xref: /linux/drivers/net/ethernet/mellanox/mlx4/crdump.c (revision 58f6259b7a08f8d47d4629609703d358b042f0fd)
1 /*
2  * Copyright (c) 2018, Mellanox Technologies. All rights reserved.
3  *
4  * This software is available to you under a choice of one of two
5  * licenses.  You may choose to be licensed under the terms of the GNU
6  * General Public License (GPL) Version 2, available from the file
7  * COPYING in the main directory of this source tree, or the
8  * OpenIB.org BSD license below:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
18  *      - Redistributions in binary form must reproduce the above
19  *        copyright notice, this list of conditions and the following
20  *        disclaimer in the documentation and/or other materials
21  *        provided with the distribution.
22  *
23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30  * SOFTWARE.
31  */
32 
33 #include "mlx4.h"
34 
35 #define BAD_ACCESS			0xBADACCE5
36 #define HEALTH_BUFFER_SIZE		0x40
37 #define CR_ENABLE_BIT			swab32(BIT(6))
38 #define CR_ENABLE_BIT_OFFSET		0xF3F04
39 #define MAX_NUM_OF_DUMPS_TO_STORE	(8)
40 
41 #define REGION_CR_SPACE "cr-space"
42 #define REGION_FW_HEALTH "fw-health"
43 
44 static const char * const region_cr_space_str = REGION_CR_SPACE;
45 static const char * const region_fw_health_str = REGION_FW_HEALTH;
46 
47 static const struct devlink_region_ops region_cr_space_ops = {
48 	.name = REGION_CR_SPACE,
49 	.destructor = &kvfree,
50 };
51 
52 static const struct devlink_region_ops region_fw_health_ops = {
53 	.name = REGION_FW_HEALTH,
54 	.destructor = &kvfree,
55 };
56 
57 /* Set to true in case cr enable bit was set to true before crdump */
58 static bool crdump_enbale_bit_set;
59 
60 static void crdump_enable_crspace_access(struct mlx4_dev *dev,
61 					 u8 __iomem *cr_space)
62 {
63 	/* Get current enable bit value */
64 	crdump_enbale_bit_set =
65 		readl(cr_space + CR_ENABLE_BIT_OFFSET) & CR_ENABLE_BIT;
66 
67 	/* Enable FW CR filter (set bit6 to 0) */
68 	if (crdump_enbale_bit_set)
69 		writel(readl(cr_space + CR_ENABLE_BIT_OFFSET) & ~CR_ENABLE_BIT,
70 		       cr_space + CR_ENABLE_BIT_OFFSET);
71 
72 	/* Enable block volatile crspace accesses */
73 	writel(swab32(1), cr_space + dev->caps.health_buffer_addrs +
74 	       HEALTH_BUFFER_SIZE);
75 }
76 
77 static void crdump_disable_crspace_access(struct mlx4_dev *dev,
78 					  u8 __iomem *cr_space)
79 {
80 	/* Disable block volatile crspace accesses */
81 	writel(0, cr_space + dev->caps.health_buffer_addrs +
82 	       HEALTH_BUFFER_SIZE);
83 
84 	/* Restore FW CR filter value (set bit6 to original value) */
85 	if (crdump_enbale_bit_set)
86 		writel(readl(cr_space + CR_ENABLE_BIT_OFFSET) | CR_ENABLE_BIT,
87 		       cr_space + CR_ENABLE_BIT_OFFSET);
88 }
89 
90 static void mlx4_crdump_collect_crspace(struct mlx4_dev *dev,
91 					u8 __iomem *cr_space,
92 					u32 id)
93 {
94 	struct mlx4_fw_crdump *crdump = &dev->persist->crdump;
95 	struct pci_dev *pdev = dev->persist->pdev;
96 	unsigned long cr_res_size;
97 	u8 *crspace_data;
98 	int offset;
99 	int err;
100 
101 	if (!crdump->region_crspace) {
102 		mlx4_err(dev, "crdump: cr-space region is NULL\n");
103 		return;
104 	}
105 
106 	/* Try to collect CR space */
107 	cr_res_size = pci_resource_len(pdev, 0);
108 	crspace_data = kvmalloc(cr_res_size, GFP_KERNEL);
109 	if (crspace_data) {
110 		for (offset = 0; offset < cr_res_size; offset += 4)
111 			*(u32 *)(crspace_data + offset) =
112 					readl(cr_space + offset);
113 
114 		err = devlink_region_snapshot_create(crdump->region_crspace,
115 						     crspace_data, id);
116 		if (err) {
117 			kvfree(crspace_data);
118 			mlx4_warn(dev, "crdump: devlink create %s snapshot id %d err %d\n",
119 				  region_cr_space_str, id, err);
120 		} else {
121 			mlx4_info(dev, "crdump: added snapshot %d to devlink region %s\n",
122 				  id, region_cr_space_str);
123 		}
124 	} else {
125 		mlx4_err(dev, "crdump: Failed to allocate crspace buffer\n");
126 	}
127 }
128 
129 static void mlx4_crdump_collect_fw_health(struct mlx4_dev *dev,
130 					  u8 __iomem *cr_space,
131 					  u32 id)
132 {
133 	struct mlx4_fw_crdump *crdump = &dev->persist->crdump;
134 	u8 *health_data;
135 	int offset;
136 	int err;
137 
138 	if (!crdump->region_fw_health) {
139 		mlx4_err(dev, "crdump: fw-health region is NULL\n");
140 		return;
141 	}
142 
143 	/* Try to collect health buffer */
144 	health_data = kvmalloc(HEALTH_BUFFER_SIZE, GFP_KERNEL);
145 	if (health_data) {
146 		u8 __iomem *health_buf_start =
147 				cr_space + dev->caps.health_buffer_addrs;
148 
149 		for (offset = 0; offset < HEALTH_BUFFER_SIZE; offset += 4)
150 			*(u32 *)(health_data + offset) =
151 					readl(health_buf_start + offset);
152 
153 		err = devlink_region_snapshot_create(crdump->region_fw_health,
154 						     health_data, id);
155 		if (err) {
156 			kvfree(health_data);
157 			mlx4_warn(dev, "crdump: devlink create %s snapshot id %d err %d\n",
158 				  region_fw_health_str, id, err);
159 		} else {
160 			mlx4_info(dev, "crdump: added snapshot %d to devlink region %s\n",
161 				  id, region_fw_health_str);
162 		}
163 	} else {
164 		mlx4_err(dev, "crdump: Failed to allocate health buffer\n");
165 	}
166 }
167 
168 int mlx4_crdump_collect(struct mlx4_dev *dev)
169 {
170 	struct devlink *devlink = priv_to_devlink(mlx4_priv(dev));
171 	struct mlx4_fw_crdump *crdump = &dev->persist->crdump;
172 	struct pci_dev *pdev = dev->persist->pdev;
173 	unsigned long cr_res_size;
174 	u8 __iomem *cr_space;
175 	int err;
176 	u32 id;
177 
178 	if (!dev->caps.health_buffer_addrs) {
179 		mlx4_info(dev, "crdump: FW doesn't support health buffer access, skipping\n");
180 		return 0;
181 	}
182 
183 	if (!crdump->snapshot_enable) {
184 		mlx4_info(dev, "crdump: devlink snapshot disabled, skipping\n");
185 		return 0;
186 	}
187 
188 	cr_res_size = pci_resource_len(pdev, 0);
189 
190 	cr_space = ioremap(pci_resource_start(pdev, 0), cr_res_size);
191 	if (!cr_space) {
192 		mlx4_err(dev, "crdump: Failed to map pci cr region\n");
193 		return -ENODEV;
194 	}
195 
196 	/* Get the available snapshot ID for the dumps */
197 	err = devlink_region_snapshot_id_get(devlink, &id);
198 	if (err) {
199 		mlx4_err(dev, "crdump: devlink get snapshot id err %d\n", err);
200 		iounmap(cr_space);
201 		return err;
202 	}
203 
204 	crdump_enable_crspace_access(dev, cr_space);
205 
206 	/* Try to capture dumps */
207 	mlx4_crdump_collect_crspace(dev, cr_space, id);
208 	mlx4_crdump_collect_fw_health(dev, cr_space, id);
209 
210 	/* Release reference on the snapshot id */
211 	devlink_region_snapshot_id_put(devlink, id);
212 
213 	crdump_disable_crspace_access(dev, cr_space);
214 
215 	iounmap(cr_space);
216 	return 0;
217 }
218 
219 int mlx4_crdump_init(struct mlx4_dev *dev)
220 {
221 	struct devlink *devlink = priv_to_devlink(mlx4_priv(dev));
222 	struct mlx4_fw_crdump *crdump = &dev->persist->crdump;
223 	struct pci_dev *pdev = dev->persist->pdev;
224 
225 	crdump->snapshot_enable = false;
226 
227 	/* Create cr-space region */
228 	crdump->region_crspace =
229 		devl_region_create(devlink,
230 				   &region_cr_space_ops,
231 				   MAX_NUM_OF_DUMPS_TO_STORE,
232 				   pci_resource_len(pdev, 0));
233 	if (IS_ERR(crdump->region_crspace))
234 		mlx4_warn(dev, "crdump: create devlink region %s err %ld\n",
235 			  region_cr_space_str,
236 			  PTR_ERR(crdump->region_crspace));
237 
238 	/* Create fw-health region */
239 	crdump->region_fw_health =
240 		devl_region_create(devlink,
241 				   &region_fw_health_ops,
242 				   MAX_NUM_OF_DUMPS_TO_STORE,
243 				   HEALTH_BUFFER_SIZE);
244 	if (IS_ERR(crdump->region_fw_health))
245 		mlx4_warn(dev, "crdump: create devlink region %s err %ld\n",
246 			  region_fw_health_str,
247 			  PTR_ERR(crdump->region_fw_health));
248 
249 	return 0;
250 }
251 
252 void mlx4_crdump_end(struct mlx4_dev *dev)
253 {
254 	struct mlx4_fw_crdump *crdump = &dev->persist->crdump;
255 
256 	devl_region_destroy(crdump->region_fw_health);
257 	devl_region_destroy(crdump->region_crspace);
258 }
259