Re: [PATCH v3 4/5] mtd: rawnand: denali_dt: add reset controlling

[Miquel Raynal]

On Fri, 2019-12-20 at 11:31:54 UTC, Masahiro Yamada wrote:
> According to the Denali NAND Flash Memory Controller User's Guide,
> this IP has two reset signals.
> 
>   rst_n:     reset most of FFs in the controller core
>   reg_rst_n: reset all FFs in the register interface, and in the
>              initialization sequencer
> 
> This commit supports controlling those reset signals.
> 
> It is possible to control them separately from the IP point of view
> although they might be often tied up together in actual SoC integration.
> 
> The IP spec says, asserting only the reg_rst_n without asserting rst_n
> will cause unpredictable behavior in the controller. So, the driver
> deasserts ->rst_reg and ->rst in this order.
> 
> Another thing that should be kept in mind is the automated initialization
> sequence (a.k.a. 'bootstrap' process) is kicked off when reg_rst_n is
> deasserted.
> 
> When the reset is deasserted, the controller issues a RESET command
> to the chip select 0, and attempts to read out the chip ID, and further
> more, ONFI parameters if it is an ONFI-compliant device. Then, the
> controller sets up the relevant registers based on the detected
> device parameters.
> 
> This process might be useful for tiny boot firmware, but is redundant
> for Linux Kernel because nand_scan_ident() probes devices and sets up
> parameters accordingly. Rather, this hardware feature is annoying
> because it ends up with misdetection due to bugs.
> 
> So, commit 0615e7ad5d52 ("mtd: nand: denali: remove Toshiba and Hynix
> specific fixup code") changed the driver to not rely on it.
> 
> However, there is no way to prevent it from running. The IP provides
> the 'bootstrap_inhibit_init' port to suppress this sequence, but it is
> usually out of software control, and dependent on SoC implementation.
> As for the Socionext UniPhier platform, LD4 always enables it. For the
> later SoCs, the bootstrap sequence runs depending on the boot mode.
> 
> I added usleep_range() to make the driver wait until the sequence
> finishes. Otherwise, the driver would fail to detect the chip due
> to the race between the driver and hardware-controlled sequence.
> 
> Signed-off-by: Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx>
> Reviewed-by: Philipp Zabel <p.zabel@xxxxxxxxxxxxxx>

Applied to https://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux.git nand/next, thanks.

Miquel