On Thu, 25 Oct 2018, Dan Carpenter wrote:
> On Thu, Oct 25, 2018 at 09:25:08AM +0100, Lee Jones wrote:
> > On Thu, 18 Oct 2018, Dan Carpenter wrote:
> >
> > > I just noticed this in review. The get_register_interruptible() should
> > > return zero on success but it instead returns the value that it read.
> > >
> > > I looked at all the places that called this directly and they check for
> > > negatives and treat greater than or equal to zero as success. This
> > > function is also called as the ->get_register() function pointer. Some
> > > of the callers of that treat all non-zero returns as errors, so it's
> > > possible that this bug causes some problems in real life.
> > >
> > > I could not find any callers that rely on the current behavior, and this
> > > makes the function align with the get_register_interruptible() in
> > > ab3100-core.c.
> > >
> > > Fixes: 47c1697508f2 ("mfd: Align ab8500 with the abx500 interface")
> > > Signed-off-by: Dan Carpenter <dan.carpen...@oracle.com>
> > > ---
> > > drivers/mfd/ab8500-core.c | 10 ++++++----
> > > 1 file changed, 6 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/mfd/ab8500-core.c b/drivers/mfd/ab8500-core.c
> > > index 30d09d177171..66458a329127 100644
> > > --- a/drivers/mfd/ab8500-core.c
> > > +++ b/drivers/mfd/ab8500-core.c
> > > @@ -252,16 +252,18 @@ static int get_register_interruptible(struct ab8500
> > > *ab8500, u8 bank,
> > > mutex_lock(&ab8500->lock);
> > >
> > > ret = ab8500->read(ab8500, addr);
> > > - if (ret < 0)
> > > + if (ret < 0) {
> > > dev_err(ab8500->dev, "failed to read reg %#x: %d\n",
> > > addr, ret);
> > > - else
> > > - *value = ret;
> > > + return ret;
> >
> > Looks like you just broke the AB8500.
> >
>
>
> Oh wow. I screwed up the locking.
>
>
> > What is it you're trying to achieve here? Apart from attempting to
> > return with the mutex still held, what semantics have you changed?
> >
>
> Sorry that wasn't clear. Here is the relevant bits from the commit
> message.
>
> > > function is also called as the ->get_register() function pointer. Some
> > > of the callers of that treat all non-zero returns as errors, so it's
> > > possible that this bug causes some problems in real life.
>
> We're returning positive non-zero values on success instead of zero.
> It's definitely a bug, but I'm not sure if it has an impact in real
> life.
Oh, I see.
Maybe:
return (ret > 0) ? 0 : ret;
--
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog
