[PATCH v2 1/9] spl: Try to get SPL boot device via board_get_int

[Tom Rini <trini@konsulko.com>]

On Mon, May 25, 2020 at 02:01:09PM +0530, Jagan Teki wrote:
> On Sat, May 23, 2020 at 2:55 AM Tom Rini <trini@konsulko.com> wrote:
> >
> > On Fri, May 22, 2020 at 09:42:22PM +0530, Jagan Teki wrote:
> > > On Wed, May 20, 2020 at 7:40 PM Tom Rini <trini@konsulko.com> wrote:
> > > >
> > > > On Wed, May 20, 2020 at 06:46:55PM +0530, Jagan Teki wrote:
> > > > > On Wed, May 20, 2020 at 6:32 PM Tom Rini <trini@konsulko.com> wrote:
> > > > > >
> > > > > > On Wed, May 20, 2020 at 12:53:32AM +0530, Jagan Teki wrote:
> > > > > >
> > > > > > > Usually, the associated board would supply spl boot device
> > > > > > > using spl_boot_device() but some boards have board driver
> > > > > > > that are possible to supply boot device via board_get_int
> > > > > > > with BOARD_SPL_BOOT_DEVICE id.
> > > > > > >
> > > > > > > This patch add support for those.
> > > > > > >
> > > > > > > Cc: Mario Six <mario.six@gdsys.cc>
> > > > > > > Cc: Tom Rini <trini@konsulko.com>
> > > > > > > Cc: Simon Glass <sjg@chromium.org>
> > > > > > > Cc: Jean-Jacques Hiblot <jjhiblot@ti.com>
> > > > > > > Signed-off-by: Jagan Teki <jagan@amarulasolutions.com>
> > > > > > > ---
> > > > > > > Changes for v2:
> > > > > > > - new patch
> > > > > > >
> > > > > > >  common/spl/spl.c | 14 +++++++++++++-
> > > > > > >  include/board.h  |  9 +++++++++
> > > > > > >  2 files changed, 22 insertions(+), 1 deletion(-)
> > > > > > >
> > > > > > > diff --git a/common/spl/spl.c b/common/spl/spl.c
> > > > > > > index fc5cbbbeba..a07b71b3c1 100644
> > > > > > > --- a/common/spl/spl.c
> > > > > > > +++ b/common/spl/spl.c
> > > > > > > @@ -9,6 +9,7 @@
> > > > > > >  #include <common.h>
> > > > > > >  #include <bloblist.h>
> > > > > > >  #include <binman_sym.h>
> > > > > > > +#include <board.h>
> > > > > > >  #include <dm.h>
> > > > > > >  #include <handoff.h>
> > > > > > >  #include <hang.h>
> > > > > > > @@ -483,9 +484,20 @@ int spl_init(void)
> > > > > > >  #define BOOT_DEVICE_NONE 0xdeadbeef
> > > > > > >  #endif
> > > > > > >
> > > > > > > +__weak u32 spl_boot_device(void)
> > > > > > > +{
> > > > > > > +     return 0;
> > > > > > > +}
> > > > > > > +
> > > > > > >  __weak void board_boot_order(u32 *spl_boot_list)
> > > > > > >  {
> > > > > > > -     spl_boot_list[0] = spl_boot_device();
> > > > > > > +     struct udevice *board;
> > > > > > > +
> > > > > > > +     if (!board_get(&board))
> > > > > > > +             board_get_int(board, BOARD_SPL_BOOT_DEVICE,
> > > > > > > +                           (int *)&spl_boot_list[0]);
> > > > > > > +     else
> > > > > > > +             spl_boot_list[0] = spl_boot_device();
> > > > > > >  }
> > > > > > >
> > > > > > >  static struct spl_image_loader *spl_ll_find_loader(uint boot_device)
> > > > > > > diff --git a/include/board.h b/include/board.h
> > > > > > > index 678b652b0a..ce4eaba38d 100644
> > > > > > > --- a/include/board.h
> > > > > > > +++ b/include/board.h
> > > > > > > @@ -211,3 +211,12 @@ static inline int board_get_fit_loadable(struct udevice *dev, int index,
> > > > > > >  }
> > > > > > >
> > > > > > >  #endif
> > > > > > > +
> > > > > > > +/**
> > > > > > > + * Common board unique identifier
> > > > > > > + *
> > > > > > > + * @BOARD_SPL_BOOT_DEVICE:   id to get SPL boot device.
> > > > > > > + */
> > > > > > > +enum common_ids {
> > > > > > > +     BOARD_SPL_BOOT_DEVICE,
> > > > > > > +};
> > > > > >
> > > > > > I don't understand why we need this abstraction.  The intention of what
> > > > > > we have today is that the generic SPL framework calls out to something
> > > > > > to ask "what are we booted from?".  Why can the board driver not just
> > > > > > supply that information?  Thanks!
> > > > >
> > > > > Yes, we can update boot-device on respective areas by probing board
> > > > > driver and assign spl_boot_list[0] by explicitly define
> > > > > spl_boot_device function, but this change bypass all these codes. Just
> > > > > like how we did on SPL fit to load the concerned image via board
> > > > > driver.
> > > >
> > > > I still don't get it, sorry.  Why is spl_boot_device() not provided by
> > > > the "board" driver to say what to boot in this case?
> > >
> > > That means, we have to add spl_boot_device in board-uclass.c ? so-that
> > > respective board driver shall use?
> >
> > Yes, or perhaps a board driver doesn't even make sense in this case and
> > the existing abstraction should be used as is?  This isn't a unique
> > problem, it's something we've been handling in SPL since the beginning.
> > In so far as we can now try and solve this problem with something
> > DM-based instead of not, it should still I believe just be the same
> > function call.
> 
> I don't understand why we have individual function calls though we
> have dm supported cases to get rid of those? then why would be the
> case for spl fit code to pump multiple images via dm board driver?
> What this patch is trying to do is fundamentally similar, like spl_fit
> is using a board driver to support multiple fit images where spl is
> using a board driver to support the desired boot device.

Because when using DM it needs to be to improve things overall, not just
duplicate them.  We have an abstraction for "figure out what device
we've booted from" already.  I don't see how your changes are improving
the situation, just making the same abstraction with more calls.

When we're dealing with SPL we need to be even more thoughtful than
usual about size impacts.  So the DM implementation of spl_boot_device()
belongs elsewhere rather than making SPL do some checks and doing DM or
not DM.  Thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200525/835a584c/attachment.sig>