linux-kernel - Re: [PATCH] KVM: SVM: Fix x2apic logical cluster mode decoding and sanity check

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <7c43a6f2-ec11-5af2-6574-68a32eb7fd41@amd.com>
Date:   Thu, 15 Sep 2022 17:47:26 -0500
From:   "Suthikulpanit, Suravee" <suravee.suthikulpanit@....com>
To:     Sean Christopherson <seanjc@...gle.com>
Cc:     linux-kernel@...r.kernel.org, kvm@...r.kernel.org,
        pbonzini@...hat.com, jon.grimm@....com,
        Maxim Levitsky <mlevitsk@...hat.com>
Subject: Re: [PATCH] KVM: SVM: Fix x2apic logical cluster mode decoding and
 sanity check



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 <mlevitsk@...hat.com>
>> Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@....com>
>> ---
>>   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