because poll() returns readable as long as there is any data in the socket buffer for
unix sockets, not respecting SO_RCVLOWAT.
- XXX: We could have used SO_PEEK_OFF to continously peek more data without changing the socket
- receive buffer, but this fucntion breaks since Linux 4.3 due to a kernel bug, which is fixed
+ XXX: We could have used SO_PEEK_OFF to continuously peek more data without changing the socket
+ receive buffer, but this function breaks since Linux 4.3 due to a kernel bug, which is fixed
in Linux 6.18 commit 7bf3a476ce43 ("af_unix: Read sk_peek_offset() again after sleeping in
unix_stream_read_generic()."). It is also not possible to detect whether the kernel is
affected by this bug at runtime.