Fix: don't flush-final for offset 0 if reader is on sub-buffer

Triggers rare sub-buffer corruption when a buffer is completely full and
a final flush is then performed.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

diff --git a/lib/ringbuffer/ring_buffer_frontend.c b/lib/ringbuffer/ring_buffer_frontend.c
index dc0357f56a3e71547c7da80a4e2bd3ce7277d0f0..2626afa31e3ea9f33a7e9ebbab98bd91d510b723 100644 (file)
--- a/lib/ringbuffer/ring_buffer_frontend.c
+++ b/lib/ringbuffer/ring_buffer_frontend.c
@@ -1424,6 +1424,19 @@ int lib_ring_buffer_try_switch_slow(enum switch_mode mode,
         */
        if (mode == SWITCH_FLUSH || off > 0) {
                if (unlikely(off == 0)) {
+                        /*
+                        * A final flush that encounters an empty
+                        * sub-buffer cannot switch buffer if a
+                        * reader is located within this sub-buffer.
+                        * Anyway, the purpose of final flushing of a
+                        * sub-buffer at offset 0 is to handle the case
+                        * of entirely empty stream.
+                        */
+                       if (unlikely(subbuf_trunc(offsets->begin, chan)
+                                       - subbuf_trunc((unsigned long)
+                                               atomic_long_read(&buf->consumed), chan)
+                                       >= chan->backend.buf_size))
+                               return -1;
                        /*
                         * The client does not save any header information.
                         * Don't switch empty subbuffer on finalize, because it