Based on kernel version 4.16.1. Page generated on 2018-04-09 11:52 EST.
1 ACPI on ARMv8 Servers 2 --------------------- 3 ACPI can be used for ARMv8 general purpose servers designed to follow 4 the ARM SBSA (Server Base System Architecture) [0] and SBBR (Server 5 Base Boot Requirements) [1] specifications. Please note that the SBBR 6 can be retrieved simply by visiting [1], but the SBSA is currently only 7 available to those with an ARM login due to ARM IP licensing concerns. 8 9 The ARMv8 kernel implements the reduced hardware model of ACPI version 10 5.1 or later. Links to the specification and all external documents 11 it refers to are managed by the UEFI Forum. The specification is 12 available at http://www.uefi.org/specifications and documents referenced 13 by the specification can be found via http://www.uefi.org/acpi. 14 15 If an ARMv8 system does not meet the requirements of the SBSA and SBBR, 16 or cannot be described using the mechanisms defined in the required ACPI 17 specifications, then ACPI may not be a good fit for the hardware. 18 19 While the documents mentioned above set out the requirements for building 20 industry-standard ARMv8 servers, they also apply to more than one operating 21 system. The purpose of this document is to describe the interaction between 22 ACPI and Linux only, on an ARMv8 system -- that is, what Linux expects of 23 ACPI and what ACPI can expect of Linux. 24 25 26 Why ACPI on ARM? 27 ---------------- 28 Before examining the details of the interface between ACPI and Linux, it is 29 useful to understand why ACPI is being used. Several technologies already 30 exist in Linux for describing non-enumerable hardware, after all. In this 31 section we summarize a blog post [2] from Grant Likely that outlines the 32 reasoning behind ACPI on ARMv8 servers. Actually, we snitch a good portion 33 of the summary text almost directly, to be honest. 34 35 The short form of the rationale for ACPI on ARM is: 36 37 -- ACPI’s byte code (AML) allows the platform to encode hardware behavior, 38 while DT explicitly does not support this. For hardware vendors, being 39 able to encode behavior is a key tool used in supporting operating 40 system releases on new hardware. 41 42 -- ACPI’s OSPM defines a power management model that constrains what the 43 platform is allowed to do into a specific model, while still providing 44 flexibility in hardware design. 45 46 -- In the enterprise server environment, ACPI has established bindings (such 47 as for RAS) which are currently used in production systems. DT does not. 48 Such bindings could be defined in DT at some point, but doing so means ARM 49 and x86 would end up using completely different code paths in both firmware 50 and the kernel. 51 52 -- Choosing a single interface to describe the abstraction between a platform 53 and an OS is important. Hardware vendors would not be required to implement 54 both DT and ACPI if they want to support multiple operating systems. And, 55 agreeing on a single interface instead of being fragmented into per OS 56 interfaces makes for better interoperability overall. 57 58 -- The new ACPI governance process works well and Linux is now at the same 59 table as hardware vendors and other OS vendors. In fact, there is no 60 longer any reason to feel that ACPI only belongs to Windows or that 61 Linux is in any way secondary to Microsoft in this arena. The move of 62 ACPI governance into the UEFI forum has significantly opened up the 63 specification development process, and currently, a large portion of the 64 changes being made to ACPI are being driven by Linux. 65 66 Key to the use of ACPI is the support model. For servers in general, the 67 responsibility for hardware behaviour cannot solely be the domain of the 68 kernel, but rather must be split between the platform and the kernel, in 69 order to allow for orderly change over time. ACPI frees the OS from needing 70 to understand all the minute details of the hardware so that the OS doesn’t 71 need to be ported to each and every device individually. It allows the 72 hardware vendors to take responsibility for power management behaviour without 73 depending on an OS release cycle which is not under their control. 74 75 ACPI is also important because hardware and OS vendors have already worked 76 out the mechanisms for supporting a general purpose computing ecosystem. The 77 infrastructure is in place, the bindings are in place, and the processes are 78 in place. DT does exactly what Linux needs it to when working with vertically 79 integrated devices, but there are no good processes for supporting what the 80 server vendors need. Linux could potentially get there with DT, but doing so 81 really just duplicates something that already works. ACPI already does what 82 the hardware vendors need, Microsoft won’t collaborate on DT, and hardware 83 vendors would still end up providing two completely separate firmware 84 interfaces -- one for Linux and one for Windows. 85 86 87 Kernel Compatibility 88 -------------------- 89 One of the primary motivations for ACPI is standardization, and using that 90 to provide backward compatibility for Linux kernels. In the server market, 91 software and hardware are often used for long periods. ACPI allows the 92 kernel and firmware to agree on a consistent abstraction that can be 93 maintained over time, even as hardware or software change. As long as the 94 abstraction is supported, systems can be updated without necessarily having 95 to replace the kernel. 96 97 When a Linux driver or subsystem is first implemented using ACPI, it by 98 definition ends up requiring a specific version of the ACPI specification 99 -- it's baseline. ACPI firmware must continue to work, even though it may 100 not be optimal, with the earliest kernel version that first provides support 101 for that baseline version of ACPI. There may be a need for additional drivers, 102 but adding new functionality (e.g., CPU power management) should not break 103 older kernel versions. Further, ACPI firmware must also work with the most 104 recent version of the kernel. 105 106 107 Relationship with Device Tree 108 ----------------------------- 109 ACPI support in drivers and subsystems for ARMv8 should never be mutually 110 exclusive with DT support at compile time. 111 112 At boot time the kernel will only use one description method depending on 113 parameters passed from the boot loader (including kernel bootargs). 114 115 Regardless of whether DT or ACPI is used, the kernel must always be capable 116 of booting with either scheme (in kernels with both schemes enabled at compile 117 time). 118 119 120 Booting using ACPI tables 121 ------------------------- 122 The only defined method for passing ACPI tables to the kernel on ARMv8 123 is via the UEFI system configuration table. Just so it is explicit, this 124 means that ACPI is only supported on platforms that boot via UEFI. 125 126 When an ARMv8 system boots, it can either have DT information, ACPI tables, 127 or in some very unusual cases, both. If no command line parameters are used, 128 the kernel will try to use DT for device enumeration; if there is no DT 129 present, the kernel will try to use ACPI tables, but only if they are present. 130 In neither is available, the kernel will not boot. If acpi=force is used 131 on the command line, the kernel will attempt to use ACPI tables first, but 132 fall back to DT if there are no ACPI tables present. The basic idea is that 133 the kernel will not fail to boot unless it absolutely has no other choice. 134 135 Processing of ACPI tables may be disabled by passing acpi=off on the kernel 136 command line; this is the default behavior. 137 138 In order for the kernel to load and use ACPI tables, the UEFI implementation 139 MUST set the ACPI_20_TABLE_GUID to point to the RSDP table (the table with 140 the ACPI signature "RSD PTR "). If this pointer is incorrect and acpi=force 141 is used, the kernel will disable ACPI and try to use DT to boot instead; the 142 kernel has, in effect, determined that ACPI tables are not present at that 143 point. 144 145 If the pointer to the RSDP table is correct, the table will be mapped into 146 the kernel by the ACPI core, using the address provided by UEFI. 147 148 The ACPI core will then locate and map in all other ACPI tables provided by 149 using the addresses in the RSDP table to find the XSDT (eXtended System 150 Description Table). The XSDT in turn provides the addresses to all other 151 ACPI tables provided by the system firmware; the ACPI core will then traverse 152 this table and map in the tables listed. 153 154 The ACPI core will ignore any provided RSDT (Root System Description Table). 155 RSDTs have been deprecated and are ignored on arm64 since they only allow 156 for 32-bit addresses. 157 158 Further, the ACPI core will only use the 64-bit address fields in the FADT 159 (Fixed ACPI Description Table). Any 32-bit address fields in the FADT will 160 be ignored on arm64. 161 162 Hardware reduced mode (see Section 4.1 of the ACPI 6.1 specification) will 163 be enforced by the ACPI core on arm64. Doing so allows the ACPI core to 164 run less complex code since it no longer has to provide support for legacy 165 hardware from other architectures. Any fields that are not to be used for 166 hardware reduced mode must be set to zero. 167 168 For the ACPI core to operate properly, and in turn provide the information 169 the kernel needs to configure devices, it expects to find the following 170 tables (all section numbers refer to the ACPI 6.1 specification): 171 172 -- RSDP (Root System Description Pointer), section 5.2.5 173 174 -- XSDT (eXtended System Description Table), section 5.2.8 175 176 -- FADT (Fixed ACPI Description Table), section 5.2.9 177 178 -- DSDT (Differentiated System Description Table), section 179 5.2.11.1 180 181 -- MADT (Multiple APIC Description Table), section 5.2.12 182 183 -- GTDT (Generic Timer Description Table), section 5.2.24 184 185 -- If PCI is supported, the MCFG (Memory mapped ConFiGuration 186 Table), section 5.2.6, specifically Table 5-31. 187 188 -- If booting without a console=<device> kernel parameter is 189 supported, the SPCR (Serial Port Console Redirection table), 190 section 5.2.6, specifically Table 5-31. 191 192 -- If necessary to describe the I/O topology, SMMUs and GIC ITSs, 193 the IORT (Input Output Remapping Table, section 5.2.6, specifically 194 Table 5-31). 195 196 -- If NUMA is supported, the SRAT (System Resource Affinity Table) 197 and SLIT (System Locality distance Information Table), sections 198 5.2.16 and 5.2.17, respectively. 199 200 If the above tables are not all present, the kernel may or may not be 201 able to boot properly since it may not be able to configure all of the 202 devices available. This list of tables is not meant to be all inclusive; 203 in some environments other tables may be needed (e.g., any of the APEI 204 tables from section 18) to support specific functionality. 205 206 207 ACPI Detection 208 -------------- 209 Drivers should determine their probe() type by checking for a null 210 value for ACPI_HANDLE, or checking .of_node, or other information in 211 the device structure. This is detailed further in the "Driver 212 Recommendations" section. 213 214 In non-driver code, if the presence of ACPI needs to be detected at 215 run time, then check the value of acpi_disabled. If CONFIG_ACPI is not 216 set, acpi_disabled will always be 1. 217 218 219 Device Enumeration 220 ------------------ 221 Device descriptions in ACPI should use standard recognized ACPI interfaces. 222 These may contain less information than is typically provided via a Device 223 Tree description for the same device. This is also one of the reasons that 224 ACPI can be useful -- the driver takes into account that it may have less 225 detailed information about the device and uses sensible defaults instead. 226 If done properly in the driver, the hardware can change and improve over 227 time without the driver having to change at all. 228 229 Clocks provide an excellent example. In DT, clocks need to be specified 230 and the drivers need to take them into account. In ACPI, the assumption 231 is that UEFI will leave the device in a reasonable default state, including 232 any clock settings. If for some reason the driver needs to change a clock 233 value, this can be done in an ACPI method; all the driver needs to do is 234 invoke the method and not concern itself with what the method needs to do 235 to change the clock. Changing the hardware can then take place over time 236 by changing what the ACPI method does, and not the driver. 237 238 In DT, the parameters needed by the driver to set up clocks as in the example 239 above are known as "bindings"; in ACPI, these are known as "Device Properties" 240 and provided to a driver via the _DSD object. 241 242 ACPI tables are described with a formal language called ASL, the ACPI 243 Source Language (section 19 of the specification). This means that there 244 are always multiple ways to describe the same thing -- including device 245 properties. For example, device properties could use an ASL construct 246 that looks like this: Name(KEY0, "value0"). An ACPI device driver would 247 then retrieve the value of the property by evaluating the KEY0 object. 248 However, using Name() this way has multiple problems: (1) ACPI limits 249 names ("KEY0") to four characters unlike DT; (2) there is no industry 250 wide registry that maintains a list of names, minimizing re-use; (3) 251 there is also no registry for the definition of property values ("value0"), 252 again making re-use difficult; and (4) how does one maintain backward 253 compatibility as new hardware comes out? The _DSD method was created 254 to solve precisely these sorts of problems; Linux drivers should ALWAYS 255 use the _DSD method for device properties and nothing else. 256 257 The _DSM object (ACPI Section 9.14.1) could also be used for conveying 258 device properties to a driver. Linux drivers should only expect it to 259 be used if _DSD cannot represent the data required, and there is no way 260 to create a new UUID for the _DSD object. Note that there is even less 261 regulation of the use of _DSM than there is of _DSD. Drivers that depend 262 on the contents of _DSM objects will be more difficult to maintain over 263 time because of this; as of this writing, the use of _DSM is the cause 264 of quite a few firmware problems and is not recommended. 265 266 Drivers should look for device properties in the _DSD object ONLY; the _DSD 267 object is described in the ACPI specification section 6.2.5, but this only 268 describes how to define the structure of an object returned via _DSD, and 269 how specific data structures are defined by specific UUIDs. Linux should 270 only use the _DSD Device Properties UUID [5]: 271 272 -- UUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301 273 274 -- http://www.uefi.org/sites/default/files/resources/_DSD-device-properties-UUID.pdf 275 276 The UEFI Forum provides a mechanism for registering device properties [4] 277 so that they may be used across all operating systems supporting ACPI. 278 Device properties that have not been registered with the UEFI Forum should 279 not be used. 280 281 Before creating new device properties, check to be sure that they have not 282 been defined before and either registered in the Linux kernel documentation 283 as DT bindings, or the UEFI Forum as device properties. While we do not want 284 to simply move all DT bindings into ACPI device properties, we can learn from 285 what has been previously defined. 286 287 If it is necessary to define a new device property, or if it makes sense to 288 synthesize the definition of a binding so it can be used in any firmware, 289 both DT bindings and ACPI device properties for device drivers have review 290 processes. Use them both. When the driver itself is submitted for review 291 to the Linux mailing lists, the device property definitions needed must be 292 submitted at the same time. A driver that supports ACPI and uses device 293 properties will not be considered complete without their definitions. Once 294 the device property has been accepted by the Linux community, it must be 295 registered with the UEFI Forum [4], which will review it again for consistency 296 within the registry. This may require iteration. The UEFI Forum, though, 297 will always be the canonical site for device property definitions. 298 299 It may make sense to provide notice to the UEFI Forum that there is the 300 intent to register a previously unused device property name as a means of 301 reserving the name for later use. Other operating system vendors will 302 also be submitting registration requests and this may help smooth the 303 process. 304 305 Once registration and review have been completed, the kernel provides an 306 interface for looking up device properties in a manner independent of 307 whether DT or ACPI is being used. This API should be used [6]; it can 308 eliminate some duplication of code paths in driver probing functions and 309 discourage divergence between DT bindings and ACPI device properties. 310 311 312 Programmable Power Control Resources 313 ------------------------------------ 314 Programmable power control resources include such resources as voltage/current 315 providers (regulators) and clock sources. 316 317 With ACPI, the kernel clock and regulator framework is not expected to be used 318 at all. 319 320 The kernel assumes that power control of these resources is represented with 321 Power Resource Objects (ACPI section 7.1). The ACPI core will then handle 322 correctly enabling and disabling resources as they are needed. In order to 323 get that to work, ACPI assumes each device has defined D-states and that these 324 can be controlled through the optional ACPI methods _PS0, _PS1, _PS2, and _PS3; 325 in ACPI, _PS0 is the method to invoke to turn a device full on, and _PS3 is for 326 turning a device full off. 327 328 There are two options for using those Power Resources. They can: 329 330 -- be managed in a _PSx method which gets called on entry to power 331 state Dx. 332 333 -- be declared separately as power resources with their own _ON and _OFF 334 methods. They are then tied back to D-states for a particular device 335 via _PRx which specifies which power resources a device needs to be on 336 while in Dx. Kernel then tracks number of devices using a power resource 337 and calls _ON/_OFF as needed. 338 339 The kernel ACPI code will also assume that the _PSx methods follow the normal 340 ACPI rules for such methods: 341 342 -- If either _PS0 or _PS3 is implemented, then the other method must also 343 be implemented. 344 345 -- If a device requires usage or setup of a power resource when on, the ASL 346 should organize that it is allocated/enabled using the _PS0 method. 347 348 -- Resources allocated or enabled in the _PS0 method should be disabled 349 or de-allocated in the _PS3 method. 350 351 -- Firmware will leave the resources in a reasonable state before handing 352 over control to the kernel. 353 354 Such code in _PSx methods will of course be very platform specific. But, 355 this allows the driver to abstract out the interface for operating the device 356 and avoid having to read special non-standard values from ACPI tables. Further, 357 abstracting the use of these resources allows the hardware to change over time 358 without requiring updates to the driver. 359 360 361 Clocks 362 ------ 363 ACPI makes the assumption that clocks are initialized by the firmware -- 364 UEFI, in this case -- to some working value before control is handed over 365 to the kernel. This has implications for devices such as UARTs, or SoC-driven 366 LCD displays, for example. 367 368 When the kernel boots, the clocks are assumed to be set to reasonable 369 working values. If for some reason the frequency needs to change -- e.g., 370 throttling for power management -- the device driver should expect that 371 process to be abstracted out into some ACPI method that can be invoked 372 (please see the ACPI specification for further recommendations on standard 373 methods to be expected). The only exceptions to this are CPU clocks where 374 CPPC provides a much richer interface than ACPI methods. If the clocks 375 are not set, there is no direct way for Linux to control them. 376 377 If an SoC vendor wants to provide fine-grained control of the system clocks, 378 they could do so by providing ACPI methods that could be invoked by Linux 379 drivers. However, this is NOT recommended and Linux drivers should NOT use 380 such methods, even if they are provided. Such methods are not currently 381 standardized in the ACPI specification, and using them could tie a kernel 382 to a very specific SoC, or tie an SoC to a very specific version of the 383 kernel, both of which we are trying to avoid. 384 385 386 Driver Recommendations 387 ---------------------- 388 DO NOT remove any DT handling when adding ACPI support for a driver. The 389 same device may be used on many different systems. 390 391 DO try to structure the driver so that it is data-driven. That is, set up 392 a struct containing internal per-device state based on defaults and whatever 393 else must be discovered by the driver probe function. Then, have the rest 394 of the driver operate off of the contents of that struct. Doing so should 395 allow most divergence between ACPI and DT functionality to be kept local to 396 the probe function instead of being scattered throughout the driver. For 397 example: 398 399 static int device_probe_dt(struct platform_device *pdev) 400 { 401 /* DT specific functionality */ 402 ... 403 } 404 405 static int device_probe_acpi(struct platform_device *pdev) 406 { 407 /* ACPI specific functionality */ 408 ... 409 } 410 411 static int device_probe(struct platform_device *pdev) 412 { 413 ... 414 struct device_node node = pdev->dev.of_node; 415 ... 416 417 if (node) 418 ret = device_probe_dt(pdev); 419 else if (ACPI_HANDLE(&pdev->dev)) 420 ret = device_probe_acpi(pdev); 421 else 422 /* other initialization */ 423 ... 424 /* Continue with any generic probe operations */ 425 ... 426 } 427 428 DO keep the MODULE_DEVICE_TABLE entries together in the driver to make it 429 clear the different names the driver is probed for, both from DT and from 430 ACPI: 431 432 static struct of_device_id virtio_mmio_match[] = { 433 { .compatible = "virtio,mmio", }, 434 { } 435 }; 436 MODULE_DEVICE_TABLE(of, virtio_mmio_match); 437 438 static const struct acpi_device_id virtio_mmio_acpi_match[] = { 439 { "LNRO0005", }, 440 { } 441 }; 442 MODULE_DEVICE_TABLE(acpi, virtio_mmio_acpi_match); 443 444 445 ASWG 446 ---- 447 The ACPI specification changes regularly. During the year 2014, for instance, 448 version 5.1 was released and version 6.0 substantially completed, with most of 449 the changes being driven by ARM-specific requirements. Proposed changes are 450 presented and discussed in the ASWG (ACPI Specification Working Group) which 451 is a part of the UEFI Forum. The current version of the ACPI specification 452 is 6.1 release in January 2016. 453 454 Participation in this group is open to all UEFI members. Please see 455 http://www.uefi.org/workinggroup for details on group membership. 456 457 It is the intent of the ARMv8 ACPI kernel code to follow the ACPI specification 458 as closely as possible, and to only implement functionality that complies with 459 the released standards from UEFI ASWG. As a practical matter, there will be 460 vendors that provide bad ACPI tables or violate the standards in some way. 461 If this is because of errors, quirks and fix-ups may be necessary, but will 462 be avoided if possible. If there are features missing from ACPI that preclude 463 it from being used on a platform, ECRs (Engineering Change Requests) should be 464 submitted to ASWG and go through the normal approval process; for those that 465 are not UEFI members, many other members of the Linux community are and would 466 likely be willing to assist in submitting ECRs. 467 468 469 Linux Code 470 ---------- 471 Individual items specific to Linux on ARM, contained in the the Linux 472 source code, are in the list that follows: 473 474 ACPI_OS_NAME This macro defines the string to be returned when 475 an ACPI method invokes the _OS method. On ARM64 476 systems, this macro will be "Linux" by default. 477 The command line parameter acpi_os=<string> 478 can be used to set it to some other value. The 479 default value for other architectures is "Microsoft 480 Windows NT", for example. 481 482 ACPI Objects 483 ------------ 484 Detailed expectations for ACPI tables and object are listed in the file 485 Documentation/arm64/acpi_object_usage.txt. 486 487 488 References 489 ---------- 490 [0] http://silver.arm.com -- document ARM-DEN-0029, or newer 491 "Server Base System Architecture", version 2.3, dated 27 Mar 2014 492 493 [1] http://infocenter.arm.com/help/topic/com.arm.doc.den0044a/Server_Base_Boot_Requirements.pdf 494 Document ARM-DEN-0044A, or newer: "Server Base Boot Requirements, System 495 Software on ARM Platforms", dated 16 Aug 2014 496 497 [2] http://www.secretlab.ca/archives/151, 10 Jan 2015, Copyright (c) 2015, 498 Linaro Ltd., written by Grant Likely. 499 500 [3] AMD ACPI for Seattle platform documentation: 501 http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2012/10/Seattle_ACPI_Guide.pdf 502 503 [4] http://www.uefi.org/acpi -- please see the link for the "ACPI _DSD Device 504 Property Registry Instructions" 505 506 [5] http://www.uefi.org/acpi -- please see the link for the "_DSD (Device 507 Specific Data) Implementation Guide" 508 509 [6] Kernel code for the unified device property interface can be found in 510 include/linux/property.h and drivers/base/property.c. 511 512 513 Authors 514 ------- 515 Al Stone <al.stone@linaro.org> 516 Graeme Gregory <graeme.gregory@linaro.org> 517 Hanjun Guo <hanjun.guo@linaro.org> 518 519 Grant Likely <grant.likely@linaro.org>, for the "Why ACPI on ARM?" section