The following three patches should be applied on top of the previous cfi_flash cleanup patches. With these in place, I can access the on-board flash on the ATNGW100 using the CFI driver, eliminating the final reason why I haven't submitted the NGW100 board patches for inclusion in mainline.
I've tested it by programming a 7MB JFFS2 image into the on-board AT49BV642D flash. It went almost successfully -- I got two erase timeouts and one programming timeout. On the second try, it worked with no problems.
It looks like I have a bug in my timing code, since "sleep 10" terminates early from time to time as well. So I feel pretty confident that the CFI driver is doing its job properly -- the timeouts are simply triggered too early every now and then.
The last patch in the series changes behaviour, so it probably needs more extensive testing than the others. But this patch is essential for the AT49BV642D to work -- without it, the geometry reversal fixup is incorrectly triggered, and the erase regions become wrong.
I think the new behaviour makes more sense, especially since the code uses the device_id to determine whether or not to apply the fixup; it's IMO wrong to depend on the device_id without also checking the manufacturer_id.
Haavard Skinnemoen (3): cfi_flash: Read whole QRY structure in one go cfi_flash: Add cmdset-specific init functions cfi_flash: Add manufacturer-specific fixups
drivers/mtd/cfi_flash.c | 377 +++++++++++++++++++++++++++++------------------ 1 files changed, 234 insertions(+), 143 deletions(-)