From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([209.51.188.92]:59582) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gvhTH-0001i3-Rr for qemu-devel@nongnu.org; Mon, 18 Feb 2019 06:52:29 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1gvhTG-00037Y-FI for qemu-devel@nongnu.org; Mon, 18 Feb 2019 06:52:27 -0500 Date: Mon, 18 Feb 2019 11:52:18 +0000 From: Daniel =?utf-8?B?UC4gQmVycmFuZ8Op?= Message-ID: <20190218115218.GF32287@redhat.com> Reply-To: Daniel =?utf-8?B?UC4gQmVycmFuZ8Op?= References: <20190218101218.4177-1-ppandit@redhat.com> <20190218123811.5b7a1d61@bahia.lan> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <20190218123811.5b7a1d61@bahia.lan> Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [Qemu-ppc] [PATCH v3] ppc: add host-serial and host-model machine attributes List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Greg Kurz Cc: P J P , David Gibson , qemu-ppc@nongnu.org, QEMU Developers , Prasad J Pandit On Mon, Feb 18, 2019 at 12:38:11PM +0100, Greg Kurz wrote: > On Mon, 18 Feb 2019 15:42:18 +0530 > P J P wrote: >=20 > > From: Prasad J Pandit > >=20 > > On ppc hosts, hypervisor shares following system attributes > >=20 > > - /proc/device-tree/system-id > > - /proc/device-tree/model > >=20 > > with a guest. This could lead to information leakage and misuse.[*] > > Add machine attributes to control such system information exposure > > to a guest. > >=20 > > [*] https://wiki.openstack.org/wiki/OSSN/OSSN-0028 > >=20 > > Reported-by: Daniel P. Berrang=C3=A9 > > Fix-suggested-by: Daniel P. Berrang=C3=A9 > > Signed-off-by: Prasad J Pandit > > --- > > hw/ppc/spapr.c | 79 ++++++++++++++++++++++++++++++++++++++--= -- > > include/hw/ppc/spapr.h | 2 ++ > > 2 files changed, 75 insertions(+), 6 deletions(-) > >=20 > > Update v3: move host-serial,host-model options to ppc sPAPR machine > > -> https://lists.gnu.org/archive/html/qemu-devel/2019-02/msg03182.h= tml =20 > >=20 > > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c > > index 0942f35bf8..666e500376 100644 > > --- a/hw/ppc/spapr.c > > +++ b/hw/ppc/spapr.c > > @@ -1249,13 +1249,30 @@ static void *spapr_build_fdt(sPAPRMachineStat= e *spapr, > > * Add info to guest to indentify which host is it being run on > > * and what is the uuid of the guest > > */ > > - if (kvmppc_get_host_model(&buf)) { > > - _FDT(fdt_setprop_string(fdt, 0, "host-model", buf)); > > - g_free(buf); > > + if (spapr->host_model && !g_str_equal(spapr->host_model, "none")= ) { > > + if (g_str_equal(spapr->host_model, "passthrough")) { > > + /* -M host-model=3Dpassthrough */ > > + if (kvmppc_get_host_model(&buf)) { > > + _FDT(fdt_setprop_string(fdt, 0, "host-model", buf)); > > + g_free(buf); > > + } > > + } else { > > + /* -M host-model=3D */ > > + _FDT(fdt_setprop_string(fdt, 0, "host-model", spapr->hos= t_model)); > > + } > > } > > - if (kvmppc_get_host_serial(&buf)) { > > - _FDT(fdt_setprop_string(fdt, 0, "host-serial", buf)); > > - g_free(buf); > > + > > + if (spapr->host_serial && !g_str_equal(spapr->host_serial, "none= ")) { > > + if (g_str_equal(spapr->host_serial, "passthrough")) { > > + /* -M host-serial=3Dpassthrough */ > > + if (kvmppc_get_host_serial(&buf)) { > > + _FDT(fdt_setprop_string(fdt, 0, "host-serial", buf))= ; > > + g_free(buf); > > + } > > + } else { > > + /* -M host-serial=3D */ > > + _FDT(fdt_setprop_string(fdt, 0, "host-serial", spapr->ho= st_serial)); > > + } > > } > > =20 > > buf =3D qemu_uuid_unparse_strdup(&qemu_uuid); > > @@ -3138,6 +3155,36 @@ static void spapr_set_ic_mode(Object *obj, con= st char *value, Error **errp) > > } > > } > > =20 > > +static char *spapr_get_host_model(Object *obj, Error **errp) > > +{ > > + sPAPRMachineState *spapr =3D SPAPR_MACHINE(obj); > > + > > + return g_strdup(spapr->host_model); > > +} > > + > > +static void spapr_set_host_model(Object *obj, const char *value, Err= or **errp) > > +{ > > + sPAPRMachineState *spapr =3D SPAPR_MACHINE(obj); > > + > > + g_free(spapr->host_model); > > + spapr->host_model =3D g_strdup(value); > > +} > > + > > +static char *spapr_get_host_serial(Object *obj, Error **errp) > > +{ > > + sPAPRMachineState *spapr =3D SPAPR_MACHINE(obj); > > + > > + return g_strdup(spapr->host_serial); > > +} > > + > > +static void spapr_set_host_serial(Object *obj, const char *value, Er= ror **errp) > > +{ > > + sPAPRMachineState *spapr =3D SPAPR_MACHINE(obj); > > + > > + g_free(spapr->host_serial); > > + spapr->host_serial =3D g_strdup(value); > > +} > > + > > static void spapr_instance_init(Object *obj) > > { > > sPAPRMachineState *spapr =3D SPAPR_MACHINE(obj); > > @@ -3183,6 +3230,20 @@ static void spapr_instance_init(Object *obj) > > object_property_set_description(obj, "ic-mode", > > "Specifies the interrupt controller mode (xics, xiv= e, dual)", > > NULL); > > + > > + spapr->host_model =3D NULL; >=20 > This isn't needed since object_initialize_with_type() already takes car= e > of zeroing the instance for us. >=20 > > + object_property_add_str(obj, "host-model", > > + spapr_get_host_model, spapr_set_host_model, > > + &error_abort); > > + object_property_set_description(obj, "host-model", > > + "Set host's model-id to use - none|passthrough|string", &err= or_abort); > > + > > + spapr->host_serial =3D NULL; >=20 > Same here. >=20 > > + object_property_add_str(obj, "host-serial", > > + spapr_get_host_serial, spapr_set_host_serial, > > + &error_abort); > > + object_property_set_description(obj, "host-serial", > > + "Set host's system-id to use - none|passthrough|string", &er= ror_abort); > > } > > =20 > > static void spapr_machine_finalizefn(Object *obj) > > @@ -4080,9 +4141,15 @@ DEFINE_SPAPR_MACHINE(4_0, "4.0", true); > > static void spapr_machine_3_1_class_options(MachineClass *mc) > > { > > sPAPRMachineClass *smc =3D SPAPR_MACHINE_CLASS(mc); > > + static GlobalProperty compat[] =3D { > > + { TYPE_SPAPR_MACHINE, "host-model", "passthrough" }, > > + { TYPE_SPAPR_MACHINE, "host-serial", "passthrough" }, > > + }; > > =20 >=20 > So... we don't fix the information leak for older machines by default ?= From > previous discussions, I understand it is for the sake of compatibility,= but > leaving the burden of securing the host to downstream or to the user st= ill > looks unsecure to me FWIW. Maintaining guest ABI compatibility has to take priority, even over fixing security issues, because we must never intentionally break guest OS/applications by silently altering guest ABI. This is one of the two reasons why machine type versioning exists (the other reason being live migration data stream). This is nothing new - we've done it before for security flaws where a fix would involve changing guest ABI. This particular security flaw is pretty minor compared to other cases that we've left unfixed by default eg Meltdown / Spectre and is easily addressed by the user if needed. Regards, Daniel --=20 |: https://berrange.com -o- https://www.flickr.com/photos/dberran= ge :| |: https://libvirt.org -o- https://fstop138.berrange.c= om :| |: https://entangle-photo.org -o- https://www.instagram.com/dberran= ge :|