Line data Source code
1 : /*
2 : * Copyright 2019 Advanced Micro Devices, Inc.
3 : *
4 : * Permission is hereby granted, free of charge, to any person obtaining a
5 : * copy of this software and associated documentation files (the "Software"),
6 : * to deal in the Software without restriction, including without limitation
7 : * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 : * and/or sell copies of the Software, and to permit persons to whom the
9 : * Software is furnished to do so, subject to the following conditions:
10 : *
11 : * The above copyright notice and this permission notice shall be included in
12 : * all copies or substantial portions of the Software.
13 : *
14 : * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 : * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 : * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 : * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 : * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 : * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 : * OTHER DEALINGS IN THE SOFTWARE.
21 : *
22 : */
23 :
24 : #include "amdgpu.h"
25 :
26 0 : static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev,
27 : void *ras_error_status,
28 : struct amdgpu_iv_entry *entry,
29 : bool reset)
30 : {
31 0 : struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
32 0 : struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
33 0 : int ret = 0;
34 :
35 0 : kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
36 0 : ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(con->umc_ecc));
37 0 : if (ret == -EOPNOTSUPP) {
38 0 : if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
39 0 : adev->umc.ras->ras_block.hw_ops->query_ras_error_count)
40 0 : adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev, ras_error_status);
41 :
42 0 : if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
43 0 : adev->umc.ras->ras_block.hw_ops->query_ras_error_address &&
44 0 : adev->umc.max_ras_err_cnt_per_query) {
45 0 : err_data->err_addr =
46 0 : kcalloc(adev->umc.max_ras_err_cnt_per_query,
47 : sizeof(struct eeprom_table_record), GFP_KERNEL);
48 :
49 : /* still call query_ras_error_address to clear error status
50 : * even NOMEM error is encountered
51 : */
52 0 : if(!err_data->err_addr)
53 0 : dev_warn(adev->dev, "Failed to alloc memory for "
54 : "umc error address record!\n");
55 :
56 : /* umc query_ras_error_address is also responsible for clearing
57 : * error status
58 : */
59 0 : adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev, ras_error_status);
60 : }
61 0 : } else if (!ret) {
62 0 : if (adev->umc.ras &&
63 0 : adev->umc.ras->ecc_info_query_ras_error_count)
64 0 : adev->umc.ras->ecc_info_query_ras_error_count(adev, ras_error_status);
65 :
66 0 : if (adev->umc.ras &&
67 0 : adev->umc.ras->ecc_info_query_ras_error_address &&
68 0 : adev->umc.max_ras_err_cnt_per_query) {
69 0 : err_data->err_addr =
70 0 : kcalloc(adev->umc.max_ras_err_cnt_per_query,
71 : sizeof(struct eeprom_table_record), GFP_KERNEL);
72 :
73 : /* still call query_ras_error_address to clear error status
74 : * even NOMEM error is encountered
75 : */
76 0 : if(!err_data->err_addr)
77 0 : dev_warn(adev->dev, "Failed to alloc memory for "
78 : "umc error address record!\n");
79 :
80 : /* umc query_ras_error_address is also responsible for clearing
81 : * error status
82 : */
83 0 : adev->umc.ras->ecc_info_query_ras_error_address(adev, ras_error_status);
84 : }
85 : }
86 :
87 : /* only uncorrectable error needs gpu reset */
88 0 : if (err_data->ue_count) {
89 0 : dev_info(adev->dev, "%ld uncorrectable hardware errors "
90 : "detected in UMC block\n",
91 : err_data->ue_count);
92 :
93 0 : if ((amdgpu_bad_page_threshold != 0) &&
94 0 : err_data->err_addr_cnt) {
95 0 : amdgpu_ras_add_bad_pages(adev, err_data->err_addr,
96 : err_data->err_addr_cnt);
97 0 : amdgpu_ras_save_bad_pages(adev);
98 :
99 0 : amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs);
100 :
101 0 : if (con->update_channel_flag == true) {
102 0 : amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap);
103 0 : con->update_channel_flag = false;
104 : }
105 : }
106 :
107 0 : if (reset)
108 0 : amdgpu_ras_reset_gpu(adev);
109 : }
110 :
111 0 : kfree(err_data->err_addr);
112 0 : return AMDGPU_RAS_SUCCESS;
113 : }
114 :
115 0 : int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
116 : void *ras_error_status,
117 : bool reset)
118 : {
119 : int ret;
120 0 : struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
121 0 : struct ras_common_if head = {
122 : .block = AMDGPU_RAS_BLOCK__UMC,
123 : };
124 0 : struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head);
125 :
126 0 : ret =
127 0 : amdgpu_umc_do_page_retirement(adev, ras_error_status, NULL, reset);
128 :
129 0 : if (ret == AMDGPU_RAS_SUCCESS && obj) {
130 0 : obj->err_data.ue_count += err_data->ue_count;
131 0 : obj->err_data.ce_count += err_data->ce_count;
132 : }
133 :
134 0 : return ret;
135 : }
136 :
137 0 : int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev,
138 : void *ras_error_status,
139 : struct amdgpu_iv_entry *entry)
140 : {
141 0 : return amdgpu_umc_do_page_retirement(adev, ras_error_status, entry, true);
142 : }
143 :
144 0 : int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
145 : {
146 : int r;
147 :
148 0 : r = amdgpu_ras_block_late_init(adev, ras_block);
149 0 : if (r)
150 : return r;
151 :
152 0 : if (amdgpu_ras_is_supported(adev, ras_block->block)) {
153 0 : r = amdgpu_irq_get(adev, &adev->gmc.ecc_irq, 0);
154 0 : if (r)
155 : goto late_fini;
156 : }
157 :
158 : /* ras init of specific umc version */
159 0 : if (adev->umc.ras &&
160 0 : adev->umc.ras->err_cnt_init)
161 0 : adev->umc.ras->err_cnt_init(adev);
162 :
163 : return 0;
164 :
165 : late_fini:
166 0 : amdgpu_ras_block_late_fini(adev, ras_block);
167 0 : return r;
168 : }
169 :
170 0 : int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev,
171 : struct amdgpu_irq_src *source,
172 : struct amdgpu_iv_entry *entry)
173 : {
174 0 : struct ras_common_if *ras_if = adev->umc.ras_if;
175 0 : struct ras_dispatch_if ih_data = {
176 : .entry = entry,
177 : };
178 :
179 0 : if (!ras_if)
180 : return 0;
181 :
182 0 : ih_data.head = *ras_if;
183 :
184 0 : amdgpu_ras_interrupt_dispatch(adev, &ih_data);
185 0 : return 0;
186 : }
187 :
188 0 : void amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
189 : uint64_t err_addr,
190 : uint64_t retired_page,
191 : uint32_t channel_index,
192 : uint32_t umc_inst)
193 : {
194 0 : struct eeprom_table_record *err_rec =
195 0 : &err_data->err_addr[err_data->err_addr_cnt];
196 :
197 0 : err_rec->address = err_addr;
198 : /* page frame address is saved */
199 0 : err_rec->retired_page = retired_page >> AMDGPU_GPU_PAGE_SHIFT;
200 0 : err_rec->ts = (uint64_t)ktime_get_real_seconds();
201 0 : err_rec->err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE;
202 0 : err_rec->cu = 0;
203 0 : err_rec->mem_channel = channel_index;
204 0 : err_rec->mcumc_id = umc_inst;
205 :
206 0 : err_data->err_addr_cnt++;
207 0 : }
|