1 // SPDX-License-Identifier: MIT
2 /*
3 * Copyright 2025 Advanced Micro Devices, Inc.
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice shall be included in
13 * all copies or substantial portions of the Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
19 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
20 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
21 * OTHER DEALINGS IN THE SOFTWARE.
22 *
23 */
24 #include "ras.h"
25 #include "ras_core_status.h"
26 #include "ras_log_ring.h"
27
28 #define RAS_LOG_MAX_QUERY_SIZE 0xC000
29 #define RAS_LOG_MEM_TEMP_SIZE 0x200
30 #define RAS_LOG_MEMPOOL_SIZE \
31 (RAS_LOG_MAX_QUERY_SIZE + RAS_LOG_MEM_TEMP_SIZE)
32
33 #define BATCH_IDX_TO_TREE_IDX(batch_idx, sn) (((batch_idx) << 8) | (sn))
34
35 static const uint64_t ras_rma_aca_reg[ACA_REG_MAX_COUNT] = {
36 [ACA_REG_IDX__CTL] = 0x1,
37 [ACA_REG_IDX__STATUS] = 0xB000000000000137,
38 [ACA_REG_IDX__ADDR] = 0x0,
39 [ACA_REG_IDX__MISC0] = 0x0,
40 [ACA_REG_IDX__CONFG] = 0x1ff00000002,
41 [ACA_REG_IDX__IPID] = 0x9600000000,
42 [ACA_REG_IDX__SYND] = 0x0,
43 };
44
ras_log_ring_get_logged_ecc_count(struct ras_core_context * ras_core)45 static uint64_t ras_log_ring_get_logged_ecc_count(struct ras_core_context *ras_core)
46 {
47 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
48 uint64_t count = 0;
49
50 if (log_ring->logged_ecc_count < 0) {
51 RAS_DEV_WARN(ras_core->dev,
52 "Error: the logged ras count should not less than 0!\n");
53 count = 0;
54 } else {
55 count = log_ring->logged_ecc_count;
56 }
57
58 if (count > RAS_LOG_MEMPOOL_SIZE)
59 RAS_DEV_WARN(ras_core->dev,
60 "Error: the logged ras count is out of range!\n");
61
62 return count;
63 }
64
ras_log_ring_add_data(struct ras_core_context * ras_core,struct ras_log_info * log,struct ras_log_batch_tag * batch_tag)65 static int ras_log_ring_add_data(struct ras_core_context *ras_core,
66 struct ras_log_info *log, struct ras_log_batch_tag *batch_tag)
67 {
68 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
69 unsigned long flags = 0;
70 int ret = 0;
71
72 if (batch_tag && (batch_tag->sub_seqno >= MAX_RECORD_PER_BATCH)) {
73 RAS_DEV_ERR(ras_core->dev,
74 "Invalid batch sub seqno:%d, batch:0x%llx\n",
75 batch_tag->sub_seqno, batch_tag->batch_id);
76 return -EINVAL;
77 }
78
79 spin_lock_irqsave(&log_ring->spin_lock, flags);
80 if (batch_tag) {
81 log->seqno =
82 BATCH_IDX_TO_TREE_IDX(batch_tag->batch_id, batch_tag->sub_seqno);
83 batch_tag->sub_seqno++;
84 } else {
85 log->seqno = BATCH_IDX_TO_TREE_IDX(log_ring->mono_upward_batch_id, 0);
86 log_ring->mono_upward_batch_id++;
87 }
88 ret = radix_tree_insert(&log_ring->ras_log_root, log->seqno, log);
89 if (!ret)
90 log_ring->logged_ecc_count++;
91 spin_unlock_irqrestore(&log_ring->spin_lock, flags);
92
93 if (ret) {
94 RAS_DEV_ERR(ras_core->dev,
95 "Failed to add ras log! seqno:0x%llx, ret:%d\n",
96 log->seqno, ret);
97 mempool_free(log, log_ring->ras_log_mempool);
98 }
99
100 return ret;
101 }
102
ras_log_ring_delete_data(struct ras_core_context * ras_core,uint32_t count)103 static int ras_log_ring_delete_data(struct ras_core_context *ras_core, uint32_t count)
104 {
105 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
106 unsigned long flags = 0;
107 uint32_t i = 0, j = 0;
108 uint64_t batch_id, idx;
109 void *data;
110 int ret = -ENODATA;
111
112 if (count > ras_log_ring_get_logged_ecc_count(ras_core))
113 return -EINVAL;
114
115 spin_lock_irqsave(&log_ring->spin_lock, flags);
116 batch_id = log_ring->last_del_batch_id;
117 while (batch_id < log_ring->mono_upward_batch_id) {
118 for (j = 0; j < MAX_RECORD_PER_BATCH; j++) {
119 idx = BATCH_IDX_TO_TREE_IDX(batch_id, j);
120 data = radix_tree_delete(&log_ring->ras_log_root, idx);
121 if (data) {
122 mempool_free(data, log_ring->ras_log_mempool);
123 log_ring->logged_ecc_count--;
124 i++;
125 }
126 }
127 batch_id = ++log_ring->last_del_batch_id;
128 if (i >= count) {
129 ret = 0;
130 break;
131 }
132 }
133 spin_unlock_irqrestore(&log_ring->spin_lock, flags);
134
135 return ret;
136 }
137
ras_log_ring_clear_log_tree(struct ras_core_context * ras_core)138 static void ras_log_ring_clear_log_tree(struct ras_core_context *ras_core)
139 {
140 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
141 uint64_t batch_id, idx;
142 unsigned long flags = 0;
143 void *data;
144 int j;
145
146 if ((log_ring->mono_upward_batch_id <= log_ring->last_del_batch_id) &&
147 !log_ring->logged_ecc_count)
148 return;
149
150 spin_lock_irqsave(&log_ring->spin_lock, flags);
151 batch_id = log_ring->last_del_batch_id;
152 while (batch_id < log_ring->mono_upward_batch_id) {
153 for (j = 0; j < MAX_RECORD_PER_BATCH; j++) {
154 idx = BATCH_IDX_TO_TREE_IDX(batch_id, j);
155 data = radix_tree_delete(&log_ring->ras_log_root, idx);
156 if (data) {
157 mempool_free(data, log_ring->ras_log_mempool);
158 log_ring->logged_ecc_count--;
159 }
160 }
161 batch_id++;
162 }
163 spin_unlock_irqrestore(&log_ring->spin_lock, flags);
164
165 }
166
ras_log_ring_sw_init(struct ras_core_context * ras_core)167 int ras_log_ring_sw_init(struct ras_core_context *ras_core)
168 {
169 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
170
171 memset(log_ring, 0, sizeof(*log_ring));
172
173 log_ring->ras_log_mempool = mempool_create_kmalloc_pool(
174 RAS_LOG_MEMPOOL_SIZE, sizeof(struct ras_log_info));
175 if (!log_ring->ras_log_mempool)
176 return -ENOMEM;
177
178 INIT_RADIX_TREE(&log_ring->ras_log_root, GFP_KERNEL);
179
180 spin_lock_init(&log_ring->spin_lock);
181
182 return 0;
183 }
184
ras_log_ring_sw_fini(struct ras_core_context * ras_core)185 int ras_log_ring_sw_fini(struct ras_core_context *ras_core)
186 {
187 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
188
189 ras_log_ring_clear_log_tree(ras_core);
190 log_ring->logged_ecc_count = 0;
191 log_ring->last_del_batch_id = 0;
192 log_ring->mono_upward_batch_id = 0;
193
194 mempool_destroy(log_ring->ras_log_mempool);
195
196 return 0;
197 }
198
ras_log_ring_create_batch_tag(struct ras_core_context * ras_core)199 struct ras_log_batch_tag *ras_log_ring_create_batch_tag(struct ras_core_context *ras_core)
200 {
201 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
202 struct ras_log_batch_tag *batch_tag;
203 unsigned long flags = 0;
204
205 batch_tag = kzalloc_obj(*batch_tag);
206 if (!batch_tag)
207 return NULL;
208
209 spin_lock_irqsave(&log_ring->spin_lock, flags);
210 batch_tag->batch_id = log_ring->mono_upward_batch_id;
211 log_ring->mono_upward_batch_id++;
212 spin_unlock_irqrestore(&log_ring->spin_lock, flags);
213
214 batch_tag->sub_seqno = 0;
215 batch_tag->timestamp = ras_core_get_utc_second_timestamp(ras_core);
216 return batch_tag;
217 }
218
ras_log_ring_destroy_batch_tag(struct ras_core_context * ras_core,struct ras_log_batch_tag * batch_tag)219 void ras_log_ring_destroy_batch_tag(struct ras_core_context *ras_core,
220 struct ras_log_batch_tag *batch_tag)
221 {
222 kfree(batch_tag);
223 }
224
ras_log_ring_add_log_event(struct ras_core_context * ras_core,enum ras_log_event event,void * data,struct ras_log_batch_tag * batch_tag)225 void ras_log_ring_add_log_event(struct ras_core_context *ras_core,
226 enum ras_log_event event, void *data, struct ras_log_batch_tag *batch_tag)
227 {
228 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
229 struct device_system_info dev_info = {0};
230 struct ras_log_info *log;
231 uint64_t socket_id;
232 void *obj;
233
234 obj = mempool_alloc_preallocated(log_ring->ras_log_mempool);
235 if (!obj ||
236 (ras_log_ring_get_logged_ecc_count(ras_core) >= RAS_LOG_MEMPOOL_SIZE)) {
237 ras_log_ring_delete_data(ras_core, RAS_LOG_MEM_TEMP_SIZE);
238 if (!obj)
239 obj = mempool_alloc_preallocated(log_ring->ras_log_mempool);
240 }
241
242 if (!obj) {
243 RAS_DEV_ERR(ras_core->dev, "ERROR: Failed to alloc ras log buffer!\n");
244 return;
245 }
246
247 log = (struct ras_log_info *)obj;
248
249 memset(log, 0, sizeof(*log));
250 log->timestamp =
251 batch_tag ? batch_tag->timestamp : ras_core_get_utc_second_timestamp(ras_core);
252 log->event = event;
253
254 if (data)
255 memcpy(&log->aca_reg, data, sizeof(log->aca_reg));
256
257 if (event == RAS_LOG_EVENT_RMA) {
258 memcpy(&log->aca_reg, ras_rma_aca_reg, sizeof(log->aca_reg));
259 ras_core_get_device_system_info(ras_core, &dev_info);
260 socket_id = dev_info.socket_id;
261 log->aca_reg.regs[ACA_REG_IDX__IPID] |= ((socket_id / 4) & 0x01);
262 log->aca_reg.regs[ACA_REG_IDX__IPID] |= (((socket_id % 4) & 0x3) << 44);
263 }
264
265 ras_log_ring_add_data(ras_core, log, batch_tag);
266 }
267
ras_log_ring_lookup_data(struct ras_core_context * ras_core,uint64_t idx)268 static struct ras_log_info *ras_log_ring_lookup_data(struct ras_core_context *ras_core,
269 uint64_t idx)
270 {
271 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
272 unsigned long flags = 0;
273 void *data;
274
275 spin_lock_irqsave(&log_ring->spin_lock, flags);
276 data = radix_tree_lookup(&log_ring->ras_log_root, idx);
277 spin_unlock_irqrestore(&log_ring->spin_lock, flags);
278
279 return (struct ras_log_info *)data;
280 }
281
ras_log_ring_get_batch_records(struct ras_core_context * ras_core,uint64_t batch_id,struct ras_log_info ** log_arr,uint32_t arr_num)282 int ras_log_ring_get_batch_records(struct ras_core_context *ras_core, uint64_t batch_id,
283 struct ras_log_info **log_arr, uint32_t arr_num)
284 {
285 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
286 uint32_t i, idx, count = 0;
287 void *data;
288
289 if ((batch_id >= log_ring->mono_upward_batch_id) ||
290 (batch_id < log_ring->last_del_batch_id))
291 return -EINVAL;
292
293 for (i = 0; i < MAX_RECORD_PER_BATCH; i++) {
294 idx = BATCH_IDX_TO_TREE_IDX(batch_id, i);
295 data = ras_log_ring_lookup_data(ras_core, idx);
296 if (data) {
297 log_arr[count++] = data;
298 if (count >= arr_num)
299 break;
300 }
301 }
302
303 return count;
304 }
305
ras_log_ring_get_batch_overview(struct ras_core_context * ras_core,struct ras_log_batch_overview * overview)306 int ras_log_ring_get_batch_overview(struct ras_core_context *ras_core,
307 struct ras_log_batch_overview *overview)
308 {
309 struct ras_log_ring *log_ring = &ras_core->ras_log_ring;
310
311 overview->logged_batch_count =
312 log_ring->mono_upward_batch_id - log_ring->last_del_batch_id;
313 overview->last_batch_id = log_ring->mono_upward_batch_id;
314 overview->first_batch_id = log_ring->last_del_batch_id;
315
316 return 0;
317 }
318