RE: [PATCH] [-next] PCI: DWC: Fix cast truncates bits from constant value

[Gustavo Pimentel <Gustavo.Pimentel@synopsys.com>]

On Mon, Sep 21, 2020 at 17:36:24, Bjorn Helgaas <helgaas@kernel.org> 
wrote:

> On Mon, Sep 21, 2020 at 11:30:48AM +0000, Gustavo Pimentel wrote:
> > On Fri, Sep 18, 2020 at 8:15:48, Gustavo Pimentel <gustavo@synopsys.com> 
> > wrote:
> > 
> > > On Thu, Sep 17, 2020 at 22:47:59, Bjorn Helgaas <helgaas@kernel.org> 
> > > wrote:
> > > 
> > > > On Thu, Sep 17, 2020 at 11:28:03PM +0200, Gustavo Pimentel wrote:
> > > > > Fixes warning given by executing "make C=2 drivers/pci/"
> > > > > 
> > > > > Sparse output:
> > > > > CHECK drivers/pci/controller/dwc/pcie-designware.c
> > > > >  drivers/pci/controller/dwc/pcie-designware.c:432:52: warning:
> > > > >  cast truncates bits from constant value (ffffffff7fffffff becomes
> > > > >  7fffffff)
> > > > > 
> > > > > Reported-by: Bjorn Helgaas <bhelgaas@google.com>
> > > > > Cc: Joao Pinto <jpinto@synopsys.com>
> > > > > Signed-off-by: Gustavo Pimentel <gustavo.pimentel@synopsys.com>
> > > > > ---
> > > > >  drivers/pci/controller/dwc/pcie-designware.c | 2 +-
> > > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > > 
> > > > > diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
> > > > > index 3c3a4d1..dcb7108 100644
> > > > > --- a/drivers/pci/controller/dwc/pcie-designware.c
> > > > > +++ b/drivers/pci/controller/dwc/pcie-designware.c
> > > > > @@ -429,7 +429,7 @@ void dw_pcie_disable_atu(struct dw_pcie *pci, int index,
> > > > >  	}
> > > > >  
> > > > >  	dw_pcie_writel_dbi(pci, PCIE_ATU_VIEWPORT, region | index);
> > > > > -	dw_pcie_writel_dbi(pci, PCIE_ATU_CR2, (u32)~PCIE_ATU_ENABLE);
> > > > > +	dw_pcie_writel_dbi(pci, PCIE_ATU_CR2, (u32)~0 & ~PCIE_ATU_ENABLE);
> > > > 
> > > > But this cure is worse than the disease.  If this is the only way to
> > > > fix the warning, I think I'd rather see the warning ;)  I'm hopeful
> > > > there's a nicer way, but I'm not a language lawyer.
> > > 
> > > I don't like it either, I tried to see if were another way a clean way 
> > > that didn't imply creating a temporary variable, but I didn't found.
> > > The issue here is that PCIE_ATU_ENABLE is defined as BIT(31) on 
> > > pcie-designware.h. The macro BIT changes its size from u32 to u64 
> > > according to the architecture and by inverting the value on the 64 bits 
> > > architecture causes the value to be transformed into 0xffffffff7fffffff.
> > > 
> > > The other possibility implies the creation of a temporary u32 variable to 
> > > overcome this issue. It's a little bit overkill, but please share your 
> > > thoughts about it.
> > > 
> > > void dw_pcie_disable_atu(struct dw_pcie *pci, int index,
> > >                          enum dw_pcie_region_type type)
> > >  {
> > > -       int region;
> > > +       u32 atu = PCIE_ATU_ENABLE;
> > > +       u32 region;
> > > 
> > >         switch (type) {
> > >         case DW_PCIE_REGION_INBOUND:
> > > @@ -429,7 +430,7 @@ void dw_pcie_disable_atu(struct dw_pcie *pci, int 
> > > index,
> > >         }
> > > 
> > >         dw_pcie_writel_dbi(pci, PCIE_ATU_VIEWPORT, region | index);
> > > -       dw_pcie_writel_dbi(pci, PCIE_ATU_CR2, (u32)~PCIE_ATU_ENABLE);
> > > +       dw_pcie_writel_dbi(pci, PCIE_ATU_CR2, ~atu);
> > >  }
> > 
> > Hi Bjorn,
> > 
> > What is your verdict on this?
> > If you prefer this approach I can send the corresponding patch.
> 
> Having a u32 temporary with no obvious reason for existence is kind of
> unpleasant, too.  Surely this is a common situation?  Maybe other
> instances just live with the warning, too?
> 
> I'd say leave this alone for now.

Hi Bjorn,

After resting about this topic, I found a easy and clean solution that 
will please for both of us.

-       dw_pcie_writel_dbi(pci, PCIE_ATU_CR2, (u32)~PCIE_ATU_ENABLE);
+       dw_pcie_writel_dbi(pci, PCIE_ATU_CR2, ~(u32)PCIE_ATU_ENABLE);

Basically I changed the cast instruction position, this way the 
PCIE_ATU_ENABLE flag defined BIT(31) on a 64 bits architecture will be 
casted into a u32 and the binary one's complement instruction will act 
upon a 32bits variable, avoiding that error.

I'll send the patch for it right away.

-Gustavo

> 
> Bjorn