When sending IPI in the X2APIC logical cluster mode, the destination
APIC ID is encoded as:
* Cluster ID = ICRH[31:16]
* Logical ID = ICRH[15:0]
Current logic incorrectly decode the ICRH, which causes VM running
with x2AVIC support to fail to boot. Therefore, fix the decoding logic.
The commit 603ccef42ce9 ("KVM: x86: SVM: fix avic_kick_target_vcpus_fast")
also added a check for multiple logical destinations before using
the fast-path. However, the same logic is already existed prior to
the commit. Therefore, remove redundant checking logic.
Fixes: 603ccef42ce9 ("KVM: x86: SVM: fix avic_kick_target_vcpus_fast")
Cc: Maxim Levitsky <[email protected]>
Signed-off-by: Suravee Suthikulpanit <[email protected]>
---
arch/x86/kvm/svm/avic.c | 19 ++++---------------
1 file changed, 4 insertions(+), 15 deletions(-)
diff --git a/arch/x86/kvm/svm/avic.c b/arch/x86/kvm/svm/avic.c
index 6919dee69f18..45ab49d1f0b8 100644
--- a/arch/x86/kvm/svm/avic.c
+++ b/arch/x86/kvm/svm/avic.c
@@ -378,8 +378,8 @@ static int avic_kick_target_vcpus_fast(struct kvm *kvm, struct kvm_lapic *source
if (apic_x2apic_mode(source)) {
/* 16 bit dest mask, 16 bit cluster id */
- bitmap = dest & 0xFFFF0000;
- cluster = (dest >> 16) << 4;
+ bitmap = dest & 0xffff;
+ cluster = (dest & 0xffff0000) >> 16;
} else if (kvm_lapic_get_reg(source, APIC_DFR) == APIC_DFR_FLAT) {
/* 8 bit dest mask*/
bitmap = dest;
@@ -387,7 +387,7 @@ static int avic_kick_target_vcpus_fast(struct kvm *kvm, struct kvm_lapic *source
} else {
/* 4 bit desk mask, 4 bit cluster id */
bitmap = dest & 0xF;
- cluster = (dest >> 4) << 2;
+ cluster = (dest & 0xf0) >> 4;
}
if (unlikely(!bitmap))
@@ -420,18 +420,7 @@ static int avic_kick_target_vcpus_fast(struct kvm *kvm, struct kvm_lapic *source
* For x2APIC logical mode, cannot leverage the index.
* Instead, calculate physical ID from logical ID in ICRH.
*/
- int cluster = (icrh & 0xffff0000) >> 16;
- int apic = ffs(icrh & 0xffff) - 1;
-
- /*
- * If the x2APIC logical ID sub-field (i.e. icrh[15:0])
- * contains anything but a single bit, we cannot use the
- * fast path, because it is limited to a single vCPU.
- */
- if (apic < 0 || icrh != (1 << apic))
- return -EINVAL;
-
- l1_physical_id = (cluster << 4) + apic;
+ l1_physical_id = (cluster << 4) + (ffs(bitmap) - 1);
}
}
--
2.34.1
On Mon, Sep 12, 2022, Suravee Suthikulpanit wrote:
> When sending IPI in the X2APIC logical cluster mode, the destination
> APIC ID is encoded as:
>
> * Cluster ID = ICRH[31:16]
> * Logical ID = ICRH[15:0]
>
> Current logic incorrectly decode the ICRH, which causes VM running
> with x2AVIC support to fail to boot. Therefore, fix the decoding logic.
There are already patches pending[1][2] for the x2AVIC bugs.
[1] https://lore.kernel.org/all/[email protected]
[2] https://lore.kernel.org/all/[email protected]
>
> The commit 603ccef42ce9 ("KVM: x86: SVM: fix avic_kick_target_vcpus_fast")
> also added a check for multiple logical destinations before using
> the fast-path. However, the same logic is already existed prior to
> the commit. Therefore, remove redundant checking logic.
>
> Fixes: 603ccef42ce9 ("KVM: x86: SVM: fix avic_kick_target_vcpus_fast")
> Cc: Maxim Levitsky <[email protected]>
> Signed-off-by: Suravee Suthikulpanit <[email protected]>
> ---
> arch/x86/kvm/svm/avic.c | 19 ++++---------------
> 1 file changed, 4 insertions(+), 15 deletions(-)
>
> diff --git a/arch/x86/kvm/svm/avic.c b/arch/x86/kvm/svm/avic.c
> index 6919dee69f18..45ab49d1f0b8 100644
> --- a/arch/x86/kvm/svm/avic.c
> +++ b/arch/x86/kvm/svm/avic.c
> @@ -378,8 +378,8 @@ static int avic_kick_target_vcpus_fast(struct kvm *kvm, struct kvm_lapic *source
>
> if (apic_x2apic_mode(source)) {
> /* 16 bit dest mask, 16 bit cluster id */
> - bitmap = dest & 0xFFFF0000;
> - cluster = (dest >> 16) << 4;
> + bitmap = dest & 0xffff;
> + cluster = (dest & 0xffff0000) >> 16;
> } else if (kvm_lapic_get_reg(source, APIC_DFR) == APIC_DFR_FLAT) {
> /* 8 bit dest mask*/
> bitmap = dest;
> @@ -387,7 +387,7 @@ static int avic_kick_target_vcpus_fast(struct kvm *kvm, struct kvm_lapic *source
> } else {
> /* 4 bit desk mask, 4 bit cluster id */
> bitmap = dest & 0xF;
> - cluster = (dest >> 4) << 2;
> + cluster = (dest & 0xf0) >> 4;
This is wrong and unrelated. The cluster needs to be shifted back by 2, i.e. multiplied
by 4, to leap over each cluster of 4 IDs.
> }
>
> if (unlikely(!bitmap))
> @@ -420,18 +420,7 @@ static int avic_kick_target_vcpus_fast(struct kvm *kvm, struct kvm_lapic *source
> * For x2APIC logical mode, cannot leverage the index.
> * Instead, calculate physical ID from logical ID in ICRH.
> */
> - int cluster = (icrh & 0xffff0000) >> 16;
> - int apic = ffs(icrh & 0xffff) - 1;
> -
> - /*
> - * If the x2APIC logical ID sub-field (i.e. icrh[15:0])
> - * contains anything but a single bit, we cannot use the
> - * fast path, because it is limited to a single vCPU.
> - */
> - if (apic < 0 || icrh != (1 << apic))
> - return -EINVAL;
> -
> - l1_physical_id = (cluster << 4) + apic;
> + l1_physical_id = (cluster << 4) + (ffs(bitmap) - 1);
> }
> }
>
> --
> 2.34.1
>
On 9/13/2022 12:29 AM, Sean Christopherson wrote:
> On Mon, Sep 12, 2022, Suravee Suthikulpanit wrote:
>> When sending IPI in the X2APIC logical cluster mode, the destination
>> APIC ID is encoded as:
>>
>> * Cluster ID = ICRH[31:16]
>> * Logical ID = ICRH[15:0]
>>
>> Current logic incorrectly decode the ICRH, which causes VM running
>> with x2AVIC support to fail to boot. Therefore, fix the decoding logic.
>
> There are already patches pending[1][2] for the x2AVIC bugs.
>
> [1] https://lore.kernel.org/all/20220903002254.2411750-9-seanjc@xxxxxxxxxx
> [2] https://lore.kernel.org/all/20220903002254.2411750-10-seanjc@xxxxxxxxxx
Ok. Thanks for the pointer.
>>
>> The commit 603ccef42ce9 ("KVM: x86: SVM: fix avic_kick_target_vcpus_fast")
>> also added a check for multiple logical destinations before using
>> the fast-path. However, the same logic is already existed prior to
>> the commit. Therefore, remove redundant checking logic.
>>
>> Fixes: 603ccef42ce9 ("KVM: x86: SVM: fix avic_kick_target_vcpus_fast")
>> Cc: Maxim Levitsky <[email protected]>
>> Signed-off-by: Suravee Suthikulpanit <[email protected]>
>> ---
>> arch/x86/kvm/svm/avic.c | 19 ++++---------------
>> 1 file changed, 4 insertions(+), 15 deletions(-)
>>
>> diff --git a/arch/x86/kvm/svm/avic.c b/arch/x86/kvm/svm/avic.c
>> index 6919dee69f18..45ab49d1f0b8 100644
>> --- a/arch/x86/kvm/svm/avic.c
>> +++ b/arch/x86/kvm/svm/avic.c
>> @@ -378,8 +378,8 @@ static int avic_kick_target_vcpus_fast(struct kvm *kvm, struct kvm_lapic *source
>>
>> if (apic_x2apic_mode(source)) {
>> /* 16 bit dest mask, 16 bit cluster id */
>> - bitmap = dest & 0xFFFF0000;
>> - cluster = (dest >> 16) << 4;
>> + bitmap = dest & 0xffff;
>> + cluster = (dest & 0xffff0000) >> 16;
>> } else if (kvm_lapic_get_reg(source, APIC_DFR) == APIC_DFR_FLAT) {
>> /* 8 bit dest mask*/
>> bitmap = dest;
>> @@ -387,7 +387,7 @@ static int avic_kick_target_vcpus_fast(struct kvm *kvm, struct kvm_lapic *source
>> } else {
>> /* 4 bit desk mask, 4 bit cluster id */
>> bitmap = dest & 0xF;
>> - cluster = (dest >> 4) << 2;
>> + cluster = (dest & 0xf0) >> 4;
>
> This is wrong and unrelated. The cluster needs to be shifted back by 2, i.e. multiplied
> by 4, to leap over each cluster of 4 IDs.
Ah, I missed the logic that was using the _cluster_ value below.
Thanks,
Suravee