From: Vitaly Kuznetsov <vkuznets@redhat.com>
To: Yuan Yao <yuan.yao@linux.intel.com>
Cc: kvm@vger.kernel.org, Paolo Bonzini <pbonzini@redhat.com>,
Sean Christopherson <seanjc@google.com>,
Maxim Levitsky <mlevitsk@redhat.com>,
Wanpeng Li <wanpengli@tencent.com>,
Jim Mattson <jmattson@google.com>,
Michael Kelley <mikelley@microsoft.com>,
Siddharth Chandrasekaran <sidcha@amazon.de>,
linux-hyperv@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4 05/37] KVM: x86: hyper-v: Handle HVCALL_FLUSH_VIRTUAL_ADDRESS_LIST{,EX} calls gently
Date: Fri, 27 May 2022 13:43:09 +0200 [thread overview]
Message-ID: <87ilprqjzm.fsf@redhat.com> (raw)
In-Reply-To: <20220527083930.okdenkvxephom5wq@yy-desk-7060>
Yuan Yao <yuan.yao@linux.intel.com> writes:
> On Wed, May 25, 2022 at 11:01:01AM +0200, Vitaly Kuznetsov wrote:
>> Currently, HVCALL_FLUSH_VIRTUAL_ADDRESS_LIST{,EX} calls are handled
>> the exact same way as HVCALL_FLUSH_VIRTUAL_ADDRESS_SPACE{,EX}: by
>> flushing the whole VPID and this is sub-optimal. Switch to handling
>> these requests with 'flush_tlb_gva()' hooks instead. Use the newly
>> introduced TLB flush fifo to queue the requests.
>>
>> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
>> ---
>> arch/x86/kvm/hyperv.c | 102 +++++++++++++++++++++++++++++++++++++-----
>> 1 file changed, 90 insertions(+), 12 deletions(-)
>>
>> diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c
>> index 762b0b699fdf..576749973727 100644
>> --- a/arch/x86/kvm/hyperv.c
>> +++ b/arch/x86/kvm/hyperv.c
>> @@ -1806,32 +1806,84 @@ static u64 kvm_get_sparse_vp_set(struct kvm *kvm, struct kvm_hv_hcall *hc,
>> sparse_banks, consumed_xmm_halves, offset);
>> }
>>
>> -static void hv_tlb_flush_enqueue(struct kvm_vcpu *vcpu)
>> +static int kvm_hv_get_tlb_flush_entries(struct kvm *kvm, struct kvm_hv_hcall *hc, u64 entries[],
>> + int consumed_xmm_halves, gpa_t offset)
>> +{
>> + return kvm_hv_get_hc_data(kvm, hc, hc->rep_cnt, hc->rep_cnt,
>> + entries, consumed_xmm_halves, offset);
>> +}
>> +
>> +static void hv_tlb_flush_enqueue(struct kvm_vcpu *vcpu, u64 *entries, int count)
>> {
>> struct kvm_vcpu_hv_tlb_flush_fifo *tlb_flush_fifo;
>> struct kvm_vcpu_hv *hv_vcpu = to_hv_vcpu(vcpu);
>> u64 entry = KVM_HV_TLB_FLUSHALL_ENTRY;
>> + unsigned long flags;
>>
>> if (!hv_vcpu)
>> return;
>>
>> tlb_flush_fifo = &hv_vcpu->tlb_flush_fifo;
>>
>> - kfifo_in_spinlocked(&tlb_flush_fifo->entries, &entry, 1, &tlb_flush_fifo->write_lock);
>> + spin_lock_irqsave(&tlb_flush_fifo->write_lock, flags);
>> +
>> + /*
>> + * All entries should fit on the fifo leaving one free for 'flush all'
>> + * entry in case another request comes in. In case there's not enough
>> + * space, just put 'flush all' entry there.
>> + */
>> + if (count && entries && count < kfifo_avail(&tlb_flush_fifo->entries)) {
>> + WARN_ON(kfifo_in(&tlb_flush_fifo->entries, entries, count) != count);
>> + goto out_unlock;
>> + }
>> +
>> + /*
>> + * Note: full fifo always contains 'flush all' entry, no need to check the
>> + * return value.
>> + */
>> + kfifo_in(&tlb_flush_fifo->entries, &entry, 1);
>> +
>> +out_unlock:
>> + spin_unlock_irqrestore(&tlb_flush_fifo->write_lock, flags);
>> }
>>
>> void kvm_hv_vcpu_flush_tlb(struct kvm_vcpu *vcpu)
>
> Where's the caller to this kvm_hv_vcpu_flush_tlb() ?
> I didn't see th caller in patch 1-22 and remains are
> self-testing patches, any thing I missed ?
No, I screwed up. It was present in
kvm_service_local_tlb_flush_requests() in previous versions but somehow
disappeared. This also means that I haven't tested this properly.
Thanks for catching this! v5 is coming to rescue!
>
>> {
>> struct kvm_vcpu_hv_tlb_flush_fifo *tlb_flush_fifo;
>> struct kvm_vcpu_hv *hv_vcpu = to_hv_vcpu(vcpu);
>> + u64 entries[KVM_HV_TLB_FLUSH_FIFO_SIZE];
>> + int i, j, count;
>> + gva_t gva;
>>
>> - kvm_vcpu_flush_tlb_guest(vcpu);
>> -
>> - if (!hv_vcpu)
>> + if (!tdp_enabled || !hv_vcpu) {
>> + kvm_vcpu_flush_tlb_guest(vcpu);
>> return;
>> + }
>>
>> tlb_flush_fifo = &hv_vcpu->tlb_flush_fifo;
>>
>> + count = kfifo_out(&tlb_flush_fifo->entries, entries, KVM_HV_TLB_FLUSH_FIFO_SIZE);
>> +
>> + for (i = 0; i < count; i++) {
>> + if (entries[i] == KVM_HV_TLB_FLUSHALL_ENTRY)
>> + goto out_flush_all;
>> +
>> + /*
>> + * Lower 12 bits of 'address' encode the number of additional
>> + * pages to flush.
>> + */
>> + gva = entries[i] & PAGE_MASK;
>> + for (j = 0; j < (entries[i] & ~PAGE_MASK) + 1; j++)
>> + static_call(kvm_x86_flush_tlb_gva)(vcpu, gva + j * PAGE_SIZE);
>> +
>> + ++vcpu->stat.tlb_flush;
>> + }
>> + goto out_empty_ring;
>> +
>> +out_flush_all:
>> + kvm_vcpu_flush_tlb_guest(vcpu);
>> +
>> +out_empty_ring:
>> kfifo_reset_out(&tlb_flush_fifo->entries);
>> }
>>
>> @@ -1841,11 +1893,21 @@ static u64 kvm_hv_flush_tlb(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc)
>> struct hv_tlb_flush_ex flush_ex;
>> struct hv_tlb_flush flush;
>> DECLARE_BITMAP(vcpu_mask, KVM_MAX_VCPUS);
>> + /*
>> + * Normally, there can be no more than 'KVM_HV_TLB_FLUSH_FIFO_SIZE'
>> + * entries on the TLB flush fifo. The last entry, however, needs to be
>> + * always left free for 'flush all' entry which gets placed when
>> + * there is not enough space to put all the requested entries.
>> + */
>> + u64 __tlb_flush_entries[KVM_HV_TLB_FLUSH_FIFO_SIZE - 1];
>> + u64 *tlb_flush_entries;
>> u64 valid_bank_mask;
>> u64 sparse_banks[KVM_HV_MAX_SPARSE_VCPU_SET_BITS];
>> struct kvm_vcpu *v;
>> unsigned long i;
>> bool all_cpus;
>> + int consumed_xmm_halves = 0;
>> + gpa_t data_offset;
>>
>> /*
>> * The Hyper-V TLFS doesn't allow more than 64 sparse banks, e.g. the
>> @@ -1861,10 +1923,12 @@ static u64 kvm_hv_flush_tlb(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc)
>> flush.address_space = hc->ingpa;
>> flush.flags = hc->outgpa;
>> flush.processor_mask = sse128_lo(hc->xmm[0]);
>> + consumed_xmm_halves = 1;
>> } else {
>> if (unlikely(kvm_read_guest(kvm, hc->ingpa,
>> &flush, sizeof(flush))))
>> return HV_STATUS_INVALID_HYPERCALL_INPUT;
>> + data_offset = sizeof(flush);
>> }
>>
>> trace_kvm_hv_flush_tlb(flush.processor_mask,
>> @@ -1888,10 +1952,12 @@ static u64 kvm_hv_flush_tlb(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc)
>> flush_ex.flags = hc->outgpa;
>> memcpy(&flush_ex.hv_vp_set,
>> &hc->xmm[0], sizeof(hc->xmm[0]));
>> + consumed_xmm_halves = 2;
>> } else {
>> if (unlikely(kvm_read_guest(kvm, hc->ingpa, &flush_ex,
>> sizeof(flush_ex))))
>> return HV_STATUS_INVALID_HYPERCALL_INPUT;
>> + data_offset = sizeof(flush_ex);
>> }
>>
>> trace_kvm_hv_flush_tlb_ex(flush_ex.hv_vp_set.valid_bank_mask,
>> @@ -1907,25 +1973,37 @@ static u64 kvm_hv_flush_tlb(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc)
>> return HV_STATUS_INVALID_HYPERCALL_INPUT;
>>
>> if (all_cpus)
>> - goto do_flush;
>> + goto read_flush_entries;
>>
>> if (!hc->var_cnt)
>> goto ret_success;
>>
>> - if (kvm_get_sparse_vp_set(kvm, hc, sparse_banks, 2,
>> - offsetof(struct hv_tlb_flush_ex,
>> - hv_vp_set.bank_contents)))
>> + if (kvm_get_sparse_vp_set(kvm, hc, sparse_banks, consumed_xmm_halves,
>> + data_offset))
>> + return HV_STATUS_INVALID_HYPERCALL_INPUT;
>> + data_offset += hc->var_cnt * sizeof(sparse_banks[0]);
>> + consumed_xmm_halves += hc->var_cnt;
>> + }
>> +
>> +read_flush_entries:
>> + if (hc->code == HVCALL_FLUSH_VIRTUAL_ADDRESS_SPACE ||
>> + hc->code == HVCALL_FLUSH_VIRTUAL_ADDRESS_SPACE_EX ||
>> + hc->rep_cnt > ARRAY_SIZE(__tlb_flush_entries)) {
>> + tlb_flush_entries = NULL;
>> + } else {
>> + if (kvm_hv_get_tlb_flush_entries(kvm, hc, __tlb_flush_entries,
>> + consumed_xmm_halves, data_offset))
>> return HV_STATUS_INVALID_HYPERCALL_INPUT;
>> + tlb_flush_entries = __tlb_flush_entries;
>> }
>>
>> -do_flush:
>> /*
>> * vcpu->arch.cr3 may not be up-to-date for running vCPUs so we can't
>> * analyze it here, flush TLB regardless of the specified address space.
>> */
>> if (all_cpus) {
>> kvm_for_each_vcpu(i, v, kvm)
>> - hv_tlb_flush_enqueue(v);
>> + hv_tlb_flush_enqueue(v, tlb_flush_entries, hc->rep_cnt);
>>
>> kvm_make_all_cpus_request(kvm, KVM_REQ_HV_TLB_FLUSH);
>> } else {
>> @@ -1935,7 +2013,7 @@ static u64 kvm_hv_flush_tlb(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc)
>> v = kvm_get_vcpu(kvm, i);
>> if (!v)
>> continue;
>> - hv_tlb_flush_enqueue(v);
>> + hv_tlb_flush_enqueue(v, tlb_flush_entries, hc->rep_cnt);
>> }
>>
>> kvm_make_vcpus_request_mask(kvm, KVM_REQ_HV_TLB_FLUSH, vcpu_mask);
>> --
>> 2.35.3
>>
>
--
Vitaly
next prev parent reply other threads:[~2022-05-27 11:48 UTC|newest]
Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-25 9:00 [PATCH v4 00/37] KVM: x86: hyper-v: Fine-grained TLB flush + L2 TLB flush features Vitaly Kuznetsov
2022-05-25 9:00 ` [PATCH v4 01/37] KVM: x86: Rename 'enable_direct_tlbflush' to 'enable_l2_tlb_flush' Vitaly Kuznetsov
2022-05-25 9:00 ` [PATCH v4 02/37] KVM: x86: hyper-v: Resurrect dedicated KVM_REQ_HV_TLB_FLUSH flag Vitaly Kuznetsov
2022-05-27 3:00 ` Yuan Yao
2022-05-25 9:00 ` [PATCH v4 03/37] KVM: x86: hyper-v: Introduce TLB flush fifo Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 04/37] KVM: x86: hyper-v: Add helper to read hypercall data for array Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 05/37] KVM: x86: hyper-v: Handle HVCALL_FLUSH_VIRTUAL_ADDRESS_LIST{,EX} calls gently Vitaly Kuznetsov
2022-05-27 8:39 ` Yuan Yao
2022-05-27 11:43 ` Vitaly Kuznetsov [this message]
2022-05-25 9:01 ` [PATCH v4 06/37] KVM: x86: hyper-v: Expose support for extended gva ranges for flush hypercalls Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 07/37] KVM: x86: Prepare kvm_hv_flush_tlb() to handle L2's GPAs Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 08/37] x86/hyperv: Introduce HV_MAX_SPARSE_VCPU_BANKS/HV_VCPUS_PER_SPARSE_BANK constants Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 09/37] KVM: x86: hyper-v: Use HV_MAX_SPARSE_VCPU_BANKS/HV_VCPUS_PER_SPARSE_BANK instead of raw '64' Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 10/37] KVM: x86: hyper-v: Don't use sparse_set_to_vcpu_mask() in kvm_hv_send_ipi() Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 11/37] KVM: x86: hyper-v: Create a separate fifo for L2 TLB flush Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 12/37] KVM: x86: hyper-v: Use preallocated buffer in 'struct kvm_vcpu_hv' instead of on-stack 'sparse_banks' Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 13/37] KVM: nVMX: Keep track of hv_vm_id/hv_vp_id when eVMCS is in use Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 14/37] KVM: nSVM: Keep track of Hyper-V hv_vm_id/hv_vp_id Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 15/37] KVM: x86: Introduce .hv_inject_synthetic_vmexit_post_tlb_flush() nested hook Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 16/37] KVM: x86: hyper-v: Introduce kvm_hv_is_tlb_flush_hcall() Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 17/37] KVM: x86: hyper-v: L2 TLB flush Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 18/37] KVM: x86: hyper-v: Introduce fast guest_hv_cpuid_has_l2_tlb_flush() check Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 19/37] x86/hyperv: Fix 'struct hv_enlightened_vmcs' definition Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 20/37] KVM: nVMX: hyper-v: Enable L2 TLB flush Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 21/37] KVM: nSVM: " Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 22/37] KVM: x86: Expose Hyper-V L2 TLB flush feature Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 23/37] KVM: selftests: Better XMM read/write helpers Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 24/37] KVM: selftests: Move HYPERV_LINUX_OS_ID definition to a common header Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 25/37] KVM: selftests: Move the function doing Hyper-V hypercall " Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 26/37] KVM: selftests: Hyper-V PV IPI selftest Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 27/37] KVM: selftests: Fill in vm->vpages_mapped bitmap in virt_map() too Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 28/37] KVM: selftests: Export vm_vaddr_unused_gap() to make it possible to request unmapped ranges Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 29/37] KVM: selftests: Export _vm_get_page_table_entry() and struct pageTableEntry/pageUpperEntry definitions Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 30/37] KVM: selftests: Hyper-V PV TLB flush selftest Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 31/37] KVM: selftests: Sync 'struct hv_enlightened_vmcs' definition with hyperv-tlfs.h Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 32/37] KVM: selftests: nVMX: Allocate Hyper-V partition assist page Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 33/37] KVM: selftests: nSVM: Allocate Hyper-V partition assist and VP assist pages Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 34/37] KVM: selftests: Sync 'struct hv_vp_assist_page' definition with hyperv-tlfs.h Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 35/37] KVM: selftests: evmcs_test: Introduce L2 TLB flush test Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 36/37] KVM: selftests: Move Hyper-V VP assist page enablement out of evmcs.h Vitaly Kuznetsov
2022-05-25 9:01 ` [PATCH v4 37/37] KVM: selftests: hyperv_svm_test: Introduce L2 TLB flush test Vitaly Kuznetsov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=87ilprqjzm.fsf@redhat.com \
--to=vkuznets@redhat.com \
--cc=jmattson@google.com \
--cc=kvm@vger.kernel.org \
--cc=linux-hyperv@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mikelley@microsoft.com \
--cc=mlevitsk@redhat.com \
--cc=pbonzini@redhat.com \
--cc=seanjc@google.com \
--cc=sidcha@amazon.de \
--cc=wanpengli@tencent.com \
--cc=yuan.yao@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).