[U-Boot] x86: SPI flash broken with SPI NOR

[Bin Meng]

Hi Vignesh,

On Wed, Jun 5, 2019 at 9:40 PM Bin Meng <bmeng.cn at gmail.com> wrote:
>
> Hi Vignesh,
>
> On Sat, Apr 27, 2019 at 12:26 AM Vignesh Raghavendra <vigneshr at ti.com> wrote:
> >
> > Bin,
> >
> > On 26/04/19 7:33 PM, Bin Meng wrote:
> > > Hi Simon,
> > >
> > > On Fri, Apr 26, 2019 at 7:26 AM Simon Glass <sjg at chromium.org> wrote:
> > >>
> > >> Hi Bin,
> > >>
> > >> I find that 'sf test 0 1000' does not work anymore since this commit:
> > >>
> > >> c4e8862308 mtd: spi: Switch to new SPI NOR framework
> > >>
> > >> Have you seen any problems? This seems to prevent proper SPI flash writing (not sure about reading).
> > >
> > > Yes, this seems to be broken. I just tested it on MinnowMax.
> > >
> >
> > What is the flash part on this board?
> >
> >
> > > 'sf probe' does not succeed anymore.
> > >
> >
> > Could you enable all debug prints in spi_mem_exec_op() in spi-mem.c:
> > https://elixir.bootlin.com/u-boot/latest/source/drivers/spi/spi-mem.c#L376
> >
> > And the post the full log here. If the log is too big please paste it to pastebin.ubuntu.com
> >
> > Also please apply http://patchwork.ozlabs.org/patch/1090121/ on top of latest tree before doing any write to flash
> >
> > Sorry for the trouble.
> >
>
> Sorry for the long delay. I finally got time to look into this.
>
> However with the latest u-boot/master, the SPI flash seems to work
> again on Intel MinnowMax. Did you recall anything changes in the
> spi-nor recently?

The thing I know is:

v2019.07-rc1: Intel ich-spi driver was broken
v2019.07-rc2: Intel ich-spi driver worked again

So I did a 'git bisect', although I used it in a reverse direction :)

This is the commit that makes the Intel ich-spi driver work again.

commit 60e2bf46784ebbd30ff29b3d3c7c97e56b11e86a
Author: Weijie Gao <weijie.gao at mediatek.com>
Date:   Fri Apr 26 17:22:19 2019 +0800

    mtd: spi-nor: fix page program issue when using spi-mem driver

    Some SPI controllers can't write nor->page_size bytes in a single step
    because their TX FIFO is too small, but when that happens we should
    make sure a WRITE_EN command before each write access and READ_SR command
    after each write access is issued.

    We should allow nor->write() to return a size that is smaller than the
    requested write size to gracefully handle this case.

    Also, the spi_nor_write_data() should return the actual number of bytes
    that were written during the spi_mem_exec_op() operation.

    This patch is a combination of two commits backported from kernel:

      commit 630d6bd8a3b4 ("mtd: spi-nor: Support controllers with limit ...")
      commit 3baa8ec88c2f ("mtd: devices: m25p80: Make sure WRITE_EN is ...")

    Cc: Vignesh R <vigneshr at ti.com>
    Signed-off-by: Weijie Gao <weijie.gao at mediatek.com>
    Acked-by: Vignesh R <vigneshr at ti.com>
    Tested-by: Shyam Saini <shyam.saini at amarulasolutions.com> # microzed
    Acked-by: Jagan Teki <jagan at amarulasolutions.com>

Hi Simon,

Could you please confirm that on the Chromebook?

Regards,
Bin