On 16-08-10 08:54:36, Mimi Zohar wrote:
> On Wed, 2016-08-10 at 19:52 +1000, Michael Ellerman wrote:
> > Thiago Jung Bauermann <bauerman at linux.vnet.ibm.com> writes:
> >
> > > Am Mittwoch, 10 August 2016, 13:41:08 schrieb Michael Ellerman:
> > >> Thiago Jung Bauermann <bauerman at linux.vnet.ibm.com> writes:
> > >> > Am Dienstag, 09 August 2016, 09:01:13 schrieb Mimi Zohar:
> > >> >> On Tue, 2016-08-09 at 20:59 +1000, Michael Ellerman wrote:
> > >> >> > Mimi Zohar <zohar at linux.vnet.ibm.com> writes:
> > >> >> > > +/* Some details preceding the binary serialized measurement list
> > >> >> > > */
> > >> >> > > +struct ima_kexec_hdr {
> > >> >> > > + unsigned short version;
> > >> >> > > + unsigned long buffer_size;
> > >> >> > > + unsigned long count;
> > >> >> > > +} __packed;
> > >> >> > > +
> > >> >> >
> > >> >> > Am I understanding it correctly that this structure is passed between
> > >> >> > kernels?
> > >> >>
> > >> >> Yes, the header prefixes the measurement list, which is being passed on
> > >> >> the same computer to the next kernel. Could the architecture (eg.
> > >> >> LE/BE) change between soft re-boots?
> > >> >
> > >> > Yes. I am able to boot a BE kernel from an LE kernel with my patches.
> > >> > Whether we want to support that or not is another question...
> > >>
> > >> Yes you must support that. BE -> LE and vice versa.
> > >
> > > I didn't test BE - LE yet, but will do.
> >
> > Thanks.
>
> Ok. There have been requests for making the binary_runtime_measurements
> architecture independent. As this was not a network facing interface,
> we left it in native format. With the kernel now consuming this data,
> it makes sense for the binary_runtime_measurements to be in an
> architecture independent format.
>
> Unfortunately, as the <securityfs>/ima/binary_runtime_measurements is
> not prefixed with any metadata, this change would need to be Kconfig
> based, but kexec would always use the architecture independent format.
>
> > >> You should also consider the possibility that the next kernel is not
> > >> Linux.
>
> Oh!
>
> > > If the next kernel is an ELF binary and it supports the kexec "calling
> > > convention", it should work too. What could possibly go wrong? I can try
> > > FreeBSD (I suppose it's an ELF kernel) and see what happens.
> >
> > At least for old style kexec (not sys_kexec_load()) I don't think it
> > even needs to be an ELF binary.
> >
> > I think there are folks working on FreeBSD (or $?BSD), so I think the
> > basic kexec part works.
> >
> > There's nothing (yet) that wants to use this measurement list obviously,
> > but it should be designed such that it could be used by an unknown
> > future kernel that knows the ABI.
> >
> > So given what you have above, you'd use something like:
> >
> > struct ima_kexec_hdr {
> > u16 version;
> > u16 _reserved0;
> > u32 _reserved1;
> > u64 buffer_size;
> > u64 count;
> > };
> >
> > cheers
>
> Thanks, I'll make this change.
I would suggest:
struct ima_kexec_hdr {
u64 buffer_size;
u64 count;
u16 version;
};
and let the compiler add the proper padding, depending on the architecture. On
32bit machine we'll have 4 bytes smaller allocations (compared to 64bit) while
retaining the same functionality.
cheers,
Petko
