[S390] ETR support.

This patch adds support for clock synchronization to an external time
reference (ETR). The external time reference sends an oscillator
signal and a synchronization signal every 2^20 microseconds to keep
the TOD clocks of all connected servers in sync. For availability
two ETR units can be connected to a machine. If the clock deviates
for more than the sync-check tolerance all cpus get a machine check
that indicates that the clock is out of sync. For the lovely details
how to get the clock back in sync see the code below.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/kernel/s390_ext.c

@@ -125,15 +125,13 @@ void do_extint(struct pt_regs *regs, unsigned short code)
 		 * Make sure that the i/o interrupt did not "overtake"
 		 * the last HZ timer interrupt.
 		 */
-		account_ticks();
+		account_ticks(S390_lowcore.int_clock);
 	kstat_cpu(smp_processor_id()).irqs[EXTERNAL_INTERRUPT]++;
         index = ext_hash(code);
 	for (p = ext_int_hash[index]; p; p = p->next) {
-		if (likely(p->code == code)) {
-			if (likely(p->handler))
+		if (likely(p->code == code))
 			p->handler(code);
 	}
-	}
 	irq_exit();
 	set_irq_regs(old_regs);
 }

arch/s390/kernel/smp.c

@@ -457,9 +457,10 @@ int __devinit start_secondary(void *cpuvoid)
         /* Setup the cpu */
         cpu_init();
 	preempt_disable();
-        /* init per CPU timer */
+	/* Enable TOD clock interrupts on the secondary cpu. */
         init_cpu_timer();
 #ifdef CONFIG_VIRT_TIMER
+	/* Enable cpu timer interrupts on the secondary cpu. */
         init_cpu_vtimer();
 #endif
 	/* Enable pfault pseudo page faults on this cpu. */

arch/s390/kernel/vtime.c

@@ -524,16 +524,15 @@ EXPORT_SYMBOL(del_virt_timer);
 void init_cpu_vtimer(void)
 {
 	struct vtimer_queue *vt_list;
-	unsigned long cr0;
 
 	/* kick the virtual timer */
 	S390_lowcore.exit_timer = VTIMER_MAX_SLICE;
 	S390_lowcore.last_update_timer = VTIMER_MAX_SLICE;
 	asm volatile ("SPT %0" : : "m" (S390_lowcore.last_update_timer));
 	asm volatile ("STCK %0" : "=m" (S390_lowcore.last_update_clock));
-	__ctl_store(cr0, 0, 0);
-	cr0 |= 0x400;
-	__ctl_load(cr0, 0, 0);
+
+	/* enable cpu timer interrupts */
+	__ctl_set_bit(0,10);
 
 	vt_list = &per_cpu(virt_cpu_timer, smp_processor_id());
 	INIT_LIST_HEAD(&vt_list->list);
@@ -572,6 +571,7 @@ void __init vtime_init(void)
 	if (register_idle_notifier(&vtimer_idle_nb))
 		panic("Couldn't register idle notifier");
 
+	/* Enable cpu timer interrupts on the boot cpu. */
 	init_cpu_vtimer();
 }
 

arch/s390/lib/delay.c

 /*
- *  arch/s390/kernel/delay.c
+ *  arch/s390/lib/delay.c
  *    Precise Delay Loops for S390
  *
  *  S390 version
@@ -13,10 +13,8 @@
 
 #include <linux/sched.h>
 #include <linux/delay.h>
-
-#ifdef CONFIG_SMP
-#include <asm/smp.h>
-#endif
+#include <linux/timex.h>
+#include <linux/irqflags.h>
 
 void __delay(unsigned long loops)
 {
@@ -31,17 +29,39 @@ void __delay(unsigned long loops)
 }
 
 /*
- * Waits for 'usecs' microseconds using the tod clock, giving up the time slice
- * of the virtual PU inbetween to avoid congestion.
+ * Waits for 'usecs' microseconds using the TOD clock comparator.
  */
 void __udelay(unsigned long usecs)
 {
-	uint64_t start_cc;
+	u64 end, time, jiffy_timer = 0;
+	unsigned long flags, cr0, mask, dummy;
+
+	local_irq_save(flags);
+	if (raw_irqs_disabled_flags(flags)) {
+		jiffy_timer = S390_lowcore.jiffy_timer;
+		S390_lowcore.jiffy_timer = -1ULL - (4096 << 12);
+		__ctl_store(cr0, 0, 0);
+		dummy = (cr0 & 0xffff00e0) | 0x00000800;
+		__ctl_load(dummy , 0, 0);
+		mask = psw_kernel_bits | PSW_MASK_WAIT | PSW_MASK_EXT;
+	} else
+		mask = psw_kernel_bits | PSW_MASK_WAIT |
+			PSW_MASK_EXT | PSW_MASK_IO;
 
-        if (usecs == 0)
-                return;
-	start_cc = get_clock();
+	end = get_clock() + ((u64) usecs << 12);
 	do {
-		cpu_relax();
-	} while (((get_clock() - start_cc)/4096) < usecs);
+		time = end < S390_lowcore.jiffy_timer ?
+			end : S390_lowcore.jiffy_timer;
+		set_clock_comparator(time);
+		trace_hardirqs_on();
+		__load_psw_mask(mask);
+		local_irq_disable();
+	} while (get_clock() < end);
+
+	if (raw_irqs_disabled_flags(flags)) {
+		__ctl_load(cr0, 0, 0);
+		S390_lowcore.jiffy_timer = jiffy_timer;
+	}
+	set_clock_comparator(S390_lowcore.jiffy_timer);
+	local_irq_restore(flags);
 }

drivers/s390/block/dasd.c

@@ -1232,6 +1232,19 @@ __dasd_process_blk_queue(struct dasd_device * device)
 		if (IS_ERR(cqr)) {
 			if (PTR_ERR(cqr) == -ENOMEM)
 				break;	/* terminate request queue loop */
+			if (PTR_ERR(cqr) == -EAGAIN) {
+				/*
+				 * The current request cannot be build right
+				 * now, we have to try later. If this request
+				 * is the head-of-queue we stop the device
+				 * for 1/2 second.
+				 */
+				if (!list_empty(&device->ccw_queue))
+					break;
+				device->stopped |= DASD_STOPPED_PENDING;
+				dasd_set_timer(device, HZ/2);
+				break;
+			}
 			DBF_DEV_EVENT(DBF_ERR, device,
 				      "CCW creation failed (rc=%ld) "
 				      "on request %p",

drivers/s390/block/dasd_eckd.c

@@ -204,37 +204,39 @@ recs_per_track(struct dasd_eckd_characteristics * rdc,
 	return 0;
 }
 
-static inline void
+static inline int
 check_XRC (struct ccw1         *de_ccw,
            struct DE_eckd_data *data,
            struct dasd_device  *device)
 {
         struct dasd_eckd_private *private;
+	int rc;
 
         private = (struct dasd_eckd_private *) device->private;
+	if (!private->rdc_data.facilities.XRC_supported)
+		return 0;
 
         /* switch on System Time Stamp - needed for XRC Support */
-        if (private->rdc_data.facilities.XRC_supported) {
-
 	data->ga_extended |= 0x08; /* switch on 'Time Stamp Valid'   */
 	data->ga_extended |= 0x02; /* switch on 'Extended Parameter' */
 
-                data->ep_sys_time = get_clock ();
+	rc = get_sync_clock(&data->ep_sys_time);
+	/* Ignore return code if sync clock is switched off. */
+	if (rc == -ENOSYS || rc == -EACCES)
+		rc = 0;
 
 	de_ccw->count = sizeof (struct DE_eckd_data);
 	de_ccw->flags |= CCW_FLAG_SLI;
-        }
-
-        return;
-
-} /* end check_XRC */
+	return rc;
+}
 
-static inline void
+static inline int
 define_extent(struct ccw1 * ccw, struct DE_eckd_data * data, int trk,
 	      int totrk, int cmd, struct dasd_device * device)
 {
 	struct dasd_eckd_private *private;
 	struct ch_t geo, beg, end;
+	int rc = 0;
 
 	private = (struct dasd_eckd_private *) device->private;
 
@@ -263,12 +265,12 @@ define_extent(struct ccw1 * ccw, struct DE_eckd_data * data, int trk,
 	case DASD_ECKD_CCW_WRITE_KD_MT:
 		data->mask.perm = 0x02;
 		data->attributes.operation = private->attrib.operation;
-                check_XRC (ccw, data, device);
+		rc = check_XRC (ccw, data, device);
 		break;
 	case DASD_ECKD_CCW_WRITE_CKD:
 	case DASD_ECKD_CCW_WRITE_CKD_MT:
 		data->attributes.operation = DASD_BYPASS_CACHE;
-                check_XRC (ccw, data, device);
+		rc = check_XRC (ccw, data, device);
 		break;
 	case DASD_ECKD_CCW_ERASE:
 	case DASD_ECKD_CCW_WRITE_HOME_ADDRESS:
@@ -276,7 +278,7 @@ define_extent(struct ccw1 * ccw, struct DE_eckd_data * data, int trk,
 		data->mask.perm = 0x3;
 		data->mask.auth = 0x1;
 		data->attributes.operation = DASD_BYPASS_CACHE;
-                check_XRC (ccw, data, device);
+		rc = check_XRC (ccw, data, device);
 		break;
 	default:
 		DEV_MESSAGE(KERN_ERR, device, "unknown opcode 0x%x", cmd);
@@ -312,6 +314,7 @@ define_extent(struct ccw1 * ccw, struct DE_eckd_data * data, int trk,
 	data->beg_ext.head = beg.head;
 	data->end_ext.cyl = end.cyl;
 	data->end_ext.head = end.head;
+	return rc;
 }
 
 static inline void
@@ -1200,7 +1203,12 @@ dasd_eckd_build_cp(struct dasd_device * device, struct request *req)
 		return cqr;
 	ccw = cqr->cpaddr;
 	/* First ccw is define extent. */
-	define_extent(ccw++, cqr->data, first_trk, last_trk, cmd, device);
+	if (define_extent(ccw++, cqr->data, first_trk,
+			  last_trk, cmd, device) == -EAGAIN) {
+		/* Clock not in sync and XRC is enabled. Try again later. */
+		dasd_sfree_request(cqr, device);
+		return ERR_PTR(-EAGAIN);
+	}
 	/* Build locate_record+read/write/ccws. */
 	idaws = (unsigned long *) (cqr->data + sizeof(struct DE_eckd_data));
 	LO_data = (struct LO_eckd_data *) (idaws + cidaw);

drivers/s390/cio/cio.c

@@ -646,7 +646,7 @@ do_IRQ (struct pt_regs *regs)
 		 * Make sure that the i/o interrupt did not "overtake"
 		 * the last HZ timer interrupt.
 		 */
-		account_ticks();
+		account_ticks(S390_lowcore.int_clock);
 	/*
 	 * Get interrupt information from lowcore
 	 */
@@ -850,6 +850,19 @@ __disable_subchannel_easy(struct subchannel_id schid, struct schib *schib)
 	return -EBUSY; /* uhm... */
 }
 
+/* we can't use the normal udelay here, since it enables external interrupts */
+
+static void udelay_reset(unsigned long usecs)
+{
+	uint64_t start_cc, end_cc;
+
+	asm volatile ("STCK %0" : "=m" (start_cc));
+	do {
+		cpu_relax();
+		asm volatile ("STCK %0" : "=m" (end_cc));
+	} while (((end_cc - start_cc)/4096) < usecs);
+}
+
 static inline int
 __clear_subchannel_easy(struct subchannel_id schid)
 {
@@ -865,7 +878,7 @@ __clear_subchannel_easy(struct subchannel_id schid)
 			if (schid_equal(&ti.schid, &schid))
 				return 0;
 		}
-		udelay(100);
+		udelay_reset(100);
 	}
 	return -EBUSY;
 }

drivers/s390/s390mach.c

@@ -15,7 +15,7 @@
 #include <linux/time.h>
 #include <linux/device.h>
 #include <linux/kthread.h>
-
+#include <asm/etr.h>
 #include <asm/lowcore.h>
 #include <asm/cio.h>
 #include "cio/cio.h"
@@ -466,6 +466,19 @@ s390_do_machine_check(struct pt_regs *regs)
 			s390_handle_damage("unable to revalidate registers.");
 	}
 
+	if (mci->cd) {
+		/* Timing facility damage */
+		s390_handle_damage("TOD clock damaged");
+	}
+
+	if (mci->ed && mci->ec) {
+		/* External damage */
+		if (S390_lowcore.external_damage_code & (1U << ED_ETR_SYNC))
+			etr_sync_check();
+		if (S390_lowcore.external_damage_code & (1U << ED_ETR_SWITCH))
+			etr_switch_to_local();
+	}
+
 	if (mci->se)
 		/* Storage error uncorrected */
 		s390_handle_damage("received storage error uncorrected "
@@ -504,7 +517,7 @@ static int
 machine_check_init(void)
 {
 	init_MUTEX_LOCKED(&m_sem);
-	ctl_clear_bit(14, 25);	/* disable external damage MCH */
+	ctl_set_bit(14, 25);	/* enable external damage MCH */
 	ctl_set_bit(14, 27);    /* enable system recovery MCH */
 #ifdef CONFIG_MACHCHK_WARNING
 	ctl_set_bit(14, 24);	/* enable warning MCH */

drivers/s390/s390mach.h

@@ -102,4 +102,7 @@ static inline int stcrw(struct crw *pcrw )
 	return ccode;
 }
 
+#define ED_ETR_SYNC	12	/* External damage ETR sync check */
+#define ED_ETR_SWITCH	13	/* External damage ETR switch to local */
+
 #endif /* __s390mach */

include/asm-s390/etr.h

+/*
+ *  include/asm-s390/etr.h
+ *
+ *  Copyright IBM Corp. 2006
+ *  Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com)
+ */
+#ifndef __S390_ETR_H
+#define __S390_ETR_H
+
+/* ETR attachment control register */
+struct etr_eacr {
+	unsigned int e0		: 1;	/* port 0 stepping control */
+	unsigned int e1		: 1;	/* port 1 stepping control */
+	unsigned int _pad0	: 5;	/* must be 00100 */
+	unsigned int dp		: 1;	/* data port control */
+	unsigned int p0		: 1;	/* port 0 change recognition control */
+	unsigned int p1		: 1;	/* port 1 change recognition control */
+	unsigned int _pad1	: 3;	/* must be 000 */
+	unsigned int ea		: 1;	/* ETR alert control */
+	unsigned int es		: 1;	/* ETR sync check control */
+	unsigned int sl		: 1;	/* switch to local control */
+} __attribute__ ((packed));
+
+/* Port state returned by steai */
+enum etr_psc {
+	etr_psc_operational = 0,
+	etr_psc_semi_operational = 1,
+	etr_psc_protocol_error =  4,
+	etr_psc_no_symbols = 8,
+	etr_psc_no_signal = 12,
+	etr_psc_pps_mode = 13
+};
+
+/* Logical port state returned by stetr */
+enum etr_lpsc {
+	etr_lpsc_operational_step = 0,
+	etr_lpsc_operational_alt = 1,
+	etr_lpsc_semi_operational = 2,
+	etr_lpsc_protocol_error =  4,
+	etr_lpsc_no_symbol_sync = 8,
+	etr_lpsc_no_signal = 12,
+	etr_lpsc_pps_mode = 13
+};
+
+/* ETR status words */
+struct etr_esw {
+	struct etr_eacr eacr;		/* attachment control register */
+	unsigned int y		: 1;	/* stepping mode */
+	unsigned int _pad0	: 5;	/* must be 00000 */
+	unsigned int p		: 1;	/* stepping port number */
+	unsigned int q		: 1;	/* data port number */
+	unsigned int psc0	: 4;	/* port 0 state code */
+	unsigned int psc1	: 4;	/* port 1 state code */
+} __attribute__ ((packed));
+
+/* Second level data register status word */
+struct etr_slsw {
+	unsigned int vv1	: 1;	/* copy of validity bit data frame 1 */
+	unsigned int vv2	: 1;	/* copy of validity bit data frame 2 */
+	unsigned int vv3	: 1;	/* copy of validity bit data frame 3 */
+	unsigned int vv4	: 1;	/* copy of validity bit data frame 4 */
+	unsigned int _pad0	: 19;	/* must by all zeroes */
+	unsigned int n		: 1;	/* EAF port number */
+	unsigned int v1		: 1;	/* validity bit ETR data frame 1 */
+	unsigned int v2		: 1;	/* validity bit ETR data frame 2 */
+	unsigned int v3		: 1;	/* validity bit ETR data frame 3 */
+	unsigned int v4		: 1;	/* validity bit ETR data frame 4 */
+	unsigned int _pad1	: 4;	/* must be 0000 */
+} __attribute__ ((packed));
+
+/* ETR data frames */
+struct etr_edf1 {
+	unsigned int u		: 1;	/* untuned bit */
+	unsigned int _pad0	: 1;	/* must be 0 */
+	unsigned int r		: 1;	/* service request bit */
+	unsigned int _pad1	: 4;	/* must be 0000 */
+	unsigned int a		: 1;	/* time adjustment bit */
+	unsigned int net_id	: 8;	/* ETR network id */
+	unsigned int etr_id	: 8;	/* id of ETR which sends data frames */
+	unsigned int etr_pn	: 8;	/* port number of ETR output port */
+} __attribute__ ((packed));
+
+struct etr_edf2 {
+	unsigned int etv	: 32;	/* Upper 32 bits of TOD. */
+} __attribute__ ((packed));
+
+struct etr_edf3 {
+	unsigned int rc		: 8;	/* failure reason code */
+	unsigned int _pad0	: 3;	/* must be 000 */
+	unsigned int c		: 1;	/* ETR coupled bit */
+	unsigned int tc		: 4;	/* ETR type code */
+	unsigned int blto	: 8;	/* biased local time offset */
+					/* (blto - 128) * 15 = minutes */
+	unsigned int buo	: 8;	/* biased utc offset */
+					/* (buo - 128) = leap seconds */
+} __attribute__ ((packed));
+
+struct etr_edf4 {
+	unsigned int ed		: 8;	/* ETS device dependent data */
+	unsigned int _pad0	: 1;	/* must be 0 */
+	unsigned int buc	: 5;	/* biased ut1 correction */
+					/* (buc - 16) * 0.1 seconds */
+	unsigned int em		: 6;	/* ETS error magnitude */
+	unsigned int dc		: 6;	/* ETS drift code */
+	unsigned int sc		: 6;	/* ETS steering code */
+} __attribute__ ((packed));
+
+/*
+ * ETR attachment information block, two formats
+ * format 1 has 4 reserved words with a size of 64 bytes
+ * format 2 has 16 reserved words with a size of 96 bytes
+ */
+struct etr_aib {
+	struct etr_esw esw;
+	struct etr_slsw slsw;
+	unsigned long long tsp;
+	struct etr_edf1 edf1;
+	struct etr_edf2 edf2;
+	struct etr_edf3 edf3;
+	struct etr_edf4 edf4;
+	unsigned int reserved[16];
+} __attribute__ ((packed,aligned(8)));
+
+/* ETR interruption parameter */
+struct etr_interruption_parameter {
+	unsigned int _pad0	: 8;
+	unsigned int pc0	: 1;	/* port 0 state change */
+	unsigned int pc1	: 1;	/* port 1 state change */
+	unsigned int _pad1	: 3;
+	unsigned int eai	: 1;	/* ETR alert indication */
+	unsigned int _pad2	: 18;
+} __attribute__ ((packed));
+
+/* Query TOD offset result */
+struct etr_ptff_qto {
+	unsigned long long physical_clock;
+	unsigned long long tod_offset;
+	unsigned long long logical_tod_offset;
+	unsigned long long tod_epoch_difference;
+} __attribute__ ((packed));
+
+/* Inline assembly helper functions */
+static inline int etr_setr(struct etr_eacr *ctrl)
+{
+	int rc = -ENOSYS;
+
+	asm volatile(
+		"	.insn	s,0xb2160000,0(%2)\n"
+		"0:	la	%0,0\n"
+		"1:\n"
+		EX_TABLE(0b,1b)
+		: "+d" (rc) : "m" (*ctrl), "a" (ctrl));
+	return rc;
+}
+
+/* Stores a format 1 aib with 64 bytes */
+static inline int etr_stetr(struct etr_aib *aib)
+{
+	int rc = -ENOSYS;
+
+	asm volatile(
+		"	.insn	s,0xb2170000,0(%2)\n"
+		"0:	la	%0,0\n"
+		"1:\n"
+		EX_TABLE(0b,1b)
+		: "+d" (rc) : "m" (*aib), "a" (aib));
+	return rc;
+}
+
+/* Stores a format 2 aib with 96 bytes for specified port */
+static inline int etr_steai(struct etr_aib *aib, unsigned int func)
+{
+	register unsigned int reg0 asm("0") = func;
+	int rc = -ENOSYS;
+
+	asm volatile(
+		"	.insn	s,0xb2b30000,0(%2)\n"
+		"0:	la	%0,0\n"
+		"1:\n"
+		EX_TABLE(0b,1b)
+		: "+d" (rc) : "m" (*aib), "a" (aib), "d" (reg0));
+	return rc;
+}
+
+/* Function codes for the steai instruction. */
+#define ETR_STEAI_STEPPING_PORT		0x10
+#define ETR_STEAI_ALTERNATE_PORT	0x11
+#define ETR_STEAI_PORT_0		0x12
+#define ETR_STEAI_PORT_1		0x13
+
+static inline int etr_ptff(void *ptff_block, unsigned int func)
+{
+	register unsigned int reg0 asm("0") = func;
+	register unsigned long reg1 asm("1") = (unsigned long) ptff_block;
+	int rc = -ENOSYS;
+
+	asm volatile(
+		"	.word	0x0104\n"
+		"	ipm	%0\n"
+		"	srl	%0,28\n"
+		: "=d" (rc), "=m" (ptff_block)
+		: "d" (reg0), "d" (reg1), "m" (ptff_block) : "cc");
+	return rc;
+}
+
+/* Function codes for the ptff instruction. */
+#define ETR_PTFF_QAF	0x00	/* query available functions */
+#define ETR_PTFF_QTO	0x01	/* query tod offset */
+#define ETR_PTFF_QSI	0x02	/* query steering information */
+#define ETR_PTFF_ATO	0x40	/* adjust tod offset */
+#define ETR_PTFF_STO	0x41	/* set tod offset */
+#define ETR_PTFF_SFS	0x42	/* set fine steering rate */
+#define ETR_PTFF_SGS	0x43	/* set gross steering rate */
+
+/* Functions needed by the machine check handler */
+extern void etr_switch_to_local(void);
+extern void etr_sync_check(void);
+
+#endif /* __S390_ETR_H */

include/asm-s390/hardirq.h

@@ -32,6 +32,6 @@ typedef struct {
 
 #define HARDIRQ_BITS	8
 
-extern void account_ticks(void);
+extern void account_ticks(u64 time);
 
 #endif /* __ASM_HARDIRQ_H */

include/asm-s390/timex.h

@@ -11,6 +11,41 @@
 #ifndef _ASM_S390_TIMEX_H
 #define _ASM_S390_TIMEX_H
 
+/* Inline functions for clock register access. */
+static inline int set_clock(__u64 time)
+{
+	int cc;
+
+	asm volatile(
+		"   sck   0(%2)\n"
+		"   ipm   %0\n"
+		"   srl   %0,28\n"
+		: "=d" (cc) : "m" (time), "a" (&time) : "cc");
+	return cc;
+}
+
+static inline int store_clock(__u64 *time)
+{
+	int cc;
+
+	asm volatile(
+		"   stck  0(%2)\n"
+		"   ipm   %0\n"
+		"   srl   %0,28\n"
+		: "=d" (cc), "=m" (*time) : "a" (time) : "cc");
+	return cc;
+}
+
+static inline void set_clock_comparator(__u64 time)
+{
+	asm volatile("sckc 0(%1)" : : "m" (time), "a" (&time));
+}
+
+static inline void store_clock_comparator(__u64 *time)
+{
+	asm volatile("stckc 0(%1)" : "=m" (*time) : "a" (time));
+}
+
 #define CLOCK_TICK_RATE	1193180 /* Underlying HZ */
 
 typedef unsigned long long cycles_t;
@@ -32,6 +67,7 @@ static inline cycles_t get_cycles(void)
 	return (cycles_t) get_clock() >> 2;
 }
 
+int get_sync_clock(unsigned long long *clock);
 void init_cpu_timer(void);
 
 #endif