Hi,
In virtualization environment, PV extensions (drivers, interrupts,
timers, etc) are enabled in the majority of use cases which is the
best option.
However, in some cases (kexec not fully working, benchmarking, etc)
we want to disable PV extensions. We have xen_nopv for that purpose
but only for XEN. For a consistent admin experience a common command
line parameter set across all PV guest implementations is a better
choice.
To achieve this introduce a new 'nopv' parameter which is usable by
most of PV guest implementation. Due to the limitation of some PV
guests(XEN PV, XEN PVH and jailhouse), 'nopv' is ignored for XEN PV
, jailhouse and XEN PVH if booting via Xen-PVH boot entry. If booting
via normal boot entry(like grub2), PVH guest has to panic itself
currently.
While analyzing the PV guest code one bug were found and fixed.
(Patches 1). It can be applied independent of the functional
changes, but is kept in the series as the functional changes
depend on them.
For compatibility reason, "xen_nopv" is keeped and mapped to "nopv",
this way also avoids an issue with xen_nopv when booting PVH guest.
Build test passes with CONFIG_HYPERVISOR_GUEST enable and disabled.
I didn't get env to test with jailhouse and Hyperv, the others work
as expected.
v6:
PATCH3 add Reviewed-by
PATCH4 remove unnecessory xen_hvm_nopv_guest_late_init() per Boris
v5:
PATCH2:
update patch description per Boris
add declaration of nopv variable in arch/x86/include/asm/hypervisor.h
which will be used in PATCH3 and PATCH4
PATCH3 update xen_parse_nopv() per Boris
PATCH4 add nopv=false per Boris
Combine PATCH5 into PATCH3
v4:
PATCH5 a new patch to add 'xen_nopv' back per Boris
v3:
Remove some unrelated patches from patchset as suggested by Tglx
PATCH1 unchanged
PATCH2 add Reviewed-by
PATCH3 add Reviewed-by
PATCH4 rewrite the patch as Jgross found an issue in old patch,
description is also updated.
v2:
PATCH3 use 'ignore_nopv' for PVH/PV guest as suggested by Jgross.
PATCH5 new added one, specifically for HVM guest
Thanks
Zhenzhong
In virtualization environment, PV extensions (drivers, interrupts,
timers, etc) are enabled in the majority of use cases which is the
best option.
However, in some cases (kexec not fully working, benchmarking)
we want to disable PV extensions. We have "xen_nopv" for that purpose
but only for XEN. For a consistent admin experience a common command
line parameter "nopv" set across all PV guest implementations is a
better choice.
There are guest types which just won't work without PV extensions,
like Xen PV, Xen PVH and jailhouse. add a "ignore_nopv" member to
struct hypervisor_x86 set to true for those guest types and call
the detect functions only if nopv is false or ignore_nopv is true.
Suggested-by: Juergen Gross <[email protected]>
Signed-off-by: Zhenzhong Duan <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Jan Kiszka <[email protected]>
Cc: Boris Ostrovsky <[email protected]>
Cc: Stefano Stabellini <[email protected]>
---
Documentation/admin-guide/kernel-parameters.txt | 5 +++++
arch/x86/include/asm/hypervisor.h | 4 ++++
arch/x86/kernel/cpu/hypervisor.c | 11 +++++++++++
arch/x86/kernel/jailhouse.c | 1 +
arch/x86/xen/enlighten_pv.c | 1 +
5 files changed, 22 insertions(+)
diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 138f666..21e08af 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -5268,6 +5268,11 @@
improve timer resolution at the expense of processing
more timer interrupts.
+ nopv= [X86,XEN,KVM,HYPER_V,VMWARE]
+ Disables the PV optimizations forcing the guest to run
+ as generic guest with no PV drivers. Currently support
+ XEN HVM, KVM, HYPER_V and VMWARE guest.
+
xirc2ps_cs= [NET,PCMCIA]
Format:
<irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]]
diff --git a/arch/x86/include/asm/hypervisor.h b/arch/x86/include/asm/hypervisor.h
index 8c5aaba..00240b0 100644
--- a/arch/x86/include/asm/hypervisor.h
+++ b/arch/x86/include/asm/hypervisor.h
@@ -52,8 +52,12 @@ struct hypervisor_x86 {
/* runtime callbacks */
struct x86_hyper_runtime runtime;
+
+ /* ignore nopv parameter */
+ bool ignore_nopv;
};
+extern bool nopv;
extern enum x86_hypervisor_type x86_hyper_type;
extern void init_hypervisor_platform(void);
static inline bool hypervisor_is_type(enum x86_hypervisor_type type)
diff --git a/arch/x86/kernel/cpu/hypervisor.c b/arch/x86/kernel/cpu/hypervisor.c
index 479ca47..337ff07 100644
--- a/arch/x86/kernel/cpu/hypervisor.c
+++ b/arch/x86/kernel/cpu/hypervisor.c
@@ -54,6 +54,14 @@
enum x86_hypervisor_type x86_hyper_type;
EXPORT_SYMBOL(x86_hyper_type);
+bool __initdata nopv;
+static __init int parse_nopv(char *arg)
+{
+ nopv = true;
+ return 0;
+}
+early_param("nopv", parse_nopv);
+
static inline const struct hypervisor_x86 * __init
detect_hypervisor_vendor(void)
{
@@ -61,6 +69,9 @@
uint32_t pri, max_pri = 0;
for (p = hypervisors; p < hypervisors + ARRAY_SIZE(hypervisors); p++) {
+ if (unlikely(nopv) && !(*p)->ignore_nopv)
+ continue;
+
pri = (*p)->detect();
if (pri > max_pri) {
max_pri = pri;
diff --git a/arch/x86/kernel/jailhouse.c b/arch/x86/kernel/jailhouse.c
index 1b2ee55..c52c4105 100644
--- a/arch/x86/kernel/jailhouse.c
+++ b/arch/x86/kernel/jailhouse.c
@@ -217,4 +217,5 @@ static bool jailhouse_x2apic_available(void)
.detect = jailhouse_detect,
.init.init_platform = jailhouse_init_platform,
.init.x2apic_available = jailhouse_x2apic_available,
+ .ignore_nopv = true,
};
diff --git a/arch/x86/xen/enlighten_pv.c b/arch/x86/xen/enlighten_pv.c
index 4722ba2..5d16824 100644
--- a/arch/x86/xen/enlighten_pv.c
+++ b/arch/x86/xen/enlighten_pv.c
@@ -1463,4 +1463,5 @@ static uint32_t __init xen_platform_pv(void)
.detect = xen_platform_pv,
.type = X86_HYPER_XEN_PV,
.runtime.pin_vcpu = xen_pin_vcpu,
+ .ignore_nopv = true,
};
--
1.8.3.1