I have a problem with socket send (or write) function on android.
There is my network lib that I use on Linux and Android. Code is written in C.
On Android, application creates a service, which loads a native code and creates the connection with the help of my network lib. Connection is the TCP socket. When I call send (or write, no difference), code hangs in this call in most cases. Sometimes, it unhangs after 10-120 seconds. Sometimes, it waits longer (until I kill the application). Data size being sent is about 40-50 bytes. First data sending (handshake, 5 bytes) never hangs (or I am just lucky). The hanging send is, usually, next after handshake packet. Time between this first handshake packet sending and hanging sending is about 10-20 seconds.
The socket is used on another thread (I use pthread), where the recv is called. But, I do not send data to Android in this time, so recv is just waiting when I call send.
I am sure that other side is waiting for the data – I see that recv on other side returns with EAGAIN every 3 seconds (I set timeout) and immediately calls recv again. Recv is waiting 10 bytes always (minimal size of packet).
I am unable to reproduce this behavior on Linux-to-Android transfer or Linux-to-Linux, only on Adnroid-to-Linux. I am able to reproduce this with two available to me different Android devices, so I don’t think this is the problem in broken hardware of one particular device.
I tried to set SO_KEEPALIVE and TCP_NODELAY options with no success.
What can issue the hang-up on send/write calls and how can I resolve this?
Socket created with this code:
int sockfd, n;
addrinfo hints, *res, *ressave;
bzero(&hints, sizeof(addrinfo));
hints.ai_family = AF_INET;
hints.ai_socktype = SOCK_STREAM;
if ((n = getaddrinfo(host, serv, &hints, &res)) != 0)
{ /* stripped error handling*/ }
ressave = res;
do
{
sockfd = socket(res->ai_family, res->ai_socktype, res->ai_protocol);
if (sockfd < 0) continue;
if (connect(sockfd, res->ai_addr, res->ai_addrlen) == 0)
{
break; /* success */
}
close(sockfd); /* ignore this one */
} while ((res = res->ai_next) != NULL);
Hanging send operation is:
mWriteMutex.lock();
mSocketMutex.lockRead();
ssize_t n = send(mSocket, pArray, size, 0);
mSocketMutex.unlock();
mWriteMutex.unlock();
The problem is solved with the help of Nikolai N Fetissov in commentaries – his right question has unblocked my mind and I found a problem in RWMutex.