[U-Boot] [PATCH] pmic: Fix pfuze100 bit definitions

[Marek Vasut]

On Thursday, November 26, 2015 at 03:35:26 PM, Przemyslaw Marczak wrote:
> Hello Marek,

Hi,

> On 11/26/2015 02:08 PM, Marek Vasut wrote:
> > On Thursday, November 26, 2015 at 01:21:36 PM, Przemyslaw Marczak wrote:
> >> Hello Marek,
> > 
> > Hi,
> > 
> >> On 11/26/2015 12:15 AM, Marek Vasut wrote:
> >>> The following patch changed the PFUZE100 swbst register bit definitions
> >>> and broke PMIC configuration on multiple boards, at least on the novena
> >>> and gw_ventana. This patch fixes it.
> >> 
> >> Ok we missed this in the review. But as I can see it broken only the two
> >> boards, you mentioned.
> >> 
> >>> commit 8fa46350a4c7dca7710362f6c871098557b934ad
> >>> Author: Peng Fan <Peng.Fan at freescale.com>
> >>> Date:   Fri Aug 7 16:43:45 2015 +0800
> >>> 
> >>>       power: regulator: add pfuze100 support
> >>> 
> >>> Signed-off-by: Marek Vasut <marex at denx.de>
> >>> Cc: Fabio Estevam <fabio.estevam at freescale.com>
> >>> Cc: Peng Fan <Peng.Fan at freescale.com>
> >>> Cc: Przemyslaw Marczak <p.marczak at samsung.com>
> >>> Cc: Tim Harvey <tharvey at gateworks.com>
> >>> Cc: Vagrant Cascadian <vagrant at aikidev.net>
> >>> ---
> >>> 
> >>>    include/power/pfuze100_pmic.h | 8 ++++----
> >>>    1 file changed, 4 insertions(+), 4 deletions(-)
> >>> 
> >>> diff --git a/include/power/pfuze100_pmic.h
> >>> b/include/power/pfuze100_pmic.h index 41cb710..cc019a9 100644
> >>> --- a/include/power/pfuze100_pmic.h
> >>> +++ b/include/power/pfuze100_pmic.h
> >>> @@ -215,10 +215,10 @@ enum {
> >>> 
> >>>    #define SWBST_VOL_MASK	0x3
> >>>    #define SWBST_MODE_MASK	0xC
> >>>    #define SWBST_MODE_SHIFT 0x2
> >>> 
> >>> -#define SWBST_MODE_OFF	0
> >>> -#define SWBST_MODE_PFM	1
> >>> -#define SWBST_MODE_AUTO	2
> >>> -#define SWBST_MODE_APS	3
> >>> +#define SWBST_MODE_OFF	(0 << 2)
> >>> +#define SWBST_MODE_PFM	(1 << 2)
> >>> +#define SWBST_MODE_AUTO	(2 << 2)
> >>> +#define SWBST_MODE_APS	(3 << 2)
> >>> 
> >>>    /*
> >>>    
> >>>     * Regulator Mode Control
> >> 
> >> The intentions are good, but this patch fixes one thing and breaks the
> >> another one, I would prefer avoid this.
> >> 
> >> 'git grep -n SWBST_MODE'
> >> 
> >> As I can see, you can fix the issue for multiple boards by update only
> >> two lines in those two boards, which you mentioned.
> >> 
> >> So why you moving back those definitions, since they are now used in
> >> more places?
> >> 
> >> The line suggested by Peng is good enough to call it 'fix' for your
> >> boards:
> >> 
> >> (SWBST_MODE_AUTO << SWBST_MODE_SHIFT)
> > 
> > OK, so instead of fixing the patch which introduced a bug, we're supposed
> > to be fixing the fallout from that. I cannot say I'm very happy with
> > this sort of handling of a bug and with the testing this particular
> > change received.
> 
> You are right, the mentioned patch breaks your boards, and we missed
> this in the review as I mentioned - sorry for that.
> 
> But for now, there is also other code based on those definitions, so you
> can not just revert only this particular change and ignore the rest -
> because it breaks the new code? Should we all work in this way?

This is even worse then -- the patch adds code which uses the changed macros,
but doesn't fix the existing users. This should not happen again and it'd be 
very nice if the author actually checked when digging in /include and changing 
some macros there if this might affect someone.

> As a custodian I'm not able to test everything, especially when I don't
> have the hardware for it. Moreover I trust people who are working for
> this project and I can imagine that they test the code.

I don't expect you to test anything in this case, other but possibly compile
testing the stuff, don't get me wrong.

> > Besides, seeing how this patch already needed another patch to make it
> > complete and how it now needs more patches to fix the boards which it
> > broke, I am really disappointed.
> 
> I can't understand what is the problem. You send new patch with two
> simple lines - it fixes your issue and doesn't break the existing PMIC
> driver. I think, this is what we need here.

I did that. And unfortunatelly, it turns out we have really no other option
now, than to fix the boards. Sigh ...