Quellcodebibliothek Statistik Leitseite products/Sources/formale Sprachen/C/Linux/drivers/char/hw_random/   (Open Source Betriebssystem Version 6.17.9©)  Datei vom 24.10.2025 mit Größe 2 kB image not shown  

SSL main.c   Interaktion und
Portierbarkeitunbekannt

 
// SPDX-License-Identifier: GPL-2.0
/*  Copyright(c) 2016-20 Intel Corporation. */

#include <cpuid.h>
#include <elf.h>
#include <errno.h>
#include <fcntl.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <sys/time.h>
#include <sys/types.h>
#include <sys/auxv.h>
#include "defines.h"
#include "../kselftest_harness.h"
#include "main.h"

static const uint64_t MAGIC = 0x1122334455667788ULL;
static const uint64_t MAGIC2 = 0x8877665544332211ULL;
vdso_sgx_enter_enclave_t vdso_sgx_enter_enclave;

/*
 * Security Information (SECINFO) data structure needed by a few SGX
 * instructions (eg. ENCLU[EACCEPT] and ENCLU[EMODPE]) holds meta-data
 * about an enclave page. &enum sgx_secinfo_page_state specifies the
 * secinfo flags used for page state.
 */

enum sgx_secinfo_page_state {
 SGX_SECINFO_PENDING = (1 << 3),
 SGX_SECINFO_MODIFIED = (1 << 4),
 SGX_SECINFO_PR = (1 << 5),
};

struct vdso_symtab {
 Elf64_Sym *elf_symtab;
 const char *elf_symstrtab;
 Elf64_Word *elf_hashtab;
};

static Elf64_Dyn *vdso_get_dyntab(void *addr)
{
 Elf64_Ehdr *ehdr = addr;
 Elf64_Phdr *phdrtab = addr + ehdr->e_phoff;
 int i;

 for (i = 0; i < ehdr->e_phnum; i++)
  if (phdrtab[i].p_type == PT_DYNAMIC)
   return addr + phdrtab[i].p_offset;

 return NULL;
}

static void *vdso_get_dyn(void *addr, Elf64_Dyn *dyntab, Elf64_Sxword tag)
{
 int i;

 for (i = 0; dyntab[i].d_tag != DT_NULL; i++)
  if (dyntab[i].d_tag == tag)
   return addr + dyntab[i].d_un.d_ptr;

 return NULL;
}

static bool vdso_get_symtab(void *addr, struct vdso_symtab *symtab)
{
 Elf64_Dyn *dyntab = vdso_get_dyntab(addr);

 symtab->elf_symtab = vdso_get_dyn(addr, dyntab, DT_SYMTAB);
 if (!symtab->elf_symtab)
  return false;

 symtab->elf_symstrtab = vdso_get_dyn(addr, dyntab, DT_STRTAB);
 if (!symtab->elf_symstrtab)
  return false;

 symtab->elf_hashtab = vdso_get_dyn(addr, dyntab, DT_HASH);
 if (!symtab->elf_hashtab)
  return false;

 return true;
}

static inline int sgx2_supported(void)
{
 unsigned int eax, ebx, ecx, edx;

 __cpuid_count(SGX_CPUID, 0x0, eax, ebx, ecx, edx);

 return eax & 0x2;
}

static unsigned long elf_sym_hash(const char *name)
{
 unsigned long h = 0, high;

 while (*name) {
  h = (h << 4) + *name++;
  high = h & 0xf0000000;

  if (high)
   h ^= high >> 24;

  h &= ~high;
 }

 return h;
}

static Elf64_Sym *vdso_symtab_get(struct vdso_symtab *symtab, const char *name)
{
 Elf64_Word bucketnum = symtab->elf_hashtab[0];
 Elf64_Word *buckettab = &symtab->elf_hashtab[2];
 Elf64_Word *chaintab = &symtab->elf_hashtab[2 + bucketnum];
 Elf64_Sym *sym;
 Elf64_Word i;

 for (i = buckettab[elf_sym_hash(name) % bucketnum]; i != STN_UNDEF;
      i = chaintab[i]) {
  sym = &symtab->elf_symtab[i];
  if (!strcmp(name, &symtab->elf_symstrtab[sym->st_name]))
   return sym;
 }

 return NULL;
}

/*
 * Return the offset in the enclave where the TCS segment can be found.
 * The first RW segment loaded is the TCS.
 */

static off_t encl_get_tcs_offset(struct encl *encl)
{
 int i;

 for (i = 0; i < encl->nr_segments; i++) {
  struct encl_segment *seg = &encl->segment_tbl[i];

  if (i == 0 && seg->prot == (PROT_READ | PROT_WRITE))
   return seg->offset;
 }

 return -1;
}

/*
 * Return the offset in the enclave where the data segment can be found.
 * The first RW segment loaded is the TCS, skip that to get info on the
 * data segment.
 */

static off_t encl_get_data_offset(struct encl *encl)
{
 int i;

 for (i = 1; i < encl->nr_segments; i++) {
  struct encl_segment *seg = &encl->segment_tbl[i];

  if (seg->prot == (PROT_READ | PROT_WRITE))
   return seg->offset;
 }

 return -1;
}

FIXTURE(enclave) {
 struct encl encl;
 struct sgx_enclave_run run;
};

static bool setup_test_encl(unsigned long heap_size, struct encl *encl,
       struct __test_metadata *_metadata)
{
 Elf64_Sym *sgx_enter_enclave_sym = NULL;
 struct vdso_symtab symtab;
 struct encl_segment *seg;
 char maps_line[256];
 FILE *maps_file;
 unsigned int i;
 void *addr;

 if (!encl_load("test_encl.elf", encl, heap_size)) {
  encl_delete(encl);
  TH_LOG("Failed to load the test enclave.");
  return false;
 }

 if (!encl_measure(encl))
  goto err;

 if (!encl_build(encl))
  goto err;

 /*
 * An enclave consumer only must do this.
 */

 for (i = 0; i < encl->nr_segments; i++) {
  struct encl_segment *seg = &encl->segment_tbl[i];

  addr = mmap((void *)encl->encl_base + seg->offset, seg->size,
       seg->prot, MAP_SHARED | MAP_FIXED, encl->fd, 0);
  EXPECT_NE(addr, MAP_FAILED);
  if (addr == MAP_FAILED)
   goto err;
 }

 /* Get vDSO base address */
 addr = (void *)getauxval(AT_SYSINFO_EHDR);
 if (!addr)
  goto err;

 if (!vdso_get_symtab(addr, &symtab))
  goto err;

 sgx_enter_enclave_sym = vdso_symtab_get(&symtab, "__vdso_sgx_enter_enclave");
 if (!sgx_enter_enclave_sym)
  goto err;

 vdso_sgx_enter_enclave = addr + sgx_enter_enclave_sym->st_value;

 return true;

err:
 for (i = 0; i < encl->nr_segments; i++) {
  seg = &encl->segment_tbl[i];

  TH_LOG("0x%016lx 0x%016lx 0x%02x", seg->offset, seg->size, seg->prot);
 }

 maps_file = fopen("/proc/self/maps""r");
 if (maps_file != NULL)  {
  while (fgets(maps_line, sizeof(maps_line), maps_file) != NULL) {
   maps_line[strlen(maps_line) - 1] = '\0';

   if (strstr(maps_line, "/dev/sgx_enclave"))
    TH_LOG("%s", maps_line);
  }

  fclose(maps_file);
 }

 TH_LOG("Failed to initialize the test enclave.");

 encl_delete(encl);

 return false;
}

FIXTURE_SETUP(enclave)
{
}

FIXTURE_TEARDOWN(enclave)
{
 encl_delete(&self->encl);
}

#define ENCL_CALL(op, run, clobbered) \
 ({ \
  int ret; \
  if ((clobbered)) \
   ret = vdso_sgx_enter_enclave((unsigned long)(op), 0, 0, \
           EENTER, 0, 0, (run)); \
  else \
   ret = sgx_enter_enclave((void *)(op), NULL, 0, EENTER, NULL, NULL, \
      (run)); \
  ret; \
 })

#define EXPECT_EEXIT(run) \
 do { \
  EXPECT_EQ((run)->function, EEXIT); \
  if ((run)->function != EEXIT) \
   TH_LOG("0x%02x 0x%02x 0x%016llx", (run)->exception_vector, \
          (run)->exception_error_code, (run)->exception_addr); \
 } while (0)

TEST_F(enclave, unclobbered_vdso)
{
 struct encl_op_get_from_buf get_op;
 struct encl_op_put_to_buf put_op;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 put_op.header.type = ENCL_OP_PUT_TO_BUFFER;
 put_op.value = MAGIC;

 EXPECT_EQ(ENCL_CALL(&put_op, &self->run, false), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);

 get_op.header.type = ENCL_OP_GET_FROM_BUFFER;
 get_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_op, &self->run, false), 0);

 EXPECT_EQ(get_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);
}

/*
 * A section metric is concatenated in a way that @low bits 12-31 define the
 * bits 12-31 of the metric and @high bits 0-19 define the bits 32-51 of the
 * metric.
 */

static unsigned long sgx_calc_section_metric(unsigned int low,
          unsigned int high)
{
 return (low & GENMASK_ULL(31, 12)) +
        ((high & GENMASK_ULL(19, 0)) << 32);
}

/*
 * Sum total available physical SGX memory across all EPC sections
 *
 * Return: total available physical SGX memory available on system
 */

static unsigned long get_total_epc_mem(void)
{
 unsigned int eax, ebx, ecx, edx;
 unsigned long total_size = 0;
 unsigned int type;
 int section = 0;

 while (true) {
  __cpuid_count(SGX_CPUID, section + SGX_CPUID_EPC, eax, ebx, ecx, edx);

  type = eax & SGX_CPUID_EPC_MASK;
  if (type == SGX_CPUID_EPC_INVALID)
   break;

  if (type != SGX_CPUID_EPC_SECTION)
   break;

  total_size += sgx_calc_section_metric(ecx, edx);

  section++;
 }

 return total_size;
}

TEST_F(enclave, unclobbered_vdso_oversubscribed)
{
 struct encl_op_get_from_buf get_op;
 struct encl_op_put_to_buf put_op;
 unsigned long total_mem;

 total_mem = get_total_epc_mem();
 ASSERT_NE(total_mem, 0);
 ASSERT_TRUE(setup_test_encl(total_mem, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 put_op.header.type = ENCL_OP_PUT_TO_BUFFER;
 put_op.value = MAGIC;

 EXPECT_EQ(ENCL_CALL(&put_op, &self->run, false), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);

 get_op.header.type = ENCL_OP_GET_FROM_BUFFER;
 get_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_op, &self->run, false), 0);

 EXPECT_EQ(get_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);
}

TEST_F_TIMEOUT(enclave, unclobbered_vdso_oversubscribed_remove, 900)
{
 struct sgx_enclave_remove_pages remove_ioc;
 struct sgx_enclave_modify_types modt_ioc;
 struct encl_op_get_from_buf get_op;
 struct encl_op_eaccept eaccept_op;
 struct encl_op_put_to_buf put_op;
 struct encl_segment *heap;
 unsigned long total_mem;
 int ret, errno_save;
 unsigned long addr;
 unsigned long i;

 /*
 * Create enclave with additional heap that is as big as all
 * available physical SGX memory.
 */

 total_mem = get_total_epc_mem();
 ASSERT_NE(total_mem, 0);
 TH_LOG("Creating an enclave with %lu bytes heap may take a while ...",
        total_mem);
 ASSERT_TRUE(setup_test_encl(total_mem, &self->encl, _metadata));

 /*
 * Hardware (SGX2) and kernel support is needed for this test. Start
 * with check that test has a chance of succeeding.
 */

 memset(&modt_ioc, 0, sizeof(modt_ioc));
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);

 if (ret == -1) {
  if (errno == ENOTTY)
   SKIP(return,
        "Kernel does not support SGX_IOC_ENCLAVE_MODIFY_TYPES ioctl()");
  else if (errno == ENODEV)
   SKIP(return"System does not support SGX2");
 }

 /*
 * Invalid parameters were provided during sanity check,
 * expect command to fail.
 */

 EXPECT_EQ(ret, -1);

 /* SGX2 is supported by kernel and hardware, test can proceed. */
 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 heap = &self->encl.segment_tbl[self->encl.nr_segments - 1];

 put_op.header.type = ENCL_OP_PUT_TO_BUFFER;
 put_op.value = MAGIC;

 EXPECT_EQ(ENCL_CALL(&put_op, &self->run, false), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);

 get_op.header.type = ENCL_OP_GET_FROM_BUFFER;
 get_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_op, &self->run, false), 0);

 EXPECT_EQ(get_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);

 /* Trim entire heap. */
 memset(&modt_ioc, 0, sizeof(modt_ioc));

 modt_ioc.offset = heap->offset;
 modt_ioc.length = heap->size;
 modt_ioc.page_type = SGX_PAGE_TYPE_TRIM;

 TH_LOG("Changing type of %zd bytes to trimmed may take a while ...",
        heap->size);
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(modt_ioc.result, 0);
 EXPECT_EQ(modt_ioc.count, heap->size);

 /* EACCEPT all removed pages. */
 addr = self->encl.encl_base + heap->offset;

 eaccept_op.flags = SGX_SECINFO_TRIM | SGX_SECINFO_MODIFIED;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 TH_LOG("Entering enclave to run EACCEPT for each page of %zd bytes may take a while ...",
        heap->size);
 for (i = 0; i < heap->size; i += 4096) {
  eaccept_op.epc_addr = addr + i;
  eaccept_op.ret = 0;

  EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

  EXPECT_EQ(self->run.exception_vector, 0);
  EXPECT_EQ(self->run.exception_error_code, 0);
  EXPECT_EQ(self->run.exception_addr, 0);
  ASSERT_EQ(eaccept_op.ret, 0);
  ASSERT_EQ(self->run.function, EEXIT);
 }

 /* Complete page removal. */
 memset(&remove_ioc, 0, sizeof(remove_ioc));

 remove_ioc.offset = heap->offset;
 remove_ioc.length = heap->size;

 TH_LOG("Removing %zd bytes from enclave may take a while ...",
        heap->size);
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_REMOVE_PAGES, &remove_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(remove_ioc.count, heap->size);
}

TEST_F(enclave, clobbered_vdso)
{
 struct encl_op_get_from_buf get_op;
 struct encl_op_put_to_buf put_op;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 put_op.header.type = ENCL_OP_PUT_TO_BUFFER;
 put_op.value = MAGIC;

 EXPECT_EQ(ENCL_CALL(&put_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);

 get_op.header.type = ENCL_OP_GET_FROM_BUFFER;
 get_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_op, &self->run, true), 0);

 EXPECT_EQ(get_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);
}

static int test_handler(long rdi, long rsi, long rdx, long ursp, long r8, long r9,
   struct sgx_enclave_run *run)
{
 run->user_data = 0;

 return 0;
}

TEST_F(enclave, clobbered_vdso_and_user_function)
{
 struct encl_op_get_from_buf get_op;
 struct encl_op_put_to_buf put_op;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 self->run.user_handler = (__u64)test_handler;
 self->run.user_data = 0xdeadbeef;

 put_op.header.type = ENCL_OP_PUT_TO_BUFFER;
 put_op.value = MAGIC;

 EXPECT_EQ(ENCL_CALL(&put_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);

 get_op.header.type = ENCL_OP_GET_FROM_BUFFER;
 get_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_op, &self->run, true), 0);

 EXPECT_EQ(get_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);
}

/*
 * Sanity check that it is possible to enter either of the two hardcoded TCS
 */

TEST_F(enclave, tcs_entry)
{
 struct encl_op_header op;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 op.type = ENCL_OP_NOP;

 EXPECT_EQ(ENCL_CALL(&op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /* Move to the next TCS. */
 self->run.tcs = self->encl.encl_base + PAGE_SIZE;

 EXPECT_EQ(ENCL_CALL(&op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
}

/*
 * Second page of .data segment is used to test changing PTE permissions.
 * This spans the local encl_buffer within the test enclave.
 *
 * 1) Start with a sanity check: a value is written to the target page within
 *    the enclave and read back to ensure target page can be written to.
 * 2) Change PTE permissions (RW -> RO) of target page within enclave.
 * 3) Repeat (1) - this time expecting a regular #PF communicated via the
 *    vDSO.
 * 4) Change PTE permissions of target page within enclave back to be RW.
 * 5) Repeat (1) by resuming enclave, now expected to be possible to write to
 *    and read from target page within enclave.
 */

TEST_F(enclave, pte_permissions)
{
 struct encl_op_get_from_addr get_addr_op;
 struct encl_op_put_to_addr put_addr_op;
 unsigned long data_start;
 int ret;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 data_start = self->encl.encl_base +
       encl_get_data_offset(&self->encl) +
       PAGE_SIZE;

 /*
 * Sanity check to ensure it is possible to write to page that will
 * have its permissions manipulated.
 */


 /* Write MAGIC to page */
 put_addr_op.value = MAGIC;
 put_addr_op.addr = data_start;
 put_addr_op.header.type = ENCL_OP_PUT_TO_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Read memory that was just written to, confirming that it is the
 * value previously written (MAGIC).
 */

 get_addr_op.value = 0;
 get_addr_op.addr = data_start;
 get_addr_op.header.type = ENCL_OP_GET_FROM_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /* Change PTE permissions of target page within the enclave */
 ret = mprotect((void *)data_start, PAGE_SIZE, PROT_READ);
 if (ret)
  perror("mprotect");

 /*
 * PTE permissions of target page changed to read-only, EPCM
 * permissions unchanged (EPCM permissions are RW), attempt to
 * write to the page, expecting a regular #PF.
 */


 put_addr_op.value = MAGIC2;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EQ(self->run.exception_vector, 14);
 EXPECT_EQ(self->run.exception_error_code, 0x7);
 EXPECT_EQ(self->run.exception_addr, data_start);

 self->run.exception_vector = 0;
 self->run.exception_error_code = 0;
 self->run.exception_addr = 0;

 /*
 * Change PTE permissions back to enable enclave to write to the
 * target page and resume enclave - do not expect any exceptions this
 * time.
 */

 ret = mprotect((void *)data_start, PAGE_SIZE, PROT_READ | PROT_WRITE);
 if (ret)
  perror("mprotect");

 EXPECT_EQ(vdso_sgx_enter_enclave((unsigned long)&put_addr_op, 0,
      0, ERESUME, 0, 0, &self->run),
   0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 get_addr_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC2);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
}

/*
 * Modifying permissions of TCS page should not be possible.
 */

TEST_F(enclave, tcs_permissions)
{
 struct sgx_enclave_restrict_permissions ioc;
 int ret, errno_save;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 memset(&ioc, 0, sizeof(ioc));

 /*
 * Ensure kernel supports needed ioctl() and system supports needed
 * commands.
 */


 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_RESTRICT_PERMISSIONS, &ioc);
 errno_save = ret == -1 ? errno : 0;

 /*
 * Invalid parameters were provided during sanity check,
 * expect command to fail.
 */

 ASSERT_EQ(ret, -1);

 /* ret == -1 */
 if (errno_save == ENOTTY)
  SKIP(return,
       "Kernel does not support SGX_IOC_ENCLAVE_RESTRICT_PERMISSIONS ioctl()");
 else if (errno_save == ENODEV)
  SKIP(return"System does not support SGX2");

 /*
 * Attempt to make TCS page read-only. This is not allowed and
 * should be prevented by the kernel.
 */

 ioc.offset = encl_get_tcs_offset(&self->encl);
 ioc.length = PAGE_SIZE;
 ioc.permissions = SGX_SECINFO_R;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_RESTRICT_PERMISSIONS, &ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, -1);
 EXPECT_EQ(errno_save, EINVAL);
 EXPECT_EQ(ioc.result, 0);
 EXPECT_EQ(ioc.count, 0);
}

/*
 * Enclave page permission test.
 *
 * Modify and restore enclave page's EPCM (enclave) permissions from
 * outside enclave (ENCLS[EMODPR] via kernel) as well as from within
 * enclave (via ENCLU[EMODPE]). Check for page fault if
 * VMA allows access but EPCM permissions do not.
 */

TEST_F(enclave, epcm_permissions)
{
 struct sgx_enclave_restrict_permissions restrict_ioc;
 struct encl_op_get_from_addr get_addr_op;
 struct encl_op_put_to_addr put_addr_op;
 struct encl_op_eaccept eaccept_op;
 struct encl_op_emodpe emodpe_op;
 unsigned long data_start;
 int ret, errno_save;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 /*
 * Ensure kernel supports needed ioctl() and system supports needed
 * commands.
 */

 memset(&restrict_ioc, 0, sizeof(restrict_ioc));

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_RESTRICT_PERMISSIONS,
      &restrict_ioc);
 errno_save = ret == -1 ? errno : 0;

 /*
 * Invalid parameters were provided during sanity check,
 * expect command to fail.
 */

 ASSERT_EQ(ret, -1);

 /* ret == -1 */
 if (errno_save == ENOTTY)
  SKIP(return,
       "Kernel does not support SGX_IOC_ENCLAVE_RESTRICT_PERMISSIONS ioctl()");
 else if (errno_save == ENODEV)
  SKIP(return"System does not support SGX2");

 /*
 * Page that will have its permissions changed is the second data
 * page in the .data segment. This forms part of the local encl_buffer
 * within the enclave.
 *
 * At start of test @data_start should have EPCM as well as PTE and
 * VMA permissions of RW.
 */


 data_start = self->encl.encl_base +
       encl_get_data_offset(&self->encl) + PAGE_SIZE;

 /*
 * Sanity check that page at @data_start is writable before making
 * any changes to page permissions.
 *
 * Start by writing MAGIC to test page.
 */

 put_addr_op.value = MAGIC;
 put_addr_op.addr = data_start;
 put_addr_op.header.type = ENCL_OP_PUT_TO_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Read memory that was just written to, confirming that
 * page is writable.
 */

 get_addr_op.value = 0;
 get_addr_op.addr = data_start;
 get_addr_op.header.type = ENCL_OP_GET_FROM_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Change EPCM permissions to read-only. Kernel still considers
 * the page writable.
 */

 memset(&restrict_ioc, 0, sizeof(restrict_ioc));

 restrict_ioc.offset = encl_get_data_offset(&self->encl) + PAGE_SIZE;
 restrict_ioc.length = PAGE_SIZE;
 restrict_ioc.permissions = SGX_SECINFO_R;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_RESTRICT_PERMISSIONS,
      &restrict_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(restrict_ioc.result, 0);
 EXPECT_EQ(restrict_ioc.count, 4096);

 /*
 * EPCM permissions changed from kernel, need to EACCEPT from enclave.
 */

 eaccept_op.epc_addr = data_start;
 eaccept_op.flags = SGX_SECINFO_R | SGX_SECINFO_REG | SGX_SECINFO_PR;
 eaccept_op.ret = 0;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 /*
 * EPCM permissions of page is now read-only, expect #PF
 * on EPCM when attempting to write to page from within enclave.
 */

 put_addr_op.value = MAGIC2;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EQ(self->run.function, ERESUME);
 EXPECT_EQ(self->run.exception_vector, 14);
 EXPECT_EQ(self->run.exception_error_code, 0x8007);
 EXPECT_EQ(self->run.exception_addr, data_start);

 self->run.exception_vector = 0;
 self->run.exception_error_code = 0;
 self->run.exception_addr = 0;

 /*
 * Received AEX but cannot return to enclave at same entrypoint,
 * need different TCS from where EPCM permission can be made writable
 * again.
 */

 self->run.tcs = self->encl.encl_base + PAGE_SIZE;

 /*
 * Enter enclave at new TCS to change EPCM permissions to be
 * writable again and thus fix the page fault that triggered the
 * AEX.
 */


 emodpe_op.epc_addr = data_start;
 emodpe_op.flags = SGX_SECINFO_R | SGX_SECINFO_W;
 emodpe_op.header.type = ENCL_OP_EMODPE;

 EXPECT_EQ(ENCL_CALL(&emodpe_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Attempt to return to main TCS to resume execution at faulting
 * instruction, PTE should continue to allow writing to the page.
 */

 self->run.tcs = self->encl.encl_base;

 /*
 * Wrong page permissions that caused original fault has
 * now been fixed via EPCM permissions.
 * Resume execution in main TCS to re-attempt the memory access.
 */

 self->run.tcs = self->encl.encl_base;

 EXPECT_EQ(vdso_sgx_enter_enclave((unsigned long)&put_addr_op, 0, 0,
      ERESUME, 0, 0,
      &self->run),
    0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 get_addr_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC2);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.user_data, 0);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
}

/*
 * Test the addition of pages to an initialized enclave via writing to
 * a page belonging to the enclave's address space but was not added
 * during enclave creation.
 */

TEST_F(enclave, augment)
{
 struct encl_op_get_from_addr get_addr_op;
 struct encl_op_put_to_addr put_addr_op;
 struct encl_op_eaccept eaccept_op;
 size_t total_size = 0;
 void *addr;
 int i;

 if (!sgx2_supported())
  SKIP(return"SGX2 not supported");

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 for (i = 0; i < self->encl.nr_segments; i++) {
  struct encl_segment *seg = &self->encl.segment_tbl[i];

  total_size += seg->size;
 }

 /*
 * Actual enclave size is expected to be larger than the loaded
 * test enclave since enclave size must be a power of 2 in bytes
 * and test_encl does not consume it all.
 */

 EXPECT_LT(total_size + PAGE_SIZE, self->encl.encl_size);

 /*
 * Create memory mapping for the page that will be added. New
 * memory mapping is for one page right after all existing
 * mappings.
 * Kernel will allow new mapping using any permissions if it
 * falls into the enclave's address range but not backed
 * by existing enclave pages.
 */

 addr = mmap((void *)self->encl.encl_base + total_size, PAGE_SIZE,
      PROT_READ | PROT_WRITE | PROT_EXEC,
      MAP_SHARED | MAP_FIXED, self->encl.fd, 0);
 EXPECT_NE(addr, MAP_FAILED);

 self->run.exception_vector = 0;
 self->run.exception_error_code = 0;
 self->run.exception_addr = 0;

 /*
 * Attempt to write to the new page from within enclave.
 * Expected to fail since page is not (yet) part of the enclave.
 * The first #PF will trigger the addition of the page to the
 * enclave, but since the new page needs an EACCEPT from within the
 * enclave before it can be used it would not be possible
 * to successfully return to the failing instruction. This is the
 * cause of the second #PF captured here having the SGX bit set,
 * it is from hardware preventing the page from being used.
 */

 put_addr_op.value = MAGIC;
 put_addr_op.addr = (unsigned long)addr;
 put_addr_op.header.type = ENCL_OP_PUT_TO_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EQ(self->run.function, ERESUME);
 EXPECT_EQ(self->run.exception_vector, 14);
 EXPECT_EQ(self->run.exception_addr, (unsigned long)addr);

 if (self->run.exception_error_code == 0x6) {
  munmap(addr, PAGE_SIZE);
  SKIP(return"Kernel does not support adding pages to initialized enclave");
 }

 EXPECT_EQ(self->run.exception_error_code, 0x8007);

 self->run.exception_vector = 0;
 self->run.exception_error_code = 0;
 self->run.exception_addr = 0;

 /* Handle AEX by running EACCEPT from new entry point. */
 self->run.tcs = self->encl.encl_base + PAGE_SIZE;

 eaccept_op.epc_addr = self->encl.encl_base + total_size;
 eaccept_op.flags = SGX_SECINFO_R | SGX_SECINFO_W | SGX_SECINFO_REG | SGX_SECINFO_PENDING;
 eaccept_op.ret = 0;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 /* Can now return to main TCS to resume execution. */
 self->run.tcs = self->encl.encl_base;

 EXPECT_EQ(vdso_sgx_enter_enclave((unsigned long)&put_addr_op, 0, 0,
      ERESUME, 0, 0,
      &self->run),
    0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Read memory from newly added page that was just written to,
 * confirming that data previously written (MAGIC) is present.
 */

 get_addr_op.value = 0;
 get_addr_op.addr = (unsigned long)addr;
 get_addr_op.header.type = ENCL_OP_GET_FROM_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 munmap(addr, PAGE_SIZE);
}

/*
 * Test for the addition of pages to an initialized enclave via a
 * pre-emptive run of EACCEPT on page to be added.
 */

TEST_F(enclave, augment_via_eaccept)
{
 struct encl_op_get_from_addr get_addr_op;
 struct encl_op_put_to_addr put_addr_op;
 struct encl_op_eaccept eaccept_op;
 size_t total_size = 0;
 void *addr;
 int i;

 if (!sgx2_supported())
  SKIP(return"SGX2 not supported");

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 for (i = 0; i < self->encl.nr_segments; i++) {
  struct encl_segment *seg = &self->encl.segment_tbl[i];

  total_size += seg->size;
 }

 /*
 * Actual enclave size is expected to be larger than the loaded
 * test enclave since enclave size must be a power of 2 in bytes while
 * test_encl does not consume it all.
 */

 EXPECT_LT(total_size + PAGE_SIZE, self->encl.encl_size);

 /*
 * mmap() a page at end of existing enclave to be used for dynamic
 * EPC page.
 *
 * Kernel will allow new mapping using any permissions if it
 * falls into the enclave's address range but not backed
 * by existing enclave pages.
 */


 addr = mmap((void *)self->encl.encl_base + total_size, PAGE_SIZE,
      PROT_READ | PROT_WRITE | PROT_EXEC, MAP_SHARED | MAP_FIXED,
      self->encl.fd, 0);
 EXPECT_NE(addr, MAP_FAILED);

 self->run.exception_vector = 0;
 self->run.exception_error_code = 0;
 self->run.exception_addr = 0;

 /*
 * Run EACCEPT on new page to trigger the #PF->EAUG->EACCEPT(again
 * without a #PF). All should be transparent to userspace.
 */

 eaccept_op.epc_addr = self->encl.encl_base + total_size;
 eaccept_op.flags = SGX_SECINFO_R | SGX_SECINFO_W | SGX_SECINFO_REG | SGX_SECINFO_PENDING;
 eaccept_op.ret = 0;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 if (self->run.exception_vector == 14 &&
     self->run.exception_error_code == 4 &&
     self->run.exception_addr == self->encl.encl_base + total_size) {
  munmap(addr, PAGE_SIZE);
  SKIP(return"Kernel does not support adding pages to initialized enclave");
 }

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 /*
 * New page should be accessible from within enclave - attempt to
 * write to it.
 */

 put_addr_op.value = MAGIC;
 put_addr_op.addr = (unsigned long)addr;
 put_addr_op.header.type = ENCL_OP_PUT_TO_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Read memory from newly added page that was just written to,
 * confirming that data previously written (MAGIC) is present.
 */

 get_addr_op.value = 0;
 get_addr_op.addr = (unsigned long)addr;
 get_addr_op.header.type = ENCL_OP_GET_FROM_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 munmap(addr, PAGE_SIZE);
}

/*
 * SGX2 page type modification test in two phases:
 * Phase 1:
 * Create a new TCS, consisting out of three new pages (stack page with regular
 * page type, SSA page with regular page type, and TCS page with TCS page
 * type) in an initialized enclave and run a simple workload within it.
 * Phase 2:
 * Remove the three pages added in phase 1, add a new regular page at the
 * same address that previously hosted the TCS page and verify that it can
 * be modified.
 */

TEST_F(enclave, tcs_create)
{
 struct encl_op_init_tcs_page init_tcs_page_op;
 struct sgx_enclave_remove_pages remove_ioc;
 struct encl_op_get_from_addr get_addr_op;
 struct sgx_enclave_modify_types modt_ioc;
 struct encl_op_put_to_addr put_addr_op;
 struct encl_op_get_from_buf get_buf_op;
 struct encl_op_put_to_buf put_buf_op;
 void *addr, *tcs, *stack_end, *ssa;
 struct encl_op_eaccept eaccept_op;
 size_t total_size = 0;
 uint64_t val_64;
 int errno_save;
 int ret, i;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl,
        _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 /*
 * Hardware (SGX2) and kernel support is needed for this test. Start
 * with check that test has a chance of succeeding.
 */

 memset(&modt_ioc, 0, sizeof(modt_ioc));
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);

 if (ret == -1) {
  if (errno == ENOTTY)
   SKIP(return,
        "Kernel does not support SGX_IOC_ENCLAVE_MODIFY_TYPES ioctl()");
  else if (errno == ENODEV)
   SKIP(return"System does not support SGX2");
 }

 /*
 * Invalid parameters were provided during sanity check,
 * expect command to fail.
 */

 EXPECT_EQ(ret, -1);

 /*
 * Add three regular pages via EAUG: one will be the TCS stack, one
 * will be the TCS SSA, and one will be the new TCS. The stack and
 * SSA will remain as regular pages, the TCS page will need its
 * type changed after populated with needed data.
 */

 for (i = 0; i < self->encl.nr_segments; i++) {
  struct encl_segment *seg = &self->encl.segment_tbl[i];

  total_size += seg->size;
 }

 /*
 * Actual enclave size is expected to be larger than the loaded
 * test enclave since enclave size must be a power of 2 in bytes while
 * test_encl does not consume it all.
 */

 EXPECT_LT(total_size + 3 * PAGE_SIZE, self->encl.encl_size);

 /*
 * mmap() three pages at end of existing enclave to be used for the
 * three new pages.
 */

 addr = mmap((void *)self->encl.encl_base + total_size, 3 * PAGE_SIZE,
      PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED,
      self->encl.fd, 0);
 EXPECT_NE(addr, MAP_FAILED);

 self->run.exception_vector = 0;
 self->run.exception_error_code = 0;
 self->run.exception_addr = 0;

 stack_end = (void *)self->encl.encl_base + total_size;
 tcs = (void *)self->encl.encl_base + total_size + PAGE_SIZE;
 ssa = (void *)self->encl.encl_base + total_size + 2 * PAGE_SIZE;

 /*
 * Run EACCEPT on each new page to trigger the
 * EACCEPT->(#PF)->EAUG->EACCEPT(again without a #PF) flow.
 */


 eaccept_op.epc_addr = (unsigned long)stack_end;
 eaccept_op.flags = SGX_SECINFO_R | SGX_SECINFO_W | SGX_SECINFO_REG | SGX_SECINFO_PENDING;
 eaccept_op.ret = 0;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 if (self->run.exception_vector == 14 &&
     self->run.exception_error_code == 4 &&
     self->run.exception_addr == (unsigned long)stack_end) {
  munmap(addr, 3 * PAGE_SIZE);
  SKIP(return"Kernel does not support adding pages to initialized enclave");
 }

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 eaccept_op.epc_addr = (unsigned long)ssa;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 eaccept_op.epc_addr = (unsigned long)tcs;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 /*
 * Three new pages added to enclave. Now populate the TCS page with
 * needed data. This should be done from within enclave. Provide
 * the function that will do the actual data population with needed
 * data.
 */


 /*
 * New TCS will use the "encl_dyn_entry" entrypoint that expects
 * stack to begin in page before TCS page.
 */

 val_64 = encl_get_entry(&self->encl, "encl_dyn_entry");
 EXPECT_NE(val_64, 0);

 init_tcs_page_op.tcs_page = (unsigned long)tcs;
 init_tcs_page_op.ssa = (unsigned long)total_size + 2 * PAGE_SIZE;
 init_tcs_page_op.entry = val_64;
 init_tcs_page_op.header.type = ENCL_OP_INIT_TCS_PAGE;

 EXPECT_EQ(ENCL_CALL(&init_tcs_page_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /* Change TCS page type to TCS. */
 memset(&modt_ioc, 0, sizeof(modt_ioc));

 modt_ioc.offset = total_size + PAGE_SIZE;
 modt_ioc.length = PAGE_SIZE;
 modt_ioc.page_type = SGX_PAGE_TYPE_TCS;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(modt_ioc.result, 0);
 EXPECT_EQ(modt_ioc.count, 4096);

 /* EACCEPT new TCS page from enclave. */
 eaccept_op.epc_addr = (unsigned long)tcs;
 eaccept_op.flags = SGX_SECINFO_TCS | SGX_SECINFO_MODIFIED;
 eaccept_op.ret = 0;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 /* Run workload from new TCS. */
 self->run.tcs = (unsigned long)tcs;

 /*
 * Simple workload to write to data buffer and read value back.
 */

 put_buf_op.header.type = ENCL_OP_PUT_TO_BUFFER;
 put_buf_op.value = MAGIC;

 EXPECT_EQ(ENCL_CALL(&put_buf_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 get_buf_op.header.type = ENCL_OP_GET_FROM_BUFFER;
 get_buf_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_buf_op, &self->run, true), 0);

 EXPECT_EQ(get_buf_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Phase 2 of test:
 * Remove pages associated with new TCS, create a regular page
 * where TCS page used to be and verify it can be used as a regular
 * page.
 */


 /* Start page removal by requesting change of page type to PT_TRIM. */
 memset(&modt_ioc, 0, sizeof(modt_ioc));

 modt_ioc.offset = total_size;
 modt_ioc.length = 3 * PAGE_SIZE;
 modt_ioc.page_type = SGX_PAGE_TYPE_TRIM;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(modt_ioc.result, 0);
 EXPECT_EQ(modt_ioc.count, 3 * PAGE_SIZE);

 /*
 * Enter enclave via TCS #1 and approve page removal by sending
 * EACCEPT for each of three removed pages.
 */

 self->run.tcs = self->encl.encl_base;

 eaccept_op.epc_addr = (unsigned long)stack_end;
 eaccept_op.flags = SGX_SECINFO_TRIM | SGX_SECINFO_MODIFIED;
 eaccept_op.ret = 0;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 eaccept_op.epc_addr = (unsigned long)tcs;
 eaccept_op.ret = 0;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 eaccept_op.epc_addr = (unsigned long)ssa;
 eaccept_op.ret = 0;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 /* Send final ioctl() to complete page removal. */
 memset(&remove_ioc, 0, sizeof(remove_ioc));

 remove_ioc.offset = total_size;
 remove_ioc.length = 3 * PAGE_SIZE;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_REMOVE_PAGES, &remove_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(remove_ioc.count, 3 * PAGE_SIZE);

 /*
 * Enter enclave via TCS #1 and access location where TCS #3 was to
 * trigger dynamic add of regular page at that location.
 */

 eaccept_op.epc_addr = (unsigned long)tcs;
 eaccept_op.flags = SGX_SECINFO_R | SGX_SECINFO_W | SGX_SECINFO_REG | SGX_SECINFO_PENDING;
 eaccept_op.ret = 0;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 /*
 * New page should be accessible from within enclave - write to it.
 */

 put_addr_op.value = MAGIC;
 put_addr_op.addr = (unsigned long)tcs;
 put_addr_op.header.type = ENCL_OP_PUT_TO_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Read memory from newly added page that was just written to,
 * confirming that data previously written (MAGIC) is present.
 */

 get_addr_op.value = 0;
 get_addr_op.addr = (unsigned long)tcs;
 get_addr_op.header.type = ENCL_OP_GET_FROM_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 munmap(addr, 3 * PAGE_SIZE);
}

/*
 * Ensure sane behavior if user requests page removal, does not run
 * EACCEPT from within enclave but still attempts to finalize page removal
 * with the SGX_IOC_ENCLAVE_REMOVE_PAGES ioctl(). The latter should fail
 * because the removal was not EACCEPTed from within the enclave.
 */

TEST_F(enclave, remove_added_page_no_eaccept)
{
 struct sgx_enclave_remove_pages remove_ioc;
 struct encl_op_get_from_addr get_addr_op;
 struct sgx_enclave_modify_types modt_ioc;
 struct encl_op_put_to_addr put_addr_op;
 unsigned long data_start;
 int ret, errno_save;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 /*
 * Hardware (SGX2) and kernel support is needed for this test. Start
 * with check that test has a chance of succeeding.
 */

 memset(&modt_ioc, 0, sizeof(modt_ioc));
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);

 if (ret == -1) {
  if (errno == ENOTTY)
   SKIP(return,
        "Kernel does not support SGX_IOC_ENCLAVE_MODIFY_TYPES ioctl()");
  else if (errno == ENODEV)
   SKIP(return"System does not support SGX2");
 }

 /*
 * Invalid parameters were provided during sanity check,
 * expect command to fail.
 */

 EXPECT_EQ(ret, -1);

 /*
 * Page that will be removed is the second data page in the .data
 * segment. This forms part of the local encl_buffer within the
 * enclave.
 */

 data_start = self->encl.encl_base +
       encl_get_data_offset(&self->encl) + PAGE_SIZE;

 /*
 * Sanity check that page at @data_start is writable before
 * removing it.
 *
 * Start by writing MAGIC to test page.
 */

 put_addr_op.value = MAGIC;
 put_addr_op.addr = data_start;
 put_addr_op.header.type = ENCL_OP_PUT_TO_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Read memory that was just written to, confirming that data
 * previously written (MAGIC) is present.
 */

 get_addr_op.value = 0;
 get_addr_op.addr = data_start;
 get_addr_op.header.type = ENCL_OP_GET_FROM_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /* Start page removal by requesting change of page type to PT_TRIM */
 memset(&modt_ioc, 0, sizeof(modt_ioc));

 modt_ioc.offset = encl_get_data_offset(&self->encl) + PAGE_SIZE;
 modt_ioc.length = PAGE_SIZE;
 modt_ioc.page_type = SGX_PAGE_TYPE_TRIM;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(modt_ioc.result, 0);
 EXPECT_EQ(modt_ioc.count, 4096);

 /* Skip EACCEPT */

 /* Send final ioctl() to complete page removal */
 memset(&remove_ioc, 0, sizeof(remove_ioc));

 remove_ioc.offset = encl_get_data_offset(&self->encl) + PAGE_SIZE;
 remove_ioc.length = PAGE_SIZE;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_REMOVE_PAGES, &remove_ioc);
 errno_save = ret == -1 ? errno : 0;

 /* Operation not permitted since EACCEPT was omitted. */
 EXPECT_EQ(ret, -1);
 EXPECT_EQ(errno_save, EPERM);
 EXPECT_EQ(remove_ioc.count, 0);
}

/*
 * Request enclave page removal but instead of correctly following with
 * EACCEPT a read attempt to page is made from within the enclave.
 */

TEST_F(enclave, remove_added_page_invalid_access)
{
 struct encl_op_get_from_addr get_addr_op;
 struct encl_op_put_to_addr put_addr_op;
 struct sgx_enclave_modify_types ioc;
 unsigned long data_start;
 int ret, errno_save;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 /*
 * Hardware (SGX2) and kernel support is needed for this test. Start
 * with check that test has a chance of succeeding.
 */

 memset(&ioc, 0, sizeof(ioc));
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &ioc);

 if (ret == -1) {
  if (errno == ENOTTY)
   SKIP(return,
        "Kernel does not support SGX_IOC_ENCLAVE_MODIFY_TYPES ioctl()");
  else if (errno == ENODEV)
   SKIP(return"System does not support SGX2");
 }

 /*
 * Invalid parameters were provided during sanity check,
 * expect command to fail.
 */

 EXPECT_EQ(ret, -1);

 /*
 * Page that will be removed is the second data page in the .data
 * segment. This forms part of the local encl_buffer within the
 * enclave.
 */

 data_start = self->encl.encl_base +
       encl_get_data_offset(&self->encl) + PAGE_SIZE;

 /*
 * Sanity check that page at @data_start is writable before
 * removing it.
 *
 * Start by writing MAGIC to test page.
 */

 put_addr_op.value = MAGIC;
 put_addr_op.addr = data_start;
 put_addr_op.header.type = ENCL_OP_PUT_TO_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Read memory that was just written to, confirming that data
 * previously written (MAGIC) is present.
 */

 get_addr_op.value = 0;
 get_addr_op.addr = data_start;
 get_addr_op.header.type = ENCL_OP_GET_FROM_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /* Start page removal by requesting change of page type to PT_TRIM. */
 memset(&ioc, 0, sizeof(ioc));

 ioc.offset = encl_get_data_offset(&self->encl) + PAGE_SIZE;
 ioc.length = PAGE_SIZE;
 ioc.page_type = SGX_PAGE_TYPE_TRIM;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(ioc.result, 0);
 EXPECT_EQ(ioc.count, 4096);

 /*
 * Read from page that was just removed.
 */

 get_addr_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 /*
 * From kernel perspective the page is present but according to SGX the
 * page should not be accessible so a #PF with SGX bit set is
 * expected.
 */


 EXPECT_EQ(self->run.function, ERESUME);
 EXPECT_EQ(self->run.exception_vector, 14);
 EXPECT_EQ(self->run.exception_error_code, 0x8005);
 EXPECT_EQ(self->run.exception_addr, data_start);
}

/*
 * Request enclave page removal and correctly follow with
 * EACCEPT but do not follow with removal ioctl() but instead a read attempt
 * to removed page is made from within the enclave.
 */

TEST_F(enclave, remove_added_page_invalid_access_after_eaccept)
{
 struct encl_op_get_from_addr get_addr_op;
 struct encl_op_put_to_addr put_addr_op;
 struct sgx_enclave_modify_types ioc;
 struct encl_op_eaccept eaccept_op;
 unsigned long data_start;
 int ret, errno_save;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 /*
 * Hardware (SGX2) and kernel support is needed for this test. Start
 * with check that test has a chance of succeeding.
 */

 memset(&ioc, 0, sizeof(ioc));
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &ioc);

 if (ret == -1) {
  if (errno == ENOTTY)
   SKIP(return,
        "Kernel does not support SGX_IOC_ENCLAVE_MODIFY_TYPES ioctl()");
  else if (errno == ENODEV)
   SKIP(return"System does not support SGX2");
 }

 /*
 * Invalid parameters were provided during sanity check,
 * expect command to fail.
 */

 EXPECT_EQ(ret, -1);

 /*
 * Page that will be removed is the second data page in the .data
 * segment. This forms part of the local encl_buffer within the
 * enclave.
 */

 data_start = self->encl.encl_base +
       encl_get_data_offset(&self->encl) + PAGE_SIZE;

 /*
 * Sanity check that page at @data_start is writable before
 * removing it.
 *
 * Start by writing MAGIC to test page.
 */

 put_addr_op.value = MAGIC;
 put_addr_op.addr = data_start;
 put_addr_op.header.type = ENCL_OP_PUT_TO_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&put_addr_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /*
 * Read memory that was just written to, confirming that data
 * previously written (MAGIC) is present.
 */

 get_addr_op.value = 0;
 get_addr_op.addr = data_start;
 get_addr_op.header.type = ENCL_OP_GET_FROM_ADDRESS;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 EXPECT_EQ(get_addr_op.value, MAGIC);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);

 /* Start page removal by requesting change of page type to PT_TRIM. */
 memset(&ioc, 0, sizeof(ioc));

 ioc.offset = encl_get_data_offset(&self->encl) + PAGE_SIZE;
 ioc.length = PAGE_SIZE;
 ioc.page_type = SGX_PAGE_TYPE_TRIM;

 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(ioc.result, 0);
 EXPECT_EQ(ioc.count, 4096);

 eaccept_op.epc_addr = (unsigned long)data_start;
 eaccept_op.ret = 0;
 eaccept_op.flags = SGX_SECINFO_TRIM | SGX_SECINFO_MODIFIED;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);

 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 /* Skip ioctl() to remove page. */

 /*
 * Read from page that was just removed.
 */

 get_addr_op.value = 0;

 EXPECT_EQ(ENCL_CALL(&get_addr_op, &self->run, true), 0);

 /*
 * From kernel perspective the page is present but according to SGX the
 * page should not be accessible so a #PF with SGX bit set is
 * expected.
 */


 EXPECT_EQ(self->run.function, ERESUME);
 EXPECT_EQ(self->run.exception_vector, 14);
 EXPECT_EQ(self->run.exception_error_code, 0x8005);
 EXPECT_EQ(self->run.exception_addr, data_start);
}

TEST_F(enclave, remove_untouched_page)
{
 struct sgx_enclave_remove_pages remove_ioc;
 struct sgx_enclave_modify_types modt_ioc;
 struct encl_op_eaccept eaccept_op;
 unsigned long data_start;
 int ret, errno_save;

 ASSERT_TRUE(setup_test_encl(ENCL_HEAP_SIZE_DEFAULT, &self->encl, _metadata));

 /*
 * Hardware (SGX2) and kernel support is needed for this test. Start
 * with check that test has a chance of succeeding.
 */

 memset(&modt_ioc, 0, sizeof(modt_ioc));
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);

 if (ret == -1) {
  if (errno == ENOTTY)
   SKIP(return,
        "Kernel does not support SGX_IOC_ENCLAVE_MODIFY_TYPES ioctl()");
  else if (errno == ENODEV)
   SKIP(return"System does not support SGX2");
 }

 /*
 * Invalid parameters were provided during sanity check,
 * expect command to fail.
 */

 EXPECT_EQ(ret, -1);

 /* SGX2 is supported by kernel and hardware, test can proceed. */
 memset(&self->run, 0, sizeof(self->run));
 self->run.tcs = self->encl.encl_base;

 data_start = self->encl.encl_base +
    encl_get_data_offset(&self->encl) + PAGE_SIZE;

 memset(&modt_ioc, 0, sizeof(modt_ioc));

 modt_ioc.offset = encl_get_data_offset(&self->encl) + PAGE_SIZE;
 modt_ioc.length = PAGE_SIZE;
 modt_ioc.page_type = SGX_PAGE_TYPE_TRIM;
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_MODIFY_TYPES, &modt_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(modt_ioc.result, 0);
 EXPECT_EQ(modt_ioc.count, 4096);

 /*
 * Enter enclave via TCS #1 and approve page removal by sending
 * EACCEPT for removed page.
 */


 eaccept_op.epc_addr = data_start;
 eaccept_op.flags = SGX_SECINFO_TRIM | SGX_SECINFO_MODIFIED;
 eaccept_op.ret = 0;
 eaccept_op.header.type = ENCL_OP_EACCEPT;

 EXPECT_EQ(ENCL_CALL(&eaccept_op, &self->run, true), 0);
 EXPECT_EEXIT(&self->run);
 EXPECT_EQ(self->run.exception_vector, 0);
 EXPECT_EQ(self->run.exception_error_code, 0);
 EXPECT_EQ(self->run.exception_addr, 0);
 EXPECT_EQ(eaccept_op.ret, 0);

 memset(&remove_ioc, 0, sizeof(remove_ioc));

 remove_ioc.offset = encl_get_data_offset(&self->encl) + PAGE_SIZE;
 remove_ioc.length = PAGE_SIZE;
 ret = ioctl(self->encl.fd, SGX_IOC_ENCLAVE_REMOVE_PAGES, &remove_ioc);
 errno_save = ret == -1 ? errno : 0;

 EXPECT_EQ(ret, 0);
 EXPECT_EQ(errno_save, 0);
 EXPECT_EQ(remove_ioc.count, 4096);
}

TEST_HARNESS_MAIN

Messung V0.5 in Prozent
C=95 H=92 G=93

[Verzeichnis aufwärts0.46unsichere VerbindungÜbersetzung europäischer Sprachen durch Browser2026-04-29]