@@ -605,6 +605,58 @@ option should only be used with a trusted device tree.
Note that the partial device tree should avoid using the phandle 65000
which is reserved by the toolstack.
+=item B<passthrough="STRING">
+
+Specify whether IOMMU mappings are enabled for the domain and hence whether
+it will be enabled for passthrough hardware. Valid values for this option
+are:
+
+=over 4
+
+=item B<disabled>
+
+IOMMU mappings are disabled for the domain and so hardware may not be
+passed through.
+
+This option is the default if no passthrough hardware is specified
+in the domain's configuration.
+
+=item B<sync_pt>
+
+This option means that IOMMU mappings will be synchronized with the
+domain's P2M table as follows:
+
+For a PV domain, all writable pages assigned to the domain are identity
+mapped by MFN in the IOMMU page table. Thus a device driver running in the
+domain may program passthrough hardware for DMA using MFN values
+(i.e. host/machine frame numbers) looked up in its P2M.
+
+For an HVM domain, all non-foreign RAM pages present in its P2M will be
+mapped by GFN in the IOMMU page table. Thus a device driver running in the
+domain may program passthrough hardware using GFN values (i.e. guest
+physical frame numbers) without any further translation.
+
+This option is the default if the domain is PV and passthrough hardware
+is specified in the configuration.
+
+This option is not available on Arm.
+
+=item B<share_pt>
+
+This option is unavailable for a PV domain. For an HVM domain, this option
+means that the IOMMU will be programmed to directly reference the domain's
+P2M table as its page table. From the point of view of a device driver
+running in the domain this is functionally equivalent to B<sync_pt> but
+places less load on the hypervisor and so should generally be selected in
+preference. However, the availability of this option is hardware specific
+and thus, if it is specified for a domain running on hardware that does
+not allow it (e.g. AMD), B<sync_pt> will be used instead.
+
+This option is the default if the domain is HVM and passthrough hardware
+is specified in the configuration.
+
+=back
+
=back
=head2 Devices
@@ -394,6 +394,11 @@
*/
#define LIBXL_HAVE_EXTENDED_VKB 1
+/*
+ * libxl_domain_create_info has libxl_passthrough enumeration.
+ */
+#define LIBXL_HAVE_CREATEINFO_PASSTHROUGH 1
+
/*
* libxl ABI compatibility
*
@@ -564,6 +564,15 @@ int libxl__domain_make(libxl__gc *gc, libxl_domain_config *d_config,
libxl_defbool_val(info->oos) ? 0 : XEN_DOMCTL_CDF_oos_off;
}
+ LOG(DETAIL, "passthrough: %s",
+ libxl_passthrough_to_string(info->passthrough));
+
+ if (info->passthrough != LIBXL_PASSTHROUGH_DISABLED)
+ create.flags |= XEN_DOMCTL_CDF_iommu;
+
+ if (info->passthrough == LIBXL_PASSTHROUGH_SYNC_PT)
+ create.iommu_opts |= XEN_DOMCTL_IOMMU_no_sharept;
+
/* Ultimately, handle is an array of 16 uint8_t, same as uuid */
libxl_uuid_copy(ctx, (libxl_uuid *)&create.handle, &info->uuid);
@@ -263,6 +263,12 @@ libxl_vkb_backend = Enumeration("vkb_backend", [
(2, "LINUX")
])
+libxl_passthrough = Enumeration("passthrough", [
+ (0, "disabled"),
+ (1, "sync_pt"),
+ (2, "share_pt"),
+ ])
+
#
# Complex libxl types
#
@@ -408,6 +414,7 @@ libxl_domain_create_info = Struct("domain_create_info",[
("pool_name", string),
("run_hotplug_scripts",libxl_defbool),
("driver_domain",libxl_defbool),
+ ("passthrough", libxl_passthrough),
], dir=DIR_IN)
libxl_domain_restore_params = Struct("domain_restore_params", [
@@ -64,11 +64,15 @@ type domain_create_flag =
| CDF_XS_DOMAIN
| CDF_IOMMU
+type domain_create_iommu_opts =
+ | IOMMU_NO_SHAREPT
+
type domctl_create_config =
{
ssidref: int32;
handle: string;
flags: domain_create_flag list;
+ iommu_opts: domain_create_iommu_opts list;
max_vcpus: int;
max_evtchn_port: int;
max_grant_frames: int;
@@ -57,10 +57,14 @@ type domain_create_flag =
| CDF_XS_DOMAIN
| CDF_IOMMU
+type domain_create_iommu_opts =
+ | IOMMU_NO_SHAREPT
+
type domctl_create_config = {
ssidref: int32;
handle: string;
flags: domain_create_flag list;
+ iommu_opts: domain_create_iommu_opts list;
max_vcpus: int;
max_evtchn_port: int;
max_grant_frames: int;
@@ -128,11 +128,12 @@ CAMLprim value stub_xc_domain_create(value xch, value config)
#define VAL_SSIDREF Field(config, 0)
#define VAL_HANDLE Field(config, 1)
#define VAL_FLAGS Field(config, 2)
-#define VAL_MAX_VCPUS Field(config, 3)
-#define VAL_MAX_EVTCHN_PORT Field(config, 4)
-#define VAL_MAX_GRANT_FRAMES Field(config, 5)
-#define VAL_MAX_MAPTRACK_FRAMES Field(config, 6)
-#define VAL_ARCH Field(config, 7)
+#define VAL_IOMMU_OPTS Field(config, 3)
+#define VAL_MAX_VCPUS Field(config, 4)
+#define VAL_MAX_EVTCHN_PORT Field(config, 5)
+#define VAL_MAX_GRANT_FRAMES Field(config, 6)
+#define VAL_MAX_MAPTRACK_FRAMES Field(config, 7)
+#define VAL_ARCH Field(config, 8)
uint32_t domid = 0;
int result;
@@ -149,6 +150,9 @@ CAMLprim value stub_xc_domain_create(value xch, value config)
for ( l = VAL_FLAGS; l != Val_none; l = Field(l, 1) )
cfg.flags |= 1u << Int_val(Field(l, 0));
+ for ( l = VAL_IOMMU_OPTS; l != Val_none; l = Field(l, 1) )
+ cfg.iommu_opts |= 1u << Int_val(Field(l, 0));
+
arch_domconfig = Field(VAL_ARCH, 0);
switch ( Tag_val(VAL_ARCH) )
{
@@ -181,6 +185,7 @@ CAMLprim value stub_xc_domain_create(value xch, value config)
#undef VAL_MAX_GRANT_FRAMES
#undef VAL_MAX_EVTCHN_PORT
#undef VAL_MAX_VCPUS
+#undef VAL_IOMMU_OPTS
#undef VAL_FLAGS
#undef VAL_HANDLE
#undef VAL_SSIDREF
@@ -1448,6 +1448,94 @@ void parse_config_data(const char *config_source,
exit(1);
}
+ if (!xlu_cfg_get_list (config, "pci", &pcis, 0, 0)) {
+ d_config->num_pcidevs = 0;
+ d_config->pcidevs = NULL;
+ for(i = 0; (buf = xlu_cfg_get_listitem (pcis, i)) != NULL; i++) {
+ libxl_device_pci *pcidev;
+
+ pcidev = ARRAY_EXTEND_INIT_NODEVID(d_config->pcidevs,
+ d_config->num_pcidevs,
+ libxl_device_pci_init);
+ pcidev->msitranslate = pci_msitranslate;
+ pcidev->power_mgmt = pci_power_mgmt;
+ pcidev->permissive = pci_permissive;
+ pcidev->seize = pci_seize;
+ /*
+ * Like other pci option, the per-device policy always follows
+ * the global policy by default.
+ */
+ pcidev->rdm_policy = b_info->u.hvm.rdm.policy;
+ e = xlu_pci_parse_bdf(config, pcidev, buf);
+ if (e) {
+ fprintf(stderr,
+ "unable to parse PCI BDF `%s' for passthrough\n",
+ buf);
+ exit(-e);
+ }
+ }
+ if (d_config->num_pcidevs && c_info->type == LIBXL_DOMAIN_TYPE_PV)
+ libxl_defbool_set(&b_info->u.pv.e820_host, true);
+ }
+
+ if (!xlu_cfg_get_list (config, "dtdev", &dtdevs, 0, 0)) {
+ d_config->num_dtdevs = 0;
+ d_config->dtdevs = NULL;
+ for (i = 0; (buf = xlu_cfg_get_listitem(dtdevs, i)) != NULL; i++) {
+ libxl_device_dtdev *dtdev;
+
+ dtdev = ARRAY_EXTEND_INIT_NODEVID(d_config->dtdevs,
+ d_config->num_dtdevs,
+ libxl_device_dtdev_init);
+
+ dtdev->path = strdup(buf);
+ if (dtdev->path == NULL) {
+ fprintf(stderr, "unable to duplicate string for dtdevs\n");
+ exit(-1);
+ }
+ }
+ }
+
+ if (!xlu_cfg_get_string(config, "passthrough", &buf, 0)) {
+ libxl_passthrough o;
+
+ e = libxl_passthrough_from_string(buf, &o);
+ if (e) {
+ fprintf(stderr,
+ "ERROR: unknown passthrough option '%s'\n",
+ buf);
+ exit(-ERROR_FAIL);
+ }
+
+ switch (o) {
+ case LIBXL_PASSTHROUGH_DISABLED:
+ if (d_config->num_pcidevs || d_config->num_dtdevs) {
+ fprintf(stderr,
+ "ERROR: passthrough disabled but devices are specified\n");
+ exit(-ERROR_FAIL);
+ }
+ break;
+ case LIBXL_PASSTHROUGH_SHARE_PT:
+ if (c_info->type == LIBXL_DOMAIN_TYPE_PV) {
+ fprintf(stderr,
+ "ERROR: passthrough=\"share_pt\" not valid for PV domain\n");
+ exit(-ERROR_FAIL);
+ }
+ break;
+ default:
+ break;
+ }
+
+ c_info->passthrough = o;
+ } else if (d_config->num_pcidevs || d_config->num_dtdevs) {
+ /*
+ * Passthrough devices are specified so set an appropriate
+ * default value.
+ */
+ c_info->passthrough = (c_info->type == LIBXL_DOMAIN_TYPE_PV) ?
+ LIBXL_PASSTHROUGH_SYNC_PT : LIBXL_PASSTHROUGH_SHARE_PT;
+ }
+
/* libxl_get_required_shadow_memory() and
* libxl_get_required_iommu_memory() must be called after final values
* (default or specified) for vcpus and memory are set, because the
@@ -1457,7 +1545,9 @@ void parse_config_data(const char *config_source,
: libxl_get_required_shadow_memory(b_info->max_memkb,
b_info->max_vcpus);
- b_info->iommu_memkb = libxl_get_required_iommu_memory(b_info->max_memkb);
+ b_info->iommu_memkb =
+ (c_info->passthrough == LIBXL_PASSTHROUGH_DISABLED) ? 0 :
+ libxl_get_required_iommu_memory(b_info->max_memkb);
xlu_cfg_get_defbool(config, "nomigrate", &b_info->disable_migrate, 0);
@@ -2281,54 +2371,6 @@ skip_vfb:
}
}
- if (!xlu_cfg_get_list (config, "pci", &pcis, 0, 0)) {
- d_config->num_pcidevs = 0;
- d_config->pcidevs = NULL;
- for(i = 0; (buf = xlu_cfg_get_listitem (pcis, i)) != NULL; i++) {
- libxl_device_pci *pcidev;
-
- pcidev = ARRAY_EXTEND_INIT_NODEVID(d_config->pcidevs,
- d_config->num_pcidevs,
- libxl_device_pci_init);
- pcidev->msitranslate = pci_msitranslate;
- pcidev->power_mgmt = pci_power_mgmt;
- pcidev->permissive = pci_permissive;
- pcidev->seize = pci_seize;
- /*
- * Like other pci option, the per-device policy always follows
- * the global policy by default.
- */
- pcidev->rdm_policy = b_info->u.hvm.rdm.policy;
- e = xlu_pci_parse_bdf(config, pcidev, buf);
- if (e) {
- fprintf(stderr,
- "unable to parse PCI BDF `%s' for passthrough\n",
- buf);
- exit(-e);
- }
- }
- if (d_config->num_pcidevs && c_info->type == LIBXL_DOMAIN_TYPE_PV)
- libxl_defbool_set(&b_info->u.pv.e820_host, true);
- }
-
- if (!xlu_cfg_get_list (config, "dtdev", &dtdevs, 0, 0)) {
- d_config->num_dtdevs = 0;
- d_config->dtdevs = NULL;
- for (i = 0; (buf = xlu_cfg_get_listitem(dtdevs, i)) != NULL; i++) {
- libxl_device_dtdev *dtdev;
-
- dtdev = ARRAY_EXTEND_INIT_NODEVID(d_config->dtdevs,
- d_config->num_dtdevs,
- libxl_device_dtdev_init);
-
- dtdev->path = strdup(buf);
- if (dtdev->path == NULL) {
- fprintf(stderr, "unable to duplicate string for dtdevs\n");
- exit(-1);
- }
- }
- }
-
if (!xlu_cfg_get_list(config, "usbctrl", &usbctrls, 0, 0)) {
d_config->num_usbctrls = 0;
d_config->usbctrls = NULL;
@@ -614,6 +614,14 @@ int arch_sanitise_domain_config(struct xen_domctl_createdomain *config)
return -EINVAL;
}
+ /* Always share P2M Table between the CPU and the IOMMU */
+ if ( config->iommu_opts & XEN_DOMCTL_IOMMU_no_sharept )
+ {
+ dprintk(XENLOG_INFO,
+ "Unsupported iommu option: XEN_DOMCTL_IOMMU_no_sharept\n");
+ return -EINVAL;
+ }
+
/* Fill in the native GIC version, passed back to the toolstack. */
if ( config->arch.gic_version == XEN_DOMCTL_CONFIG_GIC_NATIVE )
{
@@ -674,7 +682,7 @@ int arch_domain_create(struct domain *d,
ASSERT(config != NULL);
/* p2m_init relies on some value initialized by the IOMMU subsystem */
- if ( (rc = iommu_domain_init(d)) != 0 )
+ if ( (rc = iommu_domain_init(d, config->iommu_opts)) != 0 )
goto fail;
if ( (rc = p2m_init(d)) != 0 )
@@ -604,7 +604,7 @@ int arch_domain_create(struct domain *d,
if ( (rc = init_domain_irq_mapping(d)) != 0 )
goto fail;
- if ( (rc = iommu_domain_init(d)) != 0 )
+ if ( (rc = iommu_domain_init(d, config->iommu_opts)) != 0 )
goto fail;
psr_domain_init(d);
@@ -308,6 +308,13 @@ static int sanitise_domain_config(struct xen_domctl_createdomain *config)
return -EINVAL;
}
+ if ( !(config->flags & XEN_DOMCTL_CDF_iommu) && config->iommu_opts )
+ {
+ dprintk(XENLOG_INFO,
+ "IOMMU options specified but IOMMU not enabled\n");
+ return -EINVAL;
+ }
+
if ( config->max_vcpus < 1 )
{
dprintk(XENLOG_INFO, "No vCPUS\n");
@@ -515,19 +515,6 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
rover = dom;
}
- /*
- * For now, make sure the createdomain IOMMU flag is set if the
- * IOMMU is enabled. When the flag comes under toolstack control
- * this can go away.
- */
- if ( op->u.createdomain.flags & XEN_DOMCTL_CDF_iommu )
- {
- ASSERT_UNREACHABLE();
- return -EINVAL;
- }
- if ( iommu_enabled )
- op->u.createdomain.flags |= XEN_DOMCTL_CDF_iommu;
-
d = domain_create(dom, &op->u.createdomain, false);
if ( IS_ERR(d) )
{
@@ -163,7 +163,7 @@ static void __hwdom_init check_hwdom_reqs(struct domain *d)
iommu_hwdom_strict = true;
}
-int iommu_domain_init(struct domain *d)
+int iommu_domain_init(struct domain *d, unsigned int opts)
{
struct domain_iommu *hd = dom_iommu(d);
int ret = 0;
@@ -183,6 +183,15 @@ int iommu_domain_init(struct domain *d)
if ( is_hardware_domain(d) )
check_hwdom_reqs(d); /* may modify iommu_hwdom_strict */
+ /*
+ * Use shared page tables for HAP and IOMMU if the global option
+ * is enabled (from which we can infer the h/w is capable) and
+ * the domain options do not disallow it. HAP must, of course, also
+ * be enabled.
+ */
+ hd->hap_pt_share = hap_enabled(d) && iommu_hap_pt_share &&
+ !(opts & XEN_DOMCTL_IOMMU_no_sharept);
+
/*
* NB: 'relaxed' h/w domains don't need the IOMMU mappings to be kept
* in-sync with their assigned pages because all host RAM will be
@@ -191,6 +200,8 @@ int iommu_domain_init(struct domain *d)
if ( !is_hardware_domain(d) || iommu_hwdom_strict )
hd->need_sync = !iommu_use_hap_pt(d);
+ ASSERT(!(hd->need_sync && hd->hap_pt_share));
+
return 0;
}
@@ -38,7 +38,7 @@
#include "hvm/save.h"
#include "memory.h"
-#define XEN_DOMCTL_INTERFACE_VERSION 0x00000011
+#define XEN_DOMCTL_INTERFACE_VERSION 0x00000012
/*
* NB. xen_domctl.domain is an IN/OUT parameter for this operation.
@@ -70,6 +70,10 @@ struct xen_domctl_createdomain {
uint32_t flags;
+#define _XEN_DOMCTL_IOMMU_no_sharept 0
+#define XEN_DOMCTL_IOMMU_no_sharept (1U << _XEN_DOMCTL_IOMMU_no_sharept)
+ uint32_t iommu_opts;
+
/*
* Various domain limits, which impact the quantity of resources (global
* mapping space, xenheap, etc) a guest may consume.
@@ -73,7 +73,7 @@ extern unsigned int iommu_dev_iotlb_timeout;
int iommu_setup(void);
int iommu_hardware_setup(void);
-int iommu_domain_init(struct domain *d);
+int iommu_domain_init(struct domain *d, unsigned int opts);
void iommu_hwdom_init(struct domain *d);
void iommu_domain_destroy(struct domain *d);
@@ -263,9 +263,17 @@ struct domain_iommu {
DECLARE_BITMAP(features, IOMMU_FEAT_count);
/*
- * Does the guest reqire mappings to be synchonized, to maintain
- * the default dfn == pfn map. (See comment on dfn at the top of
- * include/xen/mm.h).
+ * Does the guest share HAP mapping with the IOMMU? This is always
+ * true for ARM systems and may be true for x86 systems where the
+ * the hardware is capable.
+ */
+ bool hap_pt_share;
+
+ /*
+ * Does the guest require mappings to be synchronized, to maintain
+ * the default dfn == pfn map? (See comment on dfn at the top of
+ * include/xen/mm.h). Note that hap_pt_share == false does not
+ * necessarily imply this is true.
*/
bool need_sync;
};
@@ -275,8 +283,7 @@ struct domain_iommu {
#define iommu_clear_feature(d, f) clear_bit(f, dom_iommu(d)->features)
/* Are we using the domain P2M table as its IOMMU pagetable? */
-#define iommu_use_hap_pt(d) \
- (hap_enabled(d) && is_iommu_enabled(d) && iommu_hap_pt_share)
+#define iommu_use_hap_pt(d) (dom_iommu(d)->hap_pt_share)
/* Does the IOMMU pagetable need to be kept synchronized with the P2M */
#ifdef CONFIG_HAS_PASSTHROUGH
...and hence the ability to disable IOMMU mappings, and control EPT sharing. This patch introduces a new 'libxl_passthrough' enumeration into libxl_domain_create_info. The value will be set by xl either when it parses a new 'passthrough' option in xl.cfg, or implicitly if there is passthrough hardware specified for the domain. If the value of the passthrough configuration option is 'disabled' then the XEN_DOMCTL_CDF_iommu flag will be clear in the xen_domctl_createdomain flags, thus allowing the toolstack to control whether the domain gets IOMMU mappings or not (where previously they were globally set). If the value of the passthrough configuration option is 'sync_pt' then a new 'iommu_opts' field in xen_domctl_createdomain will be set with the value XEN_DOMCTL_IOMMU_no_sharept. This will override the global default set in iommu_hap_pt_share, thus allowing the toolstack to control whether EPT sharing is used for the domain. NOTE: The 'iommu_memkb' overhead in libxl_domain_build_info will only be set to zero if passthrough is 'disabled'. It is not safe to set the overhead to zero in the 'share_pt' case because the toolstack has no means of knowing whether the hardware actually supports IOMMU page table sharing. Signed-off-by: Paul Durrant <paul.durrant@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> --- Cc: Ian Jackson <ian.jackson@eu.citrix.com> Cc: Wei Liu <wl@xen.org> Cc: Andrew Cooper <andrew.cooper3@citrix.com> Cc: George Dunlap <George.Dunlap@eu.citrix.com> Cc: Julien Grall <julien.grall@arm.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Stefano Stabellini <sstabellini@kernel.org> Cc: Tim Deegan <tim@xen.org> Cc: Anthony PERARD <anthony.perard@citrix.com> Cc: Christian Lindig <christian.lindig@citrix.com> Cc: David Scott <dave@recoil.org> Cc: Volodymyr Babchuk <Volodymyr_Babchuk@epam.com> Cc: "Roger Pau Monné" <roger.pau@citrix.com> Previously part of series https://lists.xenproject.org/archives/html/xen-devel/2019-07/msg02267.html v7: - Added missing breaks - Added missing ocaml binding changes v6: - Remove the libxl_physinfo() call since it's usefulness is limited to x86 v5: - Expand xen_domctl_createdomain flags field and hence bump interface version - Fix spelling mistakes in context line --- docs/man/xl.cfg.5.pod.in | 52 +++++++++++ tools/libxl/libxl.h | 5 + tools/libxl/libxl_create.c | 9 ++ tools/libxl/libxl_types.idl | 7 ++ tools/ocaml/libs/xc/xenctrl.ml | 4 + tools/ocaml/libs/xc/xenctrl.mli | 4 + tools/ocaml/libs/xc/xenctrl_stubs.c | 15 ++- tools/xl/xl_parse.c | 140 ++++++++++++++++++---------- xen/arch/arm/domain.c | 10 +- xen/arch/x86/domain.c | 2 +- xen/common/domain.c | 7 ++ xen/common/domctl.c | 13 --- xen/drivers/passthrough/iommu.c | 13 ++- xen/include/public/domctl.h | 6 +- xen/include/xen/iommu.h | 19 ++-- 15 files changed, 229 insertions(+), 77 deletions(-)