Xen devel - Mar 2010 - [PATCH 0/7][v5] PV extension of HVM (Hybrid) for Xen

Hi, Jeremy & Ian

Here is the fifth version of patchset to enable PV extension of HVM support
in Linux kernel of Xen.

The PV extension of HVM is started from real mode like HVM guest, but also with
a
a range of PV features(e.g. PV timer, event channel, as well as PV
drivers). So guest with this feature can takes the advantages of both H/W
virtualization and Para-Virtualization.

The first two of the patchset imported several header file from
Jeremy''s tree
and Xen tree, respect to Jeremy and Keir''s works.

The whole patchset based on Linux upstream.

You need a line like:

cpuid = [ ''0x40000002:edx=0x3'' ]

in HVM configuration file to expose hybrid feature to guest, and

CONFIG_XEN

in the guest kernel configuration file to enable the hybrid support.

And the compiled image can be used as native/pv domU/hvm guest/pv feature hvm
kernel.

Current the patchset support x86_64 only.

Current base is still on Linux 2.6.33.

Change from v4:
1. Add a new CONFIG_XEN_HVM_PV to enable the feature in kernel
2. Separate the related code form enlighted.c to hvmpv.c
3. Separate the feature "PV clocksource" from evtchn. Now we can
support HVM
guest with PV clocksource. This would be enabled by default.
4. Drop PV halt and pv drivers in this edition. We can work on that
later.
5. Update the patchset following Jeremy''s comments.

Change from v3:
1. Rebase to Linux 2.6.33 release.
2. change the name to "PV extension of HVM"
3. Some minor coding polishing.

Change from v2:
1. change the name "hybrid" to "PV featured HVM".
2. Unified the PV driver''s judgement of xen_domain() to
xen_evtchn_enabled().
3. Move the function(evtchn) initialize hypercall near the real enabling place,
rather than a unified place before function enabled.
4. Remove the reserved E820 region for grant table. Use QEmu Xen platform
device''s MMIO instead.

The major change from v1:
1. SMP support.
2. Modify the entrance point to avoid most of genernic kernel modification.
3. Binding PV timer with event channel mechanism.

--
regards
Yang, Sheng

 arch/x86/include/asm/xen/cpuid.h      |   73 +++++++++++
 arch/x86/include/asm/xen/hypercall.h  |    6 +
 arch/x86/include/asm/xen/hypervisor.h |    6 +
 arch/x86/kernel/setup.c               |    4 +
 arch/x86/xen/Kconfig                  |    8 ++
 arch/x86/xen/Makefile                 |    1 +
 arch/x86/xen/enlighten.c              |    6 +-
 arch/x86/xen/hvmpv.c                  |  219 +++++++++++++++++++++++++++++++++
 arch/x86/xen/irq.c                    |   28 ++++
 arch/x86/xen/smp.c                    |   92 +++++++++++++-
 arch/x86/xen/time.c                   |    4 +-
 arch/x86/xen/xen-ops.h                |   14 ++
 drivers/block/xen-blkfront.c          |    2 +-
 drivers/input/xen-kbdfront.c          |    2 +-
 drivers/net/xen-netfront.c            |    2 +-
 drivers/video/xen-fbfront.c           |    2 +-
 drivers/xen/events.c                  |   74 ++++++++++-
 drivers/xen/grant-table.c             |    2 +-
 drivers/xen/xenbus/xenbus_probe.c     |    4 +-
 include/xen/events.h                  |    6 +
 include/xen/hvm.h                     |   28 ++++
 include/xen/interface/hvm/hvm_op.h    |   80 ++++++++++++
 include/xen/interface/hvm/params.h    |  111 +++++++++++++++++
 include/xen/interface/xen.h           |    6 +-
 include/xen/xen.h                     |   11 ++
 25 files changed, 768 insertions(+), 23 deletions(-)

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

From: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>

Add support for hvm_op hypercall.

Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Sheng Yang <sheng@linux.intel.com>
---
 arch/x86/include/asm/xen/hypercall.h |    6 ++
 include/xen/hvm.h                    |   23 +++++++
 include/xen/interface/hvm/hvm_op.h   |   72 ++++++++++++++++++++++
 include/xen/interface/hvm/params.h   |  111 ++++++++++++++++++++++++++++++++++
 4 files changed, 212 insertions(+), 0 deletions(-)
 create mode 100644 include/xen/hvm.h
 create mode 100644 include/xen/interface/hvm/hvm_op.h
 create mode 100644 include/xen/interface/hvm/params.h

diff --git a/arch/x86/include/asm/xen/hypercall.h
b/arch/x86/include/asm/xen/hypercall.h
index 9c371e4..47c2ebb 100644
--- a/arch/x86/include/asm/xen/hypercall.h
+++ b/arch/x86/include/asm/xen/hypercall.h
@@ -417,6 +417,12 @@ HYPERVISOR_nmi_op(unsigned long op, unsigned long arg)
 	return _hypercall2(int, nmi_op, op, arg);
 }
 
+static inline unsigned long __must_check
+HYPERVISOR_hvm_op(int op, void *arg)
+{
+	return _hypercall2(unsigned long, hvm_op, op, arg);
+}
+
 static inline void
 MULTI_fpu_taskswitch(struct multicall_entry *mcl, int set)
 {
diff --git a/include/xen/hvm.h b/include/xen/hvm.h
new file mode 100644
index 0000000..4ea8887
--- /dev/null
+++ b/include/xen/hvm.h
@@ -0,0 +1,23 @@
+/* Simple wrappers around HVM functions */
+#ifndef XEN_HVM_H__
+#define XEN_HVM_H__
+
+#include <xen/interface/hvm/params.h>
+
+static inline unsigned long hvm_get_parameter(int idx)
+{
+	struct xen_hvm_param xhv;
+	int r;
+
+	xhv.domid = DOMID_SELF;
+	xhv.index = idx;
+	r = HYPERVISOR_hvm_op(HVMOP_get_param, &xhv);
+	if (r < 0) {
+		printk(KERN_ERR "cannot get hvm parameter %d: %d.\n",
+		       idx, r);
+		return 0;
+	}
+	return xhv.value;
+}
+
+#endif /* XEN_HVM_H__ */
diff --git a/include/xen/interface/hvm/hvm_op.h
b/include/xen/interface/hvm/hvm_op.h
new file mode 100644
index 0000000..7c74ba4
--- /dev/null
+++ b/include/xen/interface/hvm/hvm_op.h
@@ -0,0 +1,72 @@
+/*
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the
"Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef __XEN_PUBLIC_HVM_HVM_OP_H__
+#define __XEN_PUBLIC_HVM_HVM_OP_H__
+
+/* Get/set subcommands: extra argument == pointer to xen_hvm_param struct. */
+#define HVMOP_set_param           0
+#define HVMOP_get_param           1
+struct xen_hvm_param {
+    domid_t  domid;    /* IN */
+    uint32_t index;    /* IN */
+    uint64_t value;    /* IN/OUT */
+};
+DEFINE_GUEST_HANDLE_STRUCT(xen_hvm_param);
+
+/* Set the logical level of one of a domain''s PCI INTx wires. */
+#define HVMOP_set_pci_intx_level  2
+struct xen_hvm_set_pci_intx_level {
+    /* Domain to be updated. */
+    domid_t  domid;
+    /* PCI INTx identification in PCI topology (domain:bus:device:intx). */
+    uint8_t  domain, bus, device, intx;
+    /* Assertion level (0 = unasserted, 1 = asserted). */
+    uint8_t  level;
+};
+DEFINE_GUEST_HANDLE_STRUCT(xen_hvm_set_pci_intx_level);
+
+/* Set the logical level of one of a domain''s ISA IRQ wires. */
+#define HVMOP_set_isa_irq_level   3
+struct xen_hvm_set_isa_irq_level {
+    /* Domain to be updated. */
+    domid_t  domid;
+    /* ISA device identification, by ISA IRQ (0-15). */
+    uint8_t  isa_irq;
+    /* Assertion level (0 = unasserted, 1 = asserted). */
+    uint8_t  level;
+};
+DEFINE_GUEST_HANDLE_STRUCT(xen_hvm_set_isa_irq_level);
+
+#define HVMOP_set_pci_link_route  4
+struct xen_hvm_set_pci_link_route {
+    /* Domain to be updated. */
+    domid_t  domid;
+    /* PCI link identifier (0-3). */
+    uint8_t  link;
+    /* ISA IRQ (1-15), or 0 (disable link). */
+    uint8_t  isa_irq;
+};
+DEFINE_GUEST_HANDLE_STRUCT(xen_hvm_set_pci_link_route);
+
+/* Flushes all VCPU TLBs: @arg must be NULL. */
+#define HVMOP_flush_tlbs          5
+
+#endif /* __XEN_PUBLIC_HVM_HVM_OP_H__ */
diff --git a/include/xen/interface/hvm/params.h
b/include/xen/interface/hvm/params.h
new file mode 100644
index 0000000..15d828f
--- /dev/null
+++ b/include/xen/interface/hvm/params.h
@@ -0,0 +1,111 @@
+/*
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the
"Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef __XEN_PUBLIC_HVM_PARAMS_H__
+#define __XEN_PUBLIC_HVM_PARAMS_H__
+
+#include "hvm_op.h"
+
+/*
+ * Parameter space for HVMOP_{set,get}_param.
+ */
+
+/*
+ * How should CPU0 event-channel notifications be delivered?
+ * val[63:56] == 0: val[55:0] is a delivery GSI (Global System Interrupt).
+ * val[63:56] == 1: val[55:0] is a delivery PCI INTx line, as follows:
+ *                  Domain = val[47:32], Bus  = val[31:16],
+ *                  DevFn  = val[15: 8], IntX = val[ 1: 0]
+ * If val == 0 then CPU0 event-channel notifications are not delivered.
+ */
+#define HVM_PARAM_CALLBACK_IRQ 0
+
+/*
+ * These are not used by Xen. They are here for convenience of HVM-guest
+ * xenbus implementations.
+ */
+#define HVM_PARAM_STORE_PFN    1
+#define HVM_PARAM_STORE_EVTCHN 2
+
+#define HVM_PARAM_PAE_ENABLED  4
+
+#define HVM_PARAM_IOREQ_PFN    5
+
+#define HVM_PARAM_BUFIOREQ_PFN 6
+
+#ifdef __ia64__
+
+#define HVM_PARAM_NVRAM_FD     7
+#define HVM_PARAM_VHPT_SIZE    8
+#define HVM_PARAM_BUFPIOREQ_PFN	9
+
+#elif defined(__i386__) || defined(__x86_64__)
+
+/* Expose Viridian interfaces to this HVM guest? */
+#define HVM_PARAM_VIRIDIAN     9
+
+#endif
+
+/*
+ * Set mode for virtual timers (currently x86 only):
+ *  delay_for_missed_ticks (default):
+ *   Do not advance a vcpu''s time beyond the correct delivery time for
+ *   interrupts that have been missed due to preemption. Deliver missed
+ *   interrupts when the vcpu is rescheduled and advance the vcpu''s
virtual
+ *   time stepwise for each one.
+ *  no_delay_for_missed_ticks:
+ *   As above, missed interrupts are delivered, but guest time always tracks
+ *   wallclock (i.e., real) time while doing so.
+ *  no_missed_ticks_pending:
+ *   No missed interrupts are held pending. Instead, to ensure ticks are
+ *   delivered at some non-zero rate, if we detect missed ticks then the
+ *   internal tick alarm is not disabled if the VCPU is preempted during the
+ *   next tick period.
+ *  one_missed_tick_pending:
+ *   Missed interrupts are collapsed together and delivered as one
''late tick''.
+ *   Guest time always tracks wallclock (i.e., real) time.
+ */
+#define HVM_PARAM_TIMER_MODE   10
+#define HVMPTM_delay_for_missed_ticks    0
+#define HVMPTM_no_delay_for_missed_ticks 1
+#define HVMPTM_no_missed_ticks_pending   2
+#define HVMPTM_one_missed_tick_pending   3
+
+/* Boolean: Enable virtual HPET (high-precision event timer)? (x86-only) */
+#define HVM_PARAM_HPET_ENABLED 11
+
+/* Identity-map page directory used by Intel EPT when CR0.PG=0. */
+#define HVM_PARAM_IDENT_PT     12
+
+/* Device Model domain, defaults to 0. */
+#define HVM_PARAM_DM_DOMAIN    13
+
+/* ACPI S state: currently support S0 and S3 on x86. */
+#define HVM_PARAM_ACPI_S_STATE 14
+
+/* TSS used on Intel when CR0.PE=0. */
+#define HVM_PARAM_VM86_TSS     15
+
+/* Boolean: Enable aligning all periodic vpts to reduce interrupts */
+#define HVM_PARAM_VPT_ALIGN    16
+
+#define HVM_NR_PARAMS          17
+
+#endif /* __XEN_PUBLIC_HVM_PARAMS_H__ */
-- 
1.5.4.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

From: Keir Fraser <keir.fraser@citrix.com>

Which would be used by CPUID detection later

Signed-off-by: Keir Fraser <keir.fraser@citrix.com>
Signed-off-by: Sheng Yang <sheng@linux.intel.com>
---
 arch/x86/include/asm/xen/cpuid.h |   68 ++++++++++++++++++++++++++++++++++++++
 1 files changed, 68 insertions(+), 0 deletions(-)
 create mode 100644 arch/x86/include/asm/xen/cpuid.h

diff --git a/arch/x86/include/asm/xen/cpuid.h b/arch/x86/include/asm/xen/cpuid.h
new file mode 100644
index 0000000..8787f03
--- /dev/null
+++ b/arch/x86/include/asm/xen/cpuid.h
@@ -0,0 +1,68 @@
+/******************************************************************************
+ * arch/include/asm/xen/cpuid.h
+ *
+ * CPUID interface to Xen.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the
"Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Copyright (c) 2007 Citrix Systems, Inc.
+ *
+ * Authors:
+ *    Keir Fraser <keir.fraser@citrix.com>
+ */
+
+#ifndef __ASM_X86_XEN_CPUID_H__
+#define __ASM_X86_XEN_CPUID_H__
+
+/* Xen identification leaves start at 0x40000000. */
+#define XEN_CPUID_FIRST_LEAF 0x40000000
+#define XEN_CPUID_LEAF(i)    (XEN_CPUID_FIRST_LEAF + (i))
+
+/*
+ * Leaf 1 (0x40000000)
+ * EAX: Largest Xen-information leaf. All leaves up to an including @EAX
+ *      are supported by the Xen host.
+ * EBX-EDX: "XenVMMXenVMM" signature, allowing positive
identification
+ *      of a Xen host.
+ */
+#define XEN_CPUID_SIGNATURE_EBX 0x566e6558 /* "XenV" */
+#define XEN_CPUID_SIGNATURE_ECX 0x65584d4d /* "MMXe" */
+#define XEN_CPUID_SIGNATURE_EDX 0x4d4d566e /* "nVMM" */
+
+/*
+ * Leaf 2 (0x40000001)
+ * EAX[31:16]: Xen major version.
+ * EAX[15: 0]: Xen minor version.
+ * EBX-EDX: Reserved (currently all zeroes).
+ */
+
+/*
+ * Leaf 3 (0x40000002)
+ * EAX: Number of hypercall transfer pages. This register is always guaranteed
+ *      to specify one hypercall page.
+ * EBX: Base address of Xen-specific MSRs.
+ * ECX: Features 1. Unused bits are set to zero.
+ * EDX: Features 2. Unused bits are set to zero.
+ */
+
+/* Does the host support MMU_PT_UPDATE_PRESERVE_AD for this guest? */
+#define _XEN_CPUID_FEAT1_MMU_PT_UPDATE_PRESERVE_AD 0
+#define XEN_CPUID_FEAT1_MMU_PT_UPDATE_PRESERVE_AD  (1u<<0)
+
+#endif /* __XEN_PUBLIC_ARCH_X86_CPUID_H__ */
-- 
1.5.4.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Otherwise they would still try to enable with HVM domain type.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
---
 drivers/block/xen-blkfront.c      |    2 +-
 drivers/input/xen-kbdfront.c      |    2 +-
 drivers/net/xen-netfront.c        |    2 +-
 drivers/video/xen-fbfront.c       |    2 +-
 drivers/xen/grant-table.c         |    2 +-
 drivers/xen/xenbus/xenbus_probe.c |    4 ++--
 6 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c
index 05a31e5..32e28bd 100644
--- a/drivers/block/xen-blkfront.c
+++ b/drivers/block/xen-blkfront.c
@@ -1068,7 +1068,7 @@ static struct xenbus_driver blkfront = {
 
 static int __init xlblk_init(void)
 {
-	if (!xen_domain())
+	if (!xen_pv_domain())
 		return -ENODEV;
 
 	if (register_blkdev(XENVBD_MAJOR, DEV_NAME)) {
diff --git a/drivers/input/xen-kbdfront.c b/drivers/input/xen-kbdfront.c
index c721c0a..0c3f4ca 100644
--- a/drivers/input/xen-kbdfront.c
+++ b/drivers/input/xen-kbdfront.c
@@ -338,7 +338,7 @@ static struct xenbus_driver xenkbd_driver = {
 
 static int __init xenkbd_init(void)
 {
-	if (!xen_domain())
+	if (!xen_pv_domain())
 		return -ENODEV;
 
 	/* Nothing to do if running in dom0. */
diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c
index a869b45..e3ae126 100644
--- a/drivers/net/xen-netfront.c
+++ b/drivers/net/xen-netfront.c
@@ -1804,7 +1804,7 @@ static struct xenbus_driver netfront_driver = {
 
 static int __init netif_init(void)
 {
-	if (!xen_domain())
+	if (!xen_pv_domain())
 		return -ENODEV;
 
 	if (xen_initial_domain())
diff --git a/drivers/video/xen-fbfront.c b/drivers/video/xen-fbfront.c
index 603598f..3360470 100644
--- a/drivers/video/xen-fbfront.c
+++ b/drivers/video/xen-fbfront.c
@@ -683,7 +683,7 @@ static struct xenbus_driver xenfb_driver = {
 
 static int __init xenfb_init(void)
 {
-	if (!xen_domain())
+	if (!xen_pv_domain())
 		return -ENODEV;
 
 	/* Nothing to do if running in dom0. */
diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c
index 4c6c0bd..ade7f3d 100644
--- a/drivers/xen/grant-table.c
+++ b/drivers/xen/grant-table.c
@@ -510,7 +510,7 @@ static int __devinit gnttab_init(void)
 	unsigned int max_nr_glist_frames, nr_glist_frames;
 	unsigned int nr_init_grefs;
 
-	if (!xen_domain())
+	if (!xen_pv_domain())
 		return -ENODEV;
 
 	nr_grant_frames = 1;
diff --git a/drivers/xen/xenbus/xenbus_probe.c
b/drivers/xen/xenbus/xenbus_probe.c
index 2f7aaa9..a5712cd 100644
--- a/drivers/xen/xenbus/xenbus_probe.c
+++ b/drivers/xen/xenbus/xenbus_probe.c
@@ -786,7 +786,7 @@ static int __init xenbus_probe_init(void)
 	DPRINTK("");
 
 	err = -ENODEV;
-	if (!xen_domain())
+	if (!xen_pv_domain())
 		goto out_error;
 
 	/* Register ourselves with the kernel bus subsystem */
@@ -922,7 +922,7 @@ static void wait_for_devices(struct xenbus_driver *xendrv)
 	struct device_driver *drv = xendrv ? &xendrv->driver : NULL;
 	unsigned int seconds_waited = 0;
 
-	if (!ready_to_wait_for_devices || !xen_domain())
+	if (!ready_to_wait_for_devices || !xen_pv_domain())
 		return;
 
 	while (exists_connecting_device(drv)) {
-- 
1.5.4.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

The PV extended HVM(once known as Hybrid) is started from real mode like
HVM guest, but also with a component based PV feature selection(e.g. PV halt,
PV timer, event channel, then PV drivers). So guest can takes the advantages
of both H/W virtualization and Para-Virtualization.

This patch introduced the PV extension of HVM guest initialization.

Guest would detect the capability using CPUID 0x40000002.edx, then call
HVMOP_enable_pv hypercall to enable pv support in hypervisor.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: Yaozu (Eddie) Dong <eddie.dong@intel.com>
---
 arch/x86/include/asm/xen/cpuid.h   |    5 ++
 arch/x86/xen/Kconfig               |    8 ++
 arch/x86/xen/Makefile              |    1 +
 arch/x86/xen/hvmpv.c               |  133 ++++++++++++++++++++++++++++++++++++
 include/xen/interface/hvm/hvm_op.h |    8 ++
 include/xen/xen.h                  |    9 +++
 6 files changed, 164 insertions(+), 0 deletions(-)
 create mode 100644 arch/x86/xen/hvmpv.c

diff --git a/arch/x86/include/asm/xen/cpuid.h b/arch/x86/include/asm/xen/cpuid.h
index 8787f03..a93c851 100644
--- a/arch/x86/include/asm/xen/cpuid.h
+++ b/arch/x86/include/asm/xen/cpuid.h
@@ -65,4 +65,9 @@
 #define _XEN_CPUID_FEAT1_MMU_PT_UPDATE_PRESERVE_AD 0
 #define XEN_CPUID_FEAT1_MMU_PT_UPDATE_PRESERVE_AD  (1u<<0)
 
+#define _XEN_CPUID_FEAT2_HVM_PV 0
+#define XEN_CPUID_FEAT2_HVM_PV (1u<<0)
+#define _XEN_CPUID_FEAT2_HVM_PV_EVTCHN 1
+#define XEN_CPUID_FEAT2_HVM_PV_EVTCHN (1u<<1)
+
 #endif /* __XEN_PUBLIC_ARCH_X86_CPUID_H__ */
diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig
index b83e119..af3c7a0 100644
--- a/arch/x86/xen/Kconfig
+++ b/arch/x86/xen/Kconfig
@@ -36,3 +36,11 @@ config XEN_DEBUG_FS
 	help
 	  Enable statistics output and various tuning options in debugfs.
 	  Enabling this option may incur a significant performance overhead.
+
+config XEN_HVM_PV
+	bool "PV extension of HVM"
+	depends on XEN && X86_64
+	default n
+	help
+	  Enable the PV extension of HVM, so that HVM guest would benefit from
+	  PV features like PV timer and evtchn.
diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile
index 3bb4fc2..73bd5db 100644
--- a/arch/x86/xen/Makefile
+++ b/arch/x86/xen/Makefile
@@ -17,4 +17,5 @@ obj-y		:= enlighten.o setup.o multicalls.o mmu.o irq.o \
 obj-$(CONFIG_SMP)		+= smp.o
 obj-$(CONFIG_PARAVIRT_SPINLOCKS)+= spinlock.o
 obj-$(CONFIG_XEN_DEBUG_FS)	+= debugfs.o
+obj-$(CONFIG_XEN_HVM_PV)	+= hvmpv.o
 
diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
new file mode 100644
index 0000000..7a741ed
--- /dev/null
+++ b/arch/x86/xen/hvmpv.c
@@ -0,0 +1,133 @@
+/*
+ * PV extension of HVM implementation.
+ *
+ * Sheng Yang <sheng@linux.intel.com>, Intel Corporation, 2010
+ *
+ */
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/percpu.h>
+#include <linux/module.h>
+
+#include <xen/xen.h>
+#include <xen/features.h>
+#include <xen/events.h>
+#include <xen/hvm.h>
+#include <xen/interface/xen.h>
+#include <xen/interface/version.h>
+#include <xen/interface/memory.h>
+
+#include <asm/xen/cpuid.h>
+#include <asm/xen/hypercall.h>
+#include <asm/xen/hypervisor.h>
+
+#include "xen-ops.h"
+
+u32 xen_hvm_pv_features;
+EXPORT_SYMBOL_GPL(xen_hvm_pv_features);
+
+static const struct pv_info xen_hvm_pv_info __initdata = {
+	.paravirt_enabled = 1,
+	.shared_kernel_pmd = 0,
+	.kernel_rpl = 0,
+	.name = "Xen",
+};
+
+static void __init xen_hvm_pv_banner(void)
+{
+	unsigned version = HYPERVISOR_xen_version(XENVER_version, NULL);
+	struct xen_extraversion extra;
+	HYPERVISOR_xen_version(XENVER_extraversion, &extra);
+
+	printk(KERN_INFO "Booting PV extended HVM kernel on %s\n",
+		pv_info.name);
+	printk(KERN_INFO "Xen version: %d.%d%s\n",
+		version >> 16, version & 0xffff, extra.extraversion);
+}
+
+static int __init xen_para_available(void)
+{
+	uint32_t eax, ebx, ecx, edx;
+	cpuid(XEN_CPUID_LEAF(0), &eax, &ebx, &ecx, &edx);
+
+	if (ebx == XEN_CPUID_SIGNATURE_EBX &&
+	    ecx == XEN_CPUID_SIGNATURE_ECX &&
+	    edx == XEN_CPUID_SIGNATURE_EDX &&
+	    ((eax - XEN_CPUID_LEAF(0)) >= 2))
+		return 1;
+
+	return 0;
+}
+
+static int __init enable_hvm_pv(u64 flags)
+{
+	struct xen_hvm_pv_type a;
+
+	a.domid = DOMID_SELF;
+	a.flags = flags;
+	return HYPERVISOR_hvm_op(HVMOP_enable_pv, &a);
+}
+
+static int __init init_hvm_pv_info(void)
+{
+	uint32_t ecx, edx, pages, msr;
+	u64 pfn;
+
+	if (!xen_para_available())
+		return -EINVAL;
+
+	cpuid(XEN_CPUID_LEAF(2), &pages, &msr, &ecx, &edx);
+
+	/* Check if hvm_pv mode is supported */
+	if (!(edx & XEN_CPUID_FEAT2_HVM_PV))
+		return -ENODEV;
+
+	/* We only support 1 page of hypercall for now */
+	if (pages != 1)
+		return -ENOMEM;
+
+	pfn = __pa(hypercall_page);
+	wrmsrl(msr, pfn);
+
+	xen_setup_features();
+
+	x86_init.oem.banner = xen_hvm_pv_banner;
+	pv_info = xen_hvm_pv_info;
+
+	xen_domain_type = XEN_HVM_DOMAIN;
+
+	return 0;
+}
+
+static struct shared_info shared_info_page __page_aligned_bss;
+
+static void __init init_shared_info(void)
+{
+	struct xen_add_to_physmap xatp;
+
+	xatp.domid = DOMID_SELF;
+	xatp.idx = 0;
+	xatp.space = XENMAPSPACE_shared_info;
+	xatp.gpfn = __pa(&shared_info_page) >> PAGE_SHIFT;
+	if (HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp))
+		BUG();
+
+	HYPERVISOR_shared_info = (struct shared_info *)&shared_info_page;
+
+	per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0];
+}
+
+void __init xen_guest_init(void)
+{
+	int r;
+
+	/* Ensure the we won''t confused with others */
+	if (xen_domain_type != XEN_NATIVE)
+		return;
+
+	r = init_hvm_pv_info();
+	if (r < 0)
+		return;
+
+	init_shared_info();
+}
diff --git a/include/xen/interface/hvm/hvm_op.h
b/include/xen/interface/hvm/hvm_op.h
index 7c74ba4..3f0b118 100644
--- a/include/xen/interface/hvm/hvm_op.h
+++ b/include/xen/interface/hvm/hvm_op.h
@@ -69,4 +69,12 @@ DEFINE_GUEST_HANDLE_STRUCT(xen_hvm_set_pci_link_route);
 /* Flushes all VCPU TLBs: @arg must be NULL. */
 #define HVMOP_flush_tlbs          5
 
+#define HVMOP_enable_pv 9
+struct xen_hvm_pv_type {
+	domid_t domid;
+	uint32_t flags;
+#define HVM_PV_CLOCK  (1ull<<0)
+#define HVM_PV_EVTCHN (1ull<<1)
+};
+
 #endif /* __XEN_PUBLIC_HVM_HVM_OP_H__ */
diff --git a/include/xen/xen.h b/include/xen/xen.h
index a164024..11f5ced 100644
--- a/include/xen/xen.h
+++ b/include/xen/xen.h
@@ -19,6 +19,15 @@ extern enum xen_domain_type xen_domain_type;
 #define xen_hvm_domain()	(xen_domain() &&			\
 				 xen_domain_type == XEN_HVM_DOMAIN)
 
+#ifdef CONFIG_XEN_HVM_PV
+
+#define XEN_HVM_PV_EVTCHN_ENABLED   (1u << 1)
+extern u32 xen_hvm_pv_features;
+
+#define xen_hvm_pv_evtchn_enabled() \
+		(xen_hvm_pv_features & XEN_HVM_PV_EVTCHN_ENABLED)
+#endif /* CONFIG_XEN_HVM_PV */
+
 #ifdef CONFIG_XEN_DOM0
 #include <xen/interface/xen.h>
 #include <asm/xen/hypervisor.h>
-- 
1.5.4.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

xen_guest_init() would setup the environment.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
---
 arch/x86/include/asm/xen/hypervisor.h |    6 ++++++
 arch/x86/kernel/setup.c               |    4 ++++
 2 files changed, 10 insertions(+), 0 deletions(-)

diff --git a/arch/x86/include/asm/xen/hypervisor.h
b/arch/x86/include/asm/xen/hypervisor.h
index 396ff4c..13e089a 100644
--- a/arch/x86/include/asm/xen/hypervisor.h
+++ b/arch/x86/include/asm/xen/hypervisor.h
@@ -37,4 +37,10 @@
 extern struct shared_info *HYPERVISOR_shared_info;
 extern struct start_info *xen_start_info;
 
+#ifdef CONFIG_XEN_HVM_PV
+void __init xen_guest_init(void);
+#else
+static inline void xen_guest_init(void) {};
+#endif
+
 #endif /* _ASM_X86_XEN_HYPERVISOR_H */
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index 5d9e40c..1cc4786 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -113,6 +113,8 @@
 #endif
 #include <asm/mce.h>
 
+#include <asm/xen/hypervisor.h>
+
 /*
  * end_pfn only includes RAM, while max_pfn_mapped includes all e820 entries.
  * The direct mapping extends to max_pfn_mapped, so that we can directly access
@@ -740,6 +742,8 @@ void __init setup_arch(char **cmdline_p)
 
 	x86_init.oem.arch_setup();
 
+	xen_guest_init();
+
 	setup_memory_map();
 	parse_setup_data();
 	/* update the e820_saved too */
-- 
1.5.4.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

And enable it by default in PV extended HVM guest.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
---
 arch/x86/xen/hvmpv.c |   20 ++++++++++++++++++++
 arch/x86/xen/time.c  |    4 +++-
 2 files changed, 23 insertions(+), 1 deletions(-)

diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
index 7a741ed..284e021 100644
--- a/arch/x86/xen/hvmpv.c
+++ b/arch/x86/xen/hvmpv.c
@@ -43,6 +43,7 @@ static void __init xen_hvm_pv_banner(void)
 		pv_info.name);
 	printk(KERN_INFO "Xen version: %d.%d%s\n",
 		version >> 16, version & 0xffff, extra.extraversion);
+	printk(KERN_INFO "PV feature: PV clocksource enabled\n");
 }
 
 static int __init xen_para_available(void)
@@ -117,6 +118,22 @@ static void __init init_shared_info(void)
 	per_cpu(xen_vcpu, 0) = &HYPERVISOR_shared_info->vcpu_info[0];
 }
 
+extern struct clocksource xen_clocksource;
+
+static void __init init_pv_clocksource(void)
+{
+	if (enable_hvm_pv(HVM_PV_CLOCK))
+		BUG();
+
+	pv_time_ops.sched_clock = xen_sched_clock;
+
+	x86_platform.calibrate_tsc = xen_tsc_khz;
+	x86_platform.get_wallclock = xen_get_wallclock;
+	x86_platform.set_wallclock = xen_set_wallclock;
+
+	clocksource_register(&xen_clocksource);
+}
+
 void __init xen_guest_init(void)
 {
 	int r;
@@ -130,4 +147,7 @@ void __init xen_guest_init(void)
 		return;
 
 	init_shared_info();
+
+	/* PV clocksource would be enabled by default */
+	init_pv_clocksource();
 }
diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c
index 0d3f07c..c80eacb 100644
--- a/arch/x86/xen/time.c
+++ b/arch/x86/xen/time.c
@@ -13,6 +13,7 @@
 #include <linux/clockchips.h>
 #include <linux/kernel_stat.h>
 #include <linux/math64.h>
+#include <linux/module.h>
 
 #include <asm/pvclock.h>
 #include <asm/xen/hypervisor.h>
@@ -243,7 +244,7 @@ int xen_set_wallclock(unsigned long now)
 	return -1;
 }
 
-static struct clocksource xen_clocksource __read_mostly = {
+struct clocksource xen_clocksource __read_mostly = {
 	.name = "xen",
 	.rating = 400,
 	.read = xen_clocksource_get_cycles,
@@ -252,6 +253,7 @@ static struct clocksource xen_clocksource __read_mostly = {
 	.shift = XEN_SHIFT,
 	.flags = CLOCK_SOURCE_IS_CONTINUOUS,
 };
+EXPORT_SYMBOL_GPL(xen_clocksource);
 
 /*
    Xen clockevent implementation
-- 
1.5.4.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

We mapped each IOAPIC pin to a VIRQ, so that we can deliver interrupt through
these VIRQs.

We used X86_PLATFORM_IPI_VECTOR as the noficiation vector for hypervisor
to notify guest about the event.

The patch also enabled SMP support, then we can support IPI through evtchn as
well.

Then we don''t use IOAPIC/LAPIC, eliminated the overhead brought by
unnecessary VMExit caused by LAPIC.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
---
 arch/x86/xen/enlighten.c    |    6 +-
 arch/x86/xen/hvmpv.c        |   68 +++++++++++++++++++++++++++++++-
 arch/x86/xen/irq.c          |   28 +++++++++++++
 arch/x86/xen/smp.c          |   92 ++++++++++++++++++++++++++++++++++++++++--
 arch/x86/xen/xen-ops.h      |   14 +++++++
 drivers/xen/events.c        |   74 +++++++++++++++++++++++++++++++---
 include/xen/events.h        |    6 +++
 include/xen/hvm.h           |    5 ++
 include/xen/interface/xen.h |    6 ++-
 include/xen/xen.h           |    2 +
 10 files changed, 285 insertions(+), 16 deletions(-)

diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 36daccb..2d60e70 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -717,7 +717,7 @@ static u32 xen_safe_apic_wait_icr_idle(void)
         return 0;
 }
 
-static void set_xen_basic_apic_ops(void)
+void xen_set_basic_apic_ops(void)
 {
 	apic->read = xen_apic_read;
 	apic->write = xen_apic_write;
@@ -1026,7 +1026,7 @@ static void xen_crash_shutdown(struct pt_regs *regs)
 	xen_reboot(SHUTDOWN_crash);
 }
 
-static const struct machine_ops __initdata xen_machine_ops = {
+const struct machine_ops __initdata xen_machine_ops = {
 	.restart = xen_restart,
 	.halt = xen_machine_halt,
 	.power_off = xen_machine_halt,
@@ -1116,7 +1116,7 @@ asmlinkage void __init xen_start_kernel(void)
 	/*
 	 * set up the basic apic ops.
 	 */
-	set_xen_basic_apic_ops();
+	xen_set_basic_apic_ops();
 #endif
 
 	if (xen_feature(XENFEAT_mmu_pt_update_preserve_ad)) {
diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
index 284e021..233951b 100644
--- a/arch/x86/xen/hvmpv.c
+++ b/arch/x86/xen/hvmpv.c
@@ -17,6 +17,7 @@
 #include <xen/interface/version.h>
 #include <xen/interface/memory.h>
 
+#include <asm/reboot.h>
 #include <asm/xen/cpuid.h>
 #include <asm/xen/hypercall.h>
 #include <asm/xen/hypervisor.h>
@@ -44,6 +45,8 @@ static void __init xen_hvm_pv_banner(void)
 	printk(KERN_INFO "Xen version: %d.%d%s\n",
 		version >> 16, version & 0xffff, extra.extraversion);
 	printk(KERN_INFO "PV feature: PV clocksource enabled\n");
+	if (xen_hvm_pv_evtchn_enabled())
+		printk(KERN_INFO "PV feature: Event channel enabled\n");
 }
 
 static int __init xen_para_available(void)
@@ -83,6 +86,9 @@ static int __init init_hvm_pv_info(void)
 	if (!(edx & XEN_CPUID_FEAT2_HVM_PV))
 		return -ENODEV;
 
+	if (edx & XEN_CPUID_FEAT2_HVM_PV_EVTCHN)
+		xen_hvm_pv_features |= XEN_HVM_PV_EVTCHN_ENABLED;
+
 	/* We only support 1 page of hypercall for now */
 	if (pages != 1)
 		return -ENOMEM;
@@ -131,12 +137,42 @@ static void __init init_pv_clocksource(void)
 	x86_platform.get_wallclock = xen_get_wallclock;
 	x86_platform.set_wallclock = xen_set_wallclock;
 
-	clocksource_register(&xen_clocksource);
+	/* It would be done in xen_time_init() if evtchn enabled */
+	if (!xen_hvm_pv_evtchn_enabled())
+		clocksource_register(&xen_clocksource);
+}
+
+static int set_callback_via(uint64_t via)
+{
+	struct xen_hvm_param a;
+
+	a.domid = DOMID_SELF;
+	a.index = HVM_PARAM_CALLBACK_IRQ;
+	a.value = via;
+	return HYPERVISOR_hvm_op(HVMOP_set_param, &a);
 }
 
+void do_hvm_pv_evtchn_intr(void)
+{
+	per_cpu(irq_count, smp_processor_id())++;
+	xen_hvm_evtchn_do_upcall(get_irq_regs());
+	per_cpu(irq_count, smp_processor_id())--;
+}
+
+#ifdef CONFIG_X86_LOCAL_APIC
+static void xen_hvm_pv_evtchn_apic_write(u32 reg, u32 val)
+{
+	/* The only one reached here should be EOI */
+	WARN_ON(reg != APIC_EOI);
+}
+#endif
+
+extern struct machine_ops xen_machine_ops;
+
 void __init xen_guest_init(void)
 {
 	int r;
+	uint64_t callback_via;
 
 	/* Ensure the we won''t confused with others */
 	if (xen_domain_type != XEN_NATIVE)
@@ -150,4 +186,34 @@ void __init xen_guest_init(void)
 
 	/* PV clocksource would be enabled by default */
 	init_pv_clocksource();
+
+	if (xen_hvm_pv_evtchn_enabled()) {
+		if (enable_hvm_pv(HVM_PV_EVTCHN))
+			return;
+
+		xen_hvm_pv_init_irq_ops();
+
+		x86_init.timers.timer_init = xen_time_init;
+		x86_init.timers.setup_percpu_clockev = x86_init_noop;
+		x86_cpuinit.setup_percpu_clockev = x86_init_noop;
+
+		pv_apic_ops.startup_ipi_hook = paravirt_nop;
+#ifdef CONFIG_X86_LOCAL_APIC
+		/*
+		 * set up the basic apic ops.
+		 */
+		xen_set_basic_apic_ops();
+		apic->write = xen_hvm_pv_evtchn_apic_write;
+#endif
+
+		callback_via = HVM_CALLBACK_VECTOR(X86_PLATFORM_IPI_VECTOR);
+		set_callback_via(callback_via);
+
+		x86_platform_ipi_callback = do_hvm_pv_evtchn_intr;
+
+		disable_acpi();
+
+		xen_hvm_pv_smp_init();
+		machine_ops = xen_machine_ops;
+	}
 }
diff --git a/arch/x86/xen/irq.c b/arch/x86/xen/irq.c
index 9d30105..e325640 100644
--- a/arch/x86/xen/irq.c
+++ b/arch/x86/xen/irq.c
@@ -2,6 +2,7 @@
 
 #include <asm/x86_init.h>
 
+#include <xen/xen.h>
 #include <xen/interface/xen.h>
 #include <xen/interface/sched.h>
 #include <xen/interface/vcpu.h>
@@ -131,3 +132,30 @@ void __init xen_init_irq_ops()
 	pv_irq_ops = xen_irq_ops;
 	x86_init.irqs.intr_init = xen_init_IRQ;
 }
+
+#ifdef CONFIG_XEN_HVM_PV
+static void xen_hvm_pv_evtchn_disable(void)
+{
+	native_irq_disable();
+	xen_irq_disable();
+}
+PV_CALLEE_SAVE_REGS_THUNK(xen_hvm_pv_evtchn_disable);
+
+static void xen_hvm_pv_evtchn_enable(void)
+{
+	native_irq_enable();
+	xen_irq_enable();
+}
+PV_CALLEE_SAVE_REGS_THUNK(xen_hvm_pv_evtchn_enable);
+
+void __init xen_hvm_pv_init_irq_ops(void)
+{
+	if (xen_hvm_pv_evtchn_enabled()) {
+		pv_irq_ops.irq_disable +			PV_CALLEE_SAVE(xen_hvm_pv_evtchn_disable);
+		pv_irq_ops.irq_enable +			PV_CALLEE_SAVE(xen_hvm_pv_evtchn_enable);
+		x86_init.irqs.intr_init = xen_hvm_pv_evtchn_init_IRQ;
+	}
+}
+#endif
diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
index 563d205..a31429e 100644
--- a/arch/x86/xen/smp.c
+++ b/arch/x86/xen/smp.c
@@ -15,18 +15,24 @@
 #include <linux/sched.h>
 #include <linux/err.h>
 #include <linux/smp.h>
+#include <linux/nmi.h>
 
 #include <asm/paravirt.h>
 #include <asm/desc.h>
 #include <asm/pgtable.h>
 #include <asm/cpu.h>
+#include <asm/trampoline.h>
+#include <asm/tlbflush.h>
+#include <asm/mtrr.h>
 
 #include <xen/interface/xen.h>
 #include <xen/interface/vcpu.h>
 
 #include <asm/xen/interface.h>
 #include <asm/xen/hypercall.h>
+#include <asm/xen/hypervisor.h>
 
+#include <xen/xen.h>
 #include <xen/page.h>
 #include <xen/events.h>
 
@@ -63,8 +69,13 @@ static __cpuinit void cpu_bringup(void)
 	touch_softlockup_watchdog();
 	preempt_disable();
 
-	xen_enable_sysenter();
-	xen_enable_syscall();
+	if (xen_pv_domain()) {
+		xen_enable_sysenter();
+		xen_enable_syscall();
+	} else if (xen_hvm_pv_evtchn_enabled())
+		set_mtrr_aps_delayed_init();
+	else
+		BUG();
 
 	cpu = smp_processor_id();
 	smp_store_cpu_info(cpu);
@@ -171,7 +182,8 @@ static void __init xen_smp_prepare_boot_cpu(void)
 
 	/* We''ve switched to the "real" per-cpu gdt, so make sure
the
 	   old memory can be recycled */
-	make_lowmem_page_readwrite(xen_initial_gdt);
+	if (xen_feature(XENFEAT_writable_descriptor_tables))
+		make_lowmem_page_readwrite(xen_initial_gdt);
 
 	xen_setup_vcpu_info_placement();
 }
@@ -282,6 +294,47 @@ cpu_initialize_context(unsigned int cpu, struct task_struct
*idle)
 	return 0;
 }
 
+#ifdef CONFIG_XEN_HVM_PV
+static __cpuinit int
+hvm_pv_cpu_initialize_context(unsigned int cpu, struct task_struct *idle)
+{
+	struct vcpu_guest_context *ctxt;
+	unsigned long start_ip;
+
+	if (cpumask_test_and_set_cpu(cpu, xen_cpu_initialized_map))
+		return 0;
+
+	ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
+	if (ctxt == NULL)
+		return -ENOMEM;
+
+	early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu);
+	initial_code = (unsigned long)cpu_bringup_and_idle;
+	stack_start.sp = (void *) idle->thread.sp;
+
+	/* start_ip had better be page-aligned! */
+	start_ip = setup_trampoline();
+
+	/* only start_ip is what we want */
+	ctxt->flags = VGCF_HVM_GUEST;
+	ctxt->user_regs.eip = start_ip;
+
+	printk(KERN_INFO "Booting processor %d ip 0x%lx\n", cpu, start_ip);
+
+	if (HYPERVISOR_vcpu_op(VCPUOP_initialise, cpu, ctxt))
+		BUG();
+
+	kfree(ctxt);
+	return 0;
+}
+#else
+static int hvm_pv_cpu_initialize_context(unsigned int cpu,
+					 struct task_struct *idle)
+{
+	return 0;
+}
+#endif
+
 static int __cpuinit xen_cpu_up(unsigned int cpu)
 {
 	struct task_struct *idle = idle_task(cpu);
@@ -292,11 +345,17 @@ static int __cpuinit xen_cpu_up(unsigned int cpu)
 	irq_ctx_init(cpu);
 #else
 	clear_tsk_thread_flag(idle, TIF_FORK);
+
+	if (xen_hvm_pv_evtchn_enabled())
+		initial_gs = per_cpu_offset(cpu);
+
 	per_cpu(kernel_stack, cpu)  		(unsigned long)task_stack_page(idle) -
 		KERNEL_STACK_OFFSET + THREAD_SIZE;
 #endif
-	xen_setup_runstate_info(cpu);
+	if (xen_pv_domain())
+		xen_setup_runstate_info(cpu);
+
 	xen_setup_timer(cpu);
 	xen_init_lock_cpu(cpu);
 
@@ -305,7 +364,13 @@ static int __cpuinit xen_cpu_up(unsigned int cpu)
 	/* make sure interrupts start blocked */
 	per_cpu(xen_vcpu, cpu)->evtchn_upcall_mask = 1;
 
-	rc = cpu_initialize_context(cpu, idle);
+	if (xen_pv_domain())
+		rc = cpu_initialize_context(cpu, idle);
+	else if (xen_hvm_pv_evtchn_enabled())
+		rc = hvm_pv_cpu_initialize_context(cpu, idle);
+	else
+		BUG();
+
 	if (rc)
 		return rc;
 
@@ -480,3 +545,20 @@ void __init xen_smp_init(void)
 	xen_fill_possible_map();
 	xen_init_spinlocks();
 }
+
+#ifdef CONFIG_XEN_HVM_PV
+static void xen_hvm_pv_flush_tlb_others(const struct cpumask *cpumask,
+					struct mm_struct *mm, unsigned long va)
+{
+	/* TODO Make it more specific */
+	flush_tlb_all();
+}
+
+void __init xen_hvm_pv_smp_init(void)
+{
+	if (xen_hvm_pv_evtchn_enabled()) {
+		smp_ops = xen_smp_ops;
+		pv_mmu_ops.flush_tlb_others = xen_hvm_pv_flush_tlb_others;
+	}
+}
+#endif
diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h
index f9153a3..6f55815 100644
--- a/arch/x86/xen/xen-ops.h
+++ b/arch/x86/xen/xen-ops.h
@@ -51,6 +51,12 @@ void __init xen_time_init(void);
 unsigned long xen_get_wallclock(void);
 int xen_set_wallclock(unsigned long time);
 unsigned long long xen_sched_clock(void);
+void xen_set_basic_apic_ops(void);
+
+#ifdef CONFIG_XEN_HVM_PV
+void __init xen_hvm_pv_init_irq_ops(void);
+void __init xen_hvm_pv_evtchn_init_IRQ(void);
+#endif /* CONFIG_XEN_HVM_PV */
 
 irqreturn_t xen_debug_interrupt(int irq, void *dev_id);
 
@@ -61,9 +67,17 @@ void xen_setup_vcpu_info_placement(void);
 #ifdef CONFIG_SMP
 void xen_smp_init(void);
 
+#ifdef CONFIG_XEN_HVM_PV
+void xen_hvm_pv_smp_init(void);
+#endif /* CONFIG_XEN_HVM_PV */
+
 extern cpumask_var_t xen_cpu_initialized_map;
 #else
 static inline void xen_smp_init(void) {}
+#ifdef CONFIG_XEN_HVM_PV
+static inline void xen_hvm_pv_smp_init(void) {}
+#endif /* CONFIG_XEN_HVM_PV */
+
 #endif
 
 #ifdef CONFIG_PARAVIRT_SPINLOCKS
diff --git a/drivers/xen/events.c b/drivers/xen/events.c
index ce602dd..e4b9de6 100644
--- a/drivers/xen/events.c
+++ b/drivers/xen/events.c
@@ -32,14 +32,17 @@
 #include <asm/irq.h>
 #include <asm/idle.h>
 #include <asm/sync_bitops.h>
+#include <asm/desc.h>
 #include <asm/xen/hypercall.h>
 #include <asm/xen/hypervisor.h>
 
+#include <xen/xen.h>
 #include <xen/xen-ops.h>
 #include <xen/events.h>
 #include <xen/interface/xen.h>
 #include <xen/interface/event_channel.h>
 
+
 /*
  * This lock protects updates to the following mapping and reference-count
  * arrays. The lock does not need to be acquired to read the mapping tables.
@@ -616,17 +619,13 @@ static DEFINE_PER_CPU(unsigned, xed_nesting_count);
  * a bitset of words which contain pending event bits.  The second
  * level is a bitset of pending events themselves.
  */
-void xen_evtchn_do_upcall(struct pt_regs *regs)
+void __xen_evtchn_do_upcall(struct pt_regs *regs)
 {
 	int cpu = get_cpu();
-	struct pt_regs *old_regs = set_irq_regs(regs);
 	struct shared_info *s = HYPERVISOR_shared_info;
 	struct vcpu_info *vcpu_info = __get_cpu_var(xen_vcpu);
  	unsigned count;
 
-	exit_idle();
-	irq_enter();
-
 	do {
 		unsigned long pending_words;
 
@@ -662,10 +661,25 @@ void xen_evtchn_do_upcall(struct pt_regs *regs)
 	} while(count != 1);
 
 out:
+	put_cpu();
+}
+
+void xen_evtchn_do_upcall(struct pt_regs *regs)
+{
+	struct pt_regs *old_regs = set_irq_regs(regs);
+
+	exit_idle();
+	irq_enter();
+
+	__xen_evtchn_do_upcall(regs);
+
 	irq_exit();
 	set_irq_regs(old_regs);
+}
 
-	put_cpu();
+void xen_hvm_evtchn_do_upcall(struct pt_regs *regs)
+{
+	__xen_evtchn_do_upcall(regs);
 }
 
 /* Rebind a new event channel to an existing irq. */
@@ -944,3 +958,51 @@ void __init xen_init_IRQ(void)
 
 	irq_ctx_init(smp_processor_id());
 }
+
+void __init xen_hvm_pv_evtchn_init_IRQ(void)
+{
+	int i;
+
+	xen_init_IRQ();
+	for (i = 0; i < NR_IRQS_LEGACY; i++) {
+		struct evtchn_bind_virq bind_virq;
+		struct irq_desc *desc = irq_to_desc(i);
+		int virq, evtchn;
+
+		virq = i + VIRQ_EMUL_PIN_START;
+		bind_virq.virq = virq;
+		bind_virq.vcpu = 0;
+
+		if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
+						&bind_virq) != 0)
+			BUG();
+
+		evtchn = bind_virq.port;
+		evtchn_to_irq[evtchn] = i;
+		irq_info[i] = mk_virq_info(evtchn, virq);
+
+		desc->status = IRQ_DISABLED;
+		desc->action = NULL;
+		desc->depth = 1;
+
+		/*
+		 * 16 old-style INTA-cycle interrupts:
+		 */
+		set_irq_chip_and_handler_name(i, &xen_dynamic_chip,
+					handle_level_irq, "event");
+	}
+
+	/*
+	 * Cover the whole vector space, no vector can escape
+	 * us. (some of these will be overridden and become
+	 * ''special'' SMP interrupts)
+	 */
+	for (i = 0; i < (NR_VECTORS - FIRST_EXTERNAL_VECTOR); i++) {
+		int vector = FIRST_EXTERNAL_VECTOR + i;
+		if (vector != IA32_SYSCALL_VECTOR)
+			set_intr_gate(vector, interrupt[i]);
+	}
+
+	/* generic IPI for platform specific use, now used for HVM evtchn */
+	alloc_intr_gate(X86_PLATFORM_IPI_VECTOR, x86_platform_ipi);
+}
diff --git a/include/xen/events.h b/include/xen/events.h
index e68d59a..6b12725 100644
--- a/include/xen/events.h
+++ b/include/xen/events.h
@@ -56,4 +56,10 @@ void xen_poll_irq(int irq);
 /* Determine the IRQ which is bound to an event channel */
 unsigned irq_from_evtchn(unsigned int evtchn);
 
+void xen_evtchn_do_upcall(struct pt_regs *regs);
+
+#ifdef CONFIG_XEN_HVM_PV
+void xen_hvm_evtchn_do_upcall(struct pt_regs *regs);
+#endif
+
 #endif	/* _XEN_EVENTS_H */
diff --git a/include/xen/hvm.h b/include/xen/hvm.h
index 4ea8887..c66d788 100644
--- a/include/xen/hvm.h
+++ b/include/xen/hvm.h
@@ -20,4 +20,9 @@ static inline unsigned long hvm_get_parameter(int idx)
 	return xhv.value;
 }
 
+#define HVM_CALLBACK_VIA_TYPE_VECTOR 0x2
+#define HVM_CALLBACK_VIA_TYPE_SHIFT 56
+#define HVM_CALLBACK_VECTOR(x)
(((uint64_t)HVM_CALLBACK_VIA_TYPE_VECTOR)<<\
+				HVM_CALLBACK_VIA_TYPE_SHIFT | (x))
+
 #endif /* XEN_HVM_H__ */
diff --git a/include/xen/interface/xen.h b/include/xen/interface/xen.h
index 2befa3e..70a6c6e 100644
--- a/include/xen/interface/xen.h
+++ b/include/xen/interface/xen.h
@@ -90,7 +90,11 @@
 #define VIRQ_ARCH_6    22
 #define VIRQ_ARCH_7    23
 
-#define NR_VIRQS       24
+#define VIRQ_EMUL_PIN_START 24
+#define VIRQ_EMUL_PIN_NUM 16
+
+#define NR_VIRQS       (VIRQ_EMUL_PIN_START + VIRQ_EMUL_PIN_NUM)
+
 /*
  * MMU-UPDATE REQUESTS
  *
diff --git a/include/xen/xen.h b/include/xen/xen.h
index 11f5ced..ce2a256 100644
--- a/include/xen/xen.h
+++ b/include/xen/xen.h
@@ -26,6 +26,8 @@ extern u32 xen_hvm_pv_features;
 
 #define xen_hvm_pv_evtchn_enabled() \
 		(xen_hvm_pv_features & XEN_HVM_PV_EVTCHN_ENABLED)
+#else
+#define xen_hvm_pv_evtchn_enabled() 0
 #endif /* CONFIG_XEN_HVM_PV */
 
 #ifdef CONFIG_XEN_DOM0
-- 
1.5.4.5


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 03/04/2010 01:36 AM, Sheng Yang wrote:
> And enable it by default in PV extended HVM guest.
>
> Signed-off-by: Sheng Yang<sheng@linux.intel.com>
> ---
>   arch/x86/xen/hvmpv.c |   20 ++++++++++++++++++++
>   arch/x86/xen/time.c  |    4 +++-
>   2 files changed, 23 insertions(+), 1 deletions(-)
>
> diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
> index 7a741ed..284e021 100644
> --- a/arch/x86/xen/hvmpv.c
> +++ b/arch/x86/xen/hvmpv.c
> @@ -43,6 +43,7 @@ static void __init xen_hvm_pv_banner(void)
>   		pv_info.name);
>   	printk(KERN_INFO "Xen version: %d.%d%s\n",
>   		version>>  16, version&  0xffff, extra.extraversion);
> +	printk(KERN_INFO "PV feature: PV clocksource enabled\n");
>   }
>
>   static int __init xen_para_available(void)
> @@ -117,6 +118,22 @@ static void __init init_shared_info(void)
>   	per_cpu(xen_vcpu, 0) =&HYPERVISOR_shared_info->vcpu_info[0];
>   }
>
> +extern struct clocksource xen_clocksource;
>    
No externs in .c files.  But aside from that, I''d prefer you export a 
function from xen/time.c to do the clocksource registration and call it 
from both places (even if its a one-liner).
> +
> +static void __init init_pv_clocksource(void)
> +{
> +	if (enable_hvm_pv(HVM_PV_CLOCK))
> +		BUG();
>    
BUG is a bit severe.  Will it really never, ever fail?  And if it does, 
the consequence is hardly serious; we just fall back to emulated devices.
> +
> +	pv_time_ops.sched_clock = xen_sched_clock;
> +
> +	x86_platform.calibrate_tsc = xen_tsc_khz;
> +	x86_platform.get_wallclock = xen_get_wallclock;
> +	x86_platform.set_wallclock = xen_set_wallclock;
> +
> +	clocksource_register(&xen_clocksource);
> +}
> +
>   void __init xen_guest_init(void)
>   {
>   	int r;
> @@ -130,4 +147,7 @@ void __init xen_guest_init(void)
>   		return;
>
>   	init_shared_info();
> +
> +	/* PV clocksource would be enabled by default */
> +	init_pv_clocksource();
>   }
> diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c
> index 0d3f07c..c80eacb 100644
> --- a/arch/x86/xen/time.c
> +++ b/arch/x86/xen/time.c
> @@ -13,6 +13,7 @@
>   #include<linux/clockchips.h>
>   #include<linux/kernel_stat.h>
>   #include<linux/math64.h>
> +#include<linux/module.h>
>
>   #include<asm/pvclock.h>
>   #include<asm/xen/hypervisor.h>
> @@ -243,7 +244,7 @@ int xen_set_wallclock(unsigned long now)
>   	return -1;
>   }
>
> -static struct clocksource xen_clocksource __read_mostly = {
> +struct clocksource xen_clocksource __read_mostly = {
>   	.name = "xen",
>   	.rating = 400,
>   	.read = xen_clocksource_get_cycles,
> @@ -252,6 +253,7 @@ static struct clocksource xen_clocksource __read_mostly
= {
>   	.shift = XEN_SHIFT,
>   	.flags = CLOCK_SOURCE_IS_CONTINUOUS,
>   };
> +EXPORT_SYMBOL_GPL(xen_clocksource);
>
>   /*
>      Xen clockevent implementation
>    

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 03/04/2010 01:36 AM, Sheng Yang wrote:
> diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig
> index b83e119..af3c7a0 100644
> --- a/arch/x86/xen/Kconfig
> +++ b/arch/x86/xen/Kconfig
> @@ -36,3 +36,11 @@ config XEN_DEBUG_FS
>   	help
>   	  Enable statistics output and various tuning options in debugfs.
>   	  Enabling this option may incur a significant performance overhead.
> +
> +config XEN_HVM_PV
> +	bool "PV extension of HVM"
> +	depends on XEN&&  X86_64
> +	default n
> +	help
> +	  Enable the PV extension of HVM, so that HVM guest would benefit from
> +	  PV features like PV timer and evtchn.
>    
Unless there''s a strong reason to make this a user-visible setting,
I''d
make it a hidden option which is enabled by default when the 
dependencies are correct.
> diff --git a/arch/x86/xen/Makefile b/arch/x86/xen/Makefile
> index 3bb4fc2..73bd5db 100644
> --- a/arch/x86/xen/Makefile
> +++ b/arch/x86/xen/Makefile
> @@ -17,4 +17,5 @@ obj-y		:= enlighten.o setup.o multicalls.o mmu.o irq.o \
>   obj-$(CONFIG_SMP)		+= smp.o
>   obj-$(CONFIG_PARAVIRT_SPINLOCKS)+= spinlock.o
>   obj-$(CONFIG_XEN_DEBUG_FS)	+= debugfs.o
> +obj-$(CONFIG_XEN_HVM_PV)	+= hvmpv.o
>
> diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
> new file mode 100644
> index 0000000..7a741ed
> --- /dev/null
> +++ b/arch/x86/xen/hvmpv.c
> @@ -0,0 +1,133 @@
> +/*
> + * PV extension of HVM implementation.
> + *
> + * Sheng Yang<sheng@linux.intel.com>, Intel Corporation, 2010
> + *
> + */
> +#include<linux/kernel.h>
> +#include<linux/init.h>
> +#include<linux/percpu.h>
> +#include<linux/module.h>
> +
> +#include<xen/xen.h>
> +#include<xen/features.h>
> +#include<xen/events.h>
> +#include<xen/hvm.h>
> +#include<xen/interface/xen.h>
> +#include<xen/interface/version.h>
> +#include<xen/interface/memory.h>
> +
> +#include<asm/xen/cpuid.h>
> +#include<asm/xen/hypercall.h>
> +#include<asm/xen/hypervisor.h>
> +
> +#include "xen-ops.h"
> +
> +u32 xen_hvm_pv_features;
> +EXPORT_SYMBOL_GPL(xen_hvm_pv_features);
> +
> +static const struct pv_info xen_hvm_pv_info __initdata = {
> +	.paravirt_enabled = 1,
> +	.shared_kernel_pmd = 0,
> +	.kernel_rpl = 0,
> +	.name = "Xen",
> +};
> +
> +static void __init xen_hvm_pv_banner(void)
> +{
> +	unsigned version = HYPERVISOR_xen_version(XENVER_version, NULL);
> +	struct xen_extraversion extra;
> +	HYPERVISOR_xen_version(XENVER_extraversion,&extra);
> +
> +	printk(KERN_INFO "Booting PV extended HVM kernel on %s\n",
> +		pv_info.name);
> +	printk(KERN_INFO "Xen version: %d.%d%s\n",
> +		version>>  16, version&  0xffff, extra.extraversion);
> +}
> +
> +static int __init xen_para_available(void)
> +{
> +	uint32_t eax, ebx, ecx, edx;
> +	cpuid(XEN_CPUID_LEAF(0),&eax,&ebx,&ecx,&edx);
> +
> +	if (ebx == XEN_CPUID_SIGNATURE_EBX&&
> +	    ecx == XEN_CPUID_SIGNATURE_ECX&&
> +	    edx == XEN_CPUID_SIGNATURE_EDX&&
> +	    ((eax - XEN_CPUID_LEAF(0))>= 2))
> +		return 1;
> +
> +	return 0;
> +}
> +
> +static int __init enable_hvm_pv(u64 flags)
> +{
> +	struct xen_hvm_pv_type a;
> +
> +	a.domid = DOMID_SELF;
> +	a.flags = flags;
> +	return HYPERVISOR_hvm_op(HVMOP_enable_pv,&a);
> +}
> +
> +static int __init init_hvm_pv_info(void)
> +{
> +	uint32_t ecx, edx, pages, msr;
> +	u64 pfn;
> +
> +	if (!xen_para_available())
> +		return -EINVAL;
> +
> +	cpuid(XEN_CPUID_LEAF(2),&pages,&msr,&ecx,&edx);
> +
> +	/* Check if hvm_pv mode is supported */
> +	if (!(edx&  XEN_CPUID_FEAT2_HVM_PV))
> +		return -ENODEV;
> +
> +	/* We only support 1 page of hypercall for now */
> +	if (pages != 1)
> +		return -ENOMEM;
>    
Why does it matter?  If you''re only using hypercalls in the first page,
it doesn''t matter whether Xen exports more pages beyond that.
> +
> +	pfn = __pa(hypercall_page);
> +	wrmsrl(msr, pfn);
>    
wrmsr_safe() with error check, just in case it fails?
> +
> +	xen_setup_features();
> +
> +	x86_init.oem.banner = xen_hvm_pv_banner;
> +	pv_info = xen_hvm_pv_info;
> +
> +	xen_domain_type = XEN_HVM_DOMAIN;
>    
Set this very last, once everything else has successfully set up.
> +
> +	return 0;
> +}
> +
> +static struct shared_info shared_info_page __page_aligned_bss;
> +
> +static void __init init_shared_info(void)
> +{
> +	struct xen_add_to_physmap xatp;
> +
> +	xatp.domid = DOMID_SELF;
> +	xatp.idx = 0;
> +	xatp.space = XENMAPSPACE_shared_info;
> +	xatp.gpfn = __pa(&shared_info_page)>>  PAGE_SHIFT;
> +	if (HYPERVISOR_memory_op(XENMEM_add_to_physmap,&xatp))
> +		BUG();
>    
It isn''t too late to back out if it fails; just return an error
code.
> +
> +	HYPERVISOR_shared_info = (struct shared_info *)&shared_info_page;
> +
> +	per_cpu(xen_vcpu, 0) =&HYPERVISOR_shared_info->vcpu_info[0];
> +}
> +
> +void __init xen_guest_init(void)
> +{
> +	int r;
> +
> +	/* Ensure the we won''t confused with others */
> +	if (xen_domain_type != XEN_NATIVE)
>    
Just use xen_domain()?
> +		return;
> +
> +	r = init_hvm_pv_info();
> +	if (r<  0)
> +		return;
> +
> +	init_shared_info();
> +}
> diff --git a/include/xen/interface/hvm/hvm_op.h
b/include/xen/interface/hvm/hvm_op.h
> index 7c74ba4..3f0b118 100644
> --- a/include/xen/interface/hvm/hvm_op.h
> +++ b/include/xen/interface/hvm/hvm_op.h
> @@ -69,4 +69,12 @@ DEFINE_GUEST_HANDLE_STRUCT(xen_hvm_set_pci_link_route);
>   /* Flushes all VCPU TLBs: @arg must be NULL. */
>   #define HVMOP_flush_tlbs          5
>
> +#define HVMOP_enable_pv 9
> +struct xen_hvm_pv_type {
> +	domid_t domid;
> +	uint32_t flags;
> +#define HVM_PV_CLOCK  (1ull<<0)
> +#define HVM_PV_EVTCHN (1ull<<1)
> +};
> +
>   #endif /* __XEN_PUBLIC_HVM_HVM_OP_H__ */
> diff --git a/include/xen/xen.h b/include/xen/xen.h
> index a164024..11f5ced 100644
> --- a/include/xen/xen.h
> +++ b/include/xen/xen.h
> @@ -19,6 +19,15 @@ extern enum xen_domain_type xen_domain_type;
>   #define xen_hvm_domain()	(xen_domain()&&			\
>   				 xen_domain_type == XEN_HVM_DOMAIN)
>
> +#ifdef CONFIG_XEN_HVM_PV
> +
> +#define XEN_HVM_PV_EVTCHN_ENABLED   (1u<<  1)
> +extern u32 xen_hvm_pv_features;
> +
> +#define xen_hvm_pv_evtchn_enabled() \
> +		(xen_hvm_pv_features&  XEN_HVM_PV_EVTCHN_ENABLED)
> +#endif /* CONFIG_XEN_HVM_PV */
>    
Did you compile test this with !CONFIG_XEN_HVM_PV?  I think it would 
fail with xen_hvm_pv_evtchn_enabled() undefined.  Fortunately making 
xen_hvm_pv_evtchn_enabled() evaluate to constant 0 in the !configed case 
will clean things up.

     J

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 03/04/2010 01:36 AM, Sheng Yang wrote:
> We mapped each IOAPIC pin to a VIRQ, so that we can deliver interrupt
through
> these VIRQs.
>
> We used X86_PLATFORM_IPI_VECTOR as the noficiation vector for hypervisor
>    
"notification"
> to notify guest about the event.
>
> The patch also enabled SMP support, then we can support IPI through evtchn
as well.
>
> Then we don''t use IOAPIC/LAPIC, eliminated the overhead brought by
> unnecessary VMExit caused by LAPIC.
>    
I think you need to expand this comment to explain the consequences of 
this change on SMP bringup and timers.  I still don''t see why there are
(or should be) any differences from PV cpu bringup, aside from the tiny 
details of syscall setup (and MTRR I guess, but that''s all a noop 
anyway, right?).

Also, I think you can probably do the timer changes in a separate patch, 
as even with evtchns, an emulated timer device interrupt will still be 
delivered properly via VIRQ_EMUL_PIN_*.
> Signed-off-by: Sheng Yang<sheng@linux.intel.com>
> ---
>   arch/x86/xen/enlighten.c    |    6 +-
>   arch/x86/xen/hvmpv.c        |   68 +++++++++++++++++++++++++++++++-
>   arch/x86/xen/irq.c          |   28 +++++++++++++
>   arch/x86/xen/smp.c          |   92
++++++++++++++++++++++++++++++++++++++++--
>   arch/x86/xen/xen-ops.h      |   14 +++++++
>   drivers/xen/events.c        |   74 +++++++++++++++++++++++++++++++---
>   include/xen/events.h        |    6 +++
>   include/xen/hvm.h           |    5 ++
>   include/xen/interface/xen.h |    6 ++-
>   include/xen/xen.h           |    2 +
>   10 files changed, 285 insertions(+), 16 deletions(-)
>
> diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
> index 36daccb..2d60e70 100644
> --- a/arch/x86/xen/enlighten.c
> +++ b/arch/x86/xen/enlighten.c
> @@ -717,7 +717,7 @@ static u32 xen_safe_apic_wait_icr_idle(void)
>           return 0;
>   }
>
> -static void set_xen_basic_apic_ops(void)
> +void xen_set_basic_apic_ops(void)
>   {
>          apic->read = xen_apic_read;
>          apic->write = xen_apic_write;
> @@ -1026,7 +1026,7 @@ static void xen_crash_shutdown(struct pt_regs *regs)
>          xen_reboot(SHUTDOWN_crash);
>   }
>
> -static const struct machine_ops __initdata xen_machine_ops = {
> +const struct machine_ops __initdata xen_machine_ops = {
>          .restart = xen_restart,
>          .halt = xen_machine_halt,
>          .power_off = xen_machine_halt,
> @@ -1116,7 +1116,7 @@ asmlinkage void __init xen_start_kernel(void)
>          /*
>           * set up the basic apic ops.
>           */
> -       set_xen_basic_apic_ops();
> +       xen_set_basic_apic_ops();
>   #endif
>
>          if (xen_feature(XENFEAT_mmu_pt_update_preserve_ad)) {
> diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
> index 284e021..233951b 100644
> --- a/arch/x86/xen/hvmpv.c
> +++ b/arch/x86/xen/hvmpv.c
> @@ -17,6 +17,7 @@
>   #include<xen/interface/version.h>
>   #include<xen/interface/memory.h>
>
> +#include<asm/reboot.h>
>   #include<asm/xen/cpuid.h>
>   #include<asm/xen/hypercall.h>
>   #include<asm/xen/hypervisor.h>
> @@ -44,6 +45,8 @@ static void __init xen_hvm_pv_banner(void)
>          printk(KERN_INFO "Xen version: %d.%d%s\n",
>                  version>>  16, version&  0xffff,
extra.extraversion);
>          printk(KERN_INFO "PV feature: PV clocksource
enabled\n");
> +       if (xen_hvm_pv_evtchn_enabled())
> +               printk(KERN_INFO "PV feature: Event channel
enabled\n");
>   }
>
>   static int __init xen_para_available(void)
> @@ -83,6 +86,9 @@ static int __init init_hvm_pv_info(void)
>          if (!(edx&  XEN_CPUID_FEAT2_HVM_PV))
>                  return -ENODEV;
>
> +       if (edx&  XEN_CPUID_FEAT2_HVM_PV_EVTCHN)
> +               xen_hvm_pv_features |= XEN_HVM_PV_EVTCHN_ENABLED;
> +
>          /* We only support 1 page of hypercall for now */
>          if (pages != 1)
>                  return -ENOMEM;
> @@ -131,12 +137,42 @@ static void __init init_pv_clocksource(void)
>          x86_platform.get_wallclock = xen_get_wallclock;
>          x86_platform.set_wallclock = xen_set_wallclock;
>
> -       clocksource_register(&xen_clocksource);
> +       /* It would be done in xen_time_init() if evtchn enabled */
> +       if (!xen_hvm_pv_evtchn_enabled())
> +               clocksource_register(&xen_clocksource);
> +}
> +
> +static int set_callback_via(uint64_t via)
> +{
> +       struct xen_hvm_param a;
> +
> +       a.domid = DOMID_SELF;
> +       a.index = HVM_PARAM_CALLBACK_IRQ;
> +       a.value = via;
> +       return HYPERVISOR_hvm_op(HVMOP_set_param,&a);
>   }
>
> +void do_hvm_pv_evtchn_intr(void)
> +{
> +       per_cpu(irq_count, smp_processor_id())++;
> +       xen_hvm_evtchn_do_upcall(get_irq_regs());
> +       per_cpu(irq_count, smp_processor_id())--;
> +}
> +
> +#ifdef CONFIG_X86_LOCAL_APIC
> +static void xen_hvm_pv_evtchn_apic_write(u32 reg, u32 val)
> +{
> +       /* The only one reached here should be EOI */
> +       WARN_ON(reg != APIC_EOI);
> +}
> +#endif
> +
> +extern struct machine_ops xen_machine_ops;
> +
>   void __init xen_guest_init(void)
>   {
>          int r;
> +       uint64_t callback_via;
>
>          /* Ensure the we won''t confused with others */
>          if (xen_domain_type != XEN_NATIVE)
> @@ -150,4 +186,34 @@ void __init xen_guest_init(void)
>
>          /* PV clocksource would be enabled by default */
>          init_pv_clocksource();
> +
> +       if (xen_hvm_pv_evtchn_enabled()) {
> +               if (enable_hvm_pv(HVM_PV_EVTCHN))
> +                       return;
> +
> +               xen_hvm_pv_init_irq_ops();
> +
> +               x86_init.timers.timer_init = xen_time_init;
> +               x86_init.timers.setup_percpu_clockev = x86_init_noop;
> +               x86_cpuinit.setup_percpu_clockev = x86_init_noop;
> +
> +               pv_apic_ops.startup_ipi_hook = paravirt_nop;
> +#ifdef CONFIG_X86_LOCAL_APIC
> +               /*
> +                * set up the basic apic ops.
> +                */
> +               xen_set_basic_apic_ops();
> +               apic->write = xen_hvm_pv_evtchn_apic_write;
> +#endif
> +
> +               callback_via =
HVM_CALLBACK_VECTOR(X86_PLATFORM_IPI_VECTOR);
> +               set_callback_via(callback_via);
> +
> +               x86_platform_ipi_callback = do_hvm_pv_evtchn_intr;
> +
> +               disable_acpi();
> +
> +               xen_hvm_pv_smp_init();
> +               machine_ops = xen_machine_ops;
> +       }
>   }
> diff --git a/arch/x86/xen/irq.c b/arch/x86/xen/irq.c
> index 9d30105..e325640 100644
> --- a/arch/x86/xen/irq.c
> +++ b/arch/x86/xen/irq.c
> @@ -2,6 +2,7 @@
>
>   #include<asm/x86_init.h>
>
> +#include<xen/xen.h>
>   #include<xen/interface/xen.h>
>   #include<xen/interface/sched.h>
>   #include<xen/interface/vcpu.h>
> @@ -131,3 +132,30 @@ void __init xen_init_irq_ops()
>          pv_irq_ops = xen_irq_ops;
>          x86_init.irqs.intr_init = xen_init_IRQ;
>   }
> +
> +#ifdef CONFIG_XEN_HVM_PV
> +static void xen_hvm_pv_evtchn_disable(void)
> +{
> +       native_irq_disable();
> +       xen_irq_disable();
> +}
> +PV_CALLEE_SAVE_REGS_THUNK(xen_hvm_pv_evtchn_disable);
> +
> +static void xen_hvm_pv_evtchn_enable(void)
> +{
> +       native_irq_enable();
> +       xen_irq_enable();
> +}
> +PV_CALLEE_SAVE_REGS_THUNK(xen_hvm_pv_evtchn_enable);
> +
> +void __init xen_hvm_pv_init_irq_ops(void)
> +{
> +       if (xen_hvm_pv_evtchn_enabled()) {
> +               pv_irq_ops.irq_disable > +                      
PV_CALLEE_SAVE(xen_hvm_pv_evtchn_disable);
> +               pv_irq_ops.irq_enable > +                      
PV_CALLEE_SAVE(xen_hvm_pv_evtchn_enable);
> +               x86_init.irqs.intr_init = xen_hvm_pv_evtchn_init_IRQ;
> +       }
> +}
> +#endif
> diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
> index 563d205..a31429e 100644
> --- a/arch/x86/xen/smp.c
> +++ b/arch/x86/xen/smp.c
> @@ -15,18 +15,24 @@
>   #include<linux/sched.h>
>   #include<linux/err.h>
>   #include<linux/smp.h>
> +#include<linux/nmi.h>
>
>   #include<asm/paravirt.h>
>   #include<asm/desc.h>
>   #include<asm/pgtable.h>
>   #include<asm/cpu.h>
> +#include<asm/trampoline.h>
> +#include<asm/tlbflush.h>
> +#include<asm/mtrr.h>
>
>   #include<xen/interface/xen.h>
>   #include<xen/interface/vcpu.h>
>
>   #include<asm/xen/interface.h>
>   #include<asm/xen/hypercall.h>
> +#include<asm/xen/hypervisor.h>
>
> +#include<xen/xen.h>
>   #include<xen/page.h>
>   #include<xen/events.h>
>
> @@ -63,8 +69,13 @@ static __cpuinit void cpu_bringup(void)
>          touch_softlockup_watchdog();
>          preempt_disable();
>
> -       xen_enable_sysenter();
> -       xen_enable_syscall();
> +       if (xen_pv_domain()) {
> +               xen_enable_sysenter();
> +               xen_enable_syscall();
> +       } else if (xen_hvm_pv_evtchn_enabled())
> +               set_mtrr_aps_delayed_init();
> +       else
> +               BUG();
>
>          cpu = smp_processor_id();
>          smp_store_cpu_info(cpu);
> @@ -171,7 +182,8 @@ static void __init xen_smp_prepare_boot_cpu(void)
>
>          /* We''ve switched to the "real" per-cpu gdt, so
make sure the
>             old memory can be recycled */
> -       make_lowmem_page_readwrite(xen_initial_gdt);
> +       if (xen_feature(XENFEAT_writable_descriptor_tables))
> +               make_lowmem_page_readwrite(xen_initial_gdt);
>
>          xen_setup_vcpu_info_placement();
>   }
> @@ -282,6 +294,47 @@ cpu_initialize_context(unsigned int cpu, struct
task_struct *idle)
>          return 0;
>   }
>
> +#ifdef CONFIG_XEN_HVM_PV
>    
Rather than #ifdeffing...
> +static __cpuinit int
> +hvm_pv_cpu_initialize_context(unsigned int cpu, struct task_struct *idle)
> +{
> +       struct vcpu_guest_context *ctxt;
> +       unsigned long start_ip;
>    
Just put

	if (!xen_hvm_pv_evtchn_enabled())
		return -EOPNOTSUPP;	/* or something */

and make sure xen_hvm_pv_evtchn_enable() returns constant 0 in the 
!CONFIG_XEN_HVM_PV case.
> +
> +       if (cpumask_test_and_set_cpu(cpu, xen_cpu_initialized_map))
> +               return 0;
> +
> +       ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
> +       if (ctxt == NULL)
> +               return -ENOMEM;
> +
> +       early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu);
> +       initial_code = (unsigned long)cpu_bringup_and_idle;
> +       stack_start.sp = (void *) idle->thread.sp;
> +
> +       /* start_ip had better be page-aligned! */
> +       start_ip = setup_trampoline();
>    
Why go via trampoline?
> +
> +       /* only start_ip is what we want */
> +       ctxt->flags = VGCF_HVM_GUEST;
> +       ctxt->user_regs.eip = start_ip;
> +
> +       printk(KERN_INFO "Booting processor %d ip 0x%lx\n", cpu,
start_ip);
> +
> +       if (HYPERVISOR_vcpu_op(VCPUOP_initialise, cpu, ctxt))
> +               BUG();
> +
> +       kfree(ctxt);
> +       return 0;
> +}
> +#else
> +static int hvm_pv_cpu_initialize_context(unsigned int cpu,
> +                                        struct task_struct *idle)
> +{
> +       return 0;
> +}
> +#endif
> +
>   static int __cpuinit xen_cpu_up(unsigned int cpu)
>   {
>          struct task_struct *idle = idle_task(cpu);
> @@ -292,11 +345,17 @@ static int __cpuinit xen_cpu_up(unsigned int cpu)
>          irq_ctx_init(cpu);
>   #else
>          clear_tsk_thread_flag(idle, TIF_FORK);
> +
> +       if (xen_hvm_pv_evtchn_enabled())
> +               initial_gs = per_cpu_offset(cpu);
>    
Why does this need to be conditional?
> +
>          per_cpu(kernel_stack, cpu) >                  (unsigned
long)task_stack_page(idle) -
>                  KERNEL_STACK_OFFSET + THREAD_SIZE;
>   #endif
> -       xen_setup_runstate_info(cpu);
> +       if (xen_pv_domain())
> +               xen_setup_runstate_info(cpu);
> +
>    
Does HVM not support this?  If VCPUOP_register_runstate_memory_area 
doesn''t work on HVM, then just change xen_setup_runstate_info() to not 
BUG on hypercall failure (and presumably nothing tries to use 
runstate_info).
>          xen_setup_timer(cpu);
>          xen_init_lock_cpu(cpu);
>
> @@ -305,7 +364,13 @@ static int __cpuinit xen_cpu_up(unsigned int cpu)
>          /* make sure interrupts start blocked */
>          per_cpu(xen_vcpu, cpu)->evtchn_upcall_mask = 1;
>
> -       rc = cpu_initialize_context(cpu, idle);
> +       if (xen_pv_domain())
> +               rc = cpu_initialize_context(cpu, idle);
> +       else if (xen_hvm_pv_evtchn_enabled())
> +               rc = hvm_pv_cpu_initialize_context(cpu, idle);
> +       else
> +               BUG();
> +
>          if (rc)
>                  return rc;
>
> @@ -480,3 +545,20 @@ void __init xen_smp_init(void)
>          xen_fill_possible_map();
>          xen_init_spinlocks();
>   }
> +
> +#ifdef CONFIG_XEN_HVM_PV
> +static void xen_hvm_pv_flush_tlb_others(const struct cpumask *cpumask,
> +                                       struct mm_struct *mm, unsigned long
va)
> +{
> +       /* TODO Make it more specific */
> +       flush_tlb_all();
> +}
> +
> +void __init xen_hvm_pv_smp_init(void)
> +{
> +       if (xen_hvm_pv_evtchn_enabled()) {
> +               smp_ops = xen_smp_ops;
> +               pv_mmu_ops.flush_tlb_others = xen_hvm_pv_flush_tlb_others;
> +       }
> +}
> +#endif
> diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h
> index f9153a3..6f55815 100644
> --- a/arch/x86/xen/xen-ops.h
> +++ b/arch/x86/xen/xen-ops.h
> @@ -51,6 +51,12 @@ void __init xen_time_init(void);
>   unsigned long xen_get_wallclock(void);
>   int xen_set_wallclock(unsigned long time);
>   unsigned long long xen_sched_clock(void);
> +void xen_set_basic_apic_ops(void);
> +
> +#ifdef CONFIG_XEN_HVM_PV
> +void __init xen_hvm_pv_init_irq_ops(void);
> +void __init xen_hvm_pv_evtchn_init_IRQ(void);
> +#endif /* CONFIG_XEN_HVM_PV */
>
>   irqreturn_t xen_debug_interrupt(int irq, void *dev_id);
>
> @@ -61,9 +67,17 @@ void xen_setup_vcpu_info_placement(void);
>   #ifdef CONFIG_SMP
>   void xen_smp_init(void);
>
> +#ifdef CONFIG_XEN_HVM_PV
> +void xen_hvm_pv_smp_init(void);
> +#endif /* CONFIG_XEN_HVM_PV */
> +
>   extern cpumask_var_t xen_cpu_initialized_map;
>   #else
>   static inline void xen_smp_init(void) {}
> +#ifdef CONFIG_XEN_HVM_PV
> +static inline void xen_hvm_pv_smp_init(void) {}
> +#endif /* CONFIG_XEN_HVM_PV */
> +
>   #endif
>
>   #ifdef CONFIG_PARAVIRT_SPINLOCKS
> diff --git a/drivers/xen/events.c b/drivers/xen/events.c
> index ce602dd..e4b9de6 100644
> --- a/drivers/xen/events.c
> +++ b/drivers/xen/events.c
> @@ -32,14 +32,17 @@
>   #include<asm/irq.h>
>   #include<asm/idle.h>
>   #include<asm/sync_bitops.h>
> +#include<asm/desc.h>
>   #include<asm/xen/hypercall.h>
>   #include<asm/xen/hypervisor.h>
>
> +#include<xen/xen.h>
>   #include<xen/xen-ops.h>
>   #include<xen/events.h>
>   #include<xen/interface/xen.h>
>   #include<xen/interface/event_channel.h>
>
> +
>   /*
>    * This lock protects updates to the following mapping and
reference-count
>    * arrays. The lock does not need to be acquired to read the mapping
tables.
> @@ -616,17 +619,13 @@ static DEFINE_PER_CPU(unsigned, xed_nesting_count);
>    * a bitset of words which contain pending event bits.  The second
>    * level is a bitset of pending events themselves.
>    */
> -void xen_evtchn_do_upcall(struct pt_regs *regs)
> +void __xen_evtchn_do_upcall(struct pt_regs *regs)
>   {
>          int cpu = get_cpu();
> -       struct pt_regs *old_regs = set_irq_regs(regs);
>          struct shared_info *s = HYPERVISOR_shared_info;
>          struct vcpu_info *vcpu_info = __get_cpu_var(xen_vcpu);
>          unsigned count;
>
> -       exit_idle();
> -       irq_enter();
> -
>          do {
>                  unsigned long pending_words;
>
> @@ -662,10 +661,25 @@ void xen_evtchn_do_upcall(struct pt_regs *regs)
>          } while(count != 1);
>
>   out:
> +       put_cpu();
> +}
> +
> +void xen_evtchn_do_upcall(struct pt_regs *regs)
> +{
> +       struct pt_regs *old_regs = set_irq_regs(regs);
> +
> +       exit_idle();
> +       irq_enter();
> +
> +       __xen_evtchn_do_upcall(regs);
> +
>          irq_exit();
>          set_irq_regs(old_regs);
> +}
>
> -       put_cpu();
> +void xen_hvm_evtchn_do_upcall(struct pt_regs *regs)
> +{
> +       __xen_evtchn_do_upcall(regs);
>   }
>
>   /* Rebind a new event channel to an existing irq. */
> @@ -944,3 +958,51 @@ void __init xen_init_IRQ(void)
>
>          irq_ctx_init(smp_processor_id());
>   }
> +
> +void __init xen_hvm_pv_evtchn_init_IRQ(void)
> +{
> +       int i;
> +
> +       xen_init_IRQ();
> +       for (i = 0; i<  NR_IRQS_LEGACY; i++) {
> +               struct evtchn_bind_virq bind_virq;
> +               struct irq_desc *desc = irq_to_desc(i);
> +               int virq, evtchn;
> +
> +               virq = i + VIRQ_EMUL_PIN_START;
> +               bind_virq.virq = virq;
> +               bind_virq.vcpu = 0;
> +
> +               if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
> +&bind_virq) != 0)
> +                       BUG();
> +
> +               evtchn = bind_virq.port;
> +               evtchn_to_irq[evtchn] = i;
> +               irq_info[i] = mk_virq_info(evtchn, virq);
> +
> +               desc->status = IRQ_DISABLED;
> +               desc->action = NULL;
> +               desc->depth = 1;
> +
> +               /*
> +                * 16 old-style INTA-cycle interrupts:
> +                */
> +               set_irq_chip_and_handler_name(i,&xen_dynamic_chip,
> +                                       handle_level_irq,
"event");
> +       }
> +
> +       /*
> +        * Cover the whole vector space, no vector can escape
> +        * us. (some of these will be overridden and become
> +        * ''special'' SMP interrupts)
> +        */
> +       for (i = 0; i<  (NR_VECTORS - FIRST_EXTERNAL_VECTOR); i++) {
> +               int vector = FIRST_EXTERNAL_VECTOR + i;
> +               if (vector != IA32_SYSCALL_VECTOR)
> +                       set_intr_gate(vector, interrupt[i]);
> +       }
> +
> +       /* generic IPI for platform specific use, now used for HVM evtchn
*/
> +       alloc_intr_gate(X86_PLATFORM_IPI_VECTOR, x86_platform_ipi);
> +}
> diff --git a/include/xen/events.h b/include/xen/events.h
> index e68d59a..6b12725 100644
> --- a/include/xen/events.h
> +++ b/include/xen/events.h
> @@ -56,4 +56,10 @@ void xen_poll_irq(int irq);
>   /* Determine the IRQ which is bound to an event channel */
>   unsigned irq_from_evtchn(unsigned int evtchn);
>
> +void xen_evtchn_do_upcall(struct pt_regs *regs);
> +
> +#ifdef CONFIG_XEN_HVM_PV
> +void xen_hvm_evtchn_do_upcall(struct pt_regs *regs);
> +#endif
> +
>   #endif /* _XEN_EVENTS_H */
> diff --git a/include/xen/hvm.h b/include/xen/hvm.h
> index 4ea8887..c66d788 100644
> --- a/include/xen/hvm.h
> +++ b/include/xen/hvm.h
> @@ -20,4 +20,9 @@ static inline unsigned long hvm_get_parameter(int idx)
>          return xhv.value;
>   }
>
> +#define HVM_CALLBACK_VIA_TYPE_VECTOR 0x2
> +#define HVM_CALLBACK_VIA_TYPE_SHIFT 56
> +#define HVM_CALLBACK_VECTOR(x)
(((uint64_t)HVM_CALLBACK_VIA_TYPE_VECTOR)<<\
> +                               HVM_CALLBACK_VIA_TYPE_SHIFT | (x))
> +
>   #endif /* XEN_HVM_H__ */
> diff --git a/include/xen/interface/xen.h b/include/xen/interface/xen.h
> index 2befa3e..70a6c6e 100644
> --- a/include/xen/interface/xen.h
> +++ b/include/xen/interface/xen.h
> @@ -90,7 +90,11 @@
>   #define VIRQ_ARCH_6    22
>   #define VIRQ_ARCH_7    23
>
> -#define NR_VIRQS       24
> +#define VIRQ_EMUL_PIN_START 24
> +#define VIRQ_EMUL_PIN_NUM 16
> +
> +#define NR_VIRQS       (VIRQ_EMUL_PIN_START + VIRQ_EMUL_PIN_NUM)
> +
>   /*
>    * MMU-UPDATE REQUESTS
>    *
> diff --git a/include/xen/xen.h b/include/xen/xen.h
> index 11f5ced..ce2a256 100644
> --- a/include/xen/xen.h
> +++ b/include/xen/xen.h
> @@ -26,6 +26,8 @@ extern u32 xen_hvm_pv_features;
>
>   #define xen_hvm_pv_evtchn_enabled() \
>                  (xen_hvm_pv_features&  XEN_HVM_PV_EVTCHN_ENABLED)
> +#else
> +#define xen_hvm_pv_evtchn_enabled() 0
>   #endif /* CONFIG_XEN_HVM_PV */
>
>   #ifdef CONFIG_XEN_DOM0
> --
> 1.5.4.5
>
>    

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Thu, 4 Mar 2010, Jeremy Fitzhardinge wrote:
> 
> I think you need to expand this comment to explain the consequences of
> this change on SMP bringup and timers.  I still don''t see why
there are
> (or should be) any differences from PV cpu bringup, aside from the tiny
> details of syscall setup (and MTRR I guess, but that''s all a noop
> anyway, right?).
> 
> Also, I think you can probably do the timer changes in a separate patch,
> as even with evtchns, an emulated timer device interrupt will still be
> delivered properly via VIRQ_EMUL_PIN_*.
> 
 
One of the problems is that some VCPU initialization hypercalls at the
moment are PV specific and don''t work with HVM guests, in particular I
find that VCPUOP_initialise doesn''t work correctly for me.

But the timers shouldn''t be an issue because you should be able to
receive the VIRQ_TIMER or the emulated timer interrupt anyway.



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Thu, 2010-03-04 at 09:40 -0800, Jeremy Fitzhardinge
wrote:
> 
> > +
> > +static void __init init_pv_clocksource(void)
> > +{
> > +     if (enable_hvm_pv(HVM_PV_CLOCK))
> > +             BUG();
> >    
> 
> BUG is a bit severe.  Will it really never, ever fail?  
When running on a hypervisor without these extension, for example.

Ian.
-- 
Ian Campbell

My polyvinyl cowboy wallet was made in Hong Kong by Montgomery Clift!


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Friday 05 March 2010 02:24:24 Jeremy Fitzhardinge
wrote:
> On 03/04/2010 01:36 AM, Sheng Yang wrote:
> > We mapped each IOAPIC pin to a VIRQ, so that we can deliver interrupt
> > through these VIRQs.
> >
> > We used X86_PLATFORM_IPI_VECTOR as the noficiation vector for
hypervisor
> 
> "notification"
Oops...
> > to notify guest about the event.
> >
> > The patch also enabled SMP support, then we can support IPI through
> > evtchn as well.
> >
> > Then we don''t use IOAPIC/LAPIC, eliminated the overhead
brought by
> > unnecessary VMExit caused by LAPIC.
> 
> I think you need to expand this comment to explain the consequences of
> this change on SMP bringup and timers.  I still don''t see why
there are
> (or should be) any differences from PV cpu bringup, aside from the tiny
> details of syscall setup (and MTRR I guess, but that''s all a noop
> anyway, right?).
We follow the HVM vcpu brought up process. We don''t need to initialize
all
kinds of registers. I think leave it to existed hypervisor is the better idea.
 
> Also, I think you can probably do the timer changes in a separate patch,
> as even with evtchns, an emulated timer device interrupt will still be
> delivered properly via VIRQ_EMUL_PIN_*.
I think as long as we have evtchn, PV timer shouldn''t be a problem. And
for
evtchn, because we don''t have lapic now, we can''t use lapic
timer... So seems
PV timer is a requirement for us.
 
> > Signed-off-by: Sheng Yang<sheng@linux.intel.com>
> > ---
> >   arch/x86/xen/enlighten.c    |    6 +-
> >   arch/x86/xen/hvmpv.c        |   68 +++++++++++++++++++++++++++++++-
> >   arch/x86/xen/irq.c          |   28 +++++++++++++
> >   arch/x86/xen/smp.c          |   92
> > ++++++++++++++++++++++++++++++++++++++++-- arch/x86/xen/xen-ops.h     
|
> >  14 +++++++
> >   drivers/xen/events.c        |   74
+++++++++++++++++++++++++++++++---
> >   include/xen/events.h        |    6 +++
> >   include/xen/hvm.h           |    5 ++
> >   include/xen/interface/xen.h |    6 ++-
> >   include/xen/xen.h           |    2 +
> >   10 files changed, 285 insertions(+), 16 deletions(-)
> >
> > diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
> > index 36daccb..2d60e70 100644
> > --- a/arch/x86/xen/enlighten.c
> > +++ b/arch/x86/xen/enlighten.c
> > @@ -717,7 +717,7 @@ static u32 xen_safe_apic_wait_icr_idle(void)
> >           return 0;
> >   }
> >
> > -static void set_xen_basic_apic_ops(void)
> > +void xen_set_basic_apic_ops(void)
> >   {
> >          apic->read = xen_apic_read;
> >          apic->write = xen_apic_write;
> > @@ -1026,7 +1026,7 @@ static void xen_crash_shutdown(struct pt_regs
> > *regs) xen_reboot(SHUTDOWN_crash);
> >   }
> >
> > -static const struct machine_ops __initdata xen_machine_ops = {
> > +const struct machine_ops __initdata xen_machine_ops = {
> >          .restart = xen_restart,
> >          .halt = xen_machine_halt,
> >          .power_off = xen_machine_halt,
> > @@ -1116,7 +1116,7 @@ asmlinkage void __init xen_start_kernel(void)
> >          /*
> >           * set up the basic apic ops.
> >           */
> > -       set_xen_basic_apic_ops();
> > +       xen_set_basic_apic_ops();
> >   #endif
> >
> >          if (xen_feature(XENFEAT_mmu_pt_update_preserve_ad)) {
> > diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
> > index 284e021..233951b 100644
> > --- a/arch/x86/xen/hvmpv.c
> > +++ b/arch/x86/xen/hvmpv.c
> > @@ -17,6 +17,7 @@
> >   #include<xen/interface/version.h>
> >   #include<xen/interface/memory.h>
> >
> > +#include<asm/reboot.h>
> >   #include<asm/xen/cpuid.h>
> >   #include<asm/xen/hypercall.h>
> >   #include<asm/xen/hypervisor.h>
> > @@ -44,6 +45,8 @@ static void __init xen_hvm_pv_banner(void)
> >          printk(KERN_INFO "Xen version: %d.%d%s\n",
> >                  version>>  16, version&  0xffff,
extra.extraversion);
> >          printk(KERN_INFO "PV feature: PV clocksource
enabled\n");
> > +       if (xen_hvm_pv_evtchn_enabled())
> > +               printk(KERN_INFO "PV feature: Event channel
enabled\n");
> >   }
> >
> >   static int __init xen_para_available(void)
> > @@ -83,6 +86,9 @@ static int __init init_hvm_pv_info(void)
> >          if (!(edx&  XEN_CPUID_FEAT2_HVM_PV))
> >                  return -ENODEV;
> >
> > +       if (edx&  XEN_CPUID_FEAT2_HVM_PV_EVTCHN)
> > +               xen_hvm_pv_features |= XEN_HVM_PV_EVTCHN_ENABLED;
> > +
> >          /* We only support 1 page of hypercall for now */
> >          if (pages != 1)
> >                  return -ENOMEM;
> > @@ -131,12 +137,42 @@ static void __init init_pv_clocksource(void)
> >          x86_platform.get_wallclock = xen_get_wallclock;
> >          x86_platform.set_wallclock = xen_set_wallclock;
> >
> > -       clocksource_register(&xen_clocksource);
> > +       /* It would be done in xen_time_init() if evtchn enabled */
> > +       if (!xen_hvm_pv_evtchn_enabled())
> > +               clocksource_register(&xen_clocksource);
> > +}
> > +
> > +static int set_callback_via(uint64_t via)
> > +{
> > +       struct xen_hvm_param a;
> > +
> > +       a.domid = DOMID_SELF;
> > +       a.index = HVM_PARAM_CALLBACK_IRQ;
> > +       a.value = via;
> > +       return HYPERVISOR_hvm_op(HVMOP_set_param,&a);
> >   }
> >
> > +void do_hvm_pv_evtchn_intr(void)
> > +{
> > +       per_cpu(irq_count, smp_processor_id())++;
> > +       xen_hvm_evtchn_do_upcall(get_irq_regs());
> > +       per_cpu(irq_count, smp_processor_id())--;
> > +}
> > +
> > +#ifdef CONFIG_X86_LOCAL_APIC
> > +static void xen_hvm_pv_evtchn_apic_write(u32 reg, u32 val)
> > +{
> > +       /* The only one reached here should be EOI */
> > +       WARN_ON(reg != APIC_EOI);
> > +}
> > +#endif
> > +
> > +extern struct machine_ops xen_machine_ops;
> > +
> >   void __init xen_guest_init(void)
> >   {
> >          int r;
> > +       uint64_t callback_via;
> >
> >          /* Ensure the we won''t confused with others */
> >          if (xen_domain_type != XEN_NATIVE)
> > @@ -150,4 +186,34 @@ void __init xen_guest_init(void)
> >
> >          /* PV clocksource would be enabled by default */
> >          init_pv_clocksource();
> > +
> > +       if (xen_hvm_pv_evtchn_enabled()) {
> > +               if (enable_hvm_pv(HVM_PV_EVTCHN))
> > +                       return;
> > +
> > +               xen_hvm_pv_init_irq_ops();
> > +
> > +               x86_init.timers.timer_init = xen_time_init;
> > +               x86_init.timers.setup_percpu_clockev = x86_init_noop;
> > +               x86_cpuinit.setup_percpu_clockev = x86_init_noop;
> > +
> > +               pv_apic_ops.startup_ipi_hook = paravirt_nop;
> > +#ifdef CONFIG_X86_LOCAL_APIC
> > +               /*
> > +                * set up the basic apic ops.
> > +                */
> > +               xen_set_basic_apic_ops();
> > +               apic->write = xen_hvm_pv_evtchn_apic_write;
> > +#endif
> > +
> > +               callback_via > >
HVM_CALLBACK_VECTOR(X86_PLATFORM_IPI_VECTOR); +
> > set_callback_via(callback_via);
> > +
> > +               x86_platform_ipi_callback = do_hvm_pv_evtchn_intr;
> > +
> > +               disable_acpi();
> > +
> > +               xen_hvm_pv_smp_init();
> > +               machine_ops = xen_machine_ops;
> > +       }
> >   }
> > diff --git a/arch/x86/xen/irq.c b/arch/x86/xen/irq.c
> > index 9d30105..e325640 100644
> > --- a/arch/x86/xen/irq.c
> > +++ b/arch/x86/xen/irq.c
> > @@ -2,6 +2,7 @@
> >
> >   #include<asm/x86_init.h>
> >
> > +#include<xen/xen.h>
> >   #include<xen/interface/xen.h>
> >   #include<xen/interface/sched.h>
> >   #include<xen/interface/vcpu.h>
> > @@ -131,3 +132,30 @@ void __init xen_init_irq_ops()
> >          pv_irq_ops = xen_irq_ops;
> >          x86_init.irqs.intr_init = xen_init_IRQ;
> >   }
> > +
> > +#ifdef CONFIG_XEN_HVM_PV
> > +static void xen_hvm_pv_evtchn_disable(void)
> > +{
> > +       native_irq_disable();
> > +       xen_irq_disable();
> > +}
> > +PV_CALLEE_SAVE_REGS_THUNK(xen_hvm_pv_evtchn_disable);
> > +
> > +static void xen_hvm_pv_evtchn_enable(void)
> > +{
> > +       native_irq_enable();
> > +       xen_irq_enable();
> > +}
> > +PV_CALLEE_SAVE_REGS_THUNK(xen_hvm_pv_evtchn_enable);
> > +
> > +void __init xen_hvm_pv_init_irq_ops(void)
> > +{
> > +       if (xen_hvm_pv_evtchn_enabled()) {
> > +               pv_irq_ops.irq_disable > > +                    
PV_CALLEE_SAVE(xen_hvm_pv_evtchn_disable);
> > +               pv_irq_ops.irq_enable > > +                     
PV_CALLEE_SAVE(xen_hvm_pv_evtchn_enable);
> > +               x86_init.irqs.intr_init = xen_hvm_pv_evtchn_init_IRQ;
> > +       }
> > +}
> > +#endif
> > diff --git a/arch/x86/xen/smp.c b/arch/x86/xen/smp.c
> > index 563d205..a31429e 100644
> > --- a/arch/x86/xen/smp.c
> > +++ b/arch/x86/xen/smp.c
> > @@ -15,18 +15,24 @@
> >   #include<linux/sched.h>
> >   #include<linux/err.h>
> >   #include<linux/smp.h>
> > +#include<linux/nmi.h>
> >
> >   #include<asm/paravirt.h>
> >   #include<asm/desc.h>
> >   #include<asm/pgtable.h>
> >   #include<asm/cpu.h>
> > +#include<asm/trampoline.h>
> > +#include<asm/tlbflush.h>
> > +#include<asm/mtrr.h>
> >
> >   #include<xen/interface/xen.h>
> >   #include<xen/interface/vcpu.h>
> >
> >   #include<asm/xen/interface.h>
> >   #include<asm/xen/hypercall.h>
> > +#include<asm/xen/hypervisor.h>
> >
> > +#include<xen/xen.h>
> >   #include<xen/page.h>
> >   #include<xen/events.h>
> >
> > @@ -63,8 +69,13 @@ static __cpuinit void cpu_bringup(void)
> >          touch_softlockup_watchdog();
> >          preempt_disable();
> >
> > -       xen_enable_sysenter();
> > -       xen_enable_syscall();
> > +       if (xen_pv_domain()) {
> > +               xen_enable_sysenter();
> > +               xen_enable_syscall();
> > +       } else if (xen_hvm_pv_evtchn_enabled())
> > +               set_mtrr_aps_delayed_init();
> > +       else
> > +               BUG();
Found set_mtrr_aps_delayed_init() is harmless for PV(anyway it wouldn''t
get
the set_mtrr() when calling smp_store_cpu_info()), so make it less specific. 
> >
> >          cpu = smp_processor_id();
> >          smp_store_cpu_info(cpu);
> > @@ -171,7 +182,8 @@ static void __init xen_smp_prepare_boot_cpu(void)
> >
> >          /* We''ve switched to the "real" per-cpu
gdt, so make sure the
> >             old memory can be recycled */
> > -       make_lowmem_page_readwrite(xen_initial_gdt);
> > +       if (xen_feature(XENFEAT_writable_descriptor_tables))
> > +               make_lowmem_page_readwrite(xen_initial_gdt);
> >
> >          xen_setup_vcpu_info_placement();
> >   }
> > @@ -282,6 +294,47 @@ cpu_initialize_context(unsigned int cpu, struct
> > task_struct *idle) return 0;
> >   }
> >
> > +#ifdef CONFIG_XEN_HVM_PV
> 
> Rather than #ifdeffing...
> 
> > +static __cpuinit int
> > +hvm_pv_cpu_initialize_context(unsigned int cpu, struct task_struct
> > *idle) +{
> > +       struct vcpu_guest_context *ctxt;
> > +       unsigned long start_ip;
> 
> Just put
> 
> 	if (!xen_hvm_pv_evtchn_enabled())
> 		return -EOPNOTSUPP;	/* or something */
> 
> and make sure xen_hvm_pv_evtchn_enable() returns constant 0 in the
> !CONFIG_XEN_HVM_PV case.
 
OK. (In fact I''ve done that, but forgot to merge with the former
patch.)
> > +
> > +       if (cpumask_test_and_set_cpu(cpu, xen_cpu_initialized_map))
> > +               return 0;
> > +
> > +       ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
> > +       if (ctxt == NULL)
> > +               return -ENOMEM;
> > +
> > +       early_gdt_descr.address = (unsigned
long)get_cpu_gdt_table(cpu);
> > +       initial_code = (unsigned long)cpu_bringup_and_idle;
> > +       stack_start.sp = (void *) idle->thread.sp;
> > +
> > +       /* start_ip had better be page-aligned! */
> > +       start_ip = setup_trampoline();
> 
> Why go via trampoline?
We are still following the standard HVM brought up process. VMX/SVM would get 
the registers right.
 
> > +
> > +       /* only start_ip is what we want */
> > +       ctxt->flags = VGCF_HVM_GUEST;
> > +       ctxt->user_regs.eip = start_ip;
> > +
> > +       printk(KERN_INFO "Booting processor %d ip 0x%lx\n",
cpu,
> > start_ip); +
> > +       if (HYPERVISOR_vcpu_op(VCPUOP_initialise, cpu, ctxt))
> > +               BUG();
> > +
> > +       kfree(ctxt);
> > +       return 0;
> > +}
> > +#else
> > +static int hvm_pv_cpu_initialize_context(unsigned int cpu,
> > +                                        struct task_struct *idle)
> > +{
> > +       return 0;
> > +}
> > +#endif
> > +
> >   static int __cpuinit xen_cpu_up(unsigned int cpu)
> >   {
> >          struct task_struct *idle = idle_task(cpu);
> > @@ -292,11 +345,17 @@ static int __cpuinit xen_cpu_up(unsigned int
cpu)
> >          irq_ctx_init(cpu);
> >   #else
> >          clear_tsk_thread_flag(idle, TIF_FORK);
> > +
> > +       if (xen_hvm_pv_evtchn_enabled())
> > +               initial_gs = per_cpu_offset(cpu);
> 
> Why does this need to be conditional?
Seems it won''t affect PV... Would update it.
 
> > +
> >          per_cpu(kernel_stack, cpu) > >                 
(unsigned long)task_stack_page(idle) -
> >                  KERNEL_STACK_OFFSET + THREAD_SIZE;
> >   #endif
> > -       xen_setup_runstate_info(cpu);
> > +       if (xen_pv_domain())
> > +               xen_setup_runstate_info(cpu);
> > +
> 
> Does HVM not support this?  If VCPUOP_register_runstate_memory_area
> doesn''t work on HVM, then just change xen_setup_runstate_info() to
not
> BUG on hypercall failure (and presumably nothing tries to use
> runstate_info).
Found HVM support it...

-- 
regards
Yang, Sheng
> 
> >          xen_setup_timer(cpu);
> >          xen_init_lock_cpu(cpu);
> >
> > @@ -305,7 +364,13 @@ static int __cpuinit xen_cpu_up(unsigned int cpu)
> >          /* make sure interrupts start blocked */
> >          per_cpu(xen_vcpu, cpu)->evtchn_upcall_mask = 1;
> >
> > -       rc = cpu_initialize_context(cpu, idle);
> > +       if (xen_pv_domain())
> > +               rc = cpu_initialize_context(cpu, idle);
> > +       else if (xen_hvm_pv_evtchn_enabled())
> > +               rc = hvm_pv_cpu_initialize_context(cpu, idle);
> > +       else
> > +               BUG();
> > +
> >          if (rc)
> >                  return rc;
> >
> > @@ -480,3 +545,20 @@ void __init xen_smp_init(void)
> >          xen_fill_possible_map();
> >          xen_init_spinlocks();
> >   }
> > +
> > +#ifdef CONFIG_XEN_HVM_PV
> > +static void xen_hvm_pv_flush_tlb_others(const struct cpumask
*cpumask,
> > +                                       struct mm_struct *mm, unsigned
> > long va) +{
> > +       /* TODO Make it more specific */
> > +       flush_tlb_all();
> > +}
> > +
> > +void __init xen_hvm_pv_smp_init(void)
> > +{
> > +       if (xen_hvm_pv_evtchn_enabled()) {
> > +               smp_ops = xen_smp_ops;
> > +               pv_mmu_ops.flush_tlb_others > >
xen_hvm_pv_flush_tlb_others; +       }
> > +}
> > +#endif
> > diff --git a/arch/x86/xen/xen-ops.h b/arch/x86/xen/xen-ops.h
> > index f9153a3..6f55815 100644
> > --- a/arch/x86/xen/xen-ops.h
> > +++ b/arch/x86/xen/xen-ops.h
> > @@ -51,6 +51,12 @@ void __init xen_time_init(void);
> >   unsigned long xen_get_wallclock(void);
> >   int xen_set_wallclock(unsigned long time);
> >   unsigned long long xen_sched_clock(void);
> > +void xen_set_basic_apic_ops(void);
> > +
> > +#ifdef CONFIG_XEN_HVM_PV
> > +void __init xen_hvm_pv_init_irq_ops(void);
> > +void __init xen_hvm_pv_evtchn_init_IRQ(void);
> > +#endif /* CONFIG_XEN_HVM_PV */
> >
> >   irqreturn_t xen_debug_interrupt(int irq, void *dev_id);
> >
> > @@ -61,9 +67,17 @@ void xen_setup_vcpu_info_placement(void);
> >   #ifdef CONFIG_SMP
> >   void xen_smp_init(void);
> >
> > +#ifdef CONFIG_XEN_HVM_PV
> > +void xen_hvm_pv_smp_init(void);
> > +#endif /* CONFIG_XEN_HVM_PV */
> > +
> >   extern cpumask_var_t xen_cpu_initialized_map;
> >   #else
> >   static inline void xen_smp_init(void) {}
> > +#ifdef CONFIG_XEN_HVM_PV
> > +static inline void xen_hvm_pv_smp_init(void) {}
> > +#endif /* CONFIG_XEN_HVM_PV */
> > +
> >   #endif
> >
> >   #ifdef CONFIG_PARAVIRT_SPINLOCKS
> > diff --git a/drivers/xen/events.c b/drivers/xen/events.c
> > index ce602dd..e4b9de6 100644
> > --- a/drivers/xen/events.c
> > +++ b/drivers/xen/events.c
> > @@ -32,14 +32,17 @@
> >   #include<asm/irq.h>
> >   #include<asm/idle.h>
> >   #include<asm/sync_bitops.h>
> > +#include<asm/desc.h>
> >   #include<asm/xen/hypercall.h>
> >   #include<asm/xen/hypervisor.h>
> >
> > +#include<xen/xen.h>
> >   #include<xen/xen-ops.h>
> >   #include<xen/events.h>
> >   #include<xen/interface/xen.h>
> >   #include<xen/interface/event_channel.h>
> >
> > +
> >   /*
> >    * This lock protects updates to the following mapping and
> > reference-count * arrays. The lock does not need to be acquired to
read
> > the mapping tables. @@ -616,17 +619,13 @@ static
DEFINE_PER_CPU(unsigned,
> > xed_nesting_count); * a bitset of words which contain pending event
bits.
> >  The second * level is a bitset of pending events themselves.
> >    */
> > -void xen_evtchn_do_upcall(struct pt_regs *regs)
> > +void __xen_evtchn_do_upcall(struct pt_regs *regs)
> >   {
> >          int cpu = get_cpu();
> > -       struct pt_regs *old_regs = set_irq_regs(regs);
> >          struct shared_info *s = HYPERVISOR_shared_info;
> >          struct vcpu_info *vcpu_info = __get_cpu_var(xen_vcpu);
> >          unsigned count;
> >
> > -       exit_idle();
> > -       irq_enter();
> > -
> >          do {
> >                  unsigned long pending_words;
> >
> > @@ -662,10 +661,25 @@ void xen_evtchn_do_upcall(struct pt_regs *regs)
> >          } while(count != 1);
> >
> >   out:
> > +       put_cpu();
> > +}
> > +
> > +void xen_evtchn_do_upcall(struct pt_regs *regs)
> > +{
> > +       struct pt_regs *old_regs = set_irq_regs(regs);
> > +
> > +       exit_idle();
> > +       irq_enter();
> > +
> > +       __xen_evtchn_do_upcall(regs);
> > +
> >          irq_exit();
> >          set_irq_regs(old_regs);
> > +}
> >
> > -       put_cpu();
> > +void xen_hvm_evtchn_do_upcall(struct pt_regs *regs)
> > +{
> > +       __xen_evtchn_do_upcall(regs);
> >   }
> >
> >   /* Rebind a new event channel to an existing irq. */
> > @@ -944,3 +958,51 @@ void __init xen_init_IRQ(void)
> >
> >          irq_ctx_init(smp_processor_id());
> >   }
> > +
> > +void __init xen_hvm_pv_evtchn_init_IRQ(void)
> > +{
> > +       int i;
> > +
> > +       xen_init_IRQ();
> > +       for (i = 0; i<  NR_IRQS_LEGACY; i++) {
> > +               struct evtchn_bind_virq bind_virq;
> > +               struct irq_desc *desc = irq_to_desc(i);
> > +               int virq, evtchn;
> > +
> > +               virq = i + VIRQ_EMUL_PIN_START;
> > +               bind_virq.virq = virq;
> > +               bind_virq.vcpu = 0;
> > +
> > +               if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_virq,
> > +&bind_virq) != 0)
> > +                       BUG();
> > +
> > +               evtchn = bind_virq.port;
> > +               evtchn_to_irq[evtchn] = i;
> > +               irq_info[i] = mk_virq_info(evtchn, virq);
> > +
> > +               desc->status = IRQ_DISABLED;
> > +               desc->action = NULL;
> > +               desc->depth = 1;
> > +
> > +               /*
> > +                * 16 old-style INTA-cycle interrupts:
> > +                */
> > +               set_irq_chip_and_handler_name(i,&xen_dynamic_chip,
> > +                                       handle_level_irq,
"event");
> > +       }
> > +
> > +       /*
> > +        * Cover the whole vector space, no vector can escape
> > +        * us. (some of these will be overridden and become
> > +        * ''special'' SMP interrupts)
> > +        */
> > +       for (i = 0; i<  (NR_VECTORS - FIRST_EXTERNAL_VECTOR); i++)
{
> > +               int vector = FIRST_EXTERNAL_VECTOR + i;
> > +               if (vector != IA32_SYSCALL_VECTOR)
> > +                       set_intr_gate(vector, interrupt[i]);
> > +       }
> > +
> > +       /* generic IPI for platform specific use, now used for HVM
evtchn
> > */ +       alloc_intr_gate(X86_PLATFORM_IPI_VECTOR, x86_platform_ipi);
+}
> > diff --git a/include/xen/events.h b/include/xen/events.h
> > index e68d59a..6b12725 100644
> > --- a/include/xen/events.h
> > +++ b/include/xen/events.h
> > @@ -56,4 +56,10 @@ void xen_poll_irq(int irq);
> >   /* Determine the IRQ which is bound to an event channel */
> >   unsigned irq_from_evtchn(unsigned int evtchn);
> >
> > +void xen_evtchn_do_upcall(struct pt_regs *regs);
> > +
> > +#ifdef CONFIG_XEN_HVM_PV
> > +void xen_hvm_evtchn_do_upcall(struct pt_regs *regs);
> > +#endif
> > +
> >   #endif /* _XEN_EVENTS_H */
> > diff --git a/include/xen/hvm.h b/include/xen/hvm.h
> > index 4ea8887..c66d788 100644
> > --- a/include/xen/hvm.h
> > +++ b/include/xen/hvm.h
> > @@ -20,4 +20,9 @@ static inline unsigned long hvm_get_parameter(int
idx)
> >          return xhv.value;
> >   }
> >
> > +#define HVM_CALLBACK_VIA_TYPE_VECTOR 0x2
> > +#define HVM_CALLBACK_VIA_TYPE_SHIFT 56
> > +#define HVM_CALLBACK_VECTOR(x)
> > (((uint64_t)HVM_CALLBACK_VIA_TYPE_VECTOR)<<\ +
> >    HVM_CALLBACK_VIA_TYPE_SHIFT | (x)) +
> >   #endif /* XEN_HVM_H__ */
> > diff --git a/include/xen/interface/xen.h b/include/xen/interface/xen.h
> > index 2befa3e..70a6c6e 100644
> > --- a/include/xen/interface/xen.h
> > +++ b/include/xen/interface/xen.h
> > @@ -90,7 +90,11 @@
> >   #define VIRQ_ARCH_6    22
> >   #define VIRQ_ARCH_7    23
> >
> > -#define NR_VIRQS       24
> > +#define VIRQ_EMUL_PIN_START 24
> > +#define VIRQ_EMUL_PIN_NUM 16
> > +
> > +#define NR_VIRQS       (VIRQ_EMUL_PIN_START + VIRQ_EMUL_PIN_NUM)
> > +
> >   /*
> >    * MMU-UPDATE REQUESTS
> >    *
> > diff --git a/include/xen/xen.h b/include/xen/xen.h
> > index 11f5ced..ce2a256 100644
> > --- a/include/xen/xen.h
> > +++ b/include/xen/xen.h
> > @@ -26,6 +26,8 @@ extern u32 xen_hvm_pv_features;
> >
> >   #define xen_hvm_pv_evtchn_enabled() \
> >                  (xen_hvm_pv_features&  XEN_HVM_PV_EVTCHN_ENABLED)
> > +#else
> > +#define xen_hvm_pv_evtchn_enabled() 0
> >   #endif /* CONFIG_XEN_HVM_PV */
> >
> >   #ifdef CONFIG_XEN_DOM0
> > --
> > 1.5.4.5
> 
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Friday 05 March 2010 01:40:51 Jeremy Fitzhardinge
wrote:
> On 03/04/2010 01:36 AM, Sheng Yang wrote:
> > And enable it by default in PV extended HVM guest.
> >
> > Signed-off-by: Sheng Yang<sheng@linux.intel.com>
> > ---
> >   arch/x86/xen/hvmpv.c |   20 ++++++++++++++++++++
> >   arch/x86/xen/time.c  |    4 +++-
> >   2 files changed, 23 insertions(+), 1 deletions(-)
> >
> > diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
> > index 7a741ed..284e021 100644
> > --- a/arch/x86/xen/hvmpv.c
> > +++ b/arch/x86/xen/hvmpv.c
> > @@ -43,6 +43,7 @@ static void __init xen_hvm_pv_banner(void)
> >   		pv_info.name);
> >   	printk(KERN_INFO "Xen version: %d.%d%s\n",
> >   		version>>  16, version&  0xffff, extra.extraversion);
> > +	printk(KERN_INFO "PV feature: PV clocksource enabled\n");
> >   }
> >
> >   static int __init xen_para_available(void)
> > @@ -117,6 +118,22 @@ static void __init init_shared_info(void)
> >   	per_cpu(xen_vcpu, 0) =&HYPERVISOR_shared_info->vcpu_info[0];
> >   }
> >
> > +extern struct clocksource xen_clocksource;
> 
> No externs in .c files.  But aside from that, I''d prefer you
export a
> function from xen/time.c to do the clocksource registration and call it
> from both places (even if its a one-liner).
OK
> 
> > +
> > +static void __init init_pv_clocksource(void)
> > +{
> > +	if (enable_hvm_pv(HVM_PV_CLOCK))
> > +		BUG();
> 
> BUG is a bit severe.  Will it really never, ever fail?  And if it does,
> the consequence is hardly serious; we just fall back to emulated devices.
It shouldn''t fail. If we got a hypervisor without these extension, this
one
won''t be called. Because the feature should be detected by CPUID, and
the
hypervisor would mask the unsupported features. So the guest won''t see
that
bit if it''s not supported, and this one wouldn''t be involved.

-- 
regards
Yang, Sheng
> 
> > +
> > +	pv_time_ops.sched_clock = xen_sched_clock;
> > +
> > +	x86_platform.calibrate_tsc = xen_tsc_khz;
> > +	x86_platform.get_wallclock = xen_get_wallclock;
> > +	x86_platform.set_wallclock = xen_set_wallclock;
> > +
> > +	clocksource_register(&xen_clocksource);
> > +}
> > +
> >   void __init xen_guest_init(void)
> >   {
> >   	int r;
> > @@ -130,4 +147,7 @@ void __init xen_guest_init(void)
> >   		return;
> >
> >   	init_shared_info();
> > +
> > +	/* PV clocksource would be enabled by default */
> > +	init_pv_clocksource();
> >   }
> > diff --git a/arch/x86/xen/time.c b/arch/x86/xen/time.c
> > index 0d3f07c..c80eacb 100644
> > --- a/arch/x86/xen/time.c
> > +++ b/arch/x86/xen/time.c
> > @@ -13,6 +13,7 @@
> >   #include<linux/clockchips.h>
> >   #include<linux/kernel_stat.h>
> >   #include<linux/math64.h>
> > +#include<linux/module.h>
> >
> >   #include<asm/pvclock.h>
> >   #include<asm/xen/hypervisor.h>
> > @@ -243,7 +244,7 @@ int xen_set_wallclock(unsigned long now)
> >   	return -1;
> >   }
> >
> > -static struct clocksource xen_clocksource __read_mostly = {
> > +struct clocksource xen_clocksource __read_mostly = {
> >   	.name = "xen",
> >   	.rating = 400,
> >   	.read = xen_clocksource_get_cycles,
> > @@ -252,6 +253,7 @@ static struct clocksource xen_clocksource
> > __read_mostly = { .shift = XEN_SHIFT,
> >   	.flags = CLOCK_SOURCE_IS_CONTINUOUS,
> >   };
> > +EXPORT_SYMBOL_GPL(xen_clocksource);
> >
> >   /*
> >      Xen clockevent implementation
> 
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Fri, 2010-03-05 at 02:54 +0000, Sheng Yang wrote:
> On Friday 05 March 2010 01:40:51 Jeremy Fitzhardinge wrote:
> > On 03/04/2010 01:36 AM, Sheng Yang wrote:
> > > And enable it by default in PV extended HVM guest.
> > >
> > > Signed-off-by: Sheng Yang<sheng@linux.intel.com>
> > > ---
> > >   arch/x86/xen/hvmpv.c |   20 ++++++++++++++++++++
> > >   arch/x86/xen/time.c  |    4 +++-
> > >   2 files changed, 23 insertions(+), 1 deletions(-)
> > >
> > > diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
> > > index 7a741ed..284e021 100644
> > > --- a/arch/x86/xen/hvmpv.c
> > > +++ b/arch/x86/xen/hvmpv.c
> > > @@ -43,6 +43,7 @@ static void __init xen_hvm_pv_banner(void)
> > >   		pv_info.name);
> > >   	printk(KERN_INFO "Xen version: %d.%d%s\n",
> > >   		version>>  16, version&  0xffff,
extra.extraversion);
> > > +	printk(KERN_INFO "PV feature: PV clocksource
enabled\n");
> > >   }
> > >
> > >   static int __init xen_para_available(void)
> > > @@ -117,6 +118,22 @@ static void __init init_shared_info(void)
> > >   	per_cpu(xen_vcpu, 0)
=&HYPERVISOR_shared_info->vcpu_info[0];
> > >   }
> > >
> > > +extern struct clocksource xen_clocksource;
> > 
> > No externs in .c files.  But aside from that, I''d prefer you
export a
> > function from xen/time.c to do the clocksource registration and call
it
> > from both places (even if its a one-liner).
> 
> OK
> > 
> > > +
> > > +static void __init init_pv_clocksource(void)
> > > +{
> > > +	if (enable_hvm_pv(HVM_PV_CLOCK))
> > > +		BUG();
> > 
> > BUG is a bit severe.  Will it really never, ever fail?  And if it
does,
> > the consequence is hardly serious; we just fall back to emulated
devices.
> 
> It shouldn''t fail. If we got a hypervisor without these extension,
this one
> won''t be called. Because the feature should be detected by CPUID,
and the
> hypervisor would mask the unsupported features. So the guest won''t
see that
> bit if it''s not supported, and this one wouldn''t be
involved.
It is still possible that a hypervisor might want to turn on the general
extensions but not the PV clock extensions, and we can''t predict what
reasons we might have for doing that in the future. A BUG() is an
awfully big hammer for a failure like this and destroys any hope of
making changes in a forward/backward compatible manner in the future.
Surely it is possible to simply continue with non-PV clock?

Ian.



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Friday 05 March 2010 17:31:47 Ian Campbell wrote:
> On Fri, 2010-03-05 at 02:54 +0000, Sheng Yang wrote:
> > On Friday 05 March 2010 01:40:51 Jeremy Fitzhardinge wrote:
> > > On 03/04/2010 01:36 AM, Sheng Yang wrote:
> > > > And enable it by default in PV extended HVM guest.
> > > >
> > > > Signed-off-by: Sheng Yang<sheng@linux.intel.com>
> > > > ---
> > > >   arch/x86/xen/hvmpv.c |   20 ++++++++++++++++++++
> > > >   arch/x86/xen/time.c  |    4 +++-
> > > >   2 files changed, 23 insertions(+), 1 deletions(-)
> > > >
> > > > diff --git a/arch/x86/xen/hvmpv.c b/arch/x86/xen/hvmpv.c
> > > > index 7a741ed..284e021 100644
> > > > --- a/arch/x86/xen/hvmpv.c
> > > > +++ b/arch/x86/xen/hvmpv.c
> > > > @@ -43,6 +43,7 @@ static void __init xen_hvm_pv_banner(void)
> > > >   		pv_info.name);
> > > >   	printk(KERN_INFO "Xen version: %d.%d%s\n",
> > > >   		version>>  16, version&  0xffff,
extra.extraversion);
> > > > +	printk(KERN_INFO "PV feature: PV clocksource
enabled\n");
> > > >   }
> > > >
> > > >   static int __init xen_para_available(void)
> > > > @@ -117,6 +118,22 @@ static void __init
init_shared_info(void)
> > > >   	per_cpu(xen_vcpu, 0)
=&HYPERVISOR_shared_info->vcpu_info[0];
> > > >   }
> > > >
> > > > +extern struct clocksource xen_clocksource;
> > >
> > > No externs in .c files.  But aside from that, I''d prefer
you export a
> > > function from xen/time.c to do the clocksource registration and
call it
> > > from both places (even if its a one-liner).
> >
> > OK
> >
> > > > +
> > > > +static void __init init_pv_clocksource(void)
> > > > +{
> > > > +	if (enable_hvm_pv(HVM_PV_CLOCK))
> > > > +		BUG();
> > >
> > > BUG is a bit severe.  Will it really never, ever fail?  And if it
does,
> > > the consequence is hardly serious; we just fall back to emulated
> > > devices.
> >
> > It shouldn''t fail. If we got a hypervisor without these
extension, this
> > one won''t be called. Because the feature should be detected
by CPUID, and
> > the hypervisor would mask the unsupported features. So the guest
won''t
> > see that bit if it''s not supported, and this one
wouldn''t be involved.
> 
> It is still possible that a hypervisor might want to turn on the general
> extensions but not the PV clock extensions, and we can''t predict
what
> reasons we might have for doing that in the future. A BUG() is an
> awfully big hammer for a failure like this and destroys any hope of
> making changes in a forward/backward compatible manner in the future.
> Surely it is possible to simply continue with non-PV clock?
Evtchn can''t work with it. 

If the potentially disable pv clock is your concern, I think the better way is 
to make PV clock as a separate feature, and let PV evtchn depends on it.
I''ve
updated the patches to reflect this change.

-- 
regards
Yang, Sheng

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel