1 .. SPDX-License-Identifier: GPL-2.0
3 ==================================================
4 Booting the Linux/ppc kernel without Open Firmware
5 ==================================================
7 Copyright (c) 2005 Benjamin Herrenschmidt <benh at kernel.crashing.org>,
10 Copyright (c) 2005 Becky Bruce <becky.bruce at freescale.com>,
11 Freescale Semiconductor, FSL SOC and 32-bit additions
13 Copyright (c) 2006 MontaVista Software, Inc.
14 Flash chip node definition
19 1) Entry point for arch/arm
20 2) Entry point for arch/powerpc
21 3) Entry point for arch/x86
22 4) Entry point for arch/mips/bmips
23 5) Entry point for arch/sh
25 II - The DT block format
27 2) Device tree generalities
28 3) Device tree "structure" block
29 4) Device tree "strings" block
31 III - Required content of the device tree
32 1) Note about cells and address representation
33 2) Note about "compatible" properties
34 3) Note about "name" properties
35 4) Note about node and property names and character set
36 5) Required nodes and properties
40 d) the /memory node(s)
42 f) the /soc<SOCname> node
44 IV - "dtc", the device tree compiler
46 V - Recommendations for a bootloader
48 VI - System-on-a-chip devices and nodes
49 1) Defining child nodes of an SOC
50 2) Representing devices without a current OF specification
52 VII - Specifying interrupt information for devices
53 1) interrupts property
54 2) interrupt-parent property
55 3) OpenPIC Interrupt Controllers
56 4) ISA Interrupt Controllers
58 VIII - Specifying device power management information (sleep property)
60 IX - Specifying dma bus information
62 Appendix A - Sample SOC node for MPC8540
69 - Initial draft, no chapter III yet.
72 - Add chapter III and bits & pieces here or
73 clarifies the fact that a lot of things are
74 optional, the kernel only requires a very
75 small device tree, though it is encouraged
76 to provide an as complete one as possible.
79 - Precise that DT block has to be in RAM
81 - Define version 3 and new format version 16
82 for the DT block (version 16 needs kernel
83 patches, will be fwd separately).
84 String block now has a size, and full path
85 is replaced by unit name for more
87 linux,phandle is made optional, only nodes
88 that are referenced by other nodes need it.
89 "name" property is now automatically
90 deduced from the unit name
93 - Correct confusion between OF_DT_END and
94 OF_DT_END_NODE in structure definition.
95 - Change version 16 format to always align
96 property data to 4 bytes. Since tokens are
97 already aligned, that means no specific
98 required alignment between property size
99 and property data. The old style variable
100 alignment would make it impossible to do
101 "simple" insertion of properties using
102 memmove (thanks Milton for
103 noticing). Updated kernel patch as well
104 - Correct a few more alignment constraints
105 - Add a chapter about the device-tree
106 compiler and the textural representation of
107 the tree that can be "compiled" by dtc.
109 November 21, 2005: Rev 0.5
110 - Additions/generalizations for 32-bit
111 - Changed to reflect the new arch/powerpc
117 - Add some definitions of interrupt tree (simple/complex)
118 - Add some definitions for PCI host bridges
119 - Add some common address format examples
120 - Add definitions for standard properties and "compatible"
121 names for cells that are not already defined by the existing
123 - Compare FSL SOC use of PCI to standard and make sure no new
124 node definition required.
125 - Add more information about node definitions for SOC devices
126 that currently have no standard, like the FSL CPM.
132 During the development of the Linux/ppc64 kernel, and more
133 specifically, the addition of new platform types outside of the old
134 IBM pSeries/iSeries pair, it was decided to enforce some strict rules
135 regarding the kernel entry and bootloader <-> kernel interfaces, in
136 order to avoid the degeneration that had become the ppc32 kernel entry
137 point and the way a new platform should be added to the kernel. The
138 legacy iSeries platform breaks those rules as it predates this scheme,
139 but no new board support will be accepted in the main tree that
140 doesn't follow them properly. In addition, since the advent of the
141 arch/powerpc merged architecture for ppc32 and ppc64, new 32-bit
142 platforms and 32-bit platforms which move into arch/powerpc will be
143 required to use these rules as well.
145 The main requirement that will be defined in more detail below is
146 the presence of a device-tree whose format is defined after Open
147 Firmware specification. However, in order to make life easier
148 to embedded board vendors, the kernel doesn't require the device-tree
149 to represent every device in the system and only requires some nodes
150 and properties to be present. This will be described in detail in
151 section III, but, for example, the kernel does not require you to
152 create a node for every PCI device in the system. It is a requirement
153 to have a node for PCI host bridges in order to provide interrupt
154 routing information and memory/IO ranges, among others. It is also
155 recommended to define nodes for on chip devices and other buses that
156 don't specifically fit in an existing OF specification. This creates a
157 great flexibility in the way the kernel can then probe those and match
158 drivers to device, without having to hard code all sorts of tables. It
159 also makes it more flexible for board vendors to do minor hardware
160 upgrades without significantly impacting the kernel code or cluttering
161 it with special cases.
164 1) Entry point for arch/arm
165 ---------------------------
167 There is one single entry point to the kernel, at the start
168 of the kernel image. That entry point supports two calling
169 conventions. A summary of the interface is described here. A full
170 description of the boot requirements is documented in
171 Documentation/arm/booting.rst
173 a) ATAGS interface. Minimal information is passed from firmware
174 to the kernel with a tagged list of predefined parameters.
178 r1 : Machine type number
180 r2 : Physical address of tagged list in system RAM
182 b) Entry with a flattened device-tree block. Firmware loads the
183 physical address of the flattened device tree block (dtb) into r2,
184 r1 is not used, but it is considered good practice to use a valid
185 machine number as described in Documentation/arm/booting.rst.
189 r1 : Valid machine type number. When using a device tree,
190 a single machine type number will often be assigned to
191 represent a class or family of SoCs.
193 r2 : physical pointer to the device-tree block
194 (defined in chapter II) in RAM. Device tree can be located
195 anywhere in system RAM, but it should be aligned on a 64 bit
198 The kernel will differentiate between ATAGS and device tree booting by
199 reading the memory pointed to by r2 and looking for either the flattened
200 device tree block magic value (0xd00dfeed) or the ATAG_CORE value at
201 offset 0x4 from r2 (0x54410001).
203 2) Entry point for arch/powerpc
204 -------------------------------
206 There is one single entry point to the kernel, at the start
207 of the kernel image. That entry point supports two calling
210 a) Boot from Open Firmware. If your firmware is compatible
211 with Open Firmware (IEEE 1275) or provides an OF compatible
212 client interface API (support for "interpret" callback of
213 forth words isn't required), you can enter the kernel with:
215 r5 : OF callback pointer as defined by IEEE 1275
216 bindings to powerpc. Only the 32-bit client interface
217 is currently supported
219 r3, r4 : address & length of an initrd if any or 0
221 The MMU is either on or off; the kernel will run the
222 trampoline located in arch/powerpc/kernel/prom_init.c to
223 extract the device-tree and other information from open
224 firmware and build a flattened device-tree as described
225 in b). prom_init() will then re-enter the kernel using
226 the second method. This trampoline code runs in the
227 context of the firmware, which is supposed to handle all
228 exceptions during that time.
230 b) Direct entry with a flattened device-tree block. This entry
231 point is called by a) after the OF trampoline and can also be
232 called directly by a bootloader that does not support the Open
233 Firmware client interface. It is also used by "kexec" to
234 implement "hot" booting of a new kernel from a previous
235 running one. This method is what I will describe in more
236 details in this document, as method a) is simply standard Open
237 Firmware, and thus should be implemented according to the
238 various standard documents defining it and its binding to the
239 PowerPC platform. The entry point definition then becomes:
241 r3 : physical pointer to the device-tree block
242 (defined in chapter II) in RAM
244 r4 : physical pointer to the kernel itself. This is
245 used by the assembly code to properly disable the MMU
246 in case you are entering the kernel with MMU enabled
247 and a non-1:1 mapping.
249 r5 : NULL (as to differentiate with method a)
251 Note about SMP entry: Either your firmware puts your other
252 CPUs in some sleep loop or spin loop in ROM where you can get
253 them out via a soft reset or some other means, in which case
254 you don't need to care, or you'll have to enter the kernel
255 with all CPUs. The way to do that with method b) will be
256 described in a later revision of this document.
258 Board supports (platforms) are not exclusive config options. An
259 arbitrary set of board supports can be built in a single kernel
260 image. The kernel will "know" what set of functions to use for a
261 given platform based on the content of the device-tree. Thus, you
264 a) add your platform support as a _boolean_ option in
265 arch/powerpc/Kconfig, following the example of PPC_PSERIES,
266 PPC_PMAC and PPC_MAPLE. The later is probably a good
267 example of a board support to start from.
269 b) create your main platform file as
270 "arch/powerpc/platforms/myplatform/myboard_setup.c" and add it
271 to the Makefile under the condition of your ``CONFIG_``
272 option. This file will define a structure of type "ppc_md"
273 containing the various callbacks that the generic code will
274 use to get to your platform specific code
276 A kernel image may support multiple platforms, but only if the
277 platforms feature the same core architecture. A single kernel build
278 cannot support both configurations with Book E and configurations
279 with classic Powerpc architectures.
281 3) Entry point for arch/x86
282 ---------------------------
284 There is one single 32bit entry point to the kernel at code32_start,
285 the decompressor (the real mode entry point goes to the same 32bit
286 entry point once it switched into protected mode). That entry point
287 supports one calling convention which is documented in
288 Documentation/x86/boot.rst
289 The physical pointer to the device-tree block (defined in chapter II)
290 is passed via setup_data which requires at least boot protocol 2.09.
291 The type filed is defined as::
295 This device-tree is used as an extension to the "boot page". As such it
296 does not parse / consider data which is already covered by the boot
297 page. This includes memory size, reserved ranges, command line arguments
298 or initrd address. It simply holds information which can not be retrieved
299 otherwise like interrupt routing or a list of devices behind an I2C bus.
301 4) Entry point for arch/mips/bmips
302 ----------------------------------
304 Some bootloaders only support a single entry point, at the start of the
305 kernel image. Other bootloaders will jump to the ELF start address.
306 Both schemes are supported; CONFIG_BOOT_RAW=y and CONFIG_NO_EXCEPT_FILL=y,
307 so the first instruction immediately jumps to kernel_entry().
309 Similar to the arch/arm case (b), a DT-aware bootloader is expected to
310 set up the following registers:
316 a2 : Physical pointer to the device tree block (defined in chapter
317 II) in RAM. The device tree can be located anywhere in the first
318 512MB of the physical address space (0x00000000 - 0x1fffffff),
319 aligned on a 64 bit boundary.
321 Legacy bootloaders do not use this convention, and they do not pass in a
322 DT block. In this case, Linux will look for a builtin DTB, selected via
325 This convention is defined for 32-bit systems only, as there are not
326 currently any 64-bit BMIPS implementations.
328 5) Entry point for arch/sh
329 --------------------------
331 Device-tree-compatible SH bootloaders are expected to provide the physical
332 address of the device tree blob in r4. Since legacy bootloaders did not
333 guarantee any particular initial register state, kernels built to
334 inter-operate with old bootloaders must either use a builtin DTB or
335 select a legacy board option (something other than CONFIG_SH_DEVICE_TREE)
336 that does not use device tree. Support for the latter is being phased out
337 in favor of device tree.
340 II - The DT block format
341 ========================
344 This chapter defines the actual format of the flattened device-tree
345 passed to the kernel. The actual content of it and kernel requirements
346 are described later. You can find example of code manipulating that
347 format in various places, including arch/powerpc/kernel/prom_init.c
348 which will generate a flattened device-tree from the Open Firmware
349 representation, or the fs2dt utility which is part of the kexec tools
350 which will generate one from a filesystem representation. It is
351 expected that a bootloader like uboot provides a bit more support,
352 that will be discussed later as well.
354 Note: The block has to be in main memory. It has to be accessible in
355 both real mode and virtual mode with no mapping other than main
356 memory. If you are writing a simple flash bootloader, it should copy
357 the block to RAM before passing it to the kernel.
363 The kernel is passed the physical address pointing to an area of memory
364 that is roughly described in include/linux/of_fdt.h by the structure
367 struct boot_param_header {
368 u32 magic; /* magic word OF_DT_HEADER */
369 u32 totalsize; /* total size of DT block */
370 u32 off_dt_struct; /* offset to structure */
371 u32 off_dt_strings; /* offset to strings */
372 u32 off_mem_rsvmap; /* offset to memory reserve map
374 u32 version; /* format version */
375 u32 last_comp_version; /* last compatible version */
377 /* version 2 fields below */
378 u32 boot_cpuid_phys; /* Which physical CPU id we're
380 /* version 3 fields below */
381 u32 size_dt_strings; /* size of the strings block */
383 /* version 17 fields below */
384 u32 size_dt_struct; /* size of the DT structure block */
387 Along with the constants::
389 /* Definitions used by the flattened device tree */
390 #define OF_DT_HEADER 0xd00dfeed /* 4: version,
392 #define OF_DT_BEGIN_NODE 0x1 /* Start node: full name
394 #define OF_DT_END_NODE 0x2 /* End node */
395 #define OF_DT_PROP 0x3 /* Property: name off,
397 #define OF_DT_END 0x9
399 All values in this header are in big endian format, the various
400 fields in this header are defined more precisely below. All
401 "offset" values are in bytes from the start of the header; that is
402 from the physical base address of the device tree block.
406 This is a magic value that "marks" the beginning of the
407 device-tree block header. It contains the value 0xd00dfeed and is
408 defined by the constant OF_DT_HEADER
412 This is the total size of the DT block including the header. The
413 "DT" block should enclose all data structures defined in this
414 chapter (who are pointed to by offsets in this header). That is,
415 the device-tree structure, strings, and the memory reserve map.
419 This is an offset from the beginning of the header to the start
420 of the "structure" part the device tree. (see 2) device tree)
424 This is an offset from the beginning of the header to the start
425 of the "strings" part of the device-tree
429 This is an offset from the beginning of the header to the start
430 of the reserved memory map. This map is a list of pairs of 64-
431 bit integers. Each pair is a physical address and a size. The
432 list is terminated by an entry of size 0. This map provides the
433 kernel with a list of physical memory areas that are "reserved"
434 and thus not to be used for memory allocations, especially during
435 early initialization. The kernel needs to allocate memory during
436 boot for things like un-flattening the device-tree, allocating an
437 MMU hash table, etc... Those allocations must be done in such a
438 way to avoid overriding critical things like, on Open Firmware
439 capable machines, the RTAS instance, or on some pSeries, the TCE
440 tables used for the iommu. Typically, the reserve map should
441 contain **at least** this DT block itself (header,total_size). If
442 you are passing an initrd to the kernel, you should reserve it as
443 well. You do not need to reserve the kernel image itself. The map
444 should be 64-bit aligned.
448 This is the version of this structure. Version 1 stops
449 here. Version 2 adds an additional field boot_cpuid_phys.
450 Version 3 adds the size of the strings block, allowing the kernel
451 to reallocate it easily at boot and free up the unused flattened
452 structure after expansion. Version 16 introduces a new more
453 "compact" format for the tree itself that is however not backward
454 compatible. Version 17 adds an additional field, size_dt_struct,
455 allowing it to be reallocated or moved more easily (this is
456 particularly useful for bootloaders which need to make
457 adjustments to a device tree based on probed information). You
458 should always generate a structure of the highest version defined
459 at the time of your implementation. Currently that is version 17,
460 unless you explicitly aim at being backward compatible.
464 Last compatible version. This indicates down to what version of
465 the DT block you are backward compatible. For example, version 2
466 is backward compatible with version 1 (that is, a kernel build
467 for version 1 will be able to boot with a version 2 format). You
468 should put a 1 in this field if you generate a device tree of
469 version 1 to 3, or 16 if you generate a tree of version 16 or 17
470 using the new unit name format.
474 This field only exist on version 2 headers. It indicate which
475 physical CPU ID is calling the kernel entry point. This is used,
476 among others, by kexec. If you are on an SMP system, this value
477 should match the content of the "reg" property of the CPU node in
478 the device-tree corresponding to the CPU calling the kernel entry
479 point (see further chapters for more information on the required
480 device-tree contents)
484 This field only exists on version 3 and later headers. It
485 gives the size of the "strings" section of the device tree (which
486 starts at the offset given by off_dt_strings).
490 This field only exists on version 17 and later headers. It gives
491 the size of the "structure" section of the device tree (which
492 starts at the offset given by off_dt_struct).
494 So the typical layout of a DT block (though the various parts don't
495 need to be in that order) looks like this (addresses go from top to
499 ------------------------------
500 base -> | struct boot_param_header |
501 ------------------------------
502 | (alignment gap) (*) |
503 ------------------------------
504 | memory reserve map |
505 ------------------------------
507 ------------------------------
509 | device-tree structure |
511 ------------------------------
513 ------------------------------
515 | device-tree strings |
517 -----> ------------------------------
520 --- (base + totalsize)
522 (*) The alignment gaps are not necessarily present; their presence
523 and size are dependent on the various alignment requirements of
524 the individual data blocks.
527 2) Device tree generalities
528 ---------------------------
530 This device-tree itself is separated in two different blocks, a
531 structure block and a strings block. Both need to be aligned to a 4
534 First, let's quickly describe the device-tree concept before detailing
535 the storage format. This chapter does _not_ describe the detail of the
536 required types of nodes & properties for the kernel, this is done
537 later in chapter III.
539 The device-tree layout is strongly inherited from the definition of
540 the Open Firmware IEEE 1275 device-tree. It's basically a tree of
541 nodes, each node having two or more named properties. A property can
544 It is a tree, so each node has one and only one parent except for the
545 root node who has no parent.
547 A node has 2 names. The actual node name is generally contained in a
548 property of type "name" in the node property list whose value is a
549 zero terminated string and is mandatory for version 1 to 3 of the
550 format definition (as it is in Open Firmware). Version 16 makes it
551 optional as it can generate it from the unit name defined below.
553 There is also a "unit name" that is used to differentiate nodes with
554 the same name at the same level, it is usually made of the node
555 names, the "@" sign, and a "unit address", which definition is
556 specific to the bus type the node sits on.
558 The unit name doesn't exist as a property per-se but is included in
559 the device-tree structure. It is typically used to represent "path" in
560 the device-tree. More details about the actual format of these will be
563 The kernel generic code does not make any formal use of the
564 unit address (though some board support code may do) so the only real
565 requirement here for the unit address is to ensure uniqueness of
566 the node unit name at a given level of the tree. Nodes with no notion
567 of address and no possible sibling of the same name (like /memory or
568 /cpus) may omit the unit address in the context of this specification,
569 or use the "@0" default unit address. The unit name is used to define
570 a node "full path", which is the concatenation of all parent node
571 unit names separated with "/".
573 The root node doesn't have a defined name, and isn't required to have
574 a name property either if you are using version 3 or earlier of the
575 format. It also has no unit address (no @ symbol followed by a unit
576 address). The root node unit name is thus an empty string. The full
577 path to the root node is "/".
579 Every node which actually represents an actual device (that is, a node
580 which isn't only a virtual "container" for more nodes, like "/cpus"
581 is) is also required to have a "compatible" property indicating the
582 specific hardware and an optional list of devices it is fully
583 backwards compatible with.
585 Finally, every node that can be referenced from a property in another
586 node is required to have either a "phandle" or a "linux,phandle"
587 property. Real Open Firmware implementations provide a unique
588 "phandle" value for every node that the "prom_init()" trampoline code
589 turns into "linux,phandle" properties. However, this is made optional
590 if the flattened device tree is used directly. An example of a node
591 referencing another node via "phandle" is when laying out the
592 interrupt tree which will be described in a further version of this
595 The "phandle" property is a 32-bit value that uniquely
596 identifies a node. You are free to use whatever values or system of
597 values, internal pointers, or whatever to generate these, the only
598 requirement is that every node for which you provide that property has
599 a unique value for it.
601 Here is an example of a simple device-tree. In this example, an "o"
602 designates a node followed by the node unit name. Properties are
603 presented with their name followed by their content. "content"
604 represents an ASCII string (zero terminated) value, while <content>
605 represents a 32-bit value, specified in decimal or hexadecimal (the
606 latter prefixed 0x). The various nodes in this example will be
607 discussed in a later chapter. At this point, it is only meant to give
608 you a idea of what a device-tree looks like. I have purposefully kept
609 the "name" and "linux,phandle" properties which aren't necessary in
610 order to give you a better idea of what the tree looks like in
614 |- name = "device-tree"
615 |- model = "MyBoardName"
616 |- compatible = "MyBoardFamilyName"
617 |- #address-cells = <2>
619 |- linux,phandle = <0>
623 | | - linux,phandle = <1>
624 | | - #address-cells = <1>
625 | | - #size-cells = <0>
628 | |- name = "PowerPC,970"
629 | |- device_type = "cpu"
631 | |- clock-frequency = <0x5f5e1000>
633 | |- linux,phandle = <2>
637 | |- device_type = "memory"
638 | |- reg = <0x00000000 0x00000000 0x00000000 0x20000000>
639 | |- linux,phandle = <3>
643 |- bootargs = "root=/dev/sda2"
644 |- linux,phandle = <4>
646 This tree is almost a minimal tree. It pretty much contains the
647 minimal set of required nodes and properties to boot a linux kernel;
648 that is, some basic model information at the root, the CPUs, and the
649 physical memory layout. It also includes misc information passed
650 through /chosen, like in this example, the platform type (mandatory)
651 and the kernel command line arguments (optional).
653 The /cpus/PowerPC,970@0/64-bit property is an example of a
654 property without a value. All other properties have a value. The
655 significance of the #address-cells and #size-cells properties will be
656 explained in chapter IV which defines precisely the required nodes and
657 properties and their content.
660 3) Device tree "structure" block
661 --------------------------------
663 The structure of the device tree is a linearized tree structure. The
664 "OF_DT_BEGIN_NODE" token starts a new node, and the "OF_DT_END_NODE"
665 ends that node definition. Child nodes are simply defined before
666 "OF_DT_END_NODE" (that is nodes within the node). A 'token' is a 32
667 bit value. The tree has to be "finished" with a OF_DT_END token
669 Here's the basic structure of a single node:
671 * token OF_DT_BEGIN_NODE (that is 0x00000001)
672 * for version 1 to 3, this is the node full path as a zero
673 terminated string, starting with "/". For version 16 and later,
674 this is the node unit name only (or an empty string for the
676 * [align gap to next 4 bytes boundary]
679 * token OF_DT_PROP (that is 0x00000003)
680 * 32-bit value of property value size in bytes (or 0 if no
682 * 32-bit value of offset in string block of property name
683 * property value data if any
684 * [align gap to next 4 bytes boundary]
686 * [child nodes if any]
687 * token OF_DT_END_NODE (that is 0x00000002)
689 So the node content can be summarized as a start token, a full path,
690 a list of properties, a list of child nodes, and an end token. Every
691 child node is a full node structure itself as defined above.
693 NOTE: The above definition requires that all property definitions for
694 a particular node MUST precede any subnode definitions for that node.
695 Although the structure would not be ambiguous if properties and
696 subnodes were intermingled, the kernel parser requires that the
697 properties come first (up until at least 2.6.22). Any tools
698 manipulating a flattened tree must take care to preserve this
701 4) Device tree "strings" block
702 ------------------------------
704 In order to save space, property names, which are generally redundant,
705 are stored separately in the "strings" block. This block is simply the
706 whole bunch of zero terminated strings for all property names
707 concatenated together. The device-tree property definitions in the
708 structure block will contain offset values from the beginning of the
712 III - Required content of the device tree
713 =========================================
717 All ``linux,*`` properties defined in this document apply only
718 to a flattened device-tree. If your platform uses a real
719 implementation of Open Firmware or an implementation compatible with
720 the Open Firmware client interface, those properties will be created
721 by the trampoline code in the kernel's prom_init() file. For example,
722 that's where you'll have to add code to detect your board model and
723 set the platform number. However, when using the flattened device-tree
724 entry point, there is no prom_init() pass, and thus you have to
725 provide those properties yourself.
728 1) Note about cells and address representation
729 ----------------------------------------------
731 The general rule is documented in the various Open Firmware
732 documentations. If you choose to describe a bus with the device-tree
733 and there exist an OF bus binding, then you should follow the
734 specification. However, the kernel does not require every single
735 device or bus to be described by the device tree.
737 In general, the format of an address for a device is defined by the
738 parent bus type, based on the #address-cells and #size-cells
739 properties. Note that the parent's parent definitions of #address-cells
740 and #size-cells are not inherited so every node with children must specify
741 them. The kernel requires the root node to have those properties defining
742 addresses format for devices directly mapped on the processor bus.
744 Those 2 properties define 'cells' for representing an address and a
745 size. A "cell" is a 32-bit number. For example, if both contain 2
746 like the example tree given above, then an address and a size are both
747 composed of 2 cells, and each is a 64-bit number (cells are
748 concatenated and expected to be in big endian format). Another example
749 is the way Apple firmware defines them, with 2 cells for an address
750 and one cell for a size. Most 32-bit implementations should define
751 #address-cells and #size-cells to 1, which represents a 32-bit value.
752 Some 32-bit processors allow for physical addresses greater than 32
753 bits; these processors should define #address-cells as 2.
755 "reg" properties are always a tuple of the type "address size" where
756 the number of cells of address and size is specified by the bus
757 #address-cells and #size-cells. When a bus supports various address
758 spaces and other flags relative to a given address allocation (like
759 prefetchable, etc...) those flags are usually added to the top level
760 bits of the physical address. For example, a PCI physical address is
761 made of 3 cells, the bottom two containing the actual address itself
762 while the top cell contains address space indication, flags, and pci
763 bus & device numbers.
765 For buses that support dynamic allocation, it's the accepted practice
766 to then not provide the address in "reg" (keep it 0) though while
767 providing a flag indicating the address is dynamically allocated, and
768 then, to provide a separate "assigned-addresses" property that
769 contains the fully allocated addresses. See the PCI OF bindings for
772 In general, a simple bus with no address space bits and no dynamic
773 allocation is preferred if it reflects your hardware, as the existing
774 kernel address parsing functions will work out of the box. If you
775 define a bus type with a more complex address format, including things
776 like address space bits, you'll have to add a bus translator to the
777 prom_parse.c file of the recent kernels for your bus type.
779 The "reg" property only defines addresses and sizes (if #size-cells is
780 non-0) within a given bus. In order to translate addresses upward
781 (that is into parent bus addresses, and possibly into CPU physical
782 addresses), all buses must contain a "ranges" property. If the
783 "ranges" property is missing at a given level, it's assumed that
784 translation isn't possible, i.e., the registers are not visible on the
785 parent bus. The format of the "ranges" property for a bus is a list
788 bus address, parent bus address, size
790 "bus address" is in the format of the bus this bus node is defining,
791 that is, for a PCI bridge, it would be a PCI address. Thus, (bus
792 address, size) defines a range of addresses for child devices. "parent
793 bus address" is in the format of the parent bus of this bus. For
794 example, for a PCI host controller, that would be a CPU address. For a
795 PCI<->ISA bridge, that would be a PCI address. It defines the base
796 address in the parent bus where the beginning of that range is mapped.
798 For new 64-bit board support, I recommend either the 2/2 format or
799 Apple's 2/1 format which is slightly more compact since sizes usually
800 fit in a single 32-bit word. New 32-bit board support should use a
801 1/1 format, unless the processor supports physical addresses greater
802 than 32-bits, in which case a 2/1 format is recommended.
804 Alternatively, the "ranges" property may be empty, indicating that the
805 registers are visible on the parent bus using an identity mapping
806 translation. In other words, the parent bus address space is the same
807 as the child bus address space.
809 2) Note about "compatible" properties
810 -------------------------------------
812 These properties are optional, but recommended in devices and the root
813 node. The format of a "compatible" property is a list of concatenated
814 zero terminated strings. They allow a device to express its
815 compatibility with a family of similar devices, in some cases,
816 allowing a single driver to match against several devices regardless
817 of their actual names.
819 3) Note about "name" properties
820 -------------------------------
822 While earlier users of Open Firmware like OldWorld macintoshes tended
823 to use the actual device name for the "name" property, it's nowadays
824 considered a good practice to use a name that is closer to the device
825 class (often equal to device_type). For example, nowadays, Ethernet
826 controllers are named "ethernet", an additional "model" property
827 defining precisely the chip type/model, and "compatible" property
828 defining the family in case a single driver can driver more than one
829 of these chips. However, the kernel doesn't generally put any
830 restriction on the "name" property; it is simply considered good
831 practice to follow the standard and its evolutions as closely as
834 Note also that the new format version 16 makes the "name" property
835 optional. If it's absent for a node, then the node's unit name is then
836 used to reconstruct the name. That is, the part of the unit name
837 before the "@" sign is used (or the entire unit name if no "@" sign
840 4) Note about node and property names and character set
841 -------------------------------------------------------
843 While Open Firmware provides more flexible usage of 8859-1, this
844 specification enforces more strict rules. Nodes and properties should
845 be comprised only of ASCII characters 'a' to 'z', '0' to
846 '9', ',', '.', '_', '+', '#', '?', and '-'. Node names additionally
847 allow uppercase characters 'A' to 'Z' (property names should be
848 lowercase. The fact that vendors like Apple don't respect this rule is
849 irrelevant here). Additionally, node and property names should always
850 begin with a character in the range 'a' to 'z' (or 'A' to 'Z' for node
853 The maximum number of characters for both nodes and property names
854 is 31. In the case of node names, this is only the leftmost part of
855 a unit name (the pure "name" property), it doesn't include the unit
856 address which can extend beyond that limit.
859 5) Required nodes and properties
860 --------------------------------
861 These are all that are currently required. However, it is strongly
862 recommended that you expose PCI host bridges as documented in the
863 PCI binding to Open Firmware, and your interrupt tree as documented
864 in OF interrupt tree specification.
868 The root node requires some properties to be present:
870 - model : this is your board name/model
871 - #address-cells : address representation for "root" devices
872 - #size-cells: the size representation for "root" devices
873 - compatible : the board "family" generally finds its way here,
874 for example, if you have 2 board models with a similar layout,
875 that typically get driven by the same platform code in the
876 kernel, you would specify the exact board model in the
877 compatible property followed by an entry that represents the SoC
880 The root node is also generally where you add additional properties
881 specific to your board like the serial number if any, that sort of
882 thing. It is recommended that if you add any "custom" property whose
883 name may clash with standard defined ones, you prefix them with your
884 vendor name and a comma.
886 Additional properties for the root node:
888 - serial-number : a string representing the device's serial number
892 This node is the parent of all individual CPU nodes. It doesn't
893 have any specific requirements, though it's generally good practice
896 #address-cells = <00000001>
897 #size-cells = <00000000>
899 This defines that the "address" for a CPU is a single cell, and has
900 no meaningful size. This is not necessary but the kernel will assume
901 that format when reading the "reg" properties of a CPU node, see
904 c) The ``/cpus/*`` nodes
906 So under /cpus, you are supposed to create a node for every CPU on
907 the machine. There is no specific restriction on the name of the
908 CPU, though it's common to call it <architecture>,<core>. For
909 example, Apple uses PowerPC,G5 while IBM uses PowerPC,970FX.
910 However, the Generic Names convention suggests that it would be
911 better to simply use 'cpu' for each cpu node and use the compatible
912 property to identify the specific cpu core.
916 - device_type : has to be "cpu"
917 - reg : This is the physical CPU number, it's a single 32-bit cell
918 and is also used as-is as the unit number for constructing the
919 unit name in the full path. For example, with 2 CPUs, you would
922 /cpus/PowerPC,970FX@0
923 /cpus/PowerPC,970FX@1
925 (unit addresses do not require leading zeroes)
926 - d-cache-block-size : one cell, L1 data cache block size in bytes [#]_
927 - i-cache-block-size : one cell, L1 instruction cache block size in
929 - d-cache-size : one cell, size of L1 data cache in bytes
930 - i-cache-size : one cell, size of L1 instruction cache in bytes
932 .. [#] The cache "block" size is the size on which the cache management
933 instructions operate. Historically, this document used the cache
934 "line" size here which is incorrect. The kernel will prefer the cache
935 block size and will fallback to cache line size for backward
938 Recommended properties:
940 - timebase-frequency : a cell indicating the frequency of the
941 timebase in Hz. This is not directly used by the generic code,
942 but you are welcome to copy/paste the pSeries code for setting
943 the kernel timebase/decrementer calibration based on this
945 - clock-frequency : a cell indicating the CPU core clock frequency
946 in Hz. A new property will be defined for 64-bit values, but if
947 your frequency is < 4Ghz, one cell is enough. Here as well as
948 for the above, the common code doesn't use that property, but
949 you are welcome to re-use the pSeries or Maple one. A future
950 kernel version might provide a common function for this.
951 - d-cache-line-size : one cell, L1 data cache line size in bytes
952 if different from the block size
953 - i-cache-line-size : one cell, L1 instruction cache line size in
954 bytes if different from the block size
956 You are welcome to add any property you find relevant to your board,
957 like some information about the mechanism used to soft-reset the
958 CPUs. For example, Apple puts the GPIO number for CPU soft reset
959 lines in there as a "soft-reset" property since they start secondary
960 CPUs by soft-resetting them.
963 d) the /memory node(s)
965 To define the physical memory layout of your board, you should
966 create one or more memory node(s). You can either create a single
967 node with all memory ranges in its reg property, or you can create
968 several nodes, as you wish. The unit address (@ part) used for the
969 full path is the address of the first range of memory defined by a
970 given node. If you use a single memory node, this will typically be
975 - device_type : has to be "memory"
976 - reg : This property contains all the physical memory ranges of
977 your board. It's a list of addresses/sizes concatenated
978 together, with the number of cells of each defined by the
979 #address-cells and #size-cells of the root node. For example,
980 with both of these properties being 2 like in the example given
981 earlier, a 970 based machine with 6Gb of RAM could typically
982 have a "reg" property here that looks like::
984 00000000 00000000 00000000 80000000
985 00000001 00000000 00000001 00000000
987 That is a range starting at 0 of 0x80000000 bytes and a range
988 starting at 0x100000000 and of 0x100000000 bytes. You can see
989 that there is no memory covering the IO hole between 2Gb and
990 4Gb. Some vendors prefer splitting those ranges into smaller
991 segments, but the kernel doesn't care.
993 Additional properties:
995 - hotpluggable : The presence of this property provides an explicit
996 hint to the operating system that this memory may potentially be
997 removed later. The kernel can take this into consideration when
998 doing nonmovable allocations and when laying out memory zones.
1002 This node is a bit "special". Normally, that's where Open Firmware
1003 puts some variable environment information, like the arguments, or
1004 the default input/output devices.
1006 This specification makes a few of these mandatory, but also defines
1007 some linux-specific properties that would be normally constructed by
1008 the prom_init() trampoline when booting with an OF client interface,
1009 but that you have to provide yourself when using the flattened format.
1011 Recommended properties:
1013 - bootargs : This zero-terminated string is passed as the kernel
1015 - linux,stdout-path : This is the full path to your standard
1016 console device if any. Typically, if you have serial devices on
1017 your board, you may want to put the full path to the one set as
1018 the default console in the firmware here, for the kernel to pick
1019 it up as its own default console.
1021 Note that u-boot creates and fills in the chosen node for platforms
1024 (Note: a practice that is now obsolete was to include a property
1025 under /chosen called interrupt-controller which had a phandle value
1026 that pointed to the main interrupt controller)
1028 f) the /soc<SOCname> node
1030 This node is used to represent a system-on-a-chip (SoC) and must be
1031 present if the processor is a SoC. The top-level soc node contains
1032 information that is global to all devices on the SoC. The node name
1033 should contain a unit address for the SoC, which is the base address
1034 of the memory-mapped register set for the SoC. The name of an SoC
1035 node should start with "soc", and the remainder of the name should
1036 represent the part number for the soc. For example, the MPC8540's
1037 soc node would be called "soc8540".
1039 Required properties:
1041 - ranges : Should be defined as specified in 1) to describe the
1042 translation of SoC addresses for memory mapped SoC registers.
1043 - bus-frequency: Contains the bus frequency for the SoC node.
1044 Typically, the value of this field is filled in by the boot
1046 - compatible : Exact model of the SoC
1049 Recommended properties:
1051 - reg : This property defines the address and size of the
1052 memory-mapped registers that are used for the SOC node itself.
1053 It does not include the child device registers - these will be
1054 defined inside each child node. The address specified in the
1055 "reg" property should match the unit address of the SOC node.
1056 - #address-cells : Address representation for "soc" devices. The
1057 format of this field may vary depending on whether or not the
1058 device registers are memory mapped. For memory mapped
1059 registers, this field represents the number of cells needed to
1060 represent the address of the registers. For SOCs that do not
1061 use MMIO, a special address format should be defined that
1062 contains enough cells to represent the required information.
1063 See 1) above for more details on defining #address-cells.
1064 - #size-cells : Size representation for "soc" devices
1065 - #interrupt-cells : Defines the width of cells used to represent
1066 interrupts. Typically this value is <2>, which includes a
1067 32-bit number that represents the interrupt number, and a
1068 32-bit number that represents the interrupt sense and level.
1069 This field is only needed if the SOC contains an interrupt
1072 The SOC node may contain child nodes for each SOC device that the
1073 platform uses. Nodes should not be created for devices which exist
1074 on the SOC but are not used by a particular platform. See chapter VI
1075 for more information on how to specify devices that are part of a SOC.
1077 Example SOC node for the MPC8540::
1080 #address-cells = <1>;
1082 #interrupt-cells = <2>;
1083 device_type = "soc";
1084 ranges = <0x00000000 0xe0000000 0x00100000>
1085 reg = <0xe0000000 0x00003000>;
1086 bus-frequency = <0>;
1091 IV - "dtc", the device tree compiler
1092 ====================================
1095 dtc source code can be found at
1096 <http://git.jdl.com/gitweb/?p=dtc.git>
1100 This version is still in early development stage; the
1101 resulting device-tree "blobs" have not yet been validated with the
1102 kernel. The current generated block lacks a useful reserve map (it will
1103 be fixed to generate an empty one, it's up to the bootloader to fill
1104 it up) among others. The error handling needs work, bugs are lurking,
1107 dtc basically takes a device-tree in a given format and outputs a
1108 device-tree in another format. The currently supported formats are:
1113 - "dtb": "blob" format, that is a flattened device-tree block
1115 header all in a binary blob.
1116 - "dts": "source" format. This is a text file containing a
1117 "source" for a device-tree. The format is defined later in this
1119 - "fs" format. This is a representation equivalent to the
1120 output of /proc/device-tree, that is nodes are directories and
1121 properties are files
1126 - "dtb": "blob" format
1127 - "dts": "source" format
1128 - "asm": assembly language file. This is a file that can be
1129 sourced by gas to generate a device-tree "blob". That file can
1130 then simply be added to your Makefile. Additionally, the
1131 assembly file exports some symbols that can be used.
1134 The syntax of the dtc tool is::
1136 dtc [-I <input-format>] [-O <output-format>]
1137 [-o output-filename] [-V output_version] input_filename
1140 The "output_version" defines what version of the "blob" format will be
1141 generated. Supported versions are 1,2,3 and 16. The default is
1142 currently version 3 but that may change in the future to version 16.
1144 Additionally, dtc performs various sanity checks on the tree, like the
1145 uniqueness of linux, phandle properties, validity of strings, etc...
1147 The format of the .dts "source" file is "C" like, supports C and C++
1153 The above is the "device-tree" definition. It's the only statement
1154 supported currently at the toplevel.
1159 property1 = "string_value"; /* define a property containing a 0
1163 property2 = <0x1234abcd>; /* define a property containing a
1164 * numerical 32-bit value (hexadecimal)
1167 property3 = <0x12345678 0x12345678 0xdeadbeef>;
1168 /* define a property containing 3
1169 * numerical 32-bit values (cells) in
1172 property4 = [0x0a 0x0b 0x0c 0x0d 0xde 0xea 0xad 0xbe 0xef];
1173 /* define a property whose content is
1174 * an arbitrary array of bytes
1177 childnode@address { /* define a child node named "childnode"
1178 * whose unit name is "childnode at
1182 childprop = "hello\n"; /* define a property "childprop" of
1183 * childnode (in this case, a string)
1188 Nodes can contain other nodes etc... thus defining the hierarchical
1189 structure of the tree.
1191 Strings support common escape sequences from C: "\n", "\t", "\r",
1192 "\(octal value)", "\x(hex value)".
1194 It is also suggested that you pipe your source file through cpp (gcc
1195 preprocessor) so you can use #include's, #define for constants, etc...
1197 Finally, various options are planned but not yet implemented, like
1198 automatic generation of phandles, labels (exported to the asm file so
1199 you can point to a property content and change it easily from whatever
1200 you link the device-tree with), label or path instead of numeric value
1201 in some cells to "point" to a node (replaced by a phandle at compile
1202 time), export of reserve map address to the asm file, ability to
1203 specify reserve map content at compile time, etc...
1205 We may provide a .h include file with common definitions of that
1206 proves useful for some properties (like building PCI properties or
1207 interrupt maps) though it may be better to add a notion of struct
1208 definitions to the compiler...
1211 V - Recommendations for a bootloader
1212 ====================================
1215 Here are some various ideas/recommendations that have been proposed
1216 while all this has been defined and implemented.
1218 - The bootloader may want to be able to use the device-tree itself
1219 and may want to manipulate it (to add/edit some properties,
1220 like physical memory size or kernel arguments). At this point, 2
1221 choices can be made. Either the bootloader works directly on the
1222 flattened format, or the bootloader has its own internal tree
1223 representation with pointers (similar to the kernel one) and
1224 re-flattens the tree when booting the kernel. The former is a bit
1225 more difficult to edit/modify, the later requires probably a bit
1226 more code to handle the tree structure. Note that the structure
1227 format has been designed so it's relatively easy to "insert"
1228 properties or nodes or delete them by just memmoving things
1229 around. It contains no internal offsets or pointers for this
1232 - An example of code for iterating nodes & retrieving properties
1233 directly from the flattened tree format can be found in the kernel
1234 file drivers/of/fdt.c. Look at the of_scan_flat_dt() function,
1235 its usage in early_init_devtree(), and the corresponding various
1236 early_init_dt_scan_*() callbacks. That code can be re-used in a
1237 GPL bootloader, and as the author of that code, I would be happy
1238 to discuss possible free licensing to any vendor who wishes to
1239 integrate all or part of this code into a non-GPL bootloader.
1240 (reference needed; who is 'I' here? ---gcl Jan 31, 2011)
1244 VI - System-on-a-chip devices and nodes
1245 =======================================
1247 Many companies are now starting to develop system-on-a-chip
1248 processors, where the processor core (CPU) and many peripheral devices
1249 exist on a single piece of silicon. For these SOCs, an SOC node
1250 should be used that defines child nodes for the devices that make
1251 up the SOC. While platforms are not required to use this model in
1252 order to boot the kernel, it is highly encouraged that all SOC
1253 implementations define as complete a flat-device-tree as possible to
1254 describe the devices on the SOC. This will allow for the
1255 genericization of much of the kernel code.
1258 1) Defining child nodes of an SOC
1259 ---------------------------------
1261 Each device that is part of an SOC may have its own node entry inside
1262 the SOC node. For each device that is included in the SOC, the unit
1263 address property represents the address offset for this device's
1264 memory-mapped registers in the parent's address space. The parent's
1265 address space is defined by the "ranges" property in the top-level soc
1266 node. The "reg" property for each node that exists directly under the
1267 SOC node should contain the address mapping from the child address space
1268 to the parent SOC address space and the size of the device's
1269 memory-mapped register file.
1271 For many devices that may exist inside an SOC, there are predefined
1272 specifications for the format of the device tree node. All SOC child
1273 nodes should follow these specifications, except where noted in this
1276 See appendix A for an example partial SOC node definition for the
1280 2) Representing devices without a current OF specification
1281 ----------------------------------------------------------
1283 Currently, there are many devices on SoCs that do not have a standard
1284 representation defined as part of the Open Firmware specifications,
1285 mainly because the boards that contain these SoCs are not currently
1286 booted using Open Firmware. Binding documentation for new devices
1287 should be added to the Documentation/devicetree/bindings directory.
1288 That directory will expand as device tree support is added to more and
1292 VII - Specifying interrupt information for devices
1293 ===================================================
1295 The device tree represents the buses and devices of a hardware
1296 system in a form similar to the physical bus topology of the
1299 In addition, a logical 'interrupt tree' exists which represents the
1300 hierarchy and routing of interrupts in the hardware.
1302 The interrupt tree model is fully described in the
1303 document "Open Firmware Recommended Practice: Interrupt
1304 Mapping Version 0.9". The document is available at:
1305 <http://www.devicetree.org/open-firmware/practice/>
1307 1) interrupts property
1308 ----------------------
1310 Devices that generate interrupts to a single interrupt controller
1311 should use the conventional OF representation described in the
1312 OF interrupt mapping documentation.
1314 Each device which generates interrupts must have an 'interrupt'
1315 property. The interrupt property value is an arbitrary number of
1316 of 'interrupt specifier' values which describe the interrupt or
1317 interrupts for the device.
1319 The encoding of an interrupt specifier is determined by the
1320 interrupt domain in which the device is located in the
1321 interrupt tree. The root of an interrupt domain specifies in
1322 its #interrupt-cells property the number of 32-bit cells
1323 required to encode an interrupt specifier. See the OF interrupt
1324 mapping documentation for a detailed description of domains.
1326 For example, the binding for the OpenPIC interrupt controller
1327 specifies an #interrupt-cells value of 2 to encode the interrupt
1328 number and level/sense information. All interrupt children in an
1329 OpenPIC interrupt domain use 2 cells per interrupt in their interrupts
1332 The PCI bus binding specifies a #interrupt-cells value of 1 to encode
1333 which interrupt pin (INTA,INTB,INTC,INTD) is used.
1335 2) interrupt-parent property
1336 ----------------------------
1338 The interrupt-parent property is specified to define an explicit
1339 link between a device node and its interrupt parent in
1340 the interrupt tree. The value of interrupt-parent is the
1341 phandle of the parent node.
1343 If the interrupt-parent property is not defined for a node, its
1344 interrupt parent is assumed to be an ancestor in the node's
1345 *device tree* hierarchy.
1347 3) OpenPIC Interrupt Controllers
1348 --------------------------------
1350 OpenPIC interrupt controllers require 2 cells to encode
1351 interrupt information. The first cell defines the interrupt
1352 number. The second cell defines the sense and level
1355 Sense and level information should be encoded as follows:
1357 == ========================================
1358 0 low to high edge sensitive type enabled
1359 1 active low level sensitive type enabled
1360 2 active high level sensitive type enabled
1361 3 high to low edge sensitive type enabled
1362 == ========================================
1364 4) ISA Interrupt Controllers
1365 ----------------------------
1367 ISA PIC interrupt controllers require 2 cells to encode
1368 interrupt information. The first cell defines the interrupt
1369 number. The second cell defines the sense and level
1372 ISA PIC interrupt controllers should adhere to the ISA PIC
1373 encodings listed below:
1375 == ========================================
1376 0 active low level sensitive type enabled
1377 1 active high level sensitive type enabled
1378 2 high to low edge sensitive type enabled
1379 3 low to high edge sensitive type enabled
1380 == ========================================
1382 VIII - Specifying Device Power Management Information (sleep property)
1383 ======================================================================
1385 Devices on SOCs often have mechanisms for placing devices into low-power
1386 states that are decoupled from the devices' own register blocks. Sometimes,
1387 this information is more complicated than a cell-index property can
1388 reasonably describe. Thus, each device controlled in such a manner
1389 may contain a "sleep" property which describes these connections.
1391 The sleep property consists of one or more sleep resources, each of
1392 which consists of a phandle to a sleep controller, followed by a
1393 controller-specific sleep specifier of zero or more cells.
1395 The semantics of what type of low power modes are possible are defined
1396 by the sleep controller. Some examples of the types of low power modes
1397 that may be supported are:
1399 - Dynamic: The device may be disabled or enabled at any time.
1400 - System Suspend: The device may request to be disabled or remain
1401 awake during system suspend, but will not be disabled until then.
1402 - Permanent: The device is disabled permanently (until the next hard
1405 Some devices may share a clock domain with each other, such that they should
1406 only be suspended when none of the devices are in use. Where reasonable,
1407 such nodes should be placed on a virtual bus, where the bus has the sleep
1408 property. If the clock domain is shared among devices that cannot be
1409 reasonably grouped in this manner, then create a virtual sleep controller
1410 (similar to an interrupt nexus, except that defining a standardized
1411 sleep-map should wait until its necessity is demonstrated).
1413 IX - Specifying dma bus information
1414 ===================================
1416 Some devices may have DMA memory range shifted relatively to the beginning of
1417 RAM, or even placed outside of kernel RAM. For example, the Keystone 2 SoC
1418 worked in LPAE mode with 4G memory has:
1419 - RAM range: [0x8 0000 0000, 0x8 FFFF FFFF]
1420 - DMA range: [ 0x8000 0000, 0xFFFF FFFF]
1421 and DMA range is aliased into first 2G of RAM in HW.
1423 In such cases, DMA addresses translation should be performed between CPU phys
1424 and DMA addresses. The "dma-ranges" property is intended to be used
1425 for describing the configuration of such system in DT.
1427 In addition, each DMA master device on the DMA bus may or may not support
1428 coherent DMA operations. The "dma-coherent" property is intended to be used
1429 for identifying devices supported coherent DMA operations in DT.
1435 - dma-ranges: <prop-encoded-array> encoded as arbitrary number of triplets of
1436 (child-bus-address, parent-bus-address, length). Each triplet specified
1437 describes a contiguous DMA address range.
1438 The dma-ranges property is used to describe the direct memory access (DMA)
1439 structure of a memory-mapped bus whose device tree parent can be accessed
1440 from DMA operations originating from the bus. It provides a means of
1441 defining a mapping or translation between the physical address space of
1442 the bus and the physical address space of the parent of the bus.
1443 (for more information see the Devicetree Specification)
1449 - dma-ranges: <empty> value. if present - It means that DMA addresses
1450 translation has to be enabled for this device.
1451 - dma-coherent: Present if dma operations are coherent
1456 compatible = "ti,keystone","simple-bus";
1457 ranges = <0x0 0x0 0x0 0xc0000000>;
1458 dma-ranges = <0x80000000 0x8 0x00000000 0x80000000>;
1463 compatible = "ti,keystone-dwc3";
1470 Appendix A - Sample SOC node for MPC8540
1471 ========================================
1476 #address-cells = <1>;
1478 compatible = "fsl,mpc8540-ccsr", "simple-bus";
1479 device_type = "soc";
1480 ranges = <0x00000000 0xe0000000 0x00100000>
1481 bus-frequency = <0>;
1482 interrupt-parent = <&pic>;
1485 #address-cells = <1>;
1487 device_type = "network";
1489 compatible = "gianfar", "simple-bus";
1490 reg = <0x24000 0x1000>;
1491 local-mac-address = [ 0x00 0xE0 0x0C 0x00 0x73 0x00 ];
1492 interrupts = <0x29 2 0x30 2 0x34 2>;
1493 phy-handle = <&phy0>;
1494 sleep = <&pmc 0x00000080>;
1498 reg = <0x24520 0x20>;
1499 compatible = "fsl,gianfar-mdio";
1501 phy0: ethernet-phy@0 {
1506 phy1: ethernet-phy@1 {
1511 phy3: ethernet-phy@3 {
1519 device_type = "network";
1521 compatible = "gianfar";
1522 reg = <0x25000 0x1000>;
1523 local-mac-address = [ 0x00 0xE0 0x0C 0x00 0x73 0x01 ];
1524 interrupts = <0x13 2 0x14 2 0x18 2>;
1525 phy-handle = <&phy1>;
1526 sleep = <&pmc 0x00000040>;
1530 device_type = "network";
1532 compatible = "gianfar";
1533 reg = <0x26000 0x1000>;
1534 local-mac-address = [ 0x00 0xE0 0x0C 0x00 0x73 0x02 ];
1535 interrupts = <0x41 2>;
1536 phy-handle = <&phy3>;
1537 sleep = <&pmc 0x00000020>;
1541 #address-cells = <1>;
1543 compatible = "fsl,mpc8540-duart", "simple-bus";
1544 sleep = <&pmc 0x00000002>;
1548 device_type = "serial";
1549 compatible = "ns16550";
1550 reg = <0x4500 0x100>;
1551 clock-frequency = <0>;
1552 interrupts = <0x42 2>;
1556 device_type = "serial";
1557 compatible = "ns16550";
1558 reg = <0x4600 0x100>;
1559 clock-frequency = <0>;
1560 interrupts = <0x42 2>;
1565 interrupt-controller;
1566 #address-cells = <0>;
1567 #interrupt-cells = <2>;
1568 reg = <0x40000 0x40000>;
1569 compatible = "chrp,open-pic";
1570 device_type = "open-pic";
1574 interrupts = <0x43 2>;
1575 reg = <0x3000 0x100>;
1576 compatible = "fsl-i2c";
1578 sleep = <&pmc 0x00000004>;
1582 compatible = "fsl,mpc8540-pmc", "fsl,mpc8548-pmc";
1583 reg = <0xe0070 0x20>;