On Tue, Jul 14, 2015 at 9:02 PM, Wei Yang <weiyang@xxxxxxxxxxxxxxxxxx> wrote:
> On Tue, Jul 14, 2015 at 05:00:21PM -0500, Bjorn Helgaas wrote:
>>On Tue, Jun 30, 2015 at 09:16:44AM +0800, Wei Yang wrote:
>>> The return value of the pci_find_(ext_)capability is the position of this
>>> Cap. After previous two patches clean up, the position returned is an
>>> unsigned value. Only 0 indicates the Cap is not presented.
>>>
>>> This patch consolidates the form of check from (pos <= 0)to (!pos).
>>>
>>> Signed-off-by: Wei Yang <weiyang@xxxxxxxxxxxxxxxxxx>
>>
>>Applied to pci/misc with changelog as below.
>>
>>It seems pretty clear to me that pci_find_capability() returns either 0 or
>>a u8 value. pci_find_ext_capability() does return an int. It looks to me
>>like it can never be negative, but if you wanted it to be even more clear,
>>you could easily change just pci_find_next_ext_capability() to use a u16
>>for "pos". That would be very simple and wouldn't change any interfaces.
>
> pci_find_capability() will return either 0 or a u8 value, while in the code
> the return value is an "int" type. So for the first sight, it may not that
> immediate. The same as pci_find_ext_capability().
>
> This is the reason for patch 2/3. The purpose is to make the return type
> reflect the value it will return.
>
> Patch 3 does exactly what you said, use a u16 for "pos" in
> pci_find_next_ext_capability().
Yes. Patch 3 contains this hunk:
@@ -272,11 +272,11 @@ EXPORT_SYMBOL(pci_bus_find_capability);
* not support it. Some capabilities can occur several times, e.g., the
* vendor-specific capability, and this provides a way to find them all.
*/
-int pci_find_next_ext_capability(struct pci_dev *dev, int start, int cap)
+u16 pci_find_next_ext_capability(struct pci_dev *dev, int start, u16 cap)
{
u32 header;
int ttl;
- int pos = PCI_CFG_SPACE_SIZE;
+ u16 pos = PCI_CFG_SPACE_SIZE;
/* minimum 8 bytes per capability */
ttl = (PCI_CFG_SPACE_EXP_SIZE - PCI_CFG_SPACE_SIZE) / 8;
I think the second piece (changing "pos" from int to u16) is enough to
make it easy to analyze, and it only affects this function, so that
seems obviously fine.
The first piece (changing the return type) changes the signature of an
exported interface. That is more work, e.g., it's more of a hassle
for distros to backport if they want to preserve interfaces, so it's
not quite as obvious that it's worthwhile.
Bjorn
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
