Rework the guest exit to userspace code to generalize the concept
into what it is, a "hypercall to userspace", and provide two
implementations of it: the PortIO version currently used, but only
useable by x86, and an MMIO version that other architectures (except
s390) can use.
Signed-off-by: Andrew Jones <drjones@xxxxxxxxxx>
---
tools/testing/selftests/kvm/Makefile | 2 +-
.../selftests/kvm/cr4_cpuid_sync_test.c | 12 +-
tools/testing/selftests/kvm/dirty_log_test.c | 5 +-
.../testing/selftests/kvm/include/kvm_util.h | 76 ++++-----
tools/testing/selftests/kvm/lib/kvm_util.c | 14 --
tools/testing/selftests/kvm/lib/ucall.c | 144 ++++++++++++++++++
tools/testing/selftests/kvm/state_test.c | 23 +--
.../selftests/kvm/vmx_tsc_adjust_test.c | 17 +--
8 files changed, 214 insertions(+), 79 deletions(-)
create mode 100644 tools/testing/selftests/kvm/lib/ucall.c
diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
index 87d1a8488af8..0b254dbf6e86 100644
--- a/tools/testing/selftests/kvm/Makefile
+++ b/tools/testing/selftests/kvm/Makefile
@@ -3,7 +3,7 @@ all:
top_srcdir = ../../../../
UNAME_M := $(shell uname -m)
-LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/sparsebit.c
+LIBKVM = lib/assert.c lib/elf.c lib/io.c lib/kvm_util.c lib/ucall.c lib/sparsebit.c
LIBKVM_x86_64 = lib/x86.c lib/vmx.c
TEST_GEN_PROGS_x86_64 = set_sregs_test
diff --git a/tools/testing/selftests/kvm/cr4_cpuid_sync_test.c b/tools/testing/selftests/kvm/cr4_cpuid_sync_test.c
index 11ec358bf969..fd4f419fe9ab 100644
--- a/tools/testing/selftests/kvm/cr4_cpuid_sync_test.c
+++ b/tools/testing/selftests/kvm/cr4_cpuid_sync_test.c
@@ -67,6 +67,7 @@ int main(int argc, char *argv[])
struct kvm_vm *vm;
struct kvm_sregs sregs;
struct kvm_cpuid_entry2 *entry;
+ struct ucall uc;
int rc;
entry = kvm_get_supported_cpuid_entry(1);
@@ -87,21 +88,20 @@ int main(int argc, char *argv[])
rc = _vcpu_run(vm, VCPU_ID);
if (run->exit_reason == KVM_EXIT_IO) {
- switch (run->io.port) {
- case GUEST_PORT_SYNC:
+ switch (get_ucall(vm, VCPU_ID, &uc)) {
+ case UCALL_SYNC:
/* emulate hypervisor clearing CR4.OSXSAVE */
vcpu_sregs_get(vm, VCPU_ID, &sregs);
sregs.cr4 &= ~X86_CR4_OSXSAVE;
vcpu_sregs_set(vm, VCPU_ID, &sregs);
break;
- case GUEST_PORT_ABORT:
+ case UCALL_ABORT:
TEST_ASSERT(false, "Guest CR4 bit (OSXSAVE) unsynchronized with CPUID bit.");
break;
- case GUEST_PORT_DONE:
+ case UCALL_DONE:
goto done;
default:
- TEST_ASSERT(false, "Unknown port 0x%x.",
- run->io.port);
+ TEST_ASSERT(false, "Unknown ucall 0x%x.", uc.cmd);
}
}
}
diff --git a/tools/testing/selftests/kvm/dirty_log_test.c b/tools/testing/selftests/kvm/dirty_log_test.c
index 0c2cdc105f96..7cf3e4ae6046 100644
--- a/tools/testing/selftests/kvm/dirty_log_test.c
+++ b/tools/testing/selftests/kvm/dirty_log_test.c
@@ -110,7 +110,7 @@ void *vcpu_worker(void *data)
uint64_t loops, *guest_array, pages_count = 0;
struct kvm_vm *vm = data;
struct kvm_run *run;
- struct guest_args args;
+ struct ucall uc;
run = vcpu_state(vm, VCPU_ID);
@@ -124,9 +124,8 @@ void *vcpu_worker(void *data)
while (!READ_ONCE(host_quit)) {
/* Let the guest to dirty these random pages */
ret = _vcpu_run(vm, VCPU_ID);
- guest_args_read(vm, VCPU_ID, &args);
if (run->exit_reason == KVM_EXIT_IO &&
- args.port == GUEST_PORT_SYNC) {
+ get_ucall(vm, VCPU_ID, &uc) == UCALL_SYNC) {
pages_count += TEST_PAGES_PER_LOOP;
generate_random_array(guest_array, TEST_PAGES_PER_LOOP);
} else {
diff --git a/tools/testing/selftests/kvm/include/kvm_util.h b/tools/testing/selftests/kvm/include/kvm_util.h
index bb5a25fb82c6..138289f7d489 100644
--- a/tools/testing/selftests/kvm/include/kvm_util.h
+++ b/tools/testing/selftests/kvm/include/kvm_util.h
@@ -148,43 +148,49 @@ allocate_kvm_dirty_log(struct kvm_userspace_memory_region *region);
int vm_create_device(struct kvm_vm *vm, struct kvm_create_device *cd);
-#define GUEST_PORT_SYNC 0x1000
-#define GUEST_PORT_ABORT 0x1001
-#define GUEST_PORT_DONE 0x1002
-
-static inline void __exit_to_l0(uint16_t port, uint64_t arg0, uint64_t arg1)
-{
- __asm__ __volatile__("in %[port], %%al"
- :
- : [port]"d"(port), "D"(arg0), "S"(arg1)
- : "rax");
-}
-
-/*
- * Allows to pass three arguments to the host: port is 16bit wide,
- * arg0 & arg1 are 64bit wide
- */
-#define GUEST_SYNC_ARGS(_port, _arg0, _arg1) \
- __exit_to_l0(_port, (uint64_t) (_arg0), (uint64_t) (_arg1))
-
-#define GUEST_ASSERT(_condition) do { \
- if (!(_condition)) \
- GUEST_SYNC_ARGS(GUEST_PORT_ABORT, \
- "Failed guest assert: " \
- #_condition, __LINE__); \
- } while (0)
-
-#define GUEST_SYNC(stage) GUEST_SYNC_ARGS(GUEST_PORT_SYNC, "hello", stage)
+#define sync_global_to_guest(vm, g) ({ \
+ typeof(g) *_p = addr_gva2hva(vm, (vm_vaddr_t)&(g)); \
+ memcpy(_p, &(g), sizeof(g)); \
+})
+
+#define sync_global_from_guest(vm, g) ({ \
+ typeof(g) *_p = addr_gva2hva(vm, (vm_vaddr_t)&(g)); \
+ memcpy(&(g), _p, sizeof(g)); \
+})
+
+/* ucall implementation types */
+typedef enum {
+ UCALL_PIO,
+ UCALL_MMIO,
+} ucall_type_t;
+
+/* Common ucalls */
+enum {
+ UCALL_NONE,
+ UCALL_SYNC,
+ UCALL_ABORT,
+ UCALL_DONE,
+};
-#define GUEST_DONE() GUEST_SYNC_ARGS(GUEST_PORT_DONE, 0, 0)
+#define UCALL_MAX_ARGS 6
-struct guest_args {
- uint64_t arg0;
- uint64_t arg1;
- uint16_t port;
-} __attribute__ ((packed));
+struct ucall {
+ uint64_t cmd;
+ uint64_t args[UCALL_MAX_ARGS];
+};
-void guest_args_read(struct kvm_vm *vm, uint32_t vcpu_id,
- struct guest_args *args);
+void ucall_init(struct kvm_vm *vm, ucall_type_t type, void *arg);
+void ucall_uninit(struct kvm_vm *vm);
+void ucall(uint64_t cmd, int nargs, ...);
+uint64_t get_ucall(struct kvm_vm *vm, uint32_t vcpu_id, struct ucall *uc);
+
+#define GUEST_SYNC(stage) ucall(UCALL_SYNC, 2, "hello", stage)
+#define GUEST_DONE() ucall(UCALL_DONE, 0)
+#define GUEST_ASSERT(_condition) do { \
+ if (!(_condition)) \
+ ucall(UCALL_ABORT, 2, \
+ "Failed guest assert: " \
+ #_condition, __LINE__); \
+} while (0)
#endif /* SELFTEST_KVM_UTIL_H */
diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index e9ba389c48db..107622ce2b8c 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -1580,17 +1580,3 @@ void *addr_gva2hva(struct kvm_vm *vm, vm_vaddr_t gva)
{
return addr_gpa2hva(vm, addr_gva2gpa(vm, gva));
}
-
-void guest_args_read(struct kvm_vm *vm, uint32_t vcpu_id,
- struct guest_args *args)
-{
- struct kvm_run *run = vcpu_state(vm, vcpu_id);
- struct kvm_regs regs;
-
- memset(®s, 0, sizeof(regs));
- vcpu_regs_get(vm, vcpu_id, ®s);
-
- args->port = run->io.port;
- args->arg0 = regs.rdi;
- args->arg1 = regs.rsi;
-}
diff --git a/tools/testing/selftests/kvm/lib/ucall.c b/tools/testing/selftests/kvm/lib/ucall.c
new file mode 100644
index 000000000000..4777f9bb5194
--- /dev/null
+++ b/tools/testing/selftests/kvm/lib/ucall.c
@@ -0,0 +1,144 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * ucall support. A ucall is a "hypercall to userspace".
+ *
+ * Copyright (C) 2018, Red Hat, Inc.
+ */
+#include "kvm_util.h"
+#include "kvm_util_internal.h"
+
+#define UCALL_PIO_PORT ((uint16_t)0x1000)
+
+static ucall_type_t ucall_type;
+static vm_vaddr_t *ucall_exit_mmio_addr;
+
+static bool ucall_mmio_init(struct kvm_vm *vm, vm_paddr_t gpa)
+{
+ if (kvm_userspace_memory_region_find(vm, gpa, gpa + 1))
+ return false;
+
+ virt_pg_map(vm, gpa, gpa, 0);
+
+ ucall_exit_mmio_addr = (vm_vaddr_t *)gpa;
+ sync_global_to_guest(vm, ucall_exit_mmio_addr);
+
+ return true;
+}
+
+void ucall_init(struct kvm_vm *vm, ucall_type_t type, void *arg)
+{
+ ucall_type = type;
+ sync_global_to_guest(vm, ucall_type);
+
+ if (type == UCALL_PIO)
+ return;
+
+ if (type == UCALL_MMIO) {
+ vm_paddr_t gpa, start, end, step;
+ bool ret;
+
+ if (arg) {
+ gpa = (vm_paddr_t)arg;
+ ret = ucall_mmio_init(vm, gpa);
+ TEST_ASSERT(ret, "Can't set ucall mmio address to %lx", gpa);
+ return;
+ }
+
+ /*
+ * Find an address within the allowed virtual address space,
+ * that does _not_ have a KVM memory region associated with it.
+ * Identity mapping an address like this allows the guest to
+ * access it, but as KVM doesn't know what to do with it, it
+ * will assume it's something userspace handles and exit with
+ * KVM_EXIT_MMIO. Well, at least that's how it works for AArch64.
+ * Here we start with a guess that the addresses around two
+ * thirds of the VA space are unmapped and then work both down
+ * and up from there in 1/6 VA space sized steps.
+ */
+ start = 1ul << (vm->va_bits * 2 / 3);
+ end = 1ul << vm->va_bits;
+ step = 1ul << (vm->va_bits / 6);
+ for (gpa = start; gpa >= 0; gpa -= step) {
+ if (ucall_mmio_init(vm, gpa & ~(vm->page_size - 1)))
+ return;
+ }
+ for (gpa = start + step; gpa < end; gpa += step) {
+ if (ucall_mmio_init(vm, gpa & ~(vm->page_size - 1)))
+ return;
+ }
+ TEST_ASSERT(false, "Can't find a ucall mmio address");
+ }
+}
+
+void ucall_uninit(struct kvm_vm *vm)
+{
+ ucall_type = 0;
+ sync_global_to_guest(vm, ucall_type);
+ ucall_exit_mmio_addr = 0;
+ sync_global_to_guest(vm, ucall_exit_mmio_addr);
+}
+
+static void ucall_pio_exit(struct ucall *uc)
+{
+#ifdef __x86_64__
+ asm volatile("in %[port], %%al"
+ : : [port] "d" (UCALL_PIO_PORT), "D" (uc) : "rax");
+#endif
+}
+
+static void ucall_mmio_exit(struct ucall *uc)
+{
+ *ucall_exit_mmio_addr = (vm_vaddr_t)uc;
+}
+
+void ucall(uint64_t cmd, int nargs, ...)
+{
+ struct ucall uc = {
+ .cmd = cmd,
+ };
+ va_list va;
+ int i;
+
+ nargs = nargs <= UCALL_MAX_ARGS ? nargs : UCALL_MAX_ARGS;
+
+ va_start(va, nargs);
+ for (i = 0; i < nargs; ++i)
+ uc.args[i] = va_arg(va, uint64_t);
+ va_end(va);
+
+ switch (ucall_type) {
+ case UCALL_PIO:
+ ucall_pio_exit(&uc);
+ break;
+ case UCALL_MMIO:
+ ucall_mmio_exit(&uc);
+ break;
+ };
+}
+
+uint64_t get_ucall(struct kvm_vm *vm, uint32_t vcpu_id, struct ucall *uc)
+{
+ struct kvm_run *run = vcpu_state(vm, vcpu_id);
+
+ memset(uc, 0, sizeof(*uc));
+
+#ifdef __x86_64__
+ if (ucall_type == UCALL_PIO && run->exit_reason == KVM_EXIT_IO &&
+ run->io.port == UCALL_PIO_PORT) {
+ struct kvm_regs regs;
+ vcpu_regs_get(vm, vcpu_id, ®s);
+ memcpy(uc, addr_gva2hva(vm, (vm_vaddr_t)regs.rdi), sizeof(*uc));
+ return uc->cmd;
+ }
+#endif
+ if (ucall_type == UCALL_MMIO && run->exit_reason == KVM_EXIT_MMIO &&
+ run->mmio.phys_addr == (uint64_t)ucall_exit_mmio_addr) {
+ vm_vaddr_t gva;
+ TEST_ASSERT(run->mmio.is_write && run->mmio.len == 8,
+ "Unexpected ucall exit mmio address access");
+ gva = *(vm_vaddr_t *)run->mmio.data;
+ memcpy(uc, addr_gva2hva(vm, gva), sizeof(*uc));
+ }
+
+ return uc->cmd;
+}
diff --git a/tools/testing/selftests/kvm/state_test.c b/tools/testing/selftests/kvm/state_test.c
index 900e3e9dfb9f..cdf82735d6e7 100644
--- a/tools/testing/selftests/kvm/state_test.c
+++ b/tools/testing/selftests/kvm/state_test.c
@@ -127,6 +127,7 @@ int main(int argc, char *argv[])
struct kvm_vm *vm;
struct kvm_run *run;
struct kvm_x86_state *state;
+ struct ucall uc;
int stage;
struct kvm_cpuid_entry2 *entry = kvm_get_supported_cpuid_entry(1);
@@ -155,23 +156,23 @@ int main(int argc, char *argv[])
memset(®s1, 0, sizeof(regs1));
vcpu_regs_get(vm, VCPU_ID, ®s1);
- switch (run->io.port) {
- case GUEST_PORT_ABORT:
- TEST_ASSERT(false, "%s at %s:%d", (const char *) regs1.rdi,
- __FILE__, regs1.rsi);
+ switch (get_ucall(vm, VCPU_ID, &uc)) {
+ case UCALL_ABORT:
+ TEST_ASSERT(false, "%s at %s:%d", (const char *)uc.args[0],
+ __FILE__, uc.args[1]);
/* NOT REACHED */
- case GUEST_PORT_SYNC:
+ case UCALL_SYNC:
break;
- case GUEST_PORT_DONE:
+ case UCALL_DONE:
goto done;
default:
- TEST_ASSERT(false, "Unknown port 0x%x.", run->io.port);
+ TEST_ASSERT(false, "Unknown ucall 0x%x.", uc.cmd);
}
- /* PORT_SYNC is handled here. */
- TEST_ASSERT(!strcmp((const char *)regs1.rdi, "hello") &&
- regs1.rsi == stage, "Unexpected register values vmexit #%lx, got %lx",
- stage, (ulong) regs1.rsi);
+ /* UCALL_SYNC is handled here. */
+ TEST_ASSERT(!strcmp((const char *)uc.args[0], "hello") &&
+ uc.args[1] == stage, "Unexpected register values vmexit #%lx, got %lx",
+ stage, (ulong)uc.args[1]);
state = vcpu_save_state(vm, VCPU_ID);
kvm_vm_release(vm);
diff --git a/tools/testing/selftests/kvm/vmx_tsc_adjust_test.c b/tools/testing/selftests/kvm/vmx_tsc_adjust_test.c
index 49bcc68b0235..8d487c78796a 100644
--- a/tools/testing/selftests/kvm/vmx_tsc_adjust_test.c
+++ b/tools/testing/selftests/kvm/vmx_tsc_adjust_test.c
@@ -146,26 +146,25 @@ int main(int argc, char *argv[])
for (;;) {
volatile struct kvm_run *run = vcpu_state(vm, VCPU_ID);
- struct guest_args args;
+ struct ucall uc;
vcpu_run(vm, VCPU_ID);
- guest_args_read(vm, VCPU_ID, &args);
TEST_ASSERT(run->exit_reason == KVM_EXIT_IO,
"Got exit_reason other than KVM_EXIT_IO: %u (%s)\n",
run->exit_reason,
exit_reason_str(run->exit_reason));
- switch (args.port) {
- case GUEST_PORT_ABORT:
- TEST_ASSERT(false, "%s", (const char *) args.arg0);
+ switch (get_ucall(vm, VCPU_ID, &uc)) {
+ case UCALL_ABORT:
+ TEST_ASSERT(false, "%s", (const char *)uc.args[0]);
/* NOT REACHED */
- case GUEST_PORT_SYNC:
- report(args.arg1);
+ case UCALL_SYNC:
+ report(uc.args[1]);
break;
- case GUEST_PORT_DONE:
+ case UCALL_DONE:
goto done;
default:
- TEST_ASSERT(false, "Unknown port 0x%x.", args.port);
+ TEST_ASSERT(false, "Unknown ucall 0x%x.", uc.cmd);
}
}
--
2.17.1
