In preparation for adding SME support update the bulk of the implementation for the vector length configuration prctl() calls to be independent of vector type.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/include/asm/fpsimd.h | 6 ++--- arch/arm64/kernel/fpsimd.c | 47 ++++++++++++++++++--------------- arch/arm64/kernel/ptrace.c | 4 +-- arch/arm64/kvm/reset.c | 8 +++--- 4 files changed, 34 insertions(+), 31 deletions(-)

diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index dbb4b30a5648..cb24385e3632 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -51,8 +51,8 @@ extern void fpsimd_bind_state_to_cpu(struct user_fpsimd_state *state, extern void fpsimd_flush_task_state(struct task_struct *target); extern void fpsimd_save_and_flush_cpu_state(void);

-/* Maximum VL that SVE VL-agnostic software can transparently support */ -#define SVE_VL_ARCH_MAX 0x100 +/* Maximum VL that SVE/SME VL-agnostic software can transparently support */ +#define VL_ARCH_MAX 0x100

/* Offset of FFR in the SVE register dump */ static inline size_t sve_ffr_offset(int vl) @@ -122,7 +122,7 @@ extern void fpsimd_sync_to_sve(struct task_struct *task); extern void sve_sync_to_fpsimd(struct task_struct *task); extern void sve_sync_from_fpsimd_zeropad(struct task_struct *task);

-extern int sve_set_vector_length(struct task_struct *task, +extern int vec_set_vector_length(struct task_struct *task, enum vec_type type, unsigned long vl, unsigned long flags);

extern int sve_set_current_vl(unsigned long arg); diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 23e575c4e580..4a98cc3b1df1 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -632,7 +632,7 @@ void sve_sync_from_fpsimd_zeropad(struct task_struct *task) __fpsimd_to_sve(sst, fst, vq); }

-int sve_set_vector_length(struct task_struct *task, +int vec_set_vector_length(struct task_struct *task, enum vec_type type, unsigned long vl, unsigned long flags) { if (flags & ~(unsigned long)(PR_SVE_VL_INHERIT | @@ -643,33 +643,35 @@ int sve_set_vector_length(struct task_struct *task, return -EINVAL;

/* - * Clamp to the maximum vector length that VL-agnostic SVE code can - * work with. A flag may be assigned in the future to allow setting - * of larger vector lengths without confusing older software. + * Clamp to the maximum vector length that VL-agnostic code + * can work with. A flag may be assigned in the future to + * allow setting of larger vector lengths without confusing + * older software. */ - if (vl > SVE_VL_ARCH_MAX) - vl = SVE_VL_ARCH_MAX; + if (vl > VL_ARCH_MAX) + vl = VL_ARCH_MAX;

- vl = find_supported_vector_length(ARM64_VEC_SVE, vl); + vl = find_supported_vector_length(type, vl);

if (flags & (PR_SVE_VL_INHERIT | PR_SVE_SET_VL_ONEXEC)) - task_set_sve_vl_onexec(task, vl); + task_set_vl_onexec(task, type, vl); else /* Reset VL to system default on next exec: */ - task_set_sve_vl_onexec(task, 0); + task_set_vl_onexec(task, type, 0);

/* Only actually set the VL if not deferred: */ if (flags & PR_SVE_SET_VL_ONEXEC) goto out;

- if (vl == task_get_sve_vl(task)) + if (vl == task_get_vl(task, type)) goto out;

/* * To ensure the FPSIMD bits of the SVE vector registers are preserved, * write any live register state back to task_struct, and convert to a - * non-SVE thread. + * regular FPSIMD thread. Since the vector length can only be changed + * with a syscall we can't be in streaming mode while reconfiguring. */ if (task == current) { get_cpu_fpsimd_context(); @@ -690,10 +692,10 @@ int sve_set_vector_length(struct task_struct *task, */ sve_free(task);

- task_set_sve_vl(task, vl); + task_set_vl(task, type, vl);

out: - update_tsk_thread_flag(task, TIF_SVE_VL_INHERIT, + update_tsk_thread_flag(task, vec_vl_inherit_flag(type), flags & PR_SVE_VL_INHERIT);

return 0; @@ -701,20 +703,21 @@ int sve_set_vector_length(struct task_struct *task,

/* * Encode the current vector length and flags for return. - * This is only required for prctl(): ptrace has separate fields + * This is only required for prctl(): ptrace has separate fields. + * SVE and SME use the same bits for _ONEXEC and _INHERIT. * - * flags are as for sve_set_vector_length(). + * flags are as for vec_set_vector_length(). */ -static int sve_prctl_status(unsigned long flags) +static int vec_prctl_status(enum vec_type type, unsigned long flags) { int ret;

if (flags & PR_SVE_SET_VL_ONEXEC) - ret = task_get_sve_vl_onexec(current); + ret = task_get_vl_onexec(current, type); else - ret = task_get_sve_vl(current); + ret = task_get_vl(current, type);

- if (test_thread_flag(TIF_SVE_VL_INHERIT)) + if (test_thread_flag(vec_vl_inherit_flag(type))) ret |= PR_SVE_VL_INHERIT;

return ret; @@ -732,11 +735,11 @@ int sve_set_current_vl(unsigned long arg) if (!system_supports_sve() || is_compat_task()) return -EINVAL;

- ret = sve_set_vector_length(current, vl, flags); + ret = vec_set_vector_length(current, ARM64_VEC_SVE, vl, flags); if (ret) return ret;

- return sve_prctl_status(flags); + return vec_prctl_status(ARM64_VEC_SVE, flags); }

/* PR_SVE_GET_VL */ @@ -745,7 +748,7 @@ int sve_get_current_vl(void) if (!system_supports_sve() || is_compat_task()) return -EINVAL;

- return sve_prctl_status(0); + return vec_prctl_status(ARM64_VEC_SVE, 0); }

static void vec_probe_vqs(struct vl_info *info, diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c index 88a9034fb9b5..716dde289446 100644 --- a/arch/arm64/kernel/ptrace.c +++ b/arch/arm64/kernel/ptrace.c @@ -812,9 +812,9 @@ static int sve_set(struct task_struct *target,

/* * Apart from SVE_PT_REGS_MASK, all SVE_PT_* flags are consumed by - * sve_set_vector_length(), which will also validate them for us: + * vec_set_vector_length(), which will also validate them for us: */ - ret = sve_set_vector_length(target, header.vl, + ret = vec_set_vector_length(target, ARM64_VEC_SVE, header.vl, ((unsigned long)header.flags & ~SVE_PT_REGS_MASK) << 16); if (ret) goto out; diff --git a/arch/arm64/kvm/reset.c b/arch/arm64/kvm/reset.c index 09cd30a9aafb..0f6741c80226 100644 --- a/arch/arm64/kvm/reset.c +++ b/arch/arm64/kvm/reset.c @@ -52,10 +52,10 @@ int kvm_arm_init_sve(void) * The get_sve_reg()/set_sve_reg() ioctl interface will need * to be extended with multiple register slice support in * order to support vector lengths greater than - * SVE_VL_ARCH_MAX: + * VL_ARCH_MAX: */ - if (WARN_ON(kvm_sve_max_vl > SVE_VL_ARCH_MAX)) - kvm_sve_max_vl = SVE_VL_ARCH_MAX; + if (WARN_ON(kvm_sve_max_vl > VL_ARCH_MAX)) + kvm_sve_max_vl = VL_ARCH_MAX;

/* * Don't even try to make use of vector lengths that @@ -103,7 +103,7 @@ static int kvm_vcpu_finalize_sve(struct kvm_vcpu *vcpu) * set_sve_vls(). Double-check here just to be sure: */ if (WARN_ON(!sve_vl_valid(vl) || vl > sve_max_virtualisable_vl() || - vl > SVE_VL_ARCH_MAX)) + vl > VL_ARCH_MAX)) return -EIO;

buf = kzalloc(SVE_SIG_REGS_SIZE(sve_vq_from_vl(vl)), GFP_KERNEL);

SME introduces a new mode called streaming mode in which the SVE registers have a different vector length. Since the ptrace interface for this is based on the existing SVE interface prepare for supporting this by moving the regset specific configuration into struct and passing that around, allowing these tests to be reused for streaming mode. As we will also have to verify the interoperation of the SVE and streaming SVE regsets don't just iterate over an array.

Signed-off-by: Mark Brown broonie@kernel.org --- tools/testing/selftests/arm64/fp/sve-ptrace.c | 219 ++++++++++++------ 1 file changed, 142 insertions(+), 77 deletions(-)

diff --git a/tools/testing/selftests/arm64/fp/sve-ptrace.c b/tools/testing/selftests/arm64/fp/sve-ptrace.c index c4417bc48d4f..af798b9d232c 100644 --- a/tools/testing/selftests/arm64/fp/sve-ptrace.c +++ b/tools/testing/selftests/arm64/fp/sve-ptrace.c @@ -21,16 +21,37 @@

#include "../../kselftest.h"

-#define VL_TESTS (((SVE_VQ_MAX - SVE_VQ_MIN) + 1) * 3) -#define FPSIMD_TESTS 5 - -#define EXPECTED_TESTS (VL_TESTS + FPSIMD_TESTS) +#define ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0]))

/* <linux/elf.h> and <sys/auxv.h> don't like each other, so: */ #ifndef NT_ARM_SVE #define NT_ARM_SVE 0x405 #endif

+struct vec_type { + const char *name; + unsigned long hwcap_type; + unsigned long hwcap; + int regset; + int prctl_set; +}; + +static const struct vec_type vec_types[] = { + { + .name = "SVE", + .hwcap_type = AT_HWCAP, + .hwcap = HWCAP_SVE, + .regset = NT_ARM_SVE, + .prctl_set = PR_SVE_SET_VL, + }, +}; + +#define VL_TESTS (((SVE_VQ_MAX - SVE_VQ_MIN) + 1) * 3) +#define FLAG_TESTS 2 +#define FPSIMD_TESTS 3 + +#define EXPECTED_TESTS ((VL_TESTS + FLAG_TESTS + FPSIMD_TESTS) * ARRAY_SIZE(vec_types)) + static void fill_buf(char *buf, size_t size) { int i; @@ -59,7 +80,8 @@ static int get_fpsimd(pid_t pid, struct user_fpsimd_state *fpsimd) return ptrace(PTRACE_GETREGSET, pid, NT_PRFPREG, &iov); }

-static struct user_sve_header *get_sve(pid_t pid, void **buf, size_t *size) +static struct user_sve_header *get_sve(pid_t pid, const struct vec_type *type, + void **buf, size_t *size) { struct user_sve_header *sve; void *p; @@ -80,7 +102,7 @@ static struct user_sve_header *get_sve(pid_t pid, void **buf, size_t *size)

iov.iov_base = *buf; iov.iov_len = sz; - if (ptrace(PTRACE_GETREGSET, pid, NT_ARM_SVE, &iov)) + if (ptrace(PTRACE_GETREGSET, pid, type->regset, &iov)) goto error;

sve = *buf; @@ -96,17 +118,18 @@ static struct user_sve_header *get_sve(pid_t pid, void **buf, size_t *size) return NULL; }

-static int set_sve(pid_t pid, const struct user_sve_header *sve) +static int set_sve(pid_t pid, const struct vec_type *type, + const struct user_sve_header *sve) { struct iovec iov;

iov.iov_base = (void *)sve; iov.iov_len = sve->size; - return ptrace(PTRACE_SETREGSET, pid, NT_ARM_SVE, &iov); + return ptrace(PTRACE_SETREGSET, pid, type->regset, &iov); }

/* Validate setting and getting the inherit flag */ -static void ptrace_set_get_inherit(pid_t child) +static void ptrace_set_get_inherit(pid_t child, const struct vec_type *type) { struct user_sve_header sve; struct user_sve_header *new_sve = NULL; @@ -118,9 +141,10 @@ static void ptrace_set_get_inherit(pid_t child) sve.size = sizeof(sve); sve.vl = sve_vl_from_vq(SVE_VQ_MIN); sve.flags = SVE_PT_VL_INHERIT; - ret = set_sve(child, &sve); + ret = set_sve(child, type, &sve); if (ret != 0) { - ksft_test_result_fail("Failed to set SVE_PT_VL_INHERIT\n"); + ksft_test_result_fail("Failed to set %s SVE_PT_VL_INHERIT\n", + type->name); return; }

@@ -128,35 +152,39 @@ static void ptrace_set_get_inherit(pid_t child) * Read back the new register state and verify that we have * set the flags we expected. */ - if (!get_sve(child, (void **)&new_sve, &new_sve_size)) { - ksft_test_result_fail("Failed to read SVE flags\n"); + if (!get_sve(child, type, (void **)&new_sve, &new_sve_size)) { + ksft_test_result_fail("Failed to read %s SVE flags\n", + type->name); return; }

ksft_test_result(new_sve->flags & SVE_PT_VL_INHERIT, - "SVE_PT_VL_INHERIT set\n"); + "%s SVE_PT_VL_INHERIT set\n", type->name);

/* Now clear */ sve.flags &= ~SVE_PT_VL_INHERIT; - ret = set_sve(child, &sve); + ret = set_sve(child, type, &sve); if (ret != 0) { - ksft_test_result_fail("Failed to clear SVE_PT_VL_INHERIT\n"); + ksft_test_result_fail("Failed to clear %s SVE_PT_VL_INHERIT\n", + type->name); return; }

- if (!get_sve(child, (void **)&new_sve, &new_sve_size)) { - ksft_test_result_fail("Failed to read SVE flags\n"); + if (!get_sve(child, type, (void **)&new_sve, &new_sve_size)) { + ksft_test_result_fail("Failed to read %s SVE flags\n", + type->name); return; }

ksft_test_result(!(new_sve->flags & SVE_PT_VL_INHERIT), - "SVE_PT_VL_INHERIT cleared\n"); + "%s SVE_PT_VL_INHERIT cleared\n", type->name);

free(new_sve); }

/* Validate attempting to set the specfied VL via ptrace */ -static void ptrace_set_get_vl(pid_t child, unsigned int vl, bool *supported) +static void ptrace_set_get_vl(pid_t child, const struct vec_type *type, + unsigned int vl, bool *supported) { struct user_sve_header sve; struct user_sve_header *new_sve = NULL; @@ -166,10 +194,10 @@ static void ptrace_set_get_vl(pid_t child, unsigned int vl, bool *supported) *supported = false;

/* Check if the VL is supported in this process */ - prctl_vl = prctl(PR_SVE_SET_VL, vl); + prctl_vl = prctl(type->prctl_set, vl); if (prctl_vl == -1) - ksft_exit_fail_msg("prctl(PR_SVE_SET_VL) failed: %s (%d)\n", - strerror(errno), errno); + ksft_exit_fail_msg("prctl(PR_%s_SET_VL) failed: %s (%d)\n", + type->name, strerror(errno), errno);

/* If the VL is not supported then a supported VL will be returned */ *supported = (prctl_vl == vl); @@ -178,9 +206,10 @@ static void ptrace_set_get_vl(pid_t child, unsigned int vl, bool *supported) memset(&sve, 0, sizeof(sve)); sve.size = sizeof(sve); sve.vl = vl; - ret = set_sve(child, &sve); + ret = set_sve(child, type, &sve); if (ret != 0) { - ksft_test_result_fail("Failed to set VL %u\n", vl); + ksft_test_result_fail("Failed to set %s VL %u\n", + type->name, vl); return; }

@@ -188,12 +217,14 @@ static void ptrace_set_get_vl(pid_t child, unsigned int vl, bool *supported) * Read back the new register state and verify that we have the * same VL that we got from prctl() on ourselves. */ - if (!get_sve(child, (void **)&new_sve, &new_sve_size)) { - ksft_test_result_fail("Failed to read VL %u\n", vl); + if (!get_sve(child, type, (void **)&new_sve, &new_sve_size)) { + ksft_test_result_fail("Failed to read %s VL %u\n", + type->name, vl); return; }

- ksft_test_result(new_sve->vl = prctl_vl, "Set VL %u\n", vl); + ksft_test_result(new_sve->vl = prctl_vl, "Set %s VL %u\n", + type->name, vl);

free(new_sve); } @@ -209,7 +240,7 @@ static void check_u32(unsigned int vl, const char *reg, }

/* Access the FPSIMD registers via the SVE regset */ -static void ptrace_sve_fpsimd(pid_t child) +static void ptrace_sve_fpsimd(pid_t child, const struct vec_type *type) { void *svebuf = NULL; size_t svebufsz = 0; @@ -219,17 +250,18 @@ static void ptrace_sve_fpsimd(pid_t child) unsigned char *p;

/* New process should start with FPSIMD registers only */ - sve = get_sve(child, &svebuf, &svebufsz); + sve = get_sve(child, type, &svebuf, &svebufsz); if (!sve) { - ksft_test_result_fail("get_sve: %s\n", strerror(errno)); + ksft_test_result_fail("get_sve(%s): %s\n", + type->name, strerror(errno));

return; } else { - ksft_test_result_pass("get_sve(FPSIMD)\n"); + ksft_test_result_pass("get_sve(%s FPSIMD)\n", type->name); }

ksft_test_result((sve->flags & SVE_PT_REGS_MASK) == SVE_PT_REGS_FPSIMD, - "Set FPSIMD registers\n"); + "Set FPSIMD registers via %s\n", type->name); if ((sve->flags & SVE_PT_REGS_MASK) != SVE_PT_REGS_FPSIMD) goto out;

@@ -243,9 +275,9 @@ static void ptrace_sve_fpsimd(pid_t child) p[j] = j; }

- if (set_sve(child, sve)) { - ksft_test_result_fail("set_sve(FPSIMD): %s\n", - strerror(errno)); + if (set_sve(child, type, sve)) { + ksft_test_result_fail("set_sve(%s FPSIMD): %s\n", + type->name, strerror(errno));

goto out; } @@ -257,16 +289,20 @@ static void ptrace_sve_fpsimd(pid_t child) goto out; } if (memcmp(fpsimd, &new_fpsimd, sizeof(*fpsimd)) == 0) - ksft_test_result_pass("get_fpsimd() gave same state\n"); + ksft_test_result_pass("%s get_fpsimd() gave same state\n", + type->name); else - ksft_test_result_fail("get_fpsimd() gave different state\n"); + ksft_test_result_fail("%s get_fpsimd() gave different state\n", + type->name);

out: free(svebuf); }

/* Validate attempting to set SVE data and read SVE data */ -static void ptrace_set_sve_get_sve_data(pid_t child, unsigned int vl) +static void ptrace_set_sve_get_sve_data(pid_t child, + const struct vec_type *type, + unsigned int vl) { void *write_buf; void *read_buf = NULL; @@ -281,8 +317,8 @@ static void ptrace_set_sve_get_sve_data(pid_t child, unsigned int vl) data_size = SVE_PT_SVE_OFFSET + SVE_PT_SVE_SIZE(vq, SVE_PT_REGS_SVE); write_buf = malloc(data_size); if (!write_buf) { - ksft_test_result_fail("Error allocating %d byte buffer for VL %u\n", - data_size, vl); + ksft_test_result_fail("Error allocating %d byte buffer for %s VL %u\n", + data_size, type->name, vl); return; } write_sve = write_buf; @@ -306,23 +342,26 @@ static void ptrace_set_sve_get_sve_data(pid_t child, unsigned int vl)

/* TODO: Generate a valid FFR pattern */

- ret = set_sve(child, write_sve); + ret = set_sve(child, type, write_sve); if (ret != 0) { - ksft_test_result_fail("Failed to set VL %u data\n", vl); + ksft_test_result_fail("Failed to set %s VL %u data\n", + type->name, vl); goto out; }

/* Read the data back */ - if (!get_sve(child, (void **)&read_buf, &read_sve_size)) { - ksft_test_result_fail("Failed to read VL %u data\n", vl); + if (!get_sve(child, type, (void **)&read_buf, &read_sve_size)) { + ksft_test_result_fail("Failed to read %s VL %u data\n", + type->name, vl); goto out; } read_sve = read_buf;

/* We might read more data if there's extensions we don't know */ if (read_sve->size < write_sve->size) { - ksft_test_result_fail("Wrote %d bytes, only read %d\n", - write_sve->size, read_sve->size); + ksft_test_result_fail("%s wrote %d bytes, only read %d\n", + type->name, write_sve->size, + read_sve->size); goto out_read; }

@@ -349,7 +388,8 @@ static void ptrace_set_sve_get_sve_data(pid_t child, unsigned int vl) check_u32(vl, "FPCR", write_buf + SVE_PT_SVE_FPCR_OFFSET(vq), read_buf + SVE_PT_SVE_FPCR_OFFSET(vq), &errors);

- ksft_test_result(errors == 0, "Set and get SVE data for VL %u\n", vl); + ksft_test_result(errors == 0, "Set and get %s data for VL %u\n", + type->name, vl);

out_read: free(read_buf); @@ -358,7 +398,9 @@ static void ptrace_set_sve_get_sve_data(pid_t child, unsigned int vl) }

/* Validate attempting to set SVE data and read SVE data */ -static void ptrace_set_sve_get_fpsimd_data(pid_t child, unsigned int vl) +static void ptrace_set_sve_get_fpsimd_data(pid_t child, + const struct vec_type *type, + unsigned int vl) { void *write_buf; struct user_sve_header *write_sve; @@ -376,8 +418,8 @@ static void ptrace_set_sve_get_fpsimd_data(pid_t child, unsigned int vl) data_size = SVE_PT_SVE_OFFSET + SVE_PT_SVE_SIZE(vq, SVE_PT_REGS_SVE); write_buf = malloc(data_size); if (!write_buf) { - ksft_test_result_fail("Error allocating %d byte buffer for VL %u\n", - data_size, vl); + ksft_test_result_fail("Error allocating %d byte buffer for %s VL %u\n", + data_size, type->name, vl); return; } write_sve = write_buf; @@ -395,16 +437,17 @@ static void ptrace_set_sve_get_fpsimd_data(pid_t child, unsigned int vl) fill_buf(write_buf + SVE_PT_SVE_FPSR_OFFSET(vq), SVE_PT_SVE_FPSR_SIZE); fill_buf(write_buf + SVE_PT_SVE_FPCR_OFFSET(vq), SVE_PT_SVE_FPCR_SIZE);

- ret = set_sve(child, write_sve); + ret = set_sve(child, type, write_sve); if (ret != 0) { - ksft_test_result_fail("Failed to set VL %u data\n", vl); + ksft_test_result_fail("Failed to set %s VL %u data\n", + type->name, vl); goto out; }

/* Read the data back */ if (get_fpsimd(child, &fpsimd_state)) { - ksft_test_result_fail("Failed to read VL %u FPSIMD data\n", - vl); + ksft_test_result_fail("Failed to read %s VL %u FPSIMD data\n", + type->name, vl); goto out; }

@@ -419,7 +462,8 @@ static void ptrace_set_sve_get_fpsimd_data(pid_t child, unsigned int vl) sizeof(tmp));

if (tmp != fpsimd_state.vregs[i]) { - printf("# Mismatch in FPSIMD for VL %u Z%d\n", vl, i); + printf("# Mismatch in FPSIMD for %s VL %u Z%d\n", + type->name, vl, i); errors++; } } @@ -429,8 +473,8 @@ static void ptrace_set_sve_get_fpsimd_data(pid_t child, unsigned int vl) check_u32(vl, "FPCR", write_buf + SVE_PT_SVE_FPCR_OFFSET(vq), &fpsimd_state.fpcr, &errors);

- ksft_test_result(errors == 0, "Set and get FPSIMD data for VL %u\n", - vl); + ksft_test_result(errors == 0, "Set and get FPSIMD data for %s VL %u\n", + type->name, vl);

out: free(write_buf); @@ -440,7 +484,7 @@ static int do_parent(pid_t child) { int ret = EXIT_FAILURE; pid_t pid; - int status; + int status, i; siginfo_t si; unsigned int vq, vl; bool vl_supported; @@ -499,26 +543,47 @@ static int do_parent(pid_t child) } }

- /* FPSIMD via SVE regset */ - ptrace_sve_fpsimd(child); - - /* prctl() flags */ - ptrace_set_get_inherit(child); - - /* Step through every possible VQ */ - for (vq = SVE_VQ_MIN; vq <= SVE_VQ_MAX; vq++) { - vl = sve_vl_from_vq(vq); + for (i = 0; i < ARRAY_SIZE(vec_types); i++) { + /* FPSIMD via SVE regset */ + if (getauxval(vec_types[i].hwcap_type) & vec_types[i].hwcap) { + ptrace_sve_fpsimd(child, &vec_types[i]); + } else { + ksft_test_result_skip("%s FPSIMD get via SVE\n", + vec_types[i].name); + ksft_test_result_skip("%s FPSIMD set via SVE\n", + vec_types[i].name); + ksft_test_result_skip("%s set read via FPSIMD\n", + vec_types[i].name); + }

- /* First, try to set this vector length */ - ptrace_set_get_vl(child, vl, &vl_supported); + /* prctl() flags */ + ptrace_set_get_inherit(child, &vec_types[i]); + + /* Step through every possible VQ */ + for (vq = SVE_VQ_MIN; vq <= SVE_VQ_MAX; vq++) { + vl = sve_vl_from_vq(vq); + + /* First, try to set this vector length */ + if (getauxval(vec_types[i].hwcap_type) & + vec_types[i].hwcap) { + ptrace_set_get_vl(child, &vec_types[i], vl, + &vl_supported); + } else { + ksft_test_result_skip("%s get/set VL %d\n", + vec_types[i].name, vl); + vl_supported = false; + }

- /* If the VL is supported validate data set/get */ - if (vl_supported) { - ptrace_set_sve_get_sve_data(child, vl); - ptrace_set_sve_get_fpsimd_data(child, vl); - } else { - ksft_test_result_skip("set SVE get SVE for VL %d\n", vl); - ksft_test_result_skip("set SVE get FPSIMD for VL %d\n", vl); + /* If the VL is supported validate data set/get */ + if (vl_supported) { + ptrace_set_sve_get_sve_data(child, &vec_types[i], vl); + ptrace_set_sve_get_fpsimd_data(child, &vec_types[i], vl); + } else { + ksft_test_result_skip("%s set SVE get SVE for VL %d\n", + vec_types[i].name, vl); + ksft_test_result_skip("%s set SVE get FPSIMD for VL %d\n", + vec_types[i].name, vl); + } } }

Currently we don't have any coverage of the syscall ABI so let's add a very dumb test program which sets up register patterns, does a sysscall and then checks that the register state after the syscall matches what we expect. The program is written in an extremely simplistic fashion with the goal of making it easy to verify that it's doing what it thinks it's doing, it is not a model of how one should write actual code.

Currently we validate the general purpose, FPSIMD and SVE registers. There are other thing things that could be covered like FPCR and flags registers, these can be covered incrementally - my main focus at the minute is covering the ABI for the SVE registers.

The program repeats the tests for all possible SVE vector lengths in case some vector length specific optimisation causes issues, as well as testing FPSIMD only. It tries two syscalls, getpid() and sched_yield(), in an effort to cover both immediate return to userspace and scheduling another task though there are no guarantees which cases will be hit.

A new test directory "abi" is added to hold the test, it doesn't seem to fit well into any of the existing directories.

Signed-off-by: Mark Brown broonie@kernel.org --- tools/testing/selftests/arm64/Makefile | 2 +- tools/testing/selftests/arm64/abi/.gitignore | 1 + tools/testing/selftests/arm64/abi/Makefile | 8 + .../selftests/arm64/abi/syscall-abi-asm.S | 240 +++++++++++++ .../testing/selftests/arm64/abi/syscall-abi.c | 325 ++++++++++++++++++ 5 files changed, 575 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/arm64/abi/.gitignore create mode 100644 tools/testing/selftests/arm64/abi/Makefile create mode 100644 tools/testing/selftests/arm64/abi/syscall-abi-asm.S create mode 100644 tools/testing/selftests/arm64/abi/syscall-abi.c

diff --git a/tools/testing/selftests/arm64/Makefile b/tools/testing/selftests/arm64/Makefile index ced910fb4019..1e8d9a8f59df 100644 --- a/tools/testing/selftests/arm64/Makefile +++ b/tools/testing/selftests/arm64/Makefile @@ -4,7 +4,7 @@ ARCH ?= $(shell uname -m 2>/dev/null || echo not)

ifneq (,$(filter $(ARCH),aarch64 arm64)) -ARM64_SUBTARGETS ?= tags signal pauth fp mte bti +ARM64_SUBTARGETS ?= tags signal pauth fp mte bti abi else ARM64_SUBTARGETS := endif diff --git a/tools/testing/selftests/arm64/abi/.gitignore b/tools/testing/selftests/arm64/abi/.gitignore new file mode 100644 index 000000000000..b79cf5814c23 --- /dev/null +++ b/tools/testing/selftests/arm64/abi/.gitignore @@ -0,0 +1 @@ +syscall-abi diff --git a/tools/testing/selftests/arm64/abi/Makefile b/tools/testing/selftests/arm64/abi/Makefile new file mode 100644 index 000000000000..96eba974ac8d --- /dev/null +++ b/tools/testing/selftests/arm64/abi/Makefile @@ -0,0 +1,8 @@ +# SPDX-License-Identifier: GPL-2.0 +# Copyright (C) 2021 ARM Limited + +TEST_GEN_PROGS := syscall-abi + +include ../../lib.mk + +$(OUTPUT)/syscall-abi: syscall-abi.c syscall-abi-asm.S diff --git a/tools/testing/selftests/arm64/abi/syscall-abi-asm.S b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S new file mode 100644 index 000000000000..983467cfcee0 --- /dev/null +++ b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S @@ -0,0 +1,240 @@ +// SPDX-License-Identifier: GPL-2.0-only +// Copyright (C) 2021 ARM Limited. +// +// Assembly portion of the syscall ABI test + +// +// Load values from memory into registers, invoke a syscall and save the +// register values back to memory for later checking. The syscall to be +// invoked is configured in x8 of the input GPR data. +// +// x0: SVE VL, 0 for FP only +// +// GPRs: gpr_in, gpr_out +// FPRs: fpr_in, fpr_out +// Zn: z_in, z_out +// Pn: p_in, p_out +// FFR: ffr_in, ffr_out + +.arch_extension sve + +.globl do_syscall +do_syscall: + // Store callee saved registers x19-x29 (80 bytes) plus x0 and x1 + stp x29, x30, [sp, #-112]! + mov x29, sp + stp x0, x1, [sp, #16] + stp x19, x20, [sp, #32] + stp x21, x22, [sp, #48] + stp x23, x24, [sp, #64] + stp x25, x26, [sp, #80] + stp x27, x28, [sp, #96] + + // Load GPRs x8-x28, and save our SP/FP for later comparison + ldr x2, =gpr_in + add x2, x2, #64 + ldp x8, x9, [x2], #16 + ldp x10, x11, [x2], #16 + ldp x12, x13, [x2], #16 + ldp x14, x15, [x2], #16 + ldp x16, x17, [x2], #16 + ldp x18, x19, [x2], #16 + ldp x20, x21, [x2], #16 + ldp x22, x23, [x2], #16 + ldp x24, x25, [x2], #16 + ldp x26, x27, [x2], #16 + ldr x28, [x2], #8 + str x29, [x2], #8 // FP + str x30, [x2], #8 // LR + + // Load FPRs if we're not doing SVE + cbnz x0, 1f + ldr x2, =fpr_in + ldp q0, q1, [x2] + ldp q2, q3, [x2, #16 * 2] + ldp q4, q5, [x2, #16 * 4] + ldp q6, q7, [x2, #16 * 6] + ldp q8, q9, [x2, #16 * 8] + ldp q10, q11, [x2, #16 * 10] + ldp q12, q13, [x2, #16 * 12] + ldp q14, q15, [x2, #16 * 14] + ldp q16, q17, [x2, #16 * 16] + ldp q18, q19, [x2, #16 * 18] + ldp q20, q21, [x2, #16 * 20] + ldp q22, q23, [x2, #16 * 22] + ldp q24, q25, [x2, #16 * 24] + ldp q26, q27, [x2, #16 * 26] + ldp q28, q29, [x2, #16 * 28] + ldp q30, q31, [x2, #16 * 30] +1: + + // Load the SVE registers if we're doing SVE + cbz x0, 1f + + ldr x2, =z_in + ldr z0, [x2, #0, MUL VL] + ldr z1, [x2, #1, MUL VL] + ldr z2, [x2, #2, MUL VL] + ldr z3, [x2, #3, MUL VL] + ldr z4, [x2, #4, MUL VL] + ldr z5, [x2, #5, MUL VL] + ldr z6, [x2, #6, MUL VL] + ldr z7, [x2, #7, MUL VL] + ldr z8, [x2, #8, MUL VL] + ldr z9, [x2, #9, MUL VL] + ldr z10, [x2, #10, MUL VL] + ldr z11, [x2, #11, MUL VL] + ldr z12, [x2, #12, MUL VL] + ldr z13, [x2, #13, MUL VL] + ldr z14, [x2, #14, MUL VL] + ldr z15, [x2, #15, MUL VL] + ldr z16, [x2, #16, MUL VL] + ldr z17, [x2, #17, MUL VL] + ldr z18, [x2, #18, MUL VL] + ldr z19, [x2, #19, MUL VL] + ldr z20, [x2, #20, MUL VL] + ldr z21, [x2, #21, MUL VL] + ldr z22, [x2, #22, MUL VL] + ldr z23, [x2, #23, MUL VL] + ldr z24, [x2, #24, MUL VL] + ldr z25, [x2, #25, MUL VL] + ldr z26, [x2, #26, MUL VL] + ldr z27, [x2, #27, MUL VL] + ldr z28, [x2, #28, MUL VL] + ldr z29, [x2, #29, MUL VL] + ldr z30, [x2, #30, MUL VL] + ldr z31, [x2, #31, MUL VL] + + ldr x2, =ffr_in + ldr p0, [x2, #0] + wrffr p0.b + + ldr x2, =p_in + ldr p0, [x2, #0, MUL VL] + ldr p1, [x2, #1, MUL VL] + ldr p2, [x2, #2, MUL VL] + ldr p3, [x2, #3, MUL VL] + ldr p4, [x2, #4, MUL VL] + ldr p5, [x2, #5, MUL VL] + ldr p6, [x2, #6, MUL VL] + ldr p7, [x2, #7, MUL VL] + ldr p8, [x2, #8, MUL VL] + ldr p9, [x2, #9, MUL VL] + ldr p10, [x2, #10, MUL VL] + ldr p11, [x2, #11, MUL VL] + ldr p12, [x2, #12, MUL VL] + ldr p13, [x2, #13, MUL VL] + ldr p14, [x2, #14, MUL VL] + ldr p15, [x2, #15, MUL VL] +1: + + // Do the syscall + svc #0 + + // Save GPRs x8-x30 + ldr x2, =gpr_out + add x2, x2, #64 + stp x8, x9, [x2], #16 + stp x10, x11, [x2], #16 + stp x12, x13, [x2], #16 + stp x14, x15, [x2], #16 + stp x16, x17, [x2], #16 + stp x18, x19, [x2], #16 + stp x20, x21, [x2], #16 + stp x22, x23, [x2], #16 + stp x24, x25, [x2], #16 + stp x26, x27, [x2], #16 + stp x28, x29, [x2], #16 + str x30, [x2] + + // Restore x0 and x1 for feature checks + ldp x0, x1, [sp, #16] + + // Save FPSIMD state + ldr x2, =fpr_out + stp q0, q1, [x2] + stp q2, q3, [x2, #16 * 2] + stp q4, q5, [x2, #16 * 4] + stp q6, q7, [x2, #16 * 6] + stp q8, q9, [x2, #16 * 8] + stp q10, q11, [x2, #16 * 10] + stp q12, q13, [x2, #16 * 12] + stp q14, q15, [x2, #16 * 14] + stp q16, q17, [x2, #16 * 16] + stp q18, q19, [x2, #16 * 18] + stp q20, q21, [x2, #16 * 20] + stp q22, q23, [x2, #16 * 22] + stp q24, q25, [x2, #16 * 24] + stp q26, q27, [x2, #16 * 26] + stp q28, q29, [x2, #16 * 28] + stp q30, q31, [x2, #16 * 30] + + // Save the SVE state if we have some + cbz x0, 1f + + ldr x2, =z_out + str z0, [x2, #0, MUL VL] + str z1, [x2, #1, MUL VL] + str z2, [x2, #2, MUL VL] + str z3, [x2, #3, MUL VL] + str z4, [x2, #4, MUL VL] + str z5, [x2, #5, MUL VL] + str z6, [x2, #6, MUL VL] + str z7, [x2, #7, MUL VL] + str z8, [x2, #8, MUL VL] + str z9, [x2, #9, MUL VL] + str z10, [x2, #10, MUL VL] + str z11, [x2, #11, MUL VL] + str z12, [x2, #12, MUL VL] + str z13, [x2, #13, MUL VL] + str z14, [x2, #14, MUL VL] + str z15, [x2, #15, MUL VL] + str z16, [x2, #16, MUL VL] + str z17, [x2, #17, MUL VL] + str z18, [x2, #18, MUL VL] + str z19, [x2, #19, MUL VL] + str z20, [x2, #20, MUL VL] + str z21, [x2, #21, MUL VL] + str z22, [x2, #22, MUL VL] + str z23, [x2, #23, MUL VL] + str z24, [x2, #24, MUL VL] + str z25, [x2, #25, MUL VL] + str z26, [x2, #26, MUL VL] + str z27, [x2, #27, MUL VL] + str z28, [x2, #28, MUL VL] + str z29, [x2, #29, MUL VL] + str z30, [x2, #30, MUL VL] + str z31, [x2, #31, MUL VL] + + ldr x2, =p_out + str p0, [x2, #0, MUL VL] + str p1, [x2, #1, MUL VL] + str p2, [x2, #2, MUL VL] + str p3, [x2, #3, MUL VL] + str p4, [x2, #4, MUL VL] + str p5, [x2, #5, MUL VL] + str p6, [x2, #6, MUL VL] + str p7, [x2, #7, MUL VL] + str p8, [x2, #8, MUL VL] + str p9, [x2, #9, MUL VL] + str p10, [x2, #10, MUL VL] + str p11, [x2, #11, MUL VL] + str p12, [x2, #12, MUL VL] + str p13, [x2, #13, MUL VL] + str p14, [x2, #14, MUL VL] + str p15, [x2, #15, MUL VL] + + ldr x2, =ffr_out + rdffr p0.b + str p0, [x2, #0] +1: + + // Restore callee saved registers x19-x30 + ldp x19, x20, [sp, #32] + ldp x21, x22, [sp, #48] + ldp x23, x24, [sp, #64] + ldp x25, x26, [sp, #80] + ldp x27, x28, [sp, #96] + ldp x29, x30, [sp], #112 + + ret diff --git a/tools/testing/selftests/arm64/abi/syscall-abi.c b/tools/testing/selftests/arm64/abi/syscall-abi.c new file mode 100644 index 000000000000..d103acf1ab79 --- /dev/null +++ b/tools/testing/selftests/arm64/abi/syscall-abi.c @@ -0,0 +1,325 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (C) 2021 ARM Limited. + */ + +#include <errno.h> +#include <stdbool.h> +#include <stddef.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <unistd.h> +#include <sys/auxv.h> +#include <sys/prctl.h> +#include <asm/hwcap.h> +#include <asm/sigcontext.h> +#include <asm/unistd.h> + +#include "../../kselftest.h" + +#define ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0])) +#define NUM_VL ((SVE_VQ_MAX - SVE_VQ_MIN) + 1) + +extern void do_syscall(int sve_vl); + +static void fill_random(void *buf, size_t size) +{ + int i; + uint32_t *lbuf = buf; + + /* random() returns a 32 bit number regardless of the size of long */ + for (i = 0; i < size / sizeof(uint32_t); i++) + lbuf[i] = random(); +} + +/* + * We also repeat the test for several syscalls to try to expose different + * behaviour. + */ +static struct syscall_cfg { + int syscall_nr; + const char *name; +} syscalls[] = { + { __NR_getpid, "getpid()" }, + { __NR_sched_yield, "sched_yield()" }, +}; + +#define NUM_GPR 31 +uint64_t gpr_in[NUM_GPR]; +uint64_t gpr_out[NUM_GPR]; + +static void setup_gpr(struct syscall_cfg *cfg, int sve_vl) +{ + fill_random(gpr_in, sizeof(gpr_in)); + gpr_in[8] = cfg->syscall_nr; + memset(gpr_out, 0, sizeof(gpr_out)); +} + +static int check_gpr(struct syscall_cfg *cfg, int sve_vl) +{ + int errors = 0; + int i; + + /* + * GPR x0-x7 may be clobbered, and all others should be preserved. + */ + for (i = 9; i < ARRAY_SIZE(gpr_in); i++) { + if (gpr_in[i] != gpr_out[i]) { + ksft_print_msg("%s SVE VL %d mismatch in GPR %d: %llx != %llx\n", + cfg->name, sve_vl, i, + gpr_in[i], gpr_out[i]); + errors++; + } + } + + return errors; +} + +#define NUM_FPR 32 +uint64_t fpr_in[NUM_FPR * 2]; +uint64_t fpr_out[NUM_FPR * 2]; + +static void setup_fpr(struct syscall_cfg *cfg, int sve_vl) +{ + fill_random(fpr_in, sizeof(fpr_in)); + memset(fpr_out, 0, sizeof(fpr_out)); +} + +static int check_fpr(struct syscall_cfg *cfg, int sve_vl) +{ + int errors = 0; + int i; + + if (!sve_vl) { + for (i = 0; i < ARRAY_SIZE(fpr_in); i++) { + if (fpr_in[i] != fpr_out[i]) { + ksft_print_msg("%s Q%d/%d mismatch %llx != %llx\n", + cfg->name, + i / 2, i % 2, + fpr_in[i], fpr_out[i]); + errors++; + } + } + } + + return errors; +} + +static uint8_t z_zero[__SVE_ZREG_SIZE(SVE_VQ_MAX)]; +uint8_t z_in[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; +uint8_t z_out[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; + +static void setup_z(struct syscall_cfg *cfg, int sve_vl) +{ + fill_random(z_in, sizeof(z_in)); + fill_random(z_out, sizeof(z_out)); +} + +static int check_z(struct syscall_cfg *cfg, int sve_vl) +{ + size_t reg_size = sve_vl; + int errors = 0; + int i; + + if (!sve_vl) + return 0; + + /* + * After a syscall the low 128 bits of the Z registers should + * be preserved and the rest be zeroed. + */ + for (i = 0; i < SVE_NUM_ZREGS; i++) { + void *in = &z_in[reg_size * i]; + void *out = &z_out[reg_size * i]; + + if (memcmp(in, out, SVE_VQ_BYTES) != 0) { + ksft_print_msg("%s SVE VL %d Z%d low 128 bits changed\n", + cfg->name, sve_vl, i); + errors++; + } + + memset(out, 0, SVE_VQ_BYTES); + if (memcmp(out, z_zero, reg_size) != 0) { + ksft_print_msg("%s SVE VL %d Z%d high bits set\n", + cfg->name, sve_vl, i); + errors++; + } + } + + return errors; +} + +uint8_t p_in[SVE_NUM_PREGS * __SVE_PREG_SIZE(SVE_VQ_MAX)]; +uint8_t p_out[SVE_NUM_PREGS * __SVE_PREG_SIZE(SVE_VQ_MAX)]; + +static void setup_p(struct syscall_cfg *cfg, int sve_vl) +{ + fill_random(p_in, sizeof(p_in)); + fill_random(p_out, sizeof(p_out)); +} + +static int check_p(struct syscall_cfg *cfg, int sve_vl) +{ + size_t reg_size = sve_vq_from_vl(sve_vl) * 2; /* 1 bit per VL byte */ + + int errors = 0; + int i; + + if (!sve_vl) + return 0; + + /* After a syscall the P registers should be zeroed */ + for (i = 0; i < SVE_NUM_PREGS * reg_size; i++) + if (p_out[i]) + errors++; + if (errors) + ksft_print_msg("%s SVE VL %d predicate registers non-zero\n", + cfg->name, sve_vl); + + return errors; +} + +uint8_t ffr_in[__SVE_PREG_SIZE(SVE_VQ_MAX)]; +uint8_t ffr_out[__SVE_PREG_SIZE(SVE_VQ_MAX)]; + +static void setup_ffr(struct syscall_cfg *cfg, int sve_vl) +{ + /* + * It is only valid to set a contiguous set of bits starting + * at 0. For now since we're expecting this to be cleared by + * a syscall just set all bits. + */ + memset(ffr_in, 0xff, sizeof(ffr_in)); + fill_random(ffr_out, sizeof(ffr_out)); +} + +static int check_ffr(struct syscall_cfg *cfg, int sve_vl) +{ + size_t reg_size = sve_vq_from_vl(sve_vl) * 2; /* 1 bit per VL byte */ + int errors = 0; + int i; + + if (!sve_vl) + return 0; + + /* After a syscall the P registers should be zeroed */ + for (i = 0; i < reg_size; i++) + if (ffr_out[i]) + errors++; + if (errors) + ksft_print_msg("%s SVE VL %d FFR non-zero\n", + cfg->name, sve_vl); + + return errors; +} + +typedef void (*setup_fn)(struct syscall_cfg *cfg, int sve_vl); +typedef int (*check_fn)(struct syscall_cfg *cfg, int sve_vl); + +/* + * Each set of registers has a setup function which is called before + * the syscall to fill values in a global variable for loading by the + * test code and a check function which validates that the results are + * as expected. Vector lengths are passed everywhere, a vector length + * of 0 should be treated as do not test. + */ +static struct { + setup_fn setup; + check_fn check; +} regset[] = { + { setup_gpr, check_gpr }, + { setup_fpr, check_fpr }, + { setup_z, check_z }, + { setup_p, check_p }, + { setup_ffr, check_ffr }, +}; + +static bool do_test(struct syscall_cfg *cfg, int sve_vl) +{ + int errors = 0; + int i; + + for (i = 0; i < ARRAY_SIZE(regset); i++) + regset[i].setup(cfg, sve_vl); + + do_syscall(sve_vl); + + for (i = 0; i < ARRAY_SIZE(regset); i++) + errors += regset[i].check(cfg, sve_vl); + + return errors == 0; +} + +static void test_one_syscall(struct syscall_cfg *cfg) +{ + int sve_vq, sve_vl; + + /* FPSIMD only case */ + ksft_test_result(do_test(cfg, 0), + "%s FPSIMD\n", cfg->name); + + if (!(getauxval(AT_HWCAP) & HWCAP_SVE)) + return; + + for (sve_vq = SVE_VQ_MAX; sve_vq > 0; --sve_vq) { + sve_vl = prctl(PR_SVE_SET_VL, sve_vq * 16); + if (sve_vl == -1) + ksft_exit_fail_msg("PR_SVE_SET_VL failed: %s (%d)\n", + strerror(errno), errno); + + sve_vl &= PR_SVE_VL_LEN_MASK; + + if (sve_vq != sve_vq_from_vl(sve_vl)) + sve_vq = sve_vq_from_vl(sve_vl); + + ksft_test_result(do_test(cfg, sve_vl), + "%s SVE VL %d\n", cfg->name, sve_vl); + } +} + +int sve_count_vls(void) +{ + unsigned int vq; + int vl_count = 0; + int vl; + + if (!(getauxval(AT_HWCAP) & HWCAP_SVE)) + return 0; + + /* + * Enumerate up to SVE_VQ_MAX vector lengths + */ + for (vq = SVE_VQ_MAX; vq > 0; --vq) { + vl = prctl(PR_SVE_SET_VL, vq * 16); + if (vl == -1) + ksft_exit_fail_msg("PR_SVE_SET_VL failed: %s (%d)\n", + strerror(errno), errno); + + vl &= PR_SVE_VL_LEN_MASK; + + if (vq != sve_vq_from_vl(vl)) + vq = sve_vq_from_vl(vl); + + vl_count++; + } + + return vl_count; +} + +int main(void) +{ + int i; + + srandom(getpid()); + + ksft_print_header(); + ksft_set_plan(ARRAY_SIZE(syscalls) * (sve_count_vls() + 1)); + + for (i = 0; i < ARRAY_SIZE(syscalls); i++) + test_one_syscall(&syscalls[i]); + + ksft_print_cnts(); + + return 0; +}

The EAC1 release of the SME specification adds the FA64 feature which requires enablement at higher ELs before lower ELs can use it. Document what we require from higher ELs in our boot requirements.

Signed-off-by: Mark Brown broonie@kernel.org --- Documentation/arm64/booting.rst | 10 ++++++++++ 1 file changed, 10 insertions(+)

diff --git a/Documentation/arm64/booting.rst b/Documentation/arm64/booting.rst index 3f9d86557c5e..52d060caf8bb 100644 --- a/Documentation/arm64/booting.rst +++ b/Documentation/arm64/booting.rst @@ -340,6 +340,16 @@ Before jumping into the kernel, the following conditions must be met: - SMCR_EL2.LEN must be initialised to the same value for all CPUs the kernel will execute on.

+ For CPUs with the Scalable Matrix Extension FA64 feature (FEAT_SME_FA64) + + - If EL3 is present: + + - SMCR_EL3.FA64 (bit 31) must be initialised to 0b1. + + - If the kernel is entered at EL1 and EL2 is present: + + - SMCR_EL2.FA64 (bit 31) must be initialised to 0b1. + The requirements described above for CPU mode, caches, MMUs, architected timers, coherency and system registers apply to all CPUs. All CPUs must enter the kernel in the same exception level. Where the values documented

Provide ABI documentation for SME similar to that for SVE. Due to the very large overlap around streaming SVE mode in both implementation and interfaces documentation for streaming mode SVE is added to the SVE document rather than the SME one.

Signed-off-by: Mark Brown broonie@kernel.org --- Documentation/arm64/index.rst | 1 + Documentation/arm64/sme.rst | 430 ++++++++++++++++++++++++++++++++++ Documentation/arm64/sve.rst | 70 +++++- 3 files changed, 491 insertions(+), 10 deletions(-) create mode 100644 Documentation/arm64/sme.rst

diff --git a/Documentation/arm64/index.rst b/Documentation/arm64/index.rst index 4f840bac083e..ae21f8118830 100644 --- a/Documentation/arm64/index.rst +++ b/Documentation/arm64/index.rst @@ -21,6 +21,7 @@ ARM64 Architecture perf pointer-authentication silicon-errata + sme sve tagged-address-abi tagged-pointers diff --git a/Documentation/arm64/sme.rst b/Documentation/arm64/sme.rst new file mode 100644 index 000000000000..a0f82bac96ff --- /dev/null +++ b/Documentation/arm64/sme.rst @@ -0,0 +1,430 @@ +=================================================== +Scalable Matrix Extension support for AArch64 Linux +=================================================== + +This document outlines briefly the interface provided to userspace by Linux in +order to support use of the ARM Scalable Matrix Extension (SME). + +This is an outline of the most important features and issues only and not +intended to be exhaustive. It should be read in conjunction with the SVE +documentation in sve.rst which provides details on the Streaming SVE mode +included in SME. + +This document does not aim to describe the SME architecture or programmer's +model. To aid understanding, a minimal description of relevant programmer's +model features for SME is included in Appendix A. + + +1. General +----------- + +* PSTATE.SM, PSTATE.ZA, the streaming mode vector length, the ZA + register state and TPIDR2_EL0 are tracked per thread. + +* The presence of SME is reported to userspace via HWCAP2_SME in the aux vector + AT_HWCAP2 entry. Presence of this flag implies the presence of the SME + instructions and registers, and the Linux-specific system interfaces + described in this document. SME is reported in /proc/cpuinfo as "sme". + +* Support for the execution of SME instructions in userspace can also be + detected by reading the CPU ID register ID_AA64PFR1_EL1 using an MRS + instruction, and checking that the value of the SME field is nonzero. [3] + + It does not guarantee the presence of the system interfaces described in the + following sections: software that needs to verify that those interfaces are + present must check for HWCAP2_SME instead. + +* There are a number of optional SME features, presence of these is reported + through AT_HWCAP2 through: + + HWCAP2_SME_I16I64 + HWCAP2_SME_F64F64 + HWCAP2_SME_I8I32 + HWCAP2_SME_F16F32 + HWCAP2_SME_B16F32 + HWCAP2_SME_F32F32 + HWCAP2_SME_FA64 + + This list may be extended over time as the SME architecture evolves. + + These extensions are also reported via the CPU ID register ID_AA64SMFR0_EL1, + which userspace can read using an MRS instruction. See elf_hwcaps.txt and + cpu-feature-registers.txt for details. + +* Debuggers should restrict themselves to interacting with the target via the + NT_ARM_SVE, NT_ARM_SSVE and NT_ARM_ZA regsets. The recommended way + of detecting support for these regsets is to connect to a target process + first and then attempt a + + ptrace(PTRACE_GETREGSET, pid, NT_ARM_<regset>, &iov). + +* Whenever ZA register values are exchanged in memory between userspace and + the kernel, the register value is encoded in memory as a series of horizontal + vectors from 0 to VL/8-1 stored in the same endianness invariant format as is + used for SVE vectors. + +* On thread creation TPIDR2_EL0 is preserved unless CLONE_SETTLS is specified, + in which case it is set to 0. + +2. Vector lengths +------------------ + +SME defines a second vector length similar to the SVE vector length which is +controls the size of the streaming mode SVE vectors and the ZA matrix array. +The ZA matrix is square with each side having as many bytes as a SVE vector. + + +3. Sharing of streaming and non-streaming mode SVE state +--------------------------------------------------------- + +It is implementation defined which if any parts of the SVE state are shared +between streaming and non-streaming modes. When switching between modes +via software interfaces such as ptrace if no register content is provided as +part of switching no state will be assumed to be shared and everything will +be zeroed. + + +4. System call behaviour +------------------------- + +* On syscall PSTATE.ZA is preserved, if PSTATE.ZA==1 then the contents of the + ZA matrix are preserved. + +* On syscall PSTATE.SM will be cleared and the SVE registers will be handled + as normal. + +* Neither the SVE registers nor ZA are used to pass arguments to or receive + results from any syscall. + +* On creation fork() or clone() the newly created process will have PSTATE.SM + and PSTATE.ZA cleared. + +* All other SME state of a thread, including the currently configured vector + length, the state of the PR_SME_VL_INHERIT flag, and the deferred vector + length (if any), is preserved across all syscalls, subject to the specific + exceptions for execve() described in section 6. + + +5. Signal handling +------------------- + +* A new signal frame record za_context encodes the ZA register contents on + signal delivery. [1] + +* The signal frame record for ZA always contains basic metadata, in particular + the thread's vector length (in za_context.vl). + +* The ZA matrix may or may not be included in the record, depending on + the value of PSTATE.ZA. The registers are present if and only if: + za_context.head.size >= ZA_SIG_CONTEXT_SIZE(sve_vq_from_vl(za_context.vl)) + in which case PSTATE.ZA == 1. + +* If matrix data is present, the remainder of the record has a vl-dependent + size and layout. Macros ZA_SIG_* are defined [1] to facilitate access to + them. + +* The matrix is stored as a series of horizontal vectors in the same format as + is used for SVE vectors. + +* If the ZA context is too big to fit in sigcontext.__reserved[], then extra + space is allocated on the stack, an extra_context record is written in + __reserved[] referencing this space. za_context is then written in the + extra space. Refer to [1] for further details about this mechanism. + + +5. Signal return +----------------- + +When returning from a signal handler: + +* If there is no za_context record in the signal frame, or if the record is + present but contains no register data as desribed in the previous section, + then ZA is disabled. + +* If za_context is present in the signal frame and contains matrix data then + PSTATE.ZA is set to 1 and ZA is populated with the specified data. + +* The vector length cannot be changed via signal return. If za_context.vl in + the signal frame does not match the current vector length, the signal return + attempt is treated as illegal, resulting in a forced SIGSEGV. + + +6. prctl extensions +-------------------- + +Some new prctl() calls are added to allow programs to manage the SME vector +length: + +prctl(PR_SME_SET_VL, unsigned long arg) + + Sets the vector length of the calling thread and related flags, where + arg == vl | flags. Other threads of the calling process are unaffected. + + vl is the desired vector length, where sve_vl_valid(vl) must be true. + + flags: + + PR_SME_VL_INHERIT + + Inherit the current vector length across execve(). Otherwise, the + vector length is reset to the system default at execve(). (See + Section 9.) + + PR_SME_SET_VL_ONEXEC + + Defer the requested vector length change until the next execve() + performed by this thread. + + The effect is equivalent to implicit exceution of the following + call immediately after the next execve() (if any) by the thread: + + prctl(PR_SME_SET_VL, arg & ~PR_SME_SET_VL_ONEXEC) + + This allows launching of a new program with a different vector + length, while avoiding runtime side effects in the caller. + + Without PR_SME_SET_VL_ONEXEC, the requested change takes effect + immediately. + + + Return value: a nonnegative on success, or a negative value on error: + EINVAL: SME not supported, invalid vector length requested, or + invalid flags. + + + On success: + + * Either the calling thread's vector length or the deferred vector length + to be applied at the next execve() by the thread (dependent on whether + PR_SME_SET_VL_ONEXEC is present in arg), is set to the largest value + supported by the system that is less than or equal to vl. If vl == + SVE_VL_MAX, the value set will be the largest value supported by the + system. + + * Any previously outstanding deferred vector length change in the calling + thread is cancelled. + + * The returned value describes the resulting configuration, encoded as for + PR_SME_GET_VL. The vector length reported in this value is the new + current vector length for this thread if PR_SME_SET_VL_ONEXEC was not + present in arg; otherwise, the reported vector length is the deferred + vector length that will be applied at the next execve() by the calling + thread. + + * Changing the vector length causes all of ZA, P0..P15, FFR and all bits of + Z0..Z31 except for Z0 bits [127:0] .. Z31 bits [127:0] to become + unspecified, including both streaming and non-streaming SVE state. + Calling PR_SME_SET_VL with vl equal to the thread's current vector + length, or calling PR_SME_SET_VL with the PR_SVE_SET_VL_ONEXEC flag, + does not constitute a change to the vector length for this purpose. + + * Changing the vector length causes PSTATE.ZA and PSTATE.SM to be cleared. + Calling PR_SME_SET_VL with vl equal to the thread's current vector + length, or calling PR_SME_SET_VL with the PR_SVE_SET_VL_ONEXEC flag, + does not constitute a change to the vector length for this purpose. + + +prctl(PR_SME_GET_VL) + + Gets the vector length of the calling thread. + + The following flag may be OR-ed into the result: + + PR_SME_VL_INHERIT + + Vector length will be inherited across execve(). + + There is no way to determine whether there is an outstanding deferred + vector length change (which would only normally be the case between a + fork() or vfork() and the corresponding execve() in typical use). + + To extract the vector length from the result, bitwise and it with + PR_SME_VL_LEN_MASK. + + Return value: a nonnegative value on success, or a negative value on error: + EINVAL: SME not supported. + + +7. ptrace extensions +--------------------- + +* A new regset NT_ARM_SSVE is defined for access to streaming mode SVE + state via PTRACE_GETREGSET and PTRACE_SETREGSET, this is documented in + sve.rst. + +* A new regset NT_ARM_ZA is defined for ZA state for access to ZA state via + PTRACE_GETREGSET and PTRACE_SETREGSET. + + Refer to [2] for definitions. + +The regset data starts with struct user_za_header, containing: + + size + + Size of the complete regset, in bytes. + This depends on vl and possibly on other things in the future. + + If a call to PTRACE_GETREGSET requests less data than the value of + size, the caller can allocate a larger buffer and retry in order to + read the complete regset. + + max_size + + Maximum size in bytes that the regset can grow to for the target + thread. The regset won't grow bigger than this even if the target + thread changes its vector length etc. + + vl + + Target thread's current streaming vector length, in bytes. + + max_vl + + Maximum possible streaming vector length for the target thread. + + flags + + Zero or more of the following flags, which have the same + meaning and behaviour as the corresponding PR_SET_VL_* flags: + + SME_PT_VL_INHERIT + + SME_PT_VL_ONEXEC (SETREGSET only). + +* The effects of changing the vector length and/or flags are equivalent to + those documented for PR_SME_SET_VL. + + The caller must make a further GETREGSET call if it needs to know what VL is + actually set by SETREGSET, unless is it known in advance that the requested + VL is supported. + +* The size and layout of the payload depends on the header fields. The + SME_PT_ZA_*() macros are provided to facilitate access to the data. + +* In either case, for SETREGSET it is permissible to omit the payload, in which + case the vector length and flags are changed and PSTATE.ZA is set to 0 + (along with any consequences of those changes). If a payload is provided + then PSTATE.ZA will be set to 1. + +* For SETREGSET, if the requested VL is not supported, the effect will be the + same as if the payload were omitted, except that an EIO error is reported. + No attempt is made to translate the payload data to the correct layout + for the vector length actually set. It is up to the caller to translate the + payload layout for the actual VL and retry. + +* The effect of writing a partial, incomplete payload is unspecified. + + +8. ELF coredump extensions +--------------------------- + +* NT_ARM_SSVE notes will be added to each coredump for + each thread of the dumped process. The contents will be equivalent to the + data that would have been read if a PTRACE_GETREGSET of the corresponding + type were executed for each thread when the coredump was generated. + +* A NT_ARM_ZA note will be added to each coredump for each thread of the + dumped process. The contents will be equivalent to the data that would have + been read if a PTRACE_GETREGSET of NT_ARM_ZA were executed for each thread + when the coredump was generated. + + +9. System runtime configuration +-------------------------------- + +* To mitigate the ABI impact of expansion of the signal frame, a policy + mechanism is provided for administrators, distro maintainers and developers + to set the default vector length for userspace processes: + +/proc/sys/abi/sme_default_vector_length + + Writing the text representation of an integer to this file sets the system + default vector length to the specified value, unless the value is greater + than the maximum vector length supported by the system in which case the + default vector length is set to that maximum. + + The result can be determined by reopening the file and reading its + contents. + + At boot, the default vector length is initially set to 32 or the maximum + supported vector length, whichever is smaller and supported. This + determines the initial vector length of the init process (PID 1). + + Reading this file returns the current system default vector length. + +* At every execve() call, the new vector length of the new process is set to + the system default vector length, unless + + * PR_SME_VL_INHERIT (or equivalently SME_PT_VL_INHERIT) is set for the + calling thread, or + + * a deferred vector length change is pending, established via the + PR_SME_SET_VL_ONEXEC flag (or SME_PT_VL_ONEXEC). + +* Modifying the system default vector length does not affect the vector length + of any existing process or thread that does not make an execve() call. + + +Appendix A. SME programmer's model (informative) +================================================= + +This section provides a minimal description of the additions made by SVE to the +ARMv8-A programmer's model that are relevant to this document. + +Note: This section is for information only and not intended to be complete or +to replace any architectural specification. + +A.1. Registers +--------------- + +In A64 state, SME adds the following: + +* A new mode, streaming mode, in which a subset of the normal FPSIMD and SVE + features are available. When supported EL0 software may enter and leave + streaming mode at any time. + + For best system performance it is strongly encouraged for software to enable + streaming mode only when it is actively being used. + +* A new vector length controlling the size of ZA and the Z registers when in + streaming mode, separately to the vector length used for SVE when not in + streaming mode. There is no requirement that either the currently selected + vector length or the set of vector lengths supported for the two modes in + a given system have any relationship. The streaming mode vector length + is referred to as SVL. + +* A new ZA matrix register. This is a square matrix of SVLxSVL bits. Most + operations on ZA require that streaming mode be enabled but ZA can be + enabled without streaming mode in order to load, save and retain data. + + For best system performance it is strongly encouraged for software to enable + ZA only when it is actively being used. + +* Two new 1 bit fields in PSTATE which may be controlled via the SMSTART and + SMSTOP instructions or by access to the SVCR system register: + + * PSTATE.ZA, if this is 1 then the ZA matrix is accessible and has valid + data while if it is 0 then ZA can not be accessed. When PSTATE.ZA is + changed from 0 to 1 all bits in ZA are cleared. + + * PSTATE.SM, if this is 1 then the PE is in streaming mode. When the value + of PSTATE.SM is changed then it is implementation defined if the subset + of the floating point register bits valid in both modes may be retained. + Any other bits will be cleared. + + +References +========== + +[1] arch/arm64/include/uapi/asm/sigcontext.h + AArch64 Linux signal ABI definitions + +[2] arch/arm64/include/uapi/asm/ptrace.h + AArch64 Linux ptrace ABI definitions + +[3] Documentation/arm64/cpu-feature-registers.rst + +[4] ARM IHI0055C + http://infocenter.arm.com/help/topic/com.arm.doc.ihi0055c/IHI0055C_beta_aapc... + http://infocenter.arm.com/help/topic/com.arm.doc.subset.swdev.abi/index.html + Procedure Call Standard for the ARM 64-bit Architecture (AArch64) diff --git a/Documentation/arm64/sve.rst b/Documentation/arm64/sve.rst index 9d9a4de5bc34..93c2c2990584 100644 --- a/Documentation/arm64/sve.rst +++ b/Documentation/arm64/sve.rst @@ -7,7 +7,9 @@ Author: Dave Martin Dave.Martin@arm.com Date: 4 August 2017

This document outlines briefly the interface provided to userspace by Linux in -order to support use of the ARM Scalable Vector Extension (SVE). +order to support use of the ARM Scalable Vector Extension (SVE), including +interactions with Streaming SVE mode added by the Scalable Matrix Extension +(SME).

This is an outline of the most important features and issues only and not intended to be exhaustive. @@ -23,6 +25,10 @@ model features for SVE is included in Appendix A. * SVE registers Z0..Z31, P0..P15 and FFR and the current vector length VL, are tracked per-thread.

+* In streaming mode FFR is not accessible unless HWCAP2_SME_FA64 is present + in the system, when it is not supported and these interfaces are used to + access streaming mode FFR is read and written as zero. + * The presence of SVE is reported to userspace via HWCAP_SVE in the aux vector AT_HWCAP entry. Presence of this flag implies the presence of the SVE instructions and registers, and the Linux-specific system interfaces @@ -53,10 +59,19 @@ model features for SVE is included in Appendix A. which userspace can read using an MRS instruction. See elf_hwcaps.txt and cpu-feature-registers.txt for details.

+* On hardware that supports the SME extensions, HWCAP2_SME will also be + reported in the AT_HWCAP2 aux vector entry. Among other things SME adds + streaming mode which provides a subset of the SVE feature set using a + separate SME vector length and the same Z/V registers. See sme.rst + for more details. + * Debuggers should restrict themselves to interacting with the target via the NT_ARM_SVE regset. The recommended way of detecting support for this regset is to connect to a target process first and then attempt a - ptrace(PTRACE_GETREGSET, pid, NT_ARM_SVE, &iov). + ptrace(PTRACE_GETREGSET, pid, NT_ARM_SVE, &iov). Note that when SME is + present and streaming SVE mode is in use the FPSIMD subset of registers + will be read via NT_ARM_SVE and NT_ARM_SVE writes will exit streaming mode + in the target.

* Whenever SVE scalable register values (Zn, Pn, FFR) are exchanged in memory between userspace and the kernel, the register value is encoded in memory in @@ -126,6 +141,11 @@ the SVE instruction set architecture. are only present in fpsimd_context. For convenience, the content of V0..V31 is duplicated between sve_context and fpsimd_context.

+* The record contains a flag field which includes a flag SVE_SIG_FLAG_SM which + if set indicates that the thread is in streaming mode and the vector length + and register data (if present) describe the streaming SVE data and vector + length. + * The signal frame record for SVE always contains basic metadata, in particular the thread's vector length (in sve_context.vl).

@@ -170,6 +190,11 @@ When returning from a signal handler: the signal frame does not match the current vector length, the signal return attempt is treated as illegal, resulting in a forced SIGSEGV.

+* It is permitted to enter or leave streaming mode by setting or clearing + the SVE_SIG_FLAG_SM flag but applications should take care to ensure that + when doing so sve_context.vl and any register data are appropriate for the + vector length in the new mode. +

6. prctl extensions -------------------- @@ -265,8 +290,14 @@ prctl(PR_SVE_GET_VL) 7. ptrace extensions ---------------------

-* A new regset NT_ARM_SVE is defined for use with PTRACE_GETREGSET and - PTRACE_SETREGSET. +* New regsets NT_ARM_SVE and NT_ARM_SSVE are defined for use with + PTRACE_GETREGSET and PTRACE_SETREGSET. NT_ARM_SSVE describes the + streaming mode SVE registers and NT_ARM_SVE describes the + non-streaming mode SVE registers. + + In this description a register set is referred to as being "live" when + the target is in the appropriate streaming or non-streaming mode and is + using data beyond the subset shared with the FPSIMD Vn registers.

Refer to [2] for definitions.

@@ -297,7 +328,7 @@ The regset data starts with struct user_sve_header, containing:

flags

- either + at most one of

SVE_PT_REGS_FPSIMD

@@ -331,6 +362,10 @@ The regset data starts with struct user_sve_header, containing:

SVE_PT_VL_ONEXEC (SETREGSET only).

+ If neither FPSIMD nor SVE flags are provided then no register + payload is available, this is only possible when SME is implemented. + + * The effects of changing the vector length and/or flags are equivalent to those documented for PR_SVE_SET_VL.

@@ -346,6 +381,13 @@ The regset data starts with struct user_sve_header, containing: case only the vector length and flags are changed (along with any consequences of those changes).

+* In systems supporting SME when in streaming mode a GETREGSET for + NT_REG_SVE will return only the user_sve_header with no register data, + similarly a GETREGSET for NT_REG_SSVE will not return any register data + when not in streaming mode. + +* A GETREGSET for NT_ARM_SSVE will never return SVE_PT_REGS_FPSIMD. + * For SETREGSET, if an SVE_PT_REGS_SVE payload is present and the requested VL is not supported, the effect will be the same as if the payload were omitted, except that an EIO error is reported. No @@ -355,17 +397,25 @@ The regset data starts with struct user_sve_header, containing: unspecified. It is up to the caller to translate the payload layout for the actual VL and retry.

+* Where SME is implemented it is not possible to GETREGSET the register + state for normal SVE when in streaming mode, nor the streaming mode + register state when in normal mode, regardless of the implementation defined + behaviour of the hardware for sharing data between the two modes. + +* Any SETREGSET of NT_ARM_SVE will exit streaming mode if the target was in + streaming mode and any SETREGSET of NT_ARM_SSVE will enter streaming mode + if the target was not in streaming mode. + * The effect of writing a partial, incomplete payload is unspecified.

8. ELF coredump extensions ---------------------------

-* A NT_ARM_SVE note will be added to each coredump for each thread of the - dumped process. The contents will be equivalent to the data that would have - been read if a PTRACE_GETREGSET of NT_ARM_SVE were executed for each thread - when the coredump was generated. - +* NT_ARM_SVE and NT_ARM_SSVE notes will be added to each coredump for + each thread of the dumped process. The contents will be equivalent to the + data that would have been read if a PTRACE_GETREGSET of the corresponding + type were executed for each thread when the coredump was generated.

9. System runtime configuration --------------------------------

As with SVE rather than impose ambitious toolchain requirements for SME we manually encode the few instructions which we require in order to perform the work the kernel needs to do. That is currently:

- Vector store and load for the ZA array. - Zeroing of the whole ZA array.

This does not include the SMSTART and SMSTOP instructions which are single instructions and only used from C code, a later patch will define them as inline assembly.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/include/asm/fpsimdmacros.h | 44 +++++++++++++++++++++++++++ 1 file changed, 44 insertions(+)

diff --git a/arch/arm64/include/asm/fpsimdmacros.h b/arch/arm64/include/asm/fpsimdmacros.h index 2509d7dde55a..f9fb5f111758 100644 --- a/arch/arm64/include/asm/fpsimdmacros.h +++ b/arch/arm64/include/asm/fpsimdmacros.h @@ -93,6 +93,12 @@ .endif .endm

+.macro _sme_check_wv v + .if (\v) < 12 || (\v) > 15 + .error "Bad vector select register \v." + .endif +.endm + /* SVE instruction encodings for non-SVE-capable assemblers */ /* (pre binutils 2.28, all kernel capable clang versions support SVE) */

@@ -174,6 +180,44 @@ | (\np) .endm

+/* SME instruction encodings for non-SME-capable assemblers */ + +/* + * STR (vector from ZA array): + * STR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _sme_str_zav nw, nxbase, offset=0 + _sme_check_wv \nw + _check_general_reg \nxbase + _check_num (\offset), -0x100, 0xff + .inst 0xe1200000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +/* + * LDR (vector to ZA array): + * LDR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _sme_ldr_zav nw, nxbase, offset=0 + _sme_check_wv \nw + _check_general_reg \nxbase + _check_num (\offset), -0x100, 0xff + .inst 0xe1000000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +/* + * Zero the entire ZA array + * ZERO ZA + */ +.macro zero_za + .inst 0xc00800ff +.endm + .macro __for from:req, to:req .if (\from) == (\to) _for__body %\from

This patch introduces basic cpufeature support for discovering the presence of the Scalable Matrix Extension and reporting hwcaps for the detected features.

Signed-off-by: Mark Brown broonie@kernel.org --- Documentation/arm64/elf_hwcaps.rst | 33 +++++++++++++++++ arch/arm64/include/asm/cpu.h | 1 + arch/arm64/include/asm/cpufeature.h | 12 +++++++ arch/arm64/include/asm/fpsimd.h | 2 ++ arch/arm64/include/asm/hwcap.h | 8 +++++ arch/arm64/include/uapi/asm/hwcap.h | 8 +++++ arch/arm64/kernel/cpufeature.c | 55 +++++++++++++++++++++++++++++ arch/arm64/kernel/cpuinfo.c | 9 +++++ arch/arm64/kernel/fpsimd.c | 30 ++++++++++++++++ arch/arm64/tools/cpucaps | 2 ++ 10 files changed, 160 insertions(+)

diff --git a/Documentation/arm64/elf_hwcaps.rst b/Documentation/arm64/elf_hwcaps.rst index ec1a5a63c1d0..9f723e472668 100644 --- a/Documentation/arm64/elf_hwcaps.rst +++ b/Documentation/arm64/elf_hwcaps.rst @@ -247,6 +247,39 @@ HWCAP2_MTE Functionality implied by ID_AA64PFR1_EL1.MTE == 0b0010, as described by Documentation/arm64/memory-tagging-extension.rst.

+HWCAP2_SME + + Functionality implied by ID_AA64PFR1_EL1.SME == 0b0001, as described + by Documentation/arm64/sme.rst. + +HWCAP2_SME_I16I64 + + Functionality implied by ID_AA64SMFR0_EL1.I16I64 == 0b1111. + +HWCAP2_SME_F64F64 + + Functionality implied by ID_AA64SMFR0_EL1.F64F64 == 0b1. + +HWCAP2_SME_I8I32 + + Functionality implied by ID_AA64SMFR0_EL1.I8I32 == 0b1111. + +HWCAP2_SME_F16F32 + + Functionality implied by ID_AA64SMFR0_EL1.F16F32 == 0b1. + +HWCAP2_SME_B16F32 + + Functionality implied by ID_AA64SMFR0_EL1.B16F32 == 0b1. + +HWCAP2_SME_F32F32 + + Functionality implied by ID_AA64SMFR0_EL1.F32F32 == 0b1. + +HWCAP2_SME_FA64 + + Functionality implied by ID_AA64SMFR0_EL1.FA64 == 0b1. + 4. Unused AT_HWCAP bits -----------------------

diff --git a/arch/arm64/include/asm/cpu.h b/arch/arm64/include/asm/cpu.h index 0f6d16faa540..667b66fe1a53 100644 --- a/arch/arm64/include/asm/cpu.h +++ b/arch/arm64/include/asm/cpu.h @@ -57,6 +57,7 @@ struct cpuinfo_arm64 { u64 reg_id_aa64pfr0; u64 reg_id_aa64pfr1; u64 reg_id_aa64zfr0; + u64 reg_id_aa64smfr0;

struct cpuinfo_32bit aarch32;

diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h index ef6be92b1921..079e9f15c5eb 100644 --- a/arch/arm64/include/asm/cpufeature.h +++ b/arch/arm64/include/asm/cpufeature.h @@ -727,6 +727,18 @@ static __always_inline bool system_supports_sve(void) cpus_have_const_cap(ARM64_SVE); }

+static __always_inline bool system_supports_sme(void) +{ + return IS_ENABLED(CONFIG_ARM64_SME) && + cpus_have_const_cap(ARM64_SME); +} + +static __always_inline bool system_supports_fa64(void) +{ + return IS_ENABLED(CONFIG_ARM64_SME) && + cpus_have_const_cap(ARM64_SME_FA64); +} + static __always_inline bool system_supports_cnp(void) { return IS_ENABLED(CONFIG_ARM64_CNP) && diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index cb24385e3632..d5f2825a2412 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -74,6 +74,8 @@ extern void sve_set_vq(unsigned long vq_minus_1);

struct arm64_cpu_capabilities; extern void sve_kernel_enable(const struct arm64_cpu_capabilities *__unused); +extern void sme_kernel_enable(const struct arm64_cpu_capabilities *__unused); +extern void fa64_kernel_enable(const struct arm64_cpu_capabilities *__unused);

extern u64 read_zcr_features(void);

diff --git a/arch/arm64/include/asm/hwcap.h b/arch/arm64/include/asm/hwcap.h index 8c129db8232a..8c9de4ae9ee1 100644 --- a/arch/arm64/include/asm/hwcap.h +++ b/arch/arm64/include/asm/hwcap.h @@ -105,6 +105,14 @@ #define KERNEL_HWCAP_RNG __khwcap2_feature(RNG) #define KERNEL_HWCAP_BTI __khwcap2_feature(BTI) #define KERNEL_HWCAP_MTE __khwcap2_feature(MTE) +#define KERNEL_HWCAP_SME __khwcap2_feature(SME) +#define KERNEL_HWCAP_SME_I16I64 __khwcap2_feature(SME_I16I64) +#define KERNEL_HWCAP_SME_F64F64 __khwcap2_feature(SME_F64F64) +#define KERNEL_HWCAP_SME_I8I32 __khwcap2_feature(SME_I8I32) +#define KERNEL_HWCAP_SME_F16F32 __khwcap2_feature(SME_F16F32) +#define KERNEL_HWCAP_SME_B16F32 __khwcap2_feature(SME_B16F32) +#define KERNEL_HWCAP_SME_F32F32 __khwcap2_feature(SME_F32F32) +#define KERNEL_HWCAP_SME_FA64 __khwcap2_feature(SME_FA64)

/* * This yields a mask that user programs can use to figure out what diff --git a/arch/arm64/include/uapi/asm/hwcap.h b/arch/arm64/include/uapi/asm/hwcap.h index b8f41aa234ee..374777ab9ab2 100644 --- a/arch/arm64/include/uapi/asm/hwcap.h +++ b/arch/arm64/include/uapi/asm/hwcap.h @@ -75,5 +75,13 @@ #define HWCAP2_RNG (1 << 16) #define HWCAP2_BTI (1 << 17) #define HWCAP2_MTE (1 << 18) +#define HWCAP2_SME (1 << 20) +#define HWCAP2_SME_I16I64 (1 << 21) +#define HWCAP2_SME_F64F64 (1 << 22) +#define HWCAP2_SME_I8I32 (1 << 23) +#define HWCAP2_SME_F16F32 (1 << 24) +#define HWCAP2_SME_B16F32 (1 << 25) +#define HWCAP2_SME_F32F32 (1 << 26) +#define HWCAP2_SME_FA64 (1 << 27)

#endif /* _UAPI__ASM_HWCAP_H */ diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c index 5fd56c6b294f..0f92fa334a77 100644 --- a/arch/arm64/kernel/cpufeature.c +++ b/arch/arm64/kernel/cpufeature.c @@ -246,6 +246,7 @@ static const struct arm64_ftr_bits ftr_id_aa64pfr0[] = { };

static const struct arm64_ftr_bits ftr_id_aa64pfr1[] = { + ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64PFR1_SME_SHIFT, 4, 0), ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64PFR1_MPAMFRAC_SHIFT, 4, 0), ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64PFR1_RASFRAC_SHIFT, 4, 0), ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_MTE), @@ -278,6 +279,24 @@ static const struct arm64_ftr_bits ftr_id_aa64zfr0[] = { ARM64_FTR_END, };

+static const struct arm64_ftr_bits ftr_id_aa64smfr0[] = { + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_FA64_SHIFT, 1, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_I16I64_SHIFT, 4, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_F64F64_SHIFT, 1, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_I8I32_SHIFT, 4, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_F16F32_SHIFT, 1, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_B16F32_SHIFT, 1, 0), + ARM64_FTR_BITS(FTR_VISIBLE_IF_IS_ENABLED(CONFIG_ARM64_SME), + FTR_STRICT, FTR_EXACT, ID_AA64SMFR0_F32F32_SHIFT, 1, 0), + ARM64_FTR_END, +}; + static const struct arm64_ftr_bits ftr_id_aa64mmfr0[] = { ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64MMFR0_ECV_SHIFT, 4, 0), ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64MMFR0_FGT_SHIFT, 4, 0), @@ -624,6 +643,7 @@ static const struct __ftr_reg_entry { ARM64_FTR_REG_OVERRIDE(SYS_ID_AA64PFR1_EL1, ftr_id_aa64pfr1, &id_aa64pfr1_override), ARM64_FTR_REG(SYS_ID_AA64ZFR0_EL1, ftr_id_aa64zfr0), + ARM64_FTR_REG(SYS_ID_AA64SMFR0_EL1, ftr_id_aa64smfr0),

/* Op1 = 0, CRn = 0, CRm = 5 */ ARM64_FTR_REG(SYS_ID_AA64DFR0_EL1, ftr_id_aa64dfr0), @@ -935,6 +955,7 @@ void __init init_cpu_features(struct cpuinfo_arm64 *info) init_cpu_ftr_reg(SYS_ID_AA64PFR0_EL1, info->reg_id_aa64pfr0); init_cpu_ftr_reg(SYS_ID_AA64PFR1_EL1, info->reg_id_aa64pfr1); init_cpu_ftr_reg(SYS_ID_AA64ZFR0_EL1, info->reg_id_aa64zfr0); + init_cpu_ftr_reg(SYS_ID_AA64SMFR0_EL1, info->reg_id_aa64smfr0);

if (id_aa64pfr0_32bit_el0(info->reg_id_aa64pfr0)) init_32bit_cpu_features(&info->aarch32); @@ -2346,6 +2367,30 @@ static const struct arm64_cpu_capabilities arm64_features[] = { .matches = has_cpuid_feature, .min_field_value = 1, }, +#ifdef CONFIG_ARM64_SME + { + .desc = "Scalable Matrix Extension", + .type = ARM64_CPUCAP_SYSTEM_FEATURE, + .capability = ARM64_SME, + .sys_reg = SYS_ID_AA64PFR1_EL1, + .sign = FTR_UNSIGNED, + .field_pos = ID_AA64PFR1_SME_SHIFT, + .min_field_value = ID_AA64PFR1_SME, + .matches = has_cpuid_feature, + .cpu_enable = sme_kernel_enable, + }, + { + .desc = "FA64", + .type = ARM64_CPUCAP_SYSTEM_FEATURE, + .capability = ARM64_SME_FA64, + .sys_reg = SYS_ID_AA64SMFR0_EL1, + .sign = FTR_UNSIGNED, + .field_pos = ID_AA64SMFR0_FA64_SHIFT, + .min_field_value = ID_AA64SMFR0_FA64, + .matches = has_feature_flag, + .cpu_enable = fa64_kernel_enable, + }, +#endif /* CONFIG_ARM64_SME */ {}, };

@@ -2477,6 +2522,16 @@ static const struct arm64_cpu_capabilities arm64_elf_hwcaps[] = { #ifdef CONFIG_ARM64_MTE HWCAP_CAP(SYS_ID_AA64PFR1_EL1, ID_AA64PFR1_MTE_SHIFT, FTR_UNSIGNED, ID_AA64PFR1_MTE, CAP_HWCAP, KERNEL_HWCAP_MTE), #endif /* CONFIG_ARM64_MTE */ +#ifdef CONFIG_ARM64_SME + HWCAP_CAP(SYS_ID_AA64PFR1_EL1, ID_AA64PFR1_SME_SHIFT, FTR_UNSIGNED, ID_AA64PFR1_SME, CAP_HWCAP, KERNEL_HWCAP_SME), + HWCAP_CAP_FLAG(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_FA64_SHIFT, CAP_HWCAP, KERNEL_HWCAP_SME_FA64), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_I16I64_SHIFT, FTR_UNSIGNED, ID_AA64SMFR0_I16I64, CAP_HWCAP, KERNEL_HWCAP_SME_I16I64), + HWCAP_CAP_FLAG(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_F64F64_SHIFT, CAP_HWCAP, KERNEL_HWCAP_SME_F64F64), + HWCAP_CAP(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_I8I32_SHIFT, FTR_UNSIGNED, ID_AA64SMFR0_I8I32, CAP_HWCAP, KERNEL_HWCAP_SME_I8I32), + HWCAP_CAP_FLAG(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_F16F32_SHIFT, CAP_HWCAP, KERNEL_HWCAP_SME_F16F32), + HWCAP_CAP_FLAG(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_B16F32_SHIFT, CAP_HWCAP, KERNEL_HWCAP_SME_B16F32), + HWCAP_CAP_FLAG(SYS_ID_AA64SMFR0_EL1, ID_AA64SMFR0_F32F32_SHIFT, CAP_HWCAP, KERNEL_HWCAP_SME_F32F32), +#endif /* CONFIG_ARM64_SME */ {}, };

diff --git a/arch/arm64/kernel/cpuinfo.c b/arch/arm64/kernel/cpuinfo.c index 87731fea5e41..c3f3acb77fb8 100644 --- a/arch/arm64/kernel/cpuinfo.c +++ b/arch/arm64/kernel/cpuinfo.c @@ -94,6 +94,14 @@ static const char *const hwcap_str[] = { [KERNEL_HWCAP_RNG] = "rng", [KERNEL_HWCAP_BTI] = "bti", [KERNEL_HWCAP_MTE] = "mte", + [KERNEL_HWCAP_SME] = "sme", + [KERNEL_HWCAP_SME_I16I64] = "smei16i64", + [KERNEL_HWCAP_SME_F64F64] = "smef64f64", + [KERNEL_HWCAP_SME_I8I32] = "smei8i32", + [KERNEL_HWCAP_SME_F16F32] = "smef16f32", + [KERNEL_HWCAP_SME_B16F32] = "smeb16f32", + [KERNEL_HWCAP_SME_F32F32] = "smef32f32", + [KERNEL_HWCAP_SME_FA64] = "smefa64", };

#ifdef CONFIG_COMPAT @@ -396,6 +404,7 @@ static void __cpuinfo_store_cpu(struct cpuinfo_arm64 *info) info->reg_id_aa64pfr0 = read_cpuid(ID_AA64PFR0_EL1); info->reg_id_aa64pfr1 = read_cpuid(ID_AA64PFR1_EL1); info->reg_id_aa64zfr0 = read_cpuid(ID_AA64ZFR0_EL1); + info->reg_id_aa64smfr0 = read_cpuid(ID_AA64SMFR0_EL1);

if (id_aa64pfr1_mte(info->reg_id_aa64pfr1)) info->reg_gmid = read_cpuid(GMID_EL1); diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 4a98cc3b1df1..8dde4593ca73 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -983,6 +983,32 @@ void fpsimd_release_task(struct task_struct *dead_task)

#endif /* CONFIG_ARM64_SVE */

+#ifdef CONFIG_ARM64_SME + +void sme_kernel_enable(const struct arm64_cpu_capabilities *__always_unused p) +{ + /* Set priority for all PEs to architecturally defined minimum */ + write_sysreg_s(read_sysreg_s(SYS_SMPRI_EL1) & ~SMPRI_EL1_PRIORITY_MASK, + SYS_SMPRI_EL1); + + /* Allow SME in kernel */ + write_sysreg(read_sysreg(CPACR_EL1) | CPACR_EL1_SMEN_EL1EN, CPACR_EL1); + isb(); +} + +/* + * This must be called after sme_kernel_enable(), we rely on the + * feature table being sorted to ensure this. + */ +void fa64_kernel_enable(const struct arm64_cpu_capabilities *__always_unused p) +{ + /* Allow use of FA64 */ + write_sysreg_s(read_sysreg_s(SYS_SMCR_EL1) | SMCR_ELx_FA64_MASK, + SYS_SMCR_EL1); +} + +#endif /* CONFIG_ARM64_SVE */ + /* * Trapped SVE access * @@ -1525,6 +1551,10 @@ static int __init fpsimd_init(void) if (!cpu_have_named_feature(ASIMD)) pr_notice("Advanced SIMD is not implemented\n");

+ + if (cpu_have_named_feature(SME) && !cpu_have_named_feature(SVE)) + pr_notice("SME is implemented but not SVE\n"); + return sve_sysctl_init(); } core_initcall(fpsimd_init); diff --git a/arch/arm64/tools/cpucaps b/arch/arm64/tools/cpucaps index 49305c2e6dfd..f81723fe8482 100644 --- a/arch/arm64/tools/cpucaps +++ b/arch/arm64/tools/cpucaps @@ -39,6 +39,8 @@ HW_DBM KVM_PROTECTED_MODE MISMATCHED_CACHE_TYPE MTE +SME +SME_FA64 SPECTRE_V2 SPECTRE_V3A SPECTRE_V4

As for SVE provide a sysctl which allows the default SME vector length to be configured.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/kernel/fpsimd.c | 29 ++++++++++++++++++++++++++++- 1 file changed, 28 insertions(+), 1 deletion(-)

diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index f9b77d0b8e40..9b247d13c3fe 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -479,6 +479,30 @@ static int __init sve_sysctl_init(void) static int __init sve_sysctl_init(void) { return 0; } #endif /* ! (CONFIG_ARM64_SVE && CONFIG_SYSCTL) */

+#if defined(CONFIG_ARM64_SME) && defined(CONFIG_SYSCTL) +static struct ctl_table sme_default_vl_table[] = { + { + .procname = "sme_default_vector_length", + .mode = 0644, + .proc_handler = vec_proc_do_default_vl, + .extra1 = &vl_info[ARM64_VEC_SME], + }, + { } +}; + +static int __init sme_sysctl_init(void) +{ + if (system_supports_sme()) + if (!register_sysctl("abi", sme_default_vl_table)) + return -EINVAL; + + return 0; +} + +#else /* ! (CONFIG_ARM64_SME && CONFIG_SYSCTL) */ +static int __init sme_sysctl_init(void) { return 0; } +#endif /* ! (CONFIG_ARM64_SME && CONFIG_SYSCTL) */ + #define ZREG(sve_state, vq, n) ((char *)(sve_state) + \ (SVE_SIG_ZREG_OFFSET(vq, n) - SVE_SIG_REGS_OFFSET))

@@ -1684,6 +1708,9 @@ static int __init fpsimd_init(void) if (cpu_have_named_feature(SME) && !cpu_have_named_feature(SVE)) pr_notice("SME is implemented but not SVE\n");

- return sve_sysctl_init(); + sve_sysctl_init(); + sme_sysctl_init(); + + return 0; } core_initcall(fpsimd_init);

The Scalable Matrix Extension introduces support for a new thread specific data register TPIDR2 intended for use by libc. The kernel must save the value of TPIDR2 on context switch and should ensure that all new threads start off with a default value of 0. Add a field to the thread_struct to store TPIDR2 and context switch it with the other thread specific data.

In case there are future extensions which also use TPIDR2 we introduce system_supports_tpidr2() and use that rather than system_supports_sme() for TPIDR2 handling.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/include/asm/cpufeature.h | 5 +++++ arch/arm64/include/asm/processor.h | 1 + arch/arm64/kernel/fpsimd.c | 4 ++++ arch/arm64/kernel/process.c | 14 ++++++++++++-- 4 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h index c5fee98307cb..d67246141b7c 100644 --- a/arch/arm64/include/asm/cpufeature.h +++ b/arch/arm64/include/asm/cpufeature.h @@ -746,6 +746,11 @@ static __always_inline bool system_supports_fa64(void) cpus_have_const_cap(ARM64_SME_FA64); }

+static __always_inline bool system_supports_tpidr2(void) +{ + return system_supports_sme(); +} + static __always_inline bool system_supports_cnp(void) { return IS_ENABLED(CONFIG_ARM64_CNP) && diff --git a/arch/arm64/include/asm/processor.h b/arch/arm64/include/asm/processor.h index a62d2f8045bf..51eca2513cb5 100644 --- a/arch/arm64/include/asm/processor.h +++ b/arch/arm64/include/asm/processor.h @@ -168,6 +168,7 @@ struct thread_struct { u64 mte_ctrl; #endif u64 sctlr_user; + u64 tpidr2_el0; };

static inline unsigned int thread_get_vl(struct thread_struct *thread, diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 20c889487193..0af82c518979 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -1098,6 +1098,10 @@ void sme_kernel_enable(const struct arm64_cpu_capabilities *__always_unused p) /* Allow SME in kernel */ write_sysreg(read_sysreg(CPACR_EL1) | CPACR_EL1_SMEN_EL1EN, CPACR_EL1); isb(); + + /* Allow EL0 to access TPIDR2 */ + write_sysreg(read_sysreg(SCTLR_EL1) | SCTLR_ELx_ENTP2, SCTLR_EL1); + isb(); }

/* diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c index 40adb8cdbf5a..3f4279ad68bc 100644 --- a/arch/arm64/kernel/process.c +++ b/arch/arm64/kernel/process.c @@ -249,6 +249,8 @@ void show_regs(struct pt_regs *regs) static void tls_thread_flush(void) { write_sysreg(0, tpidr_el0); + if (system_supports_tpidr2()) + write_sysreg_s(0, SYS_TPIDR2_EL0);

if (is_compat_task()) { current->thread.uw.tp_value = 0; @@ -342,6 +344,8 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start, * out-of-sync with the saved value. */ *task_user_tls(p) = read_sysreg(tpidr_el0); + if (system_supports_tpidr2()) + p->thread.tpidr2_el0 = read_sysreg_s(SYS_TPIDR2_EL0);

if (stack_start) { if (is_compat_thread(task_thread_info(p))) @@ -352,10 +356,12 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start,

/* * If a TLS pointer was passed to clone, use it for the new - * thread. + * thread. We also reset TPIDR2 if it's in use. */ - if (clone_flags & CLONE_SETTLS) + if (clone_flags & CLONE_SETTLS) { p->thread.uw.tp_value = tls; + p->thread.tpidr2_el0 = 0; + } } else { /* * A kthread has no context to ERET to, so ensure any buggy @@ -386,6 +392,8 @@ int copy_thread(unsigned long clone_flags, unsigned long stack_start, void tls_preserve_current_state(void) { *task_user_tls(current) = read_sysreg(tpidr_el0); + if (system_supports_tpidr2() && !is_compat_task()) + current->thread.tpidr2_el0 = read_sysreg_s(SYS_TPIDR2_EL0); }

static void tls_thread_switch(struct task_struct *next) @@ -398,6 +406,8 @@ static void tls_thread_switch(struct task_struct *next) write_sysreg(0, tpidrro_el0);

write_sysreg(*task_user_tls(next), tpidr_el0); + if (system_supports_tpidr2()) + write_sysreg_s(next->thread.tpidr2_el0, SYS_TPIDR2_EL0); }

/*

When in streaming mode we need to save and restore the streaming mode SVE register state rather than the regular SVE register state. This uses the streaming mode vector length and omits FFR but is otherwise identical, if TIF_SVE is enabled when we are in streaming mode then streaming mode takes precedence.

This does not handle use of streaming SVE state with KVM, ptrace or signals. This will be updated in further patches.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/include/asm/fpsimd.h | 23 +++++- arch/arm64/include/asm/fpsimdmacros.h | 11 +++ arch/arm64/include/asm/processor.h | 10 +++ arch/arm64/kernel/entry-fpsimd.S | 9 +++ arch/arm64/kernel/fpsimd.c | 109 +++++++++++++++++++++----- arch/arm64/kvm/fpsimd.c | 3 +- 6 files changed, 142 insertions(+), 23 deletions(-)

diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index dc01fd634b75..2f7b64797736 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -46,11 +46,22 @@ extern void fpsimd_restore_current_state(void); extern void fpsimd_update_current_state(struct user_fpsimd_state const *state);

extern void fpsimd_bind_state_to_cpu(struct user_fpsimd_state *state, - void *sve_state, unsigned int sve_vl); + void *sve_state, unsigned int sve_vl, + unsigned int sme_vl);

extern void fpsimd_flush_task_state(struct task_struct *target); extern void fpsimd_save_and_flush_cpu_state(void);

+static inline bool thread_sm_enabled(struct thread_struct *thread) +{ + return system_supports_sme() && (thread->svcr & SYS_SVCR_EL0_SM_MASK); +} + +static inline bool thread_za_enabled(struct thread_struct *thread) +{ + return system_supports_sme() && (thread->svcr & SYS_SVCR_EL0_ZA_MASK); +} + /* Maximum VL that SVE/SME VL-agnostic software can transparently support */ #define VL_ARCH_MAX 0x100

@@ -62,7 +73,14 @@ static inline size_t sve_ffr_offset(int vl)

static inline void *sve_pffr(struct thread_struct *thread) { - return (char *)thread->sve_state + sve_ffr_offset(thread_get_sve_vl(thread)); + unsigned int vl; + + if (system_supports_sme() && thread_sm_enabled(thread)) + vl = thread_get_sme_vl(thread); + else + vl = thread_get_sve_vl(thread); + + return (char *)thread->sve_state + sve_ffr_offset(vl); }

extern void sve_save_state(void *state, u32 *pfpsr, int save_ffr); @@ -71,6 +89,7 @@ extern void sve_load_state(void const *state, u32 const *pfpsr, extern void sve_flush_live(bool flush_ffr, unsigned long vq_minus_1); extern unsigned int sve_get_vl(void); extern void sve_set_vq(unsigned long vq_minus_1); +extern void sme_set_vq(unsigned long vq_minus_1);

struct arm64_cpu_capabilities; extern void sve_kernel_enable(const struct arm64_cpu_capabilities *__unused); diff --git a/arch/arm64/include/asm/fpsimdmacros.h b/arch/arm64/include/asm/fpsimdmacros.h index f9fb5f111758..57bcd9ec4cb9 100644 --- a/arch/arm64/include/asm/fpsimdmacros.h +++ b/arch/arm64/include/asm/fpsimdmacros.h @@ -252,6 +252,17 @@ 921: .endm

+/* Update SMCR_EL1.LEN with the new VQ */ +.macro sme_load_vq xvqminus1, xtmp, xtmp2 + mrs_s \xtmp, SYS_SMCR_EL1 + bic \xtmp2, \xtmp, SMCR_ELx_LEN_MASK + orr \xtmp2, \xtmp2, \xvqminus1 + cmp \xtmp2, \xtmp + b.eq 921f + msr_s SYS_SMCR_EL1, \xtmp2 //self-synchronising +921: +.endm + /* Preserve the first 128-bits of Znz and zero the rest. */ .macro _sve_flush_z nz _sve_check_zreg \nz diff --git a/arch/arm64/include/asm/processor.h b/arch/arm64/include/asm/processor.h index 3c235e165725..338cb03811bd 100644 --- a/arch/arm64/include/asm/processor.h +++ b/arch/arm64/include/asm/processor.h @@ -183,6 +183,11 @@ static inline unsigned int thread_get_sve_vl(struct thread_struct *thread) return thread_get_vl(thread, ARM64_VEC_SVE); }

+static inline unsigned int thread_get_sme_vl(struct thread_struct *thread) +{ + return thread_get_vl(thread, ARM64_VEC_SME); +} + unsigned int task_get_vl(const struct task_struct *task, enum vec_type type); void task_set_vl(struct task_struct *task, enum vec_type type, unsigned long vl); @@ -196,6 +201,11 @@ static inline unsigned int task_get_sve_vl(const struct task_struct *task) return task_get_vl(task, ARM64_VEC_SVE); }

+static inline unsigned int task_get_sme_vl(const struct task_struct *task) +{ + return task_get_vl(task, ARM64_VEC_SME); +} + static inline void task_set_sve_vl(struct task_struct *task, unsigned long vl) { task_set_vl(task, ARM64_VEC_SVE, vl); diff --git a/arch/arm64/kernel/entry-fpsimd.S b/arch/arm64/kernel/entry-fpsimd.S index dc242e269f9a..b06bf42e7856 100644 --- a/arch/arm64/kernel/entry-fpsimd.S +++ b/arch/arm64/kernel/entry-fpsimd.S @@ -86,3 +86,12 @@ SYM_FUNC_START(sve_flush_live) SYM_FUNC_END(sve_flush_live)

#endif /* CONFIG_ARM64_SVE */ + +#ifdef CONFIG_ARM64_SME + +SYM_FUNC_START(sme_set_vq) + sme_load_vq x0, x1, x2 + ret +SYM_FUNC_END(sme_set_vq) + +#endif /* CONFIG_ARM64_SME */ diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index d6e6bec6b490..39a4dd0a6257 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -118,6 +118,7 @@ struct fpsimd_last_state_struct { struct user_fpsimd_state *st; void *sve_state; unsigned int sve_vl; + unsigned int sme_vl; };

static DEFINE_PER_CPU(struct fpsimd_last_state_struct, fpsimd_last_state); @@ -296,17 +297,28 @@ void task_set_vl_onexec(struct task_struct *task, enum vec_type type, task->thread.vl_onexec[type] = vl; }

+/* + * TIF_SME controls whether a task can use SME without trapping while + * in userspace, when TIF_SME is set then we must have storage + * alocated in sve_state and za_state to store the contents of both ZA + * and the SVE registers for both streaming and non-streaming modes. + * + * If both SVCR.ZA and SVCR.SM are disabled then at any point we + * may disable TIF_SME and reenable traps. + */ + + /* * TIF_SVE controls whether a task can use SVE without trapping while - * in userspace, and also the way a task's FPSIMD/SVE state is stored - * in thread_struct. + * in userspace, and also (together with TIF_SME) the way a task's + * FPSIMD/SVE state is stored in thread_struct. * * The kernel uses this flag to track whether a user task is actively * using SVE, and therefore whether full SVE register state needs to * be tracked. If not, the cheaper FPSIMD context handling code can * be used instead of the more costly SVE equivalents. * - * * TIF_SVE set: + * * TIF_SVE or SVCR.SM set: * * The task can execute SVE instructions while in userspace without * trapping to the kernel. @@ -314,7 +326,8 @@ void task_set_vl_onexec(struct task_struct *task, enum vec_type type, * When stored, Z0-Z31 (incorporating Vn in bits[127:0] or the * corresponding Zn), P0-P15 and FFR are encoded in in * task->thread.sve_state, formatted appropriately for vector - * length task->thread.sve_vl. + * length task->thread.sve_vl or, if SVCR.SM is set, + * task->thread.sme_vl. * * task->thread.sve_state must point to a valid buffer at least * sve_state_size(task) bytes in size. @@ -352,19 +365,40 @@ void task_set_vl_onexec(struct task_struct *task, enum vec_type type, */ static void task_fpsimd_load(void) { + bool restore_sve_regs = false; + bool restore_ffr; + WARN_ON(!system_supports_fpsimd()); WARN_ON(!have_cpu_fpsimd_context());

- if (IS_ENABLED(CONFIG_ARM64_SME) && test_thread_flag(TIF_SME)) - write_sysreg_s(current->thread.svcr, SYS_SVCR_EL0); - + /* Check if we should restore SVE first */ if (IS_ENABLED(CONFIG_ARM64_SVE) && test_thread_flag(TIF_SVE)) { sve_set_vq(sve_vq_from_vl(task_get_sve_vl(current)) - 1); + restore_sve_regs = true; + restore_ffr = true; + } + + /* Restore SME, override SVE register configuration if needed */ + if (system_supports_sme()) { + unsigned long sme_vl = task_get_sme_vl(current); + + if (test_thread_flag(TIF_SME)) + sme_set_vq(sve_vq_from_vl(sme_vl) - 1); + + write_sysreg_s(current->thread.svcr, SYS_SVCR_EL0); + + if (thread_sm_enabled(&current->thread)) { + restore_sve_regs = true; + restore_ffr = system_supports_fa64(); + } + } + + if (restore_sve_regs) sve_load_state(sve_pffr(&current->thread), - &current->thread.uw.fpsimd_state.fpsr, true); - } else { + &current->thread.uw.fpsimd_state.fpsr, + restore_ffr); + else fpsimd_load_state(&current->thread.uw.fpsimd_state); - } }

/* @@ -376,6 +410,9 @@ static void fpsimd_save(void) struct fpsimd_last_state_struct const *last = this_cpu_ptr(&fpsimd_last_state); /* set by fpsimd_bind_task_to_cpu() or fpsimd_bind_state_to_cpu() */ + bool save_sve_regs = false; + bool save_ffr; + unsigned int vl;

WARN_ON(!system_supports_fpsimd()); WARN_ON(!have_cpu_fpsimd_context()); @@ -383,13 +420,32 @@ static void fpsimd_save(void) if (test_thread_flag(TIF_FOREIGN_FPSTATE)) return;

- if (IS_ENABLED(CONFIG_ARM64_SME) && - test_thread_flag(TIF_SME)) + if (test_thread_flag(TIF_SVE)) { + save_sve_regs = true; + save_ffr = true; + vl = last->sve_vl; + } + + if (system_supports_sme()) { current->thread.svcr = read_sysreg_s(SYS_SVCR_EL0);

- if (IS_ENABLED(CONFIG_ARM64_SVE) && - test_thread_flag(TIF_SVE)) { - if (WARN_ON(sve_get_vl() != last->sve_vl)) { + if (thread_za_enabled(&current->thread)) { + /* ZA state managment is not implemented yet */ + force_signal_inject(SIGKILL, SI_KERNEL, 0, 0); + return; + } + + /* If we are in streaming mode override regular SVE. */ + if (thread_sm_enabled(&current->thread)) { + save_sve_regs = true; + save_ffr = system_supports_fa64(); + vl = last->sme_vl; + } + } + + if (IS_ENABLED(CONFIG_ARM64_SVE) && save_sve_regs) { + /* Get the configured VL from RDVL, will account for SM */ + if (WARN_ON(sve_get_vl() != vl)) { /* * Can't save the user regs, so current would * re-enter user with corrupt state. @@ -400,8 +456,8 @@ static void fpsimd_save(void) }

sve_save_state((char *)last->sve_state + - sve_ffr_offset(last->sve_vl), - &last->st->fpsr, true); + sve_ffr_offset(vl), + &last->st->fpsr, save_ffr); } else { fpsimd_save_state(last->st); } @@ -606,7 +662,14 @@ static void sve_to_fpsimd(struct task_struct *task) */ static size_t sve_state_size(struct task_struct const *task) { - return SVE_SIG_REGS_SIZE(sve_vq_from_vl(task_get_sve_vl(task))); + unsigned int vl = 0; + + if (system_supports_sve()) + vl = task_get_sve_vl(task); + if (system_supports_sme()) + vl = max(vl, task_get_sme_vl(task)); + + return SVE_SIG_REGS_SIZE(sve_vq_from_vl(vl)); }

/* @@ -735,7 +798,8 @@ int vec_set_vector_length(struct task_struct *task, enum vec_type type, }

fpsimd_flush_task_state(task); - if (test_and_clear_tsk_thread_flag(task, TIF_SVE)) + if (test_and_clear_tsk_thread_flag(task, TIF_SVE) || + thread_sm_enabled(&task->thread)) sve_to_fpsimd(task);

if (system_supports_sme() && type == ARM64_VEC_SME) @@ -1372,6 +1436,9 @@ void fpsimd_flush_thread(void) fpsimd_flush_thread_vl(ARM64_VEC_SVE); }

+ if (system_supports_sme()) + fpsimd_flush_thread_vl(ARM64_VEC_SME); + put_cpu_fpsimd_context(); }

@@ -1415,6 +1482,7 @@ static void fpsimd_bind_task_to_cpu(void) last->st = &current->thread.uw.fpsimd_state; last->sve_state = current->thread.sve_state; last->sve_vl = task_get_sve_vl(current); + last->sme_vl = task_get_sme_vl(current); current->thread.fpsimd_cpu = smp_processor_id();

if (system_supports_sve()) { @@ -1429,7 +1497,7 @@ static void fpsimd_bind_task_to_cpu(void) }

void fpsimd_bind_state_to_cpu(struct user_fpsimd_state *st, void *sve_state, - unsigned int sve_vl) + unsigned int sve_vl, unsigned int sme_vl) { struct fpsimd_last_state_struct *last = this_cpu_ptr(&fpsimd_last_state); @@ -1440,6 +1508,7 @@ void fpsimd_bind_state_to_cpu(struct user_fpsimd_state *st, void *sve_state, last->st = st; last->sve_state = sve_state; last->sve_vl = sve_vl; + last->sme_vl = sme_vl; }

/* diff --git a/arch/arm64/kvm/fpsimd.c b/arch/arm64/kvm/fpsimd.c index 5621020b28de..d96871002081 100644 --- a/arch/arm64/kvm/fpsimd.c +++ b/arch/arm64/kvm/fpsimd.c @@ -99,7 +99,8 @@ void kvm_arch_vcpu_ctxsync_fp(struct kvm_vcpu *vcpu) if (vcpu->arch.flags & KVM_ARM64_FP_ENABLED) { fpsimd_bind_state_to_cpu(&vcpu->arch.ctxt.fp_regs, vcpu->arch.sve_state, - vcpu->arch.sve_max_vl); + vcpu->arch.sve_max_vl, + 0);

clear_thread_flag(TIF_FOREIGN_FPSTATE); update_thread_flag(TIF_SVE, vcpu_has_sve(vcpu));

By default all SME operations in userspace will trap. When this happens we allocate storage space for the SME register state, set up the SVE registers and disable traps. We do not need to initialize ZA since the architecture guarantees that it will be zeroed when enabled and when we trap ZA is disabled.

On syscall we exit streaming mode if we were previously in it and ensure that all but the lower 128 bits of the registers are zeroed while preserving the state of ZA. This follows the aarch64 PCS for SME, ZA state is preserved over a function call and streaming mode is exited. Since the traps for SME do not distinguish between streaming mode SVE and ZA usage if ZA is in use rather than reenabling traps we instead zero the parts of the SVE registers not shared with FPSIMD and leave SME enabled, this simplifies handling SME traps. If ZA is not in use then we reenable SME traps and fall through to normal handling of SVE.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/include/asm/esr.h | 1 + arch/arm64/include/asm/exception.h | 1 + arch/arm64/include/asm/fpsimd.h | 15 +++ arch/arm64/kernel/entry-common.c | 10 ++ arch/arm64/kernel/fpsimd.c | 174 ++++++++++++++++++++++++++--- arch/arm64/kernel/process.c | 12 +- arch/arm64/kernel/syscall.c | 43 ++++++- 7 files changed, 233 insertions(+), 23 deletions(-)

diff --git a/arch/arm64/include/asm/esr.h b/arch/arm64/include/asm/esr.h index 31418bc7644a..f664f31c774b 100644 --- a/arch/arm64/include/asm/esr.h +++ b/arch/arm64/include/asm/esr.h @@ -75,6 +75,7 @@ #define ESR_ELx_IL_SHIFT (25) #define ESR_ELx_IL (UL(1) << ESR_ELx_IL_SHIFT) #define ESR_ELx_ISS_MASK (ESR_ELx_IL - 1) +#define ESR_ELx_ISS(esr) ((esr) & ESR_ELx_ISS_MASK)

/* ISS field definitions shared by different classes */ #define ESR_ELx_WNR_SHIFT (6) diff --git a/arch/arm64/include/asm/exception.h b/arch/arm64/include/asm/exception.h index 339477dca551..2add7f33b7c2 100644 --- a/arch/arm64/include/asm/exception.h +++ b/arch/arm64/include/asm/exception.h @@ -64,6 +64,7 @@ void do_debug_exception(unsigned long addr_if_watchpoint, unsigned int esr, struct pt_regs *regs); void do_fpsimd_acc(unsigned int esr, struct pt_regs *regs); void do_sve_acc(unsigned int esr, struct pt_regs *regs); +void do_sme_acc(unsigned int esr, struct pt_regs *regs); void do_fpsimd_exc(unsigned int esr, struct pt_regs *regs); void do_sysinstr(unsigned int esr, struct pt_regs *regs); void do_sp_pc_abort(unsigned long addr, unsigned int esr, struct pt_regs *regs); diff --git a/arch/arm64/include/asm/fpsimd.h b/arch/arm64/include/asm/fpsimd.h index de7947a71618..99b5b822c47f 100644 --- a/arch/arm64/include/asm/fpsimd.h +++ b/arch/arm64/include/asm/fpsimd.h @@ -282,6 +282,17 @@ static inline void sve_setup(void) { } #ifdef CONFIG_ARM64_SME

extern void __init sme_setup(void); +extern void sme_alloc(struct task_struct *task); + +static inline void sme_user_disable(void) +{ + sysreg_clear_set(cpacr_el1, CPACR_EL1_SMEN_EL0EN, 0); +} + +static inline void sme_user_enable(void) +{ + sysreg_clear_set(cpacr_el1, 0, CPACR_EL1_SMEN_EL0EN); +}

static inline void sme_smstart_sm(void) { @@ -313,6 +324,7 @@ extern int sme_get_current_vl(void); static inline void sme_setup(void) { } static inline int sme_max_vl(void) { return 0; } static inline int sme_max_virtualisable_vl(void) { return 0; } +static inline void sme_alloc(struct task_struct *task) { }

static inline void sme_smstart_sm(void) { } static inline void sme_smstop_sm(void) { } @@ -327,6 +339,9 @@ static inline int sme_get_current_vl(void) return -EINVAL; }

+static inline void sme_user_disable(void) { BUILD_BUG(); } +static inline void sme_user_enable(void) { BUILD_BUG(); } + #endif /* ! CONFIG_ARM64_SME */

/* For use by EFI runtime services calls only */ diff --git a/arch/arm64/kernel/entry-common.c b/arch/arm64/kernel/entry-common.c index 32f9796c4ffe..ed34b783244f 100644 --- a/arch/arm64/kernel/entry-common.c +++ b/arch/arm64/kernel/entry-common.c @@ -516,6 +516,13 @@ static void noinstr el0_sve_acc(struct pt_regs *regs, unsigned long esr) exit_to_user_mode(regs); }

+static void noinstr el0_sme_acc(struct pt_regs *regs, unsigned long esr) +{ + enter_from_user_mode(regs); + local_daif_restore(DAIF_PROCCTX); + do_sme_acc(esr, regs); +} + static void noinstr el0_fpsimd_exc(struct pt_regs *regs, unsigned long esr) { enter_from_user_mode(regs); @@ -624,6 +631,9 @@ asmlinkage void noinstr el0t_64_sync_handler(struct pt_regs *regs) case ESR_ELx_EC_SVE: el0_sve_acc(regs, esr); break; + case ESR_ELx_EC_SME: + el0_sme_acc(regs, esr); + break; case ESR_ELx_EC_FP_EXC64: el0_fpsimd_exc(regs, esr); break; diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index 2582374b7e48..e2b254752f5f 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -204,6 +204,12 @@ static void set_sme_default_vl(int val) set_default_vl(ARM64_VEC_SME, val); }

+static void sme_free(struct task_struct *); + +#else + +static inline void sme_free(struct task_struct *t) { } + #endif

DEFINE_PER_CPU(bool, fpsimd_context_busy); @@ -804,18 +810,22 @@ int vec_set_vector_length(struct task_struct *task, enum vec_type type, thread_sm_enabled(&task->thread)) sve_to_fpsimd(task);

- if (system_supports_sme() && type == ARM64_VEC_SME) + if (system_supports_sme() && type == ARM64_VEC_SME) { task->thread.svcr &= ~(SYS_SVCR_EL0_SM_MASK | SYS_SVCR_EL0_ZA_MASK); + clear_thread_flag(TIF_SME); + }

if (task == current) put_cpu_fpsimd_context();

/* - * Force reallocation of task SVE state to the correct size - * on next use: + * Force reallocation of task SVE and SME state to the correct + * size on next use: */ sve_free(task); + if (system_supports_sme() && type == ARM64_VEC_SME) + sme_free(task);

task_set_vl(task, type, vl);

@@ -1160,12 +1170,55 @@ void __init sve_setup(void) void fpsimd_release_task(struct task_struct *dead_task) { __sve_free(dead_task); + sme_free(dead_task); }

#endif /* CONFIG_ARM64_SVE */

#ifdef CONFIG_ARM64_SME

+/* This will move to uapi/asm/sigcontext.h when signals are implemented */ +#define ZA_SIG_REGS_SIZE(vq) ((vq * __SVE_VQ_BYTES) * (vq * __SVE_VQ_BYTES)) + +/* + * Return how many bytes of memory are required to store the full SME + * specific state (currently just ZA) for task, given task's currently + * configured vector length. + */ +size_t za_state_size(struct task_struct const *task) +{ + unsigned int vl = task_get_sme_vl(task); + + return ZA_SIG_REGS_SIZE(sve_vq_from_vl(vl)); +} + +/* + * Ensure that task->thread.za_state is allocated and sufficiently large. + * + * This function should be used only in preparation for replacing + * task->thread.za_state with new data. The memory is always zeroed + * here to prevent stale data from showing through: this is done in + * the interest of testability and predictability, the architecture + * guarantees that when ZA is enabled it will be zeroed. + */ +void sme_alloc(struct task_struct *task) +{ + if (task->thread.za_state) { + memset(task->thread.za_state, 0, za_state_size(task)); + return; + } + + /* This could potentially be up to 64K. */ + task->thread.za_state = + kzalloc(za_state_size(task), GFP_KERNEL); +} + +static void sme_free(struct task_struct *task) +{ + kfree(task->thread.za_state); + task->thread.za_state = NULL; +} + void sme_kernel_enable(const struct arm64_cpu_capabilities *__always_unused p) { /* Set priority for all PEs to architecturally defined minimum */ @@ -1275,6 +1328,29 @@ void __init sme_setup(void)

#endif /* CONFIG_ARM64_SME */

+static void sve_init_regs(void) +{ + /* + * Convert the FPSIMD state to SVE, zeroing all the state that + * is not shared with FPSIMD. If (as is likely) the current + * state is live in the registers then do this there and + * update our metadata for the current task including + * disabling the trap, otherwise update our in-memory copy. + * We are guaranteed to not be in streaming mode, we can only + * take a SVE trap when not in streaming mode and we can't be + * in streaming mode when taking a SME trap. + */ + if (!test_thread_flag(TIF_FOREIGN_FPSTATE)) { + unsigned long vq_minus_one = + sve_vq_from_vl(task_get_sve_vl(current)) - 1; + sve_set_vq(vq_minus_one); + sve_flush_live(true, vq_minus_one); + fpsimd_bind_task_to_cpu(); + } else { + fpsimd_to_sve(current); + } +} + /* * Trapped SVE access * @@ -1306,22 +1382,77 @@ void do_sve_acc(unsigned int esr, struct pt_regs *regs) WARN_ON(1); /* SVE access shouldn't have trapped */

/* - * Convert the FPSIMD state to SVE, zeroing all the state that - * is not shared with FPSIMD. If (as is likely) the current - * state is live in the registers then do this there and - * update our metadata for the current task including - * disabling the trap, otherwise update our in-memory copy. + * Even if the task can have used streaming mode we can only + * generate SVE access traps in normal SVE mode and + * transitioning out of streaming mode may discard any + * streaming mode state. Always clear the high bits to avoid + * any potential errors tracking what is properly initialised. */ + sve_init_regs(); + + put_cpu_fpsimd_context(); +} + +/* + * Trapped SME access + * + * Storage is allocated for the full SVE and SME state, the current + * FPSIMD register contents are migrated to SVE if SVE is not already + * active, and the access trap is disabled. + * + * TIF_SME should be clear on entry: otherwise, fpsimd_restore_current_state() + * would have disabled the SME access trap for userspace during + * ret_to_user, making an SVE access trap impossible in that case. + */ +void do_sme_acc(unsigned int esr, struct pt_regs *regs) +{ + /* Even if we chose not to use SME, the hardware could still trap: */ + if (unlikely(!system_supports_sme()) || WARN_ON(is_compat_task())) { + force_signal_inject(SIGILL, ILL_ILLOPC, regs->pc, 0); + return; + } + + /* + * If this not a trap due to SME being disabled then something + * is being used in the wrong mode, report as SIGILL. + */ + if (ESR_ELx_ISS(esr) != ESR_ELx_SME_ISS_SME_DISABLED) { + force_signal_inject(SIGILL, ILL_ILLOPC, regs->pc, 0); + return; + } + + sve_alloc(current); + sme_alloc(current); + if (!current->thread.sve_state || !current->thread.za_state) { + force_sig(SIGKILL); + return; + } + + get_cpu_fpsimd_context(); + + /* With TIF_SME userspace shouldn't generate any traps */ + if (test_and_set_thread_flag(TIF_SME)) + WARN_ON(1); + if (!test_thread_flag(TIF_FOREIGN_FPSTATE)) { unsigned long vq_minus_one = - sve_vq_from_vl(task_get_sve_vl(current)) - 1; - sve_set_vq(vq_minus_one); - sve_flush_live(true, vq_minus_one); + sve_vq_from_vl(task_get_sme_vl(current)) - 1; + sme_set_vq(vq_minus_one); + fpsimd_bind_task_to_cpu(); - } else { - fpsimd_to_sve(current); }

+ /* + * If SVE was not already active initialise the SVE registers, + * any non-shared state between the streaming and regular SVE + * registers is architecturally guaranteed to be zeroed when + * we enter streaming mode. We do not need to initialize ZA + * since ZA must be disabled at this point and enabling ZA is + * architecturally defined to zero ZA. + */ + if (system_supports_sve() && !test_thread_flag(TIF_SVE)) + sve_init_regs(); + put_cpu_fpsimd_context(); }

@@ -1438,8 +1569,11 @@ void fpsimd_flush_thread(void) fpsimd_flush_thread_vl(ARM64_VEC_SVE); }

- if (system_supports_sme()) + if (system_supports_sme()) { + clear_thread_flag(TIF_SME); + sme_free(current); fpsimd_flush_thread_vl(ARM64_VEC_SME); + }

put_cpu_fpsimd_context(); } @@ -1488,14 +1622,22 @@ static void fpsimd_bind_task_to_cpu(void) last->sme_vl = task_get_sme_vl(current); current->thread.fpsimd_cpu = smp_processor_id();

+ /* + * Toggle SVE and SME trapping for userspace if needed, these + * are serialsied by ret_to_user() + */ if (system_supports_sve()) { - /* Toggle SVE trapping for userspace if needed */ if (test_thread_flag(TIF_SVE)) sve_user_enable(); else sve_user_disable(); + }

- /* Serialised by exception return to user */ + if (system_supports_sme()) { + if (test_thread_flag(TIF_SME)) + sme_user_enable(); + else + sme_user_disable(); } }

diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c index 4dd35619ece6..cbd6b12c4339 100644 --- a/arch/arm64/kernel/process.c +++ b/arch/arm64/kernel/process.c @@ -298,17 +298,19 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src) BUILD_BUG_ON(!IS_ENABLED(CONFIG_THREAD_INFO_IN_TASK));

/* - * Detach src's sve_state (if any) from dst so that it does not - * get erroneously used or freed prematurely. dst's sve_state - * will be allocated on demand later on if dst uses SVE. - * For consistency, also clear TIF_SVE here: this could be done + * Detach src's sve/za_state (if any) from dst so that it does not + * get erroneously used or freed prematurely. dst's copies + * will be allocated on demand later on if dst uses SVE/SME. + * For consistency, also clear TIF_SVE/SME here: this could be done * later in copy_process(), but to avoid tripping up future - * maintainers it is best not to leave TIF_SVE and sve_state in + * maintainers it is best not to leave TIF flags and buffers in * an inconsistent state, even temporarily. */ dst->thread.sve_state = NULL; clear_tsk_thread_flag(dst, TIF_SVE);

+ dst->thread.za_state = NULL; + clear_tsk_thread_flag(dst, TIF_SME); dst->thread.svcr = 0;

/* clear any pending asynchronous tag fault raised by the parent */ diff --git a/arch/arm64/kernel/syscall.c b/arch/arm64/kernel/syscall.c index 50a0f1a38e84..c2237da046f3 100644 --- a/arch/arm64/kernel/syscall.c +++ b/arch/arm64/kernel/syscall.c @@ -158,11 +158,50 @@ static void el0_svc_common(struct pt_regs *regs, int scno, int sc_nr, syscall_trace_exit(regs); }

-static inline void sve_user_discard(void) +/* + * As per the ABI exit SME streaming mode and clear the SVE state not + * shared with FPSIMD on syscall entry. + */ +static inline void fp_user_discard(void) { + /* + * If SME is active then exit streaming mode. If ZA is active + * then flush the SVE registers but leave userspace access to + * both SVE and SME enabled, otherwise disable SME for the + * task and fall through to disabling SVE too. + * + * Since with SME traps disabled userspace can freely enter + * and exit streaming mode and SVE register state is shared + * between the two it is simpler to just leave SVE enabled + * while userspace is actively using SME than to track state + * transitions. + */ + if (system_supports_sme() && test_thread_flag(TIF_SME)) { + u64 svcr = read_sysreg_s(SYS_SVCR_EL0); + + if (svcr & SYS_SVCR_EL0_SM_MASK) + sme_smstop_sm(); + + if (svcr & SYS_SVCR_EL0_ZA_MASK) { + unsigned long sve_vq_minus_one = + sve_vq_from_vl(task_get_sve_vl(current)) - 1; + sve_flush_live(true, sve_vq_minus_one); + return; + } else { + clear_thread_flag(TIF_SME); + sme_user_disable(); + } + } + + if (!system_supports_sve()) return;

+ /* + * If SME is not active then disable SVE, the registers will + * be cleared when userspace next attempts to access them and + * we do not need to track the SVE register state until then. + */ clear_thread_flag(TIF_SVE);

/* @@ -177,7 +216,7 @@ static inline void sve_user_discard(void)

void do_el0_svc(struct pt_regs *regs) { - sve_user_discard(); + fp_user_discard(); el0_svc_common(regs, regs->regs[8], __NR_syscalls, sys_call_table); }

Implement support for ZA in signal handling in a very similar way to how we implement support for SVE registers, using a signal context structure with optional register state after it. Where present this register state stores the ZA matrix as a series of horizontal vectors numbered from 0 to VL/8 in the endinanness independent format used for vectors.

As with SVE we do not allow changes in the vector length during signal return but we do allow ZA to be enabled or disabled.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/include/uapi/asm/sigcontext.h | 41 +++++++ arch/arm64/kernel/fpsimd.c | 3 - arch/arm64/kernel/signal.c | 139 +++++++++++++++++++++++ 3 files changed, 180 insertions(+), 3 deletions(-)

diff --git a/arch/arm64/include/uapi/asm/sigcontext.h b/arch/arm64/include/uapi/asm/sigcontext.h index 3a3366d4fbc2..d45bdf2c8b26 100644 --- a/arch/arm64/include/uapi/asm/sigcontext.h +++ b/arch/arm64/include/uapi/asm/sigcontext.h @@ -140,6 +140,14 @@ struct sve_context {

#define SVE_SIG_FLAG_SM 0x1 /* Context describes streaming mode */

+#define ZA_MAGIC 0x54366345 + +struct za_context { + struct _aarch64_ctx head; + __u16 vl; + __u16 __reserved[3]; +}; + #endif /* !__ASSEMBLY__ */

#include <asm/sve_context.h> @@ -259,4 +267,37 @@ struct sve_context { #define SVE_SIG_CONTEXT_SIZE(vq) \ (SVE_SIG_REGS_OFFSET + SVE_SIG_REGS_SIZE(vq))

+/* + * If the ZA register is enabled for the thread at signal delivery then, + * za_context.head.size >= ZA_SIG_CONTEXT_SIZE(sve_vq_from_vl(za_context.vl)) + * and the register data may be accessed using the ZA_SIG_*() macros. + * + * If za_context.head.size < ZA_SIG_CONTEXT_SIZE(sve_vq_from_vl(za_context.vl)) + * then ZA was not enabled and no register data was included in which case + * ZA register was not enabled for the thread and no register data + * the ZA_SIG_*() macros should not be used except for this check. + * + * The same convention applies when returning from a signal: a caller + * will need to remove or resize the za_context block if it wants to + * enable the ZA register when it was previously non-live or vice-versa. + * This may require the caller to allocate fresh memory and/or move other + * context blocks in the signal frame. + * + * Changing the vector length during signal return is not permitted: + * za_context.vl must equal the thread's current SME vector length when + * doing a sigreturn. + */ + +#define ZA_SIG_REGS_OFFSET \ + ((sizeof(struct za_context) + (__SVE_VQ_BYTES - 1)) \ + / __SVE_VQ_BYTES * __SVE_VQ_BYTES) + +#define ZA_SIG_REGS_SIZE(vq) ((vq * __SVE_VQ_BYTES) * (vq * __SVE_VQ_BYTES)) + +#define ZA_SIG_ZAV_OFFSET(vq, n) (ZA_SIG_REGS_OFFSET + \ + (SVE_SIG_ZREG_SIZE(vq) * n)) + +#define ZA_SIG_CONTEXT_SIZE(vq) \ + (ZA_SIG_REGS_OFFSET + ZA_SIG_REGS_SIZE(vq)) + #endif /* _UAPI__ASM_SIGCONTEXT_H */ diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index e2b254752f5f..31d11dce0289 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -1177,9 +1177,6 @@ void fpsimd_release_task(struct task_struct *dead_task)

#ifdef CONFIG_ARM64_SME

-/* This will move to uapi/asm/sigcontext.h when signals are implemented */ -#define ZA_SIG_REGS_SIZE(vq) ((vq * __SVE_VQ_BYTES) * (vq * __SVE_VQ_BYTES)) - /* * Return how many bytes of memory are required to store the full SME * specific state (currently just ZA) for task, given task's currently diff --git a/arch/arm64/kernel/signal.c b/arch/arm64/kernel/signal.c index fea0e1d30449..c0007ddf0c06 100644 --- a/arch/arm64/kernel/signal.c +++ b/arch/arm64/kernel/signal.c @@ -57,6 +57,7 @@ struct rt_sigframe_user_layout { unsigned long fpsimd_offset; unsigned long esr_offset; unsigned long sve_offset; + unsigned long za_offset; unsigned long extra_offset; unsigned long end_offset; }; @@ -219,6 +220,7 @@ static int restore_fpsimd_context(struct fpsimd_context __user *ctx) struct user_ctxs { struct fpsimd_context __user *fpsimd; struct sve_context __user *sve; + struct za_context __user *za; };

#ifdef CONFIG_ARM64_SVE @@ -347,6 +349,101 @@ extern int restore_sve_fpsimd_context(struct user_ctxs *user);

#endif /* ! CONFIG_ARM64_SVE */

+#ifdef CONFIG_ARM64_SME + +static int preserve_za_context(struct za_context __user *ctx) +{ + int err = 0; + u16 reserved[ARRAY_SIZE(ctx->__reserved)]; + unsigned int vl = task_get_sme_vl(current); + unsigned int vq; + + if (thread_za_enabled(&current->thread)) + vq = sve_vq_from_vl(vl); + else + vq = 0; + + memset(reserved, 0, sizeof(reserved)); + + __put_user_error(ZA_MAGIC, &ctx->head.magic, err); + __put_user_error(round_up(SVE_SIG_CONTEXT_SIZE(vq), 16), + &ctx->head.size, err); + __put_user_error(vl, &ctx->vl, err); + BUILD_BUG_ON(sizeof(ctx->__reserved) != sizeof(reserved)); + err |= __copy_to_user(&ctx->__reserved, reserved, sizeof(reserved)); + + if (vq) { + /* + * This assumes that the ZA state has already been saved to + * the task struct by calling the function + * fpsimd_signal_preserve_current_state(). + */ + err |= __copy_to_user((char __user *)ctx + ZA_SIG_REGS_OFFSET, + current->thread.za_state, + ZA_SIG_REGS_SIZE(vq)); + } + + return err ? -EFAULT : 0; +} + +static int restore_za_context(struct user_ctxs __user *user) +{ + int err; + unsigned int vq; + struct za_context za; + + if (__copy_from_user(&za, user->za, sizeof(za))) + return -EFAULT; + + if (za.vl != task_get_sme_vl(current)) + return -EINVAL; + + if (za.head.size <= sizeof(*user->za)) { + current->thread.svcr &= ~SYS_SVCR_EL0_ZA_MASK; + return 0; + } + + vq = sve_vq_from_vl(za.vl); + + if (za.head.size < ZA_SIG_CONTEXT_SIZE(vq)) + return -EINVAL; + + /* + * Careful: we are about __copy_from_user() directly into + * thread.za_state with preemption enabled, so protection is + * needed to prevent a racing context switch from writing stale + * registers back over the new data. + */ + + fpsimd_flush_task_state(current); + /* From now, fpsimd_thread_switch() won't touch thread.sve_state */ + + sme_alloc(current); + if (!current->thread.za_state) { + current->thread.svcr &= ~SYS_SVCR_EL0_ZA_MASK; + clear_thread_flag(TIF_SME); + return -ENOMEM; + } + + err = __copy_from_user(current->thread.za_state, + (char __user const *)user->za + + ZA_SIG_REGS_OFFSET, + ZA_SIG_REGS_SIZE(vq)); + if (err) + return -EFAULT; + + set_thread_flag(TIF_SME); + current->thread.svcr |= SYS_SVCR_EL0_ZA_MASK; + + return 0; +} +#else /* ! CONFIG_ARM64_SME */ + +/* Turn any non-optimised out attempts to use these into a link error: */ +extern int preserve_za_context(void __user *ctx); +extern int restore_za_context(struct user_ctxs *user); + +#endif /* ! CONFIG_ARM64_SME */

static int parse_user_sigframe(struct user_ctxs *user, struct rt_sigframe __user *sf) @@ -361,6 +458,7 @@ static int parse_user_sigframe(struct user_ctxs *user,

user->fpsimd = NULL; user->sve = NULL; + user->za = NULL;

if (!IS_ALIGNED((unsigned long)base, 16)) goto invalid; @@ -426,6 +524,19 @@ static int parse_user_sigframe(struct user_ctxs *user, user->sve = (struct sve_context __user *)head; break;

+ case ZA_MAGIC: + if (!system_supports_sme()) + goto invalid; + + if (user->za) + goto invalid; + + if (size < sizeof(*user->za)) + goto invalid; + + user->za = (struct za_context __user *)head; + break; + case EXTRA_MAGIC: if (have_extra_context) goto invalid; @@ -549,6 +660,9 @@ static int restore_sigframe(struct pt_regs *regs, } }

+ if (err == 0 && system_supports_sme() && user.za) + err = restore_za_context(&user); + return err; }

@@ -633,6 +747,24 @@ static int setup_sigframe_layout(struct rt_sigframe_user_layout *user, return err; }

+ if (system_supports_sme()) { + unsigned int vl; + unsigned int vq = 0; + + if (add_all) + vl = sme_max_vl(); + else + vl = task_get_sme_vl(current); + + if (thread_za_enabled(&current->thread)) + vq = sve_vq_from_vl(vl); + + err = sigframe_alloc(user, &user->za_offset, + ZA_SIG_CONTEXT_SIZE(vq)); + if (err) + return err; + } + return sigframe_alloc_end(user); }

@@ -681,6 +813,13 @@ static int setup_sigframe(struct rt_sigframe_user_layout *user, err |= preserve_sve_context(sve_ctx); }

+ /* ZA state if present */ + if (system_supports_sme() && err == 0 && user->za_offset) { + struct za_context __user *za_ctx = + apply_user_offset(user, user->za_offset); + err |= preserve_za_context(za_ctx); + } + if (err == 0 && user->extra_offset) { char __user *sfp = (char __user *)user->sigframe; char __user *userp =

The ZA array can be read and written with the NT_ARM_ZA. Similarly to our interface for the SVE vector registers the regset consists of a header with information on the current vector length followed by an optional register data payload, represented as for signals as a series of horizontal vectors from 0 to VL/8 in the endianness independent format used for vectors.

On get if ZA is enabled then register data will be provided, otherwise it will be omitted. On set if register data is provided then ZA is enabled and initialized using the provided data, otherwise it is disabled.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/include/uapi/asm/ptrace.h | 56 +++++++++++ arch/arm64/kernel/ptrace.c | 144 +++++++++++++++++++++++++++ include/uapi/linux/elf.h | 1 + 3 files changed, 201 insertions(+)

diff --git a/arch/arm64/include/uapi/asm/ptrace.h b/arch/arm64/include/uapi/asm/ptrace.h index 522b925a78c1..7fa2f7036aa7 100644 --- a/arch/arm64/include/uapi/asm/ptrace.h +++ b/arch/arm64/include/uapi/asm/ptrace.h @@ -268,6 +268,62 @@ struct user_pac_generic_keys { __uint128_t apgakey; };

+/* ZA state (NT_ARM_ZA) */ + +struct user_za_header { + __u32 size; /* total meaningful regset content in bytes */ + __u32 max_size; /* maxmium possible size for this thread */ + __u16 vl; /* current vector length */ + __u16 max_vl; /* maximum possible vector length */ + __u16 flags; + __u16 __reserved; +}; + +/* + * Common ZA_PT_* flags: + * These must be kept in sync with prctl interface in <linux/prctl.h> + */ +#define ZA_PT_VL_INHERIT ((1 << 17) /* PR_SME_VL_INHERIT */ >> 16) +#define ZA_PT_VL_ONEXEC ((1 << 18) /* PR_SME_SET_VL_ONEXEC */ >> 16) + + +/* + * The remainder of the ZA state follows struct user_za_header. The + * total size of the ZA state (including header) depends on the + * metadata in the header: ZA_PT_SIZE(vq, flags) gives the total size + * of the state in bytes, including the header. + * + * Refer to <asm/sigcontext.h> for details of how to pass the correct + * "vq" argument to these macros. + */ + +/* Offset from the start of struct user_za_header to the register data */ +#define ZA_PT_ZA_OFFSET \ + ((sizeof(struct user_za_header) + (__SVE_VQ_BYTES - 1)) \ + / __SVE_VQ_BYTES * __SVE_VQ_BYTES) + +/* + * The payload starts at offset ZA_PT_ZA_OFFSET, and is of size + * ZA_PT_ZA_SIZE(vq, flags). + * + * The ZA array is stored as a sequence of horizontal vectors ZAV of SVL/8 + * bytes each, starting from vector 0. + * + * Additional data might be appended in the future. + * + * The ZA matrix is represented in memory in an endianness-invariant layout + * which differs from the layout used for the FPSIMD V-registers on big-endian + * systems: see sigcontext.h for more explanation. + */ + +#define ZA_PT_ZAV_OFFSET(vq, n) \ + (ZA_PT_ZA_OFFSET + ((vq * __SVE_VQ_BYTES) * n)) + +#define ZA_PT_ZA_SIZE(vq) ((vq * __SVE_VQ_BYTES) * (vq * __SVE_VQ_BYTES)) + +#define ZA_PT_SIZE(vq) \ + (ZA_PT_ZA_OFFSET + ZA_PT_ZA_SIZE(vq)) + #endif /* __ASSEMBLY__ */

#endif /* _UAPI__ASM_PTRACE_H */ diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c index 414126ce5897..702765f50a47 100644 --- a/arch/arm64/kernel/ptrace.c +++ b/arch/arm64/kernel/ptrace.c @@ -996,6 +996,141 @@ static int ssve_set(struct task_struct *target, ARM64_VEC_SME); }

+static int za_get(struct task_struct *target, + const struct user_regset *regset, + struct membuf to) +{ + struct user_za_header header; + unsigned int vq; + unsigned long start, end; + + if (!system_supports_sme()) + return -EINVAL; + + /* Header */ + memset(&header, 0, sizeof(header)); + + if (test_tsk_thread_flag(target, TIF_SME_VL_INHERIT)) + header.flags |= ZA_PT_VL_INHERIT; + + header.vl = task_get_sme_vl(target); + vq = sve_vq_from_vl(header.vl); + header.max_vl = sme_max_vl(); + header.max_size = ZA_PT_SIZE(vq); + + /* If ZA is not active there is only the header */ + if (thread_za_enabled(&target->thread)) + header.size = ZA_PT_SIZE(vq); + else + header.size = ZA_PT_ZA_OFFSET; + + membuf_write(&to, &header, sizeof(header)); + + BUILD_BUG_ON(ZA_PT_ZA_OFFSET != sizeof(header)); + end = ZA_PT_ZA_OFFSET; +; + if (target == current) + fpsimd_preserve_current_state(); + + /* Any register data to include? */ + if (thread_za_enabled(&target->thread)) { + start = end; + end = ZA_PT_SIZE(vq); + membuf_write(&to, target->thread.za_state, end - start); + } + + /* Zero any trailing padding */ + start = end; + end = ALIGN(header.size, SVE_VQ_BYTES); + return membuf_zero(&to, end - start); +} + +static int za_set(struct task_struct *target, + const struct user_regset *regset, + unsigned int pos, unsigned int count, + const void *kbuf, const void __user *ubuf) +{ + int ret; + struct user_za_header header; + unsigned int vq; + unsigned long start, end; + + if (!system_supports_sme()) + return -EINVAL; + + /* Header */ + if (count < sizeof(header)) + return -EINVAL; + ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, &header, + 0, sizeof(header)); + if (ret) + goto out; + + /* + * All current ZA_PT_* flags are consumed by + * vec_set_vector_length(), which will also validate them for + * us: + */ + ret = vec_set_vector_length(target, ARM64_VEC_SME, header.vl, + ((unsigned long)header.flags) << 16); + if (ret) + goto out; + + /* Actual VL set may be less than the user asked for: */ + vq = sve_vq_from_vl(task_get_sme_vl(target)); + + /* Ensure there is some SVE storage for streaming mode */ + if (!target->thread.sve_state) { + sve_alloc(target); + if (!target->thread.sve_state) { + clear_thread_flag(TIF_SME); + ret = -ENOMEM; + goto out; + } + } + + /* Allocate/reinit ZA storage */ + sme_alloc(target); + if (!target->thread.za_state) { + ret = -ENOMEM; + clear_tsk_thread_flag(target, TIF_SME); + goto out; + } + + /* If there is no data then disable ZA */ + if (!count) { + target->thread.svcr &= ~SYS_SVCR_EL0_ZA_MASK; + goto out; + } + + /* + * If setting a different VL from the requested VL and there is + * register data, the data layout will be wrong: don't even + * try to set the registers in this case. + */ + if (vq != sve_vq_from_vl(header.vl)) { + ret = -EIO; + goto out; + } + + BUILD_BUG_ON(ZA_PT_ZA_OFFSET != sizeof(header)); + start = ZA_PT_ZA_OFFSET; + end = ZA_PT_SIZE(vq); + ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf, + target->thread.za_state, + start, end); + if (ret) + goto out; + + /* Mark ZA as active and let userspace use it */ + set_tsk_thread_flag(target, TIF_SME); + target->thread.svcr |= SYS_SVCR_EL0_ZA_MASK; + +out: + fpsimd_flush_task_state(target); + return ret; +} + #endif /* CONFIG_ARM64_SME */

#ifdef CONFIG_ARM64_PTR_AUTH @@ -1217,6 +1352,7 @@ enum aarch64_regset { #endif #ifdef CONFIG_ARM64_SVE REGSET_SSVE, + REGSET_ZA, #endif #ifdef CONFIG_ARM64_PTR_AUTH REGSET_PAC_MASK, @@ -1308,6 +1444,14 @@ static const struct user_regset aarch64_regsets[] = { .regset_get = ssve_get, .set = ssve_set, }, + [REGSET_ZA] = { /* SME ZA */ + .core_note_type = NT_ARM_ZA, + .n = DIV_ROUND_UP(ZA_PT_ZA_SIZE(SVE_VQ_MAX), SVE_VQ_BYTES), + .size = SVE_VQ_BYTES, + .align = SVE_VQ_BYTES, + .regset_get = za_get, + .set = za_set, + }, #endif #ifdef CONFIG_ARM64_PTR_AUTH [REGSET_PAC_MASK] = { diff --git a/include/uapi/linux/elf.h b/include/uapi/linux/elf.h index 61502388683f..7ef574f3256a 100644 --- a/include/uapi/linux/elf.h +++ b/include/uapi/linux/elf.h @@ -428,6 +428,7 @@ typedef struct elf64_shdr { #define NT_ARM_TAGGED_ADDR_CTRL 0x409 /* arm64 tagged address control (prctl()) */ #define NT_ARM_PAC_ENABLED_KEYS 0x40a /* arm64 ptr auth enabled keys (prctl()) */ #define NT_ARM_SSVE 0x40b /* ARM Streaming SVE registers */ +#define NT_ARM_ZA 0x40c /* ARM SME ZA registers */ #define NT_ARC_V2 0x600 /* ARCv2 accumulator/extra registers */ #define NT_VMCOREDD 0x700 /* Vmcore Device Dump Note */ #define NT_MIPS_DSP 0x800 /* MIPS DSP ASE registers */

When saving and restoring the floating point state over an EFI runtime call ensure that we handle streaming mode, only handling FFR if we are not in streaming mode and ensuring that we are in normal mode over the call into runtime services.

We currently assume that ZA will not be modified by runtime services, the specification is not yet finalised so this may need updating if that changes.

Signed-off-by: Mark Brown broonie@kernel.org --- arch/arm64/kernel/fpsimd.c | 47 +++++++++++++++++++++++++++++++++----- 1 file changed, 41 insertions(+), 6 deletions(-)

diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c index ebaf174bdf82..3551ed97e0f3 100644 --- a/arch/arm64/kernel/fpsimd.c +++ b/arch/arm64/kernel/fpsimd.c @@ -1046,21 +1046,25 @@ int vec_verify_vq_map(enum vec_type type)

static void __init sve_efi_setup(void) { - struct vl_info *info = &vl_info[ARM64_VEC_SVE]; + int max_vl = 0; + int i;

if (!IS_ENABLED(CONFIG_EFI)) return;

+ for (i = 0; i < ARRAY_SIZE(vl_info); i++) + max_vl = max(vl_info[i].max_vl, max_vl); + /* * alloc_percpu() warns and prints a backtrace if this goes wrong. * This is evidence of a crippled system and we are returning void, * so no attempt is made to handle this situation here. */ - if (!sve_vl_valid(info->max_vl)) + if (!sve_vl_valid(max_vl)) goto fail;

efi_sve_state = __alloc_percpu( - SVE_SIG_REGS_SIZE(sve_vq_from_vl(info->max_vl)), SVE_VQ_BYTES); + SVE_SIG_REGS_SIZE(sve_vq_from_vl(max_vl)), SVE_VQ_BYTES); if (!efi_sve_state) goto fail;

@@ -1841,6 +1845,7 @@ EXPORT_SYMBOL(kernel_neon_end); static DEFINE_PER_CPU(struct user_fpsimd_state, efi_fpsimd_state); static DEFINE_PER_CPU(bool, efi_fpsimd_state_used); static DEFINE_PER_CPU(bool, efi_sve_state_used); +static DEFINE_PER_CPU(bool, efi_sm_state);

/* * EFI runtime services support functions @@ -1875,12 +1880,28 @@ void __efi_fpsimd_begin(void) */ if (system_supports_sve() && likely(efi_sve_state)) { char *sve_state = this_cpu_ptr(efi_sve_state); + bool ffr = true; + u64 svcr;

__this_cpu_write(efi_sve_state_used, true);

+ /* If we are in streaming mode don't touch FFR */ + if (system_supports_sme()) { + svcr = read_sysreg_s(SYS_SVCR_EL0); + + ffr = svcr & SYS_SVCR_EL0_SM_MASK; + + __this_cpu_write(efi_sm_state, ffr); + } + sve_save_state(sve_state + sve_ffr_offset(sve_max_vl()), &this_cpu_ptr(&efi_fpsimd_state)->fpsr, - true); + ffr); + + if (system_supports_sme()) + sysreg_clear_set_s(SYS_SVCR_EL0, + SYS_SVCR_EL0_SM_MASK, 0); + } else { fpsimd_save_state(this_cpu_ptr(&efi_fpsimd_state)); } @@ -1903,11 +1924,25 @@ void __efi_fpsimd_end(void) if (system_supports_sve() && likely(__this_cpu_read(efi_sve_state_used))) { char const *sve_state = this_cpu_ptr(efi_sve_state); + bool ffr = true; + + /* + * Restore streaming mode; EFI calls are + * normal function calls so should not return in + * streaming mode. + */ + if (system_supports_sme()) { + if (__this_cpu_read(efi_sm_state)) { + sysreg_clear_set_s(SYS_SVCR_EL0, + 0, + SYS_SVCR_EL0_SM_MASK); + ffr = false; + } + }

- sve_set_vq(sve_vq_from_vl(sve_get_vl()) - 1); sve_load_state(sve_state + sve_ffr_offset(sve_max_vl()), &this_cpu_ptr(&efi_fpsimd_state)->fpsr, - true); + ffr);

__this_cpu_write(efi_sve_state_used, false); } else {

The Scalable Matrix Extenions (SME) introduces additional register state with configurable vector lengths, similar to SVE but configured separately. Extend vlset to support configuring this state with a --sme or -s command line option.

Signed-off-by: Mark Brown broonie@kernel.org --- tools/testing/selftests/arm64/fp/vlset.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/arm64/fp/vlset.c b/tools/testing/selftests/arm64/fp/vlset.c index 308d27a68226..76912a581a95 100644 --- a/tools/testing/selftests/arm64/fp/vlset.c +++ b/tools/testing/selftests/arm64/fp/vlset.c @@ -22,12 +22,15 @@ static int inherit = 0; static int no_inherit = 0; static int force = 0; static unsigned long vl; +static int set_ctl = PR_SVE_SET_VL; +static int get_ctl = PR_SVE_GET_VL;

static const struct option options[] = { { "force", no_argument, NULL, 'f' }, { "inherit", no_argument, NULL, 'i' }, { "max", no_argument, NULL, 'M' }, { "no-inherit", no_argument, &no_inherit, 1 }, + { "sme", no_argument, NULL, 's' }, { "help", no_argument, NULL, '?' }, {} }; @@ -50,6 +53,9 @@ static int parse_options(int argc, char **argv) case 'M': vl = SVE_VL_MAX; break; case 'f': force = 1; break; case 'i': inherit = 1; break; + case 's': set_ctl = PR_SME_SET_VL; + get_ctl = PR_SME_GET_VL; + break; case 0: break; default: goto error; } @@ -125,14 +131,14 @@ int main(int argc, char **argv) if (inherit) flags |= PR_SVE_VL_INHERIT;

- t = prctl(PR_SVE_SET_VL, vl | flags); + t = prctl(set_ctl, vl | flags); if (t < 0) { fprintf(stderr, "%s: PR_SVE_SET_VL: %s\n", program_name, strerror(errno)); goto error; }

- t = prctl(PR_SVE_GET_VL); + t = prctl(get_ctl); if (t == -1) { fprintf(stderr, "%s: PR_SVE_GET_VL: %s\n", program_name, strerror(errno));

Provide RDVL helpers for SME and extend the main vector configuration tests to cover SME.

Signed-off-by: Mark Brown broonie@kernel.org --- tools/testing/selftests/arm64/fp/.gitignore | 1 + tools/testing/selftests/arm64/fp/Makefile | 3 ++- tools/testing/selftests/arm64/fp/rdvl-sme.c | 14 ++++++++++++++ tools/testing/selftests/arm64/fp/rdvl.S | 16 ++++++++++++++++ tools/testing/selftests/arm64/fp/rdvl.h | 1 + tools/testing/selftests/arm64/fp/vec-syscfg.c | 10 ++++++++++ 6 files changed, 44 insertions(+), 1 deletion(-) create mode 100644 tools/testing/selftests/arm64/fp/rdvl-sme.c

diff --git a/tools/testing/selftests/arm64/fp/.gitignore b/tools/testing/selftests/arm64/fp/.gitignore index b67395903b9b..885dd592807b 100644 --- a/tools/testing/selftests/arm64/fp/.gitignore +++ b/tools/testing/selftests/arm64/fp/.gitignore @@ -1,4 +1,5 @@ fpsimd-test +rdvl-sme rdvl-sve sve-probe-vls sve-ptrace diff --git a/tools/testing/selftests/arm64/fp/Makefile b/tools/testing/selftests/arm64/fp/Makefile index ba1488c7c315..11d4fc3b7115 100644 --- a/tools/testing/selftests/arm64/fp/Makefile +++ b/tools/testing/selftests/arm64/fp/Makefile @@ -3,7 +3,7 @@ CFLAGS += -I../../../../../usr/include/ TEST_GEN_PROGS := sve-ptrace sve-probe-vls vec-syscfg TEST_PROGS_EXTENDED := fpsimd-test fpsimd-stress \ - rdvl-sve \ + rdvl-sme rdvl-sve \ sve-test sve-stress \ vlset

@@ -11,6 +11,7 @@ all: $(TEST_GEN_PROGS) $(TEST_PROGS_EXTENDED)

fpsimd-test: fpsimd-test.o asm-utils.o $(CC) -nostdlib $^ -o $@ +rdvl-sme: rdvl-sme.o rdvl.o rdvl-sve: rdvl-sve.o rdvl.o sve-ptrace: sve-ptrace.o sve-probe-vls: sve-probe-vls.o rdvl.o diff --git a/tools/testing/selftests/arm64/fp/rdvl-sme.c b/tools/testing/selftests/arm64/fp/rdvl-sme.c new file mode 100644 index 000000000000..49b0b2e08bac --- /dev/null +++ b/tools/testing/selftests/arm64/fp/rdvl-sme.c @@ -0,0 +1,14 @@ +// SPDX-License-Identifier: GPL-2.0-only + +#include <stdio.h> + +#include "rdvl.h" + +int main(void) +{ + int vl = rdvl_sme(); + + printf("%d\n", vl); + + return 0; +} diff --git a/tools/testing/selftests/arm64/fp/rdvl.S b/tools/testing/selftests/arm64/fp/rdvl.S index c916c1c9defd..a9a028ba1b79 100644 --- a/tools/testing/selftests/arm64/fp/rdvl.S +++ b/tools/testing/selftests/arm64/fp/rdvl.S @@ -8,3 +8,19 @@ rdvl_sve: hint 34 // BTI C rdvl x0, #1 ret + +.globl rdvl_sme +rdvl_sme: + hint 34 // BTI C + + // Enter streaming mode + mov x16, #1 + msr S3_3_C4_C2_2, x16 + + rdvl x0, #1 + + // Leave streaming mode + mov x16, #0 + msr S3_3_C4_C2_2, x16 + + ret diff --git a/tools/testing/selftests/arm64/fp/rdvl.h b/tools/testing/selftests/arm64/fp/rdvl.h index 7c9d953fc9e7..5d323679fbc9 100644 --- a/tools/testing/selftests/arm64/fp/rdvl.h +++ b/tools/testing/selftests/arm64/fp/rdvl.h @@ -3,6 +3,7 @@ #ifndef RDVL_H #define RDVL_H

+int rdvl_sme(void); int rdvl_sve(void);

#endif diff --git a/tools/testing/selftests/arm64/fp/vec-syscfg.c b/tools/testing/selftests/arm64/fp/vec-syscfg.c index 272b888e018e..0b976eb1c1d1 100644 --- a/tools/testing/selftests/arm64/fp/vec-syscfg.c +++ b/tools/testing/selftests/arm64/fp/vec-syscfg.c @@ -53,6 +53,16 @@ static struct vec_data vec_data[] = { .prctl_set = PR_SVE_SET_VL, .default_vl_file = "/proc/sys/abi/sve_default_vector_length", }, + { + .name = "SME", + .hwcap_type = AT_HWCAP2, + .hwcap = HWCAP2_SME, + .rdvl = rdvl_sme, + .rdvl_binary = "./rdvl-sme", + .prctl_get = PR_SME_GET_VL, + .prctl_set = PR_SME_SET_VL, + .default_vl_file = "/proc/sys/abi/sme_default_vector_length", + }, };

static int stdio_read_integer(FILE *f, const char *what, int *val)

Add a stress test for context switching of the ZA register state based on the similar tests Dave Martin wrote for FPSIMD and SVE registers. The test loops indefinitely writing a data pattern to ZA then reading it back and verifying that it's what was expected.

Unlike the other tests we manually assemble the SME instructions since at present no released toolchain has SME support integrated.

Signed-off-by: Mark Brown broonie@kernel.org --- tools/testing/selftests/arm64/fp/.gitignore | 1 + tools/testing/selftests/arm64/fp/Makefile | 3 + tools/testing/selftests/arm64/fp/za-stress | 59 +++ tools/testing/selftests/arm64/fp/za-test.S | 431 ++++++++++++++++++++ 4 files changed, 494 insertions(+) create mode 100644 tools/testing/selftests/arm64/fp/za-stress create mode 100644 tools/testing/selftests/arm64/fp/za-test.S

diff --git a/tools/testing/selftests/arm64/fp/.gitignore b/tools/testing/selftests/arm64/fp/.gitignore index 73c600e1ab81..1178fecc7aa1 100644 --- a/tools/testing/selftests/arm64/fp/.gitignore +++ b/tools/testing/selftests/arm64/fp/.gitignore @@ -7,3 +7,4 @@ sve-test ssve-test vec-syscfg vlset +za-test diff --git a/tools/testing/selftests/arm64/fp/Makefile b/tools/testing/selftests/arm64/fp/Makefile index 6d9e4d1922e4..d77e9903116b 100644 --- a/tools/testing/selftests/arm64/fp/Makefile +++ b/tools/testing/selftests/arm64/fp/Makefile @@ -6,6 +6,7 @@ TEST_PROGS_EXTENDED := fpsimd-test fpsimd-stress \ rdvl-sme rdvl-sve \ sve-test sve-stress \ ssve-test ssve-stress \ + za-test za-stress \ vlset

all: $(TEST_GEN_PROGS) $(TEST_PROGS_EXTENDED) @@ -22,5 +23,7 @@ ssve-test: sve-test.S asm-utils.o $(CC) -DSSVE -nostdlib $^ -o $@ vec-syscfg: vec-syscfg.o rdvl.o vlset: vlset.o +za-test: za-test.o asm-utils.o + $(CC) -nostdlib $^ -o $@

include ../../lib.mk diff --git a/tools/testing/selftests/arm64/fp/za-stress b/tools/testing/selftests/arm64/fp/za-stress new file mode 100644 index 000000000000..5ac386b55b95 --- /dev/null +++ b/tools/testing/selftests/arm64/fp/za-stress @@ -0,0 +1,59 @@ +#!/bin/bash +# SPDX-License-Identifier: GPL-2.0-only +# Copyright (C) 2015-2019 ARM Limited. +# Original author: Dave Martin Dave.Martin@arm.com + +set -ue + +NR_CPUS=`nproc` + +pids= +logs= + +cleanup () { + trap - INT TERM CHLD + set +e + + if [ -n "$pids" ]; then + kill $pids + wait $pids + pids= + fi + + if [ -n "$logs" ]; then + cat $logs + rm $logs + logs= + fi +} + +interrupt () { + cleanup + exit 0 +} + +child_died () { + cleanup + exit 1 +} + +trap interrupt INT TERM EXIT + +for x in `seq 0 $((NR_CPUS * 4))`; do + log=`mktemp` + logs=$logs\ $log + ./za-test >$log & + pids=$pids\ $! +done + +# Wait for all child processes to be created: +sleep 10 + +while :; do + kill -USR1 $pids +done & +pids=$pids\ $! + +wait + +exit 1 diff --git a/tools/testing/selftests/arm64/fp/za-test.S b/tools/testing/selftests/arm64/fp/za-test.S new file mode 100644 index 000000000000..76bf49f1c13d --- /dev/null +++ b/tools/testing/selftests/arm64/fp/za-test.S @@ -0,0 +1,431 @@ +// SPDX-License-Identifier: GPL-2.0-only +// Copyright (C) 2021 ARM Limited. +// Original author: Mark Brown broonie@kernel.org +// +// Scalable Matrix Extension ZA context switch test +// Repeatedly writes unique test patterns into each ZA tile +// and reads them back to verify integrity. +// +// for x in `seq 1 NR_CPUS`; do sve-test & pids=$pids\ $! ; done +// (leave it running for as long as you want...) +// kill $pids + +#include <asm/unistd.h> +#include "assembler.h" +#include "asm-offsets.h" + +.arch_extension sve + +#define MAXVL 2048 +#define MAXVL_B (MAXVL / 8) + +/* + * LDR (vector to ZA array): + * LDR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _ldr_za nw, nxbase, offset=0 + .inst 0xe1000000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +/* + * STR (vector from ZA array): + * STR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _str_za nw, nxbase, offset=0 + .inst 0xe1200000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +.macro smstart + msr S0_3_C4_C7_3, xzr +.endm + +.macro smstart_sm + msr S0_3_C4_C3_3, xzr +.endm + +.macro smstop + msr S0_3_C4_C6_3, xzr +.endm + +.macro smstop_sm + msr S0_3_C4_C2_3, xzr +.endm + +// Declare some storage space to shadow ZA register contents and a +// scratch buffer for a vector. +.pushsection .text +.data +.align 4 +zaref: + .space MAXVL_B * MAXVL_B +scratch: + .space MAXVL_B +.popsection + +// Trivial memory copy: copy x2 bytes, starting at address x1, to address x0. +// Clobbers x0-x3 +function memcpy + cmp x2, #0 + b.eq 1f +0: ldrb w3, [x1], #1 + strb w3, [x0], #1 + subs x2, x2, #1 + b.ne 0b +1: ret +endfunction + +// Generate a test pattern for storage in ZA +// x0: pid +// x1: row in ZA +// x2: generation + +// These values are used to constuct a 32-bit pattern that is repeated in the +// scratch buffer as many times as will fit: +// bits 31:28 generation number (increments once per test_loop) +// bits 27:16 pid +// bits 15: 8 row number +// bits 7: 0 32-bit lane index + +function pattern + mov w3, wzr + bfi w3, w0, #16, #12 // PID + bfi w3, w1, #8, #8 // Row + bfi w3, w2, #28, #4 // Generation + + ldr x0, =scratch + mov w1, #MAXVL_B / 4 + +0: str w3, [x0], #4 + add w3, w3, #1 // Lane + subs w1, w1, #1 + b.ne 0b + + ret +endfunction + +// Get the address of shadow data for ZA horizontal vector xn +.macro _adrza xd, xn, nrtmp + ldr \xd, =zaref + smstart_sm + rdvl x\nrtmp, #1 + smstop_sm + madd \xd, x\nrtmp, \xn, \xd +.endm + +// Set up test pattern in a ZA horizontal vector +// x0: pid +// x1: row number +// x2: generation +function setup_za + mov x4, x30 + mov x12, x1 // Use x12 for vector select + + bl pattern // Get pattern in scratch buffer + _adrza x0, x12, 2 // Shadow buffer pointer to x0 and x5 + mov x5, x0 + ldr x1, =scratch + bl memcpy // length set up in x2 by _adrza + + _ldr_za 12, 5 // load vector w12 from pointer x5 + + ret x4 +endfunction + +// Trivial memory compare: compare x2 bytes starting at address x0 with +// bytes starting at address x1. +// Returns only if all bytes match; otherwise, the program is aborted. +// Clobbers x0-x5. +function memcmp + cbz x2, 2f + + stp x0, x1, [sp, #-0x20]! + str x2, [sp, #0x10] + + mov x5, #0 +0: ldrb w3, [x0, x5] + ldrb w4, [x1, x5] + add x5, x5, #1 + cmp w3, w4 + b.ne 1f + subs x2, x2, #1 + b.ne 0b + +1: ldr x2, [sp, #0x10] + ldp x0, x1, [sp], #0x20 + b.ne barf + +2: ret +endfunction + +// Verify that a ZA vector matches its shadow in memory, else abort +// x0: row number +// Clobbers x0-x7 and x12. +function check_za + mov x3, x30 + + mov x12, x0 + _adrza x5, x0, 6 // pointer to expected value in x5 + mov x4, x0 + ldr x7, =scratch // x7 is scratch + + mov x0, x7 // Poison scratch + mov x1, x6 + bl memfill_ae + + _str_za 12, 7 // save vector w12 to pointer x7 + + mov x0, x5 + mov x1, x7 + mov x2, x6 + mov x30, x3 + b memcmp +endfunction + +// Any SME register modified here can cause corruption in the main +// thread -- but *only* the locations modified here. +function irritator_handler + // Increment the irritation signal count (x23): + ldr x0, [x2, #ucontext_regs + 8 * 23] + add x0, x0, #1 + str x0, [x2, #ucontext_regs + 8 * 23] + + // Corrupt some random ZA data +#if 0 + adr x0, .text + (irritator_handler - .text) / 16 * 16 + movi v0.8b, #1 + movi v9.16b, #2 + movi v31.8b, #3 +#endif + + ret +endfunction + +function terminate_handler + mov w21, w0 + mov x20, x2 + + puts "Terminated by signal " + mov w0, w21 + bl putdec + puts ", no error, iterations=" + ldr x0, [x20, #ucontext_regs + 8 * 22] + bl putdec + puts ", signals=" + ldr x0, [x20, #ucontext_regs + 8 * 23] + bl putdecn + + mov x0, #0 + mov x8, #__NR_exit + svc #0 +endfunction + +// w0: signal number +// x1: sa_action +// w2: sa_flags +// Clobbers x0-x6,x8 +function setsignal + str x30, [sp, #-((sa_sz + 15) / 16 * 16 + 16)]! + + mov w4, w0 + mov x5, x1 + mov w6, w2 + + add x0, sp, #16 + mov x1, #sa_sz + bl memclr + + mov w0, w4 + add x1, sp, #16 + str w6, [x1, #sa_flags] + str x5, [x1, #sa_handler] + mov x2, #0 + mov x3, #sa_mask_sz + mov x8, #__NR_rt_sigaction + svc #0 + + cbz w0, 1f + + puts "sigaction failure\n" + b .Labort + +1: ldr x30, [sp], #((sa_sz + 15) / 16 * 16 + 16) + ret +endfunction + +// Main program entry point +.globl _start +function _start +_start: + puts "Streaming mode " + smstart + + // Sanity-check and report the vector length + + rdvl x19, #8 + cmp x19, #128 + b.lo 1f + cmp x19, #2048 + b.hi 1f + tst x19, #(8 - 1) + b.eq 2f + +1: puts "bad vector length: " + mov x0, x19 + bl putdecn + b .Labort + +2: puts "vector length:\t" + mov x0, x19 + bl putdec + puts " bits\n" + + // Obtain our PID, to ensure test pattern uniqueness between processes + mov x8, #__NR_getpid + svc #0 + mov x20, x0 + + puts "PID:\t" + mov x0, x20 + bl putdecn + + mov x23, #0 // Irritation signal count + + mov w0, #SIGINT + adr x1, terminate_handler + mov w2, #SA_SIGINFO + bl setsignal + + mov w0, #SIGTERM + adr x1, terminate_handler + mov w2, #SA_SIGINFO + bl setsignal + + mov w0, #SIGUSR1 + adr x1, irritator_handler + mov w2, #SA_SIGINFO + orr w2, w2, #SA_NODEFER + bl setsignal + + mov x22, #0 // generation number, increments per iteration +.Ltest_loop: + smstart_sm // printing/signals/yielding dropped out of SM + rdvl x0, #8 + cmp x0, x19 + b.ne vl_barf + + rdvl x21, #1 // Set up ZA & shadow with test pattern + smstop_sm +0: mov x0, x20 + sub x1, x21, #1 + mov x2, x22 + bl setup_za + subs x21, x21, #1 + b.ne 0b + + and x8, x22, #127 // Every 128 interations... + cbz x8, 0f + mov x8, #__NR_getpid // (otherwise minimal syscall) + b 1f +0: + mov x8, #__NR_sched_yield // ...encourage preemption +1: + svc #0 + + mrs x0, S3_3_C4_C2_2 // SVCR should have ZA=1,SM=0 + and x1, x0, #3 + cmp x1, #2 + b.ne svcr_barf + + smstart_sm + rdvl x21, #1 // Verify that the data made it through + rdvl x24, #1 // Verify that the data made it through + smstop_sm +0: sub x0, x24, x21 + bl check_za + subs x21, x21, #1 + bne 0b + + add x22, x22, #1 // Everything still working + b .Ltest_loop + +.Labort: + mov x0, #0 + mov x1, #SIGABRT + mov x8, #__NR_kill + svc #0 +endfunction + +function barf +// fpsimd.c acitivty log dump hack +// ldr w0, =0xdeadc0de +// mov w8, #__NR_exit +// svc #0 +// end hack + smstop + mov x10, x0 // expected data + mov x11, x1 // actual data + mov x12, x2 // data size + + puts "Mismatch: PID=" + mov x0, x20 + bl putdec + puts ", iteration=" + mov x0, x22 + bl putdec + puts ", row=" + mov x0, x21 + bl putdecn + puts "\tExpected [" + mov x0, x10 + mov x1, x12 + bl dumphex + puts "]\n\tGot [" + mov x0, x11 + mov x1, x12 + bl dumphex + puts "]\n" + + mov x8, #__NR_getpid + svc #0 +// fpsimd.c acitivty log dump hack +// ldr w0, =0xdeadc0de +// mov w8, #__NR_exit +// svc #0 +// ^ end of hack + mov x1, #SIGABRT + mov x8, #__NR_kill + svc #0 +// mov x8, #__NR_exit +// mov x1, #1 +// svc #0 +endfunction + +function vl_barf + mov x10, x0 + + puts "Bad active VL: " + mov x0, x10 + bl putdecn + + mov x8, #__NR_exit + mov x1, #1 + svc #0 +endfunction + +function svcr_barf + mov x10, x0 + + puts "Bad SVCR: " + mov x0, x10 + bl putdecn + + mov x8, #__NR_exit + mov x1, #1 + svc #0 +endfunction

In order to allow ptrace of streaming mode SVE registers we have added a new regset for streaming mode which in isolation offers the same ABI as regular SVE with a different vector type. Add this to the array of regsets we handle, together with additional tests for the interoperation of the two regsets.

Signed-off-by: Mark Brown broonie@kernel.org --- tools/testing/selftests/arm64/fp/sve-ptrace.c | 11 +++++++++++ 1 file changed, 11 insertions(+)

diff --git a/tools/testing/selftests/arm64/fp/sve-ptrace.c b/tools/testing/selftests/arm64/fp/sve-ptrace.c index af798b9d232c..00c35b9db27f 100644 --- a/tools/testing/selftests/arm64/fp/sve-ptrace.c +++ b/tools/testing/selftests/arm64/fp/sve-ptrace.c @@ -28,6 +28,10 @@ #define NT_ARM_SVE 0x405 #endif

+#ifndef NT_ARM_SSVE +#define NT_ARM_SSVE 0x40b +#endif + struct vec_type { const char *name; unsigned long hwcap_type; @@ -44,6 +48,13 @@ static const struct vec_type vec_types[] = { .regset = NT_ARM_SVE, .prctl_set = PR_SVE_SET_VL, }, + { + .name = "Streaming SVE", + .hwcap_type = AT_HWCAP2, + .hwcap = HWCAP2_SME, + .regset = NT_ARM_SSVE, + .prctl_set = PR_SME_SET_VL, + }, };

#define VL_TESTS (((SVE_VQ_MAX - SVE_VQ_MIN) + 1) * 3)

For every possible combination of SVE and SME vector length verify that for each possible value of SVCR after a syscall we leave streaming mode and ZA is preserved. We don't need to take account of any streaming/non streaming SVE vector length changes in the assembler code since the store instructions will handle the vector length for us. We log if the system supports FA64 and only try to set FFR in streaming mode if it does.

Signed-off-by: Mark Brown broonie@kernel.org --- .../selftests/arm64/abi/syscall-abi-asm.S | 69 +++++- .../testing/selftests/arm64/abi/syscall-abi.c | 204 ++++++++++++++++-- 2 files changed, 250 insertions(+), 23 deletions(-)

diff --git a/tools/testing/selftests/arm64/abi/syscall-abi-asm.S b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S index 983467cfcee0..bc70e04224bf 100644 --- a/tools/testing/selftests/arm64/abi/syscall-abi-asm.S +++ b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S @@ -9,15 +9,42 @@ // invoked is configured in x8 of the input GPR data. // // x0: SVE VL, 0 for FP only +// x1: SME VL // // GPRs: gpr_in, gpr_out // FPRs: fpr_in, fpr_out // Zn: z_in, z_out // Pn: p_in, p_out // FFR: ffr_in, ffr_out +// ZA: za_in, za_out +// SVCR: svcr_in, svcr_out + +#include "syscall-abi.h"

.arch_extension sve

+/* + * LDR (vector to ZA array): + * LDR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _ldr_za nw, nxbase, offset=0 + .inst 0xe1000000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + +/* + * STR (vector from ZA array): + * STR ZA[\nw, #\offset], [X\nxbase, #\offset, MUL VL] + */ +.macro _str_za nw, nxbase, offset=0 + .inst 0xe1200000 \ + | (((\nw) & 3) << 13) \ + | ((\nxbase) << 5) \ + | ((\offset) & 7) +.endm + .globl do_syscall do_syscall: // Store callee saved registers x19-x29 (80 bytes) plus x0 and x1 @@ -30,6 +57,24 @@ do_syscall: stp x25, x26, [sp, #80] stp x27, x28, [sp, #96]

+ // Set SVCR if we're doing SME + cbz x1, 1f + adrp x2, svcr_in + ldr x2, [x2, :lo12:svcr_in] + msr S3_3_C4_C2_2, x2 +1: + + // Load ZA if it's enabled - uses x12 as scratch due to SME LDR + tbz x2, #SVCR_ZA_SHIFT, 1f + mov w12, #0 + ldr x2, =za_in +2: _ldr_za 12, 2 + add x2, x2, x1 + add x12, x12, #1 + cmp x1, x12 + bne 2b +1: + // Load GPRs x8-x28, and save our SP/FP for later comparison ldr x2, =gpr_in add x2, x2, #64 @@ -68,7 +113,7 @@ do_syscall: ldp q30, q31, [x2, #16 * 30] 1:

- // Load the SVE registers if we're doing SVE + // Load the SVE registers if we're doing SVE/SME cbz x0, 1f

ldr x2, =z_in @@ -105,9 +150,13 @@ do_syscall: ldr z30, [x2, #30, MUL VL] ldr z31, [x2, #31, MUL VL]

+ // Only set a non-zero FFR, test patterns must be zero since the + // syscall should clear it - this lets us handle FA64. ldr x2, =ffr_in + cbz x2, 2f ldr p0, [x2, #0] wrffr p0.b +2:

ldr x2, =p_in ldr p0, [x2, #0, MUL VL] @@ -169,6 +218,24 @@ do_syscall: stp q28, q29, [x2, #16 * 28] stp q30, q31, [x2, #16 * 30]

+ // Save SVCR if we're doing SME + cbz x1, 1f + mrs x2, S3_3_C4_C2_2 + adrp x3, svcr_out + str x2, [x3, :lo12:svcr_out] +1: + + // Save ZA if it's enabled - uses x12 as scratch due to SME STR + tbz x2, #SVCR_ZA_SHIFT, 1f + mov w12, #0 + ldr x2, =za_out +2: _str_za 12, 2 + add x2, x2, x1 + add x12, x12, #1 + cmp x1, x12 + bne 2b +1: + // Save the SVE state if we have some cbz x0, 1f

diff --git a/tools/testing/selftests/arm64/abi/syscall-abi.c b/tools/testing/selftests/arm64/abi/syscall-abi.c index d103acf1ab79..ae38cb06a56c 100644 --- a/tools/testing/selftests/arm64/abi/syscall-abi.c +++ b/tools/testing/selftests/arm64/abi/syscall-abi.c @@ -18,10 +18,14 @@

#include "../../kselftest.h"

+#include "syscall-abi.h" + #define ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0])) #define NUM_VL ((SVE_VQ_MAX - SVE_VQ_MIN) + 1)

-extern void do_syscall(int sve_vl); +static int default_sme_vl; + +extern void do_syscall(int sve_vl, int sme_vl);

static void fill_random(void *buf, size_t size) { @@ -49,14 +53,15 @@ static struct syscall_cfg { uint64_t gpr_in[NUM_GPR]; uint64_t gpr_out[NUM_GPR];

-static void setup_gpr(struct syscall_cfg *cfg, int sve_vl) +static void setup_gpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { fill_random(gpr_in, sizeof(gpr_in)); gpr_in[8] = cfg->syscall_nr; memset(gpr_out, 0, sizeof(gpr_out)); }

-static int check_gpr(struct syscall_cfg *cfg, int sve_vl) +static int check_gpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, uint64_t svcr) { int errors = 0; int i; @@ -80,13 +85,15 @@ static int check_gpr(struct syscall_cfg *cfg, int sve_vl) uint64_t fpr_in[NUM_FPR * 2]; uint64_t fpr_out[NUM_FPR * 2];

-static void setup_fpr(struct syscall_cfg *cfg, int sve_vl) +static void setup_fpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { fill_random(fpr_in, sizeof(fpr_in)); memset(fpr_out, 0, sizeof(fpr_out)); }

-static int check_fpr(struct syscall_cfg *cfg, int sve_vl) +static int check_fpr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { int errors = 0; int i; @@ -110,13 +117,15 @@ static uint8_t z_zero[__SVE_ZREG_SIZE(SVE_VQ_MAX)]; uint8_t z_in[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; uint8_t z_out[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)];

-static void setup_z(struct syscall_cfg *cfg, int sve_vl) +static void setup_z(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { fill_random(z_in, sizeof(z_in)); fill_random(z_out, sizeof(z_out)); }

-static int check_z(struct syscall_cfg *cfg, int sve_vl) +static int check_z(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { size_t reg_size = sve_vl; int errors = 0; @@ -127,13 +136,17 @@ static int check_z(struct syscall_cfg *cfg, int sve_vl)

/* * After a syscall the low 128 bits of the Z registers should - * be preserved and the rest be zeroed. + * be preserved and the rest be zeroed, except if we were in + * streaming mode in which case the low 128 bits may also be + * cleared by the transition out of streaming mode. */ for (i = 0; i < SVE_NUM_ZREGS; i++) { void *in = &z_in[reg_size * i]; void *out = &z_out[reg_size * i];

- if (memcmp(in, out, SVE_VQ_BYTES) != 0) { + if ((memcmp(in, out, SVE_VQ_BYTES) != 0) && + !((svcr & SVCR_SM_MASK) && + memcmp(z_zero, out, SVE_VQ_BYTES) == 0)) { ksft_print_msg("%s SVE VL %d Z%d low 128 bits changed\n", cfg->name, sve_vl, i); errors++; @@ -153,13 +166,15 @@ static int check_z(struct syscall_cfg *cfg, int sve_vl) uint8_t p_in[SVE_NUM_PREGS * __SVE_PREG_SIZE(SVE_VQ_MAX)]; uint8_t p_out[SVE_NUM_PREGS * __SVE_PREG_SIZE(SVE_VQ_MAX)];

-static void setup_p(struct syscall_cfg *cfg, int sve_vl) +static void setup_p(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { fill_random(p_in, sizeof(p_in)); fill_random(p_out, sizeof(p_out)); }

-static int check_p(struct syscall_cfg *cfg, int sve_vl) +static int check_p(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { size_t reg_size = sve_vq_from_vl(sve_vl) * 2; /* 1 bit per VL byte */

@@ -183,8 +198,19 @@ static int check_p(struct syscall_cfg *cfg, int sve_vl) uint8_t ffr_in[__SVE_PREG_SIZE(SVE_VQ_MAX)]; uint8_t ffr_out[__SVE_PREG_SIZE(SVE_VQ_MAX)];

-static void setup_ffr(struct syscall_cfg *cfg, int sve_vl) +static void setup_ffr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { + /* + * If we are in streaming mode and do not have FA64 then FFR + * is unavailable. + */ + if ((svcr & SVCR_SM_MASK) && + !(getauxval(AT_HWCAP2) & HWCAP2_SME_FA64)) { + memset(&ffr_in, 0, sizeof(ffr_in)); + return; + } + /* * It is only valid to set a contiguous set of bits starting * at 0. For now since we're expecting this to be cleared by @@ -194,7 +220,8 @@ static void setup_ffr(struct syscall_cfg *cfg, int sve_vl) fill_random(ffr_out, sizeof(ffr_out)); }

-static int check_ffr(struct syscall_cfg *cfg, int sve_vl) +static int check_ffr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { size_t reg_size = sve_vq_from_vl(sve_vl) * 2; /* 1 bit per VL byte */ int errors = 0; @@ -203,6 +230,10 @@ static int check_ffr(struct syscall_cfg *cfg, int sve_vl) if (!sve_vl) return 0;

+ if ((svcr & SVCR_SM_MASK) && + !(getauxval(AT_HWCAP2) & HWCAP2_SME_FA64)) + return 0; + /* After a syscall the P registers should be zeroed */ for (i = 0; i < reg_size; i++) if (ffr_out[i]) @@ -214,8 +245,65 @@ static int check_ffr(struct syscall_cfg *cfg, int sve_vl) return errors; }

-typedef void (*setup_fn)(struct syscall_cfg *cfg, int sve_vl); -typedef int (*check_fn)(struct syscall_cfg *cfg, int sve_vl); +uint64_t svcr_in, svcr_out; + +static void setup_svcr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) +{ + svcr_in = svcr; +} + +static int check_svcr(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) +{ + int errors = 0; + + if (svcr_out & SVCR_SM_MASK) { + ksft_print_msg("%s Still in SM, SVCR %llx\n", + cfg->name, svcr_out); + errors++; + } + + if ((svcr_in & SVCR_ZA_MASK) != (svcr_out & SVCR_ZA_MASK)) { + ksft_print_msg("%s PSTATE.ZA changed, SVCR %llx != %llx\n", + cfg->name, svcr_in, svcr_out); + errors++; + } + + return errors; +} + +uint8_t za_in[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; +uint8_t za_out[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; + +static void setup_za(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) +{ + fill_random(za_in, sizeof(za_in)); + memset(za_out, 0, sizeof(za_out)); +} + +static int check_za(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) +{ + size_t reg_size = sme_vl * sme_vl; + int errors = 0; + + if (!(svcr & SVCR_ZA_MASK)) + return 0; + + if (memcmp(za_in, za_out, reg_size) != 0) { + ksft_print_msg("SME VL %d ZA does not match\n", sme_vl); + errors++; + } + + return errors; +} + +typedef void (*setup_fn)(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr); +typedef int (*check_fn)(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr);

/* * Each set of registers has a setup function which is called before @@ -233,20 +321,23 @@ static struct { { setup_z, check_z }, { setup_p, check_p }, { setup_ffr, check_ffr }, + { setup_svcr, check_svcr }, + { setup_za, check_za }, };

-static bool do_test(struct syscall_cfg *cfg, int sve_vl) +static bool do_test(struct syscall_cfg *cfg, int sve_vl, int sme_vl, + uint64_t svcr) { int errors = 0; int i;

for (i = 0; i < ARRAY_SIZE(regset); i++) - regset[i].setup(cfg, sve_vl); + regset[i].setup(cfg, sve_vl, sme_vl, svcr);

- do_syscall(sve_vl); + do_syscall(sve_vl, sme_vl);

for (i = 0; i < ARRAY_SIZE(regset); i++) - errors += regset[i].check(cfg, sve_vl); + errors += regset[i].check(cfg, sve_vl, sme_vl, svcr);

return errors == 0; } @@ -254,9 +345,10 @@ static bool do_test(struct syscall_cfg *cfg, int sve_vl) static void test_one_syscall(struct syscall_cfg *cfg) { int sve_vq, sve_vl; + int sme_vq, sme_vl;

/* FPSIMD only case */ - ksft_test_result(do_test(cfg, 0), + ksft_test_result(do_test(cfg, 0, default_sme_vl, 0), "%s FPSIMD\n", cfg->name);

if (!(getauxval(AT_HWCAP) & HWCAP_SVE)) @@ -273,8 +365,36 @@ static void test_one_syscall(struct syscall_cfg *cfg) if (sve_vq != sve_vq_from_vl(sve_vl)) sve_vq = sve_vq_from_vl(sve_vl);

- ksft_test_result(do_test(cfg, sve_vl), + ksft_test_result(do_test(cfg, sve_vl, default_sme_vl, 0), "%s SVE VL %d\n", cfg->name, sve_vl); + + if (!(getauxval(AT_HWCAP2) & HWCAP2_SME)) + continue; + + for (sme_vq = SVE_VQ_MAX; sme_vq > 0; --sme_vq) { + sme_vl = prctl(PR_SME_SET_VL, sme_vq * 16); + if (sme_vl == -1) + ksft_exit_fail_msg("PR_SME_SET_VL failed: %s (%d)\n", + strerror(errno), errno); + + sme_vl &= PR_SME_VL_LEN_MASK; + + if (sme_vq != sve_vq_from_vl(sme_vl)) + sme_vq = sve_vq_from_vl(sme_vl); + + ksft_test_result(do_test(cfg, sve_vl, sme_vl, + SVCR_ZA_MASK | SVCR_SM_MASK), + "%s SVE VL %d/SME VL %d SM+ZA\n", + cfg->name, sve_vl, sme_vl); + ksft_test_result(do_test(cfg, sve_vl, sme_vl, + SVCR_SM_MASK), + "%s SVE VL %d/SME VL %d SM\n", + cfg->name, sve_vl, sme_vl); + ksft_test_result(do_test(cfg, sve_vl, sme_vl, + SVCR_ZA_MASK), + "%s SVE VL %d/SME VL %d ZA\n", + cfg->name, sve_vl, sme_vl); + } } }

@@ -307,14 +427,54 @@ int sve_count_vls(void) return vl_count; }

+int sme_count_vls(void) +{ + unsigned int vq; + int vl_count = 0; + int vl; + + if (!(getauxval(AT_HWCAP2) & HWCAP2_SME)) + return 0; + + /* Ensure we configure a SME VL, used to flag if SVCR is set */ + default_sme_vl = 16; + + /* + * Enumerate up to SVE_VQ_MAX vector lengths + */ + for (vq = SVE_VQ_MAX; vq > 0; --vq) { + vl = prctl(PR_SME_SET_VL, vq * 16); + if (vl == -1) + ksft_exit_fail_msg("PR_SME_SET_VL failed: %s (%d)\n", + strerror(errno), errno); + + vl &= PR_SME_VL_LEN_MASK; + + if (vq != sve_vq_from_vl(vl)) + vq = sve_vq_from_vl(vl); + + vl_count++; + } + + return vl_count; +} + int main(void) { int i; + int tests = 1; /* FPSIMD */

srandom(getpid());

ksft_print_header(); - ksft_set_plan(ARRAY_SIZE(syscalls) * (sve_count_vls() + 1)); + tests += sve_count_vls(); + tests += (sve_count_vls() * sme_count_vls()) * 3; + ksft_set_plan(ARRAY_SIZE(syscalls) * tests); + + if (getauxval(AT_HWCAP2) & HWCAP2_SME_FA64) + ksft_print_msg("SME with FA64\n"); + else if (getauxval(AT_HWCAP2) & HWCAP2_SME) + ksft_print_msg("SME without FA64\n");

for (i = 0; i < ARRAY_SIZE(syscalls); i++) test_one_syscall(&syscalls[i]);