1) One thread issues scatter-gather boost::asio::async_write() for boost::array<boost::asio::const_buffer, 2> on socket A. 2) Another thread closes socket A after that. 3) Meantime (simultaneously with 2), first buffer write is completed and one of the threads running io_service::run() starts to execute second part of async_write from 1).

The problem is: second write which is issued internally by boost cannot be protected by the same mutex which is used for protected access to the socket in my code.

At first, RTFM: 1. http://www.boost.org/doc/libs/1_53_0/doc/html/boost_asio/reference/Handler.h tml ~~~~~~~~~~~~~~ asio_handler_invoke(f, &h); ~~~~~~~~~~~~~~ 2. http://www.boost.org/doc/libs/1_53_0/doc/html/boost_asio/overview/core/stran ds.html ~~~~~~~~~~~~~~ In the case of composed asynchronous operations, such as async_read() or async_read_until(), if a completion handler goes through a strand, then all intermediate handlers should also go through the same strand. This is needed to ensure thread safe access for any objects that are shared between the caller and the composed operation (in the case of async_read() it's the socket, which the caller can close() to cancel the operation). This is done by having hook functions for all intermediate handlers which forward the calls to the customisable hook associated with the final handler: ~~~~~~~~~~~~~~ Than google some more: 3. https://github.com/boostcon/2011_presentations/raw/master/mon/thinking_async hronously.pdf See template <class Handler> struct mutex_wrapper at page 95. 4. http://asio-samples.blogspot.ru/2011/06/design-journeys-with-asio.html Sorry, it's in Russian. Regards, Marat Abrarov.