Intel® Agilex™ Configuration User Guide

ID 683673
Date 12/05/2022
Public

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

5.4.2.4. Modifying the List of Application Images

The SDM uses the configuration pointer block to determine priority of application images.

The pointer block operates by taking into account the following characteristics of quad SPI flash memory:
  • On a sector erase, all the sector flash bits become 1’s.
  • A program operation can only turn 1’s into 0’s.
The pointer block contains an array of values which have the following meaning:
  • All 1’s – the entry is unused. The client can write a pointer to this entry. This is the state after a quad SPI erase operation occurs on the pointer block.
  • All 0’s – the entry has been previously used and then canceled.
  • A combination of 1's and 0's – a valid pointer to an application image.

When the configuration pointer block is erased, all entries are marked as unused. To add an application image to the list, the client finds the first unused location and writes the application image address to this location. To remove an application image from the list, the client finds the application image address in the pointer block list and writes all 0's to this address.

If the configuration pointer block runs out of space for new application images, the client compresses the pointer block by completing the following actions:

  1. Read all the valid entries from the configuration
  2. Erase the pointer block
  3. Add all previously valid entries
  4. Add the address of the new image

When using HPS to manage RSU, both the U-Boot and LIBRSU clients implement the block compression. For designs that drive RSU from FPGA logic, you can implement pointer block compression many different ways, including Nios® II code, a scripting language, or a state machine.

Pointer block compression does not occur frequently because the pointer block has up to 508 available entries.

There are two configuration pointer blocks: a primary (CPB0) and a backup (CPB1). Two blocks enable the list of application images to be protected if a power failure occurs just after erasing one of them. When a CPB is erased and re-created, the header is written last. The CPB header is checked prior to use to prevent accidental use if a power failure occurred. For more information, refer to the Configuration Pointer Block Layout topic. When compressing, the client compresses (erases and rewrites) the primary CPB completely. Once the primary CPB is valid, it is safe to modify the secondary CPB. When rewriting, the magic number at the start of a CPB is the last word written in the CPB. (After this number is written only image pointer slot values can be changed.)

When the client writes the application image to flash, it ensures that the pointers within the main image pointer of its first signature block are updated to point to the correct locations in flash. When using HPS to manage RSU, both the U-Boot and LIBRSU clients implement the required pointer updates.
Note: In order to successfully perform CPB compression, the HPS software (U-Boot or Linux) must be configured to have a QSPI erase granularity of 32 KB or less. When configured with a coarser erase granularity (like 64 KB for example), the operation fails. All supported flash devices offer erase granularities of 4 KB, 32 KB, and 64 KB, and the default for the current HPS software is 4 KB.