Aggregated FPSIMD/SVE/SME P0 Changes

Documentation/arch/arm64/cpu-feature-registers.rst

@@ -268,6 +268,8 @@ infrastructure:
      +------------------------------+---------+---------+
      | SHA3                         | [35-32] |    y    |
      +------------------------------+---------+---------+
+     | B16B16                       | [27-24] |    y    |
+     +------------------------------+---------+---------+
      | BF16                         | [23-20] |    y    |
      +------------------------------+---------+---------+
      | BitPerm                      | [19-16] |    y    |

Documentation/arch/arm64/elf_hwcaps.rst

@@ -170,6 +170,56 @@ HWCAP_PACG
     ID_AA64ISAR1_EL1.GPI == 0b0001, as described by
     Documentation/arch/arm64/pointer-authentication.rst.
 
+HWCAP_CMPBR
+    Functionality implied by ID_AA64ISAR2_EL1.CSSC == 0b0010.
+
+HWCAP_FPRCVT
+    Functionality implied by ID_AA64ISAR3_EL1.FPRCVT == 0b0001.
+
+HWCAP_F8MM8
+    Functionality implied by ID_AA64FPFR0_EL1.F8MM8 == 0b0001.
+
+HWCAP_F8MM4
+    Functionality implied by ID_AA64FPFR0_EL1.F8MM4 == 0b0001.
+
+HWCAP_SVE_F16MM
+    Functionality implied by ID_AA64PFR0_EL1.SVE == 0b0001 and
+    ID_AA64ZFR0_EL1.F16MM == 0b0001.
+
+HWCAP_SVE_ELTPERM
+    Functionality implied by ID_AA64PFR0_EL1.SVE == 0b0001 and
+    ID_AA64ZFR0_EL1.ELTPERM == 0b0001.
+
+HWCAP_SVE_AES2
+    Functionality implied by ID_AA64PFR0_EL1.SVE == 0b0001 and
+    ID_AA64ZFR0_EL1.AES == 0b0011.
+
+HWCAP_SVE_BFSCALE
+    Functionality implied by ID_AA64PFR0_EL1.SVE == 0b0001 and
+    ID_AA64ZFR0_EL1.B16B16 == 0b0010.
+
+HWCAP_SVE2P2
+    Functionality implied by ID_AA64PFR0_EL1.SVE == 0b0001 and
+    ID_AA64ZFR0_EL1.SVEver == 0b0011.
+
+HWCAP_SME2P2
+    Functionality implied by ID_AA64SMFR0_EL1.SMEver == 0b0011.
+
+HWCAP_SME_SBITPERM
+    Functionality implied by ID_AA64SMFR0_EL1.SBitPerm == 0b1.
+
+HWCAP_SME_AES
+    Functionality implied by ID_AA64SMFR0_EL1.AES == 0b1.
+
+HWCAP_SME_SFEXPA
+    Functionality implied by ID_AA64SMFR0_EL1.SFEXPA == 0b1.
+
+HWCAP_SME_STMOP
+    Functionality implied by ID_AA64SMFR0_EL1.STMOP == 0b1.
+
+HWCAP_SME_SMOP4
+    Functionality implied by ID_AA64SMFR0_EL1.SMOP4 == 0b1.
+
 HWCAP2_DCPODP
     Functionality implied by ID_AA64ISAR1_EL1.DPB == 0b0010.
 
@@ -320,12 +370,64 @@ HWCAP2_MOPS
 HWCAP2_HBC
     Functionality implied by ID_AA64ISAR2_EL1.BC == 0b0001.
 
+HWCAP2_SVE_B16B16
+    Functionality implied by ID_AA64ZFR0_EL1.B16B16 == 0b0001.
+
 HWCAP3_LS64
     Functionality implied by ID_AA64ISAR1_EL1.LS64 == 0b0001.
 
 HWCAP3_LS64_V
     Functionality implied by ID_AA64ISAR1_EL1.LS64 == 0b0010.
 
+HWCAP2_FPMR
+    Functionality implied by ID_AA64PFR2_EL1.FMR == 0b0001.
+
+HWCAP2_LUT
+    Functionality implied by ID_AA64ISAR2_EL1.LUT == 0b0001.
+
+HWCAP2_FAMINMAX
+    Functionality implied by ID_AA64ISAR3_EL1.FAMINMAX == 0b0001.
+
+HWCAP2_F8CVT
+    Functionality implied by ID_AA64FPFR0_EL1.F8CVT == 0b1.
+
+HWCAP2_F8FMA
+    Functionality implied by ID_AA64FPFR0_EL1.F8FMA == 0b1.
+
+HWCAP2_F8DP4
+    Functionality implied by ID_AA64FPFR0_EL1.F8DP4 == 0b1.
+
+HWCAP2_F8DP2
+    Functionality implied by ID_AA64FPFR0_EL1.F8DP2 == 0b1.
+
+HWCAP2_F8E4M3
+    Functionality implied by ID_AA64FPFR0_EL1.F8E4M3 == 0b1.
+
+HWCAP2_F8E5M2
+    Functionality implied by ID_AA64FPFR0_EL1.F8E5M2 == 0b1.
+
+HWCAP2_SME_LUTV2
+    Functionality implied by ID_AA64SMFR0_EL1.LUTv2 == 0b1.
+
+HWCAP2_SME_F8F16
+    Functionality implied by ID_AA64SMFR0_EL1.F8F16 == 0b1.
+
+HWCAP2_SME_F8F32
+    Functionality implied by ID_AA64SMFR0_EL1.F8F32 == 0b1.
+
+HWCAP2_SME_SF8FMA
+    Functionality implied by ID_AA64SMFR0_EL1.SF8FMA == 0b1.
+
+HWCAP2_SME_SF8DP4
+    Functionality implied by ID_AA64SMFR0_EL1.SF8DP4 == 0b1.
+
+HWCAP2_SME_SF8DP2
+    Functionality implied by ID_AA64SMFR0_EL1.SF8DP2 == 0b1.
+
+HWCAP2_SME_SF8DP4
+    Functionality implied by ID_AA64SMFR0_EL1.SF8DP4 == 0b1.
+
+
 4. Unused AT_HWCAP bits
 -----------------------
 

Documentation/arch/arm64/sme.rst

@@ -69,8 +69,8 @@ model features for SME is included in Appendix A.
   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.
+* On thread creation PSTATE.ZA and TPIDR2_EL0 are preserved unless CLONE_VM
+  is specified, in which case PSTATE.ZA is set to 0 and TPIDR2_EL0 is set to 0.
 
 2.  Vector lengths
 ------------------
@@ -115,7 +115,7 @@ be zeroed.
 5.  Signal handling
 -------------------
 
-* Signal handlers are invoked with streaming mode and ZA disabled.
+* Signal handlers are invoked with PSTATE.SM=0, PSTATE.ZA=0, and TPIDR2_EL0=0.
 
 * A new signal frame record TPIDR2_MAGIC is added formatted as a struct
   tpidr2_context to allow access to TPIDR2_EL0 from signal handlers.
@@ -241,7 +241,7 @@ prctl(PR_SME_SET_VL, unsigned long arg)
       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.
+    * Changing the vector length causes PSTATE.ZA 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.

arch/arm64/Kconfig

@@ -2244,7 +2244,6 @@ config ARM64_SME
 	bool "ARM Scalable Matrix Extension support"
 	default y
 	depends on ARM64_SVE
-	depends on BROKEN
 	help
 	  The Scalable Matrix Extension (SME) is an extension to the AArch64
 	  execution state which utilises a substantial subset of the SVE

arch/arm64/include/asm/assembler.h

@@ -766,32 +766,25 @@ alternative_endif
 .endm
 
 	/*
-	 * Check whether preempt/bh-disabled asm code should yield as soon as
-	 * it is able. This is the case if we are currently running in task
-	 * context, and either a softirq is pending, or the TIF_NEED_RESCHED
-	 * flag is set and re-enabling preemption a single time would result in
-	 * a preempt count of zero. (Note that the TIF_NEED_RESCHED flag is
-	 * stored negated in the top word of the thread_info::preempt_count
+	 * Check whether asm code should yield as soon as it is able. This is
+	 * the case if we are currently running in task context, and the
+	 * TIF_NEED_RESCHED flag is set. (Note that the TIF_NEED_RESCHED flag
+	 * is stored negated in the top word of the thread_info::preempt_count
 	 * field)
 	 */
-	.macro		cond_yield, lbl:req, tmp:req, tmp2:req
+	.macro		cond_yield, lbl:req, tmp:req, tmp2
+#ifdef CONFIG_PREEMPT_VOLUNTARY
 	get_current_task \tmp
 	ldr		\tmp, [\tmp, #TSK_TI_PREEMPT]
 	/*
 	 * If we are serving a softirq, there is no point in yielding: the
 	 * softirq will not be preempted no matter what we do, so we should
-	 * run to completion as quickly as we can.
+	 * run to completion as quickly as we can. The preempt_count field will
+	 * have BIT(SOFTIRQ_SHIFT) set in this case, so the zero check will
+	 * catch this case too.
 	 */
-	tbnz		\tmp, #SOFTIRQ_SHIFT, .Lnoyield_\@
-#ifdef CONFIG_PREEMPTION
-	sub		\tmp, \tmp, #PREEMPT_DISABLE_OFFSET
 	cbz		\tmp, \lbl
 #endif
-	adr_l		\tmp, irq_stat + IRQ_CPUSTAT_SOFTIRQ_PENDING
-	get_this_cpu_offset	\tmp2
-	ldr		w\tmp, [\tmp, \tmp2]
-	cbnz		w\tmp, \lbl	// yield on pending softirq in task context
-.Lnoyield_\@:
 	.endm
 
 /*

arch/arm64/include/asm/cpu.h

@@ -52,14 +52,17 @@ struct cpuinfo_arm64 {
 	u64		reg_id_aa64isar0;
 	u64		reg_id_aa64isar1;
 	u64		reg_id_aa64isar2;
+	u64		reg_id_aa64isar3;
 	u64		reg_id_aa64mmfr0;
 	u64		reg_id_aa64mmfr1;
 	u64		reg_id_aa64mmfr2;
 	u64		reg_id_aa64mmfr3;
 	u64		reg_id_aa64pfr0;
 	u64		reg_id_aa64pfr1;
+	u64		reg_id_aa64pfr2;
 	u64		reg_id_aa64zfr0;
 	u64		reg_id_aa64smfr0;
+	u64		reg_id_aa64fpfr0;
 
 	struct cpuinfo_32bit	aarch32;
 

arch/arm64/include/asm/cpufeature.h

@@ -766,6 +766,11 @@ static __always_inline bool system_supports_tpidr2(void)
 	return system_supports_sme();
 }
 
+static __always_inline bool system_supports_fpmr(void)
+{
+	return alternative_has_cap_unlikely(ARM64_HAS_FPMR);
+}
+
 static __always_inline bool system_supports_cnp(void)
 {
 	return alternative_has_cap_unlikely(ARM64_HAS_CNP);

arch/arm64/include/asm/fpsimd.h

@@ -79,7 +79,6 @@ extern void fpsimd_load_state(struct user_fpsimd_state *state);
 extern void fpsimd_thread_switch(struct task_struct *next);
 extern void fpsimd_flush_thread(void);
 
-extern void fpsimd_signal_preserve_current_state(void);
 extern void fpsimd_preserve_current_state(void);
 extern void fpsimd_restore_current_state(void);
 extern void fpsimd_update_current_state(struct user_fpsimd_state const *state);
@@ -90,6 +89,7 @@ struct cpu_fp_state {
 	void *sve_state;
 	void *sme_state;
 	u64 *svcr;
+	u64 *fpmr;
 	unsigned int sve_vl;
 	unsigned int sme_vl;
 	enum fp_type *fp_type;
@@ -101,6 +101,7 @@ DECLARE_PER_CPU(struct cpu_fp_state, fpsimd_last_state);
 extern void fpsimd_bind_state_to_cpu(struct cpu_fp_state *fp_state);
 
 extern void fpsimd_flush_task_state(struct task_struct *target);
+extern void fpsimd_save_and_flush_current_state(void);
 extern void fpsimd_save_and_flush_cpu_state(void);
 
 static inline bool thread_sm_enabled(struct thread_struct *thread)
@@ -113,6 +114,8 @@ static inline bool thread_za_enabled(struct thread_struct *thread)
 	return system_supports_sme() && (thread->svcr & SVCR_ZA_MASK);
 }
 
+extern void task_smstop_sm(struct task_struct *task);
+
 /* Maximum VL that SVE/SME VL-agnostic software can transparently support */
 #define VL_ARCH_MAX 0x100
 
@@ -157,6 +160,7 @@ extern void cpu_enable_sve(const struct arm64_cpu_capabilities *__unused);
 extern void cpu_enable_sme(const struct arm64_cpu_capabilities *__unused);
 extern void cpu_enable_sme2(const struct arm64_cpu_capabilities *__unused);
 extern void cpu_enable_fa64(const struct arm64_cpu_capabilities *__unused);
+extern void cpu_enable_fpmr(const struct arm64_cpu_capabilities *__unused);
 
 extern u64 read_zcr_features(void);
 extern u64 read_smcr_features(void);
@@ -202,10 +206,9 @@ struct vl_info {
 
 extern void sve_alloc(struct task_struct *task, bool flush);
 extern void fpsimd_release_task(struct task_struct *task);
-extern void fpsimd_sync_to_sve(struct task_struct *task);
 extern void fpsimd_force_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 void fpsimd_sync_from_effective_state(struct task_struct *task);
+extern void fpsimd_sync_to_effective_state_zeropad(struct task_struct *task);
 
 extern int vec_set_vector_length(struct task_struct *task, enum vec_type type,
 				 unsigned long vl, unsigned long flags);
@@ -299,14 +302,29 @@ static inline bool sve_vq_available(unsigned int vq)
 	return vq_available(ARM64_VEC_SVE, vq);
 }
 
-size_t sve_state_size(struct task_struct const *task);
+static inline size_t __sve_state_size(unsigned int sve_vl, unsigned int sme_vl)
+{
+	unsigned int vl = max(sve_vl, sme_vl);
+	return SVE_SIG_REGS_SIZE(sve_vq_from_vl(vl));
+}
+
+/*
+ * Return how many bytes of memory are required to store the full SVE
+ * state for task, given task's currently configured vector length.
+ */
+static inline size_t sve_state_size(struct task_struct const *task)
+{
+	unsigned int sve_vl = task_get_sve_vl(task);
+	unsigned int sme_vl = task_get_sme_vl(task);
+	return __sve_state_size(sve_vl, sme_vl);
+}
 
 #else /* ! CONFIG_ARM64_SVE */
 
 static inline void sve_alloc(struct task_struct *task, bool flush) { }
 static inline void fpsimd_release_task(struct task_struct *task) { }
-static inline void sve_sync_to_fpsimd(struct task_struct *task) { }
-static inline void sve_sync_from_fpsimd_zeropad(struct task_struct *task) { }
+static inline void fpsimd_sync_from_effective_state(struct task_struct *task) { }
+static inline void fpsimd_sync_to_effective_state_zeropad(struct task_struct *task) { }
 
 static inline int sve_max_virtualisable_vl(void)
 {
@@ -340,6 +358,11 @@ static inline void vec_update_vq_map(enum vec_type t) { }
 static inline int vec_verify_vq_map(enum vec_type t) { return 0; }
 static inline void sve_setup(void) { }
 
+static inline size_t __sve_state_size(unsigned int sve_vl, unsigned int sme_vl)
+{
+	return 0;
+}
+
 static inline size_t sve_state_size(struct task_struct const *task)
 {
 	return 0;
@@ -392,22 +415,24 @@ extern int sme_set_current_vl(unsigned long arg);
 extern int sme_get_current_vl(void);
 extern void sme_suspend_exit(void);
 
+static inline size_t __sme_state_size(unsigned int sme_vl)
+{
+	size_t size = ZA_SIG_REGS_SIZE(sve_vq_from_vl(sme_vl));
+
+	if (system_supports_sme2())
+		size += ZT_SIG_REG_SIZE;
+
+	return size;
+}
+
 /*
  * Return how many bytes of memory are required to store the full SME
  * specific state for task, given task's currently configured vector
  * length.
  */
 static inline size_t sme_state_size(struct task_struct const *task)
 {
-	unsigned int vl = task_get_sme_vl(task);
-	size_t size;
-
-	size = ZA_SIG_REGS_SIZE(sve_vq_from_vl(vl));
-
-	if (system_supports_sme2())
-		size += ZT_SIG_REG_SIZE;
-
-	return size;
+	return __sme_state_size(task_get_sme_vl(task));
 }
 
 #else
@@ -428,6 +453,11 @@ static inline int sme_set_current_vl(unsigned long arg) { return -EINVAL; }
 static inline int sme_get_current_vl(void) { return -EINVAL; }
 static inline void sme_suspend_exit(void) { }
 
+static inline size_t __sme_state_size(unsigned int sme_vl)
+{
+	return 0;
+}
+
 static inline size_t sme_state_size(struct task_struct const *task)
 {
 	return 0;