ring-buffer: convert to raw spinlocks
authorSteven Rostedt <srostedt@redhat.com>
Thu, 6 Nov 2008 05:09:43 +0000 (00:09 -0500)
committerIngo Molnar <mingo@elte.hu>
Thu, 6 Nov 2008 06:51:09 +0000 (07:51 +0100)
Impact: no lockdep debugging of ring buffer

The problem with running lockdep on the ring buffer is that the
ring buffer is the core infrastructure of ftrace. What happens is
that the tracer will start tracing the lockdep code while lockdep
is testing the ring buffers locks.  This can cause lockdep to
fail due to testing cases that have not fully finished their
locking transition.

This patch converts the spin locks used by the ring buffer back
into raw spin locks which lockdep does not check.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/trace/ring_buffer.c

index 151f6a74867674a4ef48bb5c6320a7accfd3daeb..a2dea5008826bc65fed9249ae416b13a531374e6 100644 (file)
@@ -154,7 +154,7 @@ static inline int test_time_stamp(u64 delta)
 struct ring_buffer_per_cpu {
        int                             cpu;
        struct ring_buffer              *buffer;
-       spinlock_t                      lock;
+       raw_spinlock_t                  lock;
        struct lock_class_key           lock_key;
        struct list_head                pages;
        struct buffer_page              *head_page;     /* read from head */
@@ -291,7 +291,7 @@ rb_allocate_cpu_buffer(struct ring_buffer *buffer, int cpu)
 
        cpu_buffer->cpu = cpu;
        cpu_buffer->buffer = buffer;
-       spin_lock_init(&cpu_buffer->lock);
+       cpu_buffer->lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED;
        INIT_LIST_HEAD(&cpu_buffer->pages);
 
        page = kzalloc_node(ALIGN(sizeof(*page), cache_line_size()),
@@ -854,7 +854,8 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
        if (write > BUF_PAGE_SIZE) {
                struct buffer_page *next_page = tail_page;
 
-               spin_lock_irqsave(&cpu_buffer->lock, flags);
+               local_irq_save(flags);
+               __raw_spin_lock(&cpu_buffer->lock);
 
                rb_inc_page(cpu_buffer, &next_page);
 
@@ -930,7 +931,8 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
                        rb_set_commit_to_write(cpu_buffer);
                }
 
-               spin_unlock_irqrestore(&cpu_buffer->lock, flags);
+               __raw_spin_unlock(&cpu_buffer->lock);
+               local_irq_restore(flags);
 
                /* fail and let the caller try again */
                return ERR_PTR(-EAGAIN);
@@ -953,7 +955,8 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
        return event;
 
  out_unlock:
-       spin_unlock_irqrestore(&cpu_buffer->lock, flags);
+       __raw_spin_unlock(&cpu_buffer->lock);
+       local_irq_restore(flags);
        return NULL;
 }
 
@@ -1524,7 +1527,8 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
        struct buffer_page *reader = NULL;
        unsigned long flags;
 
-       spin_lock_irqsave(&cpu_buffer->lock, flags);
+       local_irq_save(flags);
+       __raw_spin_lock(&cpu_buffer->lock);
 
  again:
        reader = cpu_buffer->reader_page;
@@ -1574,7 +1578,8 @@ rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
        goto again;
 
  out:
-       spin_unlock_irqrestore(&cpu_buffer->lock, flags);
+       __raw_spin_unlock(&cpu_buffer->lock);
+       local_irq_restore(flags);
 
        return reader;
 }
@@ -1815,9 +1820,11 @@ ring_buffer_read_start(struct ring_buffer *buffer, int cpu)
        atomic_inc(&cpu_buffer->record_disabled);
        synchronize_sched();
 
-       spin_lock_irqsave(&cpu_buffer->lock, flags);
+       local_irq_save(flags);
+       __raw_spin_lock(&cpu_buffer->lock);
        ring_buffer_iter_reset(iter);
-       spin_unlock_irqrestore(&cpu_buffer->lock, flags);
+       __raw_spin_unlock(&cpu_buffer->lock);
+       local_irq_restore(flags);
 
        return iter;
 }
@@ -1903,11 +1910,13 @@ void ring_buffer_reset_cpu(struct ring_buffer *buffer, int cpu)
        if (!cpu_isset(cpu, buffer->cpumask))
                return;
 
-       spin_lock_irqsave(&cpu_buffer->lock, flags);
+       local_irq_save(flags);
+       __raw_spin_lock(&cpu_buffer->lock);
 
        rb_reset_cpu(cpu_buffer);
 
-       spin_unlock_irqrestore(&cpu_buffer->lock, flags);
+       __raw_spin_unlock(&cpu_buffer->lock);
+       local_irq_restore(flags);
 }
 
 /**