Re: [Qemu-devel] [PATCH 1/4] qemu-user: Impl. setsockopt(SO_BINDTODEVICE)

[Peter Maydell]

On 12 July 2014 18:30, Joakim Tjernlund <address@hidden> wrote:
> Peter Maydell <address@hidden> wrote on 2014/07/12 17:47:56:
>> That would work with the current kernel implementation, but
>> the API definition (as described in the socket(7) manpage)
>> says we should pass a NUL-terminated string to the kernel,
>> so it's more robust for us to make sure we do so.
>
> But we emulate user space and we should not "improve" stuff on the way
> should we?

No, but in this case we don't improve anything. The unterminated
string from the guest is still truncated as a native kernel would do;
it's just that our implementation of that behaviour doesn't rely
on the specific behaviour of the host kernel in this undocumented
corner case. It's not a big deal either way, but given that the code
is pretty similar either way I prefer the approach I suggest.

thanks
-- PMM