[RE: CFI Devices in x8 mode ]

[Eric]

"Alex Lennon" <ajlennon at arcom.co.uk> wrote:

>> The 8-bit buswidth probe doesn't attempt to probe for a 
>> device type X16 any more. It should. Someone gave me a 
>> patch for that a few days ago, I think - but I can't find 
>> it. I've just committed what I think is the correct fix to
>> CVS.

You may be referring to my proposed patch:

--- cfi_probe.c.orig	Wed Mar 21 15:59:05 2001
+++ cfi_probe.c	Thu Mar 22 13:12:43 2001
@@ -235,6 +235,11 @@
 		cfi->device_type = CFI_DEVICETYPE_X8;
 		index = cfi_probe_chip(map,base,chips,cfi);
 		if (index>=0) return index;
+
+		/* if this did not work, attempt again in x16 mode */
+		cfi->device_type = CFI_DEVICETYPE_X16;
+		index = cfi_probe_chip(map,base,chips,cfi);
+		if (index>=0) return index;
 		break;			
 #endif


> Sorry - I think I'm misunderstanding - I've got an eight bit 
> device here but the code looks as though its accessing it as 
> a 16-bit device from what I can see in the spec.

[ ... ]

> If I'm reading this stuff right then the cfi_cfi_probe() 
> code doesn't seems to be taking the base offset into account ?

No, I believe it attempts to calculate the offset properly in 
cfi_probe.c cfi_check_qry_or_id():

  int osf = cfi->interleave * cfi->device_type;	// scale factor

  [ ... ]

  cfi_read(map,base+osf*0x10)==cfi_build_cmd('Q',map,cfi->interleave) 

So it appears the calculation is done properly.  What is NOT done 
properly as of the last cfi_probe.c revision is that cfi->device_type
is unconditionally set to DEVICETYPE_X16 for all 8-bit-wide busses in
cfi_probe_new_chip().  The above patch simply falls-back to probing
in DEVICETYPE_X8 mode if the X16 fails.  I have not seen David's
patch in CVS as of yet, but if it does the same it should be fine--
I cannot speak to whether or not this is the appropriate solution, but
it does work in my case (8-bit bus with 16-bit chip in X8 mode).

E



____________________________________________________________________
Get free email and a permanent address at http://www.netaddress.com/?N=1


To unsubscribe, send "unsubscribe mtd" to majordomo at infradead.org