2015-11-02 11:34:50

by Denis V. Lunev

[permalink] [raw]
Subject: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value

From: Andrey Smetanin <[email protected]>

Hypervisor Top Level Functional Specification v3/4 says
that TSC page sequence value = -1(0xFFFFFFFF) is used to
indicate that TSC page no longer reliable source of reference
timer. Unfortunately, we found that Windows Hyper-V guest
side implementation uses sequence value = 0 to indicate
that Tsc page no longer valid. This is clearly visible
inside Windows 2012R2 ntoskrnl.exe HvlGetReferenceTime()
function dissassembly:

HvlGetReferenceTime proc near
xchg ax, ax
loc_1401C3132:
mov rax, cs:HvlpReferenceTscPage
mov r9d, [rax]
test r9d, r9d
jz short loc_1401C3176
rdtsc
mov rcx, cs:HvlpReferenceTscPage
shl rdx, 20h
or rdx, rax
mov rax, [rcx+8]
mov rcx, cs:HvlpReferenceTscPage
mov r8, [rcx+10h]
mul rdx
mov rax, cs:HvlpReferenceTscPage
add rdx, r8
mov ecx, [rax]
cmp ecx, r9d
jnz short loc_1401C3132
jmp short loc_1401C3184
loc_1401C3176:
mov ecx, 40000020h
rdmsr
shl rdx, 20h
or rdx, rax
loc_1401C3184:
mov rax, rdx
retn
HvlGetReferenceTime endp

This patch aligns Tsc page invalid sequence value with
Windows Hyper-V guest implementation which is more
compatible with both Hyper-V hypervisor and KVM hypervisor.

Signed-off-by: Andrey Smetanin <[email protected]>
CC: "K. Y. Srinivasan" <[email protected]>
CC: Haiyang Zhang <[email protected]>
CC: Vitaly Kuznetsov <[email protected]>

TARGET: Microsoft Hyper-V Linux guest drivers
Signed-off-by: Denis V. Lunev <[email protected]>
---
drivers/hv/hv.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/hv/hv.c b/drivers/hv/hv.c
index 6341be8..68d0701 100644
--- a/drivers/hv/hv.c
+++ b/drivers/hv/hv.c
@@ -139,7 +139,7 @@ static cycle_t read_hv_clock_tsc(struct clocksource *arg)
cycle_t current_tick;
struct ms_hyperv_tsc_page *tsc_pg = hv_context.tsc_page;

- if (tsc_pg->tsc_sequence != -1) {
+ if (tsc_pg->tsc_sequence != 0) {
/*
* Use the tsc page to compute the value.
*/
@@ -161,7 +161,7 @@ static cycle_t read_hv_clock_tsc(struct clocksource *arg)
if (tsc_pg->tsc_sequence == sequence)
return current_tick;

- if (tsc_pg->tsc_sequence != -1)
+ if (tsc_pg->tsc_sequence != 0)
continue;
/*
* Fallback using MSR method.
--
2.1.4


2015-11-02 21:15:53

by KY Srinivasan

[permalink] [raw]
Subject: RE: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value



> -----Original Message-----
> From: Denis V. Lunev [mailto:[email protected]]
> Sent: Monday, November 2, 2015 3:34 AM
> Cc: [email protected]; [email protected]; linux-
> [email protected]; Andrey Smetanin <[email protected]>; KY
> Srinivasan <[email protected]>; Haiyang Zhang
> <[email protected]>; Vitaly Kuznetsov <[email protected]>;
> Denis V. Lunev <[email protected]>
> Subject: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value
>
> From: Andrey Smetanin <[email protected]>
>
> Hypervisor Top Level Functional Specification v3/4 says
> that TSC page sequence value = -1(0xFFFFFFFF) is used to
> indicate that TSC page no longer reliable source of reference
> timer. Unfortunately, we found that Windows Hyper-V guest
> side implementation uses sequence value = 0 to indicate
> that Tsc page no longer valid. This is clearly visible
> inside Windows 2012R2 ntoskrnl.exe HvlGetReferenceTime()
> function dissassembly:
>
> HvlGetReferenceTime proc near
> xchg ax, ax
> loc_1401C3132:
> mov rax, cs:HvlpReferenceTscPage
> mov r9d, [rax]
> test r9d, r9d
> jz short loc_1401C3176
> rdtsc
> mov rcx, cs:HvlpReferenceTscPage
> shl rdx, 20h
> or rdx, rax
> mov rax, [rcx+8]
> mov rcx, cs:HvlpReferenceTscPage
> mov r8, [rcx+10h]
> mul rdx
> mov rax, cs:HvlpReferenceTscPage
> add rdx, r8
> mov ecx, [rax]
> cmp ecx, r9d
> jnz short loc_1401C3132
> jmp short loc_1401C3184
> loc_1401C3176:
> mov ecx, 40000020h
> rdmsr
> shl rdx, 20h
> or rdx, rax
> loc_1401C3184:
> mov rax, rdx
> retn
> HvlGetReferenceTime endp
>
> This patch aligns Tsc page invalid sequence value with
> Windows Hyper-V guest implementation which is more
> compatible with both Hyper-V hypervisor and KVM hypervisor.
>
> Signed-off-by: Andrey Smetanin <[email protected]>
> CC: "K. Y. Srinivasan" <[email protected]>
Thanks Andrey; the Hyper-V team will be updating the Hyper-V documentation.

Acked-by: K. Y. Srinivasan <[email protected]>

Regards,

K. Y

> CC: Haiyang Zhang <[email protected]>
> CC: Vitaly Kuznetsov <[email protected]>
>
> TARGET: Microsoft Hyper-V Linux guest drivers
> Signed-off-by: Denis V. Lunev <[email protected]>
> ---
> drivers/hv/hv.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/hv/hv.c b/drivers/hv/hv.c
> index 6341be8..68d0701 100644
> --- a/drivers/hv/hv.c
> +++ b/drivers/hv/hv.c
> @@ -139,7 +139,7 @@ static cycle_t read_hv_clock_tsc(struct clocksource
> *arg)
> cycle_t current_tick;
> struct ms_hyperv_tsc_page *tsc_pg = hv_context.tsc_page;
>
> - if (tsc_pg->tsc_sequence != -1) {
> + if (tsc_pg->tsc_sequence != 0) {
> /*
> * Use the tsc page to compute the value.
> */
> @@ -161,7 +161,7 @@ static cycle_t read_hv_clock_tsc(struct clocksource
> *arg)
> if (tsc_pg->tsc_sequence == sequence)
> return current_tick;
>
> - if (tsc_pg->tsc_sequence != -1)
> + if (tsc_pg->tsc_sequence != 0)
> continue;
> /*
> * Fallback using MSR method.
> --
> 2.1.4

2015-11-12 07:16:37

by Denis V. Lunev

[permalink] [raw]
Subject: Re: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value

On 11/02/2015 10:42 PM, KY Srinivasan wrote:
>
>> -----Original Message-----
>> From: Denis V. Lunev [mailto:[email protected]]
>> Sent: Monday, November 2, 2015 3:34 AM
>> Cc: [email protected]; [email protected]; linux-
>> [email protected]; Andrey Smetanin <[email protected]>; KY
>> Srinivasan <[email protected]>; Haiyang Zhang
>> <[email protected]>; Vitaly Kuznetsov <[email protected]>;
>> Denis V. Lunev <[email protected]>
>> Subject: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value
>>
>> From: Andrey Smetanin <[email protected]>
>>
>> Hypervisor Top Level Functional Specification v3/4 says
>> that TSC page sequence value = -1(0xFFFFFFFF) is used to
>> indicate that TSC page no longer reliable source of reference
>> timer. Unfortunately, we found that Windows Hyper-V guest
>> side implementation uses sequence value = 0 to indicate
>> that Tsc page no longer valid. This is clearly visible
>> inside Windows 2012R2 ntoskrnl.exe HvlGetReferenceTime()
>> function dissassembly:
>>
>> HvlGetReferenceTime proc near
>> xchg ax, ax
>> loc_1401C3132:
>> mov rax, cs:HvlpReferenceTscPage
>> mov r9d, [rax]
>> test r9d, r9d
>> jz short loc_1401C3176
>> rdtsc
>> mov rcx, cs:HvlpReferenceTscPage
>> shl rdx, 20h
>> or rdx, rax
>> mov rax, [rcx+8]
>> mov rcx, cs:HvlpReferenceTscPage
>> mov r8, [rcx+10h]
>> mul rdx
>> mov rax, cs:HvlpReferenceTscPage
>> add rdx, r8
>> mov ecx, [rax]
>> cmp ecx, r9d
>> jnz short loc_1401C3132
>> jmp short loc_1401C3184
>> loc_1401C3176:
>> mov ecx, 40000020h
>> rdmsr
>> shl rdx, 20h
>> or rdx, rax
>> loc_1401C3184:
>> mov rax, rdx
>> retn
>> HvlGetReferenceTime endp
>>
>> This patch aligns Tsc page invalid sequence value with
>> Windows Hyper-V guest implementation which is more
>> compatible with both Hyper-V hypervisor and KVM hypervisor.
>>
>> Signed-off-by: Andrey Smetanin <[email protected]>
>> CC: "K. Y. Srinivasan" <[email protected]>
> Thanks Andrey; the Hyper-V team will be updating the Hyper-V documentation.
>
> Acked-by: K. Y. Srinivasan <[email protected]>
>
> Regards,
>
> K. Y
K.Y.,

can you pls clarify the state of this patch? It is a bit unclear
to me whether it is applied or not.

By the way, I also do not see the following patch
"drivers/hv: cleanup synic msrs if vmbus connect failed"
as applied in the Linux-next. You have promised to resend
it will correct author.

Thank you in advance,
Den

2015-11-12 17:35:25

by KY Srinivasan

[permalink] [raw]
Subject: RE: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value



> -----Original Message-----
> From: Denis V. Lunev [mailto:[email protected]]
> Sent: Wednesday, November 11, 2015 11:16 PM
> To: KY Srinivasan <[email protected]>
> Cc: [email protected]; [email protected]; linux-
> [email protected]; Andrey Smetanin <[email protected]>;
> Haiyang Zhang <[email protected]>; Vitaly Kuznetsov
> <[email protected]>
> Subject: Re: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value
>
> On 11/02/2015 10:42 PM, KY Srinivasan wrote:
> >
> >> -----Original Message-----
> >> From: Denis V. Lunev [mailto:[email protected]]
> >> Sent: Monday, November 2, 2015 3:34 AM
> >> Cc: [email protected]; [email protected]; linux-
> >> [email protected]; Andrey Smetanin <[email protected]>;
> KY
> >> Srinivasan <[email protected]>; Haiyang Zhang
> >> <[email protected]>; Vitaly Kuznetsov <[email protected]>;
> >> Denis V. Lunev <[email protected]>
> >> Subject: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value
> >>
> >> From: Andrey Smetanin <[email protected]>
> >>
> >> Hypervisor Top Level Functional Specification v3/4 says
> >> that TSC page sequence value = -1(0xFFFFFFFF) is used to
> >> indicate that TSC page no longer reliable source of reference
> >> timer. Unfortunately, we found that Windows Hyper-V guest
> >> side implementation uses sequence value = 0 to indicate
> >> that Tsc page no longer valid. This is clearly visible
> >> inside Windows 2012R2 ntoskrnl.exe HvlGetReferenceTime()
> >> function dissassembly:
> >>
> >> HvlGetReferenceTime proc near
> >> xchg ax, ax
> >> loc_1401C3132:
> >> mov rax, cs:HvlpReferenceTscPage
> >> mov r9d, [rax]
> >> test r9d, r9d
> >> jz short loc_1401C3176
> >> rdtsc
> >> mov rcx, cs:HvlpReferenceTscPage
> >> shl rdx, 20h
> >> or rdx, rax
> >> mov rax, [rcx+8]
> >> mov rcx, cs:HvlpReferenceTscPage
> >> mov r8, [rcx+10h]
> >> mul rdx
> >> mov rax, cs:HvlpReferenceTscPage
> >> add rdx, r8
> >> mov ecx, [rax]
> >> cmp ecx, r9d
> >> jnz short loc_1401C3132
> >> jmp short loc_1401C3184
> >> loc_1401C3176:
> >> mov ecx, 40000020h
> >> rdmsr
> >> shl rdx, 20h
> >> or rdx, rax
> >> loc_1401C3184:
> >> mov rax, rdx
> >> retn
> >> HvlGetReferenceTime endp
> >>
> >> This patch aligns Tsc page invalid sequence value with
> >> Windows Hyper-V guest implementation which is more
> >> compatible with both Hyper-V hypervisor and KVM hypervisor.
> >>
> >> Signed-off-by: Andrey Smetanin <[email protected]>
> >> CC: "K. Y. Srinivasan" <[email protected]>
> > Thanks Andrey; the Hyper-V team will be updating the Hyper-V
> documentation.
> >
> > Acked-by: K. Y. Srinivasan <[email protected]>
> >
> > Regards,
> >
> > K. Y
> K.Y.,
>
> can you pls clarify the state of this patch? It is a bit unclear
> to me whether it is applied or not.

I will be submitting this shortly.

>
> By the way, I also do not see the following patch
> "drivers/hv: cleanup synic msrs if vmbus connect failed"
> as applied in the Linux-next. You have promised to resend
> it will correct author.

It was resent on October 29. It is in Greg's queue.

Regards,

K. Y
>
> Thank you in advance,
> Den

2015-11-12 18:58:55

by Denis V. Lunev

[permalink] [raw]
Subject: Re: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value

On 11/12/2015 08:35 PM, KY Srinivasan wrote:
>
>> -----Original Message-----
>> From: Denis V. Lunev [mailto:[email protected]]
>> Sent: Wednesday, November 11, 2015 11:16 PM
>> To: KY Srinivasan <[email protected]>
>> Cc: [email protected]; [email protected]; linux-
>> [email protected]; Andrey Smetanin <[email protected]>;
>> Haiyang Zhang <[email protected]>; Vitaly Kuznetsov
>> <[email protected]>
>> Subject: Re: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value
>>
>> On 11/02/2015 10:42 PM, KY Srinivasan wrote:
>>>> -----Original Message-----
>>>> From: Denis V. Lunev [mailto:[email protected]]
>>>> Sent: Monday, November 2, 2015 3:34 AM
>>>> Cc: [email protected]; [email protected]; linux-
>>>> [email protected]; Andrey Smetanin <[email protected]>;
>> KY
>>>> Srinivasan <[email protected]>; Haiyang Zhang
>>>> <[email protected]>; Vitaly Kuznetsov <[email protected]>;
>>>> Denis V. Lunev <[email protected]>
>>>> Subject: [PATCH 1/1] drivers/hv: correct tsc page sequence invalid value
>>>>
>>>> From: Andrey Smetanin <[email protected]>
>>>>
>>>> Hypervisor Top Level Functional Specification v3/4 says
>>>> that TSC page sequence value = -1(0xFFFFFFFF) is used to
>>>> indicate that TSC page no longer reliable source of reference
>>>> timer. Unfortunately, we found that Windows Hyper-V guest
>>>> side implementation uses sequence value = 0 to indicate
>>>> that Tsc page no longer valid. This is clearly visible
>>>> inside Windows 2012R2 ntoskrnl.exe HvlGetReferenceTime()
>>>> function dissassembly:
>>>>
>>>> HvlGetReferenceTime proc near
>>>> xchg ax, ax
>>>> loc_1401C3132:
>>>> mov rax, cs:HvlpReferenceTscPage
>>>> mov r9d, [rax]
>>>> test r9d, r9d
>>>> jz short loc_1401C3176
>>>> rdtsc
>>>> mov rcx, cs:HvlpReferenceTscPage
>>>> shl rdx, 20h
>>>> or rdx, rax
>>>> mov rax, [rcx+8]
>>>> mov rcx, cs:HvlpReferenceTscPage
>>>> mov r8, [rcx+10h]
>>>> mul rdx
>>>> mov rax, cs:HvlpReferenceTscPage
>>>> add rdx, r8
>>>> mov ecx, [rax]
>>>> cmp ecx, r9d
>>>> jnz short loc_1401C3132
>>>> jmp short loc_1401C3184
>>>> loc_1401C3176:
>>>> mov ecx, 40000020h
>>>> rdmsr
>>>> shl rdx, 20h
>>>> or rdx, rax
>>>> loc_1401C3184:
>>>> mov rax, rdx
>>>> retn
>>>> HvlGetReferenceTime endp
>>>>
>>>> This patch aligns Tsc page invalid sequence value with
>>>> Windows Hyper-V guest implementation which is more
>>>> compatible with both Hyper-V hypervisor and KVM hypervisor.
>>>>
>>>> Signed-off-by: Andrey Smetanin <[email protected]>
>>>> CC: "K. Y. Srinivasan" <[email protected]>
>>> Thanks Andrey; the Hyper-V team will be updating the Hyper-V
>> documentation.
>>> Acked-by: K. Y. Srinivasan <[email protected]>
>>>
>>> Regards,
>>>
>>> K. Y
>> K.Y.,
>>
>> can you pls clarify the state of this patch? It is a bit unclear
>> to me whether it is applied or not.
> I will be submitting this shortly.
>
>> By the way, I also do not see the following patch
>> "drivers/hv: cleanup synic msrs if vmbus connect failed"
>> as applied in the Linux-next. You have promised to resend
>> it will correct author.
> It was resent on October 29. It is in Greg's queue.
>
> Regards,
>
> K. Y
>> Thank you in advance,
>> Den
thank you very much for this update :)

Den