On 11/25/2010 06:13 AM, 韦东山 wrote:
For some flash, such as SST39VF6401B, it needs 3 step to enter CFI mode.
--- src/flash/nor/cfi.c.origin Thu Nov 11 05:42:00 2010
+++ src/flash/nor/cfi.c Thu Nov 25 12:31:18 2010
@@ -2264,6 +2264,24 @@
struct cfi_flash_bank *cfi_info = bank->driver_priv;
int retval;
+ uint32_t unlock1 = 0x555;
+ uint32_t unlock2 = 0x2aa;
+
+ if (cfi_info->jedec_probe)
+ {
+ unlock1 = 0x5555;
+ unlock2 = 0x2aaa;
+ }
+
+ if ((retval = cfi_send_command(bank, 0xaa, flash_address(bank, 0,
unlock1))) != ERROR_OK)
+ {
+ return retval;
+ }
+ if ((retval = cfi_send_command(bank, 0x55, flash_address(bank, 0,
unlock2))) != ERROR_OK)
+ {
+ return retval;
+ }
+
if ((retval = cfi_send_command(bank, 0x98, flash_address(bank, 0,
address))) != ERROR_OK)
{
return retval;
Please don't apply.
These SST flashes have a broken CFI implementation. If you force them
into CFI mode that way, they report bad values in the CFI tables
(sectors are reported multiple times, summing up to more than the size
of the flash), essentially making them useless.
IMHO, the easiest way to handle those parts is to treat them as non-CFI.
The required additional workarounds in the CFI code are worse than the
required non-CFI table entries.
cu
Michael
_______________________________________________
Openocd-development mailing list
Openocd-development@lists.berlios.de
https://lists.berlios.de/mailman/listinfo/openocd-development
