ipcpd: Add congestion avoidance policies

This adds congestion avoidance policies to the unicast IPCP.  The
default policy is a multi-bit explicit congestion avoidance algorithm
based on data-center TCP congestion avoidance (DCTCP) to relay
information about the maximum queue depth that packets experienced to
the receiver. There's also a "nop" policy to disable congestion
avoidance for testing and benchmarking purposes.

The (initial) API for congestion avoidance policies is:

        void *   (* ctx_create)(void);

        void     (* ctx_destroy)(void * ctx);

These calls create / and or destroy a context for congestion control
for a specific flow. Thread-safety of the context is the
responsability of the flow allocator (operations on the ctx should be
performed under a lock).

        ca_wnd_t (* ctx_update_snd)(void * ctx,
                                    size_t len);

This is the sender call to update the context, and should be called
for every packet that is sent on the flow. The len parameter in this
API is the packet length, which allows calculating the bandwidth. It
returns an opaque union type that is used for the call to check/wait
if the congestion window is open or closed (and allowing to release
locks before waiting).

        bool     (* ctx_update_rcv)(void *     ctx,
                                    size_t     len,
                                    uint8_t    ecn,
                                    uint16_t * ece);

This is the call to update the flow congestion context on the receiver
side. It should be called for every received packet.  It gets the ecn
value from the packet and its length, and returns the ECE (explicit
congestion experienced) value to be sent to the sender in case of
congestion. The boolean returned signals whether or not a congestion
update needs to be sent.

        void     (* ctx_update_ece)(void *   ctx,
                                    uint16_t ece);

This is the call for the sending side top update the context when it
receives an ECE update from the receiver.

        void     (* wnd_wait)(ca_wnd_t wnd);

This is a (blocking) call that waits for the congestion window to
clear. It should be stateless (to avoid waiting under locks). This may
change later on if passing the context is needed for different algorithms.

        uint8_t  (* calc_ecn)(int    fd,
                              size_t len);

This is the call that intermediate IPCPs(routers) should use to update
the ECN field on passing packets.

The multi-bit ECN policy bases the value for the ECN field on the
depth of the rbuff queue packets will be sent on. I created another
call to grab the queue depth as fccntl is write-locking the
application. We can further optimize this to avoid most locking on the
rbuff.

Signed-off-by: Dimitri Staessens <[email protected]>
Signed-off-by: Sander Vrijders <[email protected]>

1 files changed, 170 insertions, 52 deletions

diff --git a/src/ipcpd/unicast/fa.c b/src/ipcpd/unicast/fa.c
index e154d785..8f268a9d 100644
--- a/src/ipcpd/unicast/fa.c
+++ b/src/ipcpd/unicast/fa.c
@@ -42,6 +42,7 @@
 #include "psched.h"
 #include "ipcp.h"
 #include "dt.h"
+#include "ca.h"
 
 #include <pthread.h>
 #include <stdlib.h>
@@ -49,9 +50,10 @@
 
 #define TIMEOUT 10000 /* nanoseconds */
 
-#define FLOW_REQ   0
-#define FLOW_REPLY 1
-#define MSGBUFSZ   2048
+#define FLOW_REQ    0
+#define FLOW_REPLY  1
+#define FLOW_UPDATE 2
+#define MSGBUFSZ    2048
 
 struct fa_msg {
         uint64_t s_addr;
@@ -59,6 +61,7 @@ struct fa_msg {
         uint32_t s_eid;
         uint8_t  code;
         int8_t   response;
+        uint16_t ece;
         /* QoS parameters from spec, aligned */
         uint8_t  availability;
         uint8_t  in_order;
@@ -75,10 +78,16 @@ struct cmd {
         struct shm_du_buff * sdb;
 };
 
+struct fa_flow {
+        int      r_eid;  /* remote endpoint id               */
+        uint64_t r_addr; /* remote address                   */
+        void *   ctx;    /* congestion avoidance context     */
+};
+
 struct {
         pthread_rwlock_t flows_lock;
-        int              r_eid[PROG_MAX_FLOWS];
-        uint64_t         r_addr[PROG_MAX_FLOWS];
+        struct fa_flow   flows[PROG_MAX_FLOWS];
+
         int              fd;
 
         struct list_head cmds;
@@ -93,22 +102,56 @@ static void packet_handler(int                  fd,
                            qoscube_t            qc,
                            struct shm_du_buff * sdb)
 {
-        pthread_rwlock_rdlock(&fa.flows_lock);
+        struct fa_flow * flow;
+        uint64_t         r_addr;
+        uint32_t         r_eid;
+        ca_wnd_t         wnd;
+        size_t           len;
 
-        if (dt_write_packet(fa.r_addr[fd], qc, fa.r_eid[fd], sdb)) {
-                pthread_rwlock_unlock(&fa.flows_lock);
+        flow = &fa.flows[fd];
+
+        pthread_rwlock_wrlock(&fa.flows_lock);
+
+        len = shm_du_buff_tail(sdb) - shm_du_buff_head(sdb);
+
+        wnd = ca_ctx_update_snd(flow->ctx, len);
+
+        r_addr = flow->r_addr;
+        r_eid  = flow->r_eid;
+
+        pthread_rwlock_unlock(&fa.flows_lock);
+
+        ca_wnd_wait(wnd);
+
+        if (dt_write_packet(r_addr, qc, r_eid, sdb)) {
                 ipcp_sdb_release(sdb);
                 log_warn("Failed to forward packet.");
                 return;
         }
+}
 
-        pthread_rwlock_unlock(&fa.flows_lock);
+static int fa_flow_init(struct fa_flow * flow)
+{
+        memset(flow, 0, sizeof(*flow));
+
+        flow->r_eid  = -1;
+        flow->r_addr = INVALID_ADDR;
+
+        flow->ctx = ca_ctx_create();
+        if (flow->ctx == NULL)
+                return -1;
+
+        return 0;
 }
 
-static void destroy_conn(int fd)
+static void fa_flow_fini(struct fa_flow * flow)
 {
-        fa.r_eid[fd]  = -1;
-        fa.r_addr[fd] = INVALID_ADDR;
+        ca_ctx_destroy(flow->ctx);
+
+        memset(flow, 0, sizeof(*flow));
+
+        flow->r_eid  = -1;
+        flow->r_addr = INVALID_ADDR;
 }
 
 static void fa_post_packet(void *               comp,
@@ -145,14 +188,15 @@ static void * fa_handle_packet(void * o)
         (void) o;
 
         while (true) {
-                struct timespec abstime;
-                int             fd;
-                uint8_t         buf[MSGBUFSZ];
-                struct fa_msg * msg;
-                qosspec_t       qs;
-                struct cmd *    cmd;
-                size_t          len;
-                size_t          msg_len;
+                struct timespec  abstime;
+                int              fd;
+                uint8_t          buf[MSGBUFSZ];
+                struct fa_msg *  msg;
+                qosspec_t        qs;
+                struct cmd *     cmd;
+                size_t           len;
+                size_t           msg_len;
+                struct fa_flow * flow;
 
                 pthread_mutex_lock(&fa.mtx);
 
@@ -232,10 +276,14 @@ static void * fa_handle_packet(void * o)
                                 continue;
                         }
 
+                        flow = &fa.flows[fd];
+
                         pthread_rwlock_wrlock(&fa.flows_lock);
 
-                        fa.r_eid[fd]  = ntoh32(msg->s_eid);
-                        fa.r_addr[fd] = ntoh64(msg->s_addr);
+                        fa_flow_init(flow);
+
+                        flow->r_eid  = ntoh32(msg->s_eid);
+                        flow->r_addr = ntoh64(msg->s_addr);
 
                         pthread_rwlock_unlock(&fa.flows_lock);
 
@@ -248,19 +296,32 @@ static void * fa_handle_packet(void * o)
                 case FLOW_REPLY:
                         assert(len >= sizeof(*msg));
 
+                        flow = &fa.flows[ntoh32(msg->r_eid)];
+
                         pthread_rwlock_wrlock(&fa.flows_lock);
 
-                        fa.r_eid[ntoh32(msg->r_eid)] = ntoh32(msg->s_eid);
+                        flow->r_eid = ntoh32(msg->s_eid);
+
+                        if (msg->response < 0)
+                                fa_flow_fini(flow);
+                        else
+                                psched_add(fa.psched, ntoh32(msg->r_eid));
+
+                        pthread_rwlock_unlock(&fa.flows_lock);
 
                         ipcp_flow_alloc_reply(ntoh32(msg->r_eid),
                                               msg->response,
                                               buf + sizeof(*msg),
                                               len - sizeof(*msg));
+                        break;
+                case FLOW_UPDATE:
+                        assert(len >= sizeof(*msg));
 
-                        if (msg->response < 0)
-                                destroy_conn(ntoh32(msg->r_eid));
-                        else
-                                psched_add(fa.psched, ntoh32(msg->r_eid));
+                        flow = &fa.flows[ntoh32(msg->r_eid)];
+
+                        pthread_rwlock_wrlock(&fa.flows_lock);
+
+                        ca_ctx_update_ece(flow->ctx, ntoh16(msg->ece));
 
                         pthread_rwlock_unlock(&fa.flows_lock);
 
@@ -275,10 +336,6 @@ static void * fa_handle_packet(void * o)
 int fa_init(void)
 {
         pthread_condattr_t cattr;
-        int                i;
-
-        for (i = 0; i < PROG_MAX_FLOWS; ++i)
-                destroy_conn(i);
 
         if (pthread_rwlock_init(&fa.flows_lock, NULL))
                 goto fail_rwlock;
@@ -383,9 +440,10 @@ int fa_alloc(int             fd,
              size_t          dlen)
 {
         struct fa_msg *      msg;
-        uint64_t             addr;
         struct shm_du_buff * sdb;
-        qoscube_t            qc;
+        struct fa_flow *     flow;
+        uint64_t             addr;
+        qoscube_t            qc = QOS_CUBE_BE;
         size_t               len;
 
         addr = dir_query(dst);
@@ -397,7 +455,9 @@ int fa_alloc(int             fd,
         if (ipcp_sdb_reserve(&sdb, len + dlen))
                 return -1;
 
-        msg               = (struct fa_msg *) shm_du_buff_head(sdb);
+        msg = (struct fa_msg *) shm_du_buff_head(sdb);
+        memset(msg, 0, sizeof(*msg));
+
         msg->code         = FLOW_REQ;
         msg->s_eid        = hton32(fd);
         msg->s_addr       = hton64(ipcpi.dt_addr);
@@ -413,17 +473,17 @@ int fa_alloc(int             fd,
         memcpy(msg + 1, dst, ipcp_dir_hash_len());
         memcpy(shm_du_buff_head(sdb) + len, data, dlen);
 
-        qc = qos_spec_to_cube(qs);
-
         if (dt_write_packet(addr, qc, fa.fd, sdb)) {
                 ipcp_sdb_release(sdb);
                 return -1;
         }
 
+        flow = &fa.flows[fd];
+
         pthread_rwlock_wrlock(&fa.flows_lock);
 
-        assert(fa.r_eid[fd] == -1);
-        fa.r_addr[fd] = addr;
+        fa_flow_init(flow);
+        flow->r_addr = addr;
 
         pthread_rwlock_unlock(&fa.flows_lock);
 
@@ -439,10 +499,13 @@ int fa_alloc_resp(int          fd,
         struct timespec      abstime;
         struct fa_msg *      msg;
         struct shm_du_buff * sdb;
-        qoscube_t            qc;
+        struct fa_flow *     flow;
+        qoscube_t            qc = QOS_CUBE_BE;
 
         clock_gettime(PTHREAD_COND_CLOCK, &abstime);
 
+        flow = &fa.flows[fd];
+
         pthread_mutex_lock(&ipcpi.alloc_lock);
 
         while (ipcpi.alloc_id != fd && ipcp_get_state() == IPCP_OPERATIONAL) {
@@ -463,33 +526,31 @@ int fa_alloc_resp(int          fd,
         pthread_mutex_unlock(&ipcpi.alloc_lock);
 
         if (ipcp_sdb_reserve(&sdb, sizeof(*msg) + len)) {
-                destroy_conn(fd);
+                fa_flow_fini(flow);
                 return -1;
         }
 
+        msg = (struct fa_msg *) shm_du_buff_head(sdb);
+        memset(msg, 0, sizeof(*msg));
+
         pthread_rwlock_wrlock(&fa.flows_lock);
 
-        msg           = (struct fa_msg *) shm_du_buff_head(sdb);
         msg->code     = FLOW_REPLY;
-        msg->r_eid    = hton32(fa.r_eid[fd]);
+        msg->r_eid    = hton32(flow->r_eid);
         msg->s_eid    = hton32(fd);
         msg->response = response;
 
         memcpy(msg + 1, data, len);
 
         if (response < 0) {
-                destroy_conn(fd);
+                fa_flow_fini(flow);
                 ipcp_sdb_release(sdb);
         } else {
                 psched_add(fa.psched, fd);
         }
 
-        ipcp_flow_get_qoscube(fd, &qc);
-
-        assert(qc >= 0 && qc < QOS_CUBE_MAX);
-
-        if (dt_write_packet(fa.r_addr[fd], qc, fa.fd, sdb)) {
-                destroy_conn(fd);
+        if (dt_write_packet(flow->r_addr, qc, fa.fd, sdb)) {
+                fa_flow_fini(flow);
                 pthread_rwlock_unlock(&fa.flows_lock);
                 ipcp_sdb_release(sdb);
                 return -1;
@@ -505,11 +566,11 @@ int fa_dealloc(int fd)
         if (ipcp_flow_fini(fd) < 0)
                 return 0;
 
-        pthread_rwlock_wrlock(&fa.flows_lock);
-
         psched_del(fa.psched, fd);
 
-        destroy_conn(fd);
+        pthread_rwlock_wrlock(&fa.flows_lock);
+
+        fa_flow_fini(&fa.flows[fd]);
 
         pthread_rwlock_unlock(&fa.flows_lock);
 
@@ -517,3 +578,60 @@ int fa_dealloc(int fd)
 
         return 0;
 }
+
+static int fa_update_remote(int      fd,
+                            uint16_t ece)
+{
+        struct fa_msg *      msg;
+        struct shm_du_buff * sdb;
+        qoscube_t            qc = QOS_CUBE_BE;
+        struct fa_flow *     flow;
+
+        if (ipcp_sdb_reserve(&sdb, sizeof(*msg))) {
+                return -1;
+        }
+
+        msg = (struct fa_msg *) shm_du_buff_head(sdb);
+
+        memset(msg, 0, sizeof(*msg));
+
+        flow = &fa.flows[fd];
+
+        pthread_rwlock_rdlock(&fa.flows_lock);
+
+        msg->code  = FLOW_UPDATE;
+        msg->r_eid = hton32(flow->r_eid);
+        msg->ece   = hton16(ece);
+
+        if (dt_write_packet(flow->r_addr, qc, fa.fd, sdb)) {
+                pthread_rwlock_unlock(&fa.flows_lock);
+                ipcp_sdb_release(sdb);
+                return -1;
+        }
+
+        pthread_rwlock_unlock(&fa.flows_lock);
+
+
+        return 0;
+}
+
+void  fa_ecn_update(int     eid,
+                    uint8_t ecn,
+                    size_t  len)
+{
+        struct fa_flow * flow;
+        bool             update;
+        uint16_t         ece;
+
+        flow = &fa.flows[eid];
+
+        pthread_rwlock_wrlock(&fa.flows_lock);
+
+        update = ca_ctx_update_rcv(flow->ctx, len, ecn, &ece);
+
+        pthread_rwlock_unlock(&fa.flows_lock);
+
+        if (update)
+                fa_update_remote(eid, ece);
+
+}