sparse warnings

sparse warnings

[Bjorn Helgaas]

Just FYI, I see the following sparse warnings (among others):

  $ make C=2 drivers/pci/

  drivers/pci/endpoint/functions/pci-epf-test.c:288:24: warning: incorrect type in argument 1 (different address spaces)
  drivers/pci/endpoint/functions/pci-epf-test.c:288:24:    expected void *to
  drivers/pci/endpoint/functions/pci-epf-test.c:288:24:    got void [noderef] <asn:2> *[assigned] dst_addr
  drivers/pci/endpoint/functions/pci-epf-test.c:288:34: warning: incorrect type in argument 2 (different address spaces)
  drivers/pci/endpoint/functions/pci-epf-test.c:288:34:    expected void const *from
  drivers/pci/endpoint/functions/pci-epf-test.c:288:34:    got void [noderef] <asn:2> *[assigned] src_addr

  drivers/pci/controller/dwc/pcie-designware.c:447:52: warning: cast truncates bits from constant value (ffffffff7fffffff becomes 7fffffff)

It'd be nice to fix these if it's practical.

There are a bunch more sparse warnings about pci_power_t in the
generic code.  Not sure what those would involve, so they're a problem
for another day.

Bjorn

Re: sparse warnings

[Bjorn Helgaas]

On Wed, Jul 29, 2020 at 02:48:15PM -0500, Bjorn Helgaas wrote:
> Just FYI, I see the following sparse warnings (among others):
> 
>   $ make C=2 drivers/pci/
> 
>   drivers/pci/endpoint/functions/pci-epf-test.c:288:24: warning: incorrect type in argument 1 (different address spaces)
>   drivers/pci/endpoint/functions/pci-epf-test.c:288:24:    expected void *to
>   drivers/pci/endpoint/functions/pci-epf-test.c:288:24:    got void [noderef] <asn:2> *[assigned] dst_addr
>   drivers/pci/endpoint/functions/pci-epf-test.c:288:34: warning: incorrect type in argument 2 (different address spaces)
>   drivers/pci/endpoint/functions/pci-epf-test.c:288:34:    expected void const *from
>   drivers/pci/endpoint/functions/pci-epf-test.c:288:34:    got void [noderef] <asn:2> *[assigned] src_addr
> 
>   drivers/pci/controller/dwc/pcie-designware.c:447:52: warning: cast truncates bits from constant value (ffffffff7fffffff becomes 7fffffff)
> 
> It'd be nice to fix these if it's practical.

Any ideas about these?

Bjorn

Re: sparse warnings

[Kishon Vijay Abraham I]

+RobH, Lorenzo

Hi Bjorn,

On 01/10/20 4:09 am, Bjorn Helgaas wrote:
> On Wed, Jul 29, 2020 at 02:48:15PM -0500, Bjorn Helgaas wrote:
>> Just FYI, I see the following sparse warnings (among others):
>>
>>   $ make C=2 drivers/pci/
>>
>>   drivers/pci/endpoint/functions/pci-epf-test.c:288:24: warning: incorrect type in argument 1 (different address spaces)
>>   drivers/pci/endpoint/functions/pci-epf-test.c:288:24:    expected void *to
>>   drivers/pci/endpoint/functions/pci-epf-test.c:288:24:    got void [noderef] <asn:2> *[assigned] dst_addr
>>   drivers/pci/endpoint/functions/pci-epf-test.c:288:34: warning: incorrect type in argument 2 (different address spaces)
>>   drivers/pci/endpoint/functions/pci-epf-test.c:288:34:    expected void const *from
>>   drivers/pci/endpoint/functions/pci-epf-test.c:288:34:    got void [noderef] <asn:2> *[assigned] src_addr
>>
>>   drivers/pci/controller/dwc/pcie-designware.c:447:52: warning: cast truncates bits from constant value (ffffffff7fffffff becomes 7fffffff)
>>
>> It'd be nice to fix these if it's practical.
> 
> Any ideas about these?

pci-epf-test here uses memcpy() to copy data from one memory mapped IO
address to other memory mapped IO address. Other places in this driver
uses memcpy_fromio() and memcpy_toio() since only one of them is memory
mapped IO address.

So the option is to either use memcpy_fromio() to copy data from IO to
temporary buffer and then use memcpy_toio() to copy data from the
temporary buffer to the memory mapped IO or create a new variant
[memcpy_fromtoio()?] that takes both memory mapped IO address.

Thanks
Kishon