KVM: selftests: arm64 exception handling and debug test
v2 cover KVM: selftests: arm64 exception handling and debug test
Hi,
These patches add a debug exception test in aarch64 KVM selftests while also adding basic exception handling support.
The structure of the exception handling is based on its x86 counterpart. Tests use the same calls to initialize exception handling and both architectures allow tests to override the handler for a particular vector, or (vector, ec) for synchronous exceptions in the arm64 case.
The debug test is similar to x86_64/debug_regs, except that the x86 one controls the debugging from outside the VM. This proposed arm64 test controls and handles debug exceptions from the inside.
Thanks,
Ricardo
v1 -> v2:
Addressed comments from Andrew and Marc (thank you very much):
- rename vm_handle_exception in all tests.
- introduce UCALL_UNHANDLED in x86 first.
- move GUEST_ASSERT_EQ to common utils header.
- handle sync and other exceptions separately: use two tables (like kvm-unit-tests).
- add two separate functions for installing sync versus other exceptions
- changes in handlers.S: use the same layout as user_pt_regs, treat the EL1t vectors as invalid, refactor the vector table creation to not use manual numbering, add comments, remove LR from the stored registers.
- changes in debug-exceptions.c: remove unused headers, use the common GUEST_ASSERT_EQ, use vcpu_run instead of _vcpu_run.
- changes in processor.h: write_sysreg with support for xzr, replace EL1 with current in macro names, define ESR_EC_MASK as ESR_EC_NUM-1.
Ricardo Koller (5):
KVM: selftests: Rename vm_handle_exception
KVM: selftests: Introduce UCALL_UNHANDLED for unhandled vector
reporting
KVM: selftests: Move GUEST_ASSERT_EQ to utils header
KVM: selftests: Add exception handling support for aarch64
KVM: selftests: Add aarch64/debug-exceptions test
tools/testing/selftests/kvm/.gitignore | 1 + tools/testing/selftests/kvm/Makefile | 3 +- .../selftests/kvm/aarch64/debug-exceptions.c | 244 ++++++++++++++++++ .../selftests/kvm/include/aarch64/processor.h | 90 ++++++- .../testing/selftests/kvm/include/kvm_util.h | 10 + .../selftests/kvm/include/x86_64/processor.h | 4 +- .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++ .../selftests/kvm/lib/x86_64/processor.c | 19 +- .../selftests/kvm/x86_64/kvm_pv_test.c | 2 +- .../selftests/kvm/x86_64/tsc_msrs_test.c | 9 - .../kvm/x86_64/userspace_msr_exit_test.c | 8 +- .../selftests/kvm/x86_64/xapic_ipi_test.c | 2 +- 13 files changed, 611 insertions(+), 35 deletions(-) create mode 100644 tools/testing/selftests/kvm/aarch64/debug-exceptions.c create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.S
Signature · 1 line
2.31.1.527.g47e6f16901-goog
PATCH v2 1/5 KVM: selftests: Rename vm_handle_exception
Rename the vm_handle_exception function to a name that indicates more clearly that it installs something: vm_install_vector_handler.
--- tools/testing/selftests/kvm/include/x86_64/processor.h | 2 +- tools/testing/selftests/kvm/lib/x86_64/processor.c | 4 ++-- tools/testing/selftests/kvm/x86_64/kvm_pv_test.c | 2 +- .../selftests/kvm/x86_64/userspace_msr_exit_test.c | 8 ++++---- tools/testing/selftests/kvm/x86_64/xapic_ipi_test.c | 2 +- 5 files changed, 9 insertions(+), 9 deletions(-)
tools/testing/selftests/kvm/include/x86_64/processor.h +1 -1
@@ -391,7 +391,7 @@ struct ex_regs {void vm_init_descriptor_tables (struct kvm_vm *vm);void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); -void vm_handle_exception (struct kvm_vm *vm,int vector, +void vm_install_vector_handler (struct kvm_vm *vm,int vector,void (*handler)(struct ex_regs *));/*
tools/testing/selftests/kvm/lib/x86_64/processor.c +2 -2
@@ -1250,8 +1250,8 @@ void vcpu_init_descriptor_tables(struct kvm_vm *vm, uint32_t vcpuid) *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; } -void vm_handle_exception (struct kvm_vm *vm,int vector, -void (*handler)(struct ex_regs *)) +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) {vm_vaddr_t *handlers = (vm_vaddr_t *)addr_gva2hva (vm, vm->handlers);
tools/testing/selftests/kvm/x86_64/kvm_pv_test.c +1 -1
@@ -227,7 +227,7 @@ int main(void)vm_init_descriptor_tables (vm);vcpu_init_descriptor_tables (vm,VCPU_ID ); -vm_handle_exception (vm,GP_VECTOR , guest_gp_handler); +vm_install_vector_handler (vm,GP_VECTOR , guest_gp_handler);enter_guest (vm);kvm_vm_free (vm);
tools/testing/selftests/kvm/x86_64/userspace_msr_exit_test.c +4 -4
@@ -574,7 +574,7 @@ static void test_msr_filter_allow(void) {vm_init_descriptor_tables (vm);vcpu_init_descriptor_tables (vm,VCPU_ID ); -vm_handle_exception (vm,GP_VECTOR , guest_gp_handler); +vm_install_vector_handler (vm,GP_VECTOR , guest_gp_handler);/* Process guest code userspace exits. */ run_guest_then_process_rdmsr (vm,MSR_IA32_XSS );
@@ -588,12 +588,12 @@ static void test_msr_filter_allow(void) {run_guest_then_process_wrmsr (vm,MSR_NON_EXISTENT );run_guest_then_process_rdmsr (vm,MSR_NON_EXISTENT ); -vm_handle_exception (vm,UD_VECTOR , guest_ud_handler); +vm_install_vector_handler (vm,UD_VECTOR , guest_ud_handler);run_guest (vm); -vm_handle_exception (vm,UD_VECTOR ,NULL ); +vm_install_vector_handler (vm,UD_VECTOR ,NULL );if (process_ucall (vm) !=UCALL_DONE ) { -vm_handle_exception (vm,GP_VECTOR , guest_fep_gp_handler); +vm_install_vector_handler (vm,GP_VECTOR , guest_fep_gp_handler);/* Process emulated rdmsr and wrmsr instructions. */ run_guest_then_process_rdmsr (vm,MSR_IA32_XSS );
tools/testing/selftests/kvm/x86_64/xapic_ipi_test.c +1 -1
@@ -462,7 +462,7 @@ int main(int argc, char *argv[])vm_init_descriptor_tables (vm);vcpu_init_descriptor_tables (vm,HALTER_VCPU_ID ); -vm_handle_exception (vm,IPI_VECTOR , guest_ipi_handler); +vm_install_vector_handler (vm,IPI_VECTOR , guest_ipi_handler);virt_pg_map (vm,APIC_DEFAULT_GPA ,APIC_DEFAULT_GPA ,0 );
Signature · 1 line
2.31.1.527.g47e6f16901-goog
Re: [PATCH v2 1/5] KVM: selftests: Rename vm_handle_exception
Quoted 6 lines · Ricardo Koller
Rename the
vm_handle_exceptionfunction to a name that indicates more clearly that it installs something:vm_install_vector_handler.Suggested-by Marc Zyngier <[email protected]>Suggested-by Andrew Jones <[email protected]>Signed-off-by Ricardo Koller <[email protected]>
Quoted 8 lines
--- tools/testing/selftests/kvm/include/x86_64/processor.h | 2 +- tools/testing/selftests/kvm/lib/x86_64/processor.c | 4 ++-- tools/testing/selftests/kvm/x86_64/kvm_pv_test.c | 2 +- .../selftests/kvm/x86_64/userspace_msr_exit_test.c | 8 ++++---- tools/testing/selftests/kvm/x86_64/xapic_ipi_test.c | 2 +- 5 files changed, 9 insertions(+), 9 deletions(-)
Re: [PATCH v2 1/5] KVM: selftests: Rename vm_handle_exception
In reply to Ricardo KollerHi Ricardo,
Quoted 6 lines · Ricardo Koller
Rename the
vm_handle_exceptionfunction to a name that indicates more clearly that it installs something:vm_install_vector_handler.Suggested-by Marc Zyngier <[email protected]>Suggested-by Andrew Jones <[email protected]>Signed-off-by Ricardo Koller <[email protected]>
Thanks
Eric
Quoted 92 lines
--- tools/testing/selftests/kvm/include/x86_64/processor.h | 2 +- tools/testing/selftests/kvm/lib/x86_64/processor.c | 4 ++-- tools/testing/selftests/kvm/x86_64/kvm_pv_test.c | 2 +- .../selftests/kvm/x86_64/userspace_msr_exit_test.c | 8 ++++---- tools/testing/selftests/kvm/x86_64/xapic_ipi_test.c | 2 +- 5 files changed, 9 insertions(+), 9 deletions(-)tools/testing/selftests/kvm/include/x86_64/processor.h +1
-1@@ -391,7 +391,7 @@ struct ex_regs {void vm_init_descriptor_tables (struct kvm_vm *vm);void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); -void vm_handle_exception (struct kvm_vm *vm,int vector, +void vm_install_vector_handler (struct kvm_vm *vm,int vector,void (*handler)(struct ex_regs *));/* tools/testing/selftests/kvm/lib/x86_64/processor.c +2
-2@@ -1250,8 +1250,8 @@ void vcpu_init_descriptor_tables(struct kvm_vm *vm, uint32_t vcpuid) *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; } -void vm_handle_exception (struct kvm_vm *vm,int vector, -void (*handler)(struct ex_regs *)) +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) {vm_vaddr_t *handlers = (vm_vaddr_t *)addr_gva2hva (vm, vm->handlers);tools/testing/selftests/kvm/x86_64/kvm_pv_test.c +1
-1@@ -227,7 +227,7 @@ int main(void)vm_init_descriptor_tables (vm);vcpu_init_descriptor_tables (vm,VCPU_ID ); -vm_handle_exception (vm,GP_VECTOR , guest_gp_handler); +vm_install_vector_handler (vm,GP_VECTOR , guest_gp_handler);enter_guest (vm);kvm_vm_free (vm);tools/testing/selftests/kvm/x86_64/userspace_msr_exit_test.c +4
-4@@ -574,7 +574,7 @@ static void test_msr_filter_allow(void) {vm_init_descriptor_tables (vm);vcpu_init_descriptor_tables (vm,VCPU_ID ); -vm_handle_exception (vm,GP_VECTOR , guest_gp_handler); +vm_install_vector_handler (vm,GP_VECTOR , guest_gp_handler);/* Process guest code userspace exits. */ run_guest_then_process_rdmsr (vm,MSR_IA32_XSS );@@ -588,12 +588,12 @@ static void test_msr_filter_allow(void) {run_guest_then_process_wrmsr (vm,MSR_NON_EXISTENT );run_guest_then_process_rdmsr (vm,MSR_NON_EXISTENT ); -vm_handle_exception (vm,UD_VECTOR , guest_ud_handler); +vm_install_vector_handler (vm,UD_VECTOR , guest_ud_handler);run_guest (vm); -vm_handle_exception (vm,UD_VECTOR ,NULL ); +vm_install_vector_handler (vm,UD_VECTOR ,NULL );if (process_ucall (vm) !=UCALL_DONE ) { -vm_handle_exception (vm,GP_VECTOR , guest_fep_gp_handler); +vm_install_vector_handler (vm,GP_VECTOR , guest_fep_gp_handler);/* Process emulated rdmsr and wrmsr instructions. */ run_guest_then_process_rdmsr (vm,MSR_IA32_XSS );tools/testing/selftests/kvm/x86_64/xapic_ipi_test.c +1
-1@@ -462,7 +462,7 @@ int main(int argc, char *argv[])vm_init_descriptor_tables (vm);vcpu_init_descriptor_tables (vm,HALTER_VCPU_ID ); -vm_handle_exception (vm,IPI_VECTOR , guest_ipi_handler); +vm_install_vector_handler (vm,IPI_VECTOR , guest_ipi_handler);virt_pg_map (vm,APIC_DEFAULT_GPA ,APIC_DEFAULT_GPA ,0 );
PATCH v2 2/5 KVM: selftests: Introduce UCALL_UNHANDLED for unhandled vector reporting
In reply to Ricardo Kollerx86, the only arch implementing exception handling, reports unhandled vectors using port IO at a specific port number. This replicates what ucall already does.
Introduce a new ucall type, UCALL_UNHANDLED, for guests to report unhandled exceptions. Then replace the x86 unhandled vector exception reporting to use it instead of port IO. This new ucall type will be used in the next commits by arm64 to report unhandled vectors as well.
Tested: Forcing a page fault in the ./x86_64/xapic_ipi_test
halter_guest_code() shows this: $ ./x86_64/xapic_ipi_test ... Unexpected vectored event in guest (vector:0xe)
--- tools/testing/selftests/kvm/include/kvm_util.h | 1 + .../selftests/kvm/include/x86_64/processor.h | 2 -- .../testing/selftests/kvm/lib/x86_64/processor.c | 15 ++++++--------- 3 files changed, 7 insertions(+), 11 deletions(-)
tools/testing/selftests/kvm/include/kvm_util.h +1 -0
@@ -347,6 +347,7 @@ enum {UCALL_SYNC ,UCALL_ABORT ,UCALL_DONE , + UCALL_UNHANDLED, };#define UCALL_MAX_ARGS 6
tools/testing/selftests/kvm/include/x86_64/processor.h +0 -2
@@ -53,8 +53,6 @@#define CPUID_PKU (1ul << 3)#define CPUID_LA57 (1ul << 16) -#define UNEXPECTED_VECTOR_PORT 0xfff0u -/* General Registers in 64-Bit Mode */ struct gpr64_regs {u64 rax;
tools/testing/selftests/kvm/lib/x86_64/processor.c +6 -9
@@ -1207,7 +1207,7 @@ static void set_idt_entry(struct kvm_vm *vm, int vector, unsigned long addr,void kvm_exit_unexpected_vector (uint32_t value) { -outl (UNEXPECTED_VECTOR_PORT , value); +ucall (UCALL_UNHANDLED ,1 , value); }void route_exception (struct ex_regs *regs)
@@ -1260,16 +1260,13 @@ void vm_install_vector_handler(struct kvm_vm *vm, int vector,void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { -if (vcpu_state (vm, vcpuid)->exit_reason ==KVM_EXIT_IO - &&vcpu_state (vm, vcpuid)->io.port ==UNEXPECTED_VECTOR_PORT - &&vcpu_state (vm, vcpuid)->io.size ==4 ) { -/* Grab pointer to io data */ -uint32_t *data = (void *)vcpu_state (vm, vcpuid) - +vcpu_state (vm, vcpuid)->io.data_offset; +struct ucall uc; +if (get_ucall (vm, vcpuid, &uc) ==UCALL_UNHANDLED ) { +uint64_t vector = uc.args[0 ];TEST_ASSERT (false, -"Unexpected vectored event in guest (vector:0x%x)" , - *data); +"Unexpected vectored event in guest (vector:0x%lx)" , + vector); } }
Signature · 1 line
2.31.1.527.g47e6f16901-goog
Re: [PATCH v2 2/5] KVM: selftests: Introduce UCALL_UNHANDLED for unhandled vector reporting
Quoted 70 lines · Ricardo Koller
x86, the only arch implementing exception handling, reports unhandled vectors using port IO at a specific port number. This replicates what ucall already does.
Introduce a new ucall type,
UCALL_UNHANDLED, for guests to report unhandled exceptions. Then replace the x86 unhandled vector exception reporting to use it instead of port IO. This new ucall type will be used in the next commits by arm64 to report unhandled vectors as well.Tested: Forcing a page fault in the ./x86_64/xapic_ipi_test
halter_guest_code() shows this: $ ./x86_64/xapic_ipi_test ... Unexpected vectored event in guest (vector:0xe)Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/include/kvm_util.h | 1 + .../selftests/kvm/include/x86_64/processor.h | 2 -- .../testing/selftests/kvm/lib/x86_64/processor.c | 15 ++++++--------- 3 files changed, 7 insertions(+), 11 deletions(-)tools/testing/selftests/kvm/include/kvm_util.h +1
-0@@ -347,6 +347,7 @@ enum {UCALL_SYNC ,UCALL_ABORT ,UCALL_DONE , + UCALL_UNHANDLED, };#define UCALL_MAX_ARGS 6tools/testing/selftests/kvm/include/x86_64/processor.h +0
-2@@ -53,8 +53,6 @@#define CPUID_PKU (1ul << 3)#define CPUID_LA57 (1ul << 16) -#define UNEXPECTED_VECTOR_PORT 0xfff0u -/* General Registers in 64-Bit Mode */ struct gpr64_regs {u64 rax;tools/testing/selftests/kvm/lib/x86_64/processor.c +1
-6@@ -1207,7 +1207,7 @@ static void set_idt_entry(struct kvm_vm *vm, int vector, unsigned long addr,void kvm_exit_unexpected_vector (uint32_t value) { -outl (UNEXPECTED_VECTOR_PORT , value); +ucall (UCALL_UNHANDLED ,1 , value); }void route_exception (struct ex_regs *regs)@@ -1260,16 +1260,13 @@ void vm_install_vector_handler(struct kvm_vm *vm, int vector,void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { -if (vcpu_state (vm, vcpuid)->exit_reason ==KVM_EXIT_IO - &&vcpu_state (vm, vcpuid)->io.port ==UNEXPECTED_VECTOR_PORT - &&vcpu_state (vm, vcpuid)->io.size ==4 ) { -/* Grab pointer to io data */ -uint32_t *data = (void *)vcpu_state (vm, vcpuid)
tools/testing/selftests/kvm/lib/x86_64/processor.c void vm_install_vector_handler(struct kvm_vm *vm, int vector, · line 1269
Quoted 10 lines
tools/testing/selftests/kvm/lib/x86_64/processor.c +5
-3- +vcpu_state (vm, vcpuid)->io.data_offset; +struct ucall uc; +if (get_ucall (vm, vcpuid, &uc) ==UCALL_UNHANDLED ) { +uint64_t vector = uc.args[0 ];TEST_ASSERT (false, -"Unexpected vectored event in guest (vector:0x%x)" , - *data); +"Unexpected vectored event in guest (vector:0x%lx)" , + vector);
nit: Could have changed this TEST_ASSERT(false, ...) to TEST_FAIL while touching it.
Quoted 6 lines
tools/testing/selftests/kvm/lib/x86_64/processor.c +0
-0} }Signature · 1 line
2.31.1.527.g47e6f16901-goog
Thanks,
drew
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Auger Eric
Reviewed-byRe: [PATCH v2 2/5] KVM: selftests: Introduce UCALL_UNHANDLED for unhandled vector reporting
Hi Ricardo,
Quoted 10 lines · Andrew Jones
Quoted 9 lines · Ricardo Koller
x86, the only arch implementing exception handling, reports unhandled vectors using port IO at a specific port number. This replicates what ucall already does.
Introduce a new ucall type,
UCALL_UNHANDLED, for guests to report unhandled exceptions. Then replace the x86 unhandled vector exception reporting to use it instead of port IO. This new ucall type will be used in the next commits by arm64 to report unhandled vectors as well.
Quoted 8 lines
Quoted 8 lines
Tested: Forcing a page fault in the ./x86_64/xapic_ipi_test
halter_guest_code() shows this: $ ./x86_64/xapic_ipi_test ... Unexpected vectored event in guest (vector:0xe)Signed-off-by Ricardo Koller <[email protected]>
Eric
Quoted 79 lines
Quoted 53 lines
--- tools/testing/selftests/kvm/include/kvm_util.h | 1 + .../selftests/kvm/include/x86_64/processor.h | 2 -- .../testing/selftests/kvm/lib/x86_64/processor.c | 15 ++++++--------- 3 files changed, 7 insertions(+), 11 deletions(-)tools/testing/selftests/kvm/include/kvm_util.h +1
-0@@ -347,6 +347,7 @@ enum {UCALL_SYNC ,UCALL_ABORT ,UCALL_DONE , + UCALL_UNHANDLED, };#define UCALL_MAX_ARGS 6tools/testing/selftests/kvm/include/x86_64/processor.h +0
-2@@ -53,8 +53,6 @@#define CPUID_PKU (1ul << 3)#define CPUID_LA57 (1ul << 16) -#define UNEXPECTED_VECTOR_PORT 0xfff0u -/* General Registers in 64-Bit Mode */ struct gpr64_regs {u64 rax;tools/testing/selftests/kvm/lib/x86_64/processor.c +1
-6@@ -1207,7 +1207,7 @@ static void set_idt_entry(struct kvm_vm *vm, int vector, unsigned long addr,void kvm_exit_unexpected_vector (uint32_t value) { -outl (UNEXPECTED_VECTOR_PORT , value); +ucall (UCALL_UNHANDLED ,1 , value); }void route_exception (struct ex_regs *regs)@@ -1260,16 +1260,13 @@ void vm_install_vector_handler(struct kvm_vm *vm, int vector,void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { -if (vcpu_state (vm, vcpuid)->exit_reason ==KVM_EXIT_IO - &&vcpu_state (vm, vcpuid)->io.port ==UNEXPECTED_VECTOR_PORT - &&vcpu_state (vm, vcpuid)->io.size ==4 ) { -/* Grab pointer to io data */ -uint32_t *data = (void *)vcpu_state (vm, vcpuid)tools/testing/selftests/kvm/lib/x86_64/processor.c void vm_install_vector_handler(struct kvm_vm *vm, int vector, · line 1269
Quoted 10 lines
tools/testing/selftests/kvm/lib/x86_64/processor.c +5
-3- +vcpu_state (vm, vcpuid)->io.data_offset; +struct ucall uc; +if (get_ucall (vm, vcpuid, &uc) ==UCALL_UNHANDLED ) { +uint64_t vector = uc.args[0 ];TEST_ASSERT (false, -"Unexpected vectored event in guest (vector:0x%x)" , - *data); +"Unexpected vectored event in guest (vector:0x%lx)" , + vector);nit: Could have changed this
TEST_ASSERT(false, ...)toTEST_FAILwhile touching it.Quoted 6 lines
tools/testing/selftests/kvm/lib/x86_64/processor.c +0
-0} }Signature · 1 line
2.31.1.527.g47e6f16901-googReviewed-by Andrew Jones <[email protected]>Thanks,
drew
PATCH v2 3/5 KVM: selftests: Move GUEST_ASSERT_EQ to utils header
In reply to Ricardo KollerMove GUEST_ASSERT_EQ to a common header, kvm_util.h, for other architectures and tests to use.
--- tools/testing/selftests/kvm/include/kvm_util.h | 9 +++++++++ tools/testing/selftests/kvm/x86_64/tsc_msrs_test.c | 9 --------- 2 files changed, 9 insertions(+), 9 deletions(-)
tools/testing/selftests/kvm/include/kvm_util.h +9 -0
@@ -388,4 +388,13 @@ uint64_t get_ucall(struct kvm_vm *vm, uint32_t vcpu_id, struct ucall *uc);#define GUEST_ASSERT_4 (_condition, arg1, arg2, arg3, arg4) \ __GUEST_ASSERT((_condition), 4, (arg1), (arg2), (arg3), (arg4)) +#define GUEST_ASSERT_EQ (a, b) do { \ + __typeof(a) _a = (a); \ + __typeof(b) _b = (b); \ + if (_a != _b) \ + ucall(UCALL_ABORT, 4, \ + "Failed guest assert: " \ + #a " == " #b, __LINE__, _a, _b); \ +} while(0) +#endif /* SELFTEST_KVM_UTIL_H */
tools/testing/selftests/kvm/x86_64/tsc_msrs_test.c +0 -9
@@ -18,15 +18,6 @@#define rounded_rdmsr (x) ROUND(rdmsr(x))#define rounded_host_rdmsr (x) ROUND(vcpu_get_msr(vm, 0, x)) -#define GUEST_ASSERT_EQ (a, b) do { \ - __typeof(a) _a = (a); \ - __typeof(b) _b = (b); \ - if (_a != _b) \ - ucall(UCALL_ABORT, 4, \ - "Failed guest assert: " \ - #a " == " #b, __LINE__, _a, _b); \ - } while(0) -static void guest_code (void ) {u64 val =0 ;
Signature · 1 line
2.31.1.527.g47e6f16901-goog
Re: [PATCH v2 3/5] KVM: selftests: Move GUEST_ASSERT_EQ to utils header
Quoted 50 lines · Ricardo Koller
Move
GUEST_ASSERT_EQto a common header,kvm_util.h, for other architectures and tests to use.Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/include/kvm_util.h | 9 +++++++++ tools/testing/selftests/kvm/x86_64/tsc_msrs_test.c | 9 --------- 2 files changed, 9 insertions(+), 9 deletions(-)tools/testing/selftests/kvm/include/kvm_util.h +9
-0@@ -388,4 +388,13 @@ uint64_t get_ucall(struct kvm_vm *vm, uint32_t vcpu_id, struct ucall *uc);#define GUEST_ASSERT_4 (_condition, arg1, arg2, arg3, arg4) \ __GUEST_ASSERT((_condition), 4, (arg1), (arg2), (arg3), (arg4)) +#define GUEST_ASSERT_EQ (a, b) do { \ + __typeof(a) _a = (a); \ + __typeof(b) _b = (b); \ + if (_a != _b) \ + ucall(UCALL_ABORT, 4, \ + "Failed guest assert: " \ + #a " == " #b, __LINE__, _a, _b); \ +} while(0) +#endif /* SELFTEST_KVM_UTIL_H */ tools/testing/selftests/kvm/x86_64/tsc_msrs_test.c +0
-9@@ -18,15 +18,6 @@#define rounded_rdmsr (x) ROUND(rdmsr(x))#define rounded_host_rdmsr (x) ROUND(vcpu_get_msr(vm, 0, x)) -#define GUEST_ASSERT_EQ (a, b) do { \ - __typeof(a) _a = (a); \ - __typeof(b) _b = (b); \ - if (_a != _b) \ - ucall(UCALL_ABORT, 4, \ - "Failed guest assert: " \ - #a " == " #b, __LINE__, _a, _b); \ - } while(0) -static void guest_code (void ) {u64 val =0 ;Signature · 1 line
2.31.1.527.g47e6f16901-goog
How about modify __GUEST_ASSERT so we can reuse it instead, like below? (I also took the opportunity to remove the unnecessary () within the comma separated statements.)
Thanks,
drew
-#define __GUEST_ASSERT(_condition, _nargs, _args...) do { \
- if (!(_condition)) \
- ucall(UCALL_ABORT, 2 + _nargs, \
- "Failed guest assert: " \
- #_condition, __LINE__, _args); \
+#define __GUEST_ASSERT(_condition, _condstr, _nargs, _args...) do { \
+ if (!(_condition)) \
+ ucall(UCALL_ABORT, 2 + _nargs, \
+ "Failed guest assert: " \
+ _condstr, __LINE__, _args); \} while (0)
#define GUEST_ASSERT(_condition) \
- __GUEST_ASSERT((_condition), 0, 0)
+ __GUEST_ASSERT(_condition, #_condition, 0, 0)
#define GUEST_ASSERT_1(_condition, arg1) \
- __GUEST_ASSERT((_condition), 1, (arg1))
+ __GUEST_ASSERT(_condition, #_condition, 1, arg1)
#define GUEST_ASSERT_2(_condition, arg1, arg2) \
- __GUEST_ASSERT((_condition), 2, (arg1), (arg2))
+ __GUEST_ASSERT(_condition, #_condition, 2, arg1, arg2)
#define GUEST_ASSERT_3(_condition, arg1, arg2, arg3) \
- __GUEST_ASSERT((_condition), 3, (arg1), (arg2), (arg3))
+ __GUEST_ASSERT(_condition, #_condition, 3, arg1, arg2, arg3)
#define GUEST_ASSERT_4(_condition, arg1, arg2, arg3, arg4) \
- __GUEST_ASSERT((_condition), 4, (arg1), (arg2), (arg3), (arg4))
-
-#define GUEST_ASSERT_EQ(a, b) do { \
- __typeof(a) _a = (a); \
- __typeof(b) _b = (b); \
- if (_a != _b) \
- ucall(UCALL_ABORT, 4, \
- "Failed guest assert: " \
- #a " == " #b, __LINE__, _a, _b); \-} while(0)
+ __GUEST_ASSERT(_condition, #_condition, 4, arg1, arg2, arg3, arg4)
+
+#define GUEST_ASSERT_EQ(a, b) __GUEST_ASSERT((a) == (b), #a " == " #b, 2, a, b)
PATCH v2 4/5 KVM: selftests: Add exception handling support for aarch64
In reply to Ricardo KollerAdd the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling vm_install_vector_handler(vector) or vm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit, UCALL_UNHANDLED.
The exception handling code is heavily inspired on kvm-unit-tests.
--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.S
tools/testing/selftests/kvm/Makefile +1 -1
@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_test
tools/testing/selftests/kvm/include/aarch64/processor.h +78 -0
94 lines not on this page · raw message
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +130 -0
130 lines not on this page · raw message
tools/testing/selftests/kvm/lib/aarch64/processor.c +124 -0
142 lines not on this page · raw message
Signature · 1 line
2.31.1.527.g47e6f16901-goog
Marc Zyngier
maintainerRe: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 152 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.The exception handling code is heavily inspired on kvm-unit-tests.
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +78
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) + +/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C + +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm); +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21) + +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +12
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7]
tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0
I'm still a bit perplexed by this. SP_EL0 is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.
Or did you have something else in mind for it?
Quoted 223 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +108
-0+ .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors: +.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label: + save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection + +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector +.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +91
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h" @@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers; +static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); +} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ { +assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + } +} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{
tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 439
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+struct handlers *handlers; + + vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler;
How about FIQ, Error? Although they are unlikely, they are valid exceptions.
Quoted 14 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +14
-0+ + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; +} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +}
tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 463
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +9
-0+ +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 170 lines · Marc Zyngier
Quoted 152 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.The exception handling code is heavily inspired on kvm-unit-tests.
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +78
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) + +/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C + +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm); +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21) + +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +12
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7]tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.
Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving sp_el0 in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp at regs->sp.
Restoring sp_el0 might be too much. So, what do you think of this v3: we keep the saving of sp_el0 into regs->sp (to keep things the same between el0 and el1) and delete the restoring of sp_el0?
Thanks,
Ricardo
Quoted 268 lines
Quoted 223 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +108
-0+ .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors: +.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label: + save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection + +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector +.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +91
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h" @@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers; +static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); +} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ { +assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + } +} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 439
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+struct handlers *handlers; + + vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler;How about FIQ, Error? Although they are unlikely, they are valid exceptions.
Quoted 14 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +14
-0+ + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; +} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +}tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 463
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +9
-0+ +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Hi Ricardo,
Quoted 12 lines · Ricardo Koller
Quoted 11 lines · Marc Zyngier
Quoted 9 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.
Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call
1 0x000000000040147b: main at debug-exceptions.c:230
2 0x000003ff8aa60de3: ?? ??:0
3 0x0000000000401517: _start at :?
Failed guest assert: hw_bp_addr == PC(hw_bp) at
aarch64/debug-exceptions.c:105
values: 0, 0x401794I guess it is not an expected result. Any known bug waiting on the list?
Thanks
Eric
Quoted 441 lines
Quoted 158 lines
Quoted 142 lines
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +78
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) + +/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C + +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm); +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21) + +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +12
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7]tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving
sp_el0in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp atregs->sp.Restoring
sp_el0might be too much. So, what do you think of this v3: we keep the saving ofsp_el0intoregs->sp(to keep things the same between el0 and el1) and delete the restoring ofsp_el0?Thanks,
RicardoQuoted 268 lines
Quoted 223 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +108
-0+ .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors: +.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label: + save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection + +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector +.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +91
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h" @@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers; +static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); +} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ { +assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + } +} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 439
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+struct handlers *handlers; + + vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler;How about FIQ, Error? Although they are unlikely, they are valid exceptions.
Quoted 14 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +14
-0+ + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; +} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +}tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 463
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +9
-0+ +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 2 lines · Auger Eric
Hi Ricardo,
Hi Eric,
Thank you very much for the test.
Quoted 28 lines · Ricardo Koller
Quoted 12 lines · Marc Zyngier
Quoted 11 lines
Quoted 9 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794
Quoted 2 lines
I guess it is not an expected result. Any known bug waiting on the list?
Not expected. That should work, or at least abort early because there is no HW breakpoints support.
I'm trying to reproduce the failure; can you help me with some questions, please?
- does your setup have support for hardware breakpoints? Can you try a 'dmesg | grep break'? I'm looking for something like 'hw-breakpoint:
found ...'. If there is no such line it's very likely that the check for "debug_ver >= 6" is not enough and the test should check for "num_breakpoints > 0".
- does it fail consistently (every single attempt)?
Thanks!
Ricardo
Quoted 448 lines
Thanks
Eric
Quoted 441 lines
Quoted 158 lines
Quoted 142 lines
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +78
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) + +/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C + +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm); +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21) + +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +12
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7]tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving
sp_el0in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp atregs->sp.Restoring
sp_el0might be too much. So, what do you think of this v3: we keep the saving ofsp_el0intoregs->sp(to keep things the same between el0 and el1) and delete the restoring ofsp_el0?Thanks,
RicardoQuoted 268 lines
Quoted 223 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +108
-0+ .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors: +.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label: + save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection + +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector +.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +91
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h" @@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers; +static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); +} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ { +assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + } +} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 439
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+struct handlers *handlers; + + vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler;How about FIQ, Error? Although they are unlikely, they are valid exceptions.
Quoted 14 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +14
-0+ + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; +} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +}tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 463
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +9
-0+ +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Hi Ricardo,
Quoted 36 lines · Ricardo Koller
Quoted 2 lines · Auger Eric
Hi Ricardo,
Hi Eric,
Thank you very much for the test.
Quoted 28 lines · Ricardo Koller
Quoted 12 lines · Marc Zyngier
Quoted 11 lines
Quoted 9 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794
Quoted 8 lines
Quoted 2 lines
I guess it is not an expected result. Any known bug waiting on the list?
Not expected. That should work, or at least abort early because there is no HW breakpoints support.
I'm trying to reproduce the failure; can you help me with some questions, please?
sure, please find the answers below.
Quoted 6 lines
- does your setup have support for hardware breakpoints? Can you try a 'dmesg | grep break'? I'm looking for something like 'hw-breakpoint:
found ...'. If there is no such line it's very likely that the check for "debug_ver>= 6" is not enough and the test should check for "num_breakpoints> 0".
[ 25.640418] hw-breakpoint: found 6 breakpoint and 4 watchpoint registers.
Quoted 1 line
- does it fail consistently (every single attempt)?
yes it does.
I will try to find some time to investigate too
Thanks
Eric
Quoted 453 lines
Thanks!
RicardoQuoted 448 lines
Thanks
Eric
Quoted 441 lines
Quoted 158 lines
Quoted 142 lines
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +78
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) + +/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C + +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm); +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21) + +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +12
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7]tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving
sp_el0in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp atregs->sp.Restoring
sp_el0might be too much. So, what do you think of this v3: we keep the saving ofsp_el0intoregs->sp(to keep things the same between el0 and el1) and delete the restoring ofsp_el0?Thanks,
RicardoQuoted 268 lines
Quoted 223 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +108
-0+ .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors: +.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label: + save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection + +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector +.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +91
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h" @@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers; +static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); +} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ { +assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + } +} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 439
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+struct handlers *handlers; + + vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler;How about FIQ, Error? Although they are unlikely, they are valid exceptions.
Quoted 14 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +14
-0+ + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; +} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +}tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 463
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +9
-0+ +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 54 lines · Auger Eric
Hi Ricardo,
Quoted 36 lines · Ricardo Koller
Quoted 2 lines · Auger Eric
Hi Ricardo,
Hi Eric,
Thank you very much for the test.
Quoted 28 lines · Ricardo Koller
Quoted 12 lines · Marc Zyngier
Quoted 11 lines
Quoted 9 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794Quoted 8 lines
Quoted 2 lines
I guess it is not an expected result. Any known bug waiting on the list?
Not expected. That should work, or at least abort early because there is no HW breakpoints support.
I'm trying to reproduce the failure; can you help me with some questions, please?
sure, please find the answers below.
Quoted 6 lines
- does your setup have support for hardware breakpoints? Can you try a 'dmesg | grep break'? I'm looking for something like 'hw-breakpoint:
found ...'. If there is no such line it's very likely that the check for "debug_ver>= 6" is not enough and the test should check for "num_breakpoints> 0".
Quoted 5 lines
[ 25.640418] hw-breakpoint: found 6 breakpoint and 4 watchpoint registers.Quoted 1 line
- does it fail consistently (every single attempt)?
yes it does.
I will try to find some time to investigate too
Thank you very much Eric.
Quoted 458 lines
Thanks
Eric
Quoted 453 lines
Thanks!
RicardoQuoted 448 lines
Thanks
Eric
Quoted 441 lines
Quoted 158 lines
Quoted 142 lines
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +78
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) + +/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C + +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm); +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21) + +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +12
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7]tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving
sp_el0in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp atregs->sp.Restoring
sp_el0might be too much. So, what do you think of this v3: we keep the saving ofsp_el0intoregs->sp(to keep things the same between el0 and el1) and delete the restoring ofsp_el0?Thanks,
RicardoQuoted 268 lines
Quoted 223 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +108
-0+ .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors: +.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label: + save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection + +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector +.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +91
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h" @@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers; +static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); +} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ { +assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + } +} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 439
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+struct handlers *handlers; + + vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler;How about FIQ, Error? Although they are unlikely, they are valid exceptions.
Quoted 14 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +14
-0+ + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; +} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +}tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 463
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +9
-0+ +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
In reply to Auger EricQuoted 28 lines · Auger Eric
Hi Ricardo,
Quoted 25 lines · Ricardo Koller
Quoted 2 lines · Auger Eric
Hi Ricardo,
Hi Eric,
Thank you very much for the test.
Quoted 17 lines · Ricardo Koller
Quoted 12 lines · Marc Zyngier
Quoted 11 lines
Quoted 9 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
Hi Eric,
I'm able to reproduce the failure you are seeing on 5.6, specifically with kernels older than this commit:
4942dc6638b0 KVM: arm64: Write arch.mdcr_el2 changes since last vcpu_load on VHE
but not yet on v5.12. Could you share the commit of the kernel you are testing, please?
Thanks!
Ricardo
Quoted 489 lines
Quoted 11 lines
Quoted 11 lines
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794Quoted 8 lines
Quoted 2 lines
I guess it is not an expected result. Any known bug waiting on the list?
Not expected. That should work, or at least abort early because there is no HW breakpoints support.
I'm trying to reproduce the failure; can you help me with some questions, please?
sure, please find the answers below.
Quoted 6 lines
- does your setup have support for hardware breakpoints? Can you try a 'dmesg | grep break'? I'm looking for something like 'hw-breakpoint:
found ...'. If there is no such line it's very likely that the check for "debug_ver>= 6" is not enough and the test should check for "num_breakpoints> 0".[ 25.640418] hw-breakpoint: found 6 breakpoint and 4 watchpoint registers.Quoted 1 line
- does it fail consistently (every single attempt)?
yes it does.
I will try to find some time to investigate too
Thanks
Eric
Quoted 453 lines
Thanks!
RicardoQuoted 448 lines
Thanks
Eric
Quoted 441 lines
Quoted 158 lines
Quoted 142 lines
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +78
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) + +/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C + +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm); +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21) + +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +12
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7]tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving
sp_el0in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp atregs->sp.Restoring
sp_el0might be too much. So, what do you think of this v3: we keep the saving ofsp_el0intoregs->sp(to keep things the same between el0 and el1) and delete the restoring ofsp_el0?Thanks,
RicardoQuoted 268 lines
Quoted 223 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +108
-0+ .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors: +.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label: + save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection + +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector +.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +91
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h" @@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers; +static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); +} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ { +assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + } +} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 439
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+struct handlers *handlers; + + vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler;How about FIQ, Error? Although they are unlikely, they are valid exceptions.
Quoted 14 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +14
-0+ + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; +} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +}tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 463
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +9
-0+ +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Hi Ricardo,
Quoted 32 lines · Ricardo Koller
Quoted 28 lines · Auger Eric
Hi Ricardo,
Quoted 25 lines · Ricardo Koller
Quoted 2 lines · Auger Eric
Hi Ricardo,
Hi Eric,
Thank you very much for the test.
Quoted 17 lines · Ricardo Koller
Quoted 12 lines · Marc Zyngier
Quoted 11 lines
Quoted 9 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
Hi Eric,
Quoted 7 lines
I'm able to reproduce the failure you are seeing on 5.6, specifically with kernels older than this commit:
4942dc6638b0KVM: arm64: Write arch.mdcr_el2 changes since lastvcpu_loadon VHEbut not yet on v5.12. Could you share the commit of the kernel you are testing, please?
my host is a 5.12 kernel (8404c9fbc84b)
Thanks
Eric
Quoted 494 lines
Thanks!
RicardoQuoted 489 lines
Quoted 11 lines
Quoted 11 lines
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794Quoted 8 lines
Quoted 2 lines
I guess it is not an expected result. Any known bug waiting on the list?
Not expected. That should work, or at least abort early because there is no HW breakpoints support.
I'm trying to reproduce the failure; can you help me with some questions, please?
sure, please find the answers below.
Quoted 6 lines
- does your setup have support for hardware breakpoints? Can you try a 'dmesg | grep break'? I'm looking for something like 'hw-breakpoint:
found ...'. If there is no such line it's very likely that the check for "debug_ver>= 6" is not enough and the test should check for "num_breakpoints> 0".[ 25.640418] hw-breakpoint: found 6 breakpoint and 4 watchpoint registers.Quoted 1 line
- does it fail consistently (every single attempt)?
yes it does.
I will try to find some time to investigate too
Thanks
Eric
Quoted 453 lines
Thanks!
RicardoQuoted 448 lines
Thanks
Eric
Quoted 441 lines
Quoted 158 lines
Quoted 142 lines
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +78
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) + +/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C + +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm); +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21) + +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +12
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7]tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving
sp_el0in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp atregs->sp.Restoring
sp_el0might be too much. So, what do you think of this v3: we keep the saving ofsp_el0intoregs->sp(to keep things the same between el0 and el1) and delete the restoring ofsp_el0?Thanks,
RicardoQuoted 268 lines
Quoted 223 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +108
-0+ .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors: +.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label: + save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection + +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector +.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +91
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h" @@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers; +static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); +} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ { +assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + } +} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 439
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+struct handlers *handlers; + + vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler;How about FIQ, Error? Although they are unlikely, they are valid exceptions.
Quoted 14 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +14
-0+ + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; +} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +}tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 463
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +9
-0+ +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Marc Zyngier
maintainerRe: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 38 lines · Auger Eric
Hi Ricardo,
Quoted 35 lines · Ricardo Koller
Quoted 28 lines · Auger Eric
Hi Ricardo,
Quoted 25 lines · Ricardo Koller
Quoted 2 lines · Auger Eric
Hi Ricardo,
Hi Eric,
Thank you very much for the test.
Quoted 17 lines · Ricardo Koller
Quoted 12 lines · Marc Zyngier
Quoted 11 lines
Quoted 9 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
Hi Eric,
I'm able to reproduce the failure you are seeing on 5.6, specifically with kernels older than this commit:
Quoted 7 lines
Quoted 5 lines
4942dc6638b0KVM: arm64: Write arch.mdcr_el2 changes since lastvcpu_loadon VHEbut not yet on v5.12. Could you share the commit of the kernel you are testing, please?
my host is a 5.12 kernel (
8404c9fbc84b)
Time to compare notes then. What HW are you using? Running VHE or not?
Thanks,
M.
Signature · 1 line
Jazz is not dead. It just smells funny...
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Hi,
Quoted 46 lines · Marc Zyngier
Quoted 45 lines · Auger Eric
Hi Ricardo,
Quoted 42 lines · Ricardo Koller
Quoted 32 lines · Auger Eric
Hi Ricardo,
Quoted 29 lines · Ricardo Koller
Quoted 2 lines · Auger Eric
Hi Ricardo,
Hi Eric,
Thank you very much for the test.
Quoted 21 lines · Ricardo Koller
Quoted 16 lines · Marc Zyngier
Quoted 15 lines
Quoted 13 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64
selftests. The exception handling defaults to an unhandled-exception
handler which aborts the test, just like x86. These handlers can be
overridden by callingvm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a
previous commit,UCALL_UNHANDLED.Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
Hi Eric,
I'm able to reproduce the failure you are seeing on 5.6, specifically with kernels older than this commit:
4942dc6638b0KVM: arm64: Write arch.mdcr_el2 changes since lastvcpu_loadon VHE
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
but not yet on v5.12. Could you share the commit of the kernel you are testing, please?
my host is a 5.12 kernel (
8404c9fbc84b)Time to compare notes then. What HW are you using? Running VHE or not?
VHE yes. Cavium Sabre system.
Thanks
Eric
Quoted 4 lines
Thanks,
M.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 49 lines · Auger Eric
Hi,
Quoted 46 lines · Marc Zyngier
Quoted 45 lines · Auger Eric
Hi Ricardo,
Quoted 42 lines · Ricardo Koller
Quoted 32 lines · Auger Eric
Hi Ricardo,
Quoted 29 lines · Ricardo Koller
Quoted 2 lines · Auger Eric
Hi Ricardo,
Hi Eric,
Thank you very much for the test.
Quoted 21 lines · Ricardo Koller
Quoted 16 lines · Marc Zyngier
Quoted 15 lines
Quoted 13 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64
selftests. The exception handling defaults to an unhandled-exception
handler which aborts the test, just like x86. These handlers can be
overridden by callingvm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a
previous commit,UCALL_UNHANDLED.Quoted 1 line
Quoted 1 line
Quoted 1 line
The exception handling code is heavily inspired on kvm-unit-tests.
running the test on 5.12 I get
Hi Eric,
I'm able to reproduce the failure you are seeing on 5.6, specifically with kernels older than this commit:
4942dc6638b0KVM: arm64: Write arch.mdcr_el2 changes since lastvcpu_loadon VHE
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
but not yet on v5.12. Could you share the commit of the kernel you are testing, please?
my host is a 5.12 kernel (
8404c9fbc84b)Time to compare notes then. What HW are you using? Running VHE or not?
VHE yes. Cavium Sabre system.
On my side. VHE (v5.12) on both QEMU (5.2.0 cpu=max) and Ampere Altra.
Quoted 8 lines
Thanks
Eric
Quoted 4 lines
Thanks,
M.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Zenghui Yu
reviewerRe: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
In reply to Auger EricHi Eric,
Quoted 11 lines · Auger Eric
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794
FYI I can also reproduce it on my VHE box. And Drew's suggestion [*] seemed to work for me. Is the ISB a requirement of architecture?
[*] https://lore.kernel.org/kvm/[email protected]/
Thanks,
Zenghui
Marc Zyngier
maintainerRe: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 15 lines · Zenghui Yu
Hi Eric,
Quoted 11 lines · Auger Eric
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794
Quoted 2 lines
FYI I can also reproduce it on my VHE box. And Drew's suggestion [*] seemed to work for me. Is the ISB a requirement of architecture?
Very much so. Given that there is no context synchronisation (such as ERET or an interrupt) in this code, the CPU is perfectly allowed to delay the system register effect as long as it can.
M.
Signature · 1 line
Jazz is not dead. It just smells funny...
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 16 lines · Marc Zyngier
Quoted 15 lines · Zenghui Yu
Hi Eric,
Quoted 11 lines · Auger Eric
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794
Quoted 10 lines
Quoted 2 lines
FYI I can also reproduce it on my VHE box. And Drew's suggestion [*] seemed to work for me. Is the ISB a requirement of architecture?
Very much so. Given that there is no context synchronisation (such as ERET or an interrupt) in this code, the CPU is perfectly allowed to delay the system register effect as long as it can.
M.Signature · 1 line
Jazz is not dead. It just smells funny...
Thank you very much Eric, Zenghui, Marc, and Andrew (for the ISB suggestion)!
As per Zenghui test, will send a V3 that includes the missing ISBs. Hopefully that will fix the issue for Eric as well. It's very interesting that the CPU seems to _always_ reorder those instructions.
Thanks!
Ricardo
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 28 lines · Ricardo Koller
Quoted 19 lines · Marc Zyngier
Quoted 15 lines · Zenghui Yu
Hi Eric,
Quoted 11 lines · Auger Eric
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794Quoted 2 lines
FYI I can also reproduce it on my VHE box. And Drew's suggestion [*] seemed to work for me. Is the ISB a requirement of architecture?
Quoted 7 lines
Very much so. Given that there is no context synchronisation (such as ERET or an interrupt) in this code, the CPU is perfectly allowed to delay the system register effect as long as it can.
M.--
Jazz is not dead. It just smells funny...
Quoted 6 lines
Thank you very much Eric, Zenghui, Marc, and Andrew (for the ISB suggestion)!
As per Zenghui test, will send a V3 that includes the missing ISBs. Hopefully that will fix the issue for Eric as well. It's very interesting that the CPU seems to _always_ reorder those instructions.
I suspect that because hitting the debug registers can be a costly operation (it mobilises a lot of resources in the CPU), there is a strong incentive to let it slide until there is an actual mandate to commit the resource.
It also means that SW can issue a bunch of these without too much overhead, and only pay the cost *once*.
Your N1 CPU seems to be less aggressive on this. Implement choice, I'd say (it probably is more aggressive than TX2 on other things). Also, QEMU will almost always hide these problems, due to the nature of TCG.
Thanks,
M.
Signature · 5 lines
Jazz is not dead. It just smells funny... _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 44 lines · Marc Zyngier
Quoted 28 lines · Ricardo Koller
Quoted 19 lines · Marc Zyngier
Quoted 15 lines · Zenghui Yu
Hi Eric,
Quoted 11 lines · Auger Eric
running the test on 5.12 I get
==== Test Assertion Failure ====
aarch64/debug-exceptions.c:232: false pid=6477 tid=6477 errno=4 - Interrupted system call1 0x000000000040147b: main at debug-exceptions.c:230 2 0x000003ff8aa60de3: ?? ??:0 3 0x0000000000401517: _start at :? Failed guest assert: hw_bp_addr == PC(hw_bp) at aarch64/debug-exceptions.c:105 values: 0, 0x401794Quoted 2 lines
FYI I can also reproduce it on my VHE box. And Drew's suggestion [*] seemed to work for me. Is the ISB a requirement of architecture?
Quoted 7 lines
Very much so. Given that there is no context synchronisation (such as ERET or an interrupt) in this code, the CPU is perfectly allowed to delay the system register effect as long as it can.
M.--
Jazz is not dead. It just smells funny...Quoted 6 lines
Thank you very much Eric, Zenghui, Marc, and Andrew (for the ISB suggestion)!
As per Zenghui test, will send a V3 that includes the missing ISBs. Hopefully that will fix the issue for Eric as well. It's very interesting that the CPU seems to _always_ reorder those instructions.
I suspect that because hitting the debug registers can be a costly operation (it mobilises a lot of resources in the CPU), there is a strong incentive to let it slide until there is an actual mandate to commit the resource.
It also means that SW can issue a bunch of these without too much overhead, and only pay the cost *once*.
Quoted 10 lines
Your N1 CPU seems to be less aggressive on this. Implement choice, I'd say (it probably is more aggressive than TX2 on other things). Also, QEMU will almost always hide these problems, due to the nature of TCG.
Thanks,
M.Signature · 1 line
Jazz is not dead. It just smells funny...
Thank you, this is very informative.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerRe: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
In reply to Ricardo KollerQuoted 4 lines · Ricardo Koller
Quoted 2 lines · Marc Zyngier
On Sat, 01 May 2021 00:24:06 +0100,
Ricardo Koller <[email protected]> wrote:
[...]
Quoted 8 lines
Quoted 8 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 2 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +2
-0+ .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.
Quoted 6 lines
Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving
sp_el0in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp atregs->sp.
We already have sp_el0 consistency by virtue of having it stored in in a sysreg.
Quoted 3 lines
Restoring
sp_el0might be too much. So, what do you think of this
v3: we keep the saving ofsp_el0intoregs->sp(to keep things the same between el0 and el1) and delete the restoring ofsp_el0?
To me, the whole purpose of saving some some context is to allow the exception handling code to run C code and introspect the interrupted state. But saving things that are not affected by the context change seems a bit pointless.
One thing I'd like to see though is to save sp_el1 as it was at the point of the exception (because that is meaningful to get the exception context -- think of an unaligned EL1 stack for example), which means correcting the value that gets saved.
So I would suggest to *only* save sp_el1, to always save it (irrespective of the exception coming from EL0 or EL1), and to save a retro-corrected value so that the handler can directly know where the previous stack pointer was.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
Quoted 36 lines · Marc Zyngier
Quoted 4 lines · Ricardo Koller
Quoted 2 lines · Marc Zyngier
On Sat, 01 May 2021 00:24:06 +0100,
Ricardo Koller <[email protected]> wrote:[...]
Quoted 8 lines
Quoted 8 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S line 22
Quoted 2 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +2
-0+ .if \vector >= 8 + mrs x1, sp_el0I'm still a bit perplexed by this.
SP_EL0is never changed, since you always run in handler mode. Therefore, saving/restoring it is only overhead. If an exception handler wants to introspect it, it is already available in the relevant system register.Quoted 6 lines
Quoted 2 lines
Or did you have something else in mind for it?
Not really. The reason for saving
sp_el0in there was just for consistency, so that handlers for both el0 and el1 exceptions could get the sp atregs->sp.We already have
sp_el0consistency by virtue of having it stored in in a sysreg.Quoted 3 lines
Restoring
sp_el0might be too much. So, what do you think of this
v3: we keep the saving ofsp_el0intoregs->sp(to keep things the same between el0 and el1) and delete the restoring ofsp_el0?To me, the whole purpose of saving some some context is to allow the exception handling code to run C code and introspect the interrupted state. But saving things that are not affected by the context change seems a bit pointless.
Quoted 4 lines
One thing I'd like to see though is to save
sp_el1as it was at the point of the exception (because that is meaningful to get the exception context -- think of an unaligned EL1 stack for example), which means correcting the value that gets saved.
Got it. Replacing:
mov x1, sp
with:
add x1, sp, #16 * 17
Quoted 5 lines
So I would suggest to *only* save
sp_el1, to always save it (irrespective of the exception coming from EL0 or EL1), and to save a retro-corrected value so that the handler can directly know where the previous stack pointer was.
Sounds good, will send a V3 accordingly.
Thanks!
Ricardo
Quoted 7 lines
Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
Re: [PATCH v2 4/5] KVM: selftests: Add exception handling support for aarch64
In reply to Ricardo KollerQuoted 92 lines · Ricardo Koller
Add the infrastructure needed to enable exception handling in aarch64 selftests. The exception handling defaults to an unhandled-exception handler which aborts the test, just like x86. These handlers can be overridden by calling
vm_install_vector_handler(vector)orvm_install_exception_handler(vector, ec). The unhandled exception reporting from the guest is done using the ucall type introduced in a previous commit,UCALL_UNHANDLED.The exception handling code is heavily inspired on kvm-unit-tests.
Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/Makefile | 2 +- .../selftests/kvm/include/aarch64/processor.h | 78 +++++++++++ .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++++++++++ 4 files changed, 333 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/kvm/lib/aarch64/handlers.Stools/testing/selftests/kvm/Makefile +1
-1@@ -35,7 +35,7 @@ endif LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c lib/test_util.c lib/guest_modes.c lib/perf_test_util.c LIBKVM_x86_64 = lib/x86_64/processor.c lib/x86_64/vmx.c lib/x86_64/svm.c lib/x86_64/ucall.c lib/x86_64/handlers.S -LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c +LIBKVM_aarch64 = lib/aarch64/processor.c lib/aarch64/ucall.c lib/aarch64/handlers.S LIBKVM_s390x = lib/s390x/processor.c lib/s390x/ucall.c lib/s390x/diag318_test_handler.c TEST_GEN_PROGS_x86_64 = x86_64/cr4_cpuid_sync_testtools/testing/selftests/kvm/include/aarch64/processor.h +38
-0@@ -8,6 +8,7 @@#define SELFTEST_KVM_PROCESSOR_H #include "kvm_util.h" +#include <linux/stringify.h> #define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \@@ -18,6 +19,7 @@#define MAIR_EL1 3, 0, 10, 2, 0#define TTBR0_EL1 3, 0, 2, 0, 0#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0/* * Default MAIR @@ -56,4 +58,80 @@ void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *inivoid aarch64_vcpu_add_default (struct kvm_vm *vm,uint32_t vcpuid,struct kvm_vcpu_init *init,void *guest_code); +struct ex_regs { +u64 regs[31 ]; +u64 sp; +u64 pc; +u64 pstate; +}; + +#define VECTOR_NUM 16 + +enum { +VECTOR_SYNC_CURRENT_SP0 , +VECTOR_IRQ_CURRENT_SP0 , +VECTOR_FIQ_CURRENT_SP0 , +VECTOR_ERROR_CURRENT_SP0 , + +VECTOR_SYNC_CURRENT , +VECTOR_IRQ_CURRENT , +VECTOR_FIQ_CURRENT , +VECTOR_ERROR_CURRENT , + +VECTOR_SYNC_LOWER_64 , +VECTOR_IRQ_LOWER_64 , +VECTOR_FIQ_LOWER_64 , +VECTOR_ERROR_LOWER_64 , + +VECTOR_SYNC_LOWER_32 , +VECTOR_IRQ_LOWER_32 , +VECTOR_FIQ_LOWER_32 , +VECTOR_ERROR_LOWER_32 , +}; + +#define VECTOR_IS_SYNC (v) ((v) == VECTOR_SYNC_CURRENT_SP0 || \ + (v) == VECTOR_SYNC_CURRENT || \ + (v) == VECTOR_SYNC_LOWER_64 || \ + (v) == VECTOR_SYNC_LOWER_32) +
tools/testing/selftests/kvm/include/aarch64/processor.h void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *ini · line 106
Quoted 10 lines
tools/testing/selftests/kvm/include/aarch64/processor.h +10
-0+/* Some common EC (Exception classes) */ +#define ESR_EC_ILLEGAL_INS 0x0e +#define ESR_EC_SVC64 0x15 +#define ESR_EC_IABORT_CURRENT 0x21 +#define ESR_EC_DABORT_CURRENT 0x25 +#define ESR_EC_SERROR 0x2f +#define ESR_EC_HW_BP_CURRENT 0x31 +#define ESR_EC_SSTEP_CURRENT 0x33 +#define ESR_EC_WP_CURRENT 0x35 +#define ESR_EC_BRK_INS 0x3C
Let's leave it to the specific unit tests to define the above EC's as they are needed. The unit tests can also then decide if the defines belong in the common header or in the unit test itself.
Quoted 7 lines
tools/testing/selftests/kvm/include/aarch64/processor.h +7
-0+ +#define ESR_EC_NUM 64 + +#define ESR_EC_SHIFT 26 +#define ESR_EC_MASK (ESR_EC_NUM - 1) + +void vm_init_descriptor_tables (struct kvm_vm *vm);
tools/testing/selftests/kvm/include/aarch64/processor.h void aarch64_vcpu_setup(struct kvm_vm *vm, int vcpuid, struct kvm_vcpu_init *ini · line 123
Quoted 10 lines
tools/testing/selftests/kvm/include/aarch64/processor.h +10
-0+void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid); + +typedef void (*handler_fn )(struct ex_regs *); +void vm_install_exception_handler (struct kvm_vm *vm, +int vector,int ec,handler_fn handler); +void vm_install_vector_handler (struct kvm_vm *vm, +int vector,handler_fn handler); + +#define SPSR_D (1 << 9) +#define SPSR_SS (1 << 21)
I think these two SPSR defines belong directly in the debug unit test, rather than the common header.
Quoted 92 lines
tools/testing/selftests/kvm/include/aarch64/processor.h +13
-0+ +#define write_sysreg (reg, val) \ +({ \ + u64 __val = (u64)(val); \ + asm volatile("msr " __stringify(reg) ", %x0" : : "rZ" (__val)); \ +}) + +#define read_sysreg (reg) \ +({ u64 val; \ + asm volatile("mrs %0, "__stringify(reg) : "=r"(val) : : "memory");\ + val; \ +}) +#endif /* SELFTEST_KVM_PROCESSOR_H */ tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +72
-0@@ -0,0 +1,130 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +.macro save_registers, vector + add sp, sp, #-16 * 17 + + stp x0, x1, [sp, #16 * 0] + stp x2, x3, [sp, #16 * 1] + stp x4, x5, [sp, #16 * 2] + stp x6, x7, [sp, #16 * 3] + stp x8, x9, [sp, #16 * 4] + stp x10, x11, [sp, #16 * 5] + stp x12, x13, [sp, #16 * 6] + stp x14, x15, [sp, #16 * 7] + stp x16, x17, [sp, #16 * 8] + stp x18, x19, [sp, #16 * 9] + stp x20, x21, [sp, #16 * 10] + stp x22, x23, [sp, #16 * 11] + stp x24, x25, [sp, #16 * 12] + stp x26, x27, [sp, #16 * 13] + stp x28, x29, [sp, #16 * 14] + + .if \vector >= 8 + mrs x1, sp_el0 + .else + /* + * This stores sp_el1 into ex_regs.sp so exception handlers can + * "look" at it. It will _not_ be used to restore the sp_el1 on + * return from the exception so handlers can not update it. + */ + mov x1, sp + .endif + stp x30, x1, [sp, #16 * 15] /* x30, SP */ + + mrs x1, elr_el1 + mrs x2, spsr_el1 + stp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ +.endm + +.macro restore_registers, vector + ldp x1, x2, [sp, #16 * 16] /* PC, PSTATE */ + msr elr_el1, x1 + msr spsr_el1, x2 + + ldp x30, x1, [sp, #16 * 15] /* x30, SP */ + .if \vector >= 8 + msr sp_el0, x1 + .endif + + ldp x28, x29, [sp, #16 * 14] + ldp x26, x27, [sp, #16 * 13] + ldp x24, x25, [sp, #16 * 12] + ldp x22, x23, [sp, #16 * 11] + ldp x20, x21, [sp, #16 * 10] + ldp x18, x19, [sp, #16 * 9] + ldp x16, x17, [sp, #16 * 8] + ldp x14, x15, [sp, #16 * 7] + ldp x12, x13, [sp, #16 * 6] + ldp x10, x11, [sp, #16 * 5] + ldp x8, x9, [sp, #16 * 4] + ldp x6, x7, [sp, #16 * 3] + ldp x4, x5, [sp, #16 * 2] + ldp x2, x3, [sp, #16 * 1] + ldp x0, x1, [sp, #16 * 0] + + add sp, sp, #16 * 17 + + eret +.endm + +.pushsection ".entry.text", "ax" +.balign 0x800 +.global vectors +vectors:
tools/testing/selftests/kvm/lib/aarch64/handlers.S line 82
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+.popsection + +.set vector, 0 + +/* + * Build an exception handler for vector and append a jump to it into + * vectors (while making sure that it's 0x80 aligned). + */ +.macro HANDLER, label +handler_\()\label:
I don't think we need the \(). Based on [1] it may be necessary *after* a macro argument to separate it from the following text. I don't think it serves any purpose before the argument though.
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ save_registers vector + mov x0, sp + mov x1, #vector + bl route_exception + restore_registers vector + +.pushsection ".entry.text", "ax" +.balign 0x80 + b handler_\()\label +.popsection
tools/testing/selftests/kvm/lib/aarch64/handlers.S line 102
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +10
-0+ +.set vector, vector + 1 +.endm + +.macro HANDLER_INVALID +.pushsection ".entry.text", "ax" +.balign 0x80 +/* This will abort so no need to save and restore registers. */ + mov x0, #vector + b kvm_exit_unexpected_vector
I'd put a 'b .' here out of paranoia.
Quoted 41 lines
tools/testing/selftests/kvm/lib/aarch64/handlers.S new file +28
-0+.popsection + +.set vector, vector + 1 +.endm + +/* + * Caution: be sure to not add anything between the declaration of vectors + * above and these macro calls that will build the vectors table below it. + */ + HANDLER_INVALID // Synchronous EL1t + HANDLER_INVALID // IRQ EL1t + HANDLER_INVALID // FIQ EL1t + HANDLER_INVALID // Error EL1t + + HANDLER el1h_sync // Synchronous EL1h + HANDLER el1h_irq // IRQ EL1h + HANDLER el1h_fiq // FIQ EL1h + HANDLER el1h_error // Error EL1h + + HANDLER el0_sync_64 // Synchronous 64-bit EL0 + HANDLER el0_irq_64 // IRQ 64-bit EL0 + HANDLER el0_fiq_64 // FIQ 64-bit EL0 + HANDLER el0_error_64 // Error 64-bit EL0 + + HANDLER el0_sync_32 // Synchronous 32-bit EL0 + HANDLER el0_irq_32 // IRQ 32-bit EL0 + HANDLER el0_fiq_32 // FIQ 32-bit EL0 + HANDLER el0_error_32 // Error 32-bit EL0tools/testing/selftests/kvm/lib/aarch64/processor.c +1
-0@@ -6,6 +6,7 @@*/ #include <linux/compiler.h> +#include <assert.h> #include "kvm_util.h" #include "../kvm_util_internal.h"
tools/testing/selftests/kvm/lib/aarch64/processor.c line 18
Quoted 4 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +1
-0@@ -14,6 +15,8 @@#define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000#define DEFAULT_ARM64_GUEST_STACK_VADDR_MIN 0xac0000 +vm_vaddr_t exception_handlers;
static? Also some functions below could maybe be static.
tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 342
Quoted 11 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +4
-0+static uint64_t page_align (struct kvm_vm *vm,uint64_t v) {return (v + vm->page_size) & ~(vm->page_size -1 );@@ -334,6 +337,127 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } +void kvm_exit_unexpected_vector (int vector) +{ +ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ );
Instead of the 'b . ' suggested above, there could be a 'while (1) ;' here.
tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 347
Quoted 5 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +5
-0+} + +void kvm_exit_unexpected_exception (int vector,uint64_t ec) +{ +ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ );
And also a 'while (1) ;' here.
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +8
-0+} +void assert_on_unhandled_exception (struct kvm_vm *vm,uint32_t vcpuid) { +struct ucall uc; + +if (get_ucall (vm, vcpuid, &uc) !=UCALL_UNHANDLED ) +return ; + +if (uc.args[2 ])/* valid_ec */ {
tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 367
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+assert (VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx, ec:0x%lx)" , + uc.args[0 ], uc.args[1 ]); + }else { +assert (!VECTOR_IS_SYNC (uc.args[0 ])); +TEST_ASSERT (false, +"Unexpected exception (vector:0x%lx)" , + uc.args[0 ]); + }
Can use TEST_FAIL for the above to TEST_ASSERT(false, ...)
Quoted 64 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +64
-0+} + +/* +* This exception handling code was heavily inspired on kvm-unit-tests. There +* is a set of default vector handlers stored in vector_handlers. These default +* vector handlers call user-installed handlers stored in exception_handlers. +* Synchronous handlers are indexed by (vector, ec), and irq handlers by +* (vector, ec=0). +*/ + +typedef void (*vector_fn )(struct ex_regs *,int vector); + +struct handlers { +vector_fn vector_handlers[VECTOR_NUM ]; +handler_fn exception_handlers[VECTOR_NUM ][ESR_EC_NUM ]; +}; + +void vcpu_init_descriptor_tables (struct kvm_vm *vm,uint32_t vcpuid) +{ +extern char vectors; + +set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); +} + +void default_sync_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; +uint64_t esr =read_sysreg (esr_el1); +uint64_t ec = (esr >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + +GUEST_ASSERT (VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][ec]) + handlers->exception_handlers[vector][ec](regs); +else +kvm_exit_unexpected_exception (vector, ec); +} + +void default_irq_handler (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +GUEST_ASSERT (!VECTOR_IS_SYNC (vector)); + +if (handlers && handlers->exception_handlers[vector][0 ]) + handlers->exception_handlers[vector][0 ](regs); +else +kvm_exit_unexpected_vector (vector); +} + +void route_exception (struct ex_regs *regs,int vector) +{ +struct handlers *handlers = (struct handlers *)exception_handlers; + +if (handlers && handlers->vector_handlers[vector]) + handlers->vector_handlers[vector](regs, vector); +else +kvm_exit_unexpected_vector (vector); +} + +void vm_init_descriptor_tables (struct kvm_vm *vm) +{ +struct handlers *handlers; +
tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...) · line 441
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +10
-0+ vm->handlers =vm_vaddr_alloc (vm,sizeof (struct handlers ), + vm->page_size,0 ,0 ); + + handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + handlers->vector_handlers[VECTOR_SYNC_CURRENT ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_CURRENT ] = default_irq_handler; + handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; + handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_irq_handler; + + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers;
I see this is derived from the x86 code, so I guess it's OK for now, but having the handlers be VM state instead of VCPU state means the VCPUs will have to always share the same handlers. I think we should make them per-vcpu instead.
Quoted 25 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +21
-0+} + +void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); +assert (ec <ESR_EC_NUM ); + handlers->exception_handlers[vector][ec] = handler; +} + +void vm_install_vector_handler (struct kvm_vm *vm,int vector, +void (*handler)(struct ex_regs *)) +{ +struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); + +assert (!VECTOR_IS_SYNC (vector)); +assert (vector <VECTOR_NUM ); + handlers->exception_handlers[vector][0 ] = handler; }Signature · 1 line
2.31.1.527.g47e6f16901-goog
Thanks,
drew
PATCH v2 5/5 KVM: selftests: Add aarch64/debug-exceptions test
In reply to Ricardo KollerCovers fundamental tests for debug exceptions. The guest installs and handle its debug exceptions itself, without KVM_SET_GUEST_DEBUG.
--- tools/testing/selftests/kvm/.gitignore | 1 + tools/testing/selftests/kvm/Makefile | 1 + .../selftests/kvm/aarch64/debug-exceptions.c | 244 ++++++++++++++++++ .../selftests/kvm/include/aarch64/processor.h | 14 +- 4 files changed, 254 insertions(+), 6 deletions(-) create mode 100644 tools/testing/selftests/kvm/aarch64/debug-exceptions.c
tools/testing/selftests/kvm/.gitignore +1 -0
@@ -1,4 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only +/aarch64/debug-exceptions /aarch64/get-reg-list /aarch64/get-reg-list-sve /aarch64/vgic_init
tools/testing/selftests/kvm/Makefile +1 -0
@@ -73,6 +73,7 @@ TEST_GEN_PROGS_x86_64 += memslot_modification_stress_test TEST_GEN_PROGS_x86_64 += set_memory_region_test TEST_GEN_PROGS_x86_64 += steal_time +TEST_GEN_PROGS_aarch64 += aarch64/debug-exceptions TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list-sve TEST_GEN_PROGS_aarch64 += aarch64/vgic_init
tools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +244 -0
244 lines not on this page · raw message
tools/testing/selftests/kvm/include/aarch64/processor.h +8 -6
@@ -14,12 +14,14 @@#define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \ KVM_REG_ARM_CORE | KVM_REG_ARM_CORE_REG(x)) -#define CPACR_EL1 3, 0, 1, 0, 2 -#define TCR_EL1 3, 0, 2, 0, 2 -#define MAIR_EL1 3, 0, 10, 2, 0 -#define TTBR0_EL1 3, 0, 2, 0, 0 -#define SCTLR_EL1 3, 0, 1, 0, 0 -#define VBAR_EL1 3, 0, 12, 0, 0 +#define CPACR_EL1 3, 0, 1, 0, 2 +#define TCR_EL1 3, 0, 2, 0, 2 +#define MAIR_EL1 3, 0, 10, 2, 0 +#define TTBR0_EL1 3, 0, 2, 0, 0 +#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0 + +#define ID_AA64DFR0_EL1 3, 0, 0, 5, 0/* * Default MAIR
Signature · 1 line
2.31.1.527.g47e6f16901-goog
Re: [PATCH v2 5/5] KVM: selftests: Add aarch64/debug-exceptions test
Quoted 66 lines · Ricardo Koller
Covers fundamental tests for debug exceptions. The guest installs and handle its debug exceptions itself, without
KVM_SET_GUEST_DEBUG.Signed-off-by Ricardo Koller <[email protected]>--- tools/testing/selftests/kvm/.gitignore | 1 + tools/testing/selftests/kvm/Makefile | 1 + .../selftests/kvm/aarch64/debug-exceptions.c | 244 ++++++++++++++++++ .../selftests/kvm/include/aarch64/processor.h | 14 +- 4 files changed, 254 insertions(+), 6 deletions(-) create mode 100644 tools/testing/selftests/kvm/aarch64/debug-exceptions.ctools/testing/selftests/kvm/.gitignore +1
-0@@ -1,4 +1,5 @@ # SPDX-License-Identifier: GPL-2.0-only +/aarch64/debug-exceptions /aarch64/get-reg-list /aarch64/get-reg-list-sve /aarch64/vgic_inittools/testing/selftests/kvm/Makefile +1
-0@@ -73,6 +73,7 @@ TEST_GEN_PROGS_x86_64 += memslot_modification_stress_test TEST_GEN_PROGS_x86_64 += set_memory_region_test TEST_GEN_PROGS_x86_64 += steal_time +TEST_GEN_PROGS_aarch64 += aarch64/debug-exceptions TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list-sve TEST_GEN_PROGS_aarch64 += aarch64/vgic_inittools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +26
-0@@ -0,0 +1,244 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <test_util.h> +#include <kvm_util.h> +#include <processor.h> + +#define VCPU_ID 0 + +#define MDSCR_KDE (1 << 13) +#define MDSCR_MDE (1 << 15) +#define MDSCR_SS (1 << 0) + +#define DBGBCR_LEN8 (0xff << 5) +#define DBGBCR_EXEC (0x0 << 3) +#define DBGBCR_EL1 (0x1 << 1) +#define DBGBCR_E (0x1 << 0) + +#define DBGWCR_LEN8 (0xff << 5) +#define DBGWCR_RD (0x1 << 3) +#define DBGWCR_WR (0x2 << 3) +#define DBGWCR_EL1 (0x1 << 1) +#define DBGWCR_E (0x1 << 0) + +extern unsigned char sw_bp, hw_bp, bp_svc, bp_brk, hw_wp, ss_start; +static volatile uint64_t sw_bp_addr, hw_bp_addr; +static volatile uint64_t wp_addr, wp_data_addr; +static volatile uint64_t svc_addr;
tools/testing/selftests/kvm/aarch64/debug-exceptions.c line 36
Quoted 10 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +10
-0+static volatile uint64_t ss_addr[4 ], ss_idx; +#define PC (v) ((uint64_t)&(v)) + +static void reset_debug_state (void ) +{ + asmvolatile ("msr daifset, #8" ); + +write_sysreg (osdlr_el1,0 ); +write_sysreg (oslar_el1,0 ); + asmvolatile ("isb" : : :"memory" );
We may want to create an isb() macro in include/aarch64/processor.h
Quoted 14 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +14
-0+ +write_sysreg (mdscr_el1,0 ); +/* This test only uses the first bp and wp slot. */ +write_sysreg (dbgbvr0_el1,0 ); +write_sysreg (dbgbcr0_el1,0 ); +write_sysreg (dbgwcr0_el1,0 ); +write_sysreg (dbgwvr0_el1,0 ); + asmvolatile ("isb" : : :"memory" ); +} + +static void install_wp (uint64_t addr) +{ +uint32_t wcr; +uint32_t mdscr;
tools/testing/selftests/kvm/aarch64/debug-exceptions.c line 60
Quoted 10 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +10
-0+ + wcr =DBGWCR_LEN8 |DBGWCR_RD |DBGWCR_WR |DBGWCR_EL1 |DBGWCR_E ; +write_sysreg (dbgwcr0_el1, wcr); +write_sysreg (dbgwvr0_el1, addr); + asmvolatile ("isb" : : :"memory" ); + + asmvolatile ("msr daifclr, #8" ); + + mdscr =read_sysreg (mdscr_el1) |MDSCR_KDE |MDSCR_MDE ; +write_sysreg (mdscr_el1, mdscr);
no isb?
Quoted 6 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +6
-0+} + +static void install_hw_bp (uint64_t addr) +{ +uint32_t bcr; +uint32_t mdscr;
tools/testing/selftests/kvm/aarch64/debug-exceptions.c line 76
Quoted 10 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +10
-0+ + bcr =DBGBCR_LEN8 |DBGBCR_EXEC |DBGBCR_EL1 |DBGBCR_E ; +write_sysreg (dbgbcr0_el1, bcr); +write_sysreg (dbgbvr0_el1, addr); + asmvolatile ("isb" : : :"memory" ); + + asmvolatile ("msr daifclr, #8" ); + + mdscr =read_sysreg (mdscr_el1) |MDSCR_KDE |MDSCR_MDE ; +write_sysreg (mdscr_el1, mdscr);
no isb?
tools/testing/selftests/kvm/aarch64/debug-exceptions.c line 86
Quoted 10 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +10
-0+} + +static void install_ss (void ) +{ +uint32_t mdscr; + + asmvolatile ("msr daifclr, #8" ); + + mdscr =read_sysreg (mdscr_el1) |MDSCR_KDE |MDSCR_SS ; +write_sysreg (mdscr_el1, mdscr);
no isb?
Quoted 186 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c new file +158
-0+} + +static volatile char write_data; + +static void guest_code (void ) +{ +GUEST_SYNC (0 ); + +/* Software-breakpoint */ + asmvolatile ("sw_bp: brk #0" ); +GUEST_ASSERT_EQ (sw_bp_addr,PC (sw_bp)); + +GUEST_SYNC (1 ); + +/* Hardware-breakpoint */ +reset_debug_state (); +install_hw_bp (PC (hw_bp)); + asmvolatile ("hw_bp: nop" ); +GUEST_ASSERT_EQ (hw_bp_addr,PC (hw_bp)); + +GUEST_SYNC (2 ); + +/* Hardware-breakpoint + svc */ +reset_debug_state (); +install_hw_bp (PC (bp_svc)); + asmvolatile ("bp_svc: svc #0" ); +GUEST_ASSERT_EQ (hw_bp_addr,PC (bp_svc)); +GUEST_ASSERT_EQ (svc_addr,PC (bp_svc) +4 ); + +GUEST_SYNC (3 ); + +/* Hardware-breakpoint + software-breakpoint */ +reset_debug_state (); +install_hw_bp (PC (bp_brk)); + asmvolatile ("bp_brk: brk #0" ); +GUEST_ASSERT_EQ (sw_bp_addr,PC (bp_brk)); +GUEST_ASSERT_EQ (hw_bp_addr,PC (bp_brk)); + +GUEST_SYNC (4 ); + +/* Watchpoint */ +reset_debug_state (); +install_wp (PC (write_data)); + write_data ='x' ; +GUEST_ASSERT_EQ (write_data,'x' ); +GUEST_ASSERT_EQ (wp_data_addr,PC (write_data)); + +GUEST_SYNC (5 ); + +/* Single-step */ +reset_debug_state (); +install_ss (); + ss_idx =0 ; + asmvolatile ("ss_start:\n" +"mrs x0, esr_el1\n" +"add x0, x0, #1\n" +"msr daifset, #8\n" + : : :"x0" ); +GUEST_ASSERT_EQ (ss_addr[0 ],PC (ss_start)); +GUEST_ASSERT_EQ (ss_addr[1 ],PC (ss_start) +4 ); +GUEST_ASSERT_EQ (ss_addr[2 ],PC (ss_start) +8 ); + +GUEST_DONE (); +} + +static void guest_sw_bp_handler (struct ex_regs *regs) +{ + sw_bp_addr = regs->pc; + regs->pc +=4 ; +} + +static void guest_hw_bp_handler (struct ex_regs *regs) +{ + hw_bp_addr = regs->pc; + regs->pstate |=SPSR_D ; +} + +static void guest_wp_handler (struct ex_regs *regs) +{ + wp_data_addr =read_sysreg (far_el1); + wp_addr = regs->pc; + regs->pstate |=SPSR_D ; +} + +static void guest_ss_handler (struct ex_regs *regs) +{ +GUEST_ASSERT_1 (ss_idx <4 , ss_idx); + ss_addr[ss_idx++] = regs->pc; + regs->pstate |=SPSR_SS ; +} + +static void guest_svc_handler (struct ex_regs *regs) +{ + svc_addr = regs->pc; +} + +static int debug_version (struct kvm_vm *vm) +{ +uint64_t id_aa64dfr0; + +get_reg (vm,VCPU_ID ,ARM64_SYS_REG (ID_AA64DFR0_EL1 ), &id_aa64dfr0); +return id_aa64dfr0 &0xf ; +} + +int main (int argc,char *argv[]) +{ +struct kvm_vm *vm; +struct ucall uc; +int stage; + + vm =vm_create_default (VCPU_ID ,0 , guest_code); +ucall_init (vm,NULL ); + +vm_init_descriptor_tables (vm); +vcpu_init_descriptor_tables (vm,VCPU_ID ); + +if (debug_version (vm) <6 ) { +print_skip ("Armv8 debug architecture not supported." ); +kvm_vm_free (vm); +exit (KSFT_SKIP ); + } + +vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +ESR_EC_BRK_INS , guest_sw_bp_handler); +vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +ESR_EC_HW_BP_CURRENT , guest_hw_bp_handler); +vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +ESR_EC_WP_CURRENT , guest_wp_handler); +vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +ESR_EC_SSTEP_CURRENT , guest_ss_handler); +vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +ESR_EC_SVC64 , guest_svc_handler); + +for (stage =0 ; stage <7 ; stage++) { +vcpu_run (vm,VCPU_ID ); + +switch (get_ucall (vm,VCPU_ID , &uc)) { +case UCALL_SYNC : +TEST_ASSERT (uc.args[1 ] == stage, +"Stage %d: Unexpected sync ucall, got %lx" , + stage, (ulong )uc.args[1 ]); +break ; +case UCALL_ABORT : +TEST_FAIL ("%s at %s:%ld\n\tvalues: %#lx, %#lx" , + (const char *)uc.args[0 ], + __FILE__, uc.args[1 ], uc.args[2 ], uc.args[3 ]); +break ; +case UCALL_DONE : + goto done; +default : +TEST_FAIL ("Unknown ucall %lu" , uc.cmd); + } + } + +done: +kvm_vm_free (vm); +return 0 ; +}tools/testing/selftests/kvm/include/aarch64/processor.h +8
-6@@ -14,12 +14,14 @@#define ARM64_CORE_REG (x) (KVM_REG_ARM64 | KVM_REG_SIZE_U64 | \ KVM_REG_ARM_CORE | KVM_REG_ARM_CORE_REG(x)) -#define CPACR_EL1 3, 0, 1, 0, 2 -#define TCR_EL1 3, 0, 2, 0, 2 -#define MAIR_EL1 3, 0, 10, 2, 0 -#define TTBR0_EL1 3, 0, 2, 0, 0 -#define SCTLR_EL1 3, 0, 1, 0, 0 -#define VBAR_EL1 3, 0, 12, 0, 0 +#define CPACR_EL1 3, 0, 1, 0, 2 +#define TCR_EL1 3, 0, 2, 0, 2 +#define MAIR_EL1 3, 0, 10, 2, 0 +#define TTBR0_EL1 3, 0, 2, 0, 0 +#define SCTLR_EL1 3, 0, 1, 0, 0 +#define VBAR_EL1 3, 0, 12, 0, 0 + +#define ID_AA64DFR0_EL1 3, 0, 0, 5, 0/* * Default MAIR Signature · 1 line
2.31.1.527.g47e6f16901-goog
Other than the potentially missing isb's it looks good to me.
Thanks,
drew
Paolo Bonzini
maintainerQuoted 10 lines · Ricardo Koller
Hi,
These patches add a debug exception test in aarch64 KVM selftests while also adding basic exception handling support.
The structure of the exception handling is based on its x86 counterpart. Tests use the same calls to initialize exception handling and both architectures allow tests to override the handler for a particular vector, or (vector, ec) for synchronous exceptions in the arm64 case.
Quoted 6 lines
The debug test is similar to x86_64/debug_regs, except that the x86 one controls the debugging from outside the VM. This proposed arm64 test controls and handles debug exceptions from the inside.
Thanks,
Ricardo
Marc, are you going to queue this in your tree?
Thanks,
Paolo
Quoted 42 lines
v1 -> v2:
Addressed comments from Andrew and Marc (thank you very much):
- renamevm_handle_exceptionin all tests.
- introduceUCALL_UNHANDLEDin x86 first.
- moveGUEST_ASSERT_EQto common utils header.
- handle sync and other exceptions separately: use two tables (like kvm-unit-tests).
- add two separate functions for installing sync versus other exceptions
- changes inhandlers.S: use the same layout asuser_pt_regs, treat the EL1t vectors as invalid, refactor the vector table creation to not use manual numbering, add comments, remove LR from the stored registers.
- changes indebug-exceptions.c: remove unused headers, use the commonGUEST_ASSERT_EQ, usevcpu_runinstead of_vcpu_run.
- changes inprocessor.h:write_sysregwith support for xzr, replace EL1 with current in macro names, defineESR_EC_MASKas ESR_EC_NUM-1.Ricardo Koller (5):
KVM: selftests: Renamevm_handle_exception
KVM: selftests: IntroduceUCALL_UNHANDLEDfor unhandled vector
reporting
KVM: selftests: MoveGUEST_ASSERT_EQto utils header
KVM: selftests: Add exception handling support for aarch64
KVM: selftests: Add aarch64/debug-exceptions testtools/testing/selftests/kvm/.gitignore | 1 + tools/testing/selftests/kvm/Makefile | 3 +- .../selftests/kvm/aarch64/debug-exceptions.c | 244 ++++++++++++++++++ .../selftests/kvm/include/aarch64/processor.h | 90 ++++++- .../testing/selftests/kvm/include/kvm_util.h | 10 + .../selftests/kvm/include/x86_64/processor.h | 4 +- .../selftests/kvm/lib/aarch64/handlers.S | 130 ++++++++++ .../selftests/kvm/lib/aarch64/processor.c | 124 +++++++++ .../selftests/kvm/lib/x86_64/processor.c | 19 +- .../selftests/kvm/x86_64/kvm_pv_test.c | 2 +- .../selftests/kvm/x86_64/tsc_msrs_test.c | 9 - .../kvm/x86_64/userspace_msr_exit_test.c | 8 +- .../selftests/kvm/x86_64/xapic_ipi_test.c | 2 +- 13 files changed, 611 insertions(+), 35 deletions(-)create mode 100644
tools/testing/selftests/kvm/aarch64/debug-exceptions.ccreate mode 100644tools/testing/selftests/kvm/lib/aarch64/handlers.S
Marc Zyngier
maintainerQuoted 12 lines · Paolo Bonzini
Quoted 10 lines · Ricardo Koller
Hi,
These patches add a debug exception test in aarch64 KVM selftests while also adding basic exception handling support.
The structure of the exception handling is based on its x86 counterpart. Tests use the same calls to initialize exception handling and both architectures allow tests to override the handler for a particular vector, or (vector, ec) for synchronous exceptions in the arm64 case.
Quoted 8 lines
Quoted 6 lines
The debug test is similar to x86_64/debug_regs, except that the x86 one controls the debugging from outside the VM. This proposed arm64 test controls and handles debug exceptions from the inside.
Thanks,
RicardoMarc, are you going to queue this in your tree?
That is my plan. I'm waiting for the fixes branch to be merged in Linus' tree to start queuing stuff on top.
Thanks,
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.