1=================== 2Userland interfaces 3=================== 4 5The DRM core exports several interfaces to applications, generally 6intended to be used through corresponding libdrm wrapper functions. In 7addition, drivers export device-specific interfaces for use by userspace 8drivers & device-aware applications through ioctls and sysfs files. 9 10External interfaces include: memory mapping, context management, DMA 11operations, AGP management, vblank control, fence management, memory 12management, and output management. 13 14Cover generic ioctls and sysfs layout here. We only need high-level 15info, since man pages should cover the rest. 16 17libdrm Device Lookup 18==================== 19 20.. kernel-doc:: drivers/gpu/drm/drm_ioctl.c 21 :doc: getunique and setversion story 22 23 24.. _drm_primary_node: 25 26Primary Nodes, DRM Master and Authentication 27============================================ 28 29.. kernel-doc:: drivers/gpu/drm/drm_auth.c 30 :doc: master and authentication 31 32.. kernel-doc:: drivers/gpu/drm/drm_auth.c 33 :export: 34 35.. kernel-doc:: include/drm/drm_auth.h 36 :internal: 37 38Open-Source Userspace Requirements 39================================== 40 41The DRM subsystem has stricter requirements than most other kernel subsystems on 42what the userspace side for new uAPI needs to look like. This section here 43explains what exactly those requirements are, and why they exist. 44 45The short summary is that any addition of DRM uAPI requires corresponding 46open-sourced userspace patches, and those patches must be reviewed and ready for 47merging into a suitable and canonical upstream project. 48 49GFX devices (both display and render/GPU side) are really complex bits of 50hardware, with userspace and kernel by necessity having to work together really 51closely. The interfaces, for rendering and modesetting, must be extremely wide 52and flexible, and therefore it is almost always impossible to precisely define 53them for every possible corner case. This in turn makes it really practically 54infeasible to differentiate between behaviour that's required by userspace, and 55which must not be changed to avoid regressions, and behaviour which is only an 56accidental artifact of the current implementation. 57 58Without access to the full source code of all userspace users that means it 59becomes impossible to change the implementation details, since userspace could 60depend upon the accidental behaviour of the current implementation in minute 61details. And debugging such regressions without access to source code is pretty 62much impossible. As a consequence this means: 63 64- The Linux kernel's "no regression" policy holds in practice only for 65 open-source userspace of the DRM subsystem. DRM developers are perfectly fine 66 if closed-source blob drivers in userspace use the same uAPI as the open 67 drivers, but they must do so in the exact same way as the open drivers. 68 Creative (ab)use of the interfaces will, and in the past routinely has, lead 69 to breakage. 70 71- Any new userspace interface must have an open-source implementation as 72 demonstration vehicle. 73 74The other reason for requiring open-source userspace is uAPI review. Since the 75kernel and userspace parts of a GFX stack must work together so closely, code 76review can only assess whether a new interface achieves its goals by looking at 77both sides. Making sure that the interface indeed covers the use-case fully 78leads to a few additional requirements: 79 80- The open-source userspace must not be a toy/test application, but the real 81 thing. Specifically it needs to handle all the usual error and corner cases. 82 These are often the places where new uAPI falls apart and hence essential to 83 assess the fitness of a proposed interface. 84 85- The userspace side must be fully reviewed and tested to the standards of that 86 userspace project. For e.g. mesa this means piglit testcases and review on the 87 mailing list. This is again to ensure that the new interface actually gets the 88 job done. 89 90- The userspace patches must be against the canonical upstream, not some vendor 91 fork. This is to make sure that no one cheats on the review and testing 92 requirements by doing a quick fork. 93 94- The kernel patch can only be merged after all the above requirements are met, 95 but it **must** be merged **before** the userspace patches land. uAPI always flows 96 from the kernel, doing things the other way round risks divergence of the uAPI 97 definitions and header files. 98 99These are fairly steep requirements, but have grown out from years of shared 100pain and experience with uAPI added hastily, and almost always regretted about 101just as fast. GFX devices change really fast, requiring a paradigm shift and 102entire new set of uAPI interfaces every few years at least. Together with the 103Linux kernel's guarantee to keep existing userspace running for 10+ years this 104is already rather painful for the DRM subsystem, with multiple different uAPIs 105for the same thing co-existing. If we add a few more complete mistakes into the 106mix every year it would be entirely unmanageable. 107 108.. _drm_render_node: 109 110Render nodes 111============ 112 113DRM core provides multiple character-devices for user-space to use. 114Depending on which device is opened, user-space can perform a different 115set of operations (mainly ioctls). The primary node is always created 116and called card<num>. Additionally, a currently unused control node, 117called controlD<num> is also created. The primary node provides all 118legacy operations and historically was the only interface used by 119userspace. With KMS, the control node was introduced. However, the 120planned KMS control interface has never been written and so the control 121node stays unused to date. 122 123With the increased use of offscreen renderers and GPGPU applications, 124clients no longer require running compositors or graphics servers to 125make use of a GPU. But the DRM API required unprivileged clients to 126authenticate to a DRM-Master prior to getting GPU access. To avoid this 127step and to grant clients GPU access without authenticating, render 128nodes were introduced. Render nodes solely serve render clients, that 129is, no modesetting or privileged ioctls can be issued on render nodes. 130Only non-global rendering commands are allowed. If a driver supports 131render nodes, it must advertise it via the DRIVER_RENDER DRM driver 132capability. If not supported, the primary node must be used for render 133clients together with the legacy drmAuth authentication procedure. 134 135If a driver advertises render node support, DRM core will create a 136separate render node called renderD<num>. There will be one render node 137per device. No ioctls except PRIME-related ioctls will be allowed on 138this node. Especially GEM_OPEN will be explicitly prohibited. Render 139nodes are designed to avoid the buffer-leaks, which occur if clients 140guess the flink names or mmap offsets on the legacy interface. 141Additionally to this basic interface, drivers must mark their 142driver-dependent render-only ioctls as DRM_RENDER_ALLOW so render 143clients can use them. Driver authors must be careful not to allow any 144privileged ioctls on render nodes. 145 146With render nodes, user-space can now control access to the render node 147via basic file-system access-modes. A running graphics server which 148authenticates clients on the privileged primary/legacy node is no longer 149required. Instead, a client can open the render node and is immediately 150granted GPU access. Communication between clients (or servers) is done 151via PRIME. FLINK from render node to legacy node is not supported. New 152clients must not use the insecure FLINK interface. 153 154Besides dropping all modeset/global ioctls, render nodes also drop the 155DRM-Master concept. There is no reason to associate render clients with 156a DRM-Master as they are independent of any graphics server. Besides, 157they must work without any running master, anyway. Drivers must be able 158to run without a master object if they support render nodes. If, on the 159other hand, a driver requires shared state between clients which is 160visible to user-space and accessible beyond open-file boundaries, they 161cannot support render nodes. 162 163.. _drm_driver_ioctl: 164 165IOCTL Support on Device Nodes 166============================= 167 168.. kernel-doc:: drivers/gpu/drm/drm_ioctl.c 169 :doc: driver specific ioctls 170 171Recommended IOCTL Return Values 172------------------------------- 173 174In theory a driver's IOCTL callback is only allowed to return very few error 175codes. In practice it's good to abuse a few more. This section documents common 176practice within the DRM subsystem: 177 178ENOENT: 179 Strictly this should only be used when a file doesn't exist e.g. when 180 calling the open() syscall. We reuse that to signal any kind of object 181 lookup failure, e.g. for unknown GEM buffer object handles, unknown KMS 182 object handles and similar cases. 183 184ENOSPC: 185 Some drivers use this to differentiate "out of kernel memory" from "out 186 of VRAM". Sometimes also applies to other limited gpu resources used for 187 rendering (e.g. when you have a special limited compression buffer). 188 Sometimes resource allocation/reservation issues in command submission 189 IOCTLs are also signalled through EDEADLK. 190 191 Simply running out of kernel/system memory is signalled through ENOMEM. 192 193EPERM/EACCESS: 194 Returned for an operation that is valid, but needs more privileges. 195 E.g. root-only or much more common, DRM master-only operations return 196 this when when called by unpriviledged clients. There's no clear 197 difference between EACCESS and EPERM. 198 199ENODEV: 200 Feature (like PRIME, modesetting, GEM) is not supported by the driver. 201 202ENXIO: 203 Remote failure, either a hardware transaction (like i2c), but also used 204 when the exporting driver of a shared dma-buf or fence doesn't support a 205 feature needed. 206 207EINTR: 208 DRM drivers assume that userspace restarts all IOCTLs. Any DRM IOCTL can 209 return EINTR and in such a case should be restarted with the IOCTL 210 parameters left unchanged. 211 212EIO: 213 The GPU died and couldn't be resurrected through a reset. Modesetting 214 hardware failures are signalled through the "link status" connector 215 property. 216 217EINVAL: 218 Catch-all for anything that is an invalid argument combination which 219 cannot work. 220 221IOCTL also use other error codes like ETIME, EFAULT, EBUSY, ENOTTY but their 222usage is in line with the common meanings. The above list tries to just document 223DRM specific patterns. Note that ENOTTY has the slightly unintuitive meaning of 224"this IOCTL does not exist", and is used exactly as such in DRM. 225 226.. kernel-doc:: include/drm/drm_ioctl.h 227 :internal: 228 229.. kernel-doc:: drivers/gpu/drm/drm_ioctl.c 230 :export: 231 232.. kernel-doc:: drivers/gpu/drm/drm_ioc32.c 233 :export: 234 235Testing and validation 236====================== 237 238Validating changes with IGT 239--------------------------- 240 241There's a collection of tests that aims to cover the whole functionality of 242DRM drivers and that can be used to check that changes to DRM drivers or the 243core don't regress existing functionality. This test suite is called IGT and 244its code can be found in https://cgit.freedesktop.org/drm/igt-gpu-tools/. 245 246To build IGT, start by installing its build dependencies. In Debian-based 247systems:: 248 249 # apt-get build-dep intel-gpu-tools 250 251And in Fedora-based systems:: 252 253 # dnf builddep intel-gpu-tools 254 255Then clone the repository:: 256 257 $ git clone git://anongit.freedesktop.org/drm/igt-gpu-tools 258 259Configure the build system and start the build:: 260 261 $ cd igt-gpu-tools && ./autogen.sh && make -j6 262 263Download the piglit dependency:: 264 265 $ ./scripts/run-tests.sh -d 266 267And run the tests:: 268 269 $ ./scripts/run-tests.sh -t kms -t core -s 270 271run-tests.sh is a wrapper around piglit that will execute the tests matching 272the -t options. A report in HTML format will be available in 273./results/html/index.html. Results can be compared with piglit. 274 275Display CRC Support 276------------------- 277 278.. kernel-doc:: drivers/gpu/drm/drm_debugfs_crc.c 279 :doc: CRC ABI 280 281.. kernel-doc:: drivers/gpu/drm/drm_debugfs_crc.c 282 :export: 283 284Debugfs Support 285--------------- 286 287.. kernel-doc:: include/drm/drm_debugfs.h 288 :internal: 289 290.. kernel-doc:: drivers/gpu/drm/drm_debugfs.c 291 :export: 292 293Sysfs Support 294============= 295 296.. kernel-doc:: drivers/gpu/drm/drm_sysfs.c 297 :doc: overview 298 299.. kernel-doc:: drivers/gpu/drm/drm_sysfs.c 300 :export: 301 302 303VBlank event handling 304===================== 305 306The DRM core exposes two vertical blank related ioctls: 307 308DRM_IOCTL_WAIT_VBLANK 309 This takes a struct drm_wait_vblank structure as its argument, and 310 it is used to block or request a signal when a specified vblank 311 event occurs. 312 313DRM_IOCTL_MODESET_CTL 314 This was only used for user-mode-settind drivers around modesetting 315 changes to allow the kernel to update the vblank interrupt after 316 mode setting, since on many devices the vertical blank counter is 317 reset to 0 at some point during modeset. Modern drivers should not 318 call this any more since with kernel mode setting it is a no-op. 319