ARM Technical Support Knowledge Articles

Avoiding spurious ECC errors in Cortex-M7 TCM

Applies to: Cortex-M7

Scenario

The Cortex-M7 processor can optionally be implemented with local private RAM, known as Tightly-Coupled Memory (TCM), occupying address ranges based at address 0x0 for Instruction TCM and at address 0x20000000 for Data TCM.

The TCMs can optionally be implemented with Error-Correcting Code (ECC) logic included. This methodology increases the word width of the memory blocks, so that a word contains both a data value (32- or 64-bits) and additional ECC bits. The ECC bit values are calculated when the memory word is written, and are verified when the word is read back.

The Cortex-M7 processor will make speculative reads to memory in a number of circumstances, and this behavior cannot be suppressed.

If the processor makes a speculative read to an uninitialized word in TCM, and ECC has been implemented in this chip, the ECC bits may not be correctly set for this word, resulting in the system attempting to perform error correction. If the uninitialized data and ECC is perceived to represent a double bit error, this could trigger the system to initiate some system-dependent major fault handling sequence.

Answer

Programmers writing boot code for a Cortex-M7 need to the check documentation of the chip which they are programming, in order to verify whether the chip contains TCM with ECC.

Where TCMs are implemented with ECC enabled in a chip, boot code should pre-initialize all uninitialized TCM addresses, for example by writing all zeros, to avoid spurious ECC fails on any speculative reads.

Rate this article

[Bad]
|
|
[Good]
Disagree? Move your mouse over the bar and click

Did you find this article helpful? Yes No

How can we improve this article?

Link to this article
Copyright © 2011 ARM Limited. All rights reserved. External (Open), Non-Confidential