On 27.06.2008 18:34, Stefan Reinauer wrote:
- Peter Stuge peter@stuge.se [080627 17:48]:
- drop erroneous "unknown flash chip" entries. They badly clash with the "multiple flash chip support" and are no longer required since it's possible to force a flash chip with Peter's last patch.
Thumbs up on this. I thought on doing this in my patch, but I wanted a flashrom mode that was completely equivalent - ie. it would scan for known vendors. But I am fine with removing this right now.
Great. This has been giving me the weirdest errors and outputs over the time, nearly rendering the tool unusable for production use.
In /dev/bios I had an extra array with { VENDOR_ID, "vendorname" } that I could use to match an unknown flash chip against a vendor.
Good idea. Do that per bus and we have a nearly perfect solution.
One idea I had a long time ago was also to always print an vendor/device ID in case the probe command changed the chip's state (ie the read ID is different from the memory read before in non-spi mode)
There is a 1/256th chance that real the vendor ID is identical to the content of the flash chip at that location. Besides that, it is possible that some chips don't leave ID mode until the next reboot, so only the initial run may yield correct results. Still, I think that approach is worthwhile.
Then we could identify the "you have a flashchip in your system, but we don't fully know it" case much better.
Indeed.
Regards, Carl-Daniel