From: Chang S. Bae chang.seok.bae@intel.com

AMX TILEDATA is a very large XSAVE feature. It could have caused nasty space XSAVE buffer space waste in two places: * Signal stacks * Kernel task_struct->fpu buffers

To avoid this waste, neither of these buffers have AMX state by default. The non-default features are called "dynamic" features.

There is an arch_prctl(ARCH_REQ_XCOMP_PERM) which allows a task to declare that it wants to use AMX or other "dynamic" XSAVE features. This arch_prctl() ensures that sufficient sigaltstack space is available before it will succeed. It also expands the task_struct buffer.

Functions of this test: * Test arch_prctl(ARCH_REQ_XCOMP_PERM). Ensure that it checks for proper sigaltstack sizing and that the sizing is enforced for future sigaltstack calls. * Ensure that ARCH_REQ_XCOMP_PERM is inherited across fork() * Ensure that TILEDATA use before the prctl() is fatal * Ensure that TILEDATA is cleared across fork()

Note: Generally, compiler support is needed to do something with AMX. Instead, directly load AMX state from userspace with a plain XSAVE. Do not depend on the compiler.

[ dhansen: bunches of cleanups ]

Signed-off-by: Chang S. Bae chang.seok.bae@intel.com Signed-off-by: Dave Hansen dave.hansen@linux.intel.com Cc: x86@kernel.org Cc: linux-kernel@vger.kernel.org Cc: linux-kselftest@vger.kernel.org

---

b/tools/testing/selftests/x86/Makefile | 2 b/tools/testing/selftests/x86/amx.c | 697 +++++++++++++++++++++++++++++++++ 2 files changed, 698 insertions(+), 1 deletion(-)

diff -puN /dev/null tools/testing/selftests/x86/amx.c --- /dev/null 2021-10-02 19:44:55.430567491 -0700 +++ b/tools/testing/selftests/x86/amx.c 2021-10-15 13:56:39.829901810 -0700 @@ -0,0 +1,697 @@ +// SPDX-License-Identifier: GPL-2.0 + +#define _GNU_SOURCE +#include <err.h> +#include <errno.h> +#include <setjmp.h> +#include <stdio.h> +#include <string.h> +#include <stdbool.h> +#include <unistd.h> +#include <x86intrin.h> + +#include <linux/futex.h> + +#include <sys/auxv.h> +#include <sys/mman.h> +#include <sys/shm.h> +#include <sys/syscall.h> +#include <sys/wait.h> + +#ifndef __x86_64__ +# error This test is 64-bit only +#endif + +#define XSAVE_HDR_OFFSET 512 +#define XSAVE_HDR_SIZE 64 + +struct xsave_buffer { + union { + struct { + char legacy[XSAVE_HDR_OFFSET]; + char header[XSAVE_HDR_SIZE]; + char extended[0]; + }; + char bytes[0]; + }; +}; + +static inline uint64_t xgetbv(uint32_t index) +{ + uint32_t eax, edx; + + asm volatile("xgetbv;" + : "=a" (eax), "=d" (edx) + : "c" (index)); + return eax + ((uint64_t)edx << 32); +} + +static inline void cpuid(uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx) +{ + asm volatile("cpuid;" + : "=a" (*eax), "=b" (*ebx), "=c" (*ecx), "=d" (*edx) + : "0" (*eax), "2" (*ecx)); +} + +static inline void xsave(struct xsave_buffer *xbuf, uint64_t rfbm) +{ + uint32_t rfbm_lo = rfbm; + uint32_t rfbm_hi = rfbm >> 32; + + asm volatile("xsave (%%rdi)" + : : "D" (xbuf), "a" (rfbm_lo), "d" (rfbm_hi) + : "memory"); +} + +static inline void xrstor(struct xsave_buffer *xbuf, uint64_t rfbm) +{ + uint32_t rfbm_lo = rfbm; + uint32_t rfbm_hi = rfbm >> 32; + + asm volatile("xrstor (%%rdi)" + : : "D" (xbuf), "a" (rfbm_lo), "d" (rfbm_hi)); +} + +/* err() exits and will not return */ +#define fatal_error(msg, ...) err(1, "[FAIL]\t" msg, ##__VA_ARGS__) + +static void sethandler(int sig, void (*handler)(int, siginfo_t *, void *), + int flags) +{ + struct sigaction sa; + + memset(&sa, 0, sizeof(sa)); + sa.sa_sigaction = handler; + sa.sa_flags = SA_SIGINFO | flags; + sigemptyset(&sa.sa_mask); + if (sigaction(sig, &sa, 0)) + fatal_error("sigaction"); +} + +static void clearhandler(int sig) +{ + struct sigaction sa; + + memset(&sa, 0, sizeof(sa)); + sa.sa_handler = SIG_DFL; + sigemptyset(&sa.sa_mask); + if (sigaction(sig, &sa, 0)) + fatal_error("sigaction"); +} + +#define XFEATURE_XTILECFG 17 +#define XFEATURE_XTILEDATA 18 +#define XFEATURE_MASK_XTILECFG (1 << XFEATURE_XTILECFG) +#define XFEATURE_MASK_XTILEDATA (1 << XFEATURE_XTILEDATA) +#define XFEATURE_MASK_XTILE (XFEATURE_MASK_XTILECFG | XFEATURE_MASK_XTILEDATA) + +#define CPUID_LEAF1_ECX_XSAVE_MASK (1 << 26) +#define CPUID_LEAF1_ECX_OSXSAVE_MASK (1 << 27) +static inline void check_cpuid_xsave(void) +{ + uint32_t eax, ebx, ecx, edx; + + /* + * CPUID.1:ECX.XSAVE[bit 26] enumerates general + * support for the XSAVE feature set, including + * XGETBV. + */ + eax = 1; + ecx = 0; + cpuid(&eax, &ebx, &ecx, &edx); + if (!(ecx & CPUID_LEAF1_ECX_XSAVE_MASK)) + fatal_error("cpuid: no CPU xsave support"); + if (!(ecx & CPUID_LEAF1_ECX_OSXSAVE_MASK)) + fatal_error("cpuid: no OS xsave support"); +} + +static uint32_t xbuf_size; + +static struct { + uint32_t xbuf_offset; + uint32_t size; +} xtiledata; + +#define CPUID_LEAF_XSTATE 0xd +#define CPUID_SUBLEAF_XSTATE_USER 0x0 +#define TILE_CPUID 0x1d +#define TILE_PALETTE_ID 0x1 + +static void check_cpuid_xtiledata(void) +{ + uint32_t eax, ebx, ecx, edx; + + eax = CPUID_LEAF_XSTATE; + ecx = CPUID_SUBLEAF_XSTATE_USER; + cpuid(&eax, &ebx, &ecx, &edx); + + /* + * EBX enumerates the size (in bytes) required by the XSAVE + * instruction for an XSAVE area containing all the user state + * components corresponding to bits currently set in XCR0. + * + * Stash that off so it can be used to allocate buffers later. + */ + xbuf_size = ebx; + + eax = CPUID_LEAF_XSTATE; + ecx = XFEATURE_XTILEDATA; + + cpuid(&eax, &ebx, &ecx, &edx); + /* + * eax: XTILEDATA state component size + * ebx: XTILEDATA state component offset in user buffer + */ + if (!eax || !ebx) + fatal_error("xstate cpuid: invalid tile data size/offset: %d/%d", + eax, ebx); + + xtiledata.size = eax; + xtiledata.xbuf_offset = ebx; +} + +/* The helpers for managing XSAVE buffer and tile states: */ + +struct xsave_buffer *alloc_xbuf(void) +{ + struct xsave_buffer *xbuf; + + /* XSAVE buffer should be 64B-aligned. */ + xbuf = aligned_alloc(64, xbuf_size); + if (!xbuf) + fatal_error("aligned_alloc()"); + return xbuf; +} + +static inline void clear_xstate_header(struct xsave_buffer *buffer) +{ + memset(&buffer->header, 0, sizeof(buffer->header)); +} + +static inline uint64_t get_xstatebv(struct xsave_buffer *buffer) +{ + /* XSTATE_BV is at the beginning of the header: */ + return *(uint64_t *)&buffer->header; +} + +static inline void set_xstatebv(struct xsave_buffer *buffer, uint64_t bv) +{ + /* XSTATE_BV is at the beginning of the header: */ + *(uint64_t *)(&buffer->header) = bv; +} + +static void set_rand_tiledata(struct xsave_buffer *xbuf) +{ + int *ptr = (int *)&xbuf->bytes[xtiledata.xbuf_offset]; + int data; + int i; + + /* + * Ensure that 'data' is never 0. This ensures that + * the registers are never in their initial configuration + * and thus never tracked as being in the init state. + */ + data = rand() | 1; + + for (i = 0; i < xtiledata.size / sizeof(int); i++, ptr++) + *ptr = data; +} + +struct xsave_buffer *stashed_xsave; + +static void init_stashed_xsave(void) +{ + stashed_xsave = alloc_xbuf(); + if (!stashed_xsave) + fatal_error("failed to allocate stashed_xsave\n"); + clear_xstate_header(stashed_xsave); +} + +static void free_stashed_xsave(void) +{ + free(stashed_xsave); +} + +/* See 'struct _fpx_sw_bytes' at sigcontext.h */ +#define SW_BYTES_OFFSET 464 +/* N.B. The struct's field name varies so read from the offset. */ +#define SW_BYTES_BV_OFFSET (SW_BYTES_OFFSET + 8) + +static inline struct _fpx_sw_bytes *get_fpx_sw_bytes(void *buffer) +{ + return (struct _fpx_sw_bytes *)(buffer + SW_BYTES_OFFSET); +} + +static inline uint64_t get_fpx_sw_bytes_features(void *buffer) +{ + return *(uint64_t *)(buffer + SW_BYTES_BV_OFFSET); +} + +/* Work around printf() being unsafe in signals: */ +#define SIGNAL_BUF_LEN 1000 +char signal_message_buffer[SIGNAL_BUF_LEN]; +void sig_print(char *msg) +{ + int left = SIGNAL_BUF_LEN - strlen(signal_message_buffer) - 1; + + strncat(signal_message_buffer, msg, left); +} + +static volatile bool noperm_signaled; +static int noperm_errs; + +/* + * Signal handler for when AMX is used but + * permission has not been obtained. + */ +static void handle_noperm(int sig, siginfo_t *si, void *ctx_void) +{ + ucontext_t *ctx = (ucontext_t *)ctx_void; + void *xbuf = ctx->uc_mcontext.fpregs; + struct _fpx_sw_bytes *sw_bytes; + uint64_t features; + + /* Reset the signal message buffer: */ + signal_message_buffer[0] = '\0'; + sig_print("\tAt SIGILL handler,\n"); + + if (si->si_code != ILL_ILLOPC) { + noperm_errs++; + sig_print("[FAIL]\tInvalid signal code.\n"); + } else { + sig_print("[OK]\tValid signal code (ILL_ILLOPC).\n"); + } + + sw_bytes = get_fpx_sw_bytes(xbuf); + /* + * Without permission, the signal XSAVE buffer should not + * have room for AMX register state (aka. xtiledata). + * Check that the size does not overlap with where xtiledata + * will reside. + * + * This also implies that no state components *PAST* + * XTILEDATA (features >=19) can be present in the buffer. + */ + if (sw_bytes->xstate_size <= xtiledata.xbuf_offset) { + sig_print("[OK]\tValid xstate size\n"); + } else { + noperm_errs++; + sig_print("[FAIL]\tInvalid xstate size\n"); + } + + features = get_fpx_sw_bytes_features(xbuf); + /* + * Without permission, the XTILEDATA feature + * bit should not be set. + */ + if ((features & XFEATURE_MASK_XTILEDATA) == 0) { + sig_print("[OK]\tValid xstate mask\n"); + } else { + noperm_errs++; + sig_print("[FAIL]\tInvalid xstate mask\n"); + } + + noperm_signaled = true; + ctx->uc_mcontext.gregs[REG_RIP] += 3; /* Skip the faulting XRSTOR */ +} + +/* Return true if XRSTOR is successful; otherwise, false. */ +static inline bool xrstor_safe(struct xsave_buffer *xbuf, uint64_t mask) +{ + noperm_signaled = false; + xrstor(xbuf, mask); + + /* Print any messages produced by the signal code: */ + printf("%s", signal_message_buffer); + /* + * Reset the buffer to make sure any future printing + * only outputs new messages: + */ + signal_message_buffer[0] = '\0'; + + if (noperm_errs) + fatal_error("saw %d errors in noperm signal handler\n", noperm_errs); + + return !noperm_signaled; +} + +/* + * Use XRSTOR to populate the XTILEDATA registers with + * random data. + * + * Return true if successful; otherwise, false. + */ +static inline bool load_rand_tiledata(struct xsave_buffer *xbuf) +{ + clear_xstate_header(xbuf); + set_xstatebv(xbuf, XFEATURE_MASK_XTILEDATA); + set_rand_tiledata(xbuf); + return xrstor_safe(xbuf, XFEATURE_MASK_XTILEDATA); +} + +/* Return XTILEDATA to its initial configuration. */ +static inline void init_xtiledata(void) +{ + clear_xstate_header(stashed_xsave); + xrstor_safe(stashed_xsave, XFEATURE_MASK_XTILEDATA); +} + +enum expected_result { FAIL_EXPECTED, SUCCESS_EXPECTED }; + +/* arch_prctl() and sigaltstack() test */ + +#define ARCH_GET_XCOMP_PERM 0x1022 +#define ARCH_REQ_XCOMP_PERM 0x1023 + +static void req_xtiledata_perm(void) +{ + syscall(SYS_arch_prctl, ARCH_REQ_XCOMP_PERM, XFEATURE_XTILEDATA); +} + +static void validate_req_xcomp_perm(enum expected_result exp) +{ + unsigned long bitmask; + long rc; + + rc = syscall(SYS_arch_prctl, ARCH_REQ_XCOMP_PERM, XFEATURE_XTILEDATA); + if (exp == FAIL_EXPECTED) { + if (rc) { + printf("[OK]\tARCH_REQ_XCOMP_PERM saw expected failure..\n"); + return; + } + + fatal_error("ARCH_REQ_XCOMP_PERM saw unexpected success.\n"); + } else if (rc) { + fatal_error("ARCH_REQ_XCOMP_PERM saw unexpected failure.\n"); + } + + rc = syscall(SYS_arch_prctl, ARCH_GET_XCOMP_PERM, &bitmask); + if (rc) { + fatal_error("prctl(ARCH_GET_XCOMP_PERM) error: %ld", rc); + } else if (bitmask & XFEATURE_MASK_XTILE) { + printf("\tARCH_REQ_XCOMP_PERM is successful.\n"); + } +} + +static void validate_xcomp_perm(enum expected_result exp) +{ + bool load_success = load_rand_tiledata(stashed_xsave); + + if (exp == FAIL_EXPECTED) { + if (load_success) { + noperm_errs++; + printf("[FAIL]\tLoad tiledata succeeded.\n"); + } else { + printf("[OK]\tLoad tiledata failed.\n"); + } + } else if (exp == SUCCESS_EXPECTED) { + if (load_success) { + printf("[OK]\tLoad tiledata succeeded.\n"); + } else { + noperm_errs++; + printf("[FAIL]\tLoad tiledata failed.\n"); + } + } +} + +#ifndef AT_MINSIGSTKSZ +# define AT_MINSIGSTKSZ 51 +#endif + +static void *alloc_altstack(unsigned int size) +{ + void *altstack; + + altstack = mmap(NULL, size, PROT_READ | PROT_WRITE, + MAP_PRIVATE | MAP_ANONYMOUS | MAP_STACK, -1, 0); + + if (altstack == MAP_FAILED) + fatal_error("mmap() for altstack"); + + return altstack; +} + +static void setup_altstack(void *addr, unsigned long size, enum expected_result exp) +{ + stack_t ss; + int rc; + + memset(&ss, 0, sizeof(ss)); + ss.ss_size = size; + ss.ss_sp = addr; + + rc = sigaltstack(&ss, NULL); + + if (exp == FAIL_EXPECTED) { + if (rc) { + printf("[OK]\tsigaltstack() failed.\n"); + } else { + fatal_error("sigaltstack() succeeded unexpectedly.\n"); + } + } else if (rc) { + fatal_error("sigaltstack()"); + } +} + +static void test_dynamic_sigaltstack(void) +{ + unsigned int small_size, enough_size; + unsigned long minsigstksz; + void *altstack; + + minsigstksz = getauxval(AT_MINSIGSTKSZ); + printf("\tAT_MINSIGSTKSZ = %lu\n", minsigstksz); + /* + * getauxval() itself can return 0 for failure or + * success. But, in this case, AT_MINSIGSTKSZ + * will always return a >=0 value if implemented. + * Just check for 0. + */ + if (minsigstksz == 0) { + printf("no support for AT_MINSIGSTKSZ, skipping sigaltstack tests\n"); + return; + } + + enough_size = minsigstksz * 2; + + altstack = alloc_altstack(enough_size); + printf("\tAllocate memory for altstack (%u bytes).\n", enough_size); + + /* + * Try setup_altstack() with a size which can not fit + * XTILEDATA. ARCH_REQ_XCOMP_PERM should fail. + */ + small_size = minsigstksz - xtiledata.size; + printf("\tAfter sigaltstack() with small size (%u bytes).\n", small_size); + setup_altstack(altstack, small_size, SUCCESS_EXPECTED); + validate_req_xcomp_perm(FAIL_EXPECTED); + + /* + * Try setup_altstack() with a size derived from + * AT_MINSIGSTKSZ. It should be more than large enough + * and thus ARCH_REQ_XCOMP_PERM should succeed. + */ + printf("\tAfter sigaltstack() with enough size (%u bytes).\n", enough_size); + setup_altstack(altstack, enough_size, SUCCESS_EXPECTED); + validate_req_xcomp_perm(SUCCESS_EXPECTED); + + /* + * Try to coerce setup_altstack() to again accept a + * too-small altstack. This ensures that big-enough + * sigaltstacks can not shrink to a too-small value + * once XTILEDATA permission is established. + */ + printf("\tThen, sigaltstack() with small size (%u bytes).\n", small_size); + setup_altstack(altstack, small_size, FAIL_EXPECTED); +} + +static void test_dynamic_state(void) +{ + pid_t parent, child, grandchild; + + parent = fork(); + if (parent < 0) { + /* fork() failed */ + fatal_error("fork"); + } else if (parent > 0) { + int status; + /* fork() succeeded. Now in the parent. */ + + wait(&status); + if (!WIFEXITED(status) || WEXITSTATUS(status)) + fatal_error("arch_prctl test parent exit"); + return; + } + /* fork() succeeded. Now in the child . */ + + printf("[RUN]\tCheck ARCH_REQ_XCOMP_PERM around process fork() and sigaltack() test.\n"); + + printf("\tFork a child.\n"); + child = fork(); + if (child < 0) { + fatal_error("fork"); + } else if (child > 0) { + int status; + + wait(&status); + if (!WIFEXITED(status) || WEXITSTATUS(status)) + fatal_error("arch_prctl test child exit"); + _exit(0); + } + + /* + * The permission request should fail without an + * XTILEDATA-compatible signal stack + */ + printf("\tTest XCOMP_PERM at child.\n"); + validate_xcomp_perm(FAIL_EXPECTED); + + /* + * Set up an XTILEDATA-compatible signal stack and + * also obtain permission to populate XTILEDATA. + */ + printf("\tTest dynamic sigaltstack at child:\n"); + test_dynamic_sigaltstack(); + + /* Ensure that XTILEDATA can be populated. */ + printf("\tTest XCOMP_PERM again at child.\n"); + validate_xcomp_perm(SUCCESS_EXPECTED); + + printf("\tFork a grandchild.\n"); + grandchild = fork(); + if (grandchild < 0) { + /* fork() failed */ + fatal_error("fork"); + } else if (!grandchild) { + /* fork() succeeded. Now in the (grand)child. */ + printf("\tTest XCOMP_PERM at grandchild.\n"); + + /* + * Ensure that the grandchild inherited + * permission and a compatible sigaltstack: + */ + validate_xcomp_perm(SUCCESS_EXPECTED); + } else { + int status; + /* fork() succeeded. Now in the parent. */ + + wait(&status); + if (!WIFEXITED(status) || WEXITSTATUS(status)) + fatal_error("fork test grandchild"); + } + + _exit(0); +} + +/* + * Save current register state and compare it to @xbuf1.' + * + * Returns false if @xbuf1 matches the registers. + * Returns true if @xbuf1 differs from the registers. + */ +static inline bool __validate_tiledata_regs(struct xsave_buffer *xbuf1) +{ + struct xsave_buffer *xbuf2; + int ret; + + xbuf2 = alloc_xbuf(); + if (!xbuf2) + fatal_error("failed to allocate XSAVE buffer\n"); + + xsave(xbuf2, XFEATURE_MASK_XTILEDATA); + ret = memcmp(&xbuf1->bytes[xtiledata.xbuf_offset], + &xbuf2->bytes[xtiledata.xbuf_offset], + xtiledata.size); + + free(xbuf2); + + if (ret == 0) + return false; + return true; +} + +static inline void validate_tiledata_regs_same(struct xsave_buffer *xbuf) +{ + int ret = __validate_tiledata_regs(xbuf); + + if (ret != 0) + fatal_error("TILEDATA registers changed"); +} + +static inline void validate_tiledata_regs_changed(struct xsave_buffer *xbuf) +{ + int ret = __validate_tiledata_regs(xbuf); + + if (ret == 0) + fatal_error("TILEDATA registers did not change"); +} + +/* tiledata inheritance test */ + +static void test_fork(void) +{ + pid_t child, grandchild; + + child = fork(); + if (child < 0) { + /* fork() failed */ + fatal_error("fork"); + } else if (child > 0) { + /* fork() succeeded. Now in the parent. */ + int status; + + wait(&status); + if (!WIFEXITED(status) || WEXITSTATUS(status)) + fatal_error("fork test child"); + return; + } + /* fork() succeeded. Now in the child. */ + printf("[RUN]\tCheck tile data inheritance.\n\tBefore fork(), load tiledata\n"); + + load_rand_tiledata(stashed_xsave); + + grandchild = fork(); + if (grandchild < 0) { + /* fork() failed */ + fatal_error("fork"); + } else if (grandchild > 0) { + /* fork() succeeded. Still in the first child. */ + int status; + + wait(&status); + if (!WIFEXITED(status) || WEXITSTATUS(status)) + fatal_error("fork test grand child"); + _exit(0); + } + /* fork() succeeded. Now in the (grand)child. */ + + /* + * TILEDATA registers are not preserved across fork(). + * Ensure that their value has changed: + */ + validate_tiledata_regs_changed(stashed_xsave); + + _exit(0); +} + +int main(void) +{ + /* Check hardware availability at first */ + check_cpuid_xsave(); + check_cpuid_xtiledata(); + + init_stashed_xsave(); + sethandler(SIGILL, handle_noperm, 0); + + test_dynamic_state(); + + /* Request permission for the following tests */ + req_xtiledata_perm(); + + test_fork(); + + clearhandler(SIGILL); + free_stashed_xsave(); + + return 0; +} diff -puN tools/testing/selftests/x86/Makefile~amx-selftest-1_3-selftest-x86-amx-Test-cases-for-the-AMX-state-management tools/testing/selftests/x86/Makefile --- a/tools/testing/selftests/x86/Makefile~amx-selftest-1_3-selftest-x86-amx-Test-cases-for-the-AMX-state-management 2021-10-15 13:56:39.829901810 -0700 +++ b/tools/testing/selftests/x86/Makefile 2021-10-15 13:56:39.829901810 -0700 @@ -18,7 +18,7 @@ TARGETS_C_32BIT_ONLY := entry_from_vm86 test_FCMOV test_FCOMI test_FISTTP \ vdso_restorer TARGETS_C_64BIT_ONLY := fsgsbase sysret_rip syscall_numbering \ - corrupt_xstate_header + corrupt_xstate_header amx # Some selftests require 32bit support enabled also on 64bit systems TARGETS_C_32BIT_NEEDED := ldt_gdt ptrace_syscall

_