Re: xen/pciback: Drop two backends, squash and cleanup some code.

Re: xen/pciback: Drop two backends, squash and cleanup some code.

[Dan Carpenter]

Hello Konrad Rzeszutek Wilk,

The patch a92336a1176b: "xen/pciback: Drop two backends, squash and
cleanup some code." from Jul 19, 2011, leads to the following static
checker warning:

	drivers/xen/xen-pciback/conf_space_capability.c:163 pm_ctrl_init()
	error: passing non negative 135 to ERR_PTR

drivers/xen/xen-pciback/conf_space_capability.c
   147  /* Ensure PMEs are disabled */
   148  static void *pm_ctrl_init(struct pci_dev *dev, int offset)
   149  {
   150          int err;
   151          u16 value;
   152  
   153          err = pci_read_config_word(dev, offset, &value);
   154          if (err)
   155                  goto out;
   156  
   157          if (value & PCI_PM_CTRL_PME_ENABLE) {
   158                  value &= ~PCI_PM_CTRL_PME_ENABLE;
   159                  err = pci_write_config_word(dev, offset, value);

The static check is complaining that pci_write_config_word() can
return PCIBIOS_BAD_REGISTER_NUMBER, but actually I think that's not
possible.

Anyway, this function is only called from
xen_pcibk_config_add_field_offset() so why are we returning a pointer
instead of just int?

   160          }
   161  
   162  out:
   163          return ERR_PTR(err);
   164  }

regards,
dan carpenter

Re: xen/pciback: Drop two backends, squash and cleanup some code.

[Konrad Rzeszutek Wilk]

On Fri, Oct 31, 2014 at 06:30:01PM +0300, Dan Carpenter wrote:
> Hello Konrad Rzeszutek Wilk,
> 
> The patch a92336a1176b: "xen/pciback: Drop two backends, squash and
> cleanup some code." from Jul 19, 2011, leads to the following static
> checker warning:
> 
> 	drivers/xen/xen-pciback/conf_space_capability.c:163 pm_ctrl_init()
> 	error: passing non negative 135 to ERR_PTR
> 
> drivers/xen/xen-pciback/conf_space_capability.c
>    147  /* Ensure PMEs are disabled */
>    148  static void *pm_ctrl_init(struct pci_dev *dev, int offset)
>    149  {
>    150          int err;
>    151          u16 value;
>    152  
>    153          err = pci_read_config_word(dev, offset, &value);
>    154          if (err)
>    155                  goto out;
>    156  
>    157          if (value & PCI_PM_CTRL_PME_ENABLE) {
>    158                  value &= ~PCI_PM_CTRL_PME_ENABLE;
>    159                  err = pci_write_config_word(dev, offset, value);
> 
> The static check is complaining that pci_write_config_word() can
> return PCIBIOS_BAD_REGISTER_NUMBER, but actually I think that's not
> possible.
> 
> Anyway, this function is only called from
> xen_pcibk_config_add_field_offset() so why are we returning a pointer
> instead of just int?

Because all the other 'init' could. And 'bar_init' for example
returns the BAR value (wrapped in 'struct pci_bar_info').
> 
>    160          }
>    161  
>    162  out:
>    163          return ERR_PTR(err);
>    164  }
> 
> regards,
> dan carpenter