xref: /linux/include/drm/drm_ras.h (revision b2207d4c4e84347a3a6b26b77e218cfc606d0109)
1 /* SPDX-License-Identifier: MIT */
2 /*
3  * Copyright © 2026 Intel Corporation
4  */
5 
6 #ifndef __DRM_RAS_H__
7 #define __DRM_RAS_H__
8 
9 #include <linux/types.h>
10 
11 #include <uapi/drm/drm_ras.h>
12 
13 /**
14  * struct drm_ras_node - A DRM RAS Node
15  */
16 struct drm_ras_node {
17 	/** @id: Unique identifier for the node. Dynamically assigned. */
18 	u32 id;
19 	/**
20 	 * @device_name: Human-readable name of the device. Given by the driver.
21 	 */
22 	const char *device_name;
23 	/** @node_name: Human-readable name of the node. Given by the driver. */
24 	const char *node_name;
25 	/** @type: Type of the node (enum drm_ras_node_type). */
26 	enum drm_ras_node_type type;
27 
28 	/* Error-Counter Related Callback and Variables */
29 
30 	/** @error_counter_range: Range of valid Error IDs for this node. */
31 	struct {
32 		/** @first: First valid Error ID. */
33 		u32 first;
34 		/** @last: Last valid Error ID. Mandatory entry. */
35 		u32 last;
36 	} error_counter_range;
37 
38 	/**
39 	 * @query_error_counter:
40 	 *
41 	 * This callback is used by drm-ras to query a specific error counter.
42 	 * Used for input check and to iterate all error counters in a node.
43 	 *
44 	 * Driver should expect query_error_counter() to be called with
45 	 * error_id from `error_counter_range.first` to
46 	 * `error_counter_range.last`.
47 	 *
48 	 * The @query_error_counter is a mandatory callback for
49 	 * error_counter_node.
50 	 *
51 	 * Returns: 0 on success,
52 	 *          -ENOENT when error_id is not supported as an indication that
53 	 *                  drm_ras should silently skip this entry. Used for
54 	 *                  supporting non-contiguous error ranges.
55 	 *                  Driver is responsible for maintaining the list of
56 	 *                  supported error IDs in the range of first to last.
57 	 *          Other negative values on errors that should terminate the
58 	 *          netlink query.
59 	 */
60 	int (*query_error_counter)(struct drm_ras_node *node, u32 error_id,
61 				   const char **name, u32 *val);
62 
63 	/**
64 	 * @clear_error_counter:
65 	 *
66 	 * This callback is used by drm_ras to clear a specific error counter.
67 	 * Driver should implement this callback to support clearing error counters
68 	 * of a node.
69 	 *
70 	 * Returns: 0 on success, negative error code on failure.
71 	 */
72 	int (*clear_error_counter)(struct drm_ras_node *node, u32 error_id);
73 
74 	/**
75 	 * @query_error_threshold:
76 	 *
77 	 * This callback is used by drm-ras to query error threshold of a
78 	 * specific counter.
79 	 *
80 	 * Driver should expect query_error_threshold() to be called with
81 	 * error_id from `error_counter_range.first` to
82 	 * `error_counter_range.last`.
83 	 *
84 	 * Returns: 0 on success, negative error code on failure.
85 	 */
86 	int (*query_error_threshold)(struct drm_ras_node *node, u32 error_id, const char **name,
87 				     u32 *threshold);
88 
89 	/**
90 	 * @set_error_threshold:
91 	 *
92 	 * This callback is used by drm-ras to set error threshold of a specific
93 	 * counter.
94 	 *
95 	 * Driver should expect set_error_threshold() to be called with error_id
96 	 * from `error_counter_range.first` to `error_counter_range.last`.
97 	 * Driver is responsible for error threshold bounds checking.
98 	 *
99 	 * Returns: 0 on success, negative error code on failure.
100 	 */
101 	int (*set_error_threshold)(struct drm_ras_node *node, u32 error_id, u32 threshold);
102 
103 	/** @priv: Driver private data */
104 	void *priv;
105 };
106 
107 struct drm_device;
108 
109 #if IS_ENABLED(CONFIG_DRM_RAS)
110 int drm_ras_node_register(struct drm_ras_node *node);
111 void drm_ras_node_unregister(struct drm_ras_node *node);
112 int drm_ras_nl_error_event(struct drm_ras_node *node, u32 error_id, const char *error_name,
113 			   u32 value);
114 #else
115 static inline int drm_ras_node_register(struct drm_ras_node *node) { return 0; }
116 static inline void drm_ras_node_unregister(struct drm_ras_node *node) { }
117 static inline int drm_ras_nl_error_event(struct drm_ras_node *node, u32 error_id,
118 					 const char *error_name, u32 value)
119 { return 0; }
120 #endif
121 
122 #endif
123