Skip to contentSkip to conversation

xsm: refactoring xsm hooks

PATCH Superseded Daniel P. Smith 3 participants 38 messages Jun 17, 2021 – Jul 12, 2021

v1 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

cover xsm: refactoring xsm hooks

Based on feedback from 2021 Xen Developers Summit the xsm-roles RFC patch set is being split into two separate patch sets. This is the first patch set and is focused purely on the clean up and refactoring of the XSM hooks.

This patch set refactors the xsm_ops wrapper hooks to use the alternative_call infrastructure. Then proceeds to move and realign the headers to remove the psuedo is/is not enable implementation. The remainder of the changes are clean up and removing no longer necessary abstractions.

Daniel P. Smith (6):
xsm: refactor xsm_ops handling
xsm: decouple xsm header inclusion selection
xsm: enabling xsm to always be included
xsm: remove xen_defualt_t from hook definitions
xsm: expanding function related macros in dummy.h
xsm: removing the XSM_ASSERT_ACTION macro

 xen/arch/arm/dm.c                     |   2 +-
 xen/arch/arm/domctl.c                 |   6 +-
 xen/arch/arm/hvm.c                    |   2 +-
 xen/arch/arm/mm.c                     |   2 +-
 xen/arch/arm/platform_hypercall.c     |   2 +-
 xen/arch/x86/cpu/mcheck/mce.c         |   2 +-
 xen/arch/x86/cpu/vpmu.c               |   2 +-
 xen/arch/x86/domctl.c                 |   8 +-
 xen/arch/x86/hvm/dm.c                 |   2 +-
 xen/arch/x86/hvm/hvm.c                |  12 +-
 xen/arch/x86/irq.c                    |   5 +-
 xen/arch/x86/mm.c                     |  20 +-
 xen/arch/x86/mm/mem_paging.c          |   2 +-
 xen/arch/x86/mm/mem_sharing.c         |   9 +-
 xen/arch/x86/mm/p2m.c                 |   2 +-
 xen/arch/x86/mm/paging.c              |   4 +-
 xen/arch/x86/mm/shadow/set.c          |   2 +-
 xen/arch/x86/msi.c                    |   3 +-
 xen/arch/x86/pci.c                    |   2 +-
 xen/arch/x86/physdev.c                |  17 +-
 xen/arch/x86/platform_hypercall.c     |  10 +-
 xen/arch/x86/pv/emul-priv-op.c        |   2 +-
 xen/arch/x86/sysctl.c                 |   4 +-
 xen/common/Kconfig                    |  55 +-
 xen/common/domain.c                   |   4 +-
 xen/common/domctl.c                   |  12 +-
 xen/common/event_channel.c            |  12 +-
 xen/common/grant_table.c              |  16 +-
 xen/common/hypfs.c                    |   2 +-
 xen/common/kernel.c                   |   2 +-
 xen/common/kexec.c                    |   2 +-
 xen/common/mem_access.c               |   2 +-
 xen/common/memory.c                   |  16 +-
 xen/common/monitor.c                  |   2 +-
 xen/common/sched/core.c               |   6 +-
 xen/common/sysctl.c                   |   8 +-
 xen/common/vm_event.c                 |   2 +-
 xen/common/xenoprof.c                 |   2 +-
 xen/drivers/char/console.c            |   2 +-
 xen/drivers/passthrough/device_tree.c |   4 +-
 xen/drivers/passthrough/pci.c         |  12 +-
 xen/include/xen/sched.h               |   2 +-
 xen/include/xsm/dummy.h               | 774 --------------------------
 xen/include/xsm/xsm-core.h            | 236 ++++++++
 xen/include/xsm/xsm.h                 | 626 +++++++--------------
 xen/xsm/Makefile                      |   4 +-
 xen/xsm/dummy.c                       |   7 +-
 xen/xsm/dummy.h                       | 697 +++++++++++++++++++++++
 xen/xsm/flask/flask_op.c              |  21 +-
 xen/xsm/silo.c                        |  18 +-
 xen/xsm/xsm_core.c                    |  54 +-
 51 files changed, 1309 insertions(+), 1413 deletions(-)
 delete mode 100644 xen/include/xsm/dummy.h
 create mode 100644 xen/include/xsm/xsm-core.h
 create mode 100644 xen/xsm/dummy.h
Signature · 1 line
2.20.1
#1 lore raw

PATCH 1/6 xsm: refactor xsm_ops handling

The assignment and setup of xsm_ops structure was refactored to make it a one-time assignment. The calling of the xsm_ops were refactored to use the alternate_call framework to reduce the need for retpolines.

---
 xen/include/xsm/xsm.h    | 206 ++++++++++++++++++++-------------------
 xen/xsm/dummy.c          |   2 -
 xen/xsm/flask/flask_op.c |  21 +---
 xen/xsm/xsm_core.c       |  50 ++++++----
 4 files changed, 138 insertions(+), 141 deletions(-)
xen/include/xsm/xsm.h +106 -100
index ad3cddbf7d..86ca045e74 100644
@@ -15,6 +15,9 @@
 #ifndef __XSM_H__
 #define __XSM_H__
 
+#ifdef CONFIG_XSM
+#include <asm/alternative.h>
+#endif
 #include <xen/sched.h>
 #include <xen/multiboot.h>
 
@@ -191,295 +194,295 @@ struct xsm_operations {
 
 #ifdef CONFIG_XSM
 
-extern struct xsm_operations *xsm_ops;
+extern struct xsm_operations xsm_ops;
 
 #ifndef XSM_NO_WRAPPERS
 
 static inline void xsm_security_domaininfo (struct domain *d,
                                         struct xen_domctl_getdomaininfo *info)
 {
-    xsm_ops->security_domaininfo(d, info);
+    alternative_vcall(xsm_ops.security_domaininfo, d, info);
 }
 
 static inline int xsm_domain_create (xsm_default_t def, struct domain *d, u32 ssidref)
 {
-    return xsm_ops->domain_create(d, ssidref);
+    return alternative_call(xsm_ops.domain_create, d, ssidref);
 }
 
 static inline int xsm_getdomaininfo (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->getdomaininfo(d);
+    return alternative_call(xsm_ops.getdomaininfo, d);
 }
 
 static inline int xsm_domctl_scheduler_op (xsm_default_t def, struct domain *d, int cmd)
 {
-    return xsm_ops->domctl_scheduler_op(d, cmd);
+    return alternative_call(xsm_ops.domctl_scheduler_op, d, cmd);
 }
 
 static inline int xsm_sysctl_scheduler_op (xsm_default_t def, int cmd)
 {
-    return xsm_ops->sysctl_scheduler_op(cmd);
+    return alternative_call(xsm_ops.sysctl_scheduler_op, cmd);
 }
 
 static inline int xsm_set_target (xsm_default_t def, struct domain *d, struct domain *e)
 {
-    return xsm_ops->set_target(d, e);
+    return alternative_call(xsm_ops.set_target, d, e);
 }
 
 static inline int xsm_domctl (xsm_default_t def, struct domain *d, int cmd)
 {
-    return xsm_ops->domctl(d, cmd);
+    return alternative_call(xsm_ops.domctl, d, cmd);
 }
 
 static inline int xsm_sysctl (xsm_default_t def, int cmd)
 {
-    return xsm_ops->sysctl(cmd);
+    return alternative_call(xsm_ops.sysctl, cmd);
 }
 
 static inline int xsm_readconsole (xsm_default_t def, uint32_t clear)
 {
-    return xsm_ops->readconsole(clear);
+    return alternative_call(xsm_ops.readconsole, clear);
 }
 
 static inline int xsm_evtchn_unbound (xsm_default_t def, struct domain *d1, struct evtchn *chn,
                                                                     domid_t id2)
 {
-    return xsm_ops->evtchn_unbound(d1, chn, id2);
+    return alternative_call(xsm_ops.evtchn_unbound, d1, chn, id2);
 }
 
 static inline int xsm_evtchn_interdomain (xsm_default_t def, struct domain *d1,
                 struct evtchn *chan1, struct domain *d2, struct evtchn *chan2)
 {
-    return xsm_ops->evtchn_interdomain(d1, chan1, d2, chan2);
+    return alternative_call(xsm_ops.evtchn_interdomain, d1, chan1, d2, chan2);
 }
 
 static inline void xsm_evtchn_close_post (struct evtchn *chn)
 {
-    xsm_ops->evtchn_close_post(chn);
+    alternative_vcall(xsm_ops.evtchn_close_post, chn);
 }
 
 static inline int xsm_evtchn_send (xsm_default_t def, struct domain *d, struct evtchn *chn)
 {
-    return xsm_ops->evtchn_send(d, chn);
+    return alternative_call(xsm_ops.evtchn_send, d, chn);
 }
 
 static inline int xsm_evtchn_status (xsm_default_t def, struct domain *d, struct evtchn *chn)
 {
-    return xsm_ops->evtchn_status(d, chn);
+    return alternative_call(xsm_ops.evtchn_status, d, chn);
 }
 
 static inline int xsm_evtchn_reset (xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->evtchn_reset(d1, d2);
+    return alternative_call(xsm_ops.evtchn_reset, d1, d2);
 }
 
 static inline int xsm_grant_mapref (xsm_default_t def, struct domain *d1, struct domain *d2,
                                                                 uint32_t flags)
 {
-    return xsm_ops->grant_mapref(d1, d2, flags);
+    return alternative_call(xsm_ops.grant_mapref, d1, d2, flags);
 }
 
 static inline int xsm_grant_unmapref (xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->grant_unmapref(d1, d2);
+    return alternative_call(xsm_ops.grant_unmapref, d1, d2);
 }
 
 static inline int xsm_grant_setup (xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->grant_setup(d1, d2);
+    return alternative_call(xsm_ops.grant_setup, d1, d2);
 }
 
 static inline int xsm_grant_transfer (xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->grant_transfer(d1, d2);
+    return alternative_call(xsm_ops.grant_transfer, d1, d2);
 }
 
 static inline int xsm_grant_copy (xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->grant_copy(d1, d2);
+    return alternative_call(xsm_ops.grant_copy, d1, d2);
 }
 
 static inline int xsm_grant_query_size (xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->grant_query_size(d1, d2);
+    return alternative_call(xsm_ops.grant_query_size, d1, d2);
 }
 
 static inline int xsm_alloc_security_domain (struct domain *d)
 {
-    return xsm_ops->alloc_security_domain(d);
+    return alternative_call(xsm_ops.alloc_security_domain, d);
 }
 
 static inline void xsm_free_security_domain (struct domain *d)
 {
-    xsm_ops->free_security_domain(d);
+    alternative_vcall(xsm_ops.free_security_domain, d);
 }
 
 static inline int xsm_alloc_security_evtchns(
     struct evtchn chn[], unsigned int nr)
 {
-    return xsm_ops->alloc_security_evtchns(chn, nr);
+    return alternative_call(xsm_ops.alloc_security_evtchns, chn, nr);
 }
 
 static inline void xsm_free_security_evtchns(
     struct evtchn chn[], unsigned int nr)
 {
-    xsm_ops->free_security_evtchns(chn, nr);
+    alternative_vcall(xsm_ops.free_security_evtchns, chn, nr);
 }
 
 static inline char *xsm_show_security_evtchn (struct domain *d, const struct evtchn *chn)
 {
-    return xsm_ops->show_security_evtchn(d, chn);
+    return alternative_call(xsm_ops.show_security_evtchn, d, chn);
 }
 
 static inline int xsm_init_hardware_domain (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->init_hardware_domain(d);
+    return alternative_call(xsm_ops.init_hardware_domain, d);
 }
 
 static inline int xsm_get_pod_target (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->get_pod_target(d);
+    return alternative_call(xsm_ops.get_pod_target, d);
 }
 
 static inline int xsm_set_pod_target (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->set_pod_target(d);
+    return alternative_call(xsm_ops.set_pod_target, d);
 }
 
 static inline int xsm_memory_exchange (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->memory_exchange(d);
+    return alternative_call(xsm_ops.memory_exchange, d);
 }
 
 static inline int xsm_memory_adjust_reservation (xsm_default_t def, struct domain *d1, struct
                                                                     domain *d2)
 {
-    return xsm_ops->memory_adjust_reservation(d1, d2);
+    return alternative_call(xsm_ops.memory_adjust_reservation, d1, d2);
 }
 
 static inline int xsm_memory_stat_reservation (xsm_default_t def, struct domain *d1,
                                                             struct domain *d2)
 {
-    return xsm_ops->memory_stat_reservation(d1, d2);
+    return alternative_call(xsm_ops.memory_stat_reservation, d1, d2);
 }
 
 static inline int xsm_memory_pin_page(xsm_default_t def, struct domain *d1, struct domain *d2,
                                       struct page_info *page)
 {
-    return xsm_ops->memory_pin_page(d1, d2, page);
+    return alternative_call(xsm_ops.memory_pin_page, d1, d2, page);
 }
 
 static inline int xsm_add_to_physmap(xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->add_to_physmap(d1, d2);
+    return alternative_call(xsm_ops.add_to_physmap, d1, d2);
 }
 
 static inline int xsm_remove_from_physmap(xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->remove_from_physmap(d1, d2);
+    return alternative_call(xsm_ops.remove_from_physmap, d1, d2);
 }
 
 static inline int xsm_map_gmfn_foreign (xsm_default_t def, struct domain *d, struct domain *t)
 {
-    return xsm_ops->map_gmfn_foreign(d, t);
+    return alternative_call(xsm_ops.map_gmfn_foreign, d, t);
 }
 
 static inline int xsm_claim_pages(xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->claim_pages(d);
+    return alternative_call(xsm_ops.claim_pages, d);
 }
 
 static inline int xsm_console_io (xsm_default_t def, struct domain *d, int cmd)
 {
-    return xsm_ops->console_io(d, cmd);
+    return alternative_call(xsm_ops.console_io, d, cmd);
 }
 
 static inline int xsm_profile (xsm_default_t def, struct domain *d, int op)
 {
-    return xsm_ops->profile(d, op);
+    return alternative_call(xsm_ops.profile, d, op);
 }
 
 static inline int xsm_kexec (xsm_default_t def)
 {
-    return xsm_ops->kexec();
+    return alternative_call(xsm_ops.kexec);
 }
 
 static inline int xsm_schedop_shutdown (xsm_default_t def, struct domain *d1, struct domain *d2)
 {
-    return xsm_ops->schedop_shutdown(d1, d2);
+    return alternative_call(xsm_ops.schedop_shutdown, d1, d2);
 }
 
 static inline char *xsm_show_irq_sid (int irq)
 {
-    return xsm_ops->show_irq_sid(irq);
+    return alternative_call(xsm_ops.show_irq_sid, irq);
 }
 
 static inline int xsm_map_domain_pirq (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->map_domain_pirq(d);
+    return alternative_call(xsm_ops.map_domain_pirq, d);
 }
 
 static inline int xsm_map_domain_irq (xsm_default_t def, struct domain *d, int irq, void *data)
 {
-    return xsm_ops->map_domain_irq(d, irq, data);
+    return alternative_call(xsm_ops.map_domain_irq, d, irq, data);
 }
 
 static inline int xsm_unmap_domain_pirq (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->unmap_domain_pirq(d);
+    return alternative_call(xsm_ops.unmap_domain_pirq, d);
 }
 
 static inline int xsm_unmap_domain_irq (xsm_default_t def, struct domain *d, int irq, void *data)
 {
-    return xsm_ops->unmap_domain_irq(d, irq, data);
+    return alternative_call(xsm_ops.unmap_domain_irq, d, irq, data);
 }
 
 static inline int xsm_bind_pt_irq(xsm_default_t def, struct domain *d,
                                   struct xen_domctl_bind_pt_irq *bind)
 {
-    return xsm_ops->bind_pt_irq(d, bind);
+    return alternative_call(xsm_ops.bind_pt_irq, d, bind);
 }
 
 static inline int xsm_unbind_pt_irq(xsm_default_t def, struct domain *d,
                                     struct xen_domctl_bind_pt_irq *bind)
 {
-    return xsm_ops->unbind_pt_irq(d, bind);
+    return alternative_call(xsm_ops.unbind_pt_irq, d, bind);
 }
 
 static inline int xsm_irq_permission (xsm_default_t def, struct domain *d, int pirq, uint8_t allow)
 {
-    return xsm_ops->irq_permission(d, pirq, allow);
+    return alternative_call(xsm_ops.irq_permission, d, pirq, allow);
 }
 
 static inline int xsm_iomem_permission (xsm_default_t def, struct domain *d, uint64_t s, uint64_t e, uint8_t allow)
 {
-    return xsm_ops->iomem_permission(d, s, e, allow);
+    return alternative_call(xsm_ops.iomem_permission, d, s, e, allow);
 }
 
 static inline int xsm_iomem_mapping (xsm_default_t def, struct domain *d, uint64_t s, uint64_t e, uint8_t allow)
 {
-    return xsm_ops->iomem_mapping(d, s, e, allow);
+    return alternative_call(xsm_ops.iomem_mapping, d, s, e, allow);
 }
 
 static inline int xsm_pci_config_permission (xsm_default_t def, struct domain *d, uint32_t machine_bdf, uint16_t start, uint16_t end, uint8_t access)
 {
-    return xsm_ops->pci_config_permission(d, machine_bdf, start, end, access);
+    return alternative_call(xsm_ops.pci_config_permission, d, machine_bdf, start, end, access);
 }
 
 #if defined(CONFIG_HAS_PASSTHROUGH) && defined(CONFIG_HAS_PCI)
 static inline int xsm_get_device_group(xsm_default_t def, uint32_t machine_bdf)
 {
-    return xsm_ops->get_device_group(machine_bdf);
+    return alternative_call(xsm_ops.get_device_group, machine_bdf);
 }
 
 static inline int xsm_assign_device(xsm_default_t def, struct domain *d, uint32_t machine_bdf)
 {
-    return xsm_ops->assign_device(d, machine_bdf);
+    return alternative_call(xsm_ops.assign_device, d, machine_bdf);
 }
 
 static inline int xsm_deassign_device(xsm_default_t def, struct domain *d, uint32_t machine_bdf)
 {
-    return xsm_ops->deassign_device(d, machine_bdf);
+    return alternative_call(xsm_ops.deassign_device, d, machine_bdf);
 }
 #endif /* HAS_PASSTHROUGH && HAS_PCI) */
 
@@ -487,240 +490,243 @@ static inline int xsm_deassign_device(xsm_default_t def, struct domain *d, uint3
 static inline int xsm_assign_dtdevice(xsm_default_t def, struct domain *d,
                                       const char *dtpath)
 {
-    return xsm_ops->assign_dtdevice(d, dtpath);
+    return alternative_call(xsm_ops.assign_dtdevice, d, dtpath);
 }
 
 static inline int xsm_deassign_dtdevice(xsm_default_t def, struct domain *d,
                                         const char *dtpath)
 {
-    return xsm_ops->deassign_dtdevice(d, dtpath);
+    return alternative_call(xsm_ops.deassign_dtdevice, d, dtpath);
 }
 
 #endif /* HAS_PASSTHROUGH && HAS_DEVICE_TREE */
 
 static inline int xsm_resource_plug_pci (xsm_default_t def, uint32_t machine_bdf)
 {
-    return xsm_ops->resource_plug_pci(machine_bdf);
+    return alternative_call(xsm_ops.resource_plug_pci, machine_bdf);
 }
 
 static inline int xsm_resource_unplug_pci (xsm_default_t def, uint32_t machine_bdf)
 {
-    return xsm_ops->resource_unplug_pci(machine_bdf);
+    return alternative_call(xsm_ops.resource_unplug_pci, machine_bdf);
 }
 
 static inline int xsm_resource_plug_core (xsm_default_t def)
 {
-    return xsm_ops->resource_plug_core();
+    return alternative_call(xsm_ops.resource_plug_core);
 }
 
 static inline int xsm_resource_unplug_core (xsm_default_t def)
 {
-    return xsm_ops->resource_unplug_core();
+    return alternative_call(xsm_ops.resource_unplug_core);
 }
 
 static inline int xsm_resource_setup_pci (xsm_default_t def, uint32_t machine_bdf)
 {
-    return xsm_ops->resource_setup_pci(machine_bdf);
+    return alternative_call(xsm_ops.resource_setup_pci, machine_bdf);
 }
 
 static inline int xsm_resource_setup_gsi (xsm_default_t def, int gsi)
 {
-    return xsm_ops->resource_setup_gsi(gsi);
+    return alternative_call(xsm_ops.resource_setup_gsi, gsi);
 }
 
 static inline int xsm_resource_setup_misc (xsm_default_t def)
 {
-    return xsm_ops->resource_setup_misc();
+    return alternative_call(xsm_ops.resource_setup_misc);
 }
 
 static inline int xsm_page_offline(xsm_default_t def, uint32_t cmd)
 {
-    return xsm_ops->page_offline(cmd);
+    return alternative_call(xsm_ops.page_offline, cmd);
 }
 
 static inline int xsm_hypfs_op(xsm_default_t def)
 {
-    return xsm_ops->hypfs_op();
+    return alternative_call(xsm_ops.hypfs_op);
 }
 
 static inline long xsm_do_xsm_op (XEN_GUEST_HANDLE_PARAM(xsm_op_t) op)
 {
-    return xsm_ops->do_xsm_op(op);
+    /* "op"(struct) is being passed by value, alternative_call does not support */
+    return xsm_ops.do_xsm_op(op);
 }
 
 #ifdef CONFIG_COMPAT
 static inline int xsm_do_compat_op (XEN_GUEST_HANDLE_PARAM(xsm_op_t) op)
 {
-    return xsm_ops->do_compat_op(op);
+    /* "op"(struct) is being passed by value, alternative_call does not support */
+    return xsm_ops.do_compat_op(op);
 }
 #endif
 
 static inline int xsm_hvm_param (xsm_default_t def, struct domain *d, unsigned long op)
 {
-    return xsm_ops->hvm_param(d, op);
+    return alternative_call(xsm_ops.hvm_param, d, op);
 }
 
 static inline int xsm_hvm_control(xsm_default_t def, struct domain *d, unsigned long op)
 {
-    return xsm_ops->hvm_control(d, op);
+    return alternative_call(xsm_ops.hvm_control, d, op);
 }
 
 static inline int xsm_hvm_param_altp2mhvm (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->hvm_param_altp2mhvm(d);
+    return alternative_call(xsm_ops.hvm_param_altp2mhvm, d);
 }
 
 static inline int xsm_hvm_altp2mhvm_op (xsm_default_t def, struct domain *d, uint64_t mode, uint32_t op)
 {
-    return xsm_ops->hvm_altp2mhvm_op(d, mode, op);
+    return alternative_call(xsm_ops.hvm_altp2mhvm_op, d, mode, op);
 }
 
 static inline int xsm_get_vnumainfo (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->get_vnumainfo(d);
+    return alternative_call(xsm_ops.get_vnumainfo, d);
 }
 
 static inline int xsm_vm_event_control (xsm_default_t def, struct domain *d, int mode, int op)
 {
-    return xsm_ops->vm_event_control(d, mode, op);
+    return alternative_call(xsm_ops.vm_event_control, d, mode, op);
 }
 
 #ifdef CONFIG_MEM_ACCESS
 static inline int xsm_mem_access (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->mem_access(d);
+    return alternative_call(xsm_ops.mem_access, d);
 }
 #endif
 
 #ifdef CONFIG_MEM_PAGING
 static inline int xsm_mem_paging (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->mem_paging(d);
+    return alternative_call(xsm_ops.mem_paging, d);
 }
 #endif
 
 #ifdef CONFIG_MEM_SHARING
 static inline int xsm_mem_sharing (xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->mem_sharing(d);
+    return alternative_call(xsm_ops.mem_sharing, d);
 }
 #endif
 
 static inline int xsm_platform_op (xsm_default_t def, uint32_t op)
 {
-    return xsm_ops->platform_op(op);
+    return alternative_call(xsm_ops.platform_op, op);
 }
 
 #ifdef CONFIG_X86
 static inline int xsm_do_mca(xsm_default_t def)
 {
-    return xsm_ops->do_mca();
+    return alternative_call(xsm_ops.do_mca);
 }
 
 static inline int xsm_shadow_control (xsm_default_t def, struct domain *d, uint32_t op)
 {
-    return xsm_ops->shadow_control(d, op);
+    return alternative_call(xsm_ops.shadow_control, d, op);
 }
 
 static inline int xsm_mem_sharing_op (xsm_default_t def, struct domain *d, struct domain *cd, int op)
 {
-    return xsm_ops->mem_sharing_op(d, cd, op);
+    return alternative_call(xsm_ops.mem_sharing_op, d, cd, op);
 }
 
 static inline int xsm_apic (xsm_default_t def, struct domain *d, int cmd)
 {
-    return xsm_ops->apic(d, cmd);
+    return alternative_call(xsm_ops.apic, d, cmd);
 }
 
 static inline int xsm_memtype (xsm_default_t def, uint32_t access)
 {
-    return xsm_ops->memtype(access);
+    return alternative_call(xsm_ops.memtype, access);
 }
 
 static inline int xsm_machine_memory_map(xsm_default_t def)
 {
-    return xsm_ops->machine_memory_map();
+    return alternative_call(xsm_ops.machine_memory_map);
 }
 
 static inline int xsm_domain_memory_map(xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->domain_memory_map(d);
+    return alternative_call(xsm_ops.domain_memory_map, d);
 }
 
 static inline int xsm_mmu_update (xsm_default_t def, struct domain *d, struct domain *t,
                                   struct domain *f, uint32_t flags)
 {
-    return xsm_ops->mmu_update(d, t, f, flags);
+    return alternative_call(xsm_ops.mmu_update, d, t, f, flags);
 }
 
 static inline int xsm_mmuext_op (xsm_default_t def, struct domain *d, struct domain *f)
 {
-    return xsm_ops->mmuext_op(d, f);
+    return alternative_call(xsm_ops.mmuext_op, d, f);
 }
 
 static inline int xsm_update_va_mapping(xsm_default_t def, struct domain *d, struct domain *f,
                                                             l1_pgentry_t pte)
 {
-    return xsm_ops->update_va_mapping(d, f, pte);
+    /* pte(struct) is being passed by value, alternative_call does not support */
+    return xsm_ops.update_va_mapping(d, f, pte);
 }
 
 static inline int xsm_priv_mapping(xsm_default_t def, struct domain *d, struct domain *t)
 {
-    return xsm_ops->priv_mapping(d, t);
+    return alternative_call(xsm_ops.priv_mapping, d, t);
 }
 
 static inline int xsm_ioport_permission (xsm_default_t def, struct domain *d, uint32_t s, uint32_t e, uint8_t allow)
 {
-    return xsm_ops->ioport_permission(d, s, e, allow);
+    return alternative_call(xsm_ops.ioport_permission, d, s, e, allow);
 }
 
 static inline int xsm_ioport_mapping (xsm_default_t def, struct domain *d, uint32_t s, uint32_t e, uint8_t allow)
 {
-    return xsm_ops->ioport_mapping(d, s, e, allow);
+    return alternative_call(xsm_ops.ioport_mapping, d, s, e, allow);
 }
 
 static inline int xsm_pmu_op (xsm_default_t def, struct domain *d, unsigned int op)
 {
-    return xsm_ops->pmu_op(d, op);
+    return alternative_call(xsm_ops.pmu_op, d, op);
 }
 
 #endif /* CONFIG_X86 */
 
 static inline int xsm_dm_op(xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->dm_op(d);
+    return alternative_call(xsm_ops.dm_op, d);
 }
 
 static inline int xsm_xen_version (xsm_default_t def, uint32_t op)
 {
-    return xsm_ops->xen_version(op);
+    return alternative_call(xsm_ops.xen_version, op);
 }
 
 static inline int xsm_domain_resource_map(xsm_default_t def, struct domain *d)
 {
-    return xsm_ops->domain_resource_map(d);
+    return alternative_call(xsm_ops.domain_resource_map, d);
 }
 
 #ifdef CONFIG_ARGO
 static inline int xsm_argo_enable(const struct domain *d)
 {
-    return xsm_ops->argo_enable(d);
+    return alternative_call(xsm_ops.argo_enable, d);
 }
 
 static inline int xsm_argo_register_single_source(const struct domain *d,
                                                   const struct domain *t)
 {
-    return xsm_ops->argo_register_single_source(d, t);
+    return alternative_call(xsm_ops.argo_register_single_source, d, t);
 }
 
 static inline int xsm_argo_register_any_source(const struct domain *d)
 {
-    return xsm_ops->argo_register_any_source(d);
+    return alternative_call(xsm_ops.argo_register_any_source, d);
 }
 
 static inline int xsm_argo_send(const struct domain *d, const struct domain *t)
 {
-    return xsm_ops->argo_send(d, t);
+    return alternative_call(xsm_ops.argo_send, d, t);
 }
 
 #endif /* CONFIG_ARGO */
xen/xsm/dummy.c +0 -2
index de44b10130..066694763a 100644
@@ -13,8 +13,6 @@
 #define XSM_NO_WRAPPERS
 #include <xsm/dummy.h>
 
-struct xsm_operations dummy_xsm_ops;
-
 #define set_to_dummy_if_null(ops, function)                            \
     do {                                                               \
         if ( !ops->function )                                          \
xen/xsm/flask/flask_op.c +1 -20
index 01e52138a1..df9fcc1d6d 100644
@@ -225,26 +225,7 @@ static int flask_security_sid(struct xen_flask_sid_context *arg)
 
 static int flask_disable(void)
 {
-    static int flask_disabled = 0;
-
-    if ( ss_initialized )
-    {
-        /* Not permitted after initial policy load. */
-        return -EINVAL;
-    }
-
-    if ( flask_disabled )
-    {
-        /* Only do this once. */
-        return -EINVAL;
-    }
-
-    printk("Flask:  Disabled at runtime.\n");
-
-    flask_disabled = 1;
-
-    /* Reset xsm_ops to the original module. */
-    xsm_ops = &dummy_xsm_ops;
+    printk("Flask:  Disabling is not supported.\n");
 
     return 0;
 }
xen/xsm/xsm_core.c +31 -19

94 lines not on this page · raw message

Signature · 1 line
2.20.1
#2 lore raw

Re: [PATCH 1/6] xsm: refactor xsm_ops handling

Quoted 5 lines · Daniel P. Smith

The assignment and setup of xsm_ops structure was refactored to make it a one-time assignment. The calling of the xsm_ops were refactored to use the alternate_call framework to reduce the need for retpolines.

I think the commit message needs a little more explanation for anyone doing code archaeology.

AFAICT, the current infrastructure has some (incomplete?) support for Flask to unload itself as the security engine, which doesn't sounds like a clever thing in general.

What we do here is make a semantic change to say that the security engine (Dummy, Flask or SILO) gets chosen once during boot, and is immutable thereafter.  This is better from a security standpoint (no accidentally unloading Flask at runtime), and allows for the use of the alternative_vcall() infrastructure to drop all the function pointer calls.

Does that about sum things up?

Quoted 19 lines
xen/xsm/flask/flask_op.c +0 -11
index 01e52138a1..df9fcc1d6d 100644
@@ -225,26 +225,7 @@ static int flask_security_sid(struct xen_flask_sid_context *arg)
 
 static int flask_disable(void)
 {
-    static int flask_disabled = 0;
-
-    if ( ss_initialized )
-    {
-        /* Not permitted after initial policy load. */
-        return -EINVAL;
-    }
-
-    if ( flask_disabled )
-    {
-        /* Only do this once. */

xen/xsm/flask/flask_op.c static int flask_security_sid(struct xen_flask_sid_context *arg) · line 228

Quoted 10 lines
xen/xsm/flask/flask_op.c +1 -9
-        return -EINVAL;
-    }
-
-    printk("Flask:  Disabled at runtime.\n");
-
-    flask_disabled = 1;
-
-    /* Reset xsm_ops to the original module. */
-    xsm_ops = &dummy_xsm_ops;
+    printk("Flask:  Disabling is not supported.\n");

Judging by this, should this patch be split up more?

I think you want to remove FLASK_DISABLE (and this function too - just return -EOPNOTSUPP in the parent) as a separate explained change (as it is a logical change in how Flask works).

The second patch wants to be the rest, which changes the indirection of xsm_ops and converts to vcall().  This is a fairly mechanical change without semantic changes.

I'm unsure if you want a 3rd patch in the middle, separating the xsm_core_init() juggling, with the "switch to using vcall()".  It might be a good idea for more easily demonstrating the changes, but I'll leave it to your judgement.

Quoted 8 lines
xen/xsm/xsm_core.c +0 -0
index 5eab21e1b1..acc1af7166 100644
 static int __init xsm_core_init(const void *policy_buffer, size_t policy_size)
 {
 #ifdef CONFIG_XSM_FLASK_POLICY

xen/xsm/xsm_core.c static int __init xsm_core_init(const void *policy_buffer, size_t policy_size) · line 92

Quoted 9 lines
xen/xsm/xsm_core.c +3 -2
@@ -87,17 +86,22 @@ static int __init xsm_core_init(const void *policy_buffer, size_t policy_size)
     }
 #endif
 
-    if ( verify(&dummy_xsm_ops) )
+    if ( xsm_ops_registered != XSM_OPS_UNREGISTERED )
     {
-        printk(XENLOG_ERR "Could not verify dummy_xsm_ops structure\n");
+        printk(XENLOG_ERR
+            "Could not init XSM, xsm_ops register already attempted\n");

Indentation.

xen/xsm/xsm_core.c static int __init xsm_core_init(const void *policy_buffer, size_t policy_size) · line 99

Quoted 8 lines
xen/xsm/xsm_core.c +4 -1
         return -EIO;
     }
 
-    xsm_ops = &dummy_xsm_ops;
+    /* install the dummy ops as default to ensure ops
+     * are defined if requested policy fails init
+     */
+    xsm_fixup_ops(&xsm_ops);

/* Comment style. */

or

/*
 * Multi-
 * line comment style.
 */

Quoted 18 lines
xen/xsm/xsm_core.c +4 -0
     switch ( xsm_bootparam )
     {
     case XSM_BOOTPARAM_DUMMY:
+        xsm_ops_registered = XSM_OPS_REGISTERED;
         break;
 
     case XSM_BOOTPARAM_FLASK:
@@ -113,6 +117,9 @@ static int __init xsm_core_init(const void *policy_buffer, size_t policy_size)
         break;
     }
 
+    if ( xsm_ops_registered != XSM_OPS_REGISTERED )
+        xsm_ops_registered = XSM_OPS_REG_FAILED;
+
     return 0;
 }
 

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 208

Quoted 6 lines
xen/xsm/xsm_core.c +2 -1
@@ -197,16 +204,21 @@ bool __init has_xsm_magic(paddr_t start)
 
 int __init register_xsm(struct xsm_operations *ops)
 {
-    if ( verify(ops) )
+    if ( xsm_ops_registered != XSM_OPS_UNREGISTERED )
+        return -EAGAIN;

I know you moved this around the function, but it really isn't -EAGAIN material any more.  It's "too late - nope".

-EEXIST is probably best for "I'm never going to tolerate another call".

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 214

Quoted 7 lines
xen/xsm/xsm_core.c +4 -1
+
+    if ( !ops )
     {
-        printk(XENLOG_ERR "Could not verify xsm_operations structure\n");
+        xsm_ops_registered = XSM_OPS_REG_FAILED;
+        printk(XENLOG_ERR "Invalid xsm_operations structure registered\n");
         return -EINVAL;

Honestly, I'd be half tempted to declare register_xsm() with __nonnull(0) and let the compiler reject any attempt to pass a NULL ops pointer.

Both callers pass a pointer to a static singleton objects.

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 218

Quoted 6 lines
xen/xsm/xsm_core.c +2 -2
     }
 
-    if ( xsm_ops != &dummy_xsm_ops )
-        return -EAGAIN;
+    /* use dummy ops for any empty ops */
+    xsm_fixup_ops(ops);

Isn't this redundant with the call in xsm_core_init(), seeing as register_xsm() must be nested within the switch statement?

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 224

Quoted 6 lines
xen/xsm/xsm_core.c +2 -1
-    xsm_ops = ops;
+    xsm_ops = *ops;
+    xsm_ops_registered = XSM_OPS_REGISTERED;
 
     return 0;
 }

Having got to the end, the xsm_core_init() vs register_xsm() dynamic is quite weird.

I think it would result in clearer code to have init_{flask,silo}() return pointers to their struct xsm_operations *, and let xsm_core_init() do the copy in to xsm_ops.  This reduces the scope of xsm_ops_state to this function only, and gets rid of at least one runtime panic() call which is dead code.

If you were to go with this approach, you'd definitely want to split into the 3-patch approach.

~Andrew

#3 lore raw

Re: [PATCH 1/6] xsm: refactor xsm_ops handling

Quoted 22 lines · Andrew Cooper

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 208

Quoted 7 lines · Daniel P. Smith
xen/xsm/xsm_core.c +2 -1
@@ -197,16 +204,21 @@ bool __init has_xsm_magic(paddr_t start)
 
 int __init register_xsm(struct xsm_operations *ops)
 {
-    if ( verify(ops) )
+    if ( xsm_ops_registered != XSM_OPS_UNREGISTERED )
+        return -EAGAIN;

I know you moved this around the function, but it really isn't -EAGAIN material any more.  It's "too late - nope".

-EEXIST is probably best for "I'm never going to tolerate another call".

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 214

Quoted 7 lines
xen/xsm/xsm_core.c +4 -1
+
+    if ( !ops )
     {
-        printk(XENLOG_ERR "Could not verify xsm_operations structure\n");
+        xsm_ops_registered = XSM_OPS_REG_FAILED;
+        printk(XENLOG_ERR "Invalid xsm_operations structure registered\n");
         return -EINVAL;
Quoted 5 lines

Honestly, I'd be half tempted to declare register_xsm() with __nonnull(0) and let the compiler reject any attempt to pass a NULL ops pointer.

Both callers pass a pointer to a static singleton objects.

Why check at all when the source of the arguments is all internal? We don't check pointers to be non-NULL elsewhere, with a few odd exceptions (which imo should all be dropped).

Jan

#4 lore raw

Re: [PATCH 1/6] xsm: refactor xsm_ops handling

Quoted 21 lines · Jan Beulich
Quoted 20 lines · Andrew Cooper

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 208

Quoted 7 lines · Daniel P. Smith
xen/xsm/xsm_core.c +2 -1
@@ -197,16 +204,21 @@ bool __init has_xsm_magic(paddr_t start)
 
 int __init register_xsm(struct xsm_operations *ops)
 {
-    if ( verify(ops) )
+    if ( xsm_ops_registered != XSM_OPS_UNREGISTERED )
+        return -EAGAIN;

I know you moved this around the function, but it really isn't -EAGAIN material any more.  It's "too late - nope".

-EEXIST is probably best for "I'm never going to tolerate another call".

Quoted 7 lines
xen/xsm/xsm_core.c +4 -1
+
+    if ( !ops )
     {
-        printk(XENLOG_ERR "Could not verify xsm_operations structure\n");
+        xsm_ops_registered = XSM_OPS_REG_FAILED;
+        printk(XENLOG_ERR "Invalid xsm_operations structure registered\n");
         return -EINVAL;
Quoted 8 lines
Quoted 5 lines

Honestly, I'd be half tempted to declare register_xsm() with __nonnull(0) and let the compiler reject any attempt to pass a NULL ops pointer.

Both callers pass a pointer to a static singleton objects.

Why check at all when the source of the arguments is all internal? We don't check pointers to be non-NULL elsewhere, with a few odd exceptions (which imo should all be dropped).

That too.  At the end of my email, I suggested an alternative approach which would remove register_xsm() entirely, and I think that is a better-still way forward.

~Andrew

#5 lore raw

Re: [PATCH 1/6] xsm: refactor xsm_ops handling

In reply to Jan Beulich
Quoted 27 lines · Jan Beulich
Quoted 26 lines · Andrew Cooper

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 208

Quoted 7 lines · Daniel P. Smith
xen/xsm/xsm_core.c +2 -1
@@ -197,16 +204,21 @@ bool __init has_xsm_magic(paddr_t start)
 
 int __init register_xsm(struct xsm_operations *ops)
 {
-    if ( verify(ops) )
+    if ( xsm_ops_registered != XSM_OPS_UNREGISTERED )
+        return -EAGAIN;

I know you moved this around the function, but it really isn't -EAGAIN material any more.  It's "too late - nope".

-EEXIST is probably best for "I'm never going to tolerate another call".

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 214

Quoted 7 lines
xen/xsm/xsm_core.c +4 -1
+
+    if ( !ops )
     {
-        printk(XENLOG_ERR "Could not verify xsm_operations structure\n");
+        xsm_ops_registered = XSM_OPS_REG_FAILED;
+        printk(XENLOG_ERR "Invalid xsm_operations structure registered\n");
         return -EINVAL;

Honestly, I'd be half tempted to declare register_xsm() with __nonnull(0) and let the compiler reject any attempt to pass a NULL ops pointer.

Quoted 5 lines
Quoted 1 line

Both callers pass a pointer to a static singleton objects.

Why check at all when the source of the arguments is all internal? We don't check pointers to be non-NULL elsewhere, with a few odd exceptions (which imo should all be dropped).

In verify(ops) there was a check on ops for NULL, I pulled that check up into here as I removed verify(). Based on Andy's comment, this function is looking like it is on the chopping block as well.

v/r,
dps

#6 lore raw

Re: [PATCH 1/6] xsm: refactor xsm_ops handling

In reply to Andrew Cooper
Quoted 14 lines · Andrew Cooper
Quoted 5 lines · Daniel P. Smith

The assignment and setup of xsm_ops structure was refactored to make it a one-time assignment. The calling of the xsm_ops were refactored to use the alternate_call framework to reduce the need for retpolines.

I think the commit message needs a little more explanation for anyone doing code archaeology.

AFAICT, the current infrastructure has some (incomplete?) support for Flask to unload itself as the security engine, which doesn't sounds like a clever thing in general.

Quoted 7 lines

What we do here is make a semantic change to say that the security engine (Dummy, Flask or SILO) gets chosen once during boot, and is immutable thereafter.  This is better from a security standpoint (no accidentally unloading Flask at runtime), and allows for the use of the alternative_vcall() infrastructure to drop all the function pointer calls.

Does that about sum things up?

ack

Quoted 56 lines
Quoted 19 lines
xen/xsm/flask/flask_op.c +0 -11
index 01e52138a1..df9fcc1d6d 100644
@@ -225,26 +225,7 @@ static int flask_security_sid(struct xen_flask_sid_context *arg)
 
 static int flask_disable(void)
 {
-    static int flask_disabled = 0;
-
-    if ( ss_initialized )
-    {
-        /* Not permitted after initial policy load. */
-        return -EINVAL;
-    }
-
-    if ( flask_disabled )
-    {
-        /* Only do this once. */

xen/xsm/flask/flask_op.c static int flask_security_sid(struct xen_flask_sid_context *arg) · line 228

Quoted 10 lines
xen/xsm/flask/flask_op.c +1 -9
-        return -EINVAL;
-    }
-
-    printk("Flask:  Disabled at runtime.\n");
-
-    flask_disabled = 1;
-
-    /* Reset xsm_ops to the original module. */
-    xsm_ops = &dummy_xsm_ops;
+    printk("Flask:  Disabling is not supported.\n");

Judging by this, should this patch be split up more?

I think you want to remove FLASK_DISABLE (and this function too - just return -EOPNOTSUPP in the parent) as a separate explained change (as it is a logical change in how Flask works).

The second patch wants to be the rest, which changes the indirection of xsm_ops and converts to vcall().  This is a fairly mechanical change without semantic changes.

I'm unsure if you want a 3rd patch in the middle, separating the xsm_core_init() juggling, with the "switch to using vcall()".  It might be a good idea for more easily demonstrating the changes, but I'll leave it to your judgement.

Quoted 11 lines
xen/xsm/xsm_core.c +0 -0
index 5eab21e1b1..acc1af7166 100644
 static int __init xsm_core_init(const void *policy_buffer, size_t policy_size)
 {
 #ifdef CONFIG_XSM_FLASK_POLICY
@@ -87,17 +86,22 @@ static int __init xsm_core_init(const void *policy_buffer, size_t policy_size)
     }
 #endif
 
Quoted 8 lines

xen/xsm/xsm_core.c static int __init xsm_core_init(const void *policy_buffer, size_t policy_size) · line 92

Quoted 6 lines
xen/xsm/xsm_core.c +3 -2
-    if ( verify(&dummy_xsm_ops) )
+    if ( xsm_ops_registered != XSM_OPS_UNREGISTERED )
     {
-        printk(XENLOG_ERR "Could not verify dummy_xsm_ops structure\n");
+        printk(XENLOG_ERR
+            "Could not init XSM, xsm_ops register already attempted\n");

Indentation.

ack

Quoted 9 lines

xen/xsm/xsm_core.c static int __init xsm_core_init(const void *policy_buffer, size_t policy_size) · line 99

Quoted 8 lines
xen/xsm/xsm_core.c +4 -1
         return -EIO;
     }
 
-    xsm_ops = &dummy_xsm_ops;
+    /* install the dummy ops as default to ensure ops
+     * are defined if requested policy fails init
+     */
+    xsm_fixup_ops(&xsm_ops);
Quoted 8 lines

/* Comment style. */

or

/*
 * Multi-
 * line comment style.
 */

ack

Quoted 28 lines
Quoted 18 lines
xen/xsm/xsm_core.c +4 -0
     switch ( xsm_bootparam )
     {
     case XSM_BOOTPARAM_DUMMY:
+        xsm_ops_registered = XSM_OPS_REGISTERED;
         break;
 
     case XSM_BOOTPARAM_FLASK:
@@ -113,6 +117,9 @@ static int __init xsm_core_init(const void *policy_buffer, size_t policy_size)
         break;
     }
 
+    if ( xsm_ops_registered != XSM_OPS_REGISTERED )
+        xsm_ops_registered = XSM_OPS_REG_FAILED;
+
     return 0;
 }
 

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 208

Quoted 6 lines
xen/xsm/xsm_core.c +2 -1
@@ -197,16 +204,21 @@ bool __init has_xsm_magic(paddr_t start)
 
 int __init register_xsm(struct xsm_operations *ops)
 {
-    if ( verify(ops) )
+    if ( xsm_ops_registered != XSM_OPS_UNREGISTERED )
+        return -EAGAIN;

I know you moved this around the function, but it really isn't -EAGAIN material any more.  It's "too late - nope".

Quoted 1 line

-EEXIST is probably best for "I'm never going to tolerate another call".

Honestly I didn't think EAGAIN was correct in the first place, so will gladly change it.

Quoted 16 lines

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 214

Quoted 7 lines
xen/xsm/xsm_core.c +4 -1
+
+    if ( !ops )
     {
-        printk(XENLOG_ERR "Could not verify xsm_operations structure\n");
+        xsm_ops_registered = XSM_OPS_REG_FAILED;
+        printk(XENLOG_ERR "Invalid xsm_operations structure registered\n");
         return -EINVAL;

Honestly, I'd be half tempted to declare register_xsm() with __nonnull(0) and let the compiler reject any attempt to pass a NULL ops pointer.

Both callers pass a pointer to a static singleton objects.

Quoted 2 lines
xen/xsm/xsm_core.c +0 -0
     }
 
Quoted 7 lines

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 218

Quoted 4 lines
xen/xsm/xsm_core.c +2 -2
-    if ( xsm_ops != &dummy_xsm_ops )
-        return -EAGAIN;
+    /* use dummy ops for any empty ops */
+    xsm_fixup_ops(ops);

Isn't this redundant with the call in xsm_core_init(), seeing as register_xsm() must be nested within the switch statement?

I don't believe so, the one in xsm_core_init is setting the default/fallback of xsm_ops var to dummy_* before attempting to register a policy module's ops. Here in register_xsm we are taking a new instance of a struct xsm_ops and ensuring every handler has a defined entry before overwriting the xsm_ops variable with passed in xsm module's ops. Now with that said, I do like your suggest down at the end.

Quoted 7 lines

xen/xsm/xsm_core.c bool __init has_xsm_magic(paddr_t start) · line 224

Quoted 6 lines
xen/xsm/xsm_core.c +2 -1
-    xsm_ops = ops;
+    xsm_ops = *ops;
+    xsm_ops_registered = XSM_OPS_REGISTERED;
 
     return 0;
 }
Quoted 8 lines

Having got to the end, the xsm_core_init() vs register_xsm() dynamic is quite weird.

I think it would result in clearer code to have init_{flask,silo}() return pointers to their struct xsm_operations *, and let xsm_core_init() do the copy in to xsm_ops.  This reduces the scope of xsm_ops_state to this function only, and gets rid of at least one runtime panic() call which is dead code.

I agree full.

Quoted 2 lines

If you were to go with this approach, you'd definitely want to split into the 3-patch approach.

v2 will have this broken out

v/r,
dps

#7 lore raw

PATCH 1/1 xen: Implement xen/alternative-call.h for use in common code

In reply to Daniel P. Smith

The alternative call infrastructure is x86-only for now, but the common iommu code has a variant and more common code wants to use the infrastructure.

Introduce CONFIG_ALTERNATIVE_CALL and a conditional implemetnation so common code can use the optimisation when available, without requiring all architectures to implement no-op stubs.

Write some documentation, which was thus far entirely absent, covering the requirements for an architecture to implement this optimsiation, and how to use the infrastructure in general code.

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
CC: Stefano Stabellini <[email protected]>
CC: Julien Grall <[email protected]>
CC: Volodymyr Babchuk <[email protected]>
CC: Bob Eshleman <[email protected]>
CC: Alistair Francis <[email protected]>
CC: Connor Davis <[email protected]>
CC: Daniel P. Smith <[email protected]>

This is a pre-requisite to "xsm: refactor xsm_ops handling" to avoid breaking the ARM build.

Build test for the XSM code:

  diff --git a/xen/xsm/xsm_core.c b/xen/xsm/xsm_core.c
  index 5eab21e1b168..592074e8f41c 100644
  --- a/xen/xsm/xsm_core.c
  +++ b/xen/xsm/xsm_core.c
  @@ -195,6 +195,16 @@ bool __init has_xsm_magic(paddr_t start)
   }
    #endif

+#include <xen/alternative-call.h>
+struct foo {
+ int (*bar)(void *);
+} foo __alt_call_maybe_initdata;
+
+int test_alternative_call(void)
+{
+ return alternative_call(foo.bar, NULL);
+}
+
int __init register_xsm(struct xsm_operations *ops)
{
if ( verify(ops) )

---
 xen/arch/x86/Kconfig               |  1 +
 xen/common/Kconfig                 |  3 ++
 xen/include/xen/alternative-call.h | 65 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 69 insertions(+)
 create mode 100644 xen/include/xen/alternative-call.h
xen/arch/x86/Kconfig +1 -0
index 9b164db64187..c91cdd83dc8a 100644
@@ -9,6 +9,7 @@ config X86
 	select ARCH_SUPPORTS_INT128
 	select CORE_PARKING
 	select HAS_ALTERNATIVE
+	select ALTERNATIVE_CALL
 	select HAS_COMPAT
 	select HAS_CPUFREQ
 	select HAS_EHCI
xen/common/Kconfig +3 -0
index 0ddd18e11af3..1594ce4e7313 100644
@@ -25,6 +25,9 @@ config GRANT_TABLE
 config HAS_ALTERNATIVE
 	bool
 
+config ALTERNATIVE_CALL
+	bool
+
 config HAS_COMPAT
 	bool
 
xen/include/xen/alternative-call.h new file +65 -0

65 lines not on this page · raw message

Signature · 1 line
2.11.0
#8 lore raw

PATCH 2/6 xsm: decouple xsm header inclusion selection

In reply to Daniel P. Smith

Multiple preprocessor defines were used as a mechanism to selective include parts of the xsm.h header file. This makes it difficult to know which portion is being included at anyone time. This commit works to simplify this by separating the core structure and functions of XSM into xsm-core.h away from the wrapper functions which remain in xsm.h and dummy.h.

---
 xen/include/xsm/dummy.h    |   2 +-
 xen/include/xsm/xsm-core.h | 262 +++++++++++++++++++++++++++++++++++++
 xen/include/xsm/xsm.h      | 240 +--------------------------------
 xen/xsm/dummy.c            |   1 -
 xen/xsm/silo.c             |   1 -
 5 files changed, 264 insertions(+), 242 deletions(-)
 create mode 100644 xen/include/xsm/xsm-core.h
xen/include/xsm/dummy.h +1 -1
index 363c6d7798..c445c5681b 100644
@@ -16,7 +16,7 @@
  */
 
 #include <xen/sched.h>
-#include <xsm/xsm.h>
+#include <xsm/xsm-core.h>
 #include <public/hvm/params.h>
 
 /* Cannot use BUILD_BUG_ON here because the expressions we check are not
xen/include/xsm/xsm-core.h new file +262 -0

262 lines not on this page · raw message

xen/include/xsm/xsm.h +1 -239

261 lines not on this page · raw message

xen/xsm/dummy.c +0 -1
index 066694763a..de4d6cf2cf 100644
@@ -10,7 +10,6 @@
  *  as published by the Free Software Foundation.
  */
 
-#define XSM_NO_WRAPPERS
 #include <xsm/dummy.h>
 
 #define set_to_dummy_if_null(ops, function)                            \
xen/xsm/silo.c +0 -1
index fc2ca5cd2d..b96dacd181 100644
@@ -17,7 +17,6 @@
  * You should have received a copy of the GNU General Public License along with
  * this program; If not, see <http://www.gnu.org/licenses/>.
  */
-#define XSM_NO_WRAPPERS
 #include <xsm/dummy.h>
 
 /*
Signature · 1 line
2.20.1
#9 lore raw

PATCH 3/6 xsm: enabling xsm to always be included

In reply to Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM
* introduces CONFIG_XSM_EVTCHN_LABELING as replacement for enabling event channel labels
* makes CONFIG_XSM_SILO AND CONFIG_XSM_FLASK default to no

---
 xen/common/Kconfig            |  55 ++++-----
 xen/include/xen/sched.h       |   2 +-
 xen/include/xsm/xsm-core.h    |  26 ----
 xen/include/xsm/xsm.h         |   8 --
 xen/xsm/Makefile              |   4 +-
 xen/xsm/dummy.c               |   4 +-
 xen/{include => }/xsm/dummy.h | 220 ++++++++++++++++------------------
 xen/xsm/silo.c                |  17 +--
 xen/xsm/xsm_core.c            |   4 -
 9 files changed, 142 insertions(+), 198 deletions(-)
 rename xen/{include => }/xsm/dummy.h (63%)
xen/common/Kconfig +26 -29
index 0ddd18e11a..203ad7ea23 100644
@@ -197,22 +197,33 @@ config XENOPROF
 
 	  If unsure, say Y.
 
-config XSM
-	bool "Xen Security Modules support"
-	default ARM
-	---help---
-	  Enables the security framework known as Xen Security Modules which
-	  allows administrators fine-grained control over a Xen domain and
-	  its capabilities by defining permissible interactions between domains,
-	  the hypervisor itself, and related resources such as memory and
-	  devices.
+menu "Xen Security Modules"
 
-	  If unsure, say N.
+choice
+	prompt "Default XSM module"
+	default XSM_SILO_DEFAULT if XSM_SILO && ARM
+	default XSM_FLASK_DEFAULT if XSM_FLASK
+	default XSM_SILO_DEFAULT if XSM_SILO
+	default XSM_DUMMY_DEFAULT
+	config XSM_DUMMY_DEFAULT
+		bool "Match non-XSM behavior"
+	config XSM_FLASK_DEFAULT
+		bool "FLux Advanced Security Kernel" if XSM_FLASK
+	config XSM_SILO_DEFAULT
+		bool "SILO" if XSM_SILO
+endchoice
+
+config XSM_EVTCHN_LABELING
+	bool "Enables security labeling of event channels"
+	default n
+	---help---
+      This enables an XSM module to label and enforce access control over
+      event channels.
 
 config XSM_FLASK
-	def_bool y
+	def_bool n
 	prompt "FLux Advanced Security Kernel support"
-	depends on XSM
+	select XSM_EVTCHN_LABELING
 	---help---
 	  Enables FLASK (FLux Advanced Security Kernel) as the access control
 	  mechanism used by the XSM framework.  This provides a mandatory access
@@ -250,9 +261,8 @@ config XSM_FLASK_POLICY
 	  If unsure, say Y.
 
 config XSM_SILO
-	def_bool y
+	def_bool n
 	prompt "SILO support"
-	depends on XSM
 	---help---
 	  Enables SILO as the access control mechanism used by the XSM framework.
 	  This is not the default module, add boot parameter xsm=silo to choose
@@ -261,25 +271,12 @@ config XSM_SILO
 
 	  If unsure, say Y.
 
-choice
-	prompt "Default XSM implementation"
-	depends on XSM
-	default XSM_SILO_DEFAULT if XSM_SILO && ARM
-	default XSM_FLASK_DEFAULT if XSM_FLASK
-	default XSM_SILO_DEFAULT if XSM_SILO
-	default XSM_DUMMY_DEFAULT
-	config XSM_DUMMY_DEFAULT
-		bool "Match non-XSM behavior"
-	config XSM_FLASK_DEFAULT
-		bool "FLux Advanced Security Kernel" if XSM_FLASK
-	config XSM_SILO_DEFAULT
-		bool "SILO" if XSM_SILO
-endchoice
+endmenu
 
 config LATE_HWDOM
 	bool "Dedicated hardware domain"
 	default n
-	depends on XSM && X86
+	depends on XSM_FLASK && X86
 	---help---
 	  Allows the creation of a dedicated hardware domain distinct from
 	  domain 0 that manages devices without needing access to other
xen/include/xen/sched.h +1 -1
index 3982167144..e92d2cdeab 100644
@@ -120,7 +120,7 @@ struct evtchn
     unsigned short notify_vcpu_id; /* VCPU for local delivery notification */
     uint32_t fifo_lastq;           /* Data for identifying last queue. */
 
-#ifdef CONFIG_XSM
+#ifdef CONFIG_XSM_EVTCHN_LABELING
     union {
 #ifdef XSM_NEED_GENERIC_EVTCHN_SSID
         /*
xen/include/xsm/xsm-core.h +0 -26
index 5297c73fe6..e3718bc62d 100644
@@ -189,8 +189,6 @@ struct xsm_operations {
 #endif
 };
 
-#ifdef CONFIG_XSM
-
 #ifdef CONFIG_MULTIBOOT
 extern int xsm_multiboot_init(unsigned long *module_map,
                               const multiboot_info_t *mbi);
@@ -235,28 +233,4 @@ extern void silo_init(void);
 static inline void silo_init(void) {}
 #endif
 
-#else /* CONFIG_XSM */
-
-#ifdef CONFIG_MULTIBOOT
-static inline int xsm_multiboot_init (unsigned long *module_map,
-                                      const multiboot_info_t *mbi)
-{
-    return 0;
-}
-#endif
-
-#ifdef CONFIG_HAS_DEVICE_TREE
-static inline int xsm_dt_init(void)
-{
-    return 0;
-}
-
-static inline bool has_xsm_magic(paddr_t start)
-{
-    return false;
-}
-#endif /* CONFIG_HAS_DEVICE_TREE */
-
-#endif /* CONFIG_XSM */
-
 #endif /* __XSM_CORE_H */
xen/include/xsm/xsm.h +0 -8
index ecbdee2c7d..69f68300e2 100644
@@ -22,8 +22,6 @@
 #include <xen/multiboot.h>
 #include <xsm/xsm-core.h>
 
-#ifdef CONFIG_XSM
-
 extern struct xsm_operations xsm_ops;
 
 static inline void xsm_security_domaininfo (struct domain *d,
@@ -559,10 +557,4 @@ static inline int xsm_argo_send(const struct domain *d, const struct domain *t)
 
 #endif /* CONFIG_ARGO */
 
-#else /* CONFIG_XSM */
-
-#include <xsm/dummy.h>
-
-#endif /* CONFIG_XSM */
-
 #endif /* __XSM_H */
xen/xsm/Makefile +2 -2
index cf0a728f1c..0059794dd5 100644
@@ -1,6 +1,6 @@
 obj-y += xsm_core.o
-obj-$(CONFIG_XSM) += xsm_policy.o
-obj-$(CONFIG_XSM) += dummy.o
+obj-y += xsm_policy.o
+obj-y += dummy.o
 obj-$(CONFIG_XSM_SILO) += silo.o
 
 obj-$(CONFIG_XSM_FLASK) += flask/
xen/xsm/dummy.c +2 -2
index de4d6cf2cf..bfd8b96f08 100644
@@ -10,12 +10,12 @@
  *  as published by the Free Software Foundation.
  */
 
-#include <xsm/dummy.h>
+#include "dummy.h"
 
 #define set_to_dummy_if_null(ops, function)                            \
     do {                                                               \
         if ( !ops->function )                                          \
-            ops->function = xsm_##function;                            \
+            ops->function = dummy_##function;                          \
     } while (0)
 
 void __init xsm_fixup_ops (struct xsm_operations *ops)
xen/xsm/dummy.h renamed from xen/include/xsm/dummy.h +102 -118

742 lines not on this page · raw message

xen/xsm/silo.c +9 -8
index b96dacd181..ebe5814efc 100644
@@ -17,7 +17,8 @@
  * You should have received a copy of the GNU General Public License along with
  * this program; If not, see <http://www.gnu.org/licenses/>.
  */
-#include <xsm/dummy.h>
+#include <xsm/xsm-core.h>
+#include "dummy.h"
 
 /*
  * Check if inter-domain communication is allowed.
@@ -43,7 +44,7 @@ static int silo_evtchn_unbound(struct domain *d1, struct evtchn *chn,
     else
     {
         if ( silo_mode_dom_check(d1, d2) )
-            rc = xsm_evtchn_unbound(d1, chn, id2);
+            rc = dummy_evtchn_unbound(d1, chn, id2);
         rcu_unlock_domain(d2);
     }
 
@@ -54,7 +55,7 @@ static int silo_evtchn_interdomain(struct domain *d1, struct evtchn *chan1,
                                    struct domain *d2, struct evtchn *chan2)
 {
     if ( silo_mode_dom_check(d1, d2) )
-        return xsm_evtchn_interdomain(d1, chan1, d2, chan2);
+        return dummy_evtchn_interdomain(d1, chan1, d2, chan2);
     return -EPERM;
 }
 
@@ -62,21 +63,21 @@ static int silo_grant_mapref(struct domain *d1, struct domain *d2,
                              uint32_t flags)
 {
     if ( silo_mode_dom_check(d1, d2) )
-        return xsm_grant_mapref(d1, d2, flags);
+        return dummy_grant_mapref(d1, d2, flags);
     return -EPERM;
 }
 
 static int silo_grant_transfer(struct domain *d1, struct domain *d2)
 {
     if ( silo_mode_dom_check(d1, d2) )
-        return xsm_grant_transfer(d1, d2);
+        return dummy_grant_transfer(d1, d2);
     return -EPERM;
 }
 
 static int silo_grant_copy(struct domain *d1, struct domain *d2)
 {
     if ( silo_mode_dom_check(d1, d2) )
-        return xsm_grant_copy(d1, d2);
+        return dummy_grant_copy(d1, d2);
     return -EPERM;
 }
 
@@ -86,14 +87,14 @@ static int silo_argo_register_single_source(const struct domain *d1,
                                             const struct domain *d2)
 {
     if ( silo_mode_dom_check(d1, d2) )
-        return xsm_argo_register_single_source(d1, d2);
+        return dummy_argo_register_single_source(d1, d2);
     return -EPERM;
 }
 
 static int silo_argo_send(const struct domain *d1, const struct domain *d2)
 {
     if ( silo_mode_dom_check(d1, d2) )
-        return xsm_argo_send(d1, d2);
+        return dummy_argo_send(d1, d2);
     return -EPERM;
 }
 
xen/xsm/xsm_core.c +0 -4
index acc1af7166..a51dc6f7dd 100644
@@ -18,8 +18,6 @@
 #include <xen/hypercall.h>
 #include <xsm/xsm.h>
 
-#ifdef CONFIG_XSM
-
 #ifdef CONFIG_MULTIBOOT
 #include <asm/setup.h>
 #endif
@@ -223,8 +221,6 @@ int __init register_xsm(struct xsm_operations *ops)
     return 0;
 }
 
-#endif
-
 long do_xsm_op (XEN_GUEST_HANDLE_PARAM(xsm_op_t) op)
 {
     return xsm_do_xsm_op(op);
Signature · 1 line
2.20.1
#10 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 39 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM
* introduces CONFIG_XSM_EVTCHN_LABELING as replacement for enabling event channel labels
* makes CONFIG_XSM_SILO AND CONFIG_XSM_FLASK default to no

---
 xen/common/Kconfig            |  55 ++++-----
 xen/include/xen/sched.h       |   2 +-
 xen/include/xsm/xsm-core.h    |  26 ----
 xen/include/xsm/xsm.h         |   8 --
 xen/xsm/Makefile              |   4 +-
 xen/xsm/dummy.c               |   4 +-
 xen/{include => }/xsm/dummy.h | 220 ++++++++++++++++------------------
 xen/xsm/silo.c                |  17 +--
 xen/xsm/xsm_core.c            |   4 -
 9 files changed, 142 insertions(+), 198 deletions(-)
 rename xen/{include => }/xsm/dummy.h (63%)
xen/common/Kconfig +1 -9
index 0ddd18e11a..203ad7ea23 100644
@@ -197,22 +197,33 @@ config XENOPROF
 
 	  If unsure, say Y.
 
-config XSM
-	bool "Xen Security Modules support"
-	default ARM
-	---help---
-	  Enables the security framework known as Xen Security Modules which
-	  allows administrators fine-grained control over a Xen domain and
-	  its capabilities by defining permissible interactions between domains,
-	  the hypervisor itself, and related resources such as memory and
-	  devices.
+menu "Xen Security Modules"

xen/common/Kconfig config XENOPROF · line 209

Quoted 10 lines
xen/common/Kconfig +8 -1
 
-	  If unsure, say N.
+choice
+	prompt "Default XSM module"
+	default XSM_SILO_DEFAULT if XSM_SILO && ARM
+	default XSM_FLASK_DEFAULT if XSM_FLASK
+	default XSM_SILO_DEFAULT if XSM_SILO
+	default XSM_DUMMY_DEFAULT
+	config XSM_DUMMY_DEFAULT
+		bool "Match non-XSM behavior"

There is no non-XSM behaviour any more.

Is it time to rename Dummy to "traditional dom0-all-powerful" or something suitable?

xen/common/Kconfig config XENOPROF · line 221

Quoted 12 lines
xen/common/Kconfig +12 -0
+	config XSM_FLASK_DEFAULT
+		bool "FLux Advanced Security Kernel" if XSM_FLASK
+	config XSM_SILO_DEFAULT
+		bool "SILO" if XSM_SILO
+endchoice
+
+config XSM_EVTCHN_LABELING
+	bool "Enables security labeling of event channels"
+	default n
+	---help---
+      This enables an XSM module to label and enforce access control over
+      event channels.

Please use help rather than ---help--- for new options (its changed in upstream Kconfig).  The indentation of the help message wants to be one tab, then two spaces.  (Yes, sadly...)

Quoted 10 lines
xen/common/Kconfig +2 -2
 config XSM_FLASK
-	def_bool y
+	def_bool n
 	prompt "FLux Advanced Security Kernel support"
-	depends on XSM
+	select XSM_EVTCHN_LABELING
 	---help---
 	  Enables FLASK (FLux Advanced Security Kernel) as the access control
 	  mechanism used by the XSM framework.  This provides a mandatory access

xen/common/Kconfig config XSM_FLASK_POLICY · line 264

Quoted 5 lines
xen/common/Kconfig +1 -1
@@ -250,9 +261,8 @@ config XSM_FLASK_POLICY
 	  If unsure, say Y.
 
 config XSM_SILO
-	def_bool y
+	def_bool n

I'm not sure we want to alter the FLASK/SILO defaults.  SILO in particular is mandatory on ARM, and without it, you're in a security unsupported configuration.

~Andrew

#11 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 53 lines · Andrew Cooper
Quoted 39 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM
* introduces CONFIG_XSM_EVTCHN_LABELING as replacement for enabling event channel labels
* makes CONFIG_XSM_SILO AND CONFIG_XSM_FLASK default to no

---
 xen/common/Kconfig            |  55 ++++-----
 xen/include/xen/sched.h       |   2 +-
 xen/include/xsm/xsm-core.h    |  26 ----
 xen/include/xsm/xsm.h         |   8 --
 xen/xsm/Makefile              |   4 +-
 xen/xsm/dummy.c               |   4 +-
 xen/{include => }/xsm/dummy.h | 220 ++++++++++++++++------------------
 xen/xsm/silo.c                |  17 +--
 xen/xsm/xsm_core.c            |   4 -
 9 files changed, 142 insertions(+), 198 deletions(-)
 rename xen/{include => }/xsm/dummy.h (63%)
xen/common/Kconfig +1 -9
index 0ddd18e11a..203ad7ea23 100644
@@ -197,22 +197,33 @@ config XENOPROF
 
 	  If unsure, say Y.
 
-config XSM
-	bool "Xen Security Modules support"
-	default ARM
-	---help---
-	  Enables the security framework known as Xen Security Modules which
-	  allows administrators fine-grained control over a Xen domain and
-	  its capabilities by defining permissible interactions between domains,
-	  the hypervisor itself, and related resources such as memory and
-	  devices.
+menu "Xen Security Modules"

xen/common/Kconfig config XENOPROF · line 209

Quoted 10 lines
xen/common/Kconfig +8 -1
 
-	  If unsure, say N.
+choice
+	prompt "Default XSM module"
+	default XSM_SILO_DEFAULT if XSM_SILO && ARM
+	default XSM_FLASK_DEFAULT if XSM_FLASK
+	default XSM_SILO_DEFAULT if XSM_SILO
+	default XSM_DUMMY_DEFAULT
+	config XSM_DUMMY_DEFAULT
+		bool "Match non-XSM behavior"

There is no non-XSM behaviour any more.

Quoted 1 line

Is it time to rename Dummy to "traditional dom0-all-powerful" or

Well, I left as dummy since that is what it has been known by thus far and additionally the subsequent patch set was going to rename this to XSM_ROLES/"XSM Role-based Access Control" For the intermediate time, I can change the wording to reflect the correct state.

Quoted 8 lines
Quoted 8 lines
xen/common/Kconfig +8 -0
+	config XSM_FLASK_DEFAULT
+		bool "FLux Advanced Security Kernel" if XSM_FLASK
+	config XSM_SILO_DEFAULT
+		bool "SILO" if XSM_SILO
+endchoice
+
+config XSM_EVTCHN_LABELING
+	bool "Enables security labeling of event channels"
Quoted 8 lines

xen/common/Kconfig config XENOPROF · line 221

Quoted 4 lines
xen/common/Kconfig +4 -0
+	default n
+	---help---
+      This enables an XSM module to label and enforce access control over
+      event channels.

Please use help rather than ---help--- for new options (its changed in upstream Kconfig).  The indentation of the help message wants to be one tab, then two spaces.  (Yes, sadly...)

ack

Quoted 12 lines
Quoted 12 lines
xen/common/Kconfig +2 -2
 config XSM_FLASK
-	def_bool y
+	def_bool n
 	prompt "FLux Advanced Security Kernel support"
-	depends on XSM
+	select XSM_EVTCHN_LABELING
 	---help---
 	  Enables FLASK (FLux Advanced Security Kernel) as the access control
 	  mechanism used by the XSM framework.  This provides a mandatory access
@@ -250,9 +261,8 @@ config XSM_FLASK_POLICY
 	  If unsure, say Y.
 
Quoted 7 lines

xen/common/Kconfig config XSM_FLASK_POLICY · line 264

Quoted 3 lines
xen/common/Kconfig +1 -1
 config XSM_SILO
-	def_bool y
+	def_bool n

I'm not sure we want to alter the FLASK/SILO defaults.  SILO in particular is mandatory on ARM, and without it, you're in a security unsupported configuration.

The intent here is the default is the classic dom0 configuration. What if I did,

def bool n
def bool y if ARM

v/r
dps

#12 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 12 lines · Daniel P. Smith
Quoted 11 lines · Andrew Cooper

xen/common/Kconfig config XSM_FLASK_POLICY · line 264

Quoted 6 lines · Daniel P. Smith
xen/common/Kconfig +1 -1
@@ -250,9 +261,8 @@ config XSM_FLASK_POLICY
 	  If unsure, say Y.
 
 config XSM_SILO
-	def_bool y
+	def_bool n

I'm not sure we want to alter the FLASK/SILO defaults.  SILO in particular is mandatory on ARM, and without it, you're in a security unsupported configuration.

Quoted 5 lines

The intent here is the default is the classic dom0 configuration. What if I did,

def bool n
def bool y if ARM

Besides it needing to be with the order of the two lines flipped, if Arm requires XSM_SILO, then I think it would better "select" it.

Jan

#13 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 13 lines · Jan Beulich
Quoted 12 lines · Daniel P. Smith
Quoted 11 lines · Andrew Cooper

xen/common/Kconfig config XSM_FLASK_POLICY · line 264

Quoted 6 lines · Daniel P. Smith
xen/common/Kconfig +1 -1
@@ -250,9 +261,8 @@ config XSM_FLASK_POLICY
  	  If unsure, say Y.
 
  config XSM_SILO
-	def_bool y
+	def_bool n

I'm not sure we want to alter the FLASK/SILO defaults.  SILO in particular is mandatory on ARM, and without it, you're in a security unsupported configuration.

Quoted 8 lines
Quoted 5 lines

The intent here is the default is the classic dom0 configuration. What if I did,

def bool n
def bool y if ARM

Besides it needing to be with the order of the two lines flipped, if Arm requires XSM_SILO, then I think it would better "select" it.

Ack, I realized that as I fixed it for the upcoming v2.

Correct me if I am wrong but if you do a "select" that means you are forcing the user to always have SILO built in, i.e. that makes it so the option cannot be disabled. There may be users who would prefer to only have Flask enabled on ARM and those users would not be able to turn SILO off.

v/r,
dps

#14 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 26 lines · Daniel P. Smith
Quoted 13 lines · Jan Beulich
Quoted 12 lines · Daniel P. Smith
Quoted 11 lines · Andrew Cooper

xen/common/Kconfig config XSM_FLASK_POLICY · line 264

Quoted 6 lines · Daniel P. Smith
xen/common/Kconfig +1 -1
@@ -250,9 +261,8 @@ config XSM_FLASK_POLICY
  	  If unsure, say Y.
 
  config XSM_SILO
-	def_bool y
+	def_bool n

I'm not sure we want to alter the FLASK/SILO defaults.  SILO in particular is mandatory on ARM, and without it, you're in a security unsupported configuration.

Quoted 8 lines
Quoted 5 lines

The intent here is the default is the classic dom0 configuration. What if I did,

def bool n
def bool y if ARM

Besides it needing to be with the order of the two lines flipped, if Arm requires XSM_SILO, then I think it would better "select" it.

Quoted 7 lines

Ack, I realized that as I fixed it for the upcoming v2.

Correct me if I am wrong but if you do a "select" that means you are forcing the user to always have SILO built in, i.e. that makes it so the option cannot be disabled. There may be users who would prefer to only have Flask enabled on ARM and those users would not be able to turn SILO off.

Yes, you're right. Problem is the (imo) malformed entry, which makes it that I couldn't see the presence of a prompt anymore in the context above. Well-formed (imo; I might also say "consistently formatted") entries with a prompt ought to look like (taking your change into account already, leaving aside whether that's really what we want)

config XSM_SILO
bool "SILO support"
default y if ARM
default n

Whether "depends" precedes or follows "default" is a less clear cut.

def_bool imo would better be used only for prompt-less entries.

Jan

#15 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

In reply to Daniel P. Smith
Quoted 4 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM

Following from what Andrew has said (including him mentioning your changing of certain Kconfig option defaults), I'm not convinced this is a good move. This still ought to serve as the overall XSM-yes-or-no setting. If internally you make said two variants match in behavior, that's a different thing.

Quoted 1 line

* introduces CONFIG_XSM_EVTCHN_LABELING as replacement for enabling event channel labels

Is this mode needed as separate functionality at all? Nothing defines XSM_NEED_GENERIC_EVTCHN_SSID anywhere. _If_ XSM went away as a separate setting, then imo this one should go away as well.

Quoted 21 lines
xen/common/Kconfig +4 -10
@@ -197,22 +197,33 @@ config XENOPROF
 
 	  If unsure, say Y.
 
-config XSM
-	bool "Xen Security Modules support"
-	default ARM
-	---help---
-	  Enables the security framework known as Xen Security Modules which
-	  allows administrators fine-grained control over a Xen domain and
-	  its capabilities by defining permissible interactions between domains,
-	  the hypervisor itself, and related resources such as memory and
-	  devices.
+menu "Xen Security Modules"
 
-	  If unsure, say N.
+choice
+	prompt "Default XSM module"
+	default XSM_SILO_DEFAULT if XSM_SILO && ARM

xen/common/Kconfig config XENOPROF · line 214

Quoted 10 lines
xen/common/Kconfig +10 -0
+	default XSM_FLASK_DEFAULT if XSM_FLASK
+	default XSM_SILO_DEFAULT if XSM_SILO
+	default XSM_DUMMY_DEFAULT
+	config XSM_DUMMY_DEFAULT
+		bool "Match non-XSM behavior"
+	config XSM_FLASK_DEFAULT
+		bool "FLux Advanced Security Kernel" if XSM_FLASK
+	config XSM_SILO_DEFAULT
+		bool "SILO" if XSM_SILO
+endchoice

This did live after the individual options it depends on for a reason, and you don't say anywhere why you need to move it up. The way you have it, with the default command line kconfig tool, users will be presented with dependent options before having chosen the settings of the dependency ones. That's because this tool, to a degree, moves linearly through the options it has parsed.

Quoted 15 lines
xen/common/Kconfig +0 -11
@@ -261,25 +271,12 @@ config XSM_SILO
 
 	  If unsure, say Y.
 
-choice
-	prompt "Default XSM implementation"
-	depends on XSM
-	default XSM_SILO_DEFAULT if XSM_SILO && ARM
-	default XSM_FLASK_DEFAULT if XSM_FLASK
-	default XSM_SILO_DEFAULT if XSM_SILO
-	default XSM_DUMMY_DEFAULT
-	config XSM_DUMMY_DEFAULT
-		bool "Match non-XSM behavior"
-	config XSM_FLASK_DEFAULT
-		bool "FLux Advanced Security Kernel" if XSM_FLASK

xen/common/Kconfig config XSM_SILO · line 279

Quoted 10 lines
xen/common/Kconfig +2 -4
-	config XSM_SILO_DEFAULT
-		bool "SILO" if XSM_SILO
-endchoice
+endmenu
 
 config LATE_HWDOM
 	bool "Dedicated hardware domain"
 	default n
-	depends on XSM && X86
+	depends on XSM_FLASK && X86

I don't think this is a compatible change. I'm not going to exclude that this is how it was meant, but as it stands LATE_HWDOM right now doesn't really require FLASK, and could e.g. also go with SILO or dummy. If you _mean_ to change this, then your description needs to say so (and ideally it would then be split out, so - if this is actually a bug - it could also be backported).

Jan

#16 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 11 lines · Jan Beulich
Quoted 4 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM

Following from what Andrew has said (including him mentioning your changing of certain Kconfig option defaults), I'm not convinced this is a good move. This still ought to serve as the overall XSM-yes-or-no setting. If internally you make said two variants match in behavior, that's a different thing.

Apologies that I did not express this clearly. What I was attempting to say is the fact of the matter is that there is no logical behavior difference between "XSM no" and "XSM yes with dummy policy". The only difference is the mechanics of how the dummy functions get called. Specifically via macro magic the dummy functions are either flipped into static inline declarations that are then included into the code where they are invoked or the macro magic has them ending up in the dummy.c XSM module where they are wrapped in macro generated functions that are set as the functions in the dummy xsm_ops structure. Thus it is always the same logic being invoked, it is just mechanics of how you get to the logic.

Quoted 36 lines
Quoted 1 line

* introduces CONFIG_XSM_EVTCHN_LABELING as replacement for enabling event channel labels

Is this mode needed as separate functionality at all? Nothing defines XSM_NEED_GENERIC_EVTCHN_SSID anywhere. _If_ XSM went away as a separate setting, then imo this one should go away as well.

Quoted 30 lines
xen/common/Kconfig +13 -10
@@ -197,22 +197,33 @@ config XENOPROF
 
 	  If unsure, say Y.
 
-config XSM
-	bool "Xen Security Modules support"
-	default ARM
-	---help---
-	  Enables the security framework known as Xen Security Modules which
-	  allows administrators fine-grained control over a Xen domain and
-	  its capabilities by defining permissible interactions between domains,
-	  the hypervisor itself, and related resources such as memory and
-	  devices.
+menu "Xen Security Modules"
 
-	  If unsure, say N.
+choice
+	prompt "Default XSM module"
+	default XSM_SILO_DEFAULT if XSM_SILO && ARM
+	default XSM_FLASK_DEFAULT if XSM_FLASK
+	default XSM_SILO_DEFAULT if XSM_SILO
+	default XSM_DUMMY_DEFAULT
+	config XSM_DUMMY_DEFAULT
+		bool "Match non-XSM behavior"
+	config XSM_FLASK_DEFAULT
+		bool "FLux Advanced Security Kernel" if XSM_FLASK
+	config XSM_SILO_DEFAULT
+		bool "SILO" if XSM_SILO
Quoted 8 lines

xen/common/Kconfig config XENOPROF · line 214

Quoted 1 line
xen/common/Kconfig +1 -0
+endchoice

This did live after the individual options it depends on for a reason, and you don't say anywhere why you need to move it up. The way you have it, with the default command line kconfig tool, users will be presented with dependent options before having chosen the settings of the dependency ones. That's because this tool, to a degree, moves linearly through the options it has parsed.

Yes, this is specifically why I moved it up. Clearly we have different approaches to how we like to interact with configurations, which is not bad thing. I personally found it awkward the other way but can easily move it back.

Quoted 20 lines
Quoted 20 lines
xen/common/Kconfig +1 -14
@@ -261,25 +271,12 @@ config XSM_SILO
 
 	  If unsure, say Y.
 
-choice
-	prompt "Default XSM implementation"
-	depends on XSM
-	default XSM_SILO_DEFAULT if XSM_SILO && ARM
-	default XSM_FLASK_DEFAULT if XSM_FLASK
-	default XSM_SILO_DEFAULT if XSM_SILO
-	default XSM_DUMMY_DEFAULT
-	config XSM_DUMMY_DEFAULT
-		bool "Match non-XSM behavior"
-	config XSM_FLASK_DEFAULT
-		bool "FLux Advanced Security Kernel" if XSM_FLASK
-	config XSM_SILO_DEFAULT
-		bool "SILO" if XSM_SILO
-endchoice
+endmenu
 
Quoted 12 lines

xen/common/Kconfig config XSM_SILO · line 279

Quoted 5 lines
xen/common/Kconfig +1 -1
 config LATE_HWDOM
 	bool "Dedicated hardware domain"
 	default n
-	depends on XSM && X86
+	depends on XSM_FLASK && X86

I don't think this is a compatible change. I'm not going to exclude that this is how it was meant, but as it stands LATE_HWDOM right now doesn't really require FLASK, and could e.g. also go with SILO or dummy. If you _mean_ to change this, then your description needs to say so (and ideally it would then be split out, so - if this is actually a bug - it could also be backported).

Actually this is the root cause that started all of this work. If you want to get technical, LATE_HWDOM does not rely on XSM at all. The issue is that you cannot use it as it was originally intended, to run Xen without a classic dom0 while still having all existing capabilities. Specifically the hardware domain does not have the ability to assign the pass-through devices for which it is in control. This is were Flask comes in to enable assigning specific privileges to labels and then constructing domains with those labels, In particular it grants the ability to do pass-through assignment to the label assigned to the hardware domain. With the upcoming XSM-Roles patch set, these privileges are assigned to roles and it becomes possible to assign the necessary roles to the hardware domain.

v/r,
dps

#17 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 13 lines · Daniel P. Smith
Quoted 11 lines · Jan Beulich
Quoted 4 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM

Following from what Andrew has said (including him mentioning your changing of certain Kconfig option defaults), I'm not convinced this is a good move. This still ought to serve as the overall XSM-yes-or-no setting. If internally you make said two variants match in behavior, that's a different thing.

Quoted 11 lines

Apologies that I did not express this clearly. What I was attempting to say is the fact of the matter is that there is no logical behavior difference between "XSM no" and "XSM yes with dummy policy". The only difference is the mechanics of how the dummy functions get called. Specifically via macro magic the dummy functions are either flipped into static inline declarations that are then included into the code where they are invoked or the macro magic has them ending up in the dummy.c XSM module where they are wrapped in macro generated functions that are set as the functions in the dummy xsm_ops structure. Thus it is always the same logic being invoked, it is just mechanics of how you get to the logic.

That's what I understood, really. What I dislike is the inline functions going away in what we currently call !XSM.

Quoted 45 lines
Quoted 36 lines
Quoted 1 line

* introduces CONFIG_XSM_EVTCHN_LABELING as replacement for enabling event channel labels

Is this mode needed as separate functionality at all? Nothing defines XSM_NEED_GENERIC_EVTCHN_SSID anywhere. _If_ XSM went away as a separate setting, then imo this one should go away as well.

Quoted 30 lines
xen/common/Kconfig +13 -10
@@ -197,22 +197,33 @@ config XENOPROF
 
 	  If unsure, say Y.
 
-config XSM
-	bool "Xen Security Modules support"
-	default ARM
-	---help---
-	  Enables the security framework known as Xen Security Modules which
-	  allows administrators fine-grained control over a Xen domain and
-	  its capabilities by defining permissible interactions between domains,
-	  the hypervisor itself, and related resources such as memory and
-	  devices.
+menu "Xen Security Modules"
 
-	  If unsure, say N.
+choice
+	prompt "Default XSM module"
+	default XSM_SILO_DEFAULT if XSM_SILO && ARM
+	default XSM_FLASK_DEFAULT if XSM_FLASK
+	default XSM_SILO_DEFAULT if XSM_SILO
+	default XSM_DUMMY_DEFAULT
+	config XSM_DUMMY_DEFAULT
+		bool "Match non-XSM behavior"
+	config XSM_FLASK_DEFAULT
+		bool "FLux Advanced Security Kernel" if XSM_FLASK
+	config XSM_SILO_DEFAULT
+		bool "SILO" if XSM_SILO
Quoted 8 lines

xen/common/Kconfig config XENOPROF · line 214

Quoted 1 line
xen/common/Kconfig +1 -0
+endchoice

This did live after the individual options it depends on for a reason, and you don't say anywhere why you need to move it up. The way you have it, with the default command line kconfig tool, users will be presented with dependent options before having chosen the settings of the dependency ones. That's because this tool, to a degree, moves linearly through the options it has parsed.

Quoted 4 lines

Yes, this is specifically why I moved it up. Clearly we have different approaches to how we like to interact with configurations, which is not bad thing. I personally found it awkward the other way but can easily move it back.

I'm having a hard time seeing how presenting dependent options ahead of their dependencies can be a good thing: The user will have made their "choice" here, while the availability of the individual settings then may change in case the depended upon options change from their defaults upon the user reacting to those prompts.

Quoted 33 lines
Quoted 20 lines
Quoted 20 lines
xen/common/Kconfig +1 -14
@@ -261,25 +271,12 @@ config XSM_SILO
 
 	  If unsure, say Y.
 
-choice
-	prompt "Default XSM implementation"
-	depends on XSM
-	default XSM_SILO_DEFAULT if XSM_SILO && ARM
-	default XSM_FLASK_DEFAULT if XSM_FLASK
-	default XSM_SILO_DEFAULT if XSM_SILO
-	default XSM_DUMMY_DEFAULT
-	config XSM_DUMMY_DEFAULT
-		bool "Match non-XSM behavior"
-	config XSM_FLASK_DEFAULT
-		bool "FLux Advanced Security Kernel" if XSM_FLASK
-	config XSM_SILO_DEFAULT
-		bool "SILO" if XSM_SILO
-endchoice
+endmenu
 
Quoted 12 lines

xen/common/Kconfig config XSM_SILO · line 279

Quoted 5 lines
xen/common/Kconfig +1 -1
 config LATE_HWDOM
 	bool "Dedicated hardware domain"
 	default n
-	depends on XSM && X86
+	depends on XSM_FLASK && X86

I don't think this is a compatible change. I'm not going to exclude that this is how it was meant, but as it stands LATE_HWDOM right now doesn't really require FLASK, and could e.g. also go with SILO or dummy. If you _mean_ to change this, then your description needs to say so (and ideally it would then be split out, so - if this is actually a bug - it could also be backported).

Quoted 12 lines

Actually this is the root cause that started all of this work. If you want to get technical, LATE_HWDOM does not rely on XSM at all. The issue is that you cannot use it as it was originally intended, to run Xen without a classic dom0 while still having all existing capabilities. Specifically the hardware domain does not have the ability to assign the pass-through devices for which it is in control. This is were Flask comes in to enable assigning specific privileges to labels and then constructing domains with those labels, In particular it grants the ability to do pass-through assignment to the label assigned to the hardware domain. With the upcoming XSM-Roles patch set, these privileges are assigned to roles and it becomes possible to assign the necessary roles to the hardware domain.

In which case this needs spelling out in the description, to justify the change in behavior (which I'm not sure I really follow / agree with just yet).

Jan

#18 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 23 lines · Jan Beulich
Quoted 11 lines · Daniel P. Smith
Quoted 10 lines · Jan Beulich
Quoted 4 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM

Following from what Andrew has said (including him mentioning your changing of certain Kconfig option defaults), I'm not convinced this is a good move. This still ought to serve as the overall XSM-yes-or-no setting. If internally you make said two variants match in behavior, that's a different thing.

Quoted 11 lines

Apologies that I did not express this clearly. What I was attempting to say is the fact of the matter is that there is no logical behavior difference between "XSM no" and "XSM yes with dummy policy". The only difference is the mechanics of how the dummy functions get called. Specifically via macro magic the dummy functions are either flipped into static inline declarations that are then included into the code where they are invoked or the macro magic has them ending up in the dummy.c XSM module where they are wrapped in macro generated functions that are set as the functions in the dummy xsm_ops structure. Thus it is always the same logic being invoked, it is just mechanics of how you get to the logic.

Quoted 2 lines

That's what I understood, really. What I dislike is the inline functions going away in what we currently call !XSM.

I'm sorry, but this is an unreasonable objection.

The mess used to create the status quo *is* the majority reason why fixing/developing XSM is so hard, and why the code is so obfuscated.  To prove this point, how many people on this email thread realise that calls using XSM_HOOK offer 0 security under xsm_default_action()?

Having xsm_default_action() forced inline isn't obviously the right move in the first place, and I doubt that you could even measure a performance difference for using real function calls.

Even if there is a marginal performance difference, and I doubt that there is, performance is far less important than de-obfuscating the code and fixing our various security mechanisms to be first-class supported citizens.

~Andrew

#19 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 34 lines · Andrew Cooper
Quoted 23 lines · Jan Beulich
Quoted 11 lines · Daniel P. Smith
Quoted 10 lines · Jan Beulich
Quoted 4 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM

Following from what Andrew has said (including him mentioning your changing of certain Kconfig option defaults), I'm not convinced this is a good move. This still ought to serve as the overall XSM-yes-or-no setting. If internally you make said two variants match in behavior, that's a different thing.

Quoted 11 lines

Apologies that I did not express this clearly. What I was attempting to say is the fact of the matter is that there is no logical behavior difference between "XSM no" and "XSM yes with dummy policy". The only difference is the mechanics of how the dummy functions get called. Specifically via macro magic the dummy functions are either flipped into static inline declarations that are then included into the code where they are invoked or the macro magic has them ending up in the dummy.c XSM module where they are wrapped in macro generated functions that are set as the functions in the dummy xsm_ops structure. Thus it is always the same logic being invoked, it is just mechanics of how you get to the logic.

Quoted 2 lines

That's what I understood, really. What I dislike is the inline functions going away in what we currently call !XSM.

I'm sorry, but this is an unreasonable objection.

The mess used to create the status quo *is* the majority reason why fixing/developing XSM is so hard, and why the code is so obfuscated.  To prove this point, how many people on this email thread realise that calls using XSM_HOOK offer 0 security under xsm_default_action()?

Quoted 8 lines

Having xsm_default_action() forced inline isn't obviously the right move in the first place, and I doubt that you could even measure a performance difference for using real function calls.

Even if there is a marginal performance difference, and I doubt that there is, performance is far less important than de-obfuscating the code and fixing our various security mechanisms to be first-class supported citizens.

What I don't understand from all you say is why you think that having an as-if-no-XSM build configuration, without any way to switch to an alternative model (i.e. the XSM=n that we have right now), is a bad thing. I don't mind the XSM=y case getting improved, but I don't see (yet) why it is a good thing to force this onto everyone.

Jan

#20 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

In reply to Jan Beulich
Quoted 10 lines · Jan Beulich
Quoted 4 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM

Following from what Andrew has said (including him mentioning your changing of certain Kconfig option defaults), I'm not convinced this is a good move. This still ought to serve as the overall XSM-yes-or-no setting. If internally you make said two variants match in behavior, that's a different thing.

I firmly disagree. There is no such thing as !XSM even in staging right now.

All over Xen, we have calls to xsm_*() functions which, even in the !XSM case, contain a non-trivial security policy.

The fact that under the hood, XSM vs !XSM creates two very different implementations of "the dom0-all-powerful model" is an error needing correcting, as it contributes a massive quantity of code complexity.

This series of Daniel's takes steps to make the code match reality, and getting rid of CONFIG_XSM is absolutely the right thing to do.  XSM is never actually absent from a build of Xen, even if you choose CONFIG_XSM=n.

I do think that the thing we currently call XSM_DUMMY wants renaming to indicate that it is "the dom0-all-powerful" security model, and I think that wants doing as part of this series.  Name suggestions welcome.

~Andrew

#21 lore raw

Re: [PATCH 3/6] xsm: enabling xsm to always be included

Quoted 12 lines · Andrew Cooper
Quoted 10 lines · Jan Beulich
Quoted 4 lines · Daniel P. Smith

The only difference between !CONFIG_XSM and CONFIG_XSM with !CONFIG_XSM_SILO and !CONFIG_XSM_FLASK is whether the XSM hooks in dummy.h are called as static inline functions or as function pointers to static functions. As such this commit,
* eliminates CONFIG_XSM

Following from what Andrew has said (including him mentioning your changing of certain Kconfig option defaults), I'm not convinced this is a good move. This still ought to serve as the overall XSM-yes-or-no setting. If internally you make said two variants match in behavior, that's a different thing.

Quoted 4 lines

I firmly disagree. There is no such thing as !XSM even in staging right now.

All over Xen, we have calls to xsm_*() functions which, even in the !XSM case, contain a non-trivial security policy.

Compared with the full-fledged XSM, I view the present xsm_default_action() as sufficiently trivial. The inline wrappers of it really only exist to allow #ifdef-ary at all the use sites to be avoided, and for the code to act like before XSM got introduced. Whether you call this !XSM or XSM_HWDOM_ALL_POWERFUL is secondary to me.

Quoted 7 lines

The fact that under the hood, XSM vs !XSM creates two very different implementations of "the dom0-all-powerful model" is an error needing correcting, as it contributes a massive quantity of code complexity.

This series of Daniel's takes steps to make the code match reality, and getting rid of CONFIG_XSM is absolutely the right thing to do.  XSM is never actually absent from a build of Xen, even if you choose CONFIG_XSM=n.

As said, what you discuss is just how to call the child. What I point out as undesirable is the going away of the inline functions, replaced by real function calls (not indirect ones thanks to alternatives patching, but still not clearly on par with the current model in terms of overhead).

Jan

#22 lore raw

PATCH 4/6 xsm: remove xen_defualt_t from hook definitions

In reply to Daniel P. Smith

With the conversion of making XSM always enabled even the dummy XSM module is being invoked through the xsm_ops dispatch which does not use passing of the default privilege. This commit removes the xen_default_t parameter from the hook definitions and all the respective call sites.

---
 xen/arch/arm/dm.c                     |   2 +-
 xen/arch/arm/domctl.c                 |   6 +-
 xen/arch/arm/hvm.c                    |   2 +-
 xen/arch/arm/mm.c                     |   2 +-
 xen/arch/arm/platform_hypercall.c     |   2 +-
 xen/arch/x86/cpu/mcheck/mce.c         |   2 +-
 xen/arch/x86/cpu/vpmu.c               |   2 +-
 xen/arch/x86/domctl.c                 |   8 +-
 xen/arch/x86/hvm/dm.c                 |   2 +-
 xen/arch/x86/hvm/hvm.c                |  12 +-
 xen/arch/x86/irq.c                    |   5 +-
 xen/arch/x86/mm.c                     |  20 +--
 xen/arch/x86/mm/mem_paging.c          |   2 +-
 xen/arch/x86/mm/mem_sharing.c         |   9 +-
 xen/arch/x86/mm/p2m.c                 |   2 +-
 xen/arch/x86/mm/paging.c              |   4 +-
 xen/arch/x86/mm/shadow/set.c          |   2 +-
 xen/arch/x86/msi.c                    |   3 +-
 xen/arch/x86/pci.c                    |   2 +-
 xen/arch/x86/physdev.c                |  17 ++-
 xen/arch/x86/platform_hypercall.c     |  10 +-
 xen/arch/x86/pv/emul-priv-op.c        |   2 +-
 xen/arch/x86/sysctl.c                 |   4 +-
 xen/common/domain.c                   |   4 +-
 xen/common/domctl.c                   |  12 +-
 xen/common/event_channel.c            |  12 +-
 xen/common/grant_table.c              |  16 +--
 xen/common/hypfs.c                    |   2 +-
 xen/common/kernel.c                   |   2 +-
 xen/common/kexec.c                    |   2 +-
 xen/common/mem_access.c               |   2 +-
 xen/common/memory.c                   |  16 +--
 xen/common/monitor.c                  |   2 +-
 xen/common/sched/core.c               |   6 +-
 xen/common/sysctl.c                   |   8 +-
 xen/common/vm_event.c                 |   2 +-
 xen/common/xenoprof.c                 |   2 +-
 xen/drivers/char/console.c            |   2 +-
 xen/drivers/passthrough/device_tree.c |   4 +-
 xen/drivers/passthrough/pci.c         |  12 +-
 xen/include/xsm/xsm.h                 | 172 +++++++++++++-------------
 41 files changed, 197 insertions(+), 203 deletions(-)
xen/arch/arm/dm.c +1 -1
index 1b3fd6bc7d..c8b89c8f47 100644
@@ -45,7 +45,7 @@ int dm_op(const struct dmop_args *op_args)
     if ( rc )
         return rc;
 
-    rc = xsm_dm_op(XSM_DM_PRIV, d);
+    rc = xsm_dm_op(d);
     if ( rc )
         goto out;
 
xen/arch/arm/domctl.c +3 -3
index b7d27f37df..e7202703ee 100644
@@ -95,11 +95,11 @@ long arch_do_domctl(struct xen_domctl *domctl, struct domain *d,
          * done by the 2 hypercalls for consistency with other
          * architectures.
          */
-        rc = xsm_map_domain_irq(XSM_HOOK, d, irq, NULL);
+        rc = xsm_map_domain_irq(d, irq, NULL);
         if ( rc )
             return rc;
 
-        rc = xsm_bind_pt_irq(XSM_HOOK, d, bind);
+        rc = xsm_bind_pt_irq(d, bind);
         if ( rc )
             return rc;
 
@@ -130,7 +130,7 @@ long arch_do_domctl(struct xen_domctl *domctl, struct domain *d,
         if ( irq != virq )
             return -EINVAL;
 
-        rc = xsm_unbind_pt_irq(XSM_HOOK, d, bind);
+        rc = xsm_unbind_pt_irq(d, bind);
         if ( rc )
             return rc;
 
xen/arch/arm/hvm.c +1 -1
index 8951b34086..cf4bd9ae09 100644
@@ -101,7 +101,7 @@ long do_hvm_op(unsigned long op, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( d == NULL )
             return -ESRCH;
 
-        rc = xsm_hvm_param(XSM_TARGET, d, op);
+        rc = xsm_hvm_param(d, op);
         if ( rc )
             goto param_fail;
 
xen/arch/arm/mm.c +1 -1
index 0e07335291..a256c89b62 100644
@@ -1446,7 +1446,7 @@ int xenmem_add_to_physmap_one(
             return -EINVAL;
         }
 
-        rc = xsm_map_gmfn_foreign(XSM_TARGET, d, od);
+        rc = xsm_map_gmfn_foreign(d, od);
         if ( rc )
         {
             put_pg_owner(od);
xen/arch/arm/platform_hypercall.c +1 -1
index 8efac7ee60..52985f8a51 100644
@@ -33,7 +33,7 @@ long do_platform_op(XEN_GUEST_HANDLE_PARAM(xen_platform_op_t) u_xenpf_op)
     if ( d == NULL )
         return -ESRCH;
 
-    ret = xsm_platform_op(XSM_PRIV, op->cmd);
+    ret = xsm_platform_op(op->cmd);
     if ( ret )
         return ret;
 
xen/arch/x86/cpu/mcheck/mce.c +1 -1
index 7f433343bc..58beb96663 100644
@@ -1376,7 +1376,7 @@ long do_mca(XEN_GUEST_HANDLE_PARAM(xen_mc_t) u_xen_mc)
     struct xen_mc_msrinject *mc_msrinject;
     struct xen_mc_mceinject *mc_mceinject;
 
-    ret = xsm_do_mca(XSM_PRIV);
+    ret = xsm_do_mca();
     if ( ret )
         return x86_mcerr("", ret);
 
xen/arch/x86/cpu/vpmu.c +1 -1
index 16e91a3694..34b536417b 100644
@@ -706,7 +706,7 @@ long do_xenpmu_op(unsigned int op, XEN_GUEST_HANDLE_PARAM(xen_pmu_params_t) arg)
     if ( !opt_vpmu_enabled || has_vlapic(current->domain) )
         return -EOPNOTSUPP;
 
-    ret = xsm_pmu_op(XSM_OTHER, current->domain, op);
+    ret = xsm_pmu_op(current->domain, op);
     if ( ret )
         return ret;
 
xen/arch/x86/domctl.c +4 -4
index 26a76d2be9..8343c59e83 100644
@@ -234,7 +234,7 @@ long arch_do_domctl(
         if ( (fp + np) <= fp || (fp + np) > MAX_IOPORTS )
             ret = -EINVAL;
         else if ( !ioports_access_permitted(currd, fp, fp + np - 1) ||
-                  xsm_ioport_permission(XSM_HOOK, d, fp, fp + np - 1, allow) )
+                  xsm_ioport_permission(d, fp, fp + np - 1, allow) )
             ret = -EPERM;
         else if ( allow )
             ret = ioports_permit_access(d, fp, fp + np - 1);
@@ -534,7 +534,7 @@ long arch_do_domctl(
         if ( !is_hvm_domain(d) )
             break;
 
-        ret = xsm_bind_pt_irq(XSM_HOOK, d, bind);
+        ret = xsm_bind_pt_irq(d, bind);
         if ( ret )
             break;
 
@@ -569,7 +569,7 @@ long arch_do_domctl(
         if ( irq <= 0 || !irq_access_permitted(currd, irq) )
             break;
 
-        ret = xsm_unbind_pt_irq(XSM_HOOK, d, bind);
+        ret = xsm_unbind_pt_irq(d, bind);
         if ( ret )
             break;
 
@@ -616,7 +616,7 @@ long arch_do_domctl(
         if ( !ioports_access_permitted(currd, fmp, fmp + np - 1) )
             break;
 
-        ret = xsm_ioport_mapping(XSM_HOOK, d, fmp, fmp + np - 1, add);
+        ret = xsm_ioport_mapping(d, fmp, fmp + np - 1, add);
         if ( ret )
             break;
 
xen/arch/x86/hvm/dm.c +1 -1
index b60b9f3364..6f996371b9 100644
@@ -370,7 +370,7 @@ int dm_op(const struct dmop_args *op_args)
     if ( !is_hvm_domain(d) )
         goto out;
 
-    rc = xsm_dm_op(XSM_DM_PRIV, d);
+    rc = xsm_dm_op(d);
     if ( rc )
         goto out;
 
xen/arch/x86/hvm/hvm.c +6 -6
index 5086773e5c..c1b0d6fca8 100644
@@ -4080,7 +4080,7 @@ static int hvm_allow_set_param(struct domain *d,
     uint64_t value;
     int rc;
 
-    rc = xsm_hvm_param(XSM_TARGET, d, HVMOP_set_param);
+    rc = xsm_hvm_param(d, HVMOP_set_param);
     if ( rc )
         return rc;
 
@@ -4227,7 +4227,7 @@ static int hvm_set_param(struct domain *d, uint32_t index, uint64_t value)
         rc = pmtimer_change_ioport(d, value);
         break;
     case HVM_PARAM_ALTP2M:
-        rc = xsm_hvm_param_altp2mhvm(XSM_PRIV, d);
+        rc = xsm_hvm_param_altp2mhvm(d);
         if ( rc )
             break;
         if ( (value > XEN_ALTP2M_limited) ||
@@ -4356,7 +4356,7 @@ static int hvm_allow_get_param(struct domain *d,
 {
     int rc;
 
-    rc = xsm_hvm_param(XSM_TARGET, d, HVMOP_get_param);
+    rc = xsm_hvm_param(d, HVMOP_get_param);
     if ( rc )
         return rc;
 
@@ -4566,7 +4566,7 @@ static int do_altp2m_op(
         goto out;
     }
 
-    if ( (rc = xsm_hvm_altp2mhvm_op(XSM_OTHER, d, mode, a.cmd)) )
+    if ( (rc = xsm_hvm_altp2mhvm_op(d, mode, a.cmd)) )
         goto out;
 
     switch ( a.cmd )
@@ -4947,7 +4947,7 @@ static int hvmop_get_mem_type(
     if ( d == NULL )
         return -ESRCH;
 
-    rc = xsm_hvm_param(XSM_TARGET, d, HVMOP_get_mem_type);
+    rc = xsm_hvm_param(d, HVMOP_get_mem_type);
     if ( rc )
         goto out;
 
@@ -5040,7 +5040,7 @@ long do_hvm_op(unsigned long op, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( unlikely(d != current->domain) )
             rc = -EOPNOTSUPP;
         else if ( is_hvm_domain(d) && paging_mode_shadow(d) )
-            rc = xsm_hvm_param(XSM_TARGET, d, op);
+            rc = xsm_hvm_param(d, op);
         if ( !rc )
             pagetable_dying(a.gpa);
 
xen/arch/x86/irq.c +2 -3
index a1693f92dd..9f79ec1b86 100644
@@ -2122,7 +2122,7 @@ int map_domain_pirq(
         return 0;
     }
 
-    ret = xsm_map_domain_irq(XSM_HOOK, d, irq, data);
+    ret = xsm_map_domain_irq(d, irq, data);
     if ( ret )
     {
         dprintk(XENLOG_G_ERR, "dom%d: could not permit access to irq %d mapping to pirq %d\n",
@@ -2342,8 +2342,7 @@ int unmap_domain_pirq(struct domain *d, int pirq)
         nr = msi_desc->msi.nvec;
     }
 
-    ret = xsm_unmap_domain_irq(XSM_HOOK, d, irq,
-                               msi_desc ? msi_desc->dev : NULL);
+    ret = xsm_unmap_domain_irq(d, irq, msi_desc ? msi_desc->dev : NULL);
     if ( ret )
         goto done;
 
xen/arch/x86/mm.c +10 -10
index 4d799032dc..33d0aa8d4b 100644
@@ -977,7 +977,7 @@ get_page_from_l1e(
          * minor hack can go away.
          */
         if ( (real_pg_owner == NULL) || (pg_owner == l1e_owner) ||
-             xsm_priv_mapping(XSM_TARGET, pg_owner, real_pg_owner) )
+             xsm_priv_mapping(pg_owner, real_pg_owner) )
         {
             gdprintk(XENLOG_WARNING,
                      "pg_owner d%d l1e_owner d%d, but real_pg_owner d%d\n",
@@ -3407,7 +3407,7 @@ long do_mmuext_op(
         return -EINVAL;
     }
 
-    rc = xsm_mmuext_op(XSM_TARGET, currd, pg_owner);
+    rc = xsm_mmuext_op(currd, pg_owner);
     if ( rc )
     {
         put_pg_owner(pg_owner);
@@ -3497,7 +3497,7 @@ long do_mmuext_op(
                 break;
             }
 
-            rc = xsm_memory_pin_page(XSM_HOOK, currd, pg_owner, page);
+            rc = xsm_memory_pin_page(currd, pg_owner, page);
             if ( !rc && unlikely(test_and_set_bit(_PGT_pinned,
                                                   &page->u.inuse.type_info)) )
             {
@@ -4006,7 +4006,7 @@ long do_mmu_update(
             }
             if ( xsm_needed != xsm_checked )
             {
-                rc = xsm_mmu_update(XSM_TARGET, d, pt_owner, pg_owner, xsm_needed);
+                rc = xsm_mmu_update(d, pt_owner, pg_owner, xsm_needed);
                 if ( rc )
                     break;
                 xsm_checked = xsm_needed;
@@ -4142,7 +4142,7 @@ long do_mmu_update(
             xsm_needed |= XSM_MMU_MACHPHYS_UPDATE;
             if ( xsm_needed != xsm_checked )
             {
-                rc = xsm_mmu_update(XSM_TARGET, d, NULL, pg_owner, xsm_needed);
+                rc = xsm_mmu_update(d, NULL, pg_owner, xsm_needed);
                 if ( rc )
                     break;
                 xsm_checked = xsm_needed;
@@ -4391,7 +4391,7 @@ static int __do_update_va_mapping(
 
     perfc_incr(calls_to_update_va);
 
-    rc = xsm_update_va_mapping(XSM_TARGET, d, pg_owner, val);
+    rc = xsm_update_va_mapping(d, pg_owner, val);
     if ( rc )
         return rc;
 
@@ -4630,7 +4630,7 @@ long arch_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( d == NULL )
             return -ESRCH;
 
-        rc = xsm_domain_memory_map(XSM_TARGET, d);
+        rc = xsm_domain_memory_map(d);
         if ( rc )
         {
             rcu_unlock_domain(d);
@@ -4697,7 +4697,7 @@ long arch_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         unsigned int i;
         bool store;
 
-        rc = xsm_machine_memory_map(XSM_PRIV);
+        rc = xsm_machine_memory_map();
         if ( rc )
             return rc;
 
@@ -4787,9 +4787,9 @@ long arch_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
             return -ESRCH;
 
         if ( cmd == XENMEM_set_pod_target )
-            rc = xsm_set_pod_target(XSM_PRIV, d);
+            rc = xsm_set_pod_target(d);
         else
-            rc = xsm_get_pod_target(XSM_PRIV, d);
+            rc = xsm_get_pod_target(d);
 
         if ( rc != 0 )
             goto pod_target_out_unlock;
xen/arch/x86/mm/mem_paging.c +1 -1
index 01281f786e..073edb9499 100644
@@ -452,7 +452,7 @@ int mem_paging_memop(XEN_GUEST_HANDLE_PARAM(xen_mem_paging_op_t) arg)
     if ( rc )
         return rc;
 
-    rc = xsm_mem_paging(XSM_DM_PRIV, d);
+    rc = xsm_mem_paging(d);
     if ( rc )
         goto out;
 
xen/arch/x86/mm/mem_sharing.c +4 -5
index 98b14f7b0a..db5f5ce8b5 100644
@@ -1883,7 +1883,7 @@ int mem_sharing_memop(XEN_GUEST_HANDLE_PARAM(xen_mem_sharing_op_t) arg)
     if ( rc )
         return rc;
 
-    rc = xsm_mem_sharing(XSM_DM_PRIV, d);
+    rc = xsm_mem_sharing(d);
     if ( rc )
         goto out;
 
@@ -1928,7 +1928,7 @@ int mem_sharing_memop(XEN_GUEST_HANDLE_PARAM(xen_mem_sharing_op_t) arg)
         if ( rc )
             goto out;
 
-        rc = xsm_mem_sharing_op(XSM_DM_PRIV, d, cd, mso.op);
+        rc = xsm_mem_sharing_op(d, cd, mso.op);
         if ( rc )
         {
             rcu_unlock_domain(cd);
@@ -1994,7 +1994,7 @@ int mem_sharing_memop(XEN_GUEST_HANDLE_PARAM(xen_mem_sharing_op_t) arg)
         if ( rc )
             goto out;
 
-        rc = xsm_mem_sharing_op(XSM_DM_PRIV, d, cd, mso.op);
+        rc = xsm_mem_sharing_op(d, cd, mso.op);
         if ( rc )
         {
             rcu_unlock_domain(cd);
@@ -2056,8 +2056,7 @@ int mem_sharing_memop(XEN_GUEST_HANDLE_PARAM(xen_mem_sharing_op_t) arg)
          * We reuse XENMEM_sharing_op_share XSM check here as this is
          * essentially the same concept repeated over multiple pages.
          */
-        rc = xsm_mem_sharing_op(XSM_DM_PRIV, d, cd,
-                                XENMEM_sharing_op_share);
+        rc = xsm_mem_sharing_op(d, cd, XENMEM_sharing_op_share);
         if ( rc )
         {
             rcu_unlock_domain(cd);
xen/arch/x86/mm/p2m.c +1 -1
index dbb1cbeb59..5cee9f5a1d 100644
@@ -2637,7 +2637,7 @@ static int p2m_add_foreign(struct domain *tdom, unsigned long fgfn,
             goto out;
     }
 
-    rc = xsm_map_gmfn_foreign(XSM_TARGET, tdom, fdom);
+    rc = xsm_map_gmfn_foreign(tdom, fdom);
     if ( rc )
         goto out;
 
xen/arch/x86/mm/paging.c +2 -2
index c304c24526..86a1ec5b80 100644
@@ -714,7 +714,7 @@ int paging_domctl(struct domain *d, struct xen_domctl_shadow_op *sc,
         return -EBUSY;
     }
 
-    rc = xsm_shadow_control(XSM_HOOK, d, sc->op);
+    rc = xsm_shadow_control(d, sc->op);
     if ( rc )
         return rc;
 
@@ -771,7 +771,7 @@ long paging_domctl_continuation(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
     if ( d == NULL )
         return -ESRCH;
 
-    ret = xsm_domctl(XSM_OTHER, d, op.cmd);
+    ret = xsm_domctl(d, op.cmd);
     if ( !ret )
     {
         if ( domctl_lock_acquire() )
xen/arch/x86/mm/shadow/set.c +1 -1
index 87e9c6eeb2..e7433ac81b 100644
@@ -117,7 +117,7 @@ shadow_get_page_from_l1e(shadow_l1e_t sl1e, struct domain *d, p2m_type_t type)
      * not own, we let it succeed anyway.
      */
     if ( owner && (d != owner) &&
-         !(res = xsm_priv_mapping(XSM_TARGET, d, owner)) )
+         !(res = xsm_priv_mapping(d, owner)) )
     {
         res = get_page_from_l1e(sl1e, d, owner);
         SHADOW_PRINTK("privileged %pd installs map of %pd's mfn %"PRI_mfn": %s\n",
xen/arch/x86/msi.c +1 -2
index 5febc0ea4b..f821d145e5 100644
@@ -1310,8 +1310,7 @@ int pci_restore_msi_state(struct pci_dev *pdev)
     if ( !use_msi )
         return -EOPNOTSUPP;
 
-    ret = xsm_resource_setup_pci(XSM_PRIV,
-                                (pdev->seg << 16) | (pdev->bus << 8) |
+    ret = xsm_resource_setup_pci((pdev->seg << 16) | (pdev->bus << 8) |
                                 pdev->devfn);
     if ( ret )
         return ret;
xen/arch/x86/pci.c +1 -1
index a9decd4f33..d5886c59e6 100644
@@ -74,7 +74,7 @@ int pci_conf_write_intercept(unsigned int seg, unsigned int bdf,
                              uint32_t *data)
 {
     struct pci_dev *pdev;
-    int rc = xsm_pci_config_permission(XSM_HOOK, current->domain, bdf,
+    int rc = xsm_pci_config_permission(current->domain, bdf,
                                        reg, reg + size - 1, 1);
 
     if ( rc < 0 )
xen/arch/x86/physdev.c +8 -9
index 23465bcd00..3f2a2035c5 100644
@@ -110,7 +110,7 @@ int physdev_map_pirq(domid_t domid, int type, int *index, int *pirq_p,
     if ( d == NULL )
         return -ESRCH;
 
-    ret = xsm_map_domain_pirq(XSM_DM_PRIV, d);
+    ret = xsm_map_domain_pirq(d);
     if ( ret )
         goto free_domain;
 
@@ -148,7 +148,7 @@ int physdev_unmap_pirq(domid_t domid, int pirq)
         return -ESRCH;
 
     if ( domid != DOMID_SELF || !is_hvm_domain(d) || !has_pirq(d) )
-        ret = xsm_unmap_domain_pirq(XSM_DM_PRIV, d);
+        ret = xsm_unmap_domain_pirq(d);
     if ( ret )
         goto free_domain;
 
@@ -355,7 +355,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         ret = -EFAULT;
         if ( copy_from_guest(&apic, arg, 1) != 0 )
             break;
-        ret = xsm_apic(XSM_PRIV, currd, cmd);
+        ret = xsm_apic(currd, cmd);
         if ( ret )
             break;
         ret = ioapic_guest_read(apic.apic_physbase, apic.reg, &apic.value);
@@ -369,7 +369,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         ret = -EFAULT;
         if ( copy_from_guest(&apic, arg, 1) != 0 )
             break;
-        ret = xsm_apic(XSM_PRIV, currd, cmd);
+        ret = xsm_apic(currd, cmd);
         if ( ret )
             break;
         ret = ioapic_guest_write(apic.apic_physbase, apic.reg, apic.value);
@@ -385,7 +385,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
 
         /* Use the APIC check since this dummy hypercall should still only
          * be called by the domain with access to program the ioapic */
-        ret = xsm_apic(XSM_PRIV, currd, cmd);
+        ret = xsm_apic(currd, cmd);
         if ( ret )
             break;
 
@@ -535,8 +535,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( copy_from_guest(&dev, arg, 1) )
             ret = -EFAULT;
         else
-            ret = xsm_resource_setup_pci(XSM_PRIV,
-                                         (dev.seg << 16) | (dev.bus << 8) |
+            ret = xsm_resource_setup_pci((dev.seg << 16) | (dev.bus << 8) |
                                          dev.devfn) ?:
                   pci_prepare_msix(dev.seg, dev.bus, dev.devfn,
                                    cmd != PHYSDEVOP_prepare_msix);
@@ -546,7 +545,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
     case PHYSDEVOP_pci_mmcfg_reserved: {
         struct physdev_pci_mmcfg_reserved info;
 
-        ret = xsm_resource_setup_misc(XSM_PRIV);
+        ret = xsm_resource_setup_misc();
         if ( ret )
             break;
 
@@ -611,7 +610,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( setup_gsi.gsi < 0 || setup_gsi.gsi >= nr_irqs_gsi )
             break;
 
-        ret = xsm_resource_setup_gsi(XSM_PRIV, setup_gsi.gsi);
+        ret = xsm_resource_setup_gsi(setup_gsi.gsi);
         if ( ret )
             break;
 
xen/arch/x86/platform_hypercall.c +5 -5
index 23fadbc782..afa97f74fd 100644
@@ -196,7 +196,7 @@ ret_t do_platform_op(XEN_GUEST_HANDLE_PARAM(xen_platform_op_t) u_xenpf_op)
     if ( op->interface_version != XENPF_INTERFACE_VERSION )
         return -EACCES;
 
-    ret = xsm_platform_op(XSM_PRIV, op->cmd);
+    ret = xsm_platform_op(op->cmd);
     if ( ret )
         return ret;
 
@@ -614,7 +614,7 @@ ret_t do_platform_op(XEN_GUEST_HANDLE_PARAM(xen_platform_op_t) u_xenpf_op)
     {
         int cpu = op->u.cpu_ol.cpuid;
 
-        ret = xsm_resource_plug_core(XSM_HOOK);
+        ret = xsm_resource_plug_core();
         if ( ret )
             break;
 
@@ -640,7 +640,7 @@ ret_t do_platform_op(XEN_GUEST_HANDLE_PARAM(xen_platform_op_t) u_xenpf_op)
     {
         int cpu = op->u.cpu_ol.cpuid;
 
-        ret = xsm_resource_unplug_core(XSM_HOOK);
+        ret = xsm_resource_unplug_core();
         if ( ret )
             break;
 
@@ -669,7 +669,7 @@ ret_t do_platform_op(XEN_GUEST_HANDLE_PARAM(xen_platform_op_t) u_xenpf_op)
     break;
 
     case XENPF_cpu_hotadd:
-        ret = xsm_resource_plug_core(XSM_HOOK);
+        ret = xsm_resource_plug_core();
         if ( ret )
             break;
 
@@ -679,7 +679,7 @@ ret_t do_platform_op(XEN_GUEST_HANDLE_PARAM(xen_platform_op_t) u_xenpf_op)
     break;
 
     case XENPF_mem_hotadd:
-        ret = xsm_resource_plug_core(XSM_HOOK);
+        ret = xsm_resource_plug_core();
         if ( ret )
             break;
 
xen/arch/x86/pv/emul-priv-op.c +1 -1
index 8889509d2a..d833b8e2e6 100644
@@ -250,7 +250,7 @@ static bool pci_cfg_ok(struct domain *currd, unsigned int start,
     }
 
     return !write ?
-           xsm_pci_config_permission(XSM_HOOK, currd, machine_bdf,
+           xsm_pci_config_permission(currd, machine_bdf,
                                      start, start + size - 1, 0) == 0 :
            pci_conf_write_intercept(0, machine_bdf, start, size, write) >= 0;
 }
xen/arch/x86/sysctl.c +2 -2
index aff52a13f3..975672360b 100644
@@ -190,8 +190,8 @@ long arch_do_sysctl(
         }
 
         if ( !ret )
-            ret = plug ? xsm_resource_plug_core(XSM_HOOK)
-                       : xsm_resource_unplug_core(XSM_HOOK);
+            ret = plug ? xsm_resource_plug_core()
+                       : xsm_resource_unplug_core();
 
         if ( !ret )
             ret = continue_hypercall_on_cpu(0, fn, hcpu);
xen/common/domain.c +2 -2
index 6b71c6d6a9..392865f0f1 100644
@@ -311,7 +311,7 @@ static int late_hwdom_init(struct domain *d)
     if ( d != hardware_domain || d->domain_id == 0 )
         return 0;
 
-    rv = xsm_init_hardware_domain(XSM_HOOK, d);
+    rv = xsm_init_hardware_domain(d);
     if ( rv )
         return rv;
 
@@ -649,7 +649,7 @@ struct domain *domain_create(domid_t domid,
         if ( !d->iomem_caps || !d->irq_caps )
             goto fail;
 
-        if ( (err = xsm_domain_create(XSM_HOOK, d, config->ssidref)) != 0 )
+        if ( (err = xsm_domain_create(d, config->ssidref)) != 0 )
             goto fail;
 
         d->controller_pause_count = 1;
xen/common/domctl.c +6 -6
index ef202c2b8c..de258ab7f7 100644
@@ -314,7 +314,7 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
             return -ESRCH;
     }
 
-    ret = xsm_domctl(XSM_OTHER, d, op->cmd);
+    ret = xsm_domctl(d, op->cmd);
     if ( ret )
         goto domctl_out_unlock_domonly;
 
@@ -553,7 +553,7 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
         if ( d == NULL )
             goto getdomaininfo_out;
 
-        ret = xsm_getdomaininfo(XSM_HOOK, d);
+        ret = xsm_getdomaininfo(d);
         if ( ret )
             goto getdomaininfo_out;
 
@@ -688,7 +688,7 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
             break;
         }
         irq = pirq_access_permitted(current->domain, pirq);
-        if ( !irq || xsm_irq_permission(XSM_HOOK, d, irq, allow) )
+        if ( !irq || xsm_irq_permission(d, irq, allow) )
             ret = -EPERM;
         else if ( allow )
             ret = irq_permit_access(d, irq);
@@ -709,7 +709,7 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
 
         if ( !iomem_access_permitted(current->domain,
                                      mfn, mfn + nr_mfns - 1) ||
-             xsm_iomem_permission(XSM_HOOK, d, mfn, mfn + nr_mfns - 1, allow) )
+             xsm_iomem_permission(d, mfn, mfn + nr_mfns - 1, allow) )
             ret = -EPERM;
         else if ( allow )
             ret = iomem_permit_access(d, mfn, mfn + nr_mfns - 1);
@@ -746,7 +746,7 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
              !iomem_access_permitted(d, mfn, mfn_end) )
             break;
 
-        ret = xsm_iomem_mapping(XSM_HOOK, d, mfn, mfn_end, add);
+        ret = xsm_iomem_mapping(d, mfn, mfn_end, add);
         if ( ret )
             break;
 
@@ -804,7 +804,7 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
 
         ret = -EOPNOTSUPP;
         if ( is_hvm_domain(e) )
-            ret = xsm_set_target(XSM_HOOK, d, e);
+            ret = xsm_set_target(d, e);
         if ( ret )
         {
             put_domain(e);
xen/common/event_channel.c +6 -6
index 5479315aae..21e7b496ef 100644
@@ -296,7 +296,7 @@ static long evtchn_alloc_unbound(evtchn_alloc_unbound_t *alloc)
         ERROR_EXIT_DOM(port, d);
     chn = evtchn_from_port(d, port);
 
-    rc = xsm_evtchn_unbound(XSM_TARGET, d, chn, alloc->remote_dom);
+    rc = xsm_evtchn_unbound(d, chn, alloc->remote_dom);
     if ( rc )
         goto out;
 
@@ -372,7 +372,7 @@ static long evtchn_bind_interdomain(evtchn_bind_interdomain_t *bind)
          (rchn->u.unbound.remote_domid != ld->domain_id) )
         ERROR_EXIT_DOM(-EINVAL, rd);
 
-    rc = xsm_evtchn_interdomain(XSM_HOOK, ld, lchn, rd, rchn);
+    rc = xsm_evtchn_interdomain(ld, lchn, rd, rchn);
     if ( rc )
         goto out;
 
@@ -760,7 +760,7 @@ int evtchn_send(struct domain *ld, unsigned int lport)
         goto out;
     }
 
-    ret = xsm_evtchn_send(XSM_HOOK, ld, lchn);
+    ret = xsm_evtchn_send(ld, lchn);
     if ( ret )
         goto out;
 
@@ -985,7 +985,7 @@ int evtchn_status(evtchn_status_t *status)
         goto out;
     }
 
-    rc = xsm_evtchn_status(XSM_TARGET, d, chn);
+    rc = xsm_evtchn_status(d, chn);
     if ( rc )
         goto out;
 
@@ -1310,7 +1310,7 @@ long do_event_channel_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( d == NULL )
             return -ESRCH;
 
-        rc = xsm_evtchn_reset(XSM_TARGET, current->domain, d);
+        rc = xsm_evtchn_reset(current->domain, d);
         if ( !rc )
             rc = evtchn_reset(d, cmd == EVTCHNOP_reset_cont);
 
@@ -1371,7 +1371,7 @@ int alloc_unbound_xen_event_channel(
         goto out;
     chn = evtchn_from_port(ld, port);
 
-    rc = xsm_evtchn_unbound(XSM_TARGET, ld, chn, remote_domid);
+    rc = xsm_evtchn_unbound(ld, chn, remote_domid);
     if ( rc )
         goto out;
 
xen/common/grant_table.c +8 -8
index ab30e2e8cf..df516f6340 100644
@@ -1063,7 +1063,7 @@ map_grant_ref(
         return;
     }
 
-    rc = xsm_grant_mapref(XSM_HOOK, ld, rd, op->flags);
+    rc = xsm_grant_mapref(ld, rd, op->flags);
     if ( rc )
     {
         rcu_unlock_domain(rd);
@@ -1403,7 +1403,7 @@ unmap_common(
         return;
     }
 
-    rc = xsm_grant_unmapref(XSM_HOOK, ld, rd);
+    rc = xsm_grant_unmapref(ld, rd);
     if ( rc )
     {
         rcu_unlock_domain(rd);
@@ -2021,7 +2021,7 @@ gnttab_setup_table(
         goto out;
     }
 
-    if ( xsm_grant_setup(XSM_TARGET, curr->domain, d) )
+    if ( xsm_grant_setup(curr->domain, d) )
     {
         op.status = GNTST_permission_denied;
         goto out;
@@ -2103,7 +2103,7 @@ gnttab_query_size(
         goto out;
     }
 
-    if ( xsm_grant_query_size(XSM_TARGET, current->domain, d) )
+    if ( xsm_grant_query_size(current->domain, d) )
     {
         op.status = GNTST_permission_denied;
         goto out;
@@ -2274,7 +2274,7 @@ gnttab_transfer(
             goto put_gfn_and_copyback;
         }
 
-        if ( xsm_grant_transfer(XSM_HOOK, d, e) )
+        if ( xsm_grant_transfer(d, e) )
         {
             gop.status = GNTST_permission_denied;
         unlock_and_copyback:
@@ -2812,7 +2812,7 @@ static int gnttab_copy_lock_domains(const struct gnttab_copy *op,
     if ( rc < 0 )
         goto error;
 
-    rc = xsm_grant_copy(XSM_HOOK, src->domain, dest->domain);
+    rc = xsm_grant_copy(src->domain, dest->domain);
     if ( rc < 0 )
     {
         rc = GNTST_permission_denied;
@@ -3231,7 +3231,7 @@ gnttab_get_status_frames(XEN_GUEST_HANDLE_PARAM(gnttab_get_status_frames_t) uop,
         op.status = GNTST_bad_domain;
         goto out1;
     }
-    rc = xsm_grant_setup(XSM_TARGET, current->domain, d);
+    rc = xsm_grant_setup(current->domain, d);
     if ( rc )
     {
         op.status = GNTST_permission_denied;
@@ -3295,7 +3295,7 @@ gnttab_get_version(XEN_GUEST_HANDLE_PARAM(gnttab_get_version_t) uop)
     if ( d == NULL )
         return -ESRCH;
 
-    rc = xsm_grant_query_size(XSM_TARGET, current->domain, d);
+    rc = xsm_grant_query_size(current->domain, d);
     if ( rc )
     {
         rcu_unlock_domain(d);
xen/common/hypfs.c +1 -1
index e71f7df479..052f3d472a 100644
@@ -679,7 +679,7 @@ long do_hypfs_op(unsigned int cmd,
     struct hypfs_entry *entry;
     static char path[XEN_HYPFS_MAX_PATHLEN];
 
-    if ( xsm_hypfs_op(XSM_PRIV) )
+    if ( xsm_hypfs_op() )
         return -EPERM;
 
     if ( cmd == XEN_HYPFS_OP_get_version )
xen/common/kernel.c +1 -1
index d77756a81e..89e01e908c 100644
@@ -459,7 +459,7 @@ __initcall(param_init);
 
 DO(xen_version)(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
 {
-    bool_t deny = !!xsm_xen_version(XSM_OTHER, cmd);
+    bool_t deny = !!xsm_xen_version(cmd);
 
     switch ( cmd )
     {
xen/common/kexec.c +1 -1
index ebeee6405a..a0d2858cd8 100644
@@ -1219,7 +1219,7 @@ static int do_kexec_op_internal(unsigned long op,
 {
     int ret = -EINVAL;
 
-    ret = xsm_kexec(XSM_PRIV);
+    ret = xsm_kexec();
     if ( ret )
         return ret;
 
xen/common/mem_access.c +1 -1
index 010e6f8dbf..2066510d3b 100644
@@ -47,7 +47,7 @@ int mem_access_memop(unsigned long cmd,
     if ( !p2m_mem_access_sanity_check(d) )
         goto out;
 
-    rc = xsm_mem_access(XSM_DM_PRIV, d);
+    rc = xsm_mem_access(d);
     if ( rc )
         goto out;
 
xen/common/memory.c +8 -8
index 72a6b70cb5..d2621bbb47 100644
@@ -609,7 +609,7 @@ static long memory_exchange(XEN_GUEST_HANDLE_PARAM(xen_memory_exchange_t) arg)
         goto fail_early;
     }
 
-    rc = xsm_memory_exchange(XSM_TARGET, d);
+    rc = xsm_memory_exchange(d);
     if ( rc )
     {
         rcu_unlock_domain(d);
@@ -1072,7 +1072,7 @@ static long xatp_permission_check(struct domain *d, unsigned int space)
          (!is_hardware_domain(d) || (d != current->domain)) )
         return -EACCES;
 
-    return xsm_add_to_physmap(XSM_TARGET, current->domain, d);
+    return xsm_add_to_physmap(current->domain, d);
 }
 
 unsigned int ioreq_server_max_frames(const struct domain *d)
@@ -1232,7 +1232,7 @@ static int acquire_resource(
     if ( rc )
         return rc;
 
-    rc = xsm_domain_resource_map(XSM_DM_PRIV, d);
+    rc = xsm_domain_resource_map(d);
     if ( rc )
         goto out;
 
@@ -1388,7 +1388,7 @@ long do_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
              && (reservation.mem_flags & XENMEMF_populate_on_demand) )
             args.memflags |= MEMF_populate_on_demand;
 
-        if ( xsm_memory_adjust_reservation(XSM_TARGET, curr_d, d) )
+        if ( xsm_memory_adjust_reservation(curr_d, d) )
         {
             rcu_unlock_domain(d);
             return start_extent;
@@ -1462,7 +1462,7 @@ long do_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( d == NULL )
             return -ESRCH;
 
-        rc = xsm_memory_stat_reservation(XSM_TARGET, curr_d, d);
+        rc = xsm_memory_stat_reservation(curr_d, d);
         if ( rc )
         {
             rcu_unlock_domain(d);
@@ -1584,7 +1584,7 @@ long do_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
             return -ESRCH;
 
         rc = paging_mode_translate(d)
-             ? xsm_remove_from_physmap(XSM_TARGET, curr_d, d)
+             ? xsm_remove_from_physmap(curr_d, d)
              : -EACCES;
         if ( rc )
         {
@@ -1631,7 +1631,7 @@ long do_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( d == NULL )
             return -EINVAL;
 
-        rc = xsm_claim_pages(XSM_PRIV, d);
+        rc = xsm_claim_pages(d);
 
         if ( !rc )
             rc = domain_set_outstanding_pages(d, reservation.nr_extents);
@@ -1662,7 +1662,7 @@ long do_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( (d = rcu_lock_domain_by_any_id(topology.domid)) == NULL )
             return -ESRCH;
 
-        rc = xsm_get_vnumainfo(XSM_TARGET, d);
+        rc = xsm_get_vnumainfo(d);
         if ( rc )
         {
             rcu_unlock_domain(d);
xen/common/monitor.c +1 -1
index d5c9ff1cbf..ff17bad733 100644
@@ -36,7 +36,7 @@ int monitor_domctl(struct domain *d, struct xen_domctl_monitor_op *mop)
     if ( unlikely(current->domain == d) ) /* no domain_pause() */
         return -EPERM;
 
-    rc = xsm_vm_event_control(XSM_PRIV, d, mop->op, mop->event);
+    rc = xsm_vm_event_control(d, mop->op, mop->event);
     if ( unlikely(rc) )
         return rc;
 
xen/common/sched/core.c +3 -3
index 6d34764d38..e5c154fe9d 100644
@@ -1944,7 +1944,7 @@ ret_t do_sched_op(int cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
         if ( d == NULL )
             break;
 
-        ret = xsm_schedop_shutdown(XSM_DM_PRIV, current->domain, d);
+        ret = xsm_schedop_shutdown(current->domain, d);
         if ( likely(!ret) )
             domain_shutdown(d, sched_remote_shutdown.reason);
 
@@ -2046,7 +2046,7 @@ long sched_adjust(struct domain *d, struct xen_domctl_scheduler_op *op)
 {
     long ret;
 
-    ret = xsm_domctl_scheduler_op(XSM_HOOK, d, op->cmd);
+    ret = xsm_domctl_scheduler_op(d, op->cmd);
     if ( ret )
         return ret;
 
@@ -2081,7 +2081,7 @@ long sched_adjust_global(struct xen_sysctl_scheduler_op *op)
     struct cpupool *pool;
     int rc;
 
-    rc = xsm_sysctl_scheduler_op(XSM_HOOK, op->cmd);
+    rc = xsm_sysctl_scheduler_op(op->cmd);
     if ( rc )
         return rc;
 
xen/common/sysctl.c +4 -4
index 3558641cd9..4e25c0e499 100644
@@ -41,7 +41,7 @@ long do_sysctl(XEN_GUEST_HANDLE_PARAM(xen_sysctl_t) u_sysctl)
     if ( op->interface_version != XEN_SYSCTL_INTERFACE_VERSION )
         return -EACCES;
 
-    ret = xsm_sysctl(XSM_PRIV, op->cmd);
+    ret = xsm_sysctl(op->cmd);
     if ( ret )
         return ret;
 
@@ -58,7 +58,7 @@ long do_sysctl(XEN_GUEST_HANDLE_PARAM(xen_sysctl_t) u_sysctl)
     switch ( op->cmd )
     {
     case XEN_SYSCTL_readconsole:
-        ret = xsm_readconsole(XSM_HOOK, op->u.readconsole.clear);
+        ret = xsm_readconsole(op->u.readconsole.clear);
         if ( ret )
             break;
 
@@ -88,7 +88,7 @@ long do_sysctl(XEN_GUEST_HANDLE_PARAM(xen_sysctl_t) u_sysctl)
             if ( num_domains == op->u.getdomaininfolist.max_domains )
                 break;
 
-            ret = xsm_getdomaininfo(XSM_HOOK, d);
+            ret = xsm_getdomaininfo(d);
             if ( ret )
                 continue;
 
@@ -191,7 +191,7 @@ long do_sysctl(XEN_GUEST_HANDLE_PARAM(xen_sysctl_t) u_sysctl)
         if ( op->u.page_offline.end < op->u.page_offline.start )
             break;
 
-        ret = xsm_page_offline(XSM_HOOK, op->u.page_offline.cmd);
+        ret = xsm_page_offline(op->u.page_offline.cmd);
         if ( ret )
             break;
 
xen/common/vm_event.c +1 -1
index 70ab3ba406..307f99fcf0 100644
@@ -584,7 +584,7 @@ int vm_event_domctl(struct domain *d, struct xen_domctl_vm_event_op *vec)
         return 0;
     }
 
-    rc = xsm_vm_event_control(XSM_PRIV, d, vec->mode, vec->op);
+    rc = xsm_vm_event_control(d, vec->mode, vec->op);
     if ( rc )
         return rc;
 
xen/common/xenoprof.c +1 -1
index 1926a92fe4..76d8b1f807 100644
@@ -737,7 +737,7 @@ ret_t do_xenoprof_op(int op, XEN_GUEST_HANDLE_PARAM(void) arg)
         return -EPERM;
     }
 
-    ret = xsm_profile(XSM_HOOK, current->domain, op);
+    ret = xsm_profile(current->domain, op);
     if ( ret )
         return ret;
 
xen/drivers/char/console.c +1 -1
index 7d0a603d03..b5d62ea4ee 100644
@@ -681,7 +681,7 @@ long do_console_io(unsigned int cmd, unsigned int count,
     long rc;
     unsigned int idx, len;
 
-    rc = xsm_console_io(XSM_OTHER, current->domain, cmd);
+    rc = xsm_console_io(current->domain, cmd);
     if ( rc )
         return rc;
 
xen/drivers/passthrough/device_tree.c +2 -2
index 999b831d90..67b03fd2a9 100644
@@ -230,7 +230,7 @@ int iommu_do_dt_domctl(struct xen_domctl *domctl, struct domain *d,
         if ( ret )
             break;
 
-        ret = xsm_assign_dtdevice(XSM_HOOK, d, dt_node_full_name(dev));
+        ret = xsm_assign_dtdevice(d, dt_node_full_name(dev));
         if ( ret )
             break;
 
@@ -284,7 +284,7 @@ int iommu_do_dt_domctl(struct xen_domctl *domctl, struct domain *d,
         if ( ret )
             break;
 
-        ret = xsm_deassign_dtdevice(XSM_HOOK, d, dt_node_full_name(dev));
+        ret = xsm_deassign_dtdevice(d, dt_node_full_name(dev));
 
         if ( d == dom_io )
             return -EINVAL;
xen/drivers/passthrough/pci.c +6 -6
index 199ce08612..1363ef8121 100644
@@ -704,7 +704,7 @@ int pci_add_device(u16 seg, u8 bus, u8 devfn,
     else
         pdev_type = "device";
 
-    ret = xsm_resource_plug_pci(XSM_PRIV, (seg << 16) | (bus << 8) | devfn);
+    ret = xsm_resource_plug_pci((seg << 16) | (bus << 8) | devfn);
     if ( ret )
         return ret;
 
@@ -814,7 +814,7 @@ int pci_remove_device(u16 seg, u8 bus, u8 devfn)
     struct pci_dev *pdev;
     int ret;
 
-    ret = xsm_resource_unplug_pci(XSM_PRIV, (seg << 16) | (bus << 8) | devfn);
+    ret = xsm_resource_unplug_pci((seg << 16) | (bus << 8) | devfn);
     if ( ret )
         return ret;
 
@@ -1477,7 +1477,7 @@ static int iommu_get_device_group(
              ((pdev->bus == bus) && (pdev->devfn == devfn)) )
             continue;
 
-        if ( xsm_get_device_group(XSM_HOOK, (seg << 16) | (pdev->bus << 8) | pdev->devfn) )
+        if ( xsm_get_device_group((seg << 16) | (pdev->bus << 8) | pdev->devfn) )
             continue;
 
         sdev_id = ops->get_device_group_id(seg, pdev->bus, pdev->devfn);
@@ -1545,7 +1545,7 @@ int iommu_do_pci_domctl(
         u32 max_sdevs;
         XEN_GUEST_HANDLE_64(uint32) sdevs;
 
-        ret = xsm_get_device_group(XSM_HOOK, domctl->u.get_device_group.machine_sbdf);
+        ret = xsm_get_device_group(domctl->u.get_device_group.machine_sbdf);
         if ( ret )
             break;
 
@@ -1596,7 +1596,7 @@ int iommu_do_pci_domctl(
 
         machine_sbdf = domctl->u.assign_device.u.pci.machine_sbdf;
 
-        ret = xsm_assign_device(XSM_HOOK, d, machine_sbdf);
+        ret = xsm_assign_device(d, machine_sbdf);
         if ( ret )
             break;
 
@@ -1641,7 +1641,7 @@ int iommu_do_pci_domctl(
 
         machine_sbdf = domctl->u.assign_device.u.pci.machine_sbdf;
 
-        ret = xsm_deassign_device(XSM_HOOK, d, machine_sbdf);
+        ret = xsm_deassign_device(d, machine_sbdf);
         if ( ret )
             break;
 
xen/include/xsm/xsm.h +85 -87

546 lines not on this page · raw message

Signature · 1 line
2.20.1
#23 lore raw

Re: [PATCH 4/6] xsm: remove xen_defualt_t from hook definitions

Quoted 6 lines · Daniel P. Smith

With the conversion of making XSM always enabled even the dummy XSM module is being invoked through the xsm_ops dispatch which does not use passing of the default privilege. This commit removes the xen_default_t parameter from the hook definitions and all the respective call sites.

I'm struggling to parse the first sentence.  Also, there's a typo in the subject.

~Andrew

#24 lore raw

Re: [PATCH 4/6] xsm: remove xen_defualt_t from hook definitions

Quoted 10 lines · Andrew Cooper
Quoted 6 lines · Daniel P. Smith

With the conversion of making XSM always enabled even the dummy XSM module is being invoked through the xsm_ops dispatch which does not use passing of the default privilege. This commit removes the xen_default_t parameter from the hook definitions and all the respective call sites.

I'm struggling to parse the first sentence.  Also, there's a typo in the subject.

I will rewrite.

v/r,
dps

#25 lore raw

Re: [PATCH 4/6] xsm: remove xen_defualt_t from hook definitions

In reply to Daniel P. Smith
Quoted 4 lines · Daniel P. Smith

With the conversion of making XSM always enabled even the dummy XSM module is being invoked through the xsm_ops dispatch which does not use passing of the default privilege. This commit removes the xen_default_t parameter from the hook definitions and all the respective call sites.

I don't think it has really become clear from the earlier patches that even in dummy mode we now have (as per my reading of the above) actual function calls, where previously everything got inlined in this case. I'm afraid I view this as another argument against the removal of XSM as a top level Kconfig option.

Jan

#26 lore raw

PATCH 5/6 xsm: expanding function related macros in dummy.h

In reply to Daniel P. Smith

With the elimination of switching how dummy.h gets included, the function declaration macros are no longer necessary. This commit expands them out to the only value for which they will ever be set. This results in function declaration lengths changing and since some definitions did not even follow the 80 column wrapping style, all function definitions were aligned with the predominate style that is used in the hypervisor code.

---
 xen/xsm/dummy.h | 274 +++++++++++++++++++++++++++---------------------
 1 file changed, 153 insertions(+), 121 deletions(-)
xen/xsm/dummy.h +153 -121
index 7e2bb09dac..0f8ea163af 100644
@@ -9,7 +9,7 @@
  *
  *
  *  Each XSM hook implementing an access check should have its first parameter
- *  preceded by XSM_DEFAULT_ARG (or use XSM_DEFAULT_VOID if it has no
+ *  preceded by (or use XSM_DEFAULT_VOID if it has no
  *  arguments). The first non-declaration statement shold be XSM_ASSERT_ACTION
  *  with the expected type of the hook, which will either define or check the
  *  value of action.
@@ -47,14 +47,12 @@ void __xsm_action_mismatch_detected(void);
  * xsm_default_t argument available, so the value from the assertion is used to
  * initialize the variable.
  */
-#define XSM_INLINE __maybe_unused
-
-#define XSM_DEFAULT_ARG /* */
 #define XSM_DEFAULT_VOID void
 #define XSM_ASSERT_ACTION(def) xsm_default_t action = def; (void)action
 
-static always_inline int xsm_default_action(
-    xsm_default_t action, struct domain *src, struct domain *target)
+static always_inline int xsm_default_action(xsm_default_t action,
+                                            struct domain *src,
+                                            struct domain *target)
 {
     switch ( action ) {
     case XSM_HOOK:
@@ -82,43 +80,43 @@ static always_inline int xsm_default_action(
     }
 }
 
-static XSM_INLINE void dummy_security_domaininfo(struct domain *d,
+static __maybe_unused void dummy_security_domaininfo(struct domain *d,
                                     struct xen_domctl_getdomaininfo *info)
 {
     return;
 }
 
-static XSM_INLINE int dummy_domain_create(XSM_DEFAULT_ARG struct domain *d, u32 ssidref)
+static __maybe_unused int dummy_domain_create(struct domain *d, u32 ssidref)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_getdomaininfo(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_getdomaininfo(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_domctl_scheduler_op(XSM_DEFAULT_ARG struct domain *d, int cmd)
+static __maybe_unused int dummy_domctl_scheduler_op(struct domain *d, int cmd)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_sysctl_scheduler_op(XSM_DEFAULT_ARG int cmd)
+static __maybe_unused int dummy_sysctl_scheduler_op(int cmd)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_set_target(XSM_DEFAULT_ARG struct domain *d, struct domain *e)
+static __maybe_unused int dummy_set_target(struct domain *d, struct domain *e)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_domctl(XSM_DEFAULT_ARG struct domain *d, int cmd)
+static __maybe_unused int dummy_domctl(struct domain *d, int cmd)
 {
     XSM_ASSERT_ACTION(XSM_OTHER);
     switch ( cmd )
@@ -135,85 +133,91 @@ static XSM_INLINE int dummy_domctl(XSM_DEFAULT_ARG struct domain *d, int cmd)
     }
 }
 
-static XSM_INLINE int dummy_sysctl(XSM_DEFAULT_ARG int cmd)
+static __maybe_unused int dummy_sysctl(int cmd)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_readconsole(XSM_DEFAULT_ARG uint32_t clear)
+static __maybe_unused int dummy_readconsole(uint32_t clear)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_alloc_security_domain(struct domain *d)
+static __maybe_unused int dummy_alloc_security_domain(struct domain *d)
 {
     return 0;
 }
 
-static XSM_INLINE void dummy_free_security_domain(struct domain *d)
+static __maybe_unused void dummy_free_security_domain(struct domain *d)
 {
     return;
 }
 
-static XSM_INLINE int dummy_grant_mapref(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2,
-                                                                uint32_t flags)
+static __maybe_unused int dummy_grant_mapref(struct domain *d1,
+                                             struct domain *d2,
+                                             uint32_t flags)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_grant_unmapref(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_grant_unmapref(struct domain *d1,
+                                               struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_grant_setup(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_grant_setup(struct domain *d1,
+                                            struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_grant_transfer(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_grant_transfer(struct domain *d1,
+                                               struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_grant_copy(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_grant_copy(struct domain *d1, struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_grant_query_size(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_grant_query_size(struct domain *d1,
+                                                 struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_memory_exchange(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_memory_exchange(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_memory_adjust_reservation(XSM_DEFAULT_ARG struct domain *d1,
-                                                            struct domain *d2)
+static __maybe_unused int dummy_memory_adjust_reservation(struct domain *d1,
+                                                          struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_memory_stat_reservation(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_memory_stat_reservation(struct domain *d1,
+                                                        struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_console_io(XSM_DEFAULT_ARG struct domain *d, int cmd)
+static __maybe_unused int dummy_console_io(struct domain *d, int cmd)
 {
     XSM_ASSERT_ACTION(XSM_OTHER);
     if ( d->is_console )
@@ -225,129 +229,140 @@ static XSM_INLINE int dummy_console_io(XSM_DEFAULT_ARG struct domain *d, int cmd
     return xsm_default_action(XSM_PRIV, d, NULL);
 }
 
-static XSM_INLINE int dummy_profile(XSM_DEFAULT_ARG struct domain *d, int op)
+static __maybe_unused int dummy_profile(struct domain *d, int op)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, d, NULL);
 }
 
-static XSM_INLINE int dummy_kexec(XSM_DEFAULT_VOID)
+static __maybe_unused int dummy_kexec(XSM_DEFAULT_VOID)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_schedop_shutdown(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_schedop_shutdown(struct domain *d1,
+                                                 struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_memory_pin_page(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2,
-                                          struct page_info *page)
+static __maybe_unused int dummy_memory_pin_page(struct domain *d1,
+                                                struct domain *d2,
+                                                struct page_info *page)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_claim_pages(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_claim_pages(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_evtchn_unbound(XSM_DEFAULT_ARG struct domain *d, struct evtchn *chn,
-                                         domid_t id2)
+static __maybe_unused int dummy_evtchn_unbound(struct domain *d,
+                                               struct evtchn *chn,
+                                               domid_t id2)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_evtchn_interdomain(XSM_DEFAULT_ARG struct domain *d1, struct evtchn
-                                *chan1, struct domain *d2, struct evtchn *chan2)
+static __maybe_unused int dummy_evtchn_interdomain(struct domain *d1,
+                                                   struct evtchn *chan1,
+                                                   struct domain *d2,
+                                                   struct evtchn *chan2)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE void dummy_evtchn_close_post(struct evtchn *chn)
+static __maybe_unused void dummy_evtchn_close_post(struct evtchn *chn)
 {
     return;
 }
 
-static XSM_INLINE int dummy_evtchn_send(XSM_DEFAULT_ARG struct domain *d, struct evtchn *chn)
+static __maybe_unused int dummy_evtchn_send(struct domain *d,
+                                            struct evtchn *chn)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, d, NULL);
 }
 
-static XSM_INLINE int dummy_evtchn_status(XSM_DEFAULT_ARG struct domain *d, struct evtchn *chn)
+static __maybe_unused int dummy_evtchn_status(struct domain *d,
+                                              struct evtchn *chn)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_evtchn_reset(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_evtchn_reset(struct domain *d1,
+                                             struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_alloc_security_evtchns(
-    struct evtchn chn[], unsigned int nr)
+static __maybe_unused int dummy_alloc_security_evtchns(struct evtchn chn[],
+                                                       unsigned int nr)
 {
     return 0;
 }
 
-static XSM_INLINE void dummy_free_security_evtchns(
-    struct evtchn chn[], unsigned int nr)
+static __maybe_unused void dummy_free_security_evtchns(struct evtchn chn[],
+                                                       unsigned int nr)
 {
     return;
 }
 
-static XSM_INLINE char *dummy_show_security_evtchn(struct domain *d, const struct evtchn *chn)
+static __maybe_unused char *dummy_show_security_evtchn(struct domain *d,
+                                                       const struct evtchn *chn)
 {
     return NULL;
 }
 
-static XSM_INLINE int dummy_init_hardware_domain(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_init_hardware_domain(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_get_pod_target(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_get_pod_target(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_set_pod_target(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_set_pod_target(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_get_vnumainfo(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_get_vnumainfo(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, current->domain, d);
 }
 
 #if defined(CONFIG_HAS_PASSTHROUGH) && defined(CONFIG_HAS_PCI)
-static XSM_INLINE int dummy_get_device_group(XSM_DEFAULT_ARG uint32_t machine_bdf)
+static __maybe_unused int dummy_get_device_group(uint32_t machine_bdf)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_assign_device(XSM_DEFAULT_ARG struct domain *d, uint32_t machine_bdf)
+static __maybe_unused int dummy_assign_device(struct domain *d,
+                                              uint32_t machine_bdf)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_deassign_device(XSM_DEFAULT_ARG struct domain *d, uint32_t machine_bdf)
+static __maybe_unused int dummy_deassign_device(struct domain *d,
+                                                uint32_t machine_bdf)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
@@ -356,15 +371,15 @@ static XSM_INLINE int dummy_deassign_device(XSM_DEFAULT_ARG struct domain *d, ui
 #endif /* HAS_PASSTHROUGH && HAS_PCI */
 
 #if defined(CONFIG_HAS_PASSTHROUGH) && defined(CONFIG_HAS_DEVICE_TREE)
-static XSM_INLINE int dummy_assign_dtdevice(XSM_DEFAULT_ARG struct domain *d,
-                                          const char *dtpath)
+static __maybe_unused int dummy_assign_dtdevice(struct domain *d,
+                                                const char *dtpath)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_deassign_dtdevice(XSM_DEFAULT_ARG struct domain *d,
-                                            const char *dtpath)
+static __maybe_unused int dummy_deassign_dtdevice(struct domain *d,
+                                                  const char *dtpath)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
@@ -372,178 +387,190 @@ static XSM_INLINE int dummy_deassign_dtdevice(XSM_DEFAULT_ARG struct domain *d,
 
 #endif /* HAS_PASSTHROUGH && HAS_DEVICE_TREE */
 
-static XSM_INLINE int dummy_resource_plug_core(XSM_DEFAULT_VOID)
+static __maybe_unused int dummy_resource_plug_core(XSM_DEFAULT_VOID)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_resource_unplug_core(XSM_DEFAULT_VOID)
+static __maybe_unused int dummy_resource_unplug_core(XSM_DEFAULT_VOID)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_resource_plug_pci(XSM_DEFAULT_ARG uint32_t machine_bdf)
+static __maybe_unused int dummy_resource_plug_pci(uint32_t machine_bdf)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_resource_unplug_pci(XSM_DEFAULT_ARG uint32_t machine_bdf)
+static __maybe_unused int dummy_resource_unplug_pci(uint32_t machine_bdf)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_resource_setup_pci(XSM_DEFAULT_ARG uint32_t machine_bdf)
+static __maybe_unused int dummy_resource_setup_pci(uint32_t machine_bdf)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_resource_setup_gsi(XSM_DEFAULT_ARG int gsi)
+static __maybe_unused int dummy_resource_setup_gsi(int gsi)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_resource_setup_misc(XSM_DEFAULT_VOID)
+static __maybe_unused int dummy_resource_setup_misc(XSM_DEFAULT_VOID)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_page_offline(XSM_DEFAULT_ARG uint32_t cmd)
+static __maybe_unused int dummy_page_offline(uint32_t cmd)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_hypfs_op(XSM_DEFAULT_VOID)
+static __maybe_unused int dummy_hypfs_op(XSM_DEFAULT_VOID)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE long dummy_do_xsm_op(XEN_GUEST_HANDLE_PARAM(xsm_op_t) op)
+static __maybe_unused long dummy_do_xsm_op(XEN_GUEST_HANDLE_PARAM(xsm_op_t) op)
 {
     return -ENOSYS;
 }
 
 #ifdef CONFIG_COMPAT
-static XSM_INLINE int dummy_do_compat_op(XEN_GUEST_HANDLE_PARAM(xsm_op_t) op)
+static __maybe_unused int dummy_do_compat_op(
+                                XEN_GUEST_HANDLE_PARAM(xsm_op_t) op)
 {
     return -ENOSYS;
 }
 #endif
 
-static XSM_INLINE char *dummy_show_irq_sid(int irq)
+static __maybe_unused char *dummy_show_irq_sid(int irq)
 {
     return NULL;
 }
 
-static XSM_INLINE int dummy_map_domain_pirq(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_map_domain_pirq(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_map_domain_irq(XSM_DEFAULT_ARG struct domain *d,
-                                         int irq, const void *data)
+static __maybe_unused int dummy_map_domain_irq(struct domain *d, int irq,
+                                               const void *data)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_unmap_domain_pirq(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_unmap_domain_pirq(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_bind_pt_irq(XSM_DEFAULT_ARG struct domain *d, struct xen_domctl_bind_pt_irq *bind)
+static __maybe_unused int dummy_bind_pt_irq(struct domain *d,
+                                            struct xen_domctl_bind_pt_irq *bind)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_unbind_pt_irq(XSM_DEFAULT_ARG struct domain *d, struct xen_domctl_bind_pt_irq *bind)
+static __maybe_unused int dummy_unbind_pt_irq(struct domain *d,
+                                        struct xen_domctl_bind_pt_irq *bind)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_unmap_domain_irq(XSM_DEFAULT_ARG struct domain *d,
-                                           int irq, const void *data)
+static __maybe_unused int dummy_unmap_domain_irq(struct domain *d, int irq,
+                                                 const void *data)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_irq_permission(XSM_DEFAULT_ARG struct domain *d, int pirq, uint8_t allow)
+static __maybe_unused int dummy_irq_permission(struct domain *d, int pirq,
+                                               uint8_t allow)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_iomem_permission(XSM_DEFAULT_ARG struct domain *d, uint64_t s, uint64_t e, uint8_t allow)
+static __maybe_unused int dummy_iomem_permission(struct domain *d, uint64_t s,
+                                                 uint64_t e, uint8_t allow)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_iomem_mapping(XSM_DEFAULT_ARG struct domain *d, uint64_t s, uint64_t e, uint8_t allow)
+static __maybe_unused int dummy_iomem_mapping(struct domain *d, uint64_t s,
+                                              uint64_t e, uint8_t allow)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_pci_config_permission(XSM_DEFAULT_ARG struct domain *d, uint32_t machine_bdf,
-                                        uint16_t start, uint16_t end,
-                                        uint8_t access)
+static __maybe_unused int dummy_pci_config_permission(struct domain *d,
+                                                      uint32_t machine_bdf,
+                                                      uint16_t start,
+                                                      uint16_t end,
+                                                      uint8_t access)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_add_to_physmap(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_add_to_physmap(struct domain *d1,
+                                               struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_remove_from_physmap(XSM_DEFAULT_ARG struct domain *d1, struct domain *d2)
+static __maybe_unused int dummy_remove_from_physmap(struct domain *d1,
+                                                    struct domain *d2)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d1, d2);
 }
 
-static XSM_INLINE int dummy_map_gmfn_foreign(XSM_DEFAULT_ARG struct domain *d, struct domain *t)
+static __maybe_unused int dummy_map_gmfn_foreign(struct domain *d,
+                                                 struct domain *t)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d, t);
 }
 
-static XSM_INLINE int dummy_hvm_param(XSM_DEFAULT_ARG struct domain *d, unsigned long op)
+static __maybe_unused int dummy_hvm_param(struct domain *d, unsigned long op)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_hvm_control(XSM_DEFAULT_ARG struct domain *d, unsigned long op)
+static __maybe_unused int dummy_hvm_control(struct domain *d, unsigned long op)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_hvm_param_altp2mhvm(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_hvm_param_altp2mhvm(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_hvm_altp2mhvm_op(XSM_DEFAULT_ARG struct domain *d, uint64_t mode, uint32_t op)
+static __maybe_unused int dummy_hvm_altp2mhvm_op(struct domain *d,
+                                                 uint64_t mode, uint32_t op)
 {
     XSM_ASSERT_ACTION(XSM_OTHER);
 
@@ -562,14 +589,15 @@ static XSM_INLINE int dummy_hvm_altp2mhvm_op(XSM_DEFAULT_ARG struct domain *d, u
     }
 }
 
-static XSM_INLINE int dummy_vm_event_control(XSM_DEFAULT_ARG struct domain *d, int mode, int op)
+static __maybe_unused int dummy_vm_event_control(struct domain *d, int mode,
+                                                 int op)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
 #ifdef CONFIG_MEM_ACCESS
-static XSM_INLINE int dummy_mem_access(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_mem_access(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, d);
@@ -577,7 +605,7 @@ static XSM_INLINE int dummy_mem_access(XSM_DEFAULT_ARG struct domain *d)
 #endif
 
 #ifdef CONFIG_MEM_PAGING
-static XSM_INLINE int dummy_mem_paging(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_mem_paging(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, d);
@@ -585,58 +613,59 @@ static XSM_INLINE int dummy_mem_paging(XSM_DEFAULT_ARG struct domain *d)
 #endif
 
 #ifdef CONFIG_MEM_SHARING
-static XSM_INLINE int dummy_mem_sharing(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_mem_sharing(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 #endif
 
-static XSM_INLINE int dummy_platform_op(XSM_DEFAULT_ARG uint32_t op)
+static __maybe_unused int dummy_platform_op(uint32_t op)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
 #ifdef CONFIG_X86
-static XSM_INLINE int dummy_do_mca(XSM_DEFAULT_VOID)
+static __maybe_unused int dummy_do_mca(XSM_DEFAULT_VOID)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_shadow_control(XSM_DEFAULT_ARG struct domain *d, uint32_t op)
+static __maybe_unused int dummy_shadow_control(struct domain *d, uint32_t op)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_mem_sharing_op(XSM_DEFAULT_ARG struct domain *d, struct domain *cd, int op)
+static __maybe_unused int dummy_mem_sharing_op(struct domain *d,
+                                               struct domain *cd, int op)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, cd);
 }
 
-static XSM_INLINE int dummy_apic(XSM_DEFAULT_ARG struct domain *d, int cmd)
+static __maybe_unused int dummy_apic(struct domain *d, int cmd)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, d, NULL);
 }
 
-static XSM_INLINE int dummy_machine_memory_map(XSM_DEFAULT_VOID)
+static __maybe_unused int dummy_machine_memory_map(XSM_DEFAULT_VOID)
 {
     XSM_ASSERT_ACTION(XSM_PRIV);
     return xsm_default_action(action, current->domain, NULL);
 }
 
-static XSM_INLINE int dummy_domain_memory_map(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_domain_memory_map(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_mmu_update(XSM_DEFAULT_ARG struct domain *d, struct domain *t,
-                                     struct domain *f, uint32_t flags)
+static __maybe_unused int dummy_mmu_update(struct domain *d, struct domain *t,
+                                           struct domain *f, uint32_t flags)
 {
     int rc = 0;
     XSM_ASSERT_ACTION(XSM_TARGET);
@@ -647,38 +676,41 @@ static XSM_INLINE int dummy_mmu_update(XSM_DEFAULT_ARG struct domain *d, struct
     return rc;
 }
 
-static XSM_INLINE int dummy_mmuext_op(XSM_DEFAULT_ARG struct domain *d, struct domain *f)
+static __maybe_unused int dummy_mmuext_op(struct domain *d, struct domain *f)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d, f);
 }
 
-static XSM_INLINE int dummy_update_va_mapping(XSM_DEFAULT_ARG struct domain *d, struct domain *f, 
-                                                            l1_pgentry_t pte)
+static __maybe_unused int dummy_update_va_mapping(struct domain *d,
+                                                  struct domain *f,
+                                                  l1_pgentry_t pte)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d, f);
 }
 
-static XSM_INLINE int dummy_priv_mapping(XSM_DEFAULT_ARG struct domain *d, struct domain *t)
+static __maybe_unused int dummy_priv_mapping(struct domain *d, struct domain *t)
 {
     XSM_ASSERT_ACTION(XSM_TARGET);
     return xsm_default_action(action, d, t);
 }
 
-static XSM_INLINE int dummy_ioport_permission(XSM_DEFAULT_ARG struct domain *d, uint32_t s, uint32_t e, uint8_t allow)
+static __maybe_unused int dummy_ioport_permission(struct domain *d, uint32_t s,
+                                                  uint32_t e, uint8_t allow)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_ioport_mapping(XSM_DEFAULT_ARG struct domain *d, uint32_t s, uint32_t e, uint8_t allow)
+static __maybe_unused int dummy_ioport_mapping(struct domain *d, uint32_t s,
+                                               uint32_t e, uint8_t allow)
 {
     XSM_ASSERT_ACTION(XSM_HOOK);
     return xsm_default_action(action, current->domain, d);
 }
 
-static XSM_INLINE int dummy_pmu_op (XSM_DEFAULT_ARG struct domain *d, unsigned int op)
+static __maybe_unused int dummy_pmu_op (struct domain *d, unsigned int op)
 {
     XSM_ASSERT_ACTION(XSM_OTHER);
     switch ( op )
@@ -695,31 +727,31 @@ static XSM_INLINE int dummy_pmu_op (XSM_DEFAULT_ARG struct domain *d, unsigned i
 
 #endif /* CONFIG_X86 */
 
-static XSM_INLINE int dummy_dm_op(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_dm_op(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, d);
 }
 
 #ifdef CONFIG_ARGO
-static XSM_INLINE int dummy_argo_enable(const struct domain *d)
+static __maybe_unused int dummy_argo_enable(const struct domain *d)
 {
     return 0;
 }
 
-static XSM_INLINE int dummy_argo_register_single_source(const struct domain *d,
-                                                      const struct domain *t)
+static __maybe_unused int dummy_argo_register_single_source(const struct domain *d,
+                                                            const struct domain *t)
 {
     return 0;
 }
 
-static XSM_INLINE int dummy_argo_register_any_source(const struct domain *d)
+static __maybe_unused int dummy_argo_register_any_source(const struct domain *d)
 {
     return 0;
 }
 
-static XSM_INLINE int dummy_argo_send(const struct domain *d,
-                                    const struct domain *t)
+static __maybe_unused int dummy_argo_send(const struct domain *d,
+                                          const struct domain *t)
 {
     return 0;
 }
@@ -727,7 +759,7 @@ static XSM_INLINE int dummy_argo_send(const struct domain *d,
 #endif /* CONFIG_ARGO */
 
 #include <public/version.h>
-static XSM_INLINE int dummy_xen_version (XSM_DEFAULT_ARG uint32_t op)
+static __maybe_unused int dummy_xen_version(uint32_t op)
 {
     XSM_ASSERT_ACTION(XSM_OTHER);
     switch ( op )
@@ -751,7 +783,7 @@ static XSM_INLINE int dummy_xen_version (XSM_DEFAULT_ARG uint32_t op)
     }
 }
 
-static XSM_INLINE int dummy_domain_resource_map(XSM_DEFAULT_ARG struct domain *d)
+static __maybe_unused int dummy_domain_resource_map(struct domain *d)
 {
     XSM_ASSERT_ACTION(XSM_DM_PRIV);
     return xsm_default_action(action, current->domain, d);
Signature · 1 line
2.20.1
#27 lore raw

Re: [PATCH 5/6] xsm: expanding function related macros in dummy.h

Quoted 14 lines · Daniel P. Smith
xen/xsm/dummy.h +1 -1
index 7e2bb09dac..0f8ea163af 100644
@@ -9,7 +9,7 @@
  *
  *
  *  Each XSM hook implementing an access check should have its first parameter
- *  preceded by XSM_DEFAULT_ARG (or use XSM_DEFAULT_VOID if it has no
+ *  preceded by (or use XSM_DEFAULT_VOID if it has no
  *  arguments). The first non-declaration statement shold be XSM_ASSERT_ACTION
  *  with the expected type of the hook, which will either define or check the
  *  value of action.

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 4 lines
xen/xsm/dummy.h +0 -1
@@ -47,14 +47,12 @@ void __xsm_action_mismatch_detected(void);
  * xsm_default_t argument available, so the value from the assertion is used to
  * initialize the variable.
  */
-#define XSM_INLINE __maybe_unused

Nothing in a header file should ever need __maybe_unused.  Now that the !XSM case has been untangled, I think this can be dropped, rather than expanded inline.

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 3 lines
xen/xsm/dummy.h +0 -2
-
-#define XSM_DEFAULT_ARG /* */
 #define XSM_DEFAULT_VOID void

XSM_DEFAULT_VOID needs to disappear too.  I can't see what it is even doing before the cleanup, because if it is missing, you'll fail the compile for using K&R style functions.

~Andrew

#28 lore raw

Re: [PATCH 5/6] xsm: expanding function related macros in dummy.h

Quoted 24 lines · Andrew Cooper
Quoted 14 lines · Daniel P. Smith
xen/xsm/dummy.h +1 -1
index 7e2bb09dac..0f8ea163af 100644
@@ -9,7 +9,7 @@
  *
  *
  *  Each XSM hook implementing an access check should have its first parameter
- *  preceded by XSM_DEFAULT_ARG (or use XSM_DEFAULT_VOID if it has no
+ *  preceded by (or use XSM_DEFAULT_VOID if it has no
  *  arguments). The first non-declaration statement shold be XSM_ASSERT_ACTION
  *  with the expected type of the hook, which will either define or check the
  *  value of action.

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 4 lines
xen/xsm/dummy.h +0 -1
@@ -47,14 +47,12 @@ void __xsm_action_mismatch_detected(void);
  * xsm_default_t argument available, so the value from the assertion is used to
  * initialize the variable.
  */
-#define XSM_INLINE __maybe_unused

Nothing in a header file should ever need __maybe_unused.  Now that the !XSM case has been untangled, I think this can be dropped, rather than expanded inline.

Quoted 7 lines

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 3 lines
xen/xsm/dummy.h +0 -2
-
-#define XSM_DEFAULT_ARG /* */
 #define XSM_DEFAULT_VOID void

XSM_DEFAULT_VOID needs to disappear too.  I can't see what it is even doing before the cleanup, because if it is missing, you'll fail the compile for using K&R style functions.

You need to look at the state before patch 3 to see its purpose. Patch 3 removed the other variant, and hence the need for this one as well, but I think it is reasonable to not clean up everything in one go (unless it would mean touching exactly the same code a 2nd time later on).

Jan

#29 lore raw

Re: [PATCH 5/6] xsm: expanding function related macros in dummy.h

Quoted 33 lines · Jan Beulich
Quoted 24 lines · Andrew Cooper
Quoted 14 lines · Daniel P. Smith
xen/xsm/dummy.h +1 -1
index 7e2bb09dac..0f8ea163af 100644
@@ -9,7 +9,7 @@
  *
  *
  *  Each XSM hook implementing an access check should have its first parameter
- *  preceded by XSM_DEFAULT_ARG (or use XSM_DEFAULT_VOID if it has no
+ *  preceded by (or use XSM_DEFAULT_VOID if it has no
  *  arguments). The first non-declaration statement shold be XSM_ASSERT_ACTION
  *  with the expected type of the hook, which will either define or check the
  *  value of action.

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 4 lines
xen/xsm/dummy.h +0 -1
@@ -47,14 +47,12 @@ void __xsm_action_mismatch_detected(void);
  * xsm_default_t argument available, so the value from the assertion is used to
  * initialize the variable.
  */
-#define XSM_INLINE __maybe_unused

Nothing in a header file should ever need __maybe_unused.  Now that the !XSM case has been untangled, I think this can be dropped, rather than expanded inline.

Quoted 7 lines

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 3 lines
xen/xsm/dummy.h +0 -2
-
-#define XSM_DEFAULT_ARG /* */
 #define XSM_DEFAULT_VOID void

XSM_DEFAULT_VOID needs to disappear too.  I can't see what it is even doing before the cleanup, because if it is missing, you'll fail the compile for using K&R style functions.

Quoted 4 lines

You need to look at the state before patch 3 to see its purpose. Patch 3 removed the other variant, and hence the need for this one as well, but I think it is reasonable to not clean up everything in one go (unless it would mean touching exactly the same code a 2nd time later on).

Albeit, having looked at the patch itself, I agree it should be dropped here together with XSM_DEFAULT_ARG, of which it is (was) a companion. But again, all provided there is agreement to remove the top level XSM option, which I personally don't think is a good idea.

Jan

#30 lore raw

Re: [PATCH 5/6] xsm: expanding function related macros in dummy.h

Quoted 39 lines · Jan Beulich
Quoted 33 lines · Jan Beulich
Quoted 24 lines · Andrew Cooper
Quoted 14 lines · Daniel P. Smith
xen/xsm/dummy.h +1 -1
index 7e2bb09dac..0f8ea163af 100644
@@ -9,7 +9,7 @@
  *
  *
  *  Each XSM hook implementing an access check should have its first parameter
- *  preceded by XSM_DEFAULT_ARG (or use XSM_DEFAULT_VOID if it has no
+ *  preceded by (or use XSM_DEFAULT_VOID if it has no
  *  arguments). The first non-declaration statement shold be XSM_ASSERT_ACTION
  *  with the expected type of the hook, which will either define or check the
  *  value of action.

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 4 lines
xen/xsm/dummy.h +0 -1
@@ -47,14 +47,12 @@ void __xsm_action_mismatch_detected(void);
  * xsm_default_t argument available, so the value from the assertion is used to
  * initialize the variable.
  */
-#define XSM_INLINE __maybe_unused

Nothing in a header file should ever need __maybe_unused.  Now that the !XSM case has been untangled, I think this can be dropped, rather than expanded inline.

Quoted 7 lines

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 3 lines
xen/xsm/dummy.h +0 -2
-
-#define XSM_DEFAULT_ARG /* */
 #define XSM_DEFAULT_VOID void

XSM_DEFAULT_VOID needs to disappear too.  I can't see what it is even doing before the cleanup, because if it is missing, you'll fail the compile for using K&R style functions.

Quoted 4 lines

You need to look at the state before patch 3 to see its purpose. Patch 3 removed the other variant, and hence the need for this one as well, but I think it is reasonable to not clean up everything in one go (unless it would mean touching exactly the same code a 2nd time later on).

Quoted 4 lines

Albeit, having looked at the patch itself, I agree it should be dropped here together with XSM_DEFAULT_ARG, of which it is (was) a companion. But again, all provided there is agreement to remove the top level XSM option, which I personally don't think is a good idea.

ack, will be removing it

v/r,
dps

#31 lore raw

Re: [PATCH 5/6] xsm: expanding function related macros in dummy.h

In reply to Andrew Cooper
Quoted 15 lines · Andrew Cooper
Quoted 14 lines · Daniel P. Smith
xen/xsm/dummy.h +1 -1
index 7e2bb09dac..0f8ea163af 100644
@@ -9,7 +9,7 @@
  *
  *
  *  Each XSM hook implementing an access check should have its first parameter
- *  preceded by XSM_DEFAULT_ARG (or use XSM_DEFAULT_VOID if it has no
+ *  preceded by (or use XSM_DEFAULT_VOID if it has no
  *  arguments). The first non-declaration statement shold be XSM_ASSERT_ACTION
  *  with the expected type of the hook, which will either define or check the
  *  value of action.
@@ -47,14 +47,12 @@ void __xsm_action_mismatch_detected(void);
Quoted 8 lines

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 4 lines
xen/xsm/dummy.h +0 -1
  * xsm_default_t argument available, so the value from the assertion is used to
  * initialize the variable.
  */
-#define XSM_INLINE __maybe_unused

Nothing in a header file should ever need __maybe_unused.  Now that the !XSM case has been untangled, I think this can be dropped, rather than expanded inline.

ack

Quoted 7 lines

xen/xsm/dummy.h void __xsm_action_mismatch_detected(void); · line 50

Quoted 3 lines
xen/xsm/dummy.h +0 -2
-
-#define XSM_DEFAULT_ARG /* */
 #define XSM_DEFAULT_VOID void

XSM_DEFAULT_VOID needs to disappear too.  I can't see what it is even doing before the cleanup, because if it is missing, you'll fail the compile for using K&R style functions.

ack, will drop

v/r,
dps

#32 lore raw

PATCH 6/6 xsm: removing the XSM_ASSERT_ACTION macro

In reply to Daniel P. Smith

With the eliminations of default priv from all the XSM hook call sites, this renders the XSM_ASSERT_ACTION macro unneeded. This commit cleans up all the dummy hooks, removing the macro.

---
 xen/xsm/dummy.h | 253 +++++++++++++++---------------------------------
 1 file changed, 80 insertions(+), 173 deletions(-)
xen/xsm/dummy.h +80 -173
index 0f8ea163af..1a9a6b2935 100644
@@ -6,13 +6,6 @@
  *  This program is free software; you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License version 2,
  *  as published by the Free Software Foundation.
- *
- *
- *  Each XSM hook implementing an access check should have its first parameter
- *  preceded by (or use XSM_DEFAULT_VOID if it has no
- *  arguments). The first non-declaration statement shold be XSM_ASSERT_ACTION
- *  with the expected type of the hook, which will either define or check the
- *  value of action.
  */
 
 #include <xen/sched.h>
@@ -48,7 +41,6 @@ void __xsm_action_mismatch_detected(void);
  * initialize the variable.
  */
 #define XSM_DEFAULT_VOID void
-#define XSM_ASSERT_ACTION(def) xsm_default_t action = def; (void)action
 
 static always_inline int xsm_default_action(xsm_default_t action,
                                             struct domain *src,
@@ -88,37 +80,31 @@ static __maybe_unused void dummy_security_domaininfo(struct domain *d,
 
 static __maybe_unused int dummy_domain_create(struct domain *d, u32 ssidref)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_getdomaininfo(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_domctl_scheduler_op(struct domain *d, int cmd)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_sysctl_scheduler_op(int cmd)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_HOOK, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_set_target(struct domain *d, struct domain *e)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_HOOK, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_domctl(struct domain *d, int cmd)
 {
-    XSM_ASSERT_ACTION(XSM_OTHER);
     switch ( cmd )
     {
     case XEN_DOMCTL_ioport_mapping:
@@ -135,14 +121,12 @@ static __maybe_unused int dummy_domctl(struct domain *d, int cmd)
 
 static __maybe_unused int dummy_sysctl(int cmd)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_readconsole(uint32_t clear)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_HOOK, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_alloc_security_domain(struct domain *d)
@@ -159,67 +143,57 @@ static __maybe_unused int dummy_grant_mapref(struct domain *d1,
                                              struct domain *d2,
                                              uint32_t flags)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_HOOK, d1, d2);
 }
 
 static __maybe_unused int dummy_grant_unmapref(struct domain *d1,
                                                struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_HOOK, d1, d2);
 }
 
 static __maybe_unused int dummy_grant_setup(struct domain *d1,
                                             struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_TARGET, d1, d2);
 }
 
 static __maybe_unused int dummy_grant_transfer(struct domain *d1,
                                                struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_HOOK, d1, d2);
 }
 
 static __maybe_unused int dummy_grant_copy(struct domain *d1, struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_HOOK, d1, d2);
 }
 
 static __maybe_unused int dummy_grant_query_size(struct domain *d1,
                                                  struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_TARGET, d1, d2);
 }
 
 static __maybe_unused int dummy_memory_exchange(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_TARGET, current->domain, d);
 }
 
 static __maybe_unused int dummy_memory_adjust_reservation(struct domain *d1,
                                                           struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_TARGET, d1, d2);
 }
 
 static __maybe_unused int dummy_memory_stat_reservation(struct domain *d1,
                                                         struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_TARGET, d1, d2);
 }
 
 static __maybe_unused int dummy_console_io(struct domain *d, int cmd)
 {
-    XSM_ASSERT_ACTION(XSM_OTHER);
     if ( d->is_console )
         return xsm_default_action(XSM_HOOK, d, NULL);
 #ifdef CONFIG_VERBOSE_DEBUG
@@ -231,43 +205,37 @@ static __maybe_unused int dummy_console_io(struct domain *d, int cmd)
 
 static __maybe_unused int dummy_profile(struct domain *d, int op)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, d, NULL);
+    return xsm_default_action(XSM_HOOK, d, NULL);
 }
 
 static __maybe_unused int dummy_kexec(XSM_DEFAULT_VOID)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_schedop_shutdown(struct domain *d1,
                                                  struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_PRIV, d1, d2);
 }
 
 static __maybe_unused int dummy_memory_pin_page(struct domain *d1,
                                                 struct domain *d2,
                                                 struct page_info *page)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_HOOK, d1, d2);
 }
 
 static __maybe_unused int dummy_claim_pages(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_PRIV, current->domain, d);
 }
 
 static __maybe_unused int dummy_evtchn_unbound(struct domain *d,
                                                struct evtchn *chn,
                                                domid_t id2)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_TARGET, current->domain, d);
 }
 
 static __maybe_unused int dummy_evtchn_interdomain(struct domain *d1,
@@ -275,8 +243,7 @@ static __maybe_unused int dummy_evtchn_interdomain(struct domain *d1,
                                                    struct domain *d2,
                                                    struct evtchn *chan2)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_HOOK, d1, d2);
 }
 
 static __maybe_unused void dummy_evtchn_close_post(struct evtchn *chn)
@@ -287,22 +254,19 @@ static __maybe_unused void dummy_evtchn_close_post(struct evtchn *chn)
 static __maybe_unused int dummy_evtchn_send(struct domain *d,
                                             struct evtchn *chn)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, d, NULL);
+    return xsm_default_action(XSM_HOOK, d, NULL);
 }
 
 static __maybe_unused int dummy_evtchn_status(struct domain *d,
                                               struct evtchn *chn)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_TARGET, current->domain, d);
 }
 
 static __maybe_unused int dummy_evtchn_reset(struct domain *d1,
                                              struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_TARGET, d1, d2);
 }
 
 static __maybe_unused int dummy_alloc_security_evtchns(struct evtchn chn[],
@@ -325,47 +289,40 @@ static __maybe_unused char *dummy_show_security_evtchn(struct domain *d,
 
 static __maybe_unused int dummy_init_hardware_domain(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_get_pod_target(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_PRIV, current->domain, d);
 }
 
 static __maybe_unused int dummy_set_pod_target(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_PRIV, current->domain, d);
 }
 
 static __maybe_unused int dummy_get_vnumainfo(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_TARGET, current->domain, d);
 }
 
 #if defined(CONFIG_HAS_PASSTHROUGH) && defined(CONFIG_HAS_PCI)
 static __maybe_unused int dummy_get_device_group(uint32_t machine_bdf)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_HOOK, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_assign_device(struct domain *d,
                                               uint32_t machine_bdf)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_deassign_device(struct domain *d,
                                                 uint32_t machine_bdf)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 #endif /* HAS_PASSTHROUGH && HAS_PCI */
@@ -374,71 +331,60 @@ static __maybe_unused int dummy_deassign_device(struct domain *d,
 static __maybe_unused int dummy_assign_dtdevice(struct domain *d,
                                                 const char *dtpath)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_deassign_dtdevice(struct domain *d,
                                                   const char *dtpath)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 #endif /* HAS_PASSTHROUGH && HAS_DEVICE_TREE */
 
 static __maybe_unused int dummy_resource_plug_core(XSM_DEFAULT_VOID)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_HOOK, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_resource_unplug_core(XSM_DEFAULT_VOID)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_HOOK, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_resource_plug_pci(uint32_t machine_bdf)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_resource_unplug_pci(uint32_t machine_bdf)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_resource_setup_pci(uint32_t machine_bdf)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_resource_setup_gsi(int gsi)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_resource_setup_misc(XSM_DEFAULT_VOID)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_page_offline(uint32_t cmd)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_HOOK, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_hypfs_op(XSM_DEFAULT_VOID)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused long dummy_do_xsm_op(XEN_GUEST_HANDLE_PARAM(xsm_op_t) op)
@@ -461,63 +407,54 @@ static __maybe_unused char *dummy_show_irq_sid(int irq)
 
 static __maybe_unused int dummy_map_domain_pirq(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, d);
 }
 
 static __maybe_unused int dummy_map_domain_irq(struct domain *d, int irq,
                                                const void *data)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_unmap_domain_pirq(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, d);
 }
 
 static __maybe_unused int dummy_bind_pt_irq(struct domain *d,
                                             struct xen_domctl_bind_pt_irq *bind)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_unbind_pt_irq(struct domain *d,
                                         struct xen_domctl_bind_pt_irq *bind)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_unmap_domain_irq(struct domain *d, int irq,
                                                  const void *data)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_irq_permission(struct domain *d, int pirq,
                                                uint8_t allow)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_iomem_permission(struct domain *d, uint64_t s,
                                                  uint64_t e, uint8_t allow)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_iomem_mapping(struct domain *d, uint64_t s,
                                               uint64_t e, uint8_t allow)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_pci_config_permission(struct domain *d,
@@ -526,54 +463,45 @@ static __maybe_unused int dummy_pci_config_permission(struct domain *d,
                                                       uint16_t end,
                                                       uint8_t access)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_add_to_physmap(struct domain *d1,
                                                struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_TARGET, d1, d2);
 }
 
 static __maybe_unused int dummy_remove_from_physmap(struct domain *d1,
                                                     struct domain *d2)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d1, d2);
+    return xsm_default_action(XSM_TARGET, d1, d2);
 }
 
 static __maybe_unused int dummy_map_gmfn_foreign(struct domain *d,
                                                  struct domain *t)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d, t);
+    return xsm_default_action(XSM_TARGET, d, t);
 }
 
 static __maybe_unused int dummy_hvm_param(struct domain *d, unsigned long op)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_TARGET, current->domain, d);
 }
 
 static __maybe_unused int dummy_hvm_control(struct domain *d, unsigned long op)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, d);
 }
 
 static __maybe_unused int dummy_hvm_param_altp2mhvm(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_PRIV, current->domain, d);
 }
 
 static __maybe_unused int dummy_hvm_altp2mhvm_op(struct domain *d,
                                                  uint64_t mode, uint32_t op)
 {
-    XSM_ASSERT_ACTION(XSM_OTHER);
-
     switch ( mode )
     {
     case XEN_ALTP2M_mixed:
@@ -592,127 +520,109 @@ static __maybe_unused int dummy_hvm_altp2mhvm_op(struct domain *d,
 static __maybe_unused int dummy_vm_event_control(struct domain *d, int mode,
                                                  int op)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_PRIV, current->domain, d);
 }
 
 #ifdef CONFIG_MEM_ACCESS
 static __maybe_unused int dummy_mem_access(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, d);
 }
 #endif
 
 #ifdef CONFIG_MEM_PAGING
 static __maybe_unused int dummy_mem_paging(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, d);
 }
 #endif
 
 #ifdef CONFIG_MEM_SHARING
 static __maybe_unused int dummy_mem_sharing(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, d);
 }
 #endif
 
 static __maybe_unused int dummy_platform_op(uint32_t op)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 #ifdef CONFIG_X86
 static __maybe_unused int dummy_do_mca(XSM_DEFAULT_VOID)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_shadow_control(struct domain *d, uint32_t op)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_mem_sharing_op(struct domain *d,
                                                struct domain *cd, int op)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, cd);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, cd);
 }
 
 static __maybe_unused int dummy_apic(struct domain *d, int cmd)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, d, NULL);
+    return xsm_default_action(XSM_PRIV, d, NULL);
 }
 
 static __maybe_unused int dummy_machine_memory_map(XSM_DEFAULT_VOID)
 {
-    XSM_ASSERT_ACTION(XSM_PRIV);
-    return xsm_default_action(action, current->domain, NULL);
+    return xsm_default_action(XSM_PRIV, current->domain, NULL);
 }
 
 static __maybe_unused int dummy_domain_memory_map(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_TARGET, current->domain, d);
 }
 
 static __maybe_unused int dummy_mmu_update(struct domain *d, struct domain *t,
                                            struct domain *f, uint32_t flags)
 {
     int rc = 0;
-    XSM_ASSERT_ACTION(XSM_TARGET);
     if ( f != dom_io )
-        rc = xsm_default_action(action, d, f);
+        rc = xsm_default_action(XSM_TARGET, d, f);
     if ( evaluate_nospec(t) && !rc )
-        rc = xsm_default_action(action, d, t);
+        rc = xsm_default_action(XSM_TARGET, d, t);
     return rc;
 }
 
 static __maybe_unused int dummy_mmuext_op(struct domain *d, struct domain *f)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d, f);
+    return xsm_default_action(XSM_TARGET, d, f);
 }
 
 static __maybe_unused int dummy_update_va_mapping(struct domain *d,
                                                   struct domain *f,
                                                   l1_pgentry_t pte)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d, f);
+    return xsm_default_action(XSM_TARGET, d, f);
 }
 
 static __maybe_unused int dummy_priv_mapping(struct domain *d, struct domain *t)
 {
-    XSM_ASSERT_ACTION(XSM_TARGET);
-    return xsm_default_action(action, d, t);
+    return xsm_default_action(XSM_TARGET, d, t);
 }
 
 static __maybe_unused int dummy_ioport_permission(struct domain *d, uint32_t s,
                                                   uint32_t e, uint8_t allow)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_ioport_mapping(struct domain *d, uint32_t s,
                                                uint32_t e, uint8_t allow)
 {
-    XSM_ASSERT_ACTION(XSM_HOOK);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_HOOK, current->domain, d);
 }
 
 static __maybe_unused int dummy_pmu_op (struct domain *d, unsigned int op)
 {
-    XSM_ASSERT_ACTION(XSM_OTHER);
     switch ( op )
     {
     case XENPMU_init:
@@ -729,8 +639,7 @@ static __maybe_unused int dummy_pmu_op (struct domain *d, unsigned int op)
 
 static __maybe_unused int dummy_dm_op(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, d);
 }
 
 #ifdef CONFIG_ARGO
@@ -761,7 +670,6 @@ static __maybe_unused int dummy_argo_send(const struct domain *d,
 #include <public/version.h>
 static __maybe_unused int dummy_xen_version(uint32_t op)
 {
-    XSM_ASSERT_ACTION(XSM_OTHER);
     switch ( op )
     {
     case XENVER_version:
@@ -785,6 +693,5 @@ static __maybe_unused int dummy_xen_version(uint32_t op)
 
 static __maybe_unused int dummy_domain_resource_map(struct domain *d)
 {
-    XSM_ASSERT_ACTION(XSM_DM_PRIV);
-    return xsm_default_action(action, current->domain, d);
+    return xsm_default_action(XSM_DM_PRIV, current->domain, d);
 }
Signature · 1 line
2.20.1
#33 lore raw
In reply to Daniel P. Smith
Quoted 5 lines · Daniel P. Smith

Based on feedback from 2021 Xen Developers Summit the xsm-roles RFC patch set is being split into two separate patch sets. This is the first patch set and is focused purely on the clean up and refactoring of the XSM hooks.

Quoted 7 lines

This patch set refactors the xsm_ops wrapper hooks to use the alternative_call infrastructure. Then proceeds to move and realign the headers to remove the psuedo is/is not enable implementation. The remainder of the changes are clean up and removing no longer necessary abstractions.

<snip>

 51 files changed, 1309 insertions(+), 1413 deletions(-)

The diffstat is great, but sadly CI says no.
https://gitlab.com/xen-project/patchew/xen/-/pipelines/323044913

The problem is that ARM doesn't have alternative_vcall().  Given how much of an improvement this ought to be for hypercalls, I don't want to lose the vcalls.

One option is to implement vcall() support on ARM, but that will leave new architectures (RISC-V on the way) with a heavy lift to get XSM to compile.

Instead, what we want to do is make vcall() a common interface, falling back to a plain function pointer call for architectures which don't implement the optimisation.  So something like:

1) Introduce CONFIG_HAS_VCALL, which is selected by X86 only right now
2) Introduce xen/vcall.h which uses CONFIG_HAS_VCALL to either include asm/vcall.h or provide the fallback implementation
3) Switch x86's current use over to this new interface

The iommu_vcall() is a red herring, not adequately documented, and needs to stay in some form.  Specifically, it needs to not become an alternative on ARM, even if ARM gains vcalls.  I'd be tempted to rework it in 4) to use the common vcall() by default, and leave ARM as the special case overriding the default behaviour, along with an explanation of why it isn't a vcall().

Obviously, name subject bikeshedding.  alternative_vcall() is a bit of a mouthful, and I don't think that alt_vcall() loses any salient information.

Thoughts?

~Andrew

#34 lore raw
Quoted 25 lines · Andrew Cooper
Quoted 5 lines · Daniel P. Smith

Based on feedback from 2021 Xen Developers Summit the xsm-roles RFC patch set is being split into two separate patch sets. This is the first patch set and is focused purely on the clean up and refactoring of the XSM hooks.

Quoted 7 lines

This patch set refactors the xsm_ops wrapper hooks to use the alternative_call infrastructure. Then proceeds to move and realign the headers to remove the psuedo is/is not enable implementation. The remainder of the changes are clean up and removing no longer necessary abstractions.

<snip>

 51 files changed, 1309 insertions(+), 1413 deletions(-)

The diffstat is great, but sadly CI says no.
https://gitlab.com/xen-project/patchew/xen/-/pipelines/323044913

The problem is that ARM doesn't have alternative_vcall().  Given how much of an improvement this ought to be for hypercalls, I don't want to lose the vcalls.

One option is to implement vcall() support on ARM, but that will leave new architectures (RISC-V on the way) with a heavy lift to get XSM to compile.

Quoted 7 lines

Instead, what we want to do is make vcall() a common interface, falling back to a plain function pointer call for architectures which don't implement the optimisation.  So something like:

1) Introduce CONFIG_HAS_VCALL, which is selected by X86 only right now
2) Introduce xen/vcall.h which uses CONFIG_HAS_VCALL to either include asm/vcall.h or provide the fallback implementation

A word on the suggested names: The 'v' in alternative_vcall() stands for "returning void", as opposed to alternative_call(). It's unclear to me what you see it stand for in the names you propose.

Jan

#35 lore raw
Quoted 29 lines · Jan Beulich
Quoted 28 lines · Andrew Cooper
Quoted 5 lines · Daniel P. Smith

Based on feedback from 2021 Xen Developers Summit the xsm-roles RFC patch set is being split into two separate patch sets. This is the first patch set and is focused purely on the clean up and refactoring of the XSM hooks.

Quoted 7 lines

This patch set refactors the xsm_ops wrapper hooks to use the alternative_call infrastructure. Then proceeds to move and realign the headers to remove the psuedo is/is not enable implementation. The remainder of the changes are clean up and removing no longer necessary abstractions.

<snip>

 51 files changed, 1309 insertions(+), 1413 deletions(-)

The diffstat is great, but sadly CI says no.
https://gitlab.com/xen-project/patchew/xen/-/pipelines/323044913

The problem is that ARM doesn't have alternative_vcall().  Given how much of an improvement this ought to be for hypercalls, I don't want to lose the vcalls.

One option is to implement vcall() support on ARM, but that will leave new architectures (RISC-V on the way) with a heavy lift to get XSM to compile.

Instead, what we want to do is make vcall() a common interface, falling back to a plain function pointer call for architectures which don't implement the optimisation.  So something like:

Quoted 6 lines
Quoted 3 lines

1) Introduce CONFIG_HAS_VCALL, which is selected by X86 only right now
2) Introduce xen/vcall.h which uses CONFIG_HAS_VCALL to either include asm/vcall.h or provide the fallback implementation

A word on the suggested names: The 'v' in alternative_vcall() stands for "returning void", as opposed to alternative_call(). It's unclear to me what you see it stand for in the names you propose.

Urgh - yet another reason to prefer the Linux static_call() infrastructure.

Would a general alt_call name be acceptable?

~Andrew

#36 lore raw
Quoted 30 lines · Andrew Cooper
Quoted 29 lines · Jan Beulich
Quoted 28 lines · Andrew Cooper
Quoted 5 lines · Daniel P. Smith

Based on feedback from 2021 Xen Developers Summit the xsm-roles RFC patch set is being split into two separate patch sets. This is the first patch set and is focused purely on the clean up and refactoring of the XSM hooks.

Quoted 7 lines

This patch set refactors the xsm_ops wrapper hooks to use the alternative_call infrastructure. Then proceeds to move and realign the headers to remove the psuedo is/is not enable implementation. The remainder of the changes are clean up and removing no longer necessary abstractions.

<snip>

 51 files changed, 1309 insertions(+), 1413 deletions(-)

The diffstat is great, but sadly CI says no.
https://gitlab.com/xen-project/patchew/xen/-/pipelines/323044913

The problem is that ARM doesn't have alternative_vcall().  Given how much of an improvement this ought to be for hypercalls, I don't want to lose the vcalls.

One option is to implement vcall() support on ARM, but that will leave new architectures (RISC-V on the way) with a heavy lift to get XSM to compile.

Instead, what we want to do is make vcall() a common interface, falling back to a plain function pointer call for architectures which don't implement the optimisation.  So something like:

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines

1) Introduce CONFIG_HAS_VCALL, which is selected by X86 only right now
2) Introduce xen/vcall.h which uses CONFIG_HAS_VCALL to either include asm/vcall.h or provide the fallback implementation

A word on the suggested names: The 'v' in alternative_vcall() stands for "returning void", as opposed to alternative_call(). It's unclear to me what you see it stand for in the names you propose.

Urgh - yet another reason to prefer the Linux static_call() infrastructure.

Which iirc wasn't there yet when I wrote ours.

Quoted 1 line

Would a general alt_call name be acceptable?

Well, it seemed a little terse to me at the time, but I'm not opposed. There's hardly anything else the "alt" there could stand for, I guess.

Jan

#37 lore raw
In reply to Andrew Cooper
Quoted 41 lines · Andrew Cooper
Quoted 5 lines · Daniel P. Smith

Based on feedback from 2021 Xen Developers Summit the xsm-roles RFC patch set is being split into two separate patch sets. This is the first patch set and is focused purely on the clean up and refactoring of the XSM hooks.

Quoted 7 lines

This patch set refactors the xsm_ops wrapper hooks to use the alternative_call infrastructure. Then proceeds to move and realign the headers to remove the psuedo is/is not enable implementation. The remainder of the changes are clean up and removing no longer necessary abstractions.

<snip>

 51 files changed, 1309 insertions(+), 1413 deletions(-)

The diffstat is great, but sadly CI says no.
https://gitlab.com/xen-project/patchew/xen/-/pipelines/323044913

The problem is that ARM doesn't have alternative_vcall().  Given how much of an improvement this ought to be for hypercalls, I don't want to lose the vcalls.

One option is to implement vcall() support on ARM, but that will leave new architectures (RISC-V on the way) with a heavy lift to get XSM to compile.

Instead, what we want to do is make vcall() a common interface, falling back to a plain function pointer call for architectures which don't implement the optimisation.  So something like:

1) Introduce CONFIG_HAS_VCALL, which is selected by X86 only right now
2) Introduce xen/vcall.h which uses CONFIG_HAS_VCALL to either include asm/vcall.h or provide the fallback implementation
3) Switch x86's current use over to this new interface

The iommu_vcall() is a red herring, not adequately documented, and needs to stay in some form.  Specifically, it needs to not become an alternative on ARM, even if ARM gains vcalls.  I'd be tempted to rework it in 4) to use the common vcall() by default, and leave ARM as the special case overriding the default behaviour, along with an explanation of why it isn't a vcall().

Quoted 4 lines

Obviously, name subject bikeshedding.  alternative_vcall() is a bit of a mouthful, and I don't think that alt_vcall() loses any salient information.

Thoughts?

I can look at spinning an attempt at what you are suggesting and submitting.

v/r,
dps