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

* git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  arch/tile: convert a BUG_ON to BUILD_BUG_ON
  arch/tile: make ptrace() work properly for TILE-Gx COMPAT mode
  arch/tile: support new info op generated by compiler
  arch/tile: minor whitespace/naming changes for string support files
  arch/tile: enable single-step support for TILE-Gx
  arch/tile: parameterize system PLs to support KVM port
  arch/tile: add Tilera's <arch/sim.h> header as an open-source header
  arch/tile: Bomb C99 comments to C89 comments in tile's <arch/sim_def.h>
  arch/tile: prevent corrupt top frame from causing backtracer runaway
  arch/tile: various top-level Makefile cleanups
  arch/tile: change lower bound on syscall error return to -4095
  arch/tile: properly export __mb_incoherent for modules
  arch/tile: provide a definition of MAP_STACK
  kmemleak: add TILE to the list of supported architectures.
  char: hvc: check for error case
  arch/tile: Add a warning if we try to allocate too much vmalloc memory.
  arch/tile: update some comments to clarify register usage.
  arch/tile: use better "punctuation" for VMSPLIT_3_5G and friends
  arch/tile: Use <asm-generic/syscalls.h>
  tile: replace some BUG_ON checks with BUILD_BUG_ON checks

arch/tile/Kconfig

@@ -96,6 +96,7 @@ config HVC_TILE
 
 config TILE
 	def_bool y
+	select HAVE_KVM if !TILEGX
 	select GENERIC_FIND_FIRST_BIT
 	select GENERIC_FIND_NEXT_BIT
 	select USE_GENERIC_SMP_HELPERS
@@ -236,9 +237,9 @@ choice
 	  If you are not absolutely sure what you are doing, leave this
 	  option alone!
 
-	config VMSPLIT_375G
+	config VMSPLIT_3_75G
 		bool "3.75G/0.25G user/kernel split (no kernel networking)"
-	config VMSPLIT_35G
+	config VMSPLIT_3_5G
 		bool "3.5G/0.5G user/kernel split"
 	config VMSPLIT_3G
 		bool "3G/1G user/kernel split"
@@ -252,8 +253,8 @@ endchoice
 
 config PAGE_OFFSET
 	hex
-	default 0xF0000000 if VMSPLIT_375G
-	default 0xE0000000 if VMSPLIT_35G
+	default 0xF0000000 if VMSPLIT_3_75G
+	default 0xE0000000 if VMSPLIT_3_5G
 	default 0xB0000000 if VMSPLIT_3G_OPT
 	default 0x80000000 if VMSPLIT_2G
 	default 0x40000000 if VMSPLIT_1G
@@ -314,6 +315,15 @@ config HARDWALL
 	bool "Hardwall support to allow access to user dynamic network"
 	default y
 
+config KERNEL_PL
+	int "Processor protection level for kernel"
+	range 1 2
+	default "1"
+	---help---
+	  This setting determines the processor protection level the
+	  kernel will be built to run at.  Generally you should use
+	  the default value here.
+
 endmenu  # Tilera-specific configuration
 
 menu "Bus options"
@@ -354,3 +364,5 @@ source "security/Kconfig"
 source "crypto/Kconfig"
 
 source "lib/Kconfig"
+
+source "arch/tile/kvm/Kconfig"

arch/tile/Makefile

@@ -26,8 +26,9 @@ $(error Set TILERA_ROOT or CROSS_COMPILE when building $(ARCH) on $(HOST_ARCH))
   endif
 endif
 
-
+ifneq ($(CONFIG_DEBUG_EXTRA_FLAGS),"")
 KBUILD_CFLAGS   += $(CONFIG_DEBUG_EXTRA_FLAGS)
+endif
 
 LIBGCC_PATH     := $(shell $(CC) $(KBUILD_CFLAGS) -print-libgcc-file-name)
 
@@ -49,6 +50,20 @@ head-y		:= arch/tile/kernel/head_$(BITS).o
 libs-y		+= arch/tile/lib/
 libs-y		+= $(LIBGCC_PATH)
 
-
 # See arch/tile/Kbuild for content of core part of the kernel
 core-y		+= arch/tile/
+
+core-$(CONFIG_KVM) += arch/tile/kvm/
+
+ifdef TILERA_ROOT
+INSTALL_PATH ?= $(TILERA_ROOT)/tile/boot
+endif
+
+install:
+	install -D -m 755 vmlinux $(INSTALL_PATH)/vmlinux-$(KERNELRELEASE)
+	install -D -m 644 .config $(INSTALL_PATH)/config-$(KERNELRELEASE)
+	install -D -m 644 System.map $(INSTALL_PATH)/System.map-$(KERNELRELEASE)
+
+define archhelp
+	echo '  install         - install kernel into $(INSTALL_PATH)'
+endef

arch/tile/include/arch/spr_def.h

@@ -12,8 +12,93 @@
  *   more details.
  */
 
+/*
+ * In addition to including the proper base SPR definition file, depending
+ * on machine architecture, this file defines several macros which allow
+ * kernel code to use protection-level dependent SPRs without worrying
+ * about which PL it's running at.  In these macros, the PL that the SPR
+ * or interrupt number applies to is replaced by K.
+ */
+
+#if CONFIG_KERNEL_PL != 1 && CONFIG_KERNEL_PL != 2
+#error CONFIG_KERNEL_PL must be 1 or 2
+#endif
+
+/* Concatenate 4 strings. */
+#define __concat4(a, b, c, d) a ## b ## c ## d
+#define _concat4(a, b, c, d)  __concat4(a, b, c, d)
+
 #ifdef __tilegx__
 #include <arch/spr_def_64.h>
+
+/* TILE-Gx dependent, protection-level dependent SPRs. */
+
+#define SPR_INTERRUPT_MASK_K \
+	_concat4(SPR_INTERRUPT_MASK_, CONFIG_KERNEL_PL,,)
+#define SPR_INTERRUPT_MASK_SET_K \
+	_concat4(SPR_INTERRUPT_MASK_SET_, CONFIG_KERNEL_PL,,)
+#define SPR_INTERRUPT_MASK_RESET_K \
+	_concat4(SPR_INTERRUPT_MASK_RESET_, CONFIG_KERNEL_PL,,)
+#define SPR_INTERRUPT_VECTOR_BASE_K \
+	_concat4(SPR_INTERRUPT_VECTOR_BASE_, CONFIG_KERNEL_PL,,)
+
+#define SPR_IPI_MASK_K \
+	_concat4(SPR_IPI_MASK_, CONFIG_KERNEL_PL,,)
+#define SPR_IPI_MASK_RESET_K \
+	_concat4(SPR_IPI_MASK_RESET_, CONFIG_KERNEL_PL,,)
+#define SPR_IPI_MASK_SET_K \
+	_concat4(SPR_IPI_MASK_SET_, CONFIG_KERNEL_PL,,)
+#define SPR_IPI_EVENT_K \
+	_concat4(SPR_IPI_EVENT_, CONFIG_KERNEL_PL,,)
+#define SPR_IPI_EVENT_RESET_K \
+	_concat4(SPR_IPI_EVENT_RESET_, CONFIG_KERNEL_PL,,)
+#define SPR_IPI_MASK_SET_K \
+	_concat4(SPR_IPI_MASK_SET_, CONFIG_KERNEL_PL,,)
+#define INT_IPI_K \
+	_concat4(INT_IPI_, CONFIG_KERNEL_PL,,)
+
+#define SPR_SINGLE_STEP_CONTROL_K \
+	_concat4(SPR_SINGLE_STEP_CONTROL_, CONFIG_KERNEL_PL,,)
+#define SPR_SINGLE_STEP_EN_K_K \
+	_concat4(SPR_SINGLE_STEP_EN_, CONFIG_KERNEL_PL, _, CONFIG_KERNEL_PL)
+#define INT_SINGLE_STEP_K \
+	_concat4(INT_SINGLE_STEP_, CONFIG_KERNEL_PL,,)
+
 #else
 #include <arch/spr_def_32.h>
+
+/* TILEPro dependent, protection-level dependent SPRs. */
+
+#define SPR_INTERRUPT_MASK_K_0 \
+	_concat4(SPR_INTERRUPT_MASK_, CONFIG_KERNEL_PL, _0,)
+#define SPR_INTERRUPT_MASK_K_1 \
+	_concat4(SPR_INTERRUPT_MASK_, CONFIG_KERNEL_PL, _1,)
+#define SPR_INTERRUPT_MASK_SET_K_0 \
+	_concat4(SPR_INTERRUPT_MASK_SET_, CONFIG_KERNEL_PL, _0,)
+#define SPR_INTERRUPT_MASK_SET_K_1 \
+	_concat4(SPR_INTERRUPT_MASK_SET_, CONFIG_KERNEL_PL, _1,)
+#define SPR_INTERRUPT_MASK_RESET_K_0 \
+	_concat4(SPR_INTERRUPT_MASK_RESET_, CONFIG_KERNEL_PL, _0,)
+#define SPR_INTERRUPT_MASK_RESET_K_1 \
+	_concat4(SPR_INTERRUPT_MASK_RESET_, CONFIG_KERNEL_PL, _1,)
+
 #endif
+
+/* Generic protection-level dependent SPRs. */
+
+#define SPR_SYSTEM_SAVE_K_0 \
+	_concat4(SPR_SYSTEM_SAVE_, CONFIG_KERNEL_PL, _0,)
+#define SPR_SYSTEM_SAVE_K_1 \
+	_concat4(SPR_SYSTEM_SAVE_, CONFIG_KERNEL_PL, _1,)
+#define SPR_SYSTEM_SAVE_K_2 \
+	_concat4(SPR_SYSTEM_SAVE_, CONFIG_KERNEL_PL, _2,)
+#define SPR_SYSTEM_SAVE_K_3 \
+	_concat4(SPR_SYSTEM_SAVE_, CONFIG_KERNEL_PL, _3,)
+#define SPR_EX_CONTEXT_K_0 \
+	_concat4(SPR_EX_CONTEXT_, CONFIG_KERNEL_PL, _0,)
+#define SPR_EX_CONTEXT_K_1 \
+	_concat4(SPR_EX_CONTEXT_, CONFIG_KERNEL_PL, _1,)
+#define SPR_INTCTRL_K_STATUS \
+	_concat4(SPR_INTCTRL_, CONFIG_KERNEL_PL, _STATUS,)
+#define INT_INTCTRL_K \
+	_concat4(INT_INTCTRL_, CONFIG_KERNEL_PL,,)

arch/tile/include/arch/spr_def_32.h

@@ -56,58 +56,93 @@
 #define SPR_EX_CONTEXT_1_1__ICS_SHIFT 2
 #define SPR_EX_CONTEXT_1_1__ICS_RMASK 0x1
 #define SPR_EX_CONTEXT_1_1__ICS_MASK  0x4
+#define SPR_EX_CONTEXT_2_0 0x4605
+#define SPR_EX_CONTEXT_2_1 0x4606
+#define SPR_EX_CONTEXT_2_1__PL_SHIFT 0
+#define SPR_EX_CONTEXT_2_1__PL_RMASK 0x3
+#define SPR_EX_CONTEXT_2_1__PL_MASK  0x3
+#define SPR_EX_CONTEXT_2_1__ICS_SHIFT 2
+#define SPR_EX_CONTEXT_2_1__ICS_RMASK 0x1
+#define SPR_EX_CONTEXT_2_1__ICS_MASK  0x4
 #define SPR_FAIL 0x4e09
 #define SPR_INTCTRL_0_STATUS 0x4a07
 #define SPR_INTCTRL_1_STATUS 0x4807
+#define SPR_INTCTRL_2_STATUS 0x4607
 #define SPR_INTERRUPT_CRITICAL_SECTION 0x4e0a
 #define SPR_INTERRUPT_MASK_0_0 0x4a08
 #define SPR_INTERRUPT_MASK_0_1 0x4a09
 #define SPR_INTERRUPT_MASK_1_0 0x4809
 #define SPR_INTERRUPT_MASK_1_1 0x480a
+#define SPR_INTERRUPT_MASK_2_0 0x4608
+#define SPR_INTERRUPT_MASK_2_1 0x4609
 #define SPR_INTERRUPT_MASK_RESET_0_0 0x4a0a
 #define SPR_INTERRUPT_MASK_RESET_0_1 0x4a0b
 #define SPR_INTERRUPT_MASK_RESET_1_0 0x480b
 #define SPR_INTERRUPT_MASK_RESET_1_1 0x480c
+#define SPR_INTERRUPT_MASK_RESET_2_0 0x460a
+#define SPR_INTERRUPT_MASK_RESET_2_1 0x460b
 #define SPR_INTERRUPT_MASK_SET_0_0 0x4a0c
 #define SPR_INTERRUPT_MASK_SET_0_1 0x4a0d
 #define SPR_INTERRUPT_MASK_SET_1_0 0x480d
 #define SPR_INTERRUPT_MASK_SET_1_1 0x480e
+#define SPR_INTERRUPT_MASK_SET_2_0 0x460c
+#define SPR_INTERRUPT_MASK_SET_2_1 0x460d
 #define SPR_MPL_DMA_CPL_SET_0 0x5800
 #define SPR_MPL_DMA_CPL_SET_1 0x5801
+#define SPR_MPL_DMA_CPL_SET_2 0x5802
 #define SPR_MPL_DMA_NOTIFY_SET_0 0x3800
 #define SPR_MPL_DMA_NOTIFY_SET_1 0x3801
+#define SPR_MPL_DMA_NOTIFY_SET_2 0x3802
 #define SPR_MPL_INTCTRL_0_SET_0 0x4a00
 #define SPR_MPL_INTCTRL_0_SET_1 0x4a01
+#define SPR_MPL_INTCTRL_0_SET_2 0x4a02
 #define SPR_MPL_INTCTRL_1_SET_0 0x4800
 #define SPR_MPL_INTCTRL_1_SET_1 0x4801
+#define SPR_MPL_INTCTRL_1_SET_2 0x4802
+#define SPR_MPL_INTCTRL_2_SET_0 0x4600
+#define SPR_MPL_INTCTRL_2_SET_1 0x4601
+#define SPR_MPL_INTCTRL_2_SET_2 0x4602
 #define SPR_MPL_SN_ACCESS_SET_0 0x0800
 #define SPR_MPL_SN_ACCESS_SET_1 0x0801
+#define SPR_MPL_SN_ACCESS_SET_2 0x0802
 #define SPR_MPL_SN_CPL_SET_0 0x5a00
 #define SPR_MPL_SN_CPL_SET_1 0x5a01
+#define SPR_MPL_SN_CPL_SET_2 0x5a02
 #define SPR_MPL_SN_FIREWALL_SET_0 0x2c00
 #define SPR_MPL_SN_FIREWALL_SET_1 0x2c01
+#define SPR_MPL_SN_FIREWALL_SET_2 0x2c02
 #define SPR_MPL_SN_NOTIFY_SET_0 0x2a00
 #define SPR_MPL_SN_NOTIFY_SET_1 0x2a01
+#define SPR_MPL_SN_NOTIFY_SET_2 0x2a02
 #define SPR_MPL_UDN_ACCESS_SET_0 0x0c00
 #define SPR_MPL_UDN_ACCESS_SET_1 0x0c01
+#define SPR_MPL_UDN_ACCESS_SET_2 0x0c02
 #define SPR_MPL_UDN_AVAIL_SET_0 0x4000
 #define SPR_MPL_UDN_AVAIL_SET_1 0x4001
+#define SPR_MPL_UDN_AVAIL_SET_2 0x4002
 #define SPR_MPL_UDN_CA_SET_0 0x3c00
 #define SPR_MPL_UDN_CA_SET_1 0x3c01
+#define SPR_MPL_UDN_CA_SET_2 0x3c02
 #define SPR_MPL_UDN_COMPLETE_SET_0 0x1400
 #define SPR_MPL_UDN_COMPLETE_SET_1 0x1401
+#define SPR_MPL_UDN_COMPLETE_SET_2 0x1402
 #define SPR_MPL_UDN_FIREWALL_SET_0 0x3000
 #define SPR_MPL_UDN_FIREWALL_SET_1 0x3001
+#define SPR_MPL_UDN_FIREWALL_SET_2 0x3002
 #define SPR_MPL_UDN_REFILL_SET_0 0x1000
 #define SPR_MPL_UDN_REFILL_SET_1 0x1001
+#define SPR_MPL_UDN_REFILL_SET_2 0x1002
 #define SPR_MPL_UDN_TIMER_SET_0 0x3600
 #define SPR_MPL_UDN_TIMER_SET_1 0x3601
+#define SPR_MPL_UDN_TIMER_SET_2 0x3602
 #define SPR_MPL_WORLD_ACCESS_SET_0 0x4e00
 #define SPR_MPL_WORLD_ACCESS_SET_1 0x4e01
+#define SPR_MPL_WORLD_ACCESS_SET_2 0x4e02
 #define SPR_PASS 0x4e0b
 #define SPR_PERF_COUNT_0 0x4205
 #define SPR_PERF_COUNT_1 0x4206
 #define SPR_PERF_COUNT_CTL 0x4207
+#define SPR_PERF_COUNT_DN_CTL 0x4210
 #define SPR_PERF_COUNT_STS 0x4208
 #define SPR_PROC_STATUS 0x4f00
 #define SPR_SIM_CONTROL 0x4e0c
@@ -124,6 +159,10 @@
 #define SPR_SYSTEM_SAVE_1_1 0x4901
 #define SPR_SYSTEM_SAVE_1_2 0x4902
 #define SPR_SYSTEM_SAVE_1_3 0x4903
+#define SPR_SYSTEM_SAVE_2_0 0x4700
+#define SPR_SYSTEM_SAVE_2_1 0x4701
+#define SPR_SYSTEM_SAVE_2_2 0x4702
+#define SPR_SYSTEM_SAVE_2_3 0x4703
 #define SPR_TILE_COORD 0x4c17
 #define SPR_TILE_RTF_HWM 0x4e10
 #define SPR_TILE_TIMER_CONTROL 0x3205

arch/tile/include/asm/backtrace.h

@@ -146,7 +146,10 @@ enum {
 
 	CALLER_SP_IN_R52_BASE = 4,
 
-	CALLER_SP_OFFSET_BASE = 8
+	CALLER_SP_OFFSET_BASE = 8,
+
+	/* Marks the entry point of certain functions. */
+	ENTRY_POINT_INFO_OP = 16
 };
 
 

arch/tile/include/asm/compat.h

@@ -216,15 +216,16 @@ struct compat_siginfo;
 struct compat_sigaltstack;
 long compat_sys_execve(const char __user *path,
 		       const compat_uptr_t __user *argv,
-		       const compat_uptr_t __user *envp);
+		       const compat_uptr_t __user *envp, struct pt_regs *);
 long compat_sys_rt_sigaction(int sig, struct compat_sigaction __user *act,
 			     struct compat_sigaction __user *oact,
 			     size_t sigsetsize);
 long compat_sys_rt_sigqueueinfo(int pid, int sig,
 				struct compat_siginfo __user *uinfo);
-long compat_sys_rt_sigreturn(void);
+long compat_sys_rt_sigreturn(struct pt_regs *);
 long compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
-			    struct compat_sigaltstack __user *uoss_ptr);
+			    struct compat_sigaltstack __user *uoss_ptr,
+			    struct pt_regs *);
 long compat_sys_truncate64(char __user *filename, u32 dummy, u32 low, u32 high);
 long compat_sys_ftruncate64(unsigned int fd, u32 dummy, u32 low, u32 high);
 long compat_sys_pread64(unsigned int fd, char __user *ubuf, size_t count,
@@ -255,4 +256,12 @@ long tile_compat_sys_ptrace(compat_long_t request, compat_long_t pid,
 /* Tilera Linux syscalls that don't have "compat" versions. */
 #define compat_sys_flush_cache sys_flush_cache
 
+/* These are the intvec_64.S trampolines. */
+long _compat_sys_execve(const char __user *path,
+			const compat_uptr_t __user *argv,
+			const compat_uptr_t __user *envp);
+long _compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
+			    struct compat_sigaltstack __user *uoss_ptr);
+long _compat_sys_rt_sigreturn(void);
+
 #endif /* _ASM_TILE_COMPAT_H */

arch/tile/include/asm/irqflags.h

@@ -47,53 +47,53 @@
 	int __n = (n); \
 	int __mask = 1 << (__n & 0x1f); \
 	if (__n < 32) \
-		__insn_mtspr(SPR_INTERRUPT_MASK_SET_1_0, __mask); \
+		__insn_mtspr(SPR_INTERRUPT_MASK_SET_K_0, __mask); \
 	else \
-		__insn_mtspr(SPR_INTERRUPT_MASK_SET_1_1, __mask); \
+		__insn_mtspr(SPR_INTERRUPT_MASK_SET_K_1, __mask); \
 } while (0)
 #define interrupt_mask_reset(n) do { \
 	int __n = (n); \
 	int __mask = 1 << (__n & 0x1f); \
 	if (__n < 32) \
-		__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1_0, __mask); \
+		__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_0, __mask); \
 	else \
-		__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1_1, __mask); \
+		__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_1, __mask); \
 } while (0)
 #define interrupt_mask_check(n) ({ \
 	int __n = (n); \
 	(((__n < 32) ? \
-	 __insn_mfspr(SPR_INTERRUPT_MASK_1_0) : \
-	 __insn_mfspr(SPR_INTERRUPT_MASK_1_1)) \
+	 __insn_mfspr(SPR_INTERRUPT_MASK_K_0) : \
+	 __insn_mfspr(SPR_INTERRUPT_MASK_K_1)) \
 	  >> (__n & 0x1f)) & 1; \
 })
 #define interrupt_mask_set_mask(mask) do { \
 	unsigned long long __m = (mask); \
-	__insn_mtspr(SPR_INTERRUPT_MASK_SET_1_0, (unsigned long)(__m)); \
-	__insn_mtspr(SPR_INTERRUPT_MASK_SET_1_1, (unsigned long)(__m>>32)); \
+	__insn_mtspr(SPR_INTERRUPT_MASK_SET_K_0, (unsigned long)(__m)); \
+	__insn_mtspr(SPR_INTERRUPT_MASK_SET_K_1, (unsigned long)(__m>>32)); \
 } while (0)
 #define interrupt_mask_reset_mask(mask) do { \
 	unsigned long long __m = (mask); \
-	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1_0, (unsigned long)(__m)); \
-	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1_1, (unsigned long)(__m>>32)); \
+	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_0, (unsigned long)(__m)); \
+	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K_1, (unsigned long)(__m>>32)); \
 } while (0)
 #else
 #define interrupt_mask_set(n) \
-	__insn_mtspr(SPR_INTERRUPT_MASK_SET_1, (1UL << (n)))
+	__insn_mtspr(SPR_INTERRUPT_MASK_SET_K, (1UL << (n)))
 #define interrupt_mask_reset(n) \
-	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1, (1UL << (n)))
+	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K, (1UL << (n)))
 #define interrupt_mask_check(n) \
-	((__insn_mfspr(SPR_INTERRUPT_MASK_1) >> (n)) & 1)
+	((__insn_mfspr(SPR_INTERRUPT_MASK_K) >> (n)) & 1)
 #define interrupt_mask_set_mask(mask) \
-	__insn_mtspr(SPR_INTERRUPT_MASK_SET_1, (mask))
+	__insn_mtspr(SPR_INTERRUPT_MASK_SET_K, (mask))
 #define interrupt_mask_reset_mask(mask) \
-	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_1, (mask))
+	__insn_mtspr(SPR_INTERRUPT_MASK_RESET_K, (mask))
 #endif
 
 /*
  * The set of interrupts we want active if irqs are enabled.
  * Note that in particular, the tile timer interrupt comes and goes
  * from this set, since we have no other way to turn off the timer.
- * Likewise, INTCTRL_1 is removed and re-added during device
+ * Likewise, INTCTRL_K is removed and re-added during device
  * interrupts, as is the the hardwall UDN_FIREWALL interrupt.
  * We use a low bit (MEM_ERROR) as our sentinel value and make sure it
  * is always claimed as an "active interrupt" so we can query that bit
@@ -170,7 +170,7 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 
 /* Return 0 or 1 to indicate whether interrupts are currently disabled. */
 #define IRQS_DISABLED(tmp)					\
-	mfspr   tmp, INTERRUPT_MASK_1;				\
+	mfspr   tmp, SPR_INTERRUPT_MASK_K;			\
 	andi    tmp, tmp, 1
 
 /* Load up a pointer to &interrupts_enabled_mask. */
@@ -185,18 +185,18 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 	moveli  tmp0, hw2_last(LINUX_MASKABLE_INTERRUPTS);	\
 	shl16insli tmp0, tmp0, hw1(LINUX_MASKABLE_INTERRUPTS);	\
 	shl16insli tmp0, tmp0, hw0(LINUX_MASKABLE_INTERRUPTS);	\
-	mtspr   INTERRUPT_MASK_SET_1, tmp0
+	mtspr   SPR_INTERRUPT_MASK_SET_K, tmp0
 
 /* Disable ALL synchronous interrupts (used by NMI entry). */
 #define IRQ_DISABLE_ALL(tmp)					\
 	movei   tmp, -1;					\
-	mtspr   INTERRUPT_MASK_SET_1, tmp
+	mtspr   SPR_INTERRUPT_MASK_SET_K, tmp
 
 /* Enable interrupts. */
 #define IRQ_ENABLE(tmp0, tmp1)					\
 	GET_INTERRUPTS_ENABLED_MASK_PTR(tmp0);			\
 	ld      tmp0, tmp0;					\
-	mtspr   INTERRUPT_MASK_RESET_1, tmp0
+	mtspr   SPR_INTERRUPT_MASK_RESET_K, tmp0
 
 #else /* !__tilegx__ */
 
@@ -210,14 +210,14 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
  * (making the original code's write of the "high" mask word idempotent).
  */
 #define IRQS_DISABLED(tmp)					\
-	mfspr   tmp, INTERRUPT_MASK_1_0;			\
+	mfspr   tmp, SPR_INTERRUPT_MASK_K_0;			\
 	shri    tmp, tmp, INT_MEM_ERROR;			\
 	andi    tmp, tmp, 1
 
 /* Load up a pointer to &interrupts_enabled_mask. */
 #define GET_INTERRUPTS_ENABLED_MASK_PTR(reg)			\
 	moveli  reg, lo16(interrupts_enabled_mask);		\
-	auli    reg, reg, ha16(interrupts_enabled_mask);\
+	auli    reg, reg, ha16(interrupts_enabled_mask);	\
 	add     reg, reg, tp
 
 /* Disable interrupts. */
@@ -227,16 +227,16 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 	 moveli tmp1, lo16(LINUX_MASKABLE_INTERRUPTS)		\
 	};							\
 	{							\
-	 mtspr  INTERRUPT_MASK_SET_1_0, tmp0;			\
+	 mtspr  SPR_INTERRUPT_MASK_SET_K_0, tmp0;		\
 	 auli   tmp1, tmp1, ha16(LINUX_MASKABLE_INTERRUPTS)	\
 	};							\
-	mtspr   INTERRUPT_MASK_SET_1_1, tmp1
+	mtspr   SPR_INTERRUPT_MASK_SET_K_1, tmp1
 
 /* Disable ALL synchronous interrupts (used by NMI entry). */
 #define IRQ_DISABLE_ALL(tmp)					\
 	movei   tmp, -1;					\
-	mtspr   INTERRUPT_MASK_SET_1_0, tmp;			\
-	mtspr   INTERRUPT_MASK_SET_1_1, tmp
+	mtspr   SPR_INTERRUPT_MASK_SET_K_0, tmp;		\
+	mtspr   SPR_INTERRUPT_MASK_SET_K_1, tmp
 
 /* Enable interrupts. */
 #define IRQ_ENABLE(tmp0, tmp1)					\
@@ -246,8 +246,8 @@ DECLARE_PER_CPU(unsigned long long, interrupts_enabled_mask);
 	 addi   tmp1, tmp0, 4					\
 	};							\
 	lw      tmp1, tmp1;					\
-	mtspr   INTERRUPT_MASK_RESET_1_0, tmp0;			\
-	mtspr   INTERRUPT_MASK_RESET_1_1, tmp1
+	mtspr   SPR_INTERRUPT_MASK_RESET_K_0, tmp0;		\
+	mtspr   SPR_INTERRUPT_MASK_RESET_K_1, tmp1
 #endif
 
 /*

arch/tile/include/asm/mman.h

@@ -23,6 +23,7 @@
 #define MAP_POPULATE	0x0040		/* populate (prefault) pagetables */
 #define MAP_NONBLOCK	0x0080		/* do not block on IO */
 #define MAP_GROWSDOWN	0x0100		/* stack-like segment */
+#define MAP_STACK	MAP_GROWSDOWN	/* provide convenience alias */
 #define MAP_LOCKED	0x0200		/* pages are locked */
 #define MAP_NORESERVE	0x0400		/* don't check for reservations */
 #define MAP_DENYWRITE	0x0800		/* ETXTBSY */

arch/tile/include/asm/page.h

@@ -199,17 +199,17 @@ static inline __attribute_const__ int get_order(unsigned long size)
  * If you want more physical memory than this then see the CONFIG_HIGHMEM
  * option in the kernel configuration.
  *
- * The top two 16MB chunks in the table below (VIRT and HV) are
- * unavailable to Linux.  Since the kernel interrupt vectors must live
- * at 0xfd000000, we map all of the bottom of RAM at this address with
- * a huge page table entry to minimize its ITLB footprint (as well as
- * at PAGE_OFFSET).  The last architected requirement is that user
- * interrupt vectors live at 0xfc000000, so we make that range of
- * memory available to user processes.  The remaining regions are sized
- * as shown; after the first four addresses, we show "typical" values,
- * since the actual addresses depend on kernel #defines.
+ * The top 16MB chunk in the table below is unavailable to Linux.  Since
+ * the kernel interrupt vectors must live at ether 0xfe000000 or 0xfd000000
+ * (depending on whether the kernel is at PL2 or Pl1), we map all of the
+ * bottom of RAM at this address with a huge page table entry to minimize
+ * its ITLB footprint (as well as at PAGE_OFFSET).  The last architected
+ * requirement is that user interrupt vectors live at 0xfc000000, so we
+ * make that range of memory available to user processes.  The remaining
+ * regions are sized as shown; the first four addresses use the PL 1
+ * values, and after that, we show "typical" values, since the actual
+ * addresses depend on kernel #defines.
  *
- * MEM_VIRT_INTRPT                 0xff000000
  * MEM_HV_INTRPT                   0xfe000000
  * MEM_SV_INTRPT (kernel code)     0xfd000000
  * MEM_USER_INTRPT (user vector)   0xfc000000
@@ -221,9 +221,14 @@ static inline __attribute_const__ int get_order(unsigned long size)
  */
 
 #define MEM_USER_INTRPT		_AC(0xfc000000, UL)
+#if CONFIG_KERNEL_PL == 1
 #define MEM_SV_INTRPT		_AC(0xfd000000, UL)
 #define MEM_HV_INTRPT		_AC(0xfe000000, UL)
-#define MEM_VIRT_INTRPT		_AC(0xff000000, UL)
+#else
+#define MEM_GUEST_INTRPT	_AC(0xfd000000, UL)
+#define MEM_SV_INTRPT		_AC(0xfe000000, UL)
+#define MEM_HV_INTRPT		_AC(0xff000000, UL)
+#endif
 
 #define INTRPT_SIZE		0x4000
 

arch/tile/include/asm/processor.h

@@ -328,18 +328,21 @@ extern int kdata_huge;
  * Note that assembly code assumes that USER_PL is zero.
  */
 #define USER_PL 0
-#define KERNEL_PL 1
+#if CONFIG_KERNEL_PL == 2
+#define GUEST_PL 1
+#endif
+#define KERNEL_PL CONFIG_KERNEL_PL
 
-/* SYSTEM_SAVE_1_0 holds the current cpu number ORed with ksp0. */
+/* SYSTEM_SAVE_K_0 holds the current cpu number ORed with ksp0. */
 #define CPU_LOG_MASK_VALUE 12
 #define CPU_MASK_VALUE ((1 << CPU_LOG_MASK_VALUE) - 1)
 #if CONFIG_NR_CPUS > CPU_MASK_VALUE
 # error Too many cpus!
 #endif
 #define raw_smp_processor_id() \
-	((int)__insn_mfspr(SPR_SYSTEM_SAVE_1_0) & CPU_MASK_VALUE)
+	((int)__insn_mfspr(SPR_SYSTEM_SAVE_K_0) & CPU_MASK_VALUE)
 #define get_current_ksp0() \
-	(__insn_mfspr(SPR_SYSTEM_SAVE_1_0) & ~CPU_MASK_VALUE)
+	(__insn_mfspr(SPR_SYSTEM_SAVE_K_0) & ~CPU_MASK_VALUE)
 #define next_current_ksp0(task) ({ \
 	unsigned long __ksp0 = task_ksp0(task); \
 	int __cpu = raw_smp_processor_id(); \

arch/tile/include/asm/ptrace.h

@@ -62,8 +62,8 @@ struct pt_regs {
 	pt_reg_t lr;		/* aliases regs[TREG_LR] */
 
 	/* Saved special registers. */
-	pt_reg_t pc;		/* stored in EX_CONTEXT_1_0 */
-	pt_reg_t ex1;		/* stored in EX_CONTEXT_1_1 (PL and ICS bit) */
+	pt_reg_t pc;		/* stored in EX_CONTEXT_K_0 */
+	pt_reg_t ex1;		/* stored in EX_CONTEXT_K_1 (PL and ICS bit) */
 	pt_reg_t faultnum;	/* fault number (INT_SWINT_1 for syscall) */
 	pt_reg_t orig_r0;	/* r0 at syscall entry, else zero */
 	pt_reg_t flags;		/* flags (see below) */

arch/tile/include/asm/syscalls.h

@@ -32,8 +32,9 @@ extern void *compat_sys_call_table[];
 
 /*
  * Note that by convention, any syscall which requires the current
- * register set takes an additional "struct pt_regs *" pointer; the
- * sys_xxx() function just adds the pointer and tail-calls to _sys_xxx().
+ * register set takes an additional "struct pt_regs *" pointer; a
+ * _sys_xxx() trampoline in intvec*.S just sets up the pointer and
+ * jumps to sys_xxx().
  */
 
 /* kernel/sys.c */
@@ -43,66 +44,17 @@ long sys32_fadvise64(int fd, u32 offset_lo, u32 offset_hi,
 int sys32_fadvise64_64(int fd, u32 offset_lo, u32 offset_hi,
 		       u32 len_lo, u32 len_hi, int advice);
 long sys_flush_cache(void);
-long sys_mmap2(unsigned long addr, unsigned long len,
-	       unsigned long prot, unsigned long flags,
-	       unsigned long fd, unsigned long pgoff);
-#ifdef __tilegx__
-long sys_mmap(unsigned long addr, unsigned long len,
-	      unsigned long prot, unsigned long flags,
-	      unsigned long fd, off_t pgoff);
+#ifndef __tilegx__  /* No mmap() in the 32-bit kernel. */
+#define sys_mmap sys_mmap
 #endif
 
-/* kernel/process.c */
-long sys_clone(unsigned long clone_flags, unsigned long newsp,
-	       void __user *parent_tid, void __user *child_tid);
-long _sys_clone(unsigned long clone_flags, unsigned long newsp,
-		void __user *parent_tid, void __user *child_tid,
-		struct pt_regs *regs);
-long sys_fork(void);
-long _sys_fork(struct pt_regs *regs);
-long sys_vfork(void);
-long _sys_vfork(struct pt_regs *regs);
-long sys_execve(const char __user *filename,
-		const char __user *const __user *argv,
-		const char __user *const __user *envp);
-long _sys_execve(const char __user *filename,
-		 const char __user *const __user *argv,
-		 const char __user *const __user *envp, struct pt_regs *regs);
-
-/* kernel/signal.c */
-long sys_sigaltstack(const stack_t __user *, stack_t __user *);
-long _sys_sigaltstack(const stack_t __user *, stack_t __user *,
-		      struct pt_regs *);
-long sys_rt_sigreturn(void);
-long _sys_rt_sigreturn(struct pt_regs *regs);
-
-/* platform-independent functions */
-long sys_rt_sigsuspend(sigset_t __user *unewset, size_t sigsetsize);
-long sys_rt_sigaction(int sig, const struct sigaction __user *act,
-		      struct sigaction __user *oact, size_t sigsetsize);
-
 #ifndef __tilegx__
 /* mm/fault.c */
-int sys_cmpxchg_badaddr(unsigned long address);
-int _sys_cmpxchg_badaddr(unsigned long address, struct pt_regs *);
+long sys_cmpxchg_badaddr(unsigned long address, struct pt_regs *);
+long _sys_cmpxchg_badaddr(unsigned long address);
 #endif
 
 #ifdef CONFIG_COMPAT
-long compat_sys_execve(const char __user *path,
-		       const compat_uptr_t __user *argv,
-		       const compat_uptr_t __user *envp);
-long _compat_sys_execve(const char __user *path,
-			const compat_uptr_t __user *argv,
-			const compat_uptr_t __user *envp,
-			struct pt_regs *regs);
-long compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
-			    struct compat_sigaltstack __user *uoss_ptr);
-long _compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
-			     struct compat_sigaltstack __user *uoss_ptr,
-			     struct pt_regs *regs);
-long compat_sys_rt_sigreturn(void);
-long _compat_sys_rt_sigreturn(struct pt_regs *regs);
-
 /* These four are not defined for 64-bit, but serve as "compat" syscalls. */
 long sys_fcntl64(unsigned int fd, unsigned int cmd, unsigned long arg);
 long sys_fstat64(unsigned long fd, struct stat64 __user *statbuf);
@@ -110,4 +62,15 @@ long sys_truncate64(const char __user *path, loff_t length);
 long sys_ftruncate64(unsigned int fd, loff_t length);
 #endif
 
+/* These are the intvec*.S trampolines. */
+long _sys_sigaltstack(const stack_t __user *, stack_t __user *);
+long _sys_rt_sigreturn(void);
+long _sys_clone(unsigned long clone_flags, unsigned long newsp,
+		void __user *parent_tid, void __user *child_tid);
+long _sys_execve(const char __user *filename,
+		 const char __user *const __user *argv,
+		 const char __user *const __user *envp);
+
+#include <asm-generic/syscalls.h>
+
 #endif /* _ASM_TILE_SYSCALLS_H */

arch/tile/include/asm/system.h

@@ -89,6 +89,10 @@
 #define get_cycles_low() __insn_mfspr(SPR_CYCLE)   /* just get all 64 bits */
 #endif
 
+#if !CHIP_HAS_MF_WAITS_FOR_VICTIMS()
+int __mb_incoherent(void);  /* Helper routine for mb_incoherent(). */
+#endif
+
 /* Fence to guarantee visibility of stores to incoherent memory. */
 static inline void
 mb_incoherent(void)
@@ -97,7 +101,6 @@ mb_incoherent(void)
 
 #if !CHIP_HAS_MF_WAITS_FOR_VICTIMS()
 	{
-		int __mb_incoherent(void);
 #if CHIP_HAS_TILE_WRITE_PENDING()
 		const unsigned long WRITE_TIMEOUT_CYCLES = 400;
 		unsigned long start = get_cycles_low();
@@ -161,7 +164,7 @@ extern struct task_struct *_switch_to(struct task_struct *prev,
 /* Helper function for _switch_to(). */
 extern struct task_struct *__switch_to(struct task_struct *prev,
 				       struct task_struct *next,
-				       unsigned long new_system_save_1_0);
+				       unsigned long new_system_save_k_0);
 
 /* Address that switched-away from tasks are at. */
 extern unsigned long get_switch_to_pc(void);
@@ -214,13 +217,6 @@ int hardwall_deactivate(struct task_struct *task);
 } while (0)
 #endif
 
-/* Invoke the simulator "syscall" mechanism (see arch/tile/kernel/entry.S). */
-extern int _sim_syscall(int syscall_num, ...);
-#define sim_syscall(syscall_num, ...) \
-	_sim_syscall(SIM_CONTROL_SYSCALL + \
-		((syscall_num) << _SIM_CONTROL_OPERATOR_BITS), \
-		## __VA_ARGS__)
-
 /*
  * Kernel threads can check to see if they need to migrate their
  * stack whenever they return from a context switch; for user

arch/tile/include/asm/traps.h

@@ -59,4 +59,8 @@ void do_hardwall_trap(struct pt_regs *, int fault_num);
 void do_breakpoint(struct pt_regs *, int fault_num);
 
 
+#ifdef __tilegx__
+void gx_singlestep_handle(struct pt_regs *, int fault_num);
+#endif
+
 #endif /* _ASM_TILE_SYSCALLS_H */

arch/tile/include/hv/hypervisor.h

@@ -1003,37 +1003,37 @@ int hv_console_write(HV_VirtAddr bytes, int len);
  *  when these occur in a client's interrupt critical section, they must
  *  be delivered through the downcall mechanism.
  *
- *  A downcall is initially delivered to the client as an INTCTRL_1
- *  interrupt.  Upon entry to the INTCTRL_1 vector, the client must
- *  immediately invoke the hv_downcall_dispatch service.  This service
- *  will not return; instead it will cause one of the client's actual
- *  downcall-handling interrupt vectors to be entered.  The EX_CONTEXT
- *  registers in the client will be set so that when the client irets,
- *  it will return to the code which was interrupted by the INTCTRL_1
- *  interrupt.
- *
- *  Under some circumstances, the firing of INTCTRL_1 can race with
+ *  A downcall is initially delivered to the client as an INTCTRL_CL
+ *  interrupt, where CL is the client's PL.  Upon entry to the INTCTRL_CL
+ *  vector, the client must immediately invoke the hv_downcall_dispatch
+ *  service.  This service will not return; instead it will cause one of
+ *  the client's actual downcall-handling interrupt vectors to be entered.
+ *  The EX_CONTEXT registers in the client will be set so that when the
+ *  client irets, it will return to the code which was interrupted by the
+ *  INTCTRL_CL interrupt.
+ *
+ *  Under some circumstances, the firing of INTCTRL_CL can race with
  *  the lowering of a device interrupt.  In such a case, the
  *  hv_downcall_dispatch service may issue an iret instruction instead
  *  of entering one of the client's actual downcall-handling interrupt
  *  vectors.  This will return execution to the location that was
- *  interrupted by INTCTRL_1.
+ *  interrupted by INTCTRL_CL.
  *
  *  Any saving of registers should be done by the actual handling
- *  vectors; no registers should be changed by the INTCTRL_1 handler.
+ *  vectors; no registers should be changed by the INTCTRL_CL handler.
  *  In particular, the client should not use a jal instruction to invoke
  *  the hv_downcall_dispatch service, as that would overwrite the client's
  *  lr register.  Note that the hv_downcall_dispatch service may overwrite
  *  one or more of the client's system save registers.
  *
- *  The client must not modify the INTCTRL_1_STATUS SPR.  The hypervisor
+ *  The client must not modify the INTCTRL_CL_STATUS SPR.  The hypervisor
  *  will set this register to cause a downcall to happen, and will clear
  *  it when no further downcalls are pending.
  *
- *  When a downcall vector is entered, the INTCTRL_1 interrupt will be
+ *  When a downcall vector is entered, the INTCTRL_CL interrupt will be
  *  masked.  When the client is done processing a downcall, and is ready
  *  to accept another, it must unmask this interrupt; if more downcalls
- *  are pending, this will cause the INTCTRL_1 vector to be reentered.
+ *  are pending, this will cause the INTCTRL_CL vector to be reentered.
  *  Currently the following interrupt vectors can be entered through a
  *  downcall:
  *

arch/tile/kernel/backtrace.c

@@ -369,6 +369,10 @@ static void find_caller_pc_and_caller_sp(CallerLocation *location,
 					/* Weird; reserved value, ignore it. */
 					continue;
 				}
+				if (info_operand & ENTRY_POINT_INFO_OP)	{
+					/* This info op is ignored by the backtracer. */
+					continue;
+				}
 
 				/* Skip info ops which are not in the
 				 * "one_ago" mode we want right now.

arch/tile/kernel/compat.c

@@ -154,8 +154,14 @@ long tile_compat_sys_msgrcv(int msqid,
 #define compat_sys_fstat64 sys_newfstat
 #define compat_sys_fstatat64 sys_newfstatat
 
-/* Pass full 64-bit values through ptrace. */
-#define compat_sys_ptrace tile_compat_sys_ptrace
+/* The native sys_ptrace dynamically handles compat binaries. */
+#define compat_sys_ptrace sys_ptrace
+
+/* Call the trampolines to manage pt_regs where necessary. */
+#define compat_sys_execve _compat_sys_execve
+#define compat_sys_sigaltstack _compat_sys_sigaltstack
+#define compat_sys_rt_sigreturn _compat_sys_rt_sigreturn
+#define sys_clone _sys_clone
 
 /*
  * Note that we can't include <linux/unistd.h> here since the header

arch/tile/kernel/compat_signal.c

@@ -256,7 +256,7 @@ int copy_siginfo_from_user32(siginfo_t *to, struct compat_siginfo __user *from)
 	return err;
 }
 
-long _compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
+long compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
 			    struct compat_sigaltstack __user *uoss_ptr,
 			    struct pt_regs *regs)
 {
@@ -291,7 +291,7 @@ long _compat_sys_sigaltstack(const struct compat_sigaltstack __user *uss_ptr,
 	return ret;
 }
 
-long _compat_sys_rt_sigreturn(struct pt_regs *regs)
+long compat_sys_rt_sigreturn(struct pt_regs *regs)
 {
 	struct compat_rt_sigframe __user *frame =
 		(struct compat_rt_sigframe __user *) compat_ptr(regs->sp);
@@ -312,7 +312,7 @@ long _compat_sys_rt_sigreturn(struct pt_regs *regs)
 	if (restore_sigcontext(regs, &frame->uc.uc_mcontext, &r0))
 		goto badframe;
 
-	if (_compat_sys_sigaltstack(&frame->uc.uc_stack, NULL, regs) != 0)
+	if (compat_sys_sigaltstack(&frame->uc.uc_stack, NULL, regs) != 0)
 		goto badframe;
 
 	return r0;

arch/tile/kernel/entry.S

@@ -15,7 +15,9 @@
 #include <linux/linkage.h>
 #include <linux/unistd.h>
 #include <asm/irqflags.h>
+#include <asm/processor.h>
 #include <arch/abi.h>
+#include <arch/spr_def.h>
 
 #ifdef __tilegx__
 #define bnzt bnezt
@@ -25,28 +27,6 @@ STD_ENTRY(current_text_addr)
 	{ move r0, lr; jrp lr }
 	STD_ENDPROC(current_text_addr)
 
-STD_ENTRY(_sim_syscall)
-	/*
-	 * Wait for r0-r9 to be ready (and lr on the off chance we
-	 * want the syscall to locate its caller), then make a magic
-	 * simulator syscall.
-	 *
-	 * We carefully stall until the registers are readable in case they
-	 * are the target of a slow load, etc. so that tile-sim will
-	 * definitely be able to read all of them inside the magic syscall.
-	 *
-	 * Technically this is wrong for r3-r9 and lr, since an interrupt
-	 * could come in and restore the registers with a slow load right
-	 * before executing the mtspr. We may need to modify tile-sim to
-	 * explicitly stall for this case, but we do not yet have
-	 * a way to implement such a stall.
-	 */
-	{ and zero, lr, r9 ; and zero, r8, r7 }
-	{ and zero, r6, r5 ; and zero, r4, r3 }
-	{ and zero, r2, r1 ; mtspr SIM_CONTROL, r0 }
-	{ jrp lr }
-	STD_ENDPROC(_sim_syscall)
-
 /*
  * Implement execve().  The i386 code has a note that forking from kernel
  * space results in no copy on write until the execve, so we should be
@@ -102,7 +82,7 @@ STD_ENTRY(KBacktraceIterator_init_current)
 STD_ENTRY(cpu_idle_on_new_stack)
 	{
 	 move sp, r1
-	 mtspr SYSTEM_SAVE_1_0, r2
+	 mtspr SPR_SYSTEM_SAVE_K_0, r2
 	}
 	jal free_thread_info
 	j cpu_idle
@@ -124,15 +104,15 @@ STD_ENTRY(smp_nap)
 STD_ENTRY(_cpu_idle)
 	{
 	 lnk r0
-	 movei r1, 1
+	 movei r1, KERNEL_PL
 	}
 	{
 	 addli r0, r0, _cpu_idle_nap - .
 	 mtspr INTERRUPT_CRITICAL_SECTION, r1
 	}
 	IRQ_ENABLE(r2, r3)             /* unmask, but still with ICS set */
-	mtspr EX_CONTEXT_1_1, r1   /* PL1, ICS clear */
-	mtspr EX_CONTEXT_1_0, r0
+	mtspr SPR_EX_CONTEXT_K_1, r1   /* Kernel PL, ICS clear */
+	mtspr SPR_EX_CONTEXT_K_0, r0
 	iret
 	.global _cpu_idle_nap
 _cpu_idle_nap:

arch/tile/kernel/head_32.S

@@ -23,6 +23,7 @@
 #include <asm/asm-offsets.h>
 #include <hv/hypervisor.h>
 #include <arch/chip.h>
+#include <arch/spr_def.h>
 
 /*
  * This module contains the entry code for kernel images. It performs the
@@ -76,7 +77,7 @@ ENTRY(_start)
 	}
 1:
 
-	/* Get our processor number and save it away in SAVE_1_0. */
+	/* Get our processor number and save it away in SAVE_K_0. */
 	jal hv_inquire_topology
 	mulll_uu r4, r1, r2        /* r1 == y, r2 == width */
 	add r4, r4, r0             /* r0 == x, so r4 == cpu == y*width + x */
@@ -124,7 +125,7 @@ ENTRY(_start)
 	lw r0, r0
 	lw sp, r1
 	or r4, sp, r4
-	mtspr SYSTEM_SAVE_1_0, r4  /* save ksp0 + cpu */
+	mtspr SPR_SYSTEM_SAVE_K_0, r4  /* save ksp0 + cpu */
 	addi sp, sp, -STACK_TOP_DELTA
 	{
 	  move lr, zero   /* stop backtraces in the called function */

arch/tile/kernel/irq.c

@@ -61,9 +61,9 @@ static DEFINE_SPINLOCK(available_irqs_lock);
 
 #if CHIP_HAS_IPI()
 /* Use SPRs to manipulate device interrupts. */
-#define mask_irqs(irq_mask) __insn_mtspr(SPR_IPI_MASK_SET_1, irq_mask)
-#define unmask_irqs(irq_mask) __insn_mtspr(SPR_IPI_MASK_RESET_1, irq_mask)
-#define clear_irqs(irq_mask) __insn_mtspr(SPR_IPI_EVENT_RESET_1, irq_mask)
+#define mask_irqs(irq_mask) __insn_mtspr(SPR_IPI_MASK_SET_K, irq_mask)
+#define unmask_irqs(irq_mask) __insn_mtspr(SPR_IPI_MASK_RESET_K, irq_mask)
+#define clear_irqs(irq_mask) __insn_mtspr(SPR_IPI_EVENT_RESET_K, irq_mask)
 #else
 /* Use HV to manipulate device interrupts. */
 #define mask_irqs(irq_mask) hv_disable_intr(irq_mask)
@@ -89,16 +89,16 @@ void tile_dev_intr(struct pt_regs *regs, int intnum)
 	 * masked by a previous interrupt.  Then, mask out the ones
 	 * we're going to handle.
 	 */
-	unsigned long masked = __insn_mfspr(SPR_IPI_MASK_1);
-	original_irqs = __insn_mfspr(SPR_IPI_EVENT_1) & ~masked;
-	__insn_mtspr(SPR_IPI_MASK_SET_1, original_irqs);
+	unsigned long masked = __insn_mfspr(SPR_IPI_MASK_K);
+	original_irqs = __insn_mfspr(SPR_IPI_EVENT_K) & ~masked;
+	__insn_mtspr(SPR_IPI_MASK_SET_K, original_irqs);
 #else
 	/*
 	 * Hypervisor performs the equivalent of the Gx code above and
 	 * then puts the pending interrupt mask into a system save reg
 	 * for us to find.
 	 */
-	original_irqs = __insn_mfspr(SPR_SYSTEM_SAVE_1_3);
+	original_irqs = __insn_mfspr(SPR_SYSTEM_SAVE_K_3);
 #endif
 	remaining_irqs = original_irqs;
 
@@ -225,7 +225,7 @@ void __cpuinit setup_irq_regs(void)
 	/* Enable interrupt delivery. */
 	unmask_irqs(~0UL);
 #if CHIP_HAS_IPI()
-	raw_local_irq_unmask(INT_IPI_1);
+	raw_local_irq_unmask(INT_IPI_K);
 #endif
 }
 

arch/tile/kernel/messaging.c

@@ -34,7 +34,7 @@ void __cpuinit init_messaging(void)
 		panic("hv_register_message_state: error %d", rc);
 
 	/* Make sure downcall interrupts will be enabled. */
-	raw_local_irq_unmask(INT_INTCTRL_1);
+	raw_local_irq_unmask(INT_INTCTRL_K);
 }
 
 void hv_message_intr(struct pt_regs *regs, int intnum)

arch/tile/kernel/process.c

@@ -214,9 +214,10 @@ int copy_thread(unsigned long clone_flags, unsigned long sp,
 	/*
 	 * Copy the callee-saved registers from the passed pt_regs struct
 	 * into the context-switch callee-saved registers area.
-	 * We have to restore the callee-saved registers since we may
-	 * be cloning a userspace task with userspace register state,
-	 * and we won't be unwinding the same kernel frames to restore them.
+	 * This way when we start the interrupt-return sequence, the
+	 * callee-save registers will be correctly in registers, which
+	 * is how we assume the compiler leaves them as we start doing
+	 * the normal return-from-interrupt path after calling C code.
 	 * Zero out the C ABI save area to mark the top of the stack.
 	 */
 	ksp = (unsigned long) childregs;
@@ -304,15 +305,25 @@ int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs)
 /* Allow user processes to access the DMA SPRs */
 void grant_dma_mpls(void)
 {
+#if CONFIG_KERNEL_PL == 2
+	__insn_mtspr(SPR_MPL_DMA_CPL_SET_1, 1);
+	__insn_mtspr(SPR_MPL_DMA_NOTIFY_SET_1, 1);
+#else
 	__insn_mtspr(SPR_MPL_DMA_CPL_SET_0, 1);
 	__insn_mtspr(SPR_MPL_DMA_NOTIFY_SET_0, 1);
+#endif
 }
 
 /* Forbid user processes from accessing the DMA SPRs */
 void restrict_dma_mpls(void)
 {
+#if CONFIG_KERNEL_PL == 2
+	__insn_mtspr(SPR_MPL_DMA_CPL_SET_2, 1);
+	__insn_mtspr(SPR_MPL_DMA_NOTIFY_SET_2, 1);
+#else
 	__insn_mtspr(SPR_MPL_DMA_CPL_SET_1, 1);
 	__insn_mtspr(SPR_MPL_DMA_NOTIFY_SET_1, 1);
+#endif
 }
 
 /* Pause the DMA engine, then save off its state registers. */
@@ -523,19 +534,14 @@ struct task_struct *__sched _switch_to(struct task_struct *prev,
 	 * Switch kernel SP, PC, and callee-saved registers.
 	 * In the context of the new task, return the old task pointer
 	 * (i.e. the task that actually called __switch_to).
-	 * Pass the value to use for SYSTEM_SAVE_1_0 when we reset our sp.
+	 * Pass the value to use for SYSTEM_SAVE_K_0 when we reset our sp.
 	 */
 	return __switch_to(prev, next, next_current_ksp0(next));
 }
 
-long _sys_fork(struct pt_regs *regs)
-{
-	return do_fork(SIGCHLD, regs->sp, regs, 0, NULL, NULL);
-}
-
-long _sys_clone(unsigned long clone_flags, unsigned long newsp,
-		void __user *parent_tidptr, void __user *child_tidptr,
-		struct pt_regs *regs)
+SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
+		void __user *, parent_tidptr, void __user *, child_tidptr,
+		struct pt_regs *, regs)
 {
 	if (!newsp)
 		newsp = regs->sp;
@@ -543,18 +549,13 @@ long _sys_clone(unsigned long clone_flags, unsigned long newsp,
 		       parent_tidptr, child_tidptr);
 }
 
-long _sys_vfork(struct pt_regs *regs)
-{
-	return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->sp,
-		       regs, 0, NULL, NULL);
-}
-
 /*
  * sys_execve() executes a new program.
  */
-long _sys_execve(const char __user *path,
-		 const char __user *const __user *argv,
-		 const char __user *const __user *envp, struct pt_regs *regs)
+SYSCALL_DEFINE4(execve, const char __user *, path,
+		const char __user *const __user *, argv,
+		const char __user *const __user *, envp,
+		struct pt_regs *, regs)
 {
 	long error;
 	char *filename;
@@ -570,9 +571,10 @@ long _sys_execve(const char __user *path,
 }
 
 #ifdef CONFIG_COMPAT
-long _compat_sys_execve(const char __user *path,
+long compat_sys_execve(const char __user *path,
 		       const compat_uptr_t __user *argv,
-			const compat_uptr_t __user *envp, struct pt_regs *regs)
+		       const compat_uptr_t __user *envp,
+		       struct pt_regs *regs)
 {
 	long error;
 	char *filename;

arch/tile/kernel/ptrace.c

@@ -31,25 +31,6 @@ void user_disable_single_step(struct task_struct *child)
 	clear_tsk_thread_flag(child, TIF_SINGLESTEP);
 }
 
-/*
- * This routine will put a word on the process's privileged stack.
- */
-static void putreg(struct task_struct *task,
-		   unsigned long addr, unsigned long value)
-{
-	unsigned int regno = addr / sizeof(unsigned long);
-	struct pt_regs *childregs = task_pt_regs(task);
-	childregs->regs[regno] = value;
-	childregs->flags |= PT_FLAGS_RESTORE_REGS;
-}
-
-static unsigned long getreg(struct task_struct *task, unsigned long addr)
-{
-	unsigned int regno = addr / sizeof(unsigned long);
-	struct pt_regs *childregs = task_pt_regs(task);
-	return childregs->regs[regno];
-}
-
 /*
  * Called by kernel/ptrace.c when detaching..
  */
@@ -66,59 +47,72 @@ void ptrace_disable(struct task_struct *child)
 
 long arch_ptrace(struct task_struct *child, long request, long addr, long data)
 {
-	unsigned long __user *datap;
+	unsigned long __user *datap = (long __user __force *)data;
 	unsigned long tmp;
 	int i;
 	long ret = -EIO;
-
-#ifdef CONFIG_COMPAT
-	if (task_thread_info(current)->status & TS_COMPAT)
-		data = (u32)data;
-	if (task_thread_info(child)->status & TS_COMPAT)
-		addr = (u32)addr;
-#endif
-	datap = (unsigned long __user __force *)data;
+	unsigned long *childregs;
+	char *childreg;
 
 	switch (request) {
 
 	case PTRACE_PEEKUSR:  /* Read register from pt_regs. */
-		if (addr & (sizeof(data)-1))
-			break;
 		if (addr < 0 || addr >= PTREGS_SIZE)
 			break;
-		tmp = getreg(child, addr);   /* Read register */
-		ret = put_user(tmp, datap);
+		childreg = (char *)task_pt_regs(child) + addr;
+#ifdef CONFIG_COMPAT
+		if (is_compat_task()) {
+			if (addr & (sizeof(compat_long_t)-1))
+				break;
+			ret = put_user(*(compat_long_t *)childreg,
+				       (compat_long_t __user *)datap);
+		} else
+#endif
+		{
+			if (addr & (sizeof(long)-1))
+				break;
+			ret = put_user(*(long *)childreg, datap);
+		}
 		break;
 
 	case PTRACE_POKEUSR:  /* Write register in pt_regs. */
-		if (addr & (sizeof(data)-1))
-			break;
 		if (addr < 0 || addr >= PTREGS_SIZE)
 			break;
-		putreg(child, addr, data);   /* Write register */
+		childreg = (char *)task_pt_regs(child) + addr;
+#ifdef CONFIG_COMPAT
+		if (is_compat_task()) {
+			if (addr & (sizeof(compat_long_t)-1))
+				break;
+			*(compat_long_t *)childreg = data;
+		} else
+#endif
+		{
+			if (addr & (sizeof(long)-1))
+				break;
+			*(long *)childreg = data;
+		}
 		ret = 0;
 		break;
 
 	case PTRACE_GETREGS:  /* Get all registers from the child. */
 		if (!access_ok(VERIFY_WRITE, datap, PTREGS_SIZE))
 			break;
-		for (i = 0; i < PTREGS_SIZE; i += sizeof(long)) {
-			ret = __put_user(getreg(child, i), datap);
+		childregs = (long *)task_pt_regs(child);
+		for (i = 0; i < sizeof(struct pt_regs)/sizeof(long); ++i) {
+			ret = __put_user(childregs[i], &datap[i]);
 			if (ret != 0)
 				break;
-			datap++;
 		}
 		break;
 
 	case PTRACE_SETREGS:  /* Set all registers in the child. */
 		if (!access_ok(VERIFY_READ, datap, PTREGS_SIZE))
 			break;
-		for (i = 0; i < PTREGS_SIZE; i += sizeof(long)) {
-			ret = __get_user(tmp, datap);
+		childregs = (long *)task_pt_regs(child);
+		for (i = 0; i < sizeof(struct pt_regs)/sizeof(long); ++i) {
+			ret = __get_user(childregs[i], &datap[i]);
 			if (ret != 0)
 				break;
-			putreg(child, i, tmp);
-			datap++;
 		}
 		break;
 

arch/tile/kernel/regs_32.S

@@ -85,7 +85,7 @@ STD_ENTRY_SECTION(__switch_to, .sched.text)
 	{
 	  /* Update sp and ksp0 simultaneously to avoid backtracer warnings. */
 	  move sp, r13
-	  mtspr SYSTEM_SAVE_1_0, r2
+	  mtspr SPR_SYSTEM_SAVE_K_0, r2
 	}
 	FOR_EACH_CALLEE_SAVED_REG(LOAD_REG)
 .L__switch_to_pc:

arch/tile/kernel/setup.c

@@ -187,11 +187,11 @@ early_param("vmalloc", parse_vmalloc);
 
 #ifdef CONFIG_HIGHMEM
 /*
- * Determine for each controller where its lowmem is mapped and how
- * much of it is mapped there.  On controller zero, the first few
- * megabytes are mapped at 0xfd000000 as code, so in principle we
- * could start our data mappings higher up, but for now we don't
- * bother, to avoid additional confusion.
+ * Determine for each controller where its lowmem is mapped and how much of
+ * it is mapped there.  On controller zero, the first few megabytes are
+ * already mapped in as code at MEM_SV_INTRPT, so in principle we could
+ * start our data mappings higher up, but for now we don't bother, to avoid
+ * additional confusion.
  *
  * One question is whether, on systems with more than 768 Mb and
  * controllers of different sizes, to map in a proportionate amount of
@@ -311,7 +311,7 @@ static void __init setup_memory(void)
 #endif
 
 	/* We are using a char to hold the cpu_2_node[] mapping */
-	BUG_ON(MAX_NUMNODES > 127);
+	BUILD_BUG_ON(MAX_NUMNODES > 127);
 
 	/* Discover the ranges of memory available to us */
 	for (i = 0; ; ++i) {
@@ -876,6 +876,9 @@ void __cpuinit setup_cpu(int boot)
 #if CHIP_HAS_SN_PROC()
 	raw_local_irq_unmask(INT_SNITLB_MISS);
 #endif
+#ifdef __tilegx__
+	raw_local_irq_unmask(INT_SINGLE_STEP_K);
+#endif
 
 	/*
 	 * Allow user access to many generic SPRs, like the cycle
@@ -893,11 +896,12 @@ void __cpuinit setup_cpu(int boot)
 #endif
 
 	/*
-	 * Set the MPL for interrupt control 0 to user level.
-	 * This includes access to the SYSTEM_SAVE and EX_CONTEXT SPRs,
-	 * as well as the PL 0 interrupt mask.
+	 * Set the MPL for interrupt control 0 & 1 to the corresponding
+	 * values.  This includes access to the SYSTEM_SAVE and EX_CONTEXT
+	 * SPRs, as well as the interrupt mask.
 	 */
 	__insn_mtspr(SPR_MPL_INTCTRL_0_SET_0, 1);
+	__insn_mtspr(SPR_MPL_INTCTRL_1_SET_1, 1);
 
 	/* Initialize IRQ support for this cpu. */
 	setup_irq_regs();
@@ -1033,7 +1037,7 @@ static void __init validate_va(void)
 	 * In addition, make sure we CAN'T use the end of memory, since
 	 * we use the last chunk of each pgd for the pgd_list.
 	 */
-	int i, fc_fd_ok = 0;
+	int i, user_kernel_ok = 0;
 	unsigned long max_va = 0;
 	unsigned long list_va =
 		((PGD_LIST_OFFSET / sizeof(pgd_t)) << PGDIR_SHIFT);
@@ -1044,13 +1048,13 @@ static void __init validate_va(void)
 			break;
 		if (range.start <= MEM_USER_INTRPT &&
 		    range.start + range.size >= MEM_HV_INTRPT)
-			fc_fd_ok = 1;
+			user_kernel_ok = 1;
 		if (range.start == 0)
 			max_va = range.size;
 		BUG_ON(range.start + range.size > list_va);
 	}
-	if (!fc_fd_ok)
-		early_panic("Hypervisor not configured for VAs 0xfc/0xfd\n");
+	if (!user_kernel_ok)
+		early_panic("Hypervisor not configured for user/kernel VAs\n");
 	if (max_va == 0)
 		early_panic("Hypervisor not configured for low VAs\n");
 	if (max_va < KERNEL_HIGH_VADDR)
@@ -1334,6 +1338,10 @@ static void __init pcpu_fc_populate_pte(unsigned long addr)
 	pte_t *pte;
 
 	BUG_ON(pgd_addr_invalid(addr));
+	if (addr < VMALLOC_START || addr >= VMALLOC_END)
+		panic("PCPU addr %#lx outside vmalloc range %#lx..%#lx;"
+		      " try increasing CONFIG_VMALLOC_RESERVE\n",
+		      addr, VMALLOC_START, VMALLOC_END);
 
 	pgd = swapper_pg_dir + pgd_index(addr);
 	pud = pud_offset(pgd, addr);

arch/tile/kernel/signal.c

@@ -41,8 +41,8 @@
 #define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
 
 
-long _sys_sigaltstack(const stack_t __user *uss,
-		      stack_t __user *uoss, struct pt_regs *regs)
+SYSCALL_DEFINE3(sigaltstack, const stack_t __user *, uss,
+		stack_t __user *, uoss, struct pt_regs *, regs)
 {
 	return do_sigaltstack(uss, uoss, regs->sp);
 }
@@ -78,7 +78,7 @@ int restore_sigcontext(struct pt_regs *regs,
 }
 
 /* sigreturn() returns long since it restores r0 in the interrupted code. */
-long _sys_rt_sigreturn(struct pt_regs *regs)
+SYSCALL_DEFINE1(rt_sigreturn, struct pt_regs *, regs)
 {
 	struct rt_sigframe __user *frame =
 		(struct rt_sigframe __user *)(regs->sp);

arch/tile/kernel/single_step.c

@@ -15,7 +15,7 @@
  * Derived from iLib's single-stepping code.
  */
 
-#ifndef __tilegx__   /* No support for single-step yet. */
+#ifndef __tilegx__   /* Hardware support for single step unavailable. */
 
 /* These functions are only used on the TILE platform */
 #include <linux/slab.h>
@@ -660,4 +660,75 @@ void single_step_once(struct pt_regs *regs)
 		regs->pc += 8;
 }
 
+#else
+#include <linux/smp.h>
+#include <linux/ptrace.h>
+#include <arch/spr_def.h>
+
+static DEFINE_PER_CPU(unsigned long, ss_saved_pc);
+
+
+/*
+ * Called directly on the occasion of an interrupt.
+ *
+ * If the process doesn't have single step set, then we use this as an
+ * opportunity to turn single step off.
+ *
+ * It has been mentioned that we could conditionally turn off single stepping
+ * on each entry into the kernel and rely on single_step_once to turn it
+ * on for the processes that matter (as we already do), but this
+ * implementation is somewhat more efficient in that we muck with registers
+ * once on a bum interrupt rather than on every entry into the kernel.
+ *
+ * If SINGLE_STEP_CONTROL_K has CANCELED set, then an interrupt occurred,
+ * so we have to run through this process again before we can say that an
+ * instruction has executed.
+ *
+ * swint will set CANCELED, but it's a legitimate instruction.  Fortunately
+ * it changes the PC.  If it hasn't changed, then we know that the interrupt
+ * wasn't generated by swint and we'll need to run this process again before
+ * we can say an instruction has executed.
+ *
+ * If either CANCELED == 0 or the PC's changed, we send out SIGTRAPs and get
+ * on with our lives.
+ */
+
+void gx_singlestep_handle(struct pt_regs *regs, int fault_num)
+{
+	unsigned long *ss_pc = &__get_cpu_var(ss_saved_pc);
+	struct thread_info *info = (void *)current_thread_info();
+	int is_single_step = test_ti_thread_flag(info, TIF_SINGLESTEP);
+	unsigned long control = __insn_mfspr(SPR_SINGLE_STEP_CONTROL_K);
+
+	if (is_single_step == 0) {
+		__insn_mtspr(SPR_SINGLE_STEP_EN_K_K, 0);
+
+	} else if ((*ss_pc != regs->pc) ||
+		   (!(control & SPR_SINGLE_STEP_CONTROL_1__CANCELED_MASK))) {
+
+		ptrace_notify(SIGTRAP);
+		control |= SPR_SINGLE_STEP_CONTROL_1__CANCELED_MASK;
+		control |= SPR_SINGLE_STEP_CONTROL_1__INHIBIT_MASK;
+		__insn_mtspr(SPR_SINGLE_STEP_CONTROL_K, control);
+	}
+}
+
+
+/*
+ * Called from need_singlestep.  Set up the control registers and the enable
+ * register, then return back.
+ */
+
+void single_step_once(struct pt_regs *regs)
+{
+	unsigned long *ss_pc = &__get_cpu_var(ss_saved_pc);
+	unsigned long control = __insn_mfspr(SPR_SINGLE_STEP_CONTROL_K);
+
+	*ss_pc = regs->pc;
+	control |= SPR_SINGLE_STEP_CONTROL_1__CANCELED_MASK;
+	control |= SPR_SINGLE_STEP_CONTROL_1__INHIBIT_MASK;
+	__insn_mtspr(SPR_SINGLE_STEP_CONTROL_K, control);
+	__insn_mtspr(SPR_SINGLE_STEP_EN_K_K, 1 << USER_PL);
+}
+
 #endif /* !__tilegx__ */

arch/tile/kernel/smp.c

@@ -212,7 +212,7 @@ void __init ipi_init(void)
 
 		tile.x = cpu_x(cpu);
 		tile.y = cpu_y(cpu);
-		if (hv_get_ipi_pte(tile, 1, &pte) != 0)
+		if (hv_get_ipi_pte(tile, KERNEL_PL, &pte) != 0)
 			panic("Failed to initialize IPI for cpu %d\n", cpu);
 
 		offset = hv_pte_get_pfn(pte) << PAGE_SHIFT;

arch/tile/kernel/stack.c

@@ -30,6 +30,10 @@
 #include <arch/abi.h>
 #include <arch/interrupts.h>
 
+#define KBT_ONGOING	0  /* Backtrace still ongoing */
+#define KBT_DONE	1  /* Backtrace cleanly completed */
+#define KBT_RUNNING	2  /* Can't run backtrace on a running task */
+#define KBT_LOOP	3  /* Backtrace entered a loop */
 
 /* Is address on the specified kernel stack? */
 static int in_kernel_stack(struct KBacktraceIterator *kbt, VirtualAddress sp)
@@ -207,11 +211,11 @@ static int KBacktraceIterator_next_item_inclusive(
 	for (;;) {
 		do {
 			if (!KBacktraceIterator_is_sigreturn(kbt))
-				return 1;
+				return KBT_ONGOING;
 		} while (backtrace_next(&kbt->it));
 
 		if (!KBacktraceIterator_restart(kbt))
-			return 0;
+			return KBT_DONE;
 	}
 }
 
@@ -264,7 +268,7 @@ void KBacktraceIterator_init(struct KBacktraceIterator *kbt,
 	kbt->pgtable = NULL;
 	kbt->verbose = 0;   /* override in caller if desired */
 	kbt->profile = 0;   /* override in caller if desired */
-	kbt->end = 0;
+	kbt->end = KBT_ONGOING;
 	kbt->new_context = 0;
 	if (is_current) {
 		HV_PhysAddr pgdir_pa = hv_inquire_context().page_table;
@@ -290,7 +294,7 @@ void KBacktraceIterator_init(struct KBacktraceIterator *kbt,
 	if (regs == NULL) {
 		if (is_current || t->state == TASK_RUNNING) {
 			/* Can't do this; we need registers */
-			kbt->end = 1;
+			kbt->end = KBT_RUNNING;
 			return;
 		}
 		pc = get_switch_to_pc();
@@ -305,26 +309,29 @@ void KBacktraceIterator_init(struct KBacktraceIterator *kbt,
 	}
 
 	backtrace_init(&kbt->it, read_memory_func, kbt, pc, lr, sp, r52);
-	kbt->end = !KBacktraceIterator_next_item_inclusive(kbt);
+	kbt->end = KBacktraceIterator_next_item_inclusive(kbt);
 }
 EXPORT_SYMBOL(KBacktraceIterator_init);
 
 int KBacktraceIterator_end(struct KBacktraceIterator *kbt)
 {
-	return kbt->end;
+	return kbt->end != KBT_ONGOING;
 }
 EXPORT_SYMBOL(KBacktraceIterator_end);
 
 void KBacktraceIterator_next(struct KBacktraceIterator *kbt)
 {
+	VirtualAddress old_pc = kbt->it.pc, old_sp = kbt->it.sp;
 	kbt->new_context = 0;
-	if (!backtrace_next(&kbt->it) &&
-	    !KBacktraceIterator_restart(kbt)) {
-			kbt->end = 1;
+	if (!backtrace_next(&kbt->it) && !KBacktraceIterator_restart(kbt)) {
+		kbt->end = KBT_DONE;
 		return;
 	}
-
-	kbt->end = !KBacktraceIterator_next_item_inclusive(kbt);
+	kbt->end = KBacktraceIterator_next_item_inclusive(kbt);
+	if (old_pc == kbt->it.pc && old_sp == kbt->it.sp) {
+		/* Trapped in a loop; give up. */
+		kbt->end = KBT_LOOP;
+	}
 }
 EXPORT_SYMBOL(KBacktraceIterator_next);
 
@@ -387,6 +394,8 @@ void tile_show_stack(struct KBacktraceIterator *kbt, int headers)
 			break;
 		}
 	}
+	if (kbt->end == KBT_LOOP)
+		pr_err("Stack dump stopped; next frame identical to this one\n");
 	if (headers)
 		pr_err("Stack dump complete\n");
 }

arch/tile/kernel/sys.c

@@ -110,6 +110,15 @@ SYSCALL_DEFINE6(mmap, unsigned long, addr, unsigned long, len,
 #define sys_sync_file_range sys_sync_file_range2
 #endif
 
+/* Call the trampolines to manage pt_regs where necessary. */
+#define sys_execve _sys_execve
+#define sys_sigaltstack _sys_sigaltstack
+#define sys_rt_sigreturn _sys_rt_sigreturn
+#define sys_clone _sys_clone
+#ifndef __tilegx__
+#define sys_cmpxchg_badaddr _sys_cmpxchg_badaddr
+#endif
+
 /*
  * Note that we can't include <linux/unistd.h> here since the header
  * guard will defeat us; <asm/unistd.h> checks for __SYSCALL as well.

arch/tile/kernel/traps.c

@@ -260,7 +260,7 @@ void __kprobes do_trap(struct pt_regs *regs, int fault_num,
 		address = regs->pc;
 		break;
 	case INT_UNALIGN_DATA:
-#ifndef __tilegx__  /* FIXME: GX: no single-step yet */
+#ifndef __tilegx__  /* Emulated support for single step debugging */
 		if (unaligned_fixup >= 0) {
 			struct single_step_state *state =
 				current_thread_info()->step_state;
@@ -278,7 +278,7 @@ void __kprobes do_trap(struct pt_regs *regs, int fault_num,
 	case INT_DOUBLE_FAULT:
 		/*
 		 * For double fault, "reason" is actually passed as
-		 * SYSTEM_SAVE_1_2, the hypervisor's double-fault info, so
+		 * SYSTEM_SAVE_K_2, the hypervisor's double-fault info, so
 		 * we can provide the original fault number rather than
 		 * the uninteresting "INT_DOUBLE_FAULT" so the user can
 		 * learn what actually struck while PL0 ICS was set.

arch/tile/kvm/Kconfig

+#
+# KVM configuration
+#
+
+source "virt/kvm/Kconfig"
+
+menuconfig VIRTUALIZATION
+	bool "Virtualization"
+	---help---
+	  Say Y here to get to see options for using your Linux host to run
+	  other operating systems inside virtual machines (guests).
+	  This option alone does not add any kernel code.
+
+	  If you say N, all options in this submenu will be skipped and
+	  disabled.
+
+if VIRTUALIZATION
+
+config KVM
+	tristate "Kernel-based Virtual Machine (KVM) support"
+	depends on HAVE_KVM && MODULES && EXPERIMENTAL
+	select PREEMPT_NOTIFIERS
+	select ANON_INODES
+	---help---
+	  Support hosting paravirtualized guest machines.
+
+	  This module provides access to the hardware capabilities through
+	  a character device node named /dev/kvm.
+
+	  To compile this as a module, choose M here: the module
+	  will be called kvm.
+
+	  If unsure, say N.
+
+source drivers/vhost/Kconfig
+source drivers/virtio/Kconfig
+
+endif # VIRTUALIZATION

arch/tile/lib/Makefile

@@ -3,8 +3,8 @@
 #
 
 lib-y = cacheflush.o checksum.o cpumask.o delay.o \
-	mb_incoherent.o uaccess.o \
-	memcpy_$(BITS).o memchr_$(BITS).o memmove_$(BITS).o memset_$(BITS).o \
+	mb_incoherent.o uaccess.o memmove.o \
+	memcpy_$(BITS).o memchr_$(BITS).o memset_$(BITS).o \
 	strchr_$(BITS).o strlen_$(BITS).o
 
 ifeq ($(CONFIG_TILEGX),y)

arch/tile/lib/atomic_32.c

@@ -300,7 +300,7 @@ void __init __init_atomic_per_cpu(void)
 #else /* ATOMIC_LOCKS_FOUND_VIA_TABLE() */
 
 	/* Validate power-of-two and "bigger than cpus" assumption */
-	BUG_ON(ATOMIC_HASH_SIZE & (ATOMIC_HASH_SIZE-1));
+	BUILD_BUG_ON(ATOMIC_HASH_SIZE & (ATOMIC_HASH_SIZE-1));
 	BUG_ON(ATOMIC_HASH_SIZE < nr_cpu_ids);
 
 	/*
@@ -314,17 +314,17 @@ void __init __init_atomic_per_cpu(void)
 	BUG_ON((unsigned long)atomic_locks % PAGE_SIZE != 0);
 
 	/* The locks must all fit on one page. */
-	BUG_ON(ATOMIC_HASH_SIZE * sizeof(int) > PAGE_SIZE);
+	BUILD_BUG_ON(ATOMIC_HASH_SIZE * sizeof(int) > PAGE_SIZE);
 
 	/*
 	 * We use the page offset of the atomic value's address as
 	 * an index into atomic_locks, excluding the low 3 bits.
 	 * That should not produce more indices than ATOMIC_HASH_SIZE.
 	 */
-	BUG_ON((PAGE_SIZE >> 3) > ATOMIC_HASH_SIZE);
+	BUILD_BUG_ON((PAGE_SIZE >> 3) > ATOMIC_HASH_SIZE);
 
 #endif /* ATOMIC_LOCKS_FOUND_VIA_TABLE() */
 
 	/* The futex code makes this assumption, so we validate it here. */
-	BUG_ON(sizeof(atomic_t) != sizeof(int));
+	BUILD_BUG_ON(sizeof(atomic_t) != sizeof(int));
 }

arch/tile/lib/exports.c

@@ -45,6 +45,9 @@ EXPORT_SYMBOL(__copy_from_user_zeroing);
 EXPORT_SYMBOL(__copy_in_user_inatomic);
 #endif
 
+/* arch/tile/lib/mb_incoherent.S */
+EXPORT_SYMBOL(__mb_incoherent);
+
 /* hypervisor glue */
 #include <hv/hypervisor.h>
 EXPORT_SYMBOL(hv_dev_open);

arch/tile/lib/memcpy_32.S

@@ -10,14 +10,16 @@
  *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
  *   NON INFRINGEMENT.  See the GNU General Public License for
  *   more details.
- *
- * This file shares the implementation of the userspace memcpy and
- * the kernel's memcpy, copy_to_user and copy_from_user.
  */
 
 #include <arch/chip.h>
 
 
+/*
+ * This file shares the implementation of the userspace memcpy and
+ * the kernel's memcpy, copy_to_user and copy_from_user.
+ */
+
 #include <linux/linkage.h>
 
 /* On TILE64, we wrap these functions via arch/tile/lib/memcpy_tile64.c */

arch/tile/lib/memset_32.c

@@ -18,6 +18,7 @@
 #include <linux/string.h>
 #include <linux/module.h>
 
+#undef memset
 
 void *memset(void *s, int c, size_t n)
 {

arch/tile/lib/strlen_32.c

@@ -16,6 +16,8 @@
 #include <linux/string.h>
 #include <linux/module.h>
 
+#undef strlen
+
 size_t strlen(const char *s)
 {
 	/* Get an aligned pointer. */

arch/tile/mm/fault.c

@@ -66,10 +66,10 @@ static noinline void force_sig_info_fault(int si_signo, int si_code,
 #ifndef __tilegx__
 /*
  * Synthesize the fault a PL0 process would get by doing a word-load of
- * an unaligned address or a high kernel address.  Called indirectly
- * from sys_cmpxchg() in kernel/intvec.S.
+ * an unaligned address or a high kernel address.
  */
-int _sys_cmpxchg_badaddr(unsigned long address, struct pt_regs *regs)
+SYSCALL_DEFINE2(cmpxchg_badaddr, unsigned long, address,
+		struct pt_regs *, regs)
 {
 	if (address >= PAGE_OFFSET)
 		force_sig_info_fault(SIGSEGV, SEGV_MAPERR, address,
@@ -563,10 +563,10 @@ static int handle_page_fault(struct pt_regs *regs,
 /*
  * When we take an ITLB or DTLB fault or access violation in the
  * supervisor while the critical section bit is set, the hypervisor is
- * reluctant to write new values into the EX_CONTEXT_1_x registers,
+ * reluctant to write new values into the EX_CONTEXT_K_x registers,
  * since that might indicate we have not yet squirreled the SPR
  * contents away and can thus safely take a recursive interrupt.
- * Accordingly, the hypervisor passes us the PC via SYSTEM_SAVE_1_2.
+ * Accordingly, the hypervisor passes us the PC via SYSTEM_SAVE_K_2.
  *
  * Note that this routine is called before homecache_tlb_defer_enter(),
  * which means that we can properly unlock any atomics that might
@@ -610,7 +610,7 @@ struct intvec_state do_page_fault_ics(struct pt_regs *regs, int fault_num,
 	 * fault.  We didn't set up a kernel stack on initial entry to
 	 * sys_cmpxchg, but instead had one set up by the fault, which
 	 * (because sys_cmpxchg never releases ICS) came to us via the
-	 * SYSTEM_SAVE_1_2 mechanism, and thus EX_CONTEXT_1_[01] are
+	 * SYSTEM_SAVE_K_2 mechanism, and thus EX_CONTEXT_K_[01] are
 	 * still referencing the original user code.  We release the
 	 * atomic lock and rewrite pt_regs so that it appears that we
 	 * came from user-space directly, and after we finish the

arch/tile/mm/homecache.c

@@ -37,6 +37,8 @@
 #include <asm/pgalloc.h>
 #include <asm/homecache.h>
 
+#include <arch/sim.h>
+
 #include "migrate.h"
 
 
@@ -217,13 +219,6 @@ static unsigned long cache_flush_length(unsigned long length)
 	return (length >= CHIP_L2_CACHE_SIZE()) ? HV_FLUSH_EVICT_L2 : length;
 }
 
-/* On the simulator, confirm lines have been evicted everywhere. */
-static void validate_lines_evicted(unsigned long pfn, size_t length)
-{
-	sim_syscall(SIM_SYSCALL_VALIDATE_LINES_EVICTED,
-		    (HV_PhysAddr)pfn << PAGE_SHIFT, length);
-}
-
 /* Flush a page out of whatever cache(s) it is in. */
 void homecache_flush_cache(struct page *page, int order)
 {
@@ -234,7 +229,7 @@ void homecache_flush_cache(struct page *page, int order)
 
 	homecache_mask(page, pages, &home_mask);
 	flush_remote(pfn, length, &home_mask, 0, 0, 0, NULL, NULL, 0);
-	validate_lines_evicted(pfn, pages * PAGE_SIZE);
+	sim_validate_lines_evicted(PFN_PHYS(pfn), pages * PAGE_SIZE);
 }
 
 

arch/tile/mm/init.c

@@ -1060,7 +1060,7 @@ void free_initmem(void)
 
 	/*
 	 * Free the pages mapped from 0xc0000000 that correspond to code
-	 * pages from 0xfd000000 that we won't use again after init.
+	 * pages from MEM_SV_INTRPT that we won't use again after init.
 	 */
 	free_init_pages("unused kernel text",
 			(unsigned long)_sinittext - text_delta,

drivers/char/hvc_tile.c

@@ -61,7 +61,8 @@ console_initcall(hvc_tile_console_init);
 
 static int __init hvc_tile_init(void)
 {
-	hvc_alloc(0, 0, &hvc_tile_get_put_ops, 128);
-	return 0;
+	struct hvc_struct *s;
+	s = hvc_alloc(0, 0, &hvc_tile_get_put_ops, 128);
+	return IS_ERR(s) ? PTR_ERR(s) : 0;
 }
 device_initcall(hvc_tile_init);

lib/Kconfig.debug

@@ -374,7 +374,7 @@ config SLUB_STATS
 config DEBUG_KMEMLEAK
 	bool "Kernel memory leak detector"
 	depends on DEBUG_KERNEL && EXPERIMENTAL && !MEMORY_HOTPLUG && \
-		(X86 || ARM || PPC || S390 || SPARC64 || SUPERH || MICROBLAZE)
+		(X86 || ARM || PPC || S390 || SPARC64 || SUPERH || MICROBLAZE || TILE)
 
 	select DEBUG_FS if SYSFS
 	select STACKTRACE if STACKTRACE_SUPPORT