xref: /linux/drivers/net/ethernet/chelsio/cxgb4/cudbg_lib.c (revision c532de5a67a70f8533d495f8f2aaa9a0491c3ad0)
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 #include "cxgb4_tc_mqprio.h"
18 
19 static const u32 t6_tp_pio_array[][IREG_NUM_ELEM] = {
20 	{0x7e40, 0x7e44, 0x020, 28}, /* t6_tp_pio_regs_20_to_3b */
21 	{0x7e40, 0x7e44, 0x040, 10}, /* t6_tp_pio_regs_40_to_49 */
22 	{0x7e40, 0x7e44, 0x050, 10}, /* t6_tp_pio_regs_50_to_59 */
23 	{0x7e40, 0x7e44, 0x060, 14}, /* t6_tp_pio_regs_60_to_6d */
24 	{0x7e40, 0x7e44, 0x06F, 1}, /* t6_tp_pio_regs_6f */
25 	{0x7e40, 0x7e44, 0x070, 6}, /* t6_tp_pio_regs_70_to_75 */
26 	{0x7e40, 0x7e44, 0x130, 18}, /* t6_tp_pio_regs_130_to_141 */
27 	{0x7e40, 0x7e44, 0x145, 19}, /* t6_tp_pio_regs_145_to_157 */
28 	{0x7e40, 0x7e44, 0x160, 1}, /* t6_tp_pio_regs_160 */
29 	{0x7e40, 0x7e44, 0x230, 25}, /* t6_tp_pio_regs_230_to_248 */
30 	{0x7e40, 0x7e44, 0x24a, 3}, /* t6_tp_pio_regs_24c */
31 	{0x7e40, 0x7e44, 0x8C0, 1} /* t6_tp_pio_regs_8c0 */
32 };
33 
34 static const u32 t5_tp_pio_array[][IREG_NUM_ELEM] = {
35 	{0x7e40, 0x7e44, 0x020, 28}, /* t5_tp_pio_regs_20_to_3b */
36 	{0x7e40, 0x7e44, 0x040, 19}, /* t5_tp_pio_regs_40_to_52 */
37 	{0x7e40, 0x7e44, 0x054, 2}, /* t5_tp_pio_regs_54_to_55 */
38 	{0x7e40, 0x7e44, 0x060, 13}, /* t5_tp_pio_regs_60_to_6c */
39 	{0x7e40, 0x7e44, 0x06F, 1}, /* t5_tp_pio_regs_6f */
40 	{0x7e40, 0x7e44, 0x120, 4}, /* t5_tp_pio_regs_120_to_123 */
41 	{0x7e40, 0x7e44, 0x12b, 2}, /* t5_tp_pio_regs_12b_to_12c */
42 	{0x7e40, 0x7e44, 0x12f, 21}, /* t5_tp_pio_regs_12f_to_143 */
43 	{0x7e40, 0x7e44, 0x145, 19}, /* t5_tp_pio_regs_145_to_157 */
44 	{0x7e40, 0x7e44, 0x230, 25}, /* t5_tp_pio_regs_230_to_248 */
45 	{0x7e40, 0x7e44, 0x8C0, 1} /* t5_tp_pio_regs_8c0 */
46 };
47 
48 static const u32 t6_tp_tm_pio_array[][IREG_NUM_ELEM] = {
49 	{0x7e18, 0x7e1c, 0x0, 12}
50 };
51 
52 static const u32 t5_tp_tm_pio_array[][IREG_NUM_ELEM] = {
53 	{0x7e18, 0x7e1c, 0x0, 12}
54 };
55 
56 static const u32 t6_tp_mib_index_array[6][IREG_NUM_ELEM] = {
57 	{0x7e50, 0x7e54, 0x0, 13},
58 	{0x7e50, 0x7e54, 0x10, 6},
59 	{0x7e50, 0x7e54, 0x18, 21},
60 	{0x7e50, 0x7e54, 0x30, 32},
61 	{0x7e50, 0x7e54, 0x50, 22},
62 	{0x7e50, 0x7e54, 0x68, 12}
63 };
64 
65 static const u32 t5_tp_mib_index_array[9][IREG_NUM_ELEM] = {
66 	{0x7e50, 0x7e54, 0x0, 13},
67 	{0x7e50, 0x7e54, 0x10, 6},
68 	{0x7e50, 0x7e54, 0x18, 8},
69 	{0x7e50, 0x7e54, 0x20, 13},
70 	{0x7e50, 0x7e54, 0x30, 16},
71 	{0x7e50, 0x7e54, 0x40, 16},
72 	{0x7e50, 0x7e54, 0x50, 16},
73 	{0x7e50, 0x7e54, 0x60, 6},
74 	{0x7e50, 0x7e54, 0x68, 4}
75 };
76 
77 static const u32 t5_sge_dbg_index_array[2][IREG_NUM_ELEM] = {
78 	{0x10cc, 0x10d0, 0x0, 16},
79 	{0x10cc, 0x10d4, 0x0, 16},
80 };
81 
82 static const u32 t6_sge_qbase_index_array[] = {
83 	/* 1 addr reg SGE_QBASE_INDEX and 4 data reg SGE_QBASE_MAP[0-3] */
84 	0x1250, 0x1240, 0x1244, 0x1248, 0x124c,
85 };
86 
87 static const u32 t5_pcie_pdbg_array[][IREG_NUM_ELEM] = {
88 	{0x5a04, 0x5a0c, 0x00, 0x20}, /* t5_pcie_pdbg_regs_00_to_20 */
89 	{0x5a04, 0x5a0c, 0x21, 0x20}, /* t5_pcie_pdbg_regs_21_to_40 */
90 	{0x5a04, 0x5a0c, 0x41, 0x10}, /* t5_pcie_pdbg_regs_41_to_50 */
91 };
92 
93 static const u32 t5_pcie_cdbg_array[][IREG_NUM_ELEM] = {
94 	{0x5a10, 0x5a18, 0x00, 0x20}, /* t5_pcie_cdbg_regs_00_to_20 */
95 	{0x5a10, 0x5a18, 0x21, 0x18}, /* t5_pcie_cdbg_regs_21_to_37 */
96 };
97 
98 static const u32 t5_pm_rx_array[][IREG_NUM_ELEM] = {
99 	{0x8FD0, 0x8FD4, 0x10000, 0x20}, /* t5_pm_rx_regs_10000_to_10020 */
100 	{0x8FD0, 0x8FD4, 0x10021, 0x0D}, /* t5_pm_rx_regs_10021_to_1002c */
101 };
102 
103 static const u32 t5_pm_tx_array[][IREG_NUM_ELEM] = {
104 	{0x8FF0, 0x8FF4, 0x10000, 0x20}, /* t5_pm_tx_regs_10000_to_10020 */
105 	{0x8FF0, 0x8FF4, 0x10021, 0x1D}, /* t5_pm_tx_regs_10021_to_1003c */
106 };
107 
108 static const u32 t5_pcie_config_array[][2] = {
109 	{0x0, 0x34},
110 	{0x3c, 0x40},
111 	{0x50, 0x64},
112 	{0x70, 0x80},
113 	{0x94, 0xa0},
114 	{0xb0, 0xb8},
115 	{0xd0, 0xd4},
116 	{0x100, 0x128},
117 	{0x140, 0x148},
118 	{0x150, 0x164},
119 	{0x170, 0x178},
120 	{0x180, 0x194},
121 	{0x1a0, 0x1b8},
122 	{0x1c0, 0x208},
123 };
124 
125 static const u32 t6_ma_ireg_array[][IREG_NUM_ELEM] = {
126 	{0x78f8, 0x78fc, 0xa000, 23}, /* t6_ma_regs_a000_to_a016 */
127 	{0x78f8, 0x78fc, 0xa400, 30}, /* t6_ma_regs_a400_to_a41e */
128 	{0x78f8, 0x78fc, 0xa800, 20} /* t6_ma_regs_a800_to_a813 */
129 };
130 
131 static const u32 t6_ma_ireg_array2[][IREG_NUM_ELEM] = {
132 	{0x78f8, 0x78fc, 0xe400, 17}, /* t6_ma_regs_e400_to_e600 */
133 	{0x78f8, 0x78fc, 0xe640, 13} /* t6_ma_regs_e640_to_e7c0 */
134 };
135 
136 static const u32 t6_up_cim_reg_array[][IREG_NUM_ELEM + 1] = {
137 	{0x7b50, 0x7b54, 0x2000, 0x20, 0}, /* up_cim_2000_to_207c */
138 	{0x7b50, 0x7b54, 0x2080, 0x1d, 0}, /* up_cim_2080_to_20fc */
139 	{0x7b50, 0x7b54, 0x00, 0x20, 0}, /* up_cim_00_to_7c */
140 	{0x7b50, 0x7b54, 0x80, 0x20, 0}, /* up_cim_80_to_fc */
141 	{0x7b50, 0x7b54, 0x100, 0x11, 0}, /* up_cim_100_to_14c */
142 	{0x7b50, 0x7b54, 0x200, 0x10, 0}, /* up_cim_200_to_23c */
143 	{0x7b50, 0x7b54, 0x240, 0x2, 0}, /* up_cim_240_to_244 */
144 	{0x7b50, 0x7b54, 0x250, 0x2, 0}, /* up_cim_250_to_254 */
145 	{0x7b50, 0x7b54, 0x260, 0x2, 0}, /* up_cim_260_to_264 */
146 	{0x7b50, 0x7b54, 0x270, 0x2, 0}, /* up_cim_270_to_274 */
147 	{0x7b50, 0x7b54, 0x280, 0x20, 0}, /* up_cim_280_to_2fc */
148 	{0x7b50, 0x7b54, 0x300, 0x20, 0}, /* up_cim_300_to_37c */
149 	{0x7b50, 0x7b54, 0x380, 0x14, 0}, /* up_cim_380_to_3cc */
150 	{0x7b50, 0x7b54, 0x4900, 0x4, 0x4}, /* up_cim_4900_to_4c60 */
151 	{0x7b50, 0x7b54, 0x4904, 0x4, 0x4}, /* up_cim_4904_to_4c64 */
152 	{0x7b50, 0x7b54, 0x4908, 0x4, 0x4}, /* up_cim_4908_to_4c68 */
153 	{0x7b50, 0x7b54, 0x4910, 0x4, 0x4}, /* up_cim_4910_to_4c70 */
154 	{0x7b50, 0x7b54, 0x4914, 0x4, 0x4}, /* up_cim_4914_to_4c74 */
155 	{0x7b50, 0x7b54, 0x4920, 0x10, 0x10}, /* up_cim_4920_to_4a10 */
156 	{0x7b50, 0x7b54, 0x4924, 0x10, 0x10}, /* up_cim_4924_to_4a14 */
157 	{0x7b50, 0x7b54, 0x4928, 0x10, 0x10}, /* up_cim_4928_to_4a18 */
158 	{0x7b50, 0x7b54, 0x492c, 0x10, 0x10}, /* up_cim_492c_to_4a1c */
159 };
160 
161 static const u32 t5_up_cim_reg_array[][IREG_NUM_ELEM + 1] = {
162 	{0x7b50, 0x7b54, 0x2000, 0x20, 0}, /* up_cim_2000_to_207c */
163 	{0x7b50, 0x7b54, 0x2080, 0x19, 0}, /* up_cim_2080_to_20ec */
164 	{0x7b50, 0x7b54, 0x00, 0x20, 0}, /* up_cim_00_to_7c */
165 	{0x7b50, 0x7b54, 0x80, 0x20, 0}, /* up_cim_80_to_fc */
166 	{0x7b50, 0x7b54, 0x100, 0x11, 0}, /* up_cim_100_to_14c */
167 	{0x7b50, 0x7b54, 0x200, 0x10, 0}, /* up_cim_200_to_23c */
168 	{0x7b50, 0x7b54, 0x240, 0x2, 0}, /* up_cim_240_to_244 */
169 	{0x7b50, 0x7b54, 0x250, 0x2, 0}, /* up_cim_250_to_254 */
170 	{0x7b50, 0x7b54, 0x260, 0x2, 0}, /* up_cim_260_to_264 */
171 	{0x7b50, 0x7b54, 0x270, 0x2, 0}, /* up_cim_270_to_274 */
172 	{0x7b50, 0x7b54, 0x280, 0x20, 0}, /* up_cim_280_to_2fc */
173 	{0x7b50, 0x7b54, 0x300, 0x20, 0}, /* up_cim_300_to_37c */
174 	{0x7b50, 0x7b54, 0x380, 0x14, 0}, /* up_cim_380_to_3cc */
175 };
176 
177 static const u32 t6_hma_ireg_array[][IREG_NUM_ELEM] = {
178 	{0x51320, 0x51324, 0xa000, 32} /* t6_hma_regs_a000_to_a01f */
179 };
180 
181 u32 cudbg_get_entity_length(struct adapter *adap, u32 entity)
182 {
183 	struct cudbg_tcam tcam_region = { 0 };
184 	u32 value, n = 0, len = 0;
185 
186 	switch (entity) {
187 	case CUDBG_REG_DUMP:
188 		switch (CHELSIO_CHIP_VERSION(adap->params.chip)) {
189 		case CHELSIO_T4:
190 			len = T4_REGMAP_SIZE;
191 			break;
192 		case CHELSIO_T5:
193 		case CHELSIO_T6:
194 			len = T5_REGMAP_SIZE;
195 			break;
196 		default:
197 			break;
198 		}
199 		break;
200 	case CUDBG_DEV_LOG:
201 		len = adap->params.devlog.size;
202 		break;
203 	case CUDBG_CIM_LA:
204 		if (is_t6(adap->params.chip)) {
205 			len = adap->params.cim_la_size / 10 + 1;
206 			len *= 10 * sizeof(u32);
207 		} else {
208 			len = adap->params.cim_la_size / 8;
209 			len *= 8 * sizeof(u32);
210 		}
211 		len += sizeof(u32); /* for reading CIM LA configuration */
212 		break;
213 	case CUDBG_CIM_MA_LA:
214 		len = 2 * CIM_MALA_SIZE * 5 * sizeof(u32);
215 		break;
216 	case CUDBG_CIM_QCFG:
217 		len = sizeof(struct cudbg_cim_qcfg);
218 		break;
219 	case CUDBG_CIM_IBQ_TP0:
220 	case CUDBG_CIM_IBQ_TP1:
221 	case CUDBG_CIM_IBQ_ULP:
222 	case CUDBG_CIM_IBQ_SGE0:
223 	case CUDBG_CIM_IBQ_SGE1:
224 	case CUDBG_CIM_IBQ_NCSI:
225 		len = CIM_IBQ_SIZE * 4 * sizeof(u32);
226 		break;
227 	case CUDBG_CIM_OBQ_ULP0:
228 		len = cudbg_cim_obq_size(adap, 0);
229 		break;
230 	case CUDBG_CIM_OBQ_ULP1:
231 		len = cudbg_cim_obq_size(adap, 1);
232 		break;
233 	case CUDBG_CIM_OBQ_ULP2:
234 		len = cudbg_cim_obq_size(adap, 2);
235 		break;
236 	case CUDBG_CIM_OBQ_ULP3:
237 		len = cudbg_cim_obq_size(adap, 3);
238 		break;
239 	case CUDBG_CIM_OBQ_SGE:
240 		len = cudbg_cim_obq_size(adap, 4);
241 		break;
242 	case CUDBG_CIM_OBQ_NCSI:
243 		len = cudbg_cim_obq_size(adap, 5);
244 		break;
245 	case CUDBG_CIM_OBQ_RXQ0:
246 		len = cudbg_cim_obq_size(adap, 6);
247 		break;
248 	case CUDBG_CIM_OBQ_RXQ1:
249 		len = cudbg_cim_obq_size(adap, 7);
250 		break;
251 	case CUDBG_EDC0:
252 		value = t4_read_reg(adap, MA_TARGET_MEM_ENABLE_A);
253 		if (value & EDRAM0_ENABLE_F) {
254 			value = t4_read_reg(adap, MA_EDRAM0_BAR_A);
255 			len = EDRAM0_SIZE_G(value);
256 		}
257 		len = cudbg_mbytes_to_bytes(len);
258 		break;
259 	case CUDBG_EDC1:
260 		value = t4_read_reg(adap, MA_TARGET_MEM_ENABLE_A);
261 		if (value & EDRAM1_ENABLE_F) {
262 			value = t4_read_reg(adap, MA_EDRAM1_BAR_A);
263 			len = EDRAM1_SIZE_G(value);
264 		}
265 		len = cudbg_mbytes_to_bytes(len);
266 		break;
267 	case CUDBG_MC0:
268 		value = t4_read_reg(adap, MA_TARGET_MEM_ENABLE_A);
269 		if (value & EXT_MEM0_ENABLE_F) {
270 			value = t4_read_reg(adap, MA_EXT_MEMORY0_BAR_A);
271 			len = EXT_MEM0_SIZE_G(value);
272 		}
273 		len = cudbg_mbytes_to_bytes(len);
274 		break;
275 	case CUDBG_MC1:
276 		value = t4_read_reg(adap, MA_TARGET_MEM_ENABLE_A);
277 		if (value & EXT_MEM1_ENABLE_F) {
278 			value = t4_read_reg(adap, MA_EXT_MEMORY1_BAR_A);
279 			len = EXT_MEM1_SIZE_G(value);
280 		}
281 		len = cudbg_mbytes_to_bytes(len);
282 		break;
283 	case CUDBG_RSS:
284 		len = t4_chip_rss_size(adap) * sizeof(u16);
285 		break;
286 	case CUDBG_RSS_VF_CONF:
287 		len = adap->params.arch.vfcount *
288 		      sizeof(struct cudbg_rss_vf_conf);
289 		break;
290 	case CUDBG_PATH_MTU:
291 		len = NMTUS * sizeof(u16);
292 		break;
293 	case CUDBG_PM_STATS:
294 		len = sizeof(struct cudbg_pm_stats);
295 		break;
296 	case CUDBG_HW_SCHED:
297 		len = sizeof(struct cudbg_hw_sched);
298 		break;
299 	case CUDBG_TP_INDIRECT:
300 		switch (CHELSIO_CHIP_VERSION(adap->params.chip)) {
301 		case CHELSIO_T5:
302 			n = sizeof(t5_tp_pio_array) +
303 			    sizeof(t5_tp_tm_pio_array) +
304 			    sizeof(t5_tp_mib_index_array);
305 			break;
306 		case CHELSIO_T6:
307 			n = sizeof(t6_tp_pio_array) +
308 			    sizeof(t6_tp_tm_pio_array) +
309 			    sizeof(t6_tp_mib_index_array);
310 			break;
311 		default:
312 			break;
313 		}
314 		n = n / (IREG_NUM_ELEM * sizeof(u32));
315 		len = sizeof(struct ireg_buf) * n;
316 		break;
317 	case CUDBG_SGE_INDIRECT:
318 		len = sizeof(struct ireg_buf) * 2 +
319 		      sizeof(struct sge_qbase_reg_field);
320 		break;
321 	case CUDBG_ULPRX_LA:
322 		len = sizeof(struct cudbg_ulprx_la);
323 		break;
324 	case CUDBG_TP_LA:
325 		len = sizeof(struct cudbg_tp_la) + TPLA_SIZE * sizeof(u64);
326 		break;
327 	case CUDBG_MEMINFO:
328 		len = sizeof(struct cudbg_ver_hdr) +
329 		      sizeof(struct cudbg_meminfo);
330 		break;
331 	case CUDBG_CIM_PIF_LA:
332 		len = sizeof(struct cudbg_cim_pif_la);
333 		len += 2 * CIM_PIFLA_SIZE * 6 * sizeof(u32);
334 		break;
335 	case CUDBG_CLK:
336 		len = sizeof(struct cudbg_clk_info);
337 		break;
338 	case CUDBG_PCIE_INDIRECT:
339 		n = sizeof(t5_pcie_pdbg_array) / (IREG_NUM_ELEM * sizeof(u32));
340 		len = sizeof(struct ireg_buf) * n * 2;
341 		break;
342 	case CUDBG_PM_INDIRECT:
343 		n = sizeof(t5_pm_rx_array) / (IREG_NUM_ELEM * sizeof(u32));
344 		len = sizeof(struct ireg_buf) * n * 2;
345 		break;
346 	case CUDBG_TID_INFO:
347 		len = sizeof(struct cudbg_tid_info_region_rev1);
348 		break;
349 	case CUDBG_PCIE_CONFIG:
350 		len = sizeof(u32) * CUDBG_NUM_PCIE_CONFIG_REGS;
351 		break;
352 	case CUDBG_DUMP_CONTEXT:
353 		len = cudbg_dump_context_size(adap);
354 		break;
355 	case CUDBG_MPS_TCAM:
356 		len = sizeof(struct cudbg_mps_tcam) *
357 		      adap->params.arch.mps_tcam_size;
358 		break;
359 	case CUDBG_VPD_DATA:
360 		len = sizeof(struct cudbg_vpd_data);
361 		break;
362 	case CUDBG_LE_TCAM:
363 		cudbg_fill_le_tcam_info(adap, &tcam_region);
364 		len = sizeof(struct cudbg_tcam) +
365 		      sizeof(struct cudbg_tid_data) * tcam_region.max_tid;
366 		break;
367 	case CUDBG_CCTRL:
368 		len = sizeof(u16) * NMTUS * NCCTRL_WIN;
369 		break;
370 	case CUDBG_MA_INDIRECT:
371 		if (CHELSIO_CHIP_VERSION(adap->params.chip) > CHELSIO_T5) {
372 			n = sizeof(t6_ma_ireg_array) /
373 			    (IREG_NUM_ELEM * sizeof(u32));
374 			len = sizeof(struct ireg_buf) * n * 2;
375 		}
376 		break;
377 	case CUDBG_ULPTX_LA:
378 		len = sizeof(struct cudbg_ver_hdr) +
379 		      sizeof(struct cudbg_ulptx_la);
380 		break;
381 	case CUDBG_UP_CIM_INDIRECT:
382 		n = 0;
383 		if (is_t5(adap->params.chip))
384 			n = sizeof(t5_up_cim_reg_array) /
385 			    ((IREG_NUM_ELEM + 1) * sizeof(u32));
386 		else if (is_t6(adap->params.chip))
387 			n = sizeof(t6_up_cim_reg_array) /
388 			    ((IREG_NUM_ELEM + 1) * sizeof(u32));
389 		len = sizeof(struct ireg_buf) * n;
390 		break;
391 	case CUDBG_PBT_TABLE:
392 		len = sizeof(struct cudbg_pbt_tables);
393 		break;
394 	case CUDBG_MBOX_LOG:
395 		len = sizeof(struct cudbg_mbox_log) * adap->mbox_log->size;
396 		break;
397 	case CUDBG_HMA_INDIRECT:
398 		if (CHELSIO_CHIP_VERSION(adap->params.chip) > CHELSIO_T5) {
399 			n = sizeof(t6_hma_ireg_array) /
400 			    (IREG_NUM_ELEM * sizeof(u32));
401 			len = sizeof(struct ireg_buf) * n;
402 		}
403 		break;
404 	case CUDBG_HMA:
405 		value = t4_read_reg(adap, MA_TARGET_MEM_ENABLE_A);
406 		if (value & HMA_MUX_F) {
407 			/* In T6, there's no MC1.  So, HMA shares MC1
408 			 * address space.
409 			 */
410 			value = t4_read_reg(adap, MA_EXT_MEMORY1_BAR_A);
411 			len = EXT_MEM1_SIZE_G(value);
412 		}
413 		len = cudbg_mbytes_to_bytes(len);
414 		break;
415 	case CUDBG_QDESC:
416 		cudbg_fill_qdesc_num_and_size(adap, NULL, &len);
417 		break;
418 	default:
419 		break;
420 	}
421 
422 	return len;
423 }
424 
425 static int cudbg_do_compression(struct cudbg_init *pdbg_init,
426 				struct cudbg_buffer *pin_buff,
427 				struct cudbg_buffer *dbg_buff)
428 {
429 	struct cudbg_buffer temp_in_buff = { 0 };
430 	int bytes_left, bytes_read, bytes;
431 	u32 offset = dbg_buff->offset;
432 	int rc;
433 
434 	temp_in_buff.offset = pin_buff->offset;
435 	temp_in_buff.data = pin_buff->data;
436 	temp_in_buff.size = pin_buff->size;
437 
438 	bytes_left = pin_buff->size;
439 	bytes_read = 0;
440 	while (bytes_left > 0) {
441 		/* Do compression in smaller chunks */
442 		bytes = min_t(unsigned long, bytes_left,
443 			      (unsigned long)CUDBG_CHUNK_SIZE);
444 		temp_in_buff.data = (char *)pin_buff->data + bytes_read;
445 		temp_in_buff.size = bytes;
446 		rc = cudbg_compress_buff(pdbg_init, &temp_in_buff, dbg_buff);
447 		if (rc)
448 			return rc;
449 		bytes_left -= bytes;
450 		bytes_read += bytes;
451 	}
452 
453 	pin_buff->size = dbg_buff->offset - offset;
454 	return 0;
455 }
456 
457 static int cudbg_write_and_release_buff(struct cudbg_init *pdbg_init,
458 					struct cudbg_buffer *pin_buff,
459 					struct cudbg_buffer *dbg_buff)
460 {
461 	int rc = 0;
462 
463 	if (pdbg_init->compress_type == CUDBG_COMPRESSION_NONE) {
464 		cudbg_update_buff(pin_buff, dbg_buff);
465 	} else {
466 		rc = cudbg_do_compression(pdbg_init, pin_buff, dbg_buff);
467 		if (rc)
468 			goto out;
469 	}
470 
471 out:
472 	cudbg_put_buff(pdbg_init, pin_buff);
473 	return rc;
474 }
475 
476 static int is_fw_attached(struct cudbg_init *pdbg_init)
477 {
478 	struct adapter *padap = pdbg_init->adap;
479 
480 	if (!(padap->flags & CXGB4_FW_OK) || padap->use_bd)
481 		return 0;
482 
483 	return 1;
484 }
485 
486 /* This function will add additional padding bytes into debug_buffer to make it
487  * 4 byte aligned.
488  */
489 void cudbg_align_debug_buffer(struct cudbg_buffer *dbg_buff,
490 			      struct cudbg_entity_hdr *entity_hdr)
491 {
492 	u8 zero_buf[4] = {0};
493 	u8 padding, remain;
494 
495 	remain = (dbg_buff->offset - entity_hdr->start_offset) % 4;
496 	padding = 4 - remain;
497 	if (remain) {
498 		memcpy(((u8 *)dbg_buff->data) + dbg_buff->offset, &zero_buf,
499 		       padding);
500 		dbg_buff->offset += padding;
501 		entity_hdr->num_pad = padding;
502 	}
503 	entity_hdr->size = dbg_buff->offset - entity_hdr->start_offset;
504 }
505 
506 struct cudbg_entity_hdr *cudbg_get_entity_hdr(void *outbuf, int i)
507 {
508 	struct cudbg_hdr *cudbg_hdr = (struct cudbg_hdr *)outbuf;
509 
510 	return (struct cudbg_entity_hdr *)
511 	       ((char *)outbuf + cudbg_hdr->hdr_len +
512 		(sizeof(struct cudbg_entity_hdr) * (i - 1)));
513 }
514 
515 static int cudbg_read_vpd_reg(struct adapter *padap, u32 addr, u32 len,
516 			      void *dest)
517 {
518 	int vaddr, rc;
519 
520 	vaddr = t4_eeprom_ptov(addr, padap->pf, EEPROMPFSIZE);
521 	if (vaddr < 0)
522 		return vaddr;
523 
524 	rc = pci_read_vpd(padap->pdev, vaddr, len, dest);
525 	if (rc < 0)
526 		return rc;
527 
528 	return 0;
529 }
530 
531 static int cudbg_mem_desc_cmp(const void *a, const void *b)
532 {
533 	return ((const struct cudbg_mem_desc *)a)->base -
534 	       ((const struct cudbg_mem_desc *)b)->base;
535 }
536 
537 int cudbg_fill_meminfo(struct adapter *padap,
538 		       struct cudbg_meminfo *meminfo_buff)
539 {
540 	struct cudbg_mem_desc *md;
541 	u32 lo, hi, used, alloc;
542 	int n, i;
543 
544 	memset(meminfo_buff->avail, 0,
545 	       ARRAY_SIZE(meminfo_buff->avail) *
546 	       sizeof(struct cudbg_mem_desc));
547 	memset(meminfo_buff->mem, 0,
548 	       (ARRAY_SIZE(cudbg_region) + 3) * sizeof(struct cudbg_mem_desc));
549 	md  = meminfo_buff->mem;
550 
551 	for (i = 0; i < ARRAY_SIZE(meminfo_buff->mem); i++) {
552 		meminfo_buff->mem[i].limit = 0;
553 		meminfo_buff->mem[i].idx = i;
554 	}
555 
556 	/* Find and sort the populated memory ranges */
557 	i = 0;
558 	lo = t4_read_reg(padap, MA_TARGET_MEM_ENABLE_A);
559 	if (lo & EDRAM0_ENABLE_F) {
560 		hi = t4_read_reg(padap, MA_EDRAM0_BAR_A);
561 		meminfo_buff->avail[i].base =
562 			cudbg_mbytes_to_bytes(EDRAM0_BASE_G(hi));
563 		meminfo_buff->avail[i].limit =
564 			meminfo_buff->avail[i].base +
565 			cudbg_mbytes_to_bytes(EDRAM0_SIZE_G(hi));
566 		meminfo_buff->avail[i].idx = 0;
567 		i++;
568 	}
569 
570 	if (lo & EDRAM1_ENABLE_F) {
571 		hi =  t4_read_reg(padap, MA_EDRAM1_BAR_A);
572 		meminfo_buff->avail[i].base =
573 			cudbg_mbytes_to_bytes(EDRAM1_BASE_G(hi));
574 		meminfo_buff->avail[i].limit =
575 			meminfo_buff->avail[i].base +
576 			cudbg_mbytes_to_bytes(EDRAM1_SIZE_G(hi));
577 		meminfo_buff->avail[i].idx = 1;
578 		i++;
579 	}
580 
581 	if (is_t5(padap->params.chip)) {
582 		if (lo & EXT_MEM0_ENABLE_F) {
583 			hi = t4_read_reg(padap, MA_EXT_MEMORY0_BAR_A);
584 			meminfo_buff->avail[i].base =
585 				cudbg_mbytes_to_bytes(EXT_MEM_BASE_G(hi));
586 			meminfo_buff->avail[i].limit =
587 				meminfo_buff->avail[i].base +
588 				cudbg_mbytes_to_bytes(EXT_MEM_SIZE_G(hi));
589 			meminfo_buff->avail[i].idx = 3;
590 			i++;
591 		}
592 
593 		if (lo & EXT_MEM1_ENABLE_F) {
594 			hi = t4_read_reg(padap, MA_EXT_MEMORY1_BAR_A);
595 			meminfo_buff->avail[i].base =
596 				cudbg_mbytes_to_bytes(EXT_MEM1_BASE_G(hi));
597 			meminfo_buff->avail[i].limit =
598 				meminfo_buff->avail[i].base +
599 				cudbg_mbytes_to_bytes(EXT_MEM1_SIZE_G(hi));
600 			meminfo_buff->avail[i].idx = 4;
601 			i++;
602 		}
603 	} else {
604 		if (lo & EXT_MEM_ENABLE_F) {
605 			hi = t4_read_reg(padap, MA_EXT_MEMORY_BAR_A);
606 			meminfo_buff->avail[i].base =
607 				cudbg_mbytes_to_bytes(EXT_MEM_BASE_G(hi));
608 			meminfo_buff->avail[i].limit =
609 				meminfo_buff->avail[i].base +
610 				cudbg_mbytes_to_bytes(EXT_MEM_SIZE_G(hi));
611 			meminfo_buff->avail[i].idx = 2;
612 			i++;
613 		}
614 
615 		if (lo & HMA_MUX_F) {
616 			hi = t4_read_reg(padap, MA_EXT_MEMORY1_BAR_A);
617 			meminfo_buff->avail[i].base =
618 				cudbg_mbytes_to_bytes(EXT_MEM1_BASE_G(hi));
619 			meminfo_buff->avail[i].limit =
620 				meminfo_buff->avail[i].base +
621 				cudbg_mbytes_to_bytes(EXT_MEM1_SIZE_G(hi));
622 			meminfo_buff->avail[i].idx = 5;
623 			i++;
624 		}
625 	}
626 
627 	if (!i) /* no memory available */
628 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
629 
630 	meminfo_buff->avail_c = i;
631 	sort(meminfo_buff->avail, i, sizeof(struct cudbg_mem_desc),
632 	     cudbg_mem_desc_cmp, NULL);
633 	(md++)->base = t4_read_reg(padap, SGE_DBQ_CTXT_BADDR_A);
634 	(md++)->base = t4_read_reg(padap, SGE_IMSG_CTXT_BADDR_A);
635 	(md++)->base = t4_read_reg(padap, SGE_FLM_CACHE_BADDR_A);
636 	(md++)->base = t4_read_reg(padap, TP_CMM_TCB_BASE_A);
637 	(md++)->base = t4_read_reg(padap, TP_CMM_MM_BASE_A);
638 	(md++)->base = t4_read_reg(padap, TP_CMM_TIMER_BASE_A);
639 	(md++)->base = t4_read_reg(padap, TP_CMM_MM_RX_FLST_BASE_A);
640 	(md++)->base = t4_read_reg(padap, TP_CMM_MM_TX_FLST_BASE_A);
641 	(md++)->base = t4_read_reg(padap, TP_CMM_MM_PS_FLST_BASE_A);
642 
643 	/* the next few have explicit upper bounds */
644 	md->base = t4_read_reg(padap, TP_PMM_TX_BASE_A);
645 	md->limit = md->base - 1 +
646 		    t4_read_reg(padap, TP_PMM_TX_PAGE_SIZE_A) *
647 		    PMTXMAXPAGE_G(t4_read_reg(padap, TP_PMM_TX_MAX_PAGE_A));
648 	md++;
649 
650 	md->base = t4_read_reg(padap, TP_PMM_RX_BASE_A);
651 	md->limit = md->base - 1 +
652 		    t4_read_reg(padap, TP_PMM_RX_PAGE_SIZE_A) *
653 		    PMRXMAXPAGE_G(t4_read_reg(padap, TP_PMM_RX_MAX_PAGE_A));
654 	md++;
655 
656 	if (t4_read_reg(padap, LE_DB_CONFIG_A) & HASHEN_F) {
657 		if (CHELSIO_CHIP_VERSION(padap->params.chip) <= CHELSIO_T5) {
658 			hi = t4_read_reg(padap, LE_DB_TID_HASHBASE_A) / 4;
659 			md->base = t4_read_reg(padap, LE_DB_HASH_TID_BASE_A);
660 		} else {
661 			hi = t4_read_reg(padap, LE_DB_HASH_TID_BASE_A);
662 			md->base = t4_read_reg(padap,
663 					       LE_DB_HASH_TBL_BASE_ADDR_A);
664 		}
665 		md->limit = 0;
666 	} else {
667 		md->base = 0;
668 		md->idx = ARRAY_SIZE(cudbg_region);  /* hide it */
669 	}
670 	md++;
671 
672 #define ulp_region(reg) do { \
673 	md->base = t4_read_reg(padap, ULP_ ## reg ## _LLIMIT_A);\
674 	(md++)->limit = t4_read_reg(padap, ULP_ ## reg ## _ULIMIT_A);\
675 } while (0)
676 
677 	ulp_region(RX_ISCSI);
678 	ulp_region(RX_TDDP);
679 	ulp_region(TX_TPT);
680 	ulp_region(RX_STAG);
681 	ulp_region(RX_RQ);
682 	ulp_region(RX_RQUDP);
683 	ulp_region(RX_PBL);
684 	ulp_region(TX_PBL);
685 #undef ulp_region
686 	md->base = 0;
687 	md->idx = ARRAY_SIZE(cudbg_region);
688 	if (!is_t4(padap->params.chip)) {
689 		u32 fifo_size = t4_read_reg(padap, SGE_DBVFIFO_SIZE_A);
690 		u32 sge_ctrl = t4_read_reg(padap, SGE_CONTROL2_A);
691 		u32 size = 0;
692 
693 		if (is_t5(padap->params.chip)) {
694 			if (sge_ctrl & VFIFO_ENABLE_F)
695 				size = DBVFIFO_SIZE_G(fifo_size);
696 		} else {
697 			size = T6_DBVFIFO_SIZE_G(fifo_size);
698 		}
699 
700 		if (size) {
701 			md->base = BASEADDR_G(t4_read_reg(padap,
702 							  SGE_DBVFIFO_BADDR_A));
703 			md->limit = md->base + (size << 2) - 1;
704 		}
705 	}
706 
707 	md++;
708 
709 	md->base = t4_read_reg(padap, ULP_RX_CTX_BASE_A);
710 	md->limit = 0;
711 	md++;
712 	md->base = t4_read_reg(padap, ULP_TX_ERR_TABLE_BASE_A);
713 	md->limit = 0;
714 	md++;
715 
716 	md->base = padap->vres.ocq.start;
717 	if (padap->vres.ocq.size)
718 		md->limit = md->base + padap->vres.ocq.size - 1;
719 	else
720 		md->idx = ARRAY_SIZE(cudbg_region);  /* hide it */
721 	md++;
722 
723 	/* add any address-space holes, there can be up to 3 */
724 	for (n = 0; n < i - 1; n++)
725 		if (meminfo_buff->avail[n].limit <
726 		    meminfo_buff->avail[n + 1].base)
727 			(md++)->base = meminfo_buff->avail[n].limit;
728 
729 	if (meminfo_buff->avail[n].limit)
730 		(md++)->base = meminfo_buff->avail[n].limit;
731 
732 	n = md - meminfo_buff->mem;
733 	meminfo_buff->mem_c = n;
734 
735 	sort(meminfo_buff->mem, n, sizeof(struct cudbg_mem_desc),
736 	     cudbg_mem_desc_cmp, NULL);
737 
738 	lo = t4_read_reg(padap, CIM_SDRAM_BASE_ADDR_A);
739 	hi = t4_read_reg(padap, CIM_SDRAM_ADDR_SIZE_A) + lo - 1;
740 	meminfo_buff->up_ram_lo = lo;
741 	meminfo_buff->up_ram_hi = hi;
742 
743 	lo = t4_read_reg(padap, CIM_EXTMEM2_BASE_ADDR_A);
744 	hi = t4_read_reg(padap, CIM_EXTMEM2_ADDR_SIZE_A) + lo - 1;
745 	meminfo_buff->up_extmem2_lo = lo;
746 	meminfo_buff->up_extmem2_hi = hi;
747 
748 	lo = t4_read_reg(padap, TP_PMM_RX_MAX_PAGE_A);
749 	for (i = 0, meminfo_buff->free_rx_cnt = 0; i < 2; i++)
750 		meminfo_buff->free_rx_cnt +=
751 			FREERXPAGECOUNT_G(t4_read_reg(padap,
752 						      TP_FLM_FREE_RX_CNT_A));
753 
754 	meminfo_buff->rx_pages_data[0] =  PMRXMAXPAGE_G(lo);
755 	meminfo_buff->rx_pages_data[1] =
756 		t4_read_reg(padap, TP_PMM_RX_PAGE_SIZE_A) >> 10;
757 	meminfo_buff->rx_pages_data[2] = (lo & PMRXNUMCHN_F) ? 2 : 1;
758 
759 	lo = t4_read_reg(padap, TP_PMM_TX_MAX_PAGE_A);
760 	hi = t4_read_reg(padap, TP_PMM_TX_PAGE_SIZE_A);
761 	for (i = 0, meminfo_buff->free_tx_cnt = 0; i < 4; i++)
762 		meminfo_buff->free_tx_cnt +=
763 			FREETXPAGECOUNT_G(t4_read_reg(padap,
764 						      TP_FLM_FREE_TX_CNT_A));
765 
766 	meminfo_buff->tx_pages_data[0] = PMTXMAXPAGE_G(lo);
767 	meminfo_buff->tx_pages_data[1] =
768 		hi >= (1 << 20) ? (hi >> 20) : (hi >> 10);
769 	meminfo_buff->tx_pages_data[2] =
770 		hi >= (1 << 20) ? 'M' : 'K';
771 	meminfo_buff->tx_pages_data[3] = 1 << PMTXNUMCHN_G(lo);
772 
773 	meminfo_buff->p_structs = t4_read_reg(padap, TP_CMM_MM_MAX_PSTRUCT_A);
774 	meminfo_buff->p_structs_free_cnt =
775 		FREEPSTRUCTCOUNT_G(t4_read_reg(padap, TP_FLM_FREE_PS_CNT_A));
776 
777 	for (i = 0; i < 4; i++) {
778 		if (CHELSIO_CHIP_VERSION(padap->params.chip) > CHELSIO_T5)
779 			lo = t4_read_reg(padap,
780 					 MPS_RX_MAC_BG_PG_CNT0_A + i * 4);
781 		else
782 			lo = t4_read_reg(padap, MPS_RX_PG_RSV0_A + i * 4);
783 		if (is_t5(padap->params.chip)) {
784 			used = T5_USED_G(lo);
785 			alloc = T5_ALLOC_G(lo);
786 		} else {
787 			used = USED_G(lo);
788 			alloc = ALLOC_G(lo);
789 		}
790 		meminfo_buff->port_used[i] = used;
791 		meminfo_buff->port_alloc[i] = alloc;
792 	}
793 
794 	for (i = 0; i < padap->params.arch.nchan; i++) {
795 		if (CHELSIO_CHIP_VERSION(padap->params.chip) > CHELSIO_T5)
796 			lo = t4_read_reg(padap,
797 					 MPS_RX_LPBK_BG_PG_CNT0_A + i * 4);
798 		else
799 			lo = t4_read_reg(padap, MPS_RX_PG_RSV4_A + i * 4);
800 		if (is_t5(padap->params.chip)) {
801 			used = T5_USED_G(lo);
802 			alloc = T5_ALLOC_G(lo);
803 		} else {
804 			used = USED_G(lo);
805 			alloc = ALLOC_G(lo);
806 		}
807 		meminfo_buff->loopback_used[i] = used;
808 		meminfo_buff->loopback_alloc[i] = alloc;
809 	}
810 
811 	return 0;
812 }
813 
814 int cudbg_collect_reg_dump(struct cudbg_init *pdbg_init,
815 			   struct cudbg_buffer *dbg_buff,
816 			   struct cudbg_error *cudbg_err)
817 {
818 	struct adapter *padap = pdbg_init->adap;
819 	struct cudbg_buffer temp_buff = { 0 };
820 	u32 buf_size = 0;
821 	int rc = 0;
822 
823 	if (is_t4(padap->params.chip))
824 		buf_size = T4_REGMAP_SIZE;
825 	else if (is_t5(padap->params.chip) || is_t6(padap->params.chip))
826 		buf_size = T5_REGMAP_SIZE;
827 
828 	rc = cudbg_get_buff(pdbg_init, dbg_buff, buf_size, &temp_buff);
829 	if (rc)
830 		return rc;
831 	t4_get_regs(padap, (void *)temp_buff.data, temp_buff.size);
832 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
833 }
834 
835 int cudbg_collect_fw_devlog(struct cudbg_init *pdbg_init,
836 			    struct cudbg_buffer *dbg_buff,
837 			    struct cudbg_error *cudbg_err)
838 {
839 	struct adapter *padap = pdbg_init->adap;
840 	struct cudbg_buffer temp_buff = { 0 };
841 	struct devlog_params *dparams;
842 	int rc = 0;
843 
844 	rc = t4_init_devlog_params(padap);
845 	if (rc < 0) {
846 		cudbg_err->sys_err = rc;
847 		return rc;
848 	}
849 
850 	dparams = &padap->params.devlog;
851 	rc = cudbg_get_buff(pdbg_init, dbg_buff, dparams->size, &temp_buff);
852 	if (rc)
853 		return rc;
854 
855 	/* Collect FW devlog */
856 	if (dparams->start != 0) {
857 		spin_lock(&padap->win0_lock);
858 		rc = t4_memory_rw(padap, padap->params.drv_memwin,
859 				  dparams->memtype, dparams->start,
860 				  dparams->size,
861 				  (__be32 *)(char *)temp_buff.data,
862 				  1);
863 		spin_unlock(&padap->win0_lock);
864 		if (rc) {
865 			cudbg_err->sys_err = rc;
866 			cudbg_put_buff(pdbg_init, &temp_buff);
867 			return rc;
868 		}
869 	}
870 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
871 }
872 
873 int cudbg_collect_cim_la(struct cudbg_init *pdbg_init,
874 			 struct cudbg_buffer *dbg_buff,
875 			 struct cudbg_error *cudbg_err)
876 {
877 	struct adapter *padap = pdbg_init->adap;
878 	struct cudbg_buffer temp_buff = { 0 };
879 	int size, rc;
880 	u32 cfg = 0;
881 
882 	if (is_t6(padap->params.chip)) {
883 		size = padap->params.cim_la_size / 10 + 1;
884 		size *= 10 * sizeof(u32);
885 	} else {
886 		size = padap->params.cim_la_size / 8;
887 		size *= 8 * sizeof(u32);
888 	}
889 
890 	size += sizeof(cfg);
891 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
892 	if (rc)
893 		return rc;
894 
895 	rc = t4_cim_read(padap, UP_UP_DBG_LA_CFG_A, 1, &cfg);
896 	if (rc) {
897 		cudbg_err->sys_err = rc;
898 		cudbg_put_buff(pdbg_init, &temp_buff);
899 		return rc;
900 	}
901 
902 	memcpy((char *)temp_buff.data, &cfg, sizeof(cfg));
903 	rc = t4_cim_read_la(padap,
904 			    (u32 *)((char *)temp_buff.data + sizeof(cfg)),
905 			    NULL);
906 	if (rc < 0) {
907 		cudbg_err->sys_err = rc;
908 		cudbg_put_buff(pdbg_init, &temp_buff);
909 		return rc;
910 	}
911 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
912 }
913 
914 int cudbg_collect_cim_ma_la(struct cudbg_init *pdbg_init,
915 			    struct cudbg_buffer *dbg_buff,
916 			    struct cudbg_error *cudbg_err)
917 {
918 	struct adapter *padap = pdbg_init->adap;
919 	struct cudbg_buffer temp_buff = { 0 };
920 	int size, rc;
921 
922 	size = 2 * CIM_MALA_SIZE * 5 * sizeof(u32);
923 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
924 	if (rc)
925 		return rc;
926 
927 	t4_cim_read_ma_la(padap,
928 			  (u32 *)temp_buff.data,
929 			  (u32 *)((char *)temp_buff.data +
930 				  5 * CIM_MALA_SIZE));
931 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
932 }
933 
934 int cudbg_collect_cim_qcfg(struct cudbg_init *pdbg_init,
935 			   struct cudbg_buffer *dbg_buff,
936 			   struct cudbg_error *cudbg_err)
937 {
938 	struct adapter *padap = pdbg_init->adap;
939 	struct cudbg_buffer temp_buff = { 0 };
940 	struct cudbg_cim_qcfg *cim_qcfg_data;
941 	int rc;
942 
943 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_cim_qcfg),
944 			    &temp_buff);
945 	if (rc)
946 		return rc;
947 
948 	cim_qcfg_data = (struct cudbg_cim_qcfg *)temp_buff.data;
949 	cim_qcfg_data->chip = padap->params.chip;
950 	rc = t4_cim_read(padap, UP_IBQ_0_RDADDR_A,
951 			 ARRAY_SIZE(cim_qcfg_data->stat), cim_qcfg_data->stat);
952 	if (rc) {
953 		cudbg_err->sys_err = rc;
954 		cudbg_put_buff(pdbg_init, &temp_buff);
955 		return rc;
956 	}
957 
958 	rc = t4_cim_read(padap, UP_OBQ_0_REALADDR_A,
959 			 ARRAY_SIZE(cim_qcfg_data->obq_wr),
960 			 cim_qcfg_data->obq_wr);
961 	if (rc) {
962 		cudbg_err->sys_err = rc;
963 		cudbg_put_buff(pdbg_init, &temp_buff);
964 		return rc;
965 	}
966 
967 	t4_read_cimq_cfg(padap, cim_qcfg_data->base, cim_qcfg_data->size,
968 			 cim_qcfg_data->thres);
969 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
970 }
971 
972 static int cudbg_read_cim_ibq(struct cudbg_init *pdbg_init,
973 			      struct cudbg_buffer *dbg_buff,
974 			      struct cudbg_error *cudbg_err, int qid)
975 {
976 	struct adapter *padap = pdbg_init->adap;
977 	struct cudbg_buffer temp_buff = { 0 };
978 	int no_of_read_words, rc = 0;
979 	u32 qsize;
980 
981 	/* collect CIM IBQ */
982 	qsize = CIM_IBQ_SIZE * 4 * sizeof(u32);
983 	rc = cudbg_get_buff(pdbg_init, dbg_buff, qsize, &temp_buff);
984 	if (rc)
985 		return rc;
986 
987 	/* t4_read_cim_ibq will return no. of read words or error */
988 	no_of_read_words = t4_read_cim_ibq(padap, qid,
989 					   (u32 *)temp_buff.data, qsize);
990 	/* no_of_read_words is less than or equal to 0 means error */
991 	if (no_of_read_words <= 0) {
992 		if (!no_of_read_words)
993 			rc = CUDBG_SYSTEM_ERROR;
994 		else
995 			rc = no_of_read_words;
996 		cudbg_err->sys_err = rc;
997 		cudbg_put_buff(pdbg_init, &temp_buff);
998 		return rc;
999 	}
1000 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1001 }
1002 
1003 int cudbg_collect_cim_ibq_tp0(struct cudbg_init *pdbg_init,
1004 			      struct cudbg_buffer *dbg_buff,
1005 			      struct cudbg_error *cudbg_err)
1006 {
1007 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 0);
1008 }
1009 
1010 int cudbg_collect_cim_ibq_tp1(struct cudbg_init *pdbg_init,
1011 			      struct cudbg_buffer *dbg_buff,
1012 			      struct cudbg_error *cudbg_err)
1013 {
1014 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 1);
1015 }
1016 
1017 int cudbg_collect_cim_ibq_ulp(struct cudbg_init *pdbg_init,
1018 			      struct cudbg_buffer *dbg_buff,
1019 			      struct cudbg_error *cudbg_err)
1020 {
1021 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 2);
1022 }
1023 
1024 int cudbg_collect_cim_ibq_sge0(struct cudbg_init *pdbg_init,
1025 			       struct cudbg_buffer *dbg_buff,
1026 			       struct cudbg_error *cudbg_err)
1027 {
1028 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 3);
1029 }
1030 
1031 int cudbg_collect_cim_ibq_sge1(struct cudbg_init *pdbg_init,
1032 			       struct cudbg_buffer *dbg_buff,
1033 			       struct cudbg_error *cudbg_err)
1034 {
1035 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 4);
1036 }
1037 
1038 int cudbg_collect_cim_ibq_ncsi(struct cudbg_init *pdbg_init,
1039 			       struct cudbg_buffer *dbg_buff,
1040 			       struct cudbg_error *cudbg_err)
1041 {
1042 	return cudbg_read_cim_ibq(pdbg_init, dbg_buff, cudbg_err, 5);
1043 }
1044 
1045 u32 cudbg_cim_obq_size(struct adapter *padap, int qid)
1046 {
1047 	u32 value;
1048 
1049 	t4_write_reg(padap, CIM_QUEUE_CONFIG_REF_A, OBQSELECT_F |
1050 		     QUENUMSELECT_V(qid));
1051 	value = t4_read_reg(padap, CIM_QUEUE_CONFIG_CTRL_A);
1052 	value = CIMQSIZE_G(value) * 64; /* size in number of words */
1053 	return value * sizeof(u32);
1054 }
1055 
1056 static int cudbg_read_cim_obq(struct cudbg_init *pdbg_init,
1057 			      struct cudbg_buffer *dbg_buff,
1058 			      struct cudbg_error *cudbg_err, int qid)
1059 {
1060 	struct adapter *padap = pdbg_init->adap;
1061 	struct cudbg_buffer temp_buff = { 0 };
1062 	int no_of_read_words, rc = 0;
1063 	u32 qsize;
1064 
1065 	/* collect CIM OBQ */
1066 	qsize =  cudbg_cim_obq_size(padap, qid);
1067 	rc = cudbg_get_buff(pdbg_init, dbg_buff, qsize, &temp_buff);
1068 	if (rc)
1069 		return rc;
1070 
1071 	/* t4_read_cim_obq will return no. of read words or error */
1072 	no_of_read_words = t4_read_cim_obq(padap, qid,
1073 					   (u32 *)temp_buff.data, qsize);
1074 	/* no_of_read_words is less than or equal to 0 means error */
1075 	if (no_of_read_words <= 0) {
1076 		if (!no_of_read_words)
1077 			rc = CUDBG_SYSTEM_ERROR;
1078 		else
1079 			rc = no_of_read_words;
1080 		cudbg_err->sys_err = rc;
1081 		cudbg_put_buff(pdbg_init, &temp_buff);
1082 		return rc;
1083 	}
1084 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1085 }
1086 
1087 int cudbg_collect_cim_obq_ulp0(struct cudbg_init *pdbg_init,
1088 			       struct cudbg_buffer *dbg_buff,
1089 			       struct cudbg_error *cudbg_err)
1090 {
1091 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 0);
1092 }
1093 
1094 int cudbg_collect_cim_obq_ulp1(struct cudbg_init *pdbg_init,
1095 			       struct cudbg_buffer *dbg_buff,
1096 			       struct cudbg_error *cudbg_err)
1097 {
1098 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 1);
1099 }
1100 
1101 int cudbg_collect_cim_obq_ulp2(struct cudbg_init *pdbg_init,
1102 			       struct cudbg_buffer *dbg_buff,
1103 			       struct cudbg_error *cudbg_err)
1104 {
1105 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 2);
1106 }
1107 
1108 int cudbg_collect_cim_obq_ulp3(struct cudbg_init *pdbg_init,
1109 			       struct cudbg_buffer *dbg_buff,
1110 			       struct cudbg_error *cudbg_err)
1111 {
1112 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 3);
1113 }
1114 
1115 int cudbg_collect_cim_obq_sge(struct cudbg_init *pdbg_init,
1116 			      struct cudbg_buffer *dbg_buff,
1117 			      struct cudbg_error *cudbg_err)
1118 {
1119 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 4);
1120 }
1121 
1122 int cudbg_collect_cim_obq_ncsi(struct cudbg_init *pdbg_init,
1123 			       struct cudbg_buffer *dbg_buff,
1124 			       struct cudbg_error *cudbg_err)
1125 {
1126 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 5);
1127 }
1128 
1129 int cudbg_collect_obq_sge_rx_q0(struct cudbg_init *pdbg_init,
1130 				struct cudbg_buffer *dbg_buff,
1131 				struct cudbg_error *cudbg_err)
1132 {
1133 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 6);
1134 }
1135 
1136 int cudbg_collect_obq_sge_rx_q1(struct cudbg_init *pdbg_init,
1137 				struct cudbg_buffer *dbg_buff,
1138 				struct cudbg_error *cudbg_err)
1139 {
1140 	return cudbg_read_cim_obq(pdbg_init, dbg_buff, cudbg_err, 7);
1141 }
1142 
1143 static int cudbg_meminfo_get_mem_index(struct adapter *padap,
1144 				       struct cudbg_meminfo *mem_info,
1145 				       u8 mem_type, u8 *idx)
1146 {
1147 	u8 i, flag;
1148 
1149 	switch (mem_type) {
1150 	case MEM_EDC0:
1151 		flag = EDC0_FLAG;
1152 		break;
1153 	case MEM_EDC1:
1154 		flag = EDC1_FLAG;
1155 		break;
1156 	case MEM_MC0:
1157 		/* Some T5 cards have both MC0 and MC1. */
1158 		flag = is_t5(padap->params.chip) ? MC0_FLAG : MC_FLAG;
1159 		break;
1160 	case MEM_MC1:
1161 		flag = MC1_FLAG;
1162 		break;
1163 	case MEM_HMA:
1164 		flag = HMA_FLAG;
1165 		break;
1166 	default:
1167 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
1168 	}
1169 
1170 	for (i = 0; i < mem_info->avail_c; i++) {
1171 		if (mem_info->avail[i].idx == flag) {
1172 			*idx = i;
1173 			return 0;
1174 		}
1175 	}
1176 
1177 	return CUDBG_STATUS_ENTITY_NOT_FOUND;
1178 }
1179 
1180 /* Fetch the @region_name's start and end from @meminfo. */
1181 static int cudbg_get_mem_region(struct adapter *padap,
1182 				struct cudbg_meminfo *meminfo,
1183 				u8 mem_type, const char *region_name,
1184 				struct cudbg_mem_desc *mem_desc)
1185 {
1186 	u8 mc, found = 0;
1187 	u32 idx = 0;
1188 	int rc, i;
1189 
1190 	rc = cudbg_meminfo_get_mem_index(padap, meminfo, mem_type, &mc);
1191 	if (rc)
1192 		return rc;
1193 
1194 	i = match_string(cudbg_region, ARRAY_SIZE(cudbg_region), region_name);
1195 	if (i < 0)
1196 		return -EINVAL;
1197 
1198 	idx = i;
1199 	for (i = 0; i < meminfo->mem_c; i++) {
1200 		if (meminfo->mem[i].idx >= ARRAY_SIZE(cudbg_region))
1201 			continue; /* Skip holes */
1202 
1203 		if (!(meminfo->mem[i].limit))
1204 			meminfo->mem[i].limit =
1205 				i < meminfo->mem_c - 1 ?
1206 				meminfo->mem[i + 1].base - 1 : ~0;
1207 
1208 		if (meminfo->mem[i].idx == idx) {
1209 			/* Check if the region exists in @mem_type memory */
1210 			if (meminfo->mem[i].base < meminfo->avail[mc].base &&
1211 			    meminfo->mem[i].limit < meminfo->avail[mc].base)
1212 				return -EINVAL;
1213 
1214 			if (meminfo->mem[i].base > meminfo->avail[mc].limit)
1215 				return -EINVAL;
1216 
1217 			memcpy(mem_desc, &meminfo->mem[i],
1218 			       sizeof(struct cudbg_mem_desc));
1219 			found = 1;
1220 			break;
1221 		}
1222 	}
1223 	if (!found)
1224 		return -EINVAL;
1225 
1226 	return 0;
1227 }
1228 
1229 /* Fetch and update the start and end of the requested memory region w.r.t 0
1230  * in the corresponding EDC/MC/HMA.
1231  */
1232 static int cudbg_get_mem_relative(struct adapter *padap,
1233 				  struct cudbg_meminfo *meminfo,
1234 				  u8 mem_type, u32 *out_base, u32 *out_end)
1235 {
1236 	u8 mc_idx;
1237 	int rc;
1238 
1239 	rc = cudbg_meminfo_get_mem_index(padap, meminfo, mem_type, &mc_idx);
1240 	if (rc)
1241 		return rc;
1242 
1243 	if (*out_base < meminfo->avail[mc_idx].base)
1244 		*out_base = 0;
1245 	else
1246 		*out_base -= meminfo->avail[mc_idx].base;
1247 
1248 	if (*out_end > meminfo->avail[mc_idx].limit)
1249 		*out_end = meminfo->avail[mc_idx].limit;
1250 	else
1251 		*out_end -= meminfo->avail[mc_idx].base;
1252 
1253 	return 0;
1254 }
1255 
1256 /* Get TX and RX Payload region */
1257 static int cudbg_get_payload_range(struct adapter *padap, u8 mem_type,
1258 				   const char *region_name,
1259 				   struct cudbg_region_info *payload)
1260 {
1261 	struct cudbg_mem_desc mem_desc = { 0 };
1262 	struct cudbg_meminfo meminfo;
1263 	int rc;
1264 
1265 	rc = cudbg_fill_meminfo(padap, &meminfo);
1266 	if (rc)
1267 		return rc;
1268 
1269 	rc = cudbg_get_mem_region(padap, &meminfo, mem_type, region_name,
1270 				  &mem_desc);
1271 	if (rc) {
1272 		payload->exist = false;
1273 		return 0;
1274 	}
1275 
1276 	payload->exist = true;
1277 	payload->start = mem_desc.base;
1278 	payload->end = mem_desc.limit;
1279 
1280 	return cudbg_get_mem_relative(padap, &meminfo, mem_type,
1281 				      &payload->start, &payload->end);
1282 }
1283 
1284 static int cudbg_memory_read(struct cudbg_init *pdbg_init, int win,
1285 			     int mtype, u32 addr, u32 len, void *hbuf)
1286 {
1287 	u32 win_pf, memoffset, mem_aperture, mem_base;
1288 	struct adapter *adap = pdbg_init->adap;
1289 	u32 pos, offset, resid;
1290 	u32 *res_buf;
1291 	u64 *buf;
1292 	int ret;
1293 
1294 	/* Argument sanity checks ...
1295 	 */
1296 	if (addr & 0x3 || (uintptr_t)hbuf & 0x3)
1297 		return -EINVAL;
1298 
1299 	buf = (u64 *)hbuf;
1300 
1301 	/* Try to do 64-bit reads.  Residual will be handled later. */
1302 	resid = len & 0x7;
1303 	len -= resid;
1304 
1305 	ret = t4_memory_rw_init(adap, win, mtype, &memoffset, &mem_base,
1306 				&mem_aperture);
1307 	if (ret)
1308 		return ret;
1309 
1310 	addr = addr + memoffset;
1311 	win_pf = is_t4(adap->params.chip) ? 0 : PFNUM_V(adap->pf);
1312 
1313 	pos = addr & ~(mem_aperture - 1);
1314 	offset = addr - pos;
1315 
1316 	/* Set up initial PCI-E Memory Window to cover the start of our
1317 	 * transfer.
1318 	 */
1319 	t4_memory_update_win(adap, win, pos | win_pf);
1320 
1321 	/* Transfer data from the adapter */
1322 	while (len > 0) {
1323 		*buf++ = le64_to_cpu((__force __le64)
1324 				     t4_read_reg64(adap, mem_base + offset));
1325 		offset += sizeof(u64);
1326 		len -= sizeof(u64);
1327 
1328 		/* If we've reached the end of our current window aperture,
1329 		 * move the PCI-E Memory Window on to the next.
1330 		 */
1331 		if (offset == mem_aperture) {
1332 			pos += mem_aperture;
1333 			offset = 0;
1334 			t4_memory_update_win(adap, win, pos | win_pf);
1335 		}
1336 	}
1337 
1338 	res_buf = (u32 *)buf;
1339 	/* Read residual in 32-bit multiples */
1340 	while (resid > sizeof(u32)) {
1341 		*res_buf++ = le32_to_cpu((__force __le32)
1342 					 t4_read_reg(adap, mem_base + offset));
1343 		offset += sizeof(u32);
1344 		resid -= sizeof(u32);
1345 
1346 		/* If we've reached the end of our current window aperture,
1347 		 * move the PCI-E Memory Window on to the next.
1348 		 */
1349 		if (offset == mem_aperture) {
1350 			pos += mem_aperture;
1351 			offset = 0;
1352 			t4_memory_update_win(adap, win, pos | win_pf);
1353 		}
1354 	}
1355 
1356 	/* Transfer residual < 32-bits */
1357 	if (resid)
1358 		t4_memory_rw_residual(adap, resid, mem_base + offset,
1359 				      (u8 *)res_buf, T4_MEMORY_READ);
1360 
1361 	return 0;
1362 }
1363 
1364 #define CUDBG_YIELD_ITERATION 256
1365 
1366 static int cudbg_read_fw_mem(struct cudbg_init *pdbg_init,
1367 			     struct cudbg_buffer *dbg_buff, u8 mem_type,
1368 			     unsigned long tot_len,
1369 			     struct cudbg_error *cudbg_err)
1370 {
1371 	static const char * const region_name[] = { "Tx payload:",
1372 						    "Rx payload:" };
1373 	unsigned long bytes, bytes_left, bytes_read = 0;
1374 	struct adapter *padap = pdbg_init->adap;
1375 	struct cudbg_buffer temp_buff = { 0 };
1376 	struct cudbg_region_info payload[2];
1377 	u32 yield_count = 0;
1378 	int rc = 0;
1379 	u8 i;
1380 
1381 	/* Get TX/RX Payload region range if they exist */
1382 	memset(payload, 0, sizeof(payload));
1383 	for (i = 0; i < ARRAY_SIZE(region_name); i++) {
1384 		rc = cudbg_get_payload_range(padap, mem_type, region_name[i],
1385 					     &payload[i]);
1386 		if (rc)
1387 			return rc;
1388 
1389 		if (payload[i].exist) {
1390 			/* Align start and end to avoid wrap around */
1391 			payload[i].start = roundup(payload[i].start,
1392 						   CUDBG_CHUNK_SIZE);
1393 			payload[i].end = rounddown(payload[i].end,
1394 						   CUDBG_CHUNK_SIZE);
1395 		}
1396 	}
1397 
1398 	bytes_left = tot_len;
1399 	while (bytes_left > 0) {
1400 		/* As MC size is huge and read through PIO access, this
1401 		 * loop will hold cpu for a longer time. OS may think that
1402 		 * the process is hanged and will generate CPU stall traces.
1403 		 * So yield the cpu regularly.
1404 		 */
1405 		yield_count++;
1406 		if (!(yield_count % CUDBG_YIELD_ITERATION))
1407 			schedule();
1408 
1409 		bytes = min_t(unsigned long, bytes_left,
1410 			      (unsigned long)CUDBG_CHUNK_SIZE);
1411 		rc = cudbg_get_buff(pdbg_init, dbg_buff, bytes, &temp_buff);
1412 		if (rc)
1413 			return rc;
1414 
1415 		for (i = 0; i < ARRAY_SIZE(payload); i++)
1416 			if (payload[i].exist &&
1417 			    bytes_read >= payload[i].start &&
1418 			    bytes_read + bytes <= payload[i].end)
1419 				/* TX and RX Payload regions can't overlap */
1420 				goto skip_read;
1421 
1422 		spin_lock(&padap->win0_lock);
1423 		rc = cudbg_memory_read(pdbg_init, MEMWIN_NIC, mem_type,
1424 				       bytes_read, bytes, temp_buff.data);
1425 		spin_unlock(&padap->win0_lock);
1426 		if (rc) {
1427 			cudbg_err->sys_err = rc;
1428 			cudbg_put_buff(pdbg_init, &temp_buff);
1429 			return rc;
1430 		}
1431 
1432 skip_read:
1433 		bytes_left -= bytes;
1434 		bytes_read += bytes;
1435 		rc = cudbg_write_and_release_buff(pdbg_init, &temp_buff,
1436 						  dbg_buff);
1437 		if (rc) {
1438 			cudbg_put_buff(pdbg_init, &temp_buff);
1439 			return rc;
1440 		}
1441 	}
1442 	return rc;
1443 }
1444 
1445 static void cudbg_t4_fwcache(struct cudbg_init *pdbg_init,
1446 			     struct cudbg_error *cudbg_err)
1447 {
1448 	struct adapter *padap = pdbg_init->adap;
1449 	int rc;
1450 
1451 	if (is_fw_attached(pdbg_init)) {
1452 		/* Flush uP dcache before reading edcX/mcX  */
1453 		rc = t4_fwcache(padap, FW_PARAM_DEV_FWCACHE_FLUSH);
1454 		if (rc)
1455 			cudbg_err->sys_warn = rc;
1456 	}
1457 }
1458 
1459 static int cudbg_mem_region_size(struct cudbg_init *pdbg_init,
1460 				 struct cudbg_error *cudbg_err,
1461 				 u8 mem_type, unsigned long *region_size)
1462 {
1463 	struct adapter *padap = pdbg_init->adap;
1464 	struct cudbg_meminfo mem_info;
1465 	u8 mc_idx;
1466 	int rc;
1467 
1468 	memset(&mem_info, 0, sizeof(struct cudbg_meminfo));
1469 	rc = cudbg_fill_meminfo(padap, &mem_info);
1470 	if (rc) {
1471 		cudbg_err->sys_err = rc;
1472 		return rc;
1473 	}
1474 
1475 	cudbg_t4_fwcache(pdbg_init, cudbg_err);
1476 	rc = cudbg_meminfo_get_mem_index(padap, &mem_info, mem_type, &mc_idx);
1477 	if (rc) {
1478 		cudbg_err->sys_err = rc;
1479 		return rc;
1480 	}
1481 
1482 	if (region_size)
1483 		*region_size = mem_info.avail[mc_idx].limit -
1484 			       mem_info.avail[mc_idx].base;
1485 
1486 	return 0;
1487 }
1488 
1489 static int cudbg_collect_mem_region(struct cudbg_init *pdbg_init,
1490 				    struct cudbg_buffer *dbg_buff,
1491 				    struct cudbg_error *cudbg_err,
1492 				    u8 mem_type)
1493 {
1494 	unsigned long size = 0;
1495 	int rc;
1496 
1497 	rc = cudbg_mem_region_size(pdbg_init, cudbg_err, mem_type, &size);
1498 	if (rc)
1499 		return rc;
1500 
1501 	return cudbg_read_fw_mem(pdbg_init, dbg_buff, mem_type, size,
1502 				 cudbg_err);
1503 }
1504 
1505 int cudbg_collect_edc0_meminfo(struct cudbg_init *pdbg_init,
1506 			       struct cudbg_buffer *dbg_buff,
1507 			       struct cudbg_error *cudbg_err)
1508 {
1509 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1510 					MEM_EDC0);
1511 }
1512 
1513 int cudbg_collect_edc1_meminfo(struct cudbg_init *pdbg_init,
1514 			       struct cudbg_buffer *dbg_buff,
1515 			       struct cudbg_error *cudbg_err)
1516 {
1517 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1518 					MEM_EDC1);
1519 }
1520 
1521 int cudbg_collect_mc0_meminfo(struct cudbg_init *pdbg_init,
1522 			      struct cudbg_buffer *dbg_buff,
1523 			      struct cudbg_error *cudbg_err)
1524 {
1525 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1526 					MEM_MC0);
1527 }
1528 
1529 int cudbg_collect_mc1_meminfo(struct cudbg_init *pdbg_init,
1530 			      struct cudbg_buffer *dbg_buff,
1531 			      struct cudbg_error *cudbg_err)
1532 {
1533 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1534 					MEM_MC1);
1535 }
1536 
1537 int cudbg_collect_hma_meminfo(struct cudbg_init *pdbg_init,
1538 			      struct cudbg_buffer *dbg_buff,
1539 			      struct cudbg_error *cudbg_err)
1540 {
1541 	return cudbg_collect_mem_region(pdbg_init, dbg_buff, cudbg_err,
1542 					MEM_HMA);
1543 }
1544 
1545 int cudbg_collect_rss(struct cudbg_init *pdbg_init,
1546 		      struct cudbg_buffer *dbg_buff,
1547 		      struct cudbg_error *cudbg_err)
1548 {
1549 	struct adapter *padap = pdbg_init->adap;
1550 	struct cudbg_buffer temp_buff = { 0 };
1551 	int rc, nentries;
1552 
1553 	nentries = t4_chip_rss_size(padap);
1554 	rc = cudbg_get_buff(pdbg_init, dbg_buff, nentries * sizeof(u16),
1555 			    &temp_buff);
1556 	if (rc)
1557 		return rc;
1558 
1559 	rc = t4_read_rss(padap, (u16 *)temp_buff.data);
1560 	if (rc) {
1561 		cudbg_err->sys_err = rc;
1562 		cudbg_put_buff(pdbg_init, &temp_buff);
1563 		return rc;
1564 	}
1565 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1566 }
1567 
1568 int cudbg_collect_rss_vf_config(struct cudbg_init *pdbg_init,
1569 				struct cudbg_buffer *dbg_buff,
1570 				struct cudbg_error *cudbg_err)
1571 {
1572 	struct adapter *padap = pdbg_init->adap;
1573 	struct cudbg_buffer temp_buff = { 0 };
1574 	struct cudbg_rss_vf_conf *vfconf;
1575 	int vf, rc, vf_count;
1576 
1577 	vf_count = padap->params.arch.vfcount;
1578 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
1579 			    vf_count * sizeof(struct cudbg_rss_vf_conf),
1580 			    &temp_buff);
1581 	if (rc)
1582 		return rc;
1583 
1584 	vfconf = (struct cudbg_rss_vf_conf *)temp_buff.data;
1585 	for (vf = 0; vf < vf_count; vf++)
1586 		t4_read_rss_vf_config(padap, vf, &vfconf[vf].rss_vf_vfl,
1587 				      &vfconf[vf].rss_vf_vfh, true);
1588 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1589 }
1590 
1591 int cudbg_collect_path_mtu(struct cudbg_init *pdbg_init,
1592 			   struct cudbg_buffer *dbg_buff,
1593 			   struct cudbg_error *cudbg_err)
1594 {
1595 	struct adapter *padap = pdbg_init->adap;
1596 	struct cudbg_buffer temp_buff = { 0 };
1597 	int rc;
1598 
1599 	rc = cudbg_get_buff(pdbg_init, dbg_buff, NMTUS * sizeof(u16),
1600 			    &temp_buff);
1601 	if (rc)
1602 		return rc;
1603 
1604 	t4_read_mtu_tbl(padap, (u16 *)temp_buff.data, NULL);
1605 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1606 }
1607 
1608 int cudbg_collect_pm_stats(struct cudbg_init *pdbg_init,
1609 			   struct cudbg_buffer *dbg_buff,
1610 			   struct cudbg_error *cudbg_err)
1611 {
1612 	struct adapter *padap = pdbg_init->adap;
1613 	struct cudbg_buffer temp_buff = { 0 };
1614 	struct cudbg_pm_stats *pm_stats_buff;
1615 	int rc;
1616 
1617 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_pm_stats),
1618 			    &temp_buff);
1619 	if (rc)
1620 		return rc;
1621 
1622 	pm_stats_buff = (struct cudbg_pm_stats *)temp_buff.data;
1623 	t4_pmtx_get_stats(padap, pm_stats_buff->tx_cnt, pm_stats_buff->tx_cyc);
1624 	t4_pmrx_get_stats(padap, pm_stats_buff->rx_cnt, pm_stats_buff->rx_cyc);
1625 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1626 }
1627 
1628 int cudbg_collect_hw_sched(struct cudbg_init *pdbg_init,
1629 			   struct cudbg_buffer *dbg_buff,
1630 			   struct cudbg_error *cudbg_err)
1631 {
1632 	struct adapter *padap = pdbg_init->adap;
1633 	struct cudbg_buffer temp_buff = { 0 };
1634 	struct cudbg_hw_sched *hw_sched_buff;
1635 	int i, rc = 0;
1636 
1637 	if (!padap->params.vpd.cclk)
1638 		return CUDBG_STATUS_CCLK_NOT_DEFINED;
1639 
1640 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_hw_sched),
1641 			    &temp_buff);
1642 
1643 	if (rc)
1644 		return rc;
1645 
1646 	hw_sched_buff = (struct cudbg_hw_sched *)temp_buff.data;
1647 	hw_sched_buff->map = t4_read_reg(padap, TP_TX_MOD_QUEUE_REQ_MAP_A);
1648 	hw_sched_buff->mode = TIMERMODE_G(t4_read_reg(padap, TP_MOD_CONFIG_A));
1649 	t4_read_pace_tbl(padap, hw_sched_buff->pace_tab);
1650 	for (i = 0; i < NTX_SCHED; ++i)
1651 		t4_get_tx_sched(padap, i, &hw_sched_buff->kbps[i],
1652 				&hw_sched_buff->ipg[i], true);
1653 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1654 }
1655 
1656 int cudbg_collect_tp_indirect(struct cudbg_init *pdbg_init,
1657 			      struct cudbg_buffer *dbg_buff,
1658 			      struct cudbg_error *cudbg_err)
1659 {
1660 	struct adapter *padap = pdbg_init->adap;
1661 	struct cudbg_buffer temp_buff = { 0 };
1662 	struct ireg_buf *ch_tp_pio;
1663 	int i, rc, n = 0;
1664 	u32 size;
1665 
1666 	if (is_t5(padap->params.chip))
1667 		n = sizeof(t5_tp_pio_array) +
1668 		    sizeof(t5_tp_tm_pio_array) +
1669 		    sizeof(t5_tp_mib_index_array);
1670 	else
1671 		n = sizeof(t6_tp_pio_array) +
1672 		    sizeof(t6_tp_tm_pio_array) +
1673 		    sizeof(t6_tp_mib_index_array);
1674 
1675 	n = n / (IREG_NUM_ELEM * sizeof(u32));
1676 	size = sizeof(struct ireg_buf) * n;
1677 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1678 	if (rc)
1679 		return rc;
1680 
1681 	ch_tp_pio = (struct ireg_buf *)temp_buff.data;
1682 
1683 	/* TP_PIO */
1684 	if (is_t5(padap->params.chip))
1685 		n = sizeof(t5_tp_pio_array) / (IREG_NUM_ELEM * sizeof(u32));
1686 	else if (is_t6(padap->params.chip))
1687 		n = sizeof(t6_tp_pio_array) / (IREG_NUM_ELEM * sizeof(u32));
1688 
1689 	for (i = 0; i < n; i++) {
1690 		struct ireg_field *tp_pio = &ch_tp_pio->tp_pio;
1691 		u32 *buff = ch_tp_pio->outbuf;
1692 
1693 		if (is_t5(padap->params.chip)) {
1694 			tp_pio->ireg_addr = t5_tp_pio_array[i][0];
1695 			tp_pio->ireg_data = t5_tp_pio_array[i][1];
1696 			tp_pio->ireg_local_offset = t5_tp_pio_array[i][2];
1697 			tp_pio->ireg_offset_range = t5_tp_pio_array[i][3];
1698 		} else if (is_t6(padap->params.chip)) {
1699 			tp_pio->ireg_addr = t6_tp_pio_array[i][0];
1700 			tp_pio->ireg_data = t6_tp_pio_array[i][1];
1701 			tp_pio->ireg_local_offset = t6_tp_pio_array[i][2];
1702 			tp_pio->ireg_offset_range = t6_tp_pio_array[i][3];
1703 		}
1704 		t4_tp_pio_read(padap, buff, tp_pio->ireg_offset_range,
1705 			       tp_pio->ireg_local_offset, true);
1706 		ch_tp_pio++;
1707 	}
1708 
1709 	/* TP_TM_PIO */
1710 	if (is_t5(padap->params.chip))
1711 		n = sizeof(t5_tp_tm_pio_array) / (IREG_NUM_ELEM * sizeof(u32));
1712 	else if (is_t6(padap->params.chip))
1713 		n = sizeof(t6_tp_tm_pio_array) / (IREG_NUM_ELEM * sizeof(u32));
1714 
1715 	for (i = 0; i < n; i++) {
1716 		struct ireg_field *tp_pio = &ch_tp_pio->tp_pio;
1717 		u32 *buff = ch_tp_pio->outbuf;
1718 
1719 		if (is_t5(padap->params.chip)) {
1720 			tp_pio->ireg_addr = t5_tp_tm_pio_array[i][0];
1721 			tp_pio->ireg_data = t5_tp_tm_pio_array[i][1];
1722 			tp_pio->ireg_local_offset = t5_tp_tm_pio_array[i][2];
1723 			tp_pio->ireg_offset_range = t5_tp_tm_pio_array[i][3];
1724 		} else if (is_t6(padap->params.chip)) {
1725 			tp_pio->ireg_addr = t6_tp_tm_pio_array[i][0];
1726 			tp_pio->ireg_data = t6_tp_tm_pio_array[i][1];
1727 			tp_pio->ireg_local_offset = t6_tp_tm_pio_array[i][2];
1728 			tp_pio->ireg_offset_range = t6_tp_tm_pio_array[i][3];
1729 		}
1730 		t4_tp_tm_pio_read(padap, buff, tp_pio->ireg_offset_range,
1731 				  tp_pio->ireg_local_offset, true);
1732 		ch_tp_pio++;
1733 	}
1734 
1735 	/* TP_MIB_INDEX */
1736 	if (is_t5(padap->params.chip))
1737 		n = sizeof(t5_tp_mib_index_array) /
1738 		    (IREG_NUM_ELEM * sizeof(u32));
1739 	else if (is_t6(padap->params.chip))
1740 		n = sizeof(t6_tp_mib_index_array) /
1741 		    (IREG_NUM_ELEM * sizeof(u32));
1742 
1743 	for (i = 0; i < n ; i++) {
1744 		struct ireg_field *tp_pio = &ch_tp_pio->tp_pio;
1745 		u32 *buff = ch_tp_pio->outbuf;
1746 
1747 		if (is_t5(padap->params.chip)) {
1748 			tp_pio->ireg_addr = t5_tp_mib_index_array[i][0];
1749 			tp_pio->ireg_data = t5_tp_mib_index_array[i][1];
1750 			tp_pio->ireg_local_offset =
1751 				t5_tp_mib_index_array[i][2];
1752 			tp_pio->ireg_offset_range =
1753 				t5_tp_mib_index_array[i][3];
1754 		} else if (is_t6(padap->params.chip)) {
1755 			tp_pio->ireg_addr = t6_tp_mib_index_array[i][0];
1756 			tp_pio->ireg_data = t6_tp_mib_index_array[i][1];
1757 			tp_pio->ireg_local_offset =
1758 				t6_tp_mib_index_array[i][2];
1759 			tp_pio->ireg_offset_range =
1760 				t6_tp_mib_index_array[i][3];
1761 		}
1762 		t4_tp_mib_read(padap, buff, tp_pio->ireg_offset_range,
1763 			       tp_pio->ireg_local_offset, true);
1764 		ch_tp_pio++;
1765 	}
1766 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1767 }
1768 
1769 static void cudbg_read_sge_qbase_indirect_reg(struct adapter *padap,
1770 					      struct sge_qbase_reg_field *qbase,
1771 					      u32 func, bool is_pf)
1772 {
1773 	u32 *buff, i;
1774 
1775 	if (is_pf) {
1776 		buff = qbase->pf_data_value[func];
1777 	} else {
1778 		buff = qbase->vf_data_value[func];
1779 		/* In SGE_QBASE_INDEX,
1780 		 * Entries 0->7 are PF0->7, Entries 8->263 are VFID0->256.
1781 		 */
1782 		func += 8;
1783 	}
1784 
1785 	t4_write_reg(padap, qbase->reg_addr, func);
1786 	for (i = 0; i < SGE_QBASE_DATA_REG_NUM; i++, buff++)
1787 		*buff = t4_read_reg(padap, qbase->reg_data[i]);
1788 }
1789 
1790 int cudbg_collect_sge_indirect(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 	struct sge_qbase_reg_field *sge_qbase;
1797 	struct ireg_buf *ch_sge_dbg;
1798 	u8 padap_running = 0;
1799 	int i, rc;
1800 	u32 size;
1801 
1802 	/* Accessing SGE_QBASE_MAP[0-3] and SGE_QBASE_INDEX regs can
1803 	 * lead to SGE missing doorbells under heavy traffic. So, only
1804 	 * collect them when adapter is idle.
1805 	 */
1806 	for_each_port(padap, i) {
1807 		padap_running = netif_running(padap->port[i]);
1808 		if (padap_running)
1809 			break;
1810 	}
1811 
1812 	size = sizeof(*ch_sge_dbg) * 2;
1813 	if (!padap_running)
1814 		size += sizeof(*sge_qbase);
1815 
1816 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1817 	if (rc)
1818 		return rc;
1819 
1820 	ch_sge_dbg = (struct ireg_buf *)temp_buff.data;
1821 	for (i = 0; i < 2; i++) {
1822 		struct ireg_field *sge_pio = &ch_sge_dbg->tp_pio;
1823 		u32 *buff = ch_sge_dbg->outbuf;
1824 
1825 		sge_pio->ireg_addr = t5_sge_dbg_index_array[i][0];
1826 		sge_pio->ireg_data = t5_sge_dbg_index_array[i][1];
1827 		sge_pio->ireg_local_offset = t5_sge_dbg_index_array[i][2];
1828 		sge_pio->ireg_offset_range = t5_sge_dbg_index_array[i][3];
1829 		t4_read_indirect(padap,
1830 				 sge_pio->ireg_addr,
1831 				 sge_pio->ireg_data,
1832 				 buff,
1833 				 sge_pio->ireg_offset_range,
1834 				 sge_pio->ireg_local_offset);
1835 		ch_sge_dbg++;
1836 	}
1837 
1838 	if (CHELSIO_CHIP_VERSION(padap->params.chip) > CHELSIO_T5 &&
1839 	    !padap_running) {
1840 		sge_qbase = (struct sge_qbase_reg_field *)ch_sge_dbg;
1841 		/* 1 addr reg SGE_QBASE_INDEX and 4 data reg
1842 		 * SGE_QBASE_MAP[0-3]
1843 		 */
1844 		sge_qbase->reg_addr = t6_sge_qbase_index_array[0];
1845 		for (i = 0; i < SGE_QBASE_DATA_REG_NUM; i++)
1846 			sge_qbase->reg_data[i] =
1847 				t6_sge_qbase_index_array[i + 1];
1848 
1849 		for (i = 0; i <= PCIE_FW_MASTER_M; i++)
1850 			cudbg_read_sge_qbase_indirect_reg(padap, sge_qbase,
1851 							  i, true);
1852 
1853 		for (i = 0; i < padap->params.arch.vfcount; i++)
1854 			cudbg_read_sge_qbase_indirect_reg(padap, sge_qbase,
1855 							  i, false);
1856 
1857 		sge_qbase->vfcount = padap->params.arch.vfcount;
1858 	}
1859 
1860 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1861 }
1862 
1863 int cudbg_collect_ulprx_la(struct cudbg_init *pdbg_init,
1864 			   struct cudbg_buffer *dbg_buff,
1865 			   struct cudbg_error *cudbg_err)
1866 {
1867 	struct adapter *padap = pdbg_init->adap;
1868 	struct cudbg_buffer temp_buff = { 0 };
1869 	struct cudbg_ulprx_la *ulprx_la_buff;
1870 	int rc;
1871 
1872 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_ulprx_la),
1873 			    &temp_buff);
1874 	if (rc)
1875 		return rc;
1876 
1877 	ulprx_la_buff = (struct cudbg_ulprx_la *)temp_buff.data;
1878 	t4_ulprx_read_la(padap, (u32 *)ulprx_la_buff->data);
1879 	ulprx_la_buff->size = ULPRX_LA_SIZE;
1880 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1881 }
1882 
1883 int cudbg_collect_tp_la(struct cudbg_init *pdbg_init,
1884 			struct cudbg_buffer *dbg_buff,
1885 			struct cudbg_error *cudbg_err)
1886 {
1887 	struct adapter *padap = pdbg_init->adap;
1888 	struct cudbg_buffer temp_buff = { 0 };
1889 	struct cudbg_tp_la *tp_la_buff;
1890 	int size, rc;
1891 
1892 	size = sizeof(struct cudbg_tp_la) + TPLA_SIZE *  sizeof(u64);
1893 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1894 	if (rc)
1895 		return rc;
1896 
1897 	tp_la_buff = (struct cudbg_tp_la *)temp_buff.data;
1898 	tp_la_buff->mode = DBGLAMODE_G(t4_read_reg(padap, TP_DBG_LA_CONFIG_A));
1899 	t4_tp_read_la(padap, (u64 *)tp_la_buff->data, NULL);
1900 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1901 }
1902 
1903 int cudbg_collect_meminfo(struct cudbg_init *pdbg_init,
1904 			  struct cudbg_buffer *dbg_buff,
1905 			  struct cudbg_error *cudbg_err)
1906 {
1907 	struct adapter *padap = pdbg_init->adap;
1908 	struct cudbg_buffer temp_buff = { 0 };
1909 	struct cudbg_meminfo *meminfo_buff;
1910 	struct cudbg_ver_hdr *ver_hdr;
1911 	int rc;
1912 
1913 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
1914 			    sizeof(struct cudbg_ver_hdr) +
1915 			    sizeof(struct cudbg_meminfo),
1916 			    &temp_buff);
1917 	if (rc)
1918 		return rc;
1919 
1920 	ver_hdr = (struct cudbg_ver_hdr *)temp_buff.data;
1921 	ver_hdr->signature = CUDBG_ENTITY_SIGNATURE;
1922 	ver_hdr->revision = CUDBG_MEMINFO_REV;
1923 	ver_hdr->size = sizeof(struct cudbg_meminfo);
1924 
1925 	meminfo_buff = (struct cudbg_meminfo *)(temp_buff.data +
1926 						sizeof(*ver_hdr));
1927 	rc = cudbg_fill_meminfo(padap, meminfo_buff);
1928 	if (rc) {
1929 		cudbg_err->sys_err = rc;
1930 		cudbg_put_buff(pdbg_init, &temp_buff);
1931 		return rc;
1932 	}
1933 
1934 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1935 }
1936 
1937 int cudbg_collect_cim_pif_la(struct cudbg_init *pdbg_init,
1938 			     struct cudbg_buffer *dbg_buff,
1939 			     struct cudbg_error *cudbg_err)
1940 {
1941 	struct cudbg_cim_pif_la *cim_pif_la_buff;
1942 	struct adapter *padap = pdbg_init->adap;
1943 	struct cudbg_buffer temp_buff = { 0 };
1944 	int size, rc;
1945 
1946 	size = sizeof(struct cudbg_cim_pif_la) +
1947 	       2 * CIM_PIFLA_SIZE * 6 * sizeof(u32);
1948 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
1949 	if (rc)
1950 		return rc;
1951 
1952 	cim_pif_la_buff = (struct cudbg_cim_pif_la *)temp_buff.data;
1953 	cim_pif_la_buff->size = CIM_PIFLA_SIZE;
1954 	t4_cim_read_pif_la(padap, (u32 *)cim_pif_la_buff->data,
1955 			   (u32 *)cim_pif_la_buff->data + 6 * CIM_PIFLA_SIZE,
1956 			   NULL, NULL);
1957 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
1958 }
1959 
1960 int cudbg_collect_clk_info(struct cudbg_init *pdbg_init,
1961 			   struct cudbg_buffer *dbg_buff,
1962 			   struct cudbg_error *cudbg_err)
1963 {
1964 	struct adapter *padap = pdbg_init->adap;
1965 	struct cudbg_buffer temp_buff = { 0 };
1966 	struct cudbg_clk_info *clk_info_buff;
1967 	u64 tp_tick_us;
1968 	int rc;
1969 
1970 	if (!padap->params.vpd.cclk)
1971 		return CUDBG_STATUS_CCLK_NOT_DEFINED;
1972 
1973 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_clk_info),
1974 			    &temp_buff);
1975 	if (rc)
1976 		return rc;
1977 
1978 	clk_info_buff = (struct cudbg_clk_info *)temp_buff.data;
1979 	clk_info_buff->cclk_ps = 1000000000 / padap->params.vpd.cclk; /* psec */
1980 	clk_info_buff->res = t4_read_reg(padap, TP_TIMER_RESOLUTION_A);
1981 	clk_info_buff->tre = TIMERRESOLUTION_G(clk_info_buff->res);
1982 	clk_info_buff->dack_re = DELAYEDACKRESOLUTION_G(clk_info_buff->res);
1983 	tp_tick_us = (clk_info_buff->cclk_ps << clk_info_buff->tre) / 1000000;
1984 
1985 	clk_info_buff->dack_timer =
1986 		(clk_info_buff->cclk_ps << clk_info_buff->dack_re) / 1000000 *
1987 		t4_read_reg(padap, TP_DACK_TIMER_A);
1988 	clk_info_buff->retransmit_min =
1989 		tp_tick_us * t4_read_reg(padap, TP_RXT_MIN_A);
1990 	clk_info_buff->retransmit_max =
1991 		tp_tick_us * t4_read_reg(padap, TP_RXT_MAX_A);
1992 	clk_info_buff->persist_timer_min =
1993 		tp_tick_us * t4_read_reg(padap, TP_PERS_MIN_A);
1994 	clk_info_buff->persist_timer_max =
1995 		tp_tick_us * t4_read_reg(padap, TP_PERS_MAX_A);
1996 	clk_info_buff->keepalive_idle_timer =
1997 		tp_tick_us * t4_read_reg(padap, TP_KEEP_IDLE_A);
1998 	clk_info_buff->keepalive_interval =
1999 		tp_tick_us * t4_read_reg(padap, TP_KEEP_INTVL_A);
2000 	clk_info_buff->initial_srtt =
2001 		tp_tick_us * INITSRTT_G(t4_read_reg(padap, TP_INIT_SRTT_A));
2002 	clk_info_buff->finwait2_timer =
2003 		tp_tick_us * t4_read_reg(padap, TP_FINWAIT2_TIMER_A);
2004 
2005 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2006 }
2007 
2008 int cudbg_collect_pcie_indirect(struct cudbg_init *pdbg_init,
2009 				struct cudbg_buffer *dbg_buff,
2010 				struct cudbg_error *cudbg_err)
2011 {
2012 	struct adapter *padap = pdbg_init->adap;
2013 	struct cudbg_buffer temp_buff = { 0 };
2014 	struct ireg_buf *ch_pcie;
2015 	int i, rc, n;
2016 	u32 size;
2017 
2018 	n = sizeof(t5_pcie_pdbg_array) / (IREG_NUM_ELEM * sizeof(u32));
2019 	size = sizeof(struct ireg_buf) * n * 2;
2020 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2021 	if (rc)
2022 		return rc;
2023 
2024 	ch_pcie = (struct ireg_buf *)temp_buff.data;
2025 	/* PCIE_PDBG */
2026 	for (i = 0; i < n; i++) {
2027 		struct ireg_field *pcie_pio = &ch_pcie->tp_pio;
2028 		u32 *buff = ch_pcie->outbuf;
2029 
2030 		pcie_pio->ireg_addr = t5_pcie_pdbg_array[i][0];
2031 		pcie_pio->ireg_data = t5_pcie_pdbg_array[i][1];
2032 		pcie_pio->ireg_local_offset = t5_pcie_pdbg_array[i][2];
2033 		pcie_pio->ireg_offset_range = t5_pcie_pdbg_array[i][3];
2034 		t4_read_indirect(padap,
2035 				 pcie_pio->ireg_addr,
2036 				 pcie_pio->ireg_data,
2037 				 buff,
2038 				 pcie_pio->ireg_offset_range,
2039 				 pcie_pio->ireg_local_offset);
2040 		ch_pcie++;
2041 	}
2042 
2043 	/* PCIE_CDBG */
2044 	n = sizeof(t5_pcie_cdbg_array) / (IREG_NUM_ELEM * sizeof(u32));
2045 	for (i = 0; i < n; i++) {
2046 		struct ireg_field *pcie_pio = &ch_pcie->tp_pio;
2047 		u32 *buff = ch_pcie->outbuf;
2048 
2049 		pcie_pio->ireg_addr = t5_pcie_cdbg_array[i][0];
2050 		pcie_pio->ireg_data = t5_pcie_cdbg_array[i][1];
2051 		pcie_pio->ireg_local_offset = t5_pcie_cdbg_array[i][2];
2052 		pcie_pio->ireg_offset_range = t5_pcie_cdbg_array[i][3];
2053 		t4_read_indirect(padap,
2054 				 pcie_pio->ireg_addr,
2055 				 pcie_pio->ireg_data,
2056 				 buff,
2057 				 pcie_pio->ireg_offset_range,
2058 				 pcie_pio->ireg_local_offset);
2059 		ch_pcie++;
2060 	}
2061 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2062 }
2063 
2064 int cudbg_collect_pm_indirect(struct cudbg_init *pdbg_init,
2065 			      struct cudbg_buffer *dbg_buff,
2066 			      struct cudbg_error *cudbg_err)
2067 {
2068 	struct adapter *padap = pdbg_init->adap;
2069 	struct cudbg_buffer temp_buff = { 0 };
2070 	struct ireg_buf *ch_pm;
2071 	int i, rc, n;
2072 	u32 size;
2073 
2074 	n = sizeof(t5_pm_rx_array) / (IREG_NUM_ELEM * sizeof(u32));
2075 	size = sizeof(struct ireg_buf) * n * 2;
2076 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2077 	if (rc)
2078 		return rc;
2079 
2080 	ch_pm = (struct ireg_buf *)temp_buff.data;
2081 	/* PM_RX */
2082 	for (i = 0; i < n; i++) {
2083 		struct ireg_field *pm_pio = &ch_pm->tp_pio;
2084 		u32 *buff = ch_pm->outbuf;
2085 
2086 		pm_pio->ireg_addr = t5_pm_rx_array[i][0];
2087 		pm_pio->ireg_data = t5_pm_rx_array[i][1];
2088 		pm_pio->ireg_local_offset = t5_pm_rx_array[i][2];
2089 		pm_pio->ireg_offset_range = t5_pm_rx_array[i][3];
2090 		t4_read_indirect(padap,
2091 				 pm_pio->ireg_addr,
2092 				 pm_pio->ireg_data,
2093 				 buff,
2094 				 pm_pio->ireg_offset_range,
2095 				 pm_pio->ireg_local_offset);
2096 		ch_pm++;
2097 	}
2098 
2099 	/* PM_TX */
2100 	n = sizeof(t5_pm_tx_array) / (IREG_NUM_ELEM * sizeof(u32));
2101 	for (i = 0; i < n; i++) {
2102 		struct ireg_field *pm_pio = &ch_pm->tp_pio;
2103 		u32 *buff = ch_pm->outbuf;
2104 
2105 		pm_pio->ireg_addr = t5_pm_tx_array[i][0];
2106 		pm_pio->ireg_data = t5_pm_tx_array[i][1];
2107 		pm_pio->ireg_local_offset = t5_pm_tx_array[i][2];
2108 		pm_pio->ireg_offset_range = t5_pm_tx_array[i][3];
2109 		t4_read_indirect(padap,
2110 				 pm_pio->ireg_addr,
2111 				 pm_pio->ireg_data,
2112 				 buff,
2113 				 pm_pio->ireg_offset_range,
2114 				 pm_pio->ireg_local_offset);
2115 		ch_pm++;
2116 	}
2117 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2118 }
2119 
2120 int cudbg_collect_tid(struct cudbg_init *pdbg_init,
2121 		      struct cudbg_buffer *dbg_buff,
2122 		      struct cudbg_error *cudbg_err)
2123 {
2124 	struct adapter *padap = pdbg_init->adap;
2125 	struct cudbg_tid_info_region_rev1 *tid1;
2126 	struct cudbg_buffer temp_buff = { 0 };
2127 	struct cudbg_tid_info_region *tid;
2128 	u32 para[2], val[2];
2129 	int rc;
2130 
2131 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
2132 			    sizeof(struct cudbg_tid_info_region_rev1),
2133 			    &temp_buff);
2134 	if (rc)
2135 		return rc;
2136 
2137 	tid1 = (struct cudbg_tid_info_region_rev1 *)temp_buff.data;
2138 	tid = &tid1->tid;
2139 	tid1->ver_hdr.signature = CUDBG_ENTITY_SIGNATURE;
2140 	tid1->ver_hdr.revision = CUDBG_TID_INFO_REV;
2141 	tid1->ver_hdr.size = sizeof(struct cudbg_tid_info_region_rev1) -
2142 			     sizeof(struct cudbg_ver_hdr);
2143 
2144 	/* If firmware is not attached/alive, use backdoor register
2145 	 * access to collect dump.
2146 	 */
2147 	if (!is_fw_attached(pdbg_init))
2148 		goto fill_tid;
2149 
2150 #define FW_PARAM_PFVF_A(param) \
2151 	(FW_PARAMS_MNEM_V(FW_PARAMS_MNEM_PFVF) | \
2152 	 FW_PARAMS_PARAM_X_V(FW_PARAMS_PARAM_PFVF_##param) | \
2153 	 FW_PARAMS_PARAM_Y_V(0) | \
2154 	 FW_PARAMS_PARAM_Z_V(0))
2155 
2156 	para[0] = FW_PARAM_PFVF_A(ETHOFLD_START);
2157 	para[1] = FW_PARAM_PFVF_A(ETHOFLD_END);
2158 	rc = t4_query_params(padap, padap->mbox, padap->pf, 0, 2, para, val);
2159 	if (rc <  0) {
2160 		cudbg_err->sys_err = rc;
2161 		cudbg_put_buff(pdbg_init, &temp_buff);
2162 		return rc;
2163 	}
2164 	tid->uotid_base = val[0];
2165 	tid->nuotids = val[1] - val[0] + 1;
2166 
2167 	if (is_t5(padap->params.chip)) {
2168 		tid->sb = t4_read_reg(padap, LE_DB_SERVER_INDEX_A) / 4;
2169 	} else if (is_t6(padap->params.chip)) {
2170 		tid1->tid_start =
2171 			t4_read_reg(padap, LE_DB_ACTIVE_TABLE_START_INDEX_A);
2172 		tid->sb = t4_read_reg(padap, LE_DB_SRVR_START_INDEX_A);
2173 
2174 		para[0] = FW_PARAM_PFVF_A(HPFILTER_START);
2175 		para[1] = FW_PARAM_PFVF_A(HPFILTER_END);
2176 		rc = t4_query_params(padap, padap->mbox, padap->pf, 0, 2,
2177 				     para, val);
2178 		if (rc < 0) {
2179 			cudbg_err->sys_err = rc;
2180 			cudbg_put_buff(pdbg_init, &temp_buff);
2181 			return rc;
2182 		}
2183 		tid->hpftid_base = val[0];
2184 		tid->nhpftids = val[1] - val[0] + 1;
2185 	}
2186 
2187 #undef FW_PARAM_PFVF_A
2188 
2189 fill_tid:
2190 	tid->ntids = padap->tids.ntids;
2191 	tid->nstids = padap->tids.nstids;
2192 	tid->stid_base = padap->tids.stid_base;
2193 	tid->hash_base = padap->tids.hash_base;
2194 
2195 	tid->natids = padap->tids.natids;
2196 	tid->nftids = padap->tids.nftids;
2197 	tid->ftid_base = padap->tids.ftid_base;
2198 	tid->aftid_base = padap->tids.aftid_base;
2199 	tid->aftid_end = padap->tids.aftid_end;
2200 
2201 	tid->sftid_base = padap->tids.sftid_base;
2202 	tid->nsftids = padap->tids.nsftids;
2203 
2204 	tid->flags = padap->flags;
2205 	tid->le_db_conf = t4_read_reg(padap, LE_DB_CONFIG_A);
2206 	tid->ip_users = t4_read_reg(padap, LE_DB_ACT_CNT_IPV4_A);
2207 	tid->ipv6_users = t4_read_reg(padap, LE_DB_ACT_CNT_IPV6_A);
2208 
2209 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2210 }
2211 
2212 int cudbg_collect_pcie_config(struct cudbg_init *pdbg_init,
2213 			      struct cudbg_buffer *dbg_buff,
2214 			      struct cudbg_error *cudbg_err)
2215 {
2216 	struct adapter *padap = pdbg_init->adap;
2217 	struct cudbg_buffer temp_buff = { 0 };
2218 	u32 size, *value, j;
2219 	int i, rc, n;
2220 
2221 	size = sizeof(u32) * CUDBG_NUM_PCIE_CONFIG_REGS;
2222 	n = sizeof(t5_pcie_config_array) / (2 * sizeof(u32));
2223 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2224 	if (rc)
2225 		return rc;
2226 
2227 	value = (u32 *)temp_buff.data;
2228 	for (i = 0; i < n; i++) {
2229 		for (j = t5_pcie_config_array[i][0];
2230 		     j <= t5_pcie_config_array[i][1]; j += 4) {
2231 			t4_hw_pci_read_cfg4(padap, j, value);
2232 			value++;
2233 		}
2234 	}
2235 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2236 }
2237 
2238 static int cudbg_sge_ctxt_check_valid(u32 *buf, int type)
2239 {
2240 	int index, bit, bit_pos = 0;
2241 
2242 	switch (type) {
2243 	case CTXT_EGRESS:
2244 		bit_pos = 176;
2245 		break;
2246 	case CTXT_INGRESS:
2247 		bit_pos = 141;
2248 		break;
2249 	case CTXT_FLM:
2250 		bit_pos = 89;
2251 		break;
2252 	}
2253 	index = bit_pos / 32;
2254 	bit =  bit_pos % 32;
2255 	return buf[index] & (1U << bit);
2256 }
2257 
2258 static int cudbg_get_ctxt_region_info(struct adapter *padap,
2259 				      struct cudbg_region_info *ctx_info,
2260 				      u8 *mem_type)
2261 {
2262 	struct cudbg_mem_desc mem_desc;
2263 	struct cudbg_meminfo meminfo;
2264 	u32 i, j, value, found;
2265 	u8 flq;
2266 	int rc;
2267 
2268 	rc = cudbg_fill_meminfo(padap, &meminfo);
2269 	if (rc)
2270 		return rc;
2271 
2272 	/* Get EGRESS and INGRESS context region size */
2273 	for (i = CTXT_EGRESS; i <= CTXT_INGRESS; i++) {
2274 		found = 0;
2275 		memset(&mem_desc, 0, sizeof(struct cudbg_mem_desc));
2276 		for (j = 0; j < ARRAY_SIZE(meminfo.avail); j++) {
2277 			rc = cudbg_get_mem_region(padap, &meminfo, j,
2278 						  cudbg_region[i],
2279 						  &mem_desc);
2280 			if (!rc) {
2281 				found = 1;
2282 				rc = cudbg_get_mem_relative(padap, &meminfo, j,
2283 							    &mem_desc.base,
2284 							    &mem_desc.limit);
2285 				if (rc) {
2286 					ctx_info[i].exist = false;
2287 					break;
2288 				}
2289 				ctx_info[i].exist = true;
2290 				ctx_info[i].start = mem_desc.base;
2291 				ctx_info[i].end = mem_desc.limit;
2292 				mem_type[i] = j;
2293 				break;
2294 			}
2295 		}
2296 		if (!found)
2297 			ctx_info[i].exist = false;
2298 	}
2299 
2300 	/* Get FLM and CNM max qid. */
2301 	value = t4_read_reg(padap, SGE_FLM_CFG_A);
2302 
2303 	/* Get number of data freelist queues */
2304 	flq = HDRSTARTFLQ_G(value);
2305 	ctx_info[CTXT_FLM].exist = true;
2306 	ctx_info[CTXT_FLM].end = (CUDBG_MAX_FL_QIDS >> flq) * SGE_CTXT_SIZE;
2307 
2308 	/* The number of CONM contexts are same as number of freelist
2309 	 * queues.
2310 	 */
2311 	ctx_info[CTXT_CNM].exist = true;
2312 	ctx_info[CTXT_CNM].end = ctx_info[CTXT_FLM].end;
2313 
2314 	return 0;
2315 }
2316 
2317 int cudbg_dump_context_size(struct adapter *padap)
2318 {
2319 	struct cudbg_region_info region_info[CTXT_CNM + 1] = { {0} };
2320 	u8 mem_type[CTXT_INGRESS + 1] = { 0 };
2321 	u32 i, size = 0;
2322 	int rc;
2323 
2324 	/* Get max valid qid for each type of queue */
2325 	rc = cudbg_get_ctxt_region_info(padap, region_info, mem_type);
2326 	if (rc)
2327 		return rc;
2328 
2329 	for (i = 0; i < CTXT_CNM; i++) {
2330 		if (!region_info[i].exist) {
2331 			if (i == CTXT_EGRESS || i == CTXT_INGRESS)
2332 				size += CUDBG_LOWMEM_MAX_CTXT_QIDS *
2333 					SGE_CTXT_SIZE;
2334 			continue;
2335 		}
2336 
2337 		size += (region_info[i].end - region_info[i].start + 1) /
2338 			SGE_CTXT_SIZE;
2339 	}
2340 	return size * sizeof(struct cudbg_ch_cntxt);
2341 }
2342 
2343 static void cudbg_read_sge_ctxt(struct cudbg_init *pdbg_init, u32 cid,
2344 				enum ctxt_type ctype, u32 *data)
2345 {
2346 	struct adapter *padap = pdbg_init->adap;
2347 	int rc = -1;
2348 
2349 	/* Under heavy traffic, the SGE Queue contexts registers will be
2350 	 * frequently accessed by firmware.
2351 	 *
2352 	 * To avoid conflicts with firmware, always ask firmware to fetch
2353 	 * the SGE Queue contexts via mailbox. On failure, fallback to
2354 	 * accessing hardware registers directly.
2355 	 */
2356 	if (is_fw_attached(pdbg_init))
2357 		rc = t4_sge_ctxt_rd(padap, padap->mbox, cid, ctype, data);
2358 	if (rc)
2359 		t4_sge_ctxt_rd_bd(padap, cid, ctype, data);
2360 }
2361 
2362 static void cudbg_get_sge_ctxt_fw(struct cudbg_init *pdbg_init, u32 max_qid,
2363 				  u8 ctxt_type,
2364 				  struct cudbg_ch_cntxt **out_buff)
2365 {
2366 	struct cudbg_ch_cntxt *buff = *out_buff;
2367 	int rc;
2368 	u32 j;
2369 
2370 	for (j = 0; j < max_qid; j++) {
2371 		cudbg_read_sge_ctxt(pdbg_init, j, ctxt_type, buff->data);
2372 		rc = cudbg_sge_ctxt_check_valid(buff->data, ctxt_type);
2373 		if (!rc)
2374 			continue;
2375 
2376 		buff->cntxt_type = ctxt_type;
2377 		buff->cntxt_id = j;
2378 		buff++;
2379 		if (ctxt_type == CTXT_FLM) {
2380 			cudbg_read_sge_ctxt(pdbg_init, j, CTXT_CNM, buff->data);
2381 			buff->cntxt_type = CTXT_CNM;
2382 			buff->cntxt_id = j;
2383 			buff++;
2384 		}
2385 	}
2386 
2387 	*out_buff = buff;
2388 }
2389 
2390 int cudbg_collect_dump_context(struct cudbg_init *pdbg_init,
2391 			       struct cudbg_buffer *dbg_buff,
2392 			       struct cudbg_error *cudbg_err)
2393 {
2394 	struct cudbg_region_info region_info[CTXT_CNM + 1] = { {0} };
2395 	struct adapter *padap = pdbg_init->adap;
2396 	u32 j, size, max_ctx_size, max_ctx_qid;
2397 	u8 mem_type[CTXT_INGRESS + 1] = { 0 };
2398 	struct cudbg_buffer temp_buff = { 0 };
2399 	struct cudbg_ch_cntxt *buff;
2400 	u8 *ctx_buf;
2401 	u8 i, k;
2402 	int rc;
2403 
2404 	/* Get max valid qid for each type of queue */
2405 	rc = cudbg_get_ctxt_region_info(padap, region_info, mem_type);
2406 	if (rc)
2407 		return rc;
2408 
2409 	rc = cudbg_dump_context_size(padap);
2410 	if (rc <= 0)
2411 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
2412 
2413 	size = rc;
2414 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2415 	if (rc)
2416 		return rc;
2417 
2418 	/* Get buffer with enough space to read the biggest context
2419 	 * region in memory.
2420 	 */
2421 	max_ctx_size = max(region_info[CTXT_EGRESS].end -
2422 			   region_info[CTXT_EGRESS].start + 1,
2423 			   region_info[CTXT_INGRESS].end -
2424 			   region_info[CTXT_INGRESS].start + 1);
2425 
2426 	ctx_buf = kvzalloc(max_ctx_size, GFP_KERNEL);
2427 	if (!ctx_buf) {
2428 		cudbg_put_buff(pdbg_init, &temp_buff);
2429 		return -ENOMEM;
2430 	}
2431 
2432 	buff = (struct cudbg_ch_cntxt *)temp_buff.data;
2433 
2434 	/* Collect EGRESS and INGRESS context data.
2435 	 * In case of failures, fallback to collecting via FW or
2436 	 * backdoor access.
2437 	 */
2438 	for (i = CTXT_EGRESS; i <= CTXT_INGRESS; i++) {
2439 		if (!region_info[i].exist) {
2440 			max_ctx_qid = CUDBG_LOWMEM_MAX_CTXT_QIDS;
2441 			cudbg_get_sge_ctxt_fw(pdbg_init, max_ctx_qid, i,
2442 					      &buff);
2443 			continue;
2444 		}
2445 
2446 		max_ctx_size = region_info[i].end - region_info[i].start + 1;
2447 		max_ctx_qid = max_ctx_size / SGE_CTXT_SIZE;
2448 
2449 		/* If firmware is not attached/alive, use backdoor register
2450 		 * access to collect dump.
2451 		 */
2452 		if (is_fw_attached(pdbg_init)) {
2453 			t4_sge_ctxt_flush(padap, padap->mbox, i);
2454 
2455 			rc = t4_memory_rw(padap, MEMWIN_NIC, mem_type[i],
2456 					  region_info[i].start, max_ctx_size,
2457 					  (__be32 *)ctx_buf, 1);
2458 		}
2459 
2460 		if (rc || !is_fw_attached(pdbg_init)) {
2461 			max_ctx_qid = CUDBG_LOWMEM_MAX_CTXT_QIDS;
2462 			cudbg_get_sge_ctxt_fw(pdbg_init, max_ctx_qid, i,
2463 					      &buff);
2464 			continue;
2465 		}
2466 
2467 		for (j = 0; j < max_ctx_qid; j++) {
2468 			__be64 *dst_off;
2469 			u64 *src_off;
2470 
2471 			src_off = (u64 *)(ctx_buf + j * SGE_CTXT_SIZE);
2472 			dst_off = (__be64 *)buff->data;
2473 
2474 			/* The data is stored in 64-bit cpu order.  Convert it
2475 			 * to big endian before parsing.
2476 			 */
2477 			for (k = 0; k < SGE_CTXT_SIZE / sizeof(u64); k++)
2478 				dst_off[k] = cpu_to_be64(src_off[k]);
2479 
2480 			rc = cudbg_sge_ctxt_check_valid(buff->data, i);
2481 			if (!rc)
2482 				continue;
2483 
2484 			buff->cntxt_type = i;
2485 			buff->cntxt_id = j;
2486 			buff++;
2487 		}
2488 	}
2489 
2490 	kvfree(ctx_buf);
2491 
2492 	/* Collect FREELIST and CONGESTION MANAGER contexts */
2493 	max_ctx_size = region_info[CTXT_FLM].end -
2494 		       region_info[CTXT_FLM].start + 1;
2495 	max_ctx_qid = max_ctx_size / SGE_CTXT_SIZE;
2496 	/* Since FLM and CONM are 1-to-1 mapped, the below function
2497 	 * will fetch both FLM and CONM contexts.
2498 	 */
2499 	cudbg_get_sge_ctxt_fw(pdbg_init, max_ctx_qid, CTXT_FLM, &buff);
2500 
2501 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2502 }
2503 
2504 static inline void cudbg_tcamxy2valmask(u64 x, u64 y, u8 *addr, u64 *mask)
2505 {
2506 	*mask = x | y;
2507 	y = (__force u64)cpu_to_be64(y);
2508 	memcpy(addr, (char *)&y + 2, ETH_ALEN);
2509 }
2510 
2511 static void cudbg_mps_rpl_backdoor(struct adapter *padap,
2512 				   struct fw_ldst_mps_rplc *mps_rplc)
2513 {
2514 	if (is_t5(padap->params.chip)) {
2515 		mps_rplc->rplc255_224 = htonl(t4_read_reg(padap,
2516 							  MPS_VF_RPLCT_MAP3_A));
2517 		mps_rplc->rplc223_192 = htonl(t4_read_reg(padap,
2518 							  MPS_VF_RPLCT_MAP2_A));
2519 		mps_rplc->rplc191_160 = htonl(t4_read_reg(padap,
2520 							  MPS_VF_RPLCT_MAP1_A));
2521 		mps_rplc->rplc159_128 = htonl(t4_read_reg(padap,
2522 							  MPS_VF_RPLCT_MAP0_A));
2523 	} else {
2524 		mps_rplc->rplc255_224 = htonl(t4_read_reg(padap,
2525 							  MPS_VF_RPLCT_MAP7_A));
2526 		mps_rplc->rplc223_192 = htonl(t4_read_reg(padap,
2527 							  MPS_VF_RPLCT_MAP6_A));
2528 		mps_rplc->rplc191_160 = htonl(t4_read_reg(padap,
2529 							  MPS_VF_RPLCT_MAP5_A));
2530 		mps_rplc->rplc159_128 = htonl(t4_read_reg(padap,
2531 							  MPS_VF_RPLCT_MAP4_A));
2532 	}
2533 	mps_rplc->rplc127_96 = htonl(t4_read_reg(padap, MPS_VF_RPLCT_MAP3_A));
2534 	mps_rplc->rplc95_64 = htonl(t4_read_reg(padap, MPS_VF_RPLCT_MAP2_A));
2535 	mps_rplc->rplc63_32 = htonl(t4_read_reg(padap, MPS_VF_RPLCT_MAP1_A));
2536 	mps_rplc->rplc31_0 = htonl(t4_read_reg(padap, MPS_VF_RPLCT_MAP0_A));
2537 }
2538 
2539 static int cudbg_collect_tcam_index(struct cudbg_init *pdbg_init,
2540 				    struct cudbg_mps_tcam *tcam, u32 idx)
2541 {
2542 	struct adapter *padap = pdbg_init->adap;
2543 	u64 tcamy, tcamx, val;
2544 	u32 ctl, data2;
2545 	int rc = 0;
2546 
2547 	if (CHELSIO_CHIP_VERSION(padap->params.chip) >= CHELSIO_T6) {
2548 		/* CtlReqID   - 1: use Host Driver Requester ID
2549 		 * CtlCmdType - 0: Read, 1: Write
2550 		 * CtlTcamSel - 0: TCAM0, 1: TCAM1
2551 		 * CtlXYBitSel- 0: Y bit, 1: X bit
2552 		 */
2553 
2554 		/* Read tcamy */
2555 		ctl = CTLREQID_V(1) | CTLCMDTYPE_V(0) | CTLXYBITSEL_V(0);
2556 		if (idx < 256)
2557 			ctl |= CTLTCAMINDEX_V(idx) | CTLTCAMSEL_V(0);
2558 		else
2559 			ctl |= CTLTCAMINDEX_V(idx - 256) | CTLTCAMSEL_V(1);
2560 
2561 		t4_write_reg(padap, MPS_CLS_TCAM_DATA2_CTL_A, ctl);
2562 		val = t4_read_reg(padap, MPS_CLS_TCAM_RDATA1_REQ_ID1_A);
2563 		tcamy = DMACH_G(val) << 32;
2564 		tcamy |= t4_read_reg(padap, MPS_CLS_TCAM_RDATA0_REQ_ID1_A);
2565 		data2 = t4_read_reg(padap, MPS_CLS_TCAM_RDATA2_REQ_ID1_A);
2566 		tcam->lookup_type = DATALKPTYPE_G(data2);
2567 
2568 		/* 0 - Outer header, 1 - Inner header
2569 		 * [71:48] bit locations are overloaded for
2570 		 * outer vs. inner lookup types.
2571 		 */
2572 		if (tcam->lookup_type && tcam->lookup_type != DATALKPTYPE_M) {
2573 			/* Inner header VNI */
2574 			tcam->vniy = (data2 & DATAVIDH2_F) | DATAVIDH1_G(data2);
2575 			tcam->vniy = (tcam->vniy << 16) | VIDL_G(val);
2576 			tcam->dip_hit = data2 & DATADIPHIT_F;
2577 		} else {
2578 			tcam->vlan_vld = data2 & DATAVIDH2_F;
2579 			tcam->ivlan = VIDL_G(val);
2580 		}
2581 
2582 		tcam->port_num = DATAPORTNUM_G(data2);
2583 
2584 		/* Read tcamx. Change the control param */
2585 		ctl |= CTLXYBITSEL_V(1);
2586 		t4_write_reg(padap, MPS_CLS_TCAM_DATA2_CTL_A, ctl);
2587 		val = t4_read_reg(padap, MPS_CLS_TCAM_RDATA1_REQ_ID1_A);
2588 		tcamx = DMACH_G(val) << 32;
2589 		tcamx |= t4_read_reg(padap, MPS_CLS_TCAM_RDATA0_REQ_ID1_A);
2590 		data2 = t4_read_reg(padap, MPS_CLS_TCAM_RDATA2_REQ_ID1_A);
2591 		if (tcam->lookup_type && tcam->lookup_type != DATALKPTYPE_M) {
2592 			/* Inner header VNI mask */
2593 			tcam->vnix = (data2 & DATAVIDH2_F) | DATAVIDH1_G(data2);
2594 			tcam->vnix = (tcam->vnix << 16) | VIDL_G(val);
2595 		}
2596 	} else {
2597 		tcamy = t4_read_reg64(padap, MPS_CLS_TCAM_Y_L(idx));
2598 		tcamx = t4_read_reg64(padap, MPS_CLS_TCAM_X_L(idx));
2599 	}
2600 
2601 	/* If no entry, return */
2602 	if (tcamx & tcamy)
2603 		return rc;
2604 
2605 	tcam->cls_lo = t4_read_reg(padap, MPS_CLS_SRAM_L(idx));
2606 	tcam->cls_hi = t4_read_reg(padap, MPS_CLS_SRAM_H(idx));
2607 
2608 	if (is_t5(padap->params.chip))
2609 		tcam->repli = (tcam->cls_lo & REPLICATE_F);
2610 	else if (is_t6(padap->params.chip))
2611 		tcam->repli = (tcam->cls_lo & T6_REPLICATE_F);
2612 
2613 	if (tcam->repli) {
2614 		struct fw_ldst_cmd ldst_cmd;
2615 		struct fw_ldst_mps_rplc mps_rplc;
2616 
2617 		memset(&ldst_cmd, 0, sizeof(ldst_cmd));
2618 		ldst_cmd.op_to_addrspace =
2619 			htonl(FW_CMD_OP_V(FW_LDST_CMD) |
2620 			      FW_CMD_REQUEST_F | FW_CMD_READ_F |
2621 			      FW_LDST_CMD_ADDRSPACE_V(FW_LDST_ADDRSPC_MPS));
2622 		ldst_cmd.cycles_to_len16 = htonl(FW_LEN16(ldst_cmd));
2623 		ldst_cmd.u.mps.rplc.fid_idx =
2624 			htons(FW_LDST_CMD_FID_V(FW_LDST_MPS_RPLC) |
2625 			      FW_LDST_CMD_IDX_V(idx));
2626 
2627 		/* If firmware is not attached/alive, use backdoor register
2628 		 * access to collect dump.
2629 		 */
2630 		if (is_fw_attached(pdbg_init))
2631 			rc = t4_wr_mbox(padap, padap->mbox, &ldst_cmd,
2632 					sizeof(ldst_cmd), &ldst_cmd);
2633 
2634 		if (rc || !is_fw_attached(pdbg_init)) {
2635 			cudbg_mps_rpl_backdoor(padap, &mps_rplc);
2636 			/* Ignore error since we collected directly from
2637 			 * reading registers.
2638 			 */
2639 			rc = 0;
2640 		} else {
2641 			mps_rplc = ldst_cmd.u.mps.rplc;
2642 		}
2643 
2644 		tcam->rplc[0] = ntohl(mps_rplc.rplc31_0);
2645 		tcam->rplc[1] = ntohl(mps_rplc.rplc63_32);
2646 		tcam->rplc[2] = ntohl(mps_rplc.rplc95_64);
2647 		tcam->rplc[3] = ntohl(mps_rplc.rplc127_96);
2648 		if (padap->params.arch.mps_rplc_size > CUDBG_MAX_RPLC_SIZE) {
2649 			tcam->rplc[4] = ntohl(mps_rplc.rplc159_128);
2650 			tcam->rplc[5] = ntohl(mps_rplc.rplc191_160);
2651 			tcam->rplc[6] = ntohl(mps_rplc.rplc223_192);
2652 			tcam->rplc[7] = ntohl(mps_rplc.rplc255_224);
2653 		}
2654 	}
2655 	cudbg_tcamxy2valmask(tcamx, tcamy, tcam->addr, &tcam->mask);
2656 	tcam->idx = idx;
2657 	tcam->rplc_size = padap->params.arch.mps_rplc_size;
2658 	return rc;
2659 }
2660 
2661 int cudbg_collect_mps_tcam(struct cudbg_init *pdbg_init,
2662 			   struct cudbg_buffer *dbg_buff,
2663 			   struct cudbg_error *cudbg_err)
2664 {
2665 	struct adapter *padap = pdbg_init->adap;
2666 	struct cudbg_buffer temp_buff = { 0 };
2667 	u32 size = 0, i, n, total_size = 0;
2668 	struct cudbg_mps_tcam *tcam;
2669 	int rc;
2670 
2671 	n = padap->params.arch.mps_tcam_size;
2672 	size = sizeof(struct cudbg_mps_tcam) * n;
2673 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2674 	if (rc)
2675 		return rc;
2676 
2677 	tcam = (struct cudbg_mps_tcam *)temp_buff.data;
2678 	for (i = 0; i < n; i++) {
2679 		rc = cudbg_collect_tcam_index(pdbg_init, tcam, i);
2680 		if (rc) {
2681 			cudbg_err->sys_err = rc;
2682 			cudbg_put_buff(pdbg_init, &temp_buff);
2683 			return rc;
2684 		}
2685 		total_size += sizeof(struct cudbg_mps_tcam);
2686 		tcam++;
2687 	}
2688 
2689 	if (!total_size) {
2690 		rc = CUDBG_SYSTEM_ERROR;
2691 		cudbg_err->sys_err = rc;
2692 		cudbg_put_buff(pdbg_init, &temp_buff);
2693 		return rc;
2694 	}
2695 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2696 }
2697 
2698 int cudbg_collect_vpd_data(struct cudbg_init *pdbg_init,
2699 			   struct cudbg_buffer *dbg_buff,
2700 			   struct cudbg_error *cudbg_err)
2701 {
2702 	struct adapter *padap = pdbg_init->adap;
2703 	struct cudbg_buffer temp_buff = { 0 };
2704 	char vpd_str[CUDBG_VPD_VER_LEN + 1];
2705 	struct cudbg_vpd_data *vpd_data;
2706 	struct vpd_params vpd = { 0 };
2707 	u32 vpd_vers, fw_vers;
2708 	int rc;
2709 
2710 	rc = t4_get_raw_vpd_params(padap, &vpd);
2711 	if (rc)
2712 		return rc;
2713 
2714 	rc = t4_get_fw_version(padap, &fw_vers);
2715 	if (rc)
2716 		return rc;
2717 
2718 	rc = cudbg_read_vpd_reg(padap, CUDBG_VPD_VER_ADDR, CUDBG_VPD_VER_LEN,
2719 				vpd_str);
2720 	if (rc)
2721 		return rc;
2722 
2723 	vpd_str[CUDBG_VPD_VER_LEN] = '\0';
2724 	rc = kstrtouint(vpd_str, 0, &vpd_vers);
2725 	if (rc)
2726 		return rc;
2727 
2728 	rc = cudbg_get_buff(pdbg_init, dbg_buff, sizeof(struct cudbg_vpd_data),
2729 			    &temp_buff);
2730 	if (rc)
2731 		return rc;
2732 
2733 	vpd_data = (struct cudbg_vpd_data *)temp_buff.data;
2734 	memcpy(vpd_data->sn, vpd.sn, SERNUM_LEN + 1);
2735 	memcpy(vpd_data->bn, vpd.pn, PN_LEN + 1);
2736 	memcpy(vpd_data->na, vpd.na, MACADDR_LEN + 1);
2737 	memcpy(vpd_data->mn, vpd.id, ID_LEN + 1);
2738 	vpd_data->scfg_vers = t4_read_reg(padap, PCIE_STATIC_SPARE2_A);
2739 	vpd_data->vpd_vers = vpd_vers;
2740 	vpd_data->fw_major = FW_HDR_FW_VER_MAJOR_G(fw_vers);
2741 	vpd_data->fw_minor = FW_HDR_FW_VER_MINOR_G(fw_vers);
2742 	vpd_data->fw_micro = FW_HDR_FW_VER_MICRO_G(fw_vers);
2743 	vpd_data->fw_build = FW_HDR_FW_VER_BUILD_G(fw_vers);
2744 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2745 }
2746 
2747 static int cudbg_read_tid(struct cudbg_init *pdbg_init, u32 tid,
2748 			  struct cudbg_tid_data *tid_data)
2749 {
2750 	struct adapter *padap = pdbg_init->adap;
2751 	int i, cmd_retry = 8;
2752 	u32 val;
2753 
2754 	/* Fill REQ_DATA regs with 0's */
2755 	for (i = 0; i < NUM_LE_DB_DBGI_REQ_DATA_INSTANCES; i++)
2756 		t4_write_reg(padap, LE_DB_DBGI_REQ_DATA_A + (i << 2), 0);
2757 
2758 	/* Write DBIG command */
2759 	val = DBGICMD_V(4) | DBGITID_V(tid);
2760 	t4_write_reg(padap, LE_DB_DBGI_REQ_TCAM_CMD_A, val);
2761 	tid_data->dbig_cmd = val;
2762 
2763 	val = DBGICMDSTRT_F | DBGICMDMODE_V(1); /* LE mode */
2764 	t4_write_reg(padap, LE_DB_DBGI_CONFIG_A, val);
2765 	tid_data->dbig_conf = val;
2766 
2767 	/* Poll the DBGICMDBUSY bit */
2768 	val = 1;
2769 	while (val) {
2770 		val = t4_read_reg(padap, LE_DB_DBGI_CONFIG_A);
2771 		val = val & DBGICMDBUSY_F;
2772 		cmd_retry--;
2773 		if (!cmd_retry)
2774 			return CUDBG_SYSTEM_ERROR;
2775 	}
2776 
2777 	/* Check RESP status */
2778 	val = t4_read_reg(padap, LE_DB_DBGI_RSP_STATUS_A);
2779 	tid_data->dbig_rsp_stat = val;
2780 	if (!(val & 1))
2781 		return CUDBG_SYSTEM_ERROR;
2782 
2783 	/* Read RESP data */
2784 	for (i = 0; i < NUM_LE_DB_DBGI_RSP_DATA_INSTANCES; i++)
2785 		tid_data->data[i] = t4_read_reg(padap,
2786 						LE_DB_DBGI_RSP_DATA_A +
2787 						(i << 2));
2788 	tid_data->tid = tid;
2789 	return 0;
2790 }
2791 
2792 static int cudbg_get_le_type(u32 tid, struct cudbg_tcam tcam_region)
2793 {
2794 	int type = LE_ET_UNKNOWN;
2795 
2796 	if (tid < tcam_region.server_start)
2797 		type = LE_ET_TCAM_CON;
2798 	else if (tid < tcam_region.filter_start)
2799 		type = LE_ET_TCAM_SERVER;
2800 	else if (tid < tcam_region.clip_start)
2801 		type = LE_ET_TCAM_FILTER;
2802 	else if (tid < tcam_region.routing_start)
2803 		type = LE_ET_TCAM_CLIP;
2804 	else if (tid < tcam_region.tid_hash_base)
2805 		type = LE_ET_TCAM_ROUTING;
2806 	else if (tid < tcam_region.max_tid)
2807 		type = LE_ET_HASH_CON;
2808 	else
2809 		type = LE_ET_INVALID_TID;
2810 
2811 	return type;
2812 }
2813 
2814 static int cudbg_is_ipv6_entry(struct cudbg_tid_data *tid_data,
2815 			       struct cudbg_tcam tcam_region)
2816 {
2817 	int ipv6 = 0;
2818 	int le_type;
2819 
2820 	le_type = cudbg_get_le_type(tid_data->tid, tcam_region);
2821 	if (tid_data->tid & 1)
2822 		return 0;
2823 
2824 	if (le_type == LE_ET_HASH_CON) {
2825 		ipv6 = tid_data->data[16] & 0x8000;
2826 	} else if (le_type == LE_ET_TCAM_CON) {
2827 		ipv6 = tid_data->data[16] & 0x8000;
2828 		if (ipv6)
2829 			ipv6 = tid_data->data[9] == 0x00C00000;
2830 	} else {
2831 		ipv6 = 0;
2832 	}
2833 	return ipv6;
2834 }
2835 
2836 void cudbg_fill_le_tcam_info(struct adapter *padap,
2837 			     struct cudbg_tcam *tcam_region)
2838 {
2839 	u32 value;
2840 
2841 	/* Get the LE regions */
2842 	value = t4_read_reg(padap, LE_DB_TID_HASHBASE_A); /* hash base index */
2843 	tcam_region->tid_hash_base = value;
2844 
2845 	/* Get routing table index */
2846 	value = t4_read_reg(padap, LE_DB_ROUTING_TABLE_INDEX_A);
2847 	tcam_region->routing_start = value;
2848 
2849 	/* Get clip table index. For T6 there is separate CLIP TCAM */
2850 	if (is_t6(padap->params.chip))
2851 		value = t4_read_reg(padap, LE_DB_CLCAM_TID_BASE_A);
2852 	else
2853 		value = t4_read_reg(padap, LE_DB_CLIP_TABLE_INDEX_A);
2854 	tcam_region->clip_start = value;
2855 
2856 	/* Get filter table index */
2857 	value = t4_read_reg(padap, LE_DB_FILTER_TABLE_INDEX_A);
2858 	tcam_region->filter_start = value;
2859 
2860 	/* Get server table index */
2861 	value = t4_read_reg(padap, LE_DB_SERVER_INDEX_A);
2862 	tcam_region->server_start = value;
2863 
2864 	/* Check whether hash is enabled and calculate the max tids */
2865 	value = t4_read_reg(padap, LE_DB_CONFIG_A);
2866 	if ((value >> HASHEN_S) & 1) {
2867 		value = t4_read_reg(padap, LE_DB_HASH_CONFIG_A);
2868 		if (CHELSIO_CHIP_VERSION(padap->params.chip) > CHELSIO_T5) {
2869 			tcam_region->max_tid = (value & 0xFFFFF) +
2870 					       tcam_region->tid_hash_base;
2871 		} else {
2872 			value = HASHTIDSIZE_G(value);
2873 			value = 1 << value;
2874 			tcam_region->max_tid = value +
2875 					       tcam_region->tid_hash_base;
2876 		}
2877 	} else { /* hash not enabled */
2878 		if (is_t6(padap->params.chip))
2879 			tcam_region->max_tid = (value & ASLIPCOMPEN_F) ?
2880 					       CUDBG_MAX_TID_COMP_EN :
2881 					       CUDBG_MAX_TID_COMP_DIS;
2882 		else
2883 			tcam_region->max_tid = CUDBG_MAX_TCAM_TID;
2884 	}
2885 
2886 	if (is_t6(padap->params.chip))
2887 		tcam_region->max_tid += CUDBG_T6_CLIP;
2888 }
2889 
2890 int cudbg_collect_le_tcam(struct cudbg_init *pdbg_init,
2891 			  struct cudbg_buffer *dbg_buff,
2892 			  struct cudbg_error *cudbg_err)
2893 {
2894 	struct adapter *padap = pdbg_init->adap;
2895 	struct cudbg_buffer temp_buff = { 0 };
2896 	struct cudbg_tcam tcam_region = { 0 };
2897 	struct cudbg_tid_data *tid_data;
2898 	u32 bytes = 0;
2899 	int rc, size;
2900 	u32 i;
2901 
2902 	cudbg_fill_le_tcam_info(padap, &tcam_region);
2903 
2904 	size = sizeof(struct cudbg_tid_data) * tcam_region.max_tid;
2905 	size += sizeof(struct cudbg_tcam);
2906 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2907 	if (rc)
2908 		return rc;
2909 
2910 	memcpy(temp_buff.data, &tcam_region, sizeof(struct cudbg_tcam));
2911 	bytes = sizeof(struct cudbg_tcam);
2912 	tid_data = (struct cudbg_tid_data *)(temp_buff.data + bytes);
2913 	/* read all tid */
2914 	for (i = 0; i < tcam_region.max_tid; ) {
2915 		rc = cudbg_read_tid(pdbg_init, i, tid_data);
2916 		if (rc) {
2917 			cudbg_err->sys_warn = CUDBG_STATUS_PARTIAL_DATA;
2918 			/* Update tcam header and exit */
2919 			tcam_region.max_tid = i;
2920 			memcpy(temp_buff.data, &tcam_region,
2921 			       sizeof(struct cudbg_tcam));
2922 			goto out;
2923 		}
2924 
2925 		if (cudbg_is_ipv6_entry(tid_data, tcam_region)) {
2926 			/* T6 CLIP TCAM: ipv6 takes 4 entries */
2927 			if (is_t6(padap->params.chip) &&
2928 			    i >= tcam_region.clip_start &&
2929 			    i < tcam_region.clip_start + CUDBG_T6_CLIP)
2930 				i += 4;
2931 			else /* Main TCAM: ipv6 takes two tids */
2932 				i += 2;
2933 		} else {
2934 			i++;
2935 		}
2936 
2937 		tid_data++;
2938 		bytes += sizeof(struct cudbg_tid_data);
2939 	}
2940 
2941 out:
2942 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2943 }
2944 
2945 int cudbg_collect_cctrl(struct cudbg_init *pdbg_init,
2946 			struct cudbg_buffer *dbg_buff,
2947 			struct cudbg_error *cudbg_err)
2948 {
2949 	struct adapter *padap = pdbg_init->adap;
2950 	struct cudbg_buffer temp_buff = { 0 };
2951 	u32 size;
2952 	int rc;
2953 
2954 	size = sizeof(u16) * NMTUS * NCCTRL_WIN;
2955 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2956 	if (rc)
2957 		return rc;
2958 
2959 	t4_read_cong_tbl(padap, (void *)temp_buff.data);
2960 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
2961 }
2962 
2963 int cudbg_collect_ma_indirect(struct cudbg_init *pdbg_init,
2964 			      struct cudbg_buffer *dbg_buff,
2965 			      struct cudbg_error *cudbg_err)
2966 {
2967 	struct adapter *padap = pdbg_init->adap;
2968 	struct cudbg_buffer temp_buff = { 0 };
2969 	struct ireg_buf *ma_indr;
2970 	int i, rc, n;
2971 	u32 size, j;
2972 
2973 	if (CHELSIO_CHIP_VERSION(padap->params.chip) < CHELSIO_T6)
2974 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
2975 
2976 	n = sizeof(t6_ma_ireg_array) / (IREG_NUM_ELEM * sizeof(u32));
2977 	size = sizeof(struct ireg_buf) * n * 2;
2978 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
2979 	if (rc)
2980 		return rc;
2981 
2982 	ma_indr = (struct ireg_buf *)temp_buff.data;
2983 	for (i = 0; i < n; i++) {
2984 		struct ireg_field *ma_fli = &ma_indr->tp_pio;
2985 		u32 *buff = ma_indr->outbuf;
2986 
2987 		ma_fli->ireg_addr = t6_ma_ireg_array[i][0];
2988 		ma_fli->ireg_data = t6_ma_ireg_array[i][1];
2989 		ma_fli->ireg_local_offset = t6_ma_ireg_array[i][2];
2990 		ma_fli->ireg_offset_range = t6_ma_ireg_array[i][3];
2991 		t4_read_indirect(padap, ma_fli->ireg_addr, ma_fli->ireg_data,
2992 				 buff, ma_fli->ireg_offset_range,
2993 				 ma_fli->ireg_local_offset);
2994 		ma_indr++;
2995 	}
2996 
2997 	n = sizeof(t6_ma_ireg_array2) / (IREG_NUM_ELEM * sizeof(u32));
2998 	for (i = 0; i < n; i++) {
2999 		struct ireg_field *ma_fli = &ma_indr->tp_pio;
3000 		u32 *buff = ma_indr->outbuf;
3001 
3002 		ma_fli->ireg_addr = t6_ma_ireg_array2[i][0];
3003 		ma_fli->ireg_data = t6_ma_ireg_array2[i][1];
3004 		ma_fli->ireg_local_offset = t6_ma_ireg_array2[i][2];
3005 		for (j = 0; j < t6_ma_ireg_array2[i][3]; j++) {
3006 			t4_read_indirect(padap, ma_fli->ireg_addr,
3007 					 ma_fli->ireg_data, buff, 1,
3008 					 ma_fli->ireg_local_offset);
3009 			buff++;
3010 			ma_fli->ireg_local_offset += 0x20;
3011 		}
3012 		ma_indr++;
3013 	}
3014 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
3015 }
3016 
3017 int cudbg_collect_ulptx_la(struct cudbg_init *pdbg_init,
3018 			   struct cudbg_buffer *dbg_buff,
3019 			   struct cudbg_error *cudbg_err)
3020 {
3021 	struct adapter *padap = pdbg_init->adap;
3022 	struct cudbg_buffer temp_buff = { 0 };
3023 	struct cudbg_ulptx_la *ulptx_la_buff;
3024 	struct cudbg_ver_hdr *ver_hdr;
3025 	u32 i, j;
3026 	int rc;
3027 
3028 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
3029 			    sizeof(struct cudbg_ver_hdr) +
3030 			    sizeof(struct cudbg_ulptx_la),
3031 			    &temp_buff);
3032 	if (rc)
3033 		return rc;
3034 
3035 	ver_hdr = (struct cudbg_ver_hdr *)temp_buff.data;
3036 	ver_hdr->signature = CUDBG_ENTITY_SIGNATURE;
3037 	ver_hdr->revision = CUDBG_ULPTX_LA_REV;
3038 	ver_hdr->size = sizeof(struct cudbg_ulptx_la);
3039 
3040 	ulptx_la_buff = (struct cudbg_ulptx_la *)(temp_buff.data +
3041 						  sizeof(*ver_hdr));
3042 	for (i = 0; i < CUDBG_NUM_ULPTX; i++) {
3043 		ulptx_la_buff->rdptr[i] = t4_read_reg(padap,
3044 						      ULP_TX_LA_RDPTR_0_A +
3045 						      0x10 * i);
3046 		ulptx_la_buff->wrptr[i] = t4_read_reg(padap,
3047 						      ULP_TX_LA_WRPTR_0_A +
3048 						      0x10 * i);
3049 		ulptx_la_buff->rddata[i] = t4_read_reg(padap,
3050 						       ULP_TX_LA_RDDATA_0_A +
3051 						       0x10 * i);
3052 		for (j = 0; j < CUDBG_NUM_ULPTX_READ; j++)
3053 			ulptx_la_buff->rd_data[i][j] =
3054 				t4_read_reg(padap,
3055 					    ULP_TX_LA_RDDATA_0_A + 0x10 * i);
3056 	}
3057 
3058 	for (i = 0; i < CUDBG_NUM_ULPTX_ASIC_READ; i++) {
3059 		t4_write_reg(padap, ULP_TX_ASIC_DEBUG_CTRL_A, 0x1);
3060 		ulptx_la_buff->rdptr_asic[i] =
3061 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_CTRL_A);
3062 		ulptx_la_buff->rddata_asic[i][0] =
3063 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_0_A);
3064 		ulptx_la_buff->rddata_asic[i][1] =
3065 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_1_A);
3066 		ulptx_la_buff->rddata_asic[i][2] =
3067 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_2_A);
3068 		ulptx_la_buff->rddata_asic[i][3] =
3069 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_3_A);
3070 		ulptx_la_buff->rddata_asic[i][4] =
3071 				t4_read_reg(padap, ULP_TX_ASIC_DEBUG_4_A);
3072 		ulptx_la_buff->rddata_asic[i][5] =
3073 				t4_read_reg(padap, PM_RX_BASE_ADDR);
3074 	}
3075 
3076 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
3077 }
3078 
3079 int cudbg_collect_up_cim_indirect(struct cudbg_init *pdbg_init,
3080 				  struct cudbg_buffer *dbg_buff,
3081 				  struct cudbg_error *cudbg_err)
3082 {
3083 	struct adapter *padap = pdbg_init->adap;
3084 	struct cudbg_buffer temp_buff = { 0 };
3085 	u32 local_offset, local_range;
3086 	struct ireg_buf *up_cim;
3087 	u32 size, j, iter;
3088 	u32 instance = 0;
3089 	int i, rc, n;
3090 
3091 	if (is_t5(padap->params.chip))
3092 		n = sizeof(t5_up_cim_reg_array) /
3093 		    ((IREG_NUM_ELEM + 1) * sizeof(u32));
3094 	else if (is_t6(padap->params.chip))
3095 		n = sizeof(t6_up_cim_reg_array) /
3096 		    ((IREG_NUM_ELEM + 1) * sizeof(u32));
3097 	else
3098 		return CUDBG_STATUS_NOT_IMPLEMENTED;
3099 
3100 	size = sizeof(struct ireg_buf) * n;
3101 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
3102 	if (rc)
3103 		return rc;
3104 
3105 	up_cim = (struct ireg_buf *)temp_buff.data;
3106 	for (i = 0; i < n; i++) {
3107 		struct ireg_field *up_cim_reg = &up_cim->tp_pio;
3108 		u32 *buff = up_cim->outbuf;
3109 
3110 		if (is_t5(padap->params.chip)) {
3111 			up_cim_reg->ireg_addr = t5_up_cim_reg_array[i][0];
3112 			up_cim_reg->ireg_data = t5_up_cim_reg_array[i][1];
3113 			up_cim_reg->ireg_local_offset =
3114 						t5_up_cim_reg_array[i][2];
3115 			up_cim_reg->ireg_offset_range =
3116 						t5_up_cim_reg_array[i][3];
3117 			instance = t5_up_cim_reg_array[i][4];
3118 		} else if (is_t6(padap->params.chip)) {
3119 			up_cim_reg->ireg_addr = t6_up_cim_reg_array[i][0];
3120 			up_cim_reg->ireg_data = t6_up_cim_reg_array[i][1];
3121 			up_cim_reg->ireg_local_offset =
3122 						t6_up_cim_reg_array[i][2];
3123 			up_cim_reg->ireg_offset_range =
3124 						t6_up_cim_reg_array[i][3];
3125 			instance = t6_up_cim_reg_array[i][4];
3126 		}
3127 
3128 		switch (instance) {
3129 		case NUM_CIM_CTL_TSCH_CHANNEL_INSTANCES:
3130 			iter = up_cim_reg->ireg_offset_range;
3131 			local_offset = 0x120;
3132 			local_range = 1;
3133 			break;
3134 		case NUM_CIM_CTL_TSCH_CHANNEL_TSCH_CLASS_INSTANCES:
3135 			iter = up_cim_reg->ireg_offset_range;
3136 			local_offset = 0x10;
3137 			local_range = 1;
3138 			break;
3139 		default:
3140 			iter = 1;
3141 			local_offset = 0;
3142 			local_range = up_cim_reg->ireg_offset_range;
3143 			break;
3144 		}
3145 
3146 		for (j = 0; j < iter; j++, buff++) {
3147 			rc = t4_cim_read(padap,
3148 					 up_cim_reg->ireg_local_offset +
3149 					 (j * local_offset), local_range, buff);
3150 			if (rc) {
3151 				cudbg_put_buff(pdbg_init, &temp_buff);
3152 				return rc;
3153 			}
3154 		}
3155 		up_cim++;
3156 	}
3157 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
3158 }
3159 
3160 int cudbg_collect_pbt_tables(struct cudbg_init *pdbg_init,
3161 			     struct cudbg_buffer *dbg_buff,
3162 			     struct cudbg_error *cudbg_err)
3163 {
3164 	struct adapter *padap = pdbg_init->adap;
3165 	struct cudbg_buffer temp_buff = { 0 };
3166 	struct cudbg_pbt_tables *pbt;
3167 	int i, rc;
3168 	u32 addr;
3169 
3170 	rc = cudbg_get_buff(pdbg_init, dbg_buff,
3171 			    sizeof(struct cudbg_pbt_tables),
3172 			    &temp_buff);
3173 	if (rc)
3174 		return rc;
3175 
3176 	pbt = (struct cudbg_pbt_tables *)temp_buff.data;
3177 	/* PBT dynamic entries */
3178 	addr = CUDBG_CHAC_PBT_ADDR;
3179 	for (i = 0; i < CUDBG_PBT_DYNAMIC_ENTRIES; i++) {
3180 		rc = t4_cim_read(padap, addr + (i * 4), 1,
3181 				 &pbt->pbt_dynamic[i]);
3182 		if (rc) {
3183 			cudbg_err->sys_err = rc;
3184 			cudbg_put_buff(pdbg_init, &temp_buff);
3185 			return rc;
3186 		}
3187 	}
3188 
3189 	/* PBT static entries */
3190 	/* static entries start when bit 6 is set */
3191 	addr = CUDBG_CHAC_PBT_ADDR + (1 << 6);
3192 	for (i = 0; i < CUDBG_PBT_STATIC_ENTRIES; i++) {
3193 		rc = t4_cim_read(padap, addr + (i * 4), 1,
3194 				 &pbt->pbt_static[i]);
3195 		if (rc) {
3196 			cudbg_err->sys_err = rc;
3197 			cudbg_put_buff(pdbg_init, &temp_buff);
3198 			return rc;
3199 		}
3200 	}
3201 
3202 	/* LRF entries */
3203 	addr = CUDBG_CHAC_PBT_LRF;
3204 	for (i = 0; i < CUDBG_LRF_ENTRIES; i++) {
3205 		rc = t4_cim_read(padap, addr + (i * 4), 1,
3206 				 &pbt->lrf_table[i]);
3207 		if (rc) {
3208 			cudbg_err->sys_err = rc;
3209 			cudbg_put_buff(pdbg_init, &temp_buff);
3210 			return rc;
3211 		}
3212 	}
3213 
3214 	/* PBT data entries */
3215 	addr = CUDBG_CHAC_PBT_DATA;
3216 	for (i = 0; i < CUDBG_PBT_DATA_ENTRIES; i++) {
3217 		rc = t4_cim_read(padap, addr + (i * 4), 1,
3218 				 &pbt->pbt_data[i]);
3219 		if (rc) {
3220 			cudbg_err->sys_err = rc;
3221 			cudbg_put_buff(pdbg_init, &temp_buff);
3222 			return rc;
3223 		}
3224 	}
3225 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
3226 }
3227 
3228 int cudbg_collect_mbox_log(struct cudbg_init *pdbg_init,
3229 			   struct cudbg_buffer *dbg_buff,
3230 			   struct cudbg_error *cudbg_err)
3231 {
3232 	struct adapter *padap = pdbg_init->adap;
3233 	struct cudbg_mbox_log *mboxlog = NULL;
3234 	struct cudbg_buffer temp_buff = { 0 };
3235 	struct mbox_cmd_log *log = NULL;
3236 	struct mbox_cmd *entry;
3237 	unsigned int entry_idx;
3238 	u16 mbox_cmds;
3239 	int i, k, rc;
3240 	u64 flit;
3241 	u32 size;
3242 
3243 	log = padap->mbox_log;
3244 	mbox_cmds = padap->mbox_log->size;
3245 	size = sizeof(struct cudbg_mbox_log) * mbox_cmds;
3246 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
3247 	if (rc)
3248 		return rc;
3249 
3250 	mboxlog = (struct cudbg_mbox_log *)temp_buff.data;
3251 	for (k = 0; k < mbox_cmds; k++) {
3252 		entry_idx = log->cursor + k;
3253 		if (entry_idx >= log->size)
3254 			entry_idx -= log->size;
3255 
3256 		entry = mbox_cmd_log_entry(log, entry_idx);
3257 		/* skip over unused entries */
3258 		if (entry->timestamp == 0)
3259 			continue;
3260 
3261 		memcpy(&mboxlog->entry, entry, sizeof(struct mbox_cmd));
3262 		for (i = 0; i < MBOX_LEN / 8; i++) {
3263 			flit = entry->cmd[i];
3264 			mboxlog->hi[i] = (u32)(flit >> 32);
3265 			mboxlog->lo[i] = (u32)flit;
3266 		}
3267 		mboxlog++;
3268 	}
3269 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
3270 }
3271 
3272 int cudbg_collect_hma_indirect(struct cudbg_init *pdbg_init,
3273 			       struct cudbg_buffer *dbg_buff,
3274 			       struct cudbg_error *cudbg_err)
3275 {
3276 	struct adapter *padap = pdbg_init->adap;
3277 	struct cudbg_buffer temp_buff = { 0 };
3278 	struct ireg_buf *hma_indr;
3279 	int i, rc, n;
3280 	u32 size;
3281 
3282 	if (CHELSIO_CHIP_VERSION(padap->params.chip) < CHELSIO_T6)
3283 		return CUDBG_STATUS_ENTITY_NOT_FOUND;
3284 
3285 	n = sizeof(t6_hma_ireg_array) / (IREG_NUM_ELEM * sizeof(u32));
3286 	size = sizeof(struct ireg_buf) * n;
3287 	rc = cudbg_get_buff(pdbg_init, dbg_buff, size, &temp_buff);
3288 	if (rc)
3289 		return rc;
3290 
3291 	hma_indr = (struct ireg_buf *)temp_buff.data;
3292 	for (i = 0; i < n; i++) {
3293 		struct ireg_field *hma_fli = &hma_indr->tp_pio;
3294 		u32 *buff = hma_indr->outbuf;
3295 
3296 		hma_fli->ireg_addr = t6_hma_ireg_array[i][0];
3297 		hma_fli->ireg_data = t6_hma_ireg_array[i][1];
3298 		hma_fli->ireg_local_offset = t6_hma_ireg_array[i][2];
3299 		hma_fli->ireg_offset_range = t6_hma_ireg_array[i][3];
3300 		t4_read_indirect(padap, hma_fli->ireg_addr, hma_fli->ireg_data,
3301 				 buff, hma_fli->ireg_offset_range,
3302 				 hma_fli->ireg_local_offset);
3303 		hma_indr++;
3304 	}
3305 	return cudbg_write_and_release_buff(pdbg_init, &temp_buff, dbg_buff);
3306 }
3307 
3308 void cudbg_fill_qdesc_num_and_size(const struct adapter *padap,
3309 				   u32 *num, u32 *size)
3310 {
3311 	u32 tot_entries = 0, tot_size = 0;
3312 
3313 	/* NIC TXQ, RXQ, FLQ, and CTRLQ */
3314 	tot_entries += MAX_ETH_QSETS * 3;
3315 	tot_entries += MAX_CTRL_QUEUES;
3316 
3317 	tot_size += MAX_ETH_QSETS * MAX_TXQ_ENTRIES * MAX_TXQ_DESC_SIZE;
3318 	tot_size += MAX_ETH_QSETS * MAX_RSPQ_ENTRIES * MAX_RXQ_DESC_SIZE;
3319 	tot_size += MAX_ETH_QSETS * MAX_RX_BUFFERS * MAX_FL_DESC_SIZE;
3320 	tot_size += MAX_CTRL_QUEUES * MAX_CTRL_TXQ_ENTRIES *
3321 		    MAX_CTRL_TXQ_DESC_SIZE;
3322 
3323 	/* FW_EVTQ and INTRQ */
3324 	tot_entries += INGQ_EXTRAS;
3325 	tot_size += INGQ_EXTRAS * MAX_RSPQ_ENTRIES * MAX_RXQ_DESC_SIZE;
3326 
3327 	/* PTP_TXQ */
3328 	tot_entries += 1;
3329 	tot_size += MAX_TXQ_ENTRIES * MAX_TXQ_DESC_SIZE;
3330 
3331 	/* ULD TXQ, RXQ, and FLQ */
3332 	tot_entries += CXGB4_TX_MAX * MAX_OFLD_QSETS;
3333 	tot_entries += CXGB4_ULD_MAX * MAX_ULD_QSETS * 2;
3334 
3335 	tot_size += CXGB4_TX_MAX * MAX_OFLD_QSETS * MAX_TXQ_ENTRIES *
3336 		    MAX_TXQ_DESC_SIZE;
3337 	tot_size += CXGB4_ULD_MAX * MAX_ULD_QSETS * MAX_RSPQ_ENTRIES *
3338 		    MAX_RXQ_DESC_SIZE;
3339 	tot_size += CXGB4_ULD_MAX * MAX_ULD_QSETS * MAX_RX_BUFFERS *
3340 		    MAX_FL_DESC_SIZE;
3341 
3342 	/* ULD CIQ */
3343 	tot_entries += CXGB4_ULD_MAX * MAX_ULD_QSETS;
3344 	tot_size += CXGB4_ULD_MAX * MAX_ULD_QSETS * SGE_MAX_IQ_SIZE *
3345 		    MAX_RXQ_DESC_SIZE;
3346 
3347 	/* ETHOFLD TXQ, RXQ, and FLQ */
3348 	tot_entries += MAX_OFLD_QSETS * 3;
3349 	tot_size += MAX_OFLD_QSETS * MAX_TXQ_ENTRIES * MAX_TXQ_DESC_SIZE;
3350 
3351 	tot_size += sizeof(struct cudbg_ver_hdr) +
3352 		    sizeof(struct cudbg_qdesc_info) +
3353 		    sizeof(struct cudbg_qdesc_entry) * tot_entries;
3354 
3355 	if (num)
3356 		*num = tot_entries;
3357 
3358 	if (size)
3359 		*size = tot_size;
3360 }
3361 
3362 int cudbg_collect_qdesc(struct cudbg_init *pdbg_init,
3363 			struct cudbg_buffer *dbg_buff,
3364 			struct cudbg_error *cudbg_err)
3365 {
3366 	u32 num_queues = 0, tot_entries = 0, size = 0;
3367 	struct adapter *padap = pdbg_init->adap;
3368 	struct cudbg_buffer temp_buff = { 0 };
3369 	struct cudbg_qdesc_entry *qdesc_entry;
3370 	struct cudbg_qdesc_info *qdesc_info;
3371 	struct cudbg_ver_hdr *ver_hdr;
3372 	struct sge *s = &padap->sge;
3373 	u32 i, j, cur_off, tot_len;
3374 	u8 *data;
3375 	int rc;
3376 
3377 	cudbg_fill_qdesc_num_and_size(padap, &tot_entries, &size);
3378 	size = min_t(u32, size, CUDBG_DUMP_BUFF_SIZE);
3379 	tot_len = size;
3380 	data = kvzalloc(size, GFP_KERNEL);
3381 	if (!data)
3382 		return -ENOMEM;
3383 
3384 	ver_hdr = (struct cudbg_ver_hdr *)data;
3385 	ver_hdr->signature = CUDBG_ENTITY_SIGNATURE;
3386 	ver_hdr->revision = CUDBG_QDESC_REV;
3387 	ver_hdr->size = sizeof(struct cudbg_qdesc_info);
3388 	size -= sizeof(*ver_hdr);
3389 
3390 	qdesc_info = (struct cudbg_qdesc_info *)(data +
3391 						 sizeof(*ver_hdr));
3392 	size -= sizeof(*qdesc_info);
3393 	qdesc_entry = (struct cudbg_qdesc_entry *)qdesc_info->data;
3394 
3395 #define QDESC_GET(q, desc, type, label) do { \
3396 	if (size <= 0) { \
3397 		goto label; \
3398 	} \
3399 	if (desc) { \
3400 		cudbg_fill_qdesc_##q(q, type, qdesc_entry); \
3401 		size -= sizeof(*qdesc_entry) + qdesc_entry->data_size; \
3402 		num_queues++; \
3403 		qdesc_entry = cudbg_next_qdesc(qdesc_entry); \
3404 	} \
3405 } while (0)
3406 
3407 #define QDESC_GET_TXQ(q, type, label) do { \
3408 	struct sge_txq *txq = (struct sge_txq *)q; \
3409 	QDESC_GET(txq, txq->desc, type, label); \
3410 } while (0)
3411 
3412 #define QDESC_GET_RXQ(q, type, label) do { \
3413 	struct sge_rspq *rxq = (struct sge_rspq *)q; \
3414 	QDESC_GET(rxq, rxq->desc, type, label); \
3415 } while (0)
3416 
3417 #define QDESC_GET_FLQ(q, type, label) do { \
3418 	struct sge_fl *flq = (struct sge_fl *)q; \
3419 	QDESC_GET(flq, flq->desc, type, label); \
3420 } while (0)
3421 
3422 	/* NIC TXQ */
3423 	for (i = 0; i < s->ethqsets; i++)
3424 		QDESC_GET_TXQ(&s->ethtxq[i].q, CUDBG_QTYPE_NIC_TXQ, out);
3425 
3426 	/* NIC RXQ */
3427 	for (i = 0; i < s->ethqsets; i++)
3428 		QDESC_GET_RXQ(&s->ethrxq[i].rspq, CUDBG_QTYPE_NIC_RXQ, out);
3429 
3430 	/* NIC FLQ */
3431 	for (i = 0; i < s->ethqsets; i++)
3432 		QDESC_GET_FLQ(&s->ethrxq[i].fl, CUDBG_QTYPE_NIC_FLQ, out);
3433 
3434 	/* NIC CTRLQ */
3435 	for (i = 0; i < padap->params.nports; i++)
3436 		QDESC_GET_TXQ(&s->ctrlq[i].q, CUDBG_QTYPE_CTRLQ, out);
3437 
3438 	/* FW_EVTQ */
3439 	QDESC_GET_RXQ(&s->fw_evtq, CUDBG_QTYPE_FWEVTQ, out);
3440 
3441 	/* INTRQ */
3442 	QDESC_GET_RXQ(&s->intrq, CUDBG_QTYPE_INTRQ, out);
3443 
3444 	/* PTP_TXQ */
3445 	QDESC_GET_TXQ(&s->ptptxq.q, CUDBG_QTYPE_PTP_TXQ, out);
3446 
3447 	/* ULD Queues */
3448 	mutex_lock(&uld_mutex);
3449 
3450 	if (s->uld_txq_info) {
3451 		struct sge_uld_txq_info *utxq;
3452 
3453 		/* ULD TXQ */
3454 		for (j = 0; j < CXGB4_TX_MAX; j++) {
3455 			if (!s->uld_txq_info[j])
3456 				continue;
3457 
3458 			utxq = s->uld_txq_info[j];
3459 			for (i = 0; i < utxq->ntxq; i++)
3460 				QDESC_GET_TXQ(&utxq->uldtxq[i].q,
3461 					      cudbg_uld_txq_to_qtype(j),
3462 					      out_unlock_uld);
3463 		}
3464 	}
3465 
3466 	if (s->uld_rxq_info) {
3467 		struct sge_uld_rxq_info *urxq;
3468 		u32 base;
3469 
3470 		/* ULD RXQ */
3471 		for (j = 0; j < CXGB4_ULD_MAX; j++) {
3472 			if (!s->uld_rxq_info[j])
3473 				continue;
3474 
3475 			urxq = s->uld_rxq_info[j];
3476 			for (i = 0; i < urxq->nrxq; i++)
3477 				QDESC_GET_RXQ(&urxq->uldrxq[i].rspq,
3478 					      cudbg_uld_rxq_to_qtype(j),
3479 					      out_unlock_uld);
3480 		}
3481 
3482 		/* ULD FLQ */
3483 		for (j = 0; j < CXGB4_ULD_MAX; j++) {
3484 			if (!s->uld_rxq_info[j])
3485 				continue;
3486 
3487 			urxq = s->uld_rxq_info[j];
3488 			for (i = 0; i < urxq->nrxq; i++)
3489 				QDESC_GET_FLQ(&urxq->uldrxq[i].fl,
3490 					      cudbg_uld_flq_to_qtype(j),
3491 					      out_unlock_uld);
3492 		}
3493 
3494 		/* ULD CIQ */
3495 		for (j = 0; j < CXGB4_ULD_MAX; j++) {
3496 			if (!s->uld_rxq_info[j])
3497 				continue;
3498 
3499 			urxq = s->uld_rxq_info[j];
3500 			base = urxq->nrxq;
3501 			for (i = 0; i < urxq->nciq; i++)
3502 				QDESC_GET_RXQ(&urxq->uldrxq[base + i].rspq,
3503 					      cudbg_uld_ciq_to_qtype(j),
3504 					      out_unlock_uld);
3505 		}
3506 	}
3507 	mutex_unlock(&uld_mutex);
3508 
3509 	if (!padap->tc_mqprio)
3510 		goto out;
3511 
3512 	mutex_lock(&padap->tc_mqprio->mqprio_mutex);
3513 	/* ETHOFLD TXQ */
3514 	if (s->eohw_txq)
3515 		for (i = 0; i < s->eoqsets; i++)
3516 			QDESC_GET_TXQ(&s->eohw_txq[i].q,
3517 				      CUDBG_QTYPE_ETHOFLD_TXQ, out_unlock_mqprio);
3518 
3519 	/* ETHOFLD RXQ and FLQ */
3520 	if (s->eohw_rxq) {
3521 		for (i = 0; i < s->eoqsets; i++)
3522 			QDESC_GET_RXQ(&s->eohw_rxq[i].rspq,
3523 				      CUDBG_QTYPE_ETHOFLD_RXQ, out_unlock_mqprio);
3524 
3525 		for (i = 0; i < s->eoqsets; i++)
3526 			QDESC_GET_FLQ(&s->eohw_rxq[i].fl,
3527 				      CUDBG_QTYPE_ETHOFLD_FLQ, out_unlock_mqprio);
3528 	}
3529 
3530 out_unlock_mqprio:
3531 	mutex_unlock(&padap->tc_mqprio->mqprio_mutex);
3532 
3533 out:
3534 	qdesc_info->qdesc_entry_size = sizeof(*qdesc_entry);
3535 	qdesc_info->num_queues = num_queues;
3536 	cur_off = 0;
3537 	while (tot_len) {
3538 		u32 chunk_size = min_t(u32, tot_len, CUDBG_CHUNK_SIZE);
3539 
3540 		rc = cudbg_get_buff(pdbg_init, dbg_buff, chunk_size,
3541 				    &temp_buff);
3542 		if (rc) {
3543 			cudbg_err->sys_warn = CUDBG_STATUS_PARTIAL_DATA;
3544 			goto out_free;
3545 		}
3546 
3547 		memcpy(temp_buff.data, data + cur_off, chunk_size);
3548 		tot_len -= chunk_size;
3549 		cur_off += chunk_size;
3550 		rc = cudbg_write_and_release_buff(pdbg_init, &temp_buff,
3551 						  dbg_buff);
3552 		if (rc) {
3553 			cudbg_put_buff(pdbg_init, &temp_buff);
3554 			cudbg_err->sys_warn = CUDBG_STATUS_PARTIAL_DATA;
3555 			goto out_free;
3556 		}
3557 	}
3558 
3559 out_free:
3560 	kvfree(data);
3561 
3562 #undef QDESC_GET_FLQ
3563 #undef QDESC_GET_RXQ
3564 #undef QDESC_GET_TXQ
3565 #undef QDESC_GET
3566 
3567 	return rc;
3568 
3569 out_unlock_uld:
3570 	mutex_unlock(&uld_mutex);
3571 	goto out;
3572 }
3573 
3574 int cudbg_collect_flash(struct cudbg_init *pdbg_init,
3575 			struct cudbg_buffer *dbg_buff,
3576 			struct cudbg_error *cudbg_err)
3577 {
3578 	struct adapter *padap = pdbg_init->adap;
3579 	u32 count = padap->params.sf_size, n;
3580 	struct cudbg_buffer temp_buff = {0};
3581 	u32 addr, i;
3582 	int rc;
3583 
3584 	addr = FLASH_EXP_ROM_START;
3585 
3586 	for (i = 0; i < count; i += SF_PAGE_SIZE) {
3587 		n = min_t(u32, count - i, SF_PAGE_SIZE);
3588 
3589 		rc = cudbg_get_buff(pdbg_init, dbg_buff, n, &temp_buff);
3590 		if (rc) {
3591 			cudbg_err->sys_warn = CUDBG_STATUS_PARTIAL_DATA;
3592 			goto out;
3593 		}
3594 		rc = t4_read_flash(padap, addr, n, (u32 *)temp_buff.data, 0);
3595 		if (rc)
3596 			goto out;
3597 
3598 		addr += (n * 4);
3599 		rc = cudbg_write_and_release_buff(pdbg_init, &temp_buff,
3600 						  dbg_buff);
3601 		if (rc) {
3602 			cudbg_err->sys_warn = CUDBG_STATUS_PARTIAL_DATA;
3603 			goto out;
3604 		}
3605 	}
3606 
3607 out:
3608 	return rc;
3609 }
3610