Lecture 19: More Nonblocking I/O

  • The outbound-file-test.cc presented earlier can be used to confirm the OutboundFile class implementation works as expected.
    • The nonblocking aspect of the doesn't really buy us anything.
    • Only one copy of the source file is being syndicated, so no harm comes from blocking, since there's nothing else to do.
  • To see why nonblocking I/O might be useful, consider the following nonblocking server implementation, presented over several slides.
    • Our server is a static file server and responds to every single request—no matter how that request is structures—with the same payload. That payload is drawn from an HTML called "expensive-server.cc.html".
  • Presented below is the portion of the server implementation that establishes the executable as a server that listens to port 12345 and sets the server socket to be nonblocking.
// expensive-server.html is expensive because it's always using the CPU, even when there's nothing to do
static const unsigned short kDefaultPort = 12345;
static const string kFileToServe("expensive-server.cc.html");
int main(int argc, char **argv) {
    int server = createServerSocket(kDefaultPort);
    assert(server != kServerSocketFailure);
    setAsNonBlocking(server);
    cout << "Static file server listening on port " << kDefaultPort << "." << endl;
}

Lecture 19: More Nonblocking I/O

  • As with all servers, our static file server loops interminably and aggressively accepts incoming connections as quickly as possible.
    • The first part of the while loop calls and immediately returns from accept. The return is immediate, because server has been configured to be nonblocking.
    • The code immediately following the accept call branches in one of two directions.
      • If accept returns a -1, we verify the -1 isn't something to be concerned about.
      • If accept surfaces a new connection , we create an new OutboundFile on its behalf and append it to the running outboundFiles list of clients currently being served.
list<OutboundFile> outboundFiles;
while (true) {
    // part 1: below
    int client = accept(server, NULL, NULL);
    if (client == -1) {
        assert(errno == EWOULDBLOCK); // sanitycheck to confirm -1 doesn't represent a true failure
    } else {
        OutboundFile obf;
        obf.initialize(kFileToServe, client);
        outboundFiles.push_back(obf);
    }
    // part 2: presented on next slide

Lecture 19: More Nonblocking I/O

  • As with all servers, our static file server loops interminably and aggressively accepts incoming connections as quickly as possible.
    • The second part executes whether or not part 1 produced a new client connection and extended the outboundFiles list.
    • It iterates over every single OutboundFile in the list and attempts to send some or all available data out to the client.
      • If sendMoreData returns true, the loop advances on to the next client via ++iter.
      • If sendMoreData returns false, the relevant OutboundFile is removed from outboundFiles before advancing. (Fortunately, erase does precisely what we want, and it returns the iterator addressing the next OutboundFile in the list.)
  list<OutboundFile> outboundFiles;
  while (true) {
    // part 1: presented and discussed on previous slide
    // part 2: below
    auto iter = outboundFiles.begin();
    while (iter != outboundFiles.end()) {
        if (iter->sendMoreData()) ++iter;
        else iter = outboundFiles.erase(iter);
    }
  }
}

Lecture 19: More Nonblocking I/O

  • The code for setAsNonblocking is fairly low-level.
    • It relies on a function called fcntl to do surgery on the relevant file session in the open file table.
    • That surgery is little more that toggles some 0 bit to a 1, as can be inferred from the last line of the three line implementation.
  • The code for setAsNonblocking and a few peer functions are presented below.
void setAsNonBlocking(int descriptor) {
    fcntl(descriptor, F_SETFL, fcntl(descriptor, F_GETFL) | O_NONBLOCK); // preserve other set flags
}

void setAsBlocking(int descriptor) {
    fcntl(descriptor, F_SETFL, fcntl(descriptor, F_GETFL) & ~O_NONBLOCK); // suppress blocking bit, preserve others
}

bool isNonBlocking(int descriptor) {
    return !isBlocking(descriptor);
}

bool isBlocking(int descriptor) {
    return (fcntl(descriptor, F_GETFL) & O_NONBLOCK) == 0;
}

Lecture 19: More Nonblocking I/O

  • We've been using the OutboundFile abstraction without understanding how it works behind the scenes.
    • We really should see the implementation (or at least part of it) so we have some sense how it works and can be implemented using nonblocking techniques.
    • The full implementation includes lots of spaghetti code.
    • In particular, true file descriptors and socket descriptors need to be treated differently in a few places—in particular, detecting when all data has been flushed out to the sink descriptor (which may be a local file, a console, or a remote client machine) isn't exactly pretty.
    • However, my implementation is decomposed well enough that I think many of the methods—the ones that I'll show in lecture, anyway—are easy to follow and provide a clear narrative.
    • At the very least, I'll convince you that the OutboundFile implementation is accessible to someone just finishing up CS110.

Lecture 19: More Nonblocking I/O

  • Here's is the condensed interface file for the OutboundFile class.
  • source and sink are nonblocking descriptors bound to the data source and recipient buffer is a reasonably sized character array that helps shovel bytes lifted from source via read calls over to the sink via write calls.
  • numBytesAvailable stores the number of meaningful characters in buffer.
  • numBytesSent tracks the portion of buffer that's been pushed to the recipient.
  • isSending tracks whether all data has been pulled from source and pushed to sink.
class OutboundFile {
    public:
        OutboundFile();
        void initialize(const std::string& source, int sink);
        bool sendMoreData();
    private:
        int source, sink;
        static const size_t kBufferSize = 128;
        char buffer[kBufferSize];
        size_t numBytesAvailable, numBytesSent;
        bool isSending;
        // private helper methods discussed later
};

Lecture 19: More Nonblocking I/O

  • source is a nonblocking file descriptor bound to some local file
    • Note that the source file is opened for reading (O_RDONLY), and the descriptor is configured to be nonblocking (O_NONBLOCK) right from the start.
    • For reasons we'll discussed, it's not super important that source be nonblocking, since it's bound to a local file.
    • But in the spirit of a nonblocking example, it's fine to make it nonblocking anyway. We just shouldn't expect very many (if any) -1's to come back from our read calls.
  • sink is explicitly converted to be nonblocking, since it might be blocking, and sink will very often be a socket descriptor that really should be nonblocking.
  • The implementations of the constructor and initialize are straightforward:
OutboundFile::OutboundFile() : isSending(false) {}
void OutboundFile::initialize(const string& source, int sink) {
    this->source = open(source.c_str(), O_RDONLY | O_NONBLOCK);
    this->sink = sink;
    setAsNonBlocking(this->sink);
    numBytesAvailable = numBytesSent = 0;
    isSending = true;
}

Lecture 19: More Nonblocking I/O

  • The first line decides if all data has been read from source and written to sink, and if so, it returns true unless it further confirms all of data written to sink has arrived at final destination, in which case it returns false to state that syndication is complete.
  • The first call to dataReadyToBeSent checks to see if buffer houses data yet to be pushed out. If not, then it attempts to readMoreData. If after reading more data the buffer is still empty—that is, a single call to read resulted in a -1/EWOULDBLOCK pair, then we return true as a statement that there's no data to be written, no need to try, but come back later to see if that changes.
  • The call to writeMoreData is an opportunity to push data out to sink.
  • The implementation of sendMoreData is less straightforward:
bool OutboundFile::sendMoreData() {
    if (!isSending) return !allDataFlushed();
    if (!dataReadyToBeSent()) {
        readMoreData();
        if (!dataReadyToBeSent()) return true;
    }
    writeMoreData();
    return true;
}