Stefano Babic | 4bf27d5 | 2013-02-23 00:53:25 +0000 | [diff] [blame] | 1 | U-Boot Falcon Mode |
| 2 | ==================== |
| 3 | |
| 4 | Introduction |
| 5 | ------------ |
| 6 | |
| 7 | This document provides an overview of how to add support for Falcon Mode |
| 8 | to a board. |
| 9 | |
| 10 | Falcon Mode is introduced to speed up the booting process, allowing |
| 11 | to boot a Linux kernel (or whatever image) without a full blown U-Boot. |
| 12 | |
| 13 | Falcon Mode relies on the SPL framework. In fact, to make booting faster, |
| 14 | U-Boot is split into two parts: the SPL (Secondary Program Loader) and U-Boot |
| 15 | image. In most implementations, SPL is used to start U-Boot when booting from |
| 16 | a mass storage, such as NAND or SD-Card. SPL has now support for other media, |
| 17 | and can generally be seen as a way to start an image performing the minimum |
| 18 | required initialization. SPL mainly initializes the RAM controller, and then |
| 19 | copies U-Boot image into the memory. |
| 20 | |
| 21 | The Falcon Mode extends this way allowing to start the Linux kernel directly |
| 22 | from SPL. A new command is added to U-Boot to prepare the parameters that SPL |
| 23 | must pass to the kernel, using ATAGS or Device Tree. |
| 24 | |
| 25 | In normal mode, these parameters are generated each time before |
| 26 | loading the kernel, passing to Linux the address in memory where |
| 27 | the parameters can be read. |
| 28 | With Falcon Mode, this snapshot can be saved into persistent storage and SPL is |
| 29 | informed to load it before running the kernel. |
| 30 | |
| 31 | To boot the kernel, these steps under a Falcon-aware U-Boot are required: |
| 32 | |
| 33 | 1. Boot the board into U-Boot. |
| 34 | Use the "spl export" command to generate the kernel parameters area or the DT. |
| 35 | U-Boot runs as when it boots the kernel, but stops before passing the control |
| 36 | to the kernel. |
| 37 | |
| 38 | 2. Save the prepared snapshot into persistent media. |
| 39 | The address where to save it must be configured into board configuration |
| 40 | file (CONFIG_CMD_SPL_NAND_OFS for NAND). |
| 41 | |
| 42 | 3. Boot the board into Falcon Mode. SPL will load the kernel and copy |
| 43 | the parameters which are saved in the persistent area to the required address. |
| 44 | |
| 45 | It is required to implement a custom mechanism to select if SPL loads U-Boot |
| 46 | or another image. |
| 47 | |
| 48 | The value of a GPIO is a simple way to operate the selection, as well as |
| 49 | reading a character from the SPL console if CONFIG_SPL_CONSOLE is set. |
| 50 | |
| 51 | Falcon Mode is generally activated by setting CONFIG_SPL_OS_BOOT. This tells |
| 52 | SPL that U-Boot is not the only available image that SPL is able to start. |
| 53 | |
| 54 | Configuration |
| 55 | ---------------------------- |
| 56 | CONFIG_CMD_SPL Enable the "spl export" command. |
| 57 | The command "spl export" is then available in U-Boot |
| 58 | mode |
| 59 | CONFIG_SYS_SPL_ARGS_ADDR Address in RAM where the parameters must be |
| 60 | copied by SPL. |
| 61 | In most cases, it is <start_of_ram> + 0x100 |
| 62 | |
| 63 | CONFIG_SYS_NAND_SPL_KERNEL_OFFS Offset in NAND where the kernel is stored |
| 64 | |
| 65 | CONFIG_CMD_SPL_NAND_OFS Offset in NAND where the parameters area was saved. |
| 66 | |
| 67 | CONFIG_CMD_SPL_WRITE_SIZE Size of the parameters area to be copied |
| 68 | |
| 69 | CONFIG_SPL_OS_BOOT Activate Falcon Mode. |
| 70 | |
| 71 | Function that a board must implement |
| 72 | ------------------------------------ |
| 73 | |
| 74 | void spl_board_prepare_for_linux(void) : optional |
| 75 | Called from SPL before starting the kernel |
| 76 | |
| 77 | spl_start_uboot() : required |
| 78 | Returns "0" if SPL should start the kernel, "1" if U-Boot |
| 79 | must be started. |
| 80 | |
| 81 | |
| 82 | Using spl command |
| 83 | ----------------- |
| 84 | |
| 85 | spl - SPL configuration |
| 86 | |
| 87 | Usage: |
| 88 | |
| 89 | spl export <img=atags|fdt> [kernel_addr] [initrd_addr] [fdt_addr ] |
| 90 | |
| 91 | img : "atags" or "fdt" |
| 92 | kernel_addr : kernel is loaded as part of the boot process, but it is not started. |
| 93 | This is the address where a kernel image is stored. |
| 94 | initrd_addr : Address of initial ramdisk |
| 95 | can be set to "-" if fdt_addr without initrd_addr is used |
| 96 | fdt_addr : in case of fdt, the address of the device tree. |
| 97 | |
| 98 | The spl export command does not write to a storage media. The user is |
| 99 | responsible to transfer the gathered information (assembled ATAGS list |
| 100 | or prepared FDT) from temporary storage in RAM into persistant storage |
| 101 | after each run of 'spl export'. Unfortunately the position of temporary |
| 102 | storage can not be predicted nor provided at commandline, it depends |
| 103 | highly on your system setup and your provided data (ATAGS or FDT). |
| 104 | However at the end of an succesful 'spl export' run it will print the |
| 105 | RAM address of temporary storage. |
| 106 | Now the user have to save the generated BLOB from that printed address |
| 107 | to the pre-defined address in persistent storage |
| 108 | (CONFIG_CMD_SPL_NAND_OFS in case of NAND). |
| 109 | The following example shows how to prepare the data for Falcon Mode on |
| 110 | twister board with ATAGS BLOB. |
| 111 | |
| 112 | The "spl export" command is prepared to work with ATAGS and FDT. However, |
| 113 | using FDT is at the moment untested. The ppc port (see a3m071 example |
| 114 | later) prepares the fdt blob with the fdt command instead. |
| 115 | |
| 116 | |
| 117 | Usage on the twister board: |
| 118 | -------------------------------- |
| 119 | |
| 120 | Using mtd names with the following (default) configuration |
| 121 | for mtdparts: |
| 122 | |
| 123 | device nand0 <omap2-nand.0>, # parts = 9 |
| 124 | #: name size offset mask_flags |
| 125 | 0: MLO 0x00080000 0x00000000 0 |
| 126 | 1: u-boot 0x00100000 0x00080000 0 |
| 127 | 2: env1 0x00040000 0x00180000 0 |
| 128 | 3: env2 0x00040000 0x001c0000 0 |
| 129 | 4: kernel 0x00600000 0x00200000 0 |
| 130 | 5: bootparms 0x00040000 0x00800000 0 |
| 131 | 6: splashimg 0x00200000 0x00840000 0 |
| 132 | 7: mini 0x02800000 0x00a40000 0 |
| 133 | 8: rootfs 0x1cdc0000 0x03240000 0 |
| 134 | |
| 135 | |
| 136 | twister => nand read 82000000 kernel |
| 137 | |
| 138 | NAND read: device 0 offset 0x200000, size 0x600000 |
| 139 | 6291456 bytes read: OK |
| 140 | |
| 141 | Now the kernel is in RAM at address 0x82000000 |
| 142 | |
| 143 | twister => spl export atags 0x82000000 |
| 144 | ## Booting kernel from Legacy Image at 82000000 ... |
| 145 | Image Name: Linux-3.5.0-rc4-14089-gda0b7f4 |
| 146 | Image Type: ARM Linux Kernel Image (uncompressed) |
| 147 | Data Size: 3654808 Bytes = 3.5 MiB |
| 148 | Load Address: 80008000 |
| 149 | Entry Point: 80008000 |
| 150 | Verifying Checksum ... OK |
| 151 | Loading Kernel Image ... OK |
| 152 | OK |
| 153 | cmdline subcommand not supported |
| 154 | bdt subcommand not supported |
| 155 | Argument image is now in RAM at: 0x80000100 |
| 156 | |
| 157 | The result can be checked at address 0x80000100: |
| 158 | |
| 159 | twister => md 0x80000100 |
| 160 | 80000100: 00000005 54410001 00000000 00000000 ......AT........ |
| 161 | 80000110: 00000000 00000067 54410009 746f6f72 ....g.....ATroot |
| 162 | 80000120: 65642f3d 666e2f76 77722073 73666e20 =/dev/nfs rw nfs |
| 163 | |
| 164 | The parameters generated with this step can be saved into NAND at the offset |
| 165 | 0x800000 (value for twister for CONFIG_CMD_SPL_NAND_OFS) |
| 166 | |
| 167 | nand erase.part bootparms |
| 168 | nand write 0x80000100 bootparms 0x4000 |
| 169 | |
| 170 | Now the parameters are stored into the NAND flash at the address |
| 171 | CONFIG_CMD_SPL_NAND_OFS (=0x800000). |
| 172 | |
| 173 | Next time, the board can be started into Falcon Mode moving the |
| 174 | setting the gpio (on twister gpio 55 is used) to kernel mode. |
| 175 | |
| 176 | The kernel is loaded directly by the SPL without passing through U-Boot. |
| 177 | |
| 178 | Example with FDT: a3m071 board |
| 179 | ------------------------------- |
| 180 | |
| 181 | To boot the Linux kernel from the SPL, the DT blob (fdt) needs to get |
| 182 | prepard/patched first. U-Boot usually inserts some dynamic values into |
| 183 | the DT binary (blob), e.g. autodetected memory size, MAC addresses, |
| 184 | clocks speeds etc. To generate this patched DT blob, you can use |
| 185 | the following command: |
| 186 | |
| 187 | 1. Load fdt blob to SDRAM: |
| 188 | => tftp 1800000 a3m071/a3m071.dtb |
| 189 | |
| 190 | 2. Set bootargs as desired for Linux booting (e.g. flash_mtd): |
| 191 | => run mtdargs addip2 addtty |
| 192 | |
| 193 | 3. Use "fdt" commands to patch the DT blob: |
| 194 | => fdt addr 1800000 |
| 195 | => fdt boardsetup |
| 196 | => fdt chosen |
| 197 | |
| 198 | 4. Display patched DT blob (optional): |
| 199 | => fdt print |
| 200 | |
| 201 | 5. Save fdt to NOR flash: |
| 202 | => erase fc060000 fc07ffff |
| 203 | => cp.b 1800000 fc060000 10000 |
| 204 | ... |
| 205 | |
| 206 | |
| 207 | Falcon Mode was presented at the RMLL 2012. Slides are available at: |
| 208 | |
| 209 | http://schedule2012.rmll.info/IMG/pdf/LSM2012_UbootFalconMode_Babic.pdf |