25eb09c3bb
replace *(.init_array .init_array.*) with KEEP(*(.init_array .init_array.*)). The KEEP statement within a linker script will instruct the linker to keep the specified section, even if no symbols inside it are referenced. This statement is used within the SECTIONS section of the linker script. This becomes relevant when garbage collection is performed at link time, enabled by passing the --gc-sections switch to the linker. The KEEP statement instructs the linker to use the specified section as a root node when creating a dependency graph, looking for unused sections. Essentially forcing the section to be marked as used. Signed-off-by: cuiziwei <cuiziwei@xiaomi.com> |
||
---|---|---|
.. | ||
configs/nsh | ||
include | ||
scripts | ||
src | ||
Kconfig | ||
README.txt |
README ====== This directory contains the port of NuttX to the Adafruit Metro M4. The Metro M4 uses a Arduino form factor and and pinout. It's powered with an ATSAMD51J19: o Cortex M4 core running at 120 MHz o Hardware DSP and floating point support o 512 KB flash, 192 KB RAM o 32-bit, 3.3V logic and power o Dual 1 MSPS DAC (A0 and A1) o Dual 1 MSPS ADC (8 analog pins) o 6 x hardware SERCOM (I2C, SPI or UART) o 16 x PWM outputs o Stereo I2S input/output with MCK pin o 10-bit Parallel capture controller (for camera/video in) o Built in crypto engines with AES (256 bit), true RNG, Pubkey controller o 64 QFN Contents ======== o STATUS o Unlocking FLASH o Serial Console o LEDs o Run from SRAM o Configurations STATUS ====== 2018-07-26: The basic port was merged into master. It is still incomplete and untested. 2018-07-29: Code complete. Clock configuration complete. Unverified SERCOM USART, SPI, I2C, Port configuration, and DMA support have been added. I still have no hardware in hand to test. 2018-07-20: Brought in the USB driver from the SAML21. It is the same USB IP with only small differences. There a a few, small open issues still to be resolved. 2018-08-01: Hardware in hand. Initial attempts to program the board using a Segger J-Link connected via SWD were unsuccessful because the Metro M4 comes with an application in FLASH and the FLASH locked. See "Unlocking FLASH with J-Link Commander" below. After unlocking the FLASH, I was able to successfully write the NuttX image. Unfortunately, the board seems to have become unusable after the first NuttX image was written to FLASH. I am unable to connect the JTAG debugger. The primary JTAG problem seems to be that it is now unable to halt the CPU. Future me: This boot-up failure was do to bad clock initialization logic that caused infinite loops during clock configuration. Unlocking and erasing the FLASH is innocuous, but the JTAG will apparently not work if the clocks are not in a good state. I would say that as a general practice, any changes to the clock configuration should be testing in SRAM first before risking the write to FLASH. 2018-08-03: Added a configuration option to run out of SRAM vs FLASH. This is a safer way to do the initial board bring-up since it does not modify the FLASH image nor does it require unlocking the FLASH pages. 2018-08-31: I finally have a new Metro M4 and have successfully debugged from SRAM (with FLASH unlocked and erased). Several errors in clock configuration logic have been corrected and it now gets through clock configuration okay. It now hangs in the low-level USART initialization. It hangs trying to enabled the SERCOM slow clock channel. The clock sequence is: 1. 32.678KHz crystal -> XOSC32K This is configured and says that XOSC32K is ready. 2. XOSCK32 -> GCLK3. This is configured and it says that is is ready (GENEN=1). 3. GCLK3 ->SERCOM slow clock channel. This hangs when I try to enable the peripheral clock. 2018-09-01: I found a workaround by substituting OSCULP32K for XOSC32 as the source to GCLK3. With that workaround, the port gets past all clock and USART configuration. A new configuration option was added, CONFIG_METRO_M4_32KHZXTAL. By default this workaround is in place. But you can enable CONFIG_METRO_M4_32KHZXTAL if you want to further study the XOSC32K problem. With that workaround (and a bunch of other fixes), the basic NSH configuration appears fully function, indicating the the board bring- up is complete. There are additional drivers ported from SAML21 which has, in most cases, identical peripherals. None of these drivers have been verified on the SAMD51, However. These include: DMAC, I2C, SPI, and USB. WARNING: If you decide to invest the time to discover whey the XOSC32K clock source is not working, be certain to use the SRAM configuration. That configuration in FLASH is most likely lock up your board irrecoverably is there are any start-up errors! Unlocking FLASH =============== Options ------- The Adafruit Metro M4 comes with a very nice bootloader resident in FLASH. so we have two options: 1. Learn to play well with others. Make NuttX coexist and work in the memory partition available to it. Or, 2. Be greedy, unlock the FLASH and overwrite the bootloader. I chose to do the last one. I used a Segger J-Link and here are the steps that I took. You can probably do these things in Atmel Studio (?) but for other debug environments, you would have to come up with the solution. Unlocking FLASH with J-Link Commander ------------------------------------- 1. Start J-Link Commander: SEGGER J-Link Commander V6.32i (Compiled Jul 24 2018 15:20:49) DLL version V6.32i, compiled Jul 24 2018 15:19:55 Connecting to J-Link via USB...O.K. Firmware: J-Link V9 compiled Apr 20 2018 16:47:26 Hardware version: V9.30 S/N: 269303123 License(s): FlashBP, GDB OEM: SEGGER-EDU VTref=3.296V Type "connect" to establish a target connection, '?' for help J-Link>con Please specify device / core. <Default>: ATSAMD51P19 Type '?' for selection dialog Device>ATSAMD51P19 Please specify target interface: J) JTAG (Default) S) SWD TIF>S Specify target interface speed [kHz]. <Default>: 4000 kHz Speed> Device "ATSAMD51P19" selected. Connecting to target via SWD Found SW-DP with ID 0x2BA01477 Scanning AP map to find all available APs ...etc. ... 2. Look at The NVM "user page" memory at address 0x00804000: J-Link>mem8 804000, 10 00804000 = 39 92 9A F6 80 FF EC AE FF FF FF FF FF FF FF FF The field NVM BOOT (also called BOOTPROT) is the field that locks the lower part of FLASH to support the boot loader. This is bits 26-29 of the NVM user page: J-Link>mem32 804000, 1 00804000 = F69A9239 In binary 11|11 01|10 1001 1010 1001 0010 0011 1001, so NVM Boot 1101. To unlock the FLASH memory reserved for the bootloader, we need to change this field to 111 so that: 11|11 11|10 10|01 1010 1001 0010 0011 1001 = F7da9239, or 00804000 = 39 92 9A FE 80 FF EC AE FF FF FF FF FF FF FF FF is read. 3. Modify the NVM "user page" I did this using the instructions for the SAMD21 found at https://roamingthings.de/use-j-link-to-change-the-boot-loader-protection-of-a-sam-d21/ We will need to create a small Motorola S-REC file to write new values into NVM. See https://en.m.wikipedia.org/wiki/SREC_(file_format) for a description of the Motorola SREC format. I wrote a small program at boards/arm/samd5e5/metro-m4-scripts/nvm.c that will generate this Motorola SREC file with the correct checksum. The file at boards/arm/samd5e5/metro-m4-scripts/nvm.c is the output of that program. J-Link>mem8 804000,10 00804000 = 39 92 9A F6 80 FF EC AE FF FF FF FF FF FF FF FF J-Link>loadfile D:\Spuda\Documents\projects\nuttx\master\nuttx\boards\metro-m4\scripts\nvm.srec Downloading file [D:\Spuda\Documents\projects\nuttx\master\nuttx\boards\metro-m4\scripts\nvm.srec]... J-Link: Flash download: Bank 1 @ 0x00804000: 1 range affected (16 bytes) J-Link: Flash download: Total time needed: 0.089s (Prepare: 0.035s, Compare: 0.011s, Erase: 0.000s, Program: 0.019s, Verify: 0.011s, Restore: 0.011s) O.K. J-Link>mem8 804000,10 00804000 = 39 92 9A FE 80 FF EC AE FF FF FF FF FF FF FF FF You will, of course, have to change the path as appropriate for your system. 4. Erase FLASH (optional) J-Link>erase Erasing device (ATSAMD51P19)... J-Link: Flash download: Total time needed: 2.596s (Prepare: 0.031s, Compare: 0.000s, Erase: 2.553s, Program: 0.000s, Verify: 0.000s, Restore: 0.012s) J-Link: Flash download: Total time needed: 0.066s (Prepare: 0.038s, Compare: 0.000s, Erase: 0.016s, Program: 0.000s, Verify: 0.000s, Restore: 0.010s) Erasing done. J-Link> Serial Console ============== An Arduino compatible serial Shield is assumed (or equivalently, and external RS-232 or serial-to-USB adapter connected on Arduino pins D0 and D1): ------ ----------------- ----------- SHIELD SAMD5E5 FUNCTION ------ ----------------- ----------- D0 PA23 SERCOM3 PAD2 RXD D1 PA22 SERCOM3 PAD0 TXD LEDs ==== The Adafruit Metro M4 has four LEDs, but only two are controllable by software: 1. The red LED on the Arduino D13 pin, and 2. A NeoPixel RGB LED. Currently, only the red LED is supported. ------ ----------------- ----------- SHIELD SAMD5E5 FUNCTION ------ ----------------- ----------- D13 PA16 GPIO output Run from SRAM ============= I bricked my first Metro M4 board because there were problems in the bring-up logic. These problems left the chip in a bad state that was repeated on each reset because the code was written into FLASH and I was unable to ever connect to it again via SWD. To make the bring-up less risky, I added a configuration option to build the code to execution entirely out of SRAM. By default, the setting CONFIG_METRO_M4_RUNFROMFLASH=y is used and the code is built to run out of FLASH. If CONFIG_METRO_M4_RUNFROMSRAM=y is selected instead, then the code is built to run out of SRAM. To use the code in this configuration, the program must be started a little differently: gdb> mon reset gdb> mon halt gdb> load nuttx << Load NuttX into SRAM gdb> file nuttx << Assuming debug symbols are enabled gdb> mon memu32 0x20000000 << Get the address of initial stack gdb> mon reg sp 0x200161c4 << Set the initial stack pointer using this address gdb> mon memu32 0x20000004 << Get the address of __start entry point gdb> mon reg pc 0x20000264 << Set the PC using this address (without bit 0 set) gdb> si << Step in just to make sure everything is okay gdb> [ set breakpoints ] gdb> c << Then continue until you hit a breakpoint Where 0x200161c4 and 0x20000264 are the values of the initial stack and the __start entry point that I read from SRAM Configurations ============== Each Adafruit Metro M4 configuration is maintained in a sub-directory and can be selected as follow: tools/configure.sh [OPTIONS] metro-m4:<subdir> Do 'tools/configure.sh -h' for the list of options. If you are building under Windows with Cygwin, you would need the -c option, for example. Before building, make sure that the PATH environmental variable includes the correct path to the directory than holds your toolchain binaries. And then build NuttX by simply typing the following. At the conclusion of the make, the nuttx binary will reside in an ELF file called, simply, nuttx. make The <subdir> that is provided above as an argument to the tools/configure.sh must be is one of configurations listed in the following paragraph. NOTES: 1. These configurations use the mconf-based configuration tool. To change any of these configurations using that tool, you should: a. Build and install the kconfig-mconf tool. See nuttx/README.txt see additional README.txt files in the NuttX tools repository. b. Execute 'make menuconfig' in nuttx/ in order to start the reconfiguration process. 2. Unless stated otherwise, all configurations generate console output of on SERCOM3 which is available on a Arduino Serial Shield (see the section "Serial Console" above). 3. Unless otherwise stated, the configurations are setup build under Linux with a generic ARM EABI toolchain: Configuration sub-directories ----------------------------- nsh: This configuration directory will built the NuttShell. See NOTES ;for common configuration above and the following: NOTES: 1. The CMCC (Cortex M Cache Controller) is enabled.