KVM: SVM: Create separate vmcbs for L1 and L2
v3 cover KVM: SVM: Create separate vmcbs for L1 and L2
svm->vmcb will now point to either a separate vmcb L1 ( not nested ) or L2 vmcb ( nested ).
Changes:
v2 -> v3
- Added vmcb switching helper.
- svm_set_nested_state always forces to L1 before determining state
to set. This is more like vmx and covers any potential L2 to L2 nested state switch.
- Moved svm->asid tracking to pre_svm_run and added ASID set dirty bit checking.
v1 -> v2
- Removed unnecessary update check of L1 save.cr3 during nested_svm_vmexit.
- Moved vmcb01_pa to svm.
- Removed get_host_vmcb() function.
- Updated vmsave/vmload corresponding vmcb state during L2 enter and exit which fixed the L2 load issue.
- Moved asid workaround to a new patch which adds asid to svm.
- Init previously uninitialized L2 vmcb save.gpat and save.cr4
Tested:
kvm-unit-tests
kvm self tests
Loaded fedora nested guest on fedora
Cathy Avery (2):
KVM: SVM: Track asid from vcpu_svm
KVM: SVM: Use a separate vmcb for the nested L2 guest
arch/x86/kvm/svm/nested.c | 125 ++++++++++++++++++-------------------- arch/x86/kvm/svm/svm.c | 58 +++++++++++------- arch/x86/kvm/svm/svm.h | 51 +++++----------- 3 files changed, 110 insertions(+), 124 deletions(-)
Signature · 1 line
2.20.1
PATCH v3 1/2 KVM: SVM: Track asid from vcpu_svm
Track asid from svm->asid to allow for vmcb assignment without regard to which level guest is running.
--- arch/x86/kvm/svm/svm.c | 16 ++++++++++++++-- arch/x86/kvm/svm/svm.h | 2 ++ 2 files changed, 16 insertions(+), 2 deletions(-)
arch/x86/kvm/svm/svm.c +14 -2
@@ -1101,6 +1101,7 @@ static void init_vmcb(struct vcpu_svm *svm) save->cr4 =0 ; } svm->asid_generation =0 ; + svm->asid =0 ; svm->nested.vmcb =0 ; svm->vcpu.arch.hflags =0 ;
@@ -1659,11 +1660,11 @@ static void new_asid(struct vcpu_svm *svm, struct svm_cpu_data *sd)if (sd->next_asid > sd->max_asid) { ++sd->asid_generation; sd->next_asid = sd->min_asid; - svm->vmcb->control.tlb_ctl =TLB_CONTROL_FLUSH_ALL_ASID ; + sd->flush_all_asid = true; } svm->asid_generation = sd->asid_generation; - svm->vmcb->control.asid = sd->next_asid++; + svm->asid = sd->next_asid++;vmcb_mark_dirty (svm->vmcb,VMCB_ASID ); }
@@ -3030,6 +3031,17 @@ static void pre_svm_run(struct vcpu_svm *svm)/* FIXME: handle wraparound of asid_generation */ if (svm->asid_generation != sd->asid_generation)new_asid (svm, sd); + +if (sd->flush_all_asid) { + svm->vmcb->control.tlb_ctl =TLB_CONTROL_FLUSH_ALL_ASID ; + sd->flush_all_asid = false; +vmcb_mark_dirty (svm->vmcb,VMCB_ASID ); + } + +if (unlikely (svm->asid != svm->vmcb->control.asid)) +vmcb_mark_dirty (svm->vmcb,VMCB_ASID ); + + svm->vmcb->control.asid = svm->asid; }static void svm_inject_nmi (struct kvm_vcpu *vcpu)
arch/x86/kvm/svm/svm.h +2 -0
@@ -104,6 +104,7 @@ struct vcpu_svm {struct vmcb *vmcb;unsigned long vmcb_pa;struct svm_cpu_data *svm_data; +u32 asid;uint64_t asid_generation;uint64_t sysenter_esp;uint64_t sysenter_eip;
@@ -164,6 +165,7 @@ struct svm_cpu_data {int cpu;u64 asid_generation; +bool flush_all_asid;u32 max_asid;u32 next_asid;u32 min_asid;
Signature · 1 line
2.20.1
PATCH v3 2/2 KVM: SVM: Use a separate vmcb for the nested L2 guest
In reply to Cathy Averysvm->vmcb will now point to either a separate vmcb L1 ( not nested ) or L2 vmcb ( nested ).
Issues:
1) There is some wholesale copying of vmcb.save and vmcb.contol areas which will need to be refined.
Tested:
kvm-unit-tests
kvm self tests
Loaded fedora nested guest on fedora
--- arch/x86/kvm/svm/nested.c | 125 ++++++++++++++++++-------------------- arch/x86/kvm/svm/svm.c | 42 +++++++------ arch/x86/kvm/svm/svm.h | 49 ++++----------- 3 files changed, 94 insertions(+), 122 deletions(-)
arch/x86/kvm/svm/nested.c +58 -67
@@ -75,12 +75,12 @@ static unsigned long nested_svm_get_tdp_cr3(struct kvm_vcpu *vcpu)static void nested_svm_init_mmu_context (struct kvm_vcpu *vcpu) {struct vcpu_svm *svm =to_svm (vcpu); -struct vmcb *hsave = svm->nested.hsave;WARN_ON (mmu_is_nested (vcpu)); vcpu->arch.mmu = &vcpu->arch.guest_mmu; -kvm_init_shadow_npt_mmu (vcpu,X86_CR0_PG , hsave->save.cr4, hsave->save.efer, +kvm_init_shadow_npt_mmu (vcpu,X86_CR0_PG , svm->vmcb01->save.cr4, + svm->vmcb01->save.efer, svm->nested.ctl.nested_cr3); vcpu->arch.mmu->get_guest_pgd = nested_svm_get_tdp_cr3; vcpu->arch.mmu->get_pdptr = nested_svm_get_tdp_pdptr;
@@ -105,7 +105,7 @@ void recalc_intercepts(struct vcpu_svm *svm)return ; c = &svm->vmcb->control; - h = &svm->nested.hsave->control; + h = &svm->vmcb01->control; g = &svm->nested.ctl; svm->nested.host_intercept_exceptions = h->intercept_exceptions;
@@ -403,7 +403,7 @@ static void nested_prepare_vmcb_control(struct vcpu_svm *svm) svm->vmcb->control.int_ctl = (svm->nested.ctl.int_ctl & ~mask) | - (svm->nested.hsave->control.int_ctl & mask); + (svm->vmcb01->control.int_ctl & mask); svm->vmcb->control.virt_ext = svm->nested.ctl.virt_ext; svm->vmcb->control.int_vector = svm->nested.ctl.int_vector;
@@ -426,12 +426,29 @@ static void nested_prepare_vmcb_control(struct vcpu_svm *svm)vmcb_mark_all_dirty (svm->vmcb); } +static void svm_switch_vmcb (struct vcpu_svm *svm,struct vmcb *vmcb,unsigned long vmcb_pa) +{ +if (svm->vmcb == vmcb) +return ; + + svm->vmcb = vmcb; + svm->vmcb_pa = vmcb_pa; +} +int enter_svm_guest_mode (struct vcpu_svm *svm,u64 vmcb_gpa,struct vmcb *nested_vmcb) {int ret; svm->nested.vmcb = vmcb_gpa; + +WARN_ON (svm->vmcb == svm->nested.vmcb02); + + svm->nested.vmcb02->control = svm->vmcb01->control; + svm->nested.vmcb02->save.cr4 = svm->vmcb01->save.cr4; + +nested_svm_vmloadsave (svm->vmcb01, svm->nested.vmcb02); +svm_switch_vmcb (svm, svm->nested.vmcb02, svm->nested.vmcb02_pa);load_nested_vmcb_control (svm, &nested_vmcb->control);nested_prepare_vmcb_save (svm, nested_vmcb);nested_prepare_vmcb_control (svm);
@@ -450,8 +467,6 @@ int nested_svm_vmrun(struct vcpu_svm *svm) {int ret;struct vmcb *nested_vmcb; -struct vmcb *hsave = svm->nested.hsave; -struct vmcb *vmcb = svm->vmcb;struct kvm_host_map map;u64 vmcb_gpa;
@@ -496,29 +511,14 @@ int nested_svm_vmrun(struct vcpu_svm *svm)kvm_clear_exception_queue (&svm->vcpu);kvm_clear_interrupt_queue (&svm->vcpu); -/* -* Save the old vmcb, so we don't need to pick what we save, but can -* restore everything when a VMEXIT occurs -*/ - hsave->save.es = vmcb->save.es; - hsave->save.cs = vmcb->save.cs; - hsave->save.ss = vmcb->save.ss; - hsave->save.ds = vmcb->save.ds; - hsave->save.gdtr = vmcb->save.gdtr; - hsave->save.idtr = vmcb->save.idtr; - hsave->save.efer = svm->vcpu.arch.efer; - hsave->save.cr0 =kvm_read_cr0 (&svm->vcpu); - hsave->save.cr4 = svm->vcpu.arch.cr4; - hsave->save.rflags =kvm_get_rflags (&svm->vcpu); - hsave->save.rip =kvm_rip_read (&svm->vcpu); - hsave->save.rsp = vmcb->save.rsp; - hsave->save.rax = vmcb->save.rax; -if (npt_enabled) - hsave->save.cr3 = vmcb->save.cr3; -else - hsave->save.cr3 =kvm_read_cr3 (&svm->vcpu); - -copy_vmcb_control_area (&hsave->control, &vmcb->control); + svm->vmcb01->save.efer = svm->vcpu.arch.efer; + svm->vmcb01->save.cr0 =kvm_read_cr0 (&svm->vcpu); + svm->vmcb01->save.cr4 = svm->vcpu.arch.cr4; + svm->vmcb01->save.rflags =kvm_get_rflags (&svm->vcpu); + svm->vmcb01->save.rip =kvm_rip_read (&svm->vcpu); + +if (!npt_enabled) + svm->vmcb01->save.cr3 =kvm_read_cr3 (&svm->vcpu); svm->nested.nested_run_pending =1 ;
@@ -564,7 +564,6 @@ int nested_svm_vmexit(struct vcpu_svm *svm) {int rc;struct vmcb *nested_vmcb; -struct vmcb *hsave = svm->nested.hsave;struct vmcb *vmcb = svm->vmcb;struct kvm_host_map map;
@@ -628,8 +627,8 @@ int nested_svm_vmexit(struct vcpu_svm *svm) nested_vmcb->control.pause_filter_thresh = svm->vmcb->control.pause_filter_thresh; -/* Restore the original control entries */ -copy_vmcb_control_area (&vmcb->control, &hsave->control); +nested_svm_vmloadsave (svm->nested.vmcb02, svm->vmcb01); +svm_switch_vmcb (svm, svm->vmcb01, svm->vmcb01_pa);/* On vmexit the GIF is set to false */ svm_set_gif (svm, false);
@@ -640,19 +639,13 @@ int nested_svm_vmexit(struct vcpu_svm *svm) svm->nested.ctl.nested_cr3 =0 ;/* Restore selected save entries */ - svm->vmcb->save.es = hsave->save.es; - svm->vmcb->save.cs = hsave->save.cs; - svm->vmcb->save.ss = hsave->save.ss; - svm->vmcb->save.ds = hsave->save.ds; - svm->vmcb->save.gdtr = hsave->save.gdtr; - svm->vmcb->save.idtr = hsave->save.idtr; -kvm_set_rflags (&svm->vcpu, hsave->save.rflags); -svm_set_efer (&svm->vcpu, hsave->save.efer); -svm_set_cr0 (&svm->vcpu, hsave->save.cr0 |X86_CR0_PE ); -svm_set_cr4 (&svm->vcpu, hsave->save.cr4); -kvm_rax_write (&svm->vcpu, hsave->save.rax); -kvm_rsp_write (&svm->vcpu, hsave->save.rsp); -kvm_rip_write (&svm->vcpu, hsave->save.rip); +kvm_set_rflags (&svm->vcpu, svm->vmcb->save.rflags); +svm_set_efer (&svm->vcpu, svm->vmcb->save.efer); +svm_set_cr0 (&svm->vcpu, svm->vmcb->save.cr0 |X86_CR0_PE ); +svm_set_cr4 (&svm->vcpu, svm->vmcb->save.cr4); +kvm_rax_write (&svm->vcpu, svm->vmcb->save.rax); +kvm_rsp_write (&svm->vcpu, svm->vmcb->save.rsp); +kvm_rip_write (&svm->vcpu, svm->vmcb->save.rip); svm->vmcb->save.dr7 =0 ; svm->vmcb->save.cpl =0 ; svm->vmcb->control.exit_int_info =0 ;
@@ -670,13 +663,10 @@ int nested_svm_vmexit(struct vcpu_svm *svm)nested_svm_uninit_mmu_context (&svm->vcpu); - rc =nested_svm_load_cr3 (&svm->vcpu, hsave->save.cr3, false); + rc =nested_svm_load_cr3 (&svm->vcpu, svm->vmcb->save.cr3, false);if (rc)return 1 ; -if (npt_enabled) - svm->vmcb->save.cr3 = hsave->save.cr3; -/* * Drop what we picked up for L2 via svm_complete_interrupts() so it * doesn't end up in L1.
@@ -694,12 +684,9 @@ int nested_svm_vmexit(struct vcpu_svm *svm)void svm_leave_nested (struct vcpu_svm *svm) {if (is_guest_mode (&svm->vcpu)) { -struct vmcb *hsave = svm->nested.hsave; -struct vmcb *vmcb = svm->vmcb; - svm->nested.nested_run_pending =0 ;leave_guest_mode (&svm->vcpu); -copy_vmcb_control_area (&vmcb->control, &hsave->control); +svm_switch_vmcb (svm, svm->vmcb01, svm->vmcb01_pa);nested_svm_uninit_mmu_context (&svm->vcpu); } }
@@ -982,7 +969,7 @@ int nested_svm_exit_special(struct vcpu_svm *svm) case SVM_EXIT_EXCP_BASE ... SVM_EXIT_EXCP_BASE + 0x1f: {u32 excp_bits =1 << (exit_code -SVM_EXIT_EXCP_BASE ); -if (get_host_vmcb (svm)->control.intercept_exceptions & excp_bits) +if (svm->vmcb01->control.intercept_exceptions & excp_bits)return NESTED_EXIT_HOST ;else if (exit_code ==SVM_EXIT_EXCP_BASE +PF_VECTOR && svm->vcpu.arch.apf.host_apf_flags)
@@ -1046,10 +1033,9 @@ static int svm_get_nested_state(struct kvm_vcpu *vcpu,if (copy_to_user (&user_vmcb->control, &svm->nested.ctl,sizeof (user_vmcb->control)))return -EFAULT ; -if (copy_to_user (&user_vmcb->save, &svm->nested.hsave->save, +if (copy_to_user (&user_vmcb->save, &svm->vmcb01->save,sizeof (user_vmcb->save)))return -EFAULT ; - out:return kvm_state.size; }
@@ -1059,7 +1045,6 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu, struct kvm_nested_state *kvm_state) {struct vcpu_svm *svm =to_svm (vcpu); -struct vmcb *hsave = svm->nested.hsave;struct vmcb __user *user_vmcb = (struct vmcb __user *) &user_kvm_nested_state->data.svm[0 ];struct vmcb_control_area ctl;
@@ -1088,10 +1073,12 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,if (is_smm (vcpu) && (kvm_state->flags &KVM_STATE_NESTED_GUEST_MODE ))return -EINVAL ; -if (!(kvm_state->flags &KVM_STATE_NESTED_GUEST_MODE )) { -svm_leave_nested (svm); +/* If in L2 force to L1 */ + +svm_leave_nested (svm); + +if (!(kvm_state->flags &KVM_STATE_NESTED_GUEST_MODE )) goto out_set_gif; - }if (!page_address_valid (vcpu, kvm_state->hdr.svm.vmcb_pa))return -EINVAL ;
@@ -1121,16 +1108,20 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,if (!(save.cr0 &X86_CR0_PG ))return -EINVAL ; + svm->nested.vmcb02->control = svm->vmcb01->control; + svm->nested.vmcb02->save = svm->vmcb01->save; + svm->vmcb01->save = save; + + svm->nested.vmcb = kvm_state->hdr.svm.vmcb_pa; +svm_switch_vmcb (svm, svm->nested.vmcb02, svm->nested.vmcb02_pa); +/* -* All checks done, we can enter guest mode. L1 control fields -* come from the nested save state. Guest state is already -* in the registers, the save area of the nested state instead -* contains saved L1 state. +* All checks done, we can enter guest mode. L2 control fields will +* be the result of a combination of L1 and userspace indicated +* L12.control. The save area of L1 vmcb now contains the userspace +* indicated L1.save. */ -copy_vmcb_control_area (&hsave->control, &svm->vmcb->control); - hsave->save = save; - svm->nested.vmcb = kvm_state->hdr.svm.vmcb_pa;load_nested_vmcb_control (svm, &ctl);nested_prepare_vmcb_control (svm);
arch/x86/kvm/svm/svm.c +23 -19
117 lines not on this page · raw message
arch/x86/kvm/svm/svm.h +13 -36
115 lines not on this page · raw message
Signature · 1 line
2.20.1
Hi Cathy,
I was going to test these patches. But it did not apply on my tree. Tried on kvm(https://git.kernel.org/pub/scm/virt/kvm/kvm.git) and Mainline
(https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git). What is your base tree?
thanks
Babu
Quoted 45 lines
-----Original Message-----
From: Cathy Avery <[email protected]>
Sent: Monday, October 26, 2020 12:42 PM
To: [email protected]; [email protected]; [email protected]
Cc: [email protected]; Huang2, Wei <[email protected]>; [email protected]; [email protected]
Subject: [PATCH v3 0/2] KVM: SVM: Create separate vmcbs for L1 and L2
svm->vmcbwill now point to either a separate vmcb L1 ( not nested ) or L2 vmcb ( nested ).Changes:
v2 -> v3
- Added vmcb switching helper.
-svm_set_nested_statealways forces to L1 before determining state to set. This is more like vmx and covers any potential L2 to L2 nested state
switch.
- Movedsvm->asidtracking topre_svm_runand added ASID set dirty bit checking.v1 -> v2
- Removed unnecessary update check of L1 save.cr3 duringnested_svm_vmexit.
- Movedvmcb01_pato svm.
- Removedget_host_vmcb()function.
- Updated vmsave/vmload corresponding vmcb state during L2 enter and exit which fixed the L2 load issue.
- Moved asid workaround to a new patch which adds asid to svm.
- Init previously uninitialized L2 vmcb save.gpat and save.cr4Tested:
kvm-unit-tests
kvm self tests
Loaded fedora nested guest on fedoraCathy Avery (2):
KVM: SVM: Track asid fromvcpu_svm
KVM: SVM: Use a separate vmcb for the nested L2 guestarch/x86/kvm/svm/nested.c | 125 ++++++++++++++++++-------------------- arch/x86/kvm/svm/svm.c | 58 +++++++++++------- arch/x86/kvm/svm/svm.h | 51 +++++----------- 3 files changed, 110 insertions(+), 124 deletions(-)--
2.20.1
That would be the master branch of git://git.kernel.org/pub/scm/virt/kvm/kvm.git where the last commit was 969df928fee43b4219646a57c7beaccccf2c0635
I was originally working off of the queue branch but there were issues with the prior commits passing the various tests.
Cathy
Quoted 54 lines · Babu Moger
Hi Cathy,
I was going to test these patches. But it did not apply on my tree. Tried on kvm(https://git.kernel.org/pub/scm/virt/kvm/kvm.git) and Mainline
(https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git). What is your base tree?
thanks
BabuQuoted 45 lines
-----Original Message-----
From: Cathy Avery <[email protected]>
Sent: Monday, October 26, 2020 12:42 PM
To: [email protected]; [email protected]; [email protected]
Cc: [email protected]; Huang2, Wei <[email protected]>; [email protected]; [email protected]
Subject: [PATCH v3 0/2] KVM: SVM: Create separate vmcbs for L1 and L2
svm->vmcbwill now point to either a separate vmcb L1 ( not nested ) or L2 vmcb ( nested ).Changes:
v2 -> v3
- Added vmcb switching helper.
-svm_set_nested_statealways forces to L1 before determining state to set. This is more like vmx and covers any potential L2 to L2 nested state
switch.
- Movedsvm->asidtracking topre_svm_runand added ASID set dirty bit checking.v1 -> v2
- Removed unnecessary update check of L1 save.cr3 duringnested_svm_vmexit.
- Movedvmcb01_pato svm.
- Removedget_host_vmcb()function.
- Updated vmsave/vmload corresponding vmcb state during L2 enter and exit which fixed the L2 load issue.
- Moved asid workaround to a new patch which adds asid to svm.
- Init previously uninitialized L2 vmcb save.gpat and save.cr4Tested:
kvm-unit-tests
kvm self tests
Loaded fedora nested guest on fedoraCathy Avery (2):
KVM: SVM: Track asid fromvcpu_svm
KVM: SVM: Use a separate vmcb for the nested L2 guestarch/x86/kvm/svm/nested.c | 125 ++++++++++++++++++-------------------- arch/x86/kvm/svm/svm.c | 58 +++++++++++------- arch/x86/kvm/svm/svm.h | 51 +++++----------- 3 files changed, 110 insertions(+), 124 deletions(-)--
2.20.1