[U-Boot] [PATCH v2] misc: fs_loader: Add support for initializing MMC

[Chee, Tien Fong]

On Wed, 2019-01-09 at 09:19 -0700, Simon Glass wrote:
> Hi Tien Fong,
> 
> On Sat, 29 Dec 2018 at 18:36, <tien.fong.chee at intel.com> wrote:
> > 
> > 
> > From: Tien Fong Chee <tien.fong.chee at intel.com>
> > 
> > Firmware loader would encounter problem if the MMC is accessed
> > before
> > initializing it. This patch would adding the support of probing
> > block
> > device and initializing MMC before the MMC is accessed by firmware
> > loader.
> > 
> > Signed-off-by: Tien Fong Chee <tien.fong.chee at intel.com>
> > ---
> > 
> > Changes in v2:
> > - Initializing MMC through probing the blk device
> > ---
> >  drivers/misc/fs_loader.c | 33 +++++++++++++++++++++++++++++++++
> >  1 file changed, 33 insertions(+)
> > 
> > diff --git a/drivers/misc/fs_loader.c b/drivers/misc/fs_loader.c
> > index 57a14a3..5858073 100644
> > --- a/drivers/misc/fs_loader.c
> > +++ b/drivers/misc/fs_loader.c
> > @@ -12,6 +12,7 @@
> >  #include <linux/string.h>
> >  #include <mapmem.h>
> >  #include <malloc.h>
> > +#include <mmc.h>
> >  #include <spl.h>
> > 
> >  DECLARE_GLOBAL_DATA_PTR;
> > @@ -252,6 +253,38 @@ static int fs_loader_ofdata_to_platdata(struct
> > udevice *dev)
> > 
> >  static int fs_loader_probe(struct udevice *dev)
> >  {
> > +#if defined(CONFIG_DM_MMC) && defined(CONFIG_BLK)
> > +       int ret;
> > +       struct device_platdata *plat = dev->platdata;
> > +
> > +       ret = mmc_initialize(NULL);
> You should not need this since the MMC device should init itself when
> probed.
Noted.
> 
> > 
> > +       if (ret) {
> > +               debug("MMC: could not initialize mmc. error: %d\n",
> > ret);
> > +
> > +               return ret;
> > +       }
> > +
> > +       if (plat->phandlepart.phandle) {
> > +               ofnode node = ofnode_get_by_phandle(plat-
> > >phandlepart.phandle);
> > +
> > +               struct udevice *mmc_dev = NULL;
> > +
> > +               ret = device_get_global_by_ofnode(node, &mmc_dev);
> You could use this?
> 
> uclass_get_device_by_ofnode(UCLASS_MMC, node, &mmc_dev)
Okay.
> 
> But actually I think you should drop any mention of MMC so that your
> code can work for any storage uclass.
I agree with you, we should made the probing generic to any storage
uclass.

But now i'm not familiar with all storage uclass yet, so my strategy is
getting one storage uclass in at a time, and restructure the probing if
there is commonality for the next storage uclass. With that, we can
control the quality and achieving the goal step by step.

What do you think?
> 
> > 
> > +               if (!ret) {
> > +                       struct mmc *mmc = mmc_get_mmc_dev(mmc_dev);
> > +                       struct udevice *dev;
> > +
> > +                       ret = blk_get_from_parent(mmc->dev, &dev);
> You are going from mmc_dev -> mmc -> mmc->dev, but unless I am
> mistaken, mmc_dev == mmc->dev. So you should be able to simplify this
> to:
> 
>                        ret = blk_get_from_parent(mmc_dev, &dev);
Okay,noted.
> 
> Also you should be able to rename mmc to something more generic, e.g.
> to make this work with SATA.
Replied in above.
> 
> > 
> > +                       if (ret) {
> > +                               debug("MMC: No block device: %d\n",
> > +                                       ret);
> > +
> > +                               return ret;
> > +                       }
> > +               }
> > +       }
> > +#endif
> > +
> >         return 0;
> >  };
> > 
> > --
> > 2.2.0
> > 
> Regards,
> Simon