Add AMD SEV and SEV-ES intra host migration support
Subsystems (4)
v4 cover Add AMD SEV and SEV-ES intra host migration support
Intra host migration provides a low-cost mechanism for userspace VMM upgrades. It is an alternative to traditional (i.e., remote) live migration. Whereas remote migration handles moving a guest to a new host, intra host migration only handles moving a guest to a new userspace VMM within a host. This can be used to update, rollback, change flags of the VMM, etc. The lower cost compared to live migration comes from the fact that the guest's memory does not need to be copied between processes. A handle to the guest memory simply gets passed to the new VMM, this could be done via /dev/shm with share=on or similar feature.
The guest state can be transferred from an old VMM to a new VMM as follows:
1. Export guest state from KVM to the old user-space VMM via a getter user-space/kernel API 2. Transfer guest state from old VMM to new VMM via IPC communication 3. Import guest state into KVM from the new user-space VMM via a setter user-space/kernel API VMMs by exporting from KVM using getters, sending that data to the new VMM, then setting it again in KVM.
In the common case for intra host migration, we can rely on the normal ioctls for passing data from one VMM to the next. SEV, SEV-ES, and other confidential compute environments make most of this information opaque, and render KVM ioctls such as "KVM_GET_REGS" irrelevant. As a result, we need the ability to pass this opaque metadata from one VMM to the next. The easiest way to do this is to leave this data in the kernel, and transfer ownership of the metadata from one KVM VM (or vCPU) to the next. For example, we need to move the SEV enabled ASID, VMSAs, and GHCB metadata from one VMM to the next. In general, we need to be able to hand off any data that would be unsafe/impossible for the kernel to hand directly to userspace (and cannot be reproduced using data that can be handed safely to userspace).
For the intra host operation the SEV required metadata, the source VM FD is sent to the target VMM. The target VMM calls the new cap ioctl with the source VM FD, KVM then moves all the SEV state to the target VM from the source VM.
V4:
* Move to seanjc@'s suggestion of source VM FD based single ioctl design.
v3:
* Fix memory leak found by dan.carpenter@
v2:
* Added marcorr@ reviewed by tag
* Renamed function introduced in 1/3
* Edited with seanjc@'s review comments
** Cleaned up WARN usage
** Userspace makes random token now
* Edited with brijesh.singh@'s review comments ** Checks for different LAUNCH_* states in send function
v1: https://lore.kernel.org/kvm/[email protected]/
Peter Gonda (2):
KVM, SEV: Add support for SEV intra host migration KVM, SEV: Add support for SEV-ES intra host migration
Documentation/virt/kvm/api.rst | 15 +++ arch/x86/include/asm/kvm_host.h | 1 + arch/x86/kvm/svm/sev.c | 160 ++++++++++++++++++++++++++++++++ arch/x86/kvm/svm/svm.c | 1 + arch/x86/kvm/svm/svm.h | 3 + arch/x86/kvm/x86.c | 5 + include/uapi/linux/kvm.h | 1 + 7 files changed, 186 insertions(+)
base-commit: a3e0b8bd99ab
Signature · 1 line
2.33.0.rc1.237.g0d66db33f3-goog
PATCH v4 1/2 KVM, SEV: Add support for SEV intra host migration
For SEV to work with intra host migration, contents of the SEV info struct such as the ASID (used to index the encryption key in the AMD SP) and the list of memory regions need to be transferred to the target VM. This change adds a commands for a target VMM to get a source SEV VM's sev info.
The target is expected to be initialized (sev_guest_init), but not launched state (sev_launch_start) when performing receive. Once the target has received, it will be in a launched state and will not need to perform the typical SEV launch commands.
--- Documentation/virt/kvm/api.rst | 15 +++++ arch/x86/include/asm/kvm_host.h | 1 + arch/x86/kvm/svm/sev.c | 108 ++++++++++++++++++++++++++++++++ arch/x86/kvm/svm/svm.c | 1 + arch/x86/kvm/svm/svm.h | 3 + arch/x86/kvm/x86.c | 5 ++ include/uapi/linux/kvm.h | 1 + 7 files changed, 134 insertions(+)
Documentation/virt/kvm/api.rst +15 -0
@@ -6701,6 +6701,21 @@ MAP_SHARED mmap will result in an -EINVAL return. When enabled the VMM may make use of the ``KVM_ARM_MTE_COPY_TAGS`` ioctl to perform a bulk copy of tags to/from the guest. +7.29 KVM_CAP_VM_MIGRATE_ENC_CONTEXT_FROM +------------------------------------- + +Architectures: x86 SEV enabled +Type: vm +Parameters: args[0] is the fd of the source vm +Returns: 0 on success + +This capability enables userspace to migrate the encryption context from the vm +indicated by the fd to the vm this is called on. + +This is intended to support intra-host migration of VMs between userspace VMMs. +in-guest workloads scheduled by the host. This allows for upgrading the VMM +process without interrupting the guest. + 8. Other capabilities. ======================
arch/x86/include/asm/kvm_host.h +1 -0
@@ -1448,6 +1448,7 @@ struct kvm_x86_ops {int (*mem_enc_reg_region)(struct kvm *kvm,struct kvm_enc_region *argp);int (*mem_enc_unreg_region)(struct kvm *kvm,struct kvm_enc_region *argp);int (*vm_copy_enc_context_from)(struct kvm *kvm,unsigned int source_fd); +int (*vm_migrate_enc_context_from)(struct kvm *kvm,unsigned int source_fd);int (*get_msr_feature)(struct kvm_msr_entry *entry);
arch/x86/kvm/svm/sev.c +108 -0
114 lines not on this page · raw message
arch/x86/kvm/svm/svm.c +1 -0
@@ -4627,6 +4627,7 @@ static struct kvm_x86_ops svm_x86_ops __initdata = { .mem_enc_unreg_region = svm_unregister_enc_region, .vm_copy_enc_context_from = svm_vm_copy_asid_from, + .vm_migrate_enc_context_from = svm_vm_migrate_from, .can_emulate_instruction = svm_can_emulate_instruction,
arch/x86/kvm/svm/svm.h +3 -0
@@ -80,6 +80,8 @@ struct kvm_sev_info {u64 ap_jump_table;/* SEV-ES AP Jump Table address */ struct kvm *enc_context_owner;/* Owner of copied encryption context */ struct misc_cg *misc_cg;/* For misc cgroup accounting */ +spinlock_t migration_lock; +bool migration_in_progress; };struct kvm_svm {
@@ -552,6 +554,7 @@ int svm_register_enc_region(struct kvm *kvm,int svm_unregister_enc_region (struct kvm *kvm,struct kvm_enc_region *range);int svm_vm_copy_asid_from (struct kvm *kvm,unsigned int source_fd); +int svm_vm_migrate_from (struct kvm *kvm,unsigned int source_fd);void pre_sev_run (struct vcpu_svm *svm,int cpu);void __init sev_set_cpu_caps (void );void __init sev_hardware_setup (void );
arch/x86/kvm/x86.c +5 -0
@@ -5655,6 +5655,11 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,if (kvm_x86_ops.vm_copy_enc_context_from) r = kvm_x86_ops.vm_copy_enc_context_from (kvm, cap->args[0 ]);return r; +case KVM_CAP_VM_MIGRATE_ENC_CONTEXT_FROM : + r = -EINVAL ; +if (kvm_x86_ops.vm_migrate_enc_context_from) + r = kvm_x86_ops.vm_migrate_enc_context_from (kvm, cap->args[0 ]); +return r;case KVM_CAP_EXIT_HYPERCALL :if (cap->args[0 ] & ~KVM_EXIT_HYPERCALL_VALID_MASK ) { r = -EINVAL ;
include/uapi/linux/kvm.h +1 -0
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {#define KVM_CAP_BINARY_STATS_FD 203#define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204#define KVM_CAP_ARM_MTE 205 +#define KVM_CAP_VM_MIGRATE_ENC_CONTEXT_FROM 206#ifdef KVM_CAP_IRQ_ROUTING
Signature · 1 line
2.33.0.rc1.237.g0d66db33f3-goog
Re: [PATCH 1/2 V4] KVM, SEV: Add support for SEV intra host migration
Quoted 109 lines · Peter Gonda
For SEV to work with intra host migration, contents of the SEV info struct such as the ASID (used to index the encryption key in the AMD SP) and the list of memory regions need to be transferred to the target VM. This change adds a commands for a target VMM to get a source SEV VM's sev info.
The target is expected to be initialized (
sev_guest_init), but not launched state (sev_launch_start) when performing receive. Once the target has received, it will be in a launched state and will not need to perform the typical SEV launch commands.Signed-off-by Peter Gonda <[email protected]>Suggested-by Sean Christopherson <[email protected]>Cc Marc Orr <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Sean Christopherson <[email protected]>Cc David Rientjes <[email protected]>Cc Dr. David Alan Gilbert <[email protected]>Cc Brijesh Singh <[email protected]>Cc Vitaly Kuznetsov <[email protected]>Cc Wanpeng Li <[email protected]>Cc Jim Mattson <[email protected]>Cc Joerg Roedel <[email protected]>Cc Thomas Gleixner <[email protected]>Cc Ingo Molnar <[email protected]>Cc Borislav Petkov <[email protected]>Cc "H. Peter Anvin" <[email protected]>--- Documentation/virt/kvm/api.rst | 15 +++++ arch/x86/include/asm/kvm_host.h | 1 + arch/x86/kvm/svm/sev.c | 108 ++++++++++++++++++++++++++++++++ arch/x86/kvm/svm/svm.c | 1 + arch/x86/kvm/svm/svm.h | 3 + arch/x86/kvm/x86.c | 5 ++ include/uapi/linux/kvm.h | 1 + 7 files changed, 134 insertions(+)Documentation/virt/kvm/api.rst +15
-0@@ -6701,6 +6701,21 @@ MAP_SHARED mmap will result in an -EINVAL return. When enabled the VMM may make use of the ``KVM_ARM_MTE_COPY_TAGS`` ioctl to perform a bulk copy of tags to/from the guest. +7.29 KVM_CAP_VM_MIGRATE_ENC_CONTEXT_FROM +------------------------------------- + +Architectures: x86 SEV enabled +Type: vm +Parameters: args[0] is the fd of the source vm +Returns: 0 on success + +This capability enables userspace to migrate the encryption context from the vm +indicated by the fd to the vm this is called on. + +This is intended to support intra-host migration of VMs between userspace VMMs. +in-guest workloads scheduled by the host. This allows for upgrading the VMM +process without interrupting the guest. + 8. Other capabilities. ======================arch/x86/include/asm/kvm_host.h +1
-0@@ -1448,6 +1448,7 @@ struct kvm_x86_ops {int (*mem_enc_reg_region)(struct kvm *kvm,struct kvm_enc_region *argp);int (*mem_enc_unreg_region)(struct kvm *kvm,struct kvm_enc_region *argp);int (*vm_copy_enc_context_from)(struct kvm *kvm,unsigned int source_fd); +int (*vm_migrate_enc_context_from)(struct kvm *kvm,unsigned int source_fd);int (*get_msr_feature)(struct kvm_msr_entry *entry);arch/x86/kvm/svm/sev.c +23
-0@@ -1501,6 +1501,114 @@ static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp)return sev_issue_cmd (kvm,SEV_CMD_RECEIVE_FINISH , &data, &argp->error); } +static int svm_sev_lock_for_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; +int ret; + +/* +* Bail if this VM is already involved in a migration to avoid deadlock +* between two VMs trying to migrate to/from each other. +*/ +spin_lock (&sev->migration_lock); +if (sev->migration_in_progress) + ret = -EBUSY ; +else { +/* +* Otherwise indicate VM is migrating and take the KVM lock. +*/ + sev->migration_in_progress = true; +mutex_lock (&kvm->lock); + ret =0 ; + } +spin_unlock (&sev->migration_lock); + +return ret;
arch/x86/kvm/svm/sev.c static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp) · line 1536
Quoted 10 lines
arch/x86/kvm/svm/sev.c +10
-0+} + +static void svm_unlock_after_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; + +mutex_unlock (&kvm->lock); +WRITE_ONCE (sev->migration_in_progress, false); +} +
This entire locking scheme seems over-complicated to me. Can we simply rely on migration_lock and get rid of migration_in_progress? I was chatting about these patches with Peter, while he worked on this new version. But he mentioned that this locking scheme had been suggested by Sean in a previous review. Sean: what do you think? My rationale was that this is called via a VM-level ioctl. So serializing the entire code path on migration_lock seems fine. But maybe I'm missing something?
Quoted 142 lines
arch/x86/kvm/svm/sev.c +75
-0+static void migrate_info_from (struct kvm_sev_info *dst, +struct kvm_sev_info *src) +{ +sev_asid_free (dst); + + dst->asid = src->asid; + dst->misc_cg = src->misc_cg; + dst->handle = src->handle; + dst->pages_locked = src->pages_locked; + + src->asid =0 ; + src->active = false; + src->handle =0 ; + src->pages_locked =0 ; + src->misc_cg =NULL ; + +INIT_LIST_HEAD (&dst->regions_list); +list_replace_init (&src->regions_list, &dst->regions_list); +} + +int svm_vm_migrate_from (struct kvm *kvm,unsigned int source_fd) +{ +struct kvm_sev_info *dst_sev = &to_kvm_svm (kvm)->sev_info; +struct file *source_kvm_file; +struct kvm *source_kvm; +int ret; + + ret =svm_sev_lock_for_migration (kvm); +if (ret) +return ret; + +if (!sev_guest (kvm) ||sev_es_guest (kvm)) { + ret = -EINVAL ; +pr_warn_ratelimited ("VM must be SEV enabled to migrate to.\n" ); + goto out_unlock; + } + +if (!list_empty (&dst_sev->regions_list)) { + ret = -EINVAL ; +pr_warn_ratelimited ( +"VM must not have encrypted regions to migrate to.\n" ); + goto out_unlock; + } + + source_kvm_file =fget (source_fd); +if (!file_is_kvm (source_kvm_file)) { + ret = -EBADF ; + goto out_fput; + } + + source_kvm = source_kvm_file->private_data; + ret =svm_sev_lock_for_migration (source_kvm); +if (ret) + goto out_fput; + +if (!sev_guest (source_kvm) ||sev_es_guest (source_kvm)) { + ret = -EINVAL ; +pr_warn_ratelimited ( +"Source VM must be SEV enabled to migrate from.\n" ); + goto out_source; + } + +migrate_info_from (dst_sev, &to_kvm_svm (source_kvm)->sev_info); + ret =0 ; + +out_source: +svm_unlock_after_migration (source_kvm); +out_fput: +if (source_kvm_file) +fput (source_kvm_file); +out_unlock: +svm_unlock_after_migration (kvm); +return ret; +} +int svm_mem_enc_op (struct kvm *kvm,void __user *argp) {struct kvm_sev_cmd sev_cmd;arch/x86/kvm/svm/svm.c +1
-0@@ -4627,6 +4627,7 @@ static struct kvm_x86_ops svm_x86_ops __initdata = { .mem_enc_unreg_region = svm_unregister_enc_region, .vm_copy_enc_context_from = svm_vm_copy_asid_from, + .vm_migrate_enc_context_from = svm_vm_migrate_from, .can_emulate_instruction = svm_can_emulate_instruction,arch/x86/kvm/svm/svm.h +3
-0@@ -80,6 +80,8 @@ struct kvm_sev_info {u64 ap_jump_table;/* SEV-ES AP Jump Table address */ struct kvm *enc_context_owner;/* Owner of copied encryption context */ struct misc_cg *misc_cg;/* For misc cgroup accounting */ +spinlock_t migration_lock; +bool migration_in_progress; };struct kvm_svm {@@ -552,6 +554,7 @@ int svm_register_enc_region(struct kvm *kvm,int svm_unregister_enc_region (struct kvm *kvm,struct kvm_enc_region *range);int svm_vm_copy_asid_from (struct kvm *kvm,unsigned int source_fd); +int svm_vm_migrate_from (struct kvm *kvm,unsigned int source_fd);void pre_sev_run (struct vcpu_svm *svm,int cpu);void __init sev_set_cpu_caps (void );void __init sev_hardware_setup (void );arch/x86/kvm/x86.c +5
-0@@ -5655,6 +5655,11 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,if (kvm_x86_ops.vm_copy_enc_context_from) r = kvm_x86_ops.vm_copy_enc_context_from (kvm, cap->args[0 ]);return r; +case KVM_CAP_VM_MIGRATE_ENC_CONTEXT_FROM : + r = -EINVAL ; +if (kvm_x86_ops.vm_migrate_enc_context_from) + r = kvm_x86_ops.vm_migrate_enc_context_from (kvm, cap->args[0 ]); +return r;case KVM_CAP_EXIT_HYPERCALL :if (cap->args[0 ] & ~KVM_EXIT_HYPERCALL_VALID_MASK ) { r = -EINVAL ;include/uapi/linux/kvm.h +1
-1@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {#define KVM_CAP_BINARY_STATS_FD 203#define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204#define KVM_CAP_ARM_MTE 205 +#define KVM_CAP_VM_MIGRATE_ENC_CONTEXT_FROM 206#ifdef KVM_CAP_IRQ_ROUTING --2.33.0.rc1.237.g0d66db33f3-goog
Re: [PATCH 1/2 V4] KVM, SEV: Add support for SEV intra host migration
Quoted 35 lines
Quoted 24 lines
arch/x86/kvm/svm/sev.c +23
-0+static int svm_sev_lock_for_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; +int ret; + +/* +* Bail if this VM is already involved in a migration to avoid deadlock +* between two VMs trying to migrate to/from each other. +*/ +spin_lock (&sev->migration_lock); +if (sev->migration_in_progress) + ret = -EBUSY ; +else { +/* +* Otherwise indicate VM is migrating and take the KVM lock. +*/ + sev->migration_in_progress = true; +mutex_lock (&kvm->lock); + ret =0 ; + } +spin_unlock (&sev->migration_lock); + +return ret;arch/x86/kvm/svm/sev.c static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp) · line 1536
Quoted 10 lines
arch/x86/kvm/svm/sev.c +10
-0+} + +static void svm_unlock_after_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; + +mutex_unlock (&kvm->lock); +WRITE_ONCE (sev->migration_in_progress, false); +} +
Quoted 9 lines
This entire locking scheme seems over-complicated to me. Can we simply rely on
migration_lockand get rid ofmigration_in_progress? I was chatting about these patches with Peter, while he worked on this new version. But he mentioned that this locking scheme had been suggested by Sean in a previous review. Sean: what do you think? My rationale was that this is called via a VM-level ioctl. So serializing the entire code path onmigration_lockseems fine. But maybe I'm missing something?
Marc I think that only having the spin lock could result in deadlocking. If userspace double migrated 2 VMs, A and B for discussion, A could grab VM_A.spin_lock then VM_A.kvm_mutex. Meanwhile B could grab VM_B.spin_lock and VM_B.kvm_mutex. Then A attempts to grab VM_B.spin_lock and we have a deadlock. If the same happens with the proposed scheme when A attempts to lock B, VM_B.spin_lock will be open but the bool will mark the VM under migration so A will unlock and bail. Sean originally proposed a global spin lock but I thought a per kvm_sev_info struct would also be safe.
Sean Christopherson
maintainerRe: [PATCH 1/2 V4] KVM, SEV: Add support for SEV intra host migration
Quoted 11 lines · Peter Gonda
Quoted 11 lines
Quoted 11 lines
arch/x86/kvm/svm/sev.c +10
-0+static int svm_sev_lock_for_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; +int ret; + +/* +* Bail if this VM is already involved in a migration to avoid deadlock +* between two VMs trying to migrate to/from each other. +*/ +spin_lock (&sev->migration_lock);
Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/svm/sev.c static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp) · line 1521
Quoted 8 lines
arch/x86/kvm/svm/sev.c +8
-0+if (sev->migration_in_progress) + ret = -EBUSY ; +else { +/* +* Otherwise indicate VM is migrating and take the KVM lock. +*/ + sev->migration_in_progress = true; +mutex_lock (&kvm->lock);
Deadlock aside, mutex_lock() can sleep, which is not allowed while holding a spinlock, i.e. this patch does not work. That's my suggestion did the crazy dance of "acquiring" a flag.
What I don't know is why on earth I suggested a global spinlock, a simple atomic should work, e.g.
if (atomic_cmpxchg_acquire(&sev->migration_in_progress, 0, 1)) return -EBUSY; mutex_lock(&kvm->lock);
and on the backend...
mutex_unlock(&kvm->lock); atomic_set_release(&sev->migration_in_progress, 0);
Quoted 26 lines
Quoted 16 lines
Quoted 5 lines
arch/x86/kvm/svm/sev.c +5
-0+ ret =0 ; + } +spin_unlock (&sev->migration_lock); + +return ret;arch/x86/kvm/svm/sev.c static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp) · line 1536
Quoted 10 lines
arch/x86/kvm/svm/sev.c +10
-0+} + +static void svm_unlock_after_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; + +mutex_unlock (&kvm->lock); +WRITE_ONCE (sev->migration_in_progress, false); +} +Quoted 8 lines
This entire locking scheme seems over-complicated to me. Can we simply rely on
migration_lockand get rid ofmigration_in_progress? I was chatting about these patches with Peter, while he worked on this new version. But he mentioned that this locking scheme had been suggested by Sean in a previous review. Sean: what do you think? My rationale was that this is called via a VM-level ioctl. So serializing the entire code path onmigration_lockseems fine. But maybe I'm missing something?
Quoted 9 lines
Marc I think that only having the spin lock could result in deadlocking. If userspace double migrated 2 VMs, A and B for discussion, A could grab VM_A.spin_lock then VM_A.kvm_mutex. Meanwhile B could grab VM_B.spin_lock and VM_B.kvm_mutex. Then A attempts to grab VM_B.spin_lock and we have a deadlock. If the same happens with the proposed scheme when A attempts to lock B, VM_B.spin_lock will be open but the bool will mark the VM under migration so A will unlock and bail. Sean originally proposed a global spin lock but I thought a per
kvm_sev_infostruct would also be safe.
Close. The issue is taking kvm->lock from both VM_A and VM_B. If userspace double migrates we'll end up with lock ordering A->B and B-A, so we need a way to guarantee one of those wins. My proposed solution is to use a flag as a sort of one-off "try lock" to detect a mean userspace.
Re: [PATCH 1/2 V4] KVM, SEV: Add support for SEV intra host migration
Quoted 76 lines · Sean Christopherson
Quoted 11 lines · Peter Gonda
Quoted 11 lines
Quoted 11 lines
arch/x86/kvm/svm/sev.c +10
-0+static int svm_sev_lock_for_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; +int ret; + +/* +* Bail if this VM is already involved in a migration to avoid deadlock +* between two VMs trying to migrate to/from each other. +*/ +spin_lock (&sev->migration_lock);Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/svm/sev.c static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp) · line 1521
Quoted 8 lines
arch/x86/kvm/svm/sev.c +8
-0+if (sev->migration_in_progress) + ret = -EBUSY ; +else { +/* +* Otherwise indicate VM is migrating and take the KVM lock. +*/ + sev->migration_in_progress = true; +mutex_lock (&kvm->lock);Deadlock aside,
mutex_lock()can sleep, which is not allowed while holding a spinlock, i.e. this patch does not work. That's my suggestion did the crazy dance of "acquiring" a flag.What I don't know is why on earth I suggested a global spinlock, a simple atomic should work, e.g.
if (atomic_cmpxchg_acquire(&sev->migration_in_progress, 0, 1)) return -EBUSY; mutex_lock(&kvm->lock);and on the backend...
mutex_unlock(&kvm->lock); atomic_set_release(&sev->migration_in_progress, 0);Quoted 26 lines
Quoted 16 lines
Quoted 5 lines
arch/x86/kvm/svm/sev.c +5
-0+ ret =0 ; + } +spin_unlock (&sev->migration_lock); + +return ret;arch/x86/kvm/svm/sev.c static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp) · line 1536
Quoted 10 lines
arch/x86/kvm/svm/sev.c +10
-0+} + +static void svm_unlock_after_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; + +mutex_unlock (&kvm->lock); +WRITE_ONCE (sev->migration_in_progress, false); +} +Quoted 8 lines
This entire locking scheme seems over-complicated to me. Can we simply rely on
migration_lockand get rid ofmigration_in_progress? I was chatting about these patches with Peter, while he worked on this new version. But he mentioned that this locking scheme had been suggested by Sean in a previous review. Sean: what do you think? My rationale was that this is called via a VM-level ioctl. So serializing the entire code path onmigration_lockseems fine. But maybe I'm missing something?Quoted 9 lines
Marc I think that only having the spin lock could result in deadlocking. If userspace double migrated 2 VMs, A and B for discussion, A could grab VM_A.spin_lock then VM_A.kvm_mutex. Meanwhile B could grab VM_B.spin_lock and VM_B.kvm_mutex. Then A attempts to grab VM_B.spin_lock and we have a deadlock. If the same happens with the proposed scheme when A attempts to lock B, VM_B.spin_lock will be open but the bool will mark the VM under migration so A will unlock and bail. Sean originally proposed a global spin lock but I thought a per
kvm_sev_infostruct would also be safe.
Quoted 4 lines
Close. The issue is taking
kvm->lockfrom bothVM_AandVM_B. If userspace double migrates we'll end up with lock orderingA->Band B-A, so we need a way to guarantee one of those wins. My proposed solution is to use a flag as a sort of one-off "try lock" to detect a mean userspace.
Got it now. Thanks to you both, for the explanation. By the way, just to make sure I completely follow, I assume that if a "double migration" occurs, then user space is mis-behaving -- correct? But presumably, we need to reason about how to respond to such mis-behavior so that buggy or malicious user-space code cannot stumble over/exploit this scenario?
Sean Christopherson
maintainerRe: [PATCH 1/2 V4] KVM, SEV: Add support for SEV intra host migration
Quoted 13 lines · Marc Orr
Quoted 12 lines · Sean Christopherson
Quoted 9 lines · Peter Gonda
Marc I think that only having the spin lock could result in deadlocking. If userspace double migrated 2 VMs, A and B for discussion, A could grab VM_A.spin_lock then VM_A.kvm_mutex. Meanwhile B could grab VM_B.spin_lock and VM_B.kvm_mutex. Then A attempts to grab VM_B.spin_lock and we have a deadlock. If the same happens with the proposed scheme when A attempts to lock B, VM_B.spin_lock will be open but the bool will mark the VM under migration so A will unlock and bail. Sean originally proposed a global spin lock but I thought a per
kvm_sev_infostruct would also be safe.
Quoted 8 lines
Quoted 4 lines
Close. The issue is taking
kvm->lockfrom bothVM_AandVM_B. If userspace double migrates we'll end up with lock orderingA->Band B-A, so we need a way to guarantee one of those wins. My proposed solution is to use a flag as a sort of one-off "try lock" to detect a mean userspace.Got it now. Thanks to you both, for the explanation. By the way, just to make sure I completely follow, I assume that if a "double migration" occurs, then user space is mis-behaving -- correct?
Yep.
Quoted 3 lines
But presumably, we need to reason about how to respond to such mis-behavior so that buggy or malicious user-space code cannot stumble over/exploit this scenario?
That's what the anti-deadlock flag is for. :-) With that in place, there's no meaningful difference between say a bad userspace doing double migrate and a bad userspace migrating from garbage, e.g. passing in a bogus fd.
Re: [PATCH 1/2 V4] KVM, SEV: Add support for SEV intra host migration
In reply to Sean ChristophersonQuoted 34 lines · Sean Christopherson
Quoted 11 lines · Peter Gonda
Quoted 11 lines
Quoted 11 lines
arch/x86/kvm/svm/sev.c +10
-0+static int svm_sev_lock_for_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; +int ret; + +/* +* Bail if this VM is already involved in a migration to avoid deadlock +* between two VMs trying to migrate to/from each other. +*/ +spin_lock (&sev->migration_lock);Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/svm/sev.c static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp) · line 1521
Quoted 8 lines
arch/x86/kvm/svm/sev.c +8
-0+if (sev->migration_in_progress) + ret = -EBUSY ; +else { +/* +* Otherwise indicate VM is migrating and take the KVM lock. +*/ + sev->migration_in_progress = true; +mutex_lock (&kvm->lock);Deadlock aside,
mutex_lock()can sleep, which is not allowed while holding a spinlock, i.e. this patch does not work. That's my suggestion did the crazy dance of "acquiring" a flag.What I don't know is why on earth I suggested a global spinlock, a simple atomic should work, e.g.
if (atomic_cmpxchg_acquire(&sev->migration_in_progress, 0, 1)) return -EBUSY; mutex_lock(&kvm->lock);
Quoted 5 lines
and on the backend...
mutex_unlock(&kvm->lock); atomic_set_release(&sev->migration_in_progress, 0);
+1 to replacing the spin lock with an atomic flag. Correctness issues aside, I think it's also cleaner. Also, I'd suggest adding a comment to source code to explain that the migration_in_progress flag is to prevent deadlock due to the "double migration" discussed previously.
Re: [PATCH 1/2 V4] KVM, SEV: Add support for SEV intra host migration
Quoted 19 lines · Marc Orr
Quoted 17 lines · Sean Christopherson
Quoted 15 lines · Peter Gonda
Quoted 15 lines
Quoted 15 lines
arch/x86/kvm/svm/sev.c +14
-0+static int svm_sev_lock_for_migration (struct kvm *kvm) +{ +struct kvm_sev_info *sev = &to_kvm_svm (kvm)->sev_info; +int ret; + +/* +* Bail if this VM is already involved in a migration to avoid deadlock +* between two VMs trying to migrate to/from each other. +*/ +spin_lock (&sev->migration_lock); +if (sev->migration_in_progress) + ret = -EBUSY ; +else { +/*
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
Quoted 4 lines
arch/x86/kvm/svm/sev.c static int sev_receive_finish(struct kvm *kvm, struct kvm_sev_cmd *argp) · line 1521
Quoted 4 lines
arch/x86/kvm/svm/sev.c +4
-0+* Otherwise indicate VM is migrating and take the KVM lock. +*/ + sev->migration_in_progress = true; +mutex_lock (&kvm->lock);Deadlock aside,
mutex_lock()can sleep, which is not allowed while holding a spinlock, i.e. this patch does not work. That's my suggestion did the crazy dance of "acquiring" a flag.
Ah, makes sense.
Quoted 11 lines
Quoted 11 lines
What I don't know is why on earth I suggested a global spinlock, a simple atomic should work, e.g.
if (atomic_cmpxchg_acquire(&sev->migration_in_progress, 0, 1)) return -EBUSY; mutex_lock(&kvm->lock);and on the backend...
Quoted 8 lines
Quoted 3 lines
mutex_unlock(&kvm->lock); atomic_set_release(&sev->migration_in_progress, 0);+1 to replacing the spin lock with an atomic flag. Correctness issues aside, I think it's also cleaner. Also, I'd suggest adding a comment to source code to explain that the
migration_in_progressflag is to prevent deadlock due to the "double migration" discussed previously.
Thanks! I've updated these locks to use the atomic. It looks much cleaner now.
PATCH v4 2/2 KVM, SEV: Add support for SEV-ES intra host migration
In reply to Peter GondaFor SEV-ES to work with intra host migration the VMSAs, GHCB metadata, and other SEV-ES info needs to be preserved along with the guest's memory.
--- arch/x86/kvm/svm/sev.c | 58 +++++++++++++++++++++++++++++++++++++++--- 1 file changed, 55 insertions(+), 3 deletions(-)
arch/x86/kvm/svm/sev.c +55 -3
@@ -1554,6 +1554,53 @@ static void migrate_info_from(struct kvm_sev_info *dst,list_replace_init (&src->regions_list, &dst->regions_list); } +static int migrate_vmsa_from (struct kvm *dst,struct kvm *src) +{ +int i, num_vcpus; +struct kvm_vcpu *dst_vcpu, *src_vcpu; +struct vcpu_svm *dst_svm, *src_svm; + + num_vcpus =atomic_read (&dst->online_vcpus); +if (num_vcpus !=atomic_read (&src->online_vcpus)) { +pr_warn_ratelimited ( +"Source and target VMs must have same number of vCPUs.\n" ); +return -EINVAL ; + } + +for (i =0 ; i < num_vcpus; ++i) { + src_vcpu = src->vcpus[i]; +if (!src_vcpu->arch.guest_state_protected) { +pr_warn_ratelimited ( +"Source ES VM vCPUs must have protected state.\n" ); +return -EINVAL ; + } + } + +for (i =0 ; i < num_vcpus; ++i) { + src_vcpu = src->vcpus[i]; + src_svm =to_svm (src_vcpu); + dst_vcpu = dst->vcpus[i]; + dst_svm =to_svm (dst_vcpu); + + dst_vcpu->vcpu_id = src_vcpu->vcpu_id; + dst_svm->vmsa = src_svm->vmsa; + src_svm->vmsa =NULL ; + dst_svm->ghcb = src_svm->ghcb; + src_svm->ghcb =NULL ; + dst_svm->vmcb->control.ghcb_gpa = + src_svm->vmcb->control.ghcb_gpa; + dst_svm->ghcb_sa = src_svm->ghcb_sa; + src_svm->ghcb_sa =NULL ; + dst_svm->ghcb_sa_len = src_svm->ghcb_sa_len; + src_svm->ghcb_sa_len =0 ; + dst_svm->ghcb_sa_sync = src_svm->ghcb_sa_sync; + src_svm->ghcb_sa_sync = false; + dst_svm->ghcb_sa_free = src_svm->ghcb_sa_free; + src_svm->ghcb_sa_free = false; + } +return 0 ; +} +int svm_vm_migrate_from (struct kvm *kvm,unsigned int source_fd) {struct kvm_sev_info *dst_sev = &to_kvm_svm (kvm)->sev_info;
@@ -1565,7 +1612,7 @@ int svm_vm_migrate_from(struct kvm *kvm, unsigned int source_fd)if (ret)return ret; -if (!sev_guest (kvm) ||sev_es_guest (kvm)) { +if (!sev_guest (kvm)) { ret = -EINVAL ;pr_warn_ratelimited ("VM must be SEV enabled to migrate to.\n" ); goto out_unlock;
@@ -1589,15 +1636,20 @@ int svm_vm_migrate_from(struct kvm *kvm, unsigned int source_fd)if (ret) goto out_fput; -if (!sev_guest (source_kvm) ||sev_es_guest (source_kvm)) { +if (!sev_guest (source_kvm)) { ret = -EINVAL ;pr_warn_ratelimited ("Source VM must be SEV enabled to migrate from.\n" ); goto out_source; } +if (sev_es_guest (kvm)) { + ret =migrate_vmsa_from (kvm, source_kvm); +if (ret) + goto out_source; + }migrate_info_from (dst_sev, &to_kvm_svm (source_kvm)->sev_info); - ret =0 ; + ret =0 ; out_source:svm_unlock_after_migration (source_kvm);
Signature · 1 line
2.33.0.rc1.237.g0d66db33f3-goog
Re: [PATCH 2/2 V4] KVM, SEV: Add support for SEV-ES intra host migration
Quoted 59 lines · Peter Gonda
For SEV-ES to work with intra host migration the VMSAs, GHCB metadata, and other SEV-ES info needs to be preserved along with the guest's memory.
Signed-off-by Peter Gonda <[email protected]>Cc Marc Orr <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Sean Christopherson <[email protected]>Cc David Rientjes <[email protected]>Cc Dr. David Alan Gilbert <[email protected]>Cc Brijesh Singh <[email protected]>Cc Vitaly Kuznetsov <[email protected]>Cc Wanpeng Li <[email protected]>Cc Jim Mattson <[email protected]>Cc Joerg Roedel <[email protected]>Cc Thomas Gleixner <[email protected]>Cc Ingo Molnar <[email protected]>Cc Borislav Petkov <[email protected]>Cc "H. Peter Anvin" <[email protected]>--- arch/x86/kvm/svm/sev.c | 58 +++++++++++++++++++++++++++++++++++++++--- 1 file changed, 55 insertions(+), 3 deletions(-)arch/x86/kvm/svm/sev.c +25
-0@@ -1554,6 +1554,53 @@ static void migrate_info_from(struct kvm_sev_info *dst,list_replace_init (&src->regions_list, &dst->regions_list); } +static int migrate_vmsa_from (struct kvm *dst,struct kvm *src) +{ +int i, num_vcpus; +struct kvm_vcpu *dst_vcpu, *src_vcpu; +struct vcpu_svm *dst_svm, *src_svm; + + num_vcpus =atomic_read (&dst->online_vcpus); +if (num_vcpus !=atomic_read (&src->online_vcpus)) { +pr_warn_ratelimited ( +"Source and target VMs must have same number of vCPUs.\n" ); +return -EINVAL ; + } + +for (i =0 ; i < num_vcpus; ++i) { + src_vcpu = src->vcpus[i]; +if (!src_vcpu->arch.guest_state_protected) { +pr_warn_ratelimited ( +"Source ES VM vCPUs must have protected state.\n" ); +return -EINVAL ; + } + } + +for (i =0 ; i < num_vcpus; ++i) { + src_vcpu = src->vcpus[i]; + src_svm =to_svm (src_vcpu);
arch/x86/kvm/svm/sev.c static void migrate_info_from(struct kvm_sev_info *dst, · line 1591
Quoted 10 lines
arch/x86/kvm/svm/sev.c +10
-0+ dst_vcpu = dst->vcpus[i]; + dst_svm =to_svm (dst_vcpu); + + dst_vcpu->vcpu_id = src_vcpu->vcpu_id; + dst_svm->vmsa = src_svm->vmsa; + src_svm->vmsa =NULL ; + dst_svm->ghcb = src_svm->ghcb; + src_svm->ghcb =NULL ; + dst_svm->vmcb->control.ghcb_gpa = + src_svm->vmcb->control.ghcb_gpa;
Should we clear src_svm->vmcb->control.ghcb_gpa? (All the other fields in srv_svm that are copied are then cleared, except for this one.) Aside: Do we really need to clear all of the fields in src_svm after they're copied over? It might be worth adding a comment to explain why we're clearing them. It's not immediately obvious to me why we're doing that.
Quoted 34 lines
arch/x86/kvm/svm/sev.c +14
-2+ dst_svm->ghcb_sa = src_svm->ghcb_sa; + src_svm->ghcb_sa =NULL ; + dst_svm->ghcb_sa_len = src_svm->ghcb_sa_len; + src_svm->ghcb_sa_len =0 ; + dst_svm->ghcb_sa_sync = src_svm->ghcb_sa_sync; + src_svm->ghcb_sa_sync = false; + dst_svm->ghcb_sa_free = src_svm->ghcb_sa_free; + src_svm->ghcb_sa_free = false; + } +return 0 ; +} +int svm_vm_migrate_from (struct kvm *kvm,unsigned int source_fd) {struct kvm_sev_info *dst_sev = &to_kvm_svm (kvm)->sev_info;@@ -1565,7 +1612,7 @@ int svm_vm_migrate_from(struct kvm *kvm, unsigned int source_fd)if (ret)return ret; -if (!sev_guest (kvm) ||sev_es_guest (kvm)) { +if (!sev_guest (kvm)) { ret = -EINVAL ;pr_warn_ratelimited ("VM must be SEV enabled to migrate to.\n" ); goto out_unlock;@@ -1589,15 +1636,20 @@ int svm_vm_migrate_from(struct kvm *kvm, unsigned int source_fd)if (ret) goto out_fput; -if (!sev_guest (source_kvm) ||sev_es_guest (source_kvm)) { +if (!sev_guest (source_kvm)) { ret = -EINVAL ;pr_warn_ratelimited ("Source VM must be SEV enabled to migrate from.\n" ); goto out_source;
arch/x86/kvm/svm/sev.c int svm_vm_migrate_from(struct kvm *kvm, unsigned int source_fd) · line 1652
Quoted 10 lines
arch/x86/kvm/svm/sev.c +6
-1} +if (sev_es_guest (kvm)) { + ret =migrate_vmsa_from (kvm, source_kvm); +if (ret) + goto out_source; + }migrate_info_from (dst_sev, &to_kvm_svm (source_kvm)->sev_info); - ret =0 ; + ret =0 ;
nit: looks like there's some space issue on this line. Since I believe this line was introduced in the previous patch, I wouldn't expect it to show up with a diff in patch #2.
Quoted 4 lines
arch/x86/kvm/svm/sev.c +0
-1out_source:svm_unlock_after_migration (source_kvm); --2.33.0.rc1.237.g0d66db33f3-goog
This patch looks good overall to me.
Re: [PATCH 2/2 V4] KVM, SEV: Add support for SEV-ES intra host migration
Quoted 70 lines · Marc Orr
Quoted 68 lines · Peter Gonda
For SEV-ES to work with intra host migration the VMSAs, GHCB metadata, and other SEV-ES info needs to be preserved along with the guest's memory.
Signed-off-by Peter Gonda <[email protected]>Cc Marc Orr <[email protected]>Cc Paolo Bonzini <[email protected]>Cc Sean Christopherson <[email protected]>Cc David Rientjes <[email protected]>Cc Dr. David Alan Gilbert <[email protected]>Cc Brijesh Singh <[email protected]>Cc Vitaly Kuznetsov <[email protected]>Cc Wanpeng Li <[email protected]>Cc Jim Mattson <[email protected]>Cc Joerg Roedel <[email protected]>Cc Thomas Gleixner <[email protected]>Cc Ingo Molnar <[email protected]>Cc Borislav Petkov <[email protected]>Cc "H. Peter Anvin" <[email protected]>--- arch/x86/kvm/svm/sev.c | 58 +++++++++++++++++++++++++++++++++++++++--- 1 file changed, 55 insertions(+), 3 deletions(-)arch/x86/kvm/svm/sev.c +34
-0@@ -1554,6 +1554,53 @@ static void migrate_info_from(struct kvm_sev_info *dst,list_replace_init (&src->regions_list, &dst->regions_list); } +static int migrate_vmsa_from (struct kvm *dst,struct kvm *src) +{ +int i, num_vcpus; +struct kvm_vcpu *dst_vcpu, *src_vcpu; +struct vcpu_svm *dst_svm, *src_svm; + + num_vcpus =atomic_read (&dst->online_vcpus); +if (num_vcpus !=atomic_read (&src->online_vcpus)) { +pr_warn_ratelimited ( +"Source and target VMs must have same number of vCPUs.\n" ); +return -EINVAL ; + } + +for (i =0 ; i < num_vcpus; ++i) { + src_vcpu = src->vcpus[i]; +if (!src_vcpu->arch.guest_state_protected) { +pr_warn_ratelimited ( +"Source ES VM vCPUs must have protected state.\n" ); +return -EINVAL ; + } + } + +for (i =0 ; i < num_vcpus; ++i) { + src_vcpu = src->vcpus[i]; + src_svm =to_svm (src_vcpu); + dst_vcpu = dst->vcpus[i]; + dst_svm =to_svm (dst_vcpu); + + dst_vcpu->vcpu_id = src_vcpu->vcpu_id; + dst_svm->vmsa = src_svm->vmsa; + src_svm->vmsa =NULL ; + dst_svm->ghcb = src_svm->ghcb; + src_svm->ghcb =NULL ; + dst_svm->vmcb->control.ghcb_gpa =
Quoted 8 lines
arch/x86/kvm/svm/sev.c static void migrate_info_from(struct kvm_sev_info *dst, · line 1591
Quoted 1 line
arch/x86/kvm/svm/sev.c +1
-0+ src_svm->vmcb->control.ghcb_gpa;Should we clear
src_svm->vmcb->control.ghcb_gpa? (All the other fields insrv_svmthat are copied are then cleared, except for this one.) Aside: Do we really need to clear all of the fields insrc_svmafter they're copied over? It might be worth adding a comment to explain why we're clearing them. It's not immediately obvious to me why we're doing that.
Cleared in the next version and added a comment. We don't want to leave references for the source VM to edit the VMSA or GHCB after the migration.
Quoted 37 lines
Quoted 37 lines
arch/x86/kvm/svm/sev.c +14
-2+ dst_svm->ghcb_sa = src_svm->ghcb_sa; + src_svm->ghcb_sa =NULL ; + dst_svm->ghcb_sa_len = src_svm->ghcb_sa_len; + src_svm->ghcb_sa_len =0 ; + dst_svm->ghcb_sa_sync = src_svm->ghcb_sa_sync; + src_svm->ghcb_sa_sync = false; + dst_svm->ghcb_sa_free = src_svm->ghcb_sa_free; + src_svm->ghcb_sa_free = false; + } +return 0 ; +} +int svm_vm_migrate_from (struct kvm *kvm,unsigned int source_fd) {struct kvm_sev_info *dst_sev = &to_kvm_svm (kvm)->sev_info;@@ -1565,7 +1612,7 @@ int svm_vm_migrate_from(struct kvm *kvm, unsigned int source_fd)if (ret)return ret; -if (!sev_guest (kvm) ||sev_es_guest (kvm)) { +if (!sev_guest (kvm)) { ret = -EINVAL ;pr_warn_ratelimited ("VM must be SEV enabled to migrate to.\n" ); goto out_unlock;@@ -1589,15 +1636,20 @@ int svm_vm_migrate_from(struct kvm *kvm, unsigned int source_fd)if (ret) goto out_fput; -if (!sev_guest (source_kvm) ||sev_es_guest (source_kvm)) { +if (!sev_guest (source_kvm)) { ret = -EINVAL ;pr_warn_ratelimited ("Source VM must be SEV enabled to migrate from.\n" ); goto out_source; }
Quoted 12 lines
arch/x86/kvm/svm/sev.c int svm_vm_migrate_from(struct kvm *kvm, unsigned int source_fd) · line 1652
Quoted 8 lines
arch/x86/kvm/svm/sev.c +6
-1+if (sev_es_guest (kvm)) { + ret =migrate_vmsa_from (kvm, source_kvm); +if (ret) + goto out_source; + }migrate_info_from (dst_sev, &to_kvm_svm (source_kvm)->sev_info); - ret =0 ; + ret =0 ;nit: looks like there's some space issue on this line. Since I believe this line was introduced in the previous patch, I wouldn't expect it to show up with a diff in patch #2.
Fixed thanks.
Quoted 9 lines
Quoted 5 lines
arch/x86/kvm/svm/sev.c +0
-1out_source:svm_unlock_after_migration (source_kvm); --2.33.0.rc1.237.g0d66db33f3-goog
This patch looks good overall to me.
Reviewed-by Marc Orr <[email protected]>