1 ===================
2 ACPI on Arm systems
3 ===================
4
5 ACPI can be used for Armv8 and Armv9 systems designed to follow
6 the BSA (Arm Base System Architecture) [0] and BBR (Arm
7 Base Boot Requirements) [1] specifications. Both BSA and BBR are publicly
8 accessible documents.
9 Arm Servers, in addition to being BSA compliant, comply with a set
10 of rules defined in SBSA (Server Base System Architecture) [2].
11
12 The Arm kernel implements the reduced hardware model of ACPI version
13 5.1 or later. Links to the specification and all external documents
14 it refers to are managed by the UEFI Forum. The specification is
15 available at http://www.uefi.org/specifications and documents referenced
16 by the specification can be found via http://www.uefi.org/acpi.
17
18 If an Arm system does not meet the requirements of the BSA and BBR,
19 or cannot be described using the mechanisms defined in the required ACPI
20 specifications, then ACPI may not be a good fit for the hardware.
21
22 While the documents mentioned above set out the requirements for building
23 industry-standard Arm systems, they also apply to more than one operating
24 system. The purpose of this document is to describe the interaction between
25 ACPI and Linux only, on an Arm system -- that is, what Linux expects of
26 ACPI and what ACPI can expect of Linux.
27
28
29 Why ACPI on Arm?
30 ----------------
31 Before examining the details of the interface between ACPI and Linux, it is
32 useful to understand why ACPI is being used. Several technologies already
33 exist in Linux for describing non-enumerable hardware, after all. In this
34 section we summarize a blog post [3] from Grant Likely that outlines the
35 reasoning behind ACPI on Arm systems. Actually, we snitch a good portion
36 of the summary text almost directly, to be honest.
37
38 The short form of the rationale for ACPI on Arm is:
39
40 - ACPI’s byte code (AML) allows the platform to encode hardware behavior,
41 while DT explicitly does not support this. For hardware vendors, being
42 able to encode behavior is a key tool used in supporting operating
43 system releases on new hardware.
44
45 - ACPI’s OSPM defines a power management model that constrains what the
46 platform is allowed to do into a specific model, while still providing
47 flexibility in hardware design.
48
49 - In the enterprise server environment, ACPI has established bindings (such
50 as for RAS) which are currently used in production systems. DT does not.
51 Such bindings could be defined in DT at some point, but doing so means Arm
52 and x86 would end up using completely different code paths in both firmware
53 and the kernel.
54
55 - Choosing a single interface to describe the abstraction between a platform
56 and an OS is important. Hardware vendors would not be required to implement
57 both DT and ACPI if they want to support multiple operating systems. And,
58 agreeing on a single interface instead of being fragmented into per OS
59 interfaces makes for better interoperability overall.
60
61 - The new ACPI governance process works well and Linux is now at the same
62 table as hardware vendors and other OS vendors. In fact, there is no
63 longer any reason to feel that ACPI only belongs to Windows or that
64 Linux is in any way secondary to Microsoft in this arena. The move of
65 ACPI governance into the UEFI forum has significantly opened up the
66 specification development process, and currently, a large portion of the
67 changes being made to ACPI are being driven by Linux.
68
69 Key to the use of ACPI is the support model. For servers in general, the
70 responsibility for hardware behaviour cannot solely be the domain of the
71 kernel, but rather must be split between the platform and the kernel, in
72 order to allow for orderly change over time. ACPI frees the OS from needing
73 to understand all the minute details of the hardware so that the OS doesn’t
74 need to be ported to each and every device individually. It allows the
75 hardware vendors to take responsibility for power management behaviour without
76 depending on an OS release cycle which is not under their control.
77
78 ACPI is also important because hardware and OS vendors have already worked
79 out the mechanisms for supporting a general purpose computing ecosystem. The
80 infrastructure is in place, the bindings are in place, and the processes are
81 in place. DT does exactly what Linux needs it to when working with vertically
82 integrated devices, but there are no good processes for supporting what the
83 server vendors need. Linux could potentially get there with DT, but doing so
84 really just duplicates something that already works. ACPI already does what
85 the hardware vendors need, Microsoft won’t collaborate on DT, and hardware
86 vendors would still end up providing two completely separate firmware
87 interfaces -- one for Linux and one for Windows.
88
89
90 Kernel Compatibility
91 --------------------
92 One of the primary motivations for ACPI is standardization, and using that
93 to provide backward compatibility for Linux kernels. In the server market,
94 software and hardware are often used for long periods. ACPI allows the
95 kernel and firmware to agree on a consistent abstraction that can be
96 maintained over time, even as hardware or software change. As long as the
97 abstraction is supported, systems can be updated without necessarily having
98 to replace the kernel.
99
100 When a Linux driver or subsystem is first implemented using ACPI, it by
101 definition ends up requiring a specific version of the ACPI specification
102 -- its baseline. ACPI firmware must continue to work, even though it may
103 not be optimal, with the earliest kernel version that first provides support
104 for that baseline version of ACPI. There may be a need for additional drivers,
105 but adding new functionality (e.g., CPU power management) should not break
106 older kernel versions. Further, ACPI firmware must also work with the most
107 recent version of the kernel.
108
109
110 Relationship with Device Tree
111 -----------------------------
112 ACPI support in drivers and subsystems for Arm should never be mutually
113 exclusive with DT support at compile time.
114
115 At boot time the kernel will only use one description method depending on
116 parameters passed from the boot loader (including kernel bootargs).
117
118 Regardless of whether DT or ACPI is used, the kernel must always be capable
119 of booting with either scheme (in kernels with both schemes enabled at compile
120 time).
121
122
123 Booting using ACPI tables
124 -------------------------
125 The only defined method for passing ACPI tables to the kernel on Arm
126 is via the UEFI system configuration table. Just so it is explicit, this
127 means that ACPI is only supported on platforms that boot via UEFI.
128
129 When an Arm system boots, it can either have DT information, ACPI tables,
130 or in some very unusual cases, both. If no command line parameters are used,
131 the kernel will try to use DT for device enumeration; if there is no DT
132 present, the kernel will try to use ACPI tables, but only if they are present.
133 If neither is available, the kernel will not boot. If acpi=force is used
134 on the command line, the kernel will attempt to use ACPI tables first, but
135 fall back to DT if there are no ACPI tables present. The basic idea is that
136 the kernel will not fail to boot unless it absolutely has no other choice.
137
138 Processing of ACPI tables may be disabled by passing acpi=off on the kernel
139 command line; this is the default behavior.
140
141 In order for the kernel to load and use ACPI tables, the UEFI implementation
142 MUST set the ACPI_20_TABLE_GUID to point to the RSDP table (the table with
143 the ACPI signature "RSD PTR "). If this pointer is incorrect and acpi=force
144 is used, the kernel will disable ACPI and try to use DT to boot instead; the
145 kernel has, in effect, determined that ACPI tables are not present at that
146 point.
147
148 If the pointer to the RSDP table is correct, the table will be mapped into
149 the kernel by the ACPI core, using the address provided by UEFI.
150
151 The ACPI core will then locate and map in all other ACPI tables provided by
152 using the addresses in the RSDP table to find the XSDT (eXtended System
153 Description Table). The XSDT in turn provides the addresses to all other
154 ACPI tables provided by the system firmware; the ACPI core will then traverse
155 this table and map in the tables listed.
156
157 The ACPI core will ignore any provided RSDT (Root System Description Table).
158 RSDTs have been deprecated and are ignored on arm64 since they only allow
159 for 32-bit addresses.
160
161 Further, the ACPI core will only use the 64-bit address fields in the FADT
162 (Fixed ACPI Description Table). Any 32-bit address fields in the FADT will
163 be ignored on arm64.
164
165 Hardware reduced mode (see Section 4.1 of the ACPI 6.1 specification) will
166 be enforced by the ACPI core on arm64. Doing so allows the ACPI core to
167 run less complex code since it no longer has to provide support for legacy
168 hardware from other architectures. Any fields that are not to be used for
169 hardware reduced mode must be set to zero.
170
171 For the ACPI core to operate properly, and in turn provide the information
172 the kernel needs to configure devices, it expects to find the following
173 tables (all section numbers refer to the ACPI 6.5 specification):
174
175 - RSDP (Root System Description Pointer), section 5.2.5
176
177 - XSDT (eXtended System Description Table), section 5.2.8
178
179 - FADT (Fixed ACPI Description Table), section 5.2.9
180
181 - DSDT (Differentiated System Description Table), section
182 5.2.11.1
183
184 - MADT (Multiple APIC Description Table), section 5.2.12
185
186 - GTDT (Generic Timer Description Table), section 5.2.24
187
188 - PPTT (Processor Properties Topology Table), section 5.2.30
189
190 - DBG2 (DeBuG port table 2), section 5.2.6, specifically Table 5-6.
191
192 - APMT (Arm Performance Monitoring unit Table), section 5.2.6, specifically Table 5-6.
193
194 - AGDI (Arm Generic diagnostic Dump and Reset Device Interface Table), section 5.2.6, specifically Table 5-6.
195
196 - If PCI is supported, the MCFG (Memory mapped ConFiGuration
197 Table), section 5.2.6, specifically Table 5-6.
198
199 - If booting without a console=<device> kernel parameter is
200 supported, the SPCR (Serial Port Console Redirection table),
201 section 5.2.6, specifically Table 5-6.
202
203 - If necessary to describe the I/O topology, SMMUs and GIC ITSs,
204 the IORT (Input Output Remapping Table, section 5.2.6, specifically
205 Table 5-6).
206
207 - If NUMA is supported, the following tables are required:
208
209 - SRAT (System Resource Affinity Table), section 5.2.16
210
211 - SLIT (System Locality distance Information Table), section 5.2.17
212
213 - If NUMA is supported, and the system contains heterogeneous memory,
214 the HMAT (Heterogeneous Memory Attribute Table), section 5.2.28.
215
216 - If the ACPI Platform Error Interfaces are required, the following
217 tables are conditionally required:
218
219 - BERT (Boot Error Record Table, section 18.3.1)
220
221 - EINJ (Error INJection table, section 18.6.1)
222
223 - ERST (Error Record Serialization Table, section 18.5)
224
225 - HEST (Hardware Error Source Table, section 18.3.2)
226
227 - SDEI (Software Delegated Exception Interface table, section 5.2.6,
228 specifically Table 5-6)
229
230 - AEST (Arm Error Source Table, section 5.2.6,
231 specifically Table 5-6)
232
233 - RAS2 (ACPI RAS2 feature table, section 5.2.21)
234
235 - If the system contains controllers using PCC channel, the
236 PCCT (Platform Communications Channel Table), section 14.1
237
238 - If the system contains a controller to capture board-level system state,
239 and communicates with the host via PCC, the PDTT (Platform Debug Trigger
240 Table), section 5.2.29.
241
242 - If NVDIMM is supported, the NFIT (NVDIMM Firmware Interface Table), section 5.2.26
243
244 - If video framebuffer is present, the BGRT (Boot Graphics Resource Table), section 5.2.23
245
246 - If IPMI is implemented, the SPMI (Server Platform Management Interface),
247 section 5.2.6, specifically Table 5-6.
248
249 - If the system contains a CXL Host Bridge, the CEDT (CXL Early Discovery
250 Table), section 5.2.6, specifically Table 5-6.
251
252 - If the system supports MPAM, the MPAM (Memory Partitioning And Monitoring table), section 5.2.6,
253 specifically Table 5-6.
254
255 - If the system lacks persistent storage, the IBFT (ISCSI Boot Firmware
256 Table), section 5.2.6, specifically Table 5-6.
257
258
259 If the above tables are not all present, the kernel may or may not be
260 able to boot properly since it may not be able to configure all of the
261 devices available. This list of tables is not meant to be all inclusive;
262 in some environments other tables may be needed (e.g., any of the APEI
263 tables from section 18) to support specific functionality.
264
265
266 ACPI Detection
267 --------------
268 Drivers should determine their probe() type by checking for a null
269 value for ACPI_HANDLE, or checking .of_node, or other information in
270 the device structure. This is detailed further in the "Driver
271 Recommendations" section.
272
273 In non-driver code, if the presence of ACPI needs to be detected at
274 run time, then check the value of acpi_disabled. If CONFIG_ACPI is not
275 set, acpi_disabled will always be 1.
276
277
278 Device Enumeration
279 ------------------
280 Device descriptions in ACPI should use standard recognized ACPI interfaces.
281 These may contain less information than is typically provided via a Device
282 Tree description for the same device. This is also one of the reasons that
283 ACPI can be useful -- the driver takes into account that it may have less
284 detailed information about the device and uses sensible defaults instead.
285 If done properly in the driver, the hardware can change and improve over
286 time without the driver having to change at all.
287
288 Clocks provide an excellent example. In DT, clocks need to be specified
289 and the drivers need to take them into account. In ACPI, the assumption
290 is that UEFI will leave the device in a reasonable default state, including
291 any clock settings. If for some reason the driver needs to change a clock
292 value, this can be done in an ACPI method; all the driver needs to do is
293 invoke the method and not concern itself with what the method needs to do
294 to change the clock. Changing the hardware can then take place over time
295 by changing what the ACPI method does, and not the driver.
296
297 In DT, the parameters needed by the driver to set up clocks as in the example
298 above are known as "bindings"; in ACPI, these are known as "Device Properties"
299 and provided to a driver via the _DSD object.
300
301 ACPI tables are described with a formal language called ASL, the ACPI
302 Source Language (section 19 of the specification). This means that there
303 are always multiple ways to describe the same thing -- including device
304 properties. For example, device properties could use an ASL construct
305 that looks like this: Name(KEY0, "value0"). An ACPI device driver would
306 then retrieve the value of the property by evaluating the KEY0 object.
307 However, using Name() this way has multiple problems: (1) ACPI limits
308 names ("KEY0") to four characters unlike DT; (2) there is no industry
309 wide registry that maintains a list of names, minimizing re-use; (3)
310 there is also no registry for the definition of property values ("value0"),
311 again making re-use difficult; and (4) how does one maintain backward
312 compatibility as new hardware comes out? The _DSD method was created
313 to solve precisely these sorts of problems; Linux drivers should ALWAYS
314 use the _DSD method for device properties and nothing else.
315
316 The _DSM object (ACPI Section 9.14.1) could also be used for conveying
317 device properties to a driver. Linux drivers should only expect it to
318 be used if _DSD cannot represent the data required, and there is no way
319 to create a new UUID for the _DSD object. Note that there is even less
320 regulation of the use of _DSM than there is of _DSD. Drivers that depend
321 on the contents of _DSM objects will be more difficult to maintain over
322 time because of this; as of this writing, the use of _DSM is the cause
323 of quite a few firmware problems and is not recommended.
324
325 Drivers should look for device properties in the _DSD object ONLY; the _DSD
326 object is described in the ACPI specification section 6.2.5, but this only
327 describes how to define the structure of an object returned via _DSD, and
328 how specific data structures are defined by specific UUIDs. Linux should
329 only use the _DSD Device Properties UUID [4]:
330
331 - UUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301
332
333 Common device properties can be registered by creating a pull request to [4] so
334 that they may be used across all operating systems supporting ACPI.
335 Device properties that have not been registered with the UEFI Forum can be used
336 but not as "uefi-" common properties.
337
338 Before creating new device properties, check to be sure that they have not
339 been defined before and either registered in the Linux kernel documentation
340 as DT bindings, or the UEFI Forum as device properties. While we do not want
341 to simply move all DT bindings into ACPI device properties, we can learn from
342 what has been previously defined.
343
344 If it is necessary to define a new device property, or if it makes sense to
345 synthesize the definition of a binding so it can be used in any firmware,
346 both DT bindings and ACPI device properties for device drivers have review
347 processes. Use them both. When the driver itself is submitted for review
348 to the Linux mailing lists, the device property definitions needed must be
349 submitted at the same time. A driver that supports ACPI and uses device
350 properties will not be considered complete without their definitions. Once
351 the device property has been accepted by the Linux community, it must be
352 registered with the UEFI Forum [4], which will review it again for consistency
353 within the registry. This may require iteration. The UEFI Forum, though,
354 will always be the canonical site for device property definitions.
355
356 It may make sense to provide notice to the UEFI Forum that there is the
357 intent to register a previously unused device property name as a means of
358 reserving the name for later use. Other operating system vendors will
359 also be submitting registration requests and this may help smooth the
360 process.
361
362 Once registration and review have been completed, the kernel provides an
363 interface for looking up device properties in a manner independent of
364 whether DT or ACPI is being used. This API should be used [5]; it can
365 eliminate some duplication of code paths in driver probing functions and
366 discourage divergence between DT bindings and ACPI device properties.
367
368
369 Programmable Power Control Resources
370 ------------------------------------
371 Programmable power control resources include such resources as voltage/current
372 providers (regulators) and clock sources.
373
374 With ACPI, the kernel clock and regulator framework is not expected to be used
375 at all.
376
377 The kernel assumes that power control of these resources is represented with
378 Power Resource Objects (ACPI section 7.1). The ACPI core will then handle
379 correctly enabling and disabling resources as they are needed. In order to
380 get that to work, ACPI assumes each device has defined D-states and that these
381 can be controlled through the optional ACPI methods _PS0, _PS1, _PS2, and _PS3;
382 in ACPI, _PS0 is the method to invoke to turn a device full on, and _PS3 is for
383 turning a device full off.
384
385 There are two options for using those Power Resources. They can:
386
387 - be managed in a _PSx method which gets called on entry to power
388 state Dx.
389
390 - be declared separately as power resources with their own _ON and _OFF
391 methods. They are then tied back to D-states for a particular device
392 via _PRx which specifies which power resources a device needs to be on
393 while in Dx. Kernel then tracks number of devices using a power resource
394 and calls _ON/_OFF as needed.
395
396 The kernel ACPI code will also assume that the _PSx methods follow the normal
397 ACPI rules for such methods:
398
399 - If either _PS0 or _PS3 is implemented, then the other method must also
400 be implemented.
401
402 - If a device requires usage or setup of a power resource when on, the ASL
403 should organize that it is allocated/enabled using the _PS0 method.
404
405 - Resources allocated or enabled in the _PS0 method should be disabled
406 or de-allocated in the _PS3 method.
407
408 - Firmware will leave the resources in a reasonable state before handing
409 over control to the kernel.
410
411 Such code in _PSx methods will of course be very platform specific. But,
412 this allows the driver to abstract out the interface for operating the device
413 and avoid having to read special non-standard values from ACPI tables. Further,
414 abstracting the use of these resources allows the hardware to change over time
415 without requiring updates to the driver.
416
417
418 Clocks
419 ------
420 ACPI makes the assumption that clocks are initialized by the firmware --
421 UEFI, in this case -- to some working value before control is handed over
422 to the kernel. This has implications for devices such as UARTs, or SoC-driven
423 LCD displays, for example.
424
425 When the kernel boots, the clocks are assumed to be set to reasonable
426 working values. If for some reason the frequency needs to change -- e.g.,
427 throttling for power management -- the device driver should expect that
428 process to be abstracted out into some ACPI method that can be invoked
429 (please see the ACPI specification for further recommendations on standard
430 methods to be expected). The only exceptions to this are CPU clocks where
431 CPPC provides a much richer interface than ACPI methods. If the clocks
432 are not set, there is no direct way for Linux to control them.
433
434 If an SoC vendor wants to provide fine-grained control of the system clocks,
435 they could do so by providing ACPI methods that could be invoked by Linux
436 drivers. However, this is NOT recommended and Linux drivers should NOT use
437 such methods, even if they are provided. Such methods are not currently
438 standardized in the ACPI specification, and using them could tie a kernel
439 to a very specific SoC, or tie an SoC to a very specific version of the
440 kernel, both of which we are trying to avoid.
441
442
443 Driver Recommendations
444 ----------------------
445 DO NOT remove any DT handling when adding ACPI support for a driver. The
446 same device may be used on many different systems.
447
448 DO try to structure the driver so that it is data-driven. That is, set up
449 a struct containing internal per-device state based on defaults and whatever
450 else must be discovered by the driver probe function. Then, have the rest
451 of the driver operate off of the contents of that struct. Doing so should
452 allow most divergence between ACPI and DT functionality to be kept local to
453 the probe function instead of being scattered throughout the driver. For
454 example::
455
456 static int device_probe_dt(struct platform_device *pdev)
457 {
458 /* DT specific functionality */
459 ...
460 }
461
462 static int device_probe_acpi(struct platform_device *pdev)
463 {
464 /* ACPI specific functionality */
465 ...
466 }
467
468 static int device_probe(struct platform_device *pdev)
469 {
470 ...
471 struct device_node node = pdev->dev.of_node;
472 ...
473
474 if (node)
475 ret = device_probe_dt(pdev);
476 else if (ACPI_HANDLE(&pdev->dev))
477 ret = device_probe_acpi(pdev);
478 else
479 /* other initialization */
480 ...
481 /* Continue with any generic probe operations */
482 ...
483 }
484
485 DO keep the MODULE_DEVICE_TABLE entries together in the driver to make it
486 clear the different names the driver is probed for, both from DT and from
487 ACPI::
488
489 static struct of_device_id virtio_mmio_match[] = {
490 { .compatible = "virtio,mmio", },
491 { }
492 };
493 MODULE_DEVICE_TABLE(of, virtio_mmio_match);
494
495 static const struct acpi_device_id virtio_mmio_acpi_match[] = {
496 { "LNRO0005", },
497 { }
498 };
499 MODULE_DEVICE_TABLE(acpi, virtio_mmio_acpi_match);
500
501
502 ASWG
503 ----
504 The ACPI specification changes regularly. During the year 2014, for instance,
505 version 5.1 was released and version 6.0 substantially completed, with most of
506 the changes being driven by Arm-specific requirements. Proposed changes are
507 presented and discussed in the ASWG (ACPI Specification Working Group) which
508 is a part of the UEFI Forum. The current version of the ACPI specification
509 is 6.5 release in August 2022.
510
511 Participation in this group is open to all UEFI members. Please see
512 http://www.uefi.org/workinggroup for details on group membership.
513
514 It is the intent of the Arm ACPI kernel code to follow the ACPI specification
515 as closely as possible, and to only implement functionality that complies with
516 the released standards from UEFI ASWG. As a practical matter, there will be
517 vendors that provide bad ACPI tables or violate the standards in some way.
518 If this is because of errors, quirks and fix-ups may be necessary, but will
519 be avoided if possible. If there are features missing from ACPI that preclude
520 it from being used on a platform, ECRs (Engineering Change Requests) should be
521 submitted to ASWG and go through the normal approval process; for those that
522 are not UEFI members, many other members of the Linux community are and would
523 likely be willing to assist in submitting ECRs.
524
525
526 Linux Code
527 ----------
528 Individual items specific to Linux on Arm, contained in the Linux
529 source code, are in the list that follows:
530
531 ACPI_OS_NAME
532 This macro defines the string to be returned when
533 an ACPI method invokes the _OS method. On Arm
534 systems, this macro will be "Linux" by default.
535 The command line parameter acpi_os=<string>
536 can be used to set it to some other value. The
537 default value for other architectures is "Microsoft
538 Windows NT", for example.
539
540 ACPI Objects
541 ------------
542 Detailed expectations for ACPI tables and object are listed in the file
543 Documentation/arch/arm64/acpi_object_usage.rst.
544
545
546 References
547 ----------
548 [0] https://developer.arm.com/documentation/den0094/latest
549 document Arm-DEN-0094: "Arm Base System Architecture", version 1.0C, dated 6 Oct 2022
550
551 [1] https://developer.arm.com/documentation/den0044/latest
552 Document Arm-DEN-0044: "Arm Base Boot Requirements", version 2.0G, dated 15 Apr 2022
553
554 [2] https://developer.arm.com/documentation/den0029/latest
555 Document Arm-DEN-0029: "Arm Server Base System Architecture", version 7.1, dated 06 Oct 2022
556
557 [3] http://www.secretlab.ca/archives/151,
558 10 Jan 2015, Copyright (c) 2015,
559 Linaro Ltd., written by Grant Likely.
560
561 [4] _DSD (Device Specific Data) Implementation Guide
562 https://github.com/UEFI/DSD-Guide/blob/main/dsd-guide.pdf
563
564 [5] Kernel code for the unified device
565 property interface can be found in
566 include/linux/property.h and drivers/base/property.c.
567
568
569 Authors
570 -------
571 - Al Stone <al.stone@linaro.org>
572 - Graeme Gregory <graeme.gregory@linaro.org>
573 - Hanjun Guo <hanjun.guo@linaro.org>
574
575 - Grant Likely <grant.likely@linaro.org>, for the "Why ACPI on ARM?" section
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.