* [PATCH v5 00/17] Mem_event and mem_access for ARM
@ 2014-09-10 13:28 Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 01/17] xen: Relocate mem_access and mem_event into common Tamas K Lengyel
` (17 more replies)
0 siblings, 18 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
The ARM virtualization extension provides 2-stage paging, a similar mechanisms
to Intel's EPT, which can be used to trace the memory accesses performed by
the guest systems. This series moves the mem_access and mem_event codebase
into Xen common, performs some code cleanup and architecture specific division
of components, then sets up the necessary infrastructure in the ARM code
to deliver the event on R/W/X traps. Finally, we turn on the compilation of
mem_access and mem_event on ARM and perform the necessary changes to the tools side.
This version of the series has been fully tested and is functional on an
Arndale board.
This PATCH series is also available at:
https://github.com/tklengyel/xen/tree/arm_memaccess5
Tamas K Lengyel (17):
xen: Relocate mem_access and mem_event into common.
xen: Relocate p2m_mem_access_resume to mem_access common
xen: Relocate struct npfec definition into common
xen: Relocate mem_event_op domctl and access_op memop into common.
xen/mem_event: Clean out superfluous white-spaces
xen/mem_event: Relax error condition on debug builds
xen/mem_event: Abstract architecture specific sanity checks
xen/mem_access: Abstract architecture specific sanity check
xen/arm: p2m type definitions and changes
xen/arm: Add set access required domctl
xen/arm: Implement domain_get_maximum_gpfn
xen/arm: Data abort exception (R/W) mem_events.
xen/arm: Instruction prefetch abort (X) mem_event handling
xen/arm: Enable the compilation of mem_access and mem_event on ARM.
xen: Extend getdomaininfo to return the domain's max_gpfn
tools/libxc: Allocate magic page for mem access on ARM
tools/tests: Enable xen-access on ARM
MAINTAINERS | 6 +
config/arm32.mk | 1 +
config/arm64.mk | 1 +
config/x86_32.mk | 2 +
config/x86_64.mk | 2 +
tools/libxc/xc_dom_arm.c | 6 +-
tools/tests/xen-access/Makefile | 9 +-
tools/tests/xen-access/xen-access.c | 60 ++-
xen/Rules.mk | 1 +
xen/arch/arm/domctl.c | 13 +
xen/arch/arm/mm.c | 2 +-
xen/arch/arm/p2m.c | 466 ++++++++++++++++++++---
xen/arch/arm/traps.c | 73 +++-
xen/arch/x86/domctl.c | 10 +-
xen/arch/x86/hvm/hvm.c | 61 +--
xen/arch/x86/mm/Makefile | 2 -
xen/arch/x86/mm/hap/nested_ept.c | 6 +-
xen/arch/x86/mm/hap/nested_hap.c | 6 +-
xen/arch/x86/mm/mem_access.c | 133 -------
xen/arch/x86/mm/mem_event.c | 705 -----------------------------------
xen/arch/x86/mm/mem_paging.c | 2 +-
xen/arch/x86/mm/mem_sharing.c | 4 +-
xen/arch/x86/mm/p2m-pod.c | 8 +-
xen/arch/x86/mm/p2m-pt.c | 10 +-
xen/arch/x86/mm/p2m.c | 32 +-
xen/arch/x86/x86_64/compat/mm.c | 8 +-
xen/arch/x86/x86_64/mm.c | 8 +-
xen/common/Makefile | 2 +
xen/common/compat/memory.c | 5 +
xen/common/domain.c | 1 +
xen/common/domctl.c | 8 +
xen/common/mem_access.c | 157 ++++++++
xen/common/mem_event.c | 723 ++++++++++++++++++++++++++++++++++++
xen/common/memory.c | 68 ++++
xen/include/asm-arm/mm.h | 1 -
xen/include/asm-arm/p2m.h | 109 +++++-
xen/include/asm-arm/processor.h | 70 +++-
xen/include/asm-x86/config.h | 6 +
xen/include/asm-x86/hvm/hvm.h | 8 +-
xen/include/asm-x86/mem_access.h | 39 --
xen/include/asm-x86/mem_event.h | 82 ----
xen/include/asm-x86/mm.h | 23 --
xen/include/asm-x86/p2m.h | 14 +-
xen/include/public/domctl.h | 1 +
xen/include/xen/mem_access.h | 63 ++++
xen/include/xen/mem_event.h | 143 +++++++
xen/include/xen/mm.h | 27 ++
xen/include/xsm/dummy.h | 26 +-
xen/include/xsm/xsm.h | 29 +-
xen/xsm/dummy.c | 7 +-
xen/xsm/flask/hooks.c | 33 +-
51 files changed, 2031 insertions(+), 1251 deletions(-)
delete mode 100644 xen/arch/x86/mm/mem_access.c
delete mode 100644 xen/arch/x86/mm/mem_event.c
create mode 100644 xen/common/mem_access.c
create mode 100644 xen/common/mem_event.c
delete mode 100644 xen/include/asm-x86/mem_access.h
delete mode 100644 xen/include/asm-x86/mem_event.h
create mode 100644 xen/include/xen/mem_access.h
create mode 100644 xen/include/xen/mem_event.h
--
2.1.0
^ permalink raw reply [flat|nested] 48+ messages in thread
* [PATCH v5 01/17] xen: Relocate mem_access and mem_event into common.
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 02/17] xen: Relocate p2m_mem_access_resume to mem_access common Tamas K Lengyel
` (16 subsequent siblings)
17 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
In preparation to add support for ARM LPAE mem_event, relocate mem_access,
mem_event and auxiliary functions into common Xen code.
This patch makes no functional changes to the X86 side, for ARM mem_event
and mem_access functions are just defined as placeholder stubs, and are
actually enabled later in the series.
Edits that are only header path adjustments:
xen/arch/x86/domctl.c
xen/arch/x86/mm/hap/nested_ept.c
xen/arch/x86/mm/hap/nested_hap.c
xen/arch/x86/mm/mem_paging.c
xen/arch/x86/mm/mem_sharing.c
xen/arch/x86/mm/p2m-pod.c
xen/arch/x86/mm/p2m-pt.c
xen/arch/x86/mm/p2m.c
xen/arch/x86/x86_64/compat/mm.c
xen/arch/x86/x86_64/mm.c
Makefile adjustments for new/removed code:
xen/common/Makefile
xen/arch/x86/mm/Makefile
Relocated prepare_ring_for_helper and destroy_ring_for_helper functions:
xen/include/xen/mm.h
xen/common/memory.c
xen/include/asm-x86/hvm/hvm.h
xen/arch/x86/hvm/hvm.c
Code movement of mem_event and mem_access:
xen/arch/x86/mm/mem_access.c -> xen/common/mem_access.c
xen/arch/x86/mm/mem_event.c -> xen/common/mem_event.c
xen/include/asm-x86/mem_access.h -> xen/include/xen/mem_access.h
xen/include/asm-x86/mem_event.h -> xen/include/xen/mem_event.h
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
Acked-by: Tim Deegan <tim@xen.org>
---
v5: Make <xen/mem-event.h> include <xen/sched.h> by default.
Style fix with grouping of #includes.
v4: Make <xen/mem_access.h> include <public/memory.h> by default.
v3: Replace asm/domain.h with xen/sched.h in mem_event.c to better
accomodate for the new code location.
Replace #ifdef CONFIG_X86 wrappers with HAS_MEM_ACCESS flags.
v2: Update MAINTAINERS.
More descriptive commit message to aid in the review process.
---
MAINTAINERS | 6 +
xen/Rules.mk | 1 +
xen/arch/x86/Rules.mk | 1 +
xen/arch/x86/domctl.c | 2 +-
xen/arch/x86/hvm/hvm.c | 61 +---
xen/arch/x86/mm/Makefile | 2 -
xen/arch/x86/mm/hap/nested_ept.c | 6 +-
xen/arch/x86/mm/hap/nested_hap.c | 6 +-
xen/arch/x86/mm/mem_access.c | 133 --------
xen/arch/x86/mm/mem_event.c | 705 ---------------------------------------
xen/arch/x86/mm/mem_paging.c | 2 +-
xen/arch/x86/mm/mem_sharing.c | 4 +-
xen/arch/x86/mm/p2m-pod.c | 8 +-
xen/arch/x86/mm/p2m-pt.c | 10 +-
xen/arch/x86/mm/p2m.c | 8 +-
xen/arch/x86/x86_64/compat/mm.c | 4 +-
xen/arch/x86/x86_64/mm.c | 4 +-
xen/common/Makefile | 2 +
xen/common/domain.c | 1 +
xen/common/mem_access.c | 133 ++++++++
xen/common/mem_event.c | 705 +++++++++++++++++++++++++++++++++++++++
xen/common/memory.c | 63 ++++
xen/include/asm-arm/mm.h | 1 -
xen/include/asm-x86/config.h | 3 +
xen/include/asm-x86/hvm/hvm.h | 6 -
xen/include/asm-x86/mem_access.h | 39 ---
xen/include/asm-x86/mem_event.h | 82 -----
xen/include/asm-x86/mm.h | 2 -
xen/include/xen/mem_access.h | 60 ++++
xen/include/xen/mem_event.h | 143 ++++++++
xen/include/xen/mm.h | 6 +
31 files changed, 1153 insertions(+), 1056 deletions(-)
delete mode 100644 xen/arch/x86/mm/mem_access.c
delete mode 100644 xen/arch/x86/mm/mem_event.c
create mode 100644 xen/common/mem_access.c
create mode 100644 xen/common/mem_event.c
delete mode 100644 xen/include/asm-x86/mem_access.h
delete mode 100644 xen/include/asm-x86/mem_event.h
create mode 100644 xen/include/xen/mem_access.h
create mode 100644 xen/include/xen/mem_event.h
diff --git a/MAINTAINERS b/MAINTAINERS
index 266e47b..f659180 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -337,6 +337,12 @@ F: xen/arch/x86/mm/mem_sharing.c
F: xen/arch/x86/mm/mem_paging.c
F: tools/memshr
+MEMORY EVENT AND ACCESS
+M: Tim Deegan <tim@xen.org>
+S: Supported
+F: xen/common/mem_event.c
+F: xen/common/mem_access.c
+
XENTRACE
M: George Dunlap <george.dunlap@eu.citrix.com>
S: Supported
diff --git a/xen/Rules.mk b/xen/Rules.mk
index b49f3c8..dc15b09 100644
--- a/xen/Rules.mk
+++ b/xen/Rules.mk
@@ -57,6 +57,7 @@ CFLAGS-$(HAS_ACPI) += -DHAS_ACPI
CFLAGS-$(HAS_GDBSX) += -DHAS_GDBSX
CFLAGS-$(HAS_PASSTHROUGH) += -DHAS_PASSTHROUGH
CFLAGS-$(HAS_DEVICE_TREE) += -DHAS_DEVICE_TREE
+CFLAGS-$(HAS_MEM_ACCESS) += -DHAS_MEM_ACCESS
CFLAGS-$(HAS_PCI) += -DHAS_PCI
CFLAGS-$(HAS_IOPORTS) += -DHAS_IOPORTS
CFLAGS-$(frame_pointer) += -fno-omit-frame-pointer -DCONFIG_FRAME_POINTER
diff --git a/xen/arch/x86/Rules.mk b/xen/arch/x86/Rules.mk
index 576985e..bd4e342 100644
--- a/xen/arch/x86/Rules.mk
+++ b/xen/arch/x86/Rules.mk
@@ -12,6 +12,7 @@ HAS_NS16550 := y
HAS_EHCI := y
HAS_KEXEC := y
HAS_GDBSX := y
+HAS_MEM_ACCESS := y
xenoprof := y
#
diff --git a/xen/arch/x86/domctl.c b/xen/arch/x86/domctl.c
index 7a5de43..26a3ea1 100644
--- a/xen/arch/x86/domctl.c
+++ b/xen/arch/x86/domctl.c
@@ -30,7 +30,7 @@
#include <xen/hypercall.h> /* for arch_do_domctl */
#include <xsm/xsm.h>
#include <xen/iommu.h>
-#include <asm/mem_event.h>
+#include <xen/mem_event.h>
#include <public/mem_event.h>
#include <asm/mem_sharing.h>
#include <asm/xstate.h>
diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index 8d905d3..9d37743 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -63,8 +63,8 @@
#include <public/hvm/ioreq.h>
#include <public/version.h>
#include <public/memory.h>
-#include <asm/mem_event.h>
-#include <asm/mem_access.h>
+#include <xen/mem_event.h>
+#include <xen/mem_access.h>
#include <public/mem_event.h>
#include <xen/rangeset.h>
#include <public/arch-x86/cpuid.h>
@@ -484,19 +484,6 @@ static void hvm_free_ioreq_gmfn(struct domain *d, unsigned long gmfn)
clear_bit(i, &d->arch.hvm_domain.ioreq_gmfn.mask);
}
-void destroy_ring_for_helper(
- void **_va, struct page_info *page)
-{
- void *va = *_va;
-
- if ( va != NULL )
- {
- unmap_domain_page_global(va);
- put_page_and_type(page);
- *_va = NULL;
- }
-}
-
static void hvm_unmap_ioreq_page(struct hvm_ioreq_server *s, bool_t buf)
{
struct hvm_ioreq_page *iorp = buf ? &s->bufioreq : &s->ioreq;
@@ -504,50 +491,6 @@ static void hvm_unmap_ioreq_page(struct hvm_ioreq_server *s, bool_t buf)
destroy_ring_for_helper(&iorp->va, iorp->page);
}
-int prepare_ring_for_helper(
- struct domain *d, unsigned long gmfn, struct page_info **_page,
- void **_va)
-{
- struct page_info *page;
- p2m_type_t p2mt;
- void *va;
-
- page = get_page_from_gfn(d, gmfn, &p2mt, P2M_UNSHARE);
- if ( p2m_is_paging(p2mt) )
- {
- if ( page )
- put_page(page);
- p2m_mem_paging_populate(d, gmfn);
- return -ENOENT;
- }
- if ( p2m_is_shared(p2mt) )
- {
- if ( page )
- put_page(page);
- return -ENOENT;
- }
- if ( !page )
- return -EINVAL;
-
- if ( !get_page_type(page, PGT_writable_page) )
- {
- put_page(page);
- return -EINVAL;
- }
-
- va = __map_domain_page_global(page);
- if ( va == NULL )
- {
- put_page_and_type(page);
- return -ENOMEM;
- }
-
- *_va = va;
- *_page = page;
-
- return 0;
-}
-
static int hvm_map_ioreq_page(
struct hvm_ioreq_server *s, bool_t buf, unsigned long gmfn)
{
diff --git a/xen/arch/x86/mm/Makefile b/xen/arch/x86/mm/Makefile
index 73dcdf4..ed4b1f8 100644
--- a/xen/arch/x86/mm/Makefile
+++ b/xen/arch/x86/mm/Makefile
@@ -6,10 +6,8 @@ obj-y += p2m.o p2m-pt.o p2m-ept.o p2m-pod.o
obj-y += guest_walk_2.o
obj-y += guest_walk_3.o
obj-$(x86_64) += guest_walk_4.o
-obj-$(x86_64) += mem_event.o
obj-$(x86_64) += mem_paging.o
obj-$(x86_64) += mem_sharing.o
-obj-$(x86_64) += mem_access.o
guest_walk_%.o: guest_walk.c Makefile
$(CC) $(CFLAGS) -DGUEST_PAGING_LEVELS=$* -c $< -o $@
diff --git a/xen/arch/x86/mm/hap/nested_ept.c b/xen/arch/x86/mm/hap/nested_ept.c
index 0d044bc..cbbc4e9 100644
--- a/xen/arch/x86/mm/hap/nested_ept.c
+++ b/xen/arch/x86/mm/hap/nested_ept.c
@@ -17,14 +17,14 @@
* this program; if not, write to the Free Software Foundation, Inc., 59 Temple
* Place - Suite 330, Boston, MA 02111-1307 USA.
*/
+#include <xen/mem_event.h>
+#include <xen/event.h>
+#include <public/mem_event.h>
#include <asm/domain.h>
#include <asm/page.h>
#include <asm/paging.h>
#include <asm/p2m.h>
-#include <asm/mem_event.h>
-#include <public/mem_event.h>
#include <asm/mem_sharing.h>
-#include <xen/event.h>
#include <asm/hap.h>
#include <asm/hvm/support.h>
diff --git a/xen/arch/x86/mm/hap/nested_hap.c b/xen/arch/x86/mm/hap/nested_hap.c
index 137a87c..a4bb835 100644
--- a/xen/arch/x86/mm/hap/nested_hap.c
+++ b/xen/arch/x86/mm/hap/nested_hap.c
@@ -19,14 +19,14 @@
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
+#include <xen/mem_event.h>
+#include <xen/event.h>
+#include <public/mem_event.h>
#include <asm/domain.h>
#include <asm/page.h>
#include <asm/paging.h>
#include <asm/p2m.h>
-#include <asm/mem_event.h>
-#include <public/mem_event.h>
#include <asm/mem_sharing.h>
-#include <xen/event.h>
#include <asm/hap.h>
#include <asm/hvm/support.h>
diff --git a/xen/arch/x86/mm/mem_access.c b/xen/arch/x86/mm/mem_access.c
deleted file mode 100644
index e8465a5..0000000
--- a/xen/arch/x86/mm/mem_access.c
+++ /dev/null
@@ -1,133 +0,0 @@
-/******************************************************************************
- * arch/x86/mm/mem_access.c
- *
- * Memory access support.
- *
- * Copyright (c) 2011 Virtuata, Inc.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2 of the License, or
- * (at your option) any later version.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
- */
-
-
-#include <xen/sched.h>
-#include <xen/guest_access.h>
-#include <xen/hypercall.h>
-#include <asm/p2m.h>
-#include <asm/mem_event.h>
-#include <xsm/xsm.h>
-
-
-int mem_access_memop(unsigned long cmd,
- XEN_GUEST_HANDLE_PARAM(xen_mem_access_op_t) arg)
-{
- long rc;
- xen_mem_access_op_t mao;
- struct domain *d;
-
- if ( copy_from_guest(&mao, arg, 1) )
- return -EFAULT;
-
- rc = rcu_lock_live_remote_domain_by_id(mao.domid, &d);
- if ( rc )
- return rc;
-
- rc = -EINVAL;
- if ( !is_hvm_domain(d) )
- goto out;
-
- rc = xsm_mem_event_op(XSM_DM_PRIV, d, XENMEM_access_op);
- if ( rc )
- goto out;
-
- rc = -ENODEV;
- if ( unlikely(!d->mem_event->access.ring_page) )
- goto out;
-
- switch ( mao.op )
- {
- case XENMEM_access_op_resume:
- p2m_mem_access_resume(d);
- rc = 0;
- break;
-
- case XENMEM_access_op_set_access:
- {
- unsigned long start_iter = cmd & ~MEMOP_CMD_MASK;
-
- rc = -EINVAL;
- if ( (mao.pfn != ~0ull) &&
- (mao.nr < start_iter ||
- ((mao.pfn + mao.nr - 1) < mao.pfn) ||
- ((mao.pfn + mao.nr - 1) > domain_get_maximum_gpfn(d))) )
- break;
-
- rc = p2m_set_mem_access(d, mao.pfn, mao.nr, start_iter,
- MEMOP_CMD_MASK, mao.access);
- if ( rc > 0 )
- {
- ASSERT(!(rc & MEMOP_CMD_MASK));
- rc = hypercall_create_continuation(__HYPERVISOR_memory_op, "lh",
- XENMEM_access_op | rc, arg);
- }
- break;
- }
-
- case XENMEM_access_op_get_access:
- {
- xenmem_access_t access;
-
- rc = -EINVAL;
- if ( (mao.pfn > domain_get_maximum_gpfn(d)) && mao.pfn != ~0ull )
- break;
-
- rc = p2m_get_mem_access(d, mao.pfn, &access);
- if ( rc != 0 )
- break;
-
- mao.access = access;
- rc = __copy_field_to_guest(arg, &mao, access) ? -EFAULT : 0;
-
- break;
- }
-
- default:
- rc = -ENOSYS;
- break;
- }
-
- out:
- rcu_unlock_domain(d);
- return rc;
-}
-
-int mem_access_send_req(struct domain *d, mem_event_request_t *req)
-{
- int rc = mem_event_claim_slot(d, &d->mem_event->access);
- if ( rc < 0 )
- return rc;
-
- mem_event_put_request(d, &d->mem_event->access, req);
-
- return 0;
-}
-
-/*
- * Local variables:
- * mode: C
- * c-file-style: "BSD"
- * c-basic-offset: 4
- * indent-tabs-mode: nil
- * End:
- */
diff --git a/xen/arch/x86/mm/mem_event.c b/xen/arch/x86/mm/mem_event.c
deleted file mode 100644
index ba7e71e..0000000
--- a/xen/arch/x86/mm/mem_event.c
+++ /dev/null
@@ -1,705 +0,0 @@
-/******************************************************************************
- * arch/x86/mm/mem_event.c
- *
- * Memory event support.
- *
- * Copyright (c) 2009 Citrix Systems, Inc. (Patrick Colp)
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2 of the License, or
- * (at your option) any later version.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
- */
-
-
-#include <asm/domain.h>
-#include <xen/event.h>
-#include <xen/wait.h>
-#include <asm/p2m.h>
-#include <asm/mem_event.h>
-#include <asm/mem_paging.h>
-#include <asm/mem_access.h>
-#include <asm/mem_sharing.h>
-#include <xsm/xsm.h>
-
-/* for public/io/ring.h macros */
-#define xen_mb() mb()
-#define xen_rmb() rmb()
-#define xen_wmb() wmb()
-
-#define mem_event_ring_lock_init(_med) spin_lock_init(&(_med)->ring_lock)
-#define mem_event_ring_lock(_med) spin_lock(&(_med)->ring_lock)
-#define mem_event_ring_unlock(_med) spin_unlock(&(_med)->ring_lock)
-
-static int mem_event_enable(
- struct domain *d,
- xen_domctl_mem_event_op_t *mec,
- struct mem_event_domain *med,
- int pause_flag,
- int param,
- xen_event_channel_notification_t notification_fn)
-{
- int rc;
- unsigned long ring_gfn = d->arch.hvm_domain.params[param];
-
- /* Only one helper at a time. If the helper crashed,
- * the ring is in an undefined state and so is the guest.
- */
- if ( med->ring_page )
- return -EBUSY;
-
- /* The parameter defaults to zero, and it should be
- * set to something */
- if ( ring_gfn == 0 )
- return -ENOSYS;
-
- mem_event_ring_lock_init(med);
- mem_event_ring_lock(med);
-
- rc = prepare_ring_for_helper(d, ring_gfn, &med->ring_pg_struct,
- &med->ring_page);
- if ( rc < 0 )
- goto err;
-
- /* Set the number of currently blocked vCPUs to 0. */
- med->blocked = 0;
-
- /* Allocate event channel */
- rc = alloc_unbound_xen_event_channel(d->vcpu[0],
- current->domain->domain_id,
- notification_fn);
- if ( rc < 0 )
- goto err;
-
- med->xen_port = mec->port = rc;
-
- /* Prepare ring buffer */
- FRONT_RING_INIT(&med->front_ring,
- (mem_event_sring_t *)med->ring_page,
- PAGE_SIZE);
-
- /* Save the pause flag for this particular ring. */
- med->pause_flag = pause_flag;
-
- /* Initialize the last-chance wait queue. */
- init_waitqueue_head(&med->wq);
-
- mem_event_ring_unlock(med);
- return 0;
-
- err:
- destroy_ring_for_helper(&med->ring_page,
- med->ring_pg_struct);
- mem_event_ring_unlock(med);
-
- return rc;
-}
-
-static unsigned int mem_event_ring_available(struct mem_event_domain *med)
-{
- int avail_req = RING_FREE_REQUESTS(&med->front_ring);
- avail_req -= med->target_producers;
- avail_req -= med->foreign_producers;
-
- BUG_ON(avail_req < 0);
-
- return avail_req;
-}
-
-/*
- * mem_event_wake_blocked() will wakeup vcpus waiting for room in the
- * ring. These vCPUs were paused on their way out after placing an event,
- * but need to be resumed where the ring is capable of processing at least
- * one event from them.
- */
-static void mem_event_wake_blocked(struct domain *d, struct mem_event_domain *med)
-{
- struct vcpu *v;
- int online = d->max_vcpus;
- unsigned int avail_req = mem_event_ring_available(med);
-
- if ( avail_req == 0 || med->blocked == 0 )
- return;
-
- /*
- * We ensure that we only have vCPUs online if there are enough free slots
- * for their memory events to be processed. This will ensure that no
- * memory events are lost (due to the fact that certain types of events
- * cannot be replayed, we need to ensure that there is space in the ring
- * for when they are hit).
- * See comment below in mem_event_put_request().
- */
- for_each_vcpu ( d, v )
- if ( test_bit(med->pause_flag, &v->pause_flags) )
- online--;
-
- ASSERT(online == (d->max_vcpus - med->blocked));
-
- /* We remember which vcpu last woke up to avoid scanning always linearly
- * from zero and starving higher-numbered vcpus under high load */
- if ( d->vcpu )
- {
- int i, j, k;
-
- for (i = med->last_vcpu_wake_up + 1, j = 0; j < d->max_vcpus; i++, j++)
- {
- k = i % d->max_vcpus;
- v = d->vcpu[k];
- if ( !v )
- continue;
-
- if ( !(med->blocked) || online >= avail_req )
- break;
-
- if ( test_and_clear_bit(med->pause_flag, &v->pause_flags) )
- {
- vcpu_unpause(v);
- online++;
- med->blocked--;
- med->last_vcpu_wake_up = k;
- }
- }
- }
-}
-
-/*
- * In the event that a vCPU attempted to place an event in the ring and
- * was unable to do so, it is queued on a wait queue. These are woken as
- * needed, and take precedence over the blocked vCPUs.
- */
-static void mem_event_wake_queued(struct domain *d, struct mem_event_domain *med)
-{
- unsigned int avail_req = mem_event_ring_available(med);
-
- if ( avail_req > 0 )
- wake_up_nr(&med->wq, avail_req);
-}
-
-/*
- * mem_event_wake() will wakeup all vcpus waiting for the ring to
- * become available. If we have queued vCPUs, they get top priority. We
- * are guaranteed that they will go through code paths that will eventually
- * call mem_event_wake() again, ensuring that any blocked vCPUs will get
- * unpaused once all the queued vCPUs have made it through.
- */
-void mem_event_wake(struct domain *d, struct mem_event_domain *med)
-{
- if (!list_empty(&med->wq.list))
- mem_event_wake_queued(d, med);
- else
- mem_event_wake_blocked(d, med);
-}
-
-static int mem_event_disable(struct domain *d, struct mem_event_domain *med)
-{
- if ( med->ring_page )
- {
- struct vcpu *v;
-
- mem_event_ring_lock(med);
-
- if ( !list_empty(&med->wq.list) )
- {
- mem_event_ring_unlock(med);
- return -EBUSY;
- }
-
- /* Free domU's event channel and leave the other one unbound */
- free_xen_event_channel(d->vcpu[0], med->xen_port);
-
- /* Unblock all vCPUs */
- for_each_vcpu ( d, v )
- {
- if ( test_and_clear_bit(med->pause_flag, &v->pause_flags) )
- {
- vcpu_unpause(v);
- med->blocked--;
- }
- }
-
- destroy_ring_for_helper(&med->ring_page,
- med->ring_pg_struct);
- mem_event_ring_unlock(med);
- }
-
- return 0;
-}
-
-static inline void mem_event_release_slot(struct domain *d,
- struct mem_event_domain *med)
-{
- /* Update the accounting */
- if ( current->domain == d )
- med->target_producers--;
- else
- med->foreign_producers--;
-
- /* Kick any waiters */
- mem_event_wake(d, med);
-}
-
-/*
- * mem_event_mark_and_pause() tags vcpu and put it to sleep.
- * The vcpu will resume execution in mem_event_wake_waiters().
- */
-void mem_event_mark_and_pause(struct vcpu *v, struct mem_event_domain *med)
-{
- if ( !test_and_set_bit(med->pause_flag, &v->pause_flags) )
- {
- vcpu_pause_nosync(v);
- med->blocked++;
- }
-}
-
-/*
- * This must be preceded by a call to claim_slot(), and is guaranteed to
- * succeed. As a side-effect however, the vCPU may be paused if the ring is
- * overly full and its continued execution would cause stalling and excessive
- * waiting. The vCPU will be automatically unpaused when the ring clears.
- */
-void mem_event_put_request(struct domain *d,
- struct mem_event_domain *med,
- mem_event_request_t *req)
-{
- mem_event_front_ring_t *front_ring;
- int free_req;
- unsigned int avail_req;
- RING_IDX req_prod;
-
- if ( current->domain != d )
- {
- req->flags |= MEM_EVENT_FLAG_FOREIGN;
- ASSERT( !(req->flags & MEM_EVENT_FLAG_VCPU_PAUSED) );
- }
-
- mem_event_ring_lock(med);
-
- /* Due to the reservations, this step must succeed. */
- front_ring = &med->front_ring;
- free_req = RING_FREE_REQUESTS(front_ring);
- ASSERT(free_req > 0);
-
- /* Copy request */
- req_prod = front_ring->req_prod_pvt;
- memcpy(RING_GET_REQUEST(front_ring, req_prod), req, sizeof(*req));
- req_prod++;
-
- /* Update ring */
- front_ring->req_prod_pvt = req_prod;
- RING_PUSH_REQUESTS(front_ring);
-
- /* We've actually *used* our reservation, so release the slot. */
- mem_event_release_slot(d, med);
-
- /* Give this vCPU a black eye if necessary, on the way out.
- * See the comments above wake_blocked() for more information
- * on how this mechanism works to avoid waiting. */
- avail_req = mem_event_ring_available(med);
- if( current->domain == d && avail_req < d->max_vcpus )
- mem_event_mark_and_pause(current, med);
-
- mem_event_ring_unlock(med);
-
- notify_via_xen_event_channel(d, med->xen_port);
-}
-
-int mem_event_get_response(struct domain *d, struct mem_event_domain *med, mem_event_response_t *rsp)
-{
- mem_event_front_ring_t *front_ring;
- RING_IDX rsp_cons;
-
- mem_event_ring_lock(med);
-
- front_ring = &med->front_ring;
- rsp_cons = front_ring->rsp_cons;
-
- if ( !RING_HAS_UNCONSUMED_RESPONSES(front_ring) )
- {
- mem_event_ring_unlock(med);
- return 0;
- }
-
- /* Copy response */
- memcpy(rsp, RING_GET_RESPONSE(front_ring, rsp_cons), sizeof(*rsp));
- rsp_cons++;
-
- /* Update ring */
- front_ring->rsp_cons = rsp_cons;
- front_ring->sring->rsp_event = rsp_cons + 1;
-
- /* Kick any waiters -- since we've just consumed an event,
- * there may be additional space available in the ring. */
- mem_event_wake(d, med);
-
- mem_event_ring_unlock(med);
-
- return 1;
-}
-
-void mem_event_cancel_slot(struct domain *d, struct mem_event_domain *med)
-{
- mem_event_ring_lock(med);
- mem_event_release_slot(d, med);
- mem_event_ring_unlock(med);
-}
-
-static int mem_event_grab_slot(struct mem_event_domain *med, int foreign)
-{
- unsigned int avail_req;
-
- if ( !med->ring_page )
- return -ENOSYS;
-
- mem_event_ring_lock(med);
-
- avail_req = mem_event_ring_available(med);
- if ( avail_req == 0 )
- {
- mem_event_ring_unlock(med);
- return -EBUSY;
- }
-
- if ( !foreign )
- med->target_producers++;
- else
- med->foreign_producers++;
-
- mem_event_ring_unlock(med);
-
- return 0;
-}
-
-/* Simple try_grab wrapper for use in the wait_event() macro. */
-static int mem_event_wait_try_grab(struct mem_event_domain *med, int *rc)
-{
- *rc = mem_event_grab_slot(med, 0);
- return *rc;
-}
-
-/* Call mem_event_grab_slot() until the ring doesn't exist, or is available. */
-static int mem_event_wait_slot(struct mem_event_domain *med)
-{
- int rc = -EBUSY;
- wait_event(med->wq, mem_event_wait_try_grab(med, &rc) != -EBUSY);
- return rc;
-}
-
-bool_t mem_event_check_ring(struct mem_event_domain *med)
-{
- return (med->ring_page != NULL);
-}
-
-/*
- * Determines whether or not the current vCPU belongs to the target domain,
- * and calls the appropriate wait function. If it is a guest vCPU, then we
- * use mem_event_wait_slot() to reserve a slot. As long as there is a ring,
- * this function will always return 0 for a guest. For a non-guest, we check
- * for space and return -EBUSY if the ring is not available.
- *
- * Return codes: -ENOSYS: the ring is not yet configured
- * -EBUSY: the ring is busy
- * 0: a spot has been reserved
- *
- */
-int __mem_event_claim_slot(struct domain *d, struct mem_event_domain *med,
- bool_t allow_sleep)
-{
- if ( (current->domain == d) && allow_sleep )
- return mem_event_wait_slot(med);
- else
- return mem_event_grab_slot(med, (current->domain != d));
-}
-
-/* Registered with Xen-bound event channel for incoming notifications. */
-static void mem_paging_notification(struct vcpu *v, unsigned int port)
-{
- if ( likely(v->domain->mem_event->paging.ring_page != NULL) )
- p2m_mem_paging_resume(v->domain);
-}
-
-/* Registered with Xen-bound event channel for incoming notifications. */
-static void mem_access_notification(struct vcpu *v, unsigned int port)
-{
- if ( likely(v->domain->mem_event->access.ring_page != NULL) )
- p2m_mem_access_resume(v->domain);
-}
-
-/* Registered with Xen-bound event channel for incoming notifications. */
-static void mem_sharing_notification(struct vcpu *v, unsigned int port)
-{
- if ( likely(v->domain->mem_event->share.ring_page != NULL) )
- mem_sharing_sharing_resume(v->domain);
-}
-
-int do_mem_event_op(int op, uint32_t domain, void *arg)
-{
- int ret;
- struct domain *d;
-
- ret = rcu_lock_live_remote_domain_by_id(domain, &d);
- if ( ret )
- return ret;
-
- ret = xsm_mem_event_op(XSM_DM_PRIV, d, op);
- if ( ret )
- goto out;
-
- switch (op)
- {
- case XENMEM_paging_op:
- ret = mem_paging_memop(d, (xen_mem_event_op_t *) arg);
- break;
- case XENMEM_sharing_op:
- ret = mem_sharing_memop(d, (xen_mem_sharing_op_t *) arg);
- break;
- default:
- ret = -ENOSYS;
- }
-
- out:
- rcu_unlock_domain(d);
- return ret;
-}
-
-/* Clean up on domain destruction */
-void mem_event_cleanup(struct domain *d)
-{
- if ( d->mem_event->paging.ring_page ) {
- /* Destroying the wait queue head means waking up all
- * queued vcpus. This will drain the list, allowing
- * the disable routine to complete. It will also drop
- * all domain refs the wait-queued vcpus are holding.
- * Finally, because this code path involves previously
- * pausing the domain (domain_kill), unpausing the
- * vcpus causes no harm. */
- destroy_waitqueue_head(&d->mem_event->paging.wq);
- (void)mem_event_disable(d, &d->mem_event->paging);
- }
- if ( d->mem_event->access.ring_page ) {
- destroy_waitqueue_head(&d->mem_event->access.wq);
- (void)mem_event_disable(d, &d->mem_event->access);
- }
- if ( d->mem_event->share.ring_page ) {
- destroy_waitqueue_head(&d->mem_event->share.wq);
- (void)mem_event_disable(d, &d->mem_event->share);
- }
-}
-
-int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
- XEN_GUEST_HANDLE_PARAM(void) u_domctl)
-{
- int rc;
-
- rc = xsm_mem_event_control(XSM_PRIV, d, mec->mode, mec->op);
- if ( rc )
- return rc;
-
- if ( unlikely(d == current->domain) )
- {
- gdprintk(XENLOG_INFO, "Tried to do a memory event op on itself.\n");
- return -EINVAL;
- }
-
- if ( unlikely(d->is_dying) )
- {
- gdprintk(XENLOG_INFO, "Ignoring memory event op on dying domain %u\n",
- d->domain_id);
- return 0;
- }
-
- if ( unlikely(d->vcpu == NULL) || unlikely(d->vcpu[0] == NULL) )
- {
- gdprintk(XENLOG_INFO,
- "Memory event op on a domain (%u) with no vcpus\n",
- d->domain_id);
- return -EINVAL;
- }
-
- rc = -ENOSYS;
-
- switch ( mec->mode )
- {
- case XEN_DOMCTL_MEM_EVENT_OP_PAGING:
- {
- struct mem_event_domain *med = &d->mem_event->paging;
- rc = -EINVAL;
-
- switch( mec->op )
- {
- case XEN_DOMCTL_MEM_EVENT_OP_PAGING_ENABLE:
- {
- struct p2m_domain *p2m = p2m_get_hostp2m(d);
-
- rc = -EOPNOTSUPP;
- /* pvh fixme: p2m_is_foreign types need addressing */
- if ( is_pvh_vcpu(current) || is_pvh_domain(hardware_domain) )
- break;
-
- rc = -ENODEV;
- /* Only HAP is supported */
- if ( !hap_enabled(d) )
- break;
-
- /* No paging if iommu is used */
- rc = -EMLINK;
- if ( unlikely(need_iommu(d)) )
- break;
-
- rc = -EXDEV;
- /* Disallow paging in a PoD guest */
- if ( p2m->pod.entry_count )
- break;
-
- rc = mem_event_enable(d, mec, med, _VPF_mem_paging,
- HVM_PARAM_PAGING_RING_PFN,
- mem_paging_notification);
- }
- break;
-
- case XEN_DOMCTL_MEM_EVENT_OP_PAGING_DISABLE:
- {
- if ( med->ring_page )
- rc = mem_event_disable(d, med);
- }
- break;
-
- default:
- rc = -ENOSYS;
- break;
- }
- }
- break;
-
- case XEN_DOMCTL_MEM_EVENT_OP_ACCESS:
- {
- struct mem_event_domain *med = &d->mem_event->access;
- rc = -EINVAL;
-
- switch( mec->op )
- {
- case XEN_DOMCTL_MEM_EVENT_OP_ACCESS_ENABLE:
- {
- rc = -ENODEV;
- /* Only HAP is supported */
- if ( !hap_enabled(d) )
- break;
-
- /* Currently only EPT is supported */
- if ( !cpu_has_vmx )
- break;
-
- rc = mem_event_enable(d, mec, med, _VPF_mem_access,
- HVM_PARAM_ACCESS_RING_PFN,
- mem_access_notification);
- }
- break;
-
- case XEN_DOMCTL_MEM_EVENT_OP_ACCESS_DISABLE:
- {
- if ( med->ring_page )
- rc = mem_event_disable(d, med);
- }
- break;
-
- default:
- rc = -ENOSYS;
- break;
- }
- }
- break;
-
- case XEN_DOMCTL_MEM_EVENT_OP_SHARING:
- {
- struct mem_event_domain *med = &d->mem_event->share;
- rc = -EINVAL;
-
- switch( mec->op )
- {
- case XEN_DOMCTL_MEM_EVENT_OP_SHARING_ENABLE:
- {
- rc = -EOPNOTSUPP;
- /* pvh fixme: p2m_is_foreign types need addressing */
- if ( is_pvh_vcpu(current) || is_pvh_domain(hardware_domain) )
- break;
-
- rc = -ENODEV;
- /* Only HAP is supported */
- if ( !hap_enabled(d) )
- break;
-
- rc = mem_event_enable(d, mec, med, _VPF_mem_sharing,
- HVM_PARAM_SHARING_RING_PFN,
- mem_sharing_notification);
- }
- break;
-
- case XEN_DOMCTL_MEM_EVENT_OP_SHARING_DISABLE:
- {
- if ( med->ring_page )
- rc = mem_event_disable(d, med);
- }
- break;
-
- default:
- rc = -ENOSYS;
- break;
- }
- }
- break;
-
- default:
- rc = -ENOSYS;
- }
-
- return rc;
-}
-
-void mem_event_vcpu_pause(struct vcpu *v)
-{
- ASSERT(v == current);
-
- atomic_inc(&v->mem_event_pause_count);
- vcpu_pause_nosync(v);
-}
-
-void mem_event_vcpu_unpause(struct vcpu *v)
-{
- int old, new, prev = v->mem_event_pause_count.counter;
-
- /* All unpause requests as a result of toolstack responses. Prevent
- * underflow of the vcpu pause count. */
- do
- {
- old = prev;
- new = old - 1;
-
- if ( new < 0 )
- {
- printk(XENLOG_G_WARNING
- "%pv mem_event: Too many unpause attempts\n", v);
- return;
- }
-
- prev = cmpxchg(&v->mem_event_pause_count.counter, old, new);
- } while ( prev != old );
-
- vcpu_unpause(v);
-}
-
-/*
- * Local variables:
- * mode: C
- * c-file-style: "BSD"
- * c-basic-offset: 4
- * indent-tabs-mode: nil
- * End:
- */
diff --git a/xen/arch/x86/mm/mem_paging.c b/xen/arch/x86/mm/mem_paging.c
index 235776d..65f6a3d 100644
--- a/xen/arch/x86/mm/mem_paging.c
+++ b/xen/arch/x86/mm/mem_paging.c
@@ -22,7 +22,7 @@
#include <asm/p2m.h>
-#include <asm/mem_event.h>
+#include <xen/mem_event.h>
int mem_paging_memop(struct domain *d, xen_mem_event_op_t *mec)
diff --git a/xen/arch/x86/mm/mem_sharing.c b/xen/arch/x86/mm/mem_sharing.c
index 79188b9..7c0fc7d 100644
--- a/xen/arch/x86/mm/mem_sharing.c
+++ b/xen/arch/x86/mm/mem_sharing.c
@@ -27,12 +27,12 @@
#include <xen/mm.h>
#include <xen/grant_table.h>
#include <xen/sched.h>
+#include <xen/rcupdate.h>
+#include <xen/mem_event.h>
#include <asm/page.h>
#include <asm/string.h>
#include <asm/p2m.h>
-#include <asm/mem_event.h>
#include <asm/atomic.h>
-#include <xen/rcupdate.h>
#include <asm/event.h>
#include <xsm/xsm.h>
diff --git a/xen/arch/x86/mm/p2m-pod.c b/xen/arch/x86/mm/p2m-pod.c
index bd4c7c8..43f507c 100644
--- a/xen/arch/x86/mm/p2m-pod.c
+++ b/xen/arch/x86/mm/p2m-pod.c
@@ -20,16 +20,16 @@
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
+#include <xen/iommu.h>
+#include <xen/mem_event.h>
+#include <xen/event.h>
+#include <public/mem_event.h>
#include <asm/domain.h>
#include <asm/page.h>
#include <asm/paging.h>
#include <asm/p2m.h>
#include <asm/hvm/vmx/vmx.h> /* ept_p2m_init() */
-#include <xen/iommu.h>
-#include <asm/mem_event.h>
-#include <public/mem_event.h>
#include <asm/mem_sharing.h>
-#include <xen/event.h>
#include <asm/hvm/nestedhvm.h>
#include <asm/hvm/svm/amd-iommu-proto.h>
diff --git a/xen/arch/x86/mm/p2m-pt.c b/xen/arch/x86/mm/p2m-pt.c
index 085ab6f..e48b63a 100644
--- a/xen/arch/x86/mm/p2m-pt.c
+++ b/xen/arch/x86/mm/p2m-pt.c
@@ -25,16 +25,16 @@
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
+#include <xen/iommu.h>
+#include <xen/mem_event.h>
+#include <xen/event.h>
+#include <xen/trace.h>
+#include <public/mem_event.h>
#include <asm/domain.h>
#include <asm/page.h>
#include <asm/paging.h>
#include <asm/p2m.h>
-#include <xen/iommu.h>
-#include <asm/mem_event.h>
-#include <public/mem_event.h>
#include <asm/mem_sharing.h>
-#include <xen/event.h>
-#include <xen/trace.h>
#include <asm/hvm/nestedhvm.h>
#include <asm/hvm/svm/amd-iommu-proto.h>
diff --git a/xen/arch/x86/mm/p2m.c b/xen/arch/x86/mm/p2m.c
index 32776c3..a9f120a 100644
--- a/xen/arch/x86/mm/p2m.c
+++ b/xen/arch/x86/mm/p2m.c
@@ -24,16 +24,16 @@
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
+#include <xen/iommu.h>
+#include <xen/mem_event.h>
+#include <xen/event.h>
+#include <public/mem_event.h>
#include <asm/domain.h>
#include <asm/page.h>
#include <asm/paging.h>
#include <asm/p2m.h>
#include <asm/hvm/vmx/vmx.h> /* ept_p2m_init() */
-#include <xen/iommu.h>
-#include <asm/mem_event.h>
-#include <public/mem_event.h>
#include <asm/mem_sharing.h>
-#include <xen/event.h>
#include <asm/hvm/nestedhvm.h>
#include <asm/hvm/svm/amd-iommu-proto.h>
#include <xsm/xsm.h>
diff --git a/xen/arch/x86/x86_64/compat/mm.c b/xen/arch/x86/x86_64/compat/mm.c
index 69c6195..c079702 100644
--- a/xen/arch/x86/x86_64/compat/mm.c
+++ b/xen/arch/x86/x86_64/compat/mm.c
@@ -1,10 +1,10 @@
#include <xen/event.h>
+#include <xen/mem_event.h>
+#include <xen/mem_access.h>
#include <xen/multicall.h>
#include <compat/memory.h>
#include <compat/xen.h>
-#include <asm/mem_event.h>
#include <asm/mem_sharing.h>
-#include <asm/mem_access.h>
int compat_set_gdt(XEN_GUEST_HANDLE_PARAM(uint) frame_list, unsigned int entries)
{
diff --git a/xen/arch/x86/x86_64/mm.c b/xen/arch/x86/x86_64/mm.c
index 4937f9a..0da6ddc 100644
--- a/xen/arch/x86/x86_64/mm.c
+++ b/xen/arch/x86/x86_64/mm.c
@@ -26,6 +26,8 @@
#include <xen/nodemask.h>
#include <xen/guest_access.h>
#include <xen/hypercall.h>
+#include <xen/mem_event.h>
+#include <xen/mem_access.h>
#include <asm/current.h>
#include <asm/asm_defns.h>
#include <asm/page.h>
@@ -35,9 +37,7 @@
#include <asm/msr.h>
#include <asm/setup.h>
#include <asm/numa.h>
-#include <asm/mem_event.h>
#include <asm/mem_sharing.h>
-#include <asm/mem_access.h>
#include <public/memory.h>
/* Parameters for PFN/MADDR compression. */
diff --git a/xen/common/Makefile b/xen/common/Makefile
index 3683ae3..b9f3387 100644
--- a/xen/common/Makefile
+++ b/xen/common/Makefile
@@ -51,6 +51,8 @@ obj-y += tmem_xen.o
obj-y += radix-tree.o
obj-y += rbtree.o
obj-y += lzo.o
+obj-$(HAS_MEM_ACCESS) += mem_access.o
+obj-$(HAS_MEM_ACCESS) += mem_event.o
obj-bin-$(CONFIG_X86) += $(foreach n,decompress bunzip2 unxz unlzma unlzo unlz4 earlycpio,$(n).init.o)
diff --git a/xen/common/domain.c b/xen/common/domain.c
index 62514b0..134bed6 100644
--- a/xen/common/domain.c
+++ b/xen/common/domain.c
@@ -15,6 +15,7 @@
#include <xen/domain.h>
#include <xen/mm.h>
#include <xen/event.h>
+#include <xen/mem_event.h>
#include <xen/time.h>
#include <xen/console.h>
#include <xen/softirq.h>
diff --git a/xen/common/mem_access.c b/xen/common/mem_access.c
new file mode 100644
index 0000000..07161a2
--- /dev/null
+++ b/xen/common/mem_access.c
@@ -0,0 +1,133 @@
+/******************************************************************************
+ * mem_access.c
+ *
+ * Memory access support.
+ *
+ * Copyright (c) 2011 Virtuata, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+
+#include <xen/sched.h>
+#include <xen/guest_access.h>
+#include <xen/hypercall.h>
+#include <asm/p2m.h>
+#include <public/memory.h>
+#include <xen/mem_event.h>
+#include <xsm/xsm.h>
+
+int mem_access_memop(unsigned long cmd,
+ XEN_GUEST_HANDLE_PARAM(xen_mem_access_op_t) arg)
+{
+ long rc;
+ xen_mem_access_op_t mao;
+ struct domain *d;
+
+ if ( copy_from_guest(&mao, arg, 1) )
+ return -EFAULT;
+
+ rc = rcu_lock_live_remote_domain_by_id(mao.domid, &d);
+ if ( rc )
+ return rc;
+
+ rc = -EINVAL;
+ if ( !is_hvm_domain(d) )
+ goto out;
+
+ rc = xsm_mem_event_op(XSM_DM_PRIV, d, XENMEM_access_op);
+ if ( rc )
+ goto out;
+
+ rc = -ENODEV;
+ if ( unlikely(!d->mem_event->access.ring_page) )
+ goto out;
+
+ switch ( mao.op )
+ {
+ case XENMEM_access_op_resume:
+ p2m_mem_access_resume(d);
+ rc = 0;
+ break;
+
+ case XENMEM_access_op_set_access:
+ {
+ unsigned long start_iter = cmd & ~MEMOP_CMD_MASK;
+
+ rc = -EINVAL;
+ if ( (mao.pfn != ~0ull) &&
+ (mao.nr < start_iter ||
+ ((mao.pfn + mao.nr - 1) < mao.pfn) ||
+ ((mao.pfn + mao.nr - 1) > domain_get_maximum_gpfn(d))) )
+ break;
+
+ rc = p2m_set_mem_access(d, mao.pfn, mao.nr, start_iter,
+ MEMOP_CMD_MASK, mao.access);
+ if ( rc > 0 )
+ {
+ ASSERT(!(rc & MEMOP_CMD_MASK));
+ rc = hypercall_create_continuation(__HYPERVISOR_memory_op, "lh",
+ XENMEM_access_op | rc, arg);
+ }
+ break;
+ }
+
+ case XENMEM_access_op_get_access:
+ {
+ xenmem_access_t access;
+
+ rc = -EINVAL;
+ if ( (mao.pfn > domain_get_maximum_gpfn(d)) && mao.pfn != ~0ull )
+ break;
+
+ rc = p2m_get_mem_access(d, mao.pfn, &access);
+ if ( rc != 0 )
+ break;
+
+ mao.access = access;
+ rc = __copy_field_to_guest(arg, &mao, access) ? -EFAULT : 0;
+
+ break;
+ }
+
+ default:
+ rc = -ENOSYS;
+ break;
+ }
+
+ out:
+ rcu_unlock_domain(d);
+ return rc;
+}
+
+int mem_access_send_req(struct domain *d, mem_event_request_t *req)
+{
+ int rc = mem_event_claim_slot(d, &d->mem_event->access);
+ if ( rc < 0 )
+ return rc;
+
+ mem_event_put_request(d, &d->mem_event->access, req);
+
+ return 0;
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/common/mem_event.c b/xen/common/mem_event.c
new file mode 100644
index 0000000..b4a23fd
--- /dev/null
+++ b/xen/common/mem_event.c
@@ -0,0 +1,705 @@
+/******************************************************************************
+ * mem_event.c
+ *
+ * Memory event support.
+ *
+ * Copyright (c) 2009 Citrix Systems, Inc. (Patrick Colp)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+
+#include <xen/sched.h>
+#include <xen/event.h>
+#include <xen/wait.h>
+#include <asm/p2m.h>
+#include <xen/mem_event.h>
+#include <xen/mem_access.h>
+#include <asm/mem_paging.h>
+#include <asm/mem_sharing.h>
+#include <xsm/xsm.h>
+
+/* for public/io/ring.h macros */
+#define xen_mb() mb()
+#define xen_rmb() rmb()
+#define xen_wmb() wmb()
+
+#define mem_event_ring_lock_init(_med) spin_lock_init(&(_med)->ring_lock)
+#define mem_event_ring_lock(_med) spin_lock(&(_med)->ring_lock)
+#define mem_event_ring_unlock(_med) spin_unlock(&(_med)->ring_lock)
+
+static int mem_event_enable(
+ struct domain *d,
+ xen_domctl_mem_event_op_t *mec,
+ struct mem_event_domain *med,
+ int pause_flag,
+ int param,
+ xen_event_channel_notification_t notification_fn)
+{
+ int rc;
+ unsigned long ring_gfn = d->arch.hvm_domain.params[param];
+
+ /* Only one helper at a time. If the helper crashed,
+ * the ring is in an undefined state and so is the guest.
+ */
+ if ( med->ring_page )
+ return -EBUSY;
+
+ /* The parameter defaults to zero, and it should be
+ * set to something */
+ if ( ring_gfn == 0 )
+ return -ENOSYS;
+
+ mem_event_ring_lock_init(med);
+ mem_event_ring_lock(med);
+
+ rc = prepare_ring_for_helper(d, ring_gfn, &med->ring_pg_struct,
+ &med->ring_page);
+ if ( rc < 0 )
+ goto err;
+
+ /* Set the number of currently blocked vCPUs to 0. */
+ med->blocked = 0;
+
+ /* Allocate event channel */
+ rc = alloc_unbound_xen_event_channel(d->vcpu[0],
+ current->domain->domain_id,
+ notification_fn);
+ if ( rc < 0 )
+ goto err;
+
+ med->xen_port = mec->port = rc;
+
+ /* Prepare ring buffer */
+ FRONT_RING_INIT(&med->front_ring,
+ (mem_event_sring_t *)med->ring_page,
+ PAGE_SIZE);
+
+ /* Save the pause flag for this particular ring. */
+ med->pause_flag = pause_flag;
+
+ /* Initialize the last-chance wait queue. */
+ init_waitqueue_head(&med->wq);
+
+ mem_event_ring_unlock(med);
+ return 0;
+
+ err:
+ destroy_ring_for_helper(&med->ring_page,
+ med->ring_pg_struct);
+ mem_event_ring_unlock(med);
+
+ return rc;
+}
+
+static unsigned int mem_event_ring_available(struct mem_event_domain *med)
+{
+ int avail_req = RING_FREE_REQUESTS(&med->front_ring);
+ avail_req -= med->target_producers;
+ avail_req -= med->foreign_producers;
+
+ BUG_ON(avail_req < 0);
+
+ return avail_req;
+}
+
+/*
+ * mem_event_wake_blocked() will wakeup vcpus waiting for room in the
+ * ring. These vCPUs were paused on their way out after placing an event,
+ * but need to be resumed where the ring is capable of processing at least
+ * one event from them.
+ */
+static void mem_event_wake_blocked(struct domain *d, struct mem_event_domain *med)
+{
+ struct vcpu *v;
+ int online = d->max_vcpus;
+ unsigned int avail_req = mem_event_ring_available(med);
+
+ if ( avail_req == 0 || med->blocked == 0 )
+ return;
+
+ /*
+ * We ensure that we only have vCPUs online if there are enough free slots
+ * for their memory events to be processed. This will ensure that no
+ * memory events are lost (due to the fact that certain types of events
+ * cannot be replayed, we need to ensure that there is space in the ring
+ * for when they are hit).
+ * See comment below in mem_event_put_request().
+ */
+ for_each_vcpu ( d, v )
+ if ( test_bit(med->pause_flag, &v->pause_flags) )
+ online--;
+
+ ASSERT(online == (d->max_vcpus - med->blocked));
+
+ /* We remember which vcpu last woke up to avoid scanning always linearly
+ * from zero and starving higher-numbered vcpus under high load */
+ if ( d->vcpu )
+ {
+ int i, j, k;
+
+ for (i = med->last_vcpu_wake_up + 1, j = 0; j < d->max_vcpus; i++, j++)
+ {
+ k = i % d->max_vcpus;
+ v = d->vcpu[k];
+ if ( !v )
+ continue;
+
+ if ( !(med->blocked) || online >= avail_req )
+ break;
+
+ if ( test_and_clear_bit(med->pause_flag, &v->pause_flags) )
+ {
+ vcpu_unpause(v);
+ online++;
+ med->blocked--;
+ med->last_vcpu_wake_up = k;
+ }
+ }
+ }
+}
+
+/*
+ * In the event that a vCPU attempted to place an event in the ring and
+ * was unable to do so, it is queued on a wait queue. These are woken as
+ * needed, and take precedence over the blocked vCPUs.
+ */
+static void mem_event_wake_queued(struct domain *d, struct mem_event_domain *med)
+{
+ unsigned int avail_req = mem_event_ring_available(med);
+
+ if ( avail_req > 0 )
+ wake_up_nr(&med->wq, avail_req);
+}
+
+/*
+ * mem_event_wake() will wakeup all vcpus waiting for the ring to
+ * become available. If we have queued vCPUs, they get top priority. We
+ * are guaranteed that they will go through code paths that will eventually
+ * call mem_event_wake() again, ensuring that any blocked vCPUs will get
+ * unpaused once all the queued vCPUs have made it through.
+ */
+void mem_event_wake(struct domain *d, struct mem_event_domain *med)
+{
+ if (!list_empty(&med->wq.list))
+ mem_event_wake_queued(d, med);
+ else
+ mem_event_wake_blocked(d, med);
+}
+
+static int mem_event_disable(struct domain *d, struct mem_event_domain *med)
+{
+ if ( med->ring_page )
+ {
+ struct vcpu *v;
+
+ mem_event_ring_lock(med);
+
+ if ( !list_empty(&med->wq.list) )
+ {
+ mem_event_ring_unlock(med);
+ return -EBUSY;
+ }
+
+ /* Free domU's event channel and leave the other one unbound */
+ free_xen_event_channel(d->vcpu[0], med->xen_port);
+
+ /* Unblock all vCPUs */
+ for_each_vcpu ( d, v )
+ {
+ if ( test_and_clear_bit(med->pause_flag, &v->pause_flags) )
+ {
+ vcpu_unpause(v);
+ med->blocked--;
+ }
+ }
+
+ destroy_ring_for_helper(&med->ring_page,
+ med->ring_pg_struct);
+ mem_event_ring_unlock(med);
+ }
+
+ return 0;
+}
+
+static inline void mem_event_release_slot(struct domain *d,
+ struct mem_event_domain *med)
+{
+ /* Update the accounting */
+ if ( current->domain == d )
+ med->target_producers--;
+ else
+ med->foreign_producers--;
+
+ /* Kick any waiters */
+ mem_event_wake(d, med);
+}
+
+/*
+ * mem_event_mark_and_pause() tags vcpu and put it to sleep.
+ * The vcpu will resume execution in mem_event_wake_waiters().
+ */
+void mem_event_mark_and_pause(struct vcpu *v, struct mem_event_domain *med)
+{
+ if ( !test_and_set_bit(med->pause_flag, &v->pause_flags) )
+ {
+ vcpu_pause_nosync(v);
+ med->blocked++;
+ }
+}
+
+/*
+ * This must be preceded by a call to claim_slot(), and is guaranteed to
+ * succeed. As a side-effect however, the vCPU may be paused if the ring is
+ * overly full and its continued execution would cause stalling and excessive
+ * waiting. The vCPU will be automatically unpaused when the ring clears.
+ */
+void mem_event_put_request(struct domain *d,
+ struct mem_event_domain *med,
+ mem_event_request_t *req)
+{
+ mem_event_front_ring_t *front_ring;
+ int free_req;
+ unsigned int avail_req;
+ RING_IDX req_prod;
+
+ if ( current->domain != d )
+ {
+ req->flags |= MEM_EVENT_FLAG_FOREIGN;
+ ASSERT( !(req->flags & MEM_EVENT_FLAG_VCPU_PAUSED) );
+ }
+
+ mem_event_ring_lock(med);
+
+ /* Due to the reservations, this step must succeed. */
+ front_ring = &med->front_ring;
+ free_req = RING_FREE_REQUESTS(front_ring);
+ ASSERT(free_req > 0);
+
+ /* Copy request */
+ req_prod = front_ring->req_prod_pvt;
+ memcpy(RING_GET_REQUEST(front_ring, req_prod), req, sizeof(*req));
+ req_prod++;
+
+ /* Update ring */
+ front_ring->req_prod_pvt = req_prod;
+ RING_PUSH_REQUESTS(front_ring);
+
+ /* We've actually *used* our reservation, so release the slot. */
+ mem_event_release_slot(d, med);
+
+ /* Give this vCPU a black eye if necessary, on the way out.
+ * See the comments above wake_blocked() for more information
+ * on how this mechanism works to avoid waiting. */
+ avail_req = mem_event_ring_available(med);
+ if( current->domain == d && avail_req < d->max_vcpus )
+ mem_event_mark_and_pause(current, med);
+
+ mem_event_ring_unlock(med);
+
+ notify_via_xen_event_channel(d, med->xen_port);
+}
+
+int mem_event_get_response(struct domain *d, struct mem_event_domain *med, mem_event_response_t *rsp)
+{
+ mem_event_front_ring_t *front_ring;
+ RING_IDX rsp_cons;
+
+ mem_event_ring_lock(med);
+
+ front_ring = &med->front_ring;
+ rsp_cons = front_ring->rsp_cons;
+
+ if ( !RING_HAS_UNCONSUMED_RESPONSES(front_ring) )
+ {
+ mem_event_ring_unlock(med);
+ return 0;
+ }
+
+ /* Copy response */
+ memcpy(rsp, RING_GET_RESPONSE(front_ring, rsp_cons), sizeof(*rsp));
+ rsp_cons++;
+
+ /* Update ring */
+ front_ring->rsp_cons = rsp_cons;
+ front_ring->sring->rsp_event = rsp_cons + 1;
+
+ /* Kick any waiters -- since we've just consumed an event,
+ * there may be additional space available in the ring. */
+ mem_event_wake(d, med);
+
+ mem_event_ring_unlock(med);
+
+ return 1;
+}
+
+void mem_event_cancel_slot(struct domain *d, struct mem_event_domain *med)
+{
+ mem_event_ring_lock(med);
+ mem_event_release_slot(d, med);
+ mem_event_ring_unlock(med);
+}
+
+static int mem_event_grab_slot(struct mem_event_domain *med, int foreign)
+{
+ unsigned int avail_req;
+
+ if ( !med->ring_page )
+ return -ENOSYS;
+
+ mem_event_ring_lock(med);
+
+ avail_req = mem_event_ring_available(med);
+ if ( avail_req == 0 )
+ {
+ mem_event_ring_unlock(med);
+ return -EBUSY;
+ }
+
+ if ( !foreign )
+ med->target_producers++;
+ else
+ med->foreign_producers++;
+
+ mem_event_ring_unlock(med);
+
+ return 0;
+}
+
+/* Simple try_grab wrapper for use in the wait_event() macro. */
+static int mem_event_wait_try_grab(struct mem_event_domain *med, int *rc)
+{
+ *rc = mem_event_grab_slot(med, 0);
+ return *rc;
+}
+
+/* Call mem_event_grab_slot() until the ring doesn't exist, or is available. */
+static int mem_event_wait_slot(struct mem_event_domain *med)
+{
+ int rc = -EBUSY;
+ wait_event(med->wq, mem_event_wait_try_grab(med, &rc) != -EBUSY);
+ return rc;
+}
+
+bool_t mem_event_check_ring(struct mem_event_domain *med)
+{
+ return (med->ring_page != NULL);
+}
+
+/*
+ * Determines whether or not the current vCPU belongs to the target domain,
+ * and calls the appropriate wait function. If it is a guest vCPU, then we
+ * use mem_event_wait_slot() to reserve a slot. As long as there is a ring,
+ * this function will always return 0 for a guest. For a non-guest, we check
+ * for space and return -EBUSY if the ring is not available.
+ *
+ * Return codes: -ENOSYS: the ring is not yet configured
+ * -EBUSY: the ring is busy
+ * 0: a spot has been reserved
+ *
+ */
+int __mem_event_claim_slot(struct domain *d, struct mem_event_domain *med,
+ bool_t allow_sleep)
+{
+ if ( (current->domain == d) && allow_sleep )
+ return mem_event_wait_slot(med);
+ else
+ return mem_event_grab_slot(med, (current->domain != d));
+}
+
+/* Registered with Xen-bound event channel for incoming notifications. */
+static void mem_paging_notification(struct vcpu *v, unsigned int port)
+{
+ if ( likely(v->domain->mem_event->paging.ring_page != NULL) )
+ p2m_mem_paging_resume(v->domain);
+}
+
+/* Registered with Xen-bound event channel for incoming notifications. */
+static void mem_access_notification(struct vcpu *v, unsigned int port)
+{
+ if ( likely(v->domain->mem_event->access.ring_page != NULL) )
+ p2m_mem_access_resume(v->domain);
+}
+
+/* Registered with Xen-bound event channel for incoming notifications. */
+static void mem_sharing_notification(struct vcpu *v, unsigned int port)
+{
+ if ( likely(v->domain->mem_event->share.ring_page != NULL) )
+ mem_sharing_sharing_resume(v->domain);
+}
+
+int do_mem_event_op(int op, uint32_t domain, void *arg)
+{
+ int ret;
+ struct domain *d;
+
+ ret = rcu_lock_live_remote_domain_by_id(domain, &d);
+ if ( ret )
+ return ret;
+
+ ret = xsm_mem_event_op(XSM_DM_PRIV, d, op);
+ if ( ret )
+ goto out;
+
+ switch (op)
+ {
+ case XENMEM_paging_op:
+ ret = mem_paging_memop(d, (xen_mem_event_op_t *) arg);
+ break;
+ case XENMEM_sharing_op:
+ ret = mem_sharing_memop(d, (xen_mem_sharing_op_t *) arg);
+ break;
+ default:
+ ret = -ENOSYS;
+ }
+
+ out:
+ rcu_unlock_domain(d);
+ return ret;
+}
+
+/* Clean up on domain destruction */
+void mem_event_cleanup(struct domain *d)
+{
+ if ( d->mem_event->paging.ring_page ) {
+ /* Destroying the wait queue head means waking up all
+ * queued vcpus. This will drain the list, allowing
+ * the disable routine to complete. It will also drop
+ * all domain refs the wait-queued vcpus are holding.
+ * Finally, because this code path involves previously
+ * pausing the domain (domain_kill), unpausing the
+ * vcpus causes no harm. */
+ destroy_waitqueue_head(&d->mem_event->paging.wq);
+ (void)mem_event_disable(d, &d->mem_event->paging);
+ }
+ if ( d->mem_event->access.ring_page ) {
+ destroy_waitqueue_head(&d->mem_event->access.wq);
+ (void)mem_event_disable(d, &d->mem_event->access);
+ }
+ if ( d->mem_event->share.ring_page ) {
+ destroy_waitqueue_head(&d->mem_event->share.wq);
+ (void)mem_event_disable(d, &d->mem_event->share);
+ }
+}
+
+int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
+ XEN_GUEST_HANDLE_PARAM(void) u_domctl)
+{
+ int rc;
+
+ rc = xsm_mem_event_control(XSM_PRIV, d, mec->mode, mec->op);
+ if ( rc )
+ return rc;
+
+ if ( unlikely(d == current->domain) )
+ {
+ gdprintk(XENLOG_INFO, "Tried to do a memory event op on itself.\n");
+ return -EINVAL;
+ }
+
+ if ( unlikely(d->is_dying) )
+ {
+ gdprintk(XENLOG_INFO, "Ignoring memory event op on dying domain %u\n",
+ d->domain_id);
+ return 0;
+ }
+
+ if ( unlikely(d->vcpu == NULL) || unlikely(d->vcpu[0] == NULL) )
+ {
+ gdprintk(XENLOG_INFO,
+ "Memory event op on a domain (%u) with no vcpus\n",
+ d->domain_id);
+ return -EINVAL;
+ }
+
+ rc = -ENOSYS;
+
+ switch ( mec->mode )
+ {
+ case XEN_DOMCTL_MEM_EVENT_OP_PAGING:
+ {
+ struct mem_event_domain *med = &d->mem_event->paging;
+ rc = -EINVAL;
+
+ switch( mec->op )
+ {
+ case XEN_DOMCTL_MEM_EVENT_OP_PAGING_ENABLE:
+ {
+ struct p2m_domain *p2m = p2m_get_hostp2m(d);
+
+ rc = -EOPNOTSUPP;
+ /* pvh fixme: p2m_is_foreign types need addressing */
+ if ( is_pvh_vcpu(current) || is_pvh_domain(hardware_domain) )
+ break;
+
+ rc = -ENODEV;
+ /* Only HAP is supported */
+ if ( !hap_enabled(d) )
+ break;
+
+ /* No paging if iommu is used */
+ rc = -EMLINK;
+ if ( unlikely(need_iommu(d)) )
+ break;
+
+ rc = -EXDEV;
+ /* Disallow paging in a PoD guest */
+ if ( p2m->pod.entry_count )
+ break;
+
+ rc = mem_event_enable(d, mec, med, _VPF_mem_paging,
+ HVM_PARAM_PAGING_RING_PFN,
+ mem_paging_notification);
+ }
+ break;
+
+ case XEN_DOMCTL_MEM_EVENT_OP_PAGING_DISABLE:
+ {
+ if ( med->ring_page )
+ rc = mem_event_disable(d, med);
+ }
+ break;
+
+ default:
+ rc = -ENOSYS;
+ break;
+ }
+ }
+ break;
+
+ case XEN_DOMCTL_MEM_EVENT_OP_ACCESS:
+ {
+ struct mem_event_domain *med = &d->mem_event->access;
+ rc = -EINVAL;
+
+ switch( mec->op )
+ {
+ case XEN_DOMCTL_MEM_EVENT_OP_ACCESS_ENABLE:
+ {
+ rc = -ENODEV;
+ /* Only HAP is supported */
+ if ( !hap_enabled(d) )
+ break;
+
+ /* Currently only EPT is supported */
+ if ( !cpu_has_vmx )
+ break;
+
+ rc = mem_event_enable(d, mec, med, _VPF_mem_access,
+ HVM_PARAM_ACCESS_RING_PFN,
+ mem_access_notification);
+ }
+ break;
+
+ case XEN_DOMCTL_MEM_EVENT_OP_ACCESS_DISABLE:
+ {
+ if ( med->ring_page )
+ rc = mem_event_disable(d, med);
+ }
+ break;
+
+ default:
+ rc = -ENOSYS;
+ break;
+ }
+ }
+ break;
+
+ case XEN_DOMCTL_MEM_EVENT_OP_SHARING:
+ {
+ struct mem_event_domain *med = &d->mem_event->share;
+ rc = -EINVAL;
+
+ switch( mec->op )
+ {
+ case XEN_DOMCTL_MEM_EVENT_OP_SHARING_ENABLE:
+ {
+ rc = -EOPNOTSUPP;
+ /* pvh fixme: p2m_is_foreign types need addressing */
+ if ( is_pvh_vcpu(current) || is_pvh_domain(hardware_domain) )
+ break;
+
+ rc = -ENODEV;
+ /* Only HAP is supported */
+ if ( !hap_enabled(d) )
+ break;
+
+ rc = mem_event_enable(d, mec, med, _VPF_mem_sharing,
+ HVM_PARAM_SHARING_RING_PFN,
+ mem_sharing_notification);
+ }
+ break;
+
+ case XEN_DOMCTL_MEM_EVENT_OP_SHARING_DISABLE:
+ {
+ if ( med->ring_page )
+ rc = mem_event_disable(d, med);
+ }
+ break;
+
+ default:
+ rc = -ENOSYS;
+ break;
+ }
+ }
+ break;
+
+ default:
+ rc = -ENOSYS;
+ }
+
+ return rc;
+}
+
+void mem_event_vcpu_pause(struct vcpu *v)
+{
+ ASSERT(v == current);
+
+ atomic_inc(&v->mem_event_pause_count);
+ vcpu_pause_nosync(v);
+}
+
+void mem_event_vcpu_unpause(struct vcpu *v)
+{
+ int old, new, prev = v->mem_event_pause_count.counter;
+
+ /* All unpause requests as a result of toolstack responses. Prevent
+ * underflow of the vcpu pause count. */
+ do
+ {
+ old = prev;
+ new = old - 1;
+
+ if ( new < 0 )
+ {
+ printk(XENLOG_G_WARNING
+ "%pv mem_event: Too many unpause attempts\n", v);
+ return;
+ }
+
+ prev = cmpxchg(&v->mem_event_pause_count.counter, old, new);
+ } while ( prev != old );
+
+ vcpu_unpause(v);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/common/memory.c b/xen/common/memory.c
index 031672a..44a8106 100644
--- a/xen/common/memory.c
+++ b/xen/common/memory.c
@@ -1108,6 +1108,69 @@ long do_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
return rc;
}
+void destroy_ring_for_helper(
+ void **_va, struct page_info *page)
+{
+ void *va = *_va;
+
+ if ( va != NULL )
+ {
+ unmap_domain_page_global(va);
+ put_page_and_type(page);
+ *_va = NULL;
+ }
+}
+
+int prepare_ring_for_helper(
+ struct domain *d, unsigned long gmfn, struct page_info **_page,
+ void **_va)
+{
+ struct page_info *page;
+ p2m_type_t p2mt;
+ void *va;
+
+ page = get_page_from_gfn(d, gmfn, &p2mt, P2M_UNSHARE);
+
+#ifdef CONFIG_MEM_PAGING
+ if ( p2m_is_paging(p2mt) )
+ {
+ if ( page )
+ put_page(page);
+ p2m_mem_paging_populate(d, gmfn);
+ return -ENOENT;
+ }
+#endif
+#ifdef CONFIG_MEM_SHARING
+ if ( p2m_is_shared(p2mt) )
+ {
+ if ( page )
+ put_page(page);
+ return -ENOENT;
+ }
+#endif
+
+ if ( !page )
+ return -EINVAL;
+
+ if ( !get_page_type(page, PGT_writable_page) )
+ {
+ put_page(page);
+ return -EINVAL;
+ }
+
+ va = __map_domain_page_global(page);
+ if ( va == NULL )
+ {
+ put_page_and_type(page);
+ return -ENOMEM;
+ }
+
+ *_va = va;
+ *_page = page;
+
+ return 0;
+}
+
/*
* Local variables:
* mode: C
diff --git a/xen/include/asm-arm/mm.h b/xen/include/asm-arm/mm.h
index 9fa80a4..7fc3b97 100644
--- a/xen/include/asm-arm/mm.h
+++ b/xen/include/asm-arm/mm.h
@@ -301,7 +301,6 @@ struct page_info *get_page_from_gva(struct domain *d, vaddr_t va,
})
static inline void put_gfn(struct domain *d, unsigned long gfn) {}
-static inline void mem_event_cleanup(struct domain *d) {}
static inline int relinquish_shared_pages(struct domain *d)
{
return 0;
diff --git a/xen/include/asm-x86/config.h b/xen/include/asm-x86/config.h
index 210ff57..8a864ce 100644
--- a/xen/include/asm-x86/config.h
+++ b/xen/include/asm-x86/config.h
@@ -67,6 +67,9 @@
#define NR_CPUS 256
#endif
+#define CONFIG_MEM_SHARING 1
+#define CONFIG_MEM_PAGING 1
+
/* Maximum we can support with current vLAPIC ID mapping. */
#define MAX_HVM_VCPUS 128
diff --git a/xen/include/asm-x86/hvm/hvm.h b/xen/include/asm-x86/hvm/hvm.h
index 1123857..74e66f8 100644
--- a/xen/include/asm-x86/hvm/hvm.h
+++ b/xen/include/asm-x86/hvm/hvm.h
@@ -226,12 +226,6 @@ int hvm_vcpu_cacheattr_init(struct vcpu *v);
void hvm_vcpu_cacheattr_destroy(struct vcpu *v);
void hvm_vcpu_reset_state(struct vcpu *v, uint16_t cs, uint16_t ip);
-/* Prepare/destroy a ring for a dom0 helper. Helper with talk
- * with Xen on behalf of this hvm domain. */
-int prepare_ring_for_helper(struct domain *d, unsigned long gmfn,
- struct page_info **_page, void **_va);
-void destroy_ring_for_helper(void **_va, struct page_info *page);
-
bool_t hvm_send_assist_req(ioreq_t *p);
void hvm_broadcast_assist_req(ioreq_t *p);
diff --git a/xen/include/asm-x86/mem_access.h b/xen/include/asm-x86/mem_access.h
deleted file mode 100644
index 5c7c5fd..0000000
--- a/xen/include/asm-x86/mem_access.h
+++ /dev/null
@@ -1,39 +0,0 @@
-/******************************************************************************
- * include/asm-x86/mem_access.h
- *
- * Memory access support.
- *
- * Copyright (c) 2011 Virtuata, Inc.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2 of the License, or
- * (at your option) any later version.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
- */
-
-#ifndef _XEN_ASM_MEM_ACCESS_H
-#define _XEN_ASM_MEM_ACCESS_H
-
-int mem_access_memop(unsigned long cmd,
- XEN_GUEST_HANDLE_PARAM(xen_mem_access_op_t) arg);
-int mem_access_send_req(struct domain *d, mem_event_request_t *req);
-
-#endif /* _XEN_ASM_MEM_ACCESS_H */
-
-/*
- * Local variables:
- * mode: C
- * c-file-style: "BSD"
- * c-basic-offset: 4
- * indent-tabs-mode: nil
- * End:
- */
diff --git a/xen/include/asm-x86/mem_event.h b/xen/include/asm-x86/mem_event.h
deleted file mode 100644
index ed4481a..0000000
--- a/xen/include/asm-x86/mem_event.h
+++ /dev/null
@@ -1,82 +0,0 @@
-/******************************************************************************
- * include/asm-x86/mem_event.h
- *
- * Common interface for memory event support.
- *
- * Copyright (c) 2009 Citrix Systems, Inc. (Patrick Colp)
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2 of the License, or
- * (at your option) any later version.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
- */
-
-
-#ifndef __MEM_EVENT_H__
-#define __MEM_EVENT_H__
-
-/* Returns whether a ring has been set up */
-bool_t mem_event_check_ring(struct mem_event_domain *med);
-
-/* Returns 0 on success, -ENOSYS if there is no ring, -EBUSY if there is no
- * available space and the caller is a foreign domain. If the guest itself
- * is the caller, -EBUSY is avoided by sleeping on a wait queue to ensure
- * that the ring does not lose future events.
- *
- * However, the allow_sleep flag can be set to false in cases in which it is ok
- * to lose future events, and thus -EBUSY can be returned to guest vcpus
- * (handle with care!).
- *
- * In general, you must follow a claim_slot() call with either put_request() or
- * cancel_slot(), both of which are guaranteed to
- * succeed.
- */
-int __mem_event_claim_slot(struct domain *d, struct mem_event_domain *med,
- bool_t allow_sleep);
-static inline int mem_event_claim_slot(struct domain *d,
- struct mem_event_domain *med)
-{
- return __mem_event_claim_slot(d, med, 1);
-}
-
-static inline int mem_event_claim_slot_nosleep(struct domain *d,
- struct mem_event_domain *med)
-{
- return __mem_event_claim_slot(d, med, 0);
-}
-
-void mem_event_cancel_slot(struct domain *d, struct mem_event_domain *med);
-
-void mem_event_put_request(struct domain *d, struct mem_event_domain *med,
- mem_event_request_t *req);
-
-int mem_event_get_response(struct domain *d, struct mem_event_domain *med,
- mem_event_response_t *rsp);
-
-int do_mem_event_op(int op, uint32_t domain, void *arg);
-int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
- XEN_GUEST_HANDLE_PARAM(void) u_domctl);
-
-void mem_event_vcpu_pause(struct vcpu *v);
-void mem_event_vcpu_unpause(struct vcpu *v);
-
-#endif /* __MEM_EVENT_H__ */
-
-
-/*
- * Local variables:
- * mode: C
- * c-file-style: "BSD"
- * c-basic-offset: 4
- * indent-tabs-mode: nil
- * End:
- */
diff --git a/xen/include/asm-x86/mm.h b/xen/include/asm-x86/mm.h
index 7b85865..ebd482d 100644
--- a/xen/include/asm-x86/mm.h
+++ b/xen/include/asm-x86/mm.h
@@ -611,8 +611,6 @@ unsigned int domain_clamp_alloc_bitsize(struct domain *d, unsigned int bits);
unsigned long domain_get_maximum_gpfn(struct domain *d);
-void mem_event_cleanup(struct domain *d);
-
extern struct domain *dom_xen, *dom_io, *dom_cow; /* for vmcoreinfo */
/* Definition of an mm lock: spinlock with extra fields for debugging */
diff --git a/xen/include/xen/mem_access.h b/xen/include/xen/mem_access.h
new file mode 100644
index 0000000..19d1a2d
--- /dev/null
+++ b/xen/include/xen/mem_access.h
@@ -0,0 +1,60 @@
+/******************************************************************************
+ * mem_access.h
+ *
+ * Memory access support.
+ *
+ * Copyright (c) 2011 Virtuata, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+#ifndef _XEN_ASM_MEM_ACCESS_H
+#define _XEN_ASM_MEM_ACCESS_H
+
+#include <public/memory.h>
+
+#ifdef HAS_MEM_ACCESS
+
+int mem_access_memop(unsigned long cmd,
+ XEN_GUEST_HANDLE_PARAM(xen_mem_access_op_t) arg);
+int mem_access_send_req(struct domain *d, mem_event_request_t *req);
+
+#else
+
+static inline
+int mem_access_memop(unsigned long cmd,
+ XEN_GUEST_HANDLE_PARAM(xen_mem_access_op_t) arg)
+{
+ return -ENOSYS;
+}
+
+static inline
+int mem_access_send_req(struct domain *d, mem_event_request_t *req)
+{
+ return -ENOSYS;
+}
+
+#endif /* HAS_MEM_ACCESS */
+
+#endif /* _XEN_ASM_MEM_ACCESS_H */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/xen/mem_event.h b/xen/include/xen/mem_event.h
new file mode 100644
index 0000000..8612b26
--- /dev/null
+++ b/xen/include/xen/mem_event.h
@@ -0,0 +1,143 @@
+/******************************************************************************
+ * mem_event.h
+ *
+ * Common interface for memory event support.
+ *
+ * Copyright (c) 2009 Citrix Systems, Inc. (Patrick Colp)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+
+
+#ifndef __MEM_EVENT_H__
+#define __MEM_EVENT_H__
+
+#include <xen/sched.h>
+
+#ifdef HAS_MEM_ACCESS
+
+/* Clean up on domain destruction */
+void mem_event_cleanup(struct domain *d);
+
+/* Returns whether a ring has been set up */
+bool_t mem_event_check_ring(struct mem_event_domain *med);
+
+/* Returns 0 on success, -ENOSYS if there is no ring, -EBUSY if there is no
+ * available space and the caller is a foreign domain. If the guest itself
+ * is the caller, -EBUSY is avoided by sleeping on a wait queue to ensure
+ * that the ring does not lose future events.
+ *
+ * However, the allow_sleep flag can be set to false in cases in which it is ok
+ * to lose future events, and thus -EBUSY can be returned to guest vcpus
+ * (handle with care!).
+ *
+ * In general, you must follow a claim_slot() call with either put_request() or
+ * cancel_slot(), both of which are guaranteed to
+ * succeed.
+ */
+int __mem_event_claim_slot(struct domain *d, struct mem_event_domain *med,
+ bool_t allow_sleep);
+static inline int mem_event_claim_slot(struct domain *d,
+ struct mem_event_domain *med)
+{
+ return __mem_event_claim_slot(d, med, 1);
+}
+
+static inline int mem_event_claim_slot_nosleep(struct domain *d,
+ struct mem_event_domain *med)
+{
+ return __mem_event_claim_slot(d, med, 0);
+}
+
+void mem_event_cancel_slot(struct domain *d, struct mem_event_domain *med);
+
+void mem_event_put_request(struct domain *d, struct mem_event_domain *med,
+ mem_event_request_t *req);
+
+int mem_event_get_response(struct domain *d, struct mem_event_domain *med,
+ mem_event_response_t *rsp);
+
+int do_mem_event_op(int op, uint32_t domain, void *arg);
+int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
+ XEN_GUEST_HANDLE_PARAM(void) u_domctl);
+
+void mem_event_vcpu_pause(struct vcpu *v);
+void mem_event_vcpu_unpause(struct vcpu *v);
+
+#else
+
+static inline void mem_event_cleanup(struct domain *d) {}
+
+static inline bool_t mem_event_check_ring(struct mem_event_domain *med)
+{
+ return 0;
+}
+
+static inline int mem_event_claim_slot(struct domain *d,
+ struct mem_event_domain *med)
+{
+ return -ENOSYS;
+}
+
+static inline int mem_event_claim_slot_nosleep(struct domain *d,
+ struct mem_event_domain *med)
+{
+ return -ENOSYS;
+}
+
+static inline
+void mem_event_cancel_slot(struct domain *d, struct mem_event_domain *med)
+{}
+
+static inline
+void mem_event_put_request(struct domain *d, struct mem_event_domain *med,
+ mem_event_request_t *req)
+{}
+
+static inline
+int mem_event_get_response(struct domain *d, struct mem_event_domain *med,
+ mem_event_response_t *rsp)
+{
+ return -ENOSYS;
+}
+
+static inline int do_mem_event_op(int op, uint32_t domain, void *arg)
+{
+ return -ENOSYS;
+}
+
+static inline
+int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
+ XEN_GUEST_HANDLE_PARAM(void) u_domctl)
+{
+ return -ENOSYS;
+}
+
+static inline void mem_event_vcpu_pause(struct vcpu *v) {}
+static inline void mem_event_vcpu_unpause(struct vcpu *v) {}
+
+#endif /* HAS_MEM_ACCESS */
+
+#endif /* __MEM_EVENT_H__ */
+
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/xen/mm.h b/xen/include/xen/mm.h
index b183189..7c0efc7 100644
--- a/xen/include/xen/mm.h
+++ b/xen/include/xen/mm.h
@@ -371,4 +371,10 @@ int guest_remove_page(struct domain *d, unsigned long gmfn);
/* TRUE if the whole page at @mfn is of the requested RAM type(s) above. */
int page_is_ram_type(unsigned long mfn, unsigned long mem_type);
+/* Prepare/destroy a ring for a dom0 helper. Helper with talk
+ * with Xen on behalf of this domain. */
+int prepare_ring_for_helper(struct domain *d, unsigned long gmfn,
+ struct page_info **_page, void **_va);
+void destroy_ring_for_helper(void **_va, struct page_info *page);
+
#endif /* __XEN_MM_H__ */
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 02/17] xen: Relocate p2m_mem_access_resume to mem_access common
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 01/17] xen: Relocate mem_access and mem_event into common Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-11 20:16 ` Julien Grall
2014-09-10 13:28 ` [PATCH v5 03/17] xen: Relocate struct npfec definition into common Tamas K Lengyel
` (15 subsequent siblings)
17 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
The p2m_mem_access resume function is not p2m specific and could be shared across ARM/x86.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v5: Style fix.
---
xen/arch/x86/mm/p2m.c | 24 ------------------------
xen/common/mem_access.c | 26 +++++++++++++++++++++++++-
xen/common/mem_event.c | 2 +-
xen/include/asm-x86/p2m.h | 2 --
xen/include/xen/mem_access.h | 3 +++
5 files changed, 29 insertions(+), 28 deletions(-)
diff --git a/xen/arch/x86/mm/p2m.c b/xen/arch/x86/mm/p2m.c
index a9f120a..bf8e537 100644
--- a/xen/arch/x86/mm/p2m.c
+++ b/xen/arch/x86/mm/p2m.c
@@ -1427,30 +1427,6 @@ bool_t p2m_mem_access_check(paddr_t gpa, unsigned long gla,
return (p2ma == p2m_access_n2rwx);
}
-void p2m_mem_access_resume(struct domain *d)
-{
- mem_event_response_t rsp;
-
- /* Pull all responses off the ring */
- while( mem_event_get_response(d, &d->mem_event->access, &rsp) )
- {
- struct vcpu *v;
-
- if ( rsp.flags & MEM_EVENT_FLAG_DUMMY )
- continue;
-
- /* Validate the vcpu_id in the response. */
- if ( (rsp.vcpu_id >= d->max_vcpus) || !d->vcpu[rsp.vcpu_id] )
- continue;
-
- v = d->vcpu[rsp.vcpu_id];
-
- /* Unpause domain */
- if ( rsp.flags & MEM_EVENT_FLAG_VCPU_PAUSED )
- mem_event_vcpu_unpause(v);
- }
-}
-
/* Set access type for a region of pfns.
* If start_pfn == -1ul, sets the default access type */
long p2m_set_mem_access(struct domain *d, unsigned long pfn, uint32_t nr,
diff --git a/xen/common/mem_access.c b/xen/common/mem_access.c
index 07161a2..c84fab3 100644
--- a/xen/common/mem_access.c
+++ b/xen/common/mem_access.c
@@ -29,6 +29,30 @@
#include <xen/mem_event.h>
#include <xsm/xsm.h>
+void mem_access_resume(struct domain *d)
+{
+ mem_event_response_t rsp;
+
+ /* Pull all responses off the ring. */
+ while ( mem_event_get_response(d, &d->mem_event->access, &rsp) )
+ {
+ struct vcpu *v;
+
+ if ( rsp.flags & MEM_EVENT_FLAG_DUMMY )
+ continue;
+
+ /* Validate the vcpu_id in the response. */
+ if ( (rsp.vcpu_id >= d->max_vcpus) || !d->vcpu[rsp.vcpu_id] )
+ continue;
+
+ v = d->vcpu[rsp.vcpu_id];
+
+ /* Unpause domain. */
+ if ( rsp.flags & MEM_EVENT_FLAG_VCPU_PAUSED )
+ mem_event_vcpu_unpause(v);
+ }
+}
+
int mem_access_memop(unsigned long cmd,
XEN_GUEST_HANDLE_PARAM(xen_mem_access_op_t) arg)
{
@@ -58,7 +82,7 @@ int mem_access_memop(unsigned long cmd,
switch ( mao.op )
{
case XENMEM_access_op_resume:
- p2m_mem_access_resume(d);
+ mem_access_resume(d);
rc = 0;
break;
diff --git a/xen/common/mem_event.c b/xen/common/mem_event.c
index b4a23fd..abafd23 100644
--- a/xen/common/mem_event.c
+++ b/xen/common/mem_event.c
@@ -430,7 +430,7 @@ static void mem_paging_notification(struct vcpu *v, unsigned int port)
static void mem_access_notification(struct vcpu *v, unsigned int port)
{
if ( likely(v->domain->mem_event->access.ring_page != NULL) )
- p2m_mem_access_resume(v->domain);
+ mem_access_resume(v->domain);
}
/* Registered with Xen-bound event channel for incoming notifications. */
diff --git a/xen/include/asm-x86/p2m.h b/xen/include/asm-x86/p2m.h
index 39f235d..2a8970b 100644
--- a/xen/include/asm-x86/p2m.h
+++ b/xen/include/asm-x86/p2m.h
@@ -601,8 +601,6 @@ void p2m_mem_paging_resume(struct domain *d);
bool_t p2m_mem_access_check(paddr_t gpa, unsigned long gla,
struct npfec npfec,
mem_event_request_t **req_ptr);
-/* Resumes the running of the VCPU, restarting the last instruction */
-void p2m_mem_access_resume(struct domain *d);
/* Set access type for a region of pfns.
* If start_pfn == -1ul, sets the default access type */
diff --git a/xen/include/xen/mem_access.h b/xen/include/xen/mem_access.h
index 19d1a2d..a58a41c 100644
--- a/xen/include/xen/mem_access.h
+++ b/xen/include/xen/mem_access.h
@@ -30,6 +30,7 @@
int mem_access_memop(unsigned long cmd,
XEN_GUEST_HANDLE_PARAM(xen_mem_access_op_t) arg);
int mem_access_send_req(struct domain *d, mem_event_request_t *req);
+void mem_access_resume(struct domain *d);
#else
@@ -46,6 +47,8 @@ int mem_access_send_req(struct domain *d, mem_event_request_t *req)
return -ENOSYS;
}
+static inline void mem_access_resume(struct domain *d) {}
+
#endif /* HAS_MEM_ACCESS */
#endif /* _XEN_ASM_MEM_ACCESS_H */
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 03/17] xen: Relocate struct npfec definition into common
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 01/17] xen: Relocate mem_access and mem_event into common Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 02/17] xen: Relocate p2m_mem_access_resume to mem_access common Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 04/17] xen: Relocate mem_event_op domctl and access_op memop " Tamas K Lengyel
` (14 subsequent siblings)
17 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Nested page fault exception code definitions can be reused on ARM as well.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
Acked-by: Jan Beulich <jbeulich@suse.com>
---
xen/include/asm-x86/hvm/hvm.h | 2 +-
xen/include/asm-x86/mm.h | 21 ---------------------
xen/include/xen/mm.h | 21 +++++++++++++++++++++
3 files changed, 22 insertions(+), 22 deletions(-)
diff --git a/xen/include/asm-x86/hvm/hvm.h b/xen/include/asm-x86/hvm/hvm.h
index 74e66f8..895f42b 100644
--- a/xen/include/asm-x86/hvm/hvm.h
+++ b/xen/include/asm-x86/hvm/hvm.h
@@ -27,7 +27,7 @@
#include <public/domctl.h>
#include <public/hvm/save.h>
#include <public/hvm/ioreq.h>
-#include <asm/mm.h>
+#include <xen/mm.h>
/* Interrupt acknowledgement sources. */
enum hvm_intsrc {
diff --git a/xen/include/asm-x86/mm.h b/xen/include/asm-x86/mm.h
index ebd482d..bafd28c 100644
--- a/xen/include/asm-x86/mm.h
+++ b/xen/include/asm-x86/mm.h
@@ -551,27 +551,6 @@ void audit_domains(void);
#endif
-/*
- * Extra fault info types which are used to further describe
- * the source of an access violation.
- */
-typedef enum {
- npfec_kind_unknown, /* must be first */
- npfec_kind_in_gpt, /* violation in guest page table */
- npfec_kind_with_gla /* violation with guest linear address */
-} npfec_kind_t;
-
-/*
- * Nested page fault exception codes.
- */
-struct npfec {
- unsigned int read_access:1;
- unsigned int write_access:1;
- unsigned int insn_fetch:1;
- unsigned int gla_valid:1;
- unsigned int kind:2; /* npfec_kind_t */
-};
-
int new_guest_cr3(unsigned long pfn);
void make_cr3(struct vcpu *v, unsigned long mfn);
void update_cr3(struct vcpu *v);
diff --git a/xen/include/xen/mm.h b/xen/include/xen/mm.h
index 7c0efc7..74a65a6 100644
--- a/xen/include/xen/mm.h
+++ b/xen/include/xen/mm.h
@@ -88,6 +88,27 @@ int assign_pages(
/* Dump info to serial console */
void arch_dump_shared_mem_info(void);
+/*
+ * Extra fault info types which are used to further describe
+ * the source of an access violation.
+ */
+typedef enum {
+ npfec_kind_unknown, /* must be first */
+ npfec_kind_in_gpt, /* violation in guest page table */
+ npfec_kind_with_gla /* violation with guest linear address */
+} npfec_kind_t;
+
+/*
+ * Nested page fault exception codes.
+ */
+struct npfec {
+ unsigned int read_access:1;
+ unsigned int write_access:1;
+ unsigned int insn_fetch:1;
+ unsigned int gla_valid:1;
+ unsigned int kind:2; /* npfec_kind_t */
+};
+
/* memflags: */
#define _MEMF_no_refcount 0
#define MEMF_no_refcount (1U<<_MEMF_no_refcount)
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 04/17] xen: Relocate mem_event_op domctl and access_op memop into common.
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (2 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 03/17] xen: Relocate struct npfec definition into common Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:44 ` Jan Beulich
2014-09-10 13:28 ` [PATCH v5 05/17] xen/mem_event: Clean out superfluous white-spaces Tamas K Lengyel
` (13 subsequent siblings)
17 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v5: Move memop compat into common as well.
Position domctl in switch relative to the domctl #.
v4: Don't remove memop handling from x86_64/compat and style fixes.
---
xen/arch/x86/domctl.c | 8 --------
xen/arch/x86/x86_64/compat/mm.c | 4 ----
xen/arch/x86/x86_64/mm.c | 4 ----
xen/common/compat/memory.c | 5 +++++
xen/common/domctl.c | 7 +++++++
xen/common/memory.c | 5 +++++
6 files changed, 17 insertions(+), 16 deletions(-)
diff --git a/xen/arch/x86/domctl.c b/xen/arch/x86/domctl.c
index 26a3ea1..166cfb3 100644
--- a/xen/arch/x86/domctl.c
+++ b/xen/arch/x86/domctl.c
@@ -1131,14 +1131,6 @@ long arch_do_domctl(
}
break;
- case XEN_DOMCTL_mem_event_op:
- {
- ret = mem_event_domctl(d, &domctl->u.mem_event_op,
- guest_handle_cast(u_domctl, void));
- copyback = 1;
- }
- break;
-
case XEN_DOMCTL_mem_sharing_op:
{
ret = mem_sharing_domctl(d, &domctl->u.mem_sharing_op);
diff --git a/xen/arch/x86/x86_64/compat/mm.c b/xen/arch/x86/x86_64/compat/mm.c
index c079702..54f25b7 100644
--- a/xen/arch/x86/x86_64/compat/mm.c
+++ b/xen/arch/x86/x86_64/compat/mm.c
@@ -198,10 +198,6 @@ int compat_arch_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
break;
}
- case XENMEM_access_op:
- rc = mem_access_memop(cmd, guest_handle_cast(arg, xen_mem_access_op_t));
- break;
-
case XENMEM_sharing_op:
{
xen_mem_sharing_op_t mso;
diff --git a/xen/arch/x86/x86_64/mm.c b/xen/arch/x86/x86_64/mm.c
index 0da6ddc..0aaa460 100644
--- a/xen/arch/x86/x86_64/mm.c
+++ b/xen/arch/x86/x86_64/mm.c
@@ -1048,10 +1048,6 @@ long subarch_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
break;
}
- case XENMEM_access_op:
- rc = mem_access_memop(cmd, guest_handle_cast(arg, xen_mem_access_op_t));
- break;
-
case XENMEM_sharing_op:
{
xen_mem_sharing_op_t mso;
diff --git a/xen/common/compat/memory.c b/xen/common/compat/memory.c
index 25dc016..43d02bc 100644
--- a/xen/common/compat/memory.c
+++ b/xen/common/compat/memory.c
@@ -4,6 +4,7 @@
#include <xen/guest_access.h>
#include <xen/sched.h>
#include <xen/event.h>
+#include <xen/mem_access.h>
#include <asm/current.h>
#include <compat/memory.h>
@@ -381,6 +382,10 @@ int compat_memory_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) compat)
break;
}
+ case XENMEM_access_op:
+ rc = mem_access_memop(cmd, guest_handle_cast(compat, xen_mem_access_op_t));
+ break;
+
case XENMEM_add_to_physmap_batch:
start_extent = end_extent;
break;
diff --git a/xen/common/domctl.c b/xen/common/domctl.c
index 222b140..638080f 100644
--- a/xen/common/domctl.c
+++ b/xen/common/domctl.c
@@ -24,6 +24,7 @@
#include <xen/bitmap.h>
#include <xen/paging.h>
#include <xen/hypercall.h>
+#include <xen/mem_event.h>
#include <asm/current.h>
#include <asm/irq.h>
#include <asm/page.h>
@@ -1126,6 +1127,12 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
}
break;
+ case XEN_DOMCTL_mem_event_op:
+ ret = mem_event_domctl(d, &op->u.mem_event_op,
+ guest_handle_cast(u_domctl, void));
+ copyback = 1;
+ break;
+
case XEN_DOMCTL_disable_migrate:
{
d->disable_migrate = op->u.disable_migrate.disable;
diff --git a/xen/common/memory.c b/xen/common/memory.c
index 44a8106..98b1ee6 100644
--- a/xen/common/memory.c
+++ b/xen/common/memory.c
@@ -25,6 +25,7 @@
#include <asm/hardirq.h>
#include <asm/p2m.h>
#include <xen/numa.h>
+#include <xen/mem_access.h>
#include <public/memory.h>
#include <xsm/xsm.h>
#include <xen/trace.h>
@@ -943,6 +944,10 @@ long do_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
break;
}
+ case XENMEM_access_op:
+ rc = mem_access_memop(cmd, guest_handle_cast(arg, xen_mem_access_op_t));
+ break;
+
case XENMEM_claim_pages:
if ( copy_from_guest(&reservation, arg, 1) )
return -EFAULT;
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 05/17] xen/mem_event: Clean out superfluous white-spaces
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (3 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 04/17] xen: Relocate mem_event_op domctl and access_op memop " Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 06/17] xen/mem_event: Relax error condition on debug builds Tamas K Lengyel
` (12 subsequent siblings)
17 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
Acked-by: Tim Deegan <tim@xen.org>
---
v2: Clean the mem_event header as well.
---
xen/common/mem_event.c | 20 ++++++++++----------
xen/include/xen/mem_event.h | 8 ++++----
2 files changed, 14 insertions(+), 14 deletions(-)
diff --git a/xen/common/mem_event.c b/xen/common/mem_event.c
index abafd23..bf1c415 100644
--- a/xen/common/mem_event.c
+++ b/xen/common/mem_event.c
@@ -57,7 +57,7 @@ static int mem_event_enable(
if ( med->ring_page )
return -EBUSY;
- /* The parameter defaults to zero, and it should be
+ /* The parameter defaults to zero, and it should be
* set to something */
if ( ring_gfn == 0 )
return -ENOSYS;
@@ -65,7 +65,7 @@ static int mem_event_enable(
mem_event_ring_lock_init(med);
mem_event_ring_lock(med);
- rc = prepare_ring_for_helper(d, ring_gfn, &med->ring_pg_struct,
+ rc = prepare_ring_for_helper(d, ring_gfn, &med->ring_pg_struct,
&med->ring_page);
if ( rc < 0 )
goto err;
@@ -97,7 +97,7 @@ static int mem_event_enable(
return 0;
err:
- destroy_ring_for_helper(&med->ring_page,
+ destroy_ring_for_helper(&med->ring_page,
med->ring_pg_struct);
mem_event_ring_unlock(med);
@@ -226,7 +226,7 @@ static int mem_event_disable(struct domain *d, struct mem_event_domain *med)
}
}
- destroy_ring_for_helper(&med->ring_page,
+ destroy_ring_for_helper(&med->ring_page,
med->ring_pg_struct);
mem_event_ring_unlock(med);
}
@@ -479,7 +479,7 @@ void mem_event_cleanup(struct domain *d)
* the disable routine to complete. It will also drop
* all domain refs the wait-queued vcpus are holding.
* Finally, because this code path involves previously
- * pausing the domain (domain_kill), unpausing the
+ * pausing the domain (domain_kill), unpausing the
* vcpus causes no harm. */
destroy_waitqueue_head(&d->mem_event->paging.wq);
(void)mem_event_disable(d, &d->mem_event->paging);
@@ -559,7 +559,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
if ( p2m->pod.entry_count )
break;
- rc = mem_event_enable(d, mec, med, _VPF_mem_paging,
+ rc = mem_event_enable(d, mec, med, _VPF_mem_paging,
HVM_PARAM_PAGING_RING_PFN,
mem_paging_notification);
}
@@ -579,7 +579,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
}
break;
- case XEN_DOMCTL_MEM_EVENT_OP_ACCESS:
+ case XEN_DOMCTL_MEM_EVENT_OP_ACCESS:
{
struct mem_event_domain *med = &d->mem_event->access;
rc = -EINVAL;
@@ -597,7 +597,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
if ( !cpu_has_vmx )
break;
- rc = mem_event_enable(d, mec, med, _VPF_mem_access,
+ rc = mem_event_enable(d, mec, med, _VPF_mem_access,
HVM_PARAM_ACCESS_RING_PFN,
mem_access_notification);
}
@@ -617,7 +617,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
}
break;
- case XEN_DOMCTL_MEM_EVENT_OP_SHARING:
+ case XEN_DOMCTL_MEM_EVENT_OP_SHARING:
{
struct mem_event_domain *med = &d->mem_event->share;
rc = -EINVAL;
@@ -636,7 +636,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
if ( !hap_enabled(d) )
break;
- rc = mem_event_enable(d, mec, med, _VPF_mem_sharing,
+ rc = mem_event_enable(d, mec, med, _VPF_mem_sharing,
HVM_PARAM_SHARING_RING_PFN,
mem_sharing_notification);
}
diff --git a/xen/include/xen/mem_event.h b/xen/include/xen/mem_event.h
index 8612b26..4f3ad8e 100644
--- a/xen/include/xen/mem_event.h
+++ b/xen/include/xen/mem_event.h
@@ -37,19 +37,19 @@ bool_t mem_event_check_ring(struct mem_event_domain *med);
/* Returns 0 on success, -ENOSYS if there is no ring, -EBUSY if there is no
* available space and the caller is a foreign domain. If the guest itself
* is the caller, -EBUSY is avoided by sleeping on a wait queue to ensure
- * that the ring does not lose future events.
+ * that the ring does not lose future events.
*
* However, the allow_sleep flag can be set to false in cases in which it is ok
* to lose future events, and thus -EBUSY can be returned to guest vcpus
- * (handle with care!).
+ * (handle with care!).
*
* In general, you must follow a claim_slot() call with either put_request() or
* cancel_slot(), both of which are guaranteed to
- * succeed.
+ * succeed.
*/
int __mem_event_claim_slot(struct domain *d, struct mem_event_domain *med,
bool_t allow_sleep);
-static inline int mem_event_claim_slot(struct domain *d,
+static inline int mem_event_claim_slot(struct domain *d,
struct mem_event_domain *med)
{
return __mem_event_claim_slot(d, med, 1);
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 06/17] xen/mem_event: Relax error condition on debug builds
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (4 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 05/17] xen/mem_event: Clean out superfluous white-spaces Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 07/17] xen/mem_event: Abstract architecture specific sanity checks Tamas K Lengyel
` (11 subsequent siblings)
17 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
A faulty tool stack can brick a debug hypervisor. Unpleasant while dev/test.
Suggested-by: Andres Lagar Cavilla <andres@lagarcavilla.org>
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v5: Make printout greppable.
v4: Add domain_id to the printout.
v3: Switch to gdprintk and print the vCPU id as well.
---
xen/common/mem_event.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/xen/common/mem_event.c b/xen/common/mem_event.c
index bf1c415..b232189 100644
--- a/xen/common/mem_event.c
+++ b/xen/common/mem_event.c
@@ -278,7 +278,11 @@ void mem_event_put_request(struct domain *d,
if ( current->domain != d )
{
req->flags |= MEM_EVENT_FLAG_FOREIGN;
- ASSERT( !(req->flags & MEM_EVENT_FLAG_VCPU_PAUSED) );
+#ifndef NDEBUG
+ if ( !(req->flags & MEM_EVENT_FLAG_VCPU_PAUSED) )
+ gdprintk(XENLOG_G_WARNING, "d%dv%d was not paused.\n",
+ d->domain_id, req->vcpu_id);
+#endif
}
mem_event_ring_lock(med);
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 07/17] xen/mem_event: Abstract architecture specific sanity checks
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (5 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 06/17] xen/mem_event: Relax error condition on debug builds Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 08/17] xen/mem_access: Abstract architecture specific sanity check Tamas K Lengyel
` (10 subsequent siblings)
17 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Move architecture specific sanity checks into its own function
which is called when enabling mem_event.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v5: Style fix.
v4: Style fix.
v2: Move sanity check function into architecture specific p2m.h
---
xen/common/mem_event.c | 7 +------
xen/include/asm-x86/p2m.h | 6 ++++++
2 files changed, 7 insertions(+), 6 deletions(-)
diff --git a/xen/common/mem_event.c b/xen/common/mem_event.c
index b232189..e22a9ab 100644
--- a/xen/common/mem_event.c
+++ b/xen/common/mem_event.c
@@ -593,12 +593,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
case XEN_DOMCTL_MEM_EVENT_OP_ACCESS_ENABLE:
{
rc = -ENODEV;
- /* Only HAP is supported */
- if ( !hap_enabled(d) )
- break;
-
- /* Currently only EPT is supported */
- if ( !cpu_has_vmx )
+ if ( !p2m_mem_event_sanity_check(d) )
break;
rc = mem_event_enable(d, mec, med, _VPF_mem_access,
diff --git a/xen/include/asm-x86/p2m.h b/xen/include/asm-x86/p2m.h
index 2a8970b..02d23d1 100644
--- a/xen/include/asm-x86/p2m.h
+++ b/xen/include/asm-x86/p2m.h
@@ -612,6 +612,12 @@ long p2m_set_mem_access(struct domain *d, unsigned long start_pfn, uint32_t nr,
int p2m_get_mem_access(struct domain *d, unsigned long pfn,
xenmem_access_t *access);
+/* Sanity check for mem_event hardware support */
+static inline bool_t p2m_mem_event_sanity_check(struct domain *d)
+{
+ return hap_enabled(d) && cpu_has_vmx;
+}
+
/*
* Internal functions, only called by other p2m code
*/
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 08/17] xen/mem_access: Abstract architecture specific sanity check
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (6 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 07/17] xen/mem_event: Abstract architecture specific sanity checks Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 09/17] xen/arm: p2m type definitions and changes Tamas K Lengyel
` (9 subsequent siblings)
17 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v5: Style fix.
v4: Style fix.
v2: Move sanity check function into architecture specific p2m.h
---
xen/common/mem_access.c | 2 +-
xen/include/asm-x86/p2m.h | 6 ++++++
2 files changed, 7 insertions(+), 1 deletion(-)
diff --git a/xen/common/mem_access.c b/xen/common/mem_access.c
index c84fab3..fc6b207 100644
--- a/xen/common/mem_access.c
+++ b/xen/common/mem_access.c
@@ -68,7 +68,7 @@ int mem_access_memop(unsigned long cmd,
return rc;
rc = -EINVAL;
- if ( !is_hvm_domain(d) )
+ if ( !p2m_mem_access_sanity_check(d) )
goto out;
rc = xsm_mem_event_op(XSM_DM_PRIV, d, XENMEM_access_op);
diff --git a/xen/include/asm-x86/p2m.h b/xen/include/asm-x86/p2m.h
index 02d23d1..6f9ed39 100644
--- a/xen/include/asm-x86/p2m.h
+++ b/xen/include/asm-x86/p2m.h
@@ -618,6 +618,12 @@ static inline bool_t p2m_mem_event_sanity_check(struct domain *d)
return hap_enabled(d) && cpu_has_vmx;
}
+/* Sanity check for mem_access hardware support */
+static inline bool_t p2m_mem_access_sanity_check(struct domain *d)
+{
+ return is_hvm_domain(d);
+}
+
/*
* Internal functions, only called by other p2m code
*/
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (7 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 08/17] xen/mem_access: Abstract architecture specific sanity check Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-11 20:25 ` Julien Grall
2014-09-11 20:49 ` Julien Grall
2014-09-10 13:28 ` [PATCH v5 10/17] xen/arm: Add set access required domctl Tamas K Lengyel
` (8 subsequent siblings)
17 siblings, 2 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Define p2m_access_t in ARM and add necessary changes for page table
construction routines to pass the default access information. Also,
define the Radix tree that will hold access permission settings as
the PTE's don't have enough software programmable bits available.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v5: #include grouping style-fix.
v4: move p2m_get_hostp2m definition here.
---
xen/arch/arm/p2m.c | 54 ++++++++++++++++++----------
xen/include/asm-arm/p2m.h | 92 ++++++++++++++++++++++++++++++++++++-----------
2 files changed, 108 insertions(+), 38 deletions(-)
diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c
index fc8c0dd..e6b4bb6 100644
--- a/xen/arch/arm/p2m.c
+++ b/xen/arch/arm/p2m.c
@@ -5,6 +5,9 @@
#include <xen/errno.h>
#include <xen/domain_page.h>
#include <xen/bitops.h>
+#include <xen/mem_event.h>
+#include <xen/mem_access.h>
+#include <public/mem_event.h>
#include <asm/flushtlb.h>
#include <asm/gic.h>
#include <asm/event.h>
@@ -229,7 +232,7 @@ int p2m_pod_decrease_reservation(struct domain *d,
}
static lpae_t mfn_to_p2m_entry(unsigned long mfn, unsigned int mattr,
- p2m_type_t t)
+ p2m_type_t t, p2m_access_t a)
{
paddr_t pa = ((paddr_t) mfn) << PAGE_SHIFT;
/* sh, xn and write bit will be defined in the following switches
@@ -347,7 +350,7 @@ static int p2m_create_table(struct domain *d, lpae_t *entry,
for ( i=0 ; i < LPAE_ENTRIES; i++ )
{
pte = mfn_to_p2m_entry(base_pfn + (i<<(level_shift-LPAE_SHIFT)),
- MATTR_MEM, t);
+ MATTR_MEM, t, p2m->default_access);
/*
* First and second level super pages set p2m.table = 0, but
@@ -367,7 +370,8 @@ static int p2m_create_table(struct domain *d, lpae_t *entry,
unmap_domain_page(p);
- pte = mfn_to_p2m_entry(page_to_mfn(page), MATTR_MEM, p2m_invalid);
+ pte = mfn_to_p2m_entry(page_to_mfn(page), MATTR_MEM, p2m_invalid,
+ p2m->default_access);
p2m_write_pte(entry, pte, flush_cache);
@@ -470,7 +474,8 @@ static int apply_one_level(struct domain *d,
paddr_t *maddr,
bool_t *flush,
int mattr,
- p2m_type_t t)
+ p2m_type_t t,
+ p2m_access_t a)
{
const paddr_t level_size = level_sizes[level];
const paddr_t level_mask = level_masks[level];
@@ -499,7 +504,7 @@ static int apply_one_level(struct domain *d,
page = alloc_domheap_pages(d, level_shift - PAGE_SHIFT, 0);
if ( page )
{
- pte = mfn_to_p2m_entry(page_to_mfn(page), mattr, t);
+ pte = mfn_to_p2m_entry(page_to_mfn(page), mattr, t, a);
if ( level < 3 )
pte.p2m.table = 0;
p2m_write_pte(entry, pte, flush_cache);
@@ -534,7 +539,7 @@ static int apply_one_level(struct domain *d,
(level == 3 || !p2m_table(orig_pte)) )
{
/* New mapping is superpage aligned, make it */
- pte = mfn_to_p2m_entry(*maddr >> PAGE_SHIFT, mattr, t);
+ pte = mfn_to_p2m_entry(*maddr >> PAGE_SHIFT, mattr, t, a);
if ( level < 3 )
pte.p2m.table = 0; /* Superpage entry */
@@ -713,7 +718,8 @@ static int apply_p2m_changes(struct domain *d,
paddr_t end_gpaddr,
paddr_t maddr,
int mattr,
- p2m_type_t t)
+ p2m_type_t t,
+ p2m_access_t a)
{
int rc, ret;
struct p2m_domain *p2m = &d->arch.p2m;
@@ -780,7 +786,7 @@ static int apply_p2m_changes(struct domain *d,
level, flush_pt, op,
start_gpaddr, end_gpaddr,
&addr, &maddr, &flush,
- mattr, t);
+ mattr, t, a);
if ( ret < 0 ) { rc = ret ; goto out; }
count += ret;
if ( ret != P2M_ONE_DESCEND ) continue;
@@ -802,7 +808,7 @@ static int apply_p2m_changes(struct domain *d,
level, flush_pt, op,
start_gpaddr, end_gpaddr,
&addr, &maddr, &flush,
- mattr, t);
+ mattr, t, a);
if ( ret < 0 ) { rc = ret ; goto out; }
count += ret;
if ( ret != P2M_ONE_DESCEND ) continue;
@@ -822,7 +828,7 @@ static int apply_p2m_changes(struct domain *d,
level, flush_pt, op,
start_gpaddr, end_gpaddr,
&addr, &maddr, &flush,
- mattr, t);
+ mattr, t, a);
if ( ret < 0 ) { rc = ret ; goto out; }
/* L3 had better have done something! We cannot descend any further */
BUG_ON(ret == P2M_ONE_DESCEND);
@@ -865,7 +871,7 @@ out:
*/
apply_p2m_changes(d, REMOVE,
start_gpaddr, addr + level_sizes[level], orig_maddr,
- mattr, p2m_invalid);
+ mattr, p2m_invalid, d->arch.p2m.default_access);
}
spin_unlock(&p2m->lock);
@@ -878,7 +884,8 @@ int p2m_populate_ram(struct domain *d,
paddr_t end)
{
return apply_p2m_changes(d, ALLOCATE, start, end,
- 0, MATTR_MEM, p2m_ram_rw);
+ 0, MATTR_MEM, p2m_ram_rw,
+ d->arch.p2m.default_access);
}
int map_mmio_regions(struct domain *d,
@@ -890,7 +897,8 @@ int map_mmio_regions(struct domain *d,
pfn_to_paddr(start_gfn),
pfn_to_paddr(start_gfn + nr),
pfn_to_paddr(mfn),
- MATTR_DEV, p2m_mmio_direct);
+ MATTR_DEV, p2m_mmio_direct,
+ d->arch.p2m.default_access);
}
int unmap_mmio_regions(struct domain *d,
@@ -902,7 +910,8 @@ int unmap_mmio_regions(struct domain *d,
pfn_to_paddr(start_gfn),
pfn_to_paddr(start_gfn + nr),
pfn_to_paddr(mfn),
- MATTR_DEV, p2m_invalid);
+ MATTR_DEV, p2m_invalid,
+ d->arch.p2m.default_access);
}
int guest_physmap_add_entry(struct domain *d,
@@ -914,7 +923,8 @@ int guest_physmap_add_entry(struct domain *d,
return apply_p2m_changes(d, INSERT,
pfn_to_paddr(gpfn),
pfn_to_paddr(gpfn + (1 << page_order)),
- pfn_to_paddr(mfn), MATTR_MEM, t);
+ pfn_to_paddr(mfn), MATTR_MEM, t,
+ d->arch.p2m.default_access);
}
void guest_physmap_remove_page(struct domain *d,
@@ -924,7 +934,8 @@ void guest_physmap_remove_page(struct domain *d,
apply_p2m_changes(d, REMOVE,
pfn_to_paddr(gpfn),
pfn_to_paddr(gpfn + (1<<page_order)),
- pfn_to_paddr(mfn), MATTR_MEM, p2m_invalid);
+ pfn_to_paddr(mfn), MATTR_MEM, p2m_invalid,
+ d->arch.p2m.default_access);
}
int p2m_alloc_table(struct domain *d)
@@ -1027,6 +1038,8 @@ void p2m_teardown(struct domain *d)
p2m_free_vmid(d);
+ radix_tree_destroy(&p2m->mem_access_settings, NULL);
+
spin_unlock(&p2m->lock);
}
@@ -1052,6 +1065,9 @@ int p2m_init(struct domain *d)
p2m->max_mapped_gfn = 0;
p2m->lowest_mapped_gfn = ULONG_MAX;
+ p2m->default_access = p2m_access_rwx;
+ radix_tree_init(&p2m->mem_access_settings);
+
err:
spin_unlock(&p2m->lock);
@@ -1066,7 +1082,8 @@ int relinquish_p2m_mapping(struct domain *d)
pfn_to_paddr(p2m->lowest_mapped_gfn),
pfn_to_paddr(p2m->max_mapped_gfn),
pfn_to_paddr(INVALID_MFN),
- MATTR_MEM, p2m_invalid);
+ MATTR_MEM, p2m_invalid,
+ d->arch.p2m.default_access);
}
int p2m_cache_flush(struct domain *d, xen_pfn_t start_mfn, xen_pfn_t end_mfn)
@@ -1080,7 +1097,8 @@ int p2m_cache_flush(struct domain *d, xen_pfn_t start_mfn, xen_pfn_t end_mfn)
pfn_to_paddr(start_mfn),
pfn_to_paddr(end_mfn),
pfn_to_paddr(INVALID_MFN),
- MATTR_MEM, p2m_invalid);
+ MATTR_MEM, p2m_invalid,
+ d->arch.p2m.default_access);
}
unsigned long gmfn_to_mfn(struct domain *d, unsigned long gpfn)
diff --git a/xen/include/asm-arm/p2m.h b/xen/include/asm-arm/p2m.h
index 08ce07b..b4ca86d 100644
--- a/xen/include/asm-arm/p2m.h
+++ b/xen/include/asm-arm/p2m.h
@@ -2,6 +2,9 @@
#define _XEN_P2M_H
#include <xen/mm.h>
+#include <xen/radix-tree.h>
+#include <public/memory.h>
+#include <public/mem_event.h>
#include <xen/p2m-common.h>
@@ -11,6 +14,48 @@ struct domain;
extern void memory_type_changed(struct domain *);
+/* List of possible type for each page in the p2m entry.
+ * The number of available bit per page in the pte for this purpose is 4 bits.
+ * So it's possible to only have 16 fields. If we run out of value in the
+ * future, it's possible to use higher value for pseudo-type and don't store
+ * them in the p2m entry.
+ */
+typedef enum {
+ p2m_invalid = 0, /* Nothing mapped here */
+ p2m_ram_rw, /* Normal read/write guest RAM */
+ p2m_ram_ro, /* Read-only; writes are silently dropped */
+ p2m_mmio_direct, /* Read/write mapping of genuine MMIO area */
+ p2m_map_foreign, /* Ram pages from foreign domain */
+ p2m_grant_map_rw, /* Read/write grant mapping */
+ p2m_grant_map_ro, /* Read-only grant mapping */
+ /* The types below are only used to decide the page attribute in the P2M */
+ p2m_iommu_map_rw, /* Read/write iommu mapping */
+ p2m_iommu_map_ro, /* Read-only iommu mapping */
+ p2m_max_real_type, /* Types after this won't be store in the p2m */
+} p2m_type_t;
+
+/*
+ * Additional access types, which are used to further restrict
+ * the permissions given by the p2m_type_t memory type. Violations
+ * caused by p2m_access_t restrictions are sent to the mem_event
+ * interface.
+ *
+ * The access permissions are soft state: when any ambigious change of page
+ * type or use occurs, or when pages are flushed, swapped, or at any other
+ * convenient type, the access permissions can get reset to the p2m_domain
+ * default.
+ */
+typedef enum {
+ p2m_access_n = 0, /* No access permissions allowed */
+ p2m_access_r = 1,
+ p2m_access_w = 2,
+ p2m_access_rw = 3,
+ p2m_access_x = 4,
+ p2m_access_rx = 5,
+ p2m_access_wx = 6,
+ p2m_access_rwx = 7
+} p2m_access_t;
+
/* Per-p2m-table state */
struct p2m_domain {
/* Lock that protects updates to the p2m */
@@ -44,27 +89,20 @@ struct p2m_domain {
* at each p2m tree level. */
unsigned long shattered[4];
} stats;
-};
-/* List of possible type for each page in the p2m entry.
- * The number of available bit per page in the pte for this purpose is 4 bits.
- * So it's possible to only have 16 fields. If we run out of value in the
- * future, it's possible to use higher value for pseudo-type and don't store
- * them in the p2m entry.
- */
-typedef enum {
- p2m_invalid = 0, /* Nothing mapped here */
- p2m_ram_rw, /* Normal read/write guest RAM */
- p2m_ram_ro, /* Read-only; writes are silently dropped */
- p2m_mmio_direct, /* Read/write mapping of genuine MMIO area */
- p2m_map_foreign, /* Ram pages from foreign domain */
- p2m_grant_map_rw, /* Read/write grant mapping */
- p2m_grant_map_ro, /* Read-only grant mapping */
- /* The types below are only used to decide the page attribute in the P2M */
- p2m_iommu_map_rw, /* Read/write iommu mapping */
- p2m_iommu_map_ro, /* Read-only iommu mapping */
- p2m_max_real_type, /* Types after this won't be store in the p2m */
-} p2m_type_t;
+ /* Default P2M access type for each page in the the domain: new pages,
+ * swapped in pages, cleared pages, and pages that are ambiquously
+ * retyped get this access type. See definition of p2m_access_t. */
+ p2m_access_t default_access;
+
+ /* If true, and an access fault comes in and there is no mem_event listener,
+ * pause domain. Otherwise, remove access restrictions. */
+ bool_t access_required;
+
+ /* Radix tree to store the p2m_access_t settings as the pte's don't have
+ * enough available bits to store this information. */
+ struct radix_tree_root mem_access_settings;
+};
#define p2m_is_foreign(_t) ((_t) == p2m_map_foreign)
#define p2m_is_ram(_t) ((_t) == p2m_ram_rw || (_t) == p2m_ram_ro)
@@ -194,6 +232,20 @@ static inline int get_page_and_type(struct page_info *page,
return rc;
}
+/* get host p2m table */
+#define p2m_get_hostp2m(d) (&((d)->arch.p2m))
+
+/* mem_event and mem_access are supported on all ARM */
+static inline bool_t p2m_mem_access_sanity_check(struct domain *d)
+{
+ return 1;
+}
+
+static inline bool_t p2m_mem_event_sanity_check(struct domain *d)
+{
+ return 1;
+}
+
#endif /* _XEN_P2M_H */
/*
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 10/17] xen/arm: Add set access required domctl
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (8 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 09/17] xen/arm: p2m type definitions and changes Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-11 20:26 ` Julien Grall
2014-09-10 13:28 ` [PATCH v5 11/17] xen/arm: Implement domain_get_maximum_gpfn Tamas K Lengyel
` (7 subsequent siblings)
17 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
xen/arch/arm/domctl.c | 13 +++++++++++++
1 file changed, 13 insertions(+)
diff --git a/xen/arch/arm/domctl.c b/xen/arch/arm/domctl.c
index 45974e7..7cf719c 100644
--- a/xen/arch/arm/domctl.c
+++ b/xen/arch/arm/domctl.c
@@ -31,6 +31,19 @@ long arch_do_domctl(struct xen_domctl *domctl, struct domain *d,
return p2m_cache_flush(d, s, e);
}
+ case XEN_DOMCTL_set_access_required:
+ {
+ struct p2m_domain* p2m;
+
+ if ( current->domain == d )
+ return -EPERM;
+
+ p2m = p2m_get_hostp2m(d);
+ p2m->access_required = domctl->u.access_required.access_required;
+ return 0;
+ }
+ break;
+
default:
return subarch_do_domctl(domctl, d, u_domctl);
}
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 11/17] xen/arm: Implement domain_get_maximum_gpfn
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (9 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 10/17] xen/arm: Add set access required domctl Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-11 20:28 ` Julien Grall
2014-09-10 13:28 ` [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events Tamas K Lengyel
` (6 subsequent siblings)
17 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
The function domain_get_maximum_gpfn is returning the maximum gpfn ever
mapped in the guest. We can use d->arch.p2m.max_mapped_gfn for this purpose.
Suggested-by: Julien Grall <julien.grall@linaro.org>
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
xen/arch/arm/mm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
index 0a243b0..e4a1e5e 100644
--- a/xen/arch/arm/mm.c
+++ b/xen/arch/arm/mm.c
@@ -954,7 +954,7 @@ int page_is_ram_type(unsigned long mfn, unsigned long mem_type)
unsigned long domain_get_maximum_gpfn(struct domain *d)
{
- return -ENOSYS;
+ return d->arch.p2m.max_mapped_gfn;
}
void share_xen_page_with_guest(struct page_info *page,
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events.
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (10 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 11/17] xen/arm: Implement domain_get_maximum_gpfn Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-11 21:19 ` Julien Grall
2014-09-10 13:28 ` [PATCH v5 13/17] xen/arm: Instruction prefetch abort (X) mem_event handling Tamas K Lengyel
` (5 subsequent siblings)
17 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
This patch enables to store, set, check and deliver LPAE R/W mem_events.
As the LPAE PTE's lack enough available software programmable bits,
we store the permissions in a Radix tree, where the key is the pfn
of a 4k page. Only settings other than p2m_access_rwx are saved
in the Radix tree.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v5: - Move p2m_set_entry's logic into apply_one_level via
a new p2m_op, MEMACCESS.
v4: - Add p2m_mem_access_radix_set function to be called
when inserting new PTE's and when updating existing entries.
- Switch p2m_mem_access_check to return bool_t.
- Use new struct npfec to pass violation info.
v3: - Add new function for updating the PTE entries, p2m_set_entry.
- Use the new struct npfec to pass violation information.
- Implement n2rwx, rx2rw and listener required routines.
v2: - Patch been split to ease the review process.
- Add definitions of data abort data fetch status codes (enum dabt_dfsc)
and only call p2m_mem_access_check for traps caused by permission violations.
- Only call p2m_write_pte in p2m_lookup if the PTE permission actually changed.
- Properly save settings in the Radix tree and pause the VCPU with
mem_event_vcpu_pause.
---
xen/arch/arm/p2m.c | 412 +++++++++++++++++++++++++++++++++++++---
xen/arch/arm/traps.c | 31 ++-
xen/include/asm-arm/p2m.h | 17 ++
xen/include/asm-arm/processor.h | 30 +++
4 files changed, 456 insertions(+), 34 deletions(-)
diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c
index e6b4bb6..e089d52 100644
--- a/xen/arch/arm/p2m.c
+++ b/xen/arch/arm/p2m.c
@@ -152,6 +152,74 @@ static lpae_t *p2m_map_first(struct p2m_domain *p2m, paddr_t addr)
return __map_domain_page(page);
}
+static void p2m_set_permission(lpae_t *e, p2m_type_t t, p2m_access_t a)
+{
+ /* First apply type permissions */
+ switch ( t )
+ {
+ case p2m_ram_rw:
+ e->p2m.xn = 0;
+ e->p2m.write = 1;
+ break;
+
+ case p2m_ram_ro:
+ e->p2m.xn = 0;
+ e->p2m.write = 0;
+ break;
+
+ case p2m_iommu_map_rw:
+ case p2m_map_foreign:
+ case p2m_grant_map_rw:
+ case p2m_mmio_direct:
+ e->p2m.xn = 1;
+ e->p2m.write = 1;
+ break;
+
+ case p2m_iommu_map_ro:
+ case p2m_grant_map_ro:
+ case p2m_invalid:
+ e->p2m.xn = 1;
+ e->p2m.write = 0;
+ break;
+
+ case p2m_max_real_type:
+ BUG();
+ break;
+ }
+
+ /* Then restrict with access permissions */
+ switch ( a )
+ {
+ case p2m_access_n:
+ e->p2m.read = e->p2m.write = 0;
+ e->p2m.xn = 1;
+ break;
+ case p2m_access_r:
+ e->p2m.write = 0;
+ e->p2m.xn = 1;
+ break;
+ case p2m_access_x:
+ e->p2m.write = 0;
+ e->p2m.read = 0;
+ break;
+ case p2m_access_rx:
+ e->p2m.write = 0;
+ break;
+ case p2m_access_w:
+ e->p2m.read = 0;
+ e->p2m.xn = 1;
+ break;
+ case p2m_access_rw:
+ e->p2m.xn = 1;
+ break;
+ case p2m_access_wx:
+ e->p2m.read = 0;
+ break;
+ case p2m_access_rwx:
+ break;
+ }
+}
+
/*
* Lookup the MFN corresponding to a domain's PFN.
*
@@ -262,37 +330,7 @@ static lpae_t mfn_to_p2m_entry(unsigned long mfn, unsigned int mattr,
break;
}
- switch (t)
- {
- case p2m_ram_rw:
- e.p2m.xn = 0;
- e.p2m.write = 1;
- break;
-
- case p2m_ram_ro:
- e.p2m.xn = 0;
- e.p2m.write = 0;
- break;
-
- case p2m_iommu_map_rw:
- case p2m_map_foreign:
- case p2m_grant_map_rw:
- case p2m_mmio_direct:
- e.p2m.xn = 1;
- e.p2m.write = 1;
- break;
-
- case p2m_iommu_map_ro:
- case p2m_grant_map_ro:
- case p2m_invalid:
- e.p2m.xn = 1;
- e.p2m.write = 0;
- break;
-
- case p2m_max_real_type:
- BUG();
- break;
- }
+ p2m_set_permission(&e, t, a);
ASSERT(!(pa & ~PAGE_MASK));
ASSERT(!(pa & ~PADDR_MASK));
@@ -384,6 +422,7 @@ enum p2m_operation {
REMOVE,
RELINQUISH,
CACHEFLUSH,
+ MEMACCESS,
};
/* Put any references on the single 4K page referenced by pte. TODO:
@@ -444,6 +483,33 @@ static bool_t is_mapping_aligned(const paddr_t start_gpaddr,
return true;
}
+static long p2m_mem_access_radix_set(struct p2m_domain *p2m, unsigned long pfn,
+ p2m_access_t a)
+{
+ long rc;
+
+ if ( p2m_access_rwx == a )
+ {
+ radix_tree_delete(&p2m->mem_access_settings, pfn);
+ return 0;
+ }
+
+ rc = radix_tree_insert(&p2m->mem_access_settings, pfn,
+ radix_tree_int_to_ptr(a));
+
+ if ( -EEXIST == rc )
+ {
+ /* If a setting existed already, change it to the new one */
+ radix_tree_replace_slot(
+ radix_tree_lookup_slot(
+ &p2m->mem_access_settings, pfn),
+ radix_tree_int_to_ptr(a));
+ rc = 0;
+ }
+
+ return rc;
+}
+
#define P2M_ONE_DESCEND 0
#define P2M_ONE_PROGRESS_NOP 0x1
#define P2M_ONE_PROGRESS 0x10
@@ -504,6 +570,10 @@ static int apply_one_level(struct domain *d,
page = alloc_domheap_pages(d, level_shift - PAGE_SHIFT, 0);
if ( page )
{
+ rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr), a);
+ if ( rc < 0 )
+ return rc;
+
pte = mfn_to_p2m_entry(page_to_mfn(page), mattr, t, a);
if ( level < 3 )
pte.p2m.table = 0;
@@ -538,6 +608,10 @@ static int apply_one_level(struct domain *d,
/* We do not handle replacing an existing table with a superpage */
(level == 3 || !p2m_table(orig_pte)) )
{
+ rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr), a);
+ if ( rc < 0 )
+ return rc;
+
/* New mapping is superpage aligned, make it */
pte = mfn_to_p2m_entry(*maddr >> PAGE_SHIFT, mattr, t, a);
if ( level < 3 )
@@ -663,6 +737,7 @@ static int apply_one_level(struct domain *d,
memset(&pte, 0x00, sizeof(pte));
p2m_write_pte(entry, pte, flush_cache);
+ radix_tree_delete(&p2m->mem_access_settings, paddr_to_pfn(*addr));
*addr += level_size;
*maddr += level_size;
@@ -707,6 +782,53 @@ static int apply_one_level(struct domain *d,
*addr += PAGE_SIZE;
return P2M_ONE_PROGRESS_NOP;
}
+
+ case MEMACCESS:
+ if ( level < 3 )
+ {
+ if ( !p2m_valid(orig_pte) )
+ {
+ (*addr)++;
+ return P2M_ONE_PROGRESS_NOP;
+ }
+
+ /* Shatter large pages as we descend */
+ if ( p2m_mapping(orig_pte) )
+ {
+ rc = p2m_create_table(d, entry,
+ level_shift - PAGE_SHIFT, flush_cache);
+ if ( rc < 0 )
+ return rc;
+
+ p2m->stats.shattered[level]++;
+ p2m->stats.mappings[level]--;
+ p2m->stats.mappings[level+1] += LPAE_ENTRIES;
+ } /* else: an existing table mapping -> descend */
+
+ return P2M_ONE_DESCEND;
+ }
+ else
+ {
+ pte = orig_pte;
+
+ if ( !p2m_table(pte) )
+ pte.bits = 0;
+
+ if ( p2m_valid(pte) )
+ {
+ ASSERT(pte.p2m.type != p2m_invalid);
+
+ rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr), a);
+ if ( rc < 0 )
+ return rc;
+
+ p2m_set_permission(&pte, pte.p2m.type, a);
+ p2m_write_pte(entry, pte, flush_cache);
+ }
+
+ (*addr)++;
+ return P2M_ONE_PROGRESS;
+ }
}
BUG(); /* Should never get here */
@@ -1135,6 +1257,234 @@ err:
return page;
}
+bool_t p2m_mem_access_check(paddr_t gpa, vaddr_t gla, struct npfec npfec)
+{
+ struct vcpu *v = current;
+ struct p2m_domain *p2m = p2m_get_hostp2m(v->domain);
+ mem_event_request_t *req = NULL;
+ xenmem_access_t xma;
+ bool_t violation;
+ int rc;
+
+ rc = p2m_get_mem_access(v->domain, paddr_to_pfn(gpa), &xma);
+ if ( rc )
+ {
+ /* No setting was found, reinject */
+ return 1;
+ }
+ else
+ {
+ /* First, handle rx2rw and n2rwx conversion automatically. */
+ if ( npfec.write_access && xma == XENMEM_access_rx2rw )
+ {
+ rc = p2m_set_mem_access(v->domain, paddr_to_pfn(gpa), 1,
+ 0, ~0, XENMEM_access_rw);
+ return 0;
+ }
+ else if ( xma == XENMEM_access_n2rwx )
+ {
+ rc = p2m_set_mem_access(v->domain, paddr_to_pfn(gpa), 1,
+ 0, ~0, XENMEM_access_rwx);
+ }
+ }
+
+ /* Otherwise, check if there is a memory event listener, and send the message along */
+ if ( !mem_event_check_ring( &v->domain->mem_event->access ) )
+ {
+ /* No listener */
+ if ( p2m->access_required )
+ {
+ gdprintk(XENLOG_INFO, "Memory access permissions failure, "
+ "no mem_event listener VCPU %d, dom %d\n",
+ v->vcpu_id, v->domain->domain_id);
+ domain_crash(v->domain);
+ }
+ else
+ {
+ /* n2rwx was already handled */
+ if ( xma != XENMEM_access_n2rwx)
+ {
+ /* A listener is not required, so clear the access
+ * restrictions. */
+ rc = p2m_set_mem_access(v->domain, paddr_to_pfn(gpa), 1,
+ 0, ~0, XENMEM_access_rwx);
+ }
+ }
+
+ /* No need to reinject */
+ return 0;
+ }
+
+ switch ( xma )
+ {
+ default:
+ case XENMEM_access_n:
+ violation = npfec.read_access || npfec.write_access || npfec.insn_fetch;
+ break;
+ case XENMEM_access_r:
+ violation = npfec.write_access || npfec.insn_fetch;
+ break;
+ case XENMEM_access_w:
+ violation = npfec.read_access || npfec.insn_fetch;
+ break;
+ case XENMEM_access_x:
+ violation = npfec.read_access || npfec.write_access;
+ break;
+ case XENMEM_access_rx:
+ violation = npfec.write_access;
+ break;
+ case XENMEM_access_wx:
+ violation = npfec.read_access;
+ break;
+ case XENMEM_access_rw:
+ violation = npfec.insn_fetch;
+ break;
+ case XENMEM_access_rwx:
+ violation = 0;
+ break;
+ }
+
+ if ( !violation )
+ return 1;
+
+ req = xzalloc(mem_event_request_t);
+ if ( req )
+ {
+ req->reason = MEM_EVENT_REASON_VIOLATION;
+ if ( xma != XENMEM_access_n2rwx )
+ req->flags |= MEM_EVENT_FLAG_VCPU_PAUSED;
+ req->gfn = gpa >> PAGE_SHIFT;
+ req->offset = gpa & ((1 << PAGE_SHIFT) - 1);
+ req->gla = gla;
+ req->gla_valid = npfec.gla_valid;
+ req->access_r = npfec.read_access;
+ req->access_w = npfec.write_access;
+ req->access_x = npfec.insn_fetch;
+ if ( npfec_kind_in_gpt == npfec.kind )
+ req->fault_in_gpt = 1;
+ if ( npfec_kind_with_gla == npfec.kind )
+ req->fault_with_gla = 1;
+ req->vcpu_id = v->vcpu_id;
+
+ mem_access_send_req(v->domain, req);
+ xfree(req);
+ }
+
+ /* Pause the current VCPU */
+ if ( xma != XENMEM_access_n2rwx )
+ mem_event_vcpu_pause(v);
+
+ return 0;
+}
+
+/* Set access type for a region of pfns.
+ * If start_pfn == -1ul, sets the default access type */
+long p2m_set_mem_access(struct domain *d, unsigned long pfn, uint32_t nr,
+ uint32_t start, uint32_t mask, xenmem_access_t access)
+{
+ struct p2m_domain *p2m = p2m_get_hostp2m(d);
+ p2m_access_t a;
+ long rc = 0;
+ paddr_t paddr;
+
+ static const p2m_access_t memaccess[] = {
+#define ACCESS(ac) [XENMEM_access_##ac] = p2m_access_##ac
+ ACCESS(n),
+ ACCESS(r),
+ ACCESS(w),
+ ACCESS(rw),
+ ACCESS(x),
+ ACCESS(rx),
+ ACCESS(wx),
+ ACCESS(rwx),
+#undef ACCESS
+ };
+
+ switch ( access )
+ {
+ case 0 ... ARRAY_SIZE(memaccess) - 1:
+ a = memaccess[access];
+ break;
+ case XENMEM_access_default:
+ a = p2m->default_access;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ /* If request to set default access */
+ if ( pfn == ~0ul )
+ {
+ p2m->default_access = a;
+ return 0;
+ }
+
+ for ( pfn += start; nr > start; ++pfn )
+ {
+ paddr = pfn_to_paddr(pfn);
+ rc = apply_p2m_changes(d, MEMACCESS, paddr, paddr+1, 0, MATTR_MEM, 0, a);
+ if ( rc < 0 )
+ break;
+
+ /* Check for continuation if it's not the last iteration. */
+ if ( nr > ++start && !(start & mask) && hypercall_preempt_check() )
+ {
+ rc = start;
+ break;
+ }
+ }
+
+ /* Flush the TLB of the domain to ensure consistency */
+ flush_tlb_domain(d);
+
+ return rc;
+}
+
+int p2m_get_mem_access(struct domain *d, unsigned long gpfn,
+ xenmem_access_t *access)
+{
+ struct p2m_domain *p2m = p2m_get_hostp2m(d);
+ void *i;
+ unsigned int index;
+
+ static const xenmem_access_t memaccess[] = {
+#define ACCESS(ac) [XENMEM_access_##ac] = XENMEM_access_##ac
+ ACCESS(n),
+ ACCESS(r),
+ ACCESS(w),
+ ACCESS(rw),
+ ACCESS(x),
+ ACCESS(rx),
+ ACCESS(wx),
+ ACCESS(rwx),
+#undef ACCESS
+ };
+
+ /* If request to get default access */
+ if ( gpfn == ~0ull )
+ {
+ *access = memaccess[p2m->default_access];
+ return 0;
+ }
+
+ spin_lock(&p2m->lock);
+
+ i = radix_tree_lookup(&p2m->mem_access_settings, gpfn);
+
+ spin_unlock(&p2m->lock);
+
+ if ( !i )
+ return -ESRCH;
+
+ index = radix_tree_ptr_to_int(i);
+
+ if ( index >= ARRAY_SIZE(memaccess) )
+ return -ERANGE;
+
+ *access = memaccess[index];
+ return 0;
+}
+
/*
* Local variables:
* mode: C
diff --git a/xen/arch/arm/traps.c b/xen/arch/arm/traps.c
index 019991f..9d93ed8 100644
--- a/xen/arch/arm/traps.c
+++ b/xen/arch/arm/traps.c
@@ -1852,11 +1852,36 @@ static void do_trap_data_abort_guest(struct cpu_user_regs *regs,
info.gva = READ_SYSREG64(FAR_EL2);
#endif
- if (dabt.s1ptw)
+ rc = gva_to_ipa(info.gva, &info.gpa);
+ if ( -EFAULT == rc )
goto bad_data_abort;
- rc = gva_to_ipa(info.gva, &info.gpa);
- if ( rc == -EFAULT )
+ switch ( dabt.dfsc )
+ {
+ case DABT_DFSC_PERMISSION_1:
+ case DABT_DFSC_PERMISSION_2:
+ case DABT_DFSC_PERMISSION_3:
+ {
+ struct npfec npfec = {
+ .read_access = 1,
+ .write_access = dabt.write,
+ .gla_valid = 1,
+ .kind = dabt.s1ptw ? npfec_kind_in_gpt : npfec_kind_with_gla
+ };
+
+ rc = p2m_mem_access_check(info.gpa, info.gva, npfec);
+
+ /* Trap was triggered by mem_access, work here is done */
+ if ( !rc )
+ return;
+ }
+ break;
+
+ default:
+ break;
+ }
+
+ if ( dabt.s1ptw )
goto bad_data_abort;
/* XXX: Decode the instruction if ISS is not valid */
diff --git a/xen/include/asm-arm/p2m.h b/xen/include/asm-arm/p2m.h
index b4ca86d..85ce984 100644
--- a/xen/include/asm-arm/p2m.h
+++ b/xen/include/asm-arm/p2m.h
@@ -246,6 +246,23 @@ static inline bool_t p2m_mem_event_sanity_check(struct domain *d)
return 1;
}
+/* Send mem event based on the access (gla is -1ull if not available). Boolean
+ * return value indicates if trap needs to be injected into guest. */
+bool_t p2m_mem_access_check(paddr_t gpa, vaddr_t gla, struct npfec npfec);
+
+/* Resumes the running of the VCPU, restarting the last instruction */
+void p2m_mem_access_resume(struct domain *d);
+
+/* Set access type for a region of pfns.
+ * If start_pfn == -1ul, sets the default access type */
+long p2m_set_mem_access(struct domain *d, unsigned long start_pfn, uint32_t nr,
+ uint32_t start, uint32_t mask, xenmem_access_t access);
+
+/* Get access type for a pfn
+ * If pfn == -1ul, gets the default access type */
+int p2m_get_mem_access(struct domain *d, unsigned long pfn,
+ xenmem_access_t *access);
+
#endif /* _XEN_P2M_H */
/*
diff --git a/xen/include/asm-arm/processor.h b/xen/include/asm-arm/processor.h
index 0cc5b6d..b844f1d 100644
--- a/xen/include/asm-arm/processor.h
+++ b/xen/include/asm-arm/processor.h
@@ -262,6 +262,36 @@ enum dabt_size {
DABT_DOUBLE_WORD = 3,
};
+/* Data abort data fetch status codes */
+enum dabt_dfsc {
+ DABT_DFSC_ADDR_SIZE_0 = 0b000000,
+ DABT_DFSC_ADDR_SIZE_1 = 0b000001,
+ DABT_DFSC_ADDR_SIZE_2 = 0b000010,
+ DABT_DFSC_ADDR_SIZE_3 = 0b000011,
+ DABT_DFSC_TRANSLATION_0 = 0b000100,
+ DABT_DFSC_TRANSLATION_1 = 0b000101,
+ DABT_DFSC_TRANSLATION_2 = 0b000110,
+ DABT_DFSC_TRANSLATION_3 = 0b000111,
+ DABT_DFSC_ACCESS_1 = 0b001001,
+ DABT_DFSC_ACCESS_2 = 0b001010,
+ DABT_DFSC_ACCESS_3 = 0b001011,
+ DABT_DFSC_PERMISSION_1 = 0b001101,
+ DABT_DFSC_PERMISSION_2 = 0b001110,
+ DABT_DFSC_PERMISSION_3 = 0b001111,
+ DABT_DFSC_SYNC_EXT = 0b010000,
+ DABT_DFSC_SYNC_PARITY = 0b011000,
+ DABT_DFSC_SYNC_EXT_TTW_0 = 0b010100,
+ DABT_DFSC_SYNC_EXT_TTW_1 = 0b010101,
+ DABT_DFSC_SYNC_EXT_TTW_2 = 0b010110,
+ DABT_DFSC_SYNC_EXT_TTW_3 = 0b010111,
+ DABT_DFSC_SYNC_PARITY_TTW_0 = 0b011100,
+ DABT_DFSC_SYNC_PARITY_TTW_1 = 0b011101,
+ DABT_DFSC_SYNC_PARITY_TTW_2 = 0b011110,
+ DABT_DFSC_SYNC_PARITY_TTW_3 = 0b011111,
+ DABT_DFSC_ALIGNMENT = 0b100001,
+ DABT_DFSC_TLB_CONFLICT = 0b110000,
+};
+
union hsr {
uint32_t bits;
struct {
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 13/17] xen/arm: Instruction prefetch abort (X) mem_event handling
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (11 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-11 21:23 ` Julien Grall
2014-09-10 13:28 ` [PATCH v5 14/17] xen/arm: Enable the compilation of mem_access and mem_event on ARM Tamas K Lengyel
` (4 subsequent siblings)
17 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Add missing structure definition for iabt and update the trap handling
mechanism to only inject the exception if the mem_access checker
decides to do so.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v4: - Don't mark instruction fetch violation as read violation.
- Use new struct npfec to pass violation info.
v2: - Add definition for instruction abort instruction fetch status codes
(enum iabt_ifsc) and only call p2m_mem_access_check for traps triggered
for permission violations.
---
xen/arch/arm/traps.c | 42 ++++++++++++++++++++++++++++++++++++++++-
xen/include/asm-arm/processor.h | 40 ++++++++++++++++++++++++++++++++++++++-
2 files changed, 80 insertions(+), 2 deletions(-)
diff --git a/xen/arch/arm/traps.c b/xen/arch/arm/traps.c
index 9d93ed8..2ba45bb 100644
--- a/xen/arch/arm/traps.c
+++ b/xen/arch/arm/traps.c
@@ -1828,7 +1828,47 @@ done:
static void do_trap_instr_abort_guest(struct cpu_user_regs *regs,
union hsr hsr)
{
- register_t addr = READ_SYSREG(FAR_EL2);
+ struct hsr_iabt iabt = hsr.iabt;
+ int rc;
+ register_t addr;
+ vaddr_t gva;
+ paddr_t gpa;
+
+#ifdef CONFIG_ARM_32
+ gva = READ_CP32(HIFAR);
+#else
+ gva = READ_SYSREG64(FAR_EL2);
+#endif
+
+ rc = gva_to_ipa(gva, &gpa);
+ if ( -EFAULT == rc )
+ return;
+
+ switch ( iabt.ifsc )
+ {
+ case IABT_IFSC_PERMISSION_1:
+ case IABT_IFSC_PERMISSION_2:
+ case IABT_IFSC_PERMISSION_3:
+ {
+ struct npfec npfec = {
+ .insn_fetch = 1,
+ .gla_valid = 1,
+ .kind = iabt.s1ptw ? npfec_kind_in_gpt : npfec_kind_with_gla
+ };
+
+ rc = p2m_mem_access_check(gpa, gva, npfec);
+
+ /* Trap was triggered by mem_access, work here is done */
+ if ( !rc )
+ return;
+ }
+ break;
+
+ default:
+ break;
+ }
+
+ addr = READ_SYSREG(FAR_EL2);
inject_iabt_exception(regs, addr, hsr.len);
}
diff --git a/xen/include/asm-arm/processor.h b/xen/include/asm-arm/processor.h
index b844f1d..044de12 100644
--- a/xen/include/asm-arm/processor.h
+++ b/xen/include/asm-arm/processor.h
@@ -292,6 +292,36 @@ enum dabt_dfsc {
DABT_DFSC_TLB_CONFLICT = 0b110000,
};
+/* Instruction abort instruction fault status codes */
+enum iabt_ifsc {
+ IABT_IFSC_ADDR_SIZE_0 = 0b000000,
+ IABT_IFSC_ADDR_SIZE_1 = 0b000001,
+ IABT_IFSC_ADDR_SIZE_2 = 0b000010,
+ IABT_IFSC_ADDR_SIZE_3 = 0b000011,
+ IABT_IFSC_TRANSLATION_0 = 0b000100,
+ IABT_IFSC_TRANSLATION_1 = 0b000101,
+ IABT_IFSC_TRANSLATION_2 = 0b000110,
+ IABT_IFSC_TRANSLATION_3 = 0b000111,
+ IABT_IFSC_ACCESS_1 = 0b001001,
+ IABT_IFSC_ACCESS_2 = 0b001010,
+ IABT_IFSC_ACCESS_3 = 0b001011,
+ IABT_IFSC_PERMISSION_1 = 0b001101,
+ IABT_IFSC_PERMISSION_2 = 0b001110,
+ IABT_IFSC_PERMISSION_3 = 0b001111,
+ IABT_IFSC_SYNC_EXT = 0b010000,
+ IABT_IFSC_SYNC_PARITY = 0b011000,
+ IABT_IFSC_SYNC_EXT_TTW_0 = 0b010100,
+ IABT_IFSC_SYNC_EXT_TTW_1 = 0b010101,
+ IABT_IFSC_SYNC_EXT_TTW_2 = 0b010110,
+ IABT_IFSC_SYNC_EXT_TTW_3 = 0b010111,
+ IABT_IFSC_SYNC_PARITY_TTW_0 = 0b011100,
+ IABT_IFSC_SYNC_PARITY_TTW_1 = 0b011101,
+ IABT_IFSC_SYNC_PARITY_TTW_2 = 0b011110,
+ IABT_IFSC_SYNC_PARITY_TTW_3 = 0b011111,
+ IABT_IFSC_ALIGNMENT = 0b100001,
+ IABT_IFSC_TLB_CONFLICT = 0b110000,
+};
+
union hsr {
uint32_t bits;
struct {
@@ -371,10 +401,18 @@ union hsr {
} sysreg; /* HSR_EC_SYSREG */
#endif
+ struct hsr_iabt {
+ unsigned long ifsc:6; /* Instruction fault status code */
+ unsigned long res0:1;
+ unsigned long s1ptw:1; /* Fault during a stage 1 translation table walk */
+ unsigned long res1:1;
+ unsigned long ea:1; /* External abort type */
+ } iabt; /* HSR_EC_INSTR_ABORT_* */
+
struct hsr_dabt {
unsigned long dfsc:6; /* Data Fault Status Code */
unsigned long write:1; /* Write / not Read */
- unsigned long s1ptw:1; /* */
+ unsigned long s1ptw:1; /* Fault during a stage 1 translation table walk */
unsigned long cache:1; /* Cache Maintenance */
unsigned long eat:1; /* External Abort Type */
#ifdef CONFIG_ARM_32
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 14/17] xen/arm: Enable the compilation of mem_access and mem_event on ARM.
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (12 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 13/17] xen/arm: Instruction prefetch abort (X) mem_event handling Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 15/17] xen: Extend getdomaininfo to return the domain's max_gpfn Tamas K Lengyel
` (3 subsequent siblings)
17 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
This patch sets up the infrastructure to support mem_access and mem_event
on ARM and turns on compilation. We define the required XSM functions.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
Acked-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
---
v3: Wrap mem_event related functions in XSM into #ifdef HAS_MEM_ACCESS
blocks.
Update XSM hooks in flask to properly wire it up on ARM.
v2: Add CONFIG_MEM_PAGING and CONFIG_MEM_SHARING definitions and
use them instead of CONFIG_X86.
Split domctl copy-back and p2m type definitions into separate
patches and move this patch to the end of the series.
---
xen/arch/arm/Rules.mk | 1 +
xen/common/mem_event.c | 19 +++++++++++++++++++
xen/include/asm-x86/config.h | 3 +++
xen/include/xsm/dummy.h | 26 ++++++++++++++------------
xen/include/xsm/xsm.h | 29 +++++++++++++++++------------
xen/xsm/dummy.c | 7 +++++--
xen/xsm/flask/hooks.c | 33 ++++++++++++++++++++-------------
7 files changed, 79 insertions(+), 39 deletions(-)
diff --git a/xen/arch/arm/Rules.mk b/xen/arch/arm/Rules.mk
index 8658176..f6781b5 100644
--- a/xen/arch/arm/Rules.mk
+++ b/xen/arch/arm/Rules.mk
@@ -10,6 +10,7 @@ HAS_DEVICE_TREE := y
HAS_VIDEO := y
HAS_ARM_HDLCD := y
HAS_PASSTHROUGH := y
+HAS_MEM_ACCESS := y
CFLAGS += -I$(BASEDIR)/include
diff --git a/xen/common/mem_event.c b/xen/common/mem_event.c
index e22a9ab..83de3b6 100644
--- a/xen/common/mem_event.c
+++ b/xen/common/mem_event.c
@@ -27,8 +27,15 @@
#include <asm/p2m.h>
#include <xen/mem_event.h>
#include <xen/mem_access.h>
+
+#ifdef CONFIG_MEM_PAGING
#include <asm/mem_paging.h>
+#endif
+
+#ifdef CONFIG_MEM_SHARING
#include <asm/mem_sharing.h>
+#endif
+
#include <xsm/xsm.h>
/* for public/io/ring.h macros */
@@ -423,12 +430,14 @@ int __mem_event_claim_slot(struct domain *d, struct mem_event_domain *med,
return mem_event_grab_slot(med, (current->domain != d));
}
+#ifdef CONFIG_MEM_PAGING
/* Registered with Xen-bound event channel for incoming notifications. */
static void mem_paging_notification(struct vcpu *v, unsigned int port)
{
if ( likely(v->domain->mem_event->paging.ring_page != NULL) )
p2m_mem_paging_resume(v->domain);
}
+#endif
/* Registered with Xen-bound event channel for incoming notifications. */
static void mem_access_notification(struct vcpu *v, unsigned int port)
@@ -437,15 +446,20 @@ static void mem_access_notification(struct vcpu *v, unsigned int port)
mem_access_resume(v->domain);
}
+#ifdef CONFIG_MEM_SHARING
/* Registered with Xen-bound event channel for incoming notifications. */
static void mem_sharing_notification(struct vcpu *v, unsigned int port)
{
if ( likely(v->domain->mem_event->share.ring_page != NULL) )
mem_sharing_sharing_resume(v->domain);
}
+#endif
int do_mem_event_op(int op, uint32_t domain, void *arg)
{
+#if !defined(CONFIG_MEM_PAGING) && !defined(CONFIG_MEM_SHARING)
+ return -ENOSYS;
+#else
int ret;
struct domain *d;
@@ -472,6 +486,7 @@ int do_mem_event_op(int op, uint32_t domain, void *arg)
out:
rcu_unlock_domain(d);
return ret;
+#endif
}
/* Clean up on domain destruction */
@@ -532,6 +547,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
switch ( mec->mode )
{
+#ifdef CONFIG_MEM_PAGING
case XEN_DOMCTL_MEM_EVENT_OP_PAGING:
{
struct mem_event_domain *med = &d->mem_event->paging;
@@ -582,6 +598,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
}
}
break;
+#endif
case XEN_DOMCTL_MEM_EVENT_OP_ACCESS:
{
@@ -616,6 +633,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
}
break;
+#ifdef CONFIG_MEM_SHARING
case XEN_DOMCTL_MEM_EVENT_OP_SHARING:
{
struct mem_event_domain *med = &d->mem_event->share;
@@ -654,6 +672,7 @@ int mem_event_domctl(struct domain *d, xen_domctl_mem_event_op_t *mec,
}
}
break;
+#endif
default:
rc = -ENOSYS;
diff --git a/xen/include/asm-x86/config.h b/xen/include/asm-x86/config.h
index 8a864ce..f8ef043 100644
--- a/xen/include/asm-x86/config.h
+++ b/xen/include/asm-x86/config.h
@@ -57,6 +57,9 @@
#define CONFIG_LATE_HWDOM 1
#endif
+#define CONFIG_MEM_SHARING 1
+#define CONFIG_MEM_PAGING 1
+
#define HZ 100
#define OPT_CONSOLE_STR "vga"
diff --git a/xen/include/xsm/dummy.h b/xen/include/xsm/dummy.h
index df55e70..f20e89c 100644
--- a/xen/include/xsm/dummy.h
+++ b/xen/include/xsm/dummy.h
@@ -513,6 +513,20 @@ static XSM_INLINE int xsm_hvm_param_nested(XSM_DEFAULT_ARG struct domain *d)
return xsm_default_action(action, current->domain, d);
}
+#ifdef HAS_MEM_ACCESS
+static XSM_INLINE int xsm_mem_event_control(XSM_DEFAULT_ARG struct domain *d, int mode, int op)
+{
+ XSM_ASSERT_ACTION(XSM_PRIV);
+ return xsm_default_action(action, current->domain, d);
+}
+
+static XSM_INLINE int xsm_mem_event_op(XSM_DEFAULT_ARG struct domain *d, int op)
+{
+ XSM_ASSERT_ACTION(XSM_DM_PRIV);
+ return xsm_default_action(action, current->domain, d);
+}
+#endif
+
#ifdef CONFIG_X86
static XSM_INLINE int xsm_do_mca(XSM_DEFAULT_VOID)
{
@@ -556,18 +570,6 @@ static XSM_INLINE int xsm_hvm_ioreq_server(XSM_DEFAULT_ARG struct domain *d, int
return xsm_default_action(action, current->domain, d);
}
-static XSM_INLINE int xsm_mem_event_control(XSM_DEFAULT_ARG struct domain *d, int mode, int op)
-{
- XSM_ASSERT_ACTION(XSM_PRIV);
- return xsm_default_action(action, current->domain, d);
-}
-
-static XSM_INLINE int xsm_mem_event_op(XSM_DEFAULT_ARG struct domain *d, int op)
-{
- XSM_ASSERT_ACTION(XSM_DM_PRIV);
- return xsm_default_action(action, current->domain, d);
-}
-
static XSM_INLINE int xsm_mem_sharing_op(XSM_DEFAULT_ARG struct domain *d, struct domain *cd, int op)
{
XSM_ASSERT_ACTION(XSM_DM_PRIV);
diff --git a/xen/include/xsm/xsm.h b/xen/include/xsm/xsm.h
index 6c1c079..4ce089f 100644
--- a/xen/include/xsm/xsm.h
+++ b/xen/include/xsm/xsm.h
@@ -141,6 +141,11 @@ struct xsm_operations {
int (*hvm_param_nested) (struct domain *d);
int (*get_vnumainfo) (struct domain *d);
+#ifdef HAS_MEM_ACCESS
+ int (*mem_event_control) (struct domain *d, int mode, int op);
+ int (*mem_event_op) (struct domain *d, int op);
+#endif
+
#ifdef CONFIG_X86
int (*do_mca) (void);
int (*shadow_control) (struct domain *d, uint32_t op);
@@ -149,8 +154,6 @@ struct xsm_operations {
int (*hvm_set_pci_link_route) (struct domain *d);
int (*hvm_inject_msi) (struct domain *d);
int (*hvm_ioreq_server) (struct domain *d, int op);
- int (*mem_event_control) (struct domain *d, int mode, int op);
- int (*mem_event_op) (struct domain *d, int op);
int (*mem_sharing_op) (struct domain *d, struct domain *cd, int op);
int (*apic) (struct domain *d, int cmd);
int (*memtype) (uint32_t access);
@@ -540,6 +543,18 @@ static inline int xsm_get_vnumainfo (xsm_default_t def, struct domain *d)
return xsm_ops->get_vnumainfo(d);
}
+#ifdef HAS_MEM_ACCESS
+static inline int xsm_mem_event_control (xsm_default_t def, struct domain *d, int mode, int op)
+{
+ return xsm_ops->mem_event_control(d, mode, op);
+}
+
+static inline int xsm_mem_event_op (xsm_default_t def, struct domain *d, int op)
+{
+ return xsm_ops->mem_event_op(d, op);
+}
+#endif
+
#ifdef CONFIG_X86
static inline int xsm_do_mca(xsm_default_t def)
{
@@ -576,16 +591,6 @@ static inline int xsm_hvm_ioreq_server (xsm_default_t def, struct domain *d, int
return xsm_ops->hvm_ioreq_server(d, op);
}
-static inline int xsm_mem_event_control (xsm_default_t def, struct domain *d, int mode, int op)
-{
- return xsm_ops->mem_event_control(d, mode, op);
-}
-
-static inline int xsm_mem_event_op (xsm_default_t def, struct domain *d, int op)
-{
- return xsm_ops->mem_event_op(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);
diff --git a/xen/xsm/dummy.c b/xen/xsm/dummy.c
index 0826a8b..8eb3050 100644
--- a/xen/xsm/dummy.c
+++ b/xen/xsm/dummy.c
@@ -118,6 +118,11 @@ void xsm_fixup_ops (struct xsm_operations *ops)
set_to_dummy_if_null(ops, remove_from_physmap);
set_to_dummy_if_null(ops, map_gmfn_foreign);
+#ifdef HAS_MEM_ACCESS
+ set_to_dummy_if_null(ops, mem_event_control);
+ set_to_dummy_if_null(ops, mem_event_op);
+#endif
+
#ifdef CONFIG_X86
set_to_dummy_if_null(ops, do_mca);
set_to_dummy_if_null(ops, shadow_control);
@@ -126,8 +131,6 @@ void xsm_fixup_ops (struct xsm_operations *ops)
set_to_dummy_if_null(ops, hvm_set_pci_link_route);
set_to_dummy_if_null(ops, hvm_inject_msi);
set_to_dummy_if_null(ops, hvm_ioreq_server);
- set_to_dummy_if_null(ops, mem_event_control);
- set_to_dummy_if_null(ops, mem_event_op);
set_to_dummy_if_null(ops, mem_sharing_op);
set_to_dummy_if_null(ops, apic);
set_to_dummy_if_null(ops, platform_op);
diff --git a/xen/xsm/flask/hooks.c b/xen/xsm/flask/hooks.c
index 00efba1..6c1e21f 100644
--- a/xen/xsm/flask/hooks.c
+++ b/xen/xsm/flask/hooks.c
@@ -576,6 +576,9 @@ static int flask_domctl(struct domain *d, int cmd)
case XEN_DOMCTL_irq_permission:
case XEN_DOMCTL_iomem_permission:
case XEN_DOMCTL_set_target:
+#ifdef HAS_MEM_ACCESS
+ case XEN_DOMCTL_mem_event_op:
+#endif
#ifdef CONFIG_X86
/* These have individual XSM hooks (arch/x86/domctl.c) */
case XEN_DOMCTL_shadow_op:
@@ -584,7 +587,6 @@ static int flask_domctl(struct domain *d, int cmd)
case XEN_DOMCTL_unbind_pt_irq:
case XEN_DOMCTL_memory_mapping:
case XEN_DOMCTL_ioport_mapping:
- case XEN_DOMCTL_mem_event_op:
/* These have individual XSM hooks (drivers/passthrough/iommu.c) */
case XEN_DOMCTL_get_device_group:
case XEN_DOMCTL_test_assign_device:
@@ -1189,6 +1191,18 @@ static int flask_deassign_device(struct domain *d, uint32_t machine_bdf)
}
#endif /* HAS_PASSTHROUGH && HAS_PCI */
+#ifdef HAS_MEM_ACCESS
+static int flask_mem_event_control(struct domain *d, int mode, int op)
+{
+ return current_has_perm(d, SECCLASS_HVM, HVM__MEM_EVENT);
+}
+
+static int flask_mem_event_op(struct domain *d, int op)
+{
+ return current_has_perm(d, SECCLASS_HVM, HVM__MEM_EVENT);
+}
+#endif /* HAS_MEM_ACCESS */
+
#ifdef CONFIG_X86
static int flask_do_mca(void)
{
@@ -1299,16 +1313,6 @@ static int flask_hvm_ioreq_server(struct domain *d, int op)
return current_has_perm(d, SECCLASS_HVM, HVM__HVMCTL);
}
-static int flask_mem_event_control(struct domain *d, int mode, int op)
-{
- return current_has_perm(d, SECCLASS_HVM, HVM__MEM_EVENT);
-}
-
-static int flask_mem_event_op(struct domain *d, int op)
-{
- return current_has_perm(d, SECCLASS_HVM, HVM__MEM_EVENT);
-}
-
static int flask_mem_sharing_op(struct domain *d, struct domain *cd, int op)
{
int rc = current_has_perm(cd, SECCLASS_HVM, HVM__MEM_SHARING);
@@ -1577,6 +1581,11 @@ static struct xsm_operations flask_ops = {
.deassign_device = flask_deassign_device,
#endif
+#ifdef HAS_MEM_ACCESS
+ .mem_event_control = flask_mem_event_control,
+ .mem_event_op = flask_mem_event_op,
+#endif
+
#ifdef CONFIG_X86
.do_mca = flask_do_mca,
.shadow_control = flask_shadow_control,
@@ -1585,8 +1594,6 @@ static struct xsm_operations flask_ops = {
.hvm_set_pci_link_route = flask_hvm_set_pci_link_route,
.hvm_inject_msi = flask_hvm_inject_msi,
.hvm_ioreq_server = flask_hvm_ioreq_server,
- .mem_event_control = flask_mem_event_control,
- .mem_event_op = flask_mem_event_op,
.mem_sharing_op = flask_mem_sharing_op,
.apic = flask_apic,
.platform_op = flask_platform_op,
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 15/17] xen: Extend getdomaininfo to return the domain's max_gpfn
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (13 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 14/17] xen/arm: Enable the compilation of mem_access and mem_event on ARM Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:48 ` Jan Beulich
2014-09-10 13:28 ` [PATCH v5 16/17] tools/libxc: Allocate magic page for mem access on ARM Tamas K Lengyel
` (2 subsequent siblings)
17 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
On ARM determining a domains max_gpfn thus far required the user to be aware
of how the guest's memory is mapped with potential holes and estimate the max
gpfn accordingly. Extending getdomaininfo to just return this value simplifies
the process for the user.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
xen/common/domctl.c | 1 +
xen/include/public/domctl.h | 1 +
2 files changed, 2 insertions(+)
diff --git a/xen/common/domctl.c b/xen/common/domctl.c
index 638080f..6a57925 100644
--- a/xen/common/domctl.c
+++ b/xen/common/domctl.c
@@ -202,6 +202,7 @@ void getdomaininfo(struct domain *d, struct xen_domctl_getdomaininfo *info)
info->tot_pages = d->tot_pages;
info->max_pages = d->max_pages;
+ info->max_gpfn = domain_get_maximum_gpfn(d);
info->outstanding_pages = d->outstanding_pages;
info->shr_pages = atomic_read(&d->shr_pages);
info->paged_pages = atomic_read(&d->paged_pages);
diff --git a/xen/include/public/domctl.h b/xen/include/public/domctl.h
index 69a8b44..66d29bb 100644
--- a/xen/include/public/domctl.h
+++ b/xen/include/public/domctl.h
@@ -102,6 +102,7 @@ struct xen_domctl_getdomaininfo {
uint32_t flags; /* XEN_DOMINF_* */
uint64_aligned_t tot_pages;
uint64_aligned_t max_pages;
+ uint64_aligned_t max_gpfn;
uint64_aligned_t outstanding_pages;
uint64_aligned_t shr_pages;
uint64_aligned_t paged_pages;
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 16/17] tools/libxc: Allocate magic page for mem access on ARM
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (14 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 15/17] xen: Extend getdomaininfo to return the domain's max_gpfn Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 17/17] tools/tests: Enable xen-access " Tamas K Lengyel
2014-09-10 13:51 ` [PATCH v5 00/17] Mem_event and mem_access for ARM Jan Beulich
17 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
Reviewed-by: Julien Grall <julien.grall@linaro.org>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
tools/libxc/xc_dom_arm.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/tools/libxc/xc_dom_arm.c b/tools/libxc/xc_dom_arm.c
index 9b31b1f..13e881e 100644
--- a/tools/libxc/xc_dom_arm.c
+++ b/tools/libxc/xc_dom_arm.c
@@ -26,9 +26,10 @@
#include "xg_private.h"
#include "xc_dom.h"
-#define NR_MAGIC_PAGES 2
+#define NR_MAGIC_PAGES 3
#define CONSOLE_PFN_OFFSET 0
#define XENSTORE_PFN_OFFSET 1
+#define MEMACCESS_PFN_OFFSET 2
#define LPAE_SHIFT 9
@@ -87,10 +88,13 @@ static int alloc_magic_pages(struct xc_dom_image *dom)
xc_clear_domain_page(dom->xch, dom->guest_domid, dom->console_pfn);
xc_clear_domain_page(dom->xch, dom->guest_domid, dom->xenstore_pfn);
+ xc_clear_domain_page(dom->xch, dom->guest_domid, base + MEMACCESS_PFN_OFFSET);
xc_hvm_param_set(dom->xch, dom->guest_domid, HVM_PARAM_CONSOLE_PFN,
dom->console_pfn);
xc_hvm_param_set(dom->xch, dom->guest_domid, HVM_PARAM_STORE_PFN,
dom->xenstore_pfn);
+ xc_hvm_param_set(dom->xch, dom->guest_domid, HVM_PARAM_ACCESS_RING_PFN,
+ base + MEMACCESS_PFN_OFFSET);
/* allocated by toolstack */
xc_hvm_param_set(dom->xch, dom->guest_domid, HVM_PARAM_CONSOLE_EVTCHN,
dom->console_evtchn);
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* [PATCH v5 17/17] tools/tests: Enable xen-access on ARM
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (15 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 16/17] tools/libxc: Allocate magic page for mem access on ARM Tamas K Lengyel
@ 2014-09-10 13:28 ` Tamas K Lengyel
2014-09-11 21:29 ` Julien Grall
2014-09-10 13:51 ` [PATCH v5 00/17] Mem_event and mem_access for ARM Jan Beulich
17 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:28 UTC (permalink / raw)
To: xen-devel
Cc: ian.campbell, tim, julien.grall, ian.jackson, stefano.stabellini,
andres, jbeulich, dgdegra, Tamas K Lengyel
On ARM the guest memory doesn't start from 0 and has multiple banks
of memory, thus we adjust the setting/unsetting routines accordingly.
We also define the ARM specific test_and_set_bit function.
Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
---
v5: - Use the new information returned by getdomaininfo, max_gpfn, to
set access permissions. On ARM this will include the potential
memory hole as well which the hypervisor just loops over.
v4: - Take into account multiple guest ram banks on ARM.
- Move HAS_MEM_ACCESS definition into config/*.mk and only compile
xen-access when it is defined.
- Pass CONFIG_X86/CONFIG_ARM flags during compilation in xen-access
Makefile.
---
config/arm32.mk | 1 +
config/arm64.mk | 1 +
config/x86_32.mk | 2 ++
config/x86_64.mk | 2 ++
tools/tests/xen-access/Makefile | 9 ++++--
tools/tests/xen-access/xen-access.c | 60 ++++++++++++++++++++++++++++---------
xen/arch/arm/Rules.mk | 1 -
xen/arch/x86/Rules.mk | 1 -
8 files changed, 59 insertions(+), 18 deletions(-)
diff --git a/config/arm32.mk b/config/arm32.mk
index aa79d22..4a7c259 100644
--- a/config/arm32.mk
+++ b/config/arm32.mk
@@ -13,6 +13,7 @@ HAS_PL011 := y
HAS_EXYNOS4210 := y
HAS_OMAP := y
HAS_NS16550 := y
+HAS_MEM_ACCESS := y
# Use only if calling $(LD) directly.
LDFLAGS_DIRECT += -EL
diff --git a/config/arm64.mk b/config/arm64.mk
index 15b57a4..ea6558d 100644
--- a/config/arm64.mk
+++ b/config/arm64.mk
@@ -8,6 +8,7 @@ CFLAGS += #-marm -march= -mcpu= etc
HAS_PL011 := y
HAS_NS16550 := y
+HAS_MEM_ACCESS := y
# Use only if calling $(LD) directly.
LDFLAGS_DIRECT += -EL
diff --git a/config/x86_32.mk b/config/x86_32.mk
index 7f76b25..5906e60 100644
--- a/config/x86_32.mk
+++ b/config/x86_32.mk
@@ -6,6 +6,8 @@ CONFIG_HVM := y
CONFIG_MIGRATE := y
CONFIG_XCUTILS := y
+HAS_MEM_ACCESS := y
+
CFLAGS += -m32 -march=i686
# Use only if calling $(LD) directly.
diff --git a/config/x86_64.mk b/config/x86_64.mk
index 11104bd..b000c4e 100644
--- a/config/x86_64.mk
+++ b/config/x86_64.mk
@@ -7,6 +7,8 @@ CONFIG_HVM := y
CONFIG_MIGRATE := y
CONFIG_XCUTILS := y
+HAS_MEM_ACCESS := y
+
CONFIG_XEN_INSTALL_SUFFIX := .gz
CFLAGS += -m64
diff --git a/tools/tests/xen-access/Makefile b/tools/tests/xen-access/Makefile
index 65eef99..5056972 100644
--- a/tools/tests/xen-access/Makefile
+++ b/tools/tests/xen-access/Makefile
@@ -7,8 +7,13 @@ CFLAGS += $(CFLAGS_libxenctrl)
CFLAGS += $(CFLAGS_libxenguest)
CFLAGS += $(CFLAGS_xeninclude)
-TARGETS-y :=
-TARGETS-$(CONFIG_X86) += xen-access
+CFLAGS-y :=
+CFLAGS-$(CONFIG_X86) := -DCONFIG_X86
+CFLAGS-$(CONFIG_ARM) := -DCONFIG_ARM
+CFLAGS += $(CFLAGS-y)
+
+TARGETS-y :=
+TARGETS-$(HAS_MEM_ACCESS) := xen-access
TARGETS := $(TARGETS-y)
.PHONY: all
diff --git a/tools/tests/xen-access/xen-access.c b/tools/tests/xen-access/xen-access.c
index 6cb382d..9158636 100644
--- a/tools/tests/xen-access/xen-access.c
+++ b/tools/tests/xen-access/xen-access.c
@@ -41,22 +41,16 @@
#include <xenctrl.h>
#include <xen/mem_event.h>
-#define DPRINTF(a, b...) fprintf(stderr, a, ## b)
-#define ERROR(a, b...) fprintf(stderr, a "\n", ## b)
-#define PERROR(a, b...) fprintf(stderr, a ": %s\n", ## b, strerror(errno))
-
-/* Spinlock and mem event definitions */
-
-#define SPIN_LOCK_UNLOCKED 0
+#ifdef CONFIG_X86
+#define START_PFN 0ULL
#define ADDR (*(volatile long *) addr)
+
/**
* test_and_set_bit - Set a bit and return its old value
* @nr: Bit to set
* @addr: Address to count from
*
- * This operation is atomic and cannot be reordered.
- * It also implies a memory barrier.
*/
static inline int test_and_set_bit(int nr, volatile void *addr)
{
@@ -69,6 +63,43 @@ static inline int test_and_set_bit(int nr, volatile void *addr)
return oldbit;
}
+#elif CONFIG_ARM
+
+#include <xen/arch-arm.h>
+
+#define PAGE_SHIFT 12
+#define START_PFN (GUEST_RAM0_BASE >> PAGE_SHIFT)
+#define BITS_PER_WORD 32
+#define BIT_MASK(nr) (1UL << ((nr) % BITS_PER_WORD))
+#define BIT_WORD(nr) ((nr) / BITS_PER_WORD)
+
+/**
+ * test_and_set_bit - Set a bit and return its old value
+ * @nr: Bit to set
+ * @addr: Address to count from
+ *
+ */
+static inline int test_and_set_bit(int nr, volatile void *addr)
+{
+ unsigned int mask = BIT_MASK(nr);
+ volatile unsigned int *p =
+ ((volatile unsigned int *)addr) + BIT_WORD(nr);
+ unsigned int old = *p;
+
+ *p = old | mask;
+ return (old & mask) != 0;
+}
+
+#endif
+
+#define DPRINTF(a, b...) fprintf(stderr, a, ## b)
+#define ERROR(a, b...) fprintf(stderr, a "\n", ## b)
+#define PERROR(a, b...) fprintf(stderr, a ": %s\n", ## b, strerror(errno))
+
+/* Spinlock and mem event definitions */
+
+#define SPIN_LOCK_UNLOCKED 0
+
typedef int spinlock_t;
static inline void spin_lock(spinlock_t *lock)
@@ -309,7 +340,7 @@ xenaccess_t *xenaccess_init(xc_interface **xch_r, domid_t domain_id)
goto err;
}
- DPRINTF("max_pages = %"PRIx64"\n", xenaccess->domain_info->max_pages);
+ DPRINTF("max_gpfn = %"PRIx64"\n", xenaccess->domain_info->max_gpfn);
return xenaccess;
@@ -492,8 +523,9 @@ int main(int argc, char *argv[])
goto exit;
}
- rc = xc_set_mem_access(xch, domain_id, default_access, 0,
- xenaccess->domain_info->max_pages);
+ rc = xc_set_mem_access(xch, domain_id, default_access, START_PFN,
+ (xenaccess->domain_info->max_gpfn - START_PFN) );
+
if ( rc < 0 )
{
ERROR("Error %d setting all memory to access type %d\n", rc,
@@ -520,8 +552,8 @@ int main(int argc, char *argv[])
/* Unregister for every event */
rc = xc_set_mem_access(xch, domain_id, XENMEM_access_rwx, ~0ull, 0);
- rc = xc_set_mem_access(xch, domain_id, XENMEM_access_rwx, 0,
- xenaccess->domain_info->max_pages);
+ rc = xc_set_mem_access(xch, domain_id, XENMEM_access_rwx, START_PFN,
+ (xenaccess->domain_info->max_gpfn - START_PFN) );
rc = xc_hvm_param_set(xch, domain_id, HVM_PARAM_MEMORY_EVENT_INT3, HVMPME_mode_disabled);
shutting_down = 1;
diff --git a/xen/arch/arm/Rules.mk b/xen/arch/arm/Rules.mk
index f6781b5..8658176 100644
--- a/xen/arch/arm/Rules.mk
+++ b/xen/arch/arm/Rules.mk
@@ -10,7 +10,6 @@ HAS_DEVICE_TREE := y
HAS_VIDEO := y
HAS_ARM_HDLCD := y
HAS_PASSTHROUGH := y
-HAS_MEM_ACCESS := y
CFLAGS += -I$(BASEDIR)/include
diff --git a/xen/arch/x86/Rules.mk b/xen/arch/x86/Rules.mk
index bd4e342..576985e 100644
--- a/xen/arch/x86/Rules.mk
+++ b/xen/arch/x86/Rules.mk
@@ -12,7 +12,6 @@ HAS_NS16550 := y
HAS_EHCI := y
HAS_KEXEC := y
HAS_GDBSX := y
-HAS_MEM_ACCESS := y
xenoprof := y
#
--
2.1.0
^ permalink raw reply related [flat|nested] 48+ messages in thread
* Re: [PATCH v5 04/17] xen: Relocate mem_event_op domctl and access_op memop into common.
2014-09-10 13:28 ` [PATCH v5 04/17] xen: Relocate mem_event_op domctl and access_op memop " Tamas K Lengyel
@ 2014-09-10 13:44 ` Jan Beulich
0 siblings, 0 replies; 48+ messages in thread
From: Jan Beulich @ 2014-09-10 13:44 UTC (permalink / raw)
To: Tamas K Lengyel
Cc: ian.campbell, tim, julien.grall, ian.jackson, xen-devel,
stefano.stabellini, andres, dgdegra
>>> On 10.09.14 at 15:28, <tklengyel@sec.in.tum.de> wrote:
> Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
Acked-by: Jan Beulich <jbeulich@suse.com>
> ---
> v5: Move memop compat into common as well.
> Position domctl in switch relative to the domctl #.
> v4: Don't remove memop handling from x86_64/compat and style fixes.
> ---
> xen/arch/x86/domctl.c | 8 --------
> xen/arch/x86/x86_64/compat/mm.c | 4 ----
> xen/arch/x86/x86_64/mm.c | 4 ----
> xen/common/compat/memory.c | 5 +++++
> xen/common/domctl.c | 7 +++++++
> xen/common/memory.c | 5 +++++
> 6 files changed, 17 insertions(+), 16 deletions(-)
>
> diff --git a/xen/arch/x86/domctl.c b/xen/arch/x86/domctl.c
> index 26a3ea1..166cfb3 100644
> --- a/xen/arch/x86/domctl.c
> +++ b/xen/arch/x86/domctl.c
> @@ -1131,14 +1131,6 @@ long arch_do_domctl(
> }
> break;
>
> - case XEN_DOMCTL_mem_event_op:
> - {
> - ret = mem_event_domctl(d, &domctl->u.mem_event_op,
> - guest_handle_cast(u_domctl, void));
> - copyback = 1;
> - }
> - break;
> -
> case XEN_DOMCTL_mem_sharing_op:
> {
> ret = mem_sharing_domctl(d, &domctl->u.mem_sharing_op);
> diff --git a/xen/arch/x86/x86_64/compat/mm.c b/xen/arch/x86/x86_64/compat/mm.c
> index c079702..54f25b7 100644
> --- a/xen/arch/x86/x86_64/compat/mm.c
> +++ b/xen/arch/x86/x86_64/compat/mm.c
> @@ -198,10 +198,6 @@ int compat_arch_memory_op(unsigned long cmd,
> XEN_GUEST_HANDLE_PARAM(void) arg)
> break;
> }
>
> - case XENMEM_access_op:
> - rc = mem_access_memop(cmd, guest_handle_cast(arg,
> xen_mem_access_op_t));
> - break;
> -
> case XENMEM_sharing_op:
> {
> xen_mem_sharing_op_t mso;
> diff --git a/xen/arch/x86/x86_64/mm.c b/xen/arch/x86/x86_64/mm.c
> index 0da6ddc..0aaa460 100644
> --- a/xen/arch/x86/x86_64/mm.c
> +++ b/xen/arch/x86/x86_64/mm.c
> @@ -1048,10 +1048,6 @@ long subarch_memory_op(unsigned long cmd,
> XEN_GUEST_HANDLE_PARAM(void) arg)
> break;
> }
>
> - case XENMEM_access_op:
> - rc = mem_access_memop(cmd, guest_handle_cast(arg,
> xen_mem_access_op_t));
> - break;
> -
> case XENMEM_sharing_op:
> {
> xen_mem_sharing_op_t mso;
> diff --git a/xen/common/compat/memory.c b/xen/common/compat/memory.c
> index 25dc016..43d02bc 100644
> --- a/xen/common/compat/memory.c
> +++ b/xen/common/compat/memory.c
> @@ -4,6 +4,7 @@
> #include <xen/guest_access.h>
> #include <xen/sched.h>
> #include <xen/event.h>
> +#include <xen/mem_access.h>
> #include <asm/current.h>
> #include <compat/memory.h>
>
> @@ -381,6 +382,10 @@ int compat_memory_op(unsigned int cmd,
> XEN_GUEST_HANDLE_PARAM(void) compat)
> break;
> }
>
> + case XENMEM_access_op:
> + rc = mem_access_memop(cmd, guest_handle_cast(compat,
> xen_mem_access_op_t));
> + break;
> +
> case XENMEM_add_to_physmap_batch:
> start_extent = end_extent;
> break;
> diff --git a/xen/common/domctl.c b/xen/common/domctl.c
> index 222b140..638080f 100644
> --- a/xen/common/domctl.c
> +++ b/xen/common/domctl.c
> @@ -24,6 +24,7 @@
> #include <xen/bitmap.h>
> #include <xen/paging.h>
> #include <xen/hypercall.h>
> +#include <xen/mem_event.h>
> #include <asm/current.h>
> #include <asm/irq.h>
> #include <asm/page.h>
> @@ -1126,6 +1127,12 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t)
> u_domctl)
> }
> break;
>
> + case XEN_DOMCTL_mem_event_op:
> + ret = mem_event_domctl(d, &op->u.mem_event_op,
> + guest_handle_cast(u_domctl, void));
> + copyback = 1;
> + break;
> +
> case XEN_DOMCTL_disable_migrate:
> {
> d->disable_migrate = op->u.disable_migrate.disable;
> diff --git a/xen/common/memory.c b/xen/common/memory.c
> index 44a8106..98b1ee6 100644
> --- a/xen/common/memory.c
> +++ b/xen/common/memory.c
> @@ -25,6 +25,7 @@
> #include <asm/hardirq.h>
> #include <asm/p2m.h>
> #include <xen/numa.h>
> +#include <xen/mem_access.h>
> #include <public/memory.h>
> #include <xsm/xsm.h>
> #include <xen/trace.h>
> @@ -943,6 +944,10 @@ long do_memory_op(unsigned long cmd,
> XEN_GUEST_HANDLE_PARAM(void) arg)
> break;
> }
>
> + case XENMEM_access_op:
> + rc = mem_access_memop(cmd, guest_handle_cast(arg,
> xen_mem_access_op_t));
> + break;
> +
> case XENMEM_claim_pages:
> if ( copy_from_guest(&reservation, arg, 1) )
> return -EFAULT;
> --
> 2.1.0
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 15/17] xen: Extend getdomaininfo to return the domain's max_gpfn
2014-09-10 13:28 ` [PATCH v5 15/17] xen: Extend getdomaininfo to return the domain's max_gpfn Tamas K Lengyel
@ 2014-09-10 13:48 ` Jan Beulich
2014-09-10 13:55 ` Tamas K Lengyel
0 siblings, 1 reply; 48+ messages in thread
From: Jan Beulich @ 2014-09-10 13:48 UTC (permalink / raw)
To: Tamas K Lengyel
Cc: ian.campbell, tim, julien.grall, ian.jackson, xen-devel,
stefano.stabellini, andres, dgdegra
>>> On 10.09.14 at 15:28, <tklengyel@sec.in.tum.de> wrote:
> On ARM determining a domains max_gpfn thus far required the user to be aware
> of how the guest's memory is mapped with potential holes and estimate the
> max
> gpfn accordingly. Extending getdomaininfo to just return this value
> simplifies
> the process for the user.
But why is XENMEM_maximum_gpfn not sufficient?
Jan
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 00/17] Mem_event and mem_access for ARM
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
` (16 preceding siblings ...)
2014-09-10 13:28 ` [PATCH v5 17/17] tools/tests: Enable xen-access " Tamas K Lengyel
@ 2014-09-10 13:51 ` Jan Beulich
2014-09-10 14:01 ` Tamas K Lengyel
17 siblings, 1 reply; 48+ messages in thread
From: Jan Beulich @ 2014-09-10 13:51 UTC (permalink / raw)
To: Tamas K Lengyel, Tim Deegan
Cc: ian.campbell, julien.grall, ian.jackson, xen-devel,
stefano.stabellini, andres, dgdegra
>>> On 10.09.14 at 15:28, <tklengyel@sec.in.tum.de> wrote:
> The ARM virtualization extension provides 2-stage paging, a similar mechanisms
> to Intel's EPT, which can be used to trace the memory accesses performed by
> the guest systems. This series moves the mem_access and mem_event codebase
> into Xen common, performs some code cleanup and architecture specific
> division
> of components, then sets up the necessary infrastructure in the ARM code
> to deliver the event on R/W/X traps. Finally, we turn on the compilation of
> mem_access and mem_event on ARM and perform the necessary changes to the
> tools side.
>
> This version of the series has been fully tested and is functional on an
> Arndale board.
>
> This PATCH series is also available at:
> https://github.com/tklengyel/xen/tree/arm_memaccess5
>
> Tamas K Lengyel (17):
> xen: Relocate mem_access and mem_event into common.
> xen: Relocate p2m_mem_access_resume to mem_access common
> xen: Relocate struct npfec definition into common
> xen: Relocate mem_event_op domctl and access_op memop into common.
> xen/mem_event: Clean out superfluous white-spaces
> xen/mem_event: Relax error condition on debug builds
> xen/mem_event: Abstract architecture specific sanity checks
> xen/mem_access: Abstract architecture specific sanity check
So I think up to here the series could go in afaic. Question is - does
this make sense without the ARM parts?
Thanks, Jan
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 15/17] xen: Extend getdomaininfo to return the domain's max_gpfn
2014-09-10 13:48 ` Jan Beulich
@ 2014-09-10 13:55 ` Tamas K Lengyel
0 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 13:55 UTC (permalink / raw)
To: Jan Beulich
Cc: Ian Campbell, Tim Deegan, Julien Grall, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Daniel De Graaf,
Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 279 bytes --]
>
>
> But why is XENMEM_maximum_gpfn not sufficient?
>
> Jan
>
Oh, you are right, that is already accessible via xc_domain_maximum_gpfn.
That makes life easier, I'll just switch xen-access to use that instead of
getdomaininfo and this patch can be dropped then.
Thanks,
Tamas
[-- Attachment #1.2: Type: text/html, Size: 633 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 00/17] Mem_event and mem_access for ARM
2014-09-10 13:51 ` [PATCH v5 00/17] Mem_event and mem_access for ARM Jan Beulich
@ 2014-09-10 14:01 ` Tamas K Lengyel
2014-09-15 22:26 ` Ian Campbell
0 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-10 14:01 UTC (permalink / raw)
To: Jan Beulich
Cc: Ian Campbell, Tim Deegan, Julien Grall, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Daniel De Graaf,
Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 1853 bytes --]
On Wed, Sep 10, 2014 at 3:51 PM, Jan Beulich <JBeulich@suse.com> wrote:
> >>> On 10.09.14 at 15:28, <tklengyel@sec.in.tum.de> wrote:
> > The ARM virtualization extension provides 2-stage paging, a similar
> mechanisms
> > to Intel's EPT, which can be used to trace the memory accesses performed
> by
> > the guest systems. This series moves the mem_access and mem_event
> codebase
> > into Xen common, performs some code cleanup and architecture specific
> > division
> > of components, then sets up the necessary infrastructure in the ARM code
> > to deliver the event on R/W/X traps. Finally, we turn on the compilation
> of
> > mem_access and mem_event on ARM and perform the necessary changes to the
> > tools side.
> >
> > This version of the series has been fully tested and is functional on an
> > Arndale board.
> >
> > This PATCH series is also available at:
> > https://github.com/tklengyel/xen/tree/arm_memaccess5
> >
> > Tamas K Lengyel (17):
> > xen: Relocate mem_access and mem_event into common.
> > xen: Relocate p2m_mem_access_resume to mem_access common
> > xen: Relocate struct npfec definition into common
> > xen: Relocate mem_event_op domctl and access_op memop into common.
> > xen/mem_event: Clean out superfluous white-spaces
> > xen/mem_event: Relax error condition on debug builds
> > xen/mem_event: Abstract architecture specific sanity checks
> > xen/mem_access: Abstract architecture specific sanity check
>
> So I think up to here the series could go in afaic. Question is - does
> this make sense without the ARM parts?
>
> Thanks, Jan
>
Without the ARM parts there is no technical need for it. On the other hand,
there are many other series pending in the same general area so having this
part merged would put everyone on the same footing to resolve conflicts and
to move forward from.
Thanks,
Tamas
[-- Attachment #1.2: Type: text/html, Size: 2522 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 02/17] xen: Relocate p2m_mem_access_resume to mem_access common
2014-09-10 13:28 ` [PATCH v5 02/17] xen: Relocate p2m_mem_access_resume to mem_access common Tamas K Lengyel
@ 2014-09-11 20:16 ` Julien Grall
2014-09-12 8:56 ` Tamas K Lengyel
0 siblings, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-11 20:16 UTC (permalink / raw)
To: Tamas K Lengyel, xen-devel
Cc: ian.campbell, tim, ian.jackson, stefano.stabellini, andres,
jbeulich, dgdegra
Hi Tamas,
On 10/09/14 06:28, Tamas K Lengyel wrote:
> /* Registered with Xen-bound event channel for incoming notifications. */
> diff --git a/xen/include/asm-x86/p2m.h b/xen/include/asm-x86/p2m.h
> index 39f235d..2a8970b 100644
> --- a/xen/include/asm-x86/p2m.h
> +++ b/xen/include/asm-x86/p2m.h
> @@ -601,8 +601,6 @@ void p2m_mem_paging_resume(struct domain *d);
> bool_t p2m_mem_access_check(paddr_t gpa, unsigned long gla,
> struct npfec npfec,
> mem_event_request_t **req_ptr);
> -/* Resumes the running of the VCPU, restarting the last instruction */
> -void p2m_mem_access_resume(struct domain *d);
>
I think this comment is useful to keep. I will move it above
mem_access_resume prototype in mem_access.h
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-10 13:28 ` [PATCH v5 09/17] xen/arm: p2m type definitions and changes Tamas K Lengyel
@ 2014-09-11 20:25 ` Julien Grall
2014-09-12 8:15 ` Tamas K Lengyel
2014-09-11 20:49 ` Julien Grall
1 sibling, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-11 20:25 UTC (permalink / raw)
To: Tamas K Lengyel, xen-devel
Cc: ian.campbell, tim, ian.jackson, stefano.stabellini, andres,
jbeulich, dgdegra
Hi Tamas,
You skipped my comments/questions on v4.
On 10/09/14 06:28, Tamas K Lengyel wrote:
> diff --git a/xen/include/asm-arm/p2m.h b/xen/include/asm-arm/p2m.h
> index 08ce07b..b4ca86d 100644
> --- a/xen/include/asm-arm/p2m.h
> +++ b/xen/include/asm-arm/p2m.h
> @@ -2,6 +2,9 @@
[..]
> +#include <public/memory.h>
> +#include <public/mem_event.h>
Why do you need those 2 includes?
>
> #include <xen/p2m-common.h>
>
> @@ -11,6 +14,48 @@ struct domain;
>
> extern void memory_type_changed(struct domain *);
>
> +/* List of possible type for each page in the p2m entry.
> + * The number of available bit per page in the pte for this purpose is 4 bits.
> + * So it's possible to only have 16 fields. If we run out of value in the
> + * future, it's possible to use higher value for pseudo-type and don't store
> + * them in the p2m entry.
> + */
> +typedef enum {
> + p2m_invalid = 0, /* Nothing mapped here */
> + p2m_ram_rw, /* Normal read/write guest RAM */
> + p2m_ram_ro, /* Read-only; writes are silently dropped */
> + p2m_mmio_direct, /* Read/write mapping of genuine MMIO area */
> + p2m_map_foreign, /* Ram pages from foreign domain */
> + p2m_grant_map_rw, /* Read/write grant mapping */
> + p2m_grant_map_ro, /* Read-only grant mapping */
> + /* The types below are only used to decide the page attribute in the P2M */
> + p2m_iommu_map_rw, /* Read/write iommu mapping */
> + p2m_iommu_map_ro, /* Read-only iommu mapping */
> + p2m_max_real_type, /* Types after this won't be store in the p2m */
> +} p2m_type_t;
Any reason to move the enum earlier? If not, I would prefer to keep at
the same place. It will be easier with git-blame to find when a new type
has been added.
> +/*
> + * Additional access types, which are used to further restrict
> + * the permissions given by the p2m_type_t memory type. Violations
> + * caused by p2m_access_t restrictions are sent to the mem_event
> + * interface.
> + *
> + * The access permissions are soft state: when any ambigious change of page
ambiguous.
[..]
> + /* Default P2M access type for each page in the the domain: new pages,
> + * swapped in pages, cleared pages, and pages that are ambiquously
Did you intend to mean ambiguously rather than ambiquously?
[..]
> +/* mem_event and mem_access are supported on all ARM */
I don't find "all ARM" clear. I would replace by "any ARM guest"
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 10/17] xen/arm: Add set access required domctl
2014-09-10 13:28 ` [PATCH v5 10/17] xen/arm: Add set access required domctl Tamas K Lengyel
@ 2014-09-11 20:26 ` Julien Grall
0 siblings, 0 replies; 48+ messages in thread
From: Julien Grall @ 2014-09-11 20:26 UTC (permalink / raw)
To: Tamas K Lengyel, xen-devel
Cc: ian.campbell, tim, ian.jackson, stefano.stabellini, andres,
jbeulich, dgdegra
Hi,
On 10/09/14 06:28, Tamas K Lengyel wrote:
> Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
I though I gave my Reviewed-by on v4... Anyway:
Reviewed-by: Julien Grall <julien.grall@linaro.org>
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 11/17] xen/arm: Implement domain_get_maximum_gpfn
2014-09-10 13:28 ` [PATCH v5 11/17] xen/arm: Implement domain_get_maximum_gpfn Tamas K Lengyel
@ 2014-09-11 20:28 ` Julien Grall
2014-09-12 8:58 ` Tamas K Lengyel
0 siblings, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-11 20:28 UTC (permalink / raw)
To: Tamas K Lengyel, xen-devel
Cc: ian.campbell, tim, ian.jackson, stefano.stabellini, andres,
jbeulich, dgdegra
Hi Tamas,
On 10/09/14 06:28, Tamas K Lengyel wrote:
> The function domain_get_maximum_gpfn is returning the maximum gpfn ever
> mapped in the guest. We can use d->arch.p2m.max_mapped_gfn for this purpose.
>
> Suggested-by: Julien Grall <julien.grall@linaro.org>
> Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
This patch has originally been written by me. Please keep my
Signed-off-by and myself as the author of this patch.
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-10 13:28 ` [PATCH v5 09/17] xen/arm: p2m type definitions and changes Tamas K Lengyel
2014-09-11 20:25 ` Julien Grall
@ 2014-09-11 20:49 ` Julien Grall
2014-09-12 8:31 ` Tamas K Lengyel
1 sibling, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-11 20:49 UTC (permalink / raw)
To: Tamas K Lengyel, xen-devel
Cc: ian.campbell, tim, ian.jackson, stefano.stabellini, andres,
jbeulich, dgdegra
Hi Tamas,
On 10/09/14 06:28, Tamas K Lengyel wrote:
> static lpae_t mfn_to_p2m_entry(unsigned long mfn, unsigned int mattr,
> - p2m_type_t t)
> + p2m_type_t t, p2m_access_t a)
> {
> paddr_t pa = ((paddr_t) mfn) << PAGE_SHIFT;
> /* sh, xn and write bit will be defined in the following switches
> @@ -347,7 +350,7 @@ static int p2m_create_table(struct domain *d, lpae_t *entry,
> for ( i=0 ; i < LPAE_ENTRIES; i++ )
> {
> pte = mfn_to_p2m_entry(base_pfn + (i<<(level_shift-LPAE_SHIFT)),
> - MATTR_MEM, t);
> + MATTR_MEM, t, p2m->default_access);
I though I've talked about it in an earlier version. I don't think we
should use the default_access to the P2M table.
Let assume a user decides to set default to another access type than
p2m_access_rwx, Xen will receive the same trap forever for the domain
because the access is not store in the radix tree (see your patch #12).
Therefore the guest will try to access back to the guest, but the page
attribute has not changed.
Also, you can't associate a PFN to this page because the table is
internal to Xen.
IHMO, I would use p2m_access_rwx for the table L1 and L2 tables. For L3,
you will have to set the permission in the radix tree.
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events.
2014-09-10 13:28 ` [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events Tamas K Lengyel
@ 2014-09-11 21:19 ` Julien Grall
2014-09-12 8:46 ` Tamas K Lengyel
0 siblings, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-11 21:19 UTC (permalink / raw)
To: Tamas K Lengyel, xen-devel
Cc: ian.campbell, tim, ian.jackson, stefano.stabellini, andres,
jbeulich, dgdegra
Hello Tamas,
You forgot to handle add the permission in the radix when the a table is
shattered.
On 10/09/14 06:28, Tamas K Lengyel wrote:
> #define P2M_ONE_DESCEND 0
> #define P2M_ONE_PROGRESS_NOP 0x1
> #define P2M_ONE_PROGRESS 0x10
> @@ -504,6 +570,10 @@ static int apply_one_level(struct domain *d,
> page = alloc_domheap_pages(d, level_shift - PAGE_SHIFT, 0);
> if ( page )
> {
> + rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr), a);
It's possible to allocate a 2M/1G mapping here. In the case of memaccess
you only want 4K mapping for granularity.
> + if ( rc < 0 )
You should free the page via free_domheap_pages if Xen fail to adds the
access type in the radix tree.
> + return rc;
> +
> pte = mfn_to_p2m_entry(page_to_mfn(page), mattr, t, a);
> if ( level < 3 )
> pte.p2m.table = 0;
> @@ -538,6 +608,10 @@ static int apply_one_level(struct domain *d,
> /* We do not handle replacing an existing table with a superpage */
> (level == 3 || !p2m_table(orig_pte)) )
> {
> + rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr), a);
> + if ( rc < 0 )
> + return rc;
> +
Same remark here about the mapping.
> /* New mapping is superpage aligned, make it */
> pte = mfn_to_p2m_entry(*maddr >> PAGE_SHIFT, mattr, t, a);
> if ( level < 3 )
> @@ -663,6 +737,7 @@ static int apply_one_level(struct domain *d,
>
> memset(&pte, 0x00, sizeof(pte));
> p2m_write_pte(entry, pte, flush_cache);
> + radix_tree_delete(&p2m->mem_access_settings, paddr_to_pfn(*addr));
>
> *addr += level_size;
> *maddr += level_size;
> @@ -707,6 +782,53 @@ static int apply_one_level(struct domain *d,
> *addr += PAGE_SIZE;
> return P2M_ONE_PROGRESS_NOP;
> }
> +
> + case MEMACCESS:
> + if ( level < 3 )
> + {
> + if ( !p2m_valid(orig_pte) )
> + {
> + (*addr)++;
Why increment by 1? You the PTE doesn't contain valid mapping you want
to skip the whole level range. ie:
*addr += level_size;
> + return P2M_ONE_PROGRESS_NOP;
> + }
> +
> + /* Shatter large pages as we descend */
> + if ( p2m_mapping(orig_pte) )
> + {
> + rc = p2m_create_table(d, entry,
> + level_shift - PAGE_SHIFT, flush_cache);
> + if ( rc < 0 )
> + return rc;
> +
> + p2m->stats.shattered[level]++;
> + p2m->stats.mappings[level]--;
> + p2m->stats.mappings[level+1] += LPAE_ENTRIES;
> + } /* else: an existing table mapping -> descend */
> +
This piece of code is exactly the same in INSERT, REMOVE and now
MEMACCESS. I would create an helper to shatter and update the stats.
> + return P2M_ONE_DESCEND;
> + }
> + else
> + {
> + pte = orig_pte;
> +
> + if ( !p2m_table(pte) )
> + pte.bits = 0;
> +
> + if ( p2m_valid(pte) )
> + {
> + ASSERT(pte.p2m.type != p2m_invalid);
> +
> + rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr), a);
> + if ( rc < 0 )
> + return rc;
> +
> + p2m_set_permission(&pte, pte.p2m.type, a);
> + p2m_write_pte(entry, pte, flush_cache);
> + }
> +
> + (*addr)++;
*addr += PAGE_SIZE;
[..]
> +/* Set access type for a region of pfns.
> + * If start_pfn == -1ul, sets the default access type */
> +long p2m_set_mem_access(struct domain *d, unsigned long pfn, uint32_t nr,
> + uint32_t start, uint32_t mask, xenmem_access_t access)
> +{
> + struct p2m_domain *p2m = p2m_get_hostp2m(d);
> + p2m_access_t a;
> + long rc = 0;
> + paddr_t paddr;
> +
> + static const p2m_access_t memaccess[] = {
> +#define ACCESS(ac) [XENMEM_access_##ac] = p2m_access_##ac
> + ACCESS(n),
> + ACCESS(r),
> + ACCESS(w),
> + ACCESS(rw),
> + ACCESS(x),
> + ACCESS(rx),
> + ACCESS(wx),
> + ACCESS(rwx),
> +#undef ACCESS
> + };
> +
> + switch ( access )
> + {
> + case 0 ... ARRAY_SIZE(memaccess) - 1:
> + a = memaccess[access];
> + break;
> + case XENMEM_access_default:
> + a = p2m->default_access;
> + break;
> + default:
> + return -EINVAL;
> + }
> +
> + /* If request to set default access */
> + if ( pfn == ~0ul )
> + {
> + p2m->default_access = a;
> + return 0;
> + }
> +
> + for ( pfn += start; nr > start; ++pfn )
> + {
> + paddr = pfn_to_paddr(pfn);
> + rc = apply_p2m_changes(d, MEMACCESS, paddr, paddr+1, 0, MATTR_MEM, 0, a);
Hmmm... why didn't you call directly apply_p2m_changes with the whole range?
[..]
> +int p2m_get_mem_access(struct domain *d, unsigned long gpfn,
> + xenmem_access_t *access)
> +{
> + struct p2m_domain *p2m = p2m_get_hostp2m(d);
> + void *i;
> + unsigned int index;
> +
> + static const xenmem_access_t memaccess[] = {
> +#define ACCESS(ac) [XENMEM_access_##ac] = XENMEM_access_##ac
> + ACCESS(n),
> + ACCESS(r),
> + ACCESS(w),
> + ACCESS(rw),
> + ACCESS(x),
> + ACCESS(rx),
> + ACCESS(wx),
> + ACCESS(rwx),
> +#undef ACCESS
> + };
> +
> + /* If request to get default access */
> + if ( gpfn == ~0ull )
> + {
> + *access = memaccess[p2m->default_access];
> + return 0;
> + }
> +
> + spin_lock(&p2m->lock);
> +
> + i = radix_tree_lookup(&p2m->mem_access_settings, gpfn);
> +
> + spin_unlock(&p2m->lock);
> +
> + if ( !i )
> + return -ESRCH;
If the gpfn is not in the radix tree, it means that either the mapping
doesn't exists or the access type is p2m_access_rwx.
You handle the former case but not the latter.
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 13/17] xen/arm: Instruction prefetch abort (X) mem_event handling
2014-09-10 13:28 ` [PATCH v5 13/17] xen/arm: Instruction prefetch abort (X) mem_event handling Tamas K Lengyel
@ 2014-09-11 21:23 ` Julien Grall
2014-09-12 8:34 ` Tamas K Lengyel
0 siblings, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-11 21:23 UTC (permalink / raw)
To: Tamas K Lengyel, xen-devel
Cc: ian.campbell, tim, ian.jackson, stefano.stabellini, andres,
jbeulich, dgdegra
Hello Tamas,
On 10/09/14 06:28, Tamas K Lengyel wrote:
> + default:
> + break;
> + }
I don't think the default: break; is useful here. I would drop it.
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 17/17] tools/tests: Enable xen-access on ARM
2014-09-10 13:28 ` [PATCH v5 17/17] tools/tests: Enable xen-access " Tamas K Lengyel
@ 2014-09-11 21:29 ` Julien Grall
2014-09-12 8:50 ` Tamas K Lengyel
2014-09-12 9:01 ` Tamas K Lengyel
0 siblings, 2 replies; 48+ messages in thread
From: Julien Grall @ 2014-09-11 21:29 UTC (permalink / raw)
To: Tamas K Lengyel, xen-devel
Cc: ian.campbell, tim, ian.jackson, stefano.stabellini, andres,
jbeulich, dgdegra
Hello Tamas,
On 10/09/14 06:28, Tamas K Lengyel wrote:
> diff --git a/config/arm32.mk b/config/arm32.mk
> index aa79d22..4a7c259 100644
> --- a/config/arm32.mk
> +++ b/config/arm32.mk
> @@ -13,6 +13,7 @@ HAS_PL011 := y
> HAS_EXYNOS4210 := y
> HAS_OMAP := y
> HAS_NS16550 := y
> +HAS_MEM_ACCESS := y
>
> # Use only if calling $(LD) directly.
> LDFLAGS_DIRECT += -EL
> diff --git a/config/arm64.mk b/config/arm64.mk
> index 15b57a4..ea6558d 100644
> --- a/config/arm64.mk
> +++ b/config/arm64.mk
> @@ -8,6 +8,7 @@ CFLAGS += #-marm -march= -mcpu= etc
>
> HAS_PL011 := y
> HAS_NS16550 := y
> +HAS_MEM_ACCESS := y
Why can't you move HAS_MEM_ACCESS directly here in patch #1?
It would make this patch simpler.
[..]
> @@ -492,8 +523,9 @@ int main(int argc, char *argv[])
> goto exit;
> }
>
> - rc = xc_set_mem_access(xch, domain_id, default_access, 0,
> - xenaccess->domain_info->max_pages);
> + rc = xc_set_mem_access(xch, domain_id, default_access, START_PFN,
> + (xenaccess->domain_info->max_gpfn - START_PFN) );
> +
max_gpfn contains the maximum GPFN mapped (it could be foreign page,
MMIO range...).
Before memaccess was used only on RAM. Now it can contains any other
page type.
I think this code has to be smarter and take into account the banks on ARM.
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-11 20:25 ` Julien Grall
@ 2014-09-12 8:15 ` Tamas K Lengyel
2014-09-12 19:23 ` Julien Grall
0 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 8:15 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 3347 bytes --]
On Thu, Sep 11, 2014 at 10:25 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hi Tamas,
>
> You skipped my comments/questions on v4.
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
>
>> diff --git a/xen/include/asm-arm/p2m.h b/xen/include/asm-arm/p2m.h
>> index 08ce07b..b4ca86d 100644
>> --- a/xen/include/asm-arm/p2m.h
>> +++ b/xen/include/asm-arm/p2m.h
>> @@ -2,6 +2,9 @@
>>
>
> [..]
>
>
> +#include <public/memory.h>
>> +#include <public/mem_event.h>
>>
>
> Why do you need those 2 includes?
>
<public/memory.h> might not be necessary in this patch yet, but it will be
required for defining the functions that mem_access will call passing
xenmem_access_t access. I can move the header include into the next patch
in the series if that is cleaner. The mem_event header is not actually
required so I'll remove it.
>
>> #include <xen/p2m-common.h>
>>
>> @@ -11,6 +14,48 @@ struct domain;
>>
>> extern void memory_type_changed(struct domain *);
>>
>> +/* List of possible type for each page in the p2m entry.
>> + * The number of available bit per page in the pte for this purpose is 4
>> bits.
>> + * So it's possible to only have 16 fields. If we run out of value in the
>> + * future, it's possible to use higher value for pseudo-type and don't
>> store
>> + * them in the p2m entry.
>> + */
>> +typedef enum {
>> + p2m_invalid = 0, /* Nothing mapped here */
>> + p2m_ram_rw, /* Normal read/write guest RAM */
>> + p2m_ram_ro, /* Read-only; writes are silently dropped */
>> + p2m_mmio_direct, /* Read/write mapping of genuine MMIO area */
>> + p2m_map_foreign, /* Ram pages from foreign domain */
>> + p2m_grant_map_rw, /* Read/write grant mapping */
>> + p2m_grant_map_ro, /* Read-only grant mapping */
>> + /* The types below are only used to decide the page attribute in the
>> P2M */
>> + p2m_iommu_map_rw, /* Read/write iommu mapping */
>> + p2m_iommu_map_ro, /* Read-only iommu mapping */
>> + p2m_max_real_type, /* Types after this won't be store in the p2m */
>> +} p2m_type_t;
>>
>
> Any reason to move the enum earlier? If not, I would prefer to keep at the
> same place. It will be easier with git-blame to find when a new type has
> been added.
>
Stylistically it made more sense to have p2m_type_t and p2m_access_t
together (as it is on the x86 as well). And they are defined here as the
p2m_domain does have a field now defining default_access with p2m_access_t.
>
> +/*
>> + * Additional access types, which are used to further restrict
>> + * the permissions given by the p2m_type_t memory type. Violations
>> + * caused by p2m_access_t restrictions are sent to the mem_event
>> + * interface.
>> + *
>> + * The access permissions are soft state: when any ambigious change of
>> page
>>
>
> ambiguous.
>
Copy-pasta but will fix.
>
> [..]
>
> + /* Default P2M access type for each page in the the domain: new
>> pages,
>> + * swapped in pages, cleared pages, and pages that are ambiquously
>>
>
> Did you intend to mean ambiguously rather than ambiquously?
>
>
Copy-pasta again but will fix. Maybe in a separate patch where I fix it
here and in the x86 side as well?
> [..]
>
> +/* mem_event and mem_access are supported on all ARM */
>>
>
> I don't find "all ARM" clear. I would replace by "any ARM guest"
>
>
Ack.
Thanks!
Tamas
[-- Attachment #1.2: Type: text/html, Size: 5585 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-11 20:49 ` Julien Grall
@ 2014-09-12 8:31 ` Tamas K Lengyel
2014-09-12 19:41 ` Julien Grall
0 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 8:31 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 2395 bytes --]
On Thu, Sep 11, 2014 at 10:49 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hi Tamas,
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
> > static lpae_t mfn_to_p2m_entry(unsigned long mfn, unsigned int mattr,
>
>> - p2m_type_t t)
>> + p2m_type_t t, p2m_access_t a)
>> {
>> paddr_t pa = ((paddr_t) mfn) << PAGE_SHIFT;
>> /* sh, xn and write bit will be defined in the following switches
>> @@ -347,7 +350,7 @@ static int p2m_create_table(struct domain *d, lpae_t
>> *entry,
>> for ( i=0 ; i < LPAE_ENTRIES; i++ )
>> {
>> pte = mfn_to_p2m_entry(base_pfn +
>> (i<<(level_shift-LPAE_SHIFT)),
>> - MATTR_MEM, t);
>> + MATTR_MEM, t, p2m->default_access);
>>
>
> I though I've talked about it in an earlier version. I don't think we
> should use the default_access to the P2M table.
>
> Let assume a user decides to set default to another access type than
> p2m_access_rwx, Xen will receive the same trap forever for the domain
> because the access is not store in the radix tree (see your patch #12).
> Therefore the guest will try to access back to the guest, but the page
> attribute has not changed.
>
Also, you can't associate a PFN to this page because the table is internal
> to Xen.
>
> IHMO, I would use p2m_access_rwx for the table L1 and L2 tables. For L3,
> you will have to set the permission in the radix tree.
>
>
I'm not sure I follow exactly what you are saying. Setting default_access
only effects pages created after the fact and I see that if a large page is
created this still causes a problem.. Is that what you mean? A better
solution would be to only allow creating 4k pages if default_access != rwx
IMHO and then store the setting in the radix tree automatically.
I was also contemplating if it would make sense if a setting was not found
in the radix tree, to check the violation against the default_access. That
way we would not have to store those entries in the radix tree that are the
same as the default_access (and rwx of course would not go in the radix
tree either). Might cut down the size of the tree.
> Regards,
>
> --
> Julien Grall
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
[-- Attachment #1.2: Type: text/html, Size: 3574 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 13/17] xen/arm: Instruction prefetch abort (X) mem_event handling
2014-09-11 21:23 ` Julien Grall
@ 2014-09-12 8:34 ` Tamas K Lengyel
0 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 8:34 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 615 bytes --]
On Thu, Sep 11, 2014 at 11:23 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hello Tamas,
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
>
>> + default:
>> + break;
>> + }
>>
>
> I don't think the default: break; is useful here. I would drop it.
>
>
I seem to recall some compilers warning me in the past if there was no
default case (in a separate project), but that might just have been an
overly restrictive setting I had.
Tamas
>
> --
> Julien Grall
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
[-- Attachment #1.2: Type: text/html, Size: 1569 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events.
2014-09-11 21:19 ` Julien Grall
@ 2014-09-12 8:46 ` Tamas K Lengyel
2014-09-12 20:35 ` Julien Grall
0 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 8:46 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 7161 bytes --]
On Thu, Sep 11, 2014 at 11:19 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hello Tamas,
>
> You forgot to handle add the permission in the radix when the a table is
> shattered.
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
>
>> #define P2M_ONE_DESCEND 0
>> #define P2M_ONE_PROGRESS_NOP 0x1
>> #define P2M_ONE_PROGRESS 0x10
>> @@ -504,6 +570,10 @@ static int apply_one_level(struct domain *d,
>> page = alloc_domheap_pages(d, level_shift - PAGE_SHIFT, 0);
>> if ( page )
>> {
>> + rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr),
>> a);
>>
>
> It's possible to allocate a 2M/1G mapping here. In the case of memaccess
> you only want 4K mapping for granularity.
>
Right, I should only set it if level==3.
>
> + if ( rc < 0 )
>>
>
> You should free the page via free_domheap_pages if Xen fail to adds the
> access type in the radix tree.
>
Ack.
> + return rc;
>> +
>> pte = mfn_to_p2m_entry(page_to_mfn(page), mattr, t, a);
>> if ( level < 3 )
>> pte.p2m.table = 0;
>> @@ -538,6 +608,10 @@ static int apply_one_level(struct domain *d,
>> /* We do not handle replacing an existing table with a
>> superpage */
>> (level == 3 || !p2m_table(orig_pte)) )
>> {
>> + rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr), a);
>> + if ( rc < 0 )
>> + return rc;
>> +
>>
>
> Same remark here about the mapping.
>
>
Ack.
> /* New mapping is superpage aligned, make it */
>> pte = mfn_to_p2m_entry(*maddr >> PAGE_SHIFT, mattr, t, a);
>> if ( level < 3 )
>> @@ -663,6 +737,7 @@ static int apply_one_level(struct domain *d,
>>
>> memset(&pte, 0x00, sizeof(pte));
>> p2m_write_pte(entry, pte, flush_cache);
>> + radix_tree_delete(&p2m->mem_access_settings,
>> paddr_to_pfn(*addr));
>>
>> *addr += level_size;
>> *maddr += level_size;
>> @@ -707,6 +782,53 @@ static int apply_one_level(struct domain *d,
>> *addr += PAGE_SIZE;
>> return P2M_ONE_PROGRESS_NOP;
>> }
>> +
>> + case MEMACCESS:
>> + if ( level < 3 )
>> + {
>> + if ( !p2m_valid(orig_pte) )
>> + {
>> + (*addr)++;
>>
>
> Why increment by 1? You the PTE doesn't contain valid mapping you want to
> skip the whole level range. ie:
>
> *addr += level_size;
>
>
It doesn't make a difference, apply_p2m_changes is called with start=paddr,
end=paddr+1 from a separate loop. So just incrementing it by one or a whole
level achieves the same effect, that is, the apply_p2m_changes loop breaks.
> + return P2M_ONE_PROGRESS_NOP;
>> + }
>> +
>> + /* Shatter large pages as we descend */
>> + if ( p2m_mapping(orig_pte) )
>> + {
>> + rc = p2m_create_table(d, entry,
>> + level_shift - PAGE_SHIFT,
>> flush_cache);
>> + if ( rc < 0 )
>> + return rc;
>> +
>> + p2m->stats.shattered[level]++;
>> + p2m->stats.mappings[level]--;
>> + p2m->stats.mappings[level+1] += LPAE_ENTRIES;
>> + } /* else: an existing table mapping -> descend */
>> +
>>
>
> This piece of code is exactly the same in INSERT, REMOVE and now
> MEMACCESS. I would create an helper to shatter and update the stats.
>
Ack.
>
> + return P2M_ONE_DESCEND;
>> + }
>> + else
>> + {
>> + pte = orig_pte;
>> +
>> + if ( !p2m_table(pte) )
>> + pte.bits = 0;
>> +
>> + if ( p2m_valid(pte) )
>> + {
>> + ASSERT(pte.p2m.type != p2m_invalid);
>> +
>> + rc = p2m_mem_access_radix_set(p2m, paddr_to_pfn(*addr),
>> a);
>> + if ( rc < 0 )
>> + return rc;
>> +
>> + p2m_set_permission(&pte, pte.p2m.type, a);
>> + p2m_write_pte(entry, pte, flush_cache);
>> + }
>> +
>> + (*addr)++;
>>
>
> *addr += PAGE_SIZE;
>
> [..]
>
>
> +/* Set access type for a region of pfns.
>> + * If start_pfn == -1ul, sets the default access type */
>> +long p2m_set_mem_access(struct domain *d, unsigned long pfn, uint32_t nr,
>> + uint32_t start, uint32_t mask, xenmem_access_t
>> access)
>> +{
>> + struct p2m_domain *p2m = p2m_get_hostp2m(d);
>> + p2m_access_t a;
>> + long rc = 0;
>> + paddr_t paddr;
>> +
>> + static const p2m_access_t memaccess[] = {
>> +#define ACCESS(ac) [XENMEM_access_##ac] = p2m_access_##ac
>> + ACCESS(n),
>> + ACCESS(r),
>> + ACCESS(w),
>> + ACCESS(rw),
>> + ACCESS(x),
>> + ACCESS(rx),
>> + ACCESS(wx),
>> + ACCESS(rwx),
>> +#undef ACCESS
>> + };
>> +
>> + switch ( access )
>> + {
>> + case 0 ... ARRAY_SIZE(memaccess) - 1:
>> + a = memaccess[access];
>> + break;
>> + case XENMEM_access_default:
>> + a = p2m->default_access;
>> + break;
>> + default:
>> + return -EINVAL;
>> + }
>> +
>> + /* If request to set default access */
>> + if ( pfn == ~0ul )
>> + {
>> + p2m->default_access = a;
>> + return 0;
>> + }
>> +
>> + for ( pfn += start; nr > start; ++pfn )
>> + {
>> + paddr = pfn_to_paddr(pfn);
>> + rc = apply_p2m_changes(d, MEMACCESS, paddr, paddr+1, 0,
>> MATTR_MEM, 0, a);
>>
>
> Hmmm... why didn't you call directly apply_p2m_changes with the whole
> range?
>
>
Because the hypercall continuation. Setting mem_access permissions needs to
be preemptible and it has its own separate routine to do that here. See
http://xenbits.xen.org/xsa/advisory-89.html for more info.
> [..]
>
>
> +int p2m_get_mem_access(struct domain *d, unsigned long gpfn,
>> + xenmem_access_t *access)
>> +{
>> + struct p2m_domain *p2m = p2m_get_hostp2m(d);
>> + void *i;
>> + unsigned int index;
>> +
>> + static const xenmem_access_t memaccess[] = {
>> +#define ACCESS(ac) [XENMEM_access_##ac] = XENMEM_access_##ac
>> + ACCESS(n),
>> + ACCESS(r),
>> + ACCESS(w),
>> + ACCESS(rw),
>> + ACCESS(x),
>> + ACCESS(rx),
>> + ACCESS(wx),
>> + ACCESS(rwx),
>> +#undef ACCESS
>> + };
>> +
>> + /* If request to get default access */
>> + if ( gpfn == ~0ull )
>> + {
>> + *access = memaccess[p2m->default_access];
>> + return 0;
>> + }
>> +
>> + spin_lock(&p2m->lock);
>> +
>> + i = radix_tree_lookup(&p2m->mem_access_settings, gpfn);
>> +
>> + spin_unlock(&p2m->lock);
>> +
>> + if ( !i )
>> + return -ESRCH;
>>
>
> If the gpfn is not in the radix tree, it means that either the mapping
> doesn't exists or the access type is p2m_access_rwx.
>
> You handle the former case but not the latter.
>
Ack.
>
> Regards,
>
> --
> Julien Grall
>
>
Thanks!
Tamas
[-- Attachment #1.2: Type: text/html, Size: 10796 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 17/17] tools/tests: Enable xen-access on ARM
2014-09-11 21:29 ` Julien Grall
@ 2014-09-12 8:50 ` Tamas K Lengyel
2014-09-12 9:01 ` Tamas K Lengyel
1 sibling, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 8:50 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 1841 bytes --]
On Thu, Sep 11, 2014 at 11:29 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hello Tamas,
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
>
>> diff --git a/config/arm32.mk b/config/arm32.mk
>> index aa79d22..4a7c259 100644
>> --- a/config/arm32.mk
>> +++ b/config/arm32.mk
>> @@ -13,6 +13,7 @@ HAS_PL011 := y
>> HAS_EXYNOS4210 := y
>> HAS_OMAP := y
>> HAS_NS16550 := y
>> +HAS_MEM_ACCESS := y
>>
>> # Use only if calling $(LD) directly.
>> LDFLAGS_DIRECT += -EL
>> diff --git a/config/arm64.mk b/config/arm64.mk
>> index 15b57a4..ea6558d 100644
>> --- a/config/arm64.mk
>> +++ b/config/arm64.mk
>> @@ -8,6 +8,7 @@ CFLAGS += #-marm -march= -mcpu= etc
>>
>> HAS_PL011 := y
>> HAS_NS16550 := y
>> +HAS_MEM_ACCESS := y
>>
>
> Why can't you move HAS_MEM_ACCESS directly here in patch #1?
>
> It would make this patch simpler.
>
>
> [..]
>
> @@ -492,8 +523,9 @@ int main(int argc, char *argv[])
>> goto exit;
>> }
>>
>> - rc = xc_set_mem_access(xch, domain_id, default_access, 0,
>> - xenaccess->domain_info->max_pages);
>> + rc = xc_set_mem_access(xch, domain_id, default_access, START_PFN,
>> + (xenaccess->domain_info->max_gpfn -
>> START_PFN) );
>> +
>>
>
> max_gpfn contains the maximum GPFN mapped (it could be foreign page, MMIO
> range...).
>
> Before memaccess was used only on RAM. Now it can contains any other page
> type.
>
> I think this code has to be smarter and take into account the banks on ARM.
>
>
This is just a sample application, nothing prevents another user of the
same API on using it this way. If this is a problem, it needs to be handled
within Xen.
> Regards,
>
> --
> Julien Grall
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
[-- Attachment #1.2: Type: text/html, Size: 3349 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 02/17] xen: Relocate p2m_mem_access_resume to mem_access common
2014-09-11 20:16 ` Julien Grall
@ 2014-09-12 8:56 ` Tamas K Lengyel
0 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 8:56 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 1076 bytes --]
On Thu, Sep 11, 2014 at 10:16 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hi Tamas,
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
>
>> /* Registered with Xen-bound event channel for incoming notifications.
>> */
>> diff --git a/xen/include/asm-x86/p2m.h b/xen/include/asm-x86/p2m.h
>> index 39f235d..2a8970b 100644
>> --- a/xen/include/asm-x86/p2m.h
>> +++ b/xen/include/asm-x86/p2m.h
>> @@ -601,8 +601,6 @@ void p2m_mem_paging_resume(struct domain *d);
>> bool_t p2m_mem_access_check(paddr_t gpa, unsigned long gla,
>> struct npfec npfec,
>> mem_event_request_t **req_ptr);
>> -/* Resumes the running of the VCPU, restarting the last instruction */
>>
> > -void p2m_mem_access_resume(struct domain *d);
> >
>
> I think this comment is useful to keep. I will move it above
> mem_access_resume prototype in mem_access.h
>
>
Ack.
> Regards,
>
> --
> Julien Grall
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
[-- Attachment #1.2: Type: text/html, Size: 1976 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 11/17] xen/arm: Implement domain_get_maximum_gpfn
2014-09-11 20:28 ` Julien Grall
@ 2014-09-12 8:58 ` Tamas K Lengyel
0 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 8:58 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 780 bytes --]
On Thu, Sep 11, 2014 at 10:28 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hi Tamas,
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
>
>> The function domain_get_maximum_gpfn is returning the maximum gpfn ever
>> mapped in the guest. We can use d->arch.p2m.max_mapped_gfn for this
>> purpose.
>>
>> Suggested-by: Julien Grall <julien.grall@linaro.org>
>> Signed-off-by: Tamas K Lengyel <tklengyel@sec.in.tum.de>
>>
>
> This patch has originally been written by me. Please keep my Signed-off-by
> and myself as the author of this patch.
>
>
Ack, wasn't sure what the protocol is here.
Thanks!
Tamas
> Regards,
>
> --
> Julien Grall
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
[-- Attachment #1.2: Type: text/html, Size: 1820 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 17/17] tools/tests: Enable xen-access on ARM
2014-09-11 21:29 ` Julien Grall
2014-09-12 8:50 ` Tamas K Lengyel
@ 2014-09-12 9:01 ` Tamas K Lengyel
1 sibling, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 9:01 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 1005 bytes --]
On Thu, Sep 11, 2014 at 11:29 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hello Tamas,
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
>
>> diff --git a/config/arm32.mk b/config/arm32.mk
>> index aa79d22..4a7c259 100644
>> --- a/config/arm32.mk
>> +++ b/config/arm32.mk
>> @@ -13,6 +13,7 @@ HAS_PL011 := y
>> HAS_EXYNOS4210 := y
>> HAS_OMAP := y
>> HAS_NS16550 := y
>> +HAS_MEM_ACCESS := y
>>
>> # Use only if calling $(LD) directly.
>> LDFLAGS_DIRECT += -EL
>> diff --git a/config/arm64.mk b/config/arm64.mk
>> index 15b57a4..ea6558d 100644
>> --- a/config/arm64.mk
>> +++ b/config/arm64.mk
>> @@ -8,6 +8,7 @@ CFLAGS += #-marm -march= -mcpu= etc
>>
>> HAS_PL011 := y
>> HAS_NS16550 := y
>> +HAS_MEM_ACCESS := y
>>
>
> Why can't you move HAS_MEM_ACCESS directly here in patch #1?
>
> It would make this patch simpler.
>
It would but having HAS_MEM_ACCESS in config/* is not really required up
until this patch. I'm fine with it either way, I just thought this is the
clean solution.
[-- Attachment #1.2: Type: text/html, Size: 1909 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-12 8:15 ` Tamas K Lengyel
@ 2014-09-12 19:23 ` Julien Grall
2014-09-12 20:25 ` Tamas K Lengyel
0 siblings, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-12 19:23 UTC (permalink / raw)
To: Tamas K Lengyel
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
H Tamas,
On 12/09/14 01:15, Tamas K Lengyel wrote:
>
>
> On Thu, Sep 11, 2014 at 10:25 PM, Julien Grall <julien.grall@linaro.org
> <mailto:julien.grall@linaro.org>> wrote:
>
> Hi Tamas,
>
> You skipped my comments/questions on v4.
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
>
> diff --git a/xen/include/asm-arm/p2m.h b/xen/include/asm-arm/p2m.h
> index 08ce07b..b4ca86d 100644
> --- a/xen/include/asm-arm/p2m.h
> +++ b/xen/include/asm-arm/p2m.h
> @@ -2,6 +2,9 @@
>
>
> [..]
>
>
> +#include <public/memory.h>
> +#include <public/mem_event.h>
>
>
> Why do you need those 2 includes?
>
>
>
> <public/memory.h> might not be necessary in this patch yet, but it will
> be required for defining the functions that mem_access will call passing
> xenmem_access_t access. I can move the header include into the next
> patch in the series if that is cleaner. The mem_event header is not
> actually required so I'll remove it.
Yes please.
>
> #include <xen/p2m-common.h>
>
> @@ -11,6 +14,48 @@ struct domain;
>
> extern void memory_type_changed(struct domain *);
>
> +/* List of possible type for each page in the p2m entry.
> + * The number of available bit per page in the pte for this
> purpose is 4 bits.
> + * So it's possible to only have 16 fields. If we run out of
> value in the
> + * future, it's possible to use higher value for pseudo-type
> and don't store
> + * them in the p2m entry.
> + */
> +typedef enum {
> + p2m_invalid = 0, /* Nothing mapped here */
> + p2m_ram_rw, /* Normal read/write guest RAM */
> + p2m_ram_ro, /* Read-only; writes are silently
> dropped */
> + p2m_mmio_direct, /* Read/write mapping of genuine MMIO
> area */
> + p2m_map_foreign, /* Ram pages from foreign domain */
> + p2m_grant_map_rw, /* Read/write grant mapping */
> + p2m_grant_map_ro, /* Read-only grant mapping */
> + /* The types below are only used to decide the page
> attribute in the P2M */
> + p2m_iommu_map_rw, /* Read/write iommu mapping */
> + p2m_iommu_map_ro, /* Read-only iommu mapping */
> + p2m_max_real_type, /* Types after this won't be store in
> the p2m */
> +} p2m_type_t;
>
>
> Any reason to move the enum earlier? If not, I would prefer to keep
> at the same place. It will be easier with git-blame to find when a
> new type has been added.
>
>
> Stylistically it made more sense to have p2m_type_t and p2m_access_t
> together (as it is on the x86 as well). And they are defined here as the
> p2m_domain does have a field now defining default_access with p2m_access_t.
If it's only for the style I wouldn't move them.
Anyway, I'll Ian/Stefano decides about this.
>
> +/*
> + * Additional access types, which are used to further restrict
> + * the permissions given by the p2m_type_t memory type. Violations
> + * caused by p2m_access_t restrictions are sent to the mem_event
> + * interface.
> + *
> + * The access permissions are soft state: when any ambigious
> change of page
>
>
> ambiguous.
>
>
> Copy-pasta but will fix.
>
>
> [..]
>
> + /* Default P2M access type for each page in the the domain:
> new pages,
> + * swapped in pages, cleared pages, and pages that are
> ambiquously
>
>
> Did you intend to mean ambiguously rather than ambiquously?
>
>
> Copy-pasta again but will fix. Maybe in a separate patch where I fix it
> here and in the x86 side as well?
I'm OK for a separate patch fixing x86 side, but there is no reason to
fix the spelling for ARM outside this patch.
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-12 8:31 ` Tamas K Lengyel
@ 2014-09-12 19:41 ` Julien Grall
2014-09-12 20:20 ` Tamas K Lengyel
0 siblings, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-12 19:41 UTC (permalink / raw)
To: Tamas K Lengyel
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
Hello Tamas,
On 12/09/14 01:31, Tamas K Lengyel wrote:
>
>
> On Thu, Sep 11, 2014 at 10:49 PM, Julien Grall <julien.grall@linaro.org
> <mailto:julien.grall@linaro.org>> wrote:
>
> Hi Tamas,
>
> On 10/09/14 06:28, Tamas K Lengyel wrote:
> > static lpae_t mfn_to_p2m_entry(unsigned long mfn, unsigned int
> mattr,
>
> - p2m_type_t t)
> + p2m_type_t t, p2m_access_t a)
> {
> paddr_t pa = ((paddr_t) mfn) << PAGE_SHIFT;
> /* sh, xn and write bit will be defined in the following
> switches
> @@ -347,7 +350,7 @@ static int p2m_create_table(struct domain
> *d, lpae_t *entry,
> for ( i=0 ; i < LPAE_ENTRIES; i++ )
> {
> pte = mfn_to_p2m_entry(base_pfn +
> (i<<(level_shift-LPAE_SHIFT)),
> - MATTR_MEM, t);
> + MATTR_MEM, t,
> p2m->default_access);
>
>
> I though I've talked about it in an earlier version. I don't think
> we should use the default_access to the P2M table.
>
> Let assume a user decides to set default to another access type than
> p2m_access_rwx, Xen will receive the same trap forever for the
> domain because the access is not store in the radix tree (see your
> patch #12). Therefore the guest will try to access back to the
> guest, but the page attribute has not changed.
>
> Also, you can't associate a PFN to this page because the table is
> internal to Xen.
>
> IHMO, I would use p2m_access_rwx for the table L1 and L2 tables. For
> L3, you will have to set the permission in the radix tree.
>
>
> I'm not sure I follow exactly what you are saying. Setting
> default_access only effects pages created after the fact and I see that
> if a large page is created this still causes a problem.. Is that what
> you mean? A better solution would be to only allow creating 4k pages if
> default_access != rwx IMHO and then store the setting in the radix tree
> automatically.
That what I was trying to say.
You can do it easily in the function is_mapping_aligned.
But for the intermediate page table (see the mfn_to_p2m_entry(...,...,
p2m_invalid), you should not use default_access but directly access_rwx.
>
> I was also contemplating if it would make sense if a setting was not
> found in the radix tree, to check the violation against the
> default_access.
The violation of the page may be different than the default access. I
think it could be a mess for the guest. The radix tree should containe
very pte where the access != rwx.
> That way we would not have to store those entries in the
> radix tree that are the same as the default_access (and rwx of course
> would not go in the radix tree either). Might cut down the size of the tree.
The default_access may change after the insertion of the PTE. How will
you handle this case? IHMO, I don't think there is simple solution.
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-12 19:41 ` Julien Grall
@ 2014-09-12 20:20 ` Tamas K Lengyel
0 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 20:20 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 3420 bytes --]
On Fri, Sep 12, 2014 at 9:41 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hello Tamas,
>
> On 12/09/14 01:31, Tamas K Lengyel wrote:
>
>>
>>
>> On Thu, Sep 11, 2014 at 10:49 PM, Julien Grall <julien.grall@linaro.org
>> <mailto:julien.grall@linaro.org>> wrote:
>>
>> Hi Tamas,
>>
>> On 10/09/14 06:28, Tamas K Lengyel wrote:
>> > static lpae_t mfn_to_p2m_entry(unsigned long mfn, unsigned int
>> mattr,
>>
>> - p2m_type_t t)
>> + p2m_type_t t, p2m_access_t a)
>> {
>> paddr_t pa = ((paddr_t) mfn) << PAGE_SHIFT;
>> /* sh, xn and write bit will be defined in the following
>> switches
>> @@ -347,7 +350,7 @@ static int p2m_create_table(struct domain
>> *d, lpae_t *entry,
>> for ( i=0 ; i < LPAE_ENTRIES; i++ )
>> {
>> pte = mfn_to_p2m_entry(base_pfn +
>> (i<<(level_shift-LPAE_SHIFT)),
>> - MATTR_MEM, t);
>> + MATTR_MEM, t,
>> p2m->default_access);
>>
>>
>> I though I've talked about it in an earlier version. I don't think
>> we should use the default_access to the P2M table.
>>
>> Let assume a user decides to set default to another access type than
>> p2m_access_rwx, Xen will receive the same trap forever for the
>> domain because the access is not store in the radix tree (see your
>> patch #12). Therefore the guest will try to access back to the
>> guest, but the page attribute has not changed.
>>
>> Also, you can't associate a PFN to this page because the table is
>> internal to Xen.
>>
>> IHMO, I would use p2m_access_rwx for the table L1 and L2 tables. For
>> L3, you will have to set the permission in the radix tree.
>>
>>
>> I'm not sure I follow exactly what you are saying. Setting
>> default_access only effects pages created after the fact and I see that
>> if a large page is created this still causes a problem.. Is that what
>> you mean? A better solution would be to only allow creating 4k pages if
>> default_access != rwx IMHO and then store the setting in the radix tree
>> automatically.
>>
>
> That what I was trying to say.
>
> You can do it easily in the function is_mapping_aligned.
>
> But for the intermediate page table (see the mfn_to_p2m_entry(...,...,
> p2m_invalid), you should not use default_access but directly access_rwx.
>
Ack.
>
>
>
>> I was also contemplating if it would make sense if a setting was not
>> found in the radix tree, to check the violation against the
>> default_access.
>>
>
> The violation of the page may be different than the default access. I
> think it could be a mess for the guest. The radix tree should containe very
> pte where the access != rwx.
>
>
> That way we would not have to store those entries in the
>> radix tree that are the same as the default_access (and rwx of course
>> would not go in the radix tree either). Might cut down the size of the
>> tree.
>>
>
> The default_access may change after the insertion of the PTE. How will you
> handle this case? IHMO, I don't think there is simple solution.
>
>
Yea, I gave it some thought today and it became quite complex so it's not
worth the trouble.
> Regards,
>
> --
> Julien Grall
>
Thanks!
Tamas
[-- Attachment #1.2: Type: text/html, Size: 5020 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 09/17] xen/arm: p2m type definitions and changes
2014-09-12 19:23 ` Julien Grall
@ 2014-09-12 20:25 ` Tamas K Lengyel
0 siblings, 0 replies; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 20:25 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 4259 bytes --]
On Fri, Sep 12, 2014 at 9:23 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> H Tamas,
>
> On 12/09/14 01:15, Tamas K Lengyel wrote:
>
>>
>>
>> On Thu, Sep 11, 2014 at 10:25 PM, Julien Grall <julien.grall@linaro.org
>> <mailto:julien.grall@linaro.org>> wrote:
>>
>> Hi Tamas,
>>
>> You skipped my comments/questions on v4.
>>
>> On 10/09/14 06:28, Tamas K Lengyel wrote:
>>
>> diff --git a/xen/include/asm-arm/p2m.h b/xen/include/asm-arm/p2m.h
>> index 08ce07b..b4ca86d 100644
>> --- a/xen/include/asm-arm/p2m.h
>> +++ b/xen/include/asm-arm/p2m.h
>> @@ -2,6 +2,9 @@
>>
>>
>> [..]
>>
>>
>> +#include <public/memory.h>
>> +#include <public/mem_event.h>
>>
>>
>> Why do you need those 2 includes?
>>
>>
>>
>> <public/memory.h> might not be necessary in this patch yet, but it will
>> be required for defining the functions that mem_access will call passing
>> xenmem_access_t access. I can move the header include into the next
>> patch in the series if that is cleaner. The mem_event header is not
>> actually required so I'll remove it.
>>
>
> Yes please.
Ack.
>
>
>
>> #include <xen/p2m-common.h>
>>
>> @@ -11,6 +14,48 @@ struct domain;
>>
>> extern void memory_type_changed(struct domain *);
>>
>> +/* List of possible type for each page in the p2m entry.
>> + * The number of available bit per page in the pte for this
>> purpose is 4 bits.
>> + * So it's possible to only have 16 fields. If we run out of
>> value in the
>> + * future, it's possible to use higher value for pseudo-type
>> and don't store
>> + * them in the p2m entry.
>> + */
>> +typedef enum {
>> + p2m_invalid = 0, /* Nothing mapped here */
>> + p2m_ram_rw, /* Normal read/write guest RAM */
>> + p2m_ram_ro, /* Read-only; writes are silently
>> dropped */
>> + p2m_mmio_direct, /* Read/write mapping of genuine MMIO
>> area */
>> + p2m_map_foreign, /* Ram pages from foreign domain */
>> + p2m_grant_map_rw, /* Read/write grant mapping */
>> + p2m_grant_map_ro, /* Read-only grant mapping */
>> + /* The types below are only used to decide the page
>> attribute in the P2M */
>> + p2m_iommu_map_rw, /* Read/write iommu mapping */
>> + p2m_iommu_map_ro, /* Read-only iommu mapping */
>> + p2m_max_real_type, /* Types after this won't be store in
>> the p2m */
>> +} p2m_type_t;
>>
>>
>> Any reason to move the enum earlier? If not, I would prefer to keep
>> at the same place. It will be easier with git-blame to find when a
>> new type has been added.
>>
>>
>> Stylistically it made more sense to have p2m_type_t and p2m_access_t
>> together (as it is on the x86 as well). And they are defined here as the
>> p2m_domain does have a field now defining default_access with
>> p2m_access_t.
>>
>
> If it's only for the style I wouldn't move them.
>
> Anyway, I'll Ian/Stefano decides about this.
>
Sure.
>
>
>> +/*
>> + * Additional access types, which are used to further restrict
>> + * the permissions given by the p2m_type_t memory type.
>> Violations
>> + * caused by p2m_access_t restrictions are sent to the mem_event
>> + * interface.
>> + *
>> + * The access permissions are soft state: when any ambigious
>> change of page
>>
>>
>> ambiguous.
>>
>>
>> Copy-pasta but will fix.
>>
>>
>> [..]
>>
>> + /* Default P2M access type for each page in the the domain:
>> new pages,
>> + * swapped in pages, cleared pages, and pages that are
>> ambiquously
>>
>>
>> Did you intend to mean ambiguously rather than ambiquously?
>>
>>
>> Copy-pasta again but will fix. Maybe in a separate patch where I fix it
>> here and in the x86 side as well?
>>
>
> I'm OK for a separate patch fixing x86 side, but there is no reason to fix
> the spelling for ARM outside this patch.
>
>
Ack, that's what I ended up doing already.
>
> Regards,
>
> --
> Julien Grall
>
Thanks!
Tamas
[-- Attachment #1.2: Type: text/html, Size: 6135 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events.
2014-09-12 8:46 ` Tamas K Lengyel
@ 2014-09-12 20:35 ` Julien Grall
2014-09-12 20:48 ` Tamas K Lengyel
0 siblings, 1 reply; 48+ messages in thread
From: Julien Grall @ 2014-09-12 20:35 UTC (permalink / raw)
To: Tamas K Lengyel
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
Hello Tamas,
On 12/09/14 01:46, Tamas K Lengyel wrote:
> /* New mapping is superpage aligned, make it */
> pte = mfn_to_p2m_entry(*maddr >> PAGE_SHIFT,
> mattr, t, a);
> if ( level < 3 )
> @@ -663,6 +737,7 @@ static int apply_one_level(struct domain *d,
>
> memset(&pte, 0x00, sizeof(pte));
> p2m_write_pte(entry, pte, flush_cache);
> + radix_tree_delete(&p2m->mem___access_settings,
> paddr_to_pfn(*addr));
>
> *addr += level_size;
> *maddr += level_size;
> @@ -707,6 +782,53 @@ static int apply_one_level(struct domain *d,
> *addr += PAGE_SIZE;
> return P2M_ONE_PROGRESS_NOP;
> }
> +
> + case MEMACCESS:
> + if ( level < 3 )
> + {
> + if ( !p2m_valid(orig_pte) )
> + {
> + (*addr)++;
>
>
> Why increment by 1? You the PTE doesn't contain valid mapping you
> want to skip the whole level range. ie:
>
> *addr += level_size;
>
>
> It doesn't make a difference, apply_p2m_changes is called with
> start=paddr, end=paddr+1 from a separate loop. So just incrementing it
> by one or a whole level achieves the same effect, that is, the
> apply_p2m_changes loop breaks.
Actually it makes a lots of difference. If you increment by 1 the
address, you will call up to level_size time your code before
effectively going to the next level entry.
This function can be called with *multiple page*.
> + return P2M_ONE_PROGRESS_NOP;
> + }
> +
> + /* Shatter large pages as we descend */
> + if ( p2m_mapping(orig_pte) )
> + {
> + rc = p2m_create_table(d, entry,
> + level_shift - PAGE_SHIFT,
> flush_cache);
> + if ( rc < 0 )
> + return rc;
> +
> + p2m->stats.shattered[level]++;
> + p2m->stats.mappings[level]--;
> + p2m->stats.mappings[level+1] += LPAE_ENTRIES;
> + } /* else: an existing table mapping -> descend */
> +
>
>
> This piece of code is exactly the same in INSERT, REMOVE and now
> MEMACCESS. I would create an helper to shatter and update the stats.
>
>
> Ack.
>
>
> + return P2M_ONE_DESCEND;
> + }
> + else
> + {
> + pte = orig_pte;
> +
> + if ( !p2m_table(pte) )
> + pte.bits = 0;
> +
> + if ( p2m_valid(pte) )
> + {
> + ASSERT(pte.p2m.type != p2m_invalid);
> +
> + rc = p2m_mem_access_radix_set(p2m,
> paddr_to_pfn(*addr), a);
> + if ( rc < 0 )
> + return rc;
> +
> + p2m_set_permission(&pte, pte.p2m.type, a);
> + p2m_write_pte(entry, pte, flush_cache);
> + }
> +
> + (*addr)++;
>
>
> *addr += PAGE_SIZE;
>
> [..]
>
>
> +/* Set access type for a region of pfns.
> + * If start_pfn == -1ul, sets the default access type */
> +long p2m_set_mem_access(struct domain *d, unsigned long pfn,
> uint32_t nr,
> + uint32_t start, uint32_t mask,
> xenmem_access_t access)
> +{
> + struct p2m_domain *p2m = p2m_get_hostp2m(d);
> + p2m_access_t a;
> + long rc = 0;
> + paddr_t paddr;
> +
> + static const p2m_access_t memaccess[] = {
> +#define ACCESS(ac) [XENMEM_access_##ac] = p2m_access_##ac
> + ACCESS(n),
> + ACCESS(r),
> + ACCESS(w),
> + ACCESS(rw),
> + ACCESS(x),
> + ACCESS(rx),
> + ACCESS(wx),
> + ACCESS(rwx),
> +#undef ACCESS
> + };
> +
> + switch ( access )
> + {
> + case 0 ... ARRAY_SIZE(memaccess) - 1:
> + a = memaccess[access];
> + break;
> + case XENMEM_access_default:
> + a = p2m->default_access;
> + break;
> + default:
> + return -EINVAL;
> + }
> +
> + /* If request to set default access */
> + if ( pfn == ~0ul )
> + {
> + p2m->default_access = a;
> + return 0;
> + }
> +
> + for ( pfn += start; nr > start; ++pfn )
> + {
> + paddr = pfn_to_paddr(pfn);
> + rc = apply_p2m_changes(d, MEMACCESS, paddr, paddr+1, 0,
> MATTR_MEM, 0, a);
>
>
> Hmmm... why didn't you call directly apply_p2m_changes with the
> whole range?
>
>
> Because the hypercall continuation. Setting mem_access permissions needs
> to be preemptible and it has its own separate routine to do that here.
> See http://xenbits.xen.org/xsa/advisory-89.html for more info.
We do have hypercall continuation in apply_p2m_changes (see for
relinquish). Please do the same for MEMACCESS rather than using your own
loop.
Hence, with your solution, the p2m lookup is taken/released at each
loop. This is inefficient.
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events.
2014-09-12 20:35 ` Julien Grall
@ 2014-09-12 20:48 ` Tamas K Lengyel
2014-09-12 21:04 ` Julien Grall
0 siblings, 1 reply; 48+ messages in thread
From: Tamas K Lengyel @ 2014-09-12 20:48 UTC (permalink / raw)
To: Julien Grall
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
[-- Attachment #1.1: Type: text/plain, Size: 6491 bytes --]
On Fri, Sep 12, 2014 at 10:35 PM, Julien Grall <julien.grall@linaro.org>
wrote:
> Hello Tamas,
>
> On 12/09/14 01:46, Tamas K Lengyel wrote:
>
>> /* New mapping is superpage aligned, make it */
>> pte = mfn_to_p2m_entry(*maddr >> PAGE_SHIFT,
>> mattr, t, a);
>> if ( level < 3 )
>> @@ -663,6 +737,7 @@ static int apply_one_level(struct domain *d,
>>
>> memset(&pte, 0x00, sizeof(pte));
>> p2m_write_pte(entry, pte, flush_cache);
>> + radix_tree_delete(&p2m->mem___access_settings,
>> paddr_to_pfn(*addr));
>>
>> *addr += level_size;
>> *maddr += level_size;
>> @@ -707,6 +782,53 @@ static int apply_one_level(struct domain *d,
>> *addr += PAGE_SIZE;
>> return P2M_ONE_PROGRESS_NOP;
>> }
>> +
>> + case MEMACCESS:
>> + if ( level < 3 )
>> + {
>> + if ( !p2m_valid(orig_pte) )
>> + {
>> + (*addr)++;
>>
>>
>> Why increment by 1? You the PTE doesn't contain valid mapping you
>> want to skip the whole level range. ie:
>>
>> *addr += level_size;
>>
>>
>> It doesn't make a difference, apply_p2m_changes is called with
>> start=paddr, end=paddr+1 from a separate loop. So just incrementing it
>> by one or a whole level achieves the same effect, that is, the
>> apply_p2m_changes loop breaks.
>>
>
> Actually it makes a lots of difference. If you increment by 1 the address,
> you will call up to level_size time your code before effectively going to
> the next level entry.
>
> This function can be called with *multiple page*.
It can be, but it isn't. It makes no difference form my perspective so I'm
fine with it either way.
>
>
> + return P2M_ONE_PROGRESS_NOP;
>> + }
>> +
>> + /* Shatter large pages as we descend */
>> + if ( p2m_mapping(orig_pte) )
>> + {
>> + rc = p2m_create_table(d, entry,
>> + level_shift - PAGE_SHIFT,
>> flush_cache);
>> + if ( rc < 0 )
>> + return rc;
>> +
>> + p2m->stats.shattered[level]++;
>> + p2m->stats.mappings[level]--;
>> + p2m->stats.mappings[level+1] += LPAE_ENTRIES;
>> + } /* else: an existing table mapping -> descend */
>> +
>>
>>
>> This piece of code is exactly the same in INSERT, REMOVE and now
>> MEMACCESS. I would create an helper to shatter and update the stats.
>>
>>
>> Ack.
>>
>>
>> + return P2M_ONE_DESCEND;
>> + }
>> + else
>> + {
>> + pte = orig_pte;
>> +
>> + if ( !p2m_table(pte) )
>> + pte.bits = 0;
>> +
>> + if ( p2m_valid(pte) )
>> + {
>> + ASSERT(pte.p2m.type != p2m_invalid);
>> +
>> + rc = p2m_mem_access_radix_set(p2m,
>> paddr_to_pfn(*addr), a);
>> + if ( rc < 0 )
>> + return rc;
>> +
>> + p2m_set_permission(&pte, pte.p2m.type, a);
>> + p2m_write_pte(entry, pte, flush_cache);
>> + }
>> +
>> + (*addr)++;
>>
>>
>> *addr += PAGE_SIZE;
>>
>> [..]
>>
>>
>> +/* Set access type for a region of pfns.
>> + * If start_pfn == -1ul, sets the default access type */
>> +long p2m_set_mem_access(struct domain *d, unsigned long pfn,
>> uint32_t nr,
>> + uint32_t start, uint32_t mask,
>> xenmem_access_t access)
>> +{
>> + struct p2m_domain *p2m = p2m_get_hostp2m(d);
>> + p2m_access_t a;
>> + long rc = 0;
>> + paddr_t paddr;
>> +
>> + static const p2m_access_t memaccess[] = {
>> +#define ACCESS(ac) [XENMEM_access_##ac] = p2m_access_##ac
>> + ACCESS(n),
>> + ACCESS(r),
>> + ACCESS(w),
>> + ACCESS(rw),
>> + ACCESS(x),
>> + ACCESS(rx),
>> + ACCESS(wx),
>> + ACCESS(rwx),
>> +#undef ACCESS
>> + };
>> +
>> + switch ( access )
>> + {
>> + case 0 ... ARRAY_SIZE(memaccess) - 1:
>> + a = memaccess[access];
>> + break;
>> + case XENMEM_access_default:
>> + a = p2m->default_access;
>> + break;
>> + default:
>> + return -EINVAL;
>> + }
>> +
>> + /* If request to set default access */
>> + if ( pfn == ~0ul )
>> + {
>> + p2m->default_access = a;
>> + return 0;
>> + }
>> +
>> + for ( pfn += start; nr > start; ++pfn )
>> + {
>> + paddr = pfn_to_paddr(pfn);
>> + rc = apply_p2m_changes(d, MEMACCESS, paddr, paddr+1, 0,
>> MATTR_MEM, 0, a);
>>
>>
>> Hmmm... why didn't you call directly apply_p2m_changes with the
>> whole range?
>>
>>
>> Because the hypercall continuation. Setting mem_access permissions needs
>> to be preemptible and it has its own separate routine to do that here.
>> See http://xenbits.xen.org/xsa/advisory-89.html for more info.
>>
>
> We do have hypercall continuation in apply_p2m_changes (see for
> relinquish). Please do the same for MEMACCESS rather than using your own
> loop.
>
> Hence, with your solution, the p2m lookup is taken/released at each loop.
> This is inefficient.
>
>
It is a bit less efficient for sure. However, the continuation setup for
mem_access is in common/mem_access for XENMEM_access_op_set_access. In
order to make use of apply_p2m_change to setup continuation I would have to
remove that code from common and make it arch specific in both ARM and x86.
Is that really a good idea just to have this optimization in ARM?
> Regards,
>
> --
> Julien Grall
>
[-- Attachment #1.2: Type: text/html, Size: 8104 bytes --]
[-- Attachment #2: Type: text/plain, Size: 126 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events.
2014-09-12 20:48 ` Tamas K Lengyel
@ 2014-09-12 21:04 ` Julien Grall
0 siblings, 0 replies; 48+ messages in thread
From: Julien Grall @ 2014-09-12 21:04 UTC (permalink / raw)
To: Tamas K Lengyel
Cc: Ian Campbell, Tim Deegan, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
Hello Tamas,
On 12/09/14 13:48, Tamas K Lengyel wrote:
>
>
> On Fri, Sep 12, 2014 at 10:35 PM, Julien Grall <julien.grall@linaro.org
> <mailto:julien.grall@linaro.org>> wrote:
>
> Hello Tamas,
>
> On 12/09/14 01:46, Tamas K Lengyel wrote:
>
> /* New mapping is superpage aligned,
> make it */
> pte = mfn_to_p2m_entry(*maddr >> PAGE_SHIFT,
> mattr, t, a);
> if ( level < 3 )
> @@ -663,6 +737,7 @@ static int apply_one_level(struct
> domain *d,
>
> memset(&pte, 0x00, sizeof(pte));
> p2m_write_pte(entry, pte, flush_cache);
> + radix_tree_delete(&p2m->mem_____access_settings,
> paddr_to_pfn(*addr));
>
> *addr += level_size;
> *maddr += level_size;
> @@ -707,6 +782,53 @@ static int apply_one_level(struct
> domain *d,
> *addr += PAGE_SIZE;
> return P2M_ONE_PROGRESS_NOP;
> }
> +
> + case MEMACCESS:
> + if ( level < 3 )
> + {
> + if ( !p2m_valid(orig_pte) )
> + {
> + (*addr)++;
>
>
> Why increment by 1? You the PTE doesn't contain valid
> mapping you
> want to skip the whole level range. ie:
>
> *addr += level_size;
>
>
> It doesn't make a difference, apply_p2m_changes is called with
> start=paddr, end=paddr+1 from a separate loop. So just
> incrementing it
> by one or a whole level achieves the same effect, that is, the
> apply_p2m_changes loop breaks.
>
>
> Actually it makes a lots of difference. If you increment by 1 the
> address, you will call up to level_size time your code before
> effectively going to the next level entry.
>
> This function can be called with *multiple page*.
>
>
> It can be, but it isn't. It makes no difference form my perspective so
> I'm fine with it either way.
Even in your case, this code will be called too often, and then make
your code slower. So please fix it, the other part of this function is
skipping the whole level in this case. There is no reason to decide to
not use it here.
> It is a bit less efficient for sure. However, the continuation setup for
> mem_access is in common/mem_access for XENMEM_access_op_set_access. In
> order to make use of apply_p2m_change to setup continuation I would have
> to remove that code from common and make it arch specific in both ARM
> and x86. Is that really a good idea just to have this optimization in ARM?
It's even far less efficient, mainly when you are applying on a batch of
PFN like xenaccess will do.
The common code won't be changed. I'm just asking to move the check in
apply_p2m_changes. It should not be a big deal.
Regards,
--
Julien Grall
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 00/17] Mem_event and mem_access for ARM
2014-09-10 14:01 ` Tamas K Lengyel
@ 2014-09-15 22:26 ` Ian Campbell
2014-09-16 8:00 ` Jan Beulich
0 siblings, 1 reply; 48+ messages in thread
From: Ian Campbell @ 2014-09-15 22:26 UTC (permalink / raw)
To: Tamas K Lengyel
Cc: Tim Deegan, Julien Grall, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Jan Beulich,
Daniel De Graaf, Tamas K Lengyel
On Wed, 2014-09-10 at 16:01 +0200, Tamas K Lengyel wrote:
> So I think up to here the series could go in afaic. Question
> is - does
> this make sense without the ARM parts?
>
> Thanks, Jan
>
>
> Without the ARM parts there is no technical need for it. On the other
> hand, there are many other series pending in the same general area so
> having this part merged would put everyone on the same footing to
> resolve conflicts and to move forward from.
I think it would be perfectly fine/useful for the generic parts to go in
before the ARM part is ready, it's harmless refactoring and it will
reduce the amount of stuff which is being needlessly resent as later
patches get reviewed (since this is quite a long series that's pretty
significant IMHO).
Also A series which moves code around can be a pain to rebase,
especially if other changes are happening in the surrounding code, I
think if it is ready it should go in.
Plus as you say it forms a basis for your's and other's ongoing
development.
Ian.
^ permalink raw reply [flat|nested] 48+ messages in thread
* Re: [PATCH v5 00/17] Mem_event and mem_access for ARM
2014-09-15 22:26 ` Ian Campbell
@ 2014-09-16 8:00 ` Jan Beulich
0 siblings, 0 replies; 48+ messages in thread
From: Jan Beulich @ 2014-09-16 8:00 UTC (permalink / raw)
To: Ian Campbell, Tamas K Lengyel
Cc: Tim Deegan, Julien Grall, Ian Jackson, xen-devel,
Stefano Stabellini, Andres Lagar-Cavilla, Daniel De Graaf,
Tamas K Lengyel
>>> On 16.09.14 at 00:26, <ian.campbell@citrix.com> wrote:
> On Wed, 2014-09-10 at 16:01 +0200, Tamas K Lengyel wrote:
>
>> So I think up to here the series could go in afaic. Question
>> is - does
>> this make sense without the ARM parts?
>>
>> Thanks, Jan
>>
>>
>> Without the ARM parts there is no technical need for it. On the other
>> hand, there are many other series pending in the same general area so
>> having this part merged would put everyone on the same footing to
>> resolve conflicts and to move forward from.
>
> I think it would be perfectly fine/useful for the generic parts to go in
> before the ARM part is ready, it's harmless refactoring and it will
> reduce the amount of stuff which is being needlessly resent as later
> patches get reviewed (since this is quite a long series that's pretty
> significant IMHO).
>
> Also A series which moves code around can be a pain to rebase,
> especially if other changes are happening in the surrounding code, I
> think if it is ready it should go in.
>
> Plus as you say it forms a basis for your's and other's ongoing
> development.
Good - all we'll have to wait for then are Tim's acks on some of the
v6 patches 1-9.
Jan
^ permalink raw reply [flat|nested] 48+ messages in thread
end of thread, other threads:[~2014-09-16 8:00 UTC | newest]
Thread overview: 48+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-09-10 13:28 [PATCH v5 00/17] Mem_event and mem_access for ARM Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 01/17] xen: Relocate mem_access and mem_event into common Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 02/17] xen: Relocate p2m_mem_access_resume to mem_access common Tamas K Lengyel
2014-09-11 20:16 ` Julien Grall
2014-09-12 8:56 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 03/17] xen: Relocate struct npfec definition into common Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 04/17] xen: Relocate mem_event_op domctl and access_op memop " Tamas K Lengyel
2014-09-10 13:44 ` Jan Beulich
2014-09-10 13:28 ` [PATCH v5 05/17] xen/mem_event: Clean out superfluous white-spaces Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 06/17] xen/mem_event: Relax error condition on debug builds Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 07/17] xen/mem_event: Abstract architecture specific sanity checks Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 08/17] xen/mem_access: Abstract architecture specific sanity check Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 09/17] xen/arm: p2m type definitions and changes Tamas K Lengyel
2014-09-11 20:25 ` Julien Grall
2014-09-12 8:15 ` Tamas K Lengyel
2014-09-12 19:23 ` Julien Grall
2014-09-12 20:25 ` Tamas K Lengyel
2014-09-11 20:49 ` Julien Grall
2014-09-12 8:31 ` Tamas K Lengyel
2014-09-12 19:41 ` Julien Grall
2014-09-12 20:20 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 10/17] xen/arm: Add set access required domctl Tamas K Lengyel
2014-09-11 20:26 ` Julien Grall
2014-09-10 13:28 ` [PATCH v5 11/17] xen/arm: Implement domain_get_maximum_gpfn Tamas K Lengyel
2014-09-11 20:28 ` Julien Grall
2014-09-12 8:58 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 12/17] xen/arm: Data abort exception (R/W) mem_events Tamas K Lengyel
2014-09-11 21:19 ` Julien Grall
2014-09-12 8:46 ` Tamas K Lengyel
2014-09-12 20:35 ` Julien Grall
2014-09-12 20:48 ` Tamas K Lengyel
2014-09-12 21:04 ` Julien Grall
2014-09-10 13:28 ` [PATCH v5 13/17] xen/arm: Instruction prefetch abort (X) mem_event handling Tamas K Lengyel
2014-09-11 21:23 ` Julien Grall
2014-09-12 8:34 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 14/17] xen/arm: Enable the compilation of mem_access and mem_event on ARM Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 15/17] xen: Extend getdomaininfo to return the domain's max_gpfn Tamas K Lengyel
2014-09-10 13:48 ` Jan Beulich
2014-09-10 13:55 ` Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 16/17] tools/libxc: Allocate magic page for mem access on ARM Tamas K Lengyel
2014-09-10 13:28 ` [PATCH v5 17/17] tools/tests: Enable xen-access " Tamas K Lengyel
2014-09-11 21:29 ` Julien Grall
2014-09-12 8:50 ` Tamas K Lengyel
2014-09-12 9:01 ` Tamas K Lengyel
2014-09-10 13:51 ` [PATCH v5 00/17] Mem_event and mem_access for ARM Jan Beulich
2014-09-10 14:01 ` Tamas K Lengyel
2014-09-15 22:26 ` Ian Campbell
2014-09-16 8:00 ` Jan Beulich
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.