drivers/pci/pcie/aer/aerdrv_core.c

   1 /*
   2  * drivers/pci/pcie/aer/aerdrv_core.c
   3  *
   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
   6  * for more details.
   7  *
   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.
  12  *
  13  * Copyright (C) 2006 Intel Corp.
  14  *      Tom Long Nguyen (tom.l.nguyen@intel.com)
  15  *      Zhang Yanmin (yanmin.zhang@intel.com)
  16  *
  17  */
  18
  19 #include <linux/module.h>
  20 #include <linux/pci.h>
  21 #include <linux/kernel.h>
  22 #include <linux/errno.h>
  23 #include <linux/pm.h>
  24 #include <linux/suspend.h>
  25 #include <linux/delay.h>
  26 #include "aerdrv.h"
  27
  28 static int forceload;
  29 static int nosourceid;
  30 module_param(forceload, bool, 0);
  31 module_param(nosourceid, bool, 0);
  32
  33 int pci_enable_pcie_error_reporting(struct pci_dev *dev)
  34 {
  35         u16 reg16 = 0;
  36         int pos;
  37
  38         pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
  39         if (!pos)
  40                 return -EIO;
  41
  42         pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
  43         if (!pos)
  44                 return -EIO;
  45
  46         pci_read_config_word(dev, pos+PCI_EXP_DEVCTL, &reg16);
  47         reg16 = reg16 |
  48                 PCI_EXP_DEVCTL_CERE |
  49                 PCI_EXP_DEVCTL_NFERE |
  50                 PCI_EXP_DEVCTL_FERE |
  51                 PCI_EXP_DEVCTL_URRE;
  52         pci_write_config_word(dev, pos+PCI_EXP_DEVCTL, reg16);
  53
  54         return 0;
  55 }
  56 EXPORT_SYMBOL_GPL(pci_enable_pcie_error_reporting);
  57
  58 int pci_disable_pcie_error_reporting(struct pci_dev *dev)
  59 {
  60         u16 reg16 = 0;
  61         int pos;
  62
  63         pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
  64         if (!pos)
  65                 return -EIO;
  66
  67         pci_read_config_word(dev, pos+PCI_EXP_DEVCTL, &reg16);
  68         reg16 = reg16 & ~(PCI_EXP_DEVCTL_CERE |
  69                         PCI_EXP_DEVCTL_NFERE |
  70                         PCI_EXP_DEVCTL_FERE |
  71                         PCI_EXP_DEVCTL_URRE);
  72         pci_write_config_word(dev, pos+PCI_EXP_DEVCTL, reg16);
  73
  74         return 0;
  75 }
  76 EXPORT_SYMBOL_GPL(pci_disable_pcie_error_reporting);
  77
  78 int pci_cleanup_aer_uncorrect_error_status(struct pci_dev *dev)
  79 {
  80         int pos;
  81         u32 status, mask;
  82
  83         pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
  84         if (!pos)
  85                 return -EIO;
  86
  87         pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
  88         pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_SEVER, &mask);
  89         if (dev->error_state == pci_channel_io_normal)
  90                 status &= ~mask; /* Clear corresponding nonfatal bits */
  91         else
  92                 status &= mask; /* Clear corresponding fatal bits */
  93         pci_write_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, status);
  94
  95         return 0;
  96 }
  97 EXPORT_SYMBOL_GPL(pci_cleanup_aer_uncorrect_error_status);
  98
  99 #if 0
 100 int pci_cleanup_aer_correct_error_status(struct pci_dev *dev)
 101 {
 102         int pos;
 103         u32 status;
 104
 105         pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
 106         if (!pos)
 107                 return -EIO;
 108
 109         pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
 110         pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS, status);
 111
 112         return 0;
 113 }
 114 #endif  /*  0  */
 115
 116 static int set_device_error_reporting(struct pci_dev *dev, void *data)
 117 {
 118         bool enable = *((bool *)data);
 119
 120         if (dev->pcie_type == PCIE_RC_PORT ||
 121             dev->pcie_type == PCIE_SW_UPSTREAM_PORT ||
 122             dev->pcie_type == PCIE_SW_DOWNSTREAM_PORT) {
 123                 if (enable)
 124                         pci_enable_pcie_error_reporting(dev);
 125                 else
 126                         pci_disable_pcie_error_reporting(dev);
 127         }
 128
 129         if (enable)
 130                 pcie_set_ecrc_checking(dev);
 131
 132         return 0;
 133 }
 134
 135 /**
 136  * set_downstream_devices_error_reporting - enable/disable the error reporting  bits on the root port and its downstream ports.
 137  * @dev: pointer to root port's pci_dev data structure
 138  * @enable: true = enable error reporting, false = disable error reporting.
 139  */
 140 static void set_downstream_devices_error_reporting(struct pci_dev *dev,
 141                                                    bool enable)
 142 {
 143         set_device_error_reporting(dev, &enable);
 144
 145         if (!dev->subordinate)
 146                 return;
 147         pci_walk_bus(dev->subordinate, set_device_error_reporting, &enable);
 148 }
 149
 150 static inline int compare_device_id(struct pci_dev *dev,
 151                         struct aer_err_info *e_info)
 152 {
 153         if (e_info->id == ((dev->bus->number << 8) | dev->devfn)) {
 154                 /*
 155                  * Device ID match
 156                  */
 157                 return 1;
 158         }
 159
 160         return 0;
 161 }
 162
 163 static int add_error_device(struct aer_err_info *e_info, struct pci_dev *dev)
 164 {
 165         if (e_info->error_dev_num < AER_MAX_MULTI_ERR_DEVICES) {
 166                 e_info->dev[e_info->error_dev_num] = dev;
 167                 e_info->error_dev_num++;
 168                 return 1;
 169         }
 170
 171         return 0;
 172 }
 173
 174
 175 #define PCI_BUS(x)      (((x) >> 8) & 0xff)
 176
 177 static int find_device_iter(struct pci_dev *dev, void *data)
 178 {
 179         int pos;
 180         u32 status;
 181         u32 mask;
 182         u16 reg16;
 183         int result;
 184         struct aer_err_info *e_info = (struct aer_err_info *)data;
 185
 186         /*
 187          * When bus id is equal to 0, it might be a bad id
 188          * reported by root port.
 189          */
 190         if (!nosourceid && (PCI_BUS(e_info->id) != 0)) {
 191                 result = compare_device_id(dev, e_info);
 192                 if (result)
 193                         add_error_device(e_info, dev);
 194
 195                 /*
 196                  * If there is no multiple error, we stop
 197                  * or continue based on the id comparing.
 198                  */
 199                 if (!e_info->multi_error_valid)
 200                         return result;
 201
 202                 /*
 203                  * If there are multiple errors and id does match,
 204                  * We need continue to search other devices under
 205                  * the root port. Return 0 means that.
 206                  */
 207                 if (result)
 208                         return 0;
 209         }
 210
 211         /*
 212          * When either
 213          *      1) nosourceid==y;
 214          *      2) bus id is equal to 0. Some ports might lose the bus
 215          *              id of error source id;
 216          *      3) There are multiple errors and prior id comparing fails;
 217          * We check AER status registers to find the initial reporter.
 218          */
 219         if (atomic_read(&dev->enable_cnt) == 0)
 220                 return 0;
 221         pos = pci_find_capability(dev, PCI_CAP_ID_EXP);
 222         if (!pos)
 223                 return 0;
 224         /* Check if AER is enabled */
 225         pci_read_config_word(dev, pos+PCI_EXP_DEVCTL, &reg16);
 226         if (!(reg16 & (
 227                 PCI_EXP_DEVCTL_CERE |
 228                 PCI_EXP_DEVCTL_NFERE |
 229                 PCI_EXP_DEVCTL_FERE |
 230                 PCI_EXP_DEVCTL_URRE)))
 231                 return 0;
 232         pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
 233         if (!pos)
 234                 return 0;
 235
 236         status = 0;
 237         mask = 0;
 238         if (e_info->severity == AER_CORRECTABLE) {
 239                 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS, &status);
 240                 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK, &mask);
 241                 if (status & ~mask) {
 242                         add_error_device(e_info, dev);
 243                         goto added;
 244                 }
 245         } else {
 246                 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS, &status);
 247                 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK, &mask);
 248                 if (status & ~mask) {
 249                         add_error_device(e_info, dev);
 250                         goto added;
 251                 }
 252         }
 253
 254         return 0;
 255
 256 added:
 257         if (e_info->multi_error_valid)
 258                 return 0;
 259         else
 260                 return 1;
 261 }
 262
 263 /**
 264  * find_source_device - search through device hierarchy for source device
 265  * @parent: pointer to Root Port pci_dev data structure
 266  * @err_info: including detailed error information such like id
 267  *
 268  * Invoked when error is detected at the Root Port.
 269  */
 270 static void find_source_device(struct pci_dev *parent,
 271                 struct aer_err_info *e_info)
 272 {
 273         struct pci_dev *dev = parent;
 274         int result;
 275
 276         /* Is Root Port an agent that sends error message? */
 277         result = find_device_iter(dev, e_info);
 278         if (result)
 279                 return;
 280
 281         pci_walk_bus(parent->subordinate, find_device_iter, e_info);
 282 }
 283
 284 static int report_error_detected(struct pci_dev *dev, void *data)
 285 {
 286         pci_ers_result_t vote;
 287         struct pci_error_handlers *err_handler;
 288         struct aer_broadcast_data *result_data;
 289         result_data = (struct aer_broadcast_data *) data;
 290
 291         dev->error_state = result_data->state;
 292
 293         if (!dev->driver ||
 294                 !dev->driver->err_handler ||
 295                 !dev->driver->err_handler->error_detected) {
 296                 if (result_data->state == pci_channel_io_frozen &&
 297                         !(dev->hdr_type & PCI_HEADER_TYPE_BRIDGE)) {
 298                         /*
 299                          * In case of fatal recovery, if one of down-
 300                          * stream device has no driver. We might be
 301                          * unable to recover because a later insmod
 302                          * of a driver for this device is unaware of
 303                          * its hw state.
 304                          */
 305                         dev_printk(KERN_DEBUG, &dev->dev, "device has %s\n",
 306                                    dev->driver ?
 307                                    "no AER-aware driver" : "no driver");
 308                 }
 309                 return 0;
 310         }
 311
 312         err_handler = dev->driver->err_handler;
 313         vote = err_handler->error_detected(dev, result_data->state);
 314         result_data->result = merge_result(result_data->result, vote);
 315         return 0;
 316 }
 317
 318 static int report_mmio_enabled(struct pci_dev *dev, void *data)
 319 {
 320         pci_ers_result_t vote;
 321         struct pci_error_handlers *err_handler;
 322         struct aer_broadcast_data *result_data;
 323         result_data = (struct aer_broadcast_data *) data;
 324
 325         if (!dev->driver ||
 326                 !dev->driver->err_handler ||
 327                 !dev->driver->err_handler->mmio_enabled)
 328                 return 0;
 329
 330         err_handler = dev->driver->err_handler;
 331         vote = err_handler->mmio_enabled(dev);
 332         result_data->result = merge_result(result_data->result, vote);
 333         return 0;
 334 }
 335
 336 static int report_slot_reset(struct pci_dev *dev, void *data)
 337 {
 338         pci_ers_result_t vote;
 339         struct pci_error_handlers *err_handler;
 340         struct aer_broadcast_data *result_data;
 341         result_data = (struct aer_broadcast_data *) data;
 342
 343         if (!dev->driver ||
 344                 !dev->driver->err_handler ||
 345                 !dev->driver->err_handler->slot_reset)
 346                 return 0;
 347
 348         err_handler = dev->driver->err_handler;
 349         vote = err_handler->slot_reset(dev);
 350         result_data->result = merge_result(result_data->result, vote);
 351         return 0;
 352 }
 353
 354 static int report_resume(struct pci_dev *dev, void *data)
 355 {
 356         struct pci_error_handlers *err_handler;
 357
 358         dev->error_state = pci_channel_io_normal;
 359
 360         if (!dev->driver ||
 361                 !dev->driver->err_handler ||
 362                 !dev->driver->err_handler->resume)
 363                 return 0;
 364
 365         err_handler = dev->driver->err_handler;
 366         err_handler->resume(dev);
 367         return 0;
 368 }
 369
 370 /**
 371  * broadcast_error_message - handle message broadcast to downstream drivers
 372  * @dev: pointer to from where in a hierarchy message is broadcasted down
 373  * @state: error state
 374  * @error_mesg: message to print
 375  * @cb: callback to be broadcasted
 376  *
 377  * Invoked during error recovery process. Once being invoked, the content
 378  * of error severity will be broadcasted to all downstream drivers in a
 379  * hierarchy in question.
 380  */
 381 static pci_ers_result_t broadcast_error_message(struct pci_dev *dev,
 382         enum pci_channel_state state,
 383         char *error_mesg,
 384         int (*cb)(struct pci_dev *, void *))
 385 {
 386         struct aer_broadcast_data result_data;
 387
 388         dev_printk(KERN_DEBUG, &dev->dev, "broadcast %s message\n", error_mesg);
 389         result_data.state = state;
 390         if (cb == report_error_detected)
 391                 result_data.result = PCI_ERS_RESULT_CAN_RECOVER;
 392         else
 393                 result_data.result = PCI_ERS_RESULT_RECOVERED;
 394
 395         if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE) {
 396                 /*
 397                  * If the error is reported by a bridge, we think this error
 398                  * is related to the downstream link of the bridge, so we
 399                  * do error recovery on all subordinates of the bridge instead
 400                  * of the bridge and clear the error status of the bridge.
 401                  */
 402                 if (cb == report_error_detected)
 403                         dev->error_state = state;
 404                 pci_walk_bus(dev->subordinate, cb, &result_data);
 405                 if (cb == report_resume) {
 406                         pci_cleanup_aer_uncorrect_error_status(dev);
 407                         dev->error_state = pci_channel_io_normal;
 408                 }
 409         } else {
 410                 /*
 411                  * If the error is reported by an end point, we think this
 412                  * error is related to the upstream link of the end point.
 413                  */
 414                 pci_walk_bus(dev->bus, cb, &result_data);
 415         }
 416
 417         return result_data.result;
 418 }
 419
 420 struct find_aer_service_data {
 421         struct pcie_port_service_driver *aer_driver;
 422         int is_downstream;
 423 };
 424
 425 static int find_aer_service_iter(struct device *device, void *data)
 426 {
 427         struct device_driver *driver;
 428         struct pcie_port_service_driver *service_driver;
 429         struct find_aer_service_data *result;
 430
 431         result = (struct find_aer_service_data *) data;
 432
 433         if (device->bus == &pcie_port_bus_type) {
 434                 struct pcie_port_data *port_data;
 435
 436                 port_data = pci_get_drvdata(to_pcie_device(device)->port);
 437                 if (port_data->port_type == PCIE_SW_DOWNSTREAM_PORT)
 438                         result->is_downstream = 1;
 439
 440                 driver = device->driver;
 441                 if (driver) {
 442                         service_driver = to_service_driver(driver);
 443                         if (service_driver->service == PCIE_PORT_SERVICE_AER) {
 444                                 result->aer_driver = service_driver;
 445                                 return 1;
 446                         }
 447                 }
 448         }
 449
 450         return 0;
 451 }
 452
 453 static void find_aer_service(struct pci_dev *dev,
 454                 struct find_aer_service_data *data)
 455 {
 456         int retval;
 457         retval = device_for_each_child(&dev->dev, data, find_aer_service_iter);
 458 }
 459
 460 static pci_ers_result_t reset_link(struct pcie_device *aerdev,
 461                 struct pci_dev *dev)
 462 {
 463         struct pci_dev *udev;
 464         pci_ers_result_t status;
 465         struct find_aer_service_data data;
 466
 467         if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE)
 468                 udev = dev;
 469         else
 470                 udev = dev->bus->self;
 471
 472         data.is_downstream = 0;
 473         data.aer_driver = NULL;
 474         find_aer_service(udev, &data);
 475
 476         /*
 477          * Use the aer driver of the error agent firstly.
 478          * If it hasn't the aer driver, use the root port's
 479          */
 480         if (!data.aer_driver || !data.aer_driver->reset_link) {
 481                 if (data.is_downstream &&
 482                         aerdev->device.driver &&
 483                         to_service_driver(aerdev->device.driver)->reset_link) {
 484                         data.aer_driver =
 485                                 to_service_driver(aerdev->device.driver);
 486                 } else {
 487                         dev_printk(KERN_DEBUG, &dev->dev, "no link-reset "
 488                                    "support\n");
 489                         return PCI_ERS_RESULT_DISCONNECT;
 490                 }
 491         }
 492
 493         status = data.aer_driver->reset_link(udev);
 494         if (status != PCI_ERS_RESULT_RECOVERED) {
 495                 dev_printk(KERN_DEBUG, &dev->dev, "link reset at upstream "
 496                            "device %s failed\n", pci_name(udev));
 497                 return PCI_ERS_RESULT_DISCONNECT;
 498         }
 499
 500         return status;
 501 }
 502
 503 /**
 504  * do_recovery - handle nonfatal/fatal error recovery process
 505  * @aerdev: pointer to a pcie_device data structure of root port
 506  * @dev: pointer to a pci_dev data structure of agent detecting an error
 507  * @severity: error severity type
 508  *
 509  * Invoked when an error is nonfatal/fatal. Once being invoked, broadcast
 510  * error detected message to all downstream drivers within a hierarchy in
 511  * question and return the returned code.
 512  */
 513 static pci_ers_result_t do_recovery(struct pcie_device *aerdev,
 514                 struct pci_dev *dev,
 515                 int severity)
 516 {
 517         pci_ers_result_t status, result = PCI_ERS_RESULT_RECOVERED;
 518         enum pci_channel_state state;
 519
 520         if (severity == AER_FATAL)
 521                 state = pci_channel_io_frozen;
 522         else
 523                 state = pci_channel_io_normal;
 524
 525         status = broadcast_error_message(dev,
 526                         state,
 527                         "error_detected",
 528                         report_error_detected);
 529
 530         if (severity == AER_FATAL) {
 531                 result = reset_link(aerdev, dev);
 532                 if (result != PCI_ERS_RESULT_RECOVERED) {
 533                         /* TODO: Should panic here? */
 534                         return result;
 535                 }
 536         }
 537
 538         if (status == PCI_ERS_RESULT_CAN_RECOVER)
 539                 status = broadcast_error_message(dev,
 540                                 state,
 541                                 "mmio_enabled",
 542                                 report_mmio_enabled);
 543
 544         if (status == PCI_ERS_RESULT_NEED_RESET) {
 545                 /*
 546                  * TODO: Should call platform-specific
 547                  * functions to reset slot before calling
 548                  * drivers' slot_reset callbacks?
 549                  */
 550                 status = broadcast_error_message(dev,
 551                                 state,
 552                                 "slot_reset",
 553                                 report_slot_reset);
 554         }
 555
 556         if (status == PCI_ERS_RESULT_RECOVERED)
 557                 broadcast_error_message(dev,
 558                                 state,
 559                                 "resume",
 560                                 report_resume);
 561
 562         return status;
 563 }
 564
 565 /**
 566  * handle_error_source - handle logging error into an event log
 567  * @aerdev: pointer to pcie_device data structure of the root port
 568  * @dev: pointer to pci_dev data structure of error source device
 569  * @info: comprehensive error information
 570  *
 571  * Invoked when an error being detected by Root Port.
 572  */
 573 static void handle_error_source(struct pcie_device *aerdev,
 574         struct pci_dev *dev,
 575         struct aer_err_info *info)
 576 {
 577         pci_ers_result_t status = 0;
 578         int pos;
 579
 580         if (info->severity == AER_CORRECTABLE) {
 581                 /*
 582                  * Correctable error does not need software intevention.
 583                  * No need to go through error recovery process.
 584                  */
 585                 pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
 586                 if (pos)
 587                         pci_write_config_dword(dev, pos + PCI_ERR_COR_STATUS,
 588                                         info->status);
 589         } else {
 590                 status = do_recovery(aerdev, dev, info->severity);
 591                 if (status == PCI_ERS_RESULT_RECOVERED) {
 592                         dev_printk(KERN_DEBUG, &dev->dev, "AER driver "
 593                                    "successfully recovered\n");
 594                 } else {
 595                         /* TODO: Should kernel panic here? */
 596                         dev_printk(KERN_DEBUG, &dev->dev, "AER driver didn't "
 597                                    "recover\n");
 598                 }
 599         }
 600 }
 601
 602 /**
 603  * aer_enable_rootport - enable Root Port's interrupts when receiving messages
 604  * @rpc: pointer to a Root Port data structure
 605  *
 606  * Invoked when PCIE bus loads AER service driver.
 607  */
 608 void aer_enable_rootport(struct aer_rpc *rpc)
 609 {
 610         struct pci_dev *pdev = rpc->rpd->port;
 611         int pos, aer_pos;
 612         u16 reg16;
 613         u32 reg32;
 614
 615         pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
 616         /* Clear PCIE Capability's Device Status */
 617         pci_read_config_word(pdev, pos+PCI_EXP_DEVSTA, &reg16);
 618         pci_write_config_word(pdev, pos+PCI_EXP_DEVSTA, reg16);
 619
 620         /* Disable system error generation in response to error messages */
 621         pci_read_config_word(pdev, pos + PCI_EXP_RTCTL, &reg16);
 622         reg16 &= ~(SYSTEM_ERROR_INTR_ON_MESG_MASK);
 623         pci_write_config_word(pdev, pos + PCI_EXP_RTCTL, reg16);
 624
 625         aer_pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ERR);
 626         /* Clear error status */
 627         pci_read_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, &reg32);
 628         pci_write_config_dword(pdev, aer_pos + PCI_ERR_ROOT_STATUS, reg32);
 629         pci_read_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, &reg32);
 630         pci_write_config_dword(pdev, aer_pos + PCI_ERR_COR_STATUS, reg32);
 631         pci_read_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, &reg32);
 632         pci_write_config_dword(pdev, aer_pos + PCI_ERR_UNCOR_STATUS, reg32);
 633
 634         /*
 635          * Enable error reporting for the root port device and downstream port
 636          * devices.
 637          */
 638         set_downstream_devices_error_reporting(pdev, true);
 639
 640         /* Enable Root Port's interrupt in response to error messages */
 641         pci_write_config_dword(pdev,
 642                 aer_pos + PCI_ERR_ROOT_COMMAND,
 643                 ROOT_PORT_INTR_ON_MESG_MASK);
 644 }
 645
 646 /**
 647  * disable_root_aer - disable Root Port's interrupts when receiving messages
 648  * @rpc: pointer to a Root Port data structure
 649  *
 650  * Invoked when PCIE bus unloads AER service driver.
 651  */
 652 static void disable_root_aer(struct aer_rpc *rpc)
 653 {
 654         struct pci_dev *pdev = rpc->rpd->port;
 655         u32 reg32;
 656         int pos;
 657
 658         /*
 659          * Disable error reporting for the root port device and downstream port
 660          * devices.
 661          */
 662         set_downstream_devices_error_reporting(pdev, false);
 663
 664         pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ERR);
 665         /* Disable Root's interrupt in response to error messages */
 666         pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_COMMAND, 0);
 667
 668         /* Clear Root's error status reg */
 669         pci_read_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, &reg32);
 670         pci_write_config_dword(pdev, pos + PCI_ERR_ROOT_STATUS, reg32);
 671 }
 672
 673 /**
 674  * get_e_source - retrieve an error source
 675  * @rpc: pointer to the root port which holds an error
 676  *
 677  * Invoked by DPC handler to consume an error.
 678  */
 679 static struct aer_err_source *get_e_source(struct aer_rpc *rpc)
 680 {
 681         struct aer_err_source *e_source;
 682         unsigned long flags;
 683
 684         /* Lock access to Root error producer/consumer index */
 685         spin_lock_irqsave(&rpc->e_lock, flags);
 686         if (rpc->prod_idx == rpc->cons_idx) {
 687                 spin_unlock_irqrestore(&rpc->e_lock, flags);
 688                 return NULL;
 689         }
 690         e_source = &rpc->e_sources[rpc->cons_idx];
 691         rpc->cons_idx++;
 692         if (rpc->cons_idx == AER_ERROR_SOURCES_MAX)
 693                 rpc->cons_idx = 0;
 694         spin_unlock_irqrestore(&rpc->e_lock, flags);
 695
 696         return e_source;
 697 }
 698
 699 /**
 700  * get_device_error_info - read error status from dev and store it to info
 701  * @dev: pointer to the device expected to have a error record
 702  * @info: pointer to structure to store the error record
 703  *
 704  * Return 1 on success, 0 on error.
 705  */
 706 static int get_device_error_info(struct pci_dev *dev, struct aer_err_info *info)
 707 {
 708         int pos, temp;
 709
 710         info->status = 0;
 711         info->tlp_header_valid = 0;
 712
 713         pos = pci_find_ext_capability(dev, PCI_EXT_CAP_ID_ERR);
 714
 715         /* The device might not support AER */
 716         if (!pos)
 717                 return 1;
 718
 719         if (info->severity == AER_CORRECTABLE) {
 720                 pci_read_config_dword(dev, pos + PCI_ERR_COR_STATUS,
 721                         &info->status);
 722                 pci_read_config_dword(dev, pos + PCI_ERR_COR_MASK,
 723                         &info->mask);
 724                 if (!(info->status & ~info->mask))
 725                         return 0;
 726         } else if (dev->hdr_type & PCI_HEADER_TYPE_BRIDGE ||
 727                 info->severity == AER_NONFATAL) {
 728
 729                 /* Link is still healthy for IO reads */
 730                 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_STATUS,
 731                         &info->status);
 732                 pci_read_config_dword(dev, pos + PCI_ERR_UNCOR_MASK,
 733                         &info->mask);
 734                 if (!(info->status & ~info->mask))
 735                         return 0;
 736
 737                 /* Get First Error Pointer */
 738                 pci_read_config_dword(dev, pos + PCI_ERR_CAP, &temp);
 739                 info->first_error = PCI_ERR_CAP_FEP(temp);
 740
 741                 if (info->status & AER_LOG_TLP_MASKS) {
 742                         info->tlp_header_valid = 1;
 743                         pci_read_config_dword(dev,
 744                                 pos + PCI_ERR_HEADER_LOG, &info->tlp.dw0);
 745                         pci_read_config_dword(dev,
 746                                 pos + PCI_ERR_HEADER_LOG + 4, &info->tlp.dw1);
 747                         pci_read_config_dword(dev,
 748                                 pos + PCI_ERR_HEADER_LOG + 8, &info->tlp.dw2);
 749                         pci_read_config_dword(dev,
 750                                 pos + PCI_ERR_HEADER_LOG + 12, &info->tlp.dw3);
 751                 }
 752         }
 753
 754         return 1;
 755 }
 756
 757 static inline void aer_process_err_devices(struct pcie_device *p_device,
 758                         struct aer_err_info *e_info)
 759 {
 760         int i;
 761
 762         if (!e_info->dev[0]) {
 763                 dev_printk(KERN_DEBUG, &p_device->port->dev,
 764                                 "can't find device of ID%04x\n",
 765                                 e_info->id);
 766         }
 767
 768         /* Report all before handle them, not to lost records by reset etc. */
 769         for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
 770                 if (get_device_error_info(e_info->dev[i], e_info))
 771                         aer_print_error(e_info->dev[i], e_info);
 772         }
 773         for (i = 0; i < e_info->error_dev_num && e_info->dev[i]; i++) {
 774                 if (get_device_error_info(e_info->dev[i], e_info))
 775                         handle_error_source(p_device, e_info->dev[i], e_info);
 776         }
 777 }
 778
 779 /**
 780  * aer_isr_one_error - consume an error detected by root port
 781  * @p_device: pointer to error root port service device
 782  * @e_src: pointer to an error source
 783  */
 784 static void aer_isr_one_error(struct pcie_device *p_device,
 785                 struct aer_err_source *e_src)
 786 {
 787         struct aer_err_info *e_info;
 788         int i;
 789
 790         /* struct aer_err_info might be big, so we allocate it with slab */
 791         e_info = kmalloc(sizeof(struct aer_err_info), GFP_KERNEL);
 792         if (e_info == NULL) {
 793                 dev_printk(KERN_DEBUG, &p_device->port->dev,
 794                         "Can't allocate mem when processing AER errors\n");
 795                 return;
 796         }
 797
 798         /*
 799          * There is a possibility that both correctable error and
 800          * uncorrectable error being logged. Report correctable error first.
 801          */
 802         for (i = 1; i & ROOT_ERR_STATUS_MASKS ; i <<= 2) {
 803                 if (i > 4)
 804                         break;
 805                 if (!(e_src->status & i))
 806                         continue;
 807
 808                 memset(e_info, 0, sizeof(struct aer_err_info));
 809
 810                 /* Init comprehensive error information */
 811                 if (i & PCI_ERR_ROOT_COR_RCV) {
 812                         e_info->id = ERR_COR_ID(e_src->id);
 813                         e_info->severity = AER_CORRECTABLE;
 814                 } else {
 815                         e_info->id = ERR_UNCOR_ID(e_src->id);
 816                         e_info->severity = ((e_src->status >> 6) & 1);
 817                 }
 818                 if (e_src->status &
 819                         (PCI_ERR_ROOT_MULTI_COR_RCV |
 820                          PCI_ERR_ROOT_MULTI_UNCOR_RCV))
 821                         e_info->multi_error_valid = 1;
 822
 823                 aer_print_port_info(p_device->port, e_info);
 824
 825                 find_source_device(p_device->port, e_info);
 826                 aer_process_err_devices(p_device, e_info);
 827         }
 828
 829         kfree(e_info);
 830 }
 831
 832 /**
 833  * aer_isr - consume errors detected by root port
 834  * @work: definition of this work item
 835  *
 836  * Invoked, as DPC, when root port records new detected error
 837  */
 838 void aer_isr(struct work_struct *work)
 839 {
 840         struct aer_rpc *rpc = container_of(work, struct aer_rpc, dpc_handler);
 841         struct pcie_device *p_device = rpc->rpd;
 842         struct aer_err_source *e_src;
 843
 844         mutex_lock(&rpc->rpc_mutex);
 845         e_src = get_e_source(rpc);
 846         while (e_src) {
 847                 aer_isr_one_error(p_device, e_src);
 848                 e_src = get_e_source(rpc);
 849         }
 850         mutex_unlock(&rpc->rpc_mutex);
 851
 852         wake_up(&rpc->wait_release);
 853 }
 854
 855 /**
 856  * aer_delete_rootport - disable root port aer and delete service data
 857  * @rpc: pointer to a root port device being deleted
 858  *
 859  * Invoked when AER service unloaded on a specific Root Port
 860  */
 861 void aer_delete_rootport(struct aer_rpc *rpc)
 862 {
 863         /* Disable root port AER itself */
 864         disable_root_aer(rpc);
 865
 866         kfree(rpc);
 867 }
 868
 869 /**
 870  * aer_init - provide AER initialization
 871  * @dev: pointer to AER pcie device
 872  *
 873  * Invoked when AER service driver is loaded.
 874  */
 875 int aer_init(struct pcie_device *dev)
 876 {
 877         if (aer_osc_setup(dev) && !forceload)
 878                 return -ENXIO;
 879
 880         return 0;
 881 }