tracing: Fix crash when it fails to alloc ring buffer
authorJing Xia <jing.xia@spreadtrum.com>
Tue, 26 Dec 2017 07:12:53 +0000 (15:12 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 2 Jan 2018 19:33:22 +0000 (20:33 +0100)
commit 24f2aaf952ee0b59f31c3a18b8b36c9e3d3c2cf5 upstream.

Double free of the ring buffer happens when it fails to alloc new
ring buffer instance for max_buffer if TRACER_MAX_TRACE is configured.
The root cause is that the pointer is not set to NULL after the buffer
is freed in allocate_trace_buffers(), and the freeing of the ring
buffer is invoked again later if the pointer is not equal to Null,
as:

instance_mkdir()
    |-allocate_trace_buffers()
        |-allocate_trace_buffer(tr, &tr->trace_buffer...)
|-allocate_trace_buffer(tr, &tr->max_buffer...)

          // allocate fail(-ENOMEM),first free
          // and the buffer pointer is not set to null
        |-ring_buffer_free(tr->trace_buffer.buffer)

       // out_free_tr
    |-free_trace_buffers()
        |-free_trace_buffer(&tr->trace_buffer);

      //if trace_buffer is not null, free again
    |-ring_buffer_free(buf->buffer)
                |-rb_free_cpu_buffer(buffer->buffers[cpu])
                    // ring_buffer_per_cpu is null, and
                    // crash in ring_buffer_per_cpu->pages

Link: http://lkml.kernel.org/r/20171226071253.8968-1-chunyan.zhang@spreadtrum.com
Fixes: 737223fbca3b1 ("tracing: Consolidate buffer allocation code")
Signed-off-by: Jing Xia <jing.xia@spreadtrum.com>
Signed-off-by: Chunyan Zhang <chunyan.zhang@spreadtrum.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kernel/trace/trace.c

index b0452eefc2150cef65f88d87e4419763e71387ea..8aef4e63ac57c705c5ad7e7f9523d61d6d2369b7 100644 (file)
@@ -6555,7 +6555,9 @@ static int allocate_trace_buffers(struct trace_array *tr, int size)
                                    allocate_snapshot ? size : 1);
        if (WARN_ON(ret)) {
                ring_buffer_free(tr->trace_buffer.buffer);
+               tr->trace_buffer.buffer = NULL;
                free_percpu(tr->trace_buffer.data);
+               tr->trace_buffer.data = NULL;
                return -ENOMEM;
        }
        tr->allocated_snapshot = allocate_snapshot;