1.. Copyright 2020 DisplayLink (UK) Ltd. 2 3=================== 4Userland interfaces 5=================== 6 7The DRM core exports several interfaces to applications, generally 8intended to be used through corresponding libdrm wrapper functions. In 9addition, drivers export device-specific interfaces for use by userspace 10drivers & device-aware applications through ioctls and sysfs files. 11 12External interfaces include: memory mapping, context management, DMA 13operations, AGP management, vblank control, fence management, memory 14management, and output management. 15 16Cover generic ioctls and sysfs layout here. We only need high-level 17info, since man pages should cover the rest. 18 19libdrm Device Lookup 20==================== 21 22.. kernel-doc:: drivers/gpu/drm/drm_ioctl.c 23 :doc: getunique and setversion story 24 25 26.. _drm_primary_node: 27 28Primary Nodes, DRM Master and Authentication 29============================================ 30 31.. kernel-doc:: drivers/gpu/drm/drm_auth.c 32 :doc: master and authentication 33 34.. kernel-doc:: drivers/gpu/drm/drm_auth.c 35 :export: 36 37.. kernel-doc:: include/drm/drm_auth.h 38 :internal: 39 40 41.. _drm_leasing: 42 43DRM Display Resource Leasing 44============================ 45 46.. kernel-doc:: drivers/gpu/drm/drm_lease.c 47 :doc: drm leasing 48 49Open-Source Userspace Requirements 50================================== 51 52The DRM subsystem has stricter requirements than most other kernel subsystems on 53what the userspace side for new uAPI needs to look like. This section here 54explains what exactly those requirements are, and why they exist. 55 56The short summary is that any addition of DRM uAPI requires corresponding 57open-sourced userspace patches, and those patches must be reviewed and ready for 58merging into a suitable and canonical upstream project. 59 60GFX devices (both display and render/GPU side) are really complex bits of 61hardware, with userspace and kernel by necessity having to work together really 62closely. The interfaces, for rendering and modesetting, must be extremely wide 63and flexible, and therefore it is almost always impossible to precisely define 64them for every possible corner case. This in turn makes it really practically 65infeasible to differentiate between behaviour that's required by userspace, and 66which must not be changed to avoid regressions, and behaviour which is only an 67accidental artifact of the current implementation. 68 69Without access to the full source code of all userspace users that means it 70becomes impossible to change the implementation details, since userspace could 71depend upon the accidental behaviour of the current implementation in minute 72details. And debugging such regressions without access to source code is pretty 73much impossible. As a consequence this means: 74 75- The Linux kernel's "no regression" policy holds in practice only for 76 open-source userspace of the DRM subsystem. DRM developers are perfectly fine 77 if closed-source blob drivers in userspace use the same uAPI as the open 78 drivers, but they must do so in the exact same way as the open drivers. 79 Creative (ab)use of the interfaces will, and in the past routinely has, lead 80 to breakage. 81 82- Any new userspace interface must have an open-source implementation as 83 demonstration vehicle. 84 85The other reason for requiring open-source userspace is uAPI review. Since the 86kernel and userspace parts of a GFX stack must work together so closely, code 87review can only assess whether a new interface achieves its goals by looking at 88both sides. Making sure that the interface indeed covers the use-case fully 89leads to a few additional requirements: 90 91- The open-source userspace must not be a toy/test application, but the real 92 thing. Specifically it needs to handle all the usual error and corner cases. 93 These are often the places where new uAPI falls apart and hence essential to 94 assess the fitness of a proposed interface. 95 96- The userspace side must be fully reviewed and tested to the standards of that 97 userspace project. For e.g. mesa this means piglit testcases and review on the 98 mailing list. This is again to ensure that the new interface actually gets the 99 job done. The userspace-side reviewer should also provide an Acked-by on the 100 kernel uAPI patch indicating that they believe the proposed uAPI is sound and 101 sufficiently documented and validated for userspace's consumption. 102 103- The userspace patches must be against the canonical upstream, not some vendor 104 fork. This is to make sure that no one cheats on the review and testing 105 requirements by doing a quick fork. 106 107- The kernel patch can only be merged after all the above requirements are met, 108 but it **must** be merged to either drm-next or drm-misc-next **before** the 109 userspace patches land. uAPI always flows from the kernel, doing things the 110 other way round risks divergence of the uAPI definitions and header files. 111 112These are fairly steep requirements, but have grown out from years of shared 113pain and experience with uAPI added hastily, and almost always regretted about 114just as fast. GFX devices change really fast, requiring a paradigm shift and 115entire new set of uAPI interfaces every few years at least. Together with the 116Linux kernel's guarantee to keep existing userspace running for 10+ years this 117is already rather painful for the DRM subsystem, with multiple different uAPIs 118for the same thing co-existing. If we add a few more complete mistakes into the 119mix every year it would be entirely unmanageable. 120 121The DRM subsystem has however no concern with independent closed-source 122userspace implementations. To officialize that position, the DRM uAPI headers 123are covered by the MIT license. 124 125.. _drm_render_node: 126 127Render nodes 128============ 129 130DRM core provides multiple character-devices for user-space to use. 131Depending on which device is opened, user-space can perform a different 132set of operations (mainly ioctls). The primary node is always created 133and called card<num>. Additionally, a currently unused control node, 134called controlD<num> is also created. The primary node provides all 135legacy operations and historically was the only interface used by 136userspace. With KMS, the control node was introduced. However, the 137planned KMS control interface has never been written and so the control 138node stays unused to date. 139 140With the increased use of offscreen renderers and GPGPU applications, 141clients no longer require running compositors or graphics servers to 142make use of a GPU. But the DRM API required unprivileged clients to 143authenticate to a DRM-Master prior to getting GPU access. To avoid this 144step and to grant clients GPU access without authenticating, render 145nodes were introduced. Render nodes solely serve render clients, that 146is, no modesetting or privileged ioctls can be issued on render nodes. 147Only non-global rendering commands are allowed. If a driver supports 148render nodes, it must advertise it via the DRIVER_RENDER DRM driver 149capability. If not supported, the primary node must be used for render 150clients together with the legacy drmAuth authentication procedure. 151 152If a driver advertises render node support, DRM core will create a 153separate render node called renderD<num>. There will be one render node 154per device. No ioctls except PRIME-related ioctls will be allowed on 155this node. Especially GEM_OPEN will be explicitly prohibited. For a 156complete list of driver-independent ioctls that can be used on render 157nodes, see the ioctls marked DRM_RENDER_ALLOW in drm_ioctl.c Render 158nodes are designed to avoid the buffer-leaks, which occur if clients 159guess the flink names or mmap offsets on the legacy interface. 160Additionally to this basic interface, drivers must mark their 161driver-dependent render-only ioctls as DRM_RENDER_ALLOW so render 162clients can use them. Driver authors must be careful not to allow any 163privileged ioctls on render nodes. 164 165With render nodes, user-space can now control access to the render node 166via basic file-system access-modes. A running graphics server which 167authenticates clients on the privileged primary/legacy node is no longer 168required. Instead, a client can open the render node and is immediately 169granted GPU access. Communication between clients (or servers) is done 170via PRIME. FLINK from render node to legacy node is not supported. New 171clients must not use the insecure FLINK interface. 172 173Besides dropping all modeset/global ioctls, render nodes also drop the 174DRM-Master concept. There is no reason to associate render clients with 175a DRM-Master as they are independent of any graphics server. Besides, 176they must work without any running master, anyway. Drivers must be able 177to run without a master object if they support render nodes. If, on the 178other hand, a driver requires shared state between clients which is 179visible to user-space and accessible beyond open-file boundaries, they 180cannot support render nodes. 181 182Device Hot-Unplug 183================= 184 185.. note:: 186 The following is the plan. Implementation is not there yet 187 (2020 May). 188 189Graphics devices (display and/or render) may be connected via USB (e.g. 190display adapters or docking stations) or Thunderbolt (e.g. eGPU). An end 191user is able to hot-unplug this kind of devices while they are being 192used, and expects that the very least the machine does not crash. Any 193damage from hot-unplugging a DRM device needs to be limited as much as 194possible and userspace must be given the chance to handle it if it wants 195to. Ideally, unplugging a DRM device still lets a desktop continue to 196run, but that is going to need explicit support throughout the whole 197graphics stack: from kernel and userspace drivers, through display 198servers, via window system protocols, and in applications and libraries. 199 200Other scenarios that should lead to the same are: unrecoverable GPU 201crash, PCI device disappearing off the bus, or forced unbind of a driver 202from the physical device. 203 204In other words, from userspace perspective everything needs to keep on 205working more or less, until userspace stops using the disappeared DRM 206device and closes it completely. Userspace will learn of the device 207disappearance from the device removed uevent, ioctls returning ENODEV 208(or driver-specific ioctls returning driver-specific things), or open() 209returning ENXIO. 210 211Only after userspace has closed all relevant DRM device and dmabuf file 212descriptors and removed all mmaps, the DRM driver can tear down its 213instance for the device that no longer exists. If the same physical 214device somehow comes back in the mean time, it shall be a new DRM 215device. 216 217Similar to PIDs, chardev minor numbers are not recycled immediately. A 218new DRM device always picks the next free minor number compared to the 219previous one allocated, and wraps around when minor numbers are 220exhausted. 221 222The goal raises at least the following requirements for the kernel and 223drivers. 224 225Requirements for KMS UAPI 226------------------------- 227 228- KMS connectors must change their status to disconnected. 229 230- Legacy modesets and pageflips, and atomic commits, both real and 231 TEST_ONLY, and any other ioctls either fail with ENODEV or fake 232 success. 233 234- Pending non-blocking KMS operations deliver the DRM events userspace 235 is expecting. This applies also to ioctls that faked success. 236 237- open() on a device node whose underlying device has disappeared will 238 fail with ENXIO. 239 240- Attempting to create a DRM lease on a disappeared DRM device will 241 fail with ENODEV. Existing DRM leases remain and work as listed 242 above. 243 244Requirements for Render and Cross-Device UAPI 245--------------------------------------------- 246 247- All GPU jobs that can no longer run must have their fences 248 force-signalled to avoid inflicting hangs on userspace. 249 The associated error code is ENODEV. 250 251- Some userspace APIs already define what should happen when the device 252 disappears (OpenGL, GL ES: `GL_KHR_robustness`_; `Vulkan`_: 253 VK_ERROR_DEVICE_LOST; etc.). DRM drivers are free to implement this 254 behaviour the way they see best, e.g. returning failures in 255 driver-specific ioctls and handling those in userspace drivers, or 256 rely on uevents, and so on. 257 258- dmabuf which point to memory that has disappeared will either fail to 259 import with ENODEV or continue to be successfully imported if it would 260 have succeeded before the disappearance. See also about memory maps 261 below for already imported dmabufs. 262 263- Attempting to import a dmabuf to a disappeared device will either fail 264 with ENODEV or succeed if it would have succeeded without the 265 disappearance. 266 267- open() on a device node whose underlying device has disappeared will 268 fail with ENXIO. 269 270.. _GL_KHR_robustness: https://www.khronos.org/registry/OpenGL/extensions/KHR/KHR_robustness.txt 271.. _Vulkan: https://www.khronos.org/vulkan/ 272 273Requirements for Memory Maps 274---------------------------- 275 276Memory maps have further requirements that apply to both existing maps 277and maps created after the device has disappeared. If the underlying 278memory disappears, the map is created or modified such that reads and 279writes will still complete successfully but the result is undefined. 280This applies to both userspace mmap()'d memory and memory pointed to by 281dmabuf which might be mapped to other devices (cross-device dmabuf 282imports). 283 284Raising SIGBUS is not an option, because userspace cannot realistically 285handle it. Signal handlers are global, which makes them extremely 286difficult to use correctly from libraries like those that Mesa produces. 287Signal handlers are not composable, you can't have different handlers 288for GPU1 and GPU2 from different vendors, and a third handler for 289mmapped regular files. Threads cause additional pain with signal 290handling as well. 291 292Device reset 293============ 294 295The GPU stack is really complex and is prone to errors, from hardware bugs, 296faulty applications and everything in between the many layers. Some errors 297require resetting the device in order to make the device usable again. This 298section describes the expectations for DRM and usermode drivers when a 299device resets and how to propagate the reset status. 300 301Device resets can not be disabled without tainting the kernel, which can lead to 302hanging the entire kernel through shrinkers/mmu_notifiers. Userspace role in 303device resets is to propagate the message to the application and apply any 304special policy for blocking guilty applications, if any. Corollary is that 305debugging a hung GPU context require hardware support to be able to preempt such 306a GPU context while it's stopped. 307 308Kernel Mode Driver 309------------------ 310 311The KMD is responsible for checking if the device needs a reset, and to perform 312it as needed. Usually a hang is detected when a job gets stuck executing. 313 314Propagation of errors to userspace has proven to be tricky since it goes in 315the opposite direction of the usual flow of commands. Because of this vendor 316independent error handling was added to the &dma_fence object, this way drivers 317can add an error code to their fences before signaling them. See function 318dma_fence_set_error() on how to do this and for examples of error codes to use. 319 320The DRM scheduler also allows setting error codes on all pending fences when 321hardware submissions are restarted after an reset. Error codes are also 322forwarded from the hardware fence to the scheduler fence to bubble up errors 323to the higher levels of the stack and eventually userspace. 324 325Fence errors can be queried by userspace through the generic SYNC_IOC_FILE_INFO 326IOCTL as well as through driver specific interfaces. 327 328Additional to setting fence errors drivers should also keep track of resets per 329context, the DRM scheduler provides the drm_sched_entity_error() function as 330helper for this use case. After a reset, KMD should reject new command 331submissions for affected contexts. 332 333User Mode Driver 334---------------- 335 336After command submission, UMD should check if the submission was accepted or 337rejected. After a reset, KMD should reject submissions, and UMD can issue an 338ioctl to the KMD to check the reset status, and this can be checked more often 339if the UMD requires it. After detecting a reset, UMD will then proceed to report 340it to the application using the appropriate API error code, as explained in the 341section below about robustness. 342 343Robustness 344---------- 345 346The only way to try to keep a graphical API context working after a reset is if 347it complies with the robustness aspects of the graphical API that it is using. 348 349Graphical APIs provide ways to applications to deal with device resets. However, 350there is no guarantee that the app will use such features correctly, and a 351userspace that doesn't support robust interfaces (like a non-robust 352OpenGL context or API without any robustness support like libva) leave the 353robustness handling entirely to the userspace driver. There is no strong 354community consensus on what the userspace driver should do in that case, 355since all reasonable approaches have some clear downsides. 356 357OpenGL 358~~~~~~ 359 360Apps using OpenGL should use the available robust interfaces, like the 361extension ``GL_ARB_robustness`` (or ``GL_EXT_robustness`` for OpenGL ES). This 362interface tells if a reset has happened, and if so, all the context state is 363considered lost and the app proceeds by creating new ones. There's no consensus 364on what to do to if robustness is not in use. 365 366Vulkan 367~~~~~~ 368 369Apps using Vulkan should check for ``VK_ERROR_DEVICE_LOST`` for submissions. 370This error code means, among other things, that a device reset has happened and 371it needs to recreate the contexts to keep going. 372 373Reporting causes of resets 374-------------------------- 375 376Apart from propagating the reset through the stack so apps can recover, it's 377really useful for driver developers to learn more about what caused the reset in 378the first place. For this, drivers can make use of devcoredump to store relevant 379information about the reset and send device wedged event with ``none`` recovery 380method (as explained in "Device Wedging" chapter) to notify userspace, so this 381information can be collected and added to user bug reports. 382 383Device Wedging 384============== 385 386Drivers can optionally make use of device wedged event (implemented as 387drm_dev_wedged_event() in DRM subsystem), which notifies userspace of 'wedged' 388(hanged/unusable) state of the DRM device through a uevent. This is useful 389especially in cases where the device is no longer operating as expected and has 390become unrecoverable from driver context. Purpose of this implementation is to 391provide drivers a generic way to recover the device with the help of userspace 392intervention, without taking any drastic measures (like resetting or 393re-enumerating the full bus, on which the underlying physical device is sitting) 394in the driver. 395 396A 'wedged' device is basically a device that is declared dead by the driver 397after exhausting all possible attempts to recover it from driver context. The 398uevent is the notification that is sent to userspace along with a hint about 399what could possibly be attempted to recover the device from userspace and bring 400it back to usable state. Different drivers may have different ideas of a 401'wedged' device depending on hardware implementation of the underlying physical 402device, and hence the vendor agnostic nature of the event. It is up to the 403drivers to decide when they see the need for device recovery and how they want 404to recover from the available methods. 405 406Driver prerequisites 407-------------------- 408 409The driver, before opting for recovery, needs to make sure that the 'wedged' 410device doesn't harm the system as a whole by taking care of the prerequisites. 411Necessary actions must include disabling DMA to system memory as well as any 412communication channels with other devices. Further, the driver must ensure 413that all dma_fences are signalled and any device state that the core kernel 414might depend on is cleaned up. All existing mmaps should be invalidated and 415page faults should be redirected to a dummy page. Once the event is sent, the 416device must be kept in 'wedged' state until the recovery is performed. New 417accesses to the device (IOCTLs) should be rejected, preferably with an error 418code that resembles the type of failure the device has encountered. This will 419signify the reason for wedging, which can be reported to the application if 420needed. 421 422Recovery 423-------- 424 425Current implementation defines four recovery methods, out of which, drivers 426can use any one, multiple or none. Method(s) of choice will be sent in the 427uevent environment as ``WEDGED=<method1>[,..,<methodN>]`` in order of less to 428more side-effects. See the section `Vendor Specific Recovery`_ 429for ``WEDGED=vendor-specific``. If driver is unsure about recovery or 430method is unknown, ``WEDGED=unknown`` will be sent instead. 431 432Userspace consumers can parse this event and attempt recovery as per the 433following expectations. 434 435 =============== ======================================== 436 Recovery method Consumer expectations 437 =============== ======================================== 438 none optional telemetry collection 439 rebind unbind + bind driver 440 bus-reset unbind + bus reset/re-enumeration + bind 441 vendor-specific vendor specific recovery method 442 unknown consumer policy 443 =============== ======================================== 444 445No Recovery 446----------- 447 448Here ``WEDGED=none`` signifies that no recovery is expected from the consumer 449but it can still try to gather telemetry information (devcoredump, syslog) for 450debug purpose in order to root cause the hang. This is useful because the first 451hang is usually the most critical one which can result in consequential hangs 452or complete wedging. 453 454Vendor Specific Recovery 455------------------------ 456 457When ``WEDGED=vendor-specific`` is sent, it indicates that the device requires 458a recovery procedure specific to the hardware vendor and is not one of the 459standardized approaches. 460 461``WEDGED=vendor-specific`` may be used to indicate different cases within a 462single vendor driver, each requiring a distinct recovery procedure. 463In such scenarios, the vendor driver must provide comprehensive documentation 464that describes each case, include additional hints to identify specific case and 465outline the corresponding recovery procedure. The documentation includes: 466 467Case - A list of all cases that sends the ``WEDGED=vendor-specific`` recovery method. 468 469Hints - Additional Information to assist the userspace consumer in identifying and 470differentiating between different cases. This can be exposed through sysfs, debugfs, 471traces, dmesg etc. 472 473Recovery Procedure - Clear instructions and guidance for recovering each case. 474This may include userspace scripts, tools needed for the recovery procedure. 475 476It is the responsibility of the admin/userspace consumer to identify the case and 477verify additional identification hints before attempting a recovery procedure. 478 479Example: If the device uses the Xe driver, then userspace consumer should refer to 480:ref:`Xe Device Wedging <xe-device-wedging>` for the detailed documentation. 481 482Task information 483---------------- 484 485The information about which application (if any) was involved in the device 486wedging is useful for userspace if they want to notify the user about what 487happened (e.g. the compositor display a message to the user "The <task name> 488caused a graphical error and the system recovered") or to implement policies 489(e.g. the daemon may "ban" an task that keeps resetting the device). If the task 490information is available, the uevent will display as ``PID=<pid>`` and 491``TASK=<task name>``. Otherwise, ``PID`` and ``TASK`` will not appear in the 492event string. 493 494The reliability of this information is driver and hardware specific, and should 495be taken with a caution regarding it's precision. To have a big picture of what 496really happened, the devcoredump file provides much more detailed information 497about the device state and about the event. 498 499Consumer prerequisites 500---------------------- 501 502It is the responsibility of the consumer to make sure that the device or its 503resources are not in use by any process before attempting recovery. With IOCTLs 504erroring out, all device memory should be unmapped and file descriptors should 505be closed to prevent leaks or undefined behaviour. The idea here is to clear the 506device of all user context beforehand and set the stage for a clean recovery. 507 508For ``WEDGED=vendor-specific`` recovery method, it is the responsibility of the 509consumer to check the driver documentation and the usecase before attempting 510a recovery. 511 512Example - rebind 513---------------- 514 515Udev rule:: 516 517 SUBSYSTEM=="drm", ENV{WEDGED}=="rebind", DEVPATH=="*/drm/card[0-9]", 518 RUN+="/path/to/rebind.sh $env{DEVPATH}" 519 520Recovery script:: 521 522 #!/bin/sh 523 524 DEVPATH=$(readlink -f /sys/$1/device) 525 DEVICE=$(basename $DEVPATH) 526 DRIVER=$(readlink -f $DEVPATH/driver) 527 528 echo -n $DEVICE > $DRIVER/unbind 529 echo -n $DEVICE > $DRIVER/bind 530 531Customization 532------------- 533 534Although basic recovery is possible with a simple script, consumers can define 535custom policies around recovery. For example, if the driver supports multiple 536recovery methods, consumers can opt for the suitable one depending on scenarios 537like repeat offences or vendor specific failures. Consumers can also choose to 538have the device available for debugging or telemetry collection and base their 539recovery decision on the findings. This is useful especially when the driver is 540unsure about recovery or method is unknown. 541 542.. _drm_driver_ioctl: 543 544IOCTL Support on Device Nodes 545============================= 546 547.. kernel-doc:: drivers/gpu/drm/drm_ioctl.c 548 :doc: driver specific ioctls 549 550Recommended IOCTL Return Values 551------------------------------- 552 553In theory a driver's IOCTL callback is only allowed to return very few error 554codes. In practice it's good to abuse a few more. This section documents common 555practice within the DRM subsystem: 556 557ENOENT: 558 Strictly this should only be used when a file doesn't exist e.g. when 559 calling the open() syscall. We reuse that to signal any kind of object 560 lookup failure, e.g. for unknown GEM buffer object handles, unknown KMS 561 object handles and similar cases. 562 563ENOSPC: 564 Some drivers use this to differentiate "out of kernel memory" from "out 565 of VRAM". Sometimes also applies to other limited gpu resources used for 566 rendering (e.g. when you have a special limited compression buffer). 567 Sometimes resource allocation/reservation issues in command submission 568 IOCTLs are also signalled through EDEADLK. 569 570 Simply running out of kernel/system memory is signalled through ENOMEM. 571 572EPERM/EACCES: 573 Returned for an operation that is valid, but needs more privileges. 574 E.g. root-only or much more common, DRM master-only operations return 575 this when called by unpriviledged clients. There's no clear 576 difference between EACCES and EPERM. 577 578ENODEV: 579 The device is not present anymore or is not yet fully initialized. 580 581EOPNOTSUPP: 582 Feature (like PRIME, modesetting, GEM) is not supported by the driver. 583 584ENXIO: 585 Remote failure, either a hardware transaction (like i2c), but also used 586 when the exporting driver of a shared dma-buf or fence doesn't support a 587 feature needed. 588 589EINTR: 590 DRM drivers assume that userspace restarts all IOCTLs. Any DRM IOCTL can 591 return EINTR and in such a case should be restarted with the IOCTL 592 parameters left unchanged. 593 594EIO: 595 The GPU died and couldn't be resurrected through a reset. Modesetting 596 hardware failures are signalled through the "link status" connector 597 property. 598 599EINVAL: 600 Catch-all for anything that is an invalid argument combination which 601 cannot work. 602 603IOCTL also use other error codes like ETIME, EFAULT, EBUSY, ENOTTY but their 604usage is in line with the common meanings. The above list tries to just document 605DRM specific patterns. Note that ENOTTY has the slightly unintuitive meaning of 606"this IOCTL does not exist", and is used exactly as such in DRM. 607 608.. kernel-doc:: include/drm/drm_ioctl.h 609 :internal: 610 611.. kernel-doc:: drivers/gpu/drm/drm_ioctl.c 612 :export: 613 614.. kernel-doc:: drivers/gpu/drm/drm_ioc32.c 615 :export: 616 617Testing and validation 618====================== 619 620Testing Requirements for userspace API 621-------------------------------------- 622 623New cross-driver userspace interface extensions, like new IOCTL, new KMS 624properties, new files in sysfs or anything else that constitutes an API change 625should have driver-agnostic testcases in IGT for that feature, if such a test 626can be reasonably made using IGT for the target hardware. 627 628Validating changes with IGT 629--------------------------- 630 631There's a collection of tests that aims to cover the whole functionality of 632DRM drivers and that can be used to check that changes to DRM drivers or the 633core don't regress existing functionality. This test suite is called IGT and 634its code and instructions to build and run can be found in 635https://gitlab.freedesktop.org/drm/igt-gpu-tools/. 636 637Using VKMS to test DRM API 638-------------------------- 639 640VKMS is a software-only model of a KMS driver that is useful for testing 641and for running compositors. VKMS aims to enable a virtual display without 642the need for a hardware display capability. These characteristics made VKMS 643a perfect tool for validating the DRM core behavior and also support the 644compositor developer. VKMS makes it possible to test DRM functions in a 645virtual machine without display, simplifying the validation of some of the 646core changes. 647 648To Validate changes in DRM API with VKMS, start setting the kernel: make 649sure to enable VKMS module; compile the kernel with the VKMS enabled and 650install it in the target machine. VKMS can be run in a Virtual Machine 651(QEMU, virtme or similar). It's recommended the use of KVM with the minimum 652of 1GB of RAM and four cores. 653 654It's possible to run the IGT-tests in a VM in two ways: 655 656 1. Use IGT inside a VM 657 2. Use IGT from the host machine and write the results in a shared directory. 658 659Following is an example of using a VM with a shared directory with 660the host machine to run igt-tests. This example uses virtme:: 661 662 $ virtme-run --rwdir /path/for/shared_dir --kdir=path/for/kernel/directory --mods=auto 663 664Run the igt-tests in the guest machine. This example runs the 'kms_flip' 665tests:: 666 667 $ /path/for/igt-gpu-tools/scripts/run-tests.sh -p -s -t "kms_flip.*" -v 668 669In this example, instead of building the igt_runner, Piglit is used 670(-p option). It creates an HTML summary of the test results and saves 671them in the folder "igt-gpu-tools/results". It executes only the igt-tests 672matching the -t option. 673 674Display CRC Support 675------------------- 676 677.. kernel-doc:: drivers/gpu/drm/drm_debugfs_crc.c 678 :doc: CRC ABI 679 680.. kernel-doc:: drivers/gpu/drm/drm_debugfs_crc.c 681 :export: 682 683Debugfs Support 684--------------- 685 686.. kernel-doc:: include/drm/drm_debugfs.h 687 :internal: 688 689.. kernel-doc:: drivers/gpu/drm/drm_debugfs.c 690 :export: 691 692Sysfs Support 693============= 694 695.. kernel-doc:: drivers/gpu/drm/drm_sysfs.c 696 :doc: overview 697 698.. kernel-doc:: drivers/gpu/drm/drm_sysfs.c 699 :export: 700 701 702VBlank event handling 703===================== 704 705The DRM core exposes two vertical blank related ioctls: 706 707:c:macro:`DRM_IOCTL_WAIT_VBLANK` 708 This takes a struct drm_wait_vblank structure as its argument, and 709 it is used to block or request a signal when a specified vblank 710 event occurs. 711 712:c:macro:`DRM_IOCTL_MODESET_CTL` 713 This was only used for user-mode-settind drivers around modesetting 714 changes to allow the kernel to update the vblank interrupt after 715 mode setting, since on many devices the vertical blank counter is 716 reset to 0 at some point during modeset. Modern drivers should not 717 call this any more since with kernel mode setting it is a no-op. 718 719Userspace API Structures 720======================== 721 722.. kernel-doc:: include/uapi/drm/drm_mode.h 723 :doc: overview 724 725.. _crtc_index: 726 727CRTC index 728---------- 729 730CRTC's have both an object ID and an index, and they are not the same thing. 731The index is used in cases where a densely packed identifier for a CRTC is 732needed, for instance a bitmask of CRTC's. The member possible_crtcs of struct 733drm_mode_get_plane is an example. 734 735:c:macro:`DRM_IOCTL_MODE_GETRESOURCES` populates a structure with an array of 736CRTC ID's, and the CRTC index is its position in this array. 737 738.. kernel-doc:: include/uapi/drm/drm.h 739 :internal: 740 741.. kernel-doc:: include/uapi/drm/drm_mode.h 742 :internal: 743 744 745dma-buf interoperability 746======================== 747 748Please see Documentation/userspace-api/dma-buf-alloc-exchange.rst for 749information on how dma-buf is integrated and exposed within DRM. 750 751 752Trace events 753============ 754 755See Documentation/trace/tracepoints.rst for information about using 756Linux Kernel Tracepoints. 757In the DRM subsystem, some events are considered stable uAPI to avoid 758breaking tools (e.g.: GPUVis, umr) relying on them. Stable means that fields 759cannot be removed, nor their formatting updated. Adding new fields is 760possible, under the normal uAPI requirements. 761 762Stable uAPI events 763------------------ 764 765From ``drivers/gpu/drm/scheduler/gpu_scheduler_trace.h`` 766 767.. kernel-doc:: drivers/gpu/drm/scheduler/gpu_scheduler_trace.h 768 :doc: uAPI trace events