Flash In-Place

This library is intended to support in-place application updates where there is insufficient available flash storage for more than one application partition.

This is aimed at the (Rp2040) Pico (W) which has only 2MBytes of flash.

This library takes a firmware image stored elsewhere in flash, then runs a simple piece of code running from RAM. This copies the new firmware into the application partition then reboots.

This is similar to the Arduino Pico approach, but it doesn’t require any bootloader changes and the critical copier code (in RAM) is much simpler as it doesn’t need to know anything about filing systems; all it does is copy blocks from one region of flash to another.


  1. Verify the source image

    The application handles this e.g. CRC32 checksum.

  2. Build a block list

    Most file systems will store larger files (especially of 2-300KBytes or more) in multiple, non-contiguous regions of flash storage. These regions are known as ‘extents’, and this information is obtained via the IFS::IFilelSystem::fgetextents() call.

    The FlashIP class uses this to build a list, which the copier code then executes from RAM. This keeps the copier code very simple.


    // Get the target partition
    auto part = *Storage::findPartition(Storage::Partition::Type::app);
    // Build the block list
    FlashIP fip;
    if(fip.addFile(part, 0, "firmware.bin")) {
        // Copy the image and reboot
  3. Write the block list to flash

    This code runs exclusively in RAM. All interrupts and DMA are disabled first. The block list is then iterated with a read/erase/program procedure.

  4. Reboot



There is no recovery available if the update operation is interrupted part-way. If physical device access is available then firmware can be re-flashed via USB.

Writing a second-stage bootloader and storing the block list in FLASH would mitigate this issue. For example, we’d reserve the first two flash sectors for this, with the application itself relocated. Each block-list entry will be invalidated once complete so the bootloader need only write incomplete blocks.


Using a compressed firmware image would require e.g. uzlib to be located in RAM. We can add a build variable to make that happen.

A quick test on the release HttpServer_ConfigNetwork image using gzip reduces the size from 341136 to 280084 bytes (82% of the original size).


Note that where updating via local network, encryption may not be necessary.

If encryption is used, the application would need to decrypt that first. Typically this is done as the file is received, before writing the data to the local file.


Used by

SoC support

  • host

  • rp2040