[U-Boot] [PATCH v2 6/8] reset: socfpga: add reset handling for old kernels

[Marek Vasut]

On 2/21/19 11:07 PM, Simon Goldschmidt wrote:
> 
> 
> Am Do., 21. Feb. 2019, 22:56 hat Marek Vasut <marex at denx.de
> <mailto:marex at denx.de>> geschrieben:
> 
>     On 2/21/19 10:43 PM, Simon Goldschmidt wrote:
>     > This adds code to take peripherals out of reset based on an
>     environment
>     > variable. This is in preparation for removing the code that does
>     this from
>     > SPL.
>     >
>     > However, some drivers even in current Linux cannot handle
>     peripheral reset,
>     > so until this works, we need a compatibility workaround.
>     >
>     > This workaround is implemented in the 'remove' callback of this reset
>     > driver, which is called on OS_PREPARE: it checks if the environment
>     > variable "socfpga_permodrst_ungate" - if it is set to 1, it
>     deasserts all
>     > peripheral resets, which is what the gen5 SPL did up to now.
>     >
>     > This is in preparation to clean up the SPL and implementing proper
>     reset
>     > handling for U-Boot.
>     >
>     > Signed-off-by: Simon Goldschmidt <simon.k.r.goldschmidt at gmail.com
>     <mailto:simon.k.r.goldschmidt at gmail.com>>
>     > ---
>     >
>     > Changes in v2:
>     > - moved from Kernel option "OLD_SOCFPGA_KERNEL_COMPAT" to environment
>     >   variable "socfpga_permodrst_ungate"
>     >
>     >  drivers/reset/reset-socfpga.c | 34 ++++++++++++++++++++++++++++++++++
>     >  1 file changed, 34 insertions(+)
>     >
>     > diff --git a/drivers/reset/reset-socfpga.c
>     b/drivers/reset/reset-socfpga.c
>     > index b2acfcd2ec..749f671e05 100644
>     > --- a/drivers/reset/reset-socfpga.c
>     > +++ b/drivers/reset/reset-socfpga.c
>     > @@ -89,6 +89,36 @@ static int socfpga_reset_probe(struct udevice *dev)
>     >       return 0;
>     >  }
>     > 
>     > +#ifndef CONFIG_SPL_BUILD
>     > +/*
>     > + * This remove function is called before starting OS to deassert
>     peripheral
>     > + * resets for Kernels that don't support this.
>     > + *
>     > + * The reset driver checks the environment variable
>     "socfpga_permodrst_ungate"
>     > + * when it is removed. If this variable is '1', all peripherals
>     in 'permodrst'
>     > + * are taken out of reset before booting into the OS.
>     > + * This should be required for gen5 systems only that are running
>     Linux kernels
>     > + * without proper peripheral reset support for all drivers used.
>     > + */
>     > +static int socfpga_reset_remove(struct udevice *dev)
>     > +{
>     > +     struct socfpga_reset_data *data = dev_get_priv(dev);
>     > +     const char *env_str;
>     > +     long val;
>     > +
>     > +     env_str = env_get("socfpga_permodrst_ungate");
>     > +     if (env_str) {
>     > +             val = simple_strtol(env_str, NULL, 0);
>     > +             if (val == 1) {
>     > +                     puts("Deasserting all peripheral resets\n");
>     > +                     writel(0, data->membase + 4);
>     > +             }
>     > +     }
>     > +
>     > +     return 0;
>     > +}
>     > +#endif
>     > +
>     >  static const struct udevice_id socfpga_reset_match[] = {
>     >       { .compatible = "altr,rst-mgr" },
>     >       { /* sentinel */ },
>     > @@ -101,4 +131,8 @@ U_BOOT_DRIVER(socfpga_reset) = {
>     >       .probe = socfpga_reset_probe,
>     >       .priv_auto_alloc_size = sizeof(struct socfpga_reset_data),
>     >       .ops = &socfpga_reset_ops,
>     > +#ifndef CONFIG_SPL_BUILD
>     > +     .remove = socfpga_reset_remove,
>     > +     .flags  = DM_FLAG_OS_PREPARE,
>     > +#endif
>     >  };
> 
>     Does this work if you use Falcon mode ? :)
> 
> 
> Probably not, but would anyone be able to use falcon mode in socfpga
> gen5 given the current SPL size? I think it didn't fit for,a last time I
> checked.

If you cut it down, why not ?

> And if it fits in the future, we probably don't need this workaraound
> any more...
> 
> I did it like that since env_get() broke compilation of SPL. But if you
> insist of making this work even for SPL, I'll check again.

SPL can also check env, so some more conditionals need to be added.
Should be a matter of some CONFIG_IS_ENABLED()

-- 
Best regards,
Marek Vasut