3 files changed, 102 insertions, 91 deletions

diff --git a/drivers/oprofile/buffer_sync.c b/drivers/oprofile/buffer_sync.c
index 908202afbae9..d969bb13a252 100644
--- a/drivers/oprofile/buffer_sync.c
+++ b/drivers/oprofile/buffer_sync.c
@@ -1,11 +1,12 @@
 /**
  * @file buffer_sync.c
  *
- * @remark Copyright 2002 OProfile authors
+ * @remark Copyright 2002-2009 OProfile authors
  * @remark Read the file COPYING
  *
  * @author John Levon <levon@movementarian.org>
  * @author Barry Kasindorf
+ * @author Robert Richter <robert.richter@amd.com>
  *
  * This is the core of the buffer management. Each
  * CPU buffer is processed and entered into the
@@ -529,6 +530,7 @@ void sync_buffer(int cpu)
         sync_buffer_state state = sb_buffer_start;
         unsigned int i;
         unsigned long available;
+        unsigned long flags;
         struct op_entry entry;
         struct op_sample *sample;
 
@@ -545,38 +547,34 @@ void sync_buffer(int cpu)
                         break;
 
                 if (is_code(sample->eip)) {
-                        switch (sample->event) {
-                        case 0:
-                        case CPU_IS_KERNEL:
+                        flags = sample->event;
+                        if (flags & TRACE_BEGIN) {
+                                state = sb_bt_start;
+                                add_trace_begin();
+                        }
+                        if (flags & KERNEL_CTX_SWITCH) {
                                 /* kernel/userspace switch */
-                                in_kernel = sample->event;
+                                in_kernel = flags & IS_KERNEL;
                                 if (state == sb_buffer_start)
                                         state = sb_sample_start;
-                                add_kernel_ctx_switch(sample->event);
-                                break;
-                        case CPU_TRACE_BEGIN:
-                                state = sb_bt_start;
-                                add_trace_begin();
-                                break;
-#ifdef CONFIG_OPROFILE_IBS
-                        case IBS_FETCH_BEGIN:
-                                add_ibs_begin(cpu, IBS_FETCH_CODE, mm);
-                                break;
-                        case IBS_OP_BEGIN:
-                                add_ibs_begin(cpu, IBS_OP_CODE, mm);
-                                break;
-#endif
-                        default:
+                                add_kernel_ctx_switch(flags & IS_KERNEL);
+                        }
+                        if (flags & USER_CTX_SWITCH) {
                                 /* userspace context switch */
                                 oldmm = mm;
-                                new = (struct task_struct *)sample->event;
+                                new = (struct task_struct *)sample->data[0];
                                 release_mm(oldmm);
                                 mm = take_tasks_mm(new);
                                 if (mm != oldmm)
                                         cookie = get_exec_dcookie(mm);
                                 add_user_ctx_switch(new, cookie);
-                                break;
                         }
+#ifdef CONFIG_OPROFILE_IBS
+                        if (flags & IBS_FETCH_BEGIN)
+                                add_ibs_begin(cpu, IBS_FETCH_CODE, mm);
+                        if (flags & IBS_OP_BEGIN)
+                                add_ibs_begin(cpu, IBS_OP_CODE, mm);
+#endif
                         continue;
                 }
 
diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 400f7fcffdbe..e859d23cfc57 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -212,6 +212,59 @@ unsigned long op_cpu_buffer_entries(int cpu)
                 + ring_buffer_entries_cpu(op_ring_buffer_write, cpu);
 }
 
+static int
+op_add_code(struct oprofile_cpu_buffer *cpu_buf, unsigned long backtrace,
+            int is_kernel, struct task_struct *task)
+{
+        struct op_entry entry;
+        struct op_sample *sample;
+        unsigned long flags;
+        int size;
+
+        flags = 0;
+
+        if (backtrace)
+                flags |= TRACE_BEGIN;
+
+        /* notice a switch from user->kernel or vice versa */
+        is_kernel = !!is_kernel;
+        if (cpu_buf->last_is_kernel != is_kernel) {
+                cpu_buf->last_is_kernel = is_kernel;
+                flags |= KERNEL_CTX_SWITCH;
+                if (is_kernel)
+                        flags |= IS_KERNEL;
+        }
+
+        /* notice a task switch */
+        if (cpu_buf->last_task != task) {
+                cpu_buf->last_task = task;
+                flags |= USER_CTX_SWITCH;
+        }
+
+        if (!flags)
+                /* nothing to do */
+                return 0;
+
+        if (flags & USER_CTX_SWITCH)
+                size = 1;
+        else
+                size = 0;
+
+        sample = op_cpu_buffer_write_reserve(&entry, size);
+        if (!sample)
+                return -ENOMEM;
+
+        sample->eip = ESCAPE_CODE;
+        sample->event = flags;
+
+        if (size)
+                sample->data[0] = (unsigned long)task;
+
+        op_cpu_buffer_write_commit(&entry);
+
+        return 0;
+}
+
 static inline int
 op_add_sample(struct oprofile_cpu_buffer *cpu_buf,
               unsigned long pc, unsigned long event)
@@ -229,26 +282,18 @@ op_add_sample(struct oprofile_cpu_buffer *cpu_buf,
         return op_cpu_buffer_write_commit(&entry);
 }
 
-static inline int
-add_code(struct oprofile_cpu_buffer *buffer, unsigned long value)
-{
-        return op_add_sample(buffer, ESCAPE_CODE, value);
-}
-
-/* This must be safe from any context. It's safe writing here
- * because of the head/tail separation of the writer and reader
- * of the CPU buffer.
+/*
+ * This must be safe from any context.
  *
  * is_kernel is needed because on some architectures you cannot
  * tell if you are in kernel or user space simply by looking at
  * pc. We tag this in the buffer by generating kernel enter/exit
  * events whenever is_kernel changes
  */
-static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
-                      int is_kernel, unsigned long event)
+static int
+log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
+           unsigned long backtrace, int is_kernel, unsigned long event)
 {
-        struct task_struct *task;
-
         cpu_buf->sample_received++;
 
         if (pc == ESCAPE_CODE) {
@@ -256,23 +301,8 @@ static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
                 return 0;
         }
 
-        is_kernel = !!is_kernel;
-
-        task = current;
-
-        /* notice a switch from user->kernel or vice versa */
-        if (cpu_buf->last_is_kernel != is_kernel) {
-                cpu_buf->last_is_kernel = is_kernel;
-                if (add_code(cpu_buf, is_kernel))
-                        goto fail;
-        }
-
-        /* notice a task switch */
-        if (cpu_buf->last_task != task) {
-                cpu_buf->last_task = task;
-                if (add_code(cpu_buf, (unsigned long)task))
-                        goto fail;
-        }
+        if (op_add_code(cpu_buf, backtrace, is_kernel, current))
+                goto fail;
 
         if (op_add_sample(cpu_buf, pc, event))
                 goto fail;
@@ -286,7 +316,6 @@ fail:
 
 static inline void oprofile_begin_trace(struct oprofile_cpu_buffer *cpu_buf)
 {
-        add_code(cpu_buf, CPU_TRACE_BEGIN);
         cpu_buf->tracing = 1;
 }
 
@@ -300,21 +329,21 @@ __oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
                           unsigned long event, int is_kernel)
 {
         struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-
-        if (!oprofile_backtrace_depth) {
-                log_sample(cpu_buf, pc, is_kernel, event);
-                return;
-        }
-
-        oprofile_begin_trace(cpu_buf);
+        unsigned long backtrace = oprofile_backtrace_depth;
 
         /*
          * if log_sample() fail we can't backtrace since we lost the
          * source of this event
          */
-        if (log_sample(cpu_buf, pc, is_kernel, event))
-                oprofile_ops.backtrace(regs, oprofile_backtrace_depth);
+        if (!log_sample(cpu_buf, pc, backtrace, is_kernel, event))
+                /* failed */
+                return;
 
+        if (!backtrace)
+                return;
+
+        oprofile_begin_trace(cpu_buf);
+        oprofile_ops.backtrace(regs, backtrace);
         oprofile_end_trace(cpu_buf);
 }
 
@@ -339,29 +368,14 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
 {
         int is_kernel = !user_mode(regs);
         struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-        struct task_struct *task;
         int fail = 0;
 
         cpu_buf->sample_received++;
 
-        /* notice a switch from user->kernel or vice versa */
-        if (cpu_buf->last_is_kernel != is_kernel) {
-                if (add_code(cpu_buf, is_kernel))
-                        goto fail;
-                cpu_buf->last_is_kernel = is_kernel;
-        }
+        /* backtraces disabled for ibs */
+        fail = fail || op_add_code(cpu_buf, 0, is_kernel, current);
 
-        /* notice a task switch */
-        if (!is_kernel) {
-                task = current;
-                if (cpu_buf->last_task != task) {
-                        if (add_code(cpu_buf, (unsigned long)task))
-                                goto fail;
-                        cpu_buf->last_task = task;
-                }
-        }
-
-        fail = fail || add_code(cpu_buf, ibs_code);
+        fail = fail || op_add_sample(cpu_buf, ESCAPE_CODE,   ibs_code);
         fail = fail || op_add_sample(cpu_buf, ibs_sample[0], ibs_sample[1]);
         fail = fail || op_add_sample(cpu_buf, ibs_sample[2], ibs_sample[3]);
         fail = fail || op_add_sample(cpu_buf, ibs_sample[4], ibs_sample[5]);
@@ -372,11 +386,8 @@ void oprofile_add_ibs_sample(struct pt_regs * const regs,
                 fail = fail || op_add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
         }
 
-        if (!fail)
-                return;
-
-fail:
-        cpu_buf->sample_lost_overflow++;
+        if (fail)
+                cpu_buf->sample_lost_overflow++;
 }
 
 #endif
@@ -384,7 +395,7 @@ fail:
 void oprofile_add_pc(unsigned long pc, int is_kernel, unsigned long event)
 {
         struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-        log_sample(cpu_buf, pc, is_kernel, event);
+        log_sample(cpu_buf, pc, 0, is_kernel, event);
 }
 
 void oprofile_add_trace(unsigned long pc)
diff --git a/drivers/oprofile/cpu_buffer.h b/drivers/oprofile/cpu_buffer.h
index d7c0545ef8b2..e634dcf2f26f 100644
--- a/drivers/oprofile/cpu_buffer.h
+++ b/drivers/oprofile/cpu_buffer.h
@@ -78,10 +78,12 @@ int op_cpu_buffer_write_commit(struct op_entry *entry);
 struct op_sample *op_cpu_buffer_read_entry(struct op_entry *entry, int cpu);
 unsigned long op_cpu_buffer_entries(int cpu);
 
-/* transient events for the CPU buffer -> event buffer */
-#define CPU_IS_KERNEL 1
-#define CPU_TRACE_BEGIN 2
-#define IBS_FETCH_BEGIN 3
-#define IBS_OP_BEGIN    4
+/* extra data flags */
+#define KERNEL_CTX_SWITCH       (1UL << 0)
+#define IS_KERNEL               (1UL << 1)
+#define TRACE_BEGIN             (1UL << 2)
+#define USER_CTX_SWITCH         (1UL << 3)
+#define IBS_FETCH_BEGIN         (1UL << 4)
+#define IBS_OP_BEGIN            (1UL << 5)
 
 #endif /* OPROFILE_CPU_BUFFER_H */