1What: /sys/bus/pci/drivers/.../bind 2What: /sys/devices/pciX/.../bind 3Date: December 2003 4Contact: linux-pci@vger.kernel.org 5Description: 6 Writing a device location to this file will cause 7 the driver to attempt to bind to the device found at 8 this location. This is useful for overriding default 9 bindings. The format for the location is: DDDD:BB:DD.F. 10 That is Domain:Bus:Device.Function and is the same as 11 found in /sys/bus/pci/devices/. For example:: 12 13 # echo 0000:00:19.0 > /sys/bus/pci/drivers/foo/bind 14 15 (Note: kernels before 2.6.28 may require echo -n). 16 17What: /sys/bus/pci/drivers/.../unbind 18What: /sys/devices/pciX/.../unbind 19Date: December 2003 20Contact: linux-pci@vger.kernel.org 21Description: 22 Writing a device location to this file will cause the 23 driver to attempt to unbind from the device found at 24 this location. This may be useful when overriding default 25 bindings. The format for the location is: DDDD:BB:DD.F. 26 That is Domain:Bus:Device.Function and is the same as 27 found in /sys/bus/pci/devices/. For example:: 28 29 # echo 0000:00:19.0 > /sys/bus/pci/drivers/foo/unbind 30 31 (Note: kernels before 2.6.28 may require echo -n). 32 33What: /sys/bus/pci/drivers/.../new_id 34What: /sys/devices/pciX/.../new_id 35Date: December 2003 36Contact: linux-pci@vger.kernel.org 37Description: 38 Writing a device ID to this file will attempt to 39 dynamically add a new device ID to a PCI device driver. 40 This may allow the driver to support more hardware than 41 was included in the driver's static device ID support 42 table at compile time. The format for the device ID is: 43 VVVV DDDD SVVV SDDD CCCC MMMM PPPP. That is Vendor ID, 44 Device ID, Subsystem Vendor ID, Subsystem Device ID, 45 Class, Class Mask, and Private Driver Data. The Vendor ID 46 and Device ID fields are required, the rest are optional. 47 Upon successfully adding an ID, the driver will probe 48 for the device and attempt to bind to it. For example:: 49 50 # echo "8086 10f5" > /sys/bus/pci/drivers/foo/new_id 51 52What: /sys/bus/pci/drivers/.../remove_id 53What: /sys/devices/pciX/.../remove_id 54Date: February 2009 55Contact: Chris Wright <chrisw@sous-sol.org> 56Description: 57 Writing a device ID to this file will remove an ID 58 that was dynamically added via the new_id sysfs entry. 59 The format for the device ID is: 60 VVVV DDDD SVVV SDDD CCCC MMMM. That is Vendor ID, Device 61 ID, Subsystem Vendor ID, Subsystem Device ID, Class, 62 and Class Mask. The Vendor ID and Device ID fields are 63 required, the rest are optional. After successfully 64 removing an ID, the driver will no longer support the 65 device. This is useful to ensure auto probing won't 66 match the driver to the device. For example:: 67 68 # echo "8086 10f5" > /sys/bus/pci/drivers/foo/remove_id 69 70What: /sys/bus/pci/rescan 71Date: January 2009 72Contact: Linux PCI developers <linux-pci@vger.kernel.org> 73Description: 74 Writing a non-zero value to this attribute will 75 force a rescan of all PCI buses in the system, and 76 re-discover previously removed devices. 77 78What: /sys/bus/pci/devices/.../msi_bus 79Date: September 2014 80Contact: Linux PCI developers <linux-pci@vger.kernel.org> 81Description: 82 Writing a zero value to this attribute disallows MSI and 83 MSI-X for any future drivers of the device. If the device 84 is a bridge, MSI and MSI-X will be disallowed for future 85 drivers of all child devices under the bridge. Drivers 86 must be reloaded for the new setting to take effect. 87 88What: /sys/bus/pci/devices/.../msi_irqs/ 89Date: September, 2011 90Contact: Neil Horman <nhorman@tuxdriver.com> 91Description: 92 The /sys/devices/.../msi_irqs directory contains a variable set 93 of files, with each file being named after a corresponding msi 94 irq vector allocated to that device. 95 96What: /sys/bus/pci/devices/.../msi_irqs/<N> 97Date: September 2011 98Contact: Neil Horman <nhorman@tuxdriver.com> 99Description: 100 This attribute indicates the mode that the irq vector named by 101 the file is in (msi vs. msix) 102 103What: /sys/bus/pci/devices/.../irq 104Date: August 2021 105Contact: Linux PCI developers <linux-pci@vger.kernel.org> 106Description: 107 If a driver has enabled MSI (not MSI-X), "irq" contains the 108 IRQ of the first MSI vector. Otherwise "irq" contains the 109 IRQ of the legacy INTx interrupt. 110 111 "irq" being set to 0 indicates that the device isn't 112 capable of generating legacy INTx interrupts. 113 114What: /sys/bus/pci/devices/.../remove 115Date: January 2009 116Contact: Linux PCI developers <linux-pci@vger.kernel.org> 117Description: 118 Writing a non-zero value to this attribute will 119 hot-remove the PCI device and any of its children. 120 121What: /sys/bus/pci/devices/.../pci_bus/.../rescan 122Date: May 2011 123Contact: Linux PCI developers <linux-pci@vger.kernel.org> 124Description: 125 Writing a non-zero value to this attribute will 126 force a rescan of the bus and all child buses, 127 and re-discover devices removed earlier from this 128 part of the device tree. 129 130What: /sys/bus/pci/devices/.../rescan 131Date: January 2009 132Contact: Linux PCI developers <linux-pci@vger.kernel.org> 133Description: 134 Writing a non-zero value to this attribute will 135 force a rescan of the device's parent bus and all 136 child buses, and re-discover devices removed earlier 137 from this part of the device tree. 138 139What: /sys/bus/pci/devices/.../reset_method 140Date: August 2021 141Contact: Amey Narkhede <ameynarkhede03@gmail.com> 142Description: 143 Some devices allow an individual function to be reset 144 without affecting other functions in the same slot. 145 146 For devices that have this support, a file named 147 reset_method is present in sysfs. Reading this file 148 gives names of the supported and enabled reset methods and 149 their ordering. Writing a space-separated list of names of 150 reset methods sets the reset methods and ordering to be 151 used when resetting the device. Writing an empty string 152 disables the ability to reset the device. Writing 153 "default" enables all supported reset methods in the 154 default ordering. 155 156What: /sys/bus/pci/devices/.../reset 157Date: July 2009 158Contact: Michael S. Tsirkin <mst@redhat.com> 159Description: 160 Some devices allow an individual function to be reset 161 without affecting other functions in the same device. 162 For devices that have this support, a file named reset 163 will be present in sysfs. Writing 1 to this file 164 will perform reset. 165 166What: /sys/bus/pci/devices/.../reset_subordinate 167Date: October 2024 168Contact: linux-pci@vger.kernel.org 169Description: 170 This is visible only for bridge devices. If you want to reset 171 all devices attached through the subordinate bus of a specific 172 bridge device, writing 1 to this will try to do it. This will 173 affect all devices attached to the system through this bridge 174 similar to writing 1 to their individual "reset" file, so use 175 with caution. 176 177What: /sys/bus/pci/devices/.../vpd 178Date: February 2008 179Contact: Ben Hutchings <bwh@kernel.org> 180Description: 181 A file named vpd in a device directory will be a 182 binary file containing the Vital Product Data for the 183 device. It should follow the VPD format defined in 184 PCI Specification 2.1 or 2.2, but users should consider 185 that some devices may have incorrectly formatted data. 186 If the underlying VPD has a writable section then the 187 corresponding section of this file will be writable. 188 189What: /sys/bus/pci/devices/.../virtfn<N> 190Date: March 2009 191Contact: Yu Zhao <yu.zhao@intel.com> 192Description: 193 This symbolic link appears when hardware supports the SR-IOV 194 capability and the Physical Function driver has enabled it. 195 The symbolic link points to the PCI device sysfs entry of the 196 Virtual Function whose index is N (0...MaxVFs-1). 197 198What: /sys/bus/pci/devices/.../dep_link 199Date: March 2009 200Contact: Yu Zhao <yu.zhao@intel.com> 201Description: 202 This symbolic link appears when hardware supports the SR-IOV 203 capability and the Physical Function driver has enabled it, 204 and this device has vendor specific dependencies with others. 205 The symbolic link points to the PCI device sysfs entry of 206 Physical Function this device depends on. 207 208What: /sys/bus/pci/devices/.../physfn 209Date: March 2009 210Contact: Yu Zhao <yu.zhao@intel.com> 211Description: 212 This symbolic link appears when a device is a Virtual Function. 213 The symbolic link points to the PCI device sysfs entry of the 214 Physical Function this device associates with. 215 216What: /sys/bus/pci/devices/.../modalias 217Date: May 2005 218Contact: Greg Kroah-Hartman <gregkh@linuxfoundation.org> 219Description: 220 This attribute indicates the PCI ID of the device object. 221 222 That is in the format: 223 pci:vXXXXXXXXdXXXXXXXXsvXXXXXXXXsdXXXXXXXXbcXXscXXiXX, 224 where: 225 226 - vXXXXXXXX contains the vendor ID; 227 - dXXXXXXXX contains the device ID; 228 - svXXXXXXXX contains the sub-vendor ID; 229 - sdXXXXXXXX contains the subsystem device ID; 230 - bcXX contains the device class; 231 - scXX contains the device subclass; 232 - iXX contains the device class programming interface. 233 234What: /sys/bus/pci/slots/.../module 235Date: June 2009 236Contact: linux-pci@vger.kernel.org 237Description: 238 This symbolic link points to the PCI hotplug controller driver 239 module that manages the hotplug slot. 240 241What: /sys/bus/pci/devices/.../label 242Date: July 2010 243Contact: Narendra K <narendra_k@dell.com>, linux-bugs@dell.com 244Description: 245 Reading this attribute will provide the firmware 246 given name (SMBIOS type 41 string or ACPI _DSM string) of 247 the PCI device. The attribute will be created only if the 248 firmware naming mechanism is implemented for the device: 249 a SMBIOS type 41 record with a non-empty reference 250 designation, or a Device Name _DSM that returns the name 251 as a string or a buffer. The value read is empty when the 252 firmware implements the _DSM but gives the device no name, 253 which per PCI Firmware r3.3, sec 4.6.7 the firmware reports 254 as a NULL string. ACPI _DSM string name will be given 255 priority if the system firmware provides SMBIOS type 41 256 string also. 257Users: 258 Userspace applications interested in knowing the 259 firmware assigned name of the PCI device. 260 261What: /sys/bus/pci/devices/.../index 262Date: July 2010 263Contact: Narendra K <narendra_k@dell.com>, linux-bugs@dell.com 264Description: 265 Reading this attribute will provide the firmware given instance 266 number of the PCI device. Depending on the platform this can 267 be for example the SMBIOS type 41 device type instance or the 268 user-defined ID (UID) on s390. The attribute will be created 269 only if the firmware has given an instance number to the PCI 270 device and that number is guaranteed to uniquely identify the 271 device in the system. 272Users: 273 Userspace applications interested in knowing the 274 firmware assigned device type instance of the PCI 275 device that can help in understanding the firmware 276 intended order of the PCI device. 277 278What: /sys/bus/pci/devices/.../acpi_index 279Date: July 2010 280Contact: Narendra K <narendra_k@dell.com>, linux-bugs@dell.com 281Description: 282 Reading this attribute will provide the firmware 283 given instance (ACPI _DSM instance number) of the PCI device. 284 The attribute will be created only if the firmware has given 285 an instance number to the PCI device. ACPI _DSM instance number 286 will be given priority if the system firmware provides SMBIOS 287 type 41 device type instance also. 288Users: 289 Userspace applications interested in knowing the 290 firmware assigned instance number of the PCI 291 device that can help in understanding the firmware 292 intended order of the PCI device. 293 294What: /sys/bus/pci/devices/.../d3cold_allowed 295Date: July 2012 296Contact: Huang Ying <ying.huang@intel.com> 297Description: 298 d3cold_allowed is bit to control whether the corresponding PCI 299 device can be put into D3Cold state. If it is cleared, the 300 device will never be put into D3Cold state. If it is set, the 301 device may be put into D3Cold state if other requirements are 302 satisfied too. Reading this attribute will show the current 303 value of d3cold_allowed bit. Writing this attribute will set 304 the value of d3cold_allowed bit. 305 306What: /sys/bus/pci/devices/.../sriov_totalvfs 307Date: November 2012 308Contact: Donald Dutile <ddutile@redhat.com> 309Description: 310 This file appears when a physical PCIe device supports SR-IOV. 311 Userspace applications can read this file to determine the 312 maximum number of Virtual Functions (VFs) a PCIe physical 313 function (PF) can support. Typically, this is the value reported 314 in the PF's SR-IOV extended capability structure's TotalVFs 315 element. Drivers have the ability at probe time to reduce the 316 value read from this file via the pci_sriov_set_totalvfs() 317 function. 318 319What: /sys/bus/pci/devices/.../sriov_numvfs 320Date: November 2012 321Contact: Donald Dutile <ddutile@redhat.com> 322Description: 323 This file appears when a physical PCIe device supports SR-IOV. 324 Userspace applications can read and write to this file to 325 determine and control the enablement or disablement of Virtual 326 Functions (VFs) on the physical function (PF). A read of this 327 file will return the number of VFs that are enabled on this PF. 328 A number written to this file will enable the specified 329 number of VFs. A userspace application would typically read the 330 file and check that the value is zero, and then write the number 331 of VFs that should be enabled on the PF; the value written 332 should be less than or equal to the value in the sriov_totalvfs 333 file. A userspace application wanting to disable the VFs would 334 write a zero to this file. The core ensures that valid values 335 are written to this file, and returns errors when values are not 336 valid. For example, writing a 2 to this file when sriov_numvfs 337 is not 0 and not 2 already will return an error. Writing a 10 338 when the value of sriov_totalvfs is 8 will return an error. 339 340What: /sys/bus/pci/devices/.../driver_override 341Date: April 2014 342Contact: Alex Williamson <alex.williamson@redhat.com> 343Description: 344 This file allows the driver for a device to be specified which 345 will override standard static and dynamic ID matching. When 346 specified, only a driver with a name matching the value written 347 to driver_override will have an opportunity to bind to the 348 device. The override is specified by writing a string to the 349 driver_override file (echo pci-stub > driver_override) and 350 may be cleared with an empty string (echo > driver_override). 351 This returns the device to standard matching rules binding. 352 Writing to driver_override does not automatically unbind the 353 device from its current driver or make any attempt to 354 automatically load the specified driver. If no driver with a 355 matching name is currently loaded in the kernel, the device 356 will not bind to any driver. This also allows devices to 357 opt-out of driver binding using a driver_override name such as 358 "none". Only a single driver may be specified in the override, 359 there is no support for parsing delimiters. 360 361What: /sys/bus/pci/devices/.../numa_node 362Date: Oct 2014 363Contact: Prarit Bhargava <prarit@redhat.com> 364Description: 365 This file contains the NUMA node to which the PCI device is 366 attached, or -1 if the node is unknown. The initial value 367 comes from an ACPI _PXM method or a similar firmware 368 source. If that is missing or incorrect, this file can be 369 written to override the node. In that case, please report 370 a firmware bug to the system vendor. Writing to this file 371 taints the kernel with TAINT_FIRMWARE_WORKAROUND, which 372 reduces the supportability of your system. 373 374What: /sys/bus/pci/devices/.../revision 375Date: November 2016 376Contact: Emil Velikov <emil.l.velikov@gmail.com> 377Description: 378 This file contains the revision field of the PCI device. 379 The value comes from device config space. The file is read only. 380 381What: /sys/bus/pci/devices/.../sriov_drivers_autoprobe 382Date: April 2017 383Contact: Bodong Wang<bodong@mellanox.com> 384Description: 385 This file is associated with the PF of a device that 386 supports SR-IOV. It determines whether newly-enabled VFs 387 are immediately bound to a driver. It initially contains 388 1, which means the kernel automatically binds VFs to a 389 compatible driver immediately after they are enabled. If 390 an application writes 0 to the file before enabling VFs, 391 the kernel will not bind VFs to a driver. 392 393 A typical use case is to write 0 to this file, then enable 394 VFs, then assign the newly-created VFs to virtual machines. 395 Note that changing this file does not affect already- 396 enabled VFs. In this scenario, the user must first disable 397 the VFs, write 0 to sriov_drivers_autoprobe, then re-enable 398 the VFs. 399 400 This is similar to /sys/bus/pci/drivers_autoprobe, but 401 affects only the VFs associated with a specific PF. 402 403What: /sys/bus/pci/devices/.../p2pmem/size 404Date: November 2017 405Contact: Logan Gunthorpe <logang@deltatee.com> 406Description: 407 If the device has any Peer-to-Peer memory registered, this 408 file contains the total amount of memory that the device 409 provides (in decimal). 410 411What: /sys/bus/pci/devices/.../p2pmem/available 412Date: November 2017 413Contact: Logan Gunthorpe <logang@deltatee.com> 414Description: 415 If the device has any Peer-to-Peer memory registered, this 416 file contains the amount of memory that has not been 417 allocated (in decimal). 418 419What: /sys/bus/pci/devices/.../p2pmem/published 420Date: November 2017 421Contact: Logan Gunthorpe <logang@deltatee.com> 422Description: 423 If the device has any Peer-to-Peer memory registered, this 424 file contains a '1' if the memory has been published for 425 use outside the driver that owns the device. 426 427What: /sys/bus/pci/devices/.../p2pmem/allocate 428Date: August 2022 429Contact: Logan Gunthorpe <logang@deltatee.com> 430Description: 431 This file allows mapping p2pmem into userspace. For each 432 mmap() call on this file, the kernel will allocate a chunk 433 of Peer-to-Peer memory for use in Peer-to-Peer transactions. 434 This memory can be used in O_DIRECT calls to NVMe backed 435 files for Peer-to-Peer copies. 436 437What: /sys/bus/pci/devices/.../link/clkpm 438 /sys/bus/pci/devices/.../link/l0s_aspm 439 /sys/bus/pci/devices/.../link/l1_aspm 440 /sys/bus/pci/devices/.../link/l1_1_aspm 441 /sys/bus/pci/devices/.../link/l1_2_aspm 442 /sys/bus/pci/devices/.../link/l1_1_pcipm 443 /sys/bus/pci/devices/.../link/l1_2_pcipm 444Date: October 2019 445Contact: Heiner Kallweit <hkallweit1@gmail.com> 446Description: If ASPM is supported for an endpoint, these files can be 447 used to disable or enable the individual power management 448 states. Write y/1/on to enable, n/0/off to disable. 449 450What: /sys/bus/pci/devices/.../power_state 451Date: November 2020 452Contact: Linux PCI developers <linux-pci@vger.kernel.org> 453Description: 454 This file contains the current PCI power state of the device. 455 The value comes from the PCI kernel device state and can be one 456 of: "unknown", "error", "D0", D1", "D2", "D3hot", "D3cold". 457 The file is read only. 458 459What: /sys/bus/pci/devices/.../sriov_vf_total_msix 460Date: January 2021 461Contact: Leon Romanovsky <leonro@nvidia.com> 462Description: 463 This file is associated with a SR-IOV physical function (PF). 464 It contains the total number of MSI-X vectors available for 465 assignment to all virtual functions (VFs) associated with PF. 466 The value will be zero if the device doesn't support this 467 functionality. For supported devices, the value will be 468 constant and won't be changed after MSI-X vectors assignment. 469 470What: /sys/bus/pci/devices/.../sriov_vf_msix_count 471Date: January 2021 472Contact: Leon Romanovsky <leonro@nvidia.com> 473Description: 474 This file is associated with a SR-IOV virtual function (VF). 475 It allows configuration of the number of MSI-X vectors for 476 the VF. This allows devices that have a global pool of MSI-X 477 vectors to optimally divide them between VFs based on VF usage. 478 479 The values accepted are: 480 * > 0 - this number will be reported as the Table Size in the 481 VF's MSI-X capability 482 * < 0 - not valid 483 * = 0 - will reset to the device default value 484 485 The file is writable if the PF is bound to a driver that 486 implements ->sriov_set_msix_vec_count(). 487 488What: /sys/bus/pci/devices/.../resourceN_resize 489Date: September 2022 490Contact: Alex Williamson <alex.williamson@redhat.com> 491Description: 492 These files provide an interface to PCIe Resizable BAR support. 493 A file is created for each BAR resource (N) supported by the 494 PCIe Resizable BAR extended capability of the device. Reading 495 each file exposes the bitmap of available resource sizes:: 496 497 # cat resource1_resize 498 00000000000001c0 499 500 The bitmap represents supported resource sizes for the BAR, 501 where bit0 = 1MB, bit1 = 2MB, bit2 = 4MB, etc. In the above 502 example the device supports 64MB, 128MB, and 256MB BAR sizes. 503 504 When writing the file, the user provides the bit position of 505 the desired resource size, for example:: 506 507 # echo 7 > resource1_resize 508 509 This indicates to set the size value corresponding to bit 7, 510 128MB. The resulting size is 2 ^ (bit# + 20). This definition 511 matches the PCIe specification of this capability. 512 513 In order to make use of resource resizing, all PCI drivers must 514 be unbound from the device and peer devices under the same 515 parent bridge may need to be soft removed. In the case of 516 VGA devices, writing a resize value will remove low level 517 console drivers from the device. Raw users of pci-sysfs 518 resourceN attributes must be terminated prior to resizing. 519 Success of the resizing operation is not guaranteed. 520 521What: /sys/bus/pci/devices/.../leds/*:enclosure:*/brightness 522What: /sys/class/leds/*:enclosure:*/brightness 523Date: August 2024 524KernelVersion: 6.12 525Description: 526 LED indications on PCIe storage enclosures which are controlled 527 through the NPEM interface (Native PCIe Enclosure Management, 528 PCIe r6.1 sec 6.28) are accessible as led class devices, both 529 below /sys/class/leds and below NPEM-capable PCI devices. 530 531 Although these led class devices could be manipulated manually, 532 in practice they are typically manipulated automatically by an 533 application such as ledmon(8). 534 535 The name of a led class device is as follows: 536 <bdf>:enclosure:<indication> 537 where: 538 539 - <bdf> is the domain, bus, device and function number 540 (e.g. 10000:02:05.0) 541 - <indication> is a short description of the LED indication 542 543 Valid indications per PCIe r6.1 table 6-27 are: 544 545 - ok (drive is functioning normally) 546 - locate (drive is being identified by an admin) 547 - fail (drive is not functioning properly) 548 - rebuild (drive is part of an array that is rebuilding) 549 - pfa (drive is predicted to fail soon) 550 - hotspare (drive is marked to be used as a replacement) 551 - ica (drive is part of an array that is degraded) 552 - ifa (drive is part of an array that is failed) 553 - idt (drive is not the right type for the connector) 554 - disabled (drive is disabled, removal is safe) 555 - specific0 to specific7 (enclosure-specific indications) 556 557 Broadly, the indications fall into one of these categories: 558 559 - to signify drive state (ok, locate, fail, idt, disabled) 560 - to signify drive role or state in a software RAID array 561 (rebuild, pfa, hotspare, ica, ifa) 562 - to signify any other role or state (specific0 to specific7) 563 564 Mandatory indications per PCIe r6.1 sec 7.9.19.2 comprise: 565 ok, locate, fail, rebuild. All others are optional. 566 A led class device is only visible if the corresponding 567 indication is supported by the device. 568 569 To manipulate the indications, write 0 (LED_OFF) or 1 (LED_ON) 570 to the "brightness" file. Note that manipulating an indication 571 may implicitly manipulate other indications at the vendor's 572 discretion. E.g. when the user lights up the "ok" indication, 573 the vendor may choose to automatically turn off the "fail" 574 indication. The current state of an indication can be 575 retrieved by reading its "brightness" file. 576 577 The PCIe Base Specification allows vendors leeway to choose 578 different colors or blinking patterns for the indications, 579 but they typically follow the IBPI standard. E.g. the "locate" 580 indication is usually presented as one or two LEDs blinking at 581 4 Hz frequency: 582 https://en.wikipedia.org/wiki/International_Blinking_Pattern_Interpretation 583 584 PCI Firmware Specification r3.3 sec 4.7 defines a DSM interface 585 to facilitate shared access by operating system and platform 586 firmware to a device's NPEM registers. The kernel will use 587 this DSM interface where available, instead of accessing NPEM 588 registers directly. The DSM interface does not support the 589 enclosure-specific indications "specific0" to "specific7", 590 hence the corresponding led class devices are unavailable if 591 the DSM interface is used. 592 593What: /sys/bus/pci/devices/.../doe_features 594Date: March 2025 595Contact: Linux PCI developers <linux-pci@vger.kernel.org> 596Description: 597 This directory contains a list of the supported Data Object 598 Exchange (DOE) features. The features are the file name. 599 The contents of each file is the raw Vendor ID and data 600 object feature values. 601 602 The value comes from the device and specifies the vendor and 603 data object type supported. The lower (RHS of the colon) is 604 the data object type in hex. The upper (LHS of the colon) 605 is the vendor ID. 606 607 As all DOE devices must support the DOE discovery feature, 608 if DOE is supported you will at least see the doe_discovery 609 file, with this contents:: 610 611 # cat doe_features/doe_discovery 612 0001:00 613 614 If the device supports other features you will see other 615 files as well. For example if CMA/SPDM and secure CMA/SPDM 616 are supported the doe_features directory will look like 617 this:: 618 619 # ls doe_features 620 0001:01 0001:02 doe_discovery 621 622What: /sys/bus/pci/devices/.../serial_number 623Date: December 2025 624Contact: Matthew Wood <thepacketgeek@gmail.com> 625Description: 626 This is visible only for PCI devices that support the serial 627 number extended capability. The file is read only and due to 628 the possible sensitivity of accessible serial numbers, admin 629 only. 630 631What: /sys/bus/pci/devices/.../tsm/ 632Contact: linux-coco@lists.linux.dev 633Description: 634 This directory only appears if a physical device function 635 supports authentication (PCIe CMA-SPDM), interface security 636 (PCIe TDISP), and is accepted for secure operation by the 637 platform TSM driver. This attribute directory appears 638 dynamically after the platform TSM driver loads. So, only after 639 the /sys/class/tsm/tsm0 device arrives can tools assume that 640 devices without a tsm/ attribute directory will never have one; 641 before that, the security capabilities of the device relative to 642 the platform TSM are unknown. See 643 Documentation/ABI/testing/sysfs-class-tsm. 644 645What: /sys/bus/pci/devices/.../tsm/connect 646Contact: linux-coco@lists.linux.dev 647Description: 648 (RW) Write the name of a TSM (TEE Security Manager) device from 649 /sys/class/tsm to this file to establish a connection with the 650 device. This typically includes an SPDM (DMTF Security 651 Protocols and Data Models) session over PCIe DOE (Data Object 652 Exchange) and may also include PCIe IDE (Integrity and Data 653 Encryption) establishment. Reads from this attribute return the 654 name of the connected TSM or the empty string if not 655 connected. A TSM device signals its readiness to accept PCI 656 connection via a KOBJ_CHANGE event. 657 658What: /sys/bus/pci/devices/.../tsm/disconnect 659Contact: linux-coco@lists.linux.dev 660Description: 661 (WO) Write the name of the TSM device that was specified 662 to 'connect' to teardown the connection. 663 664What: /sys/bus/pci/devices/.../tsm/dsm 665Contact: linux-coco@lists.linux.dev 666Description: (RO) Return PCI device name of this device's DSM (Device 667 Security Manager). When a device is in the connected state it 668 indicates that the platform TSM (TEE Security Manager) has made 669 a secure-session connection with a device's DSM. A DSM is always 670 physical function 0 and when the device supports TDISP (TEE 671 Device Interface Security Protocol) its managed functions also 672 populate this tsm/dsm attribute. The managed functions of a DSM 673 are SR-IOV (Single Root I/O Virtualization) virtual functions, 674 non-zero functions of a multi-function device, or downstream 675 endpoints depending on whether the DSM is an SR-IOV physical 676 function, function0 of a multi-function device, or an upstream 677 PCIe switch port. This is a "link" TSM attribute, see 678 Documentation/ABI/testing/sysfs-class-tsm. 679 680What: /sys/bus/pci/devices/.../tsm/bound 681Contact: linux-coco@lists.linux.dev 682Description: (RO) Return the device name of the TSM when the device is in a 683 TDISP (TEE Device Interface Security Protocol) operational state 684 (LOCKED, RUN, or ERROR, not UNLOCKED). Bound devices consume 685 platform TSM resources and depend on the device's configuration 686 (e.g. BME (Bus Master Enable) and MSE (Memory Space Enable) 687 among other settings) to remain stable for the duration of the 688 bound state. This attribute is only visible for devices that 689 support TDISP operation, and it is only populated after 690 successful connect and TSM bind. The TSM bind operation is 691 initiated by VFIO/IOMMUFD. This is a "link" TSM attribute, see 692 Documentation/ABI/testing/sysfs-class-tsm. 693 694What: /sys/bus/pci/devices/.../authenticated 695Contact: linux-pci@vger.kernel.org 696Description: 697 When the device's tsm/ directory is present device 698 authentication (PCIe CMA-SPDM) and link encryption (PCIe IDE) 699 are handled by the platform TSM (TEE Security Manager). When the 700 tsm/ directory is not present this attribute reflects only the 701 native CMA-SPDM authentication state with the kernel's 702 certificate store. 703 704 If the attribute is not present, it indicates that 705 authentication is unsupported by the device, or the TSM has no 706 available authentication methods for the device. 707 708 When present and the tsm/ attribute directory is present, the 709 authenticated attribute is an alias for the device 'connect' 710 state. See the 'tsm/connect' attribute for more details. 711