Flash Address Fetch Speculation bug migrated from MCF5223x to MCF5225x?

[scifi]

Just now I discovered a strange bug in MCF52259. My program would generate a processor exception which can be avoided when single-stepping through the offending code with the debugger. As soon as I disabled Flash Address Fetch Speculation (bit 6 in FLASHBAR), the bug disappeared.

This is exactly the scenario I went through with the MCF52235. The MCF52235 has a published errata document which documents the bug in silicon and promises to fix it in a future revision. I was hoping this bug would not show up in the MCF52259, but it appears it does. Unfortunately, there is no errata document to be found for the MCF52259 on the Freescale website.

I'm comfortable with keeping Address Fetch Speculation disabled. This is just a warning for everyone interested...

[mjbcswitzerland]

Hi

The 5225X shouldn't have the Flash Address Fetch Speculation bug.

We have a number of M5225X projects, some in production, and all have the workaround disabled. There are no known problems associated with this but I will be watching carefully in case evidence turns up of negative effects.

Is there any information from Freescale which can confirm that there are no risks involved? Have other users experienced such effects?

Regards

Mark

www.uTasker.com

[scifi]

I might have been too quick posting this. I am working with a new BDM adapter and new development tools. There could be some factors I wasn't aware of. For example, it appears that I managed to destroy the flash memory of a MCF52259. The apparent reason: the default flash loader in IAR Embedded Workbench for ColdFire 1.22 isn't good for MCF52259, since it fails to configure flash reference frequency (200 kHz) on this device.

I'll try to run some tests more carefully. I'll post any new findings.

[scifi]

I cannot reproduce the bug. It must have been something else. Sorry for the noise...