KVM: selftests: Rename vm_handle_exception in evmcs test
PATCH 1/1 KVM: selftests: Rename vm_handle_exception in evmcs test
Kernel test robot reports this:
Quoted 3 lines
/usr/bin/ld:tools/testing/selftests/kvm/x86_64/evmcs_test.c:157: undefined reference tovm_handle_exception' /usr/bin/ld: tools/testing/selftests/kvm/x86_64/evmcs_test.c:158: undefined reference tovm_handle_exception'
collect2: error: ld returned 1 exit status
Fix it by renaming vm_handle_exception to vm_install_vector_handler in evmcs_test.c.
--- tools/testing/selftests/kvm/x86_64/evmcs_test.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
tools/testing/selftests/kvm/x86_64/evmcs_test.c +2 -2
@@ -154,8 +154,8 @@ int main(int argc, char *argv[])vm_init_descriptor_tables (vm);vcpu_init_descriptor_tables (vm,VCPU_ID ); -vm_handle_exception (vm,UD_VECTOR , guest_ud_handler); -vm_handle_exception (vm,NMI_VECTOR , guest_nmi_handler); +vm_install_vector_handler (vm,UD_VECTOR , guest_ud_handler); +vm_install_vector_handler (vm,NMI_VECTOR , guest_nmi_handler);pr_info ("Running L1 which uses EVMCS to run L2\n" );
Signature · 1 line
2.32.0.rc1.229.g3e70b5a671-goog
Sean Christopherson
reviewerQuoted 10 lines · Ricardo Koller
Kernel test robot reports this:
Quoted 3 lines
/usr/bin/ld:tools/testing/selftests/kvm/x86_64/evmcs_test.c:157: undefined reference tovm_handle_exception' /usr/bin/ld: tools/testing/selftests/kvm/x86_64/evmcs_test.c:158: undefined reference tovm_handle_exception'
collect2: error: ld returned 1 exit statusFix it by renaming
vm_handle_exceptiontovm_install_vector_handlerinevmcs_test.c.Fixes a2bad6a990a4 ("KVM: selftests: Rename vm_handle_exception")
Belated code review... Can we rename the helper to vm_install_exception_handler()?
In x86, "vector" is the number of the exception and "vectoring" is the process of determining the resulting vector that gets delivered to software (e.g. when dealing with contributory faults like #GP->#PF->#DF), but the thing that's being handled is an exception.
arm appears to have similar terminology. And looking at the arm code, it's very confusing to have a helper vm_install_vector_handler() install into exception_handlers, _not_ into vector_handlers. Calling the vector_handlers "default" handlers is also confusing, as "default" usually implies the thing can be overwritten. But in this case, the "default" handler is just another layer in the routing.
The multiple layers of routing is also confusing and a bit hard to wade through for the uninitiated. The whole thing can be made more straightfoward by doing away with the intermediate routing, whacking ~50 lines of code in the process. E.g. (definitely not functional code):
tools/testing/selftests/kvm/aarch64/debug-exceptions.c +5 -5
@@ -212,15 +212,15 @@ int main(int argc, char *argv[])exit (KSFT_SKIP ); } -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_BRK_INS , guest_sw_bp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_HW_BP_CURRENT , guest_hw_bp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_WP_CURRENT , guest_wp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_SSTEP_CURRENT , guest_ss_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_SVC64 , guest_svc_handler);for (stage =0 ; stage <7 ; stage++) {
tools/testing/selftests/kvm/include/aarch64/processor.h +4 -4
@@ -110,10 +110,10 @@ 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); +void vm_install_exception_handler_ec (struct kvm_vm *vm,int vector,int ec, +handler_fn handler); +void vm_install_exception_handler (struct kvm_vm *vm,int vector, +handler_fn handler);#define write_sysreg (reg, val) \ ({ \
tools/testing/selftests/kvm/lib/aarch64/handlers.S +2 -1
@@ -93,7 +93,8 @@ handler_\label: .balign 0x80 /* This will abort so no need to save and restore registers. */ mov x0, #vector - b kvm_exit_unexpected_vector + <sean doesn't know what goes here> + b kvm_exit_unexpected_exception .popsection .set vector, vector + 1
tools/testing/selftests/kvm/lib/aarch64/processor.c +38 -80
152 lines not on this page · raw message
Quoted 7 lines · Sean Christopherson
Quoted 6 lines · Ricardo Koller
Kernel test robot reports this:
Quoted 3 lines
/usr/bin/ld:tools/testing/selftests/kvm/x86_64/evmcs_test.c:157: undefined reference tovm_handle_exception' /usr/bin/ld: tools/testing/selftests/kvm/x86_64/evmcs_test.c:158: undefined reference tovm_handle_exception'
collect2: error: ld returned 1 exit status
Quoted 6 lines
Quoted 4 lines
Fix it by renaming
vm_handle_exceptiontovm_install_vector_handlerinevmcs_test.c.Fixes a2bad6a990a4 ("KVM: selftests: Rename vm_handle_exception")Belated code review...
Thanks for the review.
Quoted 6 lines
Can we rename the helper to
vm_install_exception_handler()?In x86, "vector" is the number of the exception and "vectoring" is the process of determining the resulting vector that gets delivered to software (e.g. when dealing with contributory faults like #GP->#PF->#DF), but the thing that's being handled is an exception.
Got it. What about this renaming:
vm_handle_exception(vec) -> vm_install_exception_handler(vec)
vm_install_exception_handler(vec, ec) -> vm_install_sync_handler(vec, ec)
Quoted 8 lines
arm appears to have similar terminology. And looking at the arm code, it's very confusing to have a helper
vm_install_vector_handler()install intoexception_handlers, _not_ intovector_handlers. Calling thevector_handlers"default" handlers is also confusing, as "default" usually implies the thing can be overwritten. But in this case, the "default" handler is just another layer in the routing.
Quoted 4 lines
The multiple layers of routing is also confusing and a bit hard to wade through for the uninitiated. The whole thing can be made more straightfoward by doing away with the intermediate routing, whacking ~50 lines of code in the process. E.g. (definitely not functional code):
I think that works and it does remove a bunch of code. Just need to play with the idea and check that it can cover all cases.
For now, given that the build is broken, what about this series of patches:
1. keep this patch to fix x86 kvm selftests
2. rename both arm and x86 to vm_install_exception_handler and vm_install_sync_handler
3. restructure the internals of exception handling in arm
Alternatively, I can send 1+2 together and then 3. What do you think?
Thanks,
Ricardo
Quoted 208 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c +5
-5@@ -212,15 +212,15 @@ int main(int argc, char *argv[])exit (KSFT_SKIP ); } -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_BRK_INS , guest_sw_bp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_HW_BP_CURRENT , guest_hw_bp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_WP_CURRENT , guest_wp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_SSTEP_CURRENT , guest_ss_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_SVC64 , guest_svc_handler);for (stage =0 ; stage <7 ; stage++) {tools/testing/selftests/kvm/include/aarch64/processor.h +4
-4@@ -110,10 +110,10 @@ 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); +void vm_install_exception_handler_ec (struct kvm_vm *vm,int vector,int ec, +handler_fn handler); +void vm_install_exception_handler (struct kvm_vm *vm,int vector, +handler_fn handler);#define write_sysreg (reg, val) \ ({ \tools/testing/selftests/kvm/lib/aarch64/handlers.S +2
-1@@ -93,7 +93,8 @@ handler_\label: .balign 0x80 /* This will abort so no need to save and restore registers. */ mov x0, #vector - b kvm_exit_unexpected_vector + <sean doesn't know what goes here> + b kvm_exit_unexpected_exception .popsection .set vector, vector + 1tools/testing/selftests/kvm/lib/aarch64/processor.c +35
-75@@ -337,16 +337,9 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } -void kvm_exit_unexpected_vector (int vector) +void kvm_exit_unexpected_exception (int vector,uint64_t ec,bool valid_ec) { -ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); -while (1 ) - ; -} - -static void kvm_exit_unexpected_exception (int vector,uint64_t ec) -{ -ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +ucall (UCALL_UNHANDLED ,3 , vector, ec, valid_ec);while (1 ) ; }@@ -369,18 +362,7 @@ void assert_on_unhandled_exception(struct kvm_vm *vm, uint32_t vcpuid) } } -/* -* 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 ]; };@@ -391,80 +373,56 @@ void vcpu_init_descriptor_tables(struct kvm_vm *vm, uint32_t vcpuid)set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); } -static 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); -} - -static void default_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; +struct handler_fn *handlers = exception_handlers; +bool valid_ec; +int ec; -if (handlers && handlers->vector_handlers[vector]) - handlers->vector_handlers[vector](regs, vector); -else -kvm_exit_unexpected_vector (vector); +switch (vector) { +case VECTOR_SYNC_CURRENT : +case VECTOR_SYNC_LOWER_64 : + ec = (read_sysreg (esr_el1) >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + valid_ec = true; +break ; +case VECTOR_IRQ_CURRENT : +case VECTOR_IRQ_LOWER_64 : +case VECTOR_FIQ_CURRENT : +case VECTOR_FIQ_LOWER_64 : +case VECTOR_ERROR_CURRENT : +case VECTOR_ERROR_LOWER_64 : + ec =0 ; + valid_ec = false; +break ; +default : + goto unexpected_exception; + } + +if (handlers && handlers[vector][ec]) +return handlers[vector][ec](regs); + +unexpected_exception: +kvm_exit_unexpected_exception (vector, ec, valid_ec); }void vm_init_descriptor_tables (struct kvm_vm *vm) { -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_handler; - handlers->vector_handlers[VECTOR_FIQ_CURRENT ] = default_handler; - handlers->vector_handlers[VECTOR_ERROR_CURRENT ] = default_handler; - - handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; - handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_handler; - handlers->vector_handlers[VECTOR_FIQ_LOWER_64 ] = default_handler; - handlers->vector_handlers[VECTOR_ERROR_LOWER_64 ] = default_handler; - - *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = __exception_handlers; } -void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, -void (*handler)(struct ex_regs *)) +void vm_install_exception_handler_ec (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) { -struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); +struct handlers *handlers =addr_gva2hva (vm, vm->handlers); -assert (VECTOR_IS_SYNC (vector)); +assert (!ec == !VECTOR_IS_SYNC (vector));assert (vector <VECTOR_NUM );assert (ec <ESR_EC_NUM ); - handlers->exception_handlers[vector][ec] = handler; + exception_handlers[vector][ec] = handler; } -void vm_install_vector_handler (struct kvm_vm *vm,int vector, -void (*handler)(struct ex_regs *))
tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_init_descriptor_tables(struct kvm_vm *vm, uint32_t vcpuid) · line 428
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +3
-5+void vm_install_exception_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; +vm_install_exception_handler_ec (vm, vector,0 , handler); }
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerQuoted 49 lines · Ricardo Koller
Quoted 7 lines · Sean Christopherson
Quoted 6 lines · Ricardo Koller
Kernel test robot reports this:
Quoted 3 lines
/usr/bin/ld:tools/testing/selftests/kvm/x86_64/evmcs_test.c:157: undefined reference tovm_handle_exception' /usr/bin/ld: tools/testing/selftests/kvm/x86_64/evmcs_test.c:158: undefined reference tovm_handle_exception'
collect2: error: ld returned 1 exit statusQuoted 6 lines
Quoted 4 lines
Fix it by renaming
vm_handle_exceptiontovm_install_vector_handlerinevmcs_test.c.Fixes a2bad6a990a4 ("KVM: selftests: Rename vm_handle_exception")Belated code review...
Thanks for the review.
Quoted 6 lines
Can we rename the helper to
vm_install_exception_handler()?In x86, "vector" is the number of the exception and "vectoring" is the process of determining the resulting vector that gets delivered to software (e.g. when dealing with contributory faults like #GP->#PF->#DF), but the thing that's being handled is an exception.
Got it. What about this renaming:
vm_handle_exception(vec)->vm_install_exception_handler(vec)
vm_install_exception_handler(vec, ec)-> vm_install_sync_handler(vec,
ec)Quoted 8 lines
arm appears to have similar terminology. And looking at the arm code, it's very confusing to have a helper
vm_install_vector_handler()install intoexception_handlers, _not_ intovector_handlers. Calling thevector_handlers"default" handlers is also confusing, as "default" usually implies the thing can be overwritten. But in this case, the "default" handler is just another layer in the routing.Quoted 4 lines
The multiple layers of routing is also confusing and a bit hard to wade through for the uninitiated. The whole thing can be made more straightfoward by doing away with the intermediate routing, whacking ~50 lines of code in the process. E.g. (definitely not functional code):
I think that works and it does remove a bunch of code. Just need to play with the idea and check that it can cover all cases.
For now, given that the build is broken, what about this series of patches:
Quoted 6 lines
1. keep this patch to fix x86 kvm selftests
2. rename both arm and x86 tovm_install_exception_handlerandvm_install_sync_handler
3. restructure the internals of exception handling in armAlternatively, I can send 1+2 together and then 3. What do you think?
This is becoming a bit messy. I'd rather drop the whole series from -next, and get something that doesn't break in the middle. Please resend the series tested on top of -rc4.
Thanks,
M.
Signature · 1 line
Jazz is not dead. It just smells funny...
Sean Christopherson
reviewerQuoted 3 lines · Marc Zyngier
This is becoming a bit messy. I'd rather drop the whole series from -next, and get something that doesn't break in the middle. Please resend the series tested on top of -rc4.
That'd be my preference too. I almost asked if it could be (temporarily) dropped, but I assumed the hashes in -next are intended to be stable.
Thanks!
Marc Zyngier
maintainerQuoted 8 lines · Sean Christopherson
Quoted 3 lines · Marc Zyngier
This is becoming a bit messy. I'd rather drop the whole series from -next, and get something that doesn't break in the middle. Please resend the series tested on top of -rc4.
That'd be my preference too. I almost asked if it could be (temporarily) dropped, but I assumed the hashes in -next are intended to be stable.
I usually try and keep these commits stable.
But in this case, we end-up with a broken build at some point in the series. For such cases, I'm more than happy to drop the series and merge a clean version again (I keep each series on a separate branch for this exact purpose).
Thanks,
M.
Signature · 5 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kvmarm mailing list [email protected] https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · Sean Christopherson
Quoted 3 lines · Marc Zyngier
This is becoming a bit messy. I'd rather drop the whole series from -next, and get something that doesn't break in the middle. Please resend the series tested on top of -rc4.
That'd be my preference too. I almost asked if it could be (temporarily) dropped, but I assumed the hashes in -next are intended to be stable.
Quoted 6 lines
I usually try and keep these commits stable.
But in this case, we end-up with a broken build at some point in the series. For such cases, I'm more than happy to drop the series and merge a clean version again (I keep each series on a separate branch for this exact purpose).
Thank you both. I will send a new series with the fix and Seans suggestions (tested on the latest rc).
Apologies again for the mess.
Ricardo
Quoted 7 lines
Thanks,
M.Signature · 1 line
Without deviation from the norm, progress is not possible.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Marc Zyngier
maintainerQuoted 20 lines · Ricardo Koller
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · Sean Christopherson
Quoted 3 lines · Marc Zyngier
This is becoming a bit messy. I'd rather drop the whole series from -next, and get something that doesn't break in the middle. Please resend the series tested on top of -rc4.
That'd be my preference too. I almost asked if it could be (temporarily) dropped, but I assumed the hashes in -next are intended to be stable.
Quoted 6 lines
I usually try and keep these commits stable.
But in this case, we end-up with a broken build at some point in the series. For such cases, I'm more than happy to drop the series and merge a clean version again (I keep each series on a separate branch for this exact purpose).
Quoted 4 lines
Thank you both. I will send a new series with the fix and Seans suggestions (tested on the latest rc).
Apologies again for the mess.
No worries, that's a pretty common situation.
M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
Quoted 19 lines · Sean Christopherson
Quoted 10 lines · Ricardo Koller
Kernel test robot reports this:
Quoted 3 lines
/usr/bin/ld:tools/testing/selftests/kvm/x86_64/evmcs_test.c:157: undefined reference tovm_handle_exception' /usr/bin/ld: tools/testing/selftests/kvm/x86_64/evmcs_test.c:158: undefined reference tovm_handle_exception'
collect2: error: ld returned 1 exit statusFix it by renaming
vm_handle_exceptiontovm_install_vector_handlerinevmcs_test.c.Fixes a2bad6a990a4 ("KVM: selftests: Rename vm_handle_exception")Belated code review... Can we rename the helper to
vm_install_exception_handler()?In x86, "vector" is the number of the exception and "vectoring" is the process of determining the resulting vector that gets delivered to software (e.g. when dealing with contributory faults like #GP->#PF->#DF), but the thing that's being handled is an exception.
Quoted 6 lines
arm appears to have similar terminology. And looking at the arm code, it's very confusing to have a helper
vm_install_vector_handler()install intoexception_handlers, _not_ intovector_handlers. Calling thevector_handlers"default" handlers is also confusing, as "default" usually implies the thing can be overwritten. But in this case, the "default" handler is just another layer in the routing.
FWIW, this terminology makes sense in kvm-unit-tests (KUT) because KUT has a library function to update the default entries in vector_handlers. I based my patch on it (with names and everything) but didn't add this function to update the default entries, hence the confusion.
Quoted 5 lines
The multiple layers of routing is also confusing and a bit hard to wade through for the uninitiated. The whole thing can be made more straightfoward by doing away with the intermediate routing, whacking ~50 lines of code in the process. E.g. (definitely not functional code):
This works but it would remove the ability to replace the default sync handler with something else, like a handler that can cover all possible ec values. In this case we would have to call vm_install_exception_handler_ec 64 times. On the other hand, the tests that we are planning don't seem to need it, so I will move on with the suggestion.
Thanks,
Ricardo
Quoted 208 lines
tools/testing/selftests/kvm/aarch64/debug-exceptions.c +5
-5@@ -212,15 +212,15 @@ int main(int argc, char *argv[])exit (KSFT_SKIP ); } -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_BRK_INS , guest_sw_bp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_HW_BP_CURRENT , guest_hw_bp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_WP_CURRENT , guest_wp_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_SSTEP_CURRENT , guest_ss_handler); -vm_install_exception_handler (vm,VECTOR_SYNC_CURRENT , +vm_install_exception_handler_ec (vm,VECTOR_SYNC_CURRENT ,ESR_EC_SVC64 , guest_svc_handler);for (stage =0 ; stage <7 ; stage++) {tools/testing/selftests/kvm/include/aarch64/processor.h +4
-4@@ -110,10 +110,10 @@ 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); +void vm_install_exception_handler_ec (struct kvm_vm *vm,int vector,int ec, +handler_fn handler); +void vm_install_exception_handler (struct kvm_vm *vm,int vector, +handler_fn handler);#define write_sysreg (reg, val) \ ({ \tools/testing/selftests/kvm/lib/aarch64/handlers.S +2
-1@@ -93,7 +93,8 @@ handler_\label: .balign 0x80 /* This will abort so no need to save and restore registers. */ mov x0, #vector - b kvm_exit_unexpected_vector + <sean doesn't know what goes here> + b kvm_exit_unexpected_exception .popsection .set vector, vector + 1tools/testing/selftests/kvm/lib/aarch64/processor.c +35
-75@@ -337,16 +337,9 @@ void vcpu_args_set(struct kvm_vm *vm, uint32_t vcpuid, unsigned int num, ...)va_end (ap); } -void kvm_exit_unexpected_vector (int vector) +void kvm_exit_unexpected_exception (int vector,uint64_t ec,bool valid_ec) { -ucall (UCALL_UNHANDLED ,3 , vector,0 , false/* !valid_ec */ ); -while (1 ) - ; -} - -static void kvm_exit_unexpected_exception (int vector,uint64_t ec) -{ -ucall (UCALL_UNHANDLED ,3 , vector, ec, true/* valid_ec */ ); +ucall (UCALL_UNHANDLED ,3 , vector, ec, valid_ec);while (1 ) ; }@@ -369,18 +362,7 @@ void assert_on_unhandled_exception(struct kvm_vm *vm, uint32_t vcpuid) } } -/* -* 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 ]; };@@ -391,80 +373,56 @@ void vcpu_init_descriptor_tables(struct kvm_vm *vm, uint32_t vcpuid)set_reg (vm, vcpuid,ARM64_SYS_REG (VBAR_EL1 ), (uint64_t )&vectors); } -static 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); -} - -static void default_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; +struct handler_fn *handlers = exception_handlers; +bool valid_ec; +int ec; -if (handlers && handlers->vector_handlers[vector]) - handlers->vector_handlers[vector](regs, vector); -else -kvm_exit_unexpected_vector (vector); +switch (vector) { +case VECTOR_SYNC_CURRENT : +case VECTOR_SYNC_LOWER_64 : + ec = (read_sysreg (esr_el1) >>ESR_EC_SHIFT ) &ESR_EC_MASK ; + valid_ec = true; +break ; +case VECTOR_IRQ_CURRENT : +case VECTOR_IRQ_LOWER_64 : +case VECTOR_FIQ_CURRENT : +case VECTOR_FIQ_LOWER_64 : +case VECTOR_ERROR_CURRENT : +case VECTOR_ERROR_LOWER_64 : + ec =0 ; + valid_ec = false; +break ; +default : + goto unexpected_exception; + } + +if (handlers && handlers[vector][ec]) +return handlers[vector][ec](regs); + +unexpected_exception: +kvm_exit_unexpected_exception (vector, ec, valid_ec); }void vm_init_descriptor_tables (struct kvm_vm *vm) { -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_handler; - handlers->vector_handlers[VECTOR_FIQ_CURRENT ] = default_handler; - handlers->vector_handlers[VECTOR_ERROR_CURRENT ] = default_handler; - - handlers->vector_handlers[VECTOR_SYNC_LOWER_64 ] = default_sync_handler; - handlers->vector_handlers[VECTOR_IRQ_LOWER_64 ] = default_handler; - handlers->vector_handlers[VECTOR_FIQ_LOWER_64 ] = default_handler; - handlers->vector_handlers[VECTOR_ERROR_LOWER_64 ] = default_handler; - - *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = vm->handlers; + *(vm_vaddr_t *)addr_gva2hva (vm, (vm_vaddr_t)(&exception_handlers)) = __exception_handlers; } -void vm_install_exception_handler (struct kvm_vm *vm,int vector,int ec, -void (*handler)(struct ex_regs *)) +void vm_install_exception_handler_ec (struct kvm_vm *vm,int vector,int ec, +void (*handler)(struct ex_regs *)) { -struct handlers *handlers = (struct handlers *)addr_gva2hva (vm, vm->handlers); +struct handlers *handlers =addr_gva2hva (vm, vm->handlers); -assert (VECTOR_IS_SYNC (vector)); +assert (!ec == !VECTOR_IS_SYNC (vector));assert (vector <VECTOR_NUM );assert (ec <ESR_EC_NUM ); - handlers->exception_handlers[vector][ec] = handler; + exception_handlers[vector][ec] = handler; } -void vm_install_vector_handler (struct kvm_vm *vm,int vector, -void (*handler)(struct ex_regs *))
tools/testing/selftests/kvm/lib/aarch64/processor.c void vcpu_init_descriptor_tables(struct kvm_vm *vm, uint32_t vcpuid) · line 428
Quoted 10 lines
tools/testing/selftests/kvm/lib/aarch64/processor.c +3
-5+void vm_install_exception_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; +vm_install_exception_handler_ec (vm, vector,0 , handler); }
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
Sean Christopherson
reviewerQuoted 11 lines · Ricardo Koller
Quoted 4 lines · Sean Christopherson
The multiple layers of routing is also confusing and a bit hard to wade through for the uninitiated. The whole thing can be made more straightfoward by doing away with the intermediate routing, whacking ~50 lines of code in the process. E.g. (definitely not functional code):
This works but it would remove the ability to replace the default sync handler with something else, like a handler that can cover all possible ec values. In this case we would have to call
vm_install_exception_handler_ec64 times. On the other hand, the tests that we are planning don't seem to need it, so I will move on with the suggestion.
My objection to layering handlers is that it introduces ambiguity regarding ordering and override functionality, e.g. if a test overrides both the "default" handler and a specific exception handler, which handler will be invoked? My expectation would be that the more specific override would win, but someone else might expect that overriding the default would win.
It should be relatively easy to provide helpers to override the handler for multiple/all exceptions if we do end up with tests that want that functionality. But yeah, definitely a future problem :-)