arm64: Scalable Vector Extension (SVE) support.
Change-Id: I3568687913f583edfaa297d5cf5ac91d319d97e9
This commit is contained in:
committed by
Masamichi Takagi
parent
dac99f708c
commit
07aa96ef95
@@ -1,4 +1,4 @@
|
||||
/* fpsimd.h COPYRIGHT FUJITSU LIMITED 2016-2017 */
|
||||
/* fpsimd.h COPYRIGHT FUJITSU LIMITED 2016-2019 */
|
||||
#ifndef __HEADER_ARM64_COMMON_FPSIMD_H
|
||||
#define __HEADER_ARM64_COMMON_FPSIMD_H
|
||||
|
||||
@@ -46,12 +46,15 @@ extern void sve_alloc(struct thread *thread);
|
||||
extern void sve_save_state(void *state, unsigned int *pfpsr);
|
||||
extern void sve_load_state(void const *state, unsigned int const *pfpsr, unsigned long vq_minus_1);
|
||||
extern unsigned int sve_get_vl(void);
|
||||
extern int sve_set_thread_vl(struct thread *thread, const unsigned long vector_length, const unsigned long flags);
|
||||
extern int sve_get_thread_vl(const struct thread *thread);
|
||||
extern int sve_set_thread_vl(unsigned long arg);
|
||||
extern int sve_get_thread_vl(void);
|
||||
extern int sve_set_vector_length(struct thread *thread, unsigned long vl, unsigned long flags);
|
||||
|
||||
#define SVE_SET_VL(thread, vector_length, flags) sve_set_thread_vl(thread, vector_length, flags)
|
||||
#define SVE_GET_VL(thread) sve_get_thread_vl(thread)
|
||||
#define SVE_SET_VL(arg) sve_set_thread_vl(arg)
|
||||
#define SVE_GET_VL() sve_get_thread_vl()
|
||||
|
||||
/* Maximum VL that SVE VL-agnostic software can transparently support */
|
||||
#define SVE_VL_ARCH_MAX 0x100
|
||||
|
||||
#else /* CONFIG_ARM64_SVE */
|
||||
|
||||
@@ -80,12 +83,12 @@ static int sve_set_vector_length(struct thread *thread, unsigned long vl, unsign
|
||||
}
|
||||
|
||||
/* for prctl syscall */
|
||||
#define SVE_SET_VL(a,b,c) (-EINVAL)
|
||||
#define SVE_GET_VL(a) (-EINVAL)
|
||||
#define SVE_SET_VL(a) (-EINVAL)
|
||||
#define SVE_GET_VL() (-EINVAL)
|
||||
|
||||
#endif /* CONFIG_ARM64_SVE */
|
||||
|
||||
extern void init_sve_vl(void);
|
||||
extern void sve_setup(void);
|
||||
extern void fpsimd_save_state(struct fpsimd_state *state);
|
||||
extern void fpsimd_load_state(struct fpsimd_state *state);
|
||||
extern void thread_fpsimd_save(struct thread *thread);
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
/* prctl.h COPYRIGHT FUJITSU LIMITED 2017 */
|
||||
/* prctl.h COPYRIGHT FUJITSU LIMITED 2017-2019 */
|
||||
#ifndef __HEADER_ARM64_COMMON_PRCTL_H
|
||||
#define __HEADER_ARM64_COMMON_PRCTL_H
|
||||
|
||||
@@ -6,15 +6,12 @@
|
||||
#define PR_GET_THP_DISABLE 42
|
||||
|
||||
/* arm64 Scalable Vector Extension controls */
|
||||
#define PR_SVE_SET_VL 48 /* set task vector length */
|
||||
#define PR_SVE_SET_VL_THREAD (1 << 1) /* set just this thread */
|
||||
#define PR_SVE_SET_VL_INHERIT (1 << 2) /* inherit across exec */
|
||||
#define PR_SVE_SET_VL_ONEXEC (1 << 3) /* defer effect until exec */
|
||||
|
||||
#define PR_SVE_GET_VL 49 /* get task vector length */
|
||||
/* Decode helpers for the return value from PR_SVE_GET_VL: */
|
||||
#define PR_SVE_GET_VL_LEN(ret) ((ret) & 0x3fff) /* vector length */
|
||||
#define PR_SVE_GET_VL_INHERIT (PR_SVE_SET_VL_INHERIT << 16)
|
||||
/* For conveinence, PR_SVE_SET_VL returns the result in the same encoding */
|
||||
/* Flag values must be kept in sync with ptrace NT_ARM_SVE interface */
|
||||
#define PR_SVE_SET_VL 50 /* set task vector length */
|
||||
# define PR_SVE_SET_VL_ONEXEC (1 << 18) /* defer effect until exec */
|
||||
#define PR_SVE_GET_VL 51 /* get task vector length */
|
||||
/* Bits common to PR_SVE_SET_VL and PR_SVE_GET_VL */
|
||||
# define PR_SVE_VL_LEN_MASK 0xffff
|
||||
# define PR_SVE_VL_INHERIT (1 << 17) /* inherit across exec */
|
||||
|
||||
#endif /* !__HEADER_ARM64_COMMON_PRCTL_H */
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
/* ptrace.h COPYRIGHT FUJITSU LIMITED 2015-2017 */
|
||||
/* ptrace.h COPYRIGHT FUJITSU LIMITED 2015-2019 */
|
||||
#ifndef __HEADER_ARM64_COMMON_PTRACE_H
|
||||
#define __HEADER_ARM64_COMMON_PTRACE_H
|
||||
|
||||
@@ -46,6 +46,7 @@
|
||||
|
||||
#ifndef __ASSEMBLY__
|
||||
|
||||
#include <lwk/compiler.h>
|
||||
#include <ihk/types.h>
|
||||
|
||||
struct user_hwdebug_state {
|
||||
@@ -78,6 +79,70 @@ struct user_sve_header {
|
||||
uint16_t __reserved;
|
||||
};
|
||||
|
||||
enum aarch64_regset {
|
||||
REGSET_GPR,
|
||||
REGSET_FPR,
|
||||
REGSET_TLS,
|
||||
REGSET_HW_BREAK,
|
||||
REGSET_HW_WATCH,
|
||||
REGSET_SYSTEM_CALL,
|
||||
#ifdef CONFIG_ARM64_SVE
|
||||
REGSET_SVE,
|
||||
#endif /* CONFIG_ARM64_SVE */
|
||||
};
|
||||
|
||||
struct thread;
|
||||
struct user_regset;
|
||||
|
||||
typedef int user_regset_active_fn(struct thread *target,
|
||||
const struct user_regset *regset);
|
||||
|
||||
typedef long user_regset_get_fn(struct thread *target,
|
||||
const struct user_regset *regset,
|
||||
unsigned int pos, unsigned int count,
|
||||
void *kbuf, void __user *ubuf);
|
||||
|
||||
typedef long user_regset_set_fn(struct thread *target,
|
||||
const struct user_regset *regset,
|
||||
unsigned int pos, unsigned int count,
|
||||
const void *kbuf, const void __user *ubuf);
|
||||
|
||||
typedef int user_regset_writeback_fn(struct thread *target,
|
||||
const struct user_regset *regset,
|
||||
int immediate);
|
||||
|
||||
typedef unsigned int user_regset_get_size_fn(struct thread *target,
|
||||
const struct user_regset *regset);
|
||||
|
||||
struct user_regset {
|
||||
user_regset_get_fn *get;
|
||||
user_regset_set_fn *set;
|
||||
user_regset_active_fn *active;
|
||||
user_regset_writeback_fn *writeback;
|
||||
user_regset_get_size_fn *get_size;
|
||||
unsigned int n;
|
||||
unsigned int size;
|
||||
unsigned int align;
|
||||
unsigned int bias;
|
||||
unsigned int core_note_type;
|
||||
};
|
||||
|
||||
struct user_regset_view {
|
||||
const char *name;
|
||||
const struct user_regset *regsets;
|
||||
unsigned int n;
|
||||
uint32_t e_flags;
|
||||
uint16_t e_machine;
|
||||
uint8_t ei_osabi;
|
||||
};
|
||||
|
||||
extern const struct user_regset_view *current_user_regset_view(void);
|
||||
extern const struct user_regset *find_regset(
|
||||
const struct user_regset_view *view,
|
||||
unsigned int type);
|
||||
extern unsigned int regset_size(struct thread *target,
|
||||
const struct user_regset *regset);
|
||||
|
||||
/* Definitions for user_sve_header.flags: */
|
||||
#define SVE_PT_REGS_MASK (1 << 0)
|
||||
|
||||
@@ -85,7 +150,7 @@ struct user_sve_header {
|
||||
#define SVE_PT_REGS_SVE SVE_PT_REGS_MASK
|
||||
|
||||
#define SVE_PT_VL_THREAD PR_SVE_SET_VL_THREAD
|
||||
#define SVE_PT_VL_INHERIT PR_SVE_SET_VL_INHERIT
|
||||
#define SVE_PT_VL_INHERIT PR_SVE_VL_INHERIT
|
||||
#define SVE_PT_VL_ONEXEC PR_SVE_SET_VL_ONEXEC
|
||||
|
||||
/*
|
||||
@@ -99,7 +164,9 @@ struct user_sve_header {
|
||||
*/
|
||||
|
||||
/* Offset from the start of struct user_sve_header to the register data */
|
||||
#define SVE_PT_REGS_OFFSET ((sizeof(struct sve_context) + 15) / 16 * 16)
|
||||
#define SVE_PT_REGS_OFFSET \
|
||||
((sizeof(struct sve_context) + (SVE_VQ_BYTES - 1)) \
|
||||
/ SVE_VQ_BYTES * SVE_VQ_BYTES)
|
||||
|
||||
/*
|
||||
* The register data content and layout depends on the value of the
|
||||
@@ -174,8 +241,10 @@ struct user_sve_header {
|
||||
#define SVE_PT_SVE_FFR_OFFSET(vq) \
|
||||
__SVE_SIG_TO_PT(SVE_SIG_FFR_OFFSET(vq))
|
||||
|
||||
#define SVE_PT_SVE_FPSR_OFFSET(vq) \
|
||||
((SVE_PT_SVE_FFR_OFFSET(vq) + SVE_PT_SVE_FFR_SIZE(vq) + 15) / 16 * 16)
|
||||
#define SVE_PT_SVE_FPSR_OFFSET(vq) \
|
||||
((SVE_PT_SVE_FFR_OFFSET(vq) + SVE_PT_SVE_FFR_SIZE(vq) + \
|
||||
(SVE_VQ_BYTES - 1)) \
|
||||
/ SVE_VQ_BYTES * SVE_VQ_BYTES)
|
||||
#define SVE_PT_SVE_FPCR_OFFSET(vq) \
|
||||
(SVE_PT_SVE_FPSR_OFFSET(vq) + SVE_PT_SVE_FPSR_SIZE)
|
||||
|
||||
@@ -184,9 +253,10 @@ struct user_sve_header {
|
||||
* 128-bit boundary.
|
||||
*/
|
||||
|
||||
#define SVE_PT_SVE_SIZE(vq, flags) \
|
||||
((SVE_PT_SVE_FPCR_OFFSET(vq) + SVE_PT_SVE_FPCR_SIZE - \
|
||||
SVE_PT_SVE_OFFSET + 15) / 16 * 16)
|
||||
#define SVE_PT_SVE_SIZE(vq, flags) \
|
||||
((SVE_PT_SVE_FPCR_OFFSET(vq) + SVE_PT_SVE_FPCR_SIZE \
|
||||
- SVE_PT_SVE_OFFSET + (SVE_VQ_BYTES - 1)) \
|
||||
/ SVE_VQ_BYTES * SVE_VQ_BYTES)
|
||||
|
||||
#define SVE_PT_SIZE(vq, flags) \
|
||||
(((flags) & SVE_PT_REGS_MASK) == SVE_PT_REGS_SVE ? \
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
/* signal.h COPYRIGHT FUJITSU LIMITED 2015-2018 */
|
||||
/* signal.h COPYRIGHT FUJITSU LIMITED 2015-2019 */
|
||||
#ifndef __HEADER_ARM64_COMMON_SIGNAL_H
|
||||
#define __HEADER_ARM64_COMMON_SIGNAL_H
|
||||
|
||||
@@ -298,6 +298,7 @@ struct extra_context {
|
||||
struct _aarch64_ctx head;
|
||||
void *data; /* 16-byte aligned pointer to the extra space */
|
||||
uint32_t size; /* size in bytes of the extra space */
|
||||
uint32_t __reserved[3];
|
||||
};
|
||||
|
||||
#define SVE_MAGIC 0x53564501
|
||||
@@ -318,19 +319,25 @@ struct sve_context {
|
||||
* The SVE architecture leaves space for future expansion of the
|
||||
* vector length beyond its initial architectural limit of 2048 bits
|
||||
* (16 quadwords).
|
||||
*
|
||||
* See linux/Documentation/arm64/sve.txt for a description of the VL/VQ
|
||||
* terminology.
|
||||
*/
|
||||
#define SVE_VQ_MIN 1
|
||||
#define SVE_VQ_MAX 0x200
|
||||
#define SVE_VQ_BYTES 16 /* number of bytes per quadword */
|
||||
|
||||
#define SVE_VL_MIN (SVE_VQ_MIN * 0x10)
|
||||
#define SVE_VL_MAX (SVE_VQ_MAX * 0x10)
|
||||
#define SVE_VQ_MIN 1
|
||||
#define SVE_VQ_MAX 512
|
||||
|
||||
#define SVE_VL_MIN (SVE_VQ_MIN * SVE_VQ_BYTES)
|
||||
#define SVE_VL_MAX (SVE_VQ_MAX * SVE_VQ_BYTES)
|
||||
|
||||
#define SVE_NUM_ZREGS 32
|
||||
#define SVE_NUM_PREGS 16
|
||||
|
||||
#define sve_vl_valid(vl) \
|
||||
((vl) % 0x10 == 0 && (vl) >= SVE_VL_MIN && (vl) <= SVE_VL_MAX)
|
||||
#define sve_vq_from_vl(vl) ((vl) / 0x10)
|
||||
((vl) % SVE_VQ_BYTES == 0 && (vl) >= SVE_VL_MIN && (vl) <= SVE_VL_MAX)
|
||||
#define sve_vq_from_vl(vl) ((vl) / SVE_VQ_BYTES)
|
||||
#define sve_vl_from_vq(vq) ((vq) * SVE_VQ_BYTES)
|
||||
|
||||
/*
|
||||
* The total size of meaningful data in the SVE context in bytes,
|
||||
@@ -365,11 +372,13 @@ struct sve_context {
|
||||
* Additional data might be appended in the future.
|
||||
*/
|
||||
|
||||
#define SVE_SIG_ZREG_SIZE(vq) ((uint32_t)(vq) * 16)
|
||||
#define SVE_SIG_PREG_SIZE(vq) ((uint32_t)(vq) * 2)
|
||||
#define SVE_SIG_ZREG_SIZE(vq) ((uint32_t)(vq) * SVE_VQ_BYTES)
|
||||
#define SVE_SIG_PREG_SIZE(vq) ((uint32_t)(vq) * (SVE_VQ_BYTES / 8))
|
||||
#define SVE_SIG_FFR_SIZE(vq) SVE_SIG_PREG_SIZE(vq)
|
||||
|
||||
#define SVE_SIG_REGS_OFFSET ((sizeof(struct sve_context) + 15) / 16 * 16)
|
||||
#define SVE_SIG_REGS_OFFSET \
|
||||
((sizeof(struct sve_context) + (SVE_VQ_BYTES - 1)) \
|
||||
/ SVE_VQ_BYTES * SVE_VQ_BYTES)
|
||||
|
||||
#define SVE_SIG_ZREGS_OFFSET SVE_SIG_REGS_OFFSET
|
||||
#define SVE_SIG_ZREG_OFFSET(vq, n) \
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
/* thread_info.h COPYRIGHT FUJITSU LIMITED 2015-2018 */
|
||||
/* thread_info.h COPYRIGHT FUJITSU LIMITED 2015-2019 */
|
||||
#ifndef __HEADER_ARM64_COMMON_THREAD_INFO_H
|
||||
#define __HEADER_ARM64_COMMON_THREAD_INFO_H
|
||||
|
||||
@@ -46,9 +46,9 @@ struct thread_info {
|
||||
int cpu; /* cpu */
|
||||
struct cpu_context cpu_context; /* kernel_context */
|
||||
void *sve_state; /* SVE registers, if any */
|
||||
uint16_t sve_vl; /* SVE vector length */
|
||||
uint16_t sve_vl_onexec; /* SVE vl after next exec */
|
||||
uint16_t sve_flags; /* SVE related flags */
|
||||
unsigned int sve_vl; /* SVE vector length */
|
||||
unsigned int sve_vl_onexec; /* SVE vl after next exec */
|
||||
unsigned long sve_flags; /* SVE related flags */
|
||||
unsigned long fault_address; /* fault info */
|
||||
unsigned long fault_code; /* ESR_EL1 value */
|
||||
};
|
||||
@@ -56,7 +56,7 @@ struct thread_info {
|
||||
/* Flags for sve_flags (intentionally defined to match the prctl flags) */
|
||||
|
||||
/* Inherit sve_vl and sve_flags across execve(): */
|
||||
#define THREAD_VL_INHERIT PR_SVE_SET_VL_INHERIT
|
||||
#define THREAD_VL_INHERIT PR_SVE_VL_INHERIT
|
||||
|
||||
struct arm64_cpu_local_thread {
|
||||
struct thread_info thread_info;
|
||||
|
||||
Reference in New Issue
Block a user