s390: add support for transactional memory

Allow user-space processes to use transactional execution (TX).
If the TX facility is available user space programs can use
transactions for fine-grained serialization based on the data
objects that are referenced during a transaction. This is
useful for lockless data structures and speculative compiler
optimizations.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/include/asm/elf.h

@@ -101,6 +101,7 @@
 #define HWCAP_S390_HPAGE	128
 #define HWCAP_S390_ETF3EH	256
 #define HWCAP_S390_HIGH_GPRS	512
+#define HWCAP_S390_TE		1024
 
 /*
  * These are used to set parameters in the core dumps.

arch/s390/include/asm/lowcore.h

@@ -329,9 +329,13 @@ struct _lowcore {
 	__u8	pad_0x1338[0x1340-0x1338];	/* 0x1338 */
 	__u32	access_regs_save_area[16];	/* 0x1340 */
 	__u64	cregs_save_area[16];		/* 0x1380 */
+	__u8	pad_0x1400[0x1800-0x1400];	/* 0x1400 */
+
+	/* Transaction abort diagnostic block */
+	__u8	pgm_tdb[256];			/* 0x1800 */
 
 	/* align to the top of the prefix area */
-	__u8	pad_0x1400[0x2000-0x1400];	/* 0x1400 */
+	__u8	pad_0x1900[0x2000-0x1900];	/* 0x1900 */
 } __packed;
 
 #endif /* CONFIG_32BIT */

arch/s390/include/asm/processor.h

@@ -76,14 +76,20 @@ struct thread_struct {
 	unsigned long gmap_addr;	/* address of last gmap fault. */
 	struct per_regs per_user;	/* User specified PER registers */
 	struct per_event per_event;	/* Cause of the last PER trap */
+	unsigned long per_flags;	/* Flags to control debug behavior */
         /* pfault_wait is used to block the process on a pfault event */
 	unsigned long pfault_wait;
 	struct list_head list;
 	/* cpu runtime instrumentation */
 	struct runtime_instr_cb *ri_cb;
 	int ri_signum;
+#ifdef CONFIG_64BIT
+	unsigned char trap_tdb[256];	/* Transaction abort diagnose block */
+#endif
 };
 
+#define PER_FLAG_NO_TE		1UL	/* Flag to disable transactions. */
+
 typedef struct thread_struct thread_struct;
 
 /*

arch/s390/include/asm/ptrace.h

@@ -361,17 +361,19 @@ struct per_struct_kernel {
 	unsigned char access_id;	/* PER trap access identification */
 };
 
-#define PER_EVENT_MASK			0xE9000000UL
+#define PER_EVENT_MASK			0xEB000000UL
 
 #define PER_EVENT_BRANCH		0x80000000UL
 #define PER_EVENT_IFETCH		0x40000000UL
 #define PER_EVENT_STORE			0x20000000UL
 #define PER_EVENT_STORE_REAL		0x08000000UL
+#define PER_EVENT_TRANSACTION_END	0x02000000UL
 #define PER_EVENT_NULLIFICATION		0x01000000UL
 
-#define PER_CONTROL_MASK		0x00a00000UL
+#define PER_CONTROL_MASK		0x00e00000UL
 
 #define PER_CONTROL_BRANCH_ADDRESS	0x00800000UL
+#define PER_CONTROL_SUSPENSION		0x00400000UL
 #define PER_CONTROL_ALTERATION		0x00200000UL
 
 #endif
@@ -485,6 +487,8 @@ typedef struct
 #define PTRACE_GET_LAST_BREAK	      0x5006
 #define PTRACE_PEEK_SYSTEM_CALL       0x5007
 #define PTRACE_POKE_SYSTEM_CALL	      0x5008
+#define PTRACE_ENABLE_TE	      0x5009
+#define PTRACE_DISABLE_TE	      0x5010
 
 /*
  * PT_PROT definition is loosely based on hppa bsd definition in

arch/s390/include/asm/setup.h

@@ -80,6 +80,7 @@ extern unsigned int addressing_mode;
 #define MACHINE_FLAG_LPAR	(1UL << 12)
 #define MACHINE_FLAG_SPP	(1UL << 13)
 #define MACHINE_FLAG_TOPOLOGY	(1UL << 14)
+#define MACHINE_FLAG_TE		(1UL << 15)
 
 #define MACHINE_IS_VM		(S390_lowcore.machine_flags & MACHINE_FLAG_VM)
 #define MACHINE_IS_KVM		(S390_lowcore.machine_flags & MACHINE_FLAG_KVM)
@@ -98,6 +99,7 @@ extern unsigned int addressing_mode;
 #define MACHINE_HAS_PFMF	(0)
 #define MACHINE_HAS_SPP		(0)
 #define MACHINE_HAS_TOPOLOGY	(0)
+#define MACHINE_HAS_TE		      (0)
 #else /* CONFIG_64BIT */
 #define MACHINE_HAS_IEEE	(1)
 #define MACHINE_HAS_CSP		(1)
@@ -109,6 +111,7 @@ extern unsigned int addressing_mode;
 #define MACHINE_HAS_PFMF	(S390_lowcore.machine_flags & MACHINE_FLAG_PFMF)
 #define MACHINE_HAS_SPP		(S390_lowcore.machine_flags & MACHINE_FLAG_SPP)
 #define MACHINE_HAS_TOPOLOGY	(S390_lowcore.machine_flags & MACHINE_FLAG_TOPOLOGY)
+#define MACHINE_HAS_TE		(S390_lowcore.machine_flags & MACHINE_FLAG_TE)
 #endif /* CONFIG_64BIT */
 
 #define ZFCPDUMP_HSA_SIZE	(32UL<<20)

arch/s390/kernel/asm-offsets.c

@@ -157,6 +157,8 @@ int main(void)
 	DEFINE(__LC_LAST_BREAK, offsetof(struct _lowcore, breaking_event_addr));
 	DEFINE(__LC_VDSO_PER_CPU, offsetof(struct _lowcore, vdso_per_cpu_data));
 	DEFINE(__LC_GMAP, offsetof(struct _lowcore, gmap));
+	DEFINE(__LC_PGM_TDB, offsetof(struct _lowcore, pgm_tdb));
+	DEFINE(__THREAD_trap_tdb, offsetof(struct task_struct, thread.trap_tdb));
 	DEFINE(__GMAP_ASCE, offsetof(struct gmap, asce));
 #endif /* CONFIG_32BIT */
 	return 0;

arch/s390/kernel/dis.c

@@ -317,6 +317,9 @@ enum {
 	LONG_INSN_RISBLG,
 	LONG_INSN_RINEXT,
 	LONG_INSN_RIEMIT,
+	LONG_INSN_TABORT,
+	LONG_INSN_TBEGIN,
+	LONG_INSN_TBEGINC,
 };
 
 static char *long_insn_name[] = {
@@ -334,6 +337,9 @@ static char *long_insn_name[] = {
 	[LONG_INSN_RISBLG] = "risblk",
 	[LONG_INSN_RINEXT] = "rinext",
 	[LONG_INSN_RIEMIT] = "riemit",
+	[LONG_INSN_TABORT] = "tabort",
+	[LONG_INSN_TBEGIN] = "tbegin",
+	[LONG_INSN_TBEGINC] = "tbeginc",
 };
 
 static struct insn opcode[] = {
@@ -609,6 +615,9 @@ static struct insn opcode_b2[] = {
 	{ "lpswe", 0xb2, INSTR_S_RD },
 	{ "srnmt", 0xb9, INSTR_S_RD },
 	{ "lfas", 0xbd, INSTR_S_RD },
+	{ "etndg", 0xec, INSTR_RRE_R0 },
+	{ { 0, LONG_INSN_TABORT }, 0xfc, INSTR_S_RD },
+	{ "tend", 0xf8, INSTR_S_RD },
 #endif
 	{ "stidp", 0x02, INSTR_S_RD },
 	{ "sck", 0x04, INSTR_S_RD },
@@ -1165,6 +1174,7 @@ static struct insn opcode_e3[] = {
 	{ "stfh", 0xcb, INSTR_RXY_RRRD },
 	{ "chf", 0xcd, INSTR_RXY_RRRD },
 	{ "clhf", 0xcf, INSTR_RXY_RRRD },
+	{ "ntstg", 0x25, INSTR_RXY_RRRD },
 #endif
 	{ "lrv", 0x1e, INSTR_RXY_RRRD },
 	{ "lrvh", 0x1f, INSTR_RXY_RRRD },
@@ -1188,6 +1198,8 @@ static struct insn opcode_e5[] = {
 	{ "mvhhi", 0x44, INSTR_SIL_RDI },
 	{ "mvhi", 0x4c, INSTR_SIL_RDI },
 	{ "mvghi", 0x48, INSTR_SIL_RDI },
+	{ { 0, LONG_INSN_TBEGIN }, 0x60, INSTR_SIL_RDU },
+	{ { 0, LONG_INSN_TBEGINC }, 0x61, INSTR_SIL_RDU },
 #endif
 	{ "lasp", 0x00, INSTR_SSE_RDRD },
 	{ "tprot", 0x01, INSTR_SSE_RDRD },

arch/s390/kernel/early.c

@@ -370,6 +370,8 @@ static __init void detect_machine_facilities(void)
 		S390_lowcore.machine_flags |= MACHINE_FLAG_MVCOS;
 	if (test_facility(40))
 		S390_lowcore.machine_flags |= MACHINE_FLAG_SPP;
+	if (test_facility(50) && test_facility(73))
+		S390_lowcore.machine_flags |= MACHINE_FLAG_TE;
 #endif
 }
 
@@ -439,7 +441,6 @@ static void __init setup_boot_command_line(void)
 	append_to_cmdline(append_ipl_scpdata);
 }
 
-
 /*
  * Save ipl parameters, clear bss memory, initialize storage keys
  * and create a kernel NSS at startup if the SAVESYS= parm is defined

arch/s390/kernel/entry64.S

@@ -412,6 +412,11 @@ ENTRY(pgm_check_handler)
 1:	UPDATE_VTIME %r14,__LC_SYNC_ENTER_TIMER
 	LAST_BREAK %r14
 	lg	%r15,__LC_KERNEL_STACK
+	lg	%r14,__TI_task(%r12)
+	lghi	%r13,__LC_PGM_TDB
+	tm	__LC_PGM_ILC+2,0x02	# check for transaction abort
+	jz	2f
+	mvc	__THREAD_trap_tdb(256,%r14),0(%r13)
 2:	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	la	%r11,STACK_FRAME_OVERHEAD(%r15)
 	stmg	%r0,%r7,__PT_R0(%r11)
@@ -422,13 +427,12 @@ ENTRY(pgm_check_handler)
 	stg	%r10,__PT_ARGS(%r11)
 	tm	__LC_PGM_ILC+3,0x80	# check for per exception
 	jz	0f
-	lg	%r1,__TI_task(%r12)
 	tmhh	%r8,0x0001		# kernel per event ?
 	jz	pgm_kprobe
 	oi	__TI_flags+7(%r12),_TIF_PER_TRAP
-	mvc	__THREAD_per_address(8,%r1),__LC_PER_ADDRESS
+	mvc	__THREAD_per_address(8,%r14),__LC_PER_ADDRESS
-	mvc	__THREAD_per_cause(2,%r1),__LC_PER_CAUSE
+	mvc	__THREAD_per_cause(2,%r14),__LC_PER_CAUSE
-	mvc	__THREAD_per_paid(1,%r1),__LC_PER_PAID
+	mvc	__THREAD_per_paid(1,%r14),__LC_PER_PAID
 0:	REENABLE_IRQS
 	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
 	larl	%r1,pgm_check_table

arch/s390/kernel/processor.c

@@ -39,9 +39,9 @@ void __cpuinit cpu_init(void)
  */
 static int show_cpuinfo(struct seq_file *m, void *v)
 {
-	static const char *hwcap_str[10] = {
+	static const char *hwcap_str[11] = {
 		"esan3", "zarch", "stfle", "msa", "ldisp", "eimm", "dfp",
-		"edat", "etf3eh", "highgprs"
+		"edat", "etf3eh", "highgprs", "te"
 	};
 	unsigned long n = (unsigned long) v - 1;
 	int i;
@@ -54,7 +54,7 @@ static int show_cpuinfo(struct seq_file *m, void *v)
 			   num_online_cpus(), loops_per_jiffy/(500000/HZ),
 			   (loops_per_jiffy/(5000/HZ))%100);
 		seq_puts(m, "features\t: ");
-		for (i = 0; i < 10; i++)
+		for (i = 0; i < 11; i++)
 			if (hwcap_str[i] && (elf_hwcap & (1UL << i)))
 				seq_printf(m, "%s ", hwcap_str[i]);
 		seq_puts(m, "\n");

arch/s390/kernel/ptrace.c

@@ -42,6 +42,7 @@ enum s390_regset {
 	REGSET_GENERAL,
 	REGSET_FP,
 	REGSET_LAST_BREAK,
+	REGSET_TDB,
 	REGSET_SYSTEM_CALL,
 	REGSET_GENERAL_EXTENDED,
 };
@@ -52,6 +53,21 @@ void update_per_regs(struct task_struct *task)
 	struct thread_struct *thread = &task->thread;
 	struct per_regs old, new;
 
+	/* Take care of the enable/disable of transactional execution. */
+	if (MACHINE_HAS_TE) {
+		unsigned long cr0, cr0_new;
+
+		__ctl_store(cr0, 0, 0);
+		/* set or clear transaction execution bits 8 and 9. */
+		if (task->thread.per_flags & PER_FLAG_NO_TE)
+			cr0_new = cr0 & ~(3UL << 54);
+		else
+			cr0_new = cr0 | (3UL << 54);
+		/* Only load control register 0 if necessary. */
+		if (cr0 != cr0_new)
+			__ctl_load(cr0_new, 0, 0);
+	}
+
 	/* Copy user specified PER registers */
 	new.control = thread->per_user.control;
 	new.start = thread->per_user.start;
@@ -60,6 +76,10 @@ void update_per_regs(struct task_struct *task)
 	/* merge TIF_SINGLE_STEP into user specified PER registers. */
 	if (test_tsk_thread_flag(task, TIF_SINGLE_STEP)) {
 		new.control |= PER_EVENT_IFETCH;
+#ifdef CONFIG_64BIT
+		new.control |= PER_CONTROL_SUSPENSION;
+		new.control |= PER_EVENT_TRANSACTION_END;
+#endif
 		new.start = 0;
 		new.end = PSW_ADDR_INSN;
 	}
@@ -100,6 +120,7 @@ void ptrace_disable(struct task_struct *task)
 	memset(&task->thread.per_event, 0, sizeof(task->thread.per_event));
 	clear_tsk_thread_flag(task, TIF_SINGLE_STEP);
 	clear_tsk_thread_flag(task, TIF_PER_TRAP);
+	task->thread.per_flags = 0;
 }
 
 #ifndef CONFIG_64BIT
@@ -416,6 +437,16 @@ long arch_ptrace(struct task_struct *child, long request,
 		put_user(task_thread_info(child)->last_break,
 			 (unsigned long __user *) data);
 		return 0;
+	case PTRACE_ENABLE_TE:
+		if (!MACHINE_HAS_TE)
+			return -EIO;
+		child->thread.per_flags &= ~PER_FLAG_NO_TE;
+		return 0;
+	case PTRACE_DISABLE_TE:
+		if (!MACHINE_HAS_TE)
+			return -EIO;
+		child->thread.per_flags |= PER_FLAG_NO_TE;
+		return 0;
 	default:
 		/* Removing high order bit from addr (only for 31 bit). */
 		addr &= PSW_ADDR_INSN;
@@ -903,6 +934,28 @@ static int s390_last_break_set(struct task_struct *target,
 	return 0;
 }
 
+static int s390_tdb_get(struct task_struct *target,
+			const struct user_regset *regset,
+			unsigned int pos, unsigned int count,
+			void *kbuf, void __user *ubuf)
+{
+	struct pt_regs *regs = task_pt_regs(target);
+	unsigned char *data;
+
+	if (!(regs->int_code & 0x200))
+		return -ENODATA;
+	data = target->thread.trap_tdb;
+	return user_regset_copyout(&pos, &count, &kbuf, &ubuf, data, 0, 256);
+}
+
+static int s390_tdb_set(struct task_struct *target,
+			const struct user_regset *regset,
+			unsigned int pos, unsigned int count,
+			const void *kbuf, const void __user *ubuf)
+{
+	return 0;
+}
+
 #endif
 
 static int s390_system_call_get(struct task_struct *target,
@@ -951,6 +1004,14 @@ static const struct user_regset s390_regsets[] = {
 		.get = s390_last_break_get,
 		.set = s390_last_break_set,
 	},
+	[REGSET_TDB] = {
+		.core_note_type = NT_S390_TDB,
+		.n = 1,
+		.size = 256,
+		.align = 1,
+		.get = s390_tdb_get,
+		.set = s390_tdb_set,
+	},
 #endif
 	[REGSET_SYSTEM_CALL] = {
 		.core_note_type = NT_S390_SYSTEM_CALL,
@@ -1148,6 +1209,14 @@ static const struct user_regset s390_compat_regsets[] = {
 		.get = s390_compat_last_break_get,
 		.set = s390_compat_last_break_set,
 	},
+	[REGSET_TDB] = {
+		.core_note_type = NT_S390_TDB,
+		.n = 1,
+		.size = 256,
+		.align = 1,
+		.get = s390_tdb_get,
+		.set = s390_tdb_set,
+	},
 	[REGSET_SYSTEM_CALL] = {
 		.core_note_type = NT_S390_SYSTEM_CALL,
 		.n = 1,

arch/s390/kernel/setup.c

@@ -980,6 +980,12 @@ static void __init setup_hwcaps(void)
 	 * HWCAP_S390_HIGH_GPRS is bit 9.
 	 */
 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
+
+	/*
+	 * Transactional execution support HWCAP_S390_TE is bit 10.
+	 */
+	if (test_facility(50) && test_facility(73))
+		elf_hwcap |= HWCAP_S390_TE;
 #endif
 
 	get_cpu_id(&cpu_id);

arch/s390/kernel/traps.c

@@ -57,6 +57,23 @@ static int kstack_depth_to_print = 12;
 static int kstack_depth_to_print = 20;
 #endif /* CONFIG_64BIT */
 
+static inline void __user *get_trap_ip(struct pt_regs *regs)
+{
+#ifdef CONFIG_64BIT
+	unsigned long address;
+
+	if (regs->int_code & 0x200)
+		address = *(unsigned long *)(current->thread.trap_tdb + 24);
+	else
+		address = regs->psw.addr;
+	return (void __user *)
+		((address - (regs->int_code >> 16)) & PSW_ADDR_INSN);
+#else
+	return (void __user *)
+		((regs->psw.addr - (regs->int_code >> 16)) & PSW_ADDR_INSN);
+#endif
+}
+
 /*
  * For show_trace we have tree different stack to consider:
  *   - the panic stack which is used if the kernel stack has overflown
@@ -285,12 +302,6 @@ int is_valid_bugaddr(unsigned long addr)
 	return 1;
 }
 
-static inline void __user *get_psw_address(struct pt_regs *regs)
-{
-	return (void __user *)
-		((regs->psw.addr - (regs->int_code >> 16)) & PSW_ADDR_INSN);
-}
-
 static void __kprobes do_trap(struct pt_regs *regs,
 			      int si_signo, int si_code, char *str)
 {
@@ -304,7 +315,7 @@ static void __kprobes do_trap(struct pt_regs *regs,
 		info.si_signo = si_signo;
 		info.si_errno = 0;
 		info.si_code = si_code;
-		info.si_addr = get_psw_address(regs);
+		info.si_addr = get_trap_ip(regs);
 		force_sig_info(si_signo, &info, current);
 		report_user_fault(regs, si_signo);
         } else {
@@ -381,6 +392,11 @@ DO_ERROR_INFO(special_op_exception, SIGILL, ILL_ILLOPN,
 DO_ERROR_INFO(translation_exception, SIGILL, ILL_ILLOPN,
 	      "translation exception")
 
+#ifdef CONFIG_64BIT
+DO_ERROR_INFO(transaction_exception, SIGILL, ILL_ILLOPN,
+	      "transaction constraint exception")
+#endif
+
 static inline void do_fp_trap(struct pt_regs *regs, int fpc)
 {
 	int si_code = 0;
@@ -408,7 +424,7 @@ static void __kprobes illegal_op(struct pt_regs *regs)
 	__u16 __user *location;
 	int signal = 0;
 
-	location = get_psw_address(regs);
+	location = get_trap_ip(regs);
 
 	if (user_mode(regs)) {
 		if (get_user(*((__u16 *) opcode), (__u16 __user *) location))
@@ -476,7 +492,7 @@ void specification_exception(struct pt_regs *regs)
 	__u16 __user *location = NULL;
 	int signal = 0;
 
-	location = (__u16 __user *) get_psw_address(regs);
+	location = (__u16 __user *) get_trap_ip(regs);
 
 	if (user_mode(regs)) {
 		get_user(*((__u16 *) opcode), location);
@@ -525,7 +541,7 @@ static void data_exception(struct pt_regs *regs)
 	__u16 __user *location;
 	int signal = 0;
 
-	location = get_psw_address(regs);
+	location = get_trap_ip(regs);
 
 	if (MACHINE_HAS_IEEE)
 		asm volatile("stfpc %0" : "=m" (current->thread.fp_regs.fpc));
@@ -641,6 +657,7 @@ void __init trap_init(void)
         pgm_check_table[0x12] = &translation_exception;
         pgm_check_table[0x13] = &special_op_exception;
 #ifdef CONFIG_64BIT
+	pgm_check_table[0x18] = &transaction_exception;
 	pgm_check_table[0x38] = &do_asce_exception;
 	pgm_check_table[0x39] = &do_dat_exception;
 	pgm_check_table[0x3A] = &do_dat_exception;

include/linux/elf.h

@@ -387,6 +387,7 @@ typedef struct elf64_shdr {
 #define NT_S390_PREFIX	0x305		/* s390 prefix register */
 #define NT_S390_LAST_BREAK	0x306	/* s390 breaking event address */
 #define NT_S390_SYSTEM_CALL	0x307	/* s390 system call restart data */
+#define NT_S390_TDB	0x308		/* s390 transaction diagnostic block */
 #define NT_ARM_VFP	0x400		/* ARM VFP/NEON registers */