1 // SPDX-License-Identifier: GPL-2.0
3 * This file implements the error recovery as a core part of PCIe error
4 * reporting. When a PCIe error is delivered, an error message will be
5 * collected and printed to console, then, an error recovery procedure
6 * will be executed by following the PCI error recovery rules.
8 * Copyright (C) 2006 Intel Corp.
9 * Tom Long Nguyen (tom.l.nguyen@intel.com)
10 * Zhang Yanmin (yanmin.zhang@intel.com)
13 #define dev_fmt(fmt) "AER: " fmt
15 #include <linux/pci.h>
16 #include <linux/pm_runtime.h>
17 #include <linux/module.h>
18 #include <linux/kernel.h>
19 #include <linux/errno.h>
20 #include <linux/aer.h>
24 static pci_ers_result_t merge_result(enum pci_ers_result orig,
25 enum pci_ers_result new)
27 if (new == PCI_ERS_RESULT_NO_AER_DRIVER)
28 return PCI_ERS_RESULT_NO_AER_DRIVER;
30 if (new == PCI_ERS_RESULT_NONE)
34 case PCI_ERS_RESULT_CAN_RECOVER:
35 case PCI_ERS_RESULT_RECOVERED:
38 case PCI_ERS_RESULT_DISCONNECT:
39 if (new == PCI_ERS_RESULT_NEED_RESET)
40 orig = PCI_ERS_RESULT_NEED_RESET;
49 static int report_error_detected(struct pci_dev *dev,
50 pci_channel_state_t state,
51 enum pci_ers_result *result)
53 pci_ers_result_t vote;
54 const struct pci_error_handlers *err_handler;
56 device_lock(&dev->dev);
57 if (!pci_dev_set_io_state(dev, state) ||
59 !dev->driver->err_handler ||
60 !dev->driver->err_handler->error_detected) {
62 * If any device in the subtree does not have an error_detected
63 * callback, PCI_ERS_RESULT_NO_AER_DRIVER prevents subsequent
64 * error callbacks of "any" device in the subtree, and will
65 * exit in the disconnected error state.
67 if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE) {
68 vote = PCI_ERS_RESULT_NO_AER_DRIVER;
69 pci_info(dev, "can't recover (no error_detected callback)\n");
71 vote = PCI_ERS_RESULT_NONE;
74 err_handler = dev->driver->err_handler;
75 vote = err_handler->error_detected(dev, state);
77 pci_uevent_ers(dev, vote);
78 *result = merge_result(*result, vote);
79 device_unlock(&dev->dev);
83 static int pci_pm_runtime_get_sync(struct pci_dev *pdev, void *data)
85 pm_runtime_get_sync(&pdev->dev);
89 static int pci_pm_runtime_put(struct pci_dev *pdev, void *data)
91 pm_runtime_put(&pdev->dev);
95 static int report_frozen_detected(struct pci_dev *dev, void *data)
97 return report_error_detected(dev, pci_channel_io_frozen, data);
100 static int report_normal_detected(struct pci_dev *dev, void *data)
102 return report_error_detected(dev, pci_channel_io_normal, data);
105 static int report_mmio_enabled(struct pci_dev *dev, void *data)
107 pci_ers_result_t vote, *result = data;
108 const struct pci_error_handlers *err_handler;
110 device_lock(&dev->dev);
112 !dev->driver->err_handler ||
113 !dev->driver->err_handler->mmio_enabled)
116 err_handler = dev->driver->err_handler;
117 vote = err_handler->mmio_enabled(dev);
118 *result = merge_result(*result, vote);
120 device_unlock(&dev->dev);
124 static int report_slot_reset(struct pci_dev *dev, void *data)
126 pci_ers_result_t vote, *result = data;
127 const struct pci_error_handlers *err_handler;
129 device_lock(&dev->dev);
131 !dev->driver->err_handler ||
132 !dev->driver->err_handler->slot_reset)
135 err_handler = dev->driver->err_handler;
136 vote = err_handler->slot_reset(dev);
137 *result = merge_result(*result, vote);
139 device_unlock(&dev->dev);
143 static int report_resume(struct pci_dev *dev, void *data)
145 const struct pci_error_handlers *err_handler;
147 device_lock(&dev->dev);
148 if (!pci_dev_set_io_state(dev, pci_channel_io_normal) ||
150 !dev->driver->err_handler ||
151 !dev->driver->err_handler->resume)
154 err_handler = dev->driver->err_handler;
155 err_handler->resume(dev);
157 pci_uevent_ers(dev, PCI_ERS_RESULT_RECOVERED);
158 device_unlock(&dev->dev);
163 * pci_walk_bridge - walk bridges potentially AER affected
164 * @bridge: bridge which may be a Port or an RCEC
165 * @cb: callback to be called for each device found
166 * @userdata: arbitrary pointer to be passed to callback
168 * If the device provided is a bridge, walk the subordinate bus, including
169 * any bridged devices on buses under this bus. Call the provided callback
170 * on each device found.
172 * If the device provided has no subordinate bus, e.g., an RCEC, call the
173 * callback on the device itself.
175 static void pci_walk_bridge(struct pci_dev *bridge,
176 int (*cb)(struct pci_dev *, void *),
179 if (bridge->subordinate)
180 pci_walk_bus(bridge->subordinate, cb, userdata);
182 cb(bridge, userdata);
185 pci_ers_result_t pcie_do_recovery(struct pci_dev *dev,
186 pci_channel_state_t state,
187 pci_ers_result_t (*reset_subordinates)(struct pci_dev *pdev))
189 int type = pci_pcie_type(dev);
190 struct pci_dev *bridge;
191 pci_ers_result_t status = PCI_ERS_RESULT_CAN_RECOVER;
192 struct pci_host_bridge *host = pci_find_host_bridge(dev->bus);
195 * If the error was detected by a Root Port, Downstream Port, or
196 * RCEC, recovery runs on the device itself. For Ports, that also
197 * includes any subordinate devices.
199 * If it was detected by another device (Endpoint, etc), recovery
200 * runs on the device and anything else under the same Port, i.e.,
201 * everything under "bridge".
203 if (type == PCI_EXP_TYPE_ROOT_PORT ||
204 type == PCI_EXP_TYPE_DOWNSTREAM ||
205 type == PCI_EXP_TYPE_RC_EC)
208 bridge = pci_upstream_bridge(dev);
210 pci_walk_bridge(bridge, pci_pm_runtime_get_sync, NULL);
212 pci_dbg(bridge, "broadcast error_detected message\n");
213 if (state == pci_channel_io_frozen) {
214 pci_walk_bridge(bridge, report_frozen_detected, &status);
215 if (reset_subordinates(bridge) != PCI_ERS_RESULT_RECOVERED) {
216 pci_warn(bridge, "subordinate device reset failed\n");
220 pci_walk_bridge(bridge, report_normal_detected, &status);
223 if (status == PCI_ERS_RESULT_CAN_RECOVER) {
224 status = PCI_ERS_RESULT_RECOVERED;
225 pci_dbg(bridge, "broadcast mmio_enabled message\n");
226 pci_walk_bridge(bridge, report_mmio_enabled, &status);
229 if (status == PCI_ERS_RESULT_NEED_RESET) {
231 * TODO: Should call platform-specific
232 * functions to reset slot before calling
233 * drivers' slot_reset callbacks?
235 status = PCI_ERS_RESULT_RECOVERED;
236 pci_dbg(bridge, "broadcast slot_reset message\n");
237 pci_walk_bridge(bridge, report_slot_reset, &status);
240 if (status != PCI_ERS_RESULT_RECOVERED)
243 pci_dbg(bridge, "broadcast resume message\n");
244 pci_walk_bridge(bridge, report_resume, &status);
247 * If we have native control of AER, clear error status in the Root
248 * Port or Downstream Port that signaled the error. If the
249 * platform retained control of AER, it is responsible for clearing
250 * this status. In that case, the signaling device may not even be
253 if (host->native_aer || pcie_ports_native) {
254 pcie_clear_device_status(bridge);
255 pci_aer_clear_nonfatal_status(bridge);
258 pci_walk_bridge(bridge, pci_pm_runtime_put, NULL);
260 pci_info(bridge, "device recovery successful\n");
264 pci_walk_bridge(bridge, pci_pm_runtime_put, NULL);
266 pci_uevent_ers(bridge, PCI_ERS_RESULT_DISCONNECT);
268 /* TODO: Should kernel panic here? */
269 pci_info(bridge, "device recovery failed\n");