On Fri, May 08, 2020 at 04:29:37PM +0300, Serge Semin wrote: > Since the TX and RX DMA-slave config structures are declared on the > stack and only a part of their fields is initialized further in the > prepare Tx/Rx methods, the next dmaengine_slave_config() calls may > randomly perform an incorrect DMA channels setup depending on the > stack initial values. Lets fix this by zeroing the config-structures > before using them. Already in spi/for-next. -- With Best Regards, Andy Shevchenko
