[libvirt] [PATCH 12/15] util: pci: Introduce virPCIGetMdevTypes helper

[Michal Privoznik]

On 01/29/2018 01:24 PM, Erik Skultety wrote:
> On Fri, Jan 26, 2018 at 12:39:00PM +0100, Michal Privoznik wrote:
>> On 01/25/2018 10:23 AM, Erik Skultety wrote:
>>> This is a replacement for the existing udevPCIGetMdevTypesCap which is
>>> static to the udev backend. This simple helper constructs the sysfs path
>>> from the device's base path for each mdev type and queries the
>>> corresponding attributes of that type.
>>>
>>> Signed-off-by: Erik Skultety <eskultet at redhat.com>
>>> ---
>>>  src/libvirt_private.syms |  1 +
>>>  src/util/virpci.c        | 58 ++++++++++++++++++++++++++++++++++++++++++++++++
>>>  src/util/virpci.h        |  4 ++++
>>>  3 files changed, 63 insertions(+)
>>>
>>> diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms
>>> index 75eaf1d4c..8d4c8dd3f 100644
>>> --- a/src/libvirt_private.syms
>>> +++ b/src/libvirt_private.syms
>>> @@ -2456,6 +2456,7 @@ virPCIDeviceWaitForCleanup;
>>>  virPCIEDeviceInfoFree;
>>>  virPCIGetDeviceAddressFromSysfsLink;
>>>  virPCIGetHeaderType;
>>> +virPCIGetMdevTypes;
>>>  virPCIGetNetName;
>>>  virPCIGetPhysicalFunction;
>>>  virPCIGetVirtualFunctionIndex;
>>> diff --git a/src/util/virpci.c b/src/util/virpci.c
>>> index fe57bef32..12d7ef0e4 100644
>>> --- a/src/util/virpci.c
>>> +++ b/src/util/virpci.c
>>> @@ -3027,6 +3027,64 @@ virPCIGetVirtualFunctionInfo(const char *vf_sysfs_device_path,
>>>      return ret;
>>>  }
>>>
>>> +
>>> +int
>>> +virPCIGetMdevTypes(const char *sysfspath,
>>> +                   virMediatedDeviceTypePtr **types)
>>
>> Since this function returns size_t on success, I guess the retval should
>> be type of ssize_t at least. We are not guaranteed that size_t will fit
> 
> ssize_t wouldn't really help, since assigning size_t might overflow, so the
> only safe bet is long long, but I mean, do you really expect there to be more
> than INT_MAX mdev types for a device? That would be a lot of types to support.

In kernel, they do a lot of size_t -> ssize_t conversion in cases
where's it's next to impossible to overflow. And I believe this is one
of them.

Michal