sergiotarxz/nuttx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
a560d70e7a
nuttx/configs/esp32-core/README.txt
Gregory Nutt 23d2915179 Update README
2016-11-15 13:25:30 -06:00

619 lines
23 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

README for the Expressif ESP32 Core board (V2)
==============================================
The ESP32 is a dual-core system from Expressif with two Harvard
architecture Xtensa LX6 CPUs. All embedded memory, external memory and
peripherals are located on the data bus and/or the instruction bus of
these CPUs. With some minor exceptions, the address mapping of two CPUs
is symmetric, meaning they use the same addresses to access the same
memory. Multiple peripherals in the system can access embedded memory via
DMA.
The two CPUs are named "PRO_CPU" and "APP_CPU" (for "protocol" and
"application"), however for most purposes the two CPUs are
interchangeable.
Contents
========
o STATUS
o ESP32 Features
o ESP32 Toolchain
o Memory Map
o Serial Console
o Buttons and LEDs
o SMP
o OpenOCD for the ESP32
o Executing and Debugging from FLASH and IRAM
o Configurations
o Things to Do
STATUS
======
The basic port is underway. No testing has yet been performed.
ESP32 Features
==============
* Address Space
- Symmetric address mapping
- 4 GB (32-bit) address space for both data bus and instruction bus
- 1296 KB embedded memory address space
- 19704 KB external memory address space
- 512 KB peripheral address space
- Some embedded and external memory regions can be accessed by either
data bus or instruction bus
- 328 KB DMA address space
* Embedded Memory
- 448 KB Internal ROM
- 520 KB Internal SRAM
- 8 KB RTC FAST Memory
- 8 KB RTC SLOW Memory
* External Memory
Off-chip SPI memory can be mapped into the available address space as
external memory. Parts of the embedded memory can be used as transparent
cache for this external memory.
- Supports up to 16 MB off-Chip SPI Flash.
- Supports up to 8 MB off-Chip SPI SRAM.
* Peripherals
- 41 peripherals
* DMA
- 13 modules are capable of DMA operation
ESP32 Toolchain
===============
You must use the custom Xtensa toolchain in order to build the ESP32 Core
BSP. The steps to build toolchain with crosstool-NG on Linux are as
follows:
git clone -b xtensa-1.22.x https://github.com/espressif/crosstool-NG.git
cd crosstool-NG
./bootstrap && ./configure --prefix=$PWD && make install
./ct-ng xtensa-esp32-elf
./ct-ng build
chmod -R u+w builds/xtensa-esp32-elf
These steps are given in setup guide in ESP-IDF repository:
https://github.com/espressif/esp-idf/blob/master/docs/linux-setup.rst#alternative-step-1-compile-the-toolchain-from-source-using-crosstool-ng
NOTE: The xtensa-esp32-elf configuration is only available in the
xtensa-1.22.x branch.
Memory Map
==========
Address Mapping
----------- ---------- ---------- --------------- ---------------
BUS TYPE START LAST DESCRIPTION NOTES
----------- ---------- ---------- --------------- ---------------
0x00000000 0x3F3FFFFF Reserved
Data 0x3F400000 0x3F7FFFFF External Memory
Data 0x3F800000 0x3FBFFFFF External Memory
0x3FC00000 0x3FEFFFFF Reserved
Data 0x3FF00000 0x3FF7FFFF Peripheral
Data 0x3FF80000 0x3FFFFFFF Embedded Memory
Instruction 0x40000000 0x400C1FFF Embedded Memory
Instruction 0x400C2000 0x40BFFFFF External Memory
0x40C00000 0x4FFFFFFF Reserved
Data / 0x50000000 0x50001FFF Embedded Memory
Instruction
0x50002000 0xFFFFFFFF Reserved
Embedded Memory
----------- ---------- ---------- --------------- ---------------
BUS TYPE START LAST DESCRIPTION NOTES
----------- ---------- ---------- --------------- ---------------
Data 0x3ff80000 0x3ff81fff RTC FAST Memory PRO_CPU Only
0x3ff82000 0x3ff8ffff Reserved
Data 0x3ff90000 0x3ff9ffff Internal ROM 1
0x3ffa0000 0x3ffadfff Reserved
Data 0x3ffae000 0x3ffdffff Internal SRAM 2 DMA
Data 0x3ffe0000 0x3fffffff Internal SRAM 1 DMA
Boundary Address
----------- ---------- ---------- --------------- ---------------
BUS TYPE START LAST DESCRIPTION NOTES
----------- ---------- ---------- --------------- ---------------
Instruction 0x40000000 0x40007fff Internal ROM 0 Remap
Instruction 0x40008000 0x4005ffff Internal ROM 0
0x40060000 0x4006ffff Reserved
Instruction 0x40070000 0x4007ffff Internal SRAM 0 Cache
Instruction 0x40080000 0x4009ffff Internal SRAM 0
Instruction 0x400a0000 0x400affff Internal SRAM 1
Instruction 0x400b0000 0x400b7FFF Internal SRAM 1 Remap
Instruction 0x400b8000 0x400bffff Internal SRAM 1
Instruction 0x400c0000 0x400c1FFF RTC FAST Memory PRO_CPU Only
Data / 0x50000000 0x50001fff RTC SLOW Memory
Instruction
External Memory
----------- ---------- ---------- --------------- ---------------
BUS TYPE START LAST DESCRIPTION NOTES
----------- ---------- ---------- --------------- ---------------
Data 0x3f400000 0x3f7fffff External Flash Read
Data 0x3f800000 0x3fbfffff External SRAM Read and Write
Boundary Address
----------------
Instruction 0x400c2000 0x40bfffff 11512 KB External Flash Read
Linker Segments
------------------ ---------- ---------- ---- ----------------------------
DESCRIPTION START END ATTR LINKER SEGMENT NAME
------------------ ---------- ---------- ---- ----------------------------
FLASH mapped data: 0x3f400010 0x3fc00010 R drom0_0_seg
- .rodata
- Constructors/destructors
COMMON data RAM: 0x3ffb0000 0x40000000 RW dram0_0_seg (NOTE 1,2)
- .bss/.data
IRAM for PRO cpu: 0x40080000 0x400a0000 RX iram0_0_seg
- Interrupt Vectors
- Low level handlers
- Xtensa/Expressif libraries
RTC fast memory: 0x400c0000 0x400c2000 RWX rtc_iram_seg (PRO_CPU only)
- .rtc.text (unused?)
FLASH: 0x400d0018 0x40400018 RX iram0_2_seg (actually FLASH)
- .text
RTC slow memory: 0x50000000 0x50001000 RW rtc_slow_seg (NOTE 3)
- .rtc.data/rodata (unused?)
NOTE 1: Linker script will reserve space at the beginning of the segment
for BT and at the end for trace memory.
NOTE 2: Heap enads at the top of dram_0_seg
NOTE 3: Linker script will reserve space at the beginning of the segment
for co-processor reserve memory and at the end for ULP coprocessor
reserve memory.
Serial Console
==============
UART0 is, by default, the serial console. It connects to the on-board
CP2102 converter and is available on the USB connector USB CON8 (J1).
Buttons and LEDs
================
Buttons
-------
There are two buttons labeled Boot and EN. The EN button is not available
to software. It pulls the chip enable line that doubles as a reset line.
The BOOT button is connected to IO0. On reset it is used as a strapping
pin to determine whether the chip boots normally or into the serial
bootloader. After reset, however, the BOOT button can be used for software
input.
LEDs
----
There are several on-board LEDs for that indicate the presence of power
and USB activity. None of these are available for use by sofware.
SMP
===
The ESP32 has 2 CPUs. Support is included for testing an SMP configuration.
That configuration is still not yet ready for usage but can be enabled with
the following configuration settings:
RTOS Features -> Tasks and Scheduling
CONFIG_SPINLOCK=y
CONFIG_SMP=y
CONFIG_SMP_NCPUS=2
CONFIG_SMP_IDLETHREAD_STACKSIZE=2048
Open Issues:
1. Currently all device interrupts are handled on the PRO CPU only. Critical
sections will attempt to disable interrupts but will now disable interrupts
only on the current CPU (which may not be CPU0). Perhaps that should be a
spinlock to prohibit execution of interrupts on CPU0 when other CPUs are in
a critical section?
2. Cache Issues. I have not though about this yet, but certainly caching is
an issue in an SMP system:
- Cache coherency. Are there separate caches for each CPU? Or a single
shared cache? If the are separate then keep the caches coherent will
be an issue.
- Caching MAY interfere with spinlocks as they are currently implemented.
Waiting on a cached copy of the spinlock may result in a hang or a
failure to wait.
3. Assertions. On a fatal assertions, other CPUs need to be stopped.
OpenOCD for the ESP32
=====================
First you in need some debug environment which would be a JTAG emulator
and the ESP32 OpenOCD software which is available here:
https://github.com/espressif/openocd-esp32
OpenOCD Documentation
---------------------
There is on overiew of the use of OpenOCD here:
https://dl.espressif.com/doc/esp-idf/latest/openocd.html
This document is also available in ESP-IDF source tree in docs
directory (https://github.com/espressif/esp-idf).
OpenOCD Configuration File
--------------------------
A template ESP32 OpenOCD configuration file is provided in
ESP-IDF docs directory (esp32.cfg). Since you are not using
FreeRTOS, you will need to uncomment the line:
set ESP32_RTOS none
in the OpenOCD configuration file. You will also need to change
the source line from:
find interface/ftdi/tumpa.cfg
to reflect the physical JTAG adapter connected.
NOTE: A copy of this OpenOCD configuration file available in the NuttX
source tree at nuttx/config/esp32-core/scripts/esp32.cfg.. It has these
modifications:
- The referenced "set ESP32_RTOS none" line has been uncommented
- The "ind interface/ftdi/tumpa.cfg". This means that you will
need to specify the interface configuration file on the OpenOCD
command line.
General OpenOCD build instructions
----------------------------------
Installing OpenOCD. The sources for the ESP32-enabled variant of
OpenOCD are available from Espressifs Github. To download the source,
use the following commands:
git clone https://github.com/espressif/openocd-esp32.git
cd openocd-esp32
git submodule init
git submodule update
Then look at the README and the docs/INSTALL.txt files in the
openocd-esp32 directory for further instructions. There area
separate README files for Linux/Cygwin, OSX, and Windows. Here
is what I ended up doing (under Linux):
cd openocd-esp32
./bootstrap
./configure
make
If you do not do the install step, then you will have a localhost
version of the OpenOCD binary at openocd-esp32/src.
Starting the OpenOCD Server
---------------------------
- cd to openocd-esp32 directory
- copy the modified esp32.cfg script to this directory
Then start OpenOCD by executing a command like the following. Here
I assume that:
- You did not install OpenOCD; binararies are avalable at
openocd-esp32/src and interface scripts are in
openocd-eps32/tcl/interface
- I select the configuration for the Olimex ARM-USB-OCD
debugger.
Then the command to start OpenOCD is:
sudo ./src/openocd -s ./tcl -f tcl/interface/ftdi/olimex-arm-usb-ocd.cfg -f ./esp32.cfg
I then see:
Open On-Chip Debugger 0.10.0-dev-g3098897 (2016-11-14-12:19)
Licensed under GNU GPL v2
For bug reports, read
http://openocd.org/doc/doxygen/bugs.html
adapter speed: 200 kHz
force hard breakpoints
Info : clock speed 200 kHz
Info : JTAG tap: esp32.cpu0 tap/device found: 0x120034e5 (mfg: 0x272 (Tensilica), part: 0x2003, ver: 0x1)
Info : JTAG tap: esp32.cpu1 tap/device found: 0x120034e5 (mfg: 0x272 (Tensilica), part: 0x2003, ver: 0x1)
Info : esp32.cpu0: Debug controller was reset (pwrstat=0x5F, after clear 0x0F).
Info : esp32.cpu0: Core was reset (pwrstat=0x5F, after clear 0x0F).
Connecting a debugger to OpenOCD
--------------------------------
OpenOCD should now be ready to accept gdb connections. If you have
compiled the ESP32 toolchain using Crosstool-NG, or if you have
downloaded a precompiled toolchain from the Espressif website, you
should already have xtensa-esp32-elf-gdb, a version of gdb that can
be used for this
First, make sure the project you want to debug is compiled and
flashed into the ESP32’s SPI flash. Then, in a different console
than OpenOCD is running in, invoke gdb. For example, for the
template app, you would do this like such:
cd nuttx
xtensa-esp32-elf-gdb -ex 'target remote localhost:3333' nuttx
This should give you a gdb prompt.
Breakpoints
-----------
You can set up to 2 hardware breakpoints, which can be anywhere in the
address space. Also 2 hardware watchpoints.
The openocd esp32.cfg file currently forces gdb to use hardware
breakpoints, I believe because software breakpoints (or, at least, the
memory map for automatically choosing them) aren't implemented yet
(as of 2016-11-14).
JTAG Emulator
-------------
The documentation indicates that you need to use an external JTAG
like the TIAO USB Multi-protocol Adapter and the Flyswatter2.
The instructions at http://www.esp32.com/viewtopic.php?t=381 show
use of an FTDI C232HM-DDHSL-0 USB 2.0 high speed to MPSSE cable.
The ESP32 Core v2 board has no on board JTAG connector. It will
be necessary to make a cable or some other board to connect a JTAG
emulator. Refer to http://www.esp32.com/viewtopic.php?t=381 "How
to debug ESP32 with JTAG / OpenOCD / GDB 1st part connect the
hardware."
Relevant pin-out:
-------- ----------
PIN JTAG
LABEL FUNCTION
-------- ----------
IO14 TMS
IO12 TDI
GND GND
IO13 TCK
-------- ----------
IO15 TDO
-------- ----------
You can find the mapping of JTAG signals to ESP32 GPIO numbers in
"ESP32 Pin List" document found here:
http://espressif.com/en/support/download/documents?keys=&field_type_tid%5B%5D=13
I put the ESP32 on a prototyping board and used a standard JTAG 20-pin
connector with an older Olimex JTAG that I had. Here is how I wired
the 20-pin connector:
----------------- ----------
20-PIN JTAG ESP32 PIN
CONNECTOR LABEL
----------------- ----------
1 VREF INPUT 3V3
3 nTRST OUTPUT N/C
5 TDI OUTPUT IO12
7 TMS OUTPUT IO14
9 TCLK OUTPUT IO13
11 RTCK INPUT N/C
13 TDO INPUT IO15
15 RESET I/O N/C
17 DBGRQ OUTPUT N/C
19 5V OUTPUT N/C
------------ ----------
2 VCC INPUT 3V3
4 GND N/A GND
6 GND N/A GND
8 GND N/A GND
10 GND N/A GND
12 GND N/A GND
14 GND N/A GND
16 GND N/A GND
18 GND N/A GND
20 GND N/A GND
------------ ----------
Executing and Debugging from FLASH and IRAM
===========================================
Enable Debug Symbols
--------------------
To debug with GDB, you will need to enable symbols in the build. You do this
with 'make menuconfig' then selecting:
- "Build Setup" -> "Debug Options" -> "Generate Debug Symbols"
And, to make debugging easier, also disable optimizations. This will make
your code a lot bigger:
- "Build Setup" -> "Optimization Level" -> "Suppress Optimization"
FLASH
-----
OpenOCD currently doesn't have a FLASH driver for ESP32, so you can load
code into IRAM only via JTAG. FLASH-resident sections like .FLASH.rodata
will fail to load. The bootloader in ROM doesn't parse ELF, so any imag
which is bootloaded from FLASH has to be converted into a custom image
format first.
The tool esp-idf uses for flashing is a command line Python tool called
"esptool.py" which talks to a serial bootloader in ROM. A version is
supplied in the esp-idf codebase in components/esptool_py/esptool, the
"upstream" for that tool is here:
https://github.com/espressif/esptool/pull/121
The master branch for esptool.py is currently ESP8266-only (as of 2016-11-14),
this PR has the ESP32 support which still needs some final tidying up before
it's
merged.
To FLASH an ELF via the command line is a two step process, something like
this:
esptool.py --chip esp32 elf2image --flash_mode dio --flash_size 4MB -o ./nuttx.bin nuttx.elf
esptool.py --chip esp32 --port COMx write_flash 0x1000 bootloader.bin 0x4000 partition_table.bin 0x10000 nuttx.bin
The first step converts an ELF image into an ESP32-compatible binary
image format, and the second step flashes it (along with bootloader image and
partition table binary.)
To put the ESP32 into serial flashing mode, it needs to be reset with IO0 held
low. On the Core boards this can be accomplished by holding the button marked
"Boot" and pressing then releasing the button marked "EN". Actually, esptool.py
can enter bootloader mode automatically (via RTS/DTR control lines), but
unfortunately a timing interaction between the Windows CP2012 driver and the
hardware means this doesn't currently work on Windows.
Secondary Boot Loader / Partition Table
---------------------------------------
See https://github.com/espressif/esp-idf/tree/master/components/bootloader
and https://github.com/espressif/esp-idf/tree/master/components/partition_table.
Running from IRAM
-----------------
Running from IRAM is a good debug option. You should be able to load the ELF directly via JTAG in this case, and you may not need the bootloader. The one "gotcha" for needing the bootloader is disabling the initial watchdog, there is code in bootloader_start.c that does this.
It is possible to skip the secondary bootloader and run out of IRAM using
only the primary bootloader if your application of small enough (< 128KiB code,
<180KiB data), then you can simplify initial bring-up by avoiding second stage
bootloader. Your application will be loaded into IRAM using first stage
bootloader present in ESP32 ROM. To achieve this, you need two things:
1. Have a linker script which places all code into IRAM and all data into
IRAM/DRAM
2. Use "esptool.py" utility found in ESP-IDF to convert application .elf
file into binary format which can be loaded by first stage bootloader.
NuttX supports a configuration option, CONFIG_ESP32CORE_RUN_IRAM, that may be
selected for execution from IRAM. This option simply selects the correct
linker script for IRAM execution.
Again you would need to link the ELF file and convert it to binary format suitable
for flashing into the board. The command should to convert ELF file to binary
image looks as follows:
python esp-idf/components/esptool_py/esptool/esptool.py --chip esp32 elf2image --flash_mode "dio" --flash_freq "40m" --flash_size "2MB" -o nuttx.bin nuttx
To flash binary image to your development board, use the same esptool.py utility:
python esp-idf/components/esptool_py/esptool/esptool.py --chip esp32 --port /dev/ttyUSB0 --baud 921600 write_flash -z --flash_mode dio --flash_freq 40m --flash_size 2MB 0x1000 nuttx.bin
The argument before app.bin (0x1000) indicates the offset in flash where binary
will be written. ROM bootloader expects to find an application (or second stage
bootloader) image at offset 0x1000, so we are writing the binary there.
Clocking
--------
Right now, the NuttX port depends on the bootloader to initialize hardware,
including basic (slow) clocking. If I had the clock configuration logic,
would I be able to run directly out of IRAM without a bootloader? That
might be a simpler bring-up.
Configurations
==============
Common Configuration Information
--------------------------------
Each ESP32 core configuration is maintained in sub-directories and
can be selected as follow:
cd tools
./configure.sh esp32-core/<subdir>
cd -
make oldconfig
. ./setenv.sh
Before sourcing the setenv.sh file above, you should examine it and
perform edits as necessary so that TOOLCHAIN_BIN is the correct path to
the directory than holds your toolchain binaries.
If this is a Windows native build, then configure.bat should be used
instead of configure.sh:
configure.bat esp32-core\<subdir>
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 oldconfig
make
The <subdir> that is provided above as an argument to the
tools/configure.sh must be is one of the directories listed below.
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 on UART0 (see the "Serial Console" section above).
3. By default, these configurations assume a 40MHz crystal on-
board:
CONFIG_ESP32CORE_XTAL_40MZ=y
# CONFIG_ESP32CORE_XTAL_26MHz is not set
4. Default configurations are set to run from FLASH. You will need
to set CONFIG_ESP32CORE_RUN_IRAM=y for now (see the " Executing
and Debugging from FLASH and IRAM" section above).
To select this option, do 'make menuconfig'. Then you can find
the selection under the "Board Selection" menu as "Run from IRAM".
Configuration sub-directories
-----------------------------
nsh:
Configures the NuttShell (nsh) located at apps/examples/nsh.
NOTES:
smp:
Another NSH configuration, similar to nsh, but also enables
SMP operation.
NOTES:
Things to Do
============
1. There is no support for an interrupt stack yet.
2. There is no clock intialization logic in place. This depends on logic in
Expressif libriaries. The board comes up using that basic 40 Mhz crystal
for clocking. Getting to 80 MHz will require clocking initialization in
esp32_clockconfig.c.
3. I did not implement the lazy co-processor save logic supported by Xtensa.
That logic works like this:
a. CPENABLE is set to zero on each context switch, disabling all co-
processors.
b. If/when the task attempts to use the disabled co-processor, an
exception occurs
c. The co-processor exception handler re-enables the co-processor.
Instead, the NuttX logic saves and restores CPENABLE on each context
switch. This has disadvantages in that (1) co-processor context will
be saved and restored even if the co-processor was never used, and (2)
tasks must explicitly enable and disable co-processors.
4. Currently the Xtensa port copies register state save information from
the stack into the TCB. A more efficient alternative would be to just
save a pointer to a register state save area in the TCB. This would
add some complexity to signal handling and also also the the
up_initialstate(). But the performance improvement might be worth
the effort.
5. See SMP-related issues above
6. See OpenOCD for the ESP32 above
Reference in New Issue View Git Blame Copy Permalink