On Mon, 6 Jul 2020 14:49:43 +0200
Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > +struct mon_bin_get {
> > + struct mon_bin_hdr __user *hdr; /* Can be 48 bytes or 64. */
> > + void __user *data;
> > + size_t alloc; /* Length of data (can be zero) */
>
> is size_t a value we can pass across user/kernel boundry? Are you sure
> this isn't __kernel_size_t?
Sorry, it was my fault letting that one through.
Since currently the definitions are separate, userland uses their size_t,
and kernel uses in-kernel size_t. We have a set of MON_IOCX_GET and
MON_IOCX_GET32 with the same base number 6, but using mon_bin_get32,
so the resulting ioctl number magically matched what 32- and 64-bit
applications used. We don't even need an adaptation layer that
re-encodes the argument structure.
Not sure how to resolve this properly once we attempt to export the
structures. Something for the patch submitter to work out, I suppose.
-- Pere
