Merge branch 'oprofile/ring_buffer' into oprofile/oprofile-for-tip

7 files changed, 408 insertions, 326 deletions

diff --git a/drivers/oprofile/buffer_sync.c b/drivers/oprofile/buffer_sync.c
index b55cd23ffdef..ac014cb27915 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
@@ -268,18 +269,6 @@ lookup_dcookie(struct mm_struct *mm, unsigned long addr, off_t *offset)
         return cookie;
 }
 
-static void increment_tail(struct oprofile_cpu_buffer *b)
-{
-        unsigned long new_tail = b->tail_pos + 1;
-
-        rmb();  /* be sure fifo pointers are synchromized */
-
-        if (new_tail < b->buffer_size)
-                b->tail_pos = new_tail;
-        else
-                b->tail_pos = 0;
-}
-
 static unsigned long last_cookie = INVALID_COOKIE;
 
 static void add_cpu_switch(int i)
@@ -327,84 +316,73 @@ static void add_trace_begin(void)
         add_event_entry(TRACE_BEGIN_CODE);
 }
 
-#ifdef CONFIG_OPROFILE_IBS
-
-#define IBS_FETCH_CODE_SIZE     2
-#define IBS_OP_CODE_SIZE        5
-#define IBS_EIP(offset)                         \
-        (((struct op_sample *)&cpu_buf->buffer[(offset)])->eip)
-#define IBS_EVENT(offset)                               \
-        (((struct op_sample *)&cpu_buf->buffer[(offset)])->event)
-
-/*
- * Add IBS fetch and op entries to event buffer
- */
-static void add_ibs_begin(struct oprofile_cpu_buffer *cpu_buf, int code,
-                          struct mm_struct *mm)
+static void add_data(struct op_entry *entry, struct mm_struct *mm)
 {
-        unsigned long rip;
-        int i, count;
-        unsigned long ibs_cookie = 0;
+        unsigned long code, pc, val;
+        unsigned long cookie;
         off_t offset;
 
-        increment_tail(cpu_buf);        /* move to RIP entry */
-
-        rip = IBS_EIP(cpu_buf->tail_pos);
-
-#ifdef __LP64__
-        rip += IBS_EVENT(cpu_buf->tail_pos) << 32;
-#endif
+        if (!op_cpu_buffer_get_data(entry, &code))
+                return;
+        if (!op_cpu_buffer_get_data(entry, &pc))
+                return;
+        if (!op_cpu_buffer_get_size(entry))
+                return;
 
         if (mm) {
-                ibs_cookie = lookup_dcookie(mm, rip, &offset);
+                cookie = lookup_dcookie(mm, pc, &offset);
 
-                if (ibs_cookie == NO_COOKIE)
-                        offset = rip;
-                if (ibs_cookie == INVALID_COOKIE) {
+                if (cookie == NO_COOKIE)
+                        offset = pc;
+                if (cookie == INVALID_COOKIE) {
                         atomic_inc(&oprofile_stats.sample_lost_no_mapping);
-                        offset = rip;
+                        offset = pc;
                 }
-                if (ibs_cookie != last_cookie) {
-                        add_cookie_switch(ibs_cookie);
-                        last_cookie = ibs_cookie;
+                if (cookie != last_cookie) {
+                        add_cookie_switch(cookie);
+                        last_cookie = cookie;
                 }
         } else
-                offset = rip;
+                offset = pc;
 
         add_event_entry(ESCAPE_CODE);
         add_event_entry(code);
         add_event_entry(offset);        /* Offset from Dcookie */
 
-        /* we send the Dcookie offset, but send the raw Linear Add also*/
-        add_event_entry(IBS_EIP(cpu_buf->tail_pos));
-        add_event_entry(IBS_EVENT(cpu_buf->tail_pos));
-
-        if (code == IBS_FETCH_CODE)
-                count = IBS_FETCH_CODE_SIZE;    /*IBS FETCH is 2 int64s*/
-        else
-                count = IBS_OP_CODE_SIZE;       /*IBS OP is 5 int64s*/
-
-        for (i = 0; i < count; i++) {
-                increment_tail(cpu_buf);
-                add_event_entry(IBS_EIP(cpu_buf->tail_pos));
-                add_event_entry(IBS_EVENT(cpu_buf->tail_pos));
-        }
+        while (op_cpu_buffer_get_data(entry, &val))
+                add_event_entry(val);
 }
 
-#endif
-
-static void add_sample_entry(unsigned long offset, unsigned long event)
+static inline void add_sample_entry(unsigned long offset, unsigned long event)
 {
         add_event_entry(offset);
         add_event_entry(event);
 }
 
 
-static int add_us_sample(struct mm_struct *mm, struct op_sample *s)
+/*
+ * Add a sample to the global event buffer. If possible the
+ * sample is converted into a persistent dentry/offset pair
+ * for later lookup from userspace. Return 0 on failure.
+ */
+static int
+add_sample(struct mm_struct *mm, struct op_sample *s, int in_kernel)
 {
         unsigned long cookie;
         off_t offset;
 
+        if (in_kernel) {
+                add_sample_entry(s->eip, s->event);
+                return 1;
+        }
+
+        /* add userspace sample */
+
+        if (!mm) {
+                atomic_inc(&oprofile_stats.sample_lost_no_mm);
+                return 0;
+        }
+
         cookie = lookup_dcookie(mm, s->eip, &offset);
 
         if (cookie == INVALID_COOKIE) {
@@ -423,25 +401,6 @@ static int add_us_sample(struct mm_struct *mm, struct op_sample *s)
 }
 
 
-/* Add a sample to the global event buffer. If possible the
- * sample is converted into a persistent dentry/offset pair
- * for later lookup from userspace.
- */
-static int
-add_sample(struct mm_struct *mm, struct op_sample *s, int in_kernel)
-{
-        if (in_kernel) {
-                add_sample_entry(s->eip, s->event);
-                return 1;
-        } else if (mm) {
-                return add_us_sample(mm, s);
-        } else {
-                atomic_inc(&oprofile_stats.sample_lost_no_mm);
-        }
-        return 0;
-}
-
-
 static void release_mm(struct mm_struct *mm)
 {
         if (!mm)
@@ -466,33 +425,6 @@ static inline int is_code(unsigned long val)
 }
 
 
-/* "acquire" as many cpu buffer slots as we can */
-static unsigned long get_slots(struct oprofile_cpu_buffer *b)
-{
-        unsigned long head = b->head_pos;
-        unsigned long tail = b->tail_pos;
-
-        /*
-         * Subtle. This resets the persistent last_task
-         * and in_kernel values used for switching notes.
-         * BUT, there is a small window between reading
-         * head_pos, and this call, that means samples
-         * can appear at the new head position, but not
-         * be prefixed with the notes for switching
-         * kernel mode or a task switch. This small hole
-         * can lead to mis-attribution or samples where
-         * we don't know if it's in the kernel or not,
-         * at the start of an event buffer.
-         */
-        cpu_buffer_reset(b);
-
-        if (head >= tail)
-                return head - tail;
-
-        return head + (b->buffer_size - tail);
-}
-
-
 /* Move tasks along towards death. Any tasks on dead_tasks
  * will definitely have no remaining references in any
  * CPU buffers at this point, because we use two lists,
@@ -559,71 +491,72 @@ typedef enum {
  */
 void sync_buffer(int cpu)
 {
-        struct oprofile_cpu_buffer *cpu_buf = &per_cpu(cpu_buffer, cpu);
         struct mm_struct *mm = NULL;
+        struct mm_struct *oldmm;
+        unsigned long val;
         struct task_struct *new;
         unsigned long cookie = 0;
         int in_kernel = 1;
         sync_buffer_state state = sb_buffer_start;
-#ifndef CONFIG_OPROFILE_IBS
         unsigned int i;
         unsigned long available;
-#endif
+        unsigned long flags;
+        struct op_entry entry;
+        struct op_sample *sample;
 
         mutex_lock(&buffer_mutex);
 
         add_cpu_switch(cpu);
 
-        /* Remember, only we can modify tail_pos */
-
-#ifndef CONFIG_OPROFILE_IBS
-        available = get_slots(cpu_buf);
+        op_cpu_buffer_reset(cpu);
+        available = op_cpu_buffer_entries(cpu);
 
         for (i = 0; i < available; ++i) {
-#else
-        while (get_slots(cpu_buf)) {
-#endif
-                struct op_sample *s = &cpu_buf->buffer[cpu_buf->tail_pos];
+                sample = op_cpu_buffer_read_entry(&entry, cpu);
+                if (!sample)
+                        break;
 
-                if (is_code(s->eip)) {
-                        if (s->event <= CPU_IS_KERNEL) {
+                if (is_code(sample->eip)) {
+                        flags = sample->event;
+                        if (flags & TRACE_BEGIN) {
+                                state = sb_bt_start;
+                                add_trace_begin();
+                        }
+                        if (flags & KERNEL_CTX_SWITCH) {
                                 /* kernel/userspace switch */
-                                in_kernel = s->event;
+                                in_kernel = flags & IS_KERNEL;
                                 if (state == sb_buffer_start)
                                         state = sb_sample_start;
-                                add_kernel_ctx_switch(s->event);
-                        } else if (s->event == CPU_TRACE_BEGIN) {
-                                state = sb_bt_start;
-                                add_trace_begin();
-#ifdef CONFIG_OPROFILE_IBS
-                        } else if (s->event == IBS_FETCH_BEGIN) {
-                                state = sb_bt_start;
-                                add_ibs_begin(cpu_buf, IBS_FETCH_CODE, mm);
-                        } else if (s->event == IBS_OP_BEGIN) {
-                                state = sb_bt_start;
-                                add_ibs_begin(cpu_buf, IBS_OP_CODE, mm);
-#endif
-                        } else {
-                                struct mm_struct *oldmm = mm;
-
+                                add_kernel_ctx_switch(flags & IS_KERNEL);
+                        }
+                        if (flags & USER_CTX_SWITCH
+                            && op_cpu_buffer_get_data(&entry, &val)) {
                                 /* userspace context switch */
-                                new = (struct task_struct *)s->event;
-
+                                new = (struct task_struct *)val;
+                                oldmm = mm;
                                 release_mm(oldmm);
                                 mm = take_tasks_mm(new);
                                 if (mm != oldmm)
                                         cookie = get_exec_dcookie(mm);
                                 add_user_ctx_switch(new, cookie);
                         }
-                } else if (state >= sb_bt_start &&
-                           !add_sample(mm, s, in_kernel)) {
-                        if (state == sb_bt_start) {
-                                state = sb_bt_ignore;
-                                atomic_inc(&oprofile_stats.bt_lost_no_mapping);
-                        }
+                        if (op_cpu_buffer_get_size(&entry))
+                                add_data(&entry, mm);
+                        continue;
                 }
 
-                increment_tail(cpu_buf);
+                if (state < sb_bt_start)
+                        /* ignore sample */
+                        continue;
+
+                if (add_sample(mm, sample, in_kernel))
+                        continue;
+
+                /* ignore backtraces if failed to add a sample */
+                if (state == sb_bt_start) {
+                        state = sb_bt_ignore;
+                        atomic_inc(&oprofile_stats.bt_lost_no_mapping);
+                }
         }
         release_mm(mm);
 
diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c
index 01d38e78cde1..2e03b6d796d3 100644
--- a/drivers/oprofile/cpu_buffer.c
+++ b/drivers/oprofile/cpu_buffer.c
@@ -1,11 +1,12 @@
 /**
  * @file cpu_buffer.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 <barry.kasindorf@amd.com>
+ * @author Robert Richter <robert.richter@amd.com>
  *
  * Each CPU has a local buffer that stores PC value/event
  * pairs. We also log context switches when we notice them.
@@ -28,6 +29,25 @@
 #include "buffer_sync.h"
 #include "oprof.h"
 
+#define OP_BUFFER_FLAGS 0
+
+/*
+ * Read and write access is using spin locking. Thus, writing to the
+ * buffer by NMI handler (x86) could occur also during critical
+ * sections when reading the buffer. To avoid this, there are 2
+ * buffers for independent read and write access. Read access is in
+ * process context only, write access only in the NMI handler. If the
+ * read buffer runs empty, both buffers are swapped atomically. There
+ * is potentially a small window during swapping where the buffers are
+ * disabled and samples could be lost.
+ *
+ * Using 2 buffers is a little bit overhead, but the solution is clear
+ * and does not require changes in the ring buffer implementation. It
+ * can be changed to a single buffer solution when the ring buffer
+ * access is implemented as non-locking atomic code.
+ */
+static struct ring_buffer *op_ring_buffer_read;
+static struct ring_buffer *op_ring_buffer_write;
 DEFINE_PER_CPU(struct oprofile_cpu_buffer, cpu_buffer);
 
 static void wq_sync_buffer(struct work_struct *work);
@@ -35,19 +55,9 @@ static void wq_sync_buffer(struct work_struct *work);
 #define DEFAULT_TIMER_EXPIRE (HZ / 10)
 static int work_enabled;
 
-void free_cpu_buffers(void)
-{
-        int i;
-
-        for_each_possible_cpu(i) {
-                vfree(per_cpu(cpu_buffer, i).buffer);
-                per_cpu(cpu_buffer, i).buffer = NULL;
-        }
-}
-
 unsigned long oprofile_get_cpu_buffer_size(void)
 {
-        return fs_cpu_buffer_size;
+        return oprofile_cpu_buffer_size;
 }
 
 void oprofile_cpu_buffer_inc_smpl_lost(void)
@@ -58,26 +68,36 @@ void oprofile_cpu_buffer_inc_smpl_lost(void)
         cpu_buf->sample_lost_overflow++;
 }
 
+void free_cpu_buffers(void)
+{
+        if (op_ring_buffer_read)
+                ring_buffer_free(op_ring_buffer_read);
+        op_ring_buffer_read = NULL;
+        if (op_ring_buffer_write)
+                ring_buffer_free(op_ring_buffer_write);
+        op_ring_buffer_write = NULL;
+}
+
 int alloc_cpu_buffers(void)
 {
         int i;
 
-        unsigned long buffer_size = fs_cpu_buffer_size;
+        unsigned long buffer_size = oprofile_cpu_buffer_size;
+
+        op_ring_buffer_read = ring_buffer_alloc(buffer_size, OP_BUFFER_FLAGS);
+        if (!op_ring_buffer_read)
+                goto fail;
+        op_ring_buffer_write = ring_buffer_alloc(buffer_size, OP_BUFFER_FLAGS);
+        if (!op_ring_buffer_write)
+                goto fail;
 
         for_each_possible_cpu(i) {
                 struct oprofile_cpu_buffer *b = &per_cpu(cpu_buffer, i);
 
-                b->buffer = vmalloc_node(sizeof(struct op_sample) * buffer_size,
-                        cpu_to_node(i));
-                if (!b->buffer)
-                        goto fail;
-
                 b->last_task = NULL;
                 b->last_is_kernel = -1;
                 b->tracing = 0;
                 b->buffer_size = buffer_size;
-                b->tail_pos = 0;
-                b->head_pos = 0;
                 b->sample_received = 0;
                 b->sample_lost_overflow = 0;
                 b->backtrace_aborted = 0;
@@ -124,73 +144,156 @@ void end_cpu_work(void)
         flush_scheduled_work();
 }
 
-/* Resets the cpu buffer to a sane state. */
-void cpu_buffer_reset(struct oprofile_cpu_buffer *cpu_buf)
+/*
+ * This function prepares the cpu buffer to write a sample.
+ *
+ * Struct op_entry is used during operations on the ring buffer while
+ * struct op_sample contains the data that is stored in the ring
+ * buffer. Struct entry can be uninitialized. The function reserves a
+ * data array that is specified by size. Use
+ * op_cpu_buffer_write_commit() after preparing the sample. In case of
+ * errors a null pointer is returned, otherwise the pointer to the
+ * sample.
+ *
+ */
+struct op_sample
+*op_cpu_buffer_write_reserve(struct op_entry *entry, unsigned long size)
 {
-        /* reset these to invalid values; the next sample
-         * collected will populate the buffer with proper
-         * values to initialize the buffer
-         */
-        cpu_buf->last_is_kernel = -1;
-        cpu_buf->last_task = NULL;
+        entry->event = ring_buffer_lock_reserve
+                (op_ring_buffer_write, sizeof(struct op_sample) +
+                 size * sizeof(entry->sample->data[0]), &entry->irq_flags);
+        if (entry->event)
+                entry->sample = ring_buffer_event_data(entry->event);
+        else
+                entry->sample = NULL;
+
+        if (!entry->sample)
+                return NULL;
+
+        entry->size = size;
+        entry->data = entry->sample->data;
+
+        return entry->sample;
 }
 
-/* compute number of available slots in cpu_buffer queue */
-static unsigned long nr_available_slots(struct oprofile_cpu_buffer const *b)
+int op_cpu_buffer_write_commit(struct op_entry *entry)
 {
-        unsigned long head = b->head_pos;
-        unsigned long tail = b->tail_pos;
+        return ring_buffer_unlock_commit(op_ring_buffer_write, entry->event,
+                                         entry->irq_flags);
+}
 
-        if (tail > head)
-                return (tail - head) - 1;
+struct op_sample *op_cpu_buffer_read_entry(struct op_entry *entry, int cpu)
+{
+        struct ring_buffer_event *e;
+        e = ring_buffer_consume(op_ring_buffer_read, cpu, NULL);
+        if (e)
+                goto event;
+        if (ring_buffer_swap_cpu(op_ring_buffer_read,
+                                 op_ring_buffer_write,
+                                 cpu))
+                return NULL;
+        e = ring_buffer_consume(op_ring_buffer_read, cpu, NULL);
+        if (e)
+                goto event;
+        return NULL;
+
+event:
+        entry->event = e;
+        entry->sample = ring_buffer_event_data(e);
+        entry->size = (ring_buffer_event_length(e) - sizeof(struct op_sample))
+                / sizeof(entry->sample->data[0]);
+        entry->data = entry->sample->data;
+        return entry->sample;
+}
 
-        return tail + (b->buffer_size - head) - 1;
+unsigned long op_cpu_buffer_entries(int cpu)
+{
+        return ring_buffer_entries_cpu(op_ring_buffer_read, cpu)
+                + ring_buffer_entries_cpu(op_ring_buffer_write, cpu);
 }
 
-static void increment_head(struct oprofile_cpu_buffer *b)
+static int
+op_add_code(struct oprofile_cpu_buffer *cpu_buf, unsigned long backtrace,
+            int is_kernel, struct task_struct *task)
 {
-        unsigned long new_head = b->head_pos + 1;
+        struct op_entry entry;
+        struct op_sample *sample;
+        unsigned long flags;
+        int size;
+
+        flags = 0;
 
-        /* Ensure anything written to the slot before we
-         * increment is visible */
-        wmb();
+        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 (new_head < b->buffer_size)
-                b->head_pos = new_head;
+        if (flags & USER_CTX_SWITCH)
+                size = 1;
         else
-                b->head_pos = 0;
-}
+                size = 0;
 
-static inline void
-add_sample(struct oprofile_cpu_buffer *cpu_buf,
-           unsigned long pc, unsigned long event)
-{
-        struct op_sample *entry = &cpu_buf->buffer[cpu_buf->head_pos];
-        entry->eip = pc;
-        entry->event = event;
-        increment_head(cpu_buf);
+        sample = op_cpu_buffer_write_reserve(&entry, size);
+        if (!sample)
+                return -ENOMEM;
+
+        sample->eip = ESCAPE_CODE;
+        sample->event = flags;
+
+        if (size)
+                op_cpu_buffer_add_data(&entry, (unsigned long)task);
+
+        op_cpu_buffer_write_commit(&entry);
+
+        return 0;
 }
 
-static inline void
-add_code(struct oprofile_cpu_buffer *buffer, unsigned long value)
+static inline int
+op_add_sample(struct oprofile_cpu_buffer *cpu_buf,
+              unsigned long pc, unsigned long event)
 {
-        add_sample(buffer, ESCAPE_CODE, value);
+        struct op_entry entry;
+        struct op_sample *sample;
+
+        sample = op_cpu_buffer_write_reserve(&entry, 0);
+        if (!sample)
+                return -ENOMEM;
+
+        sample->eip = pc;
+        sample->event = event;
+
+        return op_cpu_buffer_write_commit(&entry);
 }
 
-/* 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) {
@@ -198,131 +301,115 @@ static int log_sample(struct oprofile_cpu_buffer *cpu_buf, unsigned long pc,
                 return 0;
         }
 
-        if (nr_available_slots(cpu_buf) < 3) {
-                cpu_buf->sample_lost_overflow++;
-                return 0;
-        }
-
-        is_kernel = !!is_kernel;
+        if (op_add_code(cpu_buf, backtrace, is_kernel, current))
+                goto fail;
 
-        task = current;
+        if (op_add_sample(cpu_buf, pc, event))
+                goto fail;
 
-        /* notice a switch from user->kernel or vice versa */
-        if (cpu_buf->last_is_kernel != is_kernel) {
-                cpu_buf->last_is_kernel = is_kernel;
-                add_code(cpu_buf, is_kernel);
-        }
-
-        /* notice a task switch */
-        if (cpu_buf->last_task != task) {
-                cpu_buf->last_task = task;
-                add_code(cpu_buf, (unsigned long)task);
-        }
-
-        add_sample(cpu_buf, pc, event);
         return 1;
+
+fail:
+        cpu_buf->sample_lost_overflow++;
+        return 0;
 }
 
-static int oprofile_begin_trace(struct oprofile_cpu_buffer *cpu_buf)
+static inline void oprofile_begin_trace(struct oprofile_cpu_buffer *cpu_buf)
 {
-        if (nr_available_slots(cpu_buf) < 4) {
-                cpu_buf->sample_lost_overflow++;
-                return 0;
-        }
-
-        add_code(cpu_buf, CPU_TRACE_BEGIN);
         cpu_buf->tracing = 1;
-        return 1;
 }
 
-static void oprofile_end_trace(struct oprofile_cpu_buffer *cpu_buf)
+static inline void oprofile_end_trace(struct oprofile_cpu_buffer *cpu_buf)
 {
         cpu_buf->tracing = 0;
 }
 
-void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
-                                unsigned long event, int is_kernel)
+static inline void
+__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);
+        unsigned long backtrace = oprofile_backtrace_depth;
 
-        if (!backtrace_depth) {
-                log_sample(cpu_buf, pc, is_kernel, event);
+        /*
+         * if log_sample() fail we can't backtrace since we lost the
+         * source of this event
+         */
+        if (!log_sample(cpu_buf, pc, backtrace, is_kernel, event))
+                /* failed */
                 return;
-        }
 
-        if (!oprofile_begin_trace(cpu_buf))
+        if (!backtrace)
                 return;
 
-        /* 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, backtrace_depth);
+        oprofile_begin_trace(cpu_buf);
+        oprofile_ops.backtrace(regs, backtrace);
         oprofile_end_trace(cpu_buf);
 }
 
+void oprofile_add_ext_sample(unsigned long pc, struct pt_regs * const regs,
+                             unsigned long event, int is_kernel)
+{
+        __oprofile_add_ext_sample(pc, regs, event, is_kernel);
+}
+
 void oprofile_add_sample(struct pt_regs * const regs, unsigned long event)
 {
         int is_kernel = !user_mode(regs);
         unsigned long pc = profile_pc(regs);
 
-        oprofile_add_ext_sample(pc, regs, event, is_kernel);
+        __oprofile_add_ext_sample(pc, regs, event, is_kernel);
 }
 
-#ifdef CONFIG_OPROFILE_IBS
-
-#define MAX_IBS_SAMPLE_SIZE 14
-
-void oprofile_add_ibs_sample(struct pt_regs *const regs,
-                             unsigned int *const ibs_sample, int ibs_code)
+/*
+ * Add samples with data to the ring buffer.
+ *
+ * Use oprofile_add_data(&entry, val) to add data and
+ * oprofile_write_commit(&entry) to commit the sample.
+ */
+void
+oprofile_write_reserve(struct op_entry *entry, struct pt_regs * const regs,
+                       unsigned long pc, int code, int size)
 {
+        struct op_sample *sample;
         int is_kernel = !user_mode(regs);
         struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer);
-        struct task_struct *task;
 
         cpu_buf->sample_received++;
 
-        if (nr_available_slots(cpu_buf) < MAX_IBS_SAMPLE_SIZE) {
-                /* we can't backtrace since we lost the source of this event */
-                cpu_buf->sample_lost_overflow++;
-                return;
-        }
+        /* no backtraces for samples with data */
+        if (op_add_code(cpu_buf, 0, is_kernel, current))
+                goto fail;
 
-        /* notice a switch from user->kernel or vice versa */
-        if (cpu_buf->last_is_kernel != is_kernel) {
-                cpu_buf->last_is_kernel = is_kernel;
-                add_code(cpu_buf, is_kernel);
-        }
+        sample = op_cpu_buffer_write_reserve(entry, size + 2);
+        if (!sample)
+                goto fail;
+        sample->eip = ESCAPE_CODE;
+        sample->event = 0;              /* no flags */
 
-        /* notice a task switch */
-        if (!is_kernel) {
-                task = current;
-                if (cpu_buf->last_task != task) {
-                        cpu_buf->last_task = task;
-                        add_code(cpu_buf, (unsigned long)task);
-                }
-        }
+        op_cpu_buffer_add_data(entry, code);
+        op_cpu_buffer_add_data(entry, pc);
 
-        add_code(cpu_buf, ibs_code);
-        add_sample(cpu_buf, ibs_sample[0], ibs_sample[1]);
-        add_sample(cpu_buf, ibs_sample[2], ibs_sample[3]);
-        add_sample(cpu_buf, ibs_sample[4], ibs_sample[5]);
+        return;
 
-        if (ibs_code == IBS_OP_BEGIN) {
-                add_sample(cpu_buf, ibs_sample[6], ibs_sample[7]);
-                add_sample(cpu_buf, ibs_sample[8], ibs_sample[9]);
-                add_sample(cpu_buf, ibs_sample[10], ibs_sample[11]);
-        }
+fail:
+        cpu_buf->sample_lost_overflow++;
+}
 
-        if (backtrace_depth)
-                oprofile_ops.backtrace(regs, backtrace_depth);
+int oprofile_add_data(struct op_entry *entry, unsigned long val)
+{
+        return op_cpu_buffer_add_data(entry, val);
 }
 
-#endif
+int oprofile_write_commit(struct op_entry *entry)
+{
+        return op_cpu_buffer_write_commit(entry);
+}
 
 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)
@@ -332,21 +419,21 @@ void oprofile_add_trace(unsigned long pc)
         if (!cpu_buf->tracing)
                 return;
 
-        if (nr_available_slots(cpu_buf) < 1) {
-                cpu_buf->tracing = 0;
-                cpu_buf->sample_lost_overflow++;
-                return;
-        }
+        /*
+         * broken frame can give an eip with the same value as an
+         * escape code, abort the trace if we get it
+         */
+        if (pc == ESCAPE_CODE)
+                goto fail;
 
-        /* broken frame can give an eip with the same value as an escape code,
-         * abort the trace if we get it */
-        if (pc == ESCAPE_CODE) {
-                cpu_buf->tracing = 0;
-                cpu_buf->backtrace_aborted++;
-                return;
-        }
+        if (op_add_sample(cpu_buf, pc, 0))
+                goto fail;
 
-        add_sample(cpu_buf, pc, 0);
+        return;
+fail:
+        cpu_buf->tracing = 0;
+        cpu_buf->backtrace_aborted++;
+        return;
 }
 
 /*
diff --git a/drivers/oprofile/cpu_buffer.h b/drivers/oprofile/cpu_buffer.h
index d3cc26264db5..63f81c44846a 100644
--- a/drivers/oprofile/cpu_buffer.h
+++ b/drivers/oprofile/cpu_buffer.h
@@ -1,10 +1,11 @@
 /**
  * @file cpu_buffer.h
  *
- * @remark Copyright 2002 OProfile authors
+ * @remark Copyright 2002-2009 OProfile authors
  * @remark Read the file COPYING
  *
  * @author John Levon <levon@movementarian.org>
+ * @author Robert Richter <robert.richter@amd.com>
  */
 
 #ifndef OPROFILE_CPU_BUFFER_H
@@ -15,6 +16,7 @@
 #include <linux/workqueue.h>
 #include <linux/cache.h>
 #include <linux/sched.h>
+#include <linux/ring_buffer.h>
 
 struct task_struct;
 
@@ -30,16 +32,16 @@ void end_cpu_work(void);
 struct op_sample {
         unsigned long eip;
         unsigned long event;
+        unsigned long data[0];
 };
 
+struct op_entry;
+
 struct oprofile_cpu_buffer {
-        volatile unsigned long head_pos;
-        volatile unsigned long tail_pos;
         unsigned long buffer_size;
         struct task_struct *last_task;
         int last_is_kernel;
         int tracing;
-        struct op_sample *buffer;
         unsigned long sample_received;
         unsigned long sample_lost_overflow;
         unsigned long backtrace_aborted;
@@ -50,12 +52,62 @@ struct oprofile_cpu_buffer {
 
 DECLARE_PER_CPU(struct oprofile_cpu_buffer, cpu_buffer);
 
-void cpu_buffer_reset(struct oprofile_cpu_buffer *cpu_buf);
+/*
+ * Resets the cpu buffer to a sane state.
+ *
+ * reset these to invalid values; the next sample collected will
+ * populate the buffer with proper values to initialize the buffer
+ */
+static inline void op_cpu_buffer_reset(int cpu)
+{
+        struct oprofile_cpu_buffer *cpu_buf = &per_cpu(cpu_buffer, cpu);
+
+        cpu_buf->last_is_kernel = -1;
+        cpu_buf->last_task = NULL;
+}
+
+struct op_sample
+*op_cpu_buffer_write_reserve(struct op_entry *entry, unsigned long size);
+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);
+
+/* returns the remaining free size of data in the entry */
+static inline
+int op_cpu_buffer_add_data(struct op_entry *entry, unsigned long val)
+{
+        if (!entry->size)
+                return 0;
+        *entry->data = val;
+        entry->size--;
+        entry->data++;
+        return entry->size;
+}
+
+/* returns the size of data in the entry */
+static inline
+int op_cpu_buffer_get_size(struct op_entry *entry)
+{
+        return entry->size;
+}
+
+/* returns 0 if empty or the size of data including the current value */
+static inline
+int op_cpu_buffer_get_data(struct op_entry *entry, unsigned long *val)
+{
+        int size = entry->size;
+        if (!size)
+                return 0;
+        *val = *entry->data;
+        entry->size--;
+        entry->data++;
+        return size;
+}
 
-/* 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)
 
 #endif /* OPROFILE_CPU_BUFFER_H */
diff --git a/drivers/oprofile/event_buffer.c b/drivers/oprofile/event_buffer.c
index 191a3202cecc..2b7ae366ceb1 100644
--- a/drivers/oprofile/event_buffer.c
+++ b/drivers/oprofile/event_buffer.c
@@ -73,8 +73,8 @@ int alloc_event_buffer(void)
         unsigned long flags;
 
         spin_lock_irqsave(&oprofilefs_lock, flags);
-        buffer_size = fs_buffer_size;
-        buffer_watershed = fs_buffer_watershed;
+        buffer_size = oprofile_buffer_size;
+        buffer_watershed = oprofile_buffer_watershed;
         spin_unlock_irqrestore(&oprofilefs_lock, flags);
 
         if (buffer_watershed >= buffer_size)
diff --git a/drivers/oprofile/oprof.c b/drivers/oprofile/oprof.c
index cd375907f26f..3cffce90f82a 100644
--- a/drivers/oprofile/oprof.c
+++ b/drivers/oprofile/oprof.c
@@ -23,7 +23,7 @@
 struct oprofile_operations oprofile_ops;
 
 unsigned long oprofile_started;
-unsigned long backtrace_depth;
+unsigned long oprofile_backtrace_depth;
 static unsigned long is_setup;
 static DEFINE_MUTEX(start_mutex);
 
@@ -172,7 +172,7 @@ int oprofile_set_backtrace(unsigned long val)
                 goto out;
         }
 
-        backtrace_depth = val;
+        oprofile_backtrace_depth = val;
 
 out:
         mutex_unlock(&start_mutex);
diff --git a/drivers/oprofile/oprof.h b/drivers/oprofile/oprof.h
index 5df0c21a608f..c288d3c24b50 100644
--- a/drivers/oprofile/oprof.h
+++ b/drivers/oprofile/oprof.h
@@ -21,12 +21,12 @@ void oprofile_stop(void);
 
 struct oprofile_operations;
 
-extern unsigned long fs_buffer_size;
-extern unsigned long fs_cpu_buffer_size;
-extern unsigned long fs_buffer_watershed;
+extern unsigned long oprofile_buffer_size;
+extern unsigned long oprofile_cpu_buffer_size;
+extern unsigned long oprofile_buffer_watershed;
 extern struct oprofile_operations oprofile_ops;
 extern unsigned long oprofile_started;
-extern unsigned long backtrace_depth;
+extern unsigned long oprofile_backtrace_depth;
 
 struct super_block;
 struct dentry;
diff --git a/drivers/oprofile/oprofile_files.c b/drivers/oprofile/oprofile_files.c
index cc106d503ace..5d36ffc30dd5 100644
--- a/drivers/oprofile/oprofile_files.c
+++ b/drivers/oprofile/oprofile_files.c
@@ -14,13 +14,18 @@
 #include "oprofile_stats.h"
 #include "oprof.h"
 
-unsigned long fs_buffer_size = 131072;
-unsigned long fs_cpu_buffer_size = 8192;
-unsigned long fs_buffer_watershed = 32768; /* FIXME: tune */
+#define BUFFER_SIZE_DEFAULT             131072
+#define CPU_BUFFER_SIZE_DEFAULT         8192
+#define BUFFER_WATERSHED_DEFAULT        32768   /* FIXME: tune */
+
+unsigned long oprofile_buffer_size;
+unsigned long oprofile_cpu_buffer_size;
+unsigned long oprofile_buffer_watershed;
 
 static ssize_t depth_read(struct file *file, char __user *buf, size_t count, loff_t *offset)
 {
-        return oprofilefs_ulong_to_user(backtrace_depth, buf, count, offset);
+        return oprofilefs_ulong_to_user(oprofile_backtrace_depth, buf, count,
+                                        offset);
 }
 
 
@@ -120,12 +125,17 @@ static const struct file_operations dump_fops = {
 
 void oprofile_create_files(struct super_block *sb, struct dentry *root)
 {
+        /* reinitialize default values */
+        oprofile_buffer_size =          BUFFER_SIZE_DEFAULT;
+        oprofile_cpu_buffer_size =      CPU_BUFFER_SIZE_DEFAULT;
+        oprofile_buffer_watershed =     BUFFER_WATERSHED_DEFAULT;
+
         oprofilefs_create_file(sb, root, "enable", &enable_fops);
         oprofilefs_create_file_perm(sb, root, "dump", &dump_fops, 0666);
         oprofilefs_create_file(sb, root, "buffer", &event_buffer_fops);
-        oprofilefs_create_ulong(sb, root, "buffer_size", &fs_buffer_size);
-        oprofilefs_create_ulong(sb, root, "buffer_watershed", &fs_buffer_watershed);
-        oprofilefs_create_ulong(sb, root, "cpu_buffer_size", &fs_cpu_buffer_size);
+        oprofilefs_create_ulong(sb, root, "buffer_size", &oprofile_buffer_size);
+        oprofilefs_create_ulong(sb, root, "buffer_watershed", &oprofile_buffer_watershed);
+        oprofilefs_create_ulong(sb, root, "cpu_buffer_size", &oprofile_cpu_buffer_size);
         oprofilefs_create_file(sb, root, "cpu_type", &cpu_type_fops);
         oprofilefs_create_file(sb, root, "backtrace_depth", &depth_fops);
         oprofilefs_create_file(sb, root, "pointer_size", &pointer_size_fops);