Project Logs

Collapse

• detect_GPIO_BASE.sh is broken again...

  Yann Guidon / YGDES • 01/07/2021 at 02:38 • 0 comments

  On a Pi 3 with the 2016 script, the failure is even more burning

  Why can't the RPi guys keep a consistent system or at least preserve backwards compatibility ?

  A new script is required, considering the list I recently found:

  #define PI_01_REG_BASE  0x20000000  // Pi Zero or 1
  #define PI_23_REG_BASE  0x3F000000  // Pi 2 or 3
  #define PI_4_REG_BASE   0xFE000000  // Pi 4

  And don't forget that this is the base of the peripherals block, not the start address of the GPIO block. In these cases, it's:
  

  #define PI_01_REG_BASE  0x20200000  // Pi Zero or 1
  #define PI_23_REG_BASE  0x3F200000  // Pi 2 or 3
  #define PI_4_REG_BASE   0xFE200000  // Pi 4

  thus:

  $ grep gpio /proc/iomem | grep -i -E '20200000|3F200000|FE200000' 
  00000000-00000000 : 3f200000.gpio gpio@7e200000

  It's getting tiring to have to rewrite stuff all the time...

  ---

  
  

  ---

  The latest script is :

  #!/bin/bash
  # version 20210107
  
  export GPIO_BASE=""
  GPIO_LINE=$(grep gpio /proc/iomem) && {
    echo "$GPIO_LINE" | grep -i 20200000 && export GPIO_BASE=0x20200000  # Pi Z or 1
    echo "$GPIO_LINE" | grep -i 3F200000 && export GPIO_BASE=0x3F200000  # Pi 2, 3
    echo "$GPIO_LINE" | grep -i FE200000 && export GPIO_BASE=0xFE200000  # Pi4
  }
  echo GPIO_BASE=$GPIO_BASE
  

  Now how long is it going to work ?
  

• Damned RPi3 !

  Yann Guidon / YGDES • 02/16/2017 at 17:34 • 0 comments

  Trying to turn a LED on with the lib on a RPi3 using the slim 2017-01-11-raspbian-jessie-lite.zip image.

  GPIO_on.sh and GPIO_off.sh work but not the C code.

  The culprit is found easily because detect_GPIO_BASE.sh doesn't return anything (despite being on RPi3).

  pi@raspberrypi:~/ $ cat /proc/iomem 
  00000000-3affffff : System RAM
    00008000-007ea673 : Kernel code
    00862000-009901eb : Kernel data
  3f007000-3f007eff : /soc/dma@7e007000
  3f200000-3f2000b3 : /soc/gpio@7e200000
  3f201000-3f201fff : /soc/uart@7e201000
    3f201000-3f201fff : /soc/uart@7e201000
  3f300000-3f3000ff : /soc/mmc@7e300000
  ...

  The expected keyword is not found ! Instead of bcm2708_gpio, I get /soc/gpio
  

  Another thing is that I forgot to enable the device tree... But a keyword can be added:

  grep -E '/soc/gpio|bcm2708_gpio' /proc/iomem

  Result:

  pi@raspberrypi:~/projets/files $ . detect_GPIO_BASE.sh 
  3f200000-3f2000b3 : /soc/gpio@7e200000
  GPIO_BASE=-DGPIO_BASE=(0x3f200000)
  pi@raspberrypi:~/projets/files $ echo $GPIO_BASE 
  -DGPIO_BASE=(0x3f200000)

  Seems legit... But I still have to test with the device tree.

  ---

  20180118 : don't forget to run the detection script as ROOT !

  pi@raspberrypi$ sudo detect_GPIO_BASE.sh 

  Raspbian now returns invalid values when run as user:

   $ cat /proc/iomem
  00000000-00000000 : System RAM
    00000000-00000000 : Kernel code
    00000000-00000000 : Kernel data
  00000000-00000000 : dwc_otg
  ....

  (I should update my files......)
  

• How to support the Raspberry Pi 2

  Yann Guidon / YGDES • 11/01/2016 at 17:39 • 2 comments

  The RPi2 swapped the venerable BCM2835 for a BCM2836. The Peripheral block is pretty much the same (or so we hope !) but the BCM2835 changed its base address. So that's the only thing that seems to matter for now.

  ---

  http://elinux.org/RPi_GPIO_Code_Samples#Direct_register_access:

  Note: For Raspberry Pi 2, change BCM2708_PERI_BASE to 0x3F000000 for the code to work. 

  My current approach is to provide the base pointer as a compile-time parameter because I write mostly "bare-metal" code for a specific board, which is then embedded somewhere for an indefinite while. I don't do "portable binary code" that gets distributed and executed here and there on all kinds of boards.

  Actually, "simple is best" because the situation is pretty complicated with the current software environment. A pile of legacy code is accumulating... Some distribution provide features that allow detection of the board type and revisions, some other don't. And this depends on the kernel compile flags, as well as the distribution's bells and whistles.

  Another argument in favor of a compile-time parameter is: the pointer is constant and can be optimised, which saves a bit of time and space. Nothing really important but in some cases it can do a little welcome difference.

  ---

  More food for thoughts:

  The following is a shameful copy-paste from http://www.airspayce.com/mikem/bcm2835/
  

  Physical Addresses

  The functions bcm2835_peri_read(), bcm2835_peri_write() and bcm2835_peri_set_bits() are low level peripheral register access functions. They are designed to use physical addresses as described in section 1.2.3 ARM physical addresses of the BCM2835 ARM Peripherals manual. Physical addresses range from 0x20000000 to 0x20FFFFFF for peripherals. The bus addresses for peripherals are set up to map onto the peripheral bus address range starting at 0x7E000000. Thus a peripheral advertised in the manual at bus address 0x7Ennnnnn is available at physical address 0x20nnnnnn.

  On RPI 2, the peripheral addresses are different and the bcm2835 library gets them from reading /proc/device-tree/soc/ranges. This is only availble with recent versions of the kernel on RPI 2.

  (...)

  Raspberry Pi 2 (RPI2)

  For this library to work correctly on RPI2, you MUST have the device tree support enabled in the kernel. You should also ensure you are using the latest version of Linux. The library has been tested on RPI2 with 2015-02-16-raspbian-wheezy and ArchLinuxARM-rpi-2 as of 2015-03-29.

  When device tree suport is enabled, the file /proc/device-tree/soc/ranges will appear in the file system, and the bcm2835 module relies on its presence to correctly run on RPI2 (it is optional for RPI1). Without device tree support enabled and the presence of this file, it will not work on RPI2.

  To enable device tree support:

  sudo raspi-config
  under Advanced Options - enable Device Tree
  Reboot.

  ---

  Since I don't distribute binaries, I don't care about supporting all the HW&SW combinations under the sun. I can relegate the detection to a bash script :-)
  

  Oh and in an older Raspbian running on a B+, I get the necessary info from a different /proc/ file:

  pi@pi ~/GPIO $ cat /proc/iomem 
  00000000-17ffffff : System RAM
    00008000-005bca23 : Kernel code
    ...
  20007000-20007fff : bcm2708_dma.0
    20007000-20007fff : bcm2708_dma
  20100000-201000ff : bcm2708_powerman.0
  20101098-2010109a : bcm2708-i2s.0
    20101098-2010109a : bcm2708-i2s.0
  20200000-20200fff : bcm2708_gpio
  20201000-20201fff : dev:f1
    20201000-20201fff : uart-pl011
  ....

  The base address is obtained with the following command:

  $ echo "(0x"$(grep bcm2708_gpio /proc/iomem|sed -s 's/-.*//')")"
  (0x20200000)

  Turning this into a compile-time define is straight-forward:

  $ GPIO_BASE="-DGPIO_BASE=(0x"$(grep bcm2708_gpio /proc/iomem|sed -s 's/-.*//')")"
  $ echo $GPIO_BASE 
  -DGPIO_BASE=(0x20200000)
  $ gcc $GPIO_BASE -Wall -o test_in_out test_in_out.c && sudo chown root test_in_out && sudo chmod +s test_in_out

  Then all I have to do is check if the base address has already been provided:

  // PI_GPIO.c line...

  Read more »

View all 3 project logs