| Message ID | b57280b5562893e2616257ac9c2d4525a9aeeb42.1638471124.git.thomas.lendacky@amd.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | KVM: SVM: Do not terminate SEV-ES guests on GHCB validation failure | expand |
On 12/2/21 19:52, Tom Lendacky wrote: > Currently, an SEV-ES guest is terminated if the validation of the VMGEXIT > exit code or exit parameters fails. > > The VMGEXIT instruction can be issued from userspace, even though > userspace (likely) can't update the GHCB. To prevent userspace from being > able to kill the guest, return an error through the GHCB when validation > fails rather than terminating the guest. For cases where the GHCB can't be > updated (e.g. the GHCB can't be mapped, etc.), just return back to the > guest. > > The new error codes are documented in the lasest update to the GHCB > specification. > > Fixes: 291bd20d5d88 ("KVM: SVM: Add initial support for a VMGEXIT VMEXIT") > Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> > --- > arch/x86/include/asm/sev-common.h | 11 ++++ > arch/x86/kvm/svm/sev.c | 106 +++++++++++++++++------------- > 2 files changed, 71 insertions(+), 46 deletions(-) > > diff --git a/arch/x86/include/asm/sev-common.h b/arch/x86/include/asm/sev-common.h > index 2cef6c5a52c2..6acaf5af0a3d 100644 > --- a/arch/x86/include/asm/sev-common.h > +++ b/arch/x86/include/asm/sev-common.h > @@ -73,4 +73,15 @@ > > #define GHCB_RESP_CODE(v) ((v) & GHCB_MSR_INFO_MASK) > > +/* > + * Error codes related to GHCB input that can be communicated back to the guest > + * by setting the lower 32-bits of the GHCB SW_EXITINFO1 field to 2. > + */ > +#define GHCB_ERR_NOT_REGISTERED 1 > +#define GHCB_ERR_INVALID_USAGE 2 > +#define GHCB_ERR_INVALID_SCRATCH_AREA 3 > +#define GHCB_ERR_MISSING_INPUT 4 > +#define GHCB_ERR_INVALID_INPUT 5 > +#define GHCB_ERR_INVALID_EVENT 6 > + > #endif > diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c > index 713e3daa9574..322553322202 100644 > --- a/arch/x86/kvm/svm/sev.c > +++ b/arch/x86/kvm/svm/sev.c > @@ -2353,24 +2353,29 @@ static void sev_es_sync_from_ghcb(struct vcpu_svm *svm) > memset(ghcb->save.valid_bitmap, 0, sizeof(ghcb->save.valid_bitmap)); > } > > -static int sev_es_validate_vmgexit(struct vcpu_svm *svm) > +static bool sev_es_validate_vmgexit(struct vcpu_svm *svm) > { > struct kvm_vcpu *vcpu; > struct ghcb *ghcb; > - u64 exit_code = 0; > + u64 exit_code; > + u64 reason; > > ghcb = svm->sev_es.ghcb; > > - /* Only GHCB Usage code 0 is supported */ > - if (ghcb->ghcb_usage) > - goto vmgexit_err; > - > /* > - * Retrieve the exit code now even though is may not be marked valid > + * Retrieve the exit code now even though it may not be marked valid > * as it could help with debugging. > */ > exit_code = ghcb_get_sw_exit_code(ghcb); > > + /* Only GHCB Usage code 0 is supported */ > + if (ghcb->ghcb_usage) { > + reason = GHCB_ERR_INVALID_USAGE; > + goto vmgexit_err; > + } > + > + reason = GHCB_ERR_MISSING_INPUT; > + > if (!ghcb_sw_exit_code_is_valid(ghcb) || > !ghcb_sw_exit_info_1_is_valid(ghcb) || > !ghcb_sw_exit_info_2_is_valid(ghcb)) > @@ -2449,30 +2454,34 @@ static int sev_es_validate_vmgexit(struct vcpu_svm *svm) > case SVM_VMGEXIT_UNSUPPORTED_EVENT: > break; > default: > + reason = GHCB_ERR_INVALID_EVENT; > goto vmgexit_err; > } > > - return 0; > + return true; > > vmgexit_err: > vcpu = &svm->vcpu; > > - if (ghcb->ghcb_usage) { > + if (reason == GHCB_ERR_INVALID_USAGE) { > vcpu_unimpl(vcpu, "vmgexit: ghcb usage %#x is not valid\n", > ghcb->ghcb_usage); > + } else if (reason == GHCB_ERR_INVALID_EVENT) { > + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx is not valid\n", > + exit_code); > } else { > - vcpu_unimpl(vcpu, "vmgexit: exit reason %#llx is not valid\n", > + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx input is not valid\n", > exit_code); > dump_ghcb(svm); > } > > - vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; > - vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_UNEXPECTED_EXIT_REASON; > - vcpu->run->internal.ndata = 2; > - vcpu->run->internal.data[0] = exit_code; > - vcpu->run->internal.data[1] = vcpu->arch.last_vmentry_cpu; > + /* Clear the valid entries fields */ > + memset(ghcb->save.valid_bitmap, 0, sizeof(ghcb->save.valid_bitmap)); > + > + ghcb_set_sw_exit_info_1(ghcb, 2); > + ghcb_set_sw_exit_info_2(ghcb, reason); > > - return -EINVAL; > + return false; > } > > void sev_es_unmap_ghcb(struct vcpu_svm *svm) > @@ -2531,7 +2540,7 @@ void pre_sev_run(struct vcpu_svm *svm, int cpu) > } > > #define GHCB_SCRATCH_AREA_LIMIT (16ULL * PAGE_SIZE) > -static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > +static bool setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > { > struct vmcb_control_area *control = &svm->vmcb->control; > struct ghcb *ghcb = svm->sev_es.ghcb; > @@ -2542,14 +2551,14 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > scratch_gpa_beg = ghcb_get_sw_scratch(ghcb); > if (!scratch_gpa_beg) { > pr_err("vmgexit: scratch gpa not provided\n"); > - return -EINVAL; > + goto e_scratch; > } > > scratch_gpa_end = scratch_gpa_beg + len; > if (scratch_gpa_end < scratch_gpa_beg) { > pr_err("vmgexit: scratch length (%#llx) not valid for scratch address (%#llx)\n", > len, scratch_gpa_beg); > - return -EINVAL; > + goto e_scratch; > } > > if ((scratch_gpa_beg & PAGE_MASK) == control->ghcb_gpa) { > @@ -2567,7 +2576,7 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > scratch_gpa_end > ghcb_scratch_end) { > pr_err("vmgexit: scratch area is outside of GHCB shared buffer area (%#llx - %#llx)\n", > scratch_gpa_beg, scratch_gpa_end); > - return -EINVAL; > + goto e_scratch; > } > > scratch_va = (void *)svm->sev_es.ghcb; > @@ -2580,18 +2589,18 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > if (len > GHCB_SCRATCH_AREA_LIMIT) { > pr_err("vmgexit: scratch area exceeds KVM limits (%#llx requested, %#llx limit)\n", > len, GHCB_SCRATCH_AREA_LIMIT); > - return -EINVAL; > + goto e_scratch; > } > scratch_va = kvzalloc(len, GFP_KERNEL_ACCOUNT); > if (!scratch_va) > - return -ENOMEM; > + goto e_scratch; > > if (kvm_read_guest(svm->vcpu.kvm, scratch_gpa_beg, scratch_va, len)) { > /* Unable to copy scratch area from guest */ > pr_err("vmgexit: kvm_read_guest for scratch area failed\n"); > > kvfree(scratch_va); > - return -EFAULT; > + goto e_scratch; > } > > /* > @@ -2607,7 +2616,13 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > svm->sev_es.ghcb_sa = scratch_va; > svm->sev_es.ghcb_sa_len = len; > > - return 0; > + return true; > + > +e_scratch: > + ghcb_set_sw_exit_info_1(ghcb, 2); > + ghcb_set_sw_exit_info_2(ghcb, GHCB_ERR_INVALID_SCRATCH_AREA); > + > + return false; > } > > static void set_ghcb_msr_bits(struct vcpu_svm *svm, u64 value, u64 mask, > @@ -2658,7 +2673,7 @@ static int sev_handle_vmgexit_msr_protocol(struct vcpu_svm *svm) > > ret = svm_invoke_exit_handler(vcpu, SVM_EXIT_CPUID); > if (!ret) { > - ret = -EINVAL; > + /* Error, keep GHCB MSR value as-is */ > break; > } > > @@ -2694,10 +2709,13 @@ static int sev_handle_vmgexit_msr_protocol(struct vcpu_svm *svm) > GHCB_MSR_TERM_REASON_POS); > pr_info("SEV-ES guest requested termination: %#llx:%#llx\n", > reason_set, reason_code); > - fallthrough; > + > + ret = -EINVAL; > + break; > } > default: > - ret = -EINVAL; > + /* Error, keep GHCB MSR value as-is */ > + break; > } > > trace_kvm_vmgexit_msr_protocol_exit(svm->vcpu.vcpu_id, > @@ -2721,14 +2739,18 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) > > if (!ghcb_gpa) { > vcpu_unimpl(vcpu, "vmgexit: GHCB gpa is not set\n"); > - return -EINVAL; > + > + /* Without a GHCB, just return right back to the guest */ > + return 1; > } > > if (kvm_vcpu_map(vcpu, ghcb_gpa >> PAGE_SHIFT, &svm->sev_es.ghcb_map)) { > /* Unable to map GHCB from guest */ > vcpu_unimpl(vcpu, "vmgexit: error mapping GHCB [%#llx] from guest\n", > ghcb_gpa); > - return -EINVAL; > + > + /* Without a GHCB, just return right back to the guest */ > + return 1; > } > > svm->sev_es.ghcb = svm->sev_es.ghcb_map.hva; > @@ -2738,18 +2760,17 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) > > exit_code = ghcb_get_sw_exit_code(ghcb); > > - ret = sev_es_validate_vmgexit(svm); > - if (ret) > - return ret; > + if (!sev_es_validate_vmgexit(svm)) > + return 1; > > sev_es_sync_from_ghcb(svm); > ghcb_set_sw_exit_info_1(ghcb, 0); > ghcb_set_sw_exit_info_2(ghcb, 0); > > + ret = 1; > switch (exit_code) { > case SVM_VMGEXIT_MMIO_READ: > - ret = setup_vmgexit_scratch(svm, true, control->exit_info_2); > - if (ret) > + if (!setup_vmgexit_scratch(svm, true, control->exit_info_2)) > break; > > ret = kvm_sev_es_mmio_read(vcpu, > @@ -2758,8 +2779,7 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) > svm->sev_es.ghcb_sa); > break; > case SVM_VMGEXIT_MMIO_WRITE: > - ret = setup_vmgexit_scratch(svm, false, control->exit_info_2); > - if (ret) > + if (!setup_vmgexit_scratch(svm, false, control->exit_info_2)) > break; > > ret = kvm_sev_es_mmio_write(vcpu, > @@ -2788,14 +2808,10 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) > default: > pr_err("svm: vmgexit: unsupported AP jump table request - exit_info_1=%#llx\n", > control->exit_info_1); > - ghcb_set_sw_exit_info_1(ghcb, 1); > - ghcb_set_sw_exit_info_2(ghcb, > - X86_TRAP_UD | > - SVM_EVTINJ_TYPE_EXEPT | > - SVM_EVTINJ_VALID); > + ghcb_set_sw_exit_info_1(ghcb, 2); > + ghcb_set_sw_exit_info_2(ghcb, GHCB_ERR_INVALID_INPUT); > } > > - ret = 1; > break; > } > case SVM_VMGEXIT_UNSUPPORTED_EVENT: > @@ -2815,7 +2831,6 @@ int sev_es_string_io(struct vcpu_svm *svm, int size, unsigned int port, int in) > { > int count; > int bytes; > - int r; > > if (svm->vmcb->control.exit_info_2 > INT_MAX) > return -EINVAL; > @@ -2824,9 +2839,8 @@ int sev_es_string_io(struct vcpu_svm *svm, int size, unsigned int port, int in) > if (unlikely(check_mul_overflow(count, size, &bytes))) > return -EINVAL; > > - r = setup_vmgexit_scratch(svm, in, bytes); > - if (r) > - return r; > + if (!setup_vmgexit_scratch(svm, in, bytes)) > + return 1; > > return kvm_sev_es_string_io(&svm->vcpu, size, port, svm->sev_es.ghcb_sa, > count, in); > Queued, thanks. Though it would have been nicer to split the changes in the return values (e.g. for setup_vmgexit_scratch and sev_es_validate_vmgexit) from the introduction of the new GHCB exitinfo. Paolo Paolo
On 12/2/21 1:19 PM, Paolo Bonzini wrote: > On 12/2/21 19:52, Tom Lendacky wrote: > > Queued, thanks. Though it would have been nicer to split the changes in > the return values (e.g. for setup_vmgexit_scratch and > sev_es_validate_vmgexit) from the introduction of the new GHCB exitinfo. I can still do that if it will help make things easier. Let me know. Thanks, Tom > > Paolo > > Paolo
On 12/2/21 20:39, Tom Lendacky wrote: >> Queued, thanks. Though it would have been nicer to split the changes >> in the return values (e.g. for setup_vmgexit_scratch and >> sev_es_validate_vmgexit) from the introduction of the new GHCB exitinfo. > > I can still do that if it will help make things easier. Let me know. Well, at this point I've already reviewed it. :) Paolo
On Thu, Dec 02, 2021, Tom Lendacky wrote: > diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c > index 713e3daa9574..322553322202 100644 > --- a/arch/x86/kvm/svm/sev.c > +++ b/arch/x86/kvm/svm/sev.c > @@ -2353,24 +2353,29 @@ static void sev_es_sync_from_ghcb(struct vcpu_svm *svm) > memset(ghcb->save.valid_bitmap, 0, sizeof(ghcb->save.valid_bitmap)); > } > > -static int sev_es_validate_vmgexit(struct vcpu_svm *svm) > +static bool sev_es_validate_vmgexit(struct vcpu_svm *svm) > { ... > - return 0; > + return true; > > vmgexit_err: > vcpu = &svm->vcpu; > > - if (ghcb->ghcb_usage) { > + if (reason == GHCB_ERR_INVALID_USAGE) { > vcpu_unimpl(vcpu, "vmgexit: ghcb usage %#x is not valid\n", > ghcb->ghcb_usage); > + } else if (reason == GHCB_ERR_INVALID_EVENT) { > + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx is not valid\n", > + exit_code); > } else { > - vcpu_unimpl(vcpu, "vmgexit: exit reason %#llx is not valid\n", > + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx input is not valid\n", > exit_code); > dump_ghcb(svm); > } > > - vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; > - vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_UNEXPECTED_EXIT_REASON; > - vcpu->run->internal.ndata = 2; > - vcpu->run->internal.data[0] = exit_code; > - vcpu->run->internal.data[1] = vcpu->arch.last_vmentry_cpu; > + /* Clear the valid entries fields */ > + memset(ghcb->save.valid_bitmap, 0, sizeof(ghcb->save.valid_bitmap)); > + > + ghcb_set_sw_exit_info_1(ghcb, 2); > + ghcb_set_sw_exit_info_2(ghcb, reason); > > - return -EINVAL; > + return false; I'd really prefer that this helper continue to return 0/-EINVAL, there's no hint in the function name that this return true/false. And given the usage, there's no advantage to returning true/false. On the contrary, if there's a future condition where this needs to exit to userspace, we'll end up switching this all back to int. > } > > void sev_es_unmap_ghcb(struct vcpu_svm *svm) > @@ -2531,7 +2540,7 @@ void pre_sev_run(struct vcpu_svm *svm, int cpu) > } > > #define GHCB_SCRATCH_AREA_LIMIT (16ULL * PAGE_SIZE) > -static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > +static bool setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) Same here, but now there's an actual need to return an int... > { > struct vmcb_control_area *control = &svm->vmcb->control; > struct ghcb *ghcb = svm->sev_es.ghcb; > @@ -2542,14 +2551,14 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > scratch_gpa_beg = ghcb_get_sw_scratch(ghcb); > if (!scratch_gpa_beg) { > pr_err("vmgexit: scratch gpa not provided\n"); > - return -EINVAL; > + goto e_scratch; > } > > scratch_gpa_end = scratch_gpa_beg + len; > if (scratch_gpa_end < scratch_gpa_beg) { > pr_err("vmgexit: scratch length (%#llx) not valid for scratch address (%#llx)\n", > len, scratch_gpa_beg); > - return -EINVAL; > + goto e_scratch; > } > > if ((scratch_gpa_beg & PAGE_MASK) == control->ghcb_gpa) { > @@ -2567,7 +2576,7 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > scratch_gpa_end > ghcb_scratch_end) { > pr_err("vmgexit: scratch area is outside of GHCB shared buffer area (%#llx - %#llx)\n", > scratch_gpa_beg, scratch_gpa_end); > - return -EINVAL; > + goto e_scratch; > } > > scratch_va = (void *)svm->sev_es.ghcb; > @@ -2580,18 +2589,18 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > if (len > GHCB_SCRATCH_AREA_LIMIT) { > pr_err("vmgexit: scratch area exceeds KVM limits (%#llx requested, %#llx limit)\n", > len, GHCB_SCRATCH_AREA_LIMIT); > - return -EINVAL; > + goto e_scratch; > } > scratch_va = kvzalloc(len, GFP_KERNEL_ACCOUNT); > if (!scratch_va) > - return -ENOMEM; ...because this is wrong. Failure to allocate memory should exit to userspace, not report an error to the guest. > + goto e_scratch; > > if (kvm_read_guest(svm->vcpu.kvm, scratch_gpa_beg, scratch_va, len)) { > /* Unable to copy scratch area from guest */ > pr_err("vmgexit: kvm_read_guest for scratch area failed\n"); > > kvfree(scratch_va); > - return -EFAULT; > + goto e_scratch; Same here, failure to read guest memory is a userspace issue and needs to be reported to userspace. > } > > /* IMO, this should be the patch (compile tested only). --- arch/x86/include/asm/sev-common.h | 11 +++++ arch/x86/kvm/svm/sev.c | 75 +++++++++++++++++++------------ 2 files changed, 58 insertions(+), 28 deletions(-) diff --git a/arch/x86/include/asm/sev-common.h b/arch/x86/include/asm/sev-common.h index 2cef6c5a52c2..6acaf5af0a3d 100644 --- a/arch/x86/include/asm/sev-common.h +++ b/arch/x86/include/asm/sev-common.h @@ -73,4 +73,15 @@ #define GHCB_RESP_CODE(v) ((v) & GHCB_MSR_INFO_MASK) +/* + * Error codes related to GHCB input that can be communicated back to the guest + * by setting the lower 32-bits of the GHCB SW_EXITINFO1 field to 2. + */ +#define GHCB_ERR_NOT_REGISTERED 1 +#define GHCB_ERR_INVALID_USAGE 2 +#define GHCB_ERR_INVALID_SCRATCH_AREA 3 +#define GHCB_ERR_MISSING_INPUT 4 +#define GHCB_ERR_INVALID_INPUT 5 +#define GHCB_ERR_INVALID_EVENT 6 + #endif diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c index 713e3daa9574..60c6d7b216eb 100644 --- a/arch/x86/kvm/svm/sev.c +++ b/arch/x86/kvm/svm/sev.c @@ -2357,20 +2357,25 @@ static int sev_es_validate_vmgexit(struct vcpu_svm *svm) { struct kvm_vcpu *vcpu; struct ghcb *ghcb; - u64 exit_code = 0; + u64 exit_code; + u64 reason; ghcb = svm->sev_es.ghcb; - /* Only GHCB Usage code 0 is supported */ - if (ghcb->ghcb_usage) - goto vmgexit_err; - /* - * Retrieve the exit code now even though is may not be marked valid + * Retrieve the exit code now even though it may not be marked valid * as it could help with debugging. */ exit_code = ghcb_get_sw_exit_code(ghcb); + /* Only GHCB Usage code 0 is supported */ + if (ghcb->ghcb_usage) { + reason = GHCB_ERR_INVALID_USAGE; + goto vmgexit_err; + } + + reason = GHCB_ERR_MISSING_INPUT; + if (!ghcb_sw_exit_code_is_valid(ghcb) || !ghcb_sw_exit_info_1_is_valid(ghcb) || !ghcb_sw_exit_info_2_is_valid(ghcb)) @@ -2449,6 +2454,7 @@ static int sev_es_validate_vmgexit(struct vcpu_svm *svm) case SVM_VMGEXIT_UNSUPPORTED_EVENT: break; default: + reason = GHCB_ERR_INVALID_EVENT; goto vmgexit_err; } @@ -2457,22 +2463,25 @@ static int sev_es_validate_vmgexit(struct vcpu_svm *svm) vmgexit_err: vcpu = &svm->vcpu; - if (ghcb->ghcb_usage) { + if (reason == GHCB_ERR_INVALID_USAGE) { vcpu_unimpl(vcpu, "vmgexit: ghcb usage %#x is not valid\n", ghcb->ghcb_usage); + } else if (reason == GHCB_ERR_INVALID_EVENT) { + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx is not valid\n", + exit_code); } else { - vcpu_unimpl(vcpu, "vmgexit: exit reason %#llx is not valid\n", + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx input is not valid\n", exit_code); dump_ghcb(svm); } - vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; - vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_UNEXPECTED_EXIT_REASON; - vcpu->run->internal.ndata = 2; - vcpu->run->internal.data[0] = exit_code; - vcpu->run->internal.data[1] = vcpu->arch.last_vmentry_cpu; + /* Clear the valid entries fields */ + memset(ghcb->save.valid_bitmap, 0, sizeof(ghcb->save.valid_bitmap)); - return -EINVAL; + ghcb_set_sw_exit_info_1(ghcb, 2); + ghcb_set_sw_exit_info_2(ghcb, reason); + + return 1; } void sev_es_unmap_ghcb(struct vcpu_svm *svm) @@ -2542,14 +2551,14 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) scratch_gpa_beg = ghcb_get_sw_scratch(ghcb); if (!scratch_gpa_beg) { pr_err("vmgexit: scratch gpa not provided\n"); - return -EINVAL; + goto e_scratch; } scratch_gpa_end = scratch_gpa_beg + len; if (scratch_gpa_end < scratch_gpa_beg) { pr_err("vmgexit: scratch length (%#llx) not valid for scratch address (%#llx)\n", len, scratch_gpa_beg); - return -EINVAL; + goto e_scratch; } if ((scratch_gpa_beg & PAGE_MASK) == control->ghcb_gpa) { @@ -2567,7 +2576,7 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) scratch_gpa_end > ghcb_scratch_end) { pr_err("vmgexit: scratch area is outside of GHCB shared buffer area (%#llx - %#llx)\n", scratch_gpa_beg, scratch_gpa_end); - return -EINVAL; + goto e_scratch; } scratch_va = (void *)svm->sev_es.ghcb; @@ -2580,7 +2589,7 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) if (len > GHCB_SCRATCH_AREA_LIMIT) { pr_err("vmgexit: scratch area exceeds KVM limits (%#llx requested, %#llx limit)\n", len, GHCB_SCRATCH_AREA_LIMIT); - return -EINVAL; + goto e_scratch; } scratch_va = kvzalloc(len, GFP_KERNEL_ACCOUNT); if (!scratch_va) @@ -2608,6 +2617,12 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) svm->sev_es.ghcb_sa_len = len; return 0; + +e_scratch: + ghcb_set_sw_exit_info_1(ghcb, 2); + ghcb_set_sw_exit_info_2(ghcb, GHCB_ERR_INVALID_SCRATCH_AREA); + + return 1; } static void set_ghcb_msr_bits(struct vcpu_svm *svm, u64 value, u64 mask, @@ -2658,7 +2673,7 @@ static int sev_handle_vmgexit_msr_protocol(struct vcpu_svm *svm) ret = svm_invoke_exit_handler(vcpu, SVM_EXIT_CPUID); if (!ret) { - ret = -EINVAL; + /* Error, keep GHCB MSR value as-is */ break; } @@ -2694,10 +2709,13 @@ static int sev_handle_vmgexit_msr_protocol(struct vcpu_svm *svm) GHCB_MSR_TERM_REASON_POS); pr_info("SEV-ES guest requested termination: %#llx:%#llx\n", reason_set, reason_code); - fallthrough; + + ret = -EINVAL; + break; } default: - ret = -EINVAL; + /* Error, keep GHCB MSR value as-is */ + break; } trace_kvm_vmgexit_msr_protocol_exit(svm->vcpu.vcpu_id, @@ -2721,14 +2739,18 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) if (!ghcb_gpa) { vcpu_unimpl(vcpu, "vmgexit: GHCB gpa is not set\n"); - return -EINVAL; + + /* Without a GHCB, just return right back to the guest */ + return 1; } if (kvm_vcpu_map(vcpu, ghcb_gpa >> PAGE_SHIFT, &svm->sev_es.ghcb_map)) { /* Unable to map GHCB from guest */ vcpu_unimpl(vcpu, "vmgexit: error mapping GHCB [%#llx] from guest\n", ghcb_gpa); - return -EINVAL; + + /* Without a GHCB, just return right back to the guest */ + return 1; } svm->sev_es.ghcb = svm->sev_es.ghcb_map.hva; @@ -2788,11 +2810,8 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) default: pr_err("svm: vmgexit: unsupported AP jump table request - exit_info_1=%#llx\n", control->exit_info_1); - ghcb_set_sw_exit_info_1(ghcb, 1); - ghcb_set_sw_exit_info_2(ghcb, - X86_TRAP_UD | - SVM_EVTINJ_TYPE_EXEPT | - SVM_EVTINJ_VALID); + ghcb_set_sw_exit_info_1(ghcb, 2); + ghcb_set_sw_exit_info_2(ghcb, GHCB_ERR_INVALID_INPUT); } ret = 1; base-commit: 70f433c2193fbfb5541ef98f973e087ddf2f9dfb --
On 12/3/21 10:39 AM, Sean Christopherson wrote: > On Thu, Dec 02, 2021, Tom Lendacky wrote: >> >> - return -EINVAL; >> + return false; > > I'd really prefer that this helper continue to return 0/-EINVAL, there's no hint > in the function name that this return true/false. And given the usage, there's > no advantage to returning true/false. On the contrary, if there's a future > condition where this needs to exit to userspace, we'll end up switching this all > back to int. I don't have any objection to that. > >> } >> scratch_va = kvzalloc(len, GFP_KERNEL_ACCOUNT); >> if (!scratch_va) >> - return -ENOMEM; > > ...because this is wrong. Failure to allocate memory should exit to userspace, > not report an error to the guest. > >> + goto e_scratch; >> >> if (kvm_read_guest(svm->vcpu.kvm, scratch_gpa_beg, scratch_va, len)) { >> /* Unable to copy scratch area from guest */ >> pr_err("vmgexit: kvm_read_guest for scratch area failed\n"); >> >> kvfree(scratch_va); >> - return -EFAULT; >> + goto e_scratch; > > Same here, failure to read guest memory is a userspace issue and needs to be > reported to userspace. But it could be a guest issue as well... whichever is preferred is ok by me. > >> } >> >> /* > > IMO, this should be the patch (compile tested only). I can test this, but probably won't be able to get to it until Monday. Thanks, Tom > > --- > arch/x86/include/asm/sev-common.h | 11 +++++ > arch/x86/kvm/svm/sev.c | 75 +++++++++++++++++++------------ > 2 files changed, 58 insertions(+), 28 deletions(-) > > diff --git a/arch/x86/include/asm/sev-common.h b/arch/x86/include/asm/sev-common.h > index 2cef6c5a52c2..6acaf5af0a3d 100644 > --- a/arch/x86/include/asm/sev-common.h > +++ b/arch/x86/include/asm/sev-common.h > @@ -73,4 +73,15 @@ > > #define GHCB_RESP_CODE(v) ((v) & GHCB_MSR_INFO_MASK) > > +/* > + * Error codes related to GHCB input that can be communicated back to the guest > + * by setting the lower 32-bits of the GHCB SW_EXITINFO1 field to 2. > + */ > +#define GHCB_ERR_NOT_REGISTERED 1 > +#define GHCB_ERR_INVALID_USAGE 2 > +#define GHCB_ERR_INVALID_SCRATCH_AREA 3 > +#define GHCB_ERR_MISSING_INPUT 4 > +#define GHCB_ERR_INVALID_INPUT 5 > +#define GHCB_ERR_INVALID_EVENT 6 > + > #endif > diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c > index 713e3daa9574..60c6d7b216eb 100644 > --- a/arch/x86/kvm/svm/sev.c > +++ b/arch/x86/kvm/svm/sev.c > @@ -2357,20 +2357,25 @@ static int sev_es_validate_vmgexit(struct vcpu_svm *svm) > { > struct kvm_vcpu *vcpu; > struct ghcb *ghcb; > - u64 exit_code = 0; > + u64 exit_code; > + u64 reason; > > ghcb = svm->sev_es.ghcb; > > - /* Only GHCB Usage code 0 is supported */ > - if (ghcb->ghcb_usage) > - goto vmgexit_err; > - > /* > - * Retrieve the exit code now even though is may not be marked valid > + * Retrieve the exit code now even though it may not be marked valid > * as it could help with debugging. > */ > exit_code = ghcb_get_sw_exit_code(ghcb); > > + /* Only GHCB Usage code 0 is supported */ > + if (ghcb->ghcb_usage) { > + reason = GHCB_ERR_INVALID_USAGE; > + goto vmgexit_err; > + } > + > + reason = GHCB_ERR_MISSING_INPUT; > + > if (!ghcb_sw_exit_code_is_valid(ghcb) || > !ghcb_sw_exit_info_1_is_valid(ghcb) || > !ghcb_sw_exit_info_2_is_valid(ghcb)) > @@ -2449,6 +2454,7 @@ static int sev_es_validate_vmgexit(struct vcpu_svm *svm) > case SVM_VMGEXIT_UNSUPPORTED_EVENT: > break; > default: > + reason = GHCB_ERR_INVALID_EVENT; > goto vmgexit_err; > } > > @@ -2457,22 +2463,25 @@ static int sev_es_validate_vmgexit(struct vcpu_svm *svm) > vmgexit_err: > vcpu = &svm->vcpu; > > - if (ghcb->ghcb_usage) { > + if (reason == GHCB_ERR_INVALID_USAGE) { > vcpu_unimpl(vcpu, "vmgexit: ghcb usage %#x is not valid\n", > ghcb->ghcb_usage); > + } else if (reason == GHCB_ERR_INVALID_EVENT) { > + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx is not valid\n", > + exit_code); > } else { > - vcpu_unimpl(vcpu, "vmgexit: exit reason %#llx is not valid\n", > + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx input is not valid\n", > exit_code); > dump_ghcb(svm); > } > > - vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; > - vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_UNEXPECTED_EXIT_REASON; > - vcpu->run->internal.ndata = 2; > - vcpu->run->internal.data[0] = exit_code; > - vcpu->run->internal.data[1] = vcpu->arch.last_vmentry_cpu; > + /* Clear the valid entries fields */ > + memset(ghcb->save.valid_bitmap, 0, sizeof(ghcb->save.valid_bitmap)); > > - return -EINVAL; > + ghcb_set_sw_exit_info_1(ghcb, 2); > + ghcb_set_sw_exit_info_2(ghcb, reason); > + > + return 1; > } > > void sev_es_unmap_ghcb(struct vcpu_svm *svm) > @@ -2542,14 +2551,14 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > scratch_gpa_beg = ghcb_get_sw_scratch(ghcb); > if (!scratch_gpa_beg) { > pr_err("vmgexit: scratch gpa not provided\n"); > - return -EINVAL; > + goto e_scratch; > } > > scratch_gpa_end = scratch_gpa_beg + len; > if (scratch_gpa_end < scratch_gpa_beg) { > pr_err("vmgexit: scratch length (%#llx) not valid for scratch address (%#llx)\n", > len, scratch_gpa_beg); > - return -EINVAL; > + goto e_scratch; > } > > if ((scratch_gpa_beg & PAGE_MASK) == control->ghcb_gpa) { > @@ -2567,7 +2576,7 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > scratch_gpa_end > ghcb_scratch_end) { > pr_err("vmgexit: scratch area is outside of GHCB shared buffer area (%#llx - %#llx)\n", > scratch_gpa_beg, scratch_gpa_end); > - return -EINVAL; > + goto e_scratch; > } > > scratch_va = (void *)svm->sev_es.ghcb; > @@ -2580,7 +2589,7 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > if (len > GHCB_SCRATCH_AREA_LIMIT) { > pr_err("vmgexit: scratch area exceeds KVM limits (%#llx requested, %#llx limit)\n", > len, GHCB_SCRATCH_AREA_LIMIT); > - return -EINVAL; > + goto e_scratch; > } > scratch_va = kvzalloc(len, GFP_KERNEL_ACCOUNT); > if (!scratch_va) > @@ -2608,6 +2617,12 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) > svm->sev_es.ghcb_sa_len = len; > > return 0; > + > +e_scratch: > + ghcb_set_sw_exit_info_1(ghcb, 2); > + ghcb_set_sw_exit_info_2(ghcb, GHCB_ERR_INVALID_SCRATCH_AREA); > + > + return 1; > } > > static void set_ghcb_msr_bits(struct vcpu_svm *svm, u64 value, u64 mask, > @@ -2658,7 +2673,7 @@ static int sev_handle_vmgexit_msr_protocol(struct vcpu_svm *svm) > > ret = svm_invoke_exit_handler(vcpu, SVM_EXIT_CPUID); > if (!ret) { > - ret = -EINVAL; > + /* Error, keep GHCB MSR value as-is */ > break; > } > > @@ -2694,10 +2709,13 @@ static int sev_handle_vmgexit_msr_protocol(struct vcpu_svm *svm) > GHCB_MSR_TERM_REASON_POS); > pr_info("SEV-ES guest requested termination: %#llx:%#llx\n", > reason_set, reason_code); > - fallthrough; > + > + ret = -EINVAL; > + break; > } > default: > - ret = -EINVAL; > + /* Error, keep GHCB MSR value as-is */ > + break; > } > > trace_kvm_vmgexit_msr_protocol_exit(svm->vcpu.vcpu_id, > @@ -2721,14 +2739,18 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) > > if (!ghcb_gpa) { > vcpu_unimpl(vcpu, "vmgexit: GHCB gpa is not set\n"); > - return -EINVAL; > + > + /* Without a GHCB, just return right back to the guest */ > + return 1; > } > > if (kvm_vcpu_map(vcpu, ghcb_gpa >> PAGE_SHIFT, &svm->sev_es.ghcb_map)) { > /* Unable to map GHCB from guest */ > vcpu_unimpl(vcpu, "vmgexit: error mapping GHCB [%#llx] from guest\n", > ghcb_gpa); > - return -EINVAL; > + > + /* Without a GHCB, just return right back to the guest */ > + return 1; > } > > svm->sev_es.ghcb = svm->sev_es.ghcb_map.hva; > @@ -2788,11 +2810,8 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) > default: > pr_err("svm: vmgexit: unsupported AP jump table request - exit_info_1=%#llx\n", > control->exit_info_1); > - ghcb_set_sw_exit_info_1(ghcb, 1); > - ghcb_set_sw_exit_info_2(ghcb, > - X86_TRAP_UD | > - SVM_EVTINJ_TYPE_EXEPT | > - SVM_EVTINJ_VALID); > + ghcb_set_sw_exit_info_1(ghcb, 2); > + ghcb_set_sw_exit_info_2(ghcb, GHCB_ERR_INVALID_INPUT); > } > > ret = 1; > > base-commit: 70f433c2193fbfb5541ef98f973e087ddf2f9dfb > -- >
On Fri, Dec 03, 2021, Tom Lendacky wrote: > On 12/3/21 10:39 AM, Sean Christopherson wrote: > > On Thu, Dec 02, 2021, Tom Lendacky wrote: > > > + goto e_scratch; > > > if (kvm_read_guest(svm->vcpu.kvm, scratch_gpa_beg, scratch_va, len)) { > > > /* Unable to copy scratch area from guest */ > > > pr_err("vmgexit: kvm_read_guest for scratch area failed\n"); > > > kvfree(scratch_va); > > > - return -EFAULT; > > > + goto e_scratch; > > > > Same here, failure to read guest memory is a userspace issue and needs to be > > reported to userspace. > > But it could be a guest issue as well... whichever is preferred is ok by me. Arguably, any guest issue is a violation of the guest's contract with userspace, and thus userspace needs to decide how to proceed. E.g. userspace defines what is RAM vs. MMIO and communicates that directly to the guest, KVM is not involved in deciding what is/isn't RAM nor in communicating that information to the guest. If the scratch GPA doesn't resolve to a memslot, then the guest is not honoring the memory configuration as defined by userspace. And if userspace unmaps an hva for whatever reason, then exiting to userspace with -EFAULT is absolutely the right thing to do. KVM's ABI currently sucks and doesn't provide enough information to act on the -EFAULT, but I really want to change that as there are multiple use cases, e.g. uffd and virtiofs truncation, that shouldn't require any work in KVM beyond returning -EFAULT with a small amount of metadata. KVM could define its ABI such that failure to access the scratch area is reflected into the guest, i.e. establish a contract with userspace, but IMO that's undesirable as it limits KVM's options in the future, e.g. IIRC, in the potential uffd case any failure on a uaccess needs to kick out to userspace. KVM does have several cases where it reflects these errors into the guest, e.g. kvm_pv_clock_pairing() and Hyper-V emulation, but I would prefer we change those instead of adding more code that assumes any memory failure is the guest's fault.
On 12/3/21 12:59 PM, Tom Lendacky wrote: > On 12/3/21 10:39 AM, Sean Christopherson wrote: >> On Thu, Dec 02, 2021, Tom Lendacky wrote: > >> >> IMO, this should be the patch (compile tested only). > > I can test this, but probably won't be able to get to it until Monday. I did manage to get some time to test this today. Works as expected with an invalid scratch GPA terminating the guest. Up to Paolo now with what he would like to see/do. Thanks, Tom >
On Fri, Dec 3, 2021 at 11:00 AM Tom Lendacky <thomas.lendacky@amd.com> wrote: > > On 12/3/21 10:39 AM, Sean Christopherson wrote: > > On Thu, Dec 02, 2021, Tom Lendacky wrote: > > >> > >> - return -EINVAL; > >> + return false; > > > > I'd really prefer that this helper continue to return 0/-EINVAL, there's no hint > > in the function name that this return true/false. And given the usage, there's > > no advantage to returning true/false. On the contrary, if there's a future > > condition where this needs to exit to userspace, we'll end up switching this all > > back to int. > > I don't have any objection to that. I think Sean's review makes a pretty compelling case that we should keep the int return value for `setup_vmgexit_scratch()`. In particular, failing to allocate a host kernel buffer definitely seems like a host error that should return to userspace. Though, failing to read the guest GPA seems less clear cut on who is at fault (host vs. guest), as Tom mentioned. My understanding from the commit description is that the entire point of the patch is to protect the guest from mis-behaving guest userspace code. So I would think that if we have a case like mapping the guest GPA that could fail due to the guest or the host, we should probably go ahead and use the new GHCB error codes to return back to the guest in this case. But either way, having an int return code seems like the way to go for `setup_vmgexit_scratch()`. Because if we are wrong in either direction, it's a trivial fix. It's not as obvious to me that converting `sev_es_validate_vmgexit()` to return a bool is not cleaner than returning an int. This function seems to pretty much just process the GHCB buffer as a self-contained set of bits. So it's hard to imagine how this could fail in a way where exiting to userspace is the right thing to do. That being said, I do not have a strong objection to returning an int. Sean is right that an int is definitely more future proof. And I'm sure Sean (and Tom) have much better insight into how validating the bits written into the GHCB could potentially require code that could justify exiting out to userspace.
diff --git a/arch/x86/include/asm/sev-common.h b/arch/x86/include/asm/sev-common.h index 2cef6c5a52c2..6acaf5af0a3d 100644 --- a/arch/x86/include/asm/sev-common.h +++ b/arch/x86/include/asm/sev-common.h @@ -73,4 +73,15 @@ #define GHCB_RESP_CODE(v) ((v) & GHCB_MSR_INFO_MASK) +/* + * Error codes related to GHCB input that can be communicated back to the guest + * by setting the lower 32-bits of the GHCB SW_EXITINFO1 field to 2. + */ +#define GHCB_ERR_NOT_REGISTERED 1 +#define GHCB_ERR_INVALID_USAGE 2 +#define GHCB_ERR_INVALID_SCRATCH_AREA 3 +#define GHCB_ERR_MISSING_INPUT 4 +#define GHCB_ERR_INVALID_INPUT 5 +#define GHCB_ERR_INVALID_EVENT 6 + #endif diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c index 713e3daa9574..322553322202 100644 --- a/arch/x86/kvm/svm/sev.c +++ b/arch/x86/kvm/svm/sev.c @@ -2353,24 +2353,29 @@ static void sev_es_sync_from_ghcb(struct vcpu_svm *svm) memset(ghcb->save.valid_bitmap, 0, sizeof(ghcb->save.valid_bitmap)); } -static int sev_es_validate_vmgexit(struct vcpu_svm *svm) +static bool sev_es_validate_vmgexit(struct vcpu_svm *svm) { struct kvm_vcpu *vcpu; struct ghcb *ghcb; - u64 exit_code = 0; + u64 exit_code; + u64 reason; ghcb = svm->sev_es.ghcb; - /* Only GHCB Usage code 0 is supported */ - if (ghcb->ghcb_usage) - goto vmgexit_err; - /* - * Retrieve the exit code now even though is may not be marked valid + * Retrieve the exit code now even though it may not be marked valid * as it could help with debugging. */ exit_code = ghcb_get_sw_exit_code(ghcb); + /* Only GHCB Usage code 0 is supported */ + if (ghcb->ghcb_usage) { + reason = GHCB_ERR_INVALID_USAGE; + goto vmgexit_err; + } + + reason = GHCB_ERR_MISSING_INPUT; + if (!ghcb_sw_exit_code_is_valid(ghcb) || !ghcb_sw_exit_info_1_is_valid(ghcb) || !ghcb_sw_exit_info_2_is_valid(ghcb)) @@ -2449,30 +2454,34 @@ static int sev_es_validate_vmgexit(struct vcpu_svm *svm) case SVM_VMGEXIT_UNSUPPORTED_EVENT: break; default: + reason = GHCB_ERR_INVALID_EVENT; goto vmgexit_err; } - return 0; + return true; vmgexit_err: vcpu = &svm->vcpu; - if (ghcb->ghcb_usage) { + if (reason == GHCB_ERR_INVALID_USAGE) { vcpu_unimpl(vcpu, "vmgexit: ghcb usage %#x is not valid\n", ghcb->ghcb_usage); + } else if (reason == GHCB_ERR_INVALID_EVENT) { + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx is not valid\n", + exit_code); } else { - vcpu_unimpl(vcpu, "vmgexit: exit reason %#llx is not valid\n", + vcpu_unimpl(vcpu, "vmgexit: exit code %#llx input is not valid\n", exit_code); dump_ghcb(svm); } - vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; - vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_UNEXPECTED_EXIT_REASON; - vcpu->run->internal.ndata = 2; - vcpu->run->internal.data[0] = exit_code; - vcpu->run->internal.data[1] = vcpu->arch.last_vmentry_cpu; + /* Clear the valid entries fields */ + memset(ghcb->save.valid_bitmap, 0, sizeof(ghcb->save.valid_bitmap)); + + ghcb_set_sw_exit_info_1(ghcb, 2); + ghcb_set_sw_exit_info_2(ghcb, reason); - return -EINVAL; + return false; } void sev_es_unmap_ghcb(struct vcpu_svm *svm) @@ -2531,7 +2540,7 @@ void pre_sev_run(struct vcpu_svm *svm, int cpu) } #define GHCB_SCRATCH_AREA_LIMIT (16ULL * PAGE_SIZE) -static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) +static bool setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) { struct vmcb_control_area *control = &svm->vmcb->control; struct ghcb *ghcb = svm->sev_es.ghcb; @@ -2542,14 +2551,14 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) scratch_gpa_beg = ghcb_get_sw_scratch(ghcb); if (!scratch_gpa_beg) { pr_err("vmgexit: scratch gpa not provided\n"); - return -EINVAL; + goto e_scratch; } scratch_gpa_end = scratch_gpa_beg + len; if (scratch_gpa_end < scratch_gpa_beg) { pr_err("vmgexit: scratch length (%#llx) not valid for scratch address (%#llx)\n", len, scratch_gpa_beg); - return -EINVAL; + goto e_scratch; } if ((scratch_gpa_beg & PAGE_MASK) == control->ghcb_gpa) { @@ -2567,7 +2576,7 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) scratch_gpa_end > ghcb_scratch_end) { pr_err("vmgexit: scratch area is outside of GHCB shared buffer area (%#llx - %#llx)\n", scratch_gpa_beg, scratch_gpa_end); - return -EINVAL; + goto e_scratch; } scratch_va = (void *)svm->sev_es.ghcb; @@ -2580,18 +2589,18 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) if (len > GHCB_SCRATCH_AREA_LIMIT) { pr_err("vmgexit: scratch area exceeds KVM limits (%#llx requested, %#llx limit)\n", len, GHCB_SCRATCH_AREA_LIMIT); - return -EINVAL; + goto e_scratch; } scratch_va = kvzalloc(len, GFP_KERNEL_ACCOUNT); if (!scratch_va) - return -ENOMEM; + goto e_scratch; if (kvm_read_guest(svm->vcpu.kvm, scratch_gpa_beg, scratch_va, len)) { /* Unable to copy scratch area from guest */ pr_err("vmgexit: kvm_read_guest for scratch area failed\n"); kvfree(scratch_va); - return -EFAULT; + goto e_scratch; } /* @@ -2607,7 +2616,13 @@ static int setup_vmgexit_scratch(struct vcpu_svm *svm, bool sync, u64 len) svm->sev_es.ghcb_sa = scratch_va; svm->sev_es.ghcb_sa_len = len; - return 0; + return true; + +e_scratch: + ghcb_set_sw_exit_info_1(ghcb, 2); + ghcb_set_sw_exit_info_2(ghcb, GHCB_ERR_INVALID_SCRATCH_AREA); + + return false; } static void set_ghcb_msr_bits(struct vcpu_svm *svm, u64 value, u64 mask, @@ -2658,7 +2673,7 @@ static int sev_handle_vmgexit_msr_protocol(struct vcpu_svm *svm) ret = svm_invoke_exit_handler(vcpu, SVM_EXIT_CPUID); if (!ret) { - ret = -EINVAL; + /* Error, keep GHCB MSR value as-is */ break; } @@ -2694,10 +2709,13 @@ static int sev_handle_vmgexit_msr_protocol(struct vcpu_svm *svm) GHCB_MSR_TERM_REASON_POS); pr_info("SEV-ES guest requested termination: %#llx:%#llx\n", reason_set, reason_code); - fallthrough; + + ret = -EINVAL; + break; } default: - ret = -EINVAL; + /* Error, keep GHCB MSR value as-is */ + break; } trace_kvm_vmgexit_msr_protocol_exit(svm->vcpu.vcpu_id, @@ -2721,14 +2739,18 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) if (!ghcb_gpa) { vcpu_unimpl(vcpu, "vmgexit: GHCB gpa is not set\n"); - return -EINVAL; + + /* Without a GHCB, just return right back to the guest */ + return 1; } if (kvm_vcpu_map(vcpu, ghcb_gpa >> PAGE_SHIFT, &svm->sev_es.ghcb_map)) { /* Unable to map GHCB from guest */ vcpu_unimpl(vcpu, "vmgexit: error mapping GHCB [%#llx] from guest\n", ghcb_gpa); - return -EINVAL; + + /* Without a GHCB, just return right back to the guest */ + return 1; } svm->sev_es.ghcb = svm->sev_es.ghcb_map.hva; @@ -2738,18 +2760,17 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) exit_code = ghcb_get_sw_exit_code(ghcb); - ret = sev_es_validate_vmgexit(svm); - if (ret) - return ret; + if (!sev_es_validate_vmgexit(svm)) + return 1; sev_es_sync_from_ghcb(svm); ghcb_set_sw_exit_info_1(ghcb, 0); ghcb_set_sw_exit_info_2(ghcb, 0); + ret = 1; switch (exit_code) { case SVM_VMGEXIT_MMIO_READ: - ret = setup_vmgexit_scratch(svm, true, control->exit_info_2); - if (ret) + if (!setup_vmgexit_scratch(svm, true, control->exit_info_2)) break; ret = kvm_sev_es_mmio_read(vcpu, @@ -2758,8 +2779,7 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) svm->sev_es.ghcb_sa); break; case SVM_VMGEXIT_MMIO_WRITE: - ret = setup_vmgexit_scratch(svm, false, control->exit_info_2); - if (ret) + if (!setup_vmgexit_scratch(svm, false, control->exit_info_2)) break; ret = kvm_sev_es_mmio_write(vcpu, @@ -2788,14 +2808,10 @@ int sev_handle_vmgexit(struct kvm_vcpu *vcpu) default: pr_err("svm: vmgexit: unsupported AP jump table request - exit_info_1=%#llx\n", control->exit_info_1); - ghcb_set_sw_exit_info_1(ghcb, 1); - ghcb_set_sw_exit_info_2(ghcb, - X86_TRAP_UD | - SVM_EVTINJ_TYPE_EXEPT | - SVM_EVTINJ_VALID); + ghcb_set_sw_exit_info_1(ghcb, 2); + ghcb_set_sw_exit_info_2(ghcb, GHCB_ERR_INVALID_INPUT); } - ret = 1; break; } case SVM_VMGEXIT_UNSUPPORTED_EVENT: @@ -2815,7 +2831,6 @@ int sev_es_string_io(struct vcpu_svm *svm, int size, unsigned int port, int in) { int count; int bytes; - int r; if (svm->vmcb->control.exit_info_2 > INT_MAX) return -EINVAL; @@ -2824,9 +2839,8 @@ int sev_es_string_io(struct vcpu_svm *svm, int size, unsigned int port, int in) if (unlikely(check_mul_overflow(count, size, &bytes))) return -EINVAL; - r = setup_vmgexit_scratch(svm, in, bytes); - if (r) - return r; + if (!setup_vmgexit_scratch(svm, in, bytes)) + return 1; return kvm_sev_es_string_io(&svm->vcpu, size, port, svm->sev_es.ghcb_sa, count, in);
Currently, an SEV-ES guest is terminated if the validation of the VMGEXIT exit code or exit parameters fails. The VMGEXIT instruction can be issued from userspace, even though userspace (likely) can't update the GHCB. To prevent userspace from being able to kill the guest, return an error through the GHCB when validation fails rather than terminating the guest. For cases where the GHCB can't be updated (e.g. the GHCB can't be mapped, etc.), just return back to the guest. The new error codes are documented in the lasest update to the GHCB specification. Fixes: 291bd20d5d88 ("KVM: SVM: Add initial support for a VMGEXIT VMEXIT") Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> --- arch/x86/include/asm/sev-common.h | 11 ++++ arch/x86/kvm/svm/sev.c | 106 +++++++++++++++++------------- 2 files changed, 71 insertions(+), 46 deletions(-)