[RFC] onenand detect logic

[RFC] onenand detect logic

[Nishanth Menon]

Hi Folks,
as a disclaimer, the following discussion is entirely a thought
experiment ;)

While discussing with Rohit C on supporting a different onenand on git
kernel, he enlightened me that the GPMC tracking and detecting onenand
CS based address is flaky.
Current Logic (with a highly deserving HACK alert comment) does this:
for each CS(){
   if (CONFIG_7 reg  ==0x20000000)
	then assume it is onenand.
}

Here is a alternative proposal (majority from Rohit):

for each CS(){
	if (configured as a NOR type memory){
		write 0x90 to CONFIG_7 0x0 offset /* 0x90 0r 0x98 CFI command will
ensure that the memory is in non-read state, but in onenand it is a
readonly location */
		read  manu id and device id from onenand regs offsets from CONFIG_7
		if matched as that for onenand, select same
	}
}
Caveat: does not detect 2 onenands..

Question: Any better way of doing this dynamically and a bit more elegantly?

Regards,
Nishanth Menon
		

RE: [RFC] onenand detect logic

[Kyungmin Park]

Hi,

> 
> Hi Folks,
> as a disclaimer, the following discussion is entirely a thought
> experiment ;)
> 
> While discussing with Rohit C on supporting a different onenand on git
> kernel, he enlightened me that the GPMC tracking and detecting onenand
> CS based address is flaky.
> Current Logic (with a highly deserving HACK alert comment) does this:
> for each CS(){
>    if (CONFIG_7 reg  ==0x20000000)
> 	then assume it is onenand.
> }
> 
> Here is a alternative proposal (majority from Rohit):
> 
> for each CS(){
> 	if (configured as a NOR type memory){
> 		write 0x90 to CONFIG_7 0x0 offset /* 0x90 0r 0x98 CFI command will
> ensure that the memory is in non-read state, but in onenand it is a
> readonly location */
> 		read  manu id and device id from onenand regs offsets from CONFIG_7

You should also check the maf id and device id in onenand regs additionally.
Please refer to the onenand_probe() in drivers/mtd/onenand/onenand_base.c
Is it also safe to write 0xF0 (OneNAND Reset) CFI command to NOR?

> 		if matched as that for onenand, select same
> 	}
> }
> Caveat: does not detect 2 onenands..

Why don't you add the more sdp_onenand_data if required for second onenand chip.
I mean if you find the OneNAND at matched CS, continue to find another CS for second onenand chip.

Even though, you assign the wrong CS value, onenand_scan and onenanc_probe will be failed, so don't worry about assigning the wrong
CS.

> 
> Question: Any better way of doing this dynamically and a bit more elegantly?
>