On Tue, 15 Dec 2009, Carl-Daniel Hailfinger wrote:
On 15.12.2009 23:04, Maciej Pijanka wrote:
On Tue, 15 Dec 2009, Carl-Daniel Hailfinger wrote:
We have more than one chip which gets horribly confused by probe sequences for other chips.
Sending a reset command to the detected chip (and sleeping a bit) before performing any read/write/erase action could fix a number of problems we're seeing right now. This could be achieved by adding a .reset function to struct flashchip.
Any ideas/approximations how much it will lengthen probe sequence? (ie, send detect then reset-with-wait to each probed chip)
No slowdown on probe. The idea is to run reset once after all probes are done.
Sounds good unless some of issues are related because probe-of-other-chip confused it and you issue this-chip reset routine, right? If reset is enough good to work even if chip got slightly mad already then fine otherwise an delay might be needed, i found something like that with nic3com and one atmel chip family, if given autodetect then they fail, if chip is given thus skipping autodetect but still fired this chip probe, works perfectly.
If you have patch i have that hardware running so can test if it improves anything in this case too.
best regards Maciej