Re: [PATCH] rtc: rv8803: Check return value of rv8803_write_reg

[Alexandre Belloni]

On 27/12/2018 17:28:33-0600, Kangjie Lu wrote:
> On Thu, Dec 27, 2018 at 4:31 PM Heiner Kallweit <hkallweit1@xxxxxxxxx>
> wrote:
> 
> > On 27.12.2018 21:28, Aditya Pakki wrote:
> > > In rv8803_handle_irq, rv8803_write_reg can return a failed return
> > > value when attempting to write to the bus. The fix checks the output
> > > and throws a dev_warn notifying of the failure.
> > >
> > > Signed-off-by: Aditya Pakki <pakki001@xxxxxxx>
> > > ---
> > >  drivers/rtc/rtc-rv8803.c | 9 +++++++--
> > >  1 file changed, 7 insertions(+), 2 deletions(-)
> > >
> > You seem to submit the same type of changes throughout very
> > different subsystems. And you do it w/o thinking and testing.
> > If you would have looked at rv8803_write_reg() you would have
> > seen that it prints an error in case of failure. So your
> > patch achieves nothing.
> > You got David Miller upset already and it looks like you
> > want to achieve the same with other maintainers too.
> > I'd strongly suggest that you stop sending patches until
> > you better understand the kernel code.
> >
> 
> Hello Heiner,
> 
> Thanks for your suggestion. Sure, we will try to better understand
> how the kernel works when we are preparing other patches. We recently
> found a lot of potential bugs; due to the significant workload but
> limited labor force, we may make some mistakes, but yes, we will try
> to avoid them.
> 
> One main reason we submit the patches is to seek feedback from Linux
> maintainers who know how the kernel works best.  We hope to get: (1)
> confirmation: if this is indeed a bug;

Come on, this is your job, not the maintainer job to check whether there
is indeed a bug. Else, the maintainer may as well just remove your
authorship because he did all the real work.

> (2) improvement feedback: if
> it is a bug and our fix is problematic, how can we improve it?
> 
> Taking the case in this email as an example, rv8803_write_reg could
> fail, so returning IRQ_HANDLED even when it failed doesn't seem to be
> a good practice. Would "returning IRQ_NONE upon failure" be a better
> fix?
> 
> Thanks again for your suggestion.

-- 
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com