Re: [PATCH v2] PCI: dwc: Fix W=1 build warning in dw_pcie_edma_irq_verify()

[Niklas Cassel <cassel@kernel.org>]

On Fri, Jan 03, 2025 at 04:10:56PM -0600, Bjorn Helgaas wrote:
> Can you make the subject say something about the fix instead of the
> warning?  E.g., something about fixing a potential truncation?
> 
> On Thu, Jan 02, 2025 at 12:13:40PM +0100, Niklas Cassel wrote:
> > Change dw_pcie_edma_irq_verify() to print the dma channel as %u.
> > 
> > While a DWC glue driver could theoretically initialize nr_irqs to a
> > negative value, doing so would obviously be incorrect, and the later
> > dw_edma_probe(struct dw_edma_chip *chip) call would fail, since while
> > the dw_edma_probe() call expects the caller to initialize chip->nr_irqs,
> > dw_edma_probe() verifies nr_irqs and returns failure if nr_irqs is < 1.
> > 
> > This fixes the following build warning when compiling with W=1:
> > 
> > drivers/pci/controller/dwc/pcie-designware.c: In function ‘dw_pcie_edma_detect’:
> > drivers/pci/controller/dwc/pcie-designware.c:989:50: warning: ‘%d’ directive output may be truncated writing between 1 and 11 bytes into a region of size 3 [-Wformat-truncation=]
> >   989 |                 snprintf(name, sizeof(name), "dma%d", pci->edma.nr_irqs);
> >       |                                                  ^~
> > 
> > Signed-off-by: Niklas Cassel <cassel@kernel.org>
> > ---
> > Changes since V1:
> > -Do not reject negative nr_irqs value in dw_pcie_edma_irq_verify(),
> >  as this will already be done by dw_edma_probe().
> > 
> >  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 3c683b6119c3..0a13fb4336f4 100644
> > --- a/drivers/pci/controller/dwc/pcie-designware.c
> > +++ b/drivers/pci/controller/dwc/pcie-designware.c
> > @@ -986,7 +986,7 @@ static int dw_pcie_edma_irq_verify(struct dw_pcie *pci)
> >  	}
> >  
> >  	for (; pci->edma.nr_irqs < ch_cnt; pci->edma.nr_irqs++) {
> > -		snprintf(name, sizeof(name), "dma%d", pci->edma.nr_irqs);
> > +		snprintf(name, sizeof(name), "dma%u", pci->edma.nr_irqs);
> 
> I don't understand this fix.  I guess the warning is complaining that
> sizeof(name) == 6, and "dma" takes up three bytes, so the %d has to
> fit in the remaining region of size 3?

Yes.


> 
> But I don't see how printing nr_irqs as unsigned rather than signed is
> a fix, since even an unsigned int can be longer than 3 digits.

You would need to ask GCC authors behind -Wformat-truncation why the
warning only seems to care about %d.


> 
> And I don't like using "%u" for a signed value in order to "fix"
> something.  That's asking for a future cleanup to revert the change.
>

Well, neither do I. V1 was a nicer solution IMO:
https://lore.kernel.org/linux-pci/20241220072328.351329-2-cassel@kernel.org/
But that fix was rejected by another PCI maintainer.


> What's wrong with just making the name[] buffer big enough?

Sure, I'll do that in v3.


Kind regards,
Niklas