On Tue, May 13, 2008 at 06:01:12PM -0400, Ward Vandewege wrote:
Hmm. I have a hardware-modded board with 2 chips and a switch. The mod might be a bit flakey, I see odd stuff sometimes (it will erase both chips regardless of position of switch, but it always only programs the right chip).
To make switching more reliable I suggest a different mod;
Populate R89 and R130 as before with 100k resistors but also populate Q4 and Q5 with BC847 transistors, R86 and R389 with 100k resistors and R90 and R91 with 10k resistors. Then short Q2-2 and Q2-3 together and short Q43-2 and Q43-3 together. Finally, the switch center goes to VCC, and the two outer switch connections go to the south end of R86 and R389. (The ends facing away from the flash chips.)
I have tested this with several meters of cable to the switch without issues.
This works better because the high-speed CS# signal stays in the Q4 and Q5 transistors on the mainboard, making the mod much less sensitive to interference.
So the fact that flashrom sees 2 chips might be an artifact of that problem, but maybe not...
It is not.
Switch in position 1:
# ./flashrom -m m57sli -V
..
Multiple flash chips were detected: MX25L4005 unknown SPI chip
And after flipping the switch:
# ./flashrom -m m57sli -V
..
Multiple flash chips were detected: SST25VF016B unknown SPI chip
Well, looks like you are right, no behavioural changes, this is an unpatched r3303:
Yes. This was introduced with the code that looks for multiple flash chips.
We should remove those unknown flash chip entries and have a different user interface to do what they did.
The probe function will always find the unknown flash chips. A temporary kludge is to not allow them to be detected other than as the first flash chip that is found.
//Peter