On Thu, 5 Nov 2020 07:56:08 +0530 Anant Thazhemadam wrote:
> On 05/11/20 5:54 am, Jakub Kicinski wrote:
> > On Mon, 2 Nov 2020 23:09:46 +0530 Anant Thazhemadam wrote:
> >> Currently, __usbnet_{read|write}_cmd() use usb_control_msg().
> >> However, this could lead to potential partial reads/writes being
> >> considered valid, and since most of the callers of
> >> usbnet_{read|write}_cmd() don't take partial reads/writes into account
> >> (only checking for negative error number is done), and this can lead to
> >> issues.
> >>
> >> However, the new usb_control_msg_{send|recv}() APIs don't allow partial
> >> reads and writes.
> >> Using the new APIs also relaxes the return value checking that must
> >> be done after usbnet_{read|write}_cmd() is called.
> >>
> >> Signed-off-by: Anant Thazhemadam <anant.thazhemadam@xxxxxxxxx>
> > So you're changing the semantics without updating the callers?
> >
> > I'm confused.
> >
> > Is this supposed to be applied to some tree which already has the
> > callers fixed?
> >
> > At a quick scan at least drivers/net/usb/plusb.c* would get confused
> > as it compares the return value to zero and 0 used to mean "nothing
> > transferred", now it means "all good", no?
> >
> > * I haven't looked at all the other callers
>
> I see. I checked most of the callers that directly called the functions,
> but it seems to have slipped my mind that these callers were also
> wrappers, and to check the callers for these wrapper.
> I apologize for the oversight.
> I'll perform a more in-depth analysis of all the callers, fix this mistake,
> and send in a patch series instead, that update all the callers too.
> Would that be alright?
Yes. Probably best if you rename the existing function as first patch,
then add a new one with the old name using usb_control_msg_{send|recv}()
then switch the callers one by one, and finally remove the old renamed
function.
