Re: [PATCH v2 10/22] mtd: spi-nor: Rework write_sr()

[John Garry]

On 04/10/2019 11:03, Tudor.Ambarus@xxxxxxxxxxxxx wrote:
> Hi, John,
>
> On 10/04/2019 12:39 PM, John Garry wrote:
> > External E-Mail
> >
> >
> > On 24/09/2019 08:46, Tudor.Ambarus@xxxxxxxxxxxxx wrote:
> > > +}
> > > +
> > > +/**
> > > + * spi_nor_write_sr() - Write the Status Register.
> > > + * @nor:    pointer to 'struct spi_nor'.
> > > + * @sr:        buffer to write to the Status Register.
> > > + * @len:    number of bytes to write to the Status Register.
> > > + *
> > > + * Return: 0 on success, -errno otherwise.
> > >   */
> > > -static int write_sr(struct spi_nor *nor, u8 val)
> > > +static int spi_nor_write_sr(struct spi_nor *nor, const u8 *sr, size_t len)
> > >  {
> > > -    nor->bouncebuf[0] = val;
> > > +    int ret;
> > > +
> > > +    ret = spi_nor_write_enable(nor);
> > > +    if (ret)
> > > +        return ret;
> > > +
> >
> > Hi Tudor,
> >
> > >      if (nor->spimem) {
> > >          struct spi_mem_op op =
> > >              SPI_MEM_OP(SPI_MEM_OP_CMD(SPINOR_OP_WRSR, 1),
> > >                     SPI_MEM_OP_NO_ADDR,
> > >                     SPI_MEM_OP_NO_DUMMY,
> > > -                   SPI_MEM_OP_DATA_IN(1, nor->bouncebuf, 1));
> >
> > This be SPI_MEM_OP_DATA_OUT() in the current mainline code also, right?

Hi Tudor,

> In v5.4-rc1 this is defined as SPI_MEM_OP_DATA_IN, so the Mainline code should
> fail. This looks like a bug. I didn't noticed it when doing the patch.
>
> > I'm testing my under development driver on top of v5.4-rc1, and flash_lock -u is broken.
>
> It's not clear to me, does flash_lock fail with my patches on top of v5.4-rc1?

No, I haven't tested these patches. I'm just testing my out-of-tree 
driver on top of vanilla v5.4-rc1. I'm just mentioning the issue here as 
it seemed like a reasonably appropriate place.

However it looks like I will also need to test on top of these patches.

> Or it fails when testing v5.4-rc1?
>
> Can you test v5.4-rc1 and see if flash_lock works on you flash or not?

flash_lock -u errors for my driver on top of vanilla v5.4-rc1.

When I make the change, as above, flash_lock -u and -l succeed, but I 
can still write to the flash - I need to check that more - it may be my 
buggy driver.

IIRC, it did work for my driver based on v5.3

> Please specify which flash do you use, and which controller.

The flash is n25q128a11:
[   14.917868] spi-nor spi-PRP0001:00: n25q128a11 (16384 Kbytes)

As for the driver, it's another HiSilicon SPI NOR controller driver 
which I'm developing - I eluded to it here already: 
https://lore.kernel.org/linux-mtd/c5e063e8-5025-8206-f819-6ce5228ef0fb@xxxxxxxxxx/

Cheers,
John

> Thanks for testing this!
> ta