[PATCH v3 0/2] Migrate the GPMI driver to use NAND core timings

[Miquel Raynal]

Hello Han,

Would you mind give us your input about this series as well as the
patch that follows (Use core timings instead of an empirical
derivation) ?

If you are unhappy with them I will welcome your comments.

Thank you,
Miquèl

On Fri, 19 Jan 2018 14:58:53 +0100, Miquel Raynal
<miquel.raynal at free-electrons.com> wrote:

> Hello,
> 
> The GPMI NAND controller driver has its own timings logic while the core
> also has its own that should be preferred to avoid code duplication.
> This series migrates the driver to use the ->setup_data_interface()
> hook, preferred way to handle timings negotiation between controllers
> and NAND chips.
> 
> This driver has stronger requirements in terms of respect of the ONFI
> specification than the core has, and last time I post changes to this
> section this was raised to me. Hence, the first patch effectively
> applies the same requirements to the core. Doing this can slow down
> several chips that do not follow entirely the specifications, but
> another series is coming to add the possibility for NAND vendors to
> flag certain features as "unsupported". See
> 
>     "mtd: nand: Check ONFI timings have been acked by the chip"
> 
> commit log for more details.
> 
> Thank you,
> Miquèl
> 
> 
> Changes since v2:
> =================
>   - Fix ->select_chip(): chipnr == 0 is valid.
>   - Fix ->select_chip(): hw.must_apply_timings must be reset after a
>     ->setup_data_interface().  
> 
> Changes since v1:
> =================
>   - Removed inline on non-static gpmi_enable/disable_clk() functions.
>   - Rewording some lines of the patch.
>   - Use tREA_max value provided by the core instead of a locally defined
>     value.
>   - Check ->conf in gpmi_setup_data_interface() instead of
>     ->onfi_timing_mode_default (use tRC).  
>   - Removed useless copy of hw.timing_mode.
>   - Moved gpmi_nfc_apply_timings() to ->select_chip().
>   - Canceled the useless move of one structure.
> 
> One comment is not addressed here: using one register for TIMING0/1 and
> CTRL1 that is only written in ->apply_timings(). This will be done in a
> later patch that will update the way timings are derived.
> 
> 
> Miquel Raynal (2):
>   mtd: nand: Check ONFI timings have been acked by the chip
>   mtd: nand: gpmi: Support ->setup_data_interface()
> 
>  drivers/mtd/nand/gpmi-nand/gpmi-lib.c  | 277 ++++++++++-----------------------
>  drivers/mtd/nand/gpmi-nand/gpmi-nand.c |  42 +++--
>  drivers/mtd/nand/gpmi-nand/gpmi-nand.h |  82 +++++-----
>  drivers/mtd/nand/nand_base.c           |  42 ++++-
>  4 files changed, 185 insertions(+), 258 deletions(-)
> 



-- 
Miquel Raynal, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
http://bootlin.com