ring-buffer: Make removal of ring buffer pages atomic
authorVaibhav Nagarnaik <vnagarnaik@google.com>
Fri, 4 May 2012 01:59:50 +0000 (18:59 -0700)
committerSteven Rostedt <rostedt@goodmis.org>
Wed, 16 May 2012 20:18:57 +0000 (16:18 -0400)
This patch adds the capability to remove pages from a ring buffer
without destroying any existing data in it.

This is done by removing the pages after the tail page. This makes sure
that first all the empty pages in the ring buffer are removed. If the
head page is one in the list of pages to be removed, then the page after
the removed ones is made the head page. This removes the oldest data
from the ring buffer and keeps the latest data around to be read.

To do this in a non-racey manner, tracing is stopped for a very short
time while the pages to be removed are identified and unlinked from the
ring buffer. The pages are freed after the tracing is restarted to
minimize the time needed to stop tracing.

The context in which the pages from the per-cpu ring buffer are removed
runs on the respective CPU. This minimizes the events not traced to only
NMI trace contexts.

Link: http://lkml.kernel.org/r/1336096792-25373-1-git-send-email-vnagarnaik@google.com
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Laurent Chavey <chavey@google.com>
Cc: Justin Teravest <teravest@google.com>
Cc: David Sharp <dhsharp@google.com>
Signed-off-by: Vaibhav Nagarnaik <vnagarnaik@google.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
kernel/trace/ring_buffer.c
kernel/trace/trace.c

index 2d5eb33..27ac37e 100644 (file)
@@ -23,6 +23,8 @@
 #include <asm/local.h>
 #include "trace.h"
 
+static void update_pages_handler(struct work_struct *work);
+
 /*
  * The ring buffer header is special. We must manually up keep it.
  */
@@ -470,12 +472,15 @@ struct ring_buffer_per_cpu {
        /* ring buffer pages to update, > 0 to add, < 0 to remove */
        int                             nr_pages_to_update;
        struct list_head                new_pages; /* new pages to add */
+       struct work_struct              update_pages_work;
+       struct completion               update_completion;
 };
 
 struct ring_buffer {
        unsigned                        flags;
        int                             cpus;
        atomic_t                        record_disabled;
+       atomic_t                        resize_disabled;
        cpumask_var_t                   cpumask;
 
        struct lock_class_key           *reader_lock_key;
@@ -1048,6 +1053,8 @@ rb_allocate_cpu_buffer(struct ring_buffer *buffer, int nr_pages, int cpu)
        raw_spin_lock_init(&cpu_buffer->reader_lock);
        lockdep_set_class(&cpu_buffer->reader_lock, buffer->reader_lock_key);
        cpu_buffer->lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
+       INIT_WORK(&cpu_buffer->update_pages_work, update_pages_handler);
+       init_completion(&cpu_buffer->update_completion);
 
        bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()),
                            GFP_KERNEL, cpu_to_node(cpu));
@@ -1235,32 +1242,123 @@ void ring_buffer_set_clock(struct ring_buffer *buffer,
 
 static void rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer);
 
+static inline unsigned long rb_page_entries(struct buffer_page *bpage)
+{
+       return local_read(&bpage->entries) & RB_WRITE_MASK;
+}
+
+static inline unsigned long rb_page_write(struct buffer_page *bpage)
+{
+       return local_read(&bpage->write) & RB_WRITE_MASK;
+}
+
 static void
-rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned nr_pages)
+rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned int nr_pages)
 {
-       struct buffer_page *bpage;
-       struct list_head *p;
-       unsigned i;
+       struct list_head *tail_page, *to_remove, *next_page;
+       struct buffer_page *to_remove_page, *tmp_iter_page;
+       struct buffer_page *last_page, *first_page;
+       unsigned int nr_removed;
+       unsigned long head_bit;
+       int page_entries;
+
+       head_bit = 0;
 
        raw_spin_lock_irq(&cpu_buffer->reader_lock);
-       rb_head_page_deactivate(cpu_buffer);
+       atomic_inc(&cpu_buffer->record_disabled);
+       /*
+        * We don't race with the readers since we have acquired the reader
+        * lock. We also don't race with writers after disabling recording.
+        * This makes it easy to figure out the first and the last page to be
+        * removed from the list. We unlink all the pages in between including
+        * the first and last pages. This is done in a busy loop so that we
+        * lose the least number of traces.
+        * The pages are freed after we restart recording and unlock readers.
+        */
+       tail_page = &cpu_buffer->tail_page->list;
 
-       for (i = 0; i < nr_pages; i++) {
-               if (RB_WARN_ON(cpu_buffer, list_empty(cpu_buffer->pages)))
-                       goto out;
-               p = cpu_buffer->pages->next;
-               bpage = list_entry(p, struct buffer_page, list);
-               list_del_init(&bpage->list);
-               free_buffer_page(bpage);
+       /*
+        * tail page might be on reader page, we remove the next page
+        * from the ring buffer
+        */
+       if (cpu_buffer->tail_page == cpu_buffer->reader_page)
+               tail_page = rb_list_head(tail_page->next);
+       to_remove = tail_page;
+
+       /* start of pages to remove */
+       first_page = list_entry(rb_list_head(to_remove->next),
+                               struct buffer_page, list);
+
+       for (nr_removed = 0; nr_removed < nr_pages; nr_removed++) {
+               to_remove = rb_list_head(to_remove)->next;
+               head_bit |= (unsigned long)to_remove & RB_PAGE_HEAD;
        }
-       if (RB_WARN_ON(cpu_buffer, list_empty(cpu_buffer->pages)))
-               goto out;
 
-       rb_reset_cpu(cpu_buffer);
-       rb_check_pages(cpu_buffer);
+       next_page = rb_list_head(to_remove)->next;
 
-out:
+       /*
+        * Now we remove all pages between tail_page and next_page.
+        * Make sure that we have head_bit value preserved for the
+        * next page
+        */
+       tail_page->next = (struct list_head *)((unsigned long)next_page |
+                                               head_bit);
+       next_page = rb_list_head(next_page);
+       next_page->prev = tail_page;
+
+       /* make sure pages points to a valid page in the ring buffer */
+       cpu_buffer->pages = next_page;
+
+       /* update head page */
+       if (head_bit)
+               cpu_buffer->head_page = list_entry(next_page,
+                                               struct buffer_page, list);
+
+       /*
+        * change read pointer to make sure any read iterators reset
+        * themselves
+        */
+       cpu_buffer->read = 0;
+
+       /* pages are removed, resume tracing and then free the pages */
+       atomic_dec(&cpu_buffer->record_disabled);
        raw_spin_unlock_irq(&cpu_buffer->reader_lock);
+
+       RB_WARN_ON(cpu_buffer, list_empty(cpu_buffer->pages));
+
+       /* last buffer page to remove */
+       last_page = list_entry(rb_list_head(to_remove), struct buffer_page,
+                               list);
+       tmp_iter_page = first_page;
+
+       do {
+               to_remove_page = tmp_iter_page;
+               rb_inc_page(cpu_buffer, &tmp_iter_page);
+
+               /* update the counters */
+               page_entries = rb_page_entries(to_remove_page);
+               if (page_entries) {
+                       /*
+                        * If something was added to this page, it was full
+                        * since it is not the tail page. So we deduct the
+                        * bytes consumed in ring buffer from here.
+                        * No need to update overruns, since this page is
+                        * deleted from ring buffer and its entries are
+                        * already accounted for.
+                        */
+                       local_sub(BUF_PAGE_SIZE, &cpu_buffer->entries_bytes);
+               }
+
+               /*
+                * We have already removed references to this list item, just
+                * free up the buffer_page and its page
+                */
+               free_buffer_page(to_remove_page);
+               nr_removed--;
+
+       } while (to_remove_page != last_page);
+
+       RB_WARN_ON(cpu_buffer, nr_removed);
 }
 
 static void
@@ -1272,6 +1370,8 @@ rb_insert_pages(struct ring_buffer_per_cpu *cpu_buffer,
        unsigned i;
 
        raw_spin_lock_irq(&cpu_buffer->reader_lock);
+       /* stop the writers while inserting pages */
+       atomic_inc(&cpu_buffer->record_disabled);
        rb_head_page_deactivate(cpu_buffer);
 
        for (i = 0; i < nr_pages; i++) {
@@ -1286,19 +1386,27 @@ rb_insert_pages(struct ring_buffer_per_cpu *cpu_buffer,
        rb_check_pages(cpu_buffer);
 
 out:
+       atomic_dec(&cpu_buffer->record_disabled);
        raw_spin_unlock_irq(&cpu_buffer->reader_lock);
 }
 
-static void update_pages_handler(struct ring_buffer_per_cpu *cpu_buffer)
+static void rb_update_pages(struct ring_buffer_per_cpu *cpu_buffer)
 {
        if (cpu_buffer->nr_pages_to_update > 0)
                rb_insert_pages(cpu_buffer, &cpu_buffer->new_pages,
                                cpu_buffer->nr_pages_to_update);
        else
                rb_remove_pages(cpu_buffer, -cpu_buffer->nr_pages_to_update);
+
        cpu_buffer->nr_pages += cpu_buffer->nr_pages_to_update;
-       /* reset this value */
-       cpu_buffer->nr_pages_to_update = 0;
+}
+
+static void update_pages_handler(struct work_struct *work)
+{
+       struct ring_buffer_per_cpu *cpu_buffer = container_of(work,
+                       struct ring_buffer_per_cpu, update_pages_work);
+       rb_update_pages(cpu_buffer);
+       complete(&cpu_buffer->update_completion);
 }
 
 /**
@@ -1308,14 +1416,14 @@ static void update_pages_handler(struct ring_buffer_per_cpu *cpu_buffer)
  *
  * Minimum size is 2 * BUF_PAGE_SIZE.
  *
- * Returns -1 on failure.
+ * Returns 0 on success and < 0 on failure.
  */
 int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size,
                        int cpu_id)
 {
        struct ring_buffer_per_cpu *cpu_buffer;
        unsigned nr_pages;
-       int cpu;
+       int cpu, err = 0;
 
        /*
         * Always succeed at resizing a non-existent buffer:
@@ -1330,15 +1438,18 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size,
        if (size < BUF_PAGE_SIZE * 2)
                size = BUF_PAGE_SIZE * 2;
 
-       atomic_inc(&buffer->record_disabled);
+       nr_pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE);
 
-       /* Make sure all writers are done with this buffer. */
-       synchronize_sched();
+       /*
+        * Don't succeed if resizing is disabled, as a reader might be
+        * manipulating the ring buffer and is expecting a sane state while
+        * this is true.
+        */
+       if (atomic_read(&buffer->resize_disabled))
+               return -EBUSY;
 
+       /* prevent another thread from changing buffer sizes */
        mutex_lock(&buffer->mutex);
-       get_online_cpus();
-
-       nr_pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE);
 
        if (cpu_id == RING_BUFFER_ALL_CPUS) {
                /* calculate the pages to update */
@@ -1347,33 +1458,67 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size,
 
                        cpu_buffer->nr_pages_to_update = nr_pages -
                                                        cpu_buffer->nr_pages;
-
                        /*
                         * nothing more to do for removing pages or no update
                         */
                        if (cpu_buffer->nr_pages_to_update <= 0)
                                continue;
-
                        /*
                         * to add pages, make sure all new pages can be
                         * allocated without receiving ENOMEM
                         */
                        INIT_LIST_HEAD(&cpu_buffer->new_pages);
                        if (__rb_allocate_pages(cpu_buffer->nr_pages_to_update,
-                                               &cpu_buffer->new_pages, cpu))
+                                               &cpu_buffer->new_pages, cpu)) {
                                /* not enough memory for new pages */
-                               goto no_mem;
+                               err = -ENOMEM;
+                               goto out_err;
+                       }
+               }
+
+               get_online_cpus();
+               /*
+                * Fire off all the required work handlers
+                * Look out for offline CPUs
+                */
+               for_each_buffer_cpu(buffer, cpu) {
+                       cpu_buffer = buffer->buffers[cpu];
+                       if (!cpu_buffer->nr_pages_to_update ||
+                           !cpu_online(cpu))
+                               continue;
+
+                       schedule_work_on(cpu, &cpu_buffer->update_pages_work);
+               }
+               /*
+                * This loop is for the CPUs that are not online.
+                * We can't schedule anything on them, but it's not necessary
+                * since we can change their buffer sizes without any race.
+                */
+               for_each_buffer_cpu(buffer, cpu) {
+                       cpu_buffer = buffer->buffers[cpu];
+                       if (!cpu_buffer->nr_pages_to_update ||
+                           cpu_online(cpu))
+                               continue;
+
+                       rb_update_pages(cpu_buffer);
                }
 
                /* wait for all the updates to complete */
                for_each_buffer_cpu(buffer, cpu) {
                        cpu_buffer = buffer->buffers[cpu];
-                       if (cpu_buffer->nr_pages_to_update) {
-                               update_pages_handler(cpu_buffer);
-                       }
+                       if (!cpu_buffer->nr_pages_to_update ||
+                           !cpu_online(cpu))
+                               continue;
+
+                       wait_for_completion(&cpu_buffer->update_completion);
+                       /* reset this value */
+                       cpu_buffer->nr_pages_to_update = 0;
                }
+
+               put_online_cpus();
        } else {
                cpu_buffer = buffer->buffers[cpu_id];
+
                if (nr_pages == cpu_buffer->nr_pages)
                        goto out;
 
@@ -1383,38 +1528,47 @@ int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size,
                INIT_LIST_HEAD(&cpu_buffer->new_pages);
                if (cpu_buffer->nr_pages_to_update > 0 &&
                        __rb_allocate_pages(cpu_buffer->nr_pages_to_update,
-                                               &cpu_buffer->new_pages, cpu_id))
-                       goto no_mem;
+                                           &cpu_buffer->new_pages, cpu_id)) {
+                       err = -ENOMEM;
+                       goto out_err;
+               }
 
-               update_pages_handler(cpu_buffer);
+               get_online_cpus();
+
+               if (cpu_online(cpu_id)) {
+                       schedule_work_on(cpu_id,
+                                        &cpu_buffer->update_pages_work);
+                       wait_for_completion(&cpu_buffer->update_completion);
+               } else
+                       rb_update_pages(cpu_buffer);
+
+               put_online_cpus();
+               /* reset this value */
+               cpu_buffer->nr_pages_to_update = 0;
        }
 
  out:
-       put_online_cpus();
        mutex_unlock(&buffer->mutex);
-
-       atomic_dec(&buffer->record_disabled);
-
        return size;
 
no_mem:
out_err:
        for_each_buffer_cpu(buffer, cpu) {
                struct buffer_page *bpage, *tmp;
+
                cpu_buffer = buffer->buffers[cpu];
-               /* reset this number regardless */
                cpu_buffer->nr_pages_to_update = 0;
+
                if (list_empty(&cpu_buffer->new_pages))
                        continue;
+
                list_for_each_entry_safe(bpage, tmp, &cpu_buffer->new_pages,
                                        list) {
                        list_del_init(&bpage->list);
                        free_buffer_page(bpage);
                }
        }
-       put_online_cpus();
        mutex_unlock(&buffer->mutex);
-       atomic_dec(&buffer->record_disabled);
-       return -ENOMEM;
+       return err;
 }
 EXPORT_SYMBOL_GPL(ring_buffer_resize);
 
@@ -1453,21 +1607,11 @@ rb_iter_head_event(struct ring_buffer_iter *iter)
        return __rb_page_index(iter->head_page, iter->head);
 }
 
-static inline unsigned long rb_page_write(struct buffer_page *bpage)
-{
-       return local_read(&bpage->write) & RB_WRITE_MASK;
-}
-
 static inline unsigned rb_page_commit(struct buffer_page *bpage)
 {
        return local_read(&bpage->page->commit);
 }
 
-static inline unsigned long rb_page_entries(struct buffer_page *bpage)
-{
-       return local_read(&bpage->entries) & RB_WRITE_MASK;
-}
-
 /* Size is determined by what has been committed */
 static inline unsigned rb_page_size(struct buffer_page *bpage)
 {
@@ -3492,6 +3636,7 @@ ring_buffer_read_prepare(struct ring_buffer *buffer, int cpu)
 
        iter->cpu_buffer = cpu_buffer;
 
+       atomic_inc(&buffer->resize_disabled);
        atomic_inc(&cpu_buffer->record_disabled);
 
        return iter;
@@ -3555,6 +3700,7 @@ ring_buffer_read_finish(struct ring_buffer_iter *iter)
        struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer;
 
        atomic_dec(&cpu_buffer->record_disabled);
+       atomic_dec(&cpu_buffer->buffer->resize_disabled);
        kfree(iter);
 }
 EXPORT_SYMBOL_GPL(ring_buffer_read_finish);
@@ -3662,8 +3808,12 @@ void ring_buffer_reset_cpu(struct ring_buffer *buffer, int cpu)
        if (!cpumask_test_cpu(cpu, buffer->cpumask))
                return;
 
+       atomic_inc(&buffer->resize_disabled);
        atomic_inc(&cpu_buffer->record_disabled);
 
+       /* Make sure all commits have finished */
+       synchronize_sched();
+
        raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
 
        if (RB_WARN_ON(cpu_buffer, local_read(&cpu_buffer->committing)))
@@ -3679,6 +3829,7 @@ void ring_buffer_reset_cpu(struct ring_buffer *buffer, int cpu)
        raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
 
        atomic_dec(&cpu_buffer->record_disabled);
+       atomic_dec(&buffer->resize_disabled);
 }
 EXPORT_SYMBOL_GPL(ring_buffer_reset_cpu);
 
index d1b3469..dfbd86c 100644 (file)
@@ -3076,20 +3076,10 @@ static int __tracing_resize_ring_buffer(unsigned long size, int cpu)
 
 static ssize_t tracing_resize_ring_buffer(unsigned long size, int cpu_id)
 {
-       int cpu, ret = size;
+       int ret = size;
 
        mutex_lock(&trace_types_lock);
 
-       tracing_stop();
-
-       /* disable all cpu buffers */
-       for_each_tracing_cpu(cpu) {
-               if (global_trace.data[cpu])
-                       atomic_inc(&global_trace.data[cpu]->disabled);
-               if (max_tr.data[cpu])
-                       atomic_inc(&max_tr.data[cpu]->disabled);
-       }
-
        if (cpu_id != RING_BUFFER_ALL_CPUS) {
                /* make sure, this cpu is enabled in the mask */
                if (!cpumask_test_cpu(cpu_id, tracing_buffer_mask)) {
@@ -3103,14 +3093,6 @@ static ssize_t tracing_resize_ring_buffer(unsigned long size, int cpu_id)
                ret = -ENOMEM;
 
 out:
-       for_each_tracing_cpu(cpu) {
-               if (global_trace.data[cpu])
-                       atomic_dec(&global_trace.data[cpu]->disabled);
-               if (max_tr.data[cpu])
-                       atomic_dec(&max_tr.data[cpu]->disabled);
-       }
-
-       tracing_start();
        mutex_unlock(&trace_types_lock);
 
        return ret;