PV extension of HVM (Hybrid) for Xen
Subsystems (8)
v9 cover PV extension of HVM (Hybrid) for Xen
The 9th version of patchset to enable PV extension of HVM support in Linux kernel of Xen.
The patch based on upstream 2.6.32.
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 now, and event channel, PV drivers in the future). 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=0x7' ]
in HVM configuration file to expose all PV feature to guest, and
CONFIG_XEN
in the guest kernel configuration file to enable the PV extension support. It would implicit enable CONFIG_XEN_HVM_PV if CONFIG_X86_64 is selected as well.
And the compiled image can be used as native/pv domU/hvm guest/pv feature hvm kernel.
Change from v8:
1. Delay the controversial part of evtchn.
2. Rebase to 2.6.32 so it can be picked up by Jeremy for more test easily.
Change from v7:
1. Move evtchn related macro to patch 7. So that the patch 1-6 can be cleanly applied.
Change from v6:
1. Make PV clocksource as a separate feature rather than default one. Notice PV evtchn still depends on it.
2. Some comments update.
Change from v5:
Update the comments from Jeremy.
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
Jeremy Fitzhardinge (1):
xen: add support for hvm_op
Keir Fraser (1):
xen: Import cpuid.h from Xen
Sheng Yang (5):
xen: Make pv drivers only work with xen_pv_domain() xen/hvm: Xen PV extension of HVM initialization x86/xen: The entrance for PV extension of HVM
xen: Enable PV clocksource for HVM
xen: Enable event channel of PV extension of HVM
arch/x86/include/asm/xen/cpuid.h | 75 +++++++++++ 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 | 4 + arch/x86/xen/Makefile | 1 + arch/x86/xen/enlighten.c | 6 +- arch/x86/xen/hvmpv.c | 228 +++++++++++++++++++++++++++++++++ arch/x86/xen/irq.c | 9 ++ arch/x86/xen/smp.c | 76 +++++++++++- arch/x86/xen/time.c | 12 ++- arch/x86/xen/xen-ops.h | 17 +++ 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 | 4 + 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 | 15 ++ 25 files changed, 754 insertions(+), 22 deletions(-) create mode 100644 arch/x86/include/asm/xen/cpuid.h create mode 100644 arch/x86/xen/hvmpv.c 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
PATCH v9 1/6 xen: add support for hvm_op
From: Jeremy Fitzhardinge <[email protected]>
Add support for hvm_op hypercall.
--- 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
arch/x86/include/asm/xen/hypercall.h +6 -0
@@ -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) {
include/xen/hvm.h new file +23 -0
@@ -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__ */
include/xen/interface/hvm/hvm_op.h new file +72 -0
72 lines not on this page · raw message
include/xen/interface/hvm/params.h new file +111 -0
111 lines not on this page · raw message
Signature · 1 line
1.7.0.1
PATCH v9 2/6 xen: Import cpuid.h from Xen
In reply to Sheng YangFrom: Keir Fraser <[email protected]>
Which would be used by CPUID detection later
--- 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
arch/x86/include/asm/xen/cpuid.h new file +68 -0
@@ -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 <[email protected]> +*/ + +#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__ */
Signature · 1 line
1.7.0.1
PATCH v9 3/6 xen: Make pv drivers only work with xen_pv_domain()
In reply to Sheng YangOtherwise they would still try to enable with HVM domain type.
--- 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(-)
drivers/block/xen-blkfront.c +1 -1
@@ -1067,7 +1067,7 @@ 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 )) {
drivers/input/xen-kbdfront.c +1 -1
@@ -335,7 +335,7 @@ 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. */
drivers/net/xen-netfront.c +1 -1
@@ -1803,7 +1803,7 @@ 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 ())
drivers/video/xen-fbfront.c +1 -1
@@ -680,7 +680,7 @@ 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. */
drivers/xen/grant-table.c +1 -1
@@ -509,7 +509,7 @@ 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 ;
drivers/xen/xenbus/xenbus_probe.c +2 -2
@@ -784,7 +784,7 @@ 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 */
@@ -915,7 +915,7 @@ static void wait_for_devices(struct xenbus_driver *xendrv)unsigned long timeout = jiffies +10 *HZ ;struct device_driver *drv = xendrv ? &xendrv->driver :NULL ; -if (!ready_to_wait_for_devices || !xen_domain ()) +if (!ready_to_wait_for_devices || !xen_pv_domain ())return ;while (exists_disconnected_device (drv)) {
Signature · 1 line
1.7.0.1
PATCH v9 4/6 xen/hvm: Xen PV extension of HVM initialization
In reply to Sheng YangThe 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.
--- arch/x86/include/asm/xen/cpuid.h | 5 + arch/x86/include/asm/xen/hypervisor.h | 12 +++ arch/x86/xen/Kconfig | 4 + arch/x86/xen/Makefile | 1 + arch/x86/xen/hvmpv.c | 135 +++++++++++++++++++++++++++++++++ include/xen/interface/hvm/hvm_op.h | 8 ++ 6 files changed, 165 insertions(+), 0 deletions(-) create mode 100644 arch/x86/xen/hvmpv.c
arch/x86/include/asm/xen/cpuid.h +5 -0
@@ -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_CLOCK 1 +#define XEN_CPUID_FEAT2_HVM_PV_CLOCK (1u<<1) +#endif /* __XEN_PUBLIC_ARCH_X86_CPUID_H__ */
arch/x86/include/asm/xen/hypervisor.h +12 -0
@@ -55,6 +55,18 @@ 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_CLOCK_ENABLED (1u << 0) +#define XEN_HVM_PV_EVTCHN_ENABLED (1u << 1) +extern u32 xen_hvm_pv_features; + +#define xen_hvm_pv_clock_enabled () \ + (xen_hvm_pv_features & XEN_HVM_PV_CLOCK_ENABLED) +#else +#define xen_hvm_pv_clock_enabled () 0 +#endif /* CONFIG_XEN_HVM_PV */ +#ifdef CONFIG_XEN_DOM0 #include <xen/interface/xen.h>
arch/x86/xen/Kconfig +4 -0
@@ -36,3 +36,7 @@ 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 + def_bool y + depends on XEN && X86_64
arch/x86/xen/Makefile +1 -0
@@ -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
arch/x86/xen/hvmpv.c new file +135 -0
135 lines not on this page · raw message
include/xen/interface/hvm/hvm_op.h +8 -0
@@ -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 +/* Enable PV extended HVM features. Should called by BSP */ +#define HVMOP_enable_pv 9 +struct xen_hvm_pv_type { +/* The features want to be enabled */ +uint32_t flags; +#define HVM_PV_CLOCK (1ull<<0) +}; +#endif /* __XEN_PUBLIC_HVM_HVM_OP_H__ */
Signature · 1 line
1.7.0.1
Re: [Xen-devel] [PATCH][v9 4/6] xen/hvm: Xen PV extension of HVM initialization
Quoted 41 lines · Sheng Yang
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_pvhypercall to enable pv support in hypervisor.Signed-off-by Sheng Yang<[email protected]>Signed-off-by Yaozu (Eddie) Dong<[email protected]>--- arch/x86/include/asm/xen/cpuid.h | 5 + arch/x86/include/asm/xen/hypervisor.h | 12 +++ arch/x86/xen/Kconfig | 4 + arch/x86/xen/Makefile | 1 + arch/x86/xen/hvmpv.c | 135 +++++++++++++++++++++++++++++++++ include/xen/interface/hvm/hvm_op.h | 8 ++ 6 files changed, 165 insertions(+), 0 deletions(-)create mode 100644
arch/x86/xen/hvmpv.carch/x86/include/asm/xen/cpuid.h +5
-0@@ -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_CLOCK 1 +#define XEN_CPUID_FEAT2_HVM_PV_CLOCK (1u<<1) +#endif /* __XEN_PUBLIC_ARCH_X86_CPUID_H__ */ arch/x86/include/asm/xen/hypervisor.h +0
-0
arch/x86/include/asm/xen/hypervisor.h extern enum xen_domain_type xen_domain_type; · line 61
Quoted 7 lines
arch/x86/include/asm/xen/hypervisor.h +3
-0@@ -55,6 +55,18 @@ 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_CLOCK_ENABLED (1u<< 0)
Why is this flag needed? As far as I understand it, there's no real underlying hypervisor change needed to make HVM access to pv clock possible; its just a field in the shared_info's vcpu struct after all. Even if you enable this feature unconditionally, the user can still control whether the Xen clocksource is used with the "clocksource=" kernel command-line parameter.
Also, there's nothing about this which is 64-bit specific is there?
Thanks,
J
Re: [Xen-devel] [PATCH][v9 4/6] xen/hvm: Xen PV extension of HVM initialization
Quoted 46 lines · On Saturday 13 March 2010 04:35:53 Jeremy Fitzhardinge wrote:
Quoted 45 lines · Sheng Yang
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_pvhypercall to enable pv support in hypervisor.Signed-off-by Sheng Yang<[email protected]>Signed-off-by Yaozu (Eddie) Dong<[email protected]>--- arch/x86/include/asm/xen/cpuid.h | 5 + arch/x86/include/asm/xen/hypervisor.h | 12 +++ arch/x86/xen/Kconfig | 4 + arch/x86/xen/Makefile | 1 + arch/x86/xen/hvmpv.c | 135 +++++++++++++++++++++++++++++++++ include/xen/interface/hvm/hvm_op.h | 8 ++ 6 files changed, 165 insertions(+), 0 deletions(-)create mode 100644
arch/x86/xen/hvmpv.c(quoted diff) +0
-0
b/arch/x86/include/asm/xen/cpuid.hindex 8787f03..b3a0b3a 100644arch/x86/include/asm/xen/cpuid.h +5
-0@@ -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_CLOCK 1 +#define XEN_CPUID_FEAT2_HVM_PV_CLOCK (1u<<1) +#endif /* __XEN_PUBLIC_ARCH_X86_CPUID_H__ */ (quoted diff) +0
-0
b/arch/x86/include/asm/xen/hypervisor.hindex d5b7e90..7569f64 100644arch/x86/include/asm/xen/hypervisor.h +0
-0@@ -55,6 +55,18 @@ extern enum xen_domain_type xen_domain_type;#define xen_hvm_domain () (xen_domain()&& \ xen_domain_type == XEN_HVM_DOMAIN)
Quoted 10 lines
arch/x86/include/asm/xen/hypervisor.h extern enum xen_domain_type xen_domain_type;
Quoted 3 lines
arch/x86/include/asm/xen/hypervisor.h +3
-0+#ifdef CONFIG_XEN_HVM_PV + +#define XEN_HVM_PV_CLOCK_ENABLED (1u<< 0)Why is this flag needed? As far as I understand it, there's no real underlying hypervisor change needed to make HVM access to pv clock possible; its just a field in the shared_info's vcpu struct after all. Even if you enable this feature unconditionally, the user can still control whether the Xen clocksource is used with the "clocksource=" kernel command-line parameter.
But we should make sure Xen have ability to support such kind of operation. The CPUID would show if Xen have such ability, and if it does, the feature would be enabled unconditionally. Guest kernel always enable all features it can do unconditionally, but Xen should offer the support for them.
Quoted 1 line
Also, there's nothing about this which is 64-bit specific is there?
64 bit things are mostly evtchn/interrupt related. I think no such limit here. But I think it's better to do it step by step, so leave it after evtchn solution settled.
Signature · 2 lines
regards Yang, Sheng
Re: [Xen-devel] [PATCH][v9 4/6] xen/hvm: Xen PV extension of HVM initialization
Quoted 11 lines · Sheng Yang
Quoted 10 lines
arch/x86/include/asm/xen/hypervisor.h extern enum xen_domain_type xen_domain_type;
Quoted 3 lines
arch/x86/include/asm/xen/hypervisor.h +3
-0+#ifdef CONFIG_XEN_HVM_PV + +#define XEN_HVM_PV_CLOCK_ENABLED (1u<< 0)Why is this flag needed? As far as I understand it, there's no real underlying hypervisor change needed to make HVM access to pv clock possible; its just a field in the shared_info's vcpu struct after all. Even if you enable this feature unconditionally, the user can still control whether the Xen clocksource is used with the "clocksource=" kernel command-line parameter.
Quoted 5 lines
But we should make sure Xen have ability to support such kind of operation. The CPUID would show if Xen have such ability, and if it does, the feature would be enabled unconditionally. Guest kernel always enable all features it can do unconditionally, but Xen should offer the support for them.
In my opinion once the guest knows that is running on Xen HVM (that is from xen_cpuid_base() or xen_para_available()) it should assume that the pv clocksource is available, therefore XEN_HVM_PV_CLOCK_ENABLED should not be needed.
In other words the mere presence of Xen should imply XEN_HVM_PV_CLOCK_ENABLED.
Quoted 6 lines
Quoted 1 line
Also, there's nothing about this which is 64-bit specific is there?
64 bit things are mostly evtchn/interrupt related. I think no such limit here. But I think it's better to do it step by step, so leave it after evtchn solution settled.
Do you mean write generic code now, then introduce the 64 bit limitation later? Or the other way around?
I don't have a strong opinion here so I am OK with both approaches, but I would prefer to add the limitation later (maybe we'll be able to make it work on 32 bit too...).
Re: [Xen-devel] [PATCH][v9 4/6] xen/hvm: Xen PV extension of HVM initialization
Quoted 6 lines · Stefano Stabellini
Quoted 6 lines
But we should make sure Xen have ability to support such kind of operation. The CPUID would show if Xen have such ability, and if it does, the feature would be enabled unconditionally. Guest kernel always enable all features it can do unconditionally, but Xen should offer the support for them.
Quoted 7 lines
In my opinion once the guest knows that is running on Xen HVM (that is from
xen_cpuid_base()orxen_para_available()) it should assume that the pv clocksource is available, thereforeXEN_HVM_PV_CLOCK_ENABLEDshould not be needed.
In other words the mere presence of Xen should implyXEN_HVM_PV_CLOCK_ENABLED.
The only reason why we wouldn't want to do this is if we want to withdraw this feature at some point in the future. We're stuck with it indefinitely for PV, but I don't know if that's necessarily going to be the case for HVM. On the other hand, if other - better - mechanisms become available, we can give them their own clocksource driver with a higher priority than the Xen pvclock one, and users can still select clocksources on the kernel command line.
Quoted 6 lines
Do you mean write generic code now, then introduce the 64 bit limitation later? Or the other way around?
I don't have a strong opinion here so I am OK with both approaches, but I would prefer to add the limitation later (maybe we'll be able to make it work on 32 bit too...).
Seems like making it work for both 32 and 64-bit is the easiest thing to do.
J
Re: [Xen-devel] [PATCH][v9 4/6] xen/hvm: Xen PV extension of HVM initialization
Quoted 14 lines · On Tuesday 16 March 2010 06:59:58 Jeremy Fitzhardinge wrote:
Quoted 6 lines · Stefano Stabellini
Quoted 5 lines
But we should make sure Xen have ability to support such kind of operation. The CPUID would show if Xen have such ability, and if it does, the feature would be enabled unconditionally. Guest kernel always enable all features it can do unconditionally, but Xen should offer the support for them.
Quoted 6 lines
In my opinion once the guest knows that is running on Xen HVM (that is from
xen_cpuid_base()orxen_para_available()) it should assume that the pv clocksource is available, thereforeXEN_HVM_PV_CLOCK_ENABLEDshould not be needed.
In other words the mere presence of Xen should implyXEN_HVM_PV_CLOCK_ENABLED.
Quoted 7 lines
The only reason why we wouldn't want to do this is if we want to withdraw this feature at some point in the future. We're stuck with it indefinitely for PV, but I don't know if that's necessarily going to be the case for HVM. On the other hand, if other - better - mechanisms become available, we can give them their own clocksource driver with a higher priority than the Xen pvclock one, and users can still select clocksources on the kernel command line.
So you think about adding a new XENFEAT?
Quoted 9 lines
Quoted 6 lines
Do you mean write generic code now, then introduce the 64 bit limitation later? Or the other way around?
I don't have a strong opinion here so I am OK with both approaches, but I would prefer to add the limitation later (maybe we'll be able to make it work on 32 bit too...).Seems like making it work for both 32 and 64-bit is the easiest thing to
do.
If it is, it should be fine. But I had encountered some issues on 32 bits.
Signature · 2 lines
regards Yang, Sheng
Re: [Xen-devel] [PATCH][v9 4/6] xen/hvm: Xen PV extension of HVM initialization
Quoted 25 lines · Sheng Yang
Quoted 15 lines · On Tuesday 16 March 2010 06:59:58 Jeremy Fitzhardinge wrote:
Quoted 6 lines · Stefano Stabellini
Quoted 6 lines
But we should make sure Xen have ability to support such kind of operation. The CPUID would show if Xen have such ability, and if it does, the feature would be enabled unconditionally. Guest kernel always enable all features it can do unconditionally, but Xen should offer the support for them.
Quoted 7 lines
In my opinion once the guest knows that is running on Xen HVM (that is from
xen_cpuid_base()orxen_para_available()) it should assume that the pv clocksource is available, thereforeXEN_HVM_PV_CLOCK_ENABLEDshould not be needed.
In other words the mere presence of Xen should implyXEN_HVM_PV_CLOCK_ENABLED.Quoted 8 lines
The only reason why we wouldn't want to do this is if we want to withdraw this feature at some point in the future. We're stuck with it indefinitely for PV, but I don't know if that's necessarily going to be the case for HVM. On the other hand, if other - better - mechanisms become available, we can give them their own clocksource driver with a higher priority than the Xen pvclock one, and users can still select clocksources on the kernel command line.
Quoted 2 lines
So you think about adding a new XENFEAT?
I think that's a bit arbitrary. If we need a new vcpuop to make it work properly anyway (set tsc offset), then the presence of that should be a good indicator. In other words, make it depend on its actual fixed pre-reqs, rather than a specific flag for the feature.
Quoted 5 lines
Quoted 3 lines
Seems like making it work for both 32 and 64-bit is the easiest thing to
do.If it is, it should be fine. But I had encountered some issues on 32 bits.
What kinds of issues?
J
Re: [Xen-devel] [PATCH][v9 4/6] xen/hvm: Xen PV extension of HVM initialization
Quoted 5 lines · Jeremy Fitzhardinge
If we need a new vcpuop to make it work
properly anyway (set tsc offset), then the presence of that should be a
good indicator. In other words, make it depend on its actual fixed pre-reqs, rather than a specific flag for the feature.
I don't think a new hypercall is required. Xen should simply pre-account for the TSC offset which HVM guests are subject to in __update_vcpu_system_time.
Ian.
Re: [Xen-devel] [PATCH][v9 4/6] xen/hvm: Xen PV extension of HVM initialization
Quoted 4 lines · Ian Campbell
I don't think a new hypercall is required. Xen should simply pre-account for the TSC offset which HVM guests are subject to in
__update_vcpu_system_time.
Yes, that would work too.
J
PATCH v9 5/6 x86/xen: The entrance for PV extension of HVM
In reply to Sheng Yangxen_guest_init() would setup the environment.
--- arch/x86/include/asm/xen/hypervisor.h | 6 ++++++ arch/x86/kernel/setup.c | 4 ++++ 2 files changed, 10 insertions(+), 0 deletions(-)
arch/x86/include/asm/xen/hypervisor.h +6 -0
@@ -76,4 +76,10 @@ extern u32 xen_hvm_pv_features;#define xen_initial_domain () (0)#endif /* CONFIG_XEN_DOM0 */ +#ifdef CONFIG_XEN_HVM_PV +void xen_guest_init (void ); +#else +static inline void xen_guest_init (void ) {}; +#endif +#endif /* _ASM_X86_XEN_HYPERVISOR_H */
arch/x86/kernel/setup.c +4 -0
@@ -110,6 +110,8 @@#include <asm/numa_64.h> #endif +#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
@@ -753,6 +755,8 @@ 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 */
Signature · 1 line
1.7.0.1
PATCH v9 6/6 xen: Enable PV clocksource for HVM
In reply to Sheng YangPV clocksource can provide a reliable clocksource for HVM running on Xen.
To enable it, put following line to the HVM configure file:
cpuid = [ '0x40000002:edx=0x3' ]
It would set bit 0 and bit 1 in 0x40000002:edx, which enable PV extension framework and PV clocksource.
--- arch/x86/xen/hvmpv.c | 24 ++++++++++++++++++++++++ arch/x86/xen/time.c | 12 +++++++++++- arch/x86/xen/xen-ops.h | 1 + 3 files changed, 36 insertions(+), 1 deletions(-)
arch/x86/xen/hvmpv.c +24 -0
@@ -42,6 +42,8 @@ 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); +if (xen_hvm_pv_clock_enabled ()) +printk (KERN_INFO "PV feature: PV clocksource enabled\n" ); }static int __init xen_para_available (void )
@@ -80,6 +82,9 @@ static int __init init_hvm_pv_info(void)if (!(edx &XEN_CPUID_FEAT2_HVM_PV ))return -ENODEV ; +if (edx &XEN_CPUID_FEAT2_HVM_PV_CLOCK ) + xen_hvm_pv_features |=XEN_HVM_PV_CLOCK_ENABLED ; +if (pages <1 )return -ENODEV ;
@@ -110,6 +115,23 @@ static int __init init_shared_info(void)return 0 ; } +static void __init init_pv_clocksource (void ) +{ +if (!xen_hvm_pv_clock_enabled ()) +return ; + +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; + +xen_register_clocksource (); +} +void __init xen_guest_init (void ) {int r;
@@ -132,4 +154,6 @@ void __init xen_guest_init(void) pv_info = xen_hvm_pv_info; xen_domain_type =XEN_HVM_DOMAIN ; + +init_pv_clocksource (); }
arch/x86/xen/time.c +11 -1
@@ -474,11 +474,21 @@ void xen_timer_resume(void) } } +static bool xen_clocksource_enabled; + +void xen_register_clocksource (void ) +{ +if (!xen_clocksource_enabled) { +clocksource_register (&xen_clocksource); + xen_clocksource_enabled =1 ; + } +} +__init void xen_time_init (void ) {int cpu =smp_processor_id (); -clocksource_register (&xen_clocksource); +xen_register_clocksource ();if (HYPERVISOR_vcpu_op (VCPUOP_stop_periodic_timer, cpu,NULL ) ==0 ) {/* Successfully turned off 100Hz tick, so we have the
arch/x86/xen/xen-ops.h +1 -0
@@ -46,6 +46,7 @@ cycle_t xen_clocksource_read(void);void xen_setup_cpu_clockevents (void );unsigned long xen_tsc_khz (void );void __init xen_time_init (void ); +void xen_register_clocksource (void );unsigned long xen_get_wallclock (void );int xen_set_wallclock (unsigned long time);unsigned long long xen_sched_clock (void );
Signature · 1 line
1.7.0.1
Re: [Xen-devel] [PATCH][v9 6/6] xen: Enable PV clocksource for HVM
Quoted 9 lines · Sheng Yang
PV clocksource can provide a reliable clocksource for HVM running on Xen.
To enable it, put following line to the HVM configure file:
cpuid = [ '0x40000002:edx=0x3' ]
It would set bit 0 and bit 1 in 0x40000002:edx, which enable PV extension framework and PV clocksource.
I guess my previous comment would be better made here. I don't see any reason to make the Xen pv clocksource conditional at runtime. If the user wants to disable/override it, then they can either boot with a clocksource= parameter on the kernel command line, or change it on the fly via /sys.
J
Quoted 101 lines
Signed-off-by Sheng Yang<[email protected]>--- arch/x86/xen/hvmpv.c | 24 ++++++++++++++++++++++++ arch/x86/xen/time.c | 12 +++++++++++- arch/x86/xen/xen-ops.h | 1 + 3 files changed, 36 insertions(+), 1 deletions(-)arch/x86/xen/hvmpv.c +24
-0@@ -42,6 +42,8 @@ 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); +if (xen_hvm_pv_clock_enabled ()) +printk (KERN_INFO "PV feature: PV clocksource enabled\n" ); }static int __init xen_para_available (void )@@ -80,6 +82,9 @@ static int __init init_hvm_pv_info(void)if (!(edx&XEN_CPUID_FEAT2_HVM_PV ))return -ENODEV ; +if (edx&XEN_CPUID_FEAT2_HVM_PV_CLOCK ) + xen_hvm_pv_features |=XEN_HVM_PV_CLOCK_ENABLED ; +if (pages<1 )return -ENODEV ;@@ -110,6 +115,23 @@ static int __init init_shared_info(void)return 0 ; } +static void __init init_pv_clocksource (void ) +{ +if (!xen_hvm_pv_clock_enabled ()) +return ; + +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; + +xen_register_clocksource (); +} +void __init xen_guest_init (void ) {int r;@@ -132,4 +154,6 @@ void __init xen_guest_init(void) pv_info = xen_hvm_pv_info; xen_domain_type =XEN_HVM_DOMAIN ; + +init_pv_clocksource (); }arch/x86/xen/time.c +11
-1@@ -474,11 +474,21 @@ void xen_timer_resume(void) } } +static bool xen_clocksource_enabled; + +void xen_register_clocksource (void ) +{ +if (!xen_clocksource_enabled) { +clocksource_register (&xen_clocksource); + xen_clocksource_enabled =1 ; + } +} +__init void xen_time_init (void ) {int cpu =smp_processor_id (); -clocksource_register (&xen_clocksource); +xen_register_clocksource ();if (HYPERVISOR_vcpu_op (VCPUOP_stop_periodic_timer, cpu,NULL ) ==0 ) {/* Successfully turned off 100Hz tick, so we have the arch/x86/xen/xen-ops.h +1
-0@@ -46,6 +46,7 @@ cycle_t xen_clocksource_read(void);void xen_setup_cpu_clockevents (void );unsigned long xen_tsc_khz (void );void __init xen_time_init (void ); +void xen_register_clocksource (void );unsigned long xen_get_wallclock (void );int xen_set_wallclock (unsigned long time);unsigned long long xen_sched_clock (void );