RE: [PATCH 1/2] spi: spi-mem: Add the spi_set_xfer_bpw function

[Chuanhua Han <chuanhua.han@xxxxxxx>]

> -----Original Message-----
> From: Boris Brezillon <boris.brezillon@xxxxxxxxxxx>
> Sent: 2018年9月28日 15:19
> To: Chuanhua Han <chuanhua.han@xxxxxxx>
> Cc: broonie@xxxxxxxxxx; linux-spi@xxxxxxxxxxxxxxx;
> linux-kernel@xxxxxxxxxxxxxxx; eha@xxxxxxxx
> Subject: Re: [PATCH 1/2] spi: spi-mem: Add the spi_set_xfer_bpw function
> 
> On Fri, 28 Sep 2018 06:59:58 +0000
> Chuanhua Han <chuanhua.han@xxxxxxx> wrote:
> 
> > >
> > > It's still unclear why you need to specify a bits_per_word value,
> > > but if this is needed, it's probably something you want to add to
> > > spi.c, when a message is queued.
> > To specify a specific bits_per_word to be able to use the xspi
> > (32bit) mode of the fsl_dspi module to transfer data, you can look at
> > my PATCH 2/2. Do not add a value in spis.c that takes into account
> > that the value assigned to bits_per_word is decided before the
> > transfer. Thanks for your check and reply!
> 
> I might be wrong, but that's not my understanding of ->bits_per_word.
> To me, it's something that you can use when your *device* (not
> controller) expects non-byte aligned words [1]. The spi-mem protocol is
> definitely designed to work with 1byte large words, so, as I said, I suspect
> you're abusing xfer->bits_per_word to address a controller driver issue.
> 
> [1]https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Feli
> xir.bootlin.com%2Flinux%2Flatest%2Fsource%2Finclude%2Flinux%2Fspi%2Fspi
> .h%23L114&amp;data=02%7C01%7Cchuanhua.han%40nxp.com%7C7ec61d6d
> 7ef741aba84408d62512a9e1%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0
> %7C0%7C636737159420820050&amp;sdata=iwFgXG3yFvH7Ruac7MGCoaP8h
> l2M916m9%2BZeV7nksTg%3D&amp;reserved=0
1. In the dspi driver (spi controller), bits_per_word (dspi->bits_per_word = transfer->bits_per_word) passed from the upper layer (spi-mem.c) is used. 
In this way, I can only assign the appropriate value of transfer->bits_per_word before passing to the controller, that is, the controller driver does not
know the value of bits_per_word, and it will use this value when the upper level sets what value is passed.
2. As I understand, bits_per_word does not exist for non-byte alignment, but for the need to reserve non-byte transmission mode that meets the controller.
3. In addition, now the XSPI of dspi cannot transfer data normally, so this problem needs to be solved. As for the DMA transfer mode, some colleagues will study it.