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