Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-next

Pull sparc updates from David Miller:

 1) Queued spinlocks and rwlocks for sparc64, from Babu Moger.

 2) Some const'ification from Arvind Yadav.

 3) LDC/VIO driver infrastructure changes to facilitate future upcoming
    drivers, from Jag Raman.

 4) Initialize sched_clock() et al. early so that the initial printk
    timestamps are all done while the implementation is available and
    functioning. From Pavel Tatashin.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-next: (38 commits)
  sparc: kernel: pmc: make of_device_ids const.
  sparc64: fix typo in property
  sparc64: add port_id to VIO device metadata
  sparc64: Enhance search for VIO device in MDESC
  sparc64: enhance VIO device probing
  sparc64: check if a client is allowed to register for MDESC notifications
  sparc64: remove restriction on VIO device name size
  sparc64: refactor code to obtain cfg_handle property from MDESC
  sparc64: add MDESC node name property to VIO device metadata
  sparc64: mdesc: use __GFP_REPEAT action modifier for VM allocation
  sparc64: expand MDESC interface
  sparc64: skip handshake for LDC channels in RAW mode
  sparc64: specify the device class in VIO version info. packet
  sparc64: ensure VIO operations are defined while being used
  sparc: kernel: apc: make of_device_ids const
  sparc/time: make of_device_ids const
  sparc64: broken %tick frequency on spitfire cpus
  sparc64: use prom interface to get %stick frequency
  sparc64: optimize functions that access tick
  sparc64: add hot-patched and inlined get_tick()
  ...

arch/sparc/Kconfig

@@ -83,6 +83,8 @@ config SPARC64
 	select ARCH_SUPPORTS_ATOMIC_RMW
 	select HAVE_NMI
 	select HAVE_REGS_AND_STACK_ACCESS_API
+	select ARCH_USE_QUEUED_RWLOCKS
+	select ARCH_USE_QUEUED_SPINLOCKS
 
 config ARCH_DEFCONFIG
 	string
@@ -92,6 +94,9 @@ config ARCH_DEFCONFIG
 config ARCH_PROC_KCORE_TEXT
 	def_bool y
 
+config CPU_BIG_ENDIAN
+	def_bool y
+
 config ARCH_ATU
 	bool
 	default y if SPARC64

arch/sparc/include/asm/cmpxchg_64.h

@@ -6,6 +6,17 @@
 #ifndef __ARCH_SPARC64_CMPXCHG__
 #define __ARCH_SPARC64_CMPXCHG__
 
+static inline unsigned long
+__cmpxchg_u32(volatile int *m, int old, int new)
+{
+	__asm__ __volatile__("cas [%2], %3, %0"
+			     : "=&r" (new)
+			     : "0" (new), "r" (m), "r" (old)
+			     : "memory");
+
+	return new;
+}
+
 static inline unsigned long xchg32(__volatile__ unsigned int *m, unsigned int val)
 {
 	unsigned long tmp1, tmp2;
@@ -44,10 +55,38 @@ static inline unsigned long xchg64(__volatile__ unsigned long *m, unsigned long
 
 void __xchg_called_with_bad_pointer(void);
 
+/*
+ * Use 4 byte cas instruction to achieve 2 byte xchg. Main logic
+ * here is to get the bit shift of the byte we are interested in.
+ * The XOR is handy for reversing the bits for big-endian byte order.
+ */
+static inline unsigned long
+xchg16(__volatile__ unsigned short *m, unsigned short val)
+{
+	unsigned long maddr = (unsigned long)m;
+	int bit_shift = (((unsigned long)m & 2) ^ 2) << 3;
+	unsigned int mask = 0xffff << bit_shift;
+	unsigned int *ptr = (unsigned int  *) (maddr & ~2);
+	unsigned int old32, new32, load32;
+
+	/* Read the old value */
+	load32 = *ptr;
+
+	do {
+		old32 = load32;
+		new32 = (load32 & (~mask)) | val << bit_shift;
+		load32 = __cmpxchg_u32(ptr, old32, new32);
+	} while (load32 != old32);
+
+	return (load32 & mask) >> bit_shift;
+}
+
 static inline unsigned long __xchg(unsigned long x, __volatile__ void * ptr,
 				       int size)
 {
 	switch (size) {
+	case 2:
+		return xchg16(ptr, x);
 	case 4:
 		return xchg32(ptr, x);
 	case 8:
@@ -65,10 +104,11 @@ static inline unsigned long __xchg(unsigned long x, __volatile__ void * ptr,
 
 #include <asm-generic/cmpxchg-local.h>
 
+
 static inline unsigned long
-__cmpxchg_u32(volatile int *m, int old, int new)
+__cmpxchg_u64(volatile long *m, unsigned long old, unsigned long new)
 {
-	__asm__ __volatile__("cas [%2], %3, %0"
+	__asm__ __volatile__("casx [%2], %3, %0"
 			     : "=&r" (new)
 			     : "0" (new), "r" (m), "r" (old)
 			     : "memory");
@@ -76,15 +116,31 @@ __cmpxchg_u32(volatile int *m, int old, int new)
 	return new;
 }
 
+/*
+ * Use 4 byte cas instruction to achieve 1 byte cmpxchg. Main logic
+ * here is to get the bit shift of the byte we are interested in.
+ * The XOR is handy for reversing the bits for big-endian byte order
+ */
 static inline unsigned long
-__cmpxchg_u64(volatile long *m, unsigned long old, unsigned long new)
+__cmpxchg_u8(volatile unsigned char *m, unsigned char old, unsigned char new)
 {
-	__asm__ __volatile__("casx [%2], %3, %0"
-			     : "=&r" (new)
-			     : "0" (new), "r" (m), "r" (old)
-			     : "memory");
+	unsigned long maddr = (unsigned long)m;
+	int bit_shift = (((unsigned long)m & 3) ^ 3) << 3;
+	unsigned int mask = 0xff << bit_shift;
+	unsigned int *ptr = (unsigned int *) (maddr & ~3);
+	unsigned int old32, new32, load;
+	unsigned int load32 = *ptr;
+
+	do {
+		new32 = (load32 & ~mask) | (new << bit_shift);
+		old32 = (load32 & ~mask) | (old << bit_shift);
+		load32 = __cmpxchg_u32(ptr, old32, new32);
+		if (load32 == old32)
+			return old;
+		load = (load32 & mask) >> bit_shift;
+	} while (load == old);
 
-	return new;
+	return load;
 }
 
 /* This function doesn't exist, so you'll get a linker error
@@ -95,6 +151,8 @@ static inline unsigned long
 __cmpxchg(volatile void *ptr, unsigned long old, unsigned long new, int size)
 {
 	switch (size) {
+		case 1:
+			return __cmpxchg_u8(ptr, old, new);
 		case 4:
 			return __cmpxchg_u32(ptr, old, new);
 		case 8:

arch/sparc/include/asm/ldc.h

@@ -48,6 +48,8 @@ struct ldc_channel_config {
 #define LDC_STATE_READY		0x03
 #define LDC_STATE_CONNECTED	0x04
 
+#define	LDC_PACKET_SIZE		64
+
 struct ldc_channel;
 
 /* Allocate state for a channel.  */
@@ -72,6 +74,12 @@ int ldc_connect(struct ldc_channel *lp);
 int ldc_disconnect(struct ldc_channel *lp);
 
 int ldc_state(struct ldc_channel *lp);
+void ldc_set_state(struct ldc_channel *lp, u8 state);
+int ldc_mode(struct ldc_channel *lp);
+void __ldc_print(struct ldc_channel *lp, const char *caller);
+int ldc_rx_reset(struct ldc_channel *lp);
+
+#define	ldc_print(chan)	__ldc_print(chan, __func__)
 
 /* Read and write operations.  Only valid when the link is up.  */
 int ldc_write(struct ldc_channel *lp, const void *buf,

arch/sparc/include/asm/mdesc.h

@@ -16,6 +16,7 @@ struct mdesc_handle *mdesc_grab(void);
 void mdesc_release(struct mdesc_handle *);
 
 #define MDESC_NODE_NULL		(~(u64)0)
+#define MDESC_MAX_STR_LEN	256
 
 u64 mdesc_node_by_name(struct mdesc_handle *handle,
 		       u64 from_node, const char *name);
@@ -62,15 +63,32 @@ u64 mdesc_arc_target(struct mdesc_handle *hp, u64 arc);
 void mdesc_update(void);
 
 struct mdesc_notifier_client {
-	void (*add)(struct mdesc_handle *handle, u64 node);
-	void (*remove)(struct mdesc_handle *handle, u64 node);
-
+	void (*add)(struct mdesc_handle *handle, u64 node,
+		    const char *node_name);
+	void (*remove)(struct mdesc_handle *handle, u64 node,
+		       const char *node_name);
 	const char			*node_name;
 	struct mdesc_notifier_client	*next;
 };
 
 void mdesc_register_notifier(struct mdesc_notifier_client *client);
 
+union md_node_info {
+	struct vdev_port {
+		u64 id;				/* id */
+		u64 parent_cfg_hdl;		/* parent config handle */
+		const char *name;		/* name (property) */
+	} vdev_port;
+	struct ds_port {
+		u64 id;				/* id */
+	} ds_port;
+};
+
+u64 mdesc_get_node(struct mdesc_handle *hp, const char *node_name,
+		   union md_node_info *node_info);
+int mdesc_get_node_info(struct mdesc_handle *hp, u64 node,
+			const char *node_name, union md_node_info *node_info);
+
 void mdesc_fill_in_cpu_data(cpumask_t *mask);
 void mdesc_populate_present_mask(cpumask_t *mask);
 void mdesc_get_page_sizes(cpumask_t *mask, unsigned long *pgsz_mask);

arch/sparc/include/asm/qrwlock.h

+#ifndef _ASM_SPARC_QRWLOCK_H
+#define _ASM_SPARC_QRWLOCK_H
+
+#include <asm-generic/qrwlock_types.h>
+#include <asm-generic/qrwlock.h>
+
+#endif /* _ASM_SPARC_QRWLOCK_H */

arch/sparc/include/asm/qspinlock.h

+#ifndef _ASM_SPARC_QSPINLOCK_H
+#define _ASM_SPARC_QSPINLOCK_H
+
+#include <asm-generic/qspinlock_types.h>
+#include <asm-generic/qspinlock.h>
+
+#endif /* _ASM_SPARC_QSPINLOCK_H */

arch/sparc/include/asm/spinlock_64.h

@@ -10,216 +10,12 @@
 
 #include <asm/processor.h>
 #include <asm/barrier.h>
-
-/* To get debugging spinlocks which detect and catch
- * deadlock situations, set CONFIG_DEBUG_SPINLOCK
- * and rebuild your kernel.
- */
-
-/* Because we play games to save cycles in the non-contention case, we
- * need to be extra careful about branch targets into the "spinning"
- * code.  They live in their own section, but the newer V9 branches
- * have a shorter range than the traditional 32-bit sparc branch
- * variants.  The rule is that the branches that go into and out of
- * the spinner sections must be pre-V9 branches.
- */
-
-#define arch_spin_is_locked(lp)	((lp)->lock != 0)
-
-static inline void arch_spin_unlock_wait(arch_spinlock_t *lock)
-{
-	smp_cond_load_acquire(&lock->lock, !VAL);
-}
-
-static inline void arch_spin_lock(arch_spinlock_t *lock)
-{
-	unsigned long tmp;
-
-	__asm__ __volatile__(
-"1:	ldstub		[%1], %0\n"
-"	brnz,pn		%0, 2f\n"
-"	 nop\n"
-"	.subsection	2\n"
-"2:	ldub		[%1], %0\n"
-"	brnz,pt		%0, 2b\n"
-"	 nop\n"
-"	ba,a,pt		%%xcc, 1b\n"
-"	.previous"
-	: "=&r" (tmp)
-	: "r" (lock)
-	: "memory");
-}
-
-static inline int arch_spin_trylock(arch_spinlock_t *lock)
-{
-	unsigned long result;
-
-	__asm__ __volatile__(
-"	ldstub		[%1], %0\n"
-	: "=r" (result)
-	: "r" (lock)
-	: "memory");
-
-	return (result == 0UL);
-}
-
-static inline void arch_spin_unlock(arch_spinlock_t *lock)
-{
-	__asm__ __volatile__(
-"	stb		%%g0, [%0]"
-	: /* No outputs */
-	: "r" (lock)
-	: "memory");
-}
-
-static inline void arch_spin_lock_flags(arch_spinlock_t *lock, unsigned long flags)
-{
-	unsigned long tmp1, tmp2;
-
-	__asm__ __volatile__(
-"1:	ldstub		[%2], %0\n"
-"	brnz,pn		%0, 2f\n"
-"	 nop\n"
-"	.subsection	2\n"
-"2:	rdpr		%%pil, %1\n"
-"	wrpr		%3, %%pil\n"
-"3:	ldub		[%2], %0\n"
-"	brnz,pt		%0, 3b\n"
-"	 nop\n"
-"	ba,pt		%%xcc, 1b\n"
-"	 wrpr		%1, %%pil\n"
-"	.previous"
-	: "=&r" (tmp1), "=&r" (tmp2)
-	: "r"(lock), "r"(flags)
-	: "memory");
-}
-
-/* Multi-reader locks, these are much saner than the 32-bit Sparc ones... */
-
-static inline void arch_read_lock(arch_rwlock_t *lock)
-{
-	unsigned long tmp1, tmp2;
-
-	__asm__ __volatile__ (
-"1:	ldsw		[%2], %0\n"
-"	brlz,pn		%0, 2f\n"
-"4:	 add		%0, 1, %1\n"
-"	cas		[%2], %0, %1\n"
-"	cmp		%0, %1\n"
-"	bne,pn		%%icc, 1b\n"
-"	 nop\n"
-"	.subsection	2\n"
-"2:	ldsw		[%2], %0\n"
-"	brlz,pt		%0, 2b\n"
-"	 nop\n"
-"	ba,a,pt		%%xcc, 4b\n"
-"	.previous"
-	: "=&r" (tmp1), "=&r" (tmp2)
-	: "r" (lock)
-	: "memory");
-}
-
-static inline int arch_read_trylock(arch_rwlock_t *lock)
-{
-	int tmp1, tmp2;
-
-	__asm__ __volatile__ (
-"1:	ldsw		[%2], %0\n"
-"	brlz,a,pn	%0, 2f\n"
-"	 mov		0, %0\n"
-"	add		%0, 1, %1\n"
-"	cas		[%2], %0, %1\n"
-"	cmp		%0, %1\n"
-"	bne,pn		%%icc, 1b\n"
-"	 mov		1, %0\n"
-"2:"
-	: "=&r" (tmp1), "=&r" (tmp2)
-	: "r" (lock)
-	: "memory");
-
-	return tmp1;
-}
-
-static inline void arch_read_unlock(arch_rwlock_t *lock)
-{
-	unsigned long tmp1, tmp2;
-
-	__asm__ __volatile__(
-"1:	lduw	[%2], %0\n"
-"	sub	%0, 1, %1\n"
-"	cas	[%2], %0, %1\n"
-"	cmp	%0, %1\n"
-"	bne,pn	%%xcc, 1b\n"
-"	 nop"
-	: "=&r" (tmp1), "=&r" (tmp2)
-	: "r" (lock)
-	: "memory");
-}
-
-static inline void arch_write_lock(arch_rwlock_t *lock)
-{
-	unsigned long mask, tmp1, tmp2;
-
-	mask = 0x80000000UL;
-
-	__asm__ __volatile__(
-"1:	lduw		[%2], %0\n"
-"	brnz,pn		%0, 2f\n"
-"4:	 or		%0, %3, %1\n"
-"	cas		[%2], %0, %1\n"
-"	cmp		%0, %1\n"
-"	bne,pn		%%icc, 1b\n"
-"	 nop\n"
-"	.subsection	2\n"
-"2:	lduw		[%2], %0\n"
-"	brnz,pt		%0, 2b\n"
-"	 nop\n"
-"	ba,a,pt		%%xcc, 4b\n"
-"	.previous"
-	: "=&r" (tmp1), "=&r" (tmp2)
-	: "r" (lock), "r" (mask)
-	: "memory");
-}
-
-static inline void arch_write_unlock(arch_rwlock_t *lock)
-{
-	__asm__ __volatile__(
-"	stw		%%g0, [%0]"
-	: /* no outputs */
-	: "r" (lock)
-	: "memory");
-}
-
-static inline int arch_write_trylock(arch_rwlock_t *lock)
-{
-	unsigned long mask, tmp1, tmp2, result;
-
-	mask = 0x80000000UL;
-
-	__asm__ __volatile__(
-"	mov		0, %2\n"
-"1:	lduw		[%3], %0\n"
-"	brnz,pn		%0, 2f\n"
-"	 or		%0, %4, %1\n"
-"	cas		[%3], %0, %1\n"
-"	cmp		%0, %1\n"
-"	bne,pn		%%icc, 1b\n"
-"	 nop\n"
-"	mov		1, %2\n"
-"2:"
-	: "=&r" (tmp1), "=&r" (tmp2), "=&r" (result)
-	: "r" (lock), "r" (mask)
-	: "memory");
-
-	return result;
-}
+#include <asm/qrwlock.h>
+#include <asm/qspinlock.h>
 
 #define arch_read_lock_flags(p, f) arch_read_lock(p)
 #define arch_write_lock_flags(p, f) arch_write_lock(p)
 
-#define arch_read_can_lock(rw)		(!((rw)->lock & 0x80000000UL))
-#define arch_write_can_lock(rw)	(!(rw)->lock)
-
 #define arch_spin_relax(lock)	cpu_relax()
 #define arch_read_relax(lock)	cpu_relax()
 #define arch_write_relax(lock)	cpu_relax()

arch/sparc/include/asm/spinlock_types.h

 #ifndef __SPARC_SPINLOCK_TYPES_H
 #define __SPARC_SPINLOCK_TYPES_H
 
-#ifndef __LINUX_SPINLOCK_TYPES_H
-# error "please don't include this file directly"
-#endif
+#ifdef CONFIG_QUEUED_SPINLOCKS
+#include <asm-generic/qspinlock_types.h>
+#else
 
 typedef struct {
 	volatile unsigned char lock;
 } arch_spinlock_t;
 
 #define __ARCH_SPIN_LOCK_UNLOCKED	{ 0 }
+#endif /* CONFIG_QUEUED_SPINLOCKS */
 
+#ifdef CONFIG_QUEUED_RWLOCKS
+#include <asm-generic/qrwlock_types.h>
+#else
 typedef struct {
 	volatile unsigned int lock;
 } arch_rwlock_t;
 
 #define __ARCH_RW_LOCK_UNLOCKED		{ 0 }
-
+#endif /* CONFIG_QUEUED_RWLOCKS */
 #endif

arch/sparc/include/asm/timer_64.h

@@ -9,7 +9,12 @@
 #include <linux/types.h>
 #include <linux/init.h>
 
+/* The most frequently accessed fields should be first,
+ * to fit into the same cacheline.
+ */
 struct sparc64_tick_ops {
+	unsigned long ticks_per_nsec_quotient;
+	unsigned long offset;
 	unsigned long long (*get_tick)(void);
 	int (*add_compare)(unsigned long);
 	unsigned long softint_mask;
@@ -17,6 +22,8 @@ struct sparc64_tick_ops {
 
 	void (*init_tick)(void);
 	unsigned long (*add_tick)(unsigned long);
+	unsigned long (*get_frequency)(void);
+	unsigned long frequency;
 
 	char *name;
 };
@@ -27,4 +34,64 @@ unsigned long sparc64_get_clock_tick(unsigned int cpu);
 void setup_sparc64_timer(void);
 void __init time_init(void);
 
+#define TICK_PRIV_BIT		BIT(63)
+#define TICKCMP_IRQ_BIT		BIT(63)
+
+#define HBIRD_STICKCMP_ADDR	0x1fe0000f060UL
+#define HBIRD_STICK_ADDR	0x1fe0000f070UL
+
+#define GET_TICK_NINSTR		13
+struct get_tick_patch {
+	unsigned int	addr;
+	unsigned int	tick[GET_TICK_NINSTR];
+	unsigned int	stick[GET_TICK_NINSTR];
+};
+
+extern struct get_tick_patch __get_tick_patch;
+extern struct get_tick_patch __get_tick_patch_end;
+
+static inline unsigned long get_tick(void)
+{
+	unsigned long tick, tmp1, tmp2;
+
+	__asm__ __volatile__(
+	/* read hbtick 13 instructions */
+	"661:\n"
+	"	mov	0x1fe, %1\n"
+	"	sllx	%1, 0x20, %1\n"
+	"	sethi	%%hi(0xf000), %2\n"
+	"	or	%2, 0x70, %2\n"
+	"	or	%1, %2, %1\n"	/* %1 = HBIRD_STICK_ADDR */
+	"	add	%1, 8, %2\n"
+	"	ldxa	[%2]%3, %0\n"
+	"	ldxa	[%1]%3, %1\n"
+	"	ldxa	[%2]%3, %2\n"
+	"	sub	%2, %0, %0\n"	/* don't modify %xcc */
+	"	brnz,pn	%0, 661b\n"	/* restart to save one register */
+	"	 sllx	%2, 32, %2\n"
+	"	or	%2, %1, %0\n"
+	/* Common/not patched code */
+	"	sllx	%0, 1, %0\n"
+	"	srlx	%0, 1, %0\n"	/* Clear TICK_PRIV_BIT */
+	/* Beginning of patch section */
+	"	.section .get_tick_patch, \"ax\"\n"
+	"	.word	661b\n"
+	/* read tick 2 instructions and 11 skipped */
+	"	ba	1f\n"
+	"	 rd	%%tick, %0\n"
+	"	.skip	4 * (%4 - 2)\n"
+	"1:\n"
+	/* read stick 2 instructions and 11 skipped */
+	"	ba	1f\n"
+	"	 rd	%%asr24, %0\n"
+	"	.skip	4 * (%4 - 2)\n"
+	"1:\n"
+	/* End of patch section */
+	"	.previous\n"
+	: "=&r" (tick), "=&r" (tmp1), "=&r" (tmp2)
+	: "i" (ASI_PHYS_BYPASS_EC_E), "i" (GET_TICK_NINSTR));
+
+	return tick;
+}
+
 #endif /* _SPARC64_TIMER_H */

arch/sparc/include/asm/vio.h

@@ -316,24 +316,33 @@ static inline u32 vio_dring_prev(struct vio_dring_state *dr, u32 index)
 }
 
 #define VIO_MAX_TYPE_LEN	32
+#define VIO_MAX_NAME_LEN	32
 #define VIO_MAX_COMPAT_LEN	64
 
 struct vio_dev {
 	u64			mp;
 	struct device_node	*dp;
 
+	char			node_name[VIO_MAX_NAME_LEN];
 	char			type[VIO_MAX_TYPE_LEN];
 	char			compat[VIO_MAX_COMPAT_LEN];
 	int			compat_len;
 
 	u64			dev_no;
-	u64			id;
 
+	unsigned long		port_id;
 	unsigned long		channel_id;
 
 	unsigned int		tx_irq;
 	unsigned int		rx_irq;
 	u64			rx_ino;
+	u64			tx_ino;
+
+	/* Handle to the root of "channel-devices" sub-tree in MDESC */
+	u64			cdev_handle;
+
+	/* MD specific data used to identify the vdev in MD */
+	union md_node_info	md_node_info;
 
 	struct device		dev;
 };
@@ -347,6 +356,7 @@ struct vio_driver {
 	void (*shutdown)(struct vio_dev *dev);
 	unsigned long			driver_data;
 	struct device_driver		driver;
+	bool				no_irq;
 };
 
 struct vio_version {
@@ -490,5 +500,6 @@ int vio_driver_init(struct vio_driver_state *vio, struct vio_dev *vdev,
 
 void vio_port_up(struct vio_driver_state *vio);
 int vio_set_intr(unsigned long dev_ino, int state);
+u64 vio_vdev_node(struct mdesc_handle *hp, struct vio_dev *vdev);
 
 #endif /* _SPARC64_VIO_H */

arch/sparc/kernel/apc.c

@@ -167,7 +167,7 @@ static int apc_probe(struct platform_device *op)
 	return 0;
 }
 
-static struct of_device_id apc_match[] = {
+static const struct of_device_id apc_match[] = {
 	{
 		.name = APC_OBPNAME,
 	},

arch/sparc/kernel/kernel.h

@@ -52,6 +52,9 @@ asmlinkage void do_rt_sigreturn32(struct pt_regs *regs);
 void do_signal32(struct pt_regs * regs);
 asmlinkage int do_sys32_sigstack(u32 u_ssptr, u32 u_ossptr, unsigned long sp);
 
+/* time_64.c */
+void __init time_init_early(void);
+
 /* compat_audit.c */
 extern unsigned int sparc32_dir_class[];
 extern unsigned int sparc32_chattr_class[];

arch/sparc/kernel/ldc.c

@@ -34,7 +34,6 @@
 
 static char version[] =
 	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
-#define LDC_PACKET_SIZE		64
 
 /* Packet header layout for unreliable and reliable mode frames.
  * When in RAW mode, packets are simply straight 64-byte payloads
@@ -178,6 +177,8 @@ do {	if (lp->cfg.debug & LDC_DEBUG_##TYPE) \
 		printk(KERN_INFO PFX "ID[%lu] " f, lp->id, ## a); \
 } while (0)
 
+#define	LDC_ABORT(lp)	ldc_abort((lp), __func__)
+
 static const char *state_to_str(u8 state)
 {
 	switch (state) {
@@ -196,15 +197,6 @@ static const char *state_to_str(u8 state)
 	}
 }
 
-static void ldc_set_state(struct ldc_channel *lp, u8 state)
-{
-	ldcdbg(STATE, "STATE (%s) --> (%s)\n",
-	       state_to_str(lp->state),
-	       state_to_str(state));
-
-	lp->state = state;
-}
-
 static unsigned long __advance(unsigned long off, unsigned long num_entries)
 {
 	off += LDC_PACKET_SIZE;
@@ -516,11 +508,12 @@ static int send_data_nack(struct ldc_channel *lp, struct ldc_packet *data_pkt)
 	return err;
 }
 
-static int ldc_abort(struct ldc_channel *lp)
+static int ldc_abort(struct ldc_channel *lp, const char *msg)
 {
 	unsigned long hv_err;
 
-	ldcdbg(STATE, "ABORT\n");
+	ldcdbg(STATE, "ABORT[%s]\n", msg);
+	ldc_print(lp);
 
 	/* We report but do not act upon the hypervisor errors because
 	 * there really isn't much we can do if they fail at this point.
@@ -605,7 +598,7 @@ static int process_ver_info(struct ldc_channel *lp, struct ldc_version *vp)
 		}
 	}
 	if (err)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	return 0;
 }
@@ -618,13 +611,13 @@ static int process_ver_ack(struct ldc_channel *lp, struct ldc_version *vp)
 	if (lp->hs_state == LDC_HS_GOTVERS) {
 		if (lp->ver.major != vp->major ||
 		    lp->ver.minor != vp->minor)
-			return ldc_abort(lp);
+			return LDC_ABORT(lp);
 	} else {
 		lp->ver = *vp;
 		lp->hs_state = LDC_HS_GOTVERS;
 	}
 	if (send_rts(lp))
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 	return 0;
 }
 
@@ -635,17 +628,17 @@ static int process_ver_nack(struct ldc_channel *lp, struct ldc_version *vp)
 	unsigned long new_tail;
 
 	if (vp->major == 0 && vp->minor == 0)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	vap = find_by_major(vp->major);
 	if (!vap)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	p = handshake_compose_ctrl(lp, LDC_INFO, LDC_VERS,
 					   vap, sizeof(*vap),
 					   &new_tail);
 	if (!p)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	return send_tx_packet(lp, p, new_tail);
 }
@@ -668,7 +661,7 @@ static int process_version(struct ldc_channel *lp,
 		return process_ver_nack(lp, vp);
 
 	default:
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 	}
 }
 
@@ -681,13 +674,13 @@ static int process_rts(struct ldc_channel *lp,
 	if (p->stype     != LDC_INFO	   ||
 	    lp->hs_state != LDC_HS_GOTVERS ||
 	    p->env       != lp->cfg.mode)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	lp->snd_nxt = p->seqid;
 	lp->rcv_nxt = p->seqid;
 	lp->hs_state = LDC_HS_SENTRTR;
 	if (send_rtr(lp))
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	return 0;
 }
@@ -700,7 +693,7 @@ static int process_rtr(struct ldc_channel *lp,
 
 	if (p->stype     != LDC_INFO ||
 	    p->env       != lp->cfg.mode)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	lp->snd_nxt = p->seqid;
 	lp->hs_state = LDC_HS_COMPLETE;
@@ -723,7 +716,7 @@ static int process_rdx(struct ldc_channel *lp,
 
 	if (p->stype != LDC_INFO ||
 	    !(rx_seq_ok(lp, p->seqid)))
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	lp->rcv_nxt = p->seqid;
 
@@ -750,14 +743,14 @@ static int process_control_frame(struct ldc_channel *lp,
 		return process_rdx(lp, p);
 
 	default:
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 	}
 }
 
 static int process_error_frame(struct ldc_channel *lp,
 			       struct ldc_packet *p)
 {
-	return ldc_abort(lp);
+	return LDC_ABORT(lp);
 }
 
 static int process_data_ack(struct ldc_channel *lp,
@@ -776,7 +769,7 @@ static int process_data_ack(struct ldc_channel *lp,
 			return 0;
 		}
 		if (head == lp->tx_tail)
-			return ldc_abort(lp);
+			return LDC_ABORT(lp);
 	}
 
 	return 0;
@@ -820,16 +813,21 @@ static irqreturn_t ldc_rx(int irq, void *dev_id)
 		lp->hs_state = LDC_HS_COMPLETE;
 		ldc_set_state(lp, LDC_STATE_CONNECTED);
 
+		/*
+		 * Generate an LDC_EVENT_UP event if the channel
+		 * was not already up.
+		 */
+		if (orig_state != LDC_CHANNEL_UP) {
 			event_mask |= LDC_EVENT_UP;
-
 			orig_state = lp->chan_state;
 		}
+	}
 
 	/* If we are in reset state, flush the RX queue and ignore
 	 * everything.
 	 */
 	if (lp->flags & LDC_FLAG_RESET) {
-		(void) __set_rx_head(lp, lp->rx_tail);
+		(void) ldc_rx_reset(lp);
 		goto out;
 	}
 
@@ -880,7 +878,7 @@ static irqreturn_t ldc_rx(int irq, void *dev_id)
 			break;
 
 		default:
-			err = ldc_abort(lp);
+			err = LDC_ABORT(lp);
 			break;
 		}
 
@@ -895,7 +893,7 @@ static irqreturn_t ldc_rx(int irq, void *dev_id)
 
 		err = __set_rx_head(lp, new);
 		if (err < 0) {
-			(void) ldc_abort(lp);
+			(void) LDC_ABORT(lp);
 			break;
 		}
 		if (lp->hs_state == LDC_HS_COMPLETE)
@@ -936,7 +934,14 @@ static irqreturn_t ldc_tx(int irq, void *dev_id)
 		lp->hs_state = LDC_HS_COMPLETE;
 		ldc_set_state(lp, LDC_STATE_CONNECTED);
 
+		/*
+		 * Generate an LDC_EVENT_UP event if the channel
+		 * was not already up.
+		 */
+		if (orig_state != LDC_CHANNEL_UP) {
 			event_mask |= LDC_EVENT_UP;
+			orig_state = lp->chan_state;
+		}
 	}
 
 	spin_unlock_irqrestore(&lp->lock, flags);
@@ -1342,6 +1347,14 @@ int ldc_bind(struct ldc_channel *lp)
 	lp->hs_state = LDC_HS_OPEN;
 	ldc_set_state(lp, LDC_STATE_BOUND);
 
+	if (lp->cfg.mode == LDC_MODE_RAW) {
+		/*
+		 * There is no handshake in RAW mode, so handshake
+		 * is completed.
+		 */
+		lp->hs_state = LDC_HS_COMPLETE;
+	}
+
 	spin_unlock_irqrestore(&lp->lock, flags);
 
 	return 0;
@@ -1447,12 +1460,54 @@ int ldc_state(struct ldc_channel *lp)
 }
 EXPORT_SYMBOL(ldc_state);
 
+void ldc_set_state(struct ldc_channel *lp, u8 state)
+{
+	ldcdbg(STATE, "STATE (%s) --> (%s)\n",
+	       state_to_str(lp->state),
+	       state_to_str(state));
+
+	lp->state = state;
+}
+EXPORT_SYMBOL(ldc_set_state);
+
+int ldc_mode(struct ldc_channel *lp)
+{
+	return lp->cfg.mode;
+}
+EXPORT_SYMBOL(ldc_mode);
+
+int ldc_rx_reset(struct ldc_channel *lp)
+{
+	return __set_rx_head(lp, lp->rx_tail);
+}
+
+void __ldc_print(struct ldc_channel *lp, const char *caller)
+{
+	pr_info("%s: id=0x%lx flags=0x%x state=%s cstate=0x%lx hsstate=0x%x\n"
+		"\trx_h=0x%lx rx_t=0x%lx rx_n=%ld\n"
+		"\ttx_h=0x%lx tx_t=0x%lx tx_n=%ld\n"
+		"\trcv_nxt=%u snd_nxt=%u\n",
+		caller, lp->id, lp->flags, state_to_str(lp->state),
+		lp->chan_state, lp->hs_state,
+		lp->rx_head, lp->rx_tail, lp->rx_num_entries,
+		lp->tx_head, lp->tx_tail, lp->tx_num_entries,
+		lp->rcv_nxt, lp->snd_nxt);
+}
+
 static int write_raw(struct ldc_channel *lp, const void *buf, unsigned int size)
 {
 	struct ldc_packet *p;
-	unsigned long new_tail;
+	unsigned long new_tail, hv_err;
 	int err;
 
+	hv_err = sun4v_ldc_tx_get_state(lp->id, &lp->tx_head, &lp->tx_tail,
+					&lp->chan_state);
+	if (unlikely(hv_err))
+		return -EBUSY;
+
+	if (unlikely(lp->chan_state != LDC_CHANNEL_UP))
+		return LDC_ABORT(lp);
+
 	if (size > LDC_PACKET_SIZE)
 		return -EMSGSIZE;
 
@@ -1483,7 +1538,7 @@ static int read_raw(struct ldc_channel *lp, void *buf, unsigned int size)
 					&lp->rx_tail,
 					&lp->chan_state);
 	if (hv_err)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	if (lp->chan_state == LDC_CHANNEL_DOWN ||
 	    lp->chan_state == LDC_CHANNEL_RESETTING)
@@ -1526,7 +1581,7 @@ static int write_nonraw(struct ldc_channel *lp, const void *buf,
 		return -EBUSY;
 
 	if (unlikely(lp->chan_state != LDC_CHANNEL_UP))
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	if (!tx_has_space_for(lp, size))
 		return -EAGAIN;
@@ -1592,9 +1647,9 @@ static int rx_bad_seq(struct ldc_channel *lp, struct ldc_packet *p,
 	if (err)
 		return err;
 
-	err = __set_rx_head(lp, lp->rx_tail);
+	err = ldc_rx_reset(lp);
 	if (err < 0)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	return 0;
 }
@@ -1607,7 +1662,7 @@ static int data_ack_nack(struct ldc_channel *lp, struct ldc_packet *p)
 			return err;
 	}
 	if (p->stype & LDC_NACK)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	return 0;
 }
@@ -1627,7 +1682,7 @@ static int rx_data_wait(struct ldc_channel *lp, unsigned long cur_head)
 						&lp->rx_tail,
 						&lp->chan_state);
 		if (hv_err)
-			return ldc_abort(lp);
+			return LDC_ABORT(lp);
 
 		if (lp->chan_state == LDC_CHANNEL_DOWN ||
 		    lp->chan_state == LDC_CHANNEL_RESETTING)
@@ -1650,7 +1705,7 @@ static int rx_set_head(struct ldc_channel *lp, unsigned long head)
 	int err = __set_rx_head(lp, head);
 
 	if (err < 0)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	lp->rx_head = head;
 	return 0;
@@ -1689,7 +1744,7 @@ static int read_nonraw(struct ldc_channel *lp, void *buf, unsigned int size)
 					&lp->rx_tail,
 					&lp->chan_state);
 	if (hv_err)
-		return ldc_abort(lp);
+		return LDC_ABORT(lp);
 
 	if (lp->chan_state == LDC_CHANNEL_DOWN ||
 	    lp->chan_state == LDC_CHANNEL_RESETTING)
@@ -1733,9 +1788,14 @@ static int read_nonraw(struct ldc_channel *lp, void *buf, unsigned int size)
 
 		lp->rcv_nxt = p->seqid;
 
+		/*
+		 * If this is a control-only packet, there is nothing
+		 * else to do but advance the rx queue since the packet
+		 * was already processed above.
+		 */
 		if (!(p->type & LDC_DATA)) {
 			new = rx_advance(lp, new);
-			goto no_data;
+			break;
 		}
 		if (p->stype & (LDC_ACK | LDC_NACK)) {
 			err = data_ack_nack(lp, p);
@@ -1900,6 +1960,8 @@ int ldc_read(struct ldc_channel *lp, void *buf, unsigned int size)
 	unsigned long flags;
 	int err;
 
+	ldcdbg(RX, "%s: entered size=%d\n", __func__, size);
+
 	if (!buf)
 		return -EINVAL;
 
@@ -1915,6 +1977,9 @@ int ldc_read(struct ldc_channel *lp, void *buf, unsigned int size)
 
 	spin_unlock_irqrestore(&lp->lock, flags);
 
+	ldcdbg(RX, "%s: mode=%d, head=%lu, tail=%lu rv=%d\n", __func__,
+	       lp->cfg.mode, lp->rx_head, lp->rx_tail, err);
+
 	return err;
 }
 EXPORT_SYMBOL(ldc_read);

arch/sparc/kernel/pmc.c

@@ -71,7 +71,7 @@ static int pmc_probe(struct platform_device *op)
 	return 0;
 }
 
-static struct of_device_id pmc_match[] = {
+static const struct of_device_id pmc_match[] = {
 	{
 		.name = PMC_OBPNAME,
 	},

arch/sparc/kernel/prom_64.c

@@ -381,7 +381,7 @@ bool arch_find_n_match_cpu_physical_id(struct device_node *cpun,
 	int this_cpu_id;
 
 	/* On hypervisor based platforms we interrogate the 'reg'
-	 * property.  On everything else we look for a 'upa-portis',
+	 * property.  On everything else we look for a 'upa-portid',
 	 * 'portid', or 'cpuid' property.
 	 */
 

arch/sparc/kernel/setup_64.c

@@ -365,6 +365,7 @@ void __init start_early_boot(void)
 	}
 	current_thread_info()->cpu = cpu;
 
+	time_init_early();
 	prom_init_report();
 	start_kernel();
 }

arch/sparc/kernel/time_32.c

@@ -298,7 +298,7 @@ static int clock_probe(struct platform_device *op)
 	return 0;
 }
 
-static struct of_device_id clock_match[] = {
+static const struct of_device_id clock_match[] = {
 	{
 		.name = "eeprom",
 	},

arch/sparc/kernel/time_64.c

@@ -32,7 +32,6 @@
 #include <linux/kernel_stat.h>
 #include <linux/clockchips.h>
 #include <linux/clocksource.h>
-#include <linux/of_device.h>
 #include <linux/platform_device.h>
 #include <linux/ftrace.h>
 
@@ -47,14 +46,13 @@
 #include <asm/cpudata.h>
 #include <linux/uaccess.h>
 #include <asm/irq_regs.h>
+#include <asm/cacheflush.h>
 
 #include "entry.h"
+#include "kernel.h"
 
 DEFINE_SPINLOCK(rtc_lock);
 
-#define TICK_PRIV_BIT	(1UL << 63)
-#define TICKCMP_IRQ_BIT	(1UL << 63)
-
 #ifdef CONFIG_SMP
 unsigned long profile_pc(struct pt_regs *regs)
 {
@@ -164,13 +162,44 @@ static unsigned long tick_add_tick(unsigned long adj)
 	return new_tick;
 }
 
-static struct sparc64_tick_ops tick_operations __read_mostly = {
+/* Searches for cpu clock frequency with given cpuid in OpenBoot tree */
+static unsigned long cpuid_to_freq(phandle node, int cpuid)
+{
+	bool is_cpu_node = false;
+	unsigned long freq = 0;
+	char type[128];
+
+	if (!node)
+		return freq;
+
+	if (prom_getproperty(node, "device_type", type, sizeof(type)) != -1)
+		is_cpu_node = (strcmp(type, "cpu") == 0);
+
+	/* try upa-portid then cpuid to get cpuid, see prom_64.c */
+	if (is_cpu_node && (prom_getint(node, "upa-portid") == cpuid ||
+			    prom_getint(node, "cpuid") == cpuid))
+		freq = prom_getintdefault(node, "clock-frequency", 0);
+	if (!freq)
+		freq = cpuid_to_freq(prom_getchild(node), cpuid);
+	if (!freq)
+		freq = cpuid_to_freq(prom_getsibling(node), cpuid);
+
+	return freq;
+}
+
+static unsigned long tick_get_frequency(void)
+{
+	return cpuid_to_freq(prom_root_node, hard_smp_processor_id());
+}
+
+static struct sparc64_tick_ops tick_operations __cacheline_aligned = {
 	.name		=	"tick",
 	.init_tick	=	tick_init_tick,
 	.disable_irq	=	tick_disable_irq,
 	.get_tick	=	tick_get_tick,
 	.add_tick	=	tick_add_tick,
 	.add_compare	=	tick_add_compare,
+	.get_frequency	=	tick_get_frequency,
 	.softint_mask	=	1UL << 0,
 };
 
@@ -250,6 +279,11 @@ static int stick_add_compare(unsigned long adj)
 	return ((long)(new_tick - (orig_tick+adj))) > 0L;
 }
 
+static unsigned long stick_get_frequency(void)
+{
+	return prom_getintdefault(prom_root_node, "stick-frequency", 0);
+}
+
 static struct sparc64_tick_ops stick_operations __read_mostly = {
 	.name		=	"stick",
 	.init_tick	=	stick_init_tick,
@@ -257,6 +291,7 @@ static struct sparc64_tick_ops stick_operations __read_mostly = {
 	.get_tick	=	stick_get_tick,
 	.add_tick	=	stick_add_tick,
 	.add_compare	=	stick_add_compare,
+	.get_frequency	=	stick_get_frequency,
 	.softint_mask	=	1UL << 16,
 };
 
@@ -277,9 +312,6 @@ static struct sparc64_tick_ops stick_operations __read_mostly = {
  * 2) write high
  * 3) write low
  */
-#define HBIRD_STICKCMP_ADDR	0x1fe0000f060UL
-#define HBIRD_STICK_ADDR	0x1fe0000f070UL
-
 static unsigned long __hbird_read_stick(void)
 {
 	unsigned long ret, tmp1, tmp2, tmp3;
@@ -381,6 +413,11 @@ static int hbtick_add_compare(unsigned long adj)
 	return ((long)(val2 - val)) > 0L;
 }
 
+static unsigned long hbtick_get_frequency(void)
+{
+	return prom_getintdefault(prom_root_node, "stick-frequency", 0);
+}
+
 static struct sparc64_tick_ops hbtick_operations __read_mostly = {
 	.name		=	"hbtick",
 	.init_tick	=	hbtick_init_tick,
@@ -388,11 +425,10 @@ static struct sparc64_tick_ops hbtick_operations __read_mostly = {
 	.get_tick	=	hbtick_get_tick,
 	.add_tick	=	hbtick_add_tick,
 	.add_compare	=	hbtick_add_compare,
+	.get_frequency	=	hbtick_get_frequency,
 	.softint_mask	=	1UL << 0,
 };
 
-static unsigned long timer_ticks_per_nsec_quotient __read_mostly;
-
 unsigned long cmos_regs;
 EXPORT_SYMBOL(cmos_regs);
 
@@ -582,34 +618,17 @@ static int __init clock_init(void)
  */
 fs_initcall(clock_init);
 
-/* This is gets the master TICK_INT timer going. */
-static unsigned long sparc64_init_timers(void)
+/* Return true if this is Hummingbird, aka Ultra-IIe */
+static bool is_hummingbird(void)
 {
-	struct device_node *dp;
-	unsigned long freq;
-
-	dp = of_find_node_by_path("/");
-	if (tlb_type == spitfire) {
 	unsigned long ver, manuf, impl;
 
 	__asm__ __volatile__ ("rdpr %%ver, %0"
 			      : "=&r" (ver));
 	manuf = ((ver >> 48) & 0xffff);
 	impl = ((ver >> 32) & 0xffff);
-		if (manuf == 0x17 && impl == 0x13) {
-			/* Hummingbird, aka Ultra-IIe */
-			tick_ops = &hbtick_operations;
-			freq = of_getintprop_default(dp, "stick-frequency", 0);
-		} else {
-			tick_ops = &tick_operations;
-			freq = local_cpu_data().clock_tick;
-		}
-	} else {
-		tick_ops = &stick_operations;
-		freq = of_getintprop_default(dp, "stick-frequency", 0);
-	}
 
-	return freq;
+	return (manuf == 0x17 && impl == 0x13);
 }
 
 struct freq_table {
@@ -671,12 +690,12 @@ core_initcall(register_sparc64_cpufreq_notifier);
 static int sparc64_next_event(unsigned long delta,
 			      struct clock_event_device *evt)
 {
-	return tick_ops->add_compare(delta) ? -ETIME : 0;
+	return tick_operations.add_compare(delta) ? -ETIME : 0;
 }
 
 static int sparc64_timer_shutdown(struct clock_event_device *evt)
 {
-	tick_ops->disable_irq();
+	tick_operations.disable_irq();
 	return 0;
 }
 
@@ -693,7 +712,7 @@ static DEFINE_PER_CPU(struct clock_event_device, sparc64_events);
 void __irq_entry timer_interrupt(int irq, struct pt_regs *regs)
 {
 	struct pt_regs *old_regs = set_irq_regs(regs);
-	unsigned long tick_mask = tick_ops->softint_mask;
+	unsigned long tick_mask = tick_operations.softint_mask;
 	int cpu = smp_processor_id();
 	struct clock_event_device *evt = &per_cpu(sparc64_events, cpu);
 
@@ -728,7 +747,7 @@ void setup_sparc64_timer(void)
 			     : "=r" (pstate)
 			     : "i" (PSTATE_IE));
 
-	tick_ops->init_tick();
+	tick_operations.init_tick();
 
 	/* Restore PSTATE_IE. */
 	__asm__ __volatile__("wrpr	%0, 0x0, %%pstate"
@@ -755,12 +774,10 @@ static unsigned long tb_ticks_per_usec __read_mostly;
 
 void __delay(unsigned long loops)
 {
-	unsigned long bclock, now;
+	unsigned long bclock = get_tick();
 
-	bclock = tick_ops->get_tick();
-	do {
-		now = tick_ops->get_tick();
-	} while ((now-bclock) < loops);
+	while ((get_tick() - bclock) < loops)
+		;
 }
 EXPORT_SYMBOL(__delay);
 
@@ -772,26 +789,71 @@ EXPORT_SYMBOL(udelay);
 
 static u64 clocksource_tick_read(struct clocksource *cs)
 {
-	return tick_ops->get_tick();
+	return get_tick();
+}
+
+static void __init get_tick_patch(void)
+{
+	unsigned int *addr, *instr, i;
+	struct get_tick_patch *p;
+
+	if (tlb_type == spitfire && is_hummingbird())
+		return;
+
+	for (p = &__get_tick_patch; p < &__get_tick_patch_end; p++) {
+		instr = (tlb_type == spitfire) ? p->tick : p->stick;
+		addr = (unsigned int *)(unsigned long)p->addr;
+		for (i = 0; i < GET_TICK_NINSTR; i++) {
+			addr[i] = instr[i];
+			/* ensure that address is modified before flush */
+			wmb();
+			flushi(&addr[i]);
+		}
+	}
+}
+
+static void init_tick_ops(struct sparc64_tick_ops *ops)
+{
+	unsigned long freq, quotient, tick;
+
+	freq = ops->get_frequency();
+	quotient = clocksource_hz2mult(freq, SPARC64_NSEC_PER_CYC_SHIFT);
+	tick = ops->get_tick();
+
+	ops->offset = (tick * quotient) >> SPARC64_NSEC_PER_CYC_SHIFT;
+	ops->ticks_per_nsec_quotient = quotient;
+	ops->frequency = freq;
+	tick_operations = *ops;
+	get_tick_patch();
+}
+
+void __init time_init_early(void)
+{
+	if (tlb_type == spitfire) {
+		if (is_hummingbird())
+			init_tick_ops(&hbtick_operations);
+		else
+			init_tick_ops(&tick_operations);
+	} else {
+		init_tick_ops(&stick_operations);
+	}
 }
 
 void __init time_init(void)
 {
-	unsigned long freq = sparc64_init_timers();
+	unsigned long freq;
 
+	freq = tick_operations.frequency;
 	tb_ticks_per_usec = freq / USEC_PER_SEC;
 
-	timer_ticks_per_nsec_quotient =
-		clocksource_hz2mult(freq, SPARC64_NSEC_PER_CYC_SHIFT);
-
-	clocksource_tick.name = tick_ops->name;
+	clocksource_tick.name = tick_operations.name;
 	clocksource_tick.read = clocksource_tick_read;
 
 	clocksource_register_hz(&clocksource_tick, freq);
 	printk("clocksource: mult[%x] shift[%d]\n",
 	       clocksource_tick.mult, clocksource_tick.shift);
 
-	sparc64_clockevent.name = tick_ops->name;
+	sparc64_clockevent.name = tick_operations.name;
 	clockevents_calc_mult_shift(&sparc64_clockevent, freq, 4);
 
 	sparc64_clockevent.max_delta_ns =
@@ -809,14 +871,21 @@ void __init time_init(void)
 
 unsigned long long sched_clock(void)
 {
-	unsigned long ticks = tick_ops->get_tick();
+	unsigned long quotient = tick_operations.ticks_per_nsec_quotient;
+	unsigned long offset = tick_operations.offset;
+
+	/* Use barrier so the compiler emits the loads first and overlaps load
+	 * latency with reading tick, because reading %tick/%stick is a
+	 * post-sync instruction that will flush and restart subsequent
+	 * instructions after it commits.
+	 */
+	barrier();
 
-	return (ticks * timer_ticks_per_nsec_quotient)
-		>> SPARC64_NSEC_PER_CYC_SHIFT;
+	return ((get_tick() * quotient) >> SPARC64_NSEC_PER_CYC_SHIFT) - offset;
 }
 
 int read_current_timer(unsigned long *timer_val)
 {
-	*timer_val = tick_ops->get_tick();
+	*timer_val = get_tick();
 	return 0;
 }

arch/sparc/kernel/viohs.c

@@ -223,6 +223,9 @@ static int send_rdx(struct vio_driver_state *vio)
 
 static int send_attr(struct vio_driver_state *vio)
 {
+	if (!vio->ops)
+		return -EINVAL;
+
 	return vio->ops->send_attr(vio);
 }
 
@@ -283,6 +286,7 @@ static int process_ver_info(struct vio_driver_state *vio,
 			ver.minor = vap->minor;
 		pkt->minor = ver.minor;
 		pkt->tag.stype = VIO_SUBTYPE_ACK;
+		pkt->dev_class = vio->dev_class;
 		viodbg(HS, "SEND VERSION ACK maj[%u] min[%u]\n",
 		       pkt->major, pkt->minor);
 		err = send_ctrl(vio, &pkt->tag, sizeof(*pkt));
@@ -374,6 +378,9 @@ static int process_attr(struct vio_driver_state *vio, void *pkt)
 	if (!(vio->hs_state & VIO_HS_GOTVERS))
 		return handshake_failure(vio);
 
+	if (!vio->ops)
+		return 0;
+
 	err = vio->ops->handle_attr(vio, pkt);
 	if (err < 0) {
 		return handshake_failure(vio);
@@ -388,6 +395,7 @@ static int process_attr(struct vio_driver_state *vio, void *pkt)
 			vio->hs_state |= VIO_HS_SENT_DREG;
 		}
 	}
+
 	return 0;
 }
 
@@ -647,10 +655,13 @@ int vio_control_pkt_engine(struct vio_driver_state *vio, void *pkt)
 		err = process_unknown(vio, pkt);
 		break;
 	}
+
 	if (!err &&
 	    vio->hs_state != prev_state &&
-	    (vio->hs_state & VIO_HS_COMPLETE))
+	    (vio->hs_state & VIO_HS_COMPLETE)) {
+		if (vio->ops)
 			vio->ops->handshake_complete(vio);
+	}
 
 	return err;
 }
@@ -765,7 +776,11 @@ void vio_port_up(struct vio_driver_state *vio)
 	}
 
 	if (!err) {
+		if (ldc_mode(vio->lp) == LDC_MODE_RAW)
+			ldc_set_state(vio->lp, LDC_STATE_CONNECTED);
+		else
 			err = ldc_connect(vio->lp);
+
 		if (err)
 			printk(KERN_WARNING "%s: Port %lu connect failed, "
 			       "err=%d\n",
@@ -805,8 +820,7 @@ int vio_driver_init(struct vio_driver_state *vio, struct vio_dev *vdev,
 		return -EINVAL;
 	}
 
-	if (!ops->send_attr ||
-	    !ops->handle_attr ||
+	if (!ops || !ops->send_attr || !ops->handle_attr ||
 	    !ops->handshake_complete)
 		return -EINVAL;
 

arch/sparc/kernel/vmlinux.lds.S

@@ -149,6 +149,11 @@ SECTIONS
 		*(.sun_m7_2insn_patch)
 		__sun_m7_2insn_patch_end = .;
 	}
+	.get_tick_patch : {
+		__get_tick_patch = .;
+		*(.get_tick_patch)
+		__get_tick_patch_end = .;
+	}
 	PERCPU_SECTION(SMP_CACHE_BYTES)
 
 #ifdef CONFIG_JUMP_LABEL

kernel/locking/qrwlock.c

@@ -20,6 +20,7 @@
 #include <linux/cpumask.h>
 #include <linux/percpu.h>
 #include <linux/hardirq.h>
+#include <linux/spinlock.h>
 #include <asm/qrwlock.h>
 
 /*