			ctl |= CTLTCAMINDEX_V(idx) | CTLTCAMSEL_V(0);
2134 		else
2135 			ctl |= CTLTCAMINDEX_V(idx - 256) | CTLTCAMSEL_V(1);
2136 
2137 		t4_write_reg(padap, MPS_CLS_TCAM_DATA2_CTL_A, ctl);
2138 		val = t4_read_reg(padap, MPS_CLS_TCAM_RDATA1_REQ_ID1_A);
2139 		tcamy = DMACH_G(val) << 32;
2140 		tcamy |= t4_read_reg(padap, MPS_CLS_TCAM_RDATA0_REQ_ID1_A);
2141 		data2 = t4_read_reg(padap, MPS_CLS_TCAM_RDATA2_REQ_ID1_A);
2142 		tcam->lookup_type = DATALKPTYPE_G(data2);
2143 
2144 		/* 0 - Outer header, 1 - Inner header
2145 		 * [71:48] bit locations are overloaded for
2146 		 * outer vs. inner lookup types.
2147 		 */
2148 		if (tcam->lookup_type && tcam->lookup_type != DATALKPTYPE_M) {
2149 			/* Inner header VNI */
2150 			tcam->vniy = (data2 & DATAVIDH2_F) | DATAVIDH1_G(data2);
2151 			tcam->vniy = (tcam->vniy << 16) | VIDL_G(val);
2152 			tcam->dip_hit = data2 & DATADIPHIT_F;
2153 		} else {
2154 			tcam->vlan_vld = data2 & DATAVIDH2_F;
2155 			tcam->ivlan = VIDL_G(val);
2156 		}
2157 
2158 		tcam->port_num = DATAPORTNUM_G(data2);
2159 
2160 		/* Read tcamx. Change the control param */
2161 		ctl |= CTLXYBITSEL_V(1);
2162 		t4_write_reg(padap, MPS_CLS_TCAM_DATA2_CTL_A, ctl);
2163 		val = t4_read_reg(padap, MPS_CLS_TCAM_RDATA1_REQ_ID1_A);
2164 		tcamx = DMACH_G(val) << 32;
2165 		tcamx |= t4_read_reg(padap, MPS_CLS_TCAM_RDATA0_REQ_ID1_A);
2166 		data2 = t4_read_reg(padap, MPS_CLS_TCAM_RDATA2_REQ_ID1_A);
2167 		if (tcam->lookup_type && tcam->lookup_type != DATALKPTYPE_M) {
2168 			/* Inner header VNI mask */
2169 			tcam->vnix = (data2 & DATAVIDH2_F) | DATAVIDH1_G(data2);
2170 			tcam->vnix = (tcam->vnix << 16) | VIDL_G(val);
2171 		}
2172 	} else {
2173 		tcamy = t4_read_reg64(padap, MPS_CLS_TCAM_Y_L(idx));
2174 		tcamx = t4_read_reg64(padap, MPS_CLS_TCAM_X_L(idx));
2175 	}
2176 
2177 	/* If no entry, return */
2178 	if (tcamx & tcamy)
2179 		return rc;
2180 
2181 	tcam->cls_lo = t4_read_reg(padap, MPS_CLS_SRAM_L(idx));
2182 	tcam->cls_hi = t4_read_reg(padap, MPS_CLS_SRAM_H(idx));
2183 
2184 	if (is_t5(padap->params.chip))
2185 		tcam->repli = (tcam->cls_lo & REPLICATE_F);
2186 	else if (is_t6(padap->params.chip))
2187 		tcam->repli = (tcam->cls_lo & T6_REPLICATE_F);
2188 
2189 	if (tcam->repli) {
2190 		struct fw_ldst_cmd ldst_cmd;
2191 		struct fw_ldst_mps_rplc mps_rplc;
2192 
2193 		memset(&ldst_cmd, 0, sizeof(ldst_cmd));
2194 		ldst_cmd.op_to_addrspace =
2195 			htonl(FW_CMD_OP_V(FW_LDST_CMD) |
2196 			      FW_CMD_REQUEST_F | FW_CMD_READ_F |
2197 			      FW_LDST_CMD_ADDRSPACE_V(FW_LDST_ADDRSPC_MPS));
2198 		ldst_cmd.cycles_to_len16 = htonl(FW_LEN16(ldst_cmd));
2199 		ldst_cmd.u.mps.rplc.fid_idx =
2200 			htons(FW_LDST_CMD_FID_V(FW_LDST_MPS_RPLC) |
2201 			      FW_LDST_CMD_IDX_V(idx));
2202 
2203 		/* If firmware is not attached/alive, use backdoor register
2204 		 * access to collect dump.
2205 		 */
2206 		if (is_fw_attached(pdbg_init))
2207 			rc = t4_wr_mbox(padap, padap->mbox, &ldst_cmd,
2208 					sizeof(ldst_cmd), &ldst_cmd);
2209 
2210 		if (rc || !is_fw_attached(pdbg_init)) {
2211 			cudbg_mps_rpl_backdoor(padap, &mps_rplc);
2212 			/* Ignore error since we collected directly from
2213 			 * reading registers.
2214 			 */
2215 			rc = 0;
2216 		} else {
2217 			mps_rplc = ldst_cmd.u.mps.rplc;
2218 		}
2219 
2220 		tcam->rplc[0] = ntohl(mps_rplc.rplc31_0);
2221 		tcam->rplc[1] = ntohl(mps_rplc.rplc63_32);
2222 		tcam->rplc[2] = ntohl(mps_rplc.rplc95_64);
2223 		tcam->rplc[3] = ntohl(mps_rplc.rplc127_96);
2224 		if (padap->params.arch.mps_rplc_size > CUDBG_MAX_RPLC_SIZE) {
2225 			tcam->rplc[4] = ntohl(mps_rplc.rplc159_128);
2226 			tcam->rplc[5] = ntohl(mps_rplc.rplc191_160);
2227 			tcam->rplc[6] = ntohl(mps_rplc.rplc223_192);
2228 			tcam->rplc[7] = ntohl(mps_rplc.rplc255_224);
2229 		}
2230 	}
2231 	cudbg_tcamxy2valmask(tcamx, tcamy, tcam->addr, &tcam->mask);
2232 	tcam->idx = idx;
2233 	tcam->rplc_size = padap->params.arch.mps_rplc_size;
2234 	return rc;
2235 }
2236 
2237 int cudbg_collect_mps_tcam(struct cudbg_init *pdbg_init,
2238 			   struct cudbg_buffer *dbg_buff,
2239 			   struct cudbg_error *cudbg_err)
2240 {
2241 	struct adapter *padap = pdbg_init->adap;
2242 	struct cudbg_buffer temp_buff = { 0 };
2243 	u32 size = 0, i, n, total_size = 0;
2244 	struct cudbg_mps_tcam *tcam;
2245 	int rc;
2246 
2247 	n = padap->params.arch.mps_tcam_size;
2248 	size = sizeof(struct cudbg_mps_tcam) * n;
2249 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2250 	if (rc)
2251 		return rc;
2252 
2253 	tcam = (struct cudbg_mps_tcam *)temp_buff.data;
2254 	for (i = 0; i < n; i++) {
2255 		rc = cudbg_collect_tcam_index(pdbg_init, tcam, i);
2256 		if (rc) {
2257 			cudbg_err->sys_err = rc;
2258 			cudbg_put_buff(pdbg_init, &temp_buff);
2259 			return rc;
2260 		}
2261 		total_size += sizeof(struct cudbg_mps_tcam);
2262 		tcam++;
2263 	}
2264 
2265 	if (!total_size) {
2266 		rc = CUDBG_SYSTEM_ERROR;
2267 		cudbg_err->sys_err = rc;
2268 		cudbg_put_buff(pdbg_init, &temp_buff);
2269 		return rc;
2270 	}
2271 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2272 }
2273 
2274 int cudbg_collect_vpd_data(struct cudbg_init *pdbg_init,
2275 			   struct cudbg_buffer *dbg_buff,
2276 			   struct cudbg_error *cudbg_err)
2277 {
2278 	struct adapter *padap = pdbg_init->adap;
2279 	struct cudbg_buffer temp_buff = { 0 };
2280 	char vpd_str[CUDBG_VPD_VER_LEN + 1];
2281 	u32 scfg_vers, vpd_vers, fw_vers;
2282 	struct cudbg_vpd_data *vpd_data;
2283 	struct vpd_params vpd = { 0 };
2284 	int rc, ret;
2285 
2286 	rc = t4_get_raw_vpd_params(padap, &vpd);
2287 	if (rc)
2288 		return rc;
2289 
2290 	rc = t4_get_fw_version(padap, &fw_vers);
2291 	if (rc)
2292 		return rc;
2293 
2294 	/* Serial Configuration Version is located beyond the PF's vpd size.
2295 	 * Temporarily give access to entire EEPROM to get it.
2296 	 */
2297 	rc = pci_set_vpd_size(padap->pdev, EEPROMVSIZE);
2298 	if (rc < 0)
2299 		return rc;
2300 
2301 	ret = cudbg_read_vpd_reg(padap, CUDBG_SCFG_VER_ADDR, CUDBG_SCFG_VER_LEN,
2302 				 &scfg_vers);
2303 
2304 	/* Restore back to original PF's vpd size */
2305 	rc = pci_set_vpd_size(padap->pdev, CUDBG_VPD_PF_SIZE);
2306 	if (rc < 0)
2307 		return rc;
2308 
2309 	if (ret)
2310 		return ret;
2311 
2312 	rc = cudbg_read_vpd_reg(padap, CUDBG_VPD_VER_ADDR, CUDBG_VPD_VER_LEN,
2313 				vpd_str);
2314 	if (rc)
2315 		return rc;
2316 
2317 	vpd_str[CUDBG_VPD_VER_LEN] = '\0';
2318 	rc = kstrtouint(vpd_str, 0, &vpd_vers);
2319 	if (rc)
2320 		return rc;
2321 
2322 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_vpd_data),
2323 			    &temp_buff);
2324 	if (rc)
2325 		return rc;
2326 
2327 	vpd_data = (struct cudbg_vpd_data *)temp_buff.data;
2328 	memcpy(vpd_data->sn, vpd.sn, SERNUM_LEN + 1);
2329 	memcpy(vpd_data->bn, vpd.pn, PN_LEN + 1);
2330 	memcpy(vpd_data->na, vpd.na, MACADDR_LEN + 1);
2331 	memcpy(vpd_data->mn, vpd.id, ID_LEN + 1);
2332 	vpd_data->scfg_vers = scfg_vers;
2333 	vpd_data->vpd_vers = vpd_vers;
2334 	vpd_data->fw_major = FW_HDR_FW_VER_MAJOR_G(fw_vers);
2335 	vpd_data->fw_minor = FW_HDR_FW_VER_MINOR_G(fw_vers);
2336 	vpd_data->fw_micro = FW_HDR_FW_VER_MICRO_G(fw_vers);
2337 	vpd_data->fw_build = FW_HDR_FW_VER_BUILD_G(fw_vers);
2338 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2339 }
2340 
2341 static int cudbg_read_tid(struct cudbg_init *pdbg_init, u32 tid,
2342 			  struct cudbg_tid_data *tid_data)
2343 {
2344 	struct adapter *padap = pdbg_init->adap;
2345 	int i, cmd_retry = 8;
2346 	u32 val;
2347 
2348 	/* Fill REQ_DATA regs with 0's */
2349 	for (i = 0; i < NUM_LE_DB_DBGI_REQ_DATA_INSTANCES; i++)
2350 		t4_write_reg(padap, LE_DB_DBGI_REQ_DATA_A + (i << 2), 0);
2351 
2352 	/* Write DBIG command */
2353 	val = DBGICMD_V(4) | DBGITID_V(tid);
2354 	t4_write_reg(padap, LE_DB_DBGI_REQ_TCAM_CMD_A, val);
2355 	tid_data->dbig_cmd = val;
2356 
2357 	val = DBGICMDSTRT_F | DBGICMDMODE_V(1); /* LE mode */
2358 	t4_write_reg(padap, LE_DB_DBGI_CONFIG_A, val);
2359 	tid_data->dbig_conf = val;
2360 
2361 	/* Poll the DBGICMDBUSY bit */
2362 	val = 1;
2363 	while (val) {
2364 		val = t4_read_reg(padap, LE_DB_DBGI_CONFIG_A);
2365 		val = val & DBGICMDBUSY_F;
2366 		cmd_retry--;
2367 		if (!cmd_retry)
2368 			return CUDBG_SYSTEM_ERROR;
2369 	}
2370 
2371 	/* Check RESP status */
2372 	val = t4_read_reg(padap, LE_DB_DBGI_RSP_STATUS_A);
2373 	tid_data->dbig_rsp_stat = val;
2374 	if (!(val & 1))
2375 		return CUDBG_SYSTEM_ERROR;
2376 
2377 	/* Read RESP data */
2378 	for (i = 0; i < NUM_LE_DB_DBGI_RSP_DATA_INSTANCES; i++)
2379 		tid_data->data[i] = t4_read_reg(padap,
2380 						LE_DB_DBGI_RSP_DATA_A +
2381 						(i << 2));
2382 	tid_data->tid = tid;
2383 	return 0;
2384 }
2385 
2386 static int cudbg_get_le_type(u32 tid, struct cudbg_tcam tcam_region)
2387 {
2388 	int type = LE_ET_UNKNOWN;
2389 
2390 	if (tid < tcam_region.server_start)
2391 		type = LE_ET_TCAM_CON;
2392 	else if (tid < tcam_region.filter_start)
2393 		type = LE_ET_TCAM_SERVER;
2394 	else if (tid < tcam_region.clip_start)
2395 		type = LE_ET_TCAM_FILTER;
2396 	else if (tid < tcam_region.routing_start)
2397 		type = LE_ET_TCAM_CLIP;
2398 	else if (tid < tcam_region.tid_hash_base)
2399 		type = LE_ET_TCAM_ROUTING;
2400 	else if (tid < tcam_region.max_tid)
2401 		type = LE_ET_HASH_CON;
2402 	else
2403 		type = LE_ET_INVALID_TID;
2404 
2405 	return type;
2406 }
2407 
2408 static int cudbg_is_ipv6_entry(struct cudbg_tid_data *tid_data,
2409 			       struct cudbg_tcam tcam_region)
2410 {
2411 	int ipv6 = 0;
2412 	int le_type;
2413 
2414 	le_type = cudbg_get_le_type(tid_data->tid, tcam_region);
2415 	if (tid_data->tid & 1)
2416 		return 0;
2417 
2418 	if (le_type == LE_ET_HASH_CON) {
2419 		ipv6 = tid_data->data[16] & 0x8000;
2420 	} else if (le_type == LE_ET_TCAM_CON) {
2421 		ipv6 = tid_data->data[16] & 0x8000;
2422 		if (ipv6)
2423 			ipv6 = tid_data->data[9] == 0x00C00000;
2424 	} else {
2425 		ipv6 = 0;
2426 	}
2427 	return ipv6;
2428 }
2429 
2430 void cudbg_fill_le_tcam_info(struct adapter *padap,
2431 			     struct cudbg_tcam *tcam_region)
2432 {
2433 	u32 value;
2434 
2435 	/* Get the LE regions */
2436 	value = t4_read_reg(padap, LE_DB_TID_HASHBASE_A); /* hash base index */
2437 	tcam_region->tid_hash_base = value;
2438 
2439 	/* Get routing table index */
2440 	value = t4_read_reg(padap, LE_DB_ROUTING_TABLE_INDEX_A);
2441 	tcam_region->routing_start = value;
2442 
2443 	/* Get clip table index. For T6 there is separate CLIP TCAM */
2444 	if (is_t6(padap->params.chip))
2445 		value = t4_read_reg(padap, LE_DB_CLCAM_TID_BASE_A);
2446 	else
2447 		value = t4_read_reg(padap, LE_DB_CLIP_TABLE_INDEX_A);
2448 	tcam_region->clip_start = value;
2449 
2450 	/* Get filter table index */
2451 	value = t4_read_reg(padap, LE_DB_FILTER_TABLE_INDEX_A);
2452 	tcam_region->filter_start = value;
2453 
2454 	/* Get server table index */
2455 	value = t4_read_reg(padap, LE_DB_SERVER_INDEX_A);
2456 	tcam_region->server_start = value;
2457 
2458 	/* Check whether hash is enabled and calculate the max tids */
2459 	value = t4_read_reg(padap, LE_DB_CONFIG_A);
2460 	if ((value >> HASHEN_S) & 1) {
2461 		value = t4_read_reg(padap, LE_DB_HASH_CONFIG_A);
2462 		if (CHELSIO_CHIP_VERSION(padap->params.chip) > CHELSIO_T5) {
2463 			tcam_region->max_tid = (value & 0xFFFFF) +
2464 					       tcam_region->tid_hash_base;
2465 		} else {
2466 			value = HASHTIDSIZE_G(value);
2467 			value = 1 << value;
2468 			tcam_region->max_tid = value +
2469 					       tcam_region->tid_hash_base;
2470 		}
2471 	} else { /* hash not enabled */
2472 		if (is_t6(padap->params.chip))
2473 			tcam_region->max_tid = (value & ASLIPCOMPEN_F) ?
2474 					       CUDBG_MAX_TID_COMP_EN :
2475 					       CUDBG_MAX_TID_COMP_DIS;
2476 		else
2477 			tcam_region->max_tid = CUDBG_MAX_TCAM_TID;
2478 	}
2479 
2480 	if (is_t6(padap->params.chip))
2481 		tcam_region->max_tid += CUDBG_T6_CLIP;
2482 }
2483 
2484 int cudbg_collect_le_tcam(struct cudbg_init *pdbg_init,
2485 			  struct cudbg_buffer *dbg_buff,
2486 			  struct cudbg_error *cudbg_err)
2487 {
2488 	struct adapter *padap = pdbg_init->adap;
2489 	struct cudbg_buffer temp_buff = { 0 };
2490 	struct cudbg_tcam tcam_region = { 0 };
2491 	struct cudbg_tid_data *tid_data;
2492 	u32 bytes = 0;
2493 	int rc, size;
2494 	u32 i;
2495 
2496 	cudbg_fill_le_tcam_info(padap, &tcam_region);
2497 
2498 	size = sizeof(struct cudbg_tid_data) * tcam_region.max_tid;
2499 	size += sizeof(struct cudbg_tcam);
2500 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2501 	if (rc)
2502 		return rc;
2503 
2504 	memcpy(temp_buff.data, &tcam_region, sizeof(struct cudbg_tcam));
2505 	bytes = sizeof(struct cudbg_tcam);
2506 	tid_data = (struct cudbg_tid_data *)(temp_buff.data + bytes);
2507 	/* read all tid */
2508 	for (i = 0; i < tcam_region.max_tid; ) {
2509 		rc = cudbg_read_tid(pdbg_init, i, tid_data);
2510 		if (rc) {
2511 			cudbg_err->sys_warn = CUDBG_STATUS_PARTIAL_DATA;
2512 			/* Update tcam header and exit */
2513 			tcam_region.max_tid = i;
2514 			memcpy(temp_buff.data, &tcam_region,
2515 			       sizeof(struct cudbg_tcam));
2516 			goto out;
2517 		}
2518 
2519 		if (cudbg_is_ipv6_entry(tid_data, tcam_region)) {
2520 			/* T6 CLIP TCAM: ipv6 takes 4 entries */
2521 			if (is_t6(padap->params.chip) &&
2522 			    i >= tcam_region.clip_start &&
2523 			    i < tcam_region.clip_start + CUDBG_T6_CLIP)
2524 				i += 4;
2525 			else /* Main TCAM: ipv6 takes two tids */
2526 				i += 2;
2527 		} else {
2528 			i++;
2529 		}
2530 
2531 		tid_data++;
2532 		bytes += sizeof(struct cudbg_tid_data);
2533 	}
2534 
2535 out:
2536 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2537 }
2538 
2539 int cudbg_collect_cctrl(struct cudbg_init *pdbg_init,
2540 			struct cudbg_buffer *dbg_buff,
2541 			struct cudbg_error *cudbg_err)
2542 {
2543 	struct adapter *padap = pdbg_init->adap;
2544 	struct cudbg_buffer temp_buff = { 0 };
2545 	u32 size;
2546 	int rc;
2547 
2548 	size = sizeof(u16) * NMTUS * NCCTRL_WIN;
2549 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2550 	if (rc)
2551 		return rc;
2552 
2553 	t4_read_cong_tbl(padap, (void *)temp_buff.data);
2554 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2555 }
2556 
2557 int cudbg_collect_ma_indirect(struct cudbg_init *pdbg_init,
2558 			      struct cudbg_buffer *dbg_buff,
2559 			      struct cudbg_error *cudbg_err)
2560 {
2561 	struct adapter *padap = pdbg_init->adap;
2562 	struct cudbg_buffer temp_buff = { 0 };
2563 	struct ireg_buf *ma_indr;
2564 	int i, rc, n;
2565 	u32 size, j;
2566 
2567 	if (CHELSIO_CHIP_VERSION(padap->params.chip) < CHELSIO_T6)
2568 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
2569 
2570 	n = sizeof(t6_ma_ireg_array) / (IREG_NUM_ELEM * sizeof(u32));
2571 	size = sizeof(struct ireg_buf) * n * 2;
2572 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2573 	if (rc)
2574 		return rc;
2575 
2576 	ma_indr = (struct ireg_buf *)temp_buff.data;
2577 	for (i = 0; i < n; i++) {
2578 		struct ireg_field *ma_fli = &ma_indr->tp_pio;
2579 		u32 *buff = ma_indr->outbuf;
2580 
2581 		ma_fli->ireg_addr = t6_ma_ireg_array[i][0];
2582 		ma_fli->ireg_data = t6_ma_ireg_array[i][1];
2583 		ma_fli->ireg_local_offset = t6_ma_ireg_array[i][2];
2584 		ma_fli->ireg_offset_range = t6_ma_ireg_array[i][3];
2585 		t4_read_indirect(padap, ma_fli->ireg_addr, ma_fli->ireg_data,
2586 				 buff, ma_fli->ireg_offset_range,
2587 				 ma_fli->ireg_local_offset);
2588 		ma_indr++;
2589 	}
2590 
2591 	n = sizeof(t6_ma_ireg_array2) / (IREG_NUM_ELEM * sizeof(u32));
2592 	for (i = 0; i < n; i++) {
2593 		struct ireg_field *ma_fli = &ma_indr->tp_pio;
2594 		u32 *buff = ma_indr->outbuf;
2595 
2596 		ma_fli->ireg_addr = t6_ma_ireg_array2[i][0];
2597 		ma_fli->ireg_data = t6_ma_ireg_array2[i][1];
2598 		ma_fli->ireg_local_offset = t6_ma_ireg_array2[i][2];
2599 		for (j = 0; j < t6_ma_ireg_array2[i][3]; j++) {
2600 			t4_read_indirect(padap, ma_fli->ireg_addr,
2601 					 ma_fli->ireg_data, buff, 1,
2602 					 ma_fli->ireg_local_offset);
2603 			buff++;
2604 			ma_fli->ireg_local_offset += 0x20;
2605 		}
2606 		ma_indr++;
2607 	}
2608 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2609 }
2610 
2611 int cudbg_collect_ulptx_la(struct cudbg_init *pdbg_init,
2612 			   struct cudbg_buffer *dbg_buff,
2613 			   struct cudbg_error *cudbg_err)
2614 {
2615 	struct adapter *padap = pdbg_init->adap;
2616 	struct cudbg_buffer temp_buff = { 0 };
2617 	struct cudbg_ulptx_la *ulptx_la_buff;
2618 	struct cudbg_ver_hdr *ver_hdr;
2619 	u32 i, j;
2620 	int rc;
2621 
2622 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
2623 			    sizeof(struct cudbg_ver_hdr) +
2624 			    sizeof(struct cudbg_ulptx_la),
2625 			    &temp_buff);
2626 	if (rc)
2627 		return rc;
2628 
2629 	ver_hdr = (struct cudbg_ver_hdr *)temp_buff.data;
2630 	ver_hdr->signature = CUDBG_ENTITY_SIGNATURE;
2631 	ver_hdr->revision = CUDBG_ULPTX_LA_REV;
2632 	ver_hdr->size = sizeof(struct cudbg_ulptx_la);
2633 
2634 	ulptx_la_buff = (struct cudbg_ulptx_la *)(temp_buff.data +
2635 						  sizeof(*ver_hdr));
2636 	for (i = 0; i < CUDBG_NUM_ULPTX; i++) {
2637 		ulptx_la_buff->rdptr[i] = t4_read_reg(padap,
2638 						      ULP_TX_LA_RDPTR_0_A +
2639 						      0x10 * i);
2640 		ulptx_la_buff->wrptr[i] = t4_read_reg(padap,
2641 						      ULP_TX_LA_WRPTR_0_A +
2642 						      0x10 * i);
2643 		ulptx_la_buff->rddata[i] = t4_read_reg(padap,
2644 						       ULP_TX_LA_RDDATA_0_A +
2645 						       0x10 * i);
2646 		for (j = 0; j < CUDBG_NUM_ULPTX_READ; j++)
2647 			ulptx_la_buff->rd_data[i][j] =
2648 				t4_read_reg(padap,
2649 					    ULP_TX_LA_RDDATA_0_A + 0x10 * i);
2650 	}
2651 
2652 	for (i = 0; i < CUDBG_NUM_ULPTX_ASIC_READ; i++) {
2653 		t4_write_reg(padap, ULP_TX_ASIC_DEBUG_CTRL_A, 0x1);
2654 		ulptx_la_buff->rdptr_asic[i] =
2655 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_CTRL_A);
2656 		ulptx_la_buff->rddata_asic[i][0] =
2657 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_0_A);
2658 		ulptx_la_buff->rddata_asic[i][1] =
2659 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_1_A);
2660 		ulptx_la_buff->rddata_asic[i][2] =
2661 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_2_A);
2662 		ulptx_la_buff->rddata_asic[i][3] =
2663 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_3_A);
2664 		ulptx_la_buff->rddata_asic[i][4] =
2665 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_4_A);
2666 		ulptx_la_buff->rddata_asic[i][5] =
2667 				t4_read_reg(padap, PM_RX_BASE_ADDR);
2668 	}
2669 
2670 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2671 }
2672 
2673 int cudbg_collect_up_cim_indirect(struct cudbg_init *pdbg_init,
2674 				  struct cudbg_buffer *dbg_buff,
2675 				  struct cudbg_error *cudbg_err)
2676 {
2677 	struct adapter *padap = pdbg_init->adap;
2678 	struct cudbg_buffer temp_buff = { 0 };
2679 	u32 local_offset, local_range;
2680 	struct ireg_buf *up_cim;
2681 	u32 size, j, iter;
2682 	u32 instance = 0;
2683 	int i, rc, n;
2684 
2685 	if (is_t5(padap->params.chip))
2686 		n = sizeof(t5_up_cim_reg_array) /
2687 		    ((IREG_NUM_ELEM + 1) * sizeof(u32));
2688 	else if (is_t6(padap->params.chip))
2689 		n = sizeof(t6_up_cim_reg_array) /
2690 		    ((IREG_NUM_ELEM + 1) * sizeof(u32));
2691 	else
2692 		return CUDBG_STATUS_NOT_IMPLEMENTED;
2693 
2694 	size = sizeof(struct ireg_buf) * n;
2695 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2696 	if (rc)
2697 		return rc;
2698 
2699 	up_cim = (struct ireg_buf *)temp_buff.data;
2700 	for (i = 0; i < n; i++) {
2701 		struct ireg_field *up_cim_reg = &up_cim->tp_pio;
2702 		u32 *buff = up_cim->outbuf;
2703 
2704 		if (is_t5(padap->params.chip)) {
2705 			up_cim_reg->ireg_addr = t5_up_cim_reg_array[i][0];
2706 			up_cim_reg->ireg_data = t5_up_cim_reg_array[i][1];
2707 			up_cim_reg->ireg_local_offset =
2708 						t5_up_cim_reg_array[i][2];
2709 			up_cim_reg->ireg_offset_range =
2710 						t5_up_cim_reg_array[i][3];
2711 			instance = t5_up_cim_reg_array[i][4];
2712 		} else if (is_t6(padap->params.chip)) {
2713 			up_cim_reg->ireg_addr = t6_up_cim_reg_array[i][0];
2714 			up_cim_reg->ireg_data = t6_up_cim_reg_array[i][1];
2715 			up_cim_reg->ireg_local_offset =
2716 						t6_up_cim_reg_array[i][2];
2717 			up_cim_reg->ireg_offset_range =
2718 						t6_up_cim_reg_array[i][3];
2719 			instance = t6_up_cim_reg_array[i][4];
2720 		}
2721 
2722 		switch (instance) {
2723 		case NUM_CIM_CTL_TSCH_CHANNEL_INSTANCES:
2724 			iter = up_cim_reg->ireg_offset_range;
2725 			local_offset = 0x120;
2726 			local_range = 1;
2727 			break;
2728 		case NUM_CIM_CTL_TSCH_CHANNEL_TSCH_CLASS_INSTANCES:
2729 			iter = up_cim_reg->ireg_offset_range;
2730 			local_offset = 0x10;
2731 			local_range = 1;
2732 			break;
2733 		default:
2734 			iter = 1;
2735 			local_offset = 0;
2736 			local_range = up_cim_reg->ireg_offset_range;
2737 			break;
2738 		}
2739 
2740 		for (j = 0; j < iter; j++, buff++) {
2741 			rc = t4_cim_read(padap,
2742 					 up_cim_reg->ireg_local_offset +
2743 					 (j * local_offset), local_range, buff);
2744 			if (rc) {
2745 				cudbg_put_buff(pdbg_init, &temp_buff);
2746 				return rc;
2747 			}
2748 		}
2749 		up_cim++;
2750 	}
2751 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2752 }
2753 
2754 int cudbg_collect_pbt_tables(struct cudbg_init *pdbg_init,
2755 			     struct cudbg_buffer *dbg_buff,
2756 			     struct cudbg_error *cudbg_err)
2757 {
2758 	struct adapter *padap = pdbg_init->adap;
2759 	struct cudbg_buffer temp_buff = { 0 };
2760 	struct cudbg_pbt_tables *pbt;
2761 	int i, rc;
2762 	u32 addr;
2763 
2764 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
2765 			    sizeof(struct cudbg_pbt_tables),
2766 			    &temp_buff);
2767 	if (rc)
2768 		return rc;
2769 
2770 	pbt = (struct cudbg_pbt_tables *)temp_buff.data;
2771 	/* PBT dynamic entries */
2772 	addr = CUDBG_CHAC_PBT_ADDR;
2773 	for (i = 0; i < CUDBG_PBT_DYNAMIC_ENTRIES; i++) {
2774 		rc = t4_cim_read(padap, addr + (i * 4), 1,
2775 				 &pbt->pbt_dynamic[i]);
2776 		if (rc) {
2777 			cudbg_err->sys_err = rc;
2778 			cudbg_put_buff(pdbg_init, &temp_buff);
2779 			return rc;
2780 		}
2781 	}
2782 
2783 	/* PBT static entries */
2784 	/* static entries start when bit 6 is set */
2785 	addr = CUDBG_CHAC_PBT_ADDR + (1 << 6);
2786 	for (i = 0; i < CUDBG_PBT_STATIC_ENTRIES; i++) {
2787 		rc = t4_cim_read(padap, addr + (i * 4), 1,
2788 				 &pbt->pbt_static[i]);
2789 		if (rc) {
2790 			cudbg_err->sys_err = rc;
2791 			cudbg_put_buff(pdbg_init, &temp_buff);
2792 			return rc;
2793 		}
2794 	}
2795 
2796 	/* LRF entries */
2797 	addr = CUDBG_CHAC_PBT_LRF;
2798 	for (i = 0; i < CUDBG_LRF_ENTRIES; i++) {
2799 		rc = t4_cim_read(padap, addr + (i * 4), 1,
2800 				 &pbt->lrf_table[i]);
2801 		if (rc) {
2802 			cudbg_err->sys_err = rc;
2803 			cudbg_put_buff(pdbg_init, &temp_buff);
2804 			return rc;
2805 		}
2806 	}
2807 
2808 	/* PBT data entries */
2809 	addr = CUDBG_CHAC_PBT_DATA;
2810 	for (i = 0; i < CUDBG_PBT_DATA_ENTRIES; i++) {
2811 		rc = t4_cim_read(padap, addr + (i * 4), 1,
2812 				 &pbt->pbt_data[i]);
2813 		if (rc) {
2814 			cudbg_err->sys_err = rc;
2815 			cudbg_put_buff(pdbg_init, &temp_buff);
2816 			return rc;
2817 		}
2818 	}
2819 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2820 }
2821 
2822 int cudbg_collect_mbox_log(struct cudbg_init *pdbg_init,
2823 			   struct cudbg_buffer *dbg_buff,
2824 			   struct cudbg_error *cudbg_err)
2825 {
2826 	struct adapter *padap = pdbg_init->adap;
2827 	struct cudbg_mbox_log *mboxlog = NULL;
2828 	struct cudbg_buffer temp_buff = { 0 };
2829 	struct mbox_cmd_log *log = NULL;
2830 	struct mbox_cmd *entry;
2831 	unsigned int entry_idx;
2832 	u16 mbox_cmds;
2833 	int i, k, rc;
2834 	u64 flit;
2835 	u32 size;
2836 
2837 	log = padap->mbox_log;
2838 	mbox_cmds = padap->mbox_log->size;
2839 	size = sizeof(struct cudbg_mbox_log) * mbox_cmds;
2840 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2841 	if (rc)
2842 		return rc;
2843 
2844 	mboxlog = (struct cudbg_mbox_log *)temp_buff.data;
2845 	for (k = 0; k < mbox_cmds; k++) {
2846 		entry_idx = log->cursor + k;
2847 		if (entry_idx >= log->size)
2848 			entry_idx -= log->size;
2849 
2850 		entry = mbox_cmd_log_entry(log, entry_idx);
2851 		/* skip over unused entries */
2852 		if (entry->timestamp == 0)
2853 			continue;
2854 
2855 		memcpy(&mboxlog->entry, entry, sizeof(struct mbox_cmd));
2856 		for (i = 0; i < MBOX_LEN / 8; i++) {
2857 			flit = entry->cmd[i];
2858 			mboxlog->hi[i] = (u32)(flit >> 32);
2859 			mboxlog->lo[i] = (u32)flit;
2860 		}
2861 		mboxlog++;
2862 	}
2863 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2864 }
2865 
2866 int cudbg_collect_hma_indirect(struct cudbg_init *pdbg_init,
2867 			       struct cudbg_buffer *dbg_buff,
2868 			       struct cudbg_error *cudbg_err)
2869 {
2870 	struct adapter *padap = pdbg_init->adap;
2871 	struct cudbg_buffer temp_buff = { 0 };
2872 	struct ireg_buf *hma_indr;
2873 	int i, rc, n;
2874 	u32 size;
2875 
2876 	if (CHELSIO_CHIP_VERSION(padap->params.chip) < CHELSIO_T6)
2877 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
2878 
2879 	n = sizeof(t6_hma_ireg_array) / (IREG_NUM_ELEM * sizeof(u32));
2880 	size = sizeof(struct ireg_buf) * n;
2881 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2882 	if (rc)
2883 		return rc;
2884 
2885 	hma_indr = (struct ireg_buf *)temp_buff.data;
2886 	for (i = 0; i < n; i++) {
2887 		struct ireg_field *hma_fli = &hma_indr->tp_pio;
2888 		u32 *buff = hma_indr->outbuf;
2889 
2890 		hma_fli->ireg_addr = t6_hma_ireg_array[i][0];
2891 		hma_fli->ireg_data = t6_hma_ireg_array[i][1];
2892 		hma_fli->ireg_local_offset = t6_hma_ireg_array[i][2];
2893 		hma_fli->ireg_offset_range = t6_hma_ireg_array[i][3];
2894 		t4_read_indirect(padap, hma_fli->ireg_addr, hma_fli->ireg_data,
2895 				 buff, hma_fli->ireg_offset_range,
2896 				 hma_fli->ireg_local_offset);
2897 		hma_indr++;
2898 	}
2899 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2900 }
2901 
2902 void cudbg_fill_qdesc_num_and_size(const struct adapter *padap,
2903 				   u32 *num, u32 *size)
2904 {
2905 	u32 tot_entries = 0, tot_size = 0;
2906 
2907 	/* NIC TXQ, RXQ, FLQ, and CTRLQ */
2908 	tot_entries += MAX_ETH_QSETS * 3;
2909 	tot_entries += MAX_CTRL_QUEUES;
2910 
2911 	tot_size += MAX_ETH_QSETS * MAX_TXQ_ENTRIES * MAX_TXQ_DESC_SIZE;
2912 	tot_size += MAX_ETH_QSETS * MAX_RSPQ_ENTRIES * MAX_RXQ_DESC_SIZE;
2913 	tot_size += MAX_ETH_QSETS * MAX_RX_BUFFERS * MAX_FL_DESC_SIZE;
2914 	tot_size += MAX_CTRL_QUEUES * MAX_CTRL_TXQ_ENTRIES *
2915 		    MAX_CTRL_TXQ_DESC_SIZE;
2916 
2917 	/* FW_EVTQ and INTRQ */
2918 	tot_entries += INGQ_EXTRAS;
2919 	tot_size += INGQ_EXTRAS * MAX_RSPQ_ENTRIES * MAX_RXQ_DESC_SIZE;
2920 
2921 	/* PTP_TXQ */
2922 	tot_entries += 1;
2923 	tot_size += MAX_TXQ_ENTRIES * MAX_TXQ_DESC_SIZE;
2924 
2925 	/* ULD TXQ, RXQ, and FLQ */
2926 	tot_entries += CXGB4_TX_MAX * MAX_OFLD_QSETS;
2927 	tot_entries += CXGB4_ULD_MAX * MAX_ULD_QSETS * 2;
2928 
2929 	tot_size += CXGB4_TX_MAX * MAX_OFLD_QSETS * MAX_TXQ_ENTRIES *
2930 		    MAX_TXQ_DESC_SIZE;
2931 	tot_size += CXGB4_ULD_MAX * MAX_ULD_QSETS * MAX_RSPQ_ENTRIES *
2932 		    MAX_RXQ_DESC_SIZE;
2933 	tot_size += CXGB4_ULD_MAX * MAX_ULD_QSETS * MAX_RX_BUFFERS *
2934 		    MAX_FL_DESC_SIZE;
2935 
2936 	/* ULD CIQ */
2937 	tot_entries += CXGB4_ULD_MAX * MAX_ULD_QSETS;
2938 	tot_size += CXGB4_ULD_MAX * MAX_ULD_QSETS * SGE_MAX_IQ_SIZE *
2939 		    MAX_RXQ_DESC_SIZE;
2940 
2941 	/* ETHOFLD TXQ, RXQ, and FLQ */
2942 	tot_entries += MAX_OFLD_QSETS * 3;
2943 	tot_size += MAX_OFLD_QSETS * MAX_TXQ_ENTRIES * MAX_TXQ_DESC_SIZE;
2944 
2945 	tot_size += sizeof(struct cudbg_ver_hdr) +
2946 		    sizeof(struct cudbg_qdesc_info) +
2947 		    sizeof(struct cudbg_qdesc_entry) * tot_entries;
2948 
2949 	if (num)
2950 		*num = tot_entries;
2951 
2952 	if (size)
2953 		*size = tot_size;
2954 }
2955 
2956 int cudbg_collect_qdesc(struct cudbg_init *pdbg_init,
2957 			struct cudbg_buffer *dbg_buff,
2958 			struct cudbg_error *cudbg_err)
2959 {
2960 	u32 num_queues = 0, tot_entries = 0, size = 0;
2961 	struct adapter *padap = pdbg_init->adap;
2962 	struct cudbg_buffer temp_buff = { 0 };
2963 	struct cudbg_qdesc_entry *qdesc_entry;
2964 	struct cudbg_qdesc_info *qdesc_info;
2965 	struct cudbg_ver_hdr *ver_hdr;
2966 	struct sge *s = &padap->sge;
2967 	u32 i, j, cur_off, tot_len;
2968 	u8 *data;
2969 	int rc;
2970 
2971 	cudbg_fill_qdesc_num_and_size(padap, &tot_entries, &size);
2972 	size = min_t(u32, size, CUDBG_DUMP_BUFF_SIZE);
2973 	tot_len = size;
2974 	data = kvzalloc(size, GFP_KERNEL);
2975 	if (!data)
2976 		return -ENOMEM;
2977 
2978 	ver_hdr = (struct cudbg_ver_hdr *)data;
2979 	ver_hdr->signature = CUDBG_ENTITY_SIGNATURE;
2980 	ver_hdr->revision = CUDBG_QDESC_REV;
2981 	ver_hdr->size = sizeof(struct cudbg_qdesc_info);
2982 	size -= sizeof(*ver_hdr);
2983 
2984 	qdesc_info = (struct cudbg_qdesc_info *)(data +
2985 						 sizeof(*ver_hdr));
2986 	size -= sizeof(*qdesc_info);
2987 	qdesc_entry = (struct cudbg_qdesc_entry *)qdesc_info->data;
2988 
2989 #define QDESC_GET(q, desc, type, label) do { \
2990 	if (size <= 0) { \
2991 		goto label; \
2992 	} \
2993 	if (desc) { \
2994 		cudbg_fill_qdesc_##q(q, type, qdesc_entry); \
2995 		size -= sizeof(*qdesc_entry) + qdesc_entry->data_size; \
2996 		num_queues++; \
2997 		qdesc_entry = cudbg_next_qdesc(qdesc_entry); \
2998 	} \
2999 } while (0)
3000 
3001 #define QDESC_GET_TXQ(q, type, label) do { \
3002 	struct sge_txq *txq = (struct sge_txq *)q; \
3003 	QDESC_GET(txq, txq->desc, type, label); \
3004 } while (0)
3005 
3006 #define QDESC_GET_RXQ(q, type, label) do { \
3007 	struct sge_rspq *rxq = (struct sge_rspq *)q; \
3008 	QDESC_GET(rxq, rxq->desc, type, label); \
3009 } while (0)
3010 
3011 #define QDESC_GET_FLQ(q, type, label) do { \
3012 	struct sge_fl *flq = (struct sge_fl *)q; \
3013 	QDESC_GET(flq, flq->desc, type, label); \
3014 } while (0)
3015 
3016 	/* NIC TXQ */
3017 	for (i = 0; i < s->ethqsets; i++)
3018 		QDESC_GET_TXQ(&s->ethtxq[i].q, CUDBG_QTYPE_NIC_TXQ, out);
3019 
3020 	/* NIC RXQ */
3021 	for (i = 0; i < s->ethqsets; i++)
3022 		QDESC_GET_RXQ(&s->ethrxq[i].rspq, CUDBG_QTYPE_NIC_RXQ, out);
3023 
3024 	/* NIC FLQ */
3025 	for (i = 0; i < s->ethqsets; i++)
3026 		QDESC_GET_FLQ(&s->ethrxq[i].fl, CUDBG_QTYPE_NIC_FLQ, out);
3027 
3028 	/* NIC CTRLQ */
3029 	for (i = 0; i < padap->params.nports; i++)
3030 		QDESC_GET_TXQ(&s->ctrlq[i].q, CUDBG_QTYPE_CTRLQ, out);
3031 
3032 	/* FW_EVTQ */
3033 	QDESC_GET_RXQ(&s->fw_evtq, CUDBG_QTYPE_FWEVTQ, out);
3034 
3035 	/* INTRQ */
3036 	QDESC_GET_RXQ(&s->intrq, CUDBG_QTYPE_INTRQ, out);
3037 
3038 	/* PTP_TXQ */
3039 	QDESC_GET_TXQ(&s->ptptxq.q, CUDBG_QTYPE_PTP_TXQ, out);
3040 
3041 	/* ULD Queues */
3042 	mutex_lock(&uld_mutex);
3043 
3044 	if (s->uld_txq_info) {
3045 		struct sge_uld_txq_info *utxq;
3046 
3047 		/* ULD TXQ */
3048 		for (j = 0; j < CXGB4_TX_MAX; j++) {
3049 			if (!s->uld_txq_info[j])
3050 				continue;
3051 
3052 			utxq = s->uld_txq_info[j];
3053 			for (i = 0; i < utxq->ntxq; i++)
3054 				QDESC_GET_TXQ(&utxq->uldtxq[i].q,
3055 					      cudbg_uld_txq_to_qtype(j),
3056 					      out_unlock);
3057 		}
3058 	}
3059 
3060 	if (s->uld_rxq_info) {
3061 		struct sge_uld_rxq_info *urxq;
3062 		u32 base;
3063 
3064 		/* ULD RXQ */
3065 		for (j = 0; j < CXGB4_ULD_MAX; j++) {
3066 			if (!s->uld_rxq_info[j])
3067 				continue;
3068 
3069 			urxq = s->uld_rxq_info[j];
3070 			for (i = 0; i < urxq->nrxq; i++)
3071 				QDESC_GET_RXQ(&urxq->uldrxq[i].rspq,
3072 					      cudbg_uld_rxq_to_qtype(j),
3073 					      out_unlock);
3074 		}
3075 
3076 		/* ULD FLQ */
3077 		for (j = 0; j < CXGB4_ULD_MAX; j++) {
3078 			if (!s->uld_rxq_info[j])
3079 				continue;
3080 
3081 			urxq = s->uld_rxq_info[j];
3082 			for (i = 0; i < urxq->nrxq; i++)
3083 				QDESC_GET_FLQ(&urxq->uldrxq[i].fl,
3084 					      cudbg_uld_flq_to_qtype(j),
3085 					      out_unlock);
3086 		}
3087 
3088 		/* ULD CIQ */
3089 		for (j = 0; j < CXGB4_ULD_MAX; j++) {
3090 			if (!s->uld_rxq_info[j])
3091 				continue;
3092 
3093 			urxq = s->uld_rxq_info[j];
3094 			base = urxq->nrxq;
3095 			for (i = 0; i < urxq->nciq; i++)
3096 				QDESC_GET_RXQ(&urxq->uldrxq[base + i].rspq,
3097 					      cudbg_uld_ciq_to_qtype(j),
3098 					      out_unlock);
3099 		}
3100 	}
3101 
3102 	/* ETHOFLD TXQ */
3103 	if (s->eohw_txq)
3104 		for (i = 0; i < s->eoqsets; i++)
3105 			QDESC_GET_TXQ(&s->eohw_txq[i].q,
3106 				      CUDBG_QTYPE_ETHOFLD_TXQ, out);
3107 
3108 	/* ETHOFLD RXQ and FLQ */
3109 	if (s->eohw_rxq) {
3110 		for (i = 0; i < s->eoqsets; i++)
3111 			QDESC_GET_RXQ(&s->eohw_rxq[i].rspq,
3112 				      CUDBG_QTYPE_ETHOFLD_RXQ, out);
3113 
3114 		for (i = 0; i < s->eoqsets; i++)
3115 			QDESC_GET_FLQ(&s->eohw_rxq[i].fl,
3116 				      CUDBG_QTYPE_ETHOFLD_FLQ, out);
3117 	}
3118 
3119 out_unlock:
3120 	mutex_unlock(&uld_mutex);
3121 
3122 out:
3123 	qdesc_info->qdesc_entry_size = sizeof(*qdesc_entry);
3124 	qdesc_info->num_queues = num_queues;
3125 	cur_off = 0;
3126 	while (tot_len) {
3127 		u32 chunk_size = min_t(u32, tot_len, CUDBG_CHUNK_SIZE);
3128 
3129 		rc = cudbg_get_buff(pdbg_init, dbg_buff, chunk_size,
3130 				    &temp_buff);
3131 		if (rc) {
3132 			cudbg_err->sys_warn = CUDBG_STATUS_PARTIAL_DATA;
3133 			goto out_free;
3134 		}
3135 
3136 		memcpy(temp_buff.data, data + cur_off, chunk_size);
3137 		tot_len -= chunk_size;
3138 		cur_off += chunk_size;
3139 		rc = cudbg_write_and_release_buff(pdbg_init, &temp_buff,
3140 						  dbg_buff);
3141 		if (rc) {
3142 			cudbg_put_buff(pdbg_init, &temp_buff);
3143 			cudbg_err->sys_warn = CUDBG_STATUS_PARTIAL_DATA;
3144 			goto out_free;
3145 		}
3146 	}
3147 
3148 out_free:
3149 	if (data)
3150 		kvfree(data);
3151 
3152 #undef QDESC_GET_FLQ
3153 #undef QDESC_GET_RXQ
3154 #undef QDESC_GET_TXQ
3155 #undef QDESC_GET
3156 
3157 	return rc;
3158 }
3159