1=============================== 2Adjunct Processor (AP) facility 3=============================== 4 5 6Introduction 7============ 8The Adjunct Processor (AP) facility is an IBM Z cryptographic facility comprised 9of three AP instructions and from 1 up to 256 PCIe cryptographic adapter cards. 10The AP devices provide cryptographic functions to all CPUs assigned to a 11linux system running in an IBM Z system LPAR. 12 13The AP adapter cards are exposed via the AP bus. The motivation for vfio-ap 14is to make AP cards available to KVM guests using the VFIO mediated device 15framework. This implementation relies considerably on the s390 virtualization 16facilities which do most of the hard work of providing direct access to AP 17devices. 18 19AP Architectural Overview 20========================= 21To facilitate the comprehension of the design, let's start with some 22definitions: 23 24* AP adapter 25 26 An AP adapter is an IBM Z adapter card that can perform cryptographic 27 functions. There can be from 0 to 256 adapters assigned to an LPAR. Adapters 28 assigned to the LPAR in which a linux host is running will be available to 29 the linux host. Each adapter is identified by a number from 0 to 255; however, 30 the maximum adapter number is determined by machine model and/or adapter type. 31 When installed, an AP adapter is accessed by AP instructions executed by any 32 CPU. 33 34 The AP adapter cards are assigned to a given LPAR via the system's Activation 35 Profile which can be edited via the HMC. When the linux host system is IPL'd 36 in the LPAR, the AP bus detects the AP adapter cards assigned to the LPAR and 37 creates a sysfs device for each assigned adapter. For example, if AP adapters 38 4 and 10 (0x0a) are assigned to the LPAR, the AP bus will create the following 39 sysfs device entries:: 40 41 /sys/devices/ap/card04 42 /sys/devices/ap/card0a 43 44 Symbolic links to these devices will also be created in the AP bus devices 45 sub-directory:: 46 47 /sys/bus/ap/devices/[card04] 48 /sys/bus/ap/devices/[card04] 49 50* AP domain 51 52 An adapter is partitioned into domains. An adapter can hold up to 256 domains 53 depending upon the adapter type and hardware configuration. A domain is 54 identified by a number from 0 to 255; however, the maximum domain number is 55 determined by machine model and/or adapter type.. A domain can be thought of 56 as a set of hardware registers and memory used for processing AP commands. A 57 domain can be configured with a secure private key used for clear key 58 encryption. A domain is classified in one of two ways depending upon how it 59 may be accessed: 60 61 * Usage domains are domains that are targeted by an AP instruction to 62 process an AP command. 63 64 * Control domains are domains that are changed by an AP command sent to a 65 usage domain; for example, to set the secure private key for the control 66 domain. 67 68 The AP usage and control domains are assigned to a given LPAR via the system's 69 Activation Profile which can be edited via the HMC. When a linux host system 70 is IPL'd in the LPAR, the AP bus module detects the AP usage and control 71 domains assigned to the LPAR. The domain number of each usage domain and 72 adapter number of each AP adapter are combined to create AP queue devices 73 (see AP Queue section below). The domain number of each control domain will be 74 represented in a bitmask and stored in a sysfs file 75 /sys/bus/ap/ap_control_domain_mask. The bits in the mask, from most to least 76 significant bit, correspond to domains 0-255. 77 78* AP Queue 79 80 An AP queue is the means by which an AP command is sent to a usage domain 81 inside a specific adapter. An AP queue is identified by a tuple 82 comprised of an AP adapter ID (APID) and an AP queue index (APQI). The 83 APQI corresponds to a given usage domain number within the adapter. This tuple 84 forms an AP Queue Number (APQN) uniquely identifying an AP queue. AP 85 instructions include a field containing the APQN to identify the AP queue to 86 which the AP command is to be sent for processing. 87 88 The AP bus will create a sysfs device for each APQN that can be derived from 89 the cross product of the AP adapter and usage domain numbers detected when the 90 AP bus module is loaded. For example, if adapters 4 and 10 (0x0a) and usage 91 domains 6 and 71 (0x47) are assigned to the LPAR, the AP bus will create the 92 following sysfs entries:: 93 94 /sys/devices/ap/card04/04.0006 95 /sys/devices/ap/card04/04.0047 96 /sys/devices/ap/card0a/0a.0006 97 /sys/devices/ap/card0a/0a.0047 98 99 The following symbolic links to these devices will be created in the AP bus 100 devices subdirectory:: 101 102 /sys/bus/ap/devices/[04.0006] 103 /sys/bus/ap/devices/[04.0047] 104 /sys/bus/ap/devices/[0a.0006] 105 /sys/bus/ap/devices/[0a.0047] 106 107* AP Instructions: 108 109 There are three AP instructions: 110 111 * NQAP: to enqueue an AP command-request message to a queue 112 * DQAP: to dequeue an AP command-reply message from a queue 113 * PQAP: to administer the queues 114 115 AP instructions identify the domain that is targeted to process the AP 116 command; this must be one of the usage domains. An AP command may modify a 117 domain that is not one of the usage domains, but the modified domain 118 must be one of the control domains. 119 120AP and SIE 121========== 122Let's now take a look at how AP instructions executed on a guest are interpreted 123by the hardware. 124 125A satellite control block called the Crypto Control Block (CRYCB) is attached to 126our main hardware virtualization control block. The CRYCB contains an AP Control 127Block (APCB) that has three fields to identify the adapters, usage domains and 128control domains assigned to the KVM guest: 129 130* The AP Mask (APM) field is a bit mask that identifies the AP adapters assigned 131 to the KVM guest. Each bit in the mask, from left to right, corresponds to 132 an APID from 0-255. If a bit is set, the corresponding adapter is valid for 133 use by the KVM guest. 134 135* The AP Queue Mask (AQM) field is a bit mask identifying the AP usage domains 136 assigned to the KVM guest. Each bit in the mask, from left to right, 137 corresponds to an AP queue index (APQI) from 0-255. If a bit is set, the 138 corresponding queue is valid for use by the KVM guest. 139 140* The AP Domain Mask field is a bit mask that identifies the AP control domains 141 assigned to the KVM guest. The ADM bit mask controls which domains can be 142 changed by an AP command-request message sent to a usage domain from the 143 guest. Each bit in the mask, from left to right, corresponds to a domain from 144 0-255. If a bit is set, the corresponding domain can be modified by an AP 145 command-request message sent to a usage domain. 146 147If you recall from the description of an AP Queue, AP instructions include 148an APQN to identify the AP queue to which an AP command-request message is to be 149sent (NQAP and PQAP instructions), or from which a command-reply message is to 150be received (DQAP instruction). The validity of an APQN is defined by the matrix 151calculated from the APM and AQM; it is the Cartesian product of all assigned 152adapter numbers (APM) with all assigned queue indexes (AQM). For example, if 153adapters 1 and 2 and usage domains 5 and 6 are assigned to a guest, the APQNs 154(1,5), (1,6), (2,5) and (2,6) will be valid for the guest. 155 156The APQNs can provide secure key functionality - i.e., a private key is stored 157on the adapter card for each of its domains - so each APQN must be assigned to 158at most one guest or to the linux host:: 159 160 Example 1: Valid configuration: 161 ------------------------------ 162 Guest1: adapters 1,2 domains 5,6 163 Guest2: adapter 1,2 domain 7 164 165 This is valid because both guests have a unique set of APQNs: 166 Guest1 has APQNs (1,5), (1,6), (2,5), (2,6); 167 Guest2 has APQNs (1,7), (2,7) 168 169 Example 2: Valid configuration: 170 ------------------------------ 171 Guest1: adapters 1,2 domains 5,6 172 Guest2: adapters 3,4 domains 5,6 173 174 This is also valid because both guests have a unique set of APQNs: 175 Guest1 has APQNs (1,5), (1,6), (2,5), (2,6); 176 Guest2 has APQNs (3,5), (3,6), (4,5), (4,6) 177 178 Example 3: Invalid configuration: 179 -------------------------------- 180 Guest1: adapters 1,2 domains 5,6 181 Guest2: adapter 1 domains 6,7 182 183 This is an invalid configuration because both guests have access to 184 APQN (1,6). 185 186The Design 187========== 188The design introduces three new objects: 189 1901. AP matrix device 1912. VFIO AP device driver (vfio_ap.ko) 1923. VFIO AP mediated pass-through device 193 194The VFIO AP device driver 195------------------------- 196The VFIO AP (vfio_ap) device driver serves the following purposes: 197 1981. Provides the interfaces to secure APQNs for exclusive use of KVM guests. 199 2002. Sets up the VFIO mediated device interfaces to manage a vfio_ap mediated 201 device and creates the sysfs interfaces for assigning adapters, usage 202 domains, and control domains comprising the matrix for a KVM guest. 203 2043. Configures the APM, AQM and ADM in the APCB contained in the CRYCB referenced 205 by a KVM guest's SIE state description to grant the guest access to a matrix 206 of AP devices 207 208Reserve APQNs for exclusive use of KVM guests 209--------------------------------------------- 210The following block diagram illustrates the mechanism by which APQNs are 211reserved:: 212 213 +------------------+ 214 7 remove | | 215 +--------------------> cex4queue driver | 216 | | | 217 | +------------------+ 218 | 219 | 220 | +------------------+ +----------------+ 221 | 5 register driver | | 3 create | | 222 | +----------------> Device core +----------> matrix device | 223 | | | | | | 224 | | +--------^---------+ +----------------+ 225 | | | 226 | | +-------------------+ 227 | | +-----------------------------------+ | 228 | | | 4 register AP driver | | 2 register device 229 | | | | | 230 +--------+---+-v---+ +--------+-------+-+ 231 | | | | 232 | ap_bus +--------------------- > vfio_ap driver | 233 | | 8 probe | | 234 +--------^---------+ +--^--^------------+ 235 6 edit | | | 236 apmask | +-----------------------------+ | 11 mdev create 237 aqmask | | 1 modprobe | 238 +--------+-----+---+ +----------------+-+ +----------------+ 239 | | | |10 create| mediated | 240 | admin | | VFIO device core |---------> matrix | 241 | + | | | device | 242 +------+-+---------+ +--------^---------+ +--------^-------+ 243 | | | | 244 | | 9 create vfio_ap-passthrough | | 245 | +------------------------------+ | 246 +-------------------------------------------------------------+ 247 12 assign adapter/domain/control domain 248 249The process for reserving an AP queue for use by a KVM guest is: 250 2511. The administrator loads the vfio_ap device driver 2522. The vfio-ap driver during its initialization will register a single 'matrix' 253 device with the device core. This will serve as the parent device for 254 all vfio_ap mediated devices used to configure an AP matrix for a guest. 2553. The /sys/devices/vfio_ap/matrix device is created by the device core 2564. The vfio_ap device driver will register with the AP bus for AP queue devices 257 of type 10 and higher (CEX4 and newer). The driver will provide the vfio_ap 258 driver's probe and remove callback interfaces. Devices older than CEX4 queues 259 are not supported to simplify the implementation by not needlessly 260 complicating the design by supporting older devices that will go out of 261 service in the relatively near future, and for which there are few older 262 systems around on which to test. 2635. The AP bus registers the vfio_ap device driver with the device core 2646. The administrator edits the AP adapter and queue masks to reserve AP queues 265 for use by the vfio_ap device driver. 2667. The AP bus removes the AP queues reserved for the vfio_ap driver from the 267 default zcrypt cex4queue driver. 2688. The AP bus probes the vfio_ap device driver to bind the queues reserved for 269 it. 2709. The administrator creates a passthrough type vfio_ap mediated device to be 271 used by a guest 27210. The administrator assigns the adapters, usage domains and control domains 273 to be exclusively used by a guest. 274 275Set up the VFIO mediated device interfaces 276------------------------------------------ 277The VFIO AP device driver utilizes the common interfaces of the VFIO mediated 278device core driver to: 279 280* Register an AP mediated bus driver to add a vfio_ap mediated device to and 281 remove it from a VFIO group. 282* Create and destroy a vfio_ap mediated device 283* Add a vfio_ap mediated device to and remove it from the AP mediated bus driver 284* Add a vfio_ap mediated device to and remove it from an IOMMU group 285 286The following high-level block diagram shows the main components and interfaces 287of the VFIO AP mediated device driver:: 288 289 +-------------+ 290 | | 291 | +---------+ | mdev_register_driver() +--------------+ 292 | | Mdev | +<-----------------------+ | 293 | | bus | | | vfio_mdev.ko | 294 | | driver | +----------------------->+ |<-> VFIO user 295 | +---------+ | probe()/remove() +--------------+ APIs 296 | | 297 | MDEV CORE | 298 | MODULE | 299 | mdev.ko | 300 | +---------+ | mdev_register_parent() +--------------+ 301 | |Physical | +<-----------------------+ | 302 | | device | | | vfio_ap.ko |<-> matrix 303 | |interface| +----------------------->+ | device 304 | +---------+ | callback +--------------+ 305 +-------------+ 306 307During initialization of the vfio_ap module, the matrix device is registered 308with an 'mdev_parent_ops' structure that provides the sysfs attribute 309structures, mdev functions and callback interfaces for managing the mediated 310matrix device. 311 312* sysfs attribute structures: 313 314 supported_type_groups 315 The VFIO mediated device framework supports creation of user-defined 316 mediated device types. These mediated device types are specified 317 via the 'supported_type_groups' structure when a device is registered 318 with the mediated device framework. The registration process creates the 319 sysfs structures for each mediated device type specified in the 320 'mdev_supported_types' sub-directory of the device being registered. Along 321 with the device type, the sysfs attributes of the mediated device type are 322 provided. 323 324 The VFIO AP device driver will register one mediated device type for 325 passthrough devices: 326 327 /sys/devices/vfio_ap/matrix/mdev_supported_types/vfio_ap-passthrough 328 329 Only the read-only attributes required by the VFIO mdev framework will 330 be provided:: 331 332 ... name 333 ... device_api 334 ... available_instances 335 ... device_api 336 337 Where: 338 339 * name: 340 specifies the name of the mediated device type 341 * device_api: 342 the mediated device type's API 343 * available_instances: 344 the number of vfio_ap mediated passthrough devices 345 that can be created 346 * device_api: 347 specifies the VFIO API 348 mdev_attr_groups 349 This attribute group identifies the user-defined sysfs attributes of the 350 mediated device. When a device is registered with the VFIO mediated device 351 framework, the sysfs attribute files identified in the 'mdev_attr_groups' 352 structure will be created in the vfio_ap mediated device's directory. The 353 sysfs attributes for a vfio_ap mediated device are: 354 355 assign_adapter / unassign_adapter: 356 Write-only attributes for assigning/unassigning an AP adapter to/from the 357 vfio_ap mediated device. To assign/unassign an adapter, the APID of the 358 adapter is echoed into the respective attribute file. 359 assign_domain / unassign_domain: 360 Write-only attributes for assigning/unassigning an AP usage domain to/from 361 the vfio_ap mediated device. To assign/unassign a domain, the domain 362 number of the usage domain is echoed into the respective attribute 363 file. 364 matrix: 365 A read-only file for displaying the APQNs derived from the Cartesian 366 product of the adapter and domain numbers assigned to the vfio_ap mediated 367 device. 368 guest_matrix: 369 A read-only file for displaying the APQNs derived from the Cartesian 370 product of the adapter and domain numbers assigned to the APM and AQM 371 fields respectively of the KVM guest's CRYCB. This may differ from the 372 the APQNs assigned to the vfio_ap mediated device if any APQN does not 373 reference a queue device bound to the vfio_ap device driver (i.e., the 374 queue is not in the host's AP configuration). 375 assign_control_domain / unassign_control_domain: 376 Write-only attributes for assigning/unassigning an AP control domain 377 to/from the vfio_ap mediated device. To assign/unassign a control domain, 378 the ID of the domain to be assigned/unassigned is echoed into the 379 respective attribute file. 380 control_domains: 381 A read-only file for displaying the control domain numbers assigned to the 382 vfio_ap mediated device. 383 384* functions: 385 386 create: 387 allocates the ap_matrix_mdev structure used by the vfio_ap driver to: 388 389 * Store the reference to the KVM structure for the guest using the mdev 390 * Store the AP matrix configuration for the adapters, domains, and control 391 domains assigned via the corresponding sysfs attributes files 392 * Store the AP matrix configuration for the adapters, domains and control 393 domains available to a guest. A guest may not be provided access to APQNs 394 referencing queue devices that do not exist, or are not bound to the 395 vfio_ap device driver. 396 397 remove: 398 deallocates the vfio_ap mediated device's ap_matrix_mdev structure. 399 This will be allowed only if a running guest is not using the mdev. 400 401* callback interfaces 402 403 open_device: 404 The vfio_ap driver uses this callback to register a 405 VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the matrix mdev 406 devices. The open_device callback is invoked by userspace to connect the 407 VFIO iommu group for the matrix mdev device to the MDEV bus. Access to the 408 KVM structure used to configure the KVM guest is provided via this callback. 409 The KVM structure, is used to configure the guest's access to the AP matrix 410 defined via the vfio_ap mediated device's sysfs attribute files. 411 412 close_device: 413 unregisters the VFIO_GROUP_NOTIFY_SET_KVM notifier callback function for the 414 matrix mdev device and deconfigures the guest's AP matrix. 415 416 ioctl: 417 this callback handles the VFIO_DEVICE_GET_INFO and VFIO_DEVICE_RESET ioctls 418 defined by the vfio framework. 419 420Configure the guest's AP resources 421---------------------------------- 422Configuring the AP resources for a KVM guest will be performed when the 423VFIO_GROUP_NOTIFY_SET_KVM notifier callback is invoked. The notifier 424function is called when userspace connects to KVM. The guest's AP resources are 425configured via it's APCB by: 426 427* Setting the bits in the APM corresponding to the APIDs assigned to the 428 vfio_ap mediated device via its 'assign_adapter' interface. 429* Setting the bits in the AQM corresponding to the domains assigned to the 430 vfio_ap mediated device via its 'assign_domain' interface. 431* Setting the bits in the ADM corresponding to the domain dIDs assigned to the 432 vfio_ap mediated device via its 'assign_control_domains' interface. 433 434The linux device model precludes passing a device through to a KVM guest that 435is not bound to the device driver facilitating its pass-through. Consequently, 436an APQN that does not reference a queue device bound to the vfio_ap device 437driver will not be assigned to a KVM guest's matrix. The AP architecture, 438however, does not provide a means to filter individual APQNs from the guest's 439matrix, so the adapters, domains and control domains assigned to vfio_ap 440mediated device via its sysfs 'assign_adapter', 'assign_domain' and 441'assign_control_domain' interfaces will be filtered before providing the AP 442configuration to a guest: 443 444* The APIDs of the adapters, the APQIs of the domains and the domain numbers of 445 the control domains assigned to the matrix mdev that are not also assigned to 446 the host's AP configuration will be filtered. 447 448* Each APQN derived from the Cartesian product of the APIDs and APQIs assigned 449 to the vfio_ap mdev is examined and if any one of them does not reference a 450 queue device bound to the vfio_ap device driver, the adapter will not be 451 plugged into the guest (i.e., the bit corresponding to its APID will not be 452 set in the APM of the guest's APCB). 453 454The CPU model features for AP 455----------------------------- 456The AP stack relies on the presence of the AP instructions as well as three 457facilities: The AP Facilities Test (APFT) facility; the AP Query 458Configuration Information (QCI) facility; and the AP Queue Interruption Control 459facility. These features/facilities are made available to a KVM guest via the 460following CPU model features: 461 4621. ap: Indicates whether the AP instructions are installed on the guest. This 463 feature will be enabled by KVM only if the AP instructions are installed 464 on the host. 465 4662. apft: Indicates the APFT facility is available on the guest. This facility 467 can be made available to the guest only if it is available on the host (i.e., 468 facility bit 15 is set). 469 4703. apqci: Indicates the AP QCI facility is available on the guest. This facility 471 can be made available to the guest only if it is available on the host (i.e., 472 facility bit 12 is set). 473 4744. apqi: Indicates AP Queue Interruption Control faclity is available on the 475 guest. This facility can be made available to the guest only if it is 476 available on the host (i.e., facility bit 65 is set). 477 478Note: If the user chooses to specify a CPU model different than the 'host' 479model to QEMU, the CPU model features and facilities need to be turned on 480explicitly; for example:: 481 482 /usr/bin/qemu-system-s390x ... -cpu z13,ap=on,apqci=on,apft=on,apqi=on 483 484A guest can be precluded from using AP features/facilities by turning them off 485explicitly; for example:: 486 487 /usr/bin/qemu-system-s390x ... -cpu host,ap=off,apqci=off,apft=off,apqi=off 488 489Note: If the APFT facility is turned off (apft=off) for the guest, the guest 490will not see any AP devices. The zcrypt device drivers on the guest that 491register for type 10 and newer AP devices - i.e., the cex4card and cex4queue 492device drivers - need the APFT facility to ascertain the facilities installed on 493a given AP device. If the APFT facility is not installed on the guest, then no 494adapter or domain devices will get created by the AP bus running on the 495guest because only type 10 and newer devices can be configured for guest use. 496 497Example 498======= 499Let's now provide an example to illustrate how KVM guests may be given 500access to AP facilities. For this example, we will show how to configure 501three guests such that executing the lszcrypt command on the guests would 502look like this: 503 504Guest1 505------ 506=========== ===== ============ 507CARD.DOMAIN TYPE MODE 508=========== ===== ============ 50905 CEX5C CCA-Coproc 51005.0004 CEX5C CCA-Coproc 51105.00ab CEX5C CCA-Coproc 51206 CEX5A Accelerator 51306.0004 CEX5A Accelerator 51406.00ab CEX5A Accelerator 515=========== ===== ============ 516 517Guest2 518------ 519=========== ===== ============ 520CARD.DOMAIN TYPE MODE 521=========== ===== ============ 52205 CEX5C CCA-Coproc 52305.0047 CEX5C CCA-Coproc 52405.00ff CEX5C CCA-Coproc 525=========== ===== ============ 526 527Guest3 528------ 529=========== ===== ============ 530CARD.DOMAIN TYPE MODE 531=========== ===== ============ 53206 CEX5A Accelerator 53306.0047 CEX5A Accelerator 53406.00ff CEX5A Accelerator 535=========== ===== ============ 536 537These are the steps: 538 5391. Install the vfio_ap module on the linux host. The dependency chain for the 540 vfio_ap module is: 541 * iommu 542 * s390 543 * zcrypt 544 * vfio 545 * vfio_mdev 546 * vfio_mdev_device 547 * KVM 548 549 To build the vfio_ap module, the kernel build must be configured with the 550 following Kconfig elements selected: 551 * IOMMU_SUPPORT 552 * S390 553 * ZCRYPT 554 * S390_AP_IOMMU 555 * VFIO 556 * VFIO_MDEV 557 * KVM 558 559 If using make menuconfig select the following to build the vfio_ap module:: 560 561 -> Device Drivers 562 -> IOMMU Hardware Support 563 select S390 AP IOMMU Support 564 -> VFIO Non-Privileged userspace driver framework 565 -> Mediated device driver frramework 566 -> VFIO driver for Mediated devices 567 -> I/O subsystem 568 -> VFIO support for AP devices 569 5702. Secure the AP queues to be used by the three guests so that the host can not 571 access them. To secure them, there are two sysfs files that specify 572 bitmasks marking a subset of the APQN range as usable only by the default AP 573 queue device drivers. All remaining APQNs are available for use by 574 any other device driver. The vfio_ap device driver is currently the only 575 non-default device driver. The location of the sysfs files containing the 576 masks are:: 577 578 /sys/bus/ap/apmask 579 /sys/bus/ap/aqmask 580 581 The 'apmask' is a 256-bit mask that identifies a set of AP adapter IDs 582 (APID). Each bit in the mask, from left to right, corresponds to an APID from 583 0-255. If a bit is set, the APID belongs to the subset of APQNs marked as 584 available only to the default AP queue device drivers. 585 586 The 'aqmask' is a 256-bit mask that identifies a set of AP queue indexes 587 (APQI). Each bit in the mask, from left to right, corresponds to an APQI from 588 0-255. If a bit is set, the APQI belongs to the subset of APQNs marked as 589 available only to the default AP queue device drivers. 590 591 The Cartesian product of the APIDs corresponding to the bits set in the 592 apmask and the APQIs corresponding to the bits set in the aqmask comprise 593 the subset of APQNs that can be used only by the host default device drivers. 594 All other APQNs are available to the non-default device drivers such as the 595 vfio_ap driver. 596 597 Take, for example, the following masks:: 598 599 apmask: 600 0x7d00000000000000000000000000000000000000000000000000000000000000 601 602 aqmask: 603 0x8000000000000000000000000000000000000000000000000000000000000000 604 605 The masks indicate: 606 607 * Adapters 1, 2, 3, 4, 5, and 7 are available for use by the host default 608 device drivers. 609 610 * Domain 0 is available for use by the host default device drivers 611 612 * The subset of APQNs available for use only by the default host device 613 drivers are: 614 615 (1,0), (2,0), (3,0), (4.0), (5,0) and (7,0) 616 617 * All other APQNs are available for use by the non-default device drivers. 618 619 The APQN of each AP queue device assigned to the linux host is checked by the 620 AP bus against the set of APQNs derived from the Cartesian product of APIDs 621 and APQIs marked as available to the default AP queue device drivers. If a 622 match is detected, only the default AP queue device drivers will be probed; 623 otherwise, the vfio_ap device driver will be probed. 624 625 By default, the two masks are set to reserve all APQNs for use by the default 626 AP queue device drivers. There are two ways the default masks can be changed: 627 628 1. The sysfs mask files can be edited by echoing a string into the 629 respective sysfs mask file in one of two formats: 630 631 * An absolute hex string starting with 0x - like "0x12345678" - sets 632 the mask. If the given string is shorter than the mask, it is padded 633 with 0s on the right; for example, specifying a mask value of 0x41 is 634 the same as specifying:: 635 636 0x4100000000000000000000000000000000000000000000000000000000000000 637 638 Keep in mind that the mask reads from left to right, so the mask 639 above identifies device numbers 1 and 7 (01000001). 640 641 If the string is longer than the mask, the operation is terminated with 642 an error (EINVAL). 643 644 * Individual bits in the mask can be switched on and off by specifying 645 each bit number to be switched in a comma separated list. Each bit 646 number string must be prepended with a ('+') or minus ('-') to indicate 647 the corresponding bit is to be switched on ('+') or off ('-'). Some 648 valid values are: 649 650 - "+0" switches bit 0 on 651 - "-13" switches bit 13 off 652 - "+0x41" switches bit 65 on 653 - "-0xff" switches bit 255 off 654 655 The following example: 656 657 +0,-6,+0x47,-0xf0 658 659 Switches bits 0 and 71 (0x47) on 660 661 Switches bits 6 and 240 (0xf0) off 662 663 Note that the bits not specified in the list remain as they were before 664 the operation. 665 666 2. The masks can also be changed at boot time via parameters on the kernel 667 command line like this: 668 669 ap.apmask=0xffff ap.aqmask=0x40 670 671 This would create the following masks:: 672 673 apmask: 674 0xffff000000000000000000000000000000000000000000000000000000000000 675 676 aqmask: 677 0x4000000000000000000000000000000000000000000000000000000000000000 678 679 Resulting in these two pools:: 680 681 default drivers pool: adapter 0-15, domain 1 682 alternate drivers pool: adapter 16-255, domains 0, 2-255 683 684 **Note:** 685 Changing a mask such that one or more APQNs will be taken from a vfio_ap 686 mediated device (see below) will fail with an error (EBUSY). A message 687 is logged to the kernel ring buffer which can be viewed with the 'dmesg' 688 command. The output identifies each APQN flagged as 'in use' and identifies 689 the vfio_ap mediated device to which it is assigned; for example: 690 691 Userspace may not re-assign queue 05.0054 already assigned to 62177883-f1bb-47f0-914d-32a22e3a8804 692 Userspace may not re-assign queue 04.0054 already assigned to cef03c3c-903d-4ecc-9a83-40694cb8aee4 693 694Securing the APQNs for our example 695---------------------------------- 696 To secure the AP queues 05.0004, 05.0047, 05.00ab, 05.00ff, 06.0004, 06.0047, 697 06.00ab, and 06.00ff for use by the vfio_ap device driver, the corresponding 698 APQNs can be removed from the default masks using either of the following 699 commands:: 700 701 echo -5,-6 > /sys/bus/ap/apmask 702 703 echo -4,-0x47,-0xab,-0xff > /sys/bus/ap/aqmask 704 705 Or the masks can be set as follows:: 706 707 echo 0xf9ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff \ 708 > apmask 709 710 echo 0xf7fffffffffffffffeffffffffffffffffffffffffeffffffffffffffffffffe \ 711 > aqmask 712 713 This will result in AP queues 05.0004, 05.0047, 05.00ab, 05.00ff, 06.0004, 714 06.0047, 06.00ab, and 06.00ff getting bound to the vfio_ap device driver. The 715 sysfs directory for the vfio_ap device driver will now contain symbolic links 716 to the AP queue devices bound to it:: 717 718 /sys/bus/ap 719 ... [drivers] 720 ...... [vfio_ap] 721 ......... [05.0004] 722 ......... [05.0047] 723 ......... [05.00ab] 724 ......... [05.00ff] 725 ......... [06.0004] 726 ......... [06.0047] 727 ......... [06.00ab] 728 ......... [06.00ff] 729 730 Keep in mind that only type 10 and newer adapters (i.e., CEX4 and later) 731 can be bound to the vfio_ap device driver. The reason for this is to 732 simplify the implementation by not needlessly complicating the design by 733 supporting older devices that will go out of service in the relatively near 734 future and for which there are few older systems on which to test. 735 736 The administrator, therefore, must take care to secure only AP queues that 737 can be bound to the vfio_ap device driver. The device type for a given AP 738 queue device can be read from the parent card's sysfs directory. For example, 739 to see the hardware type of the queue 05.0004: 740 741 cat /sys/bus/ap/devices/card05/hwtype 742 743 The hwtype must be 10 or higher (CEX4 or newer) in order to be bound to the 744 vfio_ap device driver. 745 7463. Create the mediated devices needed to configure the AP matrixes for the 747 three guests and to provide an interface to the vfio_ap driver for 748 use by the guests:: 749 750 /sys/devices/vfio_ap/matrix/ 751 --- [mdev_supported_types] 752 ------ [vfio_ap-passthrough] (passthrough vfio_ap mediated device type) 753 --------- create 754 --------- [devices] 755 756 To create the mediated devices for the three guests:: 757 758 uuidgen > create 759 uuidgen > create 760 uuidgen > create 761 762 or 763 764 echo $uuid1 > create 765 echo $uuid2 > create 766 echo $uuid3 > create 767 768 This will create three mediated devices in the [devices] subdirectory named 769 after the UUID written to the create attribute file. We call them $uuid1, 770 $uuid2 and $uuid3 and this is the sysfs directory structure after creation:: 771 772 /sys/devices/vfio_ap/matrix/ 773 --- [mdev_supported_types] 774 ------ [vfio_ap-passthrough] 775 --------- [devices] 776 ------------ [$uuid1] 777 --------------- assign_adapter 778 --------------- assign_control_domain 779 --------------- assign_domain 780 --------------- matrix 781 --------------- unassign_adapter 782 --------------- unassign_control_domain 783 --------------- unassign_domain 784 785 ------------ [$uuid2] 786 --------------- assign_adapter 787 --------------- assign_control_domain 788 --------------- assign_domain 789 --------------- matrix 790 --------------- unassign_adapter 791 ----------------unassign_control_domain 792 ----------------unassign_domain 793 794 ------------ [$uuid3] 795 --------------- assign_adapter 796 --------------- assign_control_domain 797 --------------- assign_domain 798 --------------- matrix 799 --------------- unassign_adapter 800 ----------------unassign_control_domain 801 ----------------unassign_domain 802 803 Note *****: The vfio_ap mdevs do not persist across reboots unless the 804 mdevctl tool is used to create and persist them. 805 8064. The administrator now needs to configure the matrixes for the mediated 807 devices $uuid1 (for Guest1), $uuid2 (for Guest2) and $uuid3 (for Guest3). 808 809 This is how the matrix is configured for Guest1:: 810 811 echo 5 > assign_adapter 812 echo 6 > assign_adapter 813 echo 4 > assign_domain 814 echo 0xab > assign_domain 815 816 Control domains can similarly be assigned using the assign_control_domain 817 sysfs file. 818 819 If a mistake is made configuring an adapter, domain or control domain, 820 you can use the unassign_xxx files to unassign the adapter, domain or 821 control domain. 822 823 To display the matrix configuration for Guest1:: 824 825 cat matrix 826 827 To display the matrix that is or will be assigned to Guest1:: 828 829 cat guest_matrix 830 831 This is how the matrix is configured for Guest2:: 832 833 echo 5 > assign_adapter 834 echo 0x47 > assign_domain 835 echo 0xff > assign_domain 836 837 This is how the matrix is configured for Guest3:: 838 839 echo 6 > assign_adapter 840 echo 0x47 > assign_domain 841 echo 0xff > assign_domain 842 843 In order to successfully assign an adapter: 844 845 * The adapter number specified must represent a value from 0 up to the 846 maximum adapter number configured for the system. If an adapter number 847 higher than the maximum is specified, the operation will terminate with 848 an error (ENODEV). 849 850 Note: The maximum adapter number can be obtained via the sysfs 851 /sys/bus/ap/ap_max_adapter_id attribute file. 852 853 * Each APQN derived from the Cartesian product of the APID of the adapter 854 being assigned and the APQIs of the domains previously assigned: 855 856 - Must only be available to the vfio_ap device driver as specified in the 857 sysfs /sys/bus/ap/apmask and /sys/bus/ap/aqmask attribute files. If even 858 one APQN is reserved for use by the host device driver, the operation 859 will terminate with an error (EADDRNOTAVAIL). 860 861 - Must NOT be assigned to another vfio_ap mediated device. If even one APQN 862 is assigned to another vfio_ap mediated device, the operation will 863 terminate with an error (EBUSY). 864 865 - Must NOT be assigned while the sysfs /sys/bus/ap/apmask and 866 sys/bus/ap/aqmask attribute files are being edited or the operation may 867 terminate with an error (EBUSY). 868 869 In order to successfully assign a domain: 870 871 * The domain number specified must represent a value from 0 up to the 872 maximum domain number configured for the system. If a domain number 873 higher than the maximum is specified, the operation will terminate with 874 an error (ENODEV). 875 876 Note: The maximum domain number can be obtained via the sysfs 877 /sys/bus/ap/ap_max_domain_id attribute file. 878 879 * Each APQN derived from the Cartesian product of the APQI of the domain 880 being assigned and the APIDs of the adapters previously assigned: 881 882 - Must only be available to the vfio_ap device driver as specified in the 883 sysfs /sys/bus/ap/apmask and /sys/bus/ap/aqmask attribute files. If even 884 one APQN is reserved for use by the host device driver, the operation 885 will terminate with an error (EADDRNOTAVAIL). 886 887 - Must NOT be assigned to another vfio_ap mediated device. If even one APQN 888 is assigned to another vfio_ap mediated device, the operation will 889 terminate with an error (EBUSY). 890 891 - Must NOT be assigned while the sysfs /sys/bus/ap/apmask and 892 sys/bus/ap/aqmask attribute files are being edited or the operation may 893 terminate with an error (EBUSY). 894 895 In order to successfully assign a control domain: 896 897 * The domain number specified must represent a value from 0 up to the maximum 898 domain number configured for the system. If a control domain number higher 899 than the maximum is specified, the operation will terminate with an 900 error (ENODEV). 901 9025. Start Guest1:: 903 904 /usr/bin/qemu-system-s390x ... -cpu host,ap=on,apqci=on,apft=on,apqi=on \ 905 -device vfio-ap,sysfsdev=/sys/devices/vfio_ap/matrix/$uuid1 ... 906 9077. Start Guest2:: 908 909 /usr/bin/qemu-system-s390x ... -cpu host,ap=on,apqci=on,apft=on,apqi=on \ 910 -device vfio-ap,sysfsdev=/sys/devices/vfio_ap/matrix/$uuid2 ... 911 9127. Start Guest3:: 913 914 /usr/bin/qemu-system-s390x ... -cpu host,ap=on,apqci=on,apft=on,apqi=on \ 915 -device vfio-ap,sysfsdev=/sys/devices/vfio_ap/matrix/$uuid3 ... 916 917When the guest is shut down, the vfio_ap mediated devices may be removed. 918 919Using our example again, to remove the vfio_ap mediated device $uuid1:: 920 921 /sys/devices/vfio_ap/matrix/ 922 --- [mdev_supported_types] 923 ------ [vfio_ap-passthrough] 924 --------- [devices] 925 ------------ [$uuid1] 926 --------------- remove 927 928:: 929 930 echo 1 > remove 931 932This will remove all of the matrix mdev device's sysfs structures including 933the mdev device itself. To recreate and reconfigure the matrix mdev device, 934all of the steps starting with step 3 will have to be performed again. Note 935that the remove will fail if a guest using the vfio_ap mdev is still running. 936 937It is not necessary to remove a vfio_ap mdev, but one may want to 938remove it if no guest will use it during the remaining lifetime of the linux 939host. If the vfio_ap mdev is removed, one may want to also reconfigure 940the pool of adapters and queues reserved for use by the default drivers. 941 942Hot plug/unplug support: 943======================== 944An adapter, domain or control domain may be hot plugged into a running KVM 945guest by assigning it to the vfio_ap mediated device being used by the guest if 946the following conditions are met: 947 948* The adapter, domain or control domain must also be assigned to the host's 949 AP configuration. 950 951* Each APQN derived from the Cartesian product comprised of the APID of the 952 adapter being assigned and the APQIs of the domains assigned must reference a 953 queue device bound to the vfio_ap device driver. 954 955* To hot plug a domain, each APQN derived from the Cartesian product 956 comprised of the APQI of the domain being assigned and the APIDs of the 957 adapters assigned must reference a queue device bound to the vfio_ap device 958 driver. 959 960An adapter, domain or control domain may be hot unplugged from a running KVM 961guest by unassigning it from the vfio_ap mediated device being used by the 962guest. 963 964Over-provisioning of AP queues for a KVM guest: 965=============================================== 966Over-provisioning is defined herein as the assignment of adapters or domains to 967a vfio_ap mediated device that do not reference AP devices in the host's AP 968configuration. The idea here is that when the adapter or domain becomes 969available, it will be automatically hot-plugged into the KVM guest using 970the vfio_ap mediated device to which it is assigned as long as each new APQN 971resulting from plugging it in references a queue device bound to the vfio_ap 972device driver. 973 974Limitations 975=========== 976Live guest migration is not supported for guests using AP devices without 977intervention by a system administrator. Before a KVM guest can be migrated, 978the vfio_ap mediated device must be removed. Unfortunately, it can not be 979removed manually (i.e., echo 1 > /sys/devices/vfio_ap/matrix/$UUID/remove) while 980the mdev is in use by a KVM guest. If the guest is being emulated by QEMU, 981its mdev can be hot unplugged from the guest in one of two ways: 982 9831. If the KVM guest was started with libvirt, you can hot unplug the mdev via 984 the following commands: 985 986 virsh detach-device <guestname> <path-to-device-xml> 987 988 For example, to hot unplug mdev 62177883-f1bb-47f0-914d-32a22e3a8804 from 989 the guest named 'my-guest': 990 991 virsh detach-device my-guest ~/config/my-guest-hostdev.xml 992 993 The contents of my-guest-hostdev.xml: 994 995.. code-block:: xml 996 997 <hostdev mode='subsystem' type='mdev' managed='no' model='vfio-ap'> 998 <source> 999 <address uuid='62177883-f1bb-47f0-914d-32a22e3a8804'/> 1000 </source> 1001 </hostdev> 1002 1003 1004 virsh qemu-monitor-command <guest-name> --hmp "device-del <device-id>" 1005 1006 For example, to hot unplug the vfio_ap mediated device identified on the 1007 qemu command line with 'id=hostdev0' from the guest named 'my-guest': 1008 1009.. code-block:: sh 1010 1011 virsh qemu-monitor-command my-guest --hmp "device_del hostdev0" 1012 10132. A vfio_ap mediated device can be hot unplugged by attaching the qemu monitor 1014 to the guest and using the following qemu monitor command: 1015 1016 (QEMU) device-del id=<device-id> 1017 1018 For example, to hot unplug the vfio_ap mediated device that was specified 1019 on the qemu command line with 'id=hostdev0' when the guest was started: 1020 1021 (QEMU) device-del id=hostdev0 1022 1023After live migration of the KVM guest completes, an AP configuration can be 1024restored to the KVM guest by hot plugging a vfio_ap mediated device on the target 1025system into the guest in one of two ways: 1026 10271. If the KVM guest was started with libvirt, you can hot plug a matrix mediated 1028 device into the guest via the following virsh commands: 1029 1030 virsh attach-device <guestname> <path-to-device-xml> 1031 1032 For example, to hot plug mdev 62177883-f1bb-47f0-914d-32a22e3a8804 into 1033 the guest named 'my-guest': 1034 1035 virsh attach-device my-guest ~/config/my-guest-hostdev.xml 1036 1037 The contents of my-guest-hostdev.xml: 1038 1039.. code-block:: xml 1040 1041 <hostdev mode='subsystem' type='mdev' managed='no' model='vfio-ap'> 1042 <source> 1043 <address uuid='62177883-f1bb-47f0-914d-32a22e3a8804'/> 1044 </source> 1045 </hostdev> 1046 1047 1048 virsh qemu-monitor-command <guest-name> --hmp \ 1049 "device_add vfio-ap,sysfsdev=<path-to-mdev>,id=<device-id>" 1050 1051 For example, to hot plug the vfio_ap mediated device 1052 62177883-f1bb-47f0-914d-32a22e3a8804 into the guest named 'my-guest' with 1053 device-id hostdev0: 1054 1055 virsh qemu-monitor-command my-guest --hmp \ 1056 "device_add vfio-ap,\ 1057 sysfsdev=/sys/devices/vfio_ap/matrix/62177883-f1bb-47f0-914d-32a22e3a8804,\ 1058 id=hostdev0" 1059 10602. A vfio_ap mediated device can be hot plugged by attaching the qemu monitor 1061 to the guest and using the following qemu monitor command: 1062 1063 (qemu) device_add "vfio-ap,sysfsdev=<path-to-mdev>,id=<device-id>" 1064 1065 For example, to plug the vfio_ap mediated device 1066 62177883-f1bb-47f0-914d-32a22e3a8804 into the guest with the device-id 1067 hostdev0: 1068 1069 (QEMU) device-add "vfio-ap,\ 1070 sysfsdev=/sys/devices/vfio_ap/matrix/62177883-f1bb-47f0-914d-32a22e3a8804,\ 1071 id=hostdev0" 1072
