On Mon, 29 Jun 2020, Johan Hovold wrote:
> On Mon, Jun 29, 2020 at 01:32:14PM +0100, Lee Jones wrote:
> > Since ddata->irqs[] is already zeroed when allocated by devm_kcalloc() and
> > dividing 0 by anything is still 0, there is no need to re-assign
> > ddata->irqs[i].* values. Instead, it should be safe to begin at 1.
> >
> > This fixes the following W=1 warning:
> >
> > drivers/mfd/sprd-sc27xx-spi.c:255 sprd_pmic_probe() debug: sval_binop_unsigned: divide by zero
> >
> > Cc: Orson Zhai <[email protected]>
> > Cc: Baolin Wang <[email protected]>
> > Cc: Chunyan Zhang <[email protected]>
> > Signed-off-by: Lee Jones <[email protected]>
> > ---
> > drivers/mfd/sprd-sc27xx-spi.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/mfd/sprd-sc27xx-spi.c b/drivers/mfd/sprd-sc27xx-spi.c
> > index c305e941e435c..694a7d429ccff 100644
> > --- a/drivers/mfd/sprd-sc27xx-spi.c
> > +++ b/drivers/mfd/sprd-sc27xx-spi.c
> > @@ -251,7 +251,7 @@ static int sprd_pmic_probe(struct spi_device *spi)
> > return -ENOMEM;
> >
> > ddata->irq_chip.irqs = ddata->irqs;
> > - for (i = 0; i < pdata->num_irqs; i++) {
> > + for (i = 1; i < pdata->num_irqs; i++) {
> > ddata->irqs[i].reg_offset = i / pdata->num_irqs;
> > ddata->irqs[i].mask = BIT(i % pdata->num_irqs);
> > }
>
> This doesn't look right either.
>
> First, the loop is never executed if num_irqs is zero.
The point of the patch is that 0 entries are never processed.
However, what I appear to have overlooked is that BIT(0 % x) is not 0,
it's 1.
> Second, the current code looks bogus too as reg_offset is always set to
> zero and mask to BIT(i)...
Heh. I wonder if/how this was tested.
I'm going to wait to hear from the authors before attempting to fix
this again.
Baolin, Could you please clarify this for us please?
--
Lee Jones [李琼斯]
Senior Technical Lead - Developer Services
Linaro.org │ Open source software for Arm SoCs
Follow Linaro: Facebook | Twitter | Blog
On Mon, Jun 29, 2020 at 10:01 PM Lee Jones <[email protected]> wrote:
>
> On Mon, 29 Jun 2020, Johan Hovold wrote:
>
> > On Mon, Jun 29, 2020 at 01:32:14PM +0100, Lee Jones wrote:
> > > Since ddata->irqs[] is already zeroed when allocated by devm_kcalloc() and
> > > dividing 0 by anything is still 0, there is no need to re-assign
> > > ddata->irqs[i].* values. Instead, it should be safe to begin at 1.
> > >
> > > This fixes the following W=1 warning:
> > >
> > > drivers/mfd/sprd-sc27xx-spi.c:255 sprd_pmic_probe() debug: sval_binop_unsigned: divide by zero
> > >
> > > Cc: Orson Zhai <[email protected]>
> > > Cc: Baolin Wang <[email protected]>
> > > Cc: Chunyan Zhang <[email protected]>
> > > Signed-off-by: Lee Jones <[email protected]>
> > > ---
> > > drivers/mfd/sprd-sc27xx-spi.c | 2 +-
> > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/mfd/sprd-sc27xx-spi.c b/drivers/mfd/sprd-sc27xx-spi.c
> > > index c305e941e435c..694a7d429ccff 100644
> > > --- a/drivers/mfd/sprd-sc27xx-spi.c
> > > +++ b/drivers/mfd/sprd-sc27xx-spi.c
> > > @@ -251,7 +251,7 @@ static int sprd_pmic_probe(struct spi_device *spi)
> > > return -ENOMEM;
> > >
> > > ddata->irq_chip.irqs = ddata->irqs;
> > > - for (i = 0; i < pdata->num_irqs; i++) {
> > > + for (i = 1; i < pdata->num_irqs; i++) {
> > > ddata->irqs[i].reg_offset = i / pdata->num_irqs;
> > > ddata->irqs[i].mask = BIT(i % pdata->num_irqs);
> > > }
> >
> > This doesn't look right either.
> >
> > First, the loop is never executed if num_irqs is zero.
>
> The point of the patch is that 0 entries are never processed.
>
> However, what I appear to have overlooked is that BIT(0 % x) is not 0,
> it's 1.
Yes.
>
> > Second, the current code looks bogus too as reg_offset is always set to
> > zero and mask to BIT(i)...
Now the result is correct, since all PMIC irq mask bits are in one
register now, which means the reg_offset is always 0 can work well.
But I think the logics still can be improved if our PMIC irq numbers
are larger than 32 in future.
>
> Heh. I wonder if/how this was tested.
>
> I'm going to wait to hear from the authors before attempting to fix
> this again.
>
> Baolin, Could you please clarify this for us please?
Yes, see above comments.
--
Baolin Wang
On Mon, Jun 29, 2020 at 10:35:06PM +0800, Baolin Wang wrote:
> On Mon, Jun 29, 2020 at 10:01 PM Lee Jones <[email protected]> wrote:
> >
> > On Mon, 29 Jun 2020, Johan Hovold wrote:
> >
> > > On Mon, Jun 29, 2020 at 01:32:14PM +0100, Lee Jones wrote:
> > > > Since ddata->irqs[] is already zeroed when allocated by devm_kcalloc() and
> > > > dividing 0 by anything is still 0, there is no need to re-assign
> > > > ddata->irqs[i].* values. Instead, it should be safe to begin at 1.
> > > >
> > > > This fixes the following W=1 warning:
> > > >
> > > > drivers/mfd/sprd-sc27xx-spi.c:255 sprd_pmic_probe() debug: sval_binop_unsigned: divide by zero
> > > >
> > > > Cc: Orson Zhai <[email protected]>
> > > > Cc: Baolin Wang <[email protected]>
> > > > Cc: Chunyan Zhang <[email protected]>
> > > > Signed-off-by: Lee Jones <[email protected]>
> > > > ---
> > > > drivers/mfd/sprd-sc27xx-spi.c | 2 +-
> > > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > > >
> > > > diff --git a/drivers/mfd/sprd-sc27xx-spi.c b/drivers/mfd/sprd-sc27xx-spi.c
> > > > index c305e941e435c..694a7d429ccff 100644
> > > > --- a/drivers/mfd/sprd-sc27xx-spi.c
> > > > +++ b/drivers/mfd/sprd-sc27xx-spi.c
> > > > @@ -251,7 +251,7 @@ static int sprd_pmic_probe(struct spi_device *spi)
> > > > return -ENOMEM;
> > > >
> > > > ddata->irq_chip.irqs = ddata->irqs;
> > > > - for (i = 0; i < pdata->num_irqs; i++) {
> > > > + for (i = 1; i < pdata->num_irqs; i++) {
> > > > ddata->irqs[i].reg_offset = i / pdata->num_irqs;
> > > > ddata->irqs[i].mask = BIT(i % pdata->num_irqs);
> > > > }
> > >
> > > This doesn't look right either.
> > >
> > > First, the loop is never executed if num_irqs is zero.
> >
> > The point of the patch is that 0 entries are never processed.
So what's the problem? There's no division by zero here.
And what compiler are you using, Lee? Seems broken.
> > > Second, the current code looks bogus too as reg_offset is always set to
> > > zero and mask to BIT(i)...
>
> Now the result is correct, since all PMIC irq mask bits are in one
> register now, which means the reg_offset is always 0 can work well.
> But I think the logics still can be improved if our PMIC irq numbers
> are larger than 32 in future.
The code is still bogus as pointed out above. Why do you bother to
divide by num_irqs at all?
And what have you guys been smoking? ;)
Johan