2 * drivers/pci/pcie/aer/aerdrv_core.c
4 * This file is subject to the terms and conditions of the GNU General Public
5 * License. See the file "COPYING" in the main directory of this archive
8 * This file implements the core part of PCI-Express AER. When an pci-express
9 * error is delivered, an error message will be collected and printed to
10 * console, then, an error recovery procedure will be executed by following
11 * the pci error recovery rules.
13 * Copyright (C) 2006 Intel Corp.
14 * Tom Long Nguyen (tom.l.nguyen@intel.com)
15 * Zhang Yanmin (yanmin.zhang@intel.com)
19 #include <linux/module.h>
20 #include <linux/pci.h>
21 #include <linux/kernel.h>
22 #include <linux/errno.h>
24 #include <linux/suspend.h>
25 #include <linux/delay.h>
29 module_param(forceload, bool, 0);
31 int pci_enable_pcie_error_reporting(struct pci_dev *dev)
36 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
40 pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
44 pci_read_config_word(dev, pos+PCI_EXP_DEVCTL, ®16);
47 PCI_EXP_DEVCTL_NFERE |
50 pci_write_config_word(dev, pos+PCI_EXP_DEVCTL,
55 int pci_disable_pcie_error_reporting(struct pci_dev *dev)
60 pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
64 pci_read_config_word(dev, pos+PCI_EXP_DEVCTL, ®16);
65 reg16 = reg16 & ~(PCI_EXP_DEVCTL_CERE |
66 PCI_EXP_DEVCTL_NFERE |
69 pci_write_config_word(dev, pos+PCI_EXP_DEVCTL,
74 int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
79 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
83 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
84 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &mask);
85 if (dev->error_state == pci_channel_io_normal)
86 status &= ~mask; /* Clear corresponding nonfatal bits */
88 status &= mask; /* Clear corresponding fatal bits */
89 pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
95 int pci_cleanup_aer_correct_error_status(struct pci_dev *dev)
100 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
104 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
105 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS, status);
112 static void set_device_error_reporting(struct pci_dev *dev, void *data)
114 bool enable = *((bool *)data);
116 if (dev->pcie_type != PCIE_RC_PORT &&
117 dev->pcie_type != PCIE_SW_UPSTREAM_PORT &&
118 dev->pcie_type != PCIE_SW_DOWNSTREAM_PORT)
122 pci_enable_pcie_error_reporting(dev);
124 pci_disable_pcie_error_reporting(dev);
128 * set_downstream_devices_error_reporting - enable/disable the error reporting bits on the root port and its downstream ports.
129 * @dev: pointer to root port's pci_dev data structure
130 * @enable: true = enable error reporting, false = disable error reporting.
132 static void set_downstream_devices_error_reporting(struct pci_dev *dev,
135 set_device_error_reporting(dev, &enable);
136 pci_walk_bus(dev->subordinate, set_device_error_reporting, &enable);
139 static int find_device_iter(struct device *device, void *data)
142 u16 id = *(unsigned long *)data;
143 u8 secondary, subordinate, d_bus = id >> 8;
145 if (device->bus == &pci_bus_type) {
146 dev = to_pci_dev(device);
147 if (id == ((dev->bus->number << 8) | dev->devfn)) {
151 *(unsigned long*)data = (unsigned long)device;
156 * If device is P2P, check if it is an upstream?
158 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
159 pci_read_config_byte(dev, PCI_SECONDARY_BUS,
161 pci_read_config_byte(dev, PCI_SUBORDINATE_BUS,
163 if (d_bus >= secondary && d_bus <= subordinate) {
164 *(unsigned long*)data = (unsigned long)device;
174 * find_source_device - search through device hierarchy for source device
175 * @parent: pointer to Root Port pci_dev data structure
176 * @id: device ID of agent who sends an error message to this Root Port
178 * Invoked when error is detected at the Root Port.
180 static struct device* find_source_device(struct pci_dev *parent, u16 id)
182 struct pci_dev *dev = parent;
183 struct device *device;
184 unsigned long device_addr;
187 /* Is Root Port an agent that sends error message? */
188 if (id == ((dev->bus->number << 8) | dev->devfn))
193 if ((status = device_for_each_child(&dev->dev,
194 &device_addr, find_device_iter))) {
195 device = (struct device*)device_addr;
196 dev = to_pci_dev(device);
197 if (id == ((dev->bus->number << 8) | dev->devfn))
205 static void report_error_detected(struct pci_dev *dev, void *data)
207 pci_ers_result_t vote;
208 struct pci_error_handlers *err_handler;
209 struct aer_broadcast_data *result_data;
210 result_data = (struct aer_broadcast_data *) data;
212 dev->error_state = result_data->state;
215 !dev->driver->err_handler ||
216 !dev->driver->err_handler->error_detected) {
217 if (result_data->state == pci_channel_io_frozen &&
218 !(dev->hdr_type & PCI_HEADER_TYPE_BRIDGE)) {
220 * In case of fatal recovery, if one of down-
221 * stream device has no driver. We might be
222 * unable to recover because a later insmod
223 * of a driver for this device is unaware of
226 dev_printk(KERN_DEBUG, &dev->dev, "device has %s\n",
228 "no AER-aware driver" : "no driver");
233 err_handler = dev->driver->err_handler;
234 vote = err_handler->error_detected(dev, result_data->state);
235 result_data->result = merge_result(result_data->result, vote);
239 static void report_mmio_enabled(struct pci_dev *dev, void *data)
241 pci_ers_result_t vote;
242 struct pci_error_handlers *err_handler;
243 struct aer_broadcast_data *result_data;
244 result_data = (struct aer_broadcast_data *) data;
247 !dev->driver->err_handler ||
248 !dev->driver->err_handler->mmio_enabled)
251 err_handler = dev->driver->err_handler;
252 vote = err_handler->mmio_enabled(dev);
253 result_data->result = merge_result(result_data->result, vote);
257 static void report_slot_reset(struct pci_dev *dev, void *data)
259 pci_ers_result_t vote;
260 struct pci_error_handlers *err_handler;
261 struct aer_broadcast_data *result_data;
262 result_data = (struct aer_broadcast_data *) data;
265 !dev->driver->err_handler ||
266 !dev->driver->err_handler->slot_reset)
269 err_handler = dev->driver->err_handler;
270 vote = err_handler->slot_reset(dev);
271 result_data->result = merge_result(result_data->result, vote);
275 static void report_resume(struct pci_dev *dev, void *data)
277 struct pci_error_handlers *err_handler;
279 dev->error_state = pci_channel_io_normal;
282 !dev->driver->err_handler ||
283 !dev->driver->err_handler->resume)
286 err_handler = dev->driver->err_handler;
287 err_handler->resume(dev);
292 * broadcast_error_message - handle message broadcast to downstream drivers
293 * @dev: pointer to from where in a hierarchy message is broadcasted down
294 * @state: error state
295 * @error_mesg: message to print
296 * @cb: callback to be broadcasted
298 * Invoked during error recovery process. Once being invoked, the content
299 * of error severity will be broadcasted to all downstream drivers in a
300 * hierarchy in question.
302 static pci_ers_result_t broadcast_error_message(struct pci_dev *dev,
303 enum pci_channel_state state,
305 void (*cb)(struct pci_dev *, void *))
307 struct aer_broadcast_data result_data;
309 dev_printk(KERN_DEBUG, &dev->dev, "broadcast %s message\n", error_mesg);
310 result_data.state = state;
311 if (cb == report_error_detected)
312 result_data.result = PCI_ERS_RESULT_CAN_RECOVER;
314 result_data.result = PCI_ERS_RESULT_RECOVERED;
316 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
318 * If the error is reported by a bridge, we think this error
319 * is related to the downstream link of the bridge, so we
320 * do error recovery on all subordinates of the bridge instead
321 * of the bridge and clear the error status of the bridge.
323 if (cb == report_error_detected)
324 dev->error_state = state;
325 pci_walk_bus(dev->subordinate, cb, &result_data);
326 if (cb == report_resume) {
327 pci_cleanup_aer_uncorrect_error_status(dev);
328 dev->error_state = pci_channel_io_normal;
333 * If the error is reported by an end point, we think this
334 * error is related to the upstream link of the end point.
336 pci_walk_bus(dev->bus, cb, &result_data);
339 return result_data.result;
342 struct find_aer_service_data {
343 struct pcie_port_service_driver *aer_driver;
347 static int find_aer_service_iter(struct device *device, void *data)
349 struct device_driver *driver;
350 struct pcie_port_service_driver *service_driver;
351 struct pcie_device *pcie_dev;
352 struct find_aer_service_data *result;
354 result = (struct find_aer_service_data *) data;
356 if (device->bus == &pcie_port_bus_type) {
357 pcie_dev = to_pcie_device(device);
358 if (pcie_dev->id.port_type == PCIE_SW_DOWNSTREAM_PORT)
359 result->is_downstream = 1;
361 driver = device->driver;
363 service_driver = to_service_driver(driver);
364 if (service_driver->id_table->service_type ==
365 PCIE_PORT_SERVICE_AER) {
366 result->aer_driver = service_driver;
375 static void find_aer_service(struct pci_dev *dev,
376 struct find_aer_service_data *data)
379 retval = device_for_each_child(&dev->dev, data, find_aer_service_iter);
382 static pci_ers_result_t reset_link(struct pcie_device *aerdev,
385 struct pci_dev *udev;
386 pci_ers_result_t status;
387 struct find_aer_service_data data;
389 if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE)
392 udev= dev->bus->self;
394 data.is_downstream = 0;
395 data.aer_driver = NULL;
396 find_aer_service(udev, &data);
399 * Use the aer driver of the error agent firstly.
400 * If it hasn't the aer driver, use the root port's
402 if (!data.aer_driver || !data.aer_driver->reset_link) {
403 if (data.is_downstream &&
404 aerdev->device.driver &&
405 to_service_driver(aerdev->device.driver)->reset_link) {
407 to_service_driver(aerdev->device.driver);
409 dev_printk(KERN_DEBUG, &dev->dev, "no link-reset "
411 return PCI_ERS_RESULT_DISCONNECT;
415 status = data.aer_driver->reset_link(udev);
416 if (status != PCI_ERS_RESULT_RECOVERED) {
417 dev_printk(KERN_DEBUG, &dev->dev, "link reset at upstream "
418 "device %s failed\n", pci_name(udev));
419 return PCI_ERS_RESULT_DISCONNECT;
426 * do_recovery - handle nonfatal/fatal error recovery process
427 * @aerdev: pointer to a pcie_device data structure of root port
428 * @dev: pointer to a pci_dev data structure of agent detecting an error
429 * @severity: error severity type
431 * Invoked when an error is nonfatal/fatal. Once being invoked, broadcast
432 * error detected message to all downstream drivers within a hierarchy in
433 * question and return the returned code.
435 static pci_ers_result_t do_recovery(struct pcie_device *aerdev,
439 pci_ers_result_t status, result = PCI_ERS_RESULT_RECOVERED;
440 enum pci_channel_state state;
442 if (severity == AER_FATAL)
443 state = pci_channel_io_frozen;
445 state = pci_channel_io_normal;
447 status = broadcast_error_message(dev,
450 report_error_detected);
452 if (severity == AER_FATAL) {
453 result = reset_link(aerdev, dev);
454 if (result != PCI_ERS_RESULT_RECOVERED) {
455 /* TODO: Should panic here? */
460 if (status == PCI_ERS_RESULT_CAN_RECOVER)
461 status = broadcast_error_message(dev,
464 report_mmio_enabled);
466 if (status == PCI_ERS_RESULT_NEED_RESET) {
468 * TODO: Should call platform-specific
469 * functions to reset slot before calling
470 * drivers' slot_reset callbacks?
472 status = broadcast_error_message(dev,
478 if (status == PCI_ERS_RESULT_RECOVERED)
479 broadcast_error_message(dev,
488 * handle_error_source - handle logging error into an event log
489 * @aerdev: pointer to pcie_device data structure of the root port
490 * @dev: pointer to pci_dev data structure of error source device
491 * @info: comprehensive error information
493 * Invoked when an error being detected by Root Port.
495 static void handle_error_source(struct pcie_device * aerdev,
497 struct aer_err_info info)
499 pci_ers_result_t status = 0;
502 if (info.severity == AER_CORRECTABLE) {
504 * Correctable error does not need software intevention.
505 * No need to go through error recovery process.
507 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
509 pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
512 status = do_recovery(aerdev, dev, info.severity);
513 if (status == PCI_ERS_RESULT_RECOVERED) {
514 dev_printk(KERN_DEBUG, &dev->dev, "AER driver "
515 "successfully recovered\n");
517 /* TODO: Should kernel panic here? */
518 dev_printk(KERN_DEBUG, &dev->dev, "AER driver didn't "
525 * aer_enable_rootport - enable Root Port's interrupts when receiving messages
526 * @rpc: pointer to a Root Port data structure
528 * Invoked when PCIE bus loads AER service driver.
530 void aer_enable_rootport(struct aer_rpc *rpc)
532 struct pci_dev *pdev = rpc->rpd->port;
537 pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
538 /* Clear PCIE Capability's Device Status */
539 pci_read_config_word(pdev, pos+PCI_EXP_DEVSTA, ®16);
540 pci_write_config_word(pdev, pos+PCI_EXP_DEVSTA, reg16);
542 /* Disable system error generation in response to error messages */
543 pci_read_config_word(pdev, pos + PCI_EXP_RTCTL, ®16);
544 reg16 &= ~(SYSTEM_ERROR_INTR_ON_MESG_MASK);
545 pci_write_config_word(pdev, pos + PCI_EXP_RTCTL, reg16);
547 aer_pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ERR);
548 /* Clear error status */
549 pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, ®32);
550 pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, reg32);
551 pci_read_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, ®32);
552 pci_write_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, reg32);
553 pci_read_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, ®32);
554 pci_write_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, reg32);
557 * Enable error reporting for the root port device and downstream port
560 set_downstream_devices_error_reporting(pdev, true);
562 /* Enable Root Port's interrupt in response to error messages */
563 pci_write_config_dword(pdev,
564 aer_pos + PCI_ERR_ROOT_COMMAND,
565 ROOT_PORT_INTR_ON_MESG_MASK);
569 * disable_root_aer - disable Root Port's interrupts when receiving messages
570 * @rpc: pointer to a Root Port data structure
572 * Invoked when PCIE bus unloads AER service driver.
574 static void disable_root_aer(struct aer_rpc *rpc)
576 struct pci_dev *pdev = rpc->rpd->port;
581 * Disable error reporting for the root port device and downstream port
584 set_downstream_devices_error_reporting(pdev, false);
586 pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ERR);
587 /* Disable Root's interrupt in response to error messages */
588 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, 0);
590 /* Clear Root's error status reg */
591 pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, ®32);
592 pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, reg32);
596 * get_e_source - retrieve an error source
597 * @rpc: pointer to the root port which holds an error
599 * Invoked by DPC handler to consume an error.
601 static struct aer_err_source* get_e_source(struct aer_rpc *rpc)
603 struct aer_err_source *e_source;
606 /* Lock access to Root error producer/consumer index */
607 spin_lock_irqsave(&rpc->e_lock, flags);
608 if (rpc->prod_idx == rpc->cons_idx) {
609 spin_unlock_irqrestore(&rpc->e_lock, flags);
612 e_source = &rpc->e_sources[rpc->cons_idx];
614 if (rpc->cons_idx == AER_ERROR_SOURCES_MAX)
616 spin_unlock_irqrestore(&rpc->e_lock, flags);
621 static int get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
625 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
627 /* The device might not support AER */
631 if (info->severity == AER_CORRECTABLE) {
632 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS,
634 if (!(info->status & ERR_CORRECTABLE_ERROR_MASK))
635 return AER_UNSUCCESS;
636 } else if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE ||
637 info->severity == AER_NONFATAL) {
639 /* Link is still healthy for IO reads */
640 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
642 if (!(info->status & ERR_UNCORRECTABLE_ERROR_MASK))
643 return AER_UNSUCCESS;
645 if (info->status & AER_LOG_TLP_MASKS) {
646 info->flags |= AER_TLP_HEADER_VALID_FLAG;
647 pci_read_config_dword(dev,
648 pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0);
649 pci_read_config_dword(dev,
650 pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1);
651 pci_read_config_dword(dev,
652 pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2);
653 pci_read_config_dword(dev,
654 pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3);
662 * aer_isr_one_error - consume an error detected by root port
663 * @p_device: pointer to error root port service device
664 * @e_src: pointer to an error source
666 static void aer_isr_one_error(struct pcie_device *p_device,
667 struct aer_err_source *e_src)
669 struct device *s_device;
670 struct aer_err_info e_info = {0, 0, 0,};
675 * There is a possibility that both correctable error and
676 * uncorrectable error being logged. Report correctable error first.
678 for (i = 1; i & ROOT_ERR_STATUS_MASKS ; i <<= 2) {
681 if (!(e_src->status & i))
684 /* Init comprehensive error information */
685 if (i & PCI_ERR_ROOT_COR_RCV) {
686 id = ERR_COR_ID(e_src->id);
687 e_info.severity = AER_CORRECTABLE;
689 id = ERR_UNCOR_ID(e_src->id);
690 e_info.severity = ((e_src->status >> 6) & 1);
693 (PCI_ERR_ROOT_MULTI_COR_RCV |
694 PCI_ERR_ROOT_MULTI_UNCOR_RCV))
695 e_info.flags |= AER_MULTI_ERROR_VALID_FLAG;
696 if (!(s_device = find_source_device(p_device->port, id))) {
697 printk(KERN_DEBUG "%s->can't find device of ID%04x\n",
701 if (get_device_error_info(to_pci_dev(s_device), &e_info) ==
703 aer_print_error(to_pci_dev(s_device), &e_info);
704 handle_error_source(p_device,
705 to_pci_dev(s_device),
712 * aer_isr - consume errors detected by root port
713 * @work: definition of this work item
715 * Invoked, as DPC, when root port records new detected error
717 void aer_isr(struct work_struct *work)
719 struct aer_rpc *rpc = container_of(work, struct aer_rpc, dpc_handler);
720 struct pcie_device *p_device = rpc->rpd;
721 struct aer_err_source *e_src;
723 mutex_lock(&rpc->rpc_mutex);
724 e_src = get_e_source(rpc);
726 aer_isr_one_error(p_device, e_src);
727 e_src = get_e_source(rpc);
729 mutex_unlock(&rpc->rpc_mutex);
731 wake_up(&rpc->wait_release);
735 * aer_delete_rootport - disable root port aer and delete service data
736 * @rpc: pointer to a root port device being deleted
738 * Invoked when AER service unloaded on a specific Root Port
740 void aer_delete_rootport(struct aer_rpc *rpc)
742 /* Disable root port AER itself */
743 disable_root_aer(rpc);
749 * aer_init - provide AER initialization
750 * @dev: pointer to AER pcie device
752 * Invoked when AER service driver is loaded.
754 int aer_init(struct pcie_device *dev)
756 if (aer_osc_setup(dev) && !forceload)
762 EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting);
763 EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting);
764 EXPORT_SYMBOL_GPL(pci_cleanup_aer_uncorrect_error_status);