Re: [RFC PATCH v4 0/2] mtd: nand: vf610_nfc: make use of ->exec_op()

[Boris Brezillon <boris.brezillon@bootlin.com>]

On Thu, 22 Feb 2018 21:29:16 +0100
Stefan Agner <stefan@agner.ch> wrote:

> Fourth revision of the rework patchset to use exec_op for NXP
> Vybrid (and others) NAND Flash Controller. The most important
> change tries to implement a nicer way of handling the endianness
> hack.
> 
> However, this currently fails oobtest and nandbiterrs test. With
> some debugging enabled it looks like this:
> [   42.930460] mtd_oobtest: writing OOBs of whole device
> [   42.935576] vf610_nfc 400e0000.nand: OP_CMD: code 0xff
> [   42.944713] vf610_nfc 400e0000.nand: OP_CMD: code 0x70
> [   42.949955] vf610_nfc 400e0000.nand: OP_CMD: code 0x80
> [   42.955254] vf610_nfc 400e0000.nand: OP_ADDR: col 2048, row 1024
> [   42.961387] vf610_nfc 400e0000.nand: OP_DATA_OUT: len 64, offset 0
> [   42.974332] vf610_nfc 400e0000.nand: OP_CMD: code 0x70
> [   42.983101] vf610_nfc_write_oob, ret -5
> [   42.986986] mtd_oobtest: error: writeoob failed at 0x0
> [   42.992311] mtd_oobtest: error: use_len 2, use_offset 0
> [   42.999054] mtd_oobtest: error -5 occurred
> [   43.003301] =================================================
> 
> It seems that when I set page_access on such granular level I
> do in the current patchset version, then it influences commands
> such as status too... I guess I have to partially reimplement
> nand_exec_prog_page_op..?
> 
> --
> Stefan
> 
> Changes in v4:
> - Rebased to nand/next
> - Simplify filling of address cycles
> - Use accessors for SRAM (vf610_nfc_rd_from_sram/vf610_nfc_wr_to_sram)
> - Use two op-parser patterns to avoid a single command reading and writing
>   in a single operation
> - Implement (read|write)_(page|oob)[_raw] to set page_access
> - Set and clear vf610_nfc_ecc_mode in ecc (read|write)_page
> - Clear/set 16-bit config when 16-bit bus is used and 8-bit access is
>   requested
> 
> Changes in v3:
> - Separate exec_op() callback addition and removal of old callbacks
> - Push data into regs in one function
> - Readd op parser
> - Implement custom read/write page for hardware ECC
> - Rely on generic ecc.write_page_raw
> - Use nand_read_oob_op instead of nand_read_page_op
> 
> Stefan Agner (2):
>   mtd: nand: vf610_nfc: make use of ->exec_op()
>   mtd: nand: vf610_nfc: remove old hooks

Since you're the first one to contribute patches after I moved the NAND
related code to a subdir, I'll ask you to change a bit the prefix for
the next iteration: "mtd: rawnand: " (SPI NAND patches will be prefixed
with "mtd: spinand: ", OneNAND patches are already prefixed with "mtd:
onenand: " and I'll keep "mtd: nand: " for everything touching the
generic/iface-agnostic NAND layer placed in drivers/mtd/nand/).

Thanks,

Boris

> 
>  drivers/mtd/nand/raw/vf610_nfc.c | 642 ++++++++++++++++++++++++---------------
>  1 file changed, 399 insertions(+), 243 deletions(-)
> 



-- 
Boris Brezillon, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com