Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 1 | ================================================= |
| 2 | FPGA Device Feature List (DFL) Framework Overview |
| 3 | ================================================= |
| 4 | |
| 5 | Authors: |
| 6 | |
| 7 | - Enno Luebbers <enno.luebbers@intel.com> |
| 8 | - Xiao Guangrong <guangrong.xiao@linux.intel.com> |
| 9 | - Wu Hao <hao.wu@intel.com> |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 10 | |
| 11 | The Device Feature List (DFL) FPGA framework (and drivers according to this |
| 12 | this framework) hides the very details of low layer hardwares and provides |
| 13 | unified interfaces to userspace. Applications could use these interfaces to |
| 14 | configure, enumerate, open and access FPGA accelerators on platforms which |
| 15 | implement the DFL in the device memory. Besides this, the DFL framework |
| 16 | enables system level management functions such as FPGA reconfiguration. |
| 17 | |
| 18 | |
| 19 | Device Feature List (DFL) Overview |
| 20 | ================================== |
| 21 | Device Feature List (DFL) defines a linked list of feature headers within the |
| 22 | device MMIO space to provide an extensible way of adding features. Software can |
| 23 | walk through these predefined data structures to enumerate FPGA features: |
| 24 | FPGA Interface Unit (FIU), Accelerated Function Unit (AFU) and Private Features, |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 25 | as illustrated below:: |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 26 | |
| 27 | Header Header Header Header |
| 28 | +----------+ +-->+----------+ +-->+----------+ +-->+----------+ |
| 29 | | Type | | | Type | | | Type | | | Type | |
| 30 | | FIU | | | Private | | | Private | | | Private | |
| 31 | +----------+ | | Feature | | | Feature | | | Feature | |
| 32 | | Next_DFH |--+ +----------+ | +----------+ | +----------+ |
| 33 | +----------+ | Next_DFH |--+ | Next_DFH |--+ | Next_DFH |--> NULL |
| 34 | | ID | +----------+ +----------+ +----------+ |
| 35 | +----------+ | ID | | ID | | ID | |
| 36 | | Next_AFU |--+ +----------+ +----------+ +----------+ |
| 37 | +----------+ | | Feature | | Feature | | Feature | |
| 38 | | Header | | | Register | | Register | | Register | |
| 39 | | Register | | | Set | | Set | | Set | |
| 40 | | Set | | +----------+ +----------+ +----------+ |
| 41 | +----------+ | Header |
| 42 | +-->+----------+ |
| 43 | | Type | |
| 44 | | AFU | |
| 45 | +----------+ |
| 46 | | Next_DFH |--> NULL |
| 47 | +----------+ |
| 48 | | GUID | |
| 49 | +----------+ |
| 50 | | Header | |
| 51 | | Register | |
| 52 | | Set | |
| 53 | +----------+ |
| 54 | |
| 55 | FPGA Interface Unit (FIU) represents a standalone functional unit for the |
| 56 | interface to FPGA, e.g. the FPGA Management Engine (FME) and Port (more |
| 57 | descriptions on FME and Port in later sections). |
| 58 | |
| 59 | Accelerated Function Unit (AFU) represents a FPGA programmable region and |
| 60 | always connects to a FIU (e.g. a Port) as its child as illustrated above. |
| 61 | |
| 62 | Private Features represent sub features of the FIU and AFU. They could be |
| 63 | various function blocks with different IDs, but all private features which |
| 64 | belong to the same FIU or AFU, must be linked to one list via the Next Device |
| 65 | Feature Header (Next_DFH) pointer. |
| 66 | |
| 67 | Each FIU, AFU and Private Feature could implement its own functional registers. |
| 68 | The functional register set for FIU and AFU, is named as Header Register Set, |
| 69 | e.g. FME Header Register Set, and the one for Private Feature, is named as |
| 70 | Feature Register Set, e.g. FME Partial Reconfiguration Feature Register Set. |
| 71 | |
| 72 | This Device Feature List provides a way of linking features together, it's |
| 73 | convenient for software to locate each feature by walking through this list, |
| 74 | and can be implemented in register regions of any FPGA device. |
| 75 | |
| 76 | |
| 77 | FIU - FME (FPGA Management Engine) |
| 78 | ================================== |
| 79 | The FPGA Management Engine performs reconfiguration and other infrastructure |
| 80 | functions. Each FPGA device only has one FME. |
| 81 | |
| 82 | User-space applications can acquire exclusive access to the FME using open(), |
| 83 | and release it using close(). |
| 84 | |
| 85 | The following functions are exposed through ioctls: |
| 86 | |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 87 | - Get driver API version (DFL_FPGA_GET_API_VERSION) |
| 88 | - Check for extensions (DFL_FPGA_CHECK_EXTENSION) |
| 89 | - Program bitstream (DFL_FPGA_FME_PORT_PR) |
Wu Hao | 77a0ef4 | 2019-08-12 10:50:04 +0800 | [diff] [blame] | 90 | - Assign port to PF (DFL_FPGA_FME_PORT_ASSIGN) |
| 91 | - Release port from PF (DFL_FPGA_FME_PORT_RELEASE) |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 92 | |
| 93 | More functions are exposed through sysfs |
| 94 | (/sys/class/fpga_region/regionX/dfl-fme.n/): |
| 95 | |
| 96 | Read bitstream ID (bitstream_id) |
| 97 | bitstream_id indicates version of the static FPGA region. |
| 98 | |
| 99 | Read bitstream metadata (bitstream_metadata) |
| 100 | bitstream_metadata includes detailed information of static FPGA region, |
| 101 | e.g. synthesis date and seed. |
| 102 | |
| 103 | Read number of ports (ports_num) |
| 104 | one FPGA device may have more than one port, this sysfs interface indicates |
| 105 | how many ports the FPGA device has. |
| 106 | |
Wu Hao | 77a0ef4 | 2019-08-12 10:50:04 +0800 | [diff] [blame] | 107 | Global error reporting management (errors/) |
| 108 | error reporting sysfs interfaces allow user to read errors detected by the |
| 109 | hardware, and clear the logged errors. |
| 110 | |
Xu Yilun | 214be74 | 2019-10-14 13:42:01 +0800 | [diff] [blame] | 111 | Power management (dfl_fme_power hwmon) |
| 112 | power management hwmon sysfs interfaces allow user to read power management |
| 113 | information (power consumption, thresholds, threshold status, limits, etc.) |
| 114 | and configure power thresholds for different throttling levels. |
| 115 | |
| 116 | Thermal management (dfl_fme_thermal hwmon) |
| 117 | thermal management hwmon sysfs interfaces allow user to read thermal |
| 118 | management information (current temperature, thresholds, threshold status, |
| 119 | etc.). |
| 120 | |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 121 | |
| 122 | FIU - PORT |
| 123 | ========== |
| 124 | A port represents the interface between the static FPGA fabric and a partially |
| 125 | reconfigurable region containing an AFU. It controls the communication from SW |
| 126 | to the accelerator and exposes features such as reset and debug. Each FPGA |
| 127 | device may have more than one port, but always one AFU per port. |
| 128 | |
| 129 | |
| 130 | AFU |
| 131 | === |
| 132 | An AFU is attached to a port FIU and exposes a fixed length MMIO region to be |
| 133 | used for accelerator-specific control registers. |
| 134 | |
| 135 | User-space applications can acquire exclusive access to an AFU attached to a |
| 136 | port by using open() on the port device node and release it using close(). |
| 137 | |
| 138 | The following functions are exposed through ioctls: |
| 139 | |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 140 | - Get driver API version (DFL_FPGA_GET_API_VERSION) |
| 141 | - Check for extensions (DFL_FPGA_CHECK_EXTENSION) |
| 142 | - Get port info (DFL_FPGA_PORT_GET_INFO) |
| 143 | - Get MMIO region info (DFL_FPGA_PORT_GET_REGION_INFO) |
| 144 | - Map DMA buffer (DFL_FPGA_PORT_DMA_MAP) |
| 145 | - Unmap DMA buffer (DFL_FPGA_PORT_DMA_UNMAP) |
| 146 | - Reset AFU (DFL_FPGA_PORT_RESET) |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 147 | |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 148 | DFL_FPGA_PORT_RESET: |
| 149 | reset the FPGA Port and its AFU. Userspace can do Port |
| 150 | reset at any time, e.g. during DMA or Partial Reconfiguration. But it should |
| 151 | never cause any system level issue, only functional failure (e.g. DMA or PR |
| 152 | operation failure) and be recoverable from the failure. |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 153 | |
| 154 | User-space applications can also mmap() accelerator MMIO regions. |
| 155 | |
| 156 | More functions are exposed through sysfs: |
| 157 | (/sys/class/fpga_region/<regionX>/<dfl-port.m>/): |
| 158 | |
| 159 | Read Accelerator GUID (afu_id) |
| 160 | afu_id indicates which PR bitstream is programmed to this AFU. |
| 161 | |
Wu Hao | 77a0ef4 | 2019-08-12 10:50:04 +0800 | [diff] [blame] | 162 | Error reporting (errors/) |
| 163 | error reporting sysfs interfaces allow user to read port/afu errors |
| 164 | detected by the hardware, and clear the logged errors. |
| 165 | |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 166 | |
| 167 | DFL Framework Overview |
| 168 | ====================== |
| 169 | |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 170 | :: |
| 171 | |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 172 | +----------+ +--------+ +--------+ +--------+ |
| 173 | | FME | | AFU | | AFU | | AFU | |
| 174 | | Module | | Module | | Module | | Module | |
| 175 | +----------+ +--------+ +--------+ +--------+ |
| 176 | +-----------------------+ |
| 177 | | FPGA Container Device | Device Feature List |
| 178 | | (FPGA Base Region) | Framework |
| 179 | +-----------------------+ |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 180 | ------------------------------------------------------------------ |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 181 | +----------------------------+ |
| 182 | | FPGA DFL Device Module | |
| 183 | | (e.g. PCIE/Platform Device)| |
| 184 | +----------------------------+ |
| 185 | +------------------------+ |
| 186 | | FPGA Hardware Device | |
| 187 | +------------------------+ |
| 188 | |
| 189 | DFL framework in kernel provides common interfaces to create container device |
| 190 | (FPGA base region), discover feature devices and their private features from the |
| 191 | given Device Feature Lists and create platform devices for feature devices |
| 192 | (e.g. FME, Port and AFU) with related resources under the container device. It |
| 193 | also abstracts operations for the private features and exposes common ops to |
| 194 | feature device drivers. |
| 195 | |
| 196 | The FPGA DFL Device could be different hardwares, e.g. PCIe device, platform |
| 197 | device and etc. Its driver module is always loaded first once the device is |
| 198 | created by the system. This driver plays an infrastructural role in the |
| 199 | driver architecture. It locates the DFLs in the device memory, handles them |
| 200 | and related resources to common interfaces from DFL framework for enumeration. |
| 201 | (Please refer to drivers/fpga/dfl.c for detailed enumeration APIs). |
| 202 | |
| 203 | The FPGA Management Engine (FME) driver is a platform driver which is loaded |
| 204 | automatically after FME platform device creation from the DFL device module. It |
| 205 | provides the key features for FPGA management, including: |
| 206 | |
| 207 | a) Expose static FPGA region information, e.g. version and metadata. |
| 208 | Users can read related information via sysfs interfaces exposed |
| 209 | by FME driver. |
| 210 | |
| 211 | b) Partial Reconfiguration. The FME driver creates FPGA manager, FPGA |
| 212 | bridges and FPGA regions during PR sub feature initialization. Once |
| 213 | it receives a DFL_FPGA_FME_PORT_PR ioctl from user, it invokes the |
| 214 | common interface function from FPGA Region to complete the partial |
| 215 | reconfiguration of the PR bitstream to the given port. |
| 216 | |
| 217 | Similar to the FME driver, the FPGA Accelerated Function Unit (AFU) driver is |
| 218 | probed once the AFU platform device is created. The main function of this module |
| 219 | is to provide an interface for userspace applications to access the individual |
| 220 | accelerators, including basic reset control on port, AFU MMIO region export, dma |
| 221 | buffer mapping service functions. |
| 222 | |
| 223 | After feature platform devices creation, matched platform drivers will be loaded |
| 224 | automatically to handle different functionalities. Please refer to next sections |
| 225 | for detailed information on functional units which have been already implemented |
| 226 | under this DFL framework. |
| 227 | |
| 228 | |
| 229 | Partial Reconfiguration |
| 230 | ======================= |
| 231 | As mentioned above, accelerators can be reconfigured through partial |
| 232 | reconfiguration of a PR bitstream file. The PR bitstream file must have been |
| 233 | generated for the exact static FPGA region and targeted reconfigurable region |
| 234 | (port) of the FPGA, otherwise, the reconfiguration operation will fail and |
| 235 | possibly cause system instability. This compatibility can be checked by |
| 236 | comparing the compatibility ID noted in the header of PR bitstream file against |
| 237 | the compat_id exposed by the target FPGA region. This check is usually done by |
| 238 | userspace before calling the reconfiguration IOCTL. |
| 239 | |
| 240 | |
Wu Hao | 77a0ef4 | 2019-08-12 10:50:04 +0800 | [diff] [blame] | 241 | FPGA virtualization - PCIe SRIOV |
| 242 | ================================ |
| 243 | This section describes the virtualization support on DFL based FPGA device to |
| 244 | enable accessing an accelerator from applications running in a virtual machine |
| 245 | (VM). This section only describes the PCIe based FPGA device with SRIOV support. |
| 246 | |
| 247 | Features supported by the particular FPGA device are exposed through Device |
| 248 | Feature Lists, as illustrated below: |
| 249 | |
| 250 | :: |
| 251 | |
| 252 | +-------------------------------+ +-------------+ |
| 253 | | PF | | VF | |
| 254 | +-------------------------------+ +-------------+ |
| 255 | ^ ^ ^ ^ |
| 256 | | | | | |
| 257 | +-----|------------|---------|--------------|-------+ |
| 258 | | | | | | | |
| 259 | | +-----+ +-------+ +-------+ +-------+ | |
| 260 | | | FME | | Port0 | | Port1 | | Port2 | | |
| 261 | | +-----+ +-------+ +-------+ +-------+ | |
| 262 | | ^ ^ ^ | |
| 263 | | | | | | |
| 264 | | +-------+ +------+ +-------+ | |
| 265 | | | AFU | | AFU | | AFU | | |
| 266 | | +-------+ +------+ +-------+ | |
| 267 | | | |
| 268 | | DFL based FPGA PCIe Device | |
| 269 | +---------------------------------------------------+ |
| 270 | |
| 271 | FME is always accessed through the physical function (PF). |
| 272 | |
| 273 | Ports (and related AFUs) are accessed via PF by default, but could be exposed |
| 274 | through virtual function (VF) devices via PCIe SRIOV. Each VF only contains |
| 275 | 1 Port and 1 AFU for isolation. Users could assign individual VFs (accelerators) |
| 276 | created via PCIe SRIOV interface, to virtual machines. |
| 277 | |
| 278 | The driver organization in virtualization case is illustrated below: |
| 279 | :: |
| 280 | |
| 281 | +-------++------++------+ | |
| 282 | | FME || FME || FME | | |
| 283 | | FPGA || FPGA || FPGA | | |
| 284 | |Manager||Bridge||Region| | |
| 285 | +-------++------++------+ | |
| 286 | +-----------------------+ +--------+ | +--------+ |
| 287 | | FME | | AFU | | | AFU | |
| 288 | | Module | | Module | | | Module | |
| 289 | +-----------------------+ +--------+ | +--------+ |
| 290 | +-----------------------+ | +-----------------------+ |
| 291 | | FPGA Container Device | | | FPGA Container Device | |
| 292 | | (FPGA Base Region) | | | (FPGA Base Region) | |
| 293 | +-----------------------+ | +-----------------------+ |
| 294 | +------------------+ | +------------------+ |
| 295 | | FPGA PCIE Module | | Virtual | FPGA PCIE Module | |
| 296 | +------------------+ Host | Machine +------------------+ |
| 297 | -------------------------------------- | ------------------------------ |
| 298 | +---------------+ | +---------------+ |
| 299 | | PCI PF Device | | | PCI VF Device | |
| 300 | +---------------+ | +---------------+ |
| 301 | |
| 302 | FPGA PCIe device driver is always loaded first once a FPGA PCIe PF or VF device |
| 303 | is detected. It: |
| 304 | |
| 305 | * Finishes enumeration on both FPGA PCIe PF and VF device using common |
| 306 | interfaces from DFL framework. |
| 307 | * Supports SRIOV. |
| 308 | |
| 309 | The FME device driver plays a management role in this driver architecture, it |
| 310 | provides ioctls to release Port from PF and assign Port to PF. After release |
| 311 | a port from PF, then it's safe to expose this port through a VF via PCIe SRIOV |
| 312 | sysfs interface. |
| 313 | |
| 314 | To enable accessing an accelerator from applications running in a VM, the |
| 315 | respective AFU's port needs to be assigned to a VF using the following steps: |
| 316 | |
| 317 | #. The PF owns all AFU ports by default. Any port that needs to be |
| 318 | reassigned to a VF must first be released through the |
| 319 | DFL_FPGA_FME_PORT_RELEASE ioctl on the FME device. |
| 320 | |
| 321 | #. Once N ports are released from PF, then user can use command below |
| 322 | to enable SRIOV and VFs. Each VF owns only one Port with AFU. |
| 323 | |
| 324 | :: |
| 325 | |
| 326 | echo N > $PCI_DEVICE_PATH/sriov_numvfs |
| 327 | |
| 328 | #. Pass through the VFs to VMs |
| 329 | |
| 330 | #. The AFU under VF is accessible from applications in VM (using the |
| 331 | same driver inside the VF). |
| 332 | |
| 333 | Note that an FME can't be assigned to a VF, thus PR and other management |
| 334 | functions are only available via the PF. |
| 335 | |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 336 | Device enumeration |
| 337 | ================== |
| 338 | This section introduces how applications enumerate the fpga device from |
| 339 | the sysfs hierarchy under /sys/class/fpga_region. |
| 340 | |
| 341 | In the example below, two DFL based FPGA devices are installed in the host. Each |
| 342 | fpga device has one FME and two ports (AFUs). |
| 343 | |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 344 | FPGA regions are created under /sys/class/fpga_region/:: |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 345 | |
| 346 | /sys/class/fpga_region/region0 |
| 347 | /sys/class/fpga_region/region1 |
| 348 | /sys/class/fpga_region/region2 |
| 349 | ... |
| 350 | |
| 351 | Application needs to search each regionX folder, if feature device is found, |
| 352 | (e.g. "dfl-port.n" or "dfl-fme.m" is found), then it's the base |
| 353 | fpga region which represents the FPGA device. |
| 354 | |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 355 | Each base region has one FME and two ports (AFUs) as child devices:: |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 356 | |
| 357 | /sys/class/fpga_region/region0/dfl-fme.0 |
| 358 | /sys/class/fpga_region/region0/dfl-port.0 |
| 359 | /sys/class/fpga_region/region0/dfl-port.1 |
| 360 | ... |
| 361 | |
| 362 | /sys/class/fpga_region/region3/dfl-fme.1 |
| 363 | /sys/class/fpga_region/region3/dfl-port.2 |
| 364 | /sys/class/fpga_region/region3/dfl-port.3 |
| 365 | ... |
| 366 | |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 367 | In general, the FME/AFU sysfs interfaces are named as follows:: |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 368 | |
| 369 | /sys/class/fpga_region/<regionX>/<dfl-fme.n>/ |
| 370 | /sys/class/fpga_region/<regionX>/<dfl-port.m>/ |
| 371 | |
| 372 | with 'n' consecutively numbering all FMEs and 'm' consecutively numbering all |
| 373 | ports. |
| 374 | |
Mauro Carvalho Chehab | c220a1f | 2019-06-12 14:52:46 -0300 | [diff] [blame] | 375 | The device nodes used for ioctl() or mmap() can be referenced through:: |
Wu Hao | c73c9ad | 2018-06-30 08:53:08 +0800 | [diff] [blame] | 376 | |
| 377 | /sys/class/fpga_region/<regionX>/<dfl-fme.n>/dev |
| 378 | /sys/class/fpga_region/<regionX>/<dfl-port.n>/dev |
| 379 | |
| 380 | |
| 381 | Add new FIUs support |
| 382 | ==================== |
| 383 | It's possible that developers made some new function blocks (FIUs) under this |
| 384 | DFL framework, then new platform device driver needs to be developed for the |
| 385 | new feature dev (FIU) following the same way as existing feature dev drivers |
| 386 | (e.g. FME and Port/AFU platform device driver). Besides that, it requires |
| 387 | modification on DFL framework enumeration code too, for new FIU type detection |
| 388 | and related platform devices creation. |
| 389 | |
| 390 | |
| 391 | Add new private features support |
| 392 | ================================ |
| 393 | In some cases, we may need to add some new private features to existing FIUs |
| 394 | (e.g. FME or Port). Developers don't need to touch enumeration code in DFL |
| 395 | framework, as each private feature will be parsed automatically and related |
| 396 | mmio resources can be found under FIU platform device created by DFL framework. |
| 397 | Developer only needs to provide a sub feature driver with matched feature id. |
| 398 | FME Partial Reconfiguration Sub Feature driver (see drivers/fpga/dfl-fme-pr.c) |
| 399 | could be a reference. |
| 400 | |
| 401 | |
| 402 | Open discussion |
| 403 | =============== |
| 404 | FME driver exports one ioctl (DFL_FPGA_FME_PORT_PR) for partial reconfiguration |
| 405 | to user now. In the future, if unified user interfaces for reconfiguration are |
| 406 | added, FME driver should switch to them from ioctl interface. |