From 489f781a5936f4b90cd5c9838e0102933f6eb2b5 Mon Sep 17 00:00:00 2001
From: Catalin Marinas <catalin.marinas@arm.com>
Date: Tue, 23 Oct 2012 14:24:21 +0100
Subject: arm64: Use pgprot_t as the last argument when invoking __ioremap()

Even if it works with since the types have the same size, the correct
type of the last __ioremap() argument is pgprot_t rather than pteval_t.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
---
 arch/arm64/include/asm/io.h | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

(limited to 'arch')

diff --git a/arch/arm64/include/asm/io.h b/arch/arm64/include/asm/io.h
index 74a2a7d304a..d10778fe91e 100644
--- a/arch/arm64/include/asm/io.h
+++ b/arch/arm64/include/asm/io.h
@@ -225,9 +225,9 @@ extern void __iounmap(volatile void __iomem *addr);
 #define PROT_DEVICE_nGnRE	(PROT_DEFAULT | PTE_XN | PTE_ATTRINDX(MT_DEVICE_nGnRE))
 #define PROT_NORMAL_NC		(PROT_DEFAULT | PTE_ATTRINDX(MT_NORMAL_NC))
 
-#define ioremap(addr, size)		__ioremap((addr), (size), PROT_DEVICE_nGnRE)
-#define ioremap_nocache(addr, size)	__ioremap((addr), (size), PROT_DEVICE_nGnRE)
-#define ioremap_wc(addr, size)		__ioremap((addr), (size), PROT_NORMAL_NC)
+#define ioremap(addr, size)		__ioremap((addr), (size), __pgprot(PROT_DEVICE_nGnRE))
+#define ioremap_nocache(addr, size)	__ioremap((addr), (size), __pgprot(PROT_DEVICE_nGnRE))
+#define ioremap_wc(addr, size)		__ioremap((addr), (size), __pgprot(PROT_NORMAL_NC))
 #define iounmap				__iounmap
 
 #define ARCH_HAS_IOREMAP_WC
-- 
cgit v1.2.3-18-g5258


From e3978cded41dc7b364e74037f56d6bc558c11fd7 Mon Sep 17 00:00:00 2001
From: Catalin Marinas <catalin.marinas@arm.com>
Date: Tue, 23 Oct 2012 14:51:16 +0100
Subject: arm64: Move PCI_IOBASE closer to MODULES_VADDR

This is to reuse the same pmd table that is sparsely populated with
the modules space.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
---
 arch/arm64/include/asm/io.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'arch')

diff --git a/arch/arm64/include/asm/io.h b/arch/arm64/include/asm/io.h
index d10778fe91e..54f6116697f 100644
--- a/arch/arm64/include/asm/io.h
+++ b/arch/arm64/include/asm/io.h
@@ -114,7 +114,7 @@ static inline u64 __raw_readq(const volatile void __iomem *addr)
  *  I/O port access primitives.
  */
 #define IO_SPACE_LIMIT		0xffff
-#define PCI_IOBASE		((void __iomem *)0xffffffbbfffe0000UL)
+#define PCI_IOBASE		((void __iomem *)(MODULES_VADDR - SZ_2M))
 
 static inline u8 inb(unsigned long addr)
 {
-- 
cgit v1.2.3-18-g5258


From 2f72d4f6a29cf84c40fc05c76020b347b4774393 Mon Sep 17 00:00:00 2001
From: Chris Zankel <chris@zankel.net>
Date: Tue, 23 Oct 2012 20:17:05 -0700
Subject: xtensa: allow multi-inclusion for uapi/unistd.h

Xtensa implements a method that allows to generate a arbitrary output
for each system call by defining the __SYSCALL(number, function, num_args).
This usually requires to include uapi/unistd.h twice. Instead of removing
the guard agains multiple inclusion entirely, allow to include unistd.h again
only if __SYSCALL is defined. Note that __SYSCALL gets always undefined at
the end of the file.

Signed-off-by: Chris Zankel <chris@zankel.net>
---
 arch/xtensa/include/asm/unistd.h      | 14 ++++----------
 arch/xtensa/include/uapi/asm/unistd.h | 14 +++-----------
 arch/xtensa/kernel/syscall.c          |  7 +++----
 3 files changed, 10 insertions(+), 25 deletions(-)

(limited to 'arch')

diff --git a/arch/xtensa/include/asm/unistd.h b/arch/xtensa/include/asm/unistd.h
index 9ef1c31d2c8..d9fa52dbfba 100644
--- a/arch/xtensa/include/asm/unistd.h
+++ b/arch/xtensa/include/asm/unistd.h
@@ -1,16 +1,8 @@
-/*
- * include/asm-xtensa/unistd.h
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- *
- * Copyright (C) 2001 - 2005 Tensilica Inc.
- */
+#ifndef _XTENSA_UNISTD_H
+#define _XTENSA_UNISTD_H
 
 #include <uapi/asm/unistd.h>
 
-
 /*
  * "Conditional" syscalls
  *
@@ -37,3 +29,5 @@
 #define __IGNORE_mmap				/* use mmap2 */
 #define __IGNORE_vfork				/* use clone */
 #define __IGNORE_fadvise64			/* use fadvise64_64 */
+
+#endif /* _XTENSA_UNISTD_H */
diff --git a/arch/xtensa/include/uapi/asm/unistd.h b/arch/xtensa/include/uapi/asm/unistd.h
index 479abaea5aa..1fd157f3121 100644
--- a/arch/xtensa/include/uapi/asm/unistd.h
+++ b/arch/xtensa/include/uapi/asm/unistd.h
@@ -1,14 +1,4 @@
-/*
- * include/asm-xtensa/unistd.h
- *
- * This file is subject to the terms and conditions of the GNU General Public
- * License.  See the file "COPYING" in the main directory of this archive
- * for more details.
- *
- * Copyright (C) 2001 - 2012 Tensilica Inc.
- */
-
-#ifndef _UAPI_XTENSA_UNISTD_H
+#if !defined(_UAPI_XTENSA_UNISTD_H) || defined(__SYSCALL)
 #define _UAPI_XTENSA_UNISTD_H
 
 #ifndef __SYSCALL
@@ -759,4 +749,6 @@ __SYSCALL(331, sys_kcmp, 5)
 
 #define SYS_XTENSA_COUNT                  5     /* count */
 
+#undef __SYSCALL
+
 #endif /* _UAPI_XTENSA_UNISTD_H */
diff --git a/arch/xtensa/kernel/syscall.c b/arch/xtensa/kernel/syscall.c
index a5c01e74d5d..5702065f472 100644
--- a/arch/xtensa/kernel/syscall.c
+++ b/arch/xtensa/kernel/syscall.c
@@ -32,10 +32,8 @@ typedef void (*syscall_t)(void);
 syscall_t sys_call_table[__NR_syscall_count] /* FIXME __cacheline_aligned */= {
 	[0 ... __NR_syscall_count - 1] = (syscall_t)&sys_ni_syscall,
 
-#undef __SYSCALL
 #define __SYSCALL(nr,symbol,nargs) [ nr ] = (syscall_t)symbol,
-#undef  __KERNEL_SYSCALLS__
-#include <asm/unistd.h>
+#include <uapi/asm/unistd.h>
 };
 
 asmlinkage long xtensa_shmat(int shmid, char __user *shmaddr, int shmflg)
@@ -49,7 +47,8 @@ asmlinkage long xtensa_shmat(int shmid, char __user *shmaddr, int shmflg)
 	return (long)ret;
 }
 
-asmlinkage long xtensa_fadvise64_64(int fd, int advice, unsigned long long offset, unsigned long long len)
+asmlinkage long xtensa_fadvise64_64(int fd, int advice,
+		unsigned long long offset, unsigned long long len)
 {
 	return sys_fadvise64_64(fd, offset, len, advice);
 }
-- 
cgit v1.2.3-18-g5258


From cddfcbcd05851a3b720055d40fc61c63ea1c2bd1 Mon Sep 17 00:00:00 2001
From: Chris Zankel <chris@zankel.net>
Date: Tue, 23 Oct 2012 20:25:37 -0700
Subject: xtensa: use physical addresses for bus addresses

Define virt_to_bus and bus_to_virt as virt_to_phys, and phys_to_virt,
       respectively.

Signed-off-by: Chris Zankel <chris@zankel.net>
---
 arch/xtensa/include/asm/io.h | 4 ++++
 1 file changed, 4 insertions(+)

(limited to 'arch')

diff --git a/arch/xtensa/include/asm/io.h b/arch/xtensa/include/asm/io.h
index e6be5b9091c..700c2e6f2d2 100644
--- a/arch/xtensa/include/asm/io.h
+++ b/arch/xtensa/include/asm/io.h
@@ -62,6 +62,10 @@ static inline void __iomem *ioremap(unsigned long offset, unsigned long size)
 static inline void iounmap(volatile void __iomem *addr)
 {
 }
+
+#define virt_to_bus     virt_to_phys
+#define bus_to_virt     phys_to_virt
+
 #endif /* CONFIG_MMU */
 
 /*
-- 
cgit v1.2.3-18-g5258


From 6ebe7da25b7879194fdb5c43ea67b383dd9014d6 Mon Sep 17 00:00:00 2001
From: Chris Zankel <chris@zankel.net>
Date: Wed, 24 Oct 2012 13:15:21 -0700
Subject: xtensa: reset windowbase/windowstart when cloning the VM

When we copy a user thread with CLONE_VM, we also have to reset
windowbase and windowstart to start a pristine stack frame. Otherwise,
overflows can happen using the address 0 as the stack pointer.
Also add a special case for vfork, which continues on the
parent stack until it calls execve. Because this could be a call8, we
need to spill the stack pointer of the previus frame (if still 'live' in
the register file).

Signed-off-by: Chris Zankel <chris@zankel.net>
---
 arch/xtensa/kernel/process.c | 26 +++++++++++++++++++++++++-
 1 file changed, 25 insertions(+), 1 deletion(-)

(limited to 'arch')

diff --git a/arch/xtensa/kernel/process.c b/arch/xtensa/kernel/process.c
index 1908f6642d3..341c151af91 100644
--- a/arch/xtensa/kernel/process.c
+++ b/arch/xtensa/kernel/process.c
@@ -220,8 +220,32 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
 	if (user_mode(regs)) {
 
 		childregs->areg[1] = usp;
+
+		/* When sharing memory with the parent thread, the child
+		   usually starts on a pristine stack, so we have to reset
+		   windowbase, windowstart and wmask.
+		   (Note that such a new thread is required to always create
+		   an initial call4 frame)
+		   The exception is vfork, where the new thread continues to
+		   run on the parent's stack until it calls execve. This could
+		   be a call8 or call12, which requires a legal stack frame
+		   of the previous caller for the overflow handlers to work.
+		   (Note that it's always legal to overflow live registers).
+		   In this case, ensure to spill at least the stack pointer
+		   of that frame. */
+
 		if (clone_flags & CLONE_VM) {
-			childregs->wmask = 1;	/* can't share live windows */
+			/* check that caller window is live and same stack */
+			int len = childregs->wmask & ~0xf;
+			if (regs->areg[1] == usp && len != 0) {
+				int callinc = (regs->areg[0] >> 30) & 3;
+				int caller_ars = XCHAL_NUM_AREGS - callinc * 4;
+				put_user(regs->areg[caller_ars+1],
+					 (unsigned __user*)(usp - 12));
+			}
+			childregs->wmask = 1;
+			childregs->windowstart = 1;
+			childregs->windowbase = 0;
 		} else {
 			int len = childregs->wmask & ~0xf;
 			memcpy(&childregs->areg[XCHAL_NUM_AREGS - len/4],
-- 
cgit v1.2.3-18-g5258


From 3306a72669a2af23b324540b08e68e82ac5b7dab Mon Sep 17 00:00:00 2001
From: Max Filippov <jcmvbkbc@gmail.com>
Date: Thu, 25 Oct 2012 11:10:50 +0400
Subject: xtensa: switch to generic kernel_thread()

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Chris Zankel <chris@zankel.net>
---
 arch/xtensa/Kconfig                 |  1 +
 arch/xtensa/include/asm/processor.h |  4 +-
 arch/xtensa/kernel/entry.S          | 41 +++++++-------------
 arch/xtensa/kernel/process.c        | 77 ++++++++++++++++++++++---------------
 arch/xtensa/kernel/xtensa_ksyms.c   |  1 -
 5 files changed, 61 insertions(+), 63 deletions(-)

(limited to 'arch')

diff --git a/arch/xtensa/Kconfig b/arch/xtensa/Kconfig
index cdcb48adee4..9525b70a13a 100644
--- a/arch/xtensa/Kconfig
+++ b/arch/xtensa/Kconfig
@@ -13,6 +13,7 @@ config XTENSA
 	select GENERIC_CPU_DEVICES
 	select MODULES_USE_ELF_RELA
 	select GENERIC_PCI_IOMAP
+	select GENERIC_KERNEL_THREAD
 	select ARCH_WANT_OPTIONAL_GPIOLIB
 	help
 	  Xtensa processors are 32-bit RISC machines designed by Tensilica
diff --git a/arch/xtensa/include/asm/processor.h b/arch/xtensa/include/asm/processor.h
index 5c371d8d452..2d630e7399c 100644
--- a/arch/xtensa/include/asm/processor.h
+++ b/arch/xtensa/include/asm/processor.h
@@ -152,6 +152,7 @@ struct thread_struct {
 
 /* Clearing a0 terminates the backtrace. */
 #define start_thread(regs, new_pc, new_sp) \
+	memset(regs, 0, sizeof(*regs)); \
 	regs->pc = new_pc; \
 	regs->ps = USER_PS_VALUE; \
 	regs->areg[1] = new_sp; \
@@ -168,9 +169,6 @@ struct mm_struct;
 /* Free all resources held by a thread. */
 #define release_thread(thread) do { } while(0)
 
-/* Create a kernel thread without removing it from tasklists */
-extern int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags);
-
 /* Copy and release all segment info associated with a VM */
 #define copy_segments(p, mm)	do { } while(0)
 #define release_segments(mm)	do { } while(0)
diff --git a/arch/xtensa/kernel/entry.S b/arch/xtensa/kernel/entry.S
index 18453067c25..d4ec1381aeb 100644
--- a/arch/xtensa/kernel/entry.S
+++ b/arch/xtensa/kernel/entry.S
@@ -1832,34 +1832,6 @@ ENTRY(system_call)
 	retw
 
 
-/*
- * Create a kernel thread
- *
- * int kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
- * a2                    a2                 a3             a4
- */
-
-ENTRY(kernel_thread)
-	entry	a1, 16
-
-	mov	a5, a2			# preserve fn over syscall
-	mov	a7, a3			# preserve args over syscall
-
-	movi	a3, _CLONE_VM | _CLONE_UNTRACED
-	movi	a2, __NR_clone
-	or	a6, a4, a3		# arg0: flags
-	mov	a3, a1			# arg1: sp
-	syscall
-
-	beq	a3, a1, 1f		# branch if parent
-	mov	a6, a7			# args
-	callx4	a5			# fn(args)
-
-	movi	a2, __NR_exit
-	syscall				# return value of fn(args) still in a6
-
-1:	retw
-
 /*
  * Do a system call from kernel instead of calling sys_execve, so we end up
  * with proper pt_regs.
@@ -1958,3 +1930,16 @@ ENTRY(ret_from_fork)
 
 	j	common_exception_return
 
+/*
+ * Kernel thread creation helper
+ * On entry, set up by copy_thread: a2 = thread_fn, a3 = thread_fn arg
+ *           left from _switch_to: a6 = prev
+ */
+ENTRY(ret_from_kernel_thread)
+
+	call4	schedule_tail
+	mov	a6, a3
+	callx4	a2
+	call4	do_exit
+
+ENDPROC(ret_from_kernel_thread)
diff --git a/arch/xtensa/kernel/process.c b/arch/xtensa/kernel/process.c
index 341c151af91..be52fe437c1 100644
--- a/arch/xtensa/kernel/process.c
+++ b/arch/xtensa/kernel/process.c
@@ -45,6 +45,7 @@
 #include <asm/regs.h>
 
 extern void ret_from_fork(void);
+extern void ret_from_kernel_thread(void);
 
 struct task_struct *current_set[NR_CPUS] = {&init_task, };
 
@@ -158,18 +159,30 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
 /*
  * Copy thread.
  *
+ * There are two modes in which this function is called:
+ * 1) Userspace thread creation,
+ *    regs != NULL, usp_thread_fn is userspace stack pointer.
+ *    It is expected to copy parent regs (in case CLONE_VM is not set
+ *    in the clone_flags) and set up passed usp in the childregs.
+ * 2) Kernel thread creation,
+ *    regs == NULL, usp_thread_fn is the function to run in the new thread
+ *    and thread_fn_arg is its parameter.
+ *    childregs are not used for the kernel threads.
+ *
  * The stack layout for the new thread looks like this:
  *
- *	+------------------------+ <- sp in childregs (= tos)
+ *	+------------------------+
  *	|       childregs        |
  *	+------------------------+ <- thread.sp = sp in dummy-frame
  *	|      dummy-frame       |    (saved in dummy-frame spill-area)
  *	+------------------------+
  *
- * We create a dummy frame to return to ret_from_fork:
- *   a0 points to ret_from_fork (simulating a call4)
+ * We create a dummy frame to return to either ret_from_fork or
+ *   ret_from_kernel_thread:
+ *   a0 points to ret_from_fork/ret_from_kernel_thread (simulating a call4)
  *   sp points to itself (thread.sp)
- *   a2, a3 are unused.
+ *   a2, a3 are unused for userspace threads,
+ *   a2 points to thread_fn, a3 holds thread_fn arg for kernel threads.
  *
  * Note: This is a pristine frame, so we don't need any spill region on top of
  *       childregs.
@@ -185,41 +198,37 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
  * involved.  Much simpler to just not copy those live frames across.
  */
 
-int copy_thread(unsigned long clone_flags, unsigned long usp,
-		unsigned long unused,
-                struct task_struct * p, struct pt_regs * regs)
+int copy_thread(unsigned long clone_flags, unsigned long usp_thread_fn,
+		unsigned long thread_fn_arg,
+		struct task_struct *p, struct pt_regs *unused)
 {
-	struct pt_regs *childregs;
-	unsigned long tos;
-	int user_mode = user_mode(regs);
+	struct pt_regs *childregs = task_pt_regs(p);
 
 #if (XTENSA_HAVE_COPROCESSORS || XTENSA_HAVE_IO_PORTS)
 	struct thread_info *ti;
 #endif
 
-	/* Set up new TSS. */
-	tos = (unsigned long)task_stack_page(p) + THREAD_SIZE;
-	if (user_mode)
-		childregs = (struct pt_regs*)(tos - PT_USER_SIZE);
-	else
-		childregs = (struct pt_regs*)tos - 1;
-
-	/* This does not copy all the regs.  In a bout of brilliance or madness,
-	   ARs beyond a0-a15 exist past the end of the struct. */
-	*childregs = *regs;
-
 	/* Create a call4 dummy-frame: a0 = 0, a1 = childregs. */
 	*((int*)childregs - 3) = (unsigned long)childregs;
 	*((int*)childregs - 4) = 0;
 
-	childregs->areg[2] = 0;
-	p->set_child_tid = p->clear_child_tid = NULL;
-	p->thread.ra = MAKE_RA_FOR_CALL((unsigned long)ret_from_fork, 0x1);
 	p->thread.sp = (unsigned long)childregs;
 
-	if (user_mode(regs)) {
+	if (!(p->flags & PF_KTHREAD)) {
+		struct pt_regs *regs = current_pt_regs();
+		unsigned long usp = usp_thread_fn ?
+			usp_thread_fn : regs->areg[1];
+
+		p->thread.ra = MAKE_RA_FOR_CALL(
+				(unsigned long)ret_from_fork, 0x1);
 
+		/* This does not copy all the regs.
+		 * In a bout of brilliance or madness,
+		 * ARs beyond a0-a15 exist past the end of the struct.
+		 */
+		*childregs = *regs;
 		childregs->areg[1] = usp;
+		childregs->areg[2] = 0;
 
 		/* When sharing memory with the parent thread, the child
 		   usually starts on a pristine stack, so we have to reset
@@ -254,11 +263,19 @@ int copy_thread(unsigned long clone_flags, unsigned long usp,
 // FIXME: we need to set THREADPTR in thread_info...
 		if (clone_flags & CLONE_SETTLS)
 			childregs->areg[2] = childregs->areg[6];
-
 	} else {
-		/* In kernel space, we start a new thread with a new stack. */
-		childregs->wmask = 1;
-		childregs->areg[1] = tos;
+		p->thread.ra = MAKE_RA_FOR_CALL(
+				(unsigned long)ret_from_kernel_thread, 1);
+
+		/* pass parameters to ret_from_kernel_thread:
+		 * a2 = thread_fn, a3 = thread_fn arg
+		 */
+		*((int *)childregs - 1) = thread_fn_arg;
+		*((int *)childregs - 2) = usp_thread_fn;
+
+		/* Childregs are only used when we're going to userspace
+		 * in which case start_thread will set them up.
+		 */
 	}
 
 #if (XTENSA_HAVE_COPROCESSORS || XTENSA_HAVE_IO_PORTS)
@@ -354,8 +371,6 @@ long xtensa_clone(unsigned long clone_flags, unsigned long newsp,
                   void __user *child_tid, long a5,
                   struct pt_regs *regs)
 {
-        if (!newsp)
-                newsp = regs->areg[1];
         return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid);
 }
 
diff --git a/arch/xtensa/kernel/xtensa_ksyms.c b/arch/xtensa/kernel/xtensa_ksyms.c
index a8b9f1fd1e1..afe058b24e6 100644
--- a/arch/xtensa/kernel/xtensa_ksyms.c
+++ b/arch/xtensa/kernel/xtensa_ksyms.c
@@ -43,7 +43,6 @@ EXPORT_SYMBOL(__strncpy_user);
 EXPORT_SYMBOL(clear_page);
 EXPORT_SYMBOL(copy_page);
 
-EXPORT_SYMBOL(kernel_thread);
 EXPORT_SYMBOL(empty_zero_page);
 
 /*
-- 
cgit v1.2.3-18-g5258


From f0a1bf0859ffa059f9009d6a17e8d46241fb2161 Mon Sep 17 00:00:00 2001
From: Max Filippov <jcmvbkbc@gmail.com>
Date: Thu, 25 Oct 2012 11:10:51 +0400
Subject: xtensa: switch to generic kernel_execve()

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Chris Zankel <chris@zankel.net>
---
 arch/xtensa/Kconfig        |  1 +
 arch/xtensa/kernel/entry.S | 18 +-----------------
 2 files changed, 2 insertions(+), 17 deletions(-)

(limited to 'arch')

diff --git a/arch/xtensa/Kconfig b/arch/xtensa/Kconfig
index 9525b70a13a..0d1f36a22c9 100644
--- a/arch/xtensa/Kconfig
+++ b/arch/xtensa/Kconfig
@@ -14,6 +14,7 @@ config XTENSA
 	select MODULES_USE_ELF_RELA
 	select GENERIC_PCI_IOMAP
 	select GENERIC_KERNEL_THREAD
+	select GENERIC_KERNEL_EXECVE
 	select ARCH_WANT_OPTIONAL_GPIOLIB
 	help
 	  Xtensa processors are 32-bit RISC machines designed by Tensilica
diff --git a/arch/xtensa/kernel/entry.S b/arch/xtensa/kernel/entry.S
index d4ec1381aeb..90bfc1dbc13 100644
--- a/arch/xtensa/kernel/entry.S
+++ b/arch/xtensa/kernel/entry.S
@@ -1832,22 +1832,6 @@ ENTRY(system_call)
 	retw
 
 
-/*
- * Do a system call from kernel instead of calling sys_execve, so we end up
- * with proper pt_regs.
- *
- * int kernel_execve(const char *fname, char *const argv[], charg *const envp[])
- * a2                        a2               a3                  a4
- */
-
-ENTRY(kernel_execve)
-	entry	a1, 16
-	mov	a6, a2			# arg0 is in a6
-	movi	a2, __NR_execve
-	syscall
-
-	retw
-
 /*
  * Task switch.
  *
@@ -1940,6 +1924,6 @@ ENTRY(ret_from_kernel_thread)
 	call4	schedule_tail
 	mov	a6, a3
 	callx4	a2
-	call4	do_exit
+	j	common_exception_return
 
 ENDPROC(ret_from_kernel_thread)
-- 
cgit v1.2.3-18-g5258


From dc241f2c1761bfdec85915f4bbf7e750663f3442 Mon Sep 17 00:00:00 2001
From: Max Filippov <jcmvbkbc@gmail.com>
Date: Thu, 25 Oct 2012 11:10:52 +0400
Subject: xtensa: switch to generic sys_execve()

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Chris Zankel <chris@zankel.net>
---
 arch/xtensa/include/asm/syscall.h     |  2 +-
 arch/xtensa/include/asm/unistd.h      |  1 +
 arch/xtensa/include/uapi/asm/unistd.h |  2 +-
 arch/xtensa/kernel/process.c          | 25 -------------------------
 4 files changed, 3 insertions(+), 27 deletions(-)

(limited to 'arch')

diff --git a/arch/xtensa/include/asm/syscall.h b/arch/xtensa/include/asm/syscall.h
index c1dacca312f..124aeee0d38 100644
--- a/arch/xtensa/include/asm/syscall.h
+++ b/arch/xtensa/include/asm/syscall.h
@@ -10,7 +10,7 @@
 
 struct pt_regs;
 struct sigaction;
-asmlinkage long xtensa_execve(char*, char**, char**, struct pt_regs*);
+asmlinkage long sys_execve(char*, char**, char**, struct pt_regs*);
 asmlinkage long xtensa_clone(unsigned long, unsigned long, struct pt_regs*);
 asmlinkage long xtensa_ptrace(long, long, long, long);
 asmlinkage long xtensa_sigreturn(struct pt_regs*);
diff --git a/arch/xtensa/include/asm/unistd.h b/arch/xtensa/include/asm/unistd.h
index d9fa52dbfba..f4e6eaa40d1 100644
--- a/arch/xtensa/include/asm/unistd.h
+++ b/arch/xtensa/include/asm/unistd.h
@@ -1,6 +1,7 @@
 #ifndef _XTENSA_UNISTD_H
 #define _XTENSA_UNISTD_H
 
+#define __ARCH_WANT_SYS_EXECVE
 #include <uapi/asm/unistd.h>
 
 /*
diff --git a/arch/xtensa/include/uapi/asm/unistd.h b/arch/xtensa/include/uapi/asm/unistd.h
index 1fd157f3121..9f36d0e3e0a 100644
--- a/arch/xtensa/include/uapi/asm/unistd.h
+++ b/arch/xtensa/include/uapi/asm/unistd.h
@@ -262,7 +262,7 @@ __SYSCALL(115, sys_sendmmsg, 4)
 #define __NR_clone 				116
 __SYSCALL(116, xtensa_clone, 5)
 #define __NR_execve 				117
-__SYSCALL(117, xtensa_execve, 3)
+__SYSCALL(117, sys_execve, 3)
 #define __NR_exit 				118
 __SYSCALL(118, sys_exit, 1)
 #define __NR_exit_group 			119
diff --git a/arch/xtensa/kernel/process.c b/arch/xtensa/kernel/process.c
index be52fe437c1..09ae7bfab9a 100644
--- a/arch/xtensa/kernel/process.c
+++ b/arch/xtensa/kernel/process.c
@@ -373,28 +373,3 @@ long xtensa_clone(unsigned long clone_flags, unsigned long newsp,
 {
         return do_fork(clone_flags, newsp, regs, 0, parent_tid, child_tid);
 }
-
-/*
- * xtensa_execve() executes a new program.
- */
-
-asmlinkage
-long xtensa_execve(const char __user *name,
-		   const char __user *const __user *argv,
-                   const char __user *const __user *envp,
-                   long a3, long a4, long a5,
-                   struct pt_regs *regs)
-{
-	long error;
-	struct filename *filename;
-
-	filename = getname(name);
-	error = PTR_ERR(filename);
-	if (IS_ERR(filename))
-		goto out;
-	error = do_execve(filename->name, argv, envp, regs);
-	putname(filename);
-out:
-	return error;
-}
-
-- 
cgit v1.2.3-18-g5258


From 156152f84e57bff4d9b5d70be6f3f16116bf6fb1 Mon Sep 17 00:00:00 2001
From: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Date: Thu, 25 Oct 2012 17:24:12 +0200
Subject: s390/mm: use pmd_large() instead of pmd_huge()

Without CONFIG_HUGETLB_PAGE, pmd_huge() will always return 0. So
pmd_large() should be used instead in places where both transparent
huge pages and hugetlbfs pages can occur.

Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---
 arch/s390/lib/uaccess_pt.c | 2 +-
 arch/s390/mm/gup.c         | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

(limited to 'arch')

diff --git a/arch/s390/lib/uaccess_pt.c b/arch/s390/lib/uaccess_pt.c
index 2d37bb861fa..9017a63dda3 100644
--- a/arch/s390/lib/uaccess_pt.c
+++ b/arch/s390/lib/uaccess_pt.c
@@ -39,7 +39,7 @@ static __always_inline unsigned long follow_table(struct mm_struct *mm,
 	pmd = pmd_offset(pud, addr);
 	if (pmd_none(*pmd))
 		return -0x10UL;
-	if (pmd_huge(*pmd)) {
+	if (pmd_large(*pmd)) {
 		if (write && (pmd_val(*pmd) & _SEGMENT_ENTRY_RO))
 			return -0x04UL;
 		return (pmd_val(*pmd) & HPAGE_MASK) + (addr & ~HPAGE_MASK);
diff --git a/arch/s390/mm/gup.c b/arch/s390/mm/gup.c
index 60acb93a468..8b8285310b5 100644
--- a/arch/s390/mm/gup.c
+++ b/arch/s390/mm/gup.c
@@ -126,7 +126,7 @@ static inline int gup_pmd_range(pud_t *pudp, pud_t pud, unsigned long addr,
 		 */
 		if (pmd_none(pmd) || pmd_trans_splitting(pmd))
 			return 0;
-		if (unlikely(pmd_huge(pmd))) {
+		if (unlikely(pmd_large(pmd))) {
 			if (!gup_huge_pmd(pmdp, pmd, addr, next,
 					  write, pages, nr))
 				return 0;
-- 
cgit v1.2.3-18-g5258


From d8e7a33df0976459674dece4b630b5d59796ece7 Mon Sep 17 00:00:00 2001
From: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Date: Thu, 25 Oct 2012 17:42:50 +0200
Subject: s390/thp: respect page protection in pmd_none() and pmd_present()

Similar to pte_none() and pte_present(), the pmd functions should also
respect page protection of huge pages, especially PROT_NONE.
This patch also simplifies massage_pgprot_pmd() by adding new definitions
for huge page protection.

Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
---
 arch/s390/include/asm/pgtable.h | 35 ++++++++++++++++++++++-------------
 1 file changed, 22 insertions(+), 13 deletions(-)

(limited to 'arch')

diff --git a/arch/s390/include/asm/pgtable.h b/arch/s390/include/asm/pgtable.h
index dd647c919a6..2d3b7cb2600 100644
--- a/arch/s390/include/asm/pgtable.h
+++ b/arch/s390/include/asm/pgtable.h
@@ -506,12 +506,15 @@ static inline int pud_bad(pud_t pud)
 
 static inline int pmd_present(pmd_t pmd)
 {
-	return (pmd_val(pmd) & _SEGMENT_ENTRY_ORIGIN) != 0UL;
+	unsigned long mask = _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO;
+	return (pmd_val(pmd) & mask) == _HPAGE_TYPE_NONE ||
+	       !(pmd_val(pmd) & _SEGMENT_ENTRY_INV);
 }
 
 static inline int pmd_none(pmd_t pmd)
 {
-	return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) != 0UL;
+	return (pmd_val(pmd) & _SEGMENT_ENTRY_INV) &&
+	       !(pmd_val(pmd) & _SEGMENT_ENTRY_RO);
 }
 
 static inline int pmd_large(pmd_t pmd)
@@ -1223,6 +1226,11 @@ static inline void __pmd_idte(unsigned long address, pmd_t *pmdp)
 }
 
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
+
+#define SEGMENT_NONE	__pgprot(_HPAGE_TYPE_NONE)
+#define SEGMENT_RO	__pgprot(_HPAGE_TYPE_RO)
+#define SEGMENT_RW	__pgprot(_HPAGE_TYPE_RW)
+
 #define __HAVE_ARCH_PGTABLE_DEPOSIT
 extern void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable);
 
@@ -1242,16 +1250,15 @@ static inline void set_pmd_at(struct mm_struct *mm, unsigned long addr,
 
 static inline unsigned long massage_pgprot_pmd(pgprot_t pgprot)
 {
-	unsigned long pgprot_pmd = 0;
-
-	if (pgprot_val(pgprot) & _PAGE_INVALID) {
-		if (pgprot_val(pgprot) & _PAGE_SWT)
-			pgprot_pmd |= _HPAGE_TYPE_NONE;
-		pgprot_pmd |= _SEGMENT_ENTRY_INV;
-	}
-	if (pgprot_val(pgprot) & _PAGE_RO)
-		pgprot_pmd |= _SEGMENT_ENTRY_RO;
-	return pgprot_pmd;
+	/*
+	 * pgprot is PAGE_NONE, PAGE_RO, or PAGE_RW (see __Pxxx / __Sxxx)
+	 * Convert to segment table entry format.
+	 */
+	if (pgprot_val(pgprot) == pgprot_val(PAGE_NONE))
+		return pgprot_val(SEGMENT_NONE);
+	if (pgprot_val(pgprot) == pgprot_val(PAGE_RO))
+		return pgprot_val(SEGMENT_RO);
+	return pgprot_val(SEGMENT_RW);
 }
 
 static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
@@ -1269,7 +1276,9 @@ static inline pmd_t pmd_mkhuge(pmd_t pmd)
 
 static inline pmd_t pmd_mkwrite(pmd_t pmd)
 {
-	pmd_val(pmd) &= ~_SEGMENT_ENTRY_RO;
+	/* Do not clobber _HPAGE_TYPE_NONE pages! */
+	if (!(pmd_val(pmd) & _SEGMENT_ENTRY_INV))
+		pmd_val(pmd) &= ~_SEGMENT_ENTRY_RO;
 	return pmd;
 }
 
-- 
cgit v1.2.3-18-g5258


From 517ffce4e1a03aea979fe3a18a3dd1761a24fafb Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Fri, 26 Oct 2012 15:18:37 -0700
Subject: sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads.

The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.

These values are 64-bit.  But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.

Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:

1) Load the top-32bits of a choosen integer register with a sentinel,
   say "-1".  This will be in the outer-most register window.

   The idea is that we're trying to see if the outer-most register
   window gets spilled, and thus the 64-bit values were truncated.

2) Load all the inputs for the montmul/montsqr/mpmul instruction,
   down to the inner-most register window.

3) Execute the opcode.

4) Traverse back up to the outer-most register window.

5) Check the sentinel, if it's still "-1" store the results.
   Otherwise retry the entire sequence.

This retry is extremely troublesome.  If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.

We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.

Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack.  Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.

So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.

Runtime detection of such a facility is tricky, and cumbersome at
best.  For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp).  Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.

With help from Andy Polyakov.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc/include/asm/compat.h         |  5 ++--
 arch/sparc/include/asm/thread_info_64.h |  5 ++++
 arch/sparc/include/asm/ttable.h         | 24 ++++++++++++-------
 arch/sparc/include/uapi/asm/unistd.h    |  6 ++++-
 arch/sparc/kernel/perf_event.c          | 22 ++++++++++++-----
 arch/sparc/kernel/process_64.c          | 42 ++++++++++++++++++---------------
 arch/sparc/kernel/ptrace_64.c           |  4 ++--
 arch/sparc/kernel/sys_sparc_64.c        |  5 ++++
 arch/sparc/kernel/systbls_64.S          |  2 ++
 arch/sparc/kernel/unaligned_64.c        | 36 ++++++++++++++++++----------
 arch/sparc/kernel/visemul.c             | 23 +++++++++++-------
 arch/sparc/kernel/winfixup.S            |  2 ++
 arch/sparc/math-emu/math_64.c           |  2 +-
 13 files changed, 117 insertions(+), 61 deletions(-)

(limited to 'arch')

diff --git a/arch/sparc/include/asm/compat.h b/arch/sparc/include/asm/compat.h
index cef99fbc0a2..830502fe62b 100644
--- a/arch/sparc/include/asm/compat.h
+++ b/arch/sparc/include/asm/compat.h
@@ -232,9 +232,10 @@ static inline void __user *arch_compat_alloc_user_space(long len)
 	struct pt_regs *regs = current_thread_info()->kregs;
 	unsigned long usp = regs->u_regs[UREG_I6];
 
-	if (!(test_thread_flag(TIF_32BIT)))
+	if (test_thread_64bit_stack(usp))
 		usp += STACK_BIAS;
-	else
+
+	if (test_thread_flag(TIF_32BIT))
 		usp &= 0xffffffffUL;
 
 	usp -= len;
diff --git a/arch/sparc/include/asm/thread_info_64.h b/arch/sparc/include/asm/thread_info_64.h
index 4e227663108..a3fe4dcc0aa 100644
--- a/arch/sparc/include/asm/thread_info_64.h
+++ b/arch/sparc/include/asm/thread_info_64.h
@@ -259,6 +259,11 @@ static inline bool test_and_clear_restore_sigmask(void)
 
 #define tsk_is_polling(t) test_tsk_thread_flag(t, TIF_POLLING_NRFLAG)
 
+#define thread32_stack_is_64bit(__SP) (((__SP) & 0x1) != 0)
+#define test_thread_64bit_stack(__SP) \
+	((test_thread_flag(TIF_32BIT) && !thread32_stack_is_64bit(__SP)) ? \
+	 false : true)
+
 #endif	/* !__ASSEMBLY__ */
 
 #endif /* __KERNEL__ */
diff --git a/arch/sparc/include/asm/ttable.h b/arch/sparc/include/asm/ttable.h
index 48f2807d326..71b5a67522a 100644
--- a/arch/sparc/include/asm/ttable.h
+++ b/arch/sparc/include/asm/ttable.h
@@ -372,7 +372,9 @@ etrap_spill_fixup_64bit:				\
 
 /* Normal 32bit spill */
 #define SPILL_2_GENERIC(ASI)				\
-	srl	%sp, 0, %sp;				\
+	and	%sp, 1, %g3;				\
+	brnz,pn	%g3, (. - (128 + 4));			\
+	 srl	%sp, 0, %sp;				\
 	stwa	%l0, [%sp + %g0] ASI;			\
 	mov	0x04, %g3;				\
 	stwa	%l1, [%sp + %g3] ASI;			\
@@ -398,14 +400,16 @@ etrap_spill_fixup_64bit:				\
 	stwa	%i6, [%g1 + %g0] ASI;			\
 	stwa	%i7, [%g1 + %g3] ASI;			\
 	saved;						\
-        retry; nop; nop;				\
+        retry;						\
 	b,a,pt	%xcc, spill_fixup_dax;			\
 	b,a,pt	%xcc, spill_fixup_mna;			\
 	b,a,pt	%xcc, spill_fixup;
 
 #define SPILL_2_GENERIC_ETRAP		\
 etrap_user_spill_32bit:			\
-	srl	%sp, 0, %sp;		\
+	and	%sp, 1, %g3;		\
+	brnz,pn	%g3, etrap_user_spill_64bit;	\
+	 srl	%sp, 0, %sp;		\
 	stwa	%l0, [%sp + 0x00] %asi;	\
 	stwa	%l1, [%sp + 0x04] %asi;	\
 	stwa	%l2, [%sp + 0x08] %asi;	\
@@ -427,7 +431,7 @@ etrap_user_spill_32bit:			\
 	ba,pt	%xcc, etrap_save;	\
 	 wrpr	%g1, %cwp;		\
 	nop; nop; nop; nop;		\
-	nop; nop; nop; nop;		\
+	nop; nop;			\
 	ba,a,pt	%xcc, etrap_spill_fixup_32bit; \
 	ba,a,pt	%xcc, etrap_spill_fixup_32bit; \
 	ba,a,pt	%xcc, etrap_spill_fixup_32bit;
@@ -592,7 +596,9 @@ user_rtt_fill_64bit:					\
 
 /* Normal 32bit fill */
 #define FILL_2_GENERIC(ASI)				\
-	srl	%sp, 0, %sp;				\
+	and	%sp, 1, %g3;				\
+	brnz,pn	%g3, (. - (128 + 4));			\
+	 srl	%sp, 0, %sp;				\
 	lduwa	[%sp + %g0] ASI, %l0;			\
 	mov	0x04, %g2;				\
 	mov	0x08, %g3;				\
@@ -616,14 +622,16 @@ user_rtt_fill_64bit:					\
 	lduwa	[%g1 + %g3] ASI, %i6;			\
 	lduwa	[%g1 + %g5] ASI, %i7;			\
 	restored;					\
-	retry; nop; nop; nop; nop;			\
+	retry; nop; nop;				\
 	b,a,pt	%xcc, fill_fixup_dax;			\
 	b,a,pt	%xcc, fill_fixup_mna;			\
 	b,a,pt	%xcc, fill_fixup;
 
 #define FILL_2_GENERIC_RTRAP				\
 user_rtt_fill_32bit:					\
-	srl	%sp, 0, %sp;				\
+	and	%sp, 1, %g3;				\
+	brnz,pn	%g3, user_rtt_fill_64bit;		\
+	 srl	%sp, 0, %sp;				\
 	lduwa	[%sp + 0x00] %asi, %l0;			\
 	lduwa	[%sp + 0x04] %asi, %l1;			\
 	lduwa	[%sp + 0x08] %asi, %l2;			\
@@ -643,7 +651,7 @@ user_rtt_fill_32bit:					\
 	ba,pt	%xcc, user_rtt_pre_restore;		\
 	 restored;					\
 	nop; nop; nop; nop; nop;			\
-	nop; nop; nop; nop; nop;			\
+	nop; nop; nop;					\
 	ba,a,pt	%xcc, user_rtt_fill_fixup;		\
 	ba,a,pt	%xcc, user_rtt_fill_fixup;		\
 	ba,a,pt	%xcc, user_rtt_fill_fixup;
diff --git a/arch/sparc/include/uapi/asm/unistd.h b/arch/sparc/include/uapi/asm/unistd.h
index 8974ef7ae92..bed86a820d0 100644
--- a/arch/sparc/include/uapi/asm/unistd.h
+++ b/arch/sparc/include/uapi/asm/unistd.h
@@ -405,8 +405,12 @@
 #define __NR_setns		337
 #define __NR_process_vm_readv	338
 #define __NR_process_vm_writev	339
+#define __NR_kern_features	340
 
-#define NR_syscalls		340
+#define NR_syscalls		341
+
+/* Bitmask values returned from kern_features system call.  */
+#define KERN_FEATURE_MIXED_MODE_STACK	0x00000001
 
 #ifdef __32bit_syscall_numbers__
 /* Sparc 32-bit only has the "setresuid32", "getresuid32" variants,
diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c
index 885a8af7406..b5c38faa4ea 100644
--- a/arch/sparc/kernel/perf_event.c
+++ b/arch/sparc/kernel/perf_event.c
@@ -1762,15 +1762,25 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry,
 
 	ufp = regs->u_regs[UREG_I6] & 0xffffffffUL;
 	do {
-		struct sparc_stackf32 *usf, sf;
 		unsigned long pc;
 
-		usf = (struct sparc_stackf32 *) ufp;
-		if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
-			break;
+		if (thread32_stack_is_64bit(ufp)) {
+			struct sparc_stackf *usf, sf;
 
-		pc = sf.callers_pc;
-		ufp = (unsigned long)sf.fp;
+			ufp += STACK_BIAS;
+			usf = (struct sparc_stackf *) ufp;
+			if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
+				break;
+			pc = sf.callers_pc & 0xffffffff;
+			ufp = ((unsigned long) sf.fp) & 0xffffffff;
+		} else {
+			struct sparc_stackf32 *usf, sf;
+			usf = (struct sparc_stackf32 *) ufp;
+			if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
+				break;
+			pc = sf.callers_pc;
+			ufp = (unsigned long)sf.fp;
+		}
 		perf_callchain_store(entry, pc);
 	} while (entry->nr < PERF_MAX_STACK_DEPTH);
 }
diff --git a/arch/sparc/kernel/process_64.c b/arch/sparc/kernel/process_64.c
index d778248ef3f..c6e0c291004 100644
--- a/arch/sparc/kernel/process_64.c
+++ b/arch/sparc/kernel/process_64.c
@@ -452,13 +452,16 @@ void flush_thread(void)
 /* It's a bit more tricky when 64-bit tasks are involved... */
 static unsigned long clone_stackframe(unsigned long csp, unsigned long psp)
 {
+	bool stack_64bit = test_thread_64bit_stack(psp);
 	unsigned long fp, distance, rval;
 
-	if (!(test_thread_flag(TIF_32BIT))) {
+	if (stack_64bit) {
 		csp += STACK_BIAS;
 		psp += STACK_BIAS;
 		__get_user(fp, &(((struct reg_window __user *)psp)->ins[6]));
 		fp += STACK_BIAS;
+		if (test_thread_flag(TIF_32BIT))
+			fp &= 0xffffffff;
 	} else
 		__get_user(fp, &(((struct reg_window32 __user *)psp)->ins[6]));
 
@@ -472,7 +475,7 @@ static unsigned long clone_stackframe(unsigned long csp, unsigned long psp)
 	rval = (csp - distance);
 	if (copy_in_user((void __user *) rval, (void __user *) psp, distance))
 		rval = 0;
-	else if (test_thread_flag(TIF_32BIT)) {
+	else if (!stack_64bit) {
 		if (put_user(((u32)csp),
 			     &(((struct reg_window32 __user *)rval)->ins[6])))
 			rval = 0;
@@ -507,18 +510,18 @@ void synchronize_user_stack(void)
 
 	flush_user_windows();
 	if ((window = get_thread_wsaved()) != 0) {
-		int winsize = sizeof(struct reg_window);
-		int bias = 0;
-
-		if (test_thread_flag(TIF_32BIT))
-			winsize = sizeof(struct reg_window32);
-		else
-			bias = STACK_BIAS;
-
 		window -= 1;
 		do {
-			unsigned long sp = (t->rwbuf_stkptrs[window] + bias);
 			struct reg_window *rwin = &t->reg_window[window];
+			int winsize = sizeof(struct reg_window);
+			unsigned long sp;
+
+			sp = t->rwbuf_stkptrs[window];
+
+			if (test_thread_64bit_stack(sp))
+				sp += STACK_BIAS;
+			else
+				winsize = sizeof(struct reg_window32);
 
 			if (!copy_to_user((char __user *)sp, rwin, winsize)) {
 				shift_window_buffer(window, get_thread_wsaved() - 1, t);
@@ -544,13 +547,6 @@ void fault_in_user_windows(void)
 {
 	struct thread_info *t = current_thread_info();
 	unsigned long window;
-	int winsize = sizeof(struct reg_window);
-	int bias = 0;
-
-	if (test_thread_flag(TIF_32BIT))
-		winsize = sizeof(struct reg_window32);
-	else
-		bias = STACK_BIAS;
 
 	flush_user_windows();
 	window = get_thread_wsaved();
@@ -558,8 +554,16 @@ void fault_in_user_windows(void)
 	if (likely(window != 0)) {
 		window -= 1;
 		do {
-			unsigned long sp = (t->rwbuf_stkptrs[window] + bias);
 			struct reg_window *rwin = &t->reg_window[window];
+			int winsize = sizeof(struct reg_window);
+			unsigned long sp;
+
+			sp = t->rwbuf_stkptrs[window];
+
+			if (test_thread_64bit_stack(sp))
+				sp += STACK_BIAS;
+			else
+				winsize = sizeof(struct reg_window32);
 
 			if (unlikely(sp & 0x7UL))
 				stack_unaligned(sp);
diff --git a/arch/sparc/kernel/ptrace_64.c b/arch/sparc/kernel/ptrace_64.c
index 484dabac704..7ff45e4ba68 100644
--- a/arch/sparc/kernel/ptrace_64.c
+++ b/arch/sparc/kernel/ptrace_64.c
@@ -151,7 +151,7 @@ static int regwindow64_get(struct task_struct *target,
 {
 	unsigned long rw_addr = regs->u_regs[UREG_I6];
 
-	if (test_tsk_thread_flag(current, TIF_32BIT)) {
+	if (!test_thread_64bit_stack(rw_addr)) {
 		struct reg_window32 win32;
 		int i;
 
@@ -176,7 +176,7 @@ static int regwindow64_set(struct task_struct *target,
 {
 	unsigned long rw_addr = regs->u_regs[UREG_I6];
 
-	if (test_tsk_thread_flag(current, TIF_32BIT)) {
+	if (!test_thread_64bit_stack(rw_addr)) {
 		struct reg_window32 win32;
 		int i;
 
diff --git a/arch/sparc/kernel/sys_sparc_64.c b/arch/sparc/kernel/sys_sparc_64.c
index 11c6c9603e7..878ef3d5fec 100644
--- a/arch/sparc/kernel/sys_sparc_64.c
+++ b/arch/sparc/kernel/sys_sparc_64.c
@@ -751,3 +751,8 @@ int kernel_execve(const char *filename,
 		      : "cc");
 	return __res;
 }
+
+asmlinkage long sys_kern_features(void)
+{
+	return KERN_FEATURE_MIXED_MODE_STACK;
+}
diff --git a/arch/sparc/kernel/systbls_64.S b/arch/sparc/kernel/systbls_64.S
index 3a58e0d66f5..45ce6be088e 100644
--- a/arch/sparc/kernel/systbls_64.S
+++ b/arch/sparc/kernel/systbls_64.S
@@ -86,6 +86,7 @@ sys_call_table32:
 	.word compat_sys_pwritev, compat_sys_rt_tgsigqueueinfo, sys_perf_event_open, compat_sys_recvmmsg, sys_fanotify_init
 /*330*/	.word sys32_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, compat_sys_open_by_handle_at, compat_sys_clock_adjtime
 	.word sys_syncfs, compat_sys_sendmmsg, sys_setns, compat_sys_process_vm_readv, compat_sys_process_vm_writev
+/*340*/	.word sys_kern_features
 
 #endif /* CONFIG_COMPAT */
 
@@ -163,3 +164,4 @@ sys_call_table:
 	.word sys_pwritev, sys_rt_tgsigqueueinfo, sys_perf_event_open, sys_recvmmsg, sys_fanotify_init
 /*330*/	.word sys_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, sys_open_by_handle_at, sys_clock_adjtime
 	.word sys_syncfs, sys_sendmmsg, sys_setns, sys_process_vm_readv, sys_process_vm_writev
+/*340*/	.word sys_kern_features
diff --git a/arch/sparc/kernel/unaligned_64.c b/arch/sparc/kernel/unaligned_64.c
index f81d038f734..8201c25e766 100644
--- a/arch/sparc/kernel/unaligned_64.c
+++ b/arch/sparc/kernel/unaligned_64.c
@@ -113,21 +113,24 @@ static inline long sign_extend_imm13(long imm)
 
 static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 {
-	unsigned long value;
+	unsigned long value, fp;
 	
 	if (reg < 16)
 		return (!reg ? 0 : regs->u_regs[reg]);
+
+	fp = regs->u_regs[UREG_FP];
+
 	if (regs->tstate & TSTATE_PRIV) {
 		struct reg_window *win;
-		win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window *)(fp + STACK_BIAS);
 		value = win->locals[reg - 16];
-	} else if (test_thread_flag(TIF_32BIT)) {
+	} else if (!test_thread_64bit_stack(fp)) {
 		struct reg_window32 __user *win32;
-		win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+		win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
 		get_user(value, &win32->locals[reg - 16]);
 	} else {
 		struct reg_window __user *win;
-		win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window __user *)(fp + STACK_BIAS);
 		get_user(value, &win->locals[reg - 16]);
 	}
 	return value;
@@ -135,19 +138,24 @@ static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 
 static unsigned long *fetch_reg_addr(unsigned int reg, struct pt_regs *regs)
 {
+	unsigned long fp;
+
 	if (reg < 16)
 		return &regs->u_regs[reg];
+
+	fp = regs->u_regs[UREG_FP];
+
 	if (regs->tstate & TSTATE_PRIV) {
 		struct reg_window *win;
-		win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window *)(fp + STACK_BIAS);
 		return &win->locals[reg - 16];
-	} else if (test_thread_flag(TIF_32BIT)) {
+	} else if (!test_thread_64bit_stack(fp)) {
 		struct reg_window32 *win32;
-		win32 = (struct reg_window32 *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+		win32 = (struct reg_window32 *)((unsigned long)((u32)fp));
 		return (unsigned long *)&win32->locals[reg - 16];
 	} else {
 		struct reg_window *win;
-		win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window *)(fp + STACK_BIAS);
 		return &win->locals[reg - 16];
 	}
 }
@@ -392,13 +400,15 @@ int handle_popc(u32 insn, struct pt_regs *regs)
 		if (rd)
 			regs->u_regs[rd] = ret;
 	} else {
-		if (test_thread_flag(TIF_32BIT)) {
+		unsigned long fp = regs->u_regs[UREG_FP];
+
+		if (!test_thread_64bit_stack(fp)) {
 			struct reg_window32 __user *win32;
-			win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+			win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
 			put_user(ret, &win32->locals[rd - 16]);
 		} else {
 			struct reg_window __user *win;
-			win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+			win = (struct reg_window __user *)(fp + STACK_BIAS);
 			put_user(ret, &win->locals[rd - 16]);
 		}
 	}
@@ -554,7 +564,7 @@ void handle_ld_nf(u32 insn, struct pt_regs *regs)
 		reg[0] = 0;
 		if ((insn & 0x780000) == 0x180000)
 			reg[1] = 0;
-	} else if (test_thread_flag(TIF_32BIT)) {
+	} else if (!test_thread_64bit_stack(regs->u_regs[UREG_FP])) {
 		put_user(0, (int __user *) reg);
 		if ((insn & 0x780000) == 0x180000)
 			put_user(0, ((int __user *) reg) + 1);
diff --git a/arch/sparc/kernel/visemul.c b/arch/sparc/kernel/visemul.c
index 08e074b7eb6..c096c624ac4 100644
--- a/arch/sparc/kernel/visemul.c
+++ b/arch/sparc/kernel/visemul.c
@@ -149,21 +149,24 @@ static inline void maybe_flush_windows(unsigned int rs1, unsigned int rs2,
 
 static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 {
-	unsigned long value;
+	unsigned long value, fp;
 	
 	if (reg < 16)
 		return (!reg ? 0 : regs->u_regs[reg]);
+
+	fp = regs->u_regs[UREG_FP];
+
 	if (regs->tstate & TSTATE_PRIV) {
 		struct reg_window *win;
-		win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window *)(fp + STACK_BIAS);
 		value = win->locals[reg - 16];
-	} else if (test_thread_flag(TIF_32BIT)) {
+	} else if (!test_thread_64bit_stack(fp)) {
 		struct reg_window32 __user *win32;
-		win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+		win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
 		get_user(value, &win32->locals[reg - 16]);
 	} else {
 		struct reg_window __user *win;
-		win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window __user *)(fp + STACK_BIAS);
 		get_user(value, &win->locals[reg - 16]);
 	}
 	return value;
@@ -172,16 +175,18 @@ static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 static inline unsigned long __user *__fetch_reg_addr_user(unsigned int reg,
 							  struct pt_regs *regs)
 {
+	unsigned long fp = regs->u_regs[UREG_FP];
+
 	BUG_ON(reg < 16);
 	BUG_ON(regs->tstate & TSTATE_PRIV);
 
-	if (test_thread_flag(TIF_32BIT)) {
+	if (!test_thread_64bit_stack(fp)) {
 		struct reg_window32 __user *win32;
-		win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+		win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
 		return (unsigned long __user *)&win32->locals[reg - 16];
 	} else {
 		struct reg_window __user *win;
-		win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+		win = (struct reg_window __user *)(fp + STACK_BIAS);
 		return &win->locals[reg - 16];
 	}
 }
@@ -204,7 +209,7 @@ static void store_reg(struct pt_regs *regs, unsigned long val, unsigned long rd)
 	} else {
 		unsigned long __user *rd_user = __fetch_reg_addr_user(rd, regs);
 
-		if (test_thread_flag(TIF_32BIT))
+		if (!test_thread_64bit_stack(regs->u_regs[UREG_FP]))
 			__put_user((u32)val, (u32 __user *)rd_user);
 		else
 			__put_user(val, rd_user);
diff --git a/arch/sparc/kernel/winfixup.S b/arch/sparc/kernel/winfixup.S
index a6b0863c27d..1e67ce95836 100644
--- a/arch/sparc/kernel/winfixup.S
+++ b/arch/sparc/kernel/winfixup.S
@@ -43,6 +43,8 @@ spill_fixup_mna:
 spill_fixup_dax:
 	TRAP_LOAD_THREAD_REG(%g6, %g1)
 	ldx	[%g6 + TI_FLAGS], %g1
+	andcc	%sp, 0x1, %g0
+	movne	%icc, 0, %g1
 	andcc	%g1, _TIF_32BIT, %g0
 	ldub	[%g6 + TI_WSAVED], %g1
 	sll	%g1, 3, %g3
diff --git a/arch/sparc/math-emu/math_64.c b/arch/sparc/math-emu/math_64.c
index 1704068da92..034aadbff03 100644
--- a/arch/sparc/math-emu/math_64.c
+++ b/arch/sparc/math-emu/math_64.c
@@ -320,7 +320,7 @@ int do_mathemu(struct pt_regs *regs, struct fpustate *f, bool illegal_insn_trap)
 					XR = 0;
 				else if (freg < 16)
 					XR = regs->u_regs[freg];
-				else if (test_thread_flag(TIF_32BIT)) {
+				else if (!test_thread_64bit_stack(regs->u_regs[UREG_FP])) {
 					struct reg_window32 __user *win32;
 					flushw_user ();
 					win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
-- 
cgit v1.2.3-18-g5258


From 270c10e00a1e557e068803a22e0556281ceb1830 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Sat, 27 Oct 2012 18:05:20 -0700
Subject: sparc64: Fix cpu strand yielding.

For atomic backoff, we just loop over an exponentially backed off
counter.  This is extremely ineffective as it doesn't actually yield
the cpu strand so that other competing strands can use the cpu core.

In cpus previous to SPARC-T4 we have to do this in a slightly hackish
way, by doing an operation with no side effects that also happens to
mark the strand as unavailable.

The mechanism we choose for this is three reads of the %ccr
(condition-code) register into %g0 (the zero register).

SPARC-T4 has an explicit "pause" instruction, and we'll make use of
that in a subsequent commit.

Yield strands also in cpu_relax().  We really should have done this a
very long time ago.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc/include/asm/backoff.h      | 5 ++++-
 arch/sparc/include/asm/processor_64.h | 5 ++++-
 2 files changed, 8 insertions(+), 2 deletions(-)

(limited to 'arch')

diff --git a/arch/sparc/include/asm/backoff.h b/arch/sparc/include/asm/backoff.h
index db3af0d30fb..64b077b3b13 100644
--- a/arch/sparc/include/asm/backoff.h
+++ b/arch/sparc/include/asm/backoff.h
@@ -13,7 +13,10 @@
 
 #define BACKOFF_SPIN(reg, tmp, label)	\
 	mov	reg, tmp; \
-88:	brnz,pt	tmp, 88b; \
+88:	rd	%ccr, %g0; \
+	rd	%ccr, %g0; \
+	rd	%ccr, %g0; \
+	brnz,pt	tmp, 88b; \
 	 sub	tmp, 1, tmp; \
 	set	BACKOFF_LIMIT, tmp; \
 	cmp	reg, tmp; \
diff --git a/arch/sparc/include/asm/processor_64.h b/arch/sparc/include/asm/processor_64.h
index 4e5a483122a..98656340946 100644
--- a/arch/sparc/include/asm/processor_64.h
+++ b/arch/sparc/include/asm/processor_64.h
@@ -196,7 +196,10 @@ extern unsigned long get_wchan(struct task_struct *task);
 #define KSTK_EIP(tsk)  (task_pt_regs(tsk)->tpc)
 #define KSTK_ESP(tsk)  (task_pt_regs(tsk)->u_regs[UREG_FP])
 
-#define cpu_relax()	barrier()
+#define cpu_relax()	asm volatile("rd	%%ccr, %%g0\n\t" \
+				     "rd	%%ccr, %%g0\n\t" \
+				     "rd	%%ccr, %%g0" \
+				     ::: "memory")
 
 /* Prefetch support.  This is tuned for UltraSPARC-III and later.
  * UltraSPARC-I will treat these as nops, and UltraSPARC-II has
-- 
cgit v1.2.3-18-g5258


From e9b9eb59ffcdee09ec96b040f85c919618f4043e Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Sat, 27 Oct 2012 23:00:41 -0700
Subject: sparc64: Use pause instruction when available.

In atomic backoff and cpu_relax(), use the pause instruction
found on SPARC-T4 and later.

It makes the cpu strand unselectable for the given number of
cycles, unless an intervening disrupting trap occurs.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc/include/asm/backoff.h      | 32 +++++++++++++++++++-------------
 arch/sparc/include/asm/processor_64.h | 13 ++++++++++---
 arch/sparc/kernel/entry.h             |  7 +++++++
 arch/sparc/kernel/setup_64.c          | 21 +++++++++++++++++++++
 arch/sparc/kernel/vmlinux.lds.S       |  5 +++++
 5 files changed, 62 insertions(+), 16 deletions(-)

(limited to 'arch')

diff --git a/arch/sparc/include/asm/backoff.h b/arch/sparc/include/asm/backoff.h
index 64b077b3b13..20f01df0871 100644
--- a/arch/sparc/include/asm/backoff.h
+++ b/arch/sparc/include/asm/backoff.h
@@ -11,19 +11,25 @@
 #define BACKOFF_LABEL(spin_label, continue_label) \
 	spin_label
 
-#define BACKOFF_SPIN(reg, tmp, label)	\
-	mov	reg, tmp; \
-88:	rd	%ccr, %g0; \
-	rd	%ccr, %g0; \
-	rd	%ccr, %g0; \
-	brnz,pt	tmp, 88b; \
-	 sub	tmp, 1, tmp; \
-	set	BACKOFF_LIMIT, tmp; \
-	cmp	reg, tmp; \
-	bg,pn	%xcc, label; \
-	 nop; \
-	ba,pt	%xcc, label; \
-	 sllx	reg, 1, reg;
+#define BACKOFF_SPIN(reg, tmp, label)		\
+	mov		reg, tmp;		\
+88:	rd		%ccr, %g0;		\
+	rd		%ccr, %g0;		\
+	rd		%ccr, %g0;		\
+	.section	.pause_patch,"ax";	\
+	.word		88b;			\
+	sllx		tmp, 7, tmp;		\
+	wr		tmp, 0, %asr27;		\
+	clr		tmp;			\
+	.previous;				\
+	brnz,pt		tmp, 88b;		\
+	 sub		tmp, 1, tmp;		\
+	set		BACKOFF_LIMIT, tmp;	\
+	cmp		reg, tmp;		\
+	bg,pn		%xcc, label;		\
+	 nop;					\
+	ba,pt		%xcc, label;		\
+	 sllx		reg, 1, reg;
 
 #else
 
diff --git a/arch/sparc/include/asm/processor_64.h b/arch/sparc/include/asm/processor_64.h
index 98656340946..9cdf52eec48 100644
--- a/arch/sparc/include/asm/processor_64.h
+++ b/arch/sparc/include/asm/processor_64.h
@@ -196,9 +196,16 @@ extern unsigned long get_wchan(struct task_struct *task);
 #define KSTK_EIP(tsk)  (task_pt_regs(tsk)->tpc)
 #define KSTK_ESP(tsk)  (task_pt_regs(tsk)->u_regs[UREG_FP])
 
-#define cpu_relax()	asm volatile("rd	%%ccr, %%g0\n\t" \
-				     "rd	%%ccr, %%g0\n\t" \
-				     "rd	%%ccr, %%g0" \
+#define cpu_relax()	asm volatile("\n99:\n\t"			\
+				     "rd	%%ccr, %%g0\n\t"	\
+				     "rd	%%ccr, %%g0\n\t"	\
+				     "rd	%%ccr, %%g0\n\t"	\
+				     ".section	.pause_patch,\"ax\"\n\t"\
+				     ".word	99b\n\t"		\
+				     "wr	%%g0, 128, %%asr27\n\t"	\
+				     "nop\n\t"				\
+				     "nop\n\t"				\
+				     ".previous"			\
 				     ::: "memory")
 
 /* Prefetch support.  This is tuned for UltraSPARC-III and later.
diff --git a/arch/sparc/kernel/entry.h b/arch/sparc/kernel/entry.h
index 0c218e4c088..51742df63c7 100644
--- a/arch/sparc/kernel/entry.h
+++ b/arch/sparc/kernel/entry.h
@@ -59,6 +59,13 @@ struct popc_6insn_patch_entry {
 extern struct popc_6insn_patch_entry __popc_6insn_patch,
 	__popc_6insn_patch_end;
 
+struct pause_patch_entry {
+	unsigned int	addr;
+	unsigned int	insns[3];
+};
+extern struct pause_patch_entry __pause_patch,
+	__pause_patch_end;
+
 extern void __init per_cpu_patch(void);
 extern void sun4v_patch_1insn_range(struct sun4v_1insn_patch_entry *,
 				    struct sun4v_1insn_patch_entry *);
diff --git a/arch/sparc/kernel/setup_64.c b/arch/sparc/kernel/setup_64.c
index 0800e71d8a8..b45cff408de 100644
--- a/arch/sparc/kernel/setup_64.c
+++ b/arch/sparc/kernel/setup_64.c
@@ -316,6 +316,25 @@ static void __init popc_patch(void)
 	}
 }
 
+static void __init pause_patch(void)
+{
+	struct pause_patch_entry *p;
+
+	p = &__pause_patch;
+	while (p < &__pause_patch_end) {
+		unsigned long i, addr = p->addr;
+
+		for (i = 0; i < 3; i++) {
+			*(unsigned int *) (addr +  (i * 4)) = p->insns[i];
+			wmb();
+			__asm__ __volatile__("flush	%0"
+					     : : "r" (addr +  (i * 4)));
+		}
+
+		p++;
+	}
+}
+
 #ifdef CONFIG_SMP
 void __init boot_cpu_id_too_large(int cpu)
 {
@@ -528,6 +547,8 @@ static void __init init_sparc64_elf_hwcap(void)
 
 	if (sparc64_elf_hwcap & AV_SPARC_POPC)
 		popc_patch();
+	if (sparc64_elf_hwcap & AV_SPARC_PAUSE)
+		pause_patch();
 }
 
 void __init setup_arch(char **cmdline_p)
diff --git a/arch/sparc/kernel/vmlinux.lds.S b/arch/sparc/kernel/vmlinux.lds.S
index 89c2c29f154..847f9f79361 100644
--- a/arch/sparc/kernel/vmlinux.lds.S
+++ b/arch/sparc/kernel/vmlinux.lds.S
@@ -132,6 +132,11 @@ SECTIONS
 		*(.popc_6insn_patch)
 		__popc_6insn_patch_end = .;
 	}
+	.pause_patch : {
+		__pause_patch = .;
+		*(.pause_patch)
+		__pause_patch_end = .;
+	}
 	PERCPU_SECTION(SMP_CACHE_BYTES)
 
 	. = ALIGN(PAGE_SIZE);
-- 
cgit v1.2.3-18-g5258


From 187818cd6a5ab6343eac47e52da2f3e40c544b98 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Sun, 28 Oct 2012 13:04:47 -0700
Subject: sparc64: Improvde documentation and readability of atomic backoff
 code.

Document what's going on in asm/backoff.h with a large and descriptive
comment.  Refer to it above the cpu_relax() definition in
asm/processor_64.h

Rename the pause patching section to have "3insn" in it's name like
the other patching sections do.

Based upon feedback from Sam Ravnborg.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc/include/asm/backoff.h      | 42 ++++++++++++++++++++++++++++++++++-
 arch/sparc/include/asm/processor_64.h |  7 +++++-
 arch/sparc/kernel/entry.h             |  4 ++--
 arch/sparc/kernel/setup_64.c          |  4 ++--
 arch/sparc/kernel/vmlinux.lds.S       |  8 +++----
 5 files changed, 55 insertions(+), 10 deletions(-)

(limited to 'arch')

diff --git a/arch/sparc/include/asm/backoff.h b/arch/sparc/include/asm/backoff.h
index 20f01df0871..4e02086b839 100644
--- a/arch/sparc/include/asm/backoff.h
+++ b/arch/sparc/include/asm/backoff.h
@@ -1,6 +1,46 @@
 #ifndef _SPARC64_BACKOFF_H
 #define _SPARC64_BACKOFF_H
 
+/* The macros in this file implement an exponential backoff facility
+ * for atomic operations.
+ *
+ * When multiple threads compete on an atomic operation, it is
+ * possible for one thread to be continually denied a successful
+ * completion of the compare-and-swap instruction.  Heavily
+ * threaded cpu implementations like Niagara can compound this
+ * problem even further.
+ *
+ * When an atomic operation fails and needs to be retried, we spin a
+ * certain number of times.  At each subsequent failure of the same
+ * operation we double the spin count, realizing an exponential
+ * backoff.
+ *
+ * When we spin, we try to use an operation that will cause the
+ * current cpu strand to block, and therefore make the core fully
+ * available to any other other runnable strands.  There are two
+ * options, based upon cpu capabilities.
+ *
+ * On all cpus prior to SPARC-T4 we do three dummy reads of the
+ * condition code register.  Each read blocks the strand for something
+ * between 40 and 50 cpu cycles.
+ *
+ * For SPARC-T4 and later we have a special "pause" instruction
+ * available.  This is implemented using writes to register %asr27.
+ * The cpu will block the number of cycles written into the register,
+ * unless a disrupting trap happens first.  SPARC-T4 specifically
+ * implements pause with a granularity of 8 cycles.  Each strand has
+ * an internal pause counter which decrements every 8 cycles.  So the
+ * chip shifts the %asr27 value down by 3 bits, and writes the result
+ * into the pause counter.  If a value smaller than 8 is written, the
+ * chip blocks for 1 cycle.
+ *
+ * To achieve the same amount of backoff as the three %ccr reads give
+ * on earlier chips, we shift the backoff value up by 7 bits.  (Three
+ * %ccr reads block for about 128 cycles, 1 << 7 == 128) We write the
+ * whole amount we want to block into the pause register, rather than
+ * loop writing 128 each time.
+ */
+
 #define BACKOFF_LIMIT	(4 * 1024)
 
 #ifdef CONFIG_SMP
@@ -16,7 +56,7 @@
 88:	rd		%ccr, %g0;		\
 	rd		%ccr, %g0;		\
 	rd		%ccr, %g0;		\
-	.section	.pause_patch,"ax";	\
+	.section	.pause_3insn_patch,"ax";\
 	.word		88b;			\
 	sllx		tmp, 7, tmp;		\
 	wr		tmp, 0, %asr27;		\
diff --git a/arch/sparc/include/asm/processor_64.h b/arch/sparc/include/asm/processor_64.h
index 9cdf52eec48..721e25f0e2e 100644
--- a/arch/sparc/include/asm/processor_64.h
+++ b/arch/sparc/include/asm/processor_64.h
@@ -196,11 +196,16 @@ extern unsigned long get_wchan(struct task_struct *task);
 #define KSTK_EIP(tsk)  (task_pt_regs(tsk)->tpc)
 #define KSTK_ESP(tsk)  (task_pt_regs(tsk)->u_regs[UREG_FP])
 
+/* Please see the commentary in asm/backoff.h for a description of
+ * what these instructions are doing and how they have been choosen.
+ * To make a long story short, we are trying to yield the current cpu
+ * strand during busy loops.
+ */
 #define cpu_relax()	asm volatile("\n99:\n\t"			\
 				     "rd	%%ccr, %%g0\n\t"	\
 				     "rd	%%ccr, %%g0\n\t"	\
 				     "rd	%%ccr, %%g0\n\t"	\
-				     ".section	.pause_patch,\"ax\"\n\t"\
+				     ".section	.pause_3insn_patch,\"ax\"\n\t"\
 				     ".word	99b\n\t"		\
 				     "wr	%%g0, 128, %%asr27\n\t"	\
 				     "nop\n\t"				\
diff --git a/arch/sparc/kernel/entry.h b/arch/sparc/kernel/entry.h
index 51742df63c7..cc3c5cb47cd 100644
--- a/arch/sparc/kernel/entry.h
+++ b/arch/sparc/kernel/entry.h
@@ -63,8 +63,8 @@ struct pause_patch_entry {
 	unsigned int	addr;
 	unsigned int	insns[3];
 };
-extern struct pause_patch_entry __pause_patch,
-	__pause_patch_end;
+extern struct pause_patch_entry __pause_3insn_patch,
+	__pause_3insn_patch_end;
 
 extern void __init per_cpu_patch(void);
 extern void sun4v_patch_1insn_range(struct sun4v_1insn_patch_entry *,
diff --git a/arch/sparc/kernel/setup_64.c b/arch/sparc/kernel/setup_64.c
index b45cff408de..0eaf0059aae 100644
--- a/arch/sparc/kernel/setup_64.c
+++ b/arch/sparc/kernel/setup_64.c
@@ -320,8 +320,8 @@ static void __init pause_patch(void)
 {
 	struct pause_patch_entry *p;
 
-	p = &__pause_patch;
-	while (p < &__pause_patch_end) {
+	p = &__pause_3insn_patch;
+	while (p < &__pause_3insn_patch_end) {
 		unsigned long i, addr = p->addr;
 
 		for (i = 0; i < 3; i++) {
diff --git a/arch/sparc/kernel/vmlinux.lds.S b/arch/sparc/kernel/vmlinux.lds.S
index 847f9f79361..0bacceb1915 100644
--- a/arch/sparc/kernel/vmlinux.lds.S
+++ b/arch/sparc/kernel/vmlinux.lds.S
@@ -132,10 +132,10 @@ SECTIONS
 		*(.popc_6insn_patch)
 		__popc_6insn_patch_end = .;
 	}
-	.pause_patch : {
-		__pause_patch = .;
-		*(.pause_patch)
-		__pause_patch_end = .;
+	.pause_3insn_patch : {
+		__pause_3insn_patch = .;
+		*(.pause_3insn_patch)
+		__pause_3insn_patch_end = .;
 	}
 	PERCPU_SECTION(SMP_CACHE_BYTES)
 
-- 
cgit v1.2.3-18-g5258


From 1df35f80f9d3bba7cd434b64c9eaff8c9109abad Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Sun, 28 Oct 2012 13:15:09 -0700
Subject: sparc: Wire up sys_kcmp.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc/include/uapi/asm/unistd.h | 3 ++-
 arch/sparc/kernel/systbls_32.S       | 1 +
 arch/sparc/kernel/systbls_64.S       | 4 ++--
 3 files changed, 5 insertions(+), 3 deletions(-)

(limited to 'arch')

diff --git a/arch/sparc/include/uapi/asm/unistd.h b/arch/sparc/include/uapi/asm/unistd.h
index bed86a820d0..cac719d1bc5 100644
--- a/arch/sparc/include/uapi/asm/unistd.h
+++ b/arch/sparc/include/uapi/asm/unistd.h
@@ -406,8 +406,9 @@
 #define __NR_process_vm_readv	338
 #define __NR_process_vm_writev	339
 #define __NR_kern_features	340
+#define __NR_kcmp		341
 
-#define NR_syscalls		341
+#define NR_syscalls		342
 
 /* Bitmask values returned from kern_features system call.  */
 #define KERN_FEATURE_MIXED_MODE_STACK	0x00000001
diff --git a/arch/sparc/kernel/systbls_32.S b/arch/sparc/kernel/systbls_32.S
index 63402f9e9f5..5147f574f12 100644
--- a/arch/sparc/kernel/systbls_32.S
+++ b/arch/sparc/kernel/systbls_32.S
@@ -85,3 +85,4 @@ sys_call_table:
 /*325*/	.long sys_pwritev, sys_rt_tgsigqueueinfo, sys_perf_event_open, sys_recvmmsg, sys_fanotify_init
 /*330*/	.long sys_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, sys_open_by_handle_at, sys_clock_adjtime
 /*335*/	.long sys_syncfs, sys_sendmmsg, sys_setns, sys_process_vm_readv, sys_process_vm_writev
+/*340*/	.long sys_ni_syscall, sys_kcmp
diff --git a/arch/sparc/kernel/systbls_64.S b/arch/sparc/kernel/systbls_64.S
index 45ce6be088e..1c9af9fa38e 100644
--- a/arch/sparc/kernel/systbls_64.S
+++ b/arch/sparc/kernel/systbls_64.S
@@ -86,7 +86,7 @@ sys_call_table32:
 	.word compat_sys_pwritev, compat_sys_rt_tgsigqueueinfo, sys_perf_event_open, compat_sys_recvmmsg, sys_fanotify_init
 /*330*/	.word sys32_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, compat_sys_open_by_handle_at, compat_sys_clock_adjtime
 	.word sys_syncfs, compat_sys_sendmmsg, sys_setns, compat_sys_process_vm_readv, compat_sys_process_vm_writev
-/*340*/	.word sys_kern_features
+/*340*/	.word sys_kern_features, sys_kcmp
 
 #endif /* CONFIG_COMPAT */
 
@@ -164,4 +164,4 @@ sys_call_table:
 	.word sys_pwritev, sys_rt_tgsigqueueinfo, sys_perf_event_open, sys_recvmmsg, sys_fanotify_init
 /*330*/	.word sys_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, sys_open_by_handle_at, sys_clock_adjtime
 	.word sys_syncfs, sys_sendmmsg, sys_setns, sys_process_vm_readv, sys_process_vm_writev
-/*340*/	.word sys_kern_features
+/*340*/	.word sys_kern_features, sys_kcmp
-- 
cgit v1.2.3-18-g5258


From 6a4dae5e138a32b45ca5218cc2b81802f9d378c3 Mon Sep 17 00:00:00 2001
From: Felipe Balbi 2 <balbi@ti.com>
Date: Tue, 23 Oct 2012 19:00:03 +0100
Subject: ARM: 7565/1: sched: stop sched_clock() during suspend

The scheduler imposes a requirement to sched_clock()
which is to stop the clock during suspend, if we don't
do that any RT thread will be rescheduled in the future
which might cause any sort of problems.

This became an issue on OMAP when we converted omap-i2c.c
to use threaded IRQs, it turned out that depending on how
much time we spent on suspend, the I2C IRQ thread would
end up being rescheduled so far in the future that I2C
transfers would timeout and, because omap_hsmmc depends
on an I2C-connected device to detect if an MMC card is
inserted in the slot, our rootfs would just vanish.

arch/arm/kernel/sched_clock.c already had an optional
implementation (sched_clock_needs_suspend()) which would
handle scheduler's requirement properly, what this patch
does is simply to make that implementation non-optional.

Note that this has the side-effect that printk timings
won't reflect the actual time spent on suspend so other
methods to measure that will have to be used.

This h