| Message ID | 20221205232341.4131240-5-vannapurve@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | KVM: selftests: selftests for fd-based private memory | expand |
On Mon, Dec 05, 2022, Vishal Annapurve wrote: > Introduce a set of APIs to execute VM with private memslots. > > Host userspace APIs for: > 1) Executing a vcpu run loop that handles MAPGPA hypercall > 2) Backing/unbacking guest private memory > > Guest APIs for: > 1) Changing memory mapping type > > Signed-off-by: Vishal Annapurve <vannapurve@google.com> > --- > tools/testing/selftests/kvm/Makefile | 1 + > .../kvm/include/x86_64/private_mem.h | 24 +++ > .../selftests/kvm/lib/x86_64/private_mem.c | 139 ++++++++++++++++++ > 3 files changed, 164 insertions(+) > create mode 100644 tools/testing/selftests/kvm/include/x86_64/private_mem.h > create mode 100644 tools/testing/selftests/kvm/lib/x86_64/private_mem.c Given that we _know_ private memory isn't always going to x86 specific, I don't want to bury any helpers in x86_64 that aren't strictly x86 only. E.g. helpers for doing Intel+AMD hypercalls is ok, but "generic" private memory helpers belong elsewhere. I experimented with extracting memslots/mmu helpers out of kvm_util.c as prep work, e.g. to avoid bloating kvm_util.c, but I couldn't come up with an obviously "good" naming scheme and/or split. At this time, the common bits are fairly small, so I think the best approach for now is to simply put vm_mem_map_shared_or_private() in kvm_util.c. > +static inline uint64_t __kvm_hypercall_map_gpa_range(uint64_t gpa, uint64_t size, > + uint64_t flags) > +{ > + return kvm_hypercall(KVM_HC_MAP_GPA_RANGE, gpa, size >> PAGE_SHIFT, flags, 0); > +} This can go in tools/testing/selftests/kvm/include/x86_64/processor.h. > +static inline void kvm_hypercall_map_gpa_range(uint64_t gpa, uint64_t size, > + uint64_t flags) > +{ > + uint64_t ret; > + > + GUEST_ASSERT_2(IS_PAGE_ALIGNED(gpa) && IS_PAGE_ALIGNED(size), gpa, size); > + > + ret = __kvm_hypercall_map_gpa_range(gpa, size, flags); > + GUEST_ASSERT_1(!ret, ret); > +} > + > +void kvm_hypercall_map_shared(uint64_t gpa, uint64_t size) > +{ > + kvm_hypercall_map_gpa_range(gpa, size, KVM_MAP_GPA_RANGE_DECRYPTED); > +} > + > +void kvm_hypercall_map_private(uint64_t gpa, uint64_t size) > +{ > + kvm_hypercall_map_gpa_range(gpa, size, KVM_MAP_GPA_RANGE_ENCRYPTED); > +} > + > +static void vm_update_private_mem(struct kvm_vm *vm, uint64_t gpa, uint64_t size, > + bool unback_mem) s/unback_mem/map_shared. "unback memory" is going to be super confusing to someone who isn't familiar with UPM. map_private would be the obvious alternative, but I like not having to invert the param in the helper. > +{ > + int restricted_fd; > + uint64_t restricted_fd_offset, guest_phys_base, fd_offset; > + struct kvm_memory_attributes attr; > + struct kvm_userspace_memory_region_ext *region_ext; > + struct kvm_userspace_memory_region *region; > + int fallocate_mode = 0; > + int ret; > + > + region_ext = kvm_userspace_memory_region_ext_find(vm, gpa, gpa + size); I forget if I've already mentioned this somewhere, but I'd prefer to use the "private" userspace_mem_region_find() and delete the existing kvm_userspace_memory_region_find(). > + TEST_ASSERT(region_ext != NULL, "Region not found"); > + region = ®ion_ext->region; > + TEST_ASSERT(region->flags & KVM_MEM_PRIVATE, > + "Can not update private memfd for non-private memslot\n"); > + restricted_fd = region_ext->restricted_fd; > + restricted_fd_offset = region_ext->restricted_offset; > + guest_phys_base = region->guest_phys_addr; > + fd_offset = restricted_fd_offset + (gpa - guest_phys_base); > + > + if (unback_mem) > + fallocate_mode = (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE); > + > + printf("restricted_fd %d fallocate_mode 0x%x for offset 0x%lx size 0x%lx\n", > + restricted_fd, fallocate_mode, fd_offset, size); Don't put prints in common helpers, except _maybe_ for error paths. It's fine for development and/or debug, but for the final product it ends up being noise 99% of the time. If you really, really want printing checked in, then pr_debug() is an option, but I would generally discourage even that for selftests. E.g. strace can give you all the information printed here without needing to rebuild the binary, and without maintenance burden. > + ret = fallocate(restricted_fd, fallocate_mode, fd_offset, size); > + TEST_ASSERT(ret == 0, "fallocate failed\n"); Use whitespace to differntiate operations/blocks. > + attr.attributes = unback_mem ? 0 : KVM_MEMORY_ATTRIBUTE_PRIVATE; > + attr.address = gpa; > + attr.size = size; > + attr.flags = 0; Add a helper to do KVM_SET_MEMORY_ATTRIBUTES, e.g. to fill the appropriate struct. > + if (unback_mem) > + printf("undoing encryption for gpa 0x%lx size 0x%lx\n", gpa, size); > + else > + printf("doing encryption for gpa 0x%lx size 0x%lx\n", gpa, size); > + > + vm_ioctl(vm, KVM_SET_MEMORY_ATTRIBUTES, &attr); > +} void vm_mem_map_shared_or_private(struct kvm_vm *vm, uint64_t gpa, uint64_t size, bool map_shared) { struct userspace_mem_region *region; uint64_t end = gpa + size - 1; off_t fd_offset; int mode, ret; region = userspace_mem_region_find(vm, gpa, gpa); TEST_ASSERT(region && region->region.flags & KVM_MEM_PRIVATE, "Private memory region not found for GPA 0x%lx", gpa); TEST_ASSERT(region == userspace_mem_region_find(vm, end, end), "Private/Shared conversions must act on a single memslot"); fd_offset = region->region.restricted_offset + (gpa - region->region.guest_phys_addr); /* To map shared, punch a hole. To map private, allocate (no flags). */ mode = map_shared ? (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE) : 0; ret = fallocate(region->region.restricted_fd, mode, fd_offset, size); TEST_ASSERT(!ret, "fallocate() failed to map %lx[%lu] %s, fd = %d, mode = %x, offset = %lx\n", gpa, size, map_shared ? "shared" : "private", region->region.restricted_fd, mode, fd_offset); vm_set_memory_attributes(vm, gpa, size, map_shared ? 0 : KVM_MEMORY_ATTRIBUTE_PRIVATE); }
On Mon, Dec 05, 2022, Vishal Annapurve wrote: > +void vcpu_run_and_handle_mapgpa(struct kvm_vm *vm, struct kvm_vcpu *vcpu) > +{ > + /* > + * Loop until the guest exits with any reason other than > + * KVM_HC_MAP_GPA_RANGE hypercall. > + */ > + > + while (true) { > + vcpu_run(vcpu); > + > + if ((vcpu->run->exit_reason == KVM_EXIT_HYPERCALL) && > + (vcpu->run->hypercall.nr == KVM_HC_MAP_GPA_RANGE)) { I get what you're trying to do, and I completely agree that we need better helpers and/or macros to reduce this type of boilerplate, but adding a one-off helper like this is going to be a net negative overall. This helper services exactly one use case, and also obfuscates what a test does. In other words, this is yet another thing that needs broad, generic support (_vcpu_run() is a very special case). E.g. something like this to make it easy for tests to run a guest and handle ucalls plus specific exits (just a strawman, I think we can do better for handling ucalls). #define vcpu_run_loop(vcpu, handlers, ucalls) \ do { \ uint32_t __exit; \ int __r = 0; \ \ while (!r) { \ vcpu_run(vcpu); \ \ __exit = vcpu->run->exit_reason; \ \ if (__exit < ARRAY_SIZE(handlers) && handlers[__exit]) \ __r = handlers[__exit](vcpu); \ else if (__exit == KVM_EXIT_IO && ucalls) \ __r = handle_exit_ucall(vcpu, ucalls, \ ARRAY_SIZE(ucalls)); \ else \ TEST_FAIL(...) \ } \ } while (0) For this series, I think it makes sense to just open code yet another test. It really doesn't end up being much code, which is partly why we haven't added helpers :-/
diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile index 2275ba861e0e..97f7d52c553b 100644 --- a/tools/testing/selftests/kvm/Makefile +++ b/tools/testing/selftests/kvm/Makefile @@ -55,6 +55,7 @@ LIBKVM_x86_64 += lib/x86_64/apic.c LIBKVM_x86_64 += lib/x86_64/handlers.S LIBKVM_x86_64 += lib/x86_64/hyperv.c LIBKVM_x86_64 += lib/x86_64/memstress.c +LIBKVM_x86_64 += lib/x86_64/private_mem.c LIBKVM_x86_64 += lib/x86_64/processor.c LIBKVM_x86_64 += lib/x86_64/svm.c LIBKVM_x86_64 += lib/x86_64/ucall.c diff --git a/tools/testing/selftests/kvm/include/x86_64/private_mem.h b/tools/testing/selftests/kvm/include/x86_64/private_mem.h new file mode 100644 index 000000000000..3aa6b4d11b28 --- /dev/null +++ b/tools/testing/selftests/kvm/include/x86_64/private_mem.h @@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Copyright (C) 2022, Google LLC. + */ + +#ifndef SELFTEST_KVM_PRIVATE_MEM_H +#define SELFTEST_KVM_PRIVATE_MEM_H + +#include <stdint.h> +#include <kvm_util.h> + +void kvm_hypercall_map_shared(uint64_t gpa, uint64_t size); +void kvm_hypercall_map_private(uint64_t gpa, uint64_t size); + +void vm_unback_private_mem(struct kvm_vm *vm, uint64_t gpa, uint64_t size); + +void vm_allocate_private_mem(struct kvm_vm *vm, uint64_t gpa, uint64_t size); + +void handle_vm_exit_map_gpa_hypercall(struct kvm_vm *vm, uint64_t gpa, + uint64_t npages, uint64_t attrs); + +void vcpu_run_and_handle_mapgpa(struct kvm_vm *vm, struct kvm_vcpu *vcpu); + +#endif /* SELFTEST_KVM_PRIVATE_MEM_H */ diff --git a/tools/testing/selftests/kvm/lib/x86_64/private_mem.c b/tools/testing/selftests/kvm/lib/x86_64/private_mem.c new file mode 100644 index 000000000000..2b97fc34ec4a --- /dev/null +++ b/tools/testing/selftests/kvm/lib/x86_64/private_mem.c @@ -0,0 +1,139 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2022, Google LLC. + */ +#define _GNU_SOURCE /* for program_invocation_name */ +#include <fcntl.h> +#include <limits.h> +#include <sched.h> +#include <signal.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <sys/ioctl.h> + +#include <linux/compiler.h> +#include <linux/kernel.h> +#include <linux/kvm_para.h> + +#include <test_util.h> +#include <kvm_util.h> +#include <private_mem.h> +#include <processor.h> + +static inline uint64_t __kvm_hypercall_map_gpa_range(uint64_t gpa, uint64_t size, + uint64_t flags) +{ + return kvm_hypercall(KVM_HC_MAP_GPA_RANGE, gpa, size >> PAGE_SHIFT, flags, 0); +} + +static inline void kvm_hypercall_map_gpa_range(uint64_t gpa, uint64_t size, + uint64_t flags) +{ + uint64_t ret; + + GUEST_ASSERT_2(IS_PAGE_ALIGNED(gpa) && IS_PAGE_ALIGNED(size), gpa, size); + + ret = __kvm_hypercall_map_gpa_range(gpa, size, flags); + GUEST_ASSERT_1(!ret, ret); +} + +void kvm_hypercall_map_shared(uint64_t gpa, uint64_t size) +{ + kvm_hypercall_map_gpa_range(gpa, size, KVM_MAP_GPA_RANGE_DECRYPTED); +} + +void kvm_hypercall_map_private(uint64_t gpa, uint64_t size) +{ + kvm_hypercall_map_gpa_range(gpa, size, KVM_MAP_GPA_RANGE_ENCRYPTED); +} + +static void vm_update_private_mem(struct kvm_vm *vm, uint64_t gpa, uint64_t size, + bool unback_mem) +{ + int restricted_fd; + uint64_t restricted_fd_offset, guest_phys_base, fd_offset; + struct kvm_memory_attributes attr; + struct kvm_userspace_memory_region_ext *region_ext; + struct kvm_userspace_memory_region *region; + int fallocate_mode = 0; + int ret; + + region_ext = kvm_userspace_memory_region_ext_find(vm, gpa, gpa + size); + TEST_ASSERT(region_ext != NULL, "Region not found"); + region = ®ion_ext->region; + TEST_ASSERT(region->flags & KVM_MEM_PRIVATE, + "Can not update private memfd for non-private memslot\n"); + restricted_fd = region_ext->restricted_fd; + restricted_fd_offset = region_ext->restricted_offset; + guest_phys_base = region->guest_phys_addr; + fd_offset = restricted_fd_offset + (gpa - guest_phys_base); + + if (unback_mem) + fallocate_mode = (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE); + + printf("restricted_fd %d fallocate_mode 0x%x for offset 0x%lx size 0x%lx\n", + restricted_fd, fallocate_mode, fd_offset, size); + ret = fallocate(restricted_fd, fallocate_mode, fd_offset, size); + TEST_ASSERT(ret == 0, "fallocate failed\n"); + attr.attributes = unback_mem ? 0 : KVM_MEMORY_ATTRIBUTE_PRIVATE; + attr.address = gpa; + attr.size = size; + attr.flags = 0; + if (unback_mem) + printf("undoing encryption for gpa 0x%lx size 0x%lx\n", gpa, size); + else + printf("doing encryption for gpa 0x%lx size 0x%lx\n", gpa, size); + + vm_ioctl(vm, KVM_SET_MEMORY_ATTRIBUTES, &attr); +} + +void vm_unback_private_mem(struct kvm_vm *vm, uint64_t gpa, uint64_t size) +{ + vm_update_private_mem(vm, gpa, size, true); +} + +void vm_allocate_private_mem(struct kvm_vm *vm, uint64_t gpa, uint64_t size) +{ + vm_update_private_mem(vm, gpa, size, false); +} + +void handle_vm_exit_map_gpa_hypercall(struct kvm_vm *vm, uint64_t gpa, + uint64_t npages, uint64_t attrs) +{ + uint64_t size; + + size = npages << MIN_PAGE_SHIFT; + pr_info("Explicit conversion off 0x%lx size 0x%lx to %s\n", gpa, size, + (attrs & KVM_MAP_GPA_RANGE_ENCRYPTED) ? "private" : "shared"); + + if (attrs & KVM_MAP_GPA_RANGE_ENCRYPTED) + vm_allocate_private_mem(vm, gpa, size); + else + vm_unback_private_mem(vm, gpa, size); +} + +void vcpu_run_and_handle_mapgpa(struct kvm_vm *vm, struct kvm_vcpu *vcpu) +{ + /* + * Loop until the guest exits with any reason other than + * KVM_HC_MAP_GPA_RANGE hypercall. + */ + + while (true) { + vcpu_run(vcpu); + + if ((vcpu->run->exit_reason == KVM_EXIT_HYPERCALL) && + (vcpu->run->hypercall.nr == KVM_HC_MAP_GPA_RANGE)) { + uint64_t gpa = vcpu->run->hypercall.args[0]; + uint64_t npages = vcpu->run->hypercall.args[1]; + uint64_t attrs = vcpu->run->hypercall.args[2]; + + handle_vm_exit_map_gpa_hypercall(vm, gpa, npages, attrs); + vcpu->run->hypercall.ret = 0; + continue; + } + + return; + } +}
Introduce a set of APIs to execute VM with private memslots. Host userspace APIs for: 1) Executing a vcpu run loop that handles MAPGPA hypercall 2) Backing/unbacking guest private memory Guest APIs for: 1) Changing memory mapping type Signed-off-by: Vishal Annapurve <vannapurve@google.com> --- tools/testing/selftests/kvm/Makefile | 1 + .../kvm/include/x86_64/private_mem.h | 24 +++ .../selftests/kvm/lib/x86_64/private_mem.c | 139 ++++++++++++++++++ 3 files changed, 164 insertions(+) create mode 100644 tools/testing/selftests/kvm/include/x86_64/private_mem.h create mode 100644 tools/testing/selftests/kvm/lib/x86_64/private_mem.c