Changeset 7b64cf0 in mainline

Timestamp:

2013-05-16T11:09:44Z (11 years ago)

Author:

Jakub Jermar <jakub@…>

Branches:

lfn, master, serial, ticket/834-toolchain-update, topic/msim-upgrade, topic/simplify-dev-export

Children:

4510e06

Parents:

2093fbe

Message:

More complete implementation of the SLIP service and protocol.

Note that the suggested EXCHANGE_PARALLEL session does not work due to
the effects of:

#508 Parallel sessions don't mix well with call forwarding

Besides of that, the ns8250 driver will not allow opening the device
more than once, which also needs to be addressed.

File:

• uspace/srv/net/slip/slip.c (modified) (9 diffs)

uspace/srv/net/slip/slip.c

@@ -36 +36 @@
 
 #include <stdio.h>
+#include <stdint.h>
 #include <loc.h>
 #include <inet/iplink_srv.h>
+#include <device/char_dev.h>
 #include <io/log.h>
 #include <errno.h>
@@ -45 +47 @@
 
 #define SLIP_MTU        1006    /* as per RFC 1055 */
+
+#define SLIP_END        0300
+#define SLIP_ESC        0333
+#define SLIP_ESC_END    0334
+#define SLIP_ESC_ESC    0335
 
 static int slip_open(iplink_srv_t *);
@@ -64 +71 @@
 };
 
+static uint8_t slip_send_buf[SLIP_MTU + 2];
+static size_t slip_send_pending;
+
+static uint8_t slip_recv_buf[SLIP_MTU + 2];
+static size_t slip_recv_pending;
+static size_t slip_recv_read;
+
 int slip_open(iplink_srv_t *srv)
 {
@@ -76 +90 @@
 }
 
+static void write_flush(async_sess_t *sess)
+{
+        size_t written = 0;
+
+        while (slip_send_pending > 0) {
+                ssize_t size;
+
+                size = char_dev_write(sess, &slip_send_buf[written],
+                    slip_send_pending);
+                if (size < 0) {
+                        log_msg(LOG_DEFAULT, LVL_ERROR,
+                            "char_dev_write() returned %d",
+                            (int) size);
+                        slip_send_pending = 0;
+                        break;
+                }
+                written += size;
+                slip_send_pending -= size;
+        }
+}
+
+static void write_buffered(async_sess_t *sess, uint8_t ch)
+{
+        if (slip_send_pending == sizeof(slip_send_buf))
+                write_flush(sess);
+        slip_send_buf[slip_send_pending++] = ch;
+}
+
 int slip_send(iplink_srv_t *srv, iplink_srv_sdu_t *sdu)
 {
+        async_sess_t *sess = (async_sess_t *) srv->arg;
+        uint8_t *data = sdu->data;
+        unsigned i;
+
         log_msg(LOG_DEFAULT, LVL_DEBUG, "slip_send()");
-        return ENOTSUP;
+
+        /*
+         * Strictly speaking, this is not prescribed by the RFC, but the RFC
+         * suggests to start with sending a SLIP_END byte as a synchronization
+         * measure for dealing with previous possible noise on the line.
+         */
+        write_buffered(sess, SLIP_END);
+
+        for (i = 0; i < sdu->size; i++) {
+                switch (data[i]) {
+                case SLIP_END:
+                        write_buffered(sess, SLIP_ESC);
+                        write_buffered(sess, SLIP_ESC_END);     
+                        break;
+                case SLIP_ESC:
+                        write_buffered(sess, SLIP_ESC);
+                        write_buffered(sess, SLIP_ESC_ESC);
+                        break;
+                default:
+                        write_buffered(sess, data[i]);
+                        break;
+                }
+        }
+        write_buffered(sess, SLIP_END);
+        write_flush(sess);
+
+        return EOK;
 }
 
@@ -110 +182 @@
         log_msg(LOG_DEFAULT, LVL_DEBUG, "slip_client_conn()");
         iplink_conn(iid, icall, &slip_iplink);
+}
+
+static uint8_t read_buffered(async_sess_t *sess)
+{
+        while (slip_recv_pending == 0) {
+                ssize_t size;
+
+                size = char_dev_read(sess, slip_recv_buf,
+                    sizeof(slip_recv_buf));
+                if (size < 0) {
+                        log_msg(LOG_DEFAULT, LVL_ERROR,
+                            "char_dev_read() returned %d", (int) size);
+                        return SLIP_END;
+                }
+                slip_recv_pending = size;
+                slip_recv_read = 0;
+        }
+        slip_recv_pending--;
+        return slip_recv_buf[slip_recv_read++];
+}
+
+static int slip_recv_fibril(void *arg)
+{
+        async_sess_t *sess = (async_sess_t *) arg;
+        static uint8_t recv_final[SLIP_MTU];
+        iplink_srv_sdu_t sdu;
+        uint8_t ch;
+        int rc;
+
+        sdu.lsrc.ipv4 = 0;
+        sdu.ldest.ipv4 = 0;
+        sdu.data = recv_final;
+
+        while (true) {
+                for (sdu.size = 0; sdu.size < sizeof(recv_final); /**/) {
+                        ch = read_buffered(sess);
+                        switch (ch) {
+                        case SLIP_END:
+                                if (sdu.size == 0) {
+                                        /*
+                                         * Discard the empty SLIP datagram.
+                                         */
+                                        break;
+                                }
+                                goto pass;
+
+                        case SLIP_ESC:
+                                ch = read_buffered(sess);
+                                if (ch == SLIP_ESC_END) {
+                                        recv_final[sdu.size++] = SLIP_END;
+                                        break;
+                                } else if (ch ==  SLIP_ESC_ESC) {
+                                        recv_final[sdu.size++] = SLIP_ESC;
+                                        break;
+                                }
+
+                                /*
+                                 * The RFC suggests to simply insert the wrongly
+                                 * escaped character into the packet so we fall
+                                 * through.
+                                 */
+ 
+                        default:
+                                recv_final[sdu.size++] = ch;
+                                break;
+                        }
+                        
+                }
+
+                /*
+                 * We have reached the limit of our MTU. Regardless of whether
+                 * the datagram is properly ended with SLIP_END, pass it along.
+                 * If the next character is really SLIP_END, nothing
+                 * catastrophic happens. The algorithm will just see an
+                 * artificially empty SLIP datagram and life will go on.
+                 */
+
+pass:
+                rc = iplink_ev_recv(&slip_iplink, &sdu);
+                if (rc != EOK) {
+                        log_msg(LOG_DEFAULT, LVL_ERROR,
+                            "iplink_ev_recv() returned %d", rc);
+                }
+        }
+
+        return 0;
 }
 
@@ -118 +276 @@
         category_id_t iplinkcid;
         async_sess_t *sess = NULL;
+        fid_t fid;
         int rc;
 
@@ -147 +306 @@
         }
 
-        sess = loc_service_connect(EXCHANGE_SERIALIZE, svcid, 0);
+        /*
+         * Create a parallel session because we will need to be able to both
+         * read and write from the char_dev.
+         */
+        sess = loc_service_connect(EXCHANGE_PARALLEL, svcid, 0);
         if (!sess) {
                 log_msg(LOG_DEFAULT, LVL_ERROR,
@@ -154 +317 @@
                 return rc;
         }
+        slip_iplink.arg = sess;
 
         rc = loc_service_register(linkstr, &linksid);
@@ -166 +330 @@
         if (rc != EOK) {
                 log_msg(LOG_DEFAULT, LVL_ERROR,
-                    "Filed to add service %d (%s) to category %d (%s).",
+                    "Failed to add service %d (%s) to category %d (%s).",
                     (int) linksid, linkstr, (int) iplinkcid, CAT_IPLINK);
                 goto fail;
         }
+
+        fid = fibril_create(slip_recv_fibril, sess);
+        if (!fid) {
+                log_msg(LOG_DEFAULT, LVL_ERROR,
+                    "Failed to create receive fibril.");
+                goto fail;
+        }
+        fibril_add_ready(fid);
 
         return EOK;