Merge git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile

Pull arch/tile updates from Chris Metcalf:
 "These are a grab bag of changes to improve debugging and respond to a
  variety of issues raised on LKML over the last couple of months"

* git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  tile: avoid a "label not used" warning in do_page_fault()
  tile: vdso: use raw_read_seqcount_begin() in vdso
  tile: force CONFIG_TILEGX if ARCH != tilepro
  tile: improve stack backtrace
  tile: fix "odd fault" warning for stack backtraces
  tile: set up initial stack top to honor STACK_TOP_DELTA
  tile: support delivering NMIs for multicore backtrace
  drivers/tty/hvc/hvc_tile.c: properly return -EAGAIN
  tile: add <asm/word-at-a-time.h> and enable support functions
  tile: use READ_ONCE() in arch_spin_is_locked()
  tile: modify arch_spin_unlock_wait() semantics

10 files changed, 133 insertions, 69 deletions

diff --git a/arch/tile/include/asm/irq.h b/arch/tile/include/asm/irq.h
index 1fe86911838b..84a924034bdb 100644
--- a/arch/tile/include/asm/irq.h
+++ b/arch/tile/include/asm/irq.h
@@ -78,4 +78,9 @@ void tile_irq_activate(unsigned int irq, int tile_irq_type);
 
 void setup_irq_regs(void);
 
+#ifdef __tilegx__
+void arch_trigger_all_cpu_backtrace(bool self);
+#define arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace
+#endif
+
 #endif /* _ASM_TILE_IRQ_H */
diff --git a/arch/tile/include/asm/processor.h b/arch/tile/include/asm/processor.h
index dd4f9f17e30a..139dfdee0134 100644
--- a/arch/tile/include/asm/processor.h
+++ b/arch/tile/include/asm/processor.h
@@ -111,8 +111,6 @@ struct thread_struct {
 	unsigned long long interrupt_mask;
 	/* User interrupt-control 0 state */
 	unsigned long intctrl_0;
-	/* Is this task currently doing a backtrace? */
-	bool in_backtrace;
 	/* Any other miscellaneous processor state bits */
 	unsigned long proc_status;
 #if !CHIP_HAS_FIXED_INTVEC_BASE()
diff --git a/arch/tile/include/asm/spinlock_32.h b/arch/tile/include/asm/spinlock_32.h
index c0a77b38d39a..b14b1ba5bf9c 100644
--- a/arch/tile/include/asm/spinlock_32.h
+++ b/arch/tile/include/asm/spinlock_32.h
@@ -41,8 +41,12 @@ static inline int arch_spin_is_locked(arch_spinlock_t *lock)
 	 * to claim the lock is held, since it will be momentarily
 	 * if not already.  There's no need to wait for a "valid"
 	 * lock->next_ticket to become available.
+	 * Use READ_ONCE() to ensure that calling this in a loop is OK.
 	 */
-	return lock->next_ticket != lock->current_ticket;
+	int curr = READ_ONCE(lock->current_ticket);
+	int next = READ_ONCE(lock->next_ticket);
+
+	return next != curr;
 }
 
 void arch_spin_lock(arch_spinlock_t *lock);
diff --git a/arch/tile/include/asm/spinlock_64.h b/arch/tile/include/asm/spinlock_64.h
index 9a12b9c7e5d3..b9718fb4e74a 100644
--- a/arch/tile/include/asm/spinlock_64.h
+++ b/arch/tile/include/asm/spinlock_64.h
@@ -18,6 +18,8 @@
 #ifndef _ASM_TILE_SPINLOCK_64_H
 #define _ASM_TILE_SPINLOCK_64_H
 
+#include <linux/compiler.h>
+
 /* Shifts and masks for the various fields in "lock". */
 #define __ARCH_SPIN_CURRENT_SHIFT	17
 #define __ARCH_SPIN_NEXT_MASK		0x7fff
@@ -44,7 +46,8 @@ static inline u32 arch_spin_next(u32 val)
 /* The lock is locked if a task would have to wait to get it. */
 static inline int arch_spin_is_locked(arch_spinlock_t *lock)
 {
-	u32 val = lock->lock;
+	/* Use READ_ONCE() to ensure that calling this in a loop is OK. */
+	u32 val = READ_ONCE(lock->lock);
 	return arch_spin_current(val) != arch_spin_next(val);
 }
 
diff --git a/arch/tile/include/asm/stack.h b/arch/tile/include/asm/stack.h
index 0e9d382a2d45..c3cb42615a9f 100644
--- a/arch/tile/include/asm/stack.h
+++ b/arch/tile/include/asm/stack.h
@@ -58,17 +58,14 @@ extern int KBacktraceIterator_end(struct KBacktraceIterator *kbt);
 /* Advance to the next frame. */
 extern void KBacktraceIterator_next(struct KBacktraceIterator *kbt);
 
+/* Dump just the contents of the pt_regs structure. */
+extern void tile_show_regs(struct pt_regs *);
+
 /*
  * Dump stack given complete register info. Use only from the
  * architecture-specific code; show_stack()
- * and dump_stack() (in entry.S) are architecture-independent entry points.
+ * and dump_stack() are architecture-independent entry points.
  */
-extern void tile_show_stack(struct KBacktraceIterator *, int headers);
-
-/* Dump stack of current process, with registers to seed the backtrace. */
-extern void dump_stack_regs(struct pt_regs *);
-
-/* Helper method for assembly dump_stack(). */
-extern void _dump_stack(int dummy, ulong pc, ulong lr, ulong sp, ulong r52);
+extern void tile_show_stack(struct KBacktraceIterator *);
 
 #endif /* _ASM_TILE_STACK_H */
diff --git a/arch/tile/include/asm/thread_info.h b/arch/tile/include/asm/thread_info.h
index f804c39a5e4d..dc1fb28d9636 100644
--- a/arch/tile/include/asm/thread_info.h
+++ b/arch/tile/include/asm/thread_info.h
@@ -42,6 +42,7 @@ struct thread_info {
 	unsigned long		unalign_jit_tmp[4]; /* temp r0..r3 storage */
 	void __user		*unalign_jit_base; /* unalign fixup JIT base */
 #endif
+	bool in_backtrace;			/* currently doing backtrace? */
 };
 
 /*
diff --git a/arch/tile/include/asm/traps.h b/arch/tile/include/asm/traps.h
index 4b99a1c3aab2..11c82270c1f5 100644
--- a/arch/tile/include/asm/traps.h
+++ b/arch/tile/include/asm/traps.h
@@ -52,6 +52,14 @@ void do_timer_interrupt(struct pt_regs *, int fault_num);
 /* kernel/messaging.c */
 void hv_message_intr(struct pt_regs *, int intnum);
 
+#define	TILE_NMI_DUMP_STACK	1	/* Dump stack for sysrq+'l' */
+
+/* kernel/process.c */
+void do_nmi_dump_stack(struct pt_regs *regs);
+
+/* kernel/traps.c */
+void do_nmi(struct pt_regs *, int fault_num, unsigned long reason);
+
 /* kernel/irq.c */
 void tile_dev_intr(struct pt_regs *, int intnum);
 
diff --git a/arch/tile/include/asm/uaccess.h b/arch/tile/include/asm/uaccess.h
index a33276bf5ca1..0a9c4265763b 100644
--- a/arch/tile/include/asm/uaccess.h
+++ b/arch/tile/include/asm/uaccess.h
@@ -65,6 +65,13 @@ static inline int is_arch_mappable_range(unsigned long addr,
 #endif
 
 /*
+ * Note that using this definition ignores is_arch_mappable_range(),
+ * so on tilepro code that uses user_addr_max() is constrained not
+ * to reference the tilepro user-interrupt region.
+ */
+#define user_addr_max() (current_thread_info()->addr_limit.seg)
+
+/*
  * Test whether a block of memory is a valid user space address.
  * Returns 0 if the range is valid, nonzero otherwise.
  */
@@ -471,62 +478,9 @@ copy_in_user(void __user *to, const void __user *from, unsigned long n)
 #endif
 
 
-/**
- * strlen_user: - Get the size of a string in user space.
- * @str: The string to measure.
- *
- * Context: User context only.  This function may sleep.
- *
- * Get the size of a NUL-terminated string in user space.
- *
- * Returns the size of the string INCLUDING the terminating NUL.
- * On exception, returns 0.
- *
- * If there is a limit on the length of a valid string, you may wish to
- * consider using strnlen_user() instead.
- */
-extern long strnlen_user_asm(const char __user *str, long n);
-static inline long __must_check strnlen_user(const char __user *str, long n)
-{
-	might_fault();
-	return strnlen_user_asm(str, n);
-}
-#define strlen_user(str) strnlen_user(str, LONG_MAX)
-
-/**
- * strncpy_from_user: - Copy a NUL terminated string from userspace, with less checking.
- * @dst:   Destination address, in kernel space.  This buffer must be at
- *         least @count bytes long.
- * @src:   Source address, in user space.
- * @count: Maximum number of bytes to copy, including the trailing NUL.
- *
- * Copies a NUL-terminated string from userspace to kernel space.
- * Caller must check the specified block with access_ok() before calling
- * this function.
- *
- * On success, returns the length of the string (not including the trailing
- * NUL).
- *
- * If access to userspace fails, returns -EFAULT (some data may have been
- * copied).
- *
- * If @count is smaller than the length of the string, copies @count bytes
- * and returns @count.
- */
-extern long strncpy_from_user_asm(char *dst, const char __user *src, long);
-static inline long __must_check __strncpy_from_user(
-	char *dst, const char __user *src, long count)
-{
-	might_fault();
-	return strncpy_from_user_asm(dst, src, count);
-}
-static inline long __must_check strncpy_from_user(
-	char *dst, const char __user *src, long count)
-{
-	if (access_ok(VERIFY_READ, src, 1))
-		return __strncpy_from_user(dst, src, count);
-	return -EFAULT;
-}
+extern long strnlen_user(const char __user *str, long n);
+extern long strlen_user(const char __user *str);
+extern long strncpy_from_user(char *dst, const char __user *src, long);
 
 /**
  * clear_user: - Zero a block of memory in user space.
diff --git a/arch/tile/include/asm/word-at-a-time.h b/arch/tile/include/asm/word-at-a-time.h
new file mode 100644
index 000000000000..9e5ce0d7b292
--- /dev/null
+++ b/arch/tile/include/asm/word-at-a-time.h
@@ -0,0 +1,36 @@
+#ifndef _ASM_WORD_AT_A_TIME_H
+#define _ASM_WORD_AT_A_TIME_H
+
+#include <asm/byteorder.h>
+
+struct word_at_a_time { /* unused */ };
+#define WORD_AT_A_TIME_CONSTANTS {}
+
+/* Generate 0x01 byte values for non-zero bytes using a SIMD instruction. */
+static inline unsigned long has_zero(unsigned long val, unsigned long *data,
+				     const struct word_at_a_time *c)
+{
+#ifdef __tilegx__
+	unsigned long mask = __insn_v1cmpeqi(val, 0);
+#else /* tilepro */
+	unsigned long mask = __insn_seqib(val, 0);
+#endif
+	*data = mask;
+	return mask;
+}
+
+/* These operations are both nops. */
+#define prep_zero_mask(val, data, c) (data)
+#define create_zero_mask(data) (data)
+
+/* And this operation just depends on endianness. */
+static inline long find_zero(unsigned long mask)
+{
+#ifdef __BIG_ENDIAN
+	return __builtin_clzl(mask) >> 3;
+#else
+	return __builtin_ctzl(mask) >> 3;
+#endif
+}
+
+#endif /* _ASM_WORD_AT_A_TIME_H */
diff --git a/arch/tile/include/hv/hypervisor.h b/arch/tile/include/hv/hypervisor.h
index e0e6af4e783b..f10b332b3b65 100644
--- a/arch/tile/include/hv/hypervisor.h
+++ b/arch/tile/include/hv/hypervisor.h
@@ -321,8 +321,11 @@
 /** hv_console_set_ipi */
 #define HV_DISPATCH_CONSOLE_SET_IPI               63
 
+/** hv_send_nmi */
+#define HV_DISPATCH_SEND_NMI                      65
+
 /** One more than the largest dispatch value */
-#define _HV_DISPATCH_END                          64
+#define _HV_DISPATCH_END                          66
 
 
 #ifndef __ASSEMBLER__
@@ -1253,6 +1256,11 @@ void hv_downcall_dispatch(void);
 #define INT_DMATLB_ACCESS_DWNCL  INT_DMA_CPL
 /** Device interrupt downcall interrupt vector */
 #define INT_DEV_INTR_DWNCL       INT_WORLD_ACCESS
+/** NMI downcall interrupt vector */
+#define INT_NMI_DWNCL            64
+
+#define HV_NMI_FLAG_FORCE    0x1  /**< Force an NMI downcall regardless of
+               the ICS bit of the client. */
 
 #ifndef __ASSEMBLER__
 
@@ -1780,6 +1788,56 @@ int hv_dev_poll(int devhdl, __hv32 events, HV_IntArg intarg);
 int hv_dev_poll_cancel(int devhdl);
 
 
+/** NMI information */
+typedef struct
+{
+  /** Result: negative error, or HV_NMI_RESULT_xxx. */
+  int result;
+
+  /** PC from interrupted remote core (if result != HV_NMI_RESULT_FAIL_HV). */
+  HV_VirtAddr pc;
+
+} HV_NMI_Info;
+
+/** NMI issued successfully. */
+#define HV_NMI_RESULT_OK        0
+
+/** NMI not issued: remote tile running at client PL with ICS set. */
+#define HV_NMI_RESULT_FAIL_ICS  1
+
+/** NMI not issued: remote tile waiting in hypervisor. */
+#define HV_NMI_RESULT_FAIL_HV   2
+
+/** Force an NMI downcall regardless of the ICS bit of the client. */
+#define HV_NMI_FLAG_FORCE    0x1
+
+/** Send an NMI interrupt request to a particular tile.
+ *
+ *  This will cause the NMI to be issued on the remote tile regardless
+ *  of the state of the client interrupt mask.  However, if the remote
+ *  tile is in the hypervisor, it will not execute the NMI, and
+ *  HV_NMI_RESULT_FAIL_HV will be returned.  Similarly, if the remote
+ *  tile is in a client interrupt critical section at the time of the
+ *  NMI, it will not execute the NMI, and HV_NMI_RESULT_FAIL_ICS will
+ *  be returned.  In this second case, however, if HV_NMI_FLAG_FORCE
+ *  is set in flags, then the remote tile will enter its NMI interrupt
+ *  vector regardless.  Forcing the NMI vector during an interrupt
+ *  critical section will mean that the client can not safely continue
+ *  execution after handling the interrupt.
+ *
+ *  @param tile Tile to which the NMI request is sent.
+ *  @param info NMI information which is defined by and interpreted by the
+ *         supervisor, is passed to the specified tile, and is
+ *         stored in the SPR register SYSTEM_SAVE_{CLIENT_PL}_2 on the
+ *         specified tile when entering the NMI handler routine.
+ *         Typically, this parameter stores the NMI type, or an aligned
+ *         VA plus some special bits, etc.
+ *  @param flags Flags (HV_NMI_FLAG_xxx).
+ *  @return Information about the requested NMI.
+ */
+HV_NMI_Info hv_send_nmi(HV_Coord tile, unsigned long info, __hv64 flags);
+
+
 /** Scatter-gather list for preada/pwritea calls. */
 typedef struct
 #if CHIP_VA_WIDTH() <= 32