linux/Documentation/PCI/pcieaer-howto.rst
<<
>>
Prefs
   1.. SPDX-License-Identifier: GPL-2.0
   2.. include:: <isonum.txt>
   3
   4===========================================================
   5The PCI Express Advanced Error Reporting Driver Guide HOWTO
   6===========================================================
   7
   8:Authors: - T. Long Nguyen <tom.l.nguyen@intel.com>
   9          - Yanmin Zhang <yanmin.zhang@intel.com>
  10
  11:Copyright: |copy| 2006 Intel Corporation
  12
  13Overview
  14===========
  15
  16About this guide
  17----------------
  18
  19This guide describes the basics of the PCI Express Advanced Error
  20Reporting (AER) driver and provides information on how to use it, as
  21well as how to enable the drivers of endpoint devices to conform with
  22PCI Express AER driver.
  23
  24
  25What is the PCI Express AER Driver?
  26-----------------------------------
  27
  28PCI Express error signaling can occur on the PCI Express link itself
  29or on behalf of transactions initiated on the link. PCI Express
  30defines two error reporting paradigms: the baseline capability and
  31the Advanced Error Reporting capability. The baseline capability is
  32required of all PCI Express components providing a minimum defined
  33set of error reporting requirements. Advanced Error Reporting
  34capability is implemented with a PCI Express advanced error reporting
  35extended capability structure providing more robust error reporting.
  36
  37The PCI Express AER driver provides the infrastructure to support PCI
  38Express Advanced Error Reporting capability. The PCI Express AER
  39driver provides three basic functions:
  40
  41  - Gathers the comprehensive error information if errors occurred.
  42  - Reports error to the users.
  43  - Performs error recovery actions.
  44
  45AER driver only attaches root ports which support PCI-Express AER
  46capability.
  47
  48
  49User Guide
  50==========
  51
  52Include the PCI Express AER Root Driver into the Linux Kernel
  53-------------------------------------------------------------
  54
  55The PCI Express AER Root driver is a Root Port service driver attached
  56to the PCI Express Port Bus driver. If a user wants to use it, the driver
  57has to be compiled. Option CONFIG_PCIEAER supports this capability. It
  58depends on CONFIG_PCIEPORTBUS, so pls. set CONFIG_PCIEPORTBUS=y and
  59CONFIG_PCIEAER = y.
  60
  61Load PCI Express AER Root Driver
  62--------------------------------
  63
  64Some systems have AER support in firmware. Enabling Linux AER support at
  65the same time the firmware handles AER may result in unpredictable
  66behavior. Therefore, Linux does not handle AER events unless the firmware
  67grants AER control to the OS via the ACPI _OSC method. See the PCI FW 3.0
  68Specification for details regarding _OSC usage.
  69
  70AER error output
  71----------------
  72
  73When a PCIe AER error is captured, an error message will be output to
  74console. If it's a correctable error, it is output as a warning.
  75Otherwise, it is printed as an error. So users could choose different
  76log level to filter out correctable error messages.
  77
  78Below shows an example::
  79
  80  0000:50:00.0: PCIe Bus Error: severity=Uncorrected (Fatal), type=Transaction Layer, id=0500(Requester ID)
  81  0000:50:00.0:   device [8086:0329] error status/mask=00100000/00000000
  82  0000:50:00.0:    [20] Unsupported Request    (First)
  83  0000:50:00.0:   TLP Header: 04000001 00200a03 05010000 00050100
  84
  85In the example, 'Requester ID' means the ID of the device who sends
  86the error message to root port. Pls. refer to pci express specs for
  87other fields.
  88
  89AER Statistics / Counters
  90-------------------------
  91
  92When PCIe AER errors are captured, the counters / statistics are also exposed
  93in the form of sysfs attributes which are documented at
  94Documentation/ABI/testing/sysfs-bus-pci-devices-aer_stats
  95
  96Developer Guide
  97===============
  98
  99To enable AER aware support requires a software driver to configure
 100the AER capability structure within its device and to provide callbacks.
 101
 102To support AER better, developers need understand how AER does work
 103firstly.
 104
 105PCI Express errors are classified into two types: correctable errors
 106and uncorrectable errors. This classification is based on the impacts
 107of those errors, which may result in degraded performance or function
 108failure.
 109
 110Correctable errors pose no impacts on the functionality of the
 111interface. The PCI Express protocol can recover without any software
 112intervention or any loss of data. These errors are detected and
 113corrected by hardware. Unlike correctable errors, uncorrectable
 114errors impact functionality of the interface. Uncorrectable errors
 115can cause a particular transaction or a particular PCI Express link
 116to be unreliable. Depending on those error conditions, uncorrectable
 117errors are further classified into non-fatal errors and fatal errors.
 118Non-fatal errors cause the particular transaction to be unreliable,
 119but the PCI Express link itself is fully functional. Fatal errors, on
 120the other hand, cause the link to be unreliable.
 121
 122When AER is enabled, a PCI Express device will automatically send an
 123error message to the PCIe root port above it when the device captures
 124an error. The Root Port, upon receiving an error reporting message,
 125internally processes and logs the error message in its PCI Express
 126capability structure. Error information being logged includes storing
 127the error reporting agent's requestor ID into the Error Source
 128Identification Registers and setting the error bits of the Root Error
 129Status Register accordingly. If AER error reporting is enabled in Root
 130Error Command Register, the Root Port generates an interrupt if an
 131error is detected.
 132
 133Note that the errors as described above are related to the PCI Express
 134hierarchy and links. These errors do not include any device specific
 135errors because device specific errors will still get sent directly to
 136the device driver.
 137
 138Configure the AER capability structure
 139--------------------------------------
 140
 141AER aware drivers of PCI Express component need change the device
 142control registers to enable AER. They also could change AER registers,
 143including mask and severity registers. Helper function
 144pci_enable_pcie_error_reporting could be used to enable AER. See
 145section 3.3.
 146
 147Provide callbacks
 148-----------------
 149
 150callback reset_link to reset pci express link
 151~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 152
 153This callback is used to reset the pci express physical link when a
 154fatal error happens. The root port aer service driver provides a
 155default reset_link function, but different upstream ports might
 156have different specifications to reset pci express link, so all
 157upstream ports should provide their own reset_link functions.
 158
 159Section 3.2.2.2 provides more detailed info on when to call
 160reset_link.
 161
 162PCI error-recovery callbacks
 163~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 164
 165The PCI Express AER Root driver uses error callbacks to coordinate
 166with downstream device drivers associated with a hierarchy in question
 167when performing error recovery actions.
 168
 169Data struct pci_driver has a pointer, err_handler, to point to
 170pci_error_handlers who consists of a couple of callback function
 171pointers. AER driver follows the rules defined in
 172pci-error-recovery.txt except pci express specific parts (e.g.
 173reset_link). Pls. refer to pci-error-recovery.txt for detailed
 174definitions of the callbacks.
 175
 176Below sections specify when to call the error callback functions.
 177
 178Correctable errors
 179~~~~~~~~~~~~~~~~~~
 180
 181Correctable errors pose no impacts on the functionality of
 182the interface. The PCI Express protocol can recover without any
 183software intervention or any loss of data. These errors do not
 184require any recovery actions. The AER driver clears the device's
 185correctable error status register accordingly and logs these errors.
 186
 187Non-correctable (non-fatal and fatal) errors
 188~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 189
 190If an error message indicates a non-fatal error, performing link reset
 191at upstream is not required. The AER driver calls error_detected(dev,
 192pci_channel_io_normal) to all drivers associated within a hierarchy in
 193question. for example::
 194
 195  EndPoint<==>DownstreamPort B<==>UpstreamPort A<==>RootPort
 196
 197If Upstream port A captures an AER error, the hierarchy consists of
 198Downstream port B and EndPoint.
 199
 200A driver may return PCI_ERS_RESULT_CAN_RECOVER,
 201PCI_ERS_RESULT_DISCONNECT, or PCI_ERS_RESULT_NEED_RESET, depending on
 202whether it can recover or the AER driver calls mmio_enabled as next.
 203
 204If an error message indicates a fatal error, kernel will broadcast
 205error_detected(dev, pci_channel_io_frozen) to all drivers within
 206a hierarchy in question. Then, performing link reset at upstream is
 207necessary. As different kinds of devices might use different approaches
 208to reset link, AER port service driver is required to provide the
 209function to reset link via callback parameter of pcie_do_recovery()
 210function. If reset_link is not NULL, recovery function will use it
 211to reset the link. If error_detected returns PCI_ERS_RESULT_CAN_RECOVER
 212and reset_link returns PCI_ERS_RESULT_RECOVERED, the error handling goes
 213to mmio_enabled.
 214
 215helper functions
 216----------------
 217::
 218
 219  int pci_enable_pcie_error_reporting(struct pci_dev *dev);
 220
 221pci_enable_pcie_error_reporting enables the device to send error
 222messages to root port when an error is detected. Note that devices
 223don't enable the error reporting by default, so device drivers need
 224call this function to enable it.
 225
 226::
 227
 228  int pci_disable_pcie_error_reporting(struct pci_dev *dev);
 229
 230pci_disable_pcie_error_reporting disables the device to send error
 231messages to root port when an error is detected.
 232
 233::
 234
 235  int pci_aer_clear_nonfatal_status(struct pci_dev *dev);`
 236
 237pci_aer_clear_nonfatal_status clears non-fatal errors in the uncorrectable
 238error status register.
 239
 240Frequent Asked Questions
 241------------------------
 242
 243Q:
 244  What happens if a PCI Express device driver does not provide an
 245  error recovery handler (pci_driver->err_handler is equal to NULL)?
 246
 247A:
 248  The devices attached with the driver won't be recovered. If the
 249  error is fatal, kernel will print out warning messages. Please refer
 250  to section 3 for more information.
 251
 252Q:
 253  What happens if an upstream port service driver does not provide
 254  callback reset_link?
 255
 256A:
 257  Fatal error recovery will fail if the errors are reported by the
 258  upstream ports who are attached by the service driver.
 259
 260Q:
 261  How does this infrastructure deal with driver that is not PCI
 262  Express aware?
 263
 264A:
 265  This infrastructure calls the error callback functions of the
 266  driver when an error happens. But if the driver is not aware of
 267  PCI Express, the device might not report its own errors to root
 268  port.
 269
 270Q:
 271  What modifications will that driver need to make it compatible
 272  with the PCI Express AER Root driver?
 273
 274A:
 275  It could call the helper functions to enable AER in devices and
 276  cleanup uncorrectable status register. Pls. refer to section 3.3.
 277
 278
 279Software error injection
 280========================
 281
 282Debugging PCIe AER error recovery code is quite difficult because it
 283is hard to trigger real hardware errors. Software based error
 284injection can be used to fake various kinds of PCIe errors.
 285
 286First you should enable PCIe AER software error injection in kernel
 287configuration, that is, following item should be in your .config.
 288
 289CONFIG_PCIEAER_INJECT=y or CONFIG_PCIEAER_INJECT=m
 290
 291After reboot with new kernel or insert the module, a device file named
 292/dev/aer_inject should be created.
 293
 294Then, you need a user space tool named aer-inject, which can be gotten
 295from:
 296
 297    https://git.kernel.org/cgit/linux/kernel/git/gong.chen/aer-inject.git/
 298
 299More information about aer-inject can be found in the document comes
 300with its source code.
 301