[PATCH 0/7] perf updates and fixes

[PATCH 0/7] perf updates and fixes

[Frederic Weisbecker]

Hi,

The series is not yet mergeable because it would break PowerPc
(hot regs snapshot API has been changed, and I don't know how
to update PowerPc for that).

But if you're fine with the ideas, I can integrate the necessary
changes to fix this, and also separate fixes and updates.

Thanks.

Frederic Weisbecker (7):
  perf: Drop the frame reliablity check
  perf: Fetch hot regs from the template caller
  x86: Unify dumpstack.h and stacktrace.h
  perf: Move perf_arch_fetch_caller_regs into a macro
  perf: Make perf_fetch_caller_regs rewind to the first caller only
  perf: Use hot regs with software sched/migrate events
  perf: Correctly align perf event tracing buffer

 arch/x86/include/asm/perf_event.h |   10 ++++++-
 arch/x86/include/asm/stacktrace.h |   45 ++++++++++++++++++++++++++++++++
 arch/x86/kernel/cpu/perf_event.c  |   18 +------------
 arch/x86/kernel/dumpstack.c       |    1 -
 arch/x86/kernel/dumpstack.h       |   47 ----------------------------------
 arch/x86/kernel/dumpstack_32.c    |    2 -
 arch/x86/kernel/dumpstack_64.c    |    2 -
 arch/x86/kernel/stacktrace.c      |    7 +++--
 include/linux/perf_event.h        |   51 ++++++++++++++----------------------
 include/trace/ftrace.h            |   23 ++++++++--------
 kernel/perf_event.c               |   10 +------
 kernel/trace/trace_event_perf.c   |   13 ++++++---
 12 files changed, 101 insertions(+), 128 deletions(-)
 delete mode 100644 arch/x86/kernel/dumpstack.h

[PATCH 1/7] perf: Drop the frame reliablity check

[Frederic Weisbecker]

It is useless now that we have a pure stack frame
walker, as given addr are always reliable.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
---
 arch/x86/kernel/cpu/perf_event.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index f571f51..9bc6550 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1597,8 +1597,7 @@ static void backtrace_address(void *data, unsigned long addr, int reliable)
 {
 	struct perf_callchain_entry *entry = data;
 
-	if (reliable)
-		callchain_store(entry, addr);
+	callchain_store(entry, addr);
 }
 
 static const struct stacktrace_ops backtrace_ops = {
-- 
1.6.2.3

[PATCH 2/7] perf: Fetch hot regs from the template caller

[Frederic Weisbecker]

Trace events can be defined from a template using
DECLARE_EVENT_CLASS/DEFINE_EVENT or directly with TRACE_EVENT.

In both cases we have a template tracepoint handler, used to
record the trace, to which we pass our ftrace event instance.

In the function level, if the class is named "foo" and the event
is named "blah", we have the following chain of calls:

perf_trace_blah() -> perf_trace_templ_foo()

In the case we have several events sharing the class "blah",
we'll have multiple users of perf_trace_templ_foo(), and it
won't be inlined by the compiler. This is usually what happens
with the DECLARE_EVENT_CLASS/DEFINE_EVENT based definition.

But if perf_trace_blah() is the only caller of perf_trace_templ_foo()
there are fair chances that it will be inlined.

The problem is that we fetch the regs from perf_trace_templ_foo()
after we rewinded the frame pointer to the second caller, we want
to reach the caller of perf_trace_blah() to get the right source
of the event. And we do this by always assuming that
perf_trace_templ_foo() is not inlined. But as shown above this
is not always true. And if it is inlined we miss the first caller,
losing the most important level of precision.

We get:
	    61.31%       ls  [kernel.kallsyms]  [k] do_softirq
                         |
                         --- do_softirq
                             irq_exit
                             do_IRQ
                             common_interrupt
                            |
                            |--25.00%-- tty_buffer_request_room

Instead of:
	    61.31%       ls  [kernel.kallsyms]  [k] __do_softirq
                         |
                         --- __do_softirq
                             do_softirq
                             irq_exit
                             do_IRQ
                             common_interrupt
                            |
                            |--25.00%-- tty_buffer_request_room

To fix this, we fetch the regs from perf_trace_blah() rather than
perf_trace_templ_foo() so that we don't have to deal with inlining
surprises.

That also bring us the advantage of having the true source of the
event even if we don't have frame pointers.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
---
 include/trace/ftrace.h |   23 ++++++++++++-----------
 1 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h
index ea6f9d4..882c648 100644
--- a/include/trace/ftrace.h
+++ b/include/trace/ftrace.h
@@ -758,13 +758,12 @@ __attribute__((section("_ftrace_events"))) event_##call = {		\
 #define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print)	\
 static notrace void							\
 perf_trace_templ_##call(struct ftrace_event_call *event_call,		\
-			    proto)					\
+			struct pt_regs *__regs, proto)			\
 {									\
 	struct ftrace_data_offsets_##call __maybe_unused __data_offsets;\
 	struct ftrace_raw_##call *entry;				\
 	u64 __addr = 0, __count = 1;					\
 	unsigned long irq_flags;					\
-	struct pt_regs *__regs;						\
 	int __entry_size;						\
 	int __data_size;						\
 	int rctx;							\
@@ -785,20 +784,22 @@ perf_trace_templ_##call(struct ftrace_event_call *event_call,		\
 									\
 	{ assign; }							\
 									\
-	__regs = &__get_cpu_var(perf_trace_regs);			\
-	perf_fetch_caller_regs(__regs, 2);				\
-									\
 	perf_trace_buf_submit(entry, __entry_size, rctx, __addr,	\
 			       __count, irq_flags, __regs);		\
 }
 
 #undef DEFINE_EVENT
-#define DEFINE_EVENT(template, call, proto, args)		\
-static notrace void perf_trace_##call(proto)			\
-{								\
-	struct ftrace_event_call *event_call = &event_##call;	\
-								\
-	perf_trace_templ_##template(event_call, args);		\
+#define DEFINE_EVENT(template, call, proto, args)			\
+static notrace void perf_trace_##call(proto)				\
+{									\
+	struct ftrace_event_call *event_call = &event_##call;		\
+	struct pt_regs *__regs = &get_cpu_var(perf_trace_regs);		\
+									\
+	perf_fetch_caller_regs(__regs, 1);				\
+									\
+	perf_trace_templ_##template(event_call, __regs, args);		\
+									\
+	put_cpu_var(perf_trace_regs);					\
 }
 
 #undef DEFINE_EVENT_PRINT
-- 
1.6.2.3

[PATCH 3/7] x86: Unify dumpstack.h and stacktrace.h

[Frederic Weisbecker]

arch/x86/include/asm/stacktrace.h and arch/x86/kernel/dumpstack.h
declare headers of objects about the same topic.
Actually most of the files that include stacktrace.h also include
dumpstack.h

Although dumpstack.h seems more reserved for internals of stack
traces, those are quite often needed to define specialized stack
trace operations. And perf event arch headers are going to need
access to such low level operations anyway. So don't continue to
bother with dumpstack.h as it's not anymore about isolated deep
internals.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
---
 arch/x86/include/asm/stacktrace.h |   46 ++++++++++++++++++++++++++++++++++++
 arch/x86/kernel/cpu/perf_event.c  |    2 -
 arch/x86/kernel/dumpstack.c       |    1 -
 arch/x86/kernel/dumpstack.h       |   47 -------------------------------------
 arch/x86/kernel/dumpstack_32.c    |    2 -
 arch/x86/kernel/dumpstack_64.c    |    2 -
 arch/x86/kernel/stacktrace.c      |    7 +++--
 7 files changed, 50 insertions(+), 57 deletions(-)
 delete mode 100644 arch/x86/kernel/dumpstack.h

diff --git a/arch/x86/include/asm/stacktrace.h b/arch/x86/include/asm/stacktrace.h
index 4dab78e..8fb70b7 100644
--- a/arch/x86/include/asm/stacktrace.h
+++ b/arch/x86/include/asm/stacktrace.h
@@ -1,3 +1,8 @@
+/*
+ *  Copyright (C) 1991, 1992  Linus Torvalds
+ *  Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs
+ */
+
 #ifndef _ASM_X86_STACKTRACE_H
 #define _ASM_X86_STACKTRACE_H
 
@@ -38,8 +43,49 @@ struct stacktrace_ops {
 	walk_stack_t	walk_stack;
 };
 
+struct pt_regs;
+struct task_struct;
+
 void dump_trace(struct task_struct *tsk, struct pt_regs *regs,
 		unsigned long *stack, unsigned long bp,
 		const struct stacktrace_ops *ops, void *data);
 
+#ifdef CONFIG_X86_32
+#define STACKSLOTS_PER_LINE 8
+#define get_bp(bp) asm("movl %%ebp, %0" : "=r" (bp) :)
+#else
+#define STACKSLOTS_PER_LINE 4
+#define get_bp(bp) asm("movq %%rbp, %0" : "=r" (bp) :)
+#endif
+
+extern void
+show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
+		unsigned long *stack, unsigned long bp, char *log_lvl);
+
+extern void
+show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
+		unsigned long *sp, unsigned long bp, char *log_lvl);
+
+extern unsigned int code_bytes;
+
+/* The form of the top of the frame on the stack */
+struct stack_frame {
+	struct stack_frame *next_frame;
+	unsigned long return_address;
+};
+
+static inline unsigned long rewind_frame_pointer(int n)
+{
+	struct stack_frame *frame;
+
+	get_bp(frame);
+
+#ifdef CONFIG_FRAME_POINTER
+	while (n--)
+		frame = frame->next_frame;
+#endif
+
+	return (unsigned long)frame;
+}
+
 #endif /* _ASM_X86_STACKTRACE_H */
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 9bc6550..c3f203e 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1608,8 +1608,6 @@ static const struct stacktrace_ops backtrace_ops = {
 	.walk_stack		= print_context_stack_bp,
 };
 
-#include "../dumpstack.h"
-
 static void
 perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry)
 {
diff --git a/arch/x86/kernel/dumpstack.c b/arch/x86/kernel/dumpstack.c
index 6d81755..b5329df 100644
--- a/arch/x86/kernel/dumpstack.c
+++ b/arch/x86/kernel/dumpstack.c
@@ -18,7 +18,6 @@
 
 #include <asm/stacktrace.h>
 
-#include "dumpstack.h"
 
 int panic_on_unrecovered_nmi;
 int panic_on_io_nmi;
diff --git a/arch/x86/kernel/dumpstack.h b/arch/x86/kernel/dumpstack.h
deleted file mode 100644
index 29e5f7c..0000000
--- a/arch/x86/kernel/dumpstack.h
+++ /dev/null
@@ -1,47 +0,0 @@
-/*
- *  Copyright (C) 1991, 1992  Linus Torvalds
- *  Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs
- */
-
-#ifndef DUMPSTACK_H
-#define DUMPSTACK_H
-
-#ifdef CONFIG_X86_32
-#define STACKSLOTS_PER_LINE 8
-#define get_bp(bp) asm("movl %%ebp, %0" : "=r" (bp) :)
-#else
-#define STACKSLOTS_PER_LINE 4
-#define get_bp(bp) asm("movq %%rbp, %0" : "=r" (bp) :)
-#endif
-
-extern void
-show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
-		unsigned long *stack, unsigned long bp, char *log_lvl);
-
-extern void
-show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs,
-		unsigned long *sp, unsigned long bp, char *log_lvl);
-
-extern unsigned int code_bytes;
-
-/* The form of the top of the frame on the stack */
-struct stack_frame {
-	struct stack_frame *next_frame;
-	unsigned long return_address;
-};
-
-static inline unsigned long rewind_frame_pointer(int n)
-{
-	struct stack_frame *frame;
-
-	get_bp(frame);
-
-#ifdef CONFIG_FRAME_POINTER
-	while (n--)
-		frame = frame->next_frame;
-#endif
-
-	return (unsigned long)frame;
-}
-
-#endif /* DUMPSTACK_H */
diff --git a/arch/x86/kernel/dumpstack_32.c b/arch/x86/kernel/dumpstack_32.c
index 11540a1..0f6376f 100644
--- a/arch/x86/kernel/dumpstack_32.c
+++ b/arch/x86/kernel/dumpstack_32.c
@@ -16,8 +16,6 @@
 
 #include <asm/stacktrace.h>
 
-#include "dumpstack.h"
-
 
 void dump_trace(struct task_struct *task, struct pt_regs *regs,
 		unsigned long *stack, unsigned long bp,
diff --git a/arch/x86/kernel/dumpstack_64.c b/arch/x86/kernel/dumpstack_64.c
index 272c9f1..77fc784 100644
--- a/arch/x86/kernel/dumpstack_64.c
+++ b/arch/x86/kernel/dumpstack_64.c
@@ -16,8 +16,6 @@
 
 #include <asm/stacktrace.h>
 
-#include "dumpstack.h"
-
 #define N_EXCEPTION_STACKS_END \
 		(N_EXCEPTION_STACKS + DEBUG_STKSZ/EXCEPTION_STKSZ - 2)
 
diff --git a/arch/x86/kernel/stacktrace.c b/arch/x86/kernel/stacktrace.c
index 922eefb..ea54d02 100644
--- a/arch/x86/kernel/stacktrace.c
+++ b/arch/x86/kernel/stacktrace.c
@@ -96,12 +96,13 @@ EXPORT_SYMBOL_GPL(save_stack_trace_tsk);
 
 /* Userspace stacktrace - based on kernel/trace/trace_sysprof.c */
 
-struct stack_frame {
+struct stack_frame_user {
 	const void __user	*next_fp;
 	unsigned long		ret_addr;
 };
 
-static int copy_stack_frame(const void __user *fp, struct stack_frame *frame)
+static int
+copy_stack_frame(const void __user *fp, struct stack_frame_user *frame)
 {
 	int ret;
 
@@ -126,7 +127,7 @@ static inline void __save_stack_trace_user(struct stack_trace *trace)
 		trace->entries[trace->nr_entries++] = regs->ip;
 
 	while (trace->nr_entries < trace->max_entries) {
-		struct stack_frame frame;
+		struct stack_frame_user frame;
 
 		frame.next_fp = NULL;
 		frame.ret_addr = 0;
-- 
1.6.2.3

[PATCH 4/7] perf: Move perf_arch_fetch_caller_regs into a macro

[Frederic Weisbecker]

Having perf_arch_fetch_caller_regs() as a weak function overridable
by macros makes it hard to export its symbol for modules, as we
need to export the symbol in another file than the generic weak
definition to deal with compiler bugs.

Currently it's not a problem because we can define the symbol in
trace_event_perf.c as only trace events use it. But we are going
to make some generic software events to use this new facility,
so the symbol must be available even if trace events are not built.

This patch fixes the problem by making perf_arch_fetch_caller_regs
a macro that archs can define in their <asm/perf_event.h>

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: Archs <linux-arch@vger.kernel.org>
---
 arch/x86/include/asm/perf_event.h |   10 +++++++++-
 arch/x86/kernel/cpu/perf_event.c  |   13 -------------
 include/linux/perf_event.h        |    8 +++++---
 kernel/perf_event.c               |    7 -------
 kernel/trace/trace_event_perf.c   |    2 --
 5 files changed, 14 insertions(+), 26 deletions(-)

diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 124dddd..4bf3d37 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -155,7 +155,15 @@ extern void perf_events_lapic_init(void);
 
 #define perf_instruction_pointer(regs)	((regs)->ip)
 
-#else
+#include <asm/stacktrace.h>
+
+#define perf_arch_fetch_caller_regs(regs, ip, skip)	\
+	(regs)->ip = ip;				\
+	(regs)->bp = rewind_frame_pointer(skip);	\
+	(regs)->cs = __KERNEL_CS;			\
+	local_save_flags((regs)->flags);
+
+#else /* !CONFIG_PERF_EVENTS */
 static inline void init_hw_perf_events(void)		{ }
 static inline void perf_events_lapic_init(void)	{ }
 #endif
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index c3f203e..43fae02 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1691,16 +1691,3 @@ struct perf_callchain_entry *perf_callchain(struct pt_regs *regs)
 	return entry;
 }
 
-#ifdef CONFIG_EVENT_TRACING
-void perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int skip)
-{
-	regs->ip = ip;
-	/*
-	 * perf_arch_fetch_caller_regs adds another call, we need to increment
-	 * the skip level
-	 */
-	regs->bp = rewind_frame_pointer(skip + 1);
-	regs->cs = __KERNEL_CS;
-	local_save_flags(regs->flags);
-}
-#endif
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 2bccb7b..76b680f 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -867,8 +867,10 @@ perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr)
 		__perf_sw_event(event_id, nr, nmi, regs, addr);
 }
 
-extern void
-perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int skip);
+#ifndef perf_arch_fetch_caller_regs
+static inline void
+perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int skip){ }
+#endif
 
 /*
  * Take a snapshot of the regs. Skip ip and frame pointer to
@@ -902,7 +904,7 @@ static inline void perf_fetch_caller_regs(struct pt_regs *regs, int skip)
 		ip = 0;
 	}
 
-	return perf_arch_fetch_caller_regs(regs, ip, skip);
+	perf_arch_fetch_caller_regs(regs, ip, skip);
 }
 
 extern void __perf_event_mmap(struct vm_area_struct *vma);
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 455393e..bbed6f0 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -2790,13 +2790,6 @@ __weak struct perf_callchain_entry *perf_callchain(struct pt_regs *regs)
 	return NULL;
 }
 
-#ifdef CONFIG_EVENT_TRACING
-__weak
-void perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int skip)
-{
-}
-#endif
-
 /*
  * Output
  */
diff --git a/kernel/trace/trace_event_perf.c b/kernel/trace/trace_event_perf.c
index 81f691e..8e9edcd 100644
--- a/kernel/trace/trace_event_perf.c
+++ b/kernel/trace/trace_event_perf.c
@@ -12,8 +12,6 @@
 DEFINE_PER_CPU(struct pt_regs, perf_trace_regs);
 EXPORT_PER_CPU_SYMBOL_GPL(perf_trace_regs);
 
-EXPORT_SYMBOL_GPL(perf_arch_fetch_caller_regs);
-
 static char *perf_trace_buf;
 static char *perf_trace_buf_nmi;
 
-- 
1.6.2.3

[PATCH 5/7] perf: Make perf_fetch_caller_regs rewind to the first caller only

[Frederic Weisbecker]

Trace events now only need to rewind the regs to the immediate
caller, so we don't need anymore to implement the support for
further stack walking.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: Archs <linux-arch@vger.kernel.org>
---
 arch/x86/include/asm/perf_event.h |    6 +++---
 arch/x86/include/asm/stacktrace.h |    5 ++---
 include/linux/perf_event.h        |   30 +++++-------------------------
 include/trace/ftrace.h            |    2 +-
 4 files changed, 11 insertions(+), 32 deletions(-)

diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 4bf3d37..1df2317 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -157,9 +157,9 @@ extern void perf_events_lapic_init(void);
 
 #include <asm/stacktrace.h>
 
-#define perf_arch_fetch_caller_regs(regs, ip, skip)	\
-	(regs)->ip = ip;				\
-	(regs)->bp = rewind_frame_pointer(skip);	\
+#define perf_arch_fetch_caller_regs(regs, __ip)		\
+	(regs)->ip = __ip;				\
+	(regs)->bp = caller_frame_pointer();		\
 	(regs)->cs = __KERNEL_CS;			\
 	local_save_flags((regs)->flags);
 
diff --git a/arch/x86/include/asm/stacktrace.h b/arch/x86/include/asm/stacktrace.h
index 8fb70b7..62c9897 100644
--- a/arch/x86/include/asm/stacktrace.h
+++ b/arch/x86/include/asm/stacktrace.h
@@ -74,15 +74,14 @@ struct stack_frame {
 	unsigned long return_address;
 };
 
-static inline unsigned long rewind_frame_pointer(int n)
+static inline unsigned long caller_frame_pointer(void)
 {
 	struct stack_frame *frame;
 
 	get_bp(frame);
 
 #ifdef CONFIG_FRAME_POINTER
-	while (n--)
-		frame = frame->next_frame;
+	frame = frame->next_frame;
 #endif
 
 	return (unsigned long)frame;
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 76b680f..3b59cf7 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -869,42 +869,22 @@ perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr)
 
 #ifndef perf_arch_fetch_caller_regs
 static inline void
-perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip, int skip){ }
+perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip) { }
 #endif
 
 /*
- * Take a snapshot of the regs. Skip ip and frame pointer to
- * the nth caller. We only need a few of the regs:
+ * Take a snapshot of the regs. Rewind ip and frame pointer to
+ * the caller. We only need a few of the regs:
  * - ip for PERF_SAMPLE_IP
  * - cs for user_mode() tests
  * - bp for callchains
  * - eflags, for future purposes, just in case
  */
-static inline void perf_fetch_caller_regs(struct pt_regs *regs, int skip)
+static inline void perf_fetch_caller_regs(struct pt_regs *regs)
 {
-	unsigned long ip;
-
 	memset(regs, 0, sizeof(*regs));
 
-	switch (skip) {
-	case 1 :
-		ip = CALLER_ADDR0;
-		break;
-	case 2 :
-		ip = CALLER_ADDR1;
-		break;
-	case 3 :
-		ip = CALLER_ADDR2;
-		break;
-	case 4:
-		ip = CALLER_ADDR3;
-		break;
-	/* No need to support further for now */
-	default:
-		ip = 0;
-	}
-
-	perf_arch_fetch_caller_regs(regs, ip, skip);
+	perf_arch_fetch_caller_regs(regs, CALLER_ADDR0);
 }
 
 extern void __perf_event_mmap(struct vm_area_struct *vma);
diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h
index 882c648..82e9977 100644
--- a/include/trace/ftrace.h
+++ b/include/trace/ftrace.h
@@ -795,7 +795,7 @@ static notrace void perf_trace_##call(proto)				\
 	struct ftrace_event_call *event_call = &event_##call;		\
 	struct pt_regs *__regs = &get_cpu_var(perf_trace_regs);		\
 									\
-	perf_fetch_caller_regs(__regs, 1);				\
+	perf_fetch_caller_regs(__regs);					\
 									\
 	perf_trace_templ_##template(event_call, __regs, args);		\
 									\
-- 
1.6.2.3

[PATCH 6/7] perf: Use hot regs with software sched/migrate events

[Frederic Weisbecker]

Scheduler's task migration events don't work because they always
pass NULL regs perf_sw_event(). The event hence gets filtered
in perf_swevent_add().

Scheduler's context switches events use task_pt_regs() to get
the context when the event occured which is a wrong thing to
do as this won't give us the place in the kernel where we went
to sleep but the place where we left userspace. The result is
even more wrong if we switch from a kernel thread.

Use the hot regs snapshot for both events as they belong to the
non-interrupt/exception based events family. Unlike page faults
or so that provide the regs matching the exact origin of the event,
we need to save the current context.

This makes the task migration event working and fix the context
switch callchains and origin ip.

Example: perf record -a -e cs

Before:

    10.91%      ksoftirqd/0                  0  [k] 0000000000000000
                |
                --- (nil)
                    perf_callchain
                    perf_prepare_sample
                    __perf_event_overflow
                    perf_swevent_overflow
                    perf_swevent_add
                    perf_swevent_ctx_event
                    do_perf_sw_event
                    __perf_sw_event
                    perf_event_task_sched_out
                    schedule
                    run_ksoftirqd
                    kthread
                    kernel_thread_helper

After:

    23.77%  hald-addon-stor  [kernel.kallsyms]  [k] schedule
            |
            --- schedule
               |
               |--60.00%-- schedule_timeout
               |          wait_for_common
               |          wait_for_completion
               |          blk_execute_rq
               |          scsi_execute
               |          scsi_execute_req
               |          sr_test_unit_ready
               |          |
               |          |--66.67%-- sr_media_change
               |          |          media_changed
               |          |          cdrom_media_changed
               |          |          sr_block_media_changed
               |          |          check_disk_change
               |          |          cdrom_open

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
---
 include/linux/perf_event.h |   21 ++++++++++++++-------
 kernel/perf_event.c        |    3 +--
 2 files changed, 15 insertions(+), 9 deletions(-)

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 3b59cf7..613b419 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -860,13 +860,6 @@ extern atomic_t perf_swevent_enabled[PERF_COUNT_SW_MAX];
 
 extern void __perf_sw_event(u32, u64, int, struct pt_regs *, u64);
 
-static inline void
-perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr)
-{
-	if (atomic_read(&perf_swevent_enabled[event_id]))
-		__perf_sw_event(event_id, nr, nmi, regs, addr);
-}
-
 #ifndef perf_arch_fetch_caller_regs
 static inline void
 perf_arch_fetch_caller_regs(struct pt_regs *regs, unsigned long ip) { }
@@ -887,6 +880,20 @@ static inline void perf_fetch_caller_regs(struct pt_regs *regs)
 	perf_arch_fetch_caller_regs(regs, CALLER_ADDR0);
 }
 
+static inline void
+perf_sw_event(u32 event_id, u64 nr, int nmi, struct pt_regs *regs, u64 addr)
+{
+	if (atomic_read(&perf_swevent_enabled[event_id])) {
+		struct pt_regs caller_regs;
+
+		if (!regs) {
+			perf_fetch_caller_regs(&caller_regs);
+			regs = &caller_regs;
+		}
+		__perf_sw_event(event_id, nr, nmi, regs, addr);
+	}
+}
+
 extern void __perf_event_mmap(struct vm_area_struct *vma);
 
 static inline void perf_event_mmap(struct vm_area_struct *vma)
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index bbed6f0..8b392c3 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -1167,8 +1167,7 @@ void perf_event_task_sched_out(struct task_struct *task,
 	struct pt_regs *regs;
 	int do_switch = 1;
 
-	regs = task_pt_regs(task);
-	perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, 1, regs, 0);
+	perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, 1, NULL, 0);
 
 	if (likely(!ctx || !cpuctx->task_ctx))
 		return;
-- 
1.6.2.3

[PATCH 7/7] perf: Correctly align perf event tracing buffer

[Frederic Weisbecker]

The trace event buffer used by perf to record raw sample events
is typed as an array of char and may then not be aligned to 8
by alloc_percpu().

But we need it to be aligned to 8 in sparc64 because we cast
this buffer into a random structure type built by the TRACE_EVENT()
macro to store the traces. So if a random 64 bits field is accessed
inside, it may be not under an expected good alignment.

Use an array of long instead to force the appropriate alignment, and
perform a compile time check to ensure the size in byte of the buffer
is a multiple of sizeof(long) so that its actual size doesn't get
shrinked under us.

This fixes unaligned accesses reported while using perf lock
in sparc 64.

Suggested-by: David Miller <davem@davemloft.net>
Suggested-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: Tejun Heo <htejun@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
---
 kernel/trace/trace_event_perf.c |   11 +++++++++--
 1 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/trace_event_perf.c b/kernel/trace/trace_event_perf.c
index 8e9edcd..30314f5 100644
--- a/kernel/trace/trace_event_perf.c
+++ b/kernel/trace/trace_event_perf.c
@@ -15,7 +15,12 @@ EXPORT_PER_CPU_SYMBOL_GPL(perf_trace_regs);
 static char *perf_trace_buf;
 static char *perf_trace_buf_nmi;
 
-typedef typeof(char [PERF_MAX_TRACE_SIZE]) perf_trace_t ;
+/*
+ * Force it to be aligned to unsigned long to avoid misaligned accesses
+ * suprises
+ */
+typedef typeof(unsigned long [PERF_MAX_TRACE_SIZE / sizeof(unsigned long)])
+	perf_trace_t;
 
 /* Count the events in use (per event id, not per instance) */
 static int	total_ref_count;
@@ -128,6 +133,8 @@ __kprobes void *perf_trace_buf_prepare(int size, unsigned short type,
 	char *trace_buf, *raw_data;
 	int pc, cpu;
 
+	BUILD_BUG_ON(PERF_MAX_TRACE_SIZE % sizeof(unsigned long));
+
 	pc = preempt_count();
 
 	/* Protect the per cpu buffer, begin the rcu read side */
@@ -150,7 +157,7 @@ __kprobes void *perf_trace_buf_prepare(int size, unsigned short type,
 	raw_data = per_cpu_ptr(trace_buf, cpu);
 
 	/* zero the dead bytes from align to not leak stack to user */
-	*(u64 *)(&raw_data[size - sizeof(u64)]) = 0ULL;
+	memset(&raw_data[size - sizeof(u64)], 0, sizeof(u64));
 
 	entry = (struct trace_entry *)raw_data;
 	tracing_generic_entry_update(entry, *irq_flags, pc);
-- 
1.6.2.3

Re: [PATCH 0/7] perf updates and fixes

[Paul Mackerras]

On Fri, Mar 26, 2010 at 02:52:35AM +0100, Frederic Weisbecker wrote:

> The series is not yet mergeable because it would break PowerPc
> (hot regs snapshot API has been changed, and I don't know how
> to update PowerPc for that).
> 
> But if you're fine with the ideas, I can integrate the necessary
> changes to fix this, and also separate fixes and updates.

The patch below adds the necessary stuff for powerpc.  You could fold
it into your "perf: Move perf_arch_fetch_caller_regs into a macro"
patch, or keep it as a separate patch in the series (though that would
make preserving bisectability more difficult).

There is a little difficulty in that you first create a 3-argument
form of perf_arch_fetch_caller_regs and then remove one argument in a
later patch, whereas my patch below just creates the 2-argument form.
I'm sure you can resolve that one way or another.

The patch to add the old-style perf_arch_fetch_caller_regs for powerpc
is sitting in the tip/perf/urgent branch but hasn't gone anywhere.
I suppose we need to get Ingo to pull that branch into tip/perf/core
and then include a revert patch in the series that switches to the new
interface.  Alternatively, if the urgent branch isn't append-only then
we could disappear it (the urgent branch would need to be rewound by
one commit).

Paul.
----------
[PATCH] perf/powerpc: Implement perf_arch_fetch_caller_regs for powerpc

This adds the ability to get a hot register snapshot for powerpc,
enabling us to get meaningful call chains for tracepoints and context
switch events.

Signed-off-by: Paul Mackerras <paulus@samba.org>
---
 arch/powerpc/include/asm/perf_event.h |   12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/arch/powerpc/include/asm/perf_event.h b/arch/powerpc/include/asm/perf_event.h
index e6d4ce6..5c16b89 100644
--- a/arch/powerpc/include/asm/perf_event.h
+++ b/arch/powerpc/include/asm/perf_event.h
@@ -21,3 +21,15 @@
 #ifdef CONFIG_FSL_EMB_PERF_EVENT
 #include <asm/perf_event_fsl_emb.h>
 #endif
+
+#ifdef CONFIG_PERF_EVENTS
+#include <asm/ptrace.h>
+#include <asm/reg.h>
+
+#define perf_arch_fetch_caller_regs(regs, __ip)			\
+	do {							\
+		(regs)->nip = __ip;				\
+		(regs)->gpr[1] = *(unsigned long *)__get_SP();	\
+		asm volatile("mfmsr %0" : "=r" ((regs)->msr));	\
+	} while (0)
+#endif

Re: [PATCH 0/7] perf updates and fixes

[Ingo Molnar]

* Paul Mackerras <paulus@samba.org> wrote:

> On Fri, Mar 26, 2010 at 02:52:35AM +0100, Frederic Weisbecker wrote:
> 
> > The series is not yet mergeable because it would break PowerPc (hot regs 
> > snapshot API has been changed, and I don't know how to update PowerPc for 
> > that).
> > 
> > But if you're fine with the ideas, I can integrate the necessary changes 
> > to fix this, and also separate fixes and updates.
> 
> The patch below adds the necessary stuff for powerpc.  You could fold it 
> into your "perf: Move perf_arch_fetch_caller_regs into a macro" patch, or 
> keep it as a separate patch in the series (though that would make preserving 
> bisectability more difficult).

Since the series needs a resend anyway folding back ought to be fine i think.

I'm wondering whether this should get into tip:perf/urgent - or in 
tip:perf/core for 2.6.35.

It fixes sw event call trace ugliness, but is that a 2.6.34 regression? Is 
there any other aspect of the series that points towards accelerating this 
into .34?

Thanks,

	Ingo

Re: [PATCH 0/7] perf updates and fixes

[Frederic Weisbecker]

On Fri, Mar 26, 2010 at 08:58:17AM +0100, Ingo Molnar wrote:
> 
> * Paul Mackerras <paulus@samba.org> wrote:
> 
> > On Fri, Mar 26, 2010 at 02:52:35AM +0100, Frederic Weisbecker wrote:
> > 
> > > The series is not yet mergeable because it would break PowerPc (hot regs 
> > > snapshot API has been changed, and I don't know how to update PowerPc for 
> > > that).
> > > 
> > > But if you're fine with the ideas, I can integrate the necessary changes 
> > > to fix this, and also separate fixes and updates.
> > 
> > The patch below adds the necessary stuff for powerpc.  You could fold it 
> > into your "perf: Move perf_arch_fetch_caller_regs into a macro" patch, or 
> > keep it as a separate patch in the series (though that would make preserving 
> > bisectability more difficult).
> 
> Since the series needs a resend anyway folding back ought to be fine i think.
> 
> I'm wondering whether this should get into tip:perf/urgent - or in 
> tip:perf/core for 2.6.35.
> 
> It fixes sw event call trace ugliness, but is that a 2.6.34 regression? Is 
> there any other aspect of the series that points towards accelerating this 
> into .34?


Let's have a look:

	perf: Correctly align perf event tracing buffer

Should probably go into urgent. The change is not invasive at
all. It doesn't fix a regression but it's still an important fix.

	The rest

It depends. The whole bunch is rather invasive.
The callchains of context switches never worked correctly
I think. I couldn't tell if the cpu migration has ever worked.
If it ever did, then it's a regression fix but in the middle of
too much hot regs improvements. So I can cook a specific fix for
the cpu migration event to work, and keep the rest for perf/core.

Re: [PATCH 0/7] perf updates and fixes

[Frederic Weisbecker]

On Fri, Mar 26, 2010 at 05:02:30PM +1100, Paul Mackerras wrote:
> On Fri, Mar 26, 2010 at 02:52:35AM +0100, Frederic Weisbecker wrote:
> 
> > The series is not yet mergeable because it would break PowerPc
> > (hot regs snapshot API has been changed, and I don't know how
> > to update PowerPc for that).
> > 
> > But if you're fine with the ideas, I can integrate the necessary
> > changes to fix this, and also separate fixes and updates.
> 
> The patch below adds the necessary stuff for powerpc.  You could fold
> it into your "perf: Move perf_arch_fetch_caller_regs into a macro"
> patch, or keep it as a separate patch in the series (though that would
> make preserving bisectability more difficult).
> 
> There is a little difficulty in that you first create a 3-argument
> form of perf_arch_fetch_caller_regs and then remove one argument in a
> later patch, whereas my patch below just creates the 2-argument form.
> I'm sure you can resolve that one way or another.
> 
> The patch to add the old-style perf_arch_fetch_caller_regs for powerpc
> is sitting in the tip/perf/urgent branch but hasn't gone anywhere.
> I suppose we need to get Ingo to pull that branch into tip/perf/core
> and then include a revert patch in the series that switches to the new
> interface.  Alternatively, if the urgent branch isn't append-only then
> we could disappear it (the urgent branch would need to be rewound by
> one commit).



Thanks a lot, that's axactly what I needed.
I'll sort it out to not break things in the middle :)



> 
> Paul.
> ----------
> [PATCH] perf/powerpc: Implement perf_arch_fetch_caller_regs for powerpc
> 
> This adds the ability to get a hot register snapshot for powerpc,
> enabling us to get meaningful call chains for tracepoints and context
> switch events.
> 
> Signed-off-by: Paul Mackerras <paulus@samba.org>
> ---
>  arch/powerpc/include/asm/perf_event.h |   12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 
> diff --git a/arch/powerpc/include/asm/perf_event.h b/arch/powerpc/include/asm/perf_event.h
> index e6d4ce6..5c16b89 100644
> --- a/arch/powerpc/include/asm/perf_event.h
> +++ b/arch/powerpc/include/asm/perf_event.h
> @@ -21,3 +21,15 @@
>  #ifdef CONFIG_FSL_EMB_PERF_EVENT
>  #include <asm/perf_event_fsl_emb.h>
>  #endif
> +
> +#ifdef CONFIG_PERF_EVENTS
> +#include <asm/ptrace.h>
> +#include <asm/reg.h>
> +
> +#define perf_arch_fetch_caller_regs(regs, __ip)			\
> +	do {							\
> +		(regs)->nip = __ip;				\
> +		(regs)->gpr[1] = *(unsigned long *)__get_SP();	\
> +		asm volatile("mfmsr %0" : "=r" ((regs)->msr));	\
> +	} while (0)
> +#endif

[BUG perf] perf_fetch_caller_regs / rewind_frame_pointer can panic

[Eric Dumazet]

Hello

Current linux-2.6 tree panics on my dev machine

64 bit kernel, 32bit user land
CONFIG_FRAME_POINTER=y

perf timechart record &

Instant crash

Call Trace:
 perf_trace_sched_switch+0xd5/0x120
 schedule+0x6b5/0x860
 retint_careful+0xd/0x21
 
RIP ffffffff81010955 perf_arch_fetch_caller_regs+0x15/0x40
CR2: 00000000d21f1422


rewind_frame_pointer() is probably wrong.

No test performed to check frame is in current stack, or
that (!user_mode_vm(regs))


static inline unsigned long rewind_frame_pointer(int n)
{
	struct stack_frame *frame;

	get_bp(frame);

#ifdef CONFIG_FRAME_POINTER
	while (n--)
		frame = frame->next_frame;
#endif

	return (unsigned long)frame;
}

Re: [BUG perf] perf_fetch_caller_regs / rewind_frame_pointer can panic

[Frederic Weisbecker]

On Thu, Apr 08, 2010 at 11:57:20AM +0200, Eric Dumazet wrote:
> Hello
> 
> Current linux-2.6 tree panics on my dev machine
> 
> 64 bit kernel, 32bit user land
> CONFIG_FRAME_POINTER=y
> 
> perf timechart record &
> 
> Instant crash
> 
> Call Trace:
>  perf_trace_sched_switch+0xd5/0x120
>  schedule+0x6b5/0x860
>  retint_careful+0xd/0x21
>  
> RIP ffffffff81010955 perf_arch_fetch_caller_regs+0x15/0x40
> CR2: 00000000d21f1422
> 
> 
> rewind_frame_pointer() is probably wrong.
> 
> No test performed to check frame is in current stack, or
> that (!user_mode_vm(regs))



user_mode_vm() can not work here as we are actually filling
regs from scratch.

But we indeed need to have a safe dereference to avoid such
crashes. A simple probe_kernel_address() should do the trick.

This API is going to change for the next cycle as it won't need
to rewind further than the first caller. So I'm going to do a
rough probe_kernel_address() fix for the current version. The next
one won't have this problem.



> 
> 
> static inline unsigned long rewind_frame_pointer(int n)
> {
> 	struct stack_frame *frame;
> 
> 	get_bp(frame);
> 
> #ifdef CONFIG_FRAME_POINTER
> 	while (n--)
> 		frame = frame->next_frame;
> #endif
> 
> 	return (unsigned long)frame;
> }
> 
> 
> 

[PATCH] perf: Fix unsafe frame rewinding with hot regs fetching

[Frederic Weisbecker]

On Thu, Apr 08, 2010 at 11:57:20AM +0200, Eric Dumazet wrote:
> Hello
> 
> Current linux-2.6 tree panics on my dev machine
> 
> 64 bit kernel, 32bit user land
> CONFIG_FRAME_POINTER=y
> 
> perf timechart record &
> 
> Instant crash
> 
> Call Trace:
>  perf_trace_sched_switch+0xd5/0x120
>  schedule+0x6b5/0x860
>  retint_careful+0xd/0x21
>  
> RIP ffffffff81010955 perf_arch_fetch_caller_regs+0x15/0x40
> CR2: 00000000d21f1422
> 
> 
> rewind_frame_pointer() is probably wrong.
> 
> No test performed to check frame is in current stack, or
> that (!user_mode_vm(regs))
> 
> 
> static inline unsigned long rewind_frame_pointer(int n)
> {
> 	struct stack_frame *frame;
> 
> 	get_bp(frame);
> 
> #ifdef CONFIG_FRAME_POINTER
> 	while (n--)
> 		frame = frame->next_frame;
> #endif
> 
> 	return (unsigned long)frame;
> }
> 
> 
> 


Can you please test this fix?

Thanks.

---
>From 60d5c4e8498efc4a01abceef54ad3bc91993bf41 Mon Sep 17 00:00:00 2001
From: Frederic Weisbecker <fweisbec@gmail.com>
Date: Thu, 8 Apr 2010 14:05:50 +0200
Subject: [PATCH] perf: Fix unsafe frame rewinding with hot regs fetching

When we fetch the hot regs and rewind to the nth caller, it
might happen that we dereference a frame pointer outside the
kernel stack boundaries, like in this example:

	perf_trace_sched_switch+0xd5/0x120
        schedule+0x6b5/0x860
        retint_careful+0xd/0x21

Since we directly dereference a userspace frame pointer here while
rewinding behind retint_careful, this may end up in a crash.

Fix this by simply using probe_kernel_address() when we rewind the
frame pointer.

This issue will have a much more proper fix in the next version of the
perf_arch_fetch_caller_regs() API that will only need to rewind to the
first caller.

Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: David Miller <davem@davemloft.net>
Cc: Archs <linux-arch@vger.kernel.org>
---
 arch/x86/kernel/dumpstack.h |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/dumpstack.h b/arch/x86/kernel/dumpstack.h
index e39e771..e1a93be 100644
--- a/arch/x86/kernel/dumpstack.h
+++ b/arch/x86/kernel/dumpstack.h
@@ -14,6 +14,8 @@
 #define get_bp(bp) asm("movq %%rbp, %0" : "=r" (bp) :)
 #endif
 
+#include <linux/uaccess.h>
+
 extern void
 show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
 		unsigned long *stack, unsigned long bp, char *log_lvl);
@@ -42,8 +44,10 @@ static inline unsigned long rewind_frame_pointer(int n)
 	get_bp(frame);
 
 #ifdef CONFIG_FRAME_POINTER
-	while (n--)
-		frame = frame->next_frame;
+	while (n--) {
+		if (probe_kernel_address(&frame->next_frame, frame))
+			break;
+	}
 #endif
 
 	return (unsigned long)frame;
-- 
1.6.2.3

Re: [PATCH] perf: Fix unsafe frame rewinding with hot regs fetching

[Eric Dumazet]

Le jeudi 08 avril 2010 à 14:32 +0200, Frederic Weisbecker a écrit :


> 
> Can you please test this fix?
> 
> Thanks.
> 
> ---
> From 60d5c4e8498efc4a01abceef54ad3bc91993bf41 Mon Sep 17 00:00:00 2001
> From: Frederic Weisbecker <fweisbec@gmail.com>
> Date: Thu, 8 Apr 2010 14:05:50 +0200
> Subject: [PATCH] perf: Fix unsafe frame rewinding with hot regs fetching
> 
> When we fetch the hot regs and rewind to the nth caller, it
> might happen that we dereference a frame pointer outside the
> kernel stack boundaries, like in this example:
> 
> 	perf_trace_sched_switch+0xd5/0x120
>         schedule+0x6b5/0x860
>         retint_careful+0xd/0x21
> 
> Since we directly dereference a userspace frame pointer here while
> rewinding behind retint_careful, this may end up in a crash.
> 
> Fix this by simply using probe_kernel_address() when we rewind the
> frame pointer.
> 
> This issue will have a much more proper fix in the next version of the
> perf_arch_fetch_caller_regs() API that will only need to rewind to the
> first caller.
> 
> Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
> Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
> Cc: Paul Mackerras <paulus@samba.org>
> Cc: David Miller <davem@davemloft.net>
> Cc: Archs <linux-arch@vger.kernel.org>
> ---
>  arch/x86/kernel/dumpstack.h |    8 ++++++--
>  1 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/kernel/dumpstack.h b/arch/x86/kernel/dumpstack.h
> index e39e771..e1a93be 100644
> --- a/arch/x86/kernel/dumpstack.h
> +++ b/arch/x86/kernel/dumpstack.h
> @@ -14,6 +14,8 @@
>  #define get_bp(bp) asm("movq %%rbp, %0" : "=r" (bp) :)
>  #endif
>  
> +#include <linux/uaccess.h>
> +
>  extern void
>  show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
>  		unsigned long *stack, unsigned long bp, char *log_lvl);
> @@ -42,8 +44,10 @@ static inline unsigned long rewind_frame_pointer(int n)
>  	get_bp(frame);
>  
>  #ifdef CONFIG_FRAME_POINTER
> -	while (n--)
> -		frame = frame->next_frame;
> +	while (n--) {
> +		if (probe_kernel_address(&frame->next_frame, frame))
> +			break;
> +	}
>  #endif
>  
>  	return (unsigned long)frame;

Thanks, no more crash :)

Tested-by: Eric Dumazet <eric.dumazet@gmail.com>

[GIT PULL] perf fix

[Frederic Weisbecker]

Ingo,

Please pull the perf/urgent branch that can be found at:

git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git
	perf/urgent

Thanks,
	Frederic
---

Frederic Weisbecker (1):
      perf: Fix unsafe frame rewinding with hot regs fetching


 arch/x86/kernel/dumpstack.h |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

---
commit ab285f2b5290d92b7ec1a6f9aad54308dadf6157
Author: Frederic Weisbecker <fweisbec@gmail.com>
Date:   Thu Apr 8 14:05:50 2010 +0200

    perf: Fix unsafe frame rewinding with hot regs fetching
    
    When we fetch the hot regs and rewind to the nth caller, it
    might happen that we dereference a frame pointer outside the
    kernel stack boundaries, like in this example:
    
    	perf_trace_sched_switch+0xd5/0x120
            schedule+0x6b5/0x860
            retint_careful+0xd/0x21
    
    Since we directly dereference a userspace frame pointer here while
    rewinding behind retint_careful, this may end up in a crash.
    
    Fix this by simply using probe_kernel_address() when we rewind the
    frame pointer.
    
    This issue will have a much more proper fix in the next version of the
    perf_arch_fetch_caller_regs() API that will only need to rewind to the
    first caller.
    
    Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
    Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
    Tested-by: Eric Dumazet <eric.dumazet@gmail.com>
    Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
    Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
    Cc: Paul Mackerras <paulus@samba.org>
    Cc: David Miller <davem@davemloft.net>
    Cc: Archs <linux-arch@vger.kernel.org>

diff --git a/arch/x86/kernel/dumpstack.h b/arch/x86/kernel/dumpstack.h
index e39e771..e1a93be 100644
--- a/arch/x86/kernel/dumpstack.h
+++ b/arch/x86/kernel/dumpstack.h
@@ -14,6 +14,8 @@
 #define get_bp(bp) asm("movq %%rbp, %0" : "=r" (bp) :)
 #endif
 
+#include <linux/uaccess.h>
+
 extern void
 show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
 		unsigned long *stack, unsigned long bp, char *log_lvl);
@@ -42,8 +44,10 @@ static inline unsigned long rewind_frame_pointer(int n)
 	get_bp(frame);
 
 #ifdef CONFIG_FRAME_POINTER
-	while (n--)
-		frame = frame->next_frame;
+	while (n--) {
+		if (probe_kernel_address(&frame->next_frame, frame))
+			break;
+	}
 #endif
 
 	return (unsigned long)frame;

Re: [GIT PULL] perf fix

[Ingo Molnar]

* Frederic Weisbecker <fweisbec@gmail.com> wrote:

> Ingo,
> 
> Please pull the perf/urgent branch that can be found at:
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git
> 	perf/urgent
> 
> Thanks,
> 	Frederic
> ---
> 
> Frederic Weisbecker (1):
>       perf: Fix unsafe frame rewinding with hot regs fetching
> 
> 
>  arch/x86/kernel/dumpstack.h |    8 ++++++--
>  1 files changed, 6 insertions(+), 2 deletions(-)

Pulled, thanks a lot Frederic!

	Ingo

Re: [GIT PULL] perf fix

[pr-tracker-bot]

The pull request you sent on Sun, 7 Apr 2024 09:55:27 +0200:

> git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-2024-04-07

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/e2948effa9063ae8bd0e7dd65ca4a0fa068c71a6

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/prtracker.html

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf/urgent Git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-2024-04-07

   # HEAD: 312be9fc2234c8acfb8148a9f4c358b70d358dee perf/x86/intel/ds: Don't clear ->pebs_data_cfg for the last PEBS event

Fix a combined PEBS events bug on x86 Intel CPUs.

 Thanks,

	Ingo

------------------>
Kan Liang (1):
      perf/x86/intel/ds: Don't clear ->pebs_data_cfg for the last PEBS event


 arch/x86/events/intel/ds.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/x86/events/intel/ds.c b/arch/x86/events/intel/ds.c
index 2641ba620f12..e010bfed8417 100644
--- a/arch/x86/events/intel/ds.c
+++ b/arch/x86/events/intel/ds.c
@@ -1237,11 +1237,11 @@ pebs_update_state(bool needed_cb, struct cpu_hw_events *cpuc,
 	struct pmu *pmu = event->pmu;
 
 	/*
-	 * Make sure we get updated with the first PEBS
-	 * event. It will trigger also during removal, but
-	 * that does not hurt:
+	 * Make sure we get updated with the first PEBS event.
+	 * During removal, ->pebs_data_cfg is still valid for
+	 * the last PEBS event. Don't clear it.
 	 */
-	if (cpuc->n_pebs == 1)
+	if ((cpuc->n_pebs == 1) && add)
 		cpuc->pebs_data_cfg = PEBS_UPDATE_DS_SW;
 
 	if (needed_cb != pebs_needs_sched_cb(cpuc)) {

Re: [GIT PULL] perf fix

[pr-tracker-bot]

The pull request you sent on Thu, 24 Jun 2021 09:01:40 +0200:

> git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-2021-06-24

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/df5011000496355d8508d56cd3ce45b1196b8b43

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/prtracker.html

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf/urgent git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-2021-06-24

   # HEAD: 7f049fbdd57f6ea71dc741d903c19c73b2f70950 perf/x86/intel/lbr: Zero the xstate buffer on allocation

An LBR buffer fix for code that probably only worked accidentally.

 Thanks,

	Ingo

------------------>
Thomas Gleixner (1):
      perf/x86/intel/lbr: Zero the xstate buffer on allocation


 arch/x86/events/intel/lbr.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/x86/events/intel/lbr.c b/arch/x86/events/intel/lbr.c
index 4409d2cccfda..e8453de7a964 100644
--- a/arch/x86/events/intel/lbr.c
+++ b/arch/x86/events/intel/lbr.c
@@ -731,7 +731,8 @@ void reserve_lbr_buffers(void)
 		if (!kmem_cache || cpuc->lbr_xsave)
 			continue;
 
-		cpuc->lbr_xsave = kmem_cache_alloc_node(kmem_cache, GFP_KERNEL,
+		cpuc->lbr_xsave = kmem_cache_alloc_node(kmem_cache,
+							GFP_KERNEL | __GFP_ZERO,
 							cpu_to_node(cpu));
 	}
 }

Re: [GIT PULL] perf fix

[pr-tracker-bot]

The pull request you sent on Sun, 11 Oct 2020 10:00:15 +0200:

> git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-2020-10-11

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/aa5c3a2911e6d5feaf0c3c7af5bbc652f129d8b2

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/prtracker.html

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf/urgent git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-2020-10-11

   # HEAD: 6d6b8b9f4fceab7266ca03d194f60ec72bd4b654 perf: Fix task_function_call() error handling

Fix an error handling bug that can cause a lockup if a CPU is offline. (doh ...)

 Thanks,

	Ingo

------------------>
Kajol Jain (1):
      perf: Fix task_function_call() error handling


 kernel/events/core.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index 7ed5248f0445..e8bf92202542 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -99,7 +99,7 @@ static void remote_function(void *data)
  * retry due to any failures in smp_call_function_single(), such as if the
  * task_cpu() goes offline concurrently.
  *
- * returns @func return value or -ESRCH when the process isn't running
+ * returns @func return value or -ESRCH or -ENXIO when the process isn't running
  */
 static int
 task_function_call(struct task_struct *p, remote_function_f func, void *info)
@@ -115,7 +115,8 @@ task_function_call(struct task_struct *p, remote_function_f func, void *info)
 	for (;;) {
 		ret = smp_call_function_single(task_cpu(p), remote_function,
 					       &data, 1);
-		ret = !ret ? data.ret : -EAGAIN;
+		if (!ret)
+			ret = data.ret;
 
 		if (ret != -EAGAIN)
 			break;

Re: [GIT PULL] perf fix

[pr-tracker-bot]

The pull request you sent on Sun, 28 Jun 2020 20:33:23 +0200:

> git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-2020-06-28

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/ae71d4bf0074a81cc04255c96e3de0a49b1d95fa

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.wiki.kernel.org/userdoc/prtracker

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf/urgent git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-2020-06-28

   # HEAD: 16accae3d97f97d7f61c4ee5d0002bccdef59088 perf/x86/rapl: Fix RAPL config variable bug

A single Kbuild dependency fix.

Signed-off-by: Ingo Molnar <mingo@kernel.org>
 Thanks,

	Ingo

------------------>
Stephane Eranian (1):
      perf/x86/rapl: Fix RAPL config variable bug


 arch/x86/events/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/events/Makefile b/arch/x86/events/Makefile
index 12c42eba77ec..9933c0e8e97a 100644
--- a/arch/x86/events/Makefile
+++ b/arch/x86/events/Makefile
@@ -1,6 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-only
 obj-y					+= core.o probe.o
-obj-$(PERF_EVENTS_INTEL_RAPL)		+= rapl.o
+obj-$(CONFIG_PERF_EVENTS_INTEL_RAPL)	+= rapl.o
 obj-y					+= amd/
 obj-$(CONFIG_X86_LOCAL_APIC)            += msr.o
 obj-$(CONFIG_CPU_SUP_INTEL)		+= intel/

Re: [GIT PULL] perf fix

[pr-tracker-bot]

The pull request you sent on Thu, 12 Sep 2019 10:39:03 +0200:

> git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/6dcf6a4eb95a8a78c181cd27132f41aca36aeb94

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.wiki.kernel.org/userdoc/prtracker

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: 310aa0a25b338b3100c94880c9a69bec8ce8c3ae perf/hw_breakpoint: Fix arch_hw_breakpoint use-before-initialization

Fix an initialization bug in the hw-breakpoints, which triggered on the 
ARM platform.

 Thanks,

	Ingo

------------------>
Mark-PK Tsai (1):
      perf/hw_breakpoint: Fix arch_hw_breakpoint use-before-initialization


 kernel/events/hw_breakpoint.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c
index c5cd852fe86b..3cc8416ec844 100644
--- a/kernel/events/hw_breakpoint.c
+++ b/kernel/events/hw_breakpoint.c
@@ -413,7 +413,7 @@ static int hw_breakpoint_parse(struct perf_event *bp,
 
 int register_perf_hw_breakpoint(struct perf_event *bp)
 {
-	struct arch_hw_breakpoint hw;
+	struct arch_hw_breakpoint hw = { };
 	int err;
 
 	err = reserve_bp_slot(bp);
@@ -461,7 +461,7 @@ int
 modify_user_hw_breakpoint_check(struct perf_event *bp, struct perf_event_attr *attr,
 			        bool check)
 {
-	struct arch_hw_breakpoint hw;
+	struct arch_hw_breakpoint hw = { };
 	int err;
 
 	err = hw_breakpoint_parse(bp, attr, &hw);

Re: [GIT PULL] perf fix

[pr-tracker-bot]

The pull request you sent on Sat, 27 Apr 2019 16:38:18 +0200:

> git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/5084991663650a238dd9d1981e09eef00a90594c

Thank you!

-- 
Deet-doot-dot, I am a bot.
https://korg.wiki.kernel.org/userdoc/prtracker

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: 82c99f7a81f28f8c1be5f701c8377d14c4075b10 perf/x86/intel: Update KBL Package C-state events to also include PC8/PC9/PC10 counters

A cstate event enumeration fix for Kaby/Coffee Lake CPUs.

 Thanks,

	Ingo

------------------>
Harry Pan (1):
      perf/x86/intel: Update KBL Package C-state events to also include PC8/PC9/PC10 counters


 arch/x86/events/intel/cstate.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c
index 94a4b7fc75d0..d41de9af7a39 100644
--- a/arch/x86/events/intel/cstate.c
+++ b/arch/x86/events/intel/cstate.c
@@ -76,15 +76,15 @@
  *			       Scope: Package (physical package)
  *	MSR_PKG_C8_RESIDENCY:  Package C8 Residency Counter.
  *			       perf code: 0x04
- *			       Available model: HSW ULT,CNL
+ *			       Available model: HSW ULT,KBL,CNL
  *			       Scope: Package (physical package)
  *	MSR_PKG_C9_RESIDENCY:  Package C9 Residency Counter.
  *			       perf code: 0x05
- *			       Available model: HSW ULT,CNL
+ *			       Available model: HSW ULT,KBL,CNL
  *			       Scope: Package (physical package)
  *	MSR_PKG_C10_RESIDENCY: Package C10 Residency Counter.
  *			       perf code: 0x06
- *			       Available model: HSW ULT,GLM,CNL
+ *			       Available model: HSW ULT,KBL,GLM,CNL
  *			       Scope: Package (physical package)
  *
  */
@@ -566,8 +566,8 @@ static const struct x86_cpu_id intel_cstates_match[] __initconst = {
 	X86_CSTATES_MODEL(INTEL_FAM6_SKYLAKE_DESKTOP, snb_cstates),
 	X86_CSTATES_MODEL(INTEL_FAM6_SKYLAKE_X, snb_cstates),
 
-	X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_MOBILE,  snb_cstates),
-	X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_DESKTOP, snb_cstates),
+	X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_MOBILE,  hswult_cstates),
+	X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_DESKTOP, hswult_cstates),
 
 	X86_CSTATES_MODEL(INTEL_FAM6_CANNONLAKE_MOBILE, cnl_cstates),
 

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: 1289e0e29857e606a70a0200bf7849ae38d3493a perf/x86/rapl: Fix Haswell and Broadwell server RAPL event

An Intel RAPL events fix.

 Thanks,

	Ingo

------------------>
Kan Liang (1):
      perf/x86/rapl: Fix Haswell and Broadwell server RAPL event


 arch/x86/events/intel/rapl.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/x86/events/intel/rapl.c b/arch/x86/events/intel/rapl.c
index 005908ee9333..a2efb490f743 100644
--- a/arch/x86/events/intel/rapl.c
+++ b/arch/x86/events/intel/rapl.c
@@ -755,14 +755,14 @@ static const struct x86_cpu_id rapl_cpu_match[] __initconst = {
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_IVYBRIDGE_X, snbep_rapl_init),
 
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_HASWELL_CORE, hsw_rapl_init),
-	X86_RAPL_MODEL_MATCH(INTEL_FAM6_HASWELL_X,    hsw_rapl_init),
+	X86_RAPL_MODEL_MATCH(INTEL_FAM6_HASWELL_X,    hsx_rapl_init),
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_HASWELL_ULT,  hsw_rapl_init),
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_HASWELL_GT3E, hsw_rapl_init),
 
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_BROADWELL_CORE,   hsw_rapl_init),
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_BROADWELL_GT3E,   hsw_rapl_init),
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_BROADWELL_X,	  hsx_rapl_init),
-	X86_RAPL_MODEL_MATCH(INTEL_FAM6_BROADWELL_XEON_D, hsw_rapl_init),
+	X86_RAPL_MODEL_MATCH(INTEL_FAM6_BROADWELL_XEON_D, hsx_rapl_init),
 
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_XEON_PHI_KNL, knl_rapl_init),
 	X86_RAPL_MODEL_MATCH(INTEL_FAM6_XEON_PHI_KNM, knl_rapl_init),

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: 2eece390bf68ec8f733d7e4a3ba8a5ea350082ae perf/x86/intel/bts: Fix exclusive event reference leak

A fix for a misplaced permission check that can leave perf PT or LBR disabled (on 
Intel CPUs) permanently until the next reboot.

 Thanks,

	Ingo

------------------>
Alexander Shishkin (1):
      perf/x86/intel/bts: Fix exclusive event reference leak


 arch/x86/events/intel/bts.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/x86/events/intel/bts.c b/arch/x86/events/intel/bts.c
index 16076eb34699..141e07b06216 100644
--- a/arch/x86/events/intel/bts.c
+++ b/arch/x86/events/intel/bts.c
@@ -546,9 +546,6 @@ static int bts_event_init(struct perf_event *event)
 	if (event->attr.type != bts_pmu.type)
 		return -ENOENT;
 
-	if (x86_add_exclusive(x86_lbr_exclusive_bts))
-		return -EBUSY;
-
 	/*
 	 * BTS leaks kernel addresses even when CPL0 tracing is
 	 * disabled, so disallow intel_bts driver for unprivileged
@@ -562,6 +559,9 @@ static int bts_event_init(struct perf_event *event)
 	    !capable(CAP_SYS_ADMIN))
 		return -EACCES;
 
+	if (x86_add_exclusive(x86_lbr_exclusive_bts))
+		return -EBUSY;
+
 	ret = x86_reserve_hardware();
 	if (ret) {
 		x86_del_exclusive(x86_lbr_exclusive_bts);

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: 64aee2a965cf2954a038b5522f11d2cd2f0f8f3e perf/core: Fix group {cpu,task} validation

A single fix to not allow nonsensical event groups that result in kernel warnings.

 Thanks,

	Ingo

------------------>
Mark Rutland (1):
      perf/core: Fix group {cpu,task} validation


 kernel/events/core.c | 39 +++++++++++++++++++--------------------
 1 file changed, 19 insertions(+), 20 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index ee20d4c546b5..3504125871d2 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -10032,28 +10032,27 @@ SYSCALL_DEFINE5(perf_event_open,
 			goto err_context;
 
 		/*
-		 * Do not allow to attach to a group in a different
-		 * task or CPU context:
+		 * Make sure we're both events for the same CPU;
+		 * grouping events for different CPUs is broken; since
+		 * you can never concurrently schedule them anyhow.
 		 */
-		if (move_group) {
-			/*
-			 * Make sure we're both on the same task, or both
-			 * per-cpu events.
-			 */
-			if (group_leader->ctx->task != ctx->task)
-				goto err_context;
+		if (group_leader->cpu != event->cpu)
+			goto err_context;
 
-			/*
-			 * Make sure we're both events for the same CPU;
-			 * grouping events for different CPUs is broken; since
-			 * you can never concurrently schedule them anyhow.
-			 */
-			if (group_leader->cpu != event->cpu)
-				goto err_context;
-		} else {
-			if (group_leader->ctx != ctx)
-				goto err_context;
-		}
+		/*
+		 * Make sure we're both on the same task, or both
+		 * per-CPU events.
+		 */
+		if (group_leader->ctx->task != ctx->task)
+			goto err_context;
+
+		/*
+		 * Do not allow to attach to a group in a different task
+		 * or CPU context. If we're moving SW events, we'll fix
+		 * this up later, so allow that.
+		 */
+		if (!move_group && group_leader->ctx != ctx)
+			goto err_context;
 
 		/*
 		 * Only a group leader can be exclusive or pinned

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: eb019503569c8c701f1e9c70e848d99c6680839b perf/x86: Fix bogus kernel printk, again

A single printk message fix.

 Thanks,

	Ingo

------------------>
Vegard Nossum (1):
      perf/x86: Fix bogus kernel printk, again


 arch/x86/events/core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c
index 26ced536005a..91eac39625be 100644
--- a/arch/x86/events/core.c
+++ b/arch/x86/events/core.c
@@ -263,7 +263,7 @@ static bool check_hw_exists(void)
 
 msr_fail:
 	pr_cont("Broken PMU hardware detected, using software events only.\n");
-	pr_info("%sFailed to access perfctr msr (MSR %x is %Lx)\n",
+	printk("%sFailed to access perfctr msr (MSR %x is %Lx)\n",
 		boot_cpu_has(X86_FEATURE_HYPERVISOR) ? KERN_INFO : KERN_ERR,
 		reg, val_new);
 

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: 93193230f7b8bd484d1749f4044d1214d521c804 perf/core: Change the default paranoia level to 2

Bump up the default perf paranoia level to 2 because otherwise perf can be used by 
local users to defeat KASLR, amongst other things.

 Thanks,

	Ingo

------------------>
Andy Lutomirski (1):
      perf/core: Change the default paranoia level to 2


 Documentation/sysctl/kernel.txt | 2 +-
 kernel/events/core.c            | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/Documentation/sysctl/kernel.txt b/Documentation/sysctl/kernel.txt
index 57653a44b128..fcddfd5ded99 100644
--- a/Documentation/sysctl/kernel.txt
+++ b/Documentation/sysctl/kernel.txt
@@ -645,7 +645,7 @@ allowed to execute.
 perf_event_paranoid:
 
 Controls use of the performance events system by unprivileged
-users (without CAP_SYS_ADMIN).  The default value is 1.
+users (without CAP_SYS_ADMIN).  The default value is 2.
 
  -1: Allow use of (almost) all events by all users
 >=0: Disallow raw tracepoint access by users without CAP_IOC_LOCK
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 4e2ebf6f2f1f..c0ded2416615 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -351,7 +351,7 @@ static struct srcu_struct pmus_srcu;
  *   1 - disallow cpu events for unpriv
  *   2 - disallow kernel profiling for unpriv
  */
-int sysctl_perf_event_paranoid __read_mostly = 1;
+int sysctl_perf_event_paranoid __read_mostly = 2;
 
 /* Minimum for 512 kiB + 1 user control page */
 int sysctl_perf_event_mlock __read_mostly = 512 + (PAGE_SIZE / 1024); /* 'free' kiB per user */

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: a19cad6d66823ddd54b0e7c88d7bddd307cb1161 Merge tag 'perf-urgent-for-mingo-20160418' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent

A single tooling fix for a user-triggerable segfault.

 Thanks,

	Ingo

------------------>
Adrian Hunter (1):
      perf intel-pt: Fix segfault tracing transactions


 tools/perf/util/intel-pt.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c
index 407f11b97c8d..617578440989 100644
--- a/tools/perf/util/intel-pt.c
+++ b/tools/perf/util/intel-pt.c
@@ -1130,7 +1130,7 @@ static int intel_pt_synth_transaction_sample(struct intel_pt_queue *ptq)
 		pr_err("Intel Processor Trace: failed to deliver transaction event, error %d\n",
 		       ret);
 
-	if (pt->synth_opts.callchain)
+	if (pt->synth_opts.last_branch)
 		intel_pt_reset_last_branch_rb(ptq);
 
 	return ret;

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: d525211f9d1be8b523ec7633f080f2116f5ea536 perf: Fix irq_work 'tail' recursion

A perf kernel side fix for a fuzzer triggered lockup.

 Thanks,

	Ingo

------------------>
Peter Zijlstra (1):
      perf: Fix irq_work 'tail' recursion


 kernel/events/core.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index 453ef61311d4..2fabc0627165 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -4574,6 +4574,13 @@ static void perf_pending_event(struct irq_work *entry)
 {
 	struct perf_event *event = container_of(entry,
 			struct perf_event, pending);
+	int rctx;
+
+	rctx = perf_swevent_get_recursion_context();
+	/*
+	 * If we 'fail' here, that's OK, it means recursion is already disabled
+	 * and we won't recurse 'further'.
+	 */
 
 	if (event->pending_disable) {
 		event->pending_disable = 0;
@@ -4584,6 +4591,9 @@ static void perf_pending_event(struct irq_work *entry)
 		event->pending_wakeup = 0;
 		perf_event_wakeup(event);
 	}
+
+	if (rctx >= 0)
+		perf_swevent_put_recursion_context(rctx);
 }
 
 /*

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   # HEAD: cf30d52e2d11523c42048ab89ed4215b5021526a perf/x86: Fix constraint table end marker bug

An x86/intel event constraint fix.

 Thanks,

	Ingo

------------------>
Maria Dimakopoulou (1):
      perf/x86: Fix constraint table end marker bug


 arch/x86/kernel/cpu/perf_event.h | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index fd00bb2..c1a8618 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -262,11 +262,20 @@ struct cpu_hw_events {
 	__EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVENT_MASK, \
 			  HWEIGHT(n), 0, PERF_X86_EVENT_PEBS_ST_HSW)
 
-#define EVENT_CONSTRAINT_END		\
-	EVENT_CONSTRAINT(0, 0, 0)
+/*
+ * We define the end marker as having a weight of -1
+ * to enable blacklisting of events using a counter bitmask
+ * of zero and thus a weight of zero.
+ * The end marker has a weight that cannot possibly be
+ * obtained from counting the bits in the bitmask.
+ */
+#define EVENT_CONSTRAINT_END { .weight = -1 }
 
+/*
+ * Check for end marker with weight == -1
+ */
 #define for_each_event_constraint(e, c)	\
-	for ((e) = (c); (e)->weight; (e)++)
+	for ((e) = (c); (e)->weight != -1; (e)++)
 
 /*
  * Extra registers for specific events.

[GIT PULL] perf fix

[Ingo Molnar]

* Markus Trippelsdorf <markus@trippelsdorf.de> wrote:

> On 2013.09.28 at 20:03 +0200, Ingo Molnar wrote:
> > Andi Kleen (1):
> >       perf symbols: Demangle cloned functions
> 
> The commit above e95ab53645:
> 
>  commit de95ab53645a2f0015e0f68ee723f18dce2b8b51
>  Author: Andi Kleen <ak@linux.intel.com>
>  Date:   Thu Sep 12 08:16:38 2013 -0700
> 
>      perf symbols: Demangle cloned functions
> 
> breaks "perf top" on my machine. I just see a gray screen with no text
> at all. Sometimes the following error messages are printed:
>  *** Error in `perf': invalid fastbin entry (free): 0x00000000029b18c0 ***
>  *** Error in `perf': malloc(): memory corruption (fast): 0x0000000000ee0b10 ***
> 
> Reverting the commit "fixes" the issue.

Sorry about that.

Looking at the commit it fails on several quality levels:

 - sloppy structure: the code is unreadable, it should have been a 
                     separate helper function

 - sloppy flow:      why the heck does it do strchr() twice? Not to 
                     mention the fragile way of how the new flow tries 
                     to merge with the old instead of cleanly separating.

 - sloppy types:     it casts a const char * over into char *

 - sloppy style:     variable block not separated from statements, 
                     meaningless temporary variable names, etc. etc.

 - and it's not even a regression fix to begin with!

So the right resolution is a revert. Andi, please send a fixed patch with 
all of these issues and the crash fixed.

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   HEAD: 14951f22f1cc8375ead345d2ca08455e91f2152b Revert "perf symbols: Demangle cloned functions"

It fixes the 'perf top' regression Markus Trippelsdorf reported.

 Thanks,

	Ingo

------------------>
Ingo Molnar (1):
      Revert "perf symbols: Demangle cloned functions"


 tools/perf/util/symbol-elf.c | 27 +--------------------------
 1 file changed, 1 insertion(+), 26 deletions(-)

diff --git a/tools/perf/util/symbol-elf.c b/tools/perf/util/symbol-elf.c
index d2a888e..a9c829b 100644
--- a/tools/perf/util/symbol-elf.c
+++ b/tools/perf/util/symbol-elf.c
@@ -928,33 +928,8 @@ int dso__load_sym(struct dso *dso, struct map *map,
 		 * to it...
 		 */
 		if (symbol_conf.demangle) {
-			/*
-			 * The demangler doesn't deal with cloned functions.
-			 * XXXX.clone.NUM or similar
-			 * Strip the dot part and readd it later.
-			 */
-			char *p = (char *)elf_name, *dot;
-			dot = strchr(elf_name, '.');
-			if (dot) {
-				p = strdup(elf_name);
-				if (!p)
-					goto new_symbol;
-				dot = strchr(p, '.');
-				*dot = 0;
-			}
-
-			demangled = bfd_demangle(NULL, p,
+			demangled = bfd_demangle(NULL, elf_name,
 						 DMGL_PARAMS | DMGL_ANSI);
-			if (dot)
-				*dot = '.';
-			if (demangled && dot) {
-				demangled = realloc(demangled, strlen(demangled) + strlen(dot) + 1);
-				if (!demangled)
-					goto new_symbol;
-				strcpy(demangled + (dot - p), dot);
-			}
-			if (p != elf_name)
-				free(p);
 			if (demangled != NULL)
 				elf_name = demangled;
 		}

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   HEAD: ae23bff1d71f8b416ed740bc458df67355c77c92 perf: Prevent race in unthrottling code

This is a leftover fix from the v3.11 cycle.

 Thanks,

	Ingo

------------------>
Jiri Olsa (1):
      perf: Prevent race in unthrottling code


 kernel/events/core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index f86599e..258eaaf 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2712,7 +2712,7 @@ static void perf_adjust_freq_unthr_context(struct perf_event_context *ctx,
 
 		hwc = &event->hw;
 
-		if (needs_unthr && hwc->interrupts == MAX_INTERRUPTS) {
+		if (hwc->interrupts == MAX_INTERRUPTS) {
 			hwc->interrupts = 0;
 			perf_log_throttle(event, 1);
 			event->pmu->start(event, 0);

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   HEAD: 2976b10f05bd7f6dab9f9e7524451ddfed656a89 perf: Disable monitoring on setuid processes for regular users

One more fix for a recently discovered bug.

 Thanks,

	Ingo

------------------>
Stephane Eranian (1):
      perf: Disable monitoring on setuid processes for regular users


 fs/exec.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/fs/exec.c b/fs/exec.c
index 6430195..ffd7a81 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1135,13 +1135,6 @@ void setup_new_exec(struct linux_binprm * bprm)
 			set_dumpable(current->mm, suid_dumpable);
 	}
 
-	/*
-	 * Flush performance counters when crossing a
-	 * security domain:
-	 */
-	if (!get_dumpable(current->mm))
-		perf_event_exit_task(current);
-
 	/* An exec changes our domain. We are no longer part of the thread
 	   group */
 
@@ -1205,6 +1198,15 @@ void install_exec_creds(struct linux_binprm *bprm)
 
 	commit_creds(bprm->cred);
 	bprm->cred = NULL;
+
+	/*
+	 * Disable monitoring for regular users
+	 * when executing setuid binaries. Must
+	 * wait until new credentials are committed
+	 * by commit_creds() above
+	 */
+	if (get_dumpable(current->mm) != SUID_DUMP_USER)
+		perf_event_exit_task(current);
 	/*
 	 * cred_guard_mutex must be held at least to this point to prevent
 	 * ptrace_attach() from altering our determination of the task's

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   HEAD: c79aa0d96548aee50570209eb2d45c8f4ac49230 events: Protect access via task_subsys_state_check()

This fix adds missing RCU read protection.

 Thanks,

	Ingo

------------------>
Paul E. McKenney (1):
      events: Protect access via task_subsys_state_check()


 kernel/events/core.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index 4d3124b..9fcb094 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -4596,6 +4596,7 @@ void perf_event_comm(struct task_struct *task)
 	struct perf_event_context *ctx;
 	int ctxn;
 
+	rcu_read_lock();
 	for_each_task_context_nr(ctxn) {
 		ctx = task->perf_event_ctxp[ctxn];
 		if (!ctx)
@@ -4603,6 +4604,7 @@ void perf_event_comm(struct task_struct *task)
 
 		perf_event_enable_on_exec(ctx);
 	}
+	rcu_read_unlock();
 
 	if (!atomic_read(&nr_comm_events))
 		return;

Re: [GIT PULL] perf fix

[Ingo Molnar]

* Josh Boyer <jwboyer@gmail.com> wrote:

> > This pull request you replied to is the v3.4 era fixes tree, 
> > with one remaining fix in it.
> 
> I see.  The forest of tip trees apparently confuses me still.  
> I'll figure it out eventually.

The topic tree layout for single-topic trees is pretty simple 
and straightforward - but the situation you met here was 
arguably a weird corner case:

 X/urgent are the fixes that go to Linus
 X/core   are the development patches for the next merge window

Where 'X' can be one of: perf, sched, timer, irq - the main 
subsystem trees we maintain. (x86 is a multi-topic tree, with 
intuitively named topic trees, such as x86/reboot, x86/asm or 
x86/mm.)

All of them are test-merged into tip:master - this is the one 
that you will typically use, the topic layout is for maintainers 
and for power-contributors/submaintaners who are sending Git 
pull requests to us.

at the beginning of a merge window (i.e. right now) there might 
be fixes pending in perf/urgent that did not make it to v3.4. 
Instead of merging them into perf/core I tend to send them to 
Linus as a standalone tree.

The rest of perf/core, once the initial one or two sets of 
commits get pulled by Linus, morphs into perf/urgent, fairly 
early in the merge window.

Thus there's a new perf/urgent and an empty perf/core, and the 
cycle starts again.

You met this cycle switch period to the day (the chance is only 
1:90 for that, consider yourself lucky ;-), which created the 
impression of a confusing fixes workflow.

Thanks,

	Ingo

Re: [GIT PULL] perf fix

[Josh Boyer]

On Wed, May 23, 2012 at 4:04 PM, Ingo Molnar <mingo@kernel.org> wrote:
>
> * Josh Boyer <jwboyer@gmail.com> wrote:
>> Perf doesn't even build in Linus' current tree.
>
> Yes, that's a regression in some build environments. (it builds
> fine in our environments, that's why it wasn't noticed.)
>
>> [...]  You might want to include
>>
>> http://git.kernel.org/?p=linux/kernel/git/tip/tip.git;a=commitdiff;h=8784eb7497b928de33a715e26da2bd51d424bec1
>>
>> with your pull request.
>
> I sent that (and other fixes) to Linus too, earlier today.

I saw that.  Thank you.

> This pull request you replied to is the v3.4 era fixes tree,
> with one remaining fix in it.

I see.  The forest of tip trees apparently confuses me still.  I'll
figure it out eventually.

josh

Re: [GIT PULL] perf fix

[Ingo Molnar]

* Josh Boyer <jwboyer@gmail.com> wrote:

> On Wed, May 23, 2012 at 2:50 PM, Ingo Molnar <mingo@kernel.org> wrote:
> > Linus,
> >
> > Please pull the latest perf-urgent-for-linus git tree from:
> >
> >   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus
> >
> >   HEAD: 5bcdf5e4fee3c45e1281c25e4941f2163cb28c65 perf/x86: Update event scheduling constraints for AMD family 15h models
> >
> > Leftover AMD PMU driver fix fix from the end of the v3.4
> > stabilization cycle.
> 
> Perf doesn't even build in Linus' current tree.

Yes, that's a regression in some build environments. (it builds 
fine in our environments, that's why it wasn't noticed.)

> [...]  You might want to include
> 
> http://git.kernel.org/?p=linux/kernel/git/tip/tip.git;a=commitdiff;h=8784eb7497b928de33a715e26da2bd51d424bec1
> 
> with your pull request.

I sent that (and other fixes) to Linus too, earlier today.

This pull request you replied to is the v3.4 era fixes tree, 
with one remaining fix in it.

Thanks,

	Ingo

Re: [GIT PULL] perf fix

[Josh Boyer]

On Wed, May 23, 2012 at 2:56 PM, Josh Boyer <jwboyer@gmail.com> wrote:
> On Wed, May 23, 2012 at 2:50 PM, Ingo Molnar <mingo@kernel.org> wrote:
>> Linus,
>>
>> Please pull the latest perf-urgent-for-linus git tree from:
>>
>>   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus
>>
>>   HEAD: 5bcdf5e4fee3c45e1281c25e4941f2163cb28c65 perf/x86: Update event scheduling constraints for AMD family 15h models
>>
>> Leftover AMD PMU driver fix fix from the end of the v3.4
>> stabilization cycle.
>
> Perf doesn't even build in Linus' current tree.  You might want to include
>
> http://git.kernel.org/?p=linux/kernel/git/tip/tip.git;a=commitdiff;h=8784eb7497b928de33a715e26da2bd51d424bec1

Possibly this one too:

http://git.kernel.org/?p=linux/kernel/git/tip/tip.git;a=commitdiff;h=e326e7524531c5c6de4a8cf8eeed60b4a9f47637

josh

Re: [GIT PULL] perf fix

[Josh Boyer]

On Wed, May 23, 2012 at 2:50 PM, Ingo Molnar <mingo@kernel.org> wrote:
> Linus,
>
> Please pull the latest perf-urgent-for-linus git tree from:
>
>   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus
>
>   HEAD: 5bcdf5e4fee3c45e1281c25e4941f2163cb28c65 perf/x86: Update event scheduling constraints for AMD family 15h models
>
> Leftover AMD PMU driver fix fix from the end of the v3.4
> stabilization cycle.

Perf doesn't even build in Linus' current tree.  You might want to include

http://git.kernel.org/?p=linux/kernel/git/tip/tip.git;a=commitdiff;h=8784eb7497b928de33a715e26da2bd51d424bec1

with your pull request.

josh

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   HEAD: 5bcdf5e4fee3c45e1281c25e4941f2163cb28c65 perf/x86: Update event scheduling constraints for AMD family 15h models

Leftover AMD PMU driver fix fix from the end of the v3.4 
stabilization cycle.

 Thanks,

	Ingo

------------------>
Robert Richter (1):
      perf/x86: Update event scheduling constraints for AMD family 15h models


 arch/x86/kernel/cpu/perf_event_amd.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c
index 95e7fe1..9edc786 100644
--- a/arch/x86/kernel/cpu/perf_event_amd.c
+++ b/arch/x86/kernel/cpu/perf_event_amd.c
@@ -493,6 +493,7 @@ static __initconst const struct x86_pmu amd_pmu = {
  * 0x023	DE	PERF_CTL[2:0]
  * 0x02D	LS	PERF_CTL[3]
  * 0x02E	LS	PERF_CTL[3,0]
+ * 0x031	LS	PERF_CTL[2:0] (**)
  * 0x043	CU	PERF_CTL[2:0]
  * 0x045	CU	PERF_CTL[2:0]
  * 0x046	CU	PERF_CTL[2:0]
@@ -506,10 +507,12 @@ static __initconst const struct x86_pmu amd_pmu = {
  * 0x0DD	LS	PERF_CTL[5:0]
  * 0x0DE	LS	PERF_CTL[5:0]
  * 0x0DF	LS	PERF_CTL[5:0]
+ * 0x1C0	EX	PERF_CTL[5:3]
  * 0x1D6	EX	PERF_CTL[5:0]
  * 0x1D8	EX	PERF_CTL[5:0]
  *
- * (*) depending on the umask all FPU counters may be used
+ * (*)  depending on the umask all FPU counters may be used
+ * (**) only one unitmask enabled at a time
  */
 
 static struct event_constraint amd_f15_PMC0  = EVENT_CONSTRAINT(0, 0x01, 0);
@@ -559,6 +562,12 @@ amd_get_event_constraints_f15h(struct cpu_hw_events *cpuc, struct perf_event *ev
 			return &amd_f15_PMC3;
 		case 0x02E:
 			return &amd_f15_PMC30;
+		case 0x031:
+			if (hweight_long(hwc->config & ARCH_PERFMON_EVENTSEL_UMASK) <= 1)
+				return &amd_f15_PMC20;
+			return &emptyconstraint;
+		case 0x1C0:
+			return &amd_f15_PMC53;
 		default:
 			return &amd_f15_PMC50;
 		}

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

   HEAD: 1018faa6cf23b256bf25919ef203cd7c129f06f2 perf/x86/kvm: Fix Host-Only/Guest-Only counting with SVM disabled

 Thanks,

	Ingo

------------------>
Joerg Roedel (1):
      perf/x86/kvm: Fix Host-Only/Guest-Only counting with SVM disabled


 arch/x86/include/asm/perf_event.h    |    8 +++++++
 arch/x86/kernel/cpu/perf_event.h     |    8 +++++-
 arch/x86/kernel/cpu/perf_event_amd.c |   37 ++++++++++++++++++++++++++++++++-
 arch/x86/kvm/svm.c                   |    5 ++++
 4 files changed, 54 insertions(+), 4 deletions(-)

diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 096c975..461ce43 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -242,4 +242,12 @@ static inline void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap)
 static inline void perf_events_lapic_init(void)	{ }
 #endif
 
+#if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD)
+ extern void amd_pmu_enable_virt(void);
+ extern void amd_pmu_disable_virt(void);
+#else
+ static inline void amd_pmu_enable_virt(void) { }
+ static inline void amd_pmu_disable_virt(void) { }
+#endif
+
 #endif /* _ASM_X86_PERF_EVENT_H */
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index 8944062..c30c807 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -147,7 +147,9 @@ struct cpu_hw_events {
 	/*
 	 * AMD specific bits
 	 */
-	struct amd_nb		*amd_nb;
+	struct amd_nb			*amd_nb;
+	/* Inverted mask of bits to clear in the perf_ctr ctrl registers */
+	u64				perf_ctr_virt_mask;
 
 	void				*kfree_on_online;
 };
@@ -417,9 +419,11 @@ void x86_pmu_disable_all(void);
 static inline void __x86_pmu_enable_event(struct hw_perf_event *hwc,
 					  u64 enable_mask)
 {
+	u64 disable_mask = __this_cpu_read(cpu_hw_events.perf_ctr_virt_mask);
+
 	if (hwc->extra_reg.reg)
 		wrmsrl(hwc->extra_reg.reg, hwc->extra_reg.config);
-	wrmsrl(hwc->config_base, hwc->config | enable_mask);
+	wrmsrl(hwc->config_base, (hwc->config | enable_mask) & ~disable_mask);
 }
 
 void x86_pmu_enable_all(int added);
diff --git a/arch/x86/kernel/cpu/perf_event_amd.c b/arch/x86/kernel/cpu/perf_event_amd.c
index 0397b23..67250a5 100644
--- a/arch/x86/kernel/cpu/perf_event_amd.c
+++ b/arch/x86/kernel/cpu/perf_event_amd.c
@@ -1,4 +1,5 @@
 #include <linux/perf_event.h>
+#include <linux/export.h>
 #include <linux/types.h>
 #include <linux/init.h>
 #include <linux/slab.h>
@@ -357,7 +358,9 @@ static void amd_pmu_cpu_starting(int cpu)
 	struct amd_nb *nb;
 	int i, nb_id;
 
-	if (boot_cpu_data.x86_max_cores < 2)
+	cpuc->perf_ctr_virt_mask = AMD_PERFMON_EVENTSEL_HOSTONLY;
+
+	if (boot_cpu_data.x86_max_cores < 2 || boot_cpu_data.x86 == 0x15)
 		return;
 
 	nb_id = amd_get_nb_id(cpu);
@@ -587,9 +590,9 @@ static __initconst const struct x86_pmu amd_pmu_f15h = {
 	.put_event_constraints	= amd_put_event_constraints,
 
 	.cpu_prepare		= amd_pmu_cpu_prepare,
-	.cpu_starting		= amd_pmu_cpu_starting,
 	.cpu_dead		= amd_pmu_cpu_dead,
 #endif
+	.cpu_starting		= amd_pmu_cpu_starting,
 };
 
 __init int amd_pmu_init(void)
@@ -621,3 +624,33 @@ __init int amd_pmu_init(void)
 
 	return 0;
 }
+
+void amd_pmu_enable_virt(void)
+{
+	struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
+
+	cpuc->perf_ctr_virt_mask = 0;
+
+	/* Reload all events */
+	x86_pmu_disable_all();
+	x86_pmu_enable_all(0);
+}
+EXPORT_SYMBOL_GPL(amd_pmu_enable_virt);
+
+void amd_pmu_disable_virt(void)
+{
+	struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
+
+	/*
+	 * We only mask out the Host-only bit so that host-only counting works
+	 * when SVM is disabled. If someone sets up a guest-only counter when
+	 * SVM is disabled the Guest-only bits still gets set and the counter
+	 * will not count anything.
+	 */
+	cpuc->perf_ctr_virt_mask = AMD_PERFMON_EVENTSEL_HOSTONLY;
+
+	/* Reload all events */
+	x86_pmu_disable_all();
+	x86_pmu_enable_all(0);
+}
+EXPORT_SYMBOL_GPL(amd_pmu_disable_virt);
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index 5fa553b..e385214 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -29,6 +29,7 @@
 #include <linux/ftrace_event.h>
 #include <linux/slab.h>
 
+#include <asm/perf_event.h>
 #include <asm/tlbflush.h>
 #include <asm/desc.h>
 #include <asm/kvm_para.h>
@@ -575,6 +576,8 @@ static void svm_hardware_disable(void *garbage)
 		wrmsrl(MSR_AMD64_TSC_RATIO, TSC_RATIO_DEFAULT);
 
 	cpu_svm_disable();
+
+	amd_pmu_disable_virt();
 }
 
 static int svm_hardware_enable(void *garbage)
@@ -622,6 +625,8 @@ static int svm_hardware_enable(void *garbage)
 
 	svm_init_erratum_383();
 
+	amd_pmu_enable_virt();
+
 	return 0;
 }
 

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

 Thanks,

	Ingo

------------------>
Will Deacon (1):
      perf events: Fix ring_buffer_wakeup() brown paperbag bug


 kernel/events/core.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index d3b9df5..58690af 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -3558,9 +3558,13 @@ static void ring_buffer_wakeup(struct perf_event *event)
 
 	rcu_read_lock();
 	rb = rcu_dereference(event->rb);
-	list_for_each_entry_rcu(event, &rb->event_list, rb_entry) {
+	if (!rb)
+		goto unlock;
+
+	list_for_each_entry_rcu(event, &rb->event_list, rb_entry)
 		wake_up_all(&event->waitq);
-	}
+
+unlock:
 	rcu_read_unlock();
 }
 

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://tesla.tglx.de/git/linux-2.6-tip.git perf-urgent-for-linus

 Thanks,

	Ingo

------------------>
Jiri Olsa (1):
      perf tools: Fix raw sample reading


 tools/perf/util/evsel.c |    7 +++++--
 1 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index c5748c5..e389815 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -449,6 +449,8 @@ int perf_event__parse_sample(const union perf_event *event, u64 type,
 	}
 
 	if (type & PERF_SAMPLE_RAW) {
+		const u64 *pdata;
+
 		u.val64 = *array;
 		if (WARN_ONCE(swapped,
 			      "Endianness of raw data not corrected!\n")) {
@@ -462,11 +464,12 @@ int perf_event__parse_sample(const union perf_event *event, u64 type,
 			return -EFAULT;
 
 		data->raw_size = u.val32[0];
+		pdata = (void *) array + sizeof(u32);
 
-		if (sample_overlap(event, &u.val32[1], data->raw_size))
+		if (sample_overlap(event, pdata, data->raw_size))
 			return -EFAULT;
 
-		data->raw_data = &u.val32[1];
+		data->raw_data = (void *) pdata;
 	}
 
 	return 0;

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git perf-urgent-for-linus

 Thanks,

	Ingo

------------------>
David Ahern (1):
      perf tools, x86: Fix 32-bit compile on 64-bit system


 tools/perf/Makefile |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index 940257b..c168366 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -52,7 +52,10 @@ ifeq ($(ARCH),i386)
 endif
 ifeq ($(ARCH),x86_64)
 	ARCH := x86
-	IS_X86_64 := $(shell echo __x86_64__ | ${CC} -E -xc - | tail -n 1)
+	IS_X86_64 := 0
+	ifeq (, $(findstring m32,$(EXTRA_CFLAGS)))
+		IS_X86_64 := $(shell echo __x86_64__ | ${CC} -E -xc - | tail -n 1)
+	endif
 	ifeq (${IS_X86_64}, 1)
 		RAW_ARCH := x86_64
 		ARCH_CFLAGS := -DARCH_X86_64

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-fixes-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git perf-fixes-for-linus

 Thanks,

	Ingo

------------------>
Arnaldo Carvalho de Melo (1):
      perf record: Add a fallback to the reference relocation symbol


 tools/perf/builtin-record.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 3b8b638..f1411e9 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -563,6 +563,9 @@ static int __cmd_record(int argc, const char **argv)
 
 	err = event__synthesize_kernel_mmap(process_synthesized_event,
 					    session, "_text");
+	if (err < 0)
+		err = event__synthesize_kernel_mmap(process_synthesized_event,
+						    session, "_stext");
 	if (err < 0) {
 		pr_err("Couldn't record kernel reference relocation symbol.\n");
 		return err;

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-fixes-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git perf-fixes-for-linus

 Thanks,

	Ingo

------------------>
Tejun Heo (1):
      perf: Fix resource leak in failure path of perf_event_open()


 kernel/perf_event.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 2f3fbf8..3d1552d 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -4897,7 +4897,7 @@ err_fput_free_put_context:
 
 err_free_put_context:
 	if (err < 0)
-		kfree(event);
+		free_event(event);
 
 err_put_context:
 	if (err < 0)

[GIT PULL] perf fix

[Ingo Molnar]

Linus,

Please pull the latest perf-fixes-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip.git perf-fixes-for-linus

 Thanks,

	Ingo

------------------>
Frederic Weisbecker (1):
      perf: Fix unsafe frame rewinding with hot regs fetching


 arch/x86/kernel/dumpstack.h |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/dumpstack.h b/arch/x86/kernel/dumpstack.h
index e39e771..e1a93be 100644
--- a/arch/x86/kernel/dumpstack.h
+++ b/arch/x86/kernel/dumpstack.h
@@ -14,6 +14,8 @@
 #define get_bp(bp) asm("movq %%rbp, %0" : "=r" (bp) :)
 #endif
 
+#include <linux/uaccess.h>
+
 extern void
 show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
 		unsigned long *stack, unsigned long bp, char *log_lvl);
@@ -42,8 +44,10 @@ static inline unsigned long rewind_frame_pointer(int n)
 	get_bp(frame);
 
 #ifdef CONFIG_FRAME_POINTER
-	while (n--)
-		frame = frame->next_frame;
+	while (n--) {
+		if (probe_kernel_address(&frame->next_frame, frame))
+			break;
+	}
 #endif
 
 	return (unsigned long)frame;