s390/time: add support for the TOD clock epoch extension

The TOD epoch extension adds 8 epoch bits to the TOD clock to provide
a continuous clock after 2042/09/17. The store-clock-extended (STCKE)
instruction will store the epoch index in the first byte of the
16 bytes stored by the instruction. The read_boot_clock64 and the
read_presistent_clock64 functions need to take the additional bits
into account to give the correct result after 2042/09/17.

The clock-comparator register will stay 64 bit wide. The comparison
of the clock-comparator with the TOD clock is limited to bytes
1 to 8 of the extended TOD format. To deal with the overflow problem
due to an epoch change the clock-comparator sign control in CR0 can
be used to switch the comparison of the 64-bit TOD clock with the
clock-comparator to a signed comparison.

The decision between the signed vs. unsigned clock-comparator
comparisons is done at boot time. Only if the TOD clock is in the
second half of a 142 year epoch the signed comparison is used.
This solves the epoch overflow issue as long as the machine is
booted at least once in an epoch.

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

arch/s390/include/asm/lowcore.h

@@ -95,46 +95,46 @@ struct lowcore {
 	__u64	int_clock;			/* 0x0310 */
 	__u64	mcck_clock;			/* 0x0318 */
 	__u64	clock_comparator;		/* 0x0320 */
+	__u64	boot_clock[2];			/* 0x0328 */
 
 	/* Current process. */
-	__u64	current_task;			/* 0x0328 */
-	__u8	pad_0x318[0x320-0x318];		/* 0x0330 */
-	__u64	kernel_stack;			/* 0x0338 */
+	__u64	current_task;			/* 0x0338 */
+	__u64	kernel_stack;			/* 0x0340 */
 
 	/* Interrupt, panic and restart stack. */
-	__u64	async_stack;			/* 0x0340 */
-	__u64	panic_stack;			/* 0x0348 */
-	__u64	restart_stack;			/* 0x0350 */
+	__u64	async_stack;			/* 0x0348 */
+	__u64	panic_stack;			/* 0x0350 */
+	__u64	restart_stack;			/* 0x0358 */
 
 	/* Restart function and parameter. */
-	__u64	restart_fn;			/* 0x0358 */
-	__u64	restart_data;			/* 0x0360 */
-	__u64	restart_source;			/* 0x0368 */
+	__u64	restart_fn;			/* 0x0360 */
+	__u64	restart_data;			/* 0x0368 */
+	__u64	restart_source;			/* 0x0370 */
 
 	/* Address space pointer. */
-	__u64	kernel_asce;			/* 0x0370 */
-	__u64	user_asce;			/* 0x0378 */
+	__u64	kernel_asce;			/* 0x0378 */
+	__u64	user_asce;			/* 0x0380 */
 
 	/*
 	 * The lpp and current_pid fields form a
 	 * 64-bit value that is set as program
 	 * parameter with the LPP instruction.
 	 */
-	__u32	lpp;				/* 0x0380 */
-	__u32	current_pid;			/* 0x0384 */
+	__u32	lpp;				/* 0x0388 */
+	__u32	current_pid;			/* 0x038c */
 
 	/* SMP info area */
-	__u32	cpu_nr;				/* 0x0388 */
-	__u32	softirq_pending;		/* 0x038c */
-	__u64	percpu_offset;			/* 0x0390 */
-	__u64	vdso_per_cpu_data;		/* 0x0398 */
-	__u64	machine_flags;			/* 0x03a0 */
-	__u32	preempt_count;			/* 0x03a8 */
-	__u8	pad_0x03ac[0x03b0-0x03ac];	/* 0x03ac */
-	__u64	gmap;				/* 0x03b0 */
-	__u32	spinlock_lockval;		/* 0x03b8 */
-	__u32	fpu_flags;			/* 0x03bc */
-	__u8	pad_0x03c0[0x0400-0x03c0];	/* 0x03c0 */
+	__u32	cpu_nr;				/* 0x0390 */
+	__u32	softirq_pending;		/* 0x0394 */
+	__u64	percpu_offset;			/* 0x0398 */
+	__u64	vdso_per_cpu_data;		/* 0x03a0 */
+	__u64	machine_flags;			/* 0x03a8 */
+	__u32	preempt_count;			/* 0x03b0 */
+	__u8	pad_0x03b4[0x03b8-0x03b4];	/* 0x03b4 */
+	__u64	gmap;				/* 0x03b8 */
+	__u32	spinlock_lockval;		/* 0x03c0 */
+	__u32	fpu_flags;			/* 0x03c4 */
+	__u8	pad_0x03c8[0x0400-0x03c8];	/* 0x03c8 */
 
 	/* Per cpu primary space access list */
 	__u32	paste[16];			/* 0x0400 */

arch/s390/include/asm/setup.h

@@ -32,6 +32,7 @@
 #define MACHINE_FLAG_TLB_GUEST	_BITUL(14)
 #define MACHINE_FLAG_NX		_BITUL(15)
 #define MACHINE_FLAG_GS		_BITUL(16)
+#define MACHINE_FLAG_SCC	_BITUL(17)
 
 #define LPP_MAGIC		_BITUL(31)
 #define LPP_PFAULT_PID_MASK	_AC(0xffffffff, UL)
@@ -72,6 +73,7 @@ extern void detect_memory_memblock(void);
 #define MACHINE_HAS_TLB_GUEST	(S390_lowcore.machine_flags & MACHINE_FLAG_TLB_GUEST)
 #define MACHINE_HAS_NX		(S390_lowcore.machine_flags & MACHINE_FLAG_NX)
 #define MACHINE_HAS_GS		(S390_lowcore.machine_flags & MACHINE_FLAG_GS)
+#define MACHINE_HAS_SCC		(S390_lowcore.machine_flags & MACHINE_FLAG_SCC)
 
 /*
  * Console mode. Override with conmode=

arch/s390/include/asm/timex.h

@@ -15,6 +15,8 @@
 /* The value of the TOD clock for 1.1.1970. */
 #define TOD_UNIX_EPOCH 0x7d91048bca000000ULL
 
+extern u64 clock_comparator_max;
+
 /* Inline functions for clock register access. */
 static inline int set_tod_clock(__u64 time)
 {
@@ -126,7 +128,7 @@ static inline unsigned long long local_tick_disable(void)
 	unsigned long long old;
 
 	old = S390_lowcore.clock_comparator;
-	S390_lowcore.clock_comparator = -1ULL;
+	S390_lowcore.clock_comparator = clock_comparator_max;
 	set_clock_comparator(S390_lowcore.clock_comparator);
 	return old;
 }
@@ -178,20 +180,20 @@ int get_phys_clock(unsigned long long *clock);
 void init_cpu_timer(void);
 unsigned long long monotonic_clock(void);
 
-extern u64 sched_clock_base_cc;
+extern unsigned char tod_clock_base[16] __aligned(8);
 
 /**
  * get_clock_monotonic - returns current time in clock rate units
  *
  * The caller must ensure that preemption is disabled.
- * The clock and sched_clock_base get changed via stop_machine.
+ * The clock and tod_clock_base get changed via stop_machine.
  * Therefore preemption must be disabled when calling this
  * function, otherwise the returned value is not guaranteed to
  * be monotonic.
  */
 static inline unsigned long long get_tod_clock_monotonic(void)
 {
-	return get_tod_clock() - sched_clock_base_cc;
+	return get_tod_clock() - *(unsigned long long *) &tod_clock_base[1];
 }
 
 /**
@@ -218,4 +220,32 @@ static inline unsigned long long tod_to_ns(unsigned long long todval)
 	return ((todval >> 9) * 125) + (((todval & 0x1ff) * 125) >> 9);
 }
 
+/**
+ * tod_after - compare two 64 bit TOD values
+ * @a: first 64 bit TOD timestamp
+ * @b: second 64 bit TOD timestamp
+ *
+ * Returns: true if a is later than b
+ */
+static inline int tod_after(unsigned long long a, unsigned long long b)
+{
+	if (MACHINE_HAS_SCC)
+		return (long long) a > (long long) b;
+	return a > b;
+}
+
+/**
+ * tod_after_eq - compare two 64 bit TOD values
+ * @a: first 64 bit TOD timestamp
+ * @b: second 64 bit TOD timestamp
+ *
+ * Returns: true if a is later than b
+ */
+static inline int tod_after_eq(unsigned long long a, unsigned long long b)
+{
+	if (MACHINE_HAS_SCC)
+		return (long long) a >= (long long) b;
+	return a >= b;
+}
+
 #endif

arch/s390/kernel/asm-offsets.c

@@ -158,6 +158,7 @@ int main(void)
 	OFFSET(__LC_LAST_UPDATE_CLOCK, lowcore, last_update_clock);
 	OFFSET(__LC_INT_CLOCK, lowcore, int_clock);
 	OFFSET(__LC_MCCK_CLOCK, lowcore, mcck_clock);
+	OFFSET(__LC_BOOT_CLOCK, lowcore, boot_clock);
 	OFFSET(__LC_CURRENT, lowcore, current_task);
 	OFFSET(__LC_KERNEL_STACK, lowcore, kernel_stack);
 	OFFSET(__LC_ASYNC_STACK, lowcore, async_stack);

arch/s390/kernel/debug.c

@@ -866,7 +866,8 @@ static inline void
 debug_finish_entry(debug_info_t * id, debug_entry_t* active, int level,
 			int exception)
 {
-	active->id.stck = get_tod_clock_fast() - sched_clock_base_cc;
+	active->id.stck = get_tod_clock_fast() -
+		*(unsigned long long *) &tod_clock_base[1];
 	active->id.fields.cpuid = smp_processor_id();
 	active->caller = __builtin_return_address(0);
 	active->id.fields.exception = exception;
@@ -1455,15 +1456,15 @@ int
 debug_dflt_header_fn(debug_info_t * id, struct debug_view *view,
 			 int area, debug_entry_t * entry, char *out_buf)
 {
-	unsigned long sec, usec;
+	unsigned long base, sec, usec;
 	char *except_str;
 	unsigned long caller;
 	int rc = 0;
 	unsigned int level;
 
 	level = entry->id.fields.level;
-	sec = (entry->id.stck >> 12) + (sched_clock_base_cc >> 12);
-	sec = sec - (TOD_UNIX_EPOCH >> 12);
+	base = (*(unsigned long *) &tod_clock_base[0]) >> 4;
+	sec = (entry->id.stck >> 12) + base - (TOD_UNIX_EPOCH >> 12);
 	usec = do_div(sec, USEC_PER_SEC);
 
 	if (entry->id.fields.exception)

arch/s390/kernel/early.c

@@ -53,8 +53,9 @@ static void __init reset_tod_clock(void)
 	if (set_tod_clock(TOD_UNIX_EPOCH) != 0 || store_tod_clock(&time) != 0)
 		disabled_wait(0);
 
-	sched_clock_base_cc = TOD_UNIX_EPOCH;
-	S390_lowcore.last_update_clock = sched_clock_base_cc;
+	memset(tod_clock_base, 0, 16);
+	*(__u64 *) &tod_clock_base[1] = TOD_UNIX_EPOCH;
+	S390_lowcore.last_update_clock = TOD_UNIX_EPOCH;
 }
 
 #ifdef CONFIG_SHARED_KERNEL
@@ -165,8 +166,8 @@ static noinline __init void create_kernel_nss(void)
 	}
 
 	/* re-initialize cputime accounting. */
-	sched_clock_base_cc = get_tod_clock();
-	S390_lowcore.last_update_clock = sched_clock_base_cc;
+	get_tod_clock_ext(tod_clock_base);
+	S390_lowcore.last_update_clock = *(__u64 *) &tod_clock_base[1];
 	S390_lowcore.last_update_timer = 0x7fffffffffffffffULL;
 	S390_lowcore.user_timer = 0;
 	S390_lowcore.system_timer = 0;
@@ -387,6 +388,12 @@ static __init void detect_machine_facilities(void)
 	}
 	if (test_facility(133))
 		S390_lowcore.machine_flags |= MACHINE_FLAG_GS;
+	if (test_facility(139) && (tod_clock_base[1] & 0x80)) {
+		/* Enabled signed clock comparator comparisons */
+		S390_lowcore.machine_flags |= MACHINE_FLAG_SCC;
+		clock_comparator_max = -1ULL >> 1;
+		__ctl_set_bit(0, 53);
+	}
 }
 
 static inline void save_vector_registers(void)

arch/s390/kernel/head.S

@@ -302,7 +302,8 @@ ENTRY(startup_kdump)
 	xc	0xe00(256),0xe00
 	xc	0xf00(256),0xf00
 	lctlg	%c0,%c15,0x200(%r0)	# initialize control registers
-	stck	__LC_LAST_UPDATE_CLOCK
+	stcke	__LC_BOOT_CLOCK
+	mvc	__LC_LAST_UPDATE_CLOCK(8),__LC_BOOT_CLOCK+1
 	spt	6f-.LPG0(%r13)
 	mvc	__LC_LAST_UPDATE_TIMER(8),6f-.LPG0(%r13)
 	l	%r15,.Lstack-.LPG0(%r13)

arch/s390/kernel/head64.S

@@ -21,8 +21,8 @@ ENTRY(startup_continue)
 	xc	__LC_LPP+1(7,0),__LC_LPP+1	# clear lpp and current_pid
 	mvi	__LC_LPP,0x80			#   and set LPP_MAGIC
 	.insn	s,0xb2800000,__LC_LPP		# load program parameter
-0:	larl	%r1,sched_clock_base_cc
-	mvc	0(8,%r1),__LC_LAST_UPDATE_CLOCK
+0:	larl	%r1,tod_clock_base
+	mvc	0(16,%r1),__LC_BOOT_CLOCK
 	larl	%r13,.LPG1		# get base
 	lctlg	%c0,%c15,.Lctl-.LPG1(%r13)	# load control registers
 	lg	%r12,.Lparmaddr-.LPG1(%r13)	# pointer to parameter area

arch/s390/kernel/irq.c

@@ -105,7 +105,8 @@ void do_IRQ(struct pt_regs *regs, int irq)
 
 	old_regs = set_irq_regs(regs);
 	irq_enter();
-	if (S390_lowcore.int_clock >= S390_lowcore.clock_comparator)
+	if (tod_after_eq(S390_lowcore.int_clock,
+			 S390_lowcore.clock_comparator))
 		/* Serve timer interrupts first. */
 		clock_comparator_work();
 	generic_handle_irq(irq);

arch/s390/kernel/setup.c

@@ -323,7 +323,7 @@ static void __init setup_lowcore(void)
 	lc->io_new_psw.mask = PSW_KERNEL_BITS |
 		PSW_MASK_DAT | PSW_MASK_MCHECK;
 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
-	lc->clock_comparator = -1ULL;
+	lc->clock_comparator = clock_comparator_max;
 	lc->kernel_stack = ((unsigned long) &init_thread_union)
 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 	lc->async_stack = (unsigned long)

arch/s390/kernel/time.c

@@ -51,8 +51,15 @@
 #include <asm/cio.h>
 #include "entry.h"
 
-u64 sched_clock_base_cc = -1;	/* Force to data section. */
-EXPORT_SYMBOL_GPL(sched_clock_base_cc);
+unsigned char tod_clock_base[16] __aligned(8) = {
+	/* Force to data section. */
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
+};
+EXPORT_SYMBOL_GPL(tod_clock_base);
+
+u64 clock_comparator_max = -1ULL;
+EXPORT_SYMBOL_GPL(clock_comparator_max);
 
 static DEFINE_PER_CPU(struct clock_event_device, comparators);
 
@@ -75,7 +82,7 @@ void __init time_early_init(void)
 	struct ptff_qui qui;
 
 	/* Initialize TOD steering parameters */
-	tod_steering_end = sched_clock_base_cc;
+	tod_steering_end = *(unsigned long long *) &tod_clock_base[1];
 	vdso_data->ts_end = tod_steering_end;
 
 	if (!test_facility(28))
@@ -111,22 +118,27 @@ unsigned long long monotonic_clock(void)
 }
 EXPORT_SYMBOL(monotonic_clock);
 
-static void tod_to_timeval(__u64 todval, struct timespec64 *xt)
+static void ext_to_timespec64(unsigned char *clk, struct timespec64 *xt)
 {
-	unsigned long long sec;
+	unsigned long long high, low, rem, sec, nsec;
+
+	/* Split extendnd TOD clock to micro-seconds and sub-micro-seconds */
+	high = (*(unsigned long long *) clk) >> 4;
+	low = (*(unsigned long long *)&clk[7]) << 4;
+	/* Calculate seconds and nano-seconds */
+	sec = high;
+	rem = do_div(sec, 1000000);
+	nsec = (((low >> 32) + (rem << 32)) * 1000) >> 32;
 
-	sec = todval >> 12;
-	do_div(sec, 1000000);
 	xt->tv_sec = sec;
-	todval -= (sec * 1000000) << 12;
-	xt->tv_nsec = ((todval * 1000) >> 12);
+	xt->tv_nsec = nsec;
 }
 
 void clock_comparator_work(void)
 {
 	struct clock_event_device *cd;
 
-	S390_lowcore.clock_comparator = -1ULL;
+	S390_lowcore.clock_comparator = clock_comparator_max;
 	cd = this_cpu_ptr(&comparators);
 	cd->event_handler(cd);
 }
@@ -148,7 +160,7 @@ void init_cpu_timer(void)
 	struct clock_event_device *cd;
 	int cpu;
 
-	S390_lowcore.clock_comparator = -1ULL;
+	S390_lowcore.clock_comparator = clock_comparator_max;
 	set_clock_comparator(S390_lowcore.clock_comparator);
 
 	cpu = smp_processor_id();
@@ -179,7 +191,7 @@ static void clock_comparator_interrupt(struct ext_code ext_code,
 				       unsigned long param64)
 {
 	inc_irq_stat(IRQEXT_CLK);
-	if (S390_lowcore.clock_comparator == -1ULL)
+	if (S390_lowcore.clock_comparator == clock_comparator_max)
 		set_clock_comparator(S390_lowcore.clock_comparator);
 }
 
@@ -197,18 +209,28 @@ static void stp_reset(void);
 
 void read_persistent_clock64(struct timespec64 *ts)
 {
-	__u64 clock;
+	unsigned char clk[STORE_CLOCK_EXT_SIZE];
+	__u64 delta;
 
-	clock = get_tod_clock() - initial_leap_seconds;
-	tod_to_timeval(clock - TOD_UNIX_EPOCH, ts);
+	delta = initial_leap_seconds + TOD_UNIX_EPOCH;
+	get_tod_clock_ext(clk);
+	*(__u64 *) &clk[1] -= delta;
+	if (*(__u64 *) &clk[1] > delta)
+		clk[0]--;
+	ext_to_timespec64(clk, ts);
 }
 
 void read_boot_clock64(struct timespec64 *ts)
 {
-	__u64 clock;
+	unsigned char clk[STORE_CLOCK_EXT_SIZE];
+	__u64 delta;
 
-	clock = sched_clock_base_cc - initial_leap_seconds;
-	tod_to_timeval(clock - TOD_UNIX_EPOCH, ts);
+	delta = initial_leap_seconds + TOD_UNIX_EPOCH;
+	memcpy(clk, tod_clock_base, 16);
+	*(__u64 *) &clk[1] -= delta;
+	if (*(__u64 *) &clk[1] > delta)
+		clk[0]--;
+	ext_to_timespec64(clk, ts);
 }
 
 static u64 read_tod_clock(struct clocksource *cs)
@@ -406,7 +428,10 @@ static void clock_sync_global(unsigned long long delta)
 	struct ptff_qto qto;
 
 	/* Fixup the monotonic sched clock. */
-	sched_clock_base_cc += delta;
+	*(unsigned long long *) &tod_clock_base[1] += delta;
+	if (*(unsigned long long *) &tod_clock_base[1] < delta)
+		/* Epoch overflow */
+		tod_clock_base[0]++;
 	/* Adjust TOD steering parameters. */
 	vdso_data->tb_update_count++;
 	now = get_tod_clock();
@@ -437,7 +462,7 @@ static void clock_sync_global(unsigned long long delta)
 static void clock_sync_local(unsigned long long delta)
 {
 	/* Add the delta to the clock comparator. */
-	if (S390_lowcore.clock_comparator != -1ULL) {
+	if (S390_lowcore.clock_comparator != clock_comparator_max) {
 		S390_lowcore.clock_comparator += delta;
 		set_clock_comparator(S390_lowcore.clock_comparator);
 	}

arch/s390/lib/delay.c

@@ -57,7 +57,7 @@ static void __udelay_enabled(unsigned long long usecs)
 	end = get_tod_clock_fast() + (usecs << 12);
 	do {
 		clock_saved = 0;
-		if (end < S390_lowcore.clock_comparator) {
+		if (tod_after(S390_lowcore.clock_comparator, end)) {
 			clock_saved = local_tick_disable();
 			set_clock_comparator(end);
 		}