7 The following documentation is relevant to 2.4.18-rmk6 and beyond.
9 In order to boot ARM Linux, you require a boot loader, which is a small
10 program that runs before the main kernel. The boot loader is expected
11 to initialise various devices, and eventually call the Linux kernel,
12 passing information to the kernel.
14 Essentially, the boot loader should provide (as a minimum) the
17 1. Setup and initialise the RAM.
18 2. Initialise one serial port.
19 3. Detect the machine type.
20 4. Setup the kernel tagged list.
22 6. Call the kernel image.
25 1. Setup and initialise RAM
26 ---------------------------
28 Existing boot loaders: MANDATORY
29 New boot loaders: MANDATORY
31 The boot loader is expected to find and initialise all RAM that the
32 kernel will use for volatile data storage in the system. It performs
33 this in a machine dependent manner. (It may use internal algorithms
34 to automatically locate and size all RAM, or it may use knowledge of
35 the RAM in the machine, or any other method the boot loader designer
39 2. Initialise one serial port
40 -----------------------------
42 Existing boot loaders: OPTIONAL, RECOMMENDED
43 New boot loaders: OPTIONAL, RECOMMENDED
45 The boot loader should initialise and enable one serial port on the
46 target. This allows the kernel serial driver to automatically detect
47 which serial port it should use for the kernel console (generally
48 used for debugging purposes, or communication with the target.)
50 As an alternative, the boot loader can pass the relevant 'console='
51 option to the kernel via the tagged lists specifying the port, and
52 serial format options as described in
54 Documentation/kernel-parameters.txt.
57 3. Detect the machine type
58 --------------------------
60 Existing boot loaders: OPTIONAL
61 New boot loaders: MANDATORY
63 The boot loader should detect the machine type its running on by some
64 method. Whether this is a hard coded value or some algorithm that
65 looks at the connected hardware is beyond the scope of this document.
66 The boot loader must ultimately be able to provide a MACH_TYPE_xxx
67 value to the kernel. (see linux/arch/arm/tools/mach-types).
72 Existing boot loaders: OPTIONAL, HIGHLY RECOMMENDED
73 New boot loaders: MANDATORY
75 The boot loader must provide either a tagged list or a dtb image for
76 passing configuration data to the kernel. The physical address of the
77 boot data is passed to the kernel in register r2.
79 4a. Setup the kernel tagged list
80 --------------------------------
82 The boot loader must create and initialise the kernel tagged list.
83 A valid tagged list starts with ATAG_CORE and ends with ATAG_NONE.
84 The ATAG_CORE tag may or may not be empty. An empty ATAG_CORE tag
85 has the size field set to '2' (0x00000002). The ATAG_NONE must set
86 the size field to zero.
88 Any number of tags can be placed in the list. It is undefined
89 whether a repeated tag appends to the information carried by the
90 previous tag, or whether it replaces the information in its
91 entirety; some tags behave as the former, others the latter.
93 The boot loader must pass at a minimum the size and location of
94 the system memory, and root filesystem location. Therefore, the
95 minimum tagged list should look:
98 base -> | ATAG_CORE | |
100 | ATAG_MEM | | increasing address
105 The tagged list should be stored in system RAM.
107 The tagged list must be placed in a region of memory where neither
108 the kernel decompressor nor initrd 'bootp' program will overwrite
109 it. The recommended placement is in the first 16KiB of RAM.
111 4b. Setup the device tree
112 -------------------------
114 The boot loader must load a device tree image (dtb) into system ram
115 at a 64bit aligned address and initialize it with the boot data. The
116 dtb format is documented in Documentation/devicetree/booting-without-of.txt.
117 The kernel will look for the dtb magic value of 0xd00dfeed at the dtb
118 physical address to determine if a dtb has been passed instead of a
121 The boot loader must pass at a minimum the size and location of the
122 system memory, and the root filesystem location. The dtb must be
123 placed in a region of memory where the kernel decompressor will not
124 overwrite it, whilst remaining within the region which will be covered
125 by the kernel's low-memory mapping.
127 A safe location is just above the 128MiB boundary from start of RAM.
132 Existing boot loaders: OPTIONAL
133 New boot loaders: OPTIONAL
135 If an initramfs is in use then, as with the dtb, it must be placed in
136 a region of memory where the kernel decompressor will not overwrite it
137 while also with the region which will be covered by the kernel's
140 A safe location is just above the device tree blob which itself will
141 be loaded just above the 128MiB boundary from the start of RAM as
144 6. Calling the kernel image
145 ---------------------------
147 Existing boot loaders: MANDATORY
148 New boot loaders: MANDATORY
150 There are two options for calling the kernel zImage. If the zImage
151 is stored in flash, and is linked correctly to be run from flash,
152 then it is legal for the boot loader to call the zImage in flash
155 The zImage may also be placed in system RAM and called there. The
156 kernel should be placed in the first 128MiB of RAM. It is recommended
157 that it is loaded above 32MiB in order to avoid the need to relocate
158 prior to decompression, which will make the boot process slightly
161 When booting a raw (non-zImage) kernel the constraints are tighter.
162 In this case the kernel must be loaded at an offset into system equal
163 to TEXT_OFFSET - PAGE_OFFSET.
165 In any case, the following conditions must be met:
167 - Quiesce all DMA capable devices so that memory does not get
168 corrupted by bogus network packets or disk data. This will save
169 you many hours of debug.
171 - CPU register settings
173 r1 = machine type number discovered in (3) above.
174 r2 = physical address of tagged list in system RAM, or
175 physical address of device tree block (dtb) in system RAM
178 All forms of interrupts must be disabled (IRQs and FIQs)
180 For CPUs which do not include the ARM virtualization extensions, the
181 CPU must be in SVC mode. (A special exception exists for Angel)
183 CPUs which include support for the virtualization extensions can be
184 entered in HYP mode in order to enable the kernel to make full use of
185 these extensions. This is the recommended boot method for such CPUs,
186 unless the virtualisations are already in use by a pre-installed
189 If the kernel is not entered in HYP mode for any reason, it must be
194 Instruction cache may be on or off.
195 Data cache must be off.
197 If the kernel is entered in HYP mode, the above requirements apply to
198 the HYP mode configuration in addition to the ordinary PL1 (privileged
199 kernel modes) configuration. In addition, all traps into the
200 hypervisor must be disabled, and PL1 access must be granted for all
201 peripherals and CPU resources for which this is architecturally
202 possible. Except for entering in HYP mode, the system configuration
203 should be such that a kernel which does not include support for the
204 virtualization extensions can boot correctly without extra help.
206 - The boot loader is expected to call the kernel image by jumping
207 directly to the first instruction of the kernel image.
209 On CPUs supporting the ARM instruction set, the entry must be
210 made in ARM state, even for a Thumb-2 kernel.
212 On CPUs supporting only the Thumb instruction set such as
213 Cortex-M class CPUs, the entry must be made in Thumb state.