summaryrefslogtreecommitdiffstats
path: root/drivers/s390/net/ctcm_fsms.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/s390/net/ctcm_fsms.c')
-rw-r--r--drivers/s390/net/ctcm_fsms.c2347
1 files changed, 2347 insertions, 0 deletions
diff --git a/drivers/s390/net/ctcm_fsms.c b/drivers/s390/net/ctcm_fsms.c
new file mode 100644
index 00000000000..2a106f3a076
--- /dev/null
+++ b/drivers/s390/net/ctcm_fsms.c
@@ -0,0 +1,2347 @@
+/*
+ * drivers/s390/net/ctcm_fsms.c
+ *
+ * Copyright IBM Corp. 2001, 2007
+ * Authors: Fritz Elfert (felfert@millenux.com)
+ * Peter Tiedemann (ptiedem@de.ibm.com)
+ * MPC additions :
+ * Belinda Thompson (belindat@us.ibm.com)
+ * Andy Richter (richtera@us.ibm.com)
+ */
+
+#undef DEBUG
+#undef DEBUGDATA
+#undef DEBUGCCW
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/interrupt.h>
+#include <linux/timer.h>
+#include <linux/bitops.h>
+
+#include <linux/signal.h>
+#include <linux/string.h>
+
+#include <linux/ip.h>
+#include <linux/if_arp.h>
+#include <linux/tcp.h>
+#include <linux/skbuff.h>
+#include <linux/ctype.h>
+#include <net/dst.h>
+
+#include <linux/io.h>
+#include <asm/ccwdev.h>
+#include <asm/ccwgroup.h>
+#include <linux/uaccess.h>
+
+#include <asm/idals.h>
+
+#include "fsm.h"
+#include "cu3088.h"
+
+#include "ctcm_dbug.h"
+#include "ctcm_main.h"
+#include "ctcm_fsms.h"
+
+const char *dev_state_names[] = {
+ [DEV_STATE_STOPPED] = "Stopped",
+ [DEV_STATE_STARTWAIT_RXTX] = "StartWait RXTX",
+ [DEV_STATE_STARTWAIT_RX] = "StartWait RX",
+ [DEV_STATE_STARTWAIT_TX] = "StartWait TX",
+ [DEV_STATE_STOPWAIT_RXTX] = "StopWait RXTX",
+ [DEV_STATE_STOPWAIT_RX] = "StopWait RX",
+ [DEV_STATE_STOPWAIT_TX] = "StopWait TX",
+ [DEV_STATE_RUNNING] = "Running",
+};
+
+const char *dev_event_names[] = {
+ [DEV_EVENT_START] = "Start",
+ [DEV_EVENT_STOP] = "Stop",
+ [DEV_EVENT_RXUP] = "RX up",
+ [DEV_EVENT_TXUP] = "TX up",
+ [DEV_EVENT_RXDOWN] = "RX down",
+ [DEV_EVENT_TXDOWN] = "TX down",
+ [DEV_EVENT_RESTART] = "Restart",
+};
+
+const char *ctc_ch_event_names[] = {
+ [CTC_EVENT_IO_SUCCESS] = "ccw_device success",
+ [CTC_EVENT_IO_EBUSY] = "ccw_device busy",
+ [CTC_EVENT_IO_ENODEV] = "ccw_device enodev",
+ [CTC_EVENT_IO_UNKNOWN] = "ccw_device unknown",
+ [CTC_EVENT_ATTNBUSY] = "Status ATTN & BUSY",
+ [CTC_EVENT_ATTN] = "Status ATTN",
+ [CTC_EVENT_BUSY] = "Status BUSY",
+ [CTC_EVENT_UC_RCRESET] = "Unit check remote reset",
+ [CTC_EVENT_UC_RSRESET] = "Unit check remote system reset",
+ [CTC_EVENT_UC_TXTIMEOUT] = "Unit check TX timeout",
+ [CTC_EVENT_UC_TXPARITY] = "Unit check TX parity",
+ [CTC_EVENT_UC_HWFAIL] = "Unit check Hardware failure",
+ [CTC_EVENT_UC_RXPARITY] = "Unit check RX parity",
+ [CTC_EVENT_UC_ZERO] = "Unit check ZERO",
+ [CTC_EVENT_UC_UNKNOWN] = "Unit check Unknown",
+ [CTC_EVENT_SC_UNKNOWN] = "SubChannel check Unknown",
+ [CTC_EVENT_MC_FAIL] = "Machine check failure",
+ [CTC_EVENT_MC_GOOD] = "Machine check operational",
+ [CTC_EVENT_IRQ] = "IRQ normal",
+ [CTC_EVENT_FINSTAT] = "IRQ final",
+ [CTC_EVENT_TIMER] = "Timer",
+ [CTC_EVENT_START] = "Start",
+ [CTC_EVENT_STOP] = "Stop",
+ /*
+ * additional MPC events
+ */
+ [CTC_EVENT_SEND_XID] = "XID Exchange",
+ [CTC_EVENT_RSWEEP_TIMER] = "MPC Group Sweep Timer",
+};
+
+const char *ctc_ch_state_names[] = {
+ [CTC_STATE_IDLE] = "Idle",
+ [CTC_STATE_STOPPED] = "Stopped",
+ [CTC_STATE_STARTWAIT] = "StartWait",
+ [CTC_STATE_STARTRETRY] = "StartRetry",
+ [CTC_STATE_SETUPWAIT] = "SetupWait",
+ [CTC_STATE_RXINIT] = "RX init",
+ [CTC_STATE_TXINIT] = "TX init",
+ [CTC_STATE_RX] = "RX",
+ [CTC_STATE_TX] = "TX",
+ [CTC_STATE_RXIDLE] = "RX idle",
+ [CTC_STATE_TXIDLE] = "TX idle",
+ [CTC_STATE_RXERR] = "RX error",
+ [CTC_STATE_TXERR] = "TX error",
+ [CTC_STATE_TERM] = "Terminating",
+ [CTC_STATE_DTERM] = "Restarting",
+ [CTC_STATE_NOTOP] = "Not operational",
+ /*
+ * additional MPC states
+ */
+ [CH_XID0_PENDING] = "Pending XID0 Start",
+ [CH_XID0_INPROGRESS] = "In XID0 Negotiations ",
+ [CH_XID7_PENDING] = "Pending XID7 P1 Start",
+ [CH_XID7_PENDING1] = "Active XID7 P1 Exchange ",
+ [CH_XID7_PENDING2] = "Pending XID7 P2 Start ",
+ [CH_XID7_PENDING3] = "Active XID7 P2 Exchange ",
+ [CH_XID7_PENDING4] = "XID7 Complete - Pending READY ",
+};
+
+static void ctcm_action_nop(fsm_instance *fi, int event, void *arg);
+
+/*
+ * ----- static ctcm actions for channel statemachine -----
+ *
+*/
+static void chx_txdone(fsm_instance *fi, int event, void *arg);
+static void chx_rx(fsm_instance *fi, int event, void *arg);
+static void chx_rxidle(fsm_instance *fi, int event, void *arg);
+static void chx_firstio(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_setmode(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_start(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_haltio(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_stopped(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_stop(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_fail(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_setuperr(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_restart(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_rxiniterr(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_rxinitfail(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_rxdisc(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_txiniterr(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_txretry(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_iofatal(fsm_instance *fi, int event, void *arg);
+
+/*
+ * ----- static ctcmpc actions for ctcmpc channel statemachine -----
+ *
+*/
+static void ctcmpc_chx_txdone(fsm_instance *fi, int event, void *arg);
+static void ctcmpc_chx_rx(fsm_instance *fi, int event, void *arg);
+static void ctcmpc_chx_firstio(fsm_instance *fi, int event, void *arg);
+/* shared :
+static void ctcm_chx_setmode(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_start(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_haltio(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_stopped(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_stop(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_fail(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_setuperr(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_restart(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_rxiniterr(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_rxinitfail(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_rxdisc(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_txiniterr(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_txretry(fsm_instance *fi, int event, void *arg);
+static void ctcm_chx_iofatal(fsm_instance *fi, int event, void *arg);
+*/
+static void ctcmpc_chx_attn(fsm_instance *fsm, int event, void *arg);
+static void ctcmpc_chx_attnbusy(fsm_instance *, int, void *);
+static void ctcmpc_chx_resend(fsm_instance *, int, void *);
+static void ctcmpc_chx_send_sweep(fsm_instance *fsm, int event, void *arg);
+
+/**
+ * Check return code of a preceeding ccw_device call, halt_IO etc...
+ *
+ * ch : The channel, the error belongs to.
+ * Returns the error code (!= 0) to inspect.
+ */
+void ctcm_ccw_check_rc(struct channel *ch, int rc, char *msg)
+{
+ CTCM_DBF_TEXT_(ERROR, CTC_DBF_ERROR,
+ "ccw error %s (%s): %04x\n", ch->id, msg, rc);
+ switch (rc) {
+ case -EBUSY:
+ ctcm_pr_warn("%s (%s): Busy !\n", ch->id, msg);
+ fsm_event(ch->fsm, CTC_EVENT_IO_EBUSY, ch);
+ break;
+ case -ENODEV:
+ ctcm_pr_emerg("%s (%s): Invalid device called for IO\n",
+ ch->id, msg);
+ fsm_event(ch->fsm, CTC_EVENT_IO_ENODEV, ch);
+ break;
+ default:
+ ctcm_pr_emerg("%s (%s): Unknown error in do_IO %04x\n",
+ ch->id, msg, rc);
+ fsm_event(ch->fsm, CTC_EVENT_IO_UNKNOWN, ch);
+ }
+}
+
+void ctcm_purge_skb_queue(struct sk_buff_head *q)
+{
+ struct sk_buff *skb;
+
+ CTCM_DBF_TEXT(TRACE, 3, __FUNCTION__);
+
+ while ((skb = skb_dequeue(q))) {
+ atomic_dec(&skb->users);
+ dev_kfree_skb_any(skb);
+ }
+}
+
+/**
+ * NOP action for statemachines
+ */
+static void ctcm_action_nop(fsm_instance *fi, int event, void *arg)
+{
+}
+
+/*
+ * Actions for channel - statemachines.
+ */
+
+/**
+ * Normal data has been send. Free the corresponding
+ * skb (it's in io_queue), reset dev->tbusy and
+ * revert to idle state.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void chx_txdone(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct sk_buff *skb;
+ int first = 1;
+ int i;
+ unsigned long duration;
+ struct timespec done_stamp = current_kernel_time(); /* xtime */
+
+ duration =
+ (done_stamp.tv_sec - ch->prof.send_stamp.tv_sec) * 1000000 +
+ (done_stamp.tv_nsec - ch->prof.send_stamp.tv_nsec) / 1000;
+ if (duration > ch->prof.tx_time)
+ ch->prof.tx_time = duration;
+
+ if (ch->irb->scsw.count != 0)
+ ctcm_pr_debug("%s: TX not complete, remaining %d bytes\n",
+ dev->name, ch->irb->scsw.count);
+ fsm_deltimer(&ch->timer);
+ while ((skb = skb_dequeue(&ch->io_queue))) {
+ priv->stats.tx_packets++;
+ priv->stats.tx_bytes += skb->len - LL_HEADER_LENGTH;
+ if (first) {
+ priv->stats.tx_bytes += 2;
+ first = 0;
+ }
+ atomic_dec(&skb->users);
+ dev_kfree_skb_irq(skb);
+ }
+ spin_lock(&ch->collect_lock);
+ clear_normalized_cda(&ch->ccw[4]);
+ if (ch->collect_len > 0) {
+ int rc;
+
+ if (ctcm_checkalloc_buffer(ch)) {
+ spin_unlock(&ch->collect_lock);
+ return;
+ }
+ ch->trans_skb->data = ch->trans_skb_data;
+ skb_reset_tail_pointer(ch->trans_skb);
+ ch->trans_skb->len = 0;
+ if (ch->prof.maxmulti < (ch->collect_len + 2))
+ ch->prof.maxmulti = ch->collect_len + 2;
+ if (ch->prof.maxcqueue < skb_queue_len(&ch->collect_queue))
+ ch->prof.maxcqueue = skb_queue_len(&ch->collect_queue);
+ *((__u16 *)skb_put(ch->trans_skb, 2)) = ch->collect_len + 2;
+ i = 0;
+ while ((skb = skb_dequeue(&ch->collect_queue))) {
+ skb_copy_from_linear_data(skb,
+ skb_put(ch->trans_skb, skb->len), skb->len);
+ priv->stats.tx_packets++;
+ priv->stats.tx_bytes += skb->len - LL_HEADER_LENGTH;
+ atomic_dec(&skb->users);
+ dev_kfree_skb_irq(skb);
+ i++;
+ }
+ ch->collect_len = 0;
+ spin_unlock(&ch->collect_lock);
+ ch->ccw[1].count = ch->trans_skb->len;
+ fsm_addtimer(&ch->timer, CTCM_TIME_5_SEC, CTC_EVENT_TIMER, ch);
+ ch->prof.send_stamp = current_kernel_time(); /* xtime */
+ rc = ccw_device_start(ch->cdev, &ch->ccw[0],
+ (unsigned long)ch, 0xff, 0);
+ ch->prof.doios_multi++;
+ if (rc != 0) {
+ priv->stats.tx_dropped += i;
+ priv->stats.tx_errors += i;
+ fsm_deltimer(&ch->timer);
+ ctcm_ccw_check_rc(ch, rc, "chained TX");
+ }
+ } else {
+ spin_unlock(&ch->collect_lock);
+ fsm_newstate(fi, CTC_STATE_TXIDLE);
+ }
+ ctcm_clear_busy_do(dev);
+}
+
+/**
+ * Initial data is sent.
+ * Notify device statemachine that we are up and
+ * running.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+void ctcm_chx_txidle(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCM_DBF_TEXT(TRACE, 6, __FUNCTION__);
+ fsm_deltimer(&ch->timer);
+ fsm_newstate(fi, CTC_STATE_TXIDLE);
+ fsm_event(priv->fsm, DEV_EVENT_TXUP, ch->netdev);
+}
+
+/**
+ * Got normal data, check for sanity, queue it up, allocate new buffer
+ * trigger bottom half, and initiate next read.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void chx_rx(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ int len = ch->max_bufsize - ch->irb->scsw.count;
+ struct sk_buff *skb = ch->trans_skb;
+ __u16 block_len = *((__u16 *)skb->data);
+ int check_len;
+ int rc;
+
+ fsm_deltimer(&ch->timer);
+ if (len < 8) {
+ ctcm_pr_debug("%s: got packet with length %d < 8\n",
+ dev->name, len);
+ priv->stats.rx_dropped++;
+ priv->stats.rx_length_errors++;
+ goto again;
+ }
+ if (len > ch->max_bufsize) {
+ ctcm_pr_debug("%s: got packet with length %d > %d\n",
+ dev->name, len, ch->max_bufsize);
+ priv->stats.rx_dropped++;
+ priv->stats.rx_length_errors++;
+ goto again;
+ }
+
+ /*
+ * VM TCP seems to have a bug sending 2 trailing bytes of garbage.
+ */
+ switch (ch->protocol) {
+ case CTCM_PROTO_S390:
+ case CTCM_PROTO_OS390:
+ check_len = block_len + 2;
+ break;
+ default:
+ check_len = block_len;
+ break;
+ }
+ if ((len < block_len) || (len > check_len)) {
+ ctcm_pr_debug("%s: got block length %d != rx length %d\n",
+ dev->name, block_len, len);
+ if (do_debug)
+ ctcmpc_dump_skb(skb, 0);
+
+ *((__u16 *)skb->data) = len;
+ priv->stats.rx_dropped++;
+ priv->stats.rx_length_errors++;
+ goto again;
+ }
+ block_len -= 2;
+ if (block_len > 0) {
+ *((__u16 *)skb->data) = block_len;
+ ctcm_unpack_skb(ch, skb);
+ }
+ again:
+ skb->data = ch->trans_skb_data;
+ skb_reset_tail_pointer(skb);
+ skb->len = 0;
+ if (ctcm_checkalloc_buffer(ch))
+ return;
+ ch->ccw[1].count = ch->max_bufsize;
+ rc = ccw_device_start(ch->cdev, &ch->ccw[0],
+ (unsigned long)ch, 0xff, 0);
+ if (rc != 0)
+ ctcm_ccw_check_rc(ch, rc, "normal RX");
+}
+
+/**
+ * Initialize connection by sending a __u16 of value 0.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void chx_firstio(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ int rc;
+
+ CTCM_DBF_TEXT(TRACE, 6, __FUNCTION__);
+
+ if (fsm_getstate(fi) == CTC_STATE_TXIDLE)
+ ctcm_pr_debug("%s: remote side issued READ?, init.\n", ch->id);
+ fsm_deltimer(&ch->timer);
+ if (ctcm_checkalloc_buffer(ch))
+ return;
+ if ((fsm_getstate(fi) == CTC_STATE_SETUPWAIT) &&
+ (ch->protocol == CTCM_PROTO_OS390)) {
+ /* OS/390 resp. z/OS */
+ if (CHANNEL_DIRECTION(ch->flags) == READ) {
+ *((__u16 *)ch->trans_skb->data) = CTCM_INITIAL_BLOCKLEN;
+ fsm_addtimer(&ch->timer, CTCM_TIME_5_SEC,
+ CTC_EVENT_TIMER, ch);
+ chx_rxidle(fi, event, arg);
+ } else {
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ fsm_newstate(fi, CTC_STATE_TXIDLE);
+ fsm_event(priv->fsm, DEV_EVENT_TXUP, dev);
+ }
+ return;
+ }
+
+ /*
+ * Don't setup a timer for receiving the initial RX frame
+ * if in compatibility mode, since VM TCP delays the initial
+ * frame until it has some data to send.
+ */
+ if ((CHANNEL_DIRECTION(ch->flags) == WRITE) ||
+ (ch->protocol != CTCM_PROTO_S390))
+ fsm_addtimer(&ch->timer, CTCM_TIME_5_SEC, CTC_EVENT_TIMER, ch);
+
+ *((__u16 *)ch->trans_skb->data) = CTCM_INITIAL_BLOCKLEN;
+ ch->ccw[1].count = 2; /* Transfer only length */
+
+ fsm_newstate(fi, (CHANNEL_DIRECTION(ch->flags) == READ)
+ ? CTC_STATE_RXINIT : CTC_STATE_TXINIT);
+ rc = ccw_device_start(ch->cdev, &ch->ccw[0],
+ (unsigned long)ch, 0xff, 0);
+ if (rc != 0) {
+ fsm_deltimer(&ch->timer);
+ fsm_newstate(fi, CTC_STATE_SETUPWAIT);
+ ctcm_ccw_check_rc(ch, rc, "init IO");
+ }
+ /*
+ * If in compatibility mode since we don't setup a timer, we
+ * also signal RX channel up immediately. This enables us
+ * to send packets early which in turn usually triggers some
+ * reply from VM TCP which brings up the RX channel to it's
+ * final state.
+ */
+ if ((CHANNEL_DIRECTION(ch->flags) == READ) &&
+ (ch->protocol == CTCM_PROTO_S390)) {
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ fsm_event(priv->fsm, DEV_EVENT_RXUP, dev);
+ }
+}
+
+/**
+ * Got initial data, check it. If OK,
+ * notify device statemachine that we are up and
+ * running.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void chx_rxidle(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ __u16 buflen;
+ int rc;
+
+ CTCM_DBF_TEXT(TRACE, 6, __FUNCTION__);
+ fsm_deltimer(&ch->timer);
+ buflen = *((__u16 *)ch->trans_skb->data);
+ if (do_debug)
+ ctcm_pr_debug("%s: Initial RX count %d\n", dev->name, buflen);
+
+ if (buflen >= CTCM_INITIAL_BLOCKLEN) {
+ if (ctcm_checkalloc_buffer(ch))
+ return;
+ ch->ccw[1].count = ch->max_bufsize;
+ fsm_newstate(fi, CTC_STATE_RXIDLE);
+ rc = ccw_device_start(ch->cdev, &ch->ccw[0],
+ (unsigned long)ch, 0xff, 0);
+ if (rc != 0) {
+ fsm_newstate(fi, CTC_STATE_RXINIT);
+ ctcm_ccw_check_rc(ch, rc, "initial RX");
+ } else
+ fsm_event(priv->fsm, DEV_EVENT_RXUP, dev);
+ } else {
+ if (do_debug)
+ ctcm_pr_debug("%s: Initial RX count %d not %d\n",
+ dev->name, buflen, CTCM_INITIAL_BLOCKLEN);
+ chx_firstio(fi, event, arg);
+ }
+}
+
+/**
+ * Set channel into extended mode.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_setmode(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ int rc;
+ unsigned long saveflags = 0;
+ int timeout = CTCM_TIME_5_SEC;
+
+ fsm_deltimer(&ch->timer);
+ if (IS_MPC(ch)) {
+ timeout = 1500;
+ if (do_debug)
+ ctcm_pr_debug("ctcm enter: %s(): cp=%i ch=0x%p id=%s\n",
+ __FUNCTION__, smp_processor_id(), ch, ch->id);
+ }
+ fsm_addtimer(&ch->timer, timeout, CTC_EVENT_TIMER, ch);
+ fsm_newstate(fi, CTC_STATE_SETUPWAIT);
+ if (do_debug_ccw && IS_MPC(ch))
+ ctcmpc_dumpit((char *)&ch->ccw[6], sizeof(struct ccw1) * 2);
+
+ if (event == CTC_EVENT_TIMER) /* only for timer not yet locked */
+ spin_lock_irqsave(get_ccwdev_lock(ch->cdev), saveflags);
+ /* Such conditional locking is undeterministic in
+ * static view. => ignore sparse warnings here. */
+
+ rc = ccw_device_start(ch->cdev, &ch->ccw[6],
+ (unsigned long)ch, 0xff, 0);
+ if (event == CTC_EVENT_TIMER) /* see above comments */
+ spin_unlock_irqrestore(get_ccwdev_lock(ch->cdev), saveflags);
+ if (rc != 0) {
+ fsm_deltimer(&ch->timer);
+ fsm_newstate(fi, CTC_STATE_STARTWAIT);
+ ctcm_ccw_check_rc(ch, rc, "set Mode");
+ } else
+ ch->retry = 0;
+}
+
+/**
+ * Setup channel.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_start(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ int rc;
+ struct net_device *dev;
+ unsigned long saveflags;
+
+ CTCM_DBF_TEXT(TRACE, 5, __FUNCTION__);
+ if (ch == NULL) {
+ ctcm_pr_warn("chx_start ch=NULL\n");
+ return;
+ }
+ if (ch->netdev == NULL) {
+ ctcm_pr_warn("chx_start dev=NULL, id=%s\n", ch->id);
+ return;
+ }
+ dev = ch->netdev;
+
+ if (do_debug)
+ ctcm_pr_debug("%s: %s channel start\n", dev->name,
+ (CHANNEL_DIRECTION(ch->flags) == READ) ? "RX" : "TX");
+
+ if (ch->trans_skb != NULL) {
+ clear_normalized_cda(&ch->ccw[1]);
+ dev_kfree_skb(ch->trans_skb);
+ ch->trans_skb = NULL;
+ }
+ if (CHANNEL_DIRECTION(ch->flags) == READ) {
+ ch->ccw[1].cmd_code = CCW_CMD_READ;
+ ch->ccw[1].flags = CCW_FLAG_SLI;
+ ch->ccw[1].count = 0;
+ } else {
+ ch->ccw[1].cmd_code = CCW_CMD_WRITE;
+ ch->ccw[1].flags = CCW_FLAG_SLI | CCW_FLAG_CC;
+ ch->ccw[1].count = 0;
+ }
+ if (ctcm_checkalloc_buffer(ch)) {
+ ctcm_pr_notice("%s: %s trans_skb allocation delayed "
+ "until first transfer\n", dev->name,
+ (CHANNEL_DIRECTION(ch->flags) == READ) ? "RX" : "TX");
+ }
+
+ ch->ccw[0].cmd_code = CCW_CMD_PREPARE;
+ ch->ccw[0].flags = CCW_FLAG_SLI | CCW_FLAG_CC;
+ ch->ccw[0].count = 0;
+ ch->ccw[0].cda = 0;
+ ch->ccw[2].cmd_code = CCW_CMD_NOOP; /* jointed CE + DE */
+ ch->ccw[2].flags = CCW_FLAG_SLI;
+ ch->ccw[2].count = 0;
+ ch->ccw[2].cda = 0;
+ memcpy(&ch->ccw[3], &ch->ccw[0], sizeof(struct ccw1) * 3);
+ ch->ccw[4].cda = 0;
+ ch->ccw[4].flags &= ~CCW_FLAG_IDA;
+
+ fsm_newstate(fi, CTC_STATE_STARTWAIT);
+ fsm_addtimer(&ch->timer, 1000, CTC_EVENT_TIMER, ch);
+ spin_lock_irqsave(get_ccwdev_lock(ch->cdev), saveflags);
+ rc = ccw_device_halt(ch->cdev, (unsigned long)ch);
+ spin_unlock_irqrestore(get_ccwdev_lock(ch->cdev), saveflags);
+ if (rc != 0) {
+ if (rc != -EBUSY)
+ fsm_deltimer(&ch->timer);
+ ctcm_ccw_check_rc(ch, rc, "initial HaltIO");
+ }
+}
+
+/**
+ * Shutdown a channel.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_haltio(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ unsigned long saveflags = 0;
+ int rc;
+ int oldstate;
+
+ CTCM_DBF_TEXT(TRACE, 2, __FUNCTION__);
+ fsm_deltimer(&ch->timer);
+ if (IS_MPC(ch))
+ fsm_deltimer(&ch->sweep_timer);
+
+ fsm_addtimer(&ch->timer, CTCM_TIME_5_SEC, CTC_EVENT_TIMER, ch);
+
+ if (event == CTC_EVENT_STOP) /* only for STOP not yet locked */
+ spin_lock_irqsave(get_ccwdev_lock(ch->cdev), saveflags);
+ /* Such conditional locking is undeterministic in
+ * static view. => ignore sparse warnings here. */
+ oldstate = fsm_getstate(fi);
+ fsm_newstate(fi, CTC_STATE_TERM);
+ rc = ccw_device_halt(ch->cdev, (unsigned long)ch);
+
+ if (event == CTC_EVENT_STOP)
+ spin_unlock_irqrestore(get_ccwdev_lock(ch->cdev), saveflags);
+ /* see remark above about conditional locking */
+
+ if (rc != 0 && rc != -EBUSY) {
+ fsm_deltimer(&ch->timer);
+ if (event != CTC_EVENT_STOP) {
+ fsm_newstate(fi, oldstate);
+ ctcm_ccw_check_rc(ch, rc, (char *)__FUNCTION__);
+ }
+ }
+}
+
+/**
+ * Cleanup helper for chx_fail and chx_stopped
+ * cleanup channels queue and notify interface statemachine.
+ *
+ * fi An instance of a channel statemachine.
+ * state The next state (depending on caller).
+ * ch The channel to operate on.
+ */
+static void ctcm_chx_cleanup(fsm_instance *fi, int state,
+ struct channel *ch)
+{
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCM_DBF_TEXT(TRACE, 3, __FUNCTION__);
+
+ fsm_deltimer(&ch->timer);
+ if (IS_MPC(ch))
+ fsm_deltimer(&ch->sweep_timer);
+
+ fsm_newstate(fi, state);
+ if (state == CTC_STATE_STOPPED && ch->trans_skb != NULL) {
+ clear_normalized_cda(&ch->ccw[1]);
+ dev_kfree_skb_any(ch->trans_skb);
+ ch->trans_skb = NULL;
+ }
+
+ ch->th_seg = 0x00;
+ ch->th_seq_num = 0x00;
+ if (CHANNEL_DIRECTION(ch->flags) == READ) {
+ skb_queue_purge(&ch->io_queue);
+ fsm_event(priv->fsm, DEV_EVENT_RXDOWN, dev);
+ } else {
+ ctcm_purge_skb_queue(&ch->io_queue);
+ if (IS_MPC(ch))
+ ctcm_purge_skb_queue(&ch->sweep_queue);
+ spin_lock(&ch->collect_lock);
+ ctcm_purge_skb_queue(&ch->collect_queue);
+ ch->collect_len = 0;
+ spin_unlock(&ch->collect_lock);
+ fsm_event(priv->fsm, DEV_EVENT_TXDOWN, dev);
+ }
+}
+
+/**
+ * A channel has successfully been halted.
+ * Cleanup it's queue and notify interface statemachine.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_stopped(fsm_instance *fi, int event, void *arg)
+{
+ CTCM_DBF_TEXT(TRACE, 3, __FUNCTION__);
+ ctcm_chx_cleanup(fi, CTC_STATE_STOPPED, arg);
+}
+
+/**
+ * A stop command from device statemachine arrived and we are in
+ * not operational mode. Set state to stopped.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_stop(fsm_instance *fi, int event, void *arg)
+{
+ fsm_newstate(fi, CTC_STATE_STOPPED);
+}
+
+/**
+ * A machine check for no path, not operational status or gone device has
+ * happened.
+ * Cleanup queue and notify interface statemachine.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_fail(fsm_instance *fi, int event, void *arg)
+{
+ CTCM_DBF_TEXT(TRACE, 3, __FUNCTION__);
+ ctcm_chx_cleanup(fi, CTC_STATE_NOTOP, arg);
+}
+
+/**
+ * Handle error during setup of channel.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_setuperr(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ /*
+ * Special case: Got UC_RCRESET on setmode.
+ * This means that remote side isn't setup. In this case
+ * simply retry after some 10 secs...
+ */
+ if ((fsm_getstate(fi) == CTC_STATE_SETUPWAIT) &&
+ ((event == CTC_EVENT_UC_RCRESET) ||
+ (event == CTC_EVENT_UC_RSRESET))) {
+ fsm_newstate(fi, CTC_STATE_STARTRETRY);
+ fsm_deltimer(&ch->timer);
+ fsm_addtimer(&ch->timer, CTCM_TIME_5_SEC, CTC_EVENT_TIMER, ch);
+ if (!IS_MPC(ch) && (CHANNEL_DIRECTION(ch->flags) == READ)) {
+ int rc = ccw_device_halt(ch->cdev, (unsigned long)ch);
+ if (rc != 0)
+ ctcm_ccw_check_rc(ch, rc,
+ "HaltIO in chx_setuperr");
+ }
+ return;
+ }
+
+ CTCM_DBF_TEXT_(ERROR, CTC_DBF_CRIT,
+ "%s : %s error during %s channel setup state=%s\n",
+ dev->name, ctc_ch_event_names[event],
+ (CHANNEL_DIRECTION(ch->flags) == READ) ? "RX" : "TX",
+ fsm_getstate_str(fi));
+
+ if (CHANNEL_DIRECTION(ch->flags) == READ) {
+ fsm_newstate(fi, CTC_STATE_RXERR);
+ fsm_event(priv->fsm, DEV_EVENT_RXDOWN, dev);
+ } else {
+ fsm_newstate(fi, CTC_STATE_TXERR);
+ fsm_event(priv->fsm, DEV_EVENT_TXDOWN, dev);
+ }
+}
+
+/**
+ * Restart a channel after an error.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_restart(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ unsigned long saveflags = 0;
+ int oldstate;
+ int rc;
+
+ CTCM_DBF_TEXT(TRACE, CTC_DBF_NOTICE, __FUNCTION__);
+ fsm_deltimer(&ch->timer);
+ ctcm_pr_debug("%s: %s channel restart\n", dev->name,
+ (CHANNEL_DIRECTION(ch->flags) == READ) ? "RX" : "TX");
+ fsm_addtimer(&ch->timer, CTCM_TIME_5_SEC, CTC_EVENT_TIMER, ch);
+ oldstate = fsm_getstate(fi);
+ fsm_newstate(fi, CTC_STATE_STARTWAIT);
+ if (event == CTC_EVENT_TIMER) /* only for timer not yet locked */
+ spin_lock_irqsave(get_ccwdev_lock(ch->cdev), saveflags);
+ /* Such conditional locking is a known problem for
+ * sparse because its undeterministic in static view.
+ * Warnings should be ignored here. */
+ rc = ccw_device_halt(ch->cdev, (unsigned long)ch);
+ if (event == CTC_EVENT_TIMER)
+ spin_unlock_irqrestore(get_ccwdev_lock(ch->cdev), saveflags);
+ if (rc != 0) {
+ if (rc != -EBUSY) {
+ fsm_deltimer(&ch->timer);
+ fsm_newstate(fi, oldstate);
+ }
+ ctcm_ccw_check_rc(ch, rc, "HaltIO in ctcm_chx_restart");
+ }
+}
+
+/**
+ * Handle error during RX initial handshake (exchange of
+ * 0-length block header)
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_rxiniterr(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCM_DBF_TEXT(SETUP, 3, __FUNCTION__);
+ if (event == CTC_EVENT_TIMER) {
+ if (!IS_MPCDEV(dev))
+ /* TODO : check if MPC deletes timer somewhere */
+ fsm_deltimer(&ch->timer);
+ ctcm_pr_debug("%s: Timeout during RX init handshake\n",
+ dev->name);
+ if (ch->retry++ < 3)
+ ctcm_chx_restart(fi, event, arg);
+ else {
+ fsm_newstate(fi, CTC_STATE_RXERR);
+ fsm_event(priv->fsm, DEV_EVENT_RXDOWN, dev);
+ }
+ } else
+ ctcm_pr_warn("%s: Error during RX init handshake\n", dev->name);
+}
+
+/**
+ * Notify device statemachine if we gave up initialization
+ * of RX channel.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_rxinitfail(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCM_DBF_TEXT(SETUP, 3, __FUNCTION__);
+ fsm_newstate(fi, CTC_STATE_RXERR);
+ ctcm_pr_warn("%s: RX busy. Initialization failed\n", dev->name);
+ fsm_event(priv->fsm, DEV_EVENT_RXDOWN, dev);
+}
+
+/**
+ * Handle RX Unit check remote reset (remote disconnected)
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_rxdisc(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct channel *ch2;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCM_DBF_DEV_NAME(TRACE, dev, "Got remote disconnect, re-initializing");
+ fsm_deltimer(&ch->timer);
+ if (do_debug)
+ ctcm_pr_debug("%s: Got remote disconnect, "
+ "re-initializing ...\n", dev->name);
+ /*
+ * Notify device statemachine
+ */
+ fsm_event(priv->fsm, DEV_EVENT_RXDOWN, dev);
+ fsm_event(priv->fsm, DEV_EVENT_TXDOWN, dev);
+
+ fsm_newstate(fi, CTC_STATE_DTERM);
+ ch2 = priv->channel[WRITE];
+ fsm_newstate(ch2->fsm, CTC_STATE_DTERM);
+
+ ccw_device_halt(ch->cdev, (unsigned long)ch);
+ ccw_device_halt(ch2->cdev, (unsigned long)ch2);
+}
+
+/**
+ * Handle error during TX channel initialization.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_txiniterr(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ if (event == CTC_EVENT_TIMER) {
+ fsm_deltimer(&ch->timer);
+ CTCM_DBF_DEV_NAME(ERROR, dev,
+ "Timeout during TX init handshake");
+ if (ch->retry++ < 3)
+ ctcm_chx_restart(fi, event, arg);
+ else {
+ fsm_newstate(fi, CTC_STATE_TXERR);
+ fsm_event(priv->fsm, DEV_EVENT_TXDOWN, dev);
+ }
+ } else {
+ CTCM_DBF_TEXT_(ERROR, CTC_DBF_ERROR,
+ "%s : %s error during channel setup state=%s",
+ dev->name, ctc_ch_event_names[event],
+ fsm_getstate_str(fi));
+
+ ctcm_pr_warn("%s: Error during TX init handshake\n", dev->name);
+ }
+}
+
+/**
+ * Handle TX timeout by retrying operation.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_txretry(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct sk_buff *skb;
+
+ if (do_debug)
+ ctcm_pr_debug("ctcmpc enter: %s(): cp=%i ch=0x%p id=%s\n",
+ __FUNCTION__, smp_processor_id(), ch, ch->id);
+
+ fsm_deltimer(&ch->timer);
+ if (ch->retry++ > 3) {
+ struct mpc_group *gptr = priv->mpcg;
+ ctcm_pr_debug("%s: TX retry failed, restarting channel\n",
+ dev->name);
+ fsm_event(priv->fsm, DEV_EVENT_TXDOWN, dev);
+ /* call restart if not MPC or if MPC and mpcg fsm is ready.
+ use gptr as mpc indicator */
+ if (!(gptr && (fsm_getstate(gptr->fsm) != MPCG_STATE_READY)))
+ ctcm_chx_restart(fi, event, arg);
+ goto done;
+ }
+
+ ctcm_pr_debug("%s: TX retry %d\n", dev->name, ch->retry);
+ skb = skb_peek(&ch->io_queue);
+ if (skb) {
+ int rc = 0;
+ unsigned long saveflags = 0;
+ clear_normalized_cda(&ch->ccw[4]);
+ ch->ccw[4].count = skb->len;
+ if (set_normalized_cda(&ch->ccw[4], skb->data)) {
+ ctcm_pr_debug("%s: IDAL alloc failed, chan restart\n",
+ dev->name);
+ fsm_event(priv->fsm, DEV_EVENT_TXDOWN, dev);
+ ctcm_chx_restart(fi, event, arg);
+ goto done;
+ }
+ fsm_addtimer(&ch->timer, 1000, CTC_EVENT_TIMER, ch);
+ if (event == CTC_EVENT_TIMER) /* for TIMER not yet locked */
+ spin_lock_irqsave(get_ccwdev_lock(ch->cdev), saveflags);
+ /* Such conditional locking is a known problem for
+ * sparse because its undeterministic in static view.
+ * Warnings should be ignored here. */
+ if (do_debug_ccw)
+ ctcmpc_dumpit((char *)&ch->ccw[3],
+ sizeof(struct ccw1) * 3);
+
+ rc = ccw_device_start(ch->cdev, &ch->ccw[3],
+ (unsigned long)ch, 0xff, 0);
+ if (event == CTC_EVENT_TIMER)
+ spin_unlock_irqrestore(get_ccwdev_lock(ch->cdev),
+ saveflags);
+ if (rc != 0) {
+ fsm_deltimer(&ch->timer);
+ ctcm_ccw_check_rc(ch, rc, "TX in chx_txretry");
+ ctcm_purge_skb_queue(&ch->io_queue);
+ }
+ }
+done:
+ return;
+}
+
+/**
+ * Handle fatal errors during an I/O command.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcm_chx_iofatal(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCM_DBF_TEXT(TRACE, 3, __FUNCTION__);
+ fsm_deltimer(&ch->timer);
+ ctcm_pr_warn("%s %s : unrecoverable channel error\n",
+ CTC_DRIVER_NAME, dev->name);
+ if (IS_MPC(ch)) {
+ priv->stats.tx_dropped++;
+ priv->stats.tx_errors++;
+ }
+
+ if (CHANNEL_DIRECTION(ch->flags) == READ) {
+ ctcm_pr_debug("%s: RX I/O error\n", dev->name);
+ fsm_newstate(fi, CTC_STATE_RXERR);
+ fsm_event(priv->fsm, DEV_EVENT_RXDOWN, dev);
+ } else {
+ ctcm_pr_debug("%s: TX I/O error\n", dev->name);
+ fsm_newstate(fi, CTC_STATE_TXERR);
+ fsm_event(priv->fsm, DEV_EVENT_TXDOWN, dev);
+ }
+}
+
+/*
+ * The ctcm statemachine for a channel.
+ */
+const fsm_node ch_fsm[] = {
+ { CTC_STATE_STOPPED, CTC_EVENT_STOP, ctcm_action_nop },
+ { CTC_STATE_STOPPED, CTC_EVENT_START, ctcm_chx_start },
+ { CTC_STATE_STOPPED, CTC_EVENT_FINSTAT, ctcm_action_nop },
+ { CTC_STATE_STOPPED, CTC_EVENT_MC_FAIL, ctcm_action_nop },
+
+ { CTC_STATE_NOTOP, CTC_EVENT_STOP, ctcm_chx_stop },
+ { CTC_STATE_NOTOP, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_NOTOP, CTC_EVENT_FINSTAT, ctcm_action_nop },
+ { CTC_STATE_NOTOP, CTC_EVENT_MC_FAIL, ctcm_action_nop },
+ { CTC_STATE_NOTOP, CTC_EVENT_MC_GOOD, ctcm_chx_start },
+
+ { CTC_STATE_STARTWAIT, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_FINSTAT, ctcm_chx_setmode },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_TIMER, ctcm_chx_setuperr },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_STARTRETRY, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_STARTRETRY, CTC_EVENT_TIMER, ctcm_chx_setmode },
+ { CTC_STATE_STARTRETRY, CTC_EVENT_FINSTAT, ctcm_action_nop },
+ { CTC_STATE_STARTRETRY, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_FINSTAT, chx_firstio },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_TIMER, ctcm_chx_setmode },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_RXINIT, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_RXINIT, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_RXINIT, CTC_EVENT_FINSTAT, chx_rxidle },
+ { CTC_STATE_RXINIT, CTC_EVENT_UC_RCRESET, ctcm_chx_rxiniterr },
+ { CTC_STATE_RXINIT, CTC_EVENT_UC_RSRESET, ctcm_chx_rxiniterr },
+ { CTC_STATE_RXINIT, CTC_EVENT_TIMER, ctcm_chx_rxiniterr },
+ { CTC_STATE_RXINIT, CTC_EVENT_ATTNBUSY, ctcm_chx_rxinitfail },
+ { CTC_STATE_RXINIT, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_RXINIT, CTC_EVENT_UC_ZERO, chx_firstio },
+ { CTC_STATE_RXINIT, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_RXIDLE, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_RXIDLE, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_RXIDLE, CTC_EVENT_FINSTAT, chx_rx },
+ { CTC_STATE_RXIDLE, CTC_EVENT_UC_RCRESET, ctcm_chx_rxdisc },
+ { CTC_STATE_RXIDLE, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_RXIDLE, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_RXIDLE, CTC_EVENT_UC_ZERO, chx_rx },
+
+ { CTC_STATE_TXINIT, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TXINIT, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_TXINIT, CTC_EVENT_FINSTAT, ctcm_chx_txidle },
+ { CTC_STATE_TXINIT, CTC_EVENT_UC_RCRESET, ctcm_chx_txiniterr },
+ { CTC_STATE_TXINIT, CTC_EVENT_UC_RSRESET, ctcm_chx_txiniterr },
+ { CTC_STATE_TXINIT, CTC_EVENT_TIMER, ctcm_chx_txiniterr },
+ { CTC_STATE_TXINIT, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_TXINIT, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_TXIDLE, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TXIDLE, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_TXIDLE, CTC_EVENT_FINSTAT, chx_firstio },
+ { CTC_STATE_TXIDLE, CTC_EVENT_UC_RCRESET, ctcm_action_nop },
+ { CTC_STATE_TXIDLE, CTC_EVENT_UC_RSRESET, ctcm_action_nop },
+ { CTC_STATE_TXIDLE, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_TXIDLE, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_TERM, CTC_EVENT_STOP, ctcm_action_nop },
+ { CTC_STATE_TERM, CTC_EVENT_START, ctcm_chx_restart },
+ { CTC_STATE_TERM, CTC_EVENT_FINSTAT, ctcm_chx_stopped },
+ { CTC_STATE_TERM, CTC_EVENT_UC_RCRESET, ctcm_action_nop },
+ { CTC_STATE_TERM, CTC_EVENT_UC_RSRESET, ctcm_action_nop },
+ { CTC_STATE_TERM, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_DTERM, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_DTERM, CTC_EVENT_START, ctcm_chx_restart },
+ { CTC_STATE_DTERM, CTC_EVENT_FINSTAT, ctcm_chx_setmode },
+ { CTC_STATE_DTERM, CTC_EVENT_UC_RCRESET, ctcm_action_nop },
+ { CTC_STATE_DTERM, CTC_EVENT_UC_RSRESET, ctcm_action_nop },
+ { CTC_STATE_DTERM, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_TX, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TX, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_TX, CTC_EVENT_FINSTAT, chx_txdone },
+ { CTC_STATE_TX, CTC_EVENT_UC_RCRESET, ctcm_chx_txretry },
+ { CTC_STATE_TX, CTC_EVENT_UC_RSRESET, ctcm_chx_txretry },
+ { CTC_STATE_TX, CTC_EVENT_TIMER, ctcm_chx_txretry },
+ { CTC_STATE_TX, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_TX, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_RXERR, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TXERR, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TXERR, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_RXERR, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+};
+
+int ch_fsm_len = ARRAY_SIZE(ch_fsm);
+
+/*
+ * MPC actions for mpc channel statemachine
+ * handling of MPC protocol requires extra
+ * statemachine and actions which are prefixed ctcmpc_ .
+ * The ctc_ch_states and ctc_ch_state_names,
+ * ctc_ch_events and ctc_ch_event_names share the ctcm definitions
+ * which are expanded by some elements.
+ */
+
+/*
+ * Actions for mpc channel statemachine.
+ */
+
+/**
+ * Normal data has been send. Free the corresponding
+ * skb (it's in io_queue), reset dev->tbusy and
+ * revert to idle state.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcmpc_chx_txdone(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct mpc_group *grp = priv->mpcg;
+ struct sk_buff *skb;
+ int first = 1;
+ int i;
+ struct timespec done_stamp;
+ __u32 data_space;
+ unsigned long duration;
+ struct sk_buff *peekskb;
+ int rc;
+ struct th_header *header;
+ struct pdu *p_header;
+
+ if (do_debug)
+ ctcm_pr_debug("%s cp:%i enter: %s()\n",
+ dev->name, smp_processor_id(), __FUNCTION__);
+
+ done_stamp = current_kernel_time(); /* xtime */
+ duration = (done_stamp.tv_sec - ch->prof.send_stamp.tv_sec) * 1000000
+ + (done_stamp.tv_nsec - ch->prof.send_stamp.tv_nsec) / 1000;
+ if (duration > ch->prof.tx_time)
+ ch->prof.tx_time = duration;
+
+ if (ch->irb->scsw.count != 0)
+ ctcm_pr_debug("%s: TX not complete, remaining %d bytes\n",
+ dev->name, ch->irb->scsw.count);
+ fsm_deltimer(&ch->timer);
+ while ((skb = skb_dequeue(&ch->io_queue))) {
+ priv->stats.tx_packets++;
+ priv->stats.tx_bytes += skb->len - TH_HEADER_LENGTH;
+ if (first) {
+ priv->stats.tx_bytes += 2;
+ first = 0;
+ }
+ atomic_dec(&skb->users);
+ dev_kfree_skb_irq(skb);
+ }
+ spin_lock(&ch->collect_lock);
+ clear_normalized_cda(&ch->ccw[4]);
+
+ if ((ch->collect_len <= 0) || (grp->in_sweep != 0)) {
+ spin_unlock(&ch->collect_lock);
+ fsm_newstate(fi, CTC_STATE_TXIDLE);
+ goto done;
+ }
+
+ if (ctcm_checkalloc_buffer(ch)) {
+ spin_unlock(&ch->collect_lock);
+ goto done;
+ }
+ ch->trans_skb->data = ch->trans_skb_data;
+ skb_reset_tail_pointer(ch->trans_skb);
+ ch->trans_skb->len = 0;
+ if (ch->prof.maxmulti < (ch->collect_len + TH_HEADER_LENGTH))
+ ch->prof.maxmulti = ch->collect_len + TH_HEADER_LENGTH;
+ if (ch->prof.maxcqueue < skb_queue_len(&ch->collect_queue))
+ ch->prof.maxcqueue = skb_queue_len(&ch->collect_queue);
+ i = 0;
+
+ if (do_debug_data)
+ ctcm_pr_debug("ctcmpc: %s() building "
+ "trans_skb from collect_q \n", __FUNCTION__);
+
+ data_space = grp->group_max_buflen - TH_HEADER_LENGTH;
+
+ if (do_debug_data)
+ ctcm_pr_debug("ctcmpc: %s() building trans_skb from collect_q"
+ " data_space:%04x\n", __FUNCTION__, data_space);
+ p_header = NULL;
+ while ((skb = skb_dequeue(&ch->collect_queue))) {
+ memcpy(skb_put(ch->trans_skb, skb->len), skb->data, skb->len);
+ p_header = (struct pdu *)
+ (skb_tail_pointer(ch->trans_skb) - skb->len);
+ p_header->pdu_flag = 0x00;
+ if (skb->protocol == ntohs(ETH_P_SNAP))
+ p_header->pdu_flag |= 0x60;
+ else
+ p_header->pdu_flag |= 0x20;
+
+ if (do_debug_data) {
+ ctcm_pr_debug("ctcmpc: %s()trans_skb len:%04x \n",
+ __FUNCTION__, ch->trans_skb->len);
+ ctcm_pr_debug("ctcmpc: %s() pdu header and data"
+ " for up to 32 bytes sent to vtam\n",
+ __FUNCTION__);
+ ctcmpc_dumpit((char *)p_header,
+ min_t(int, skb->len, 32));
+ }
+ ch->collect_len -= skb->len;
+ data_space -= skb->len;
+ priv->stats.tx_packets++;
+ priv->stats.tx_bytes += skb->len;
+ atomic_dec(&skb->users);
+ dev_kfree_skb_any(skb);
+ peekskb = skb_peek(&ch->collect_queue);
+ if (peekskb->len > data_space)
+ break;
+ i++;
+ }
+ /* p_header points to the last one we handled */
+ if (p_header)
+ p_header->pdu_flag |= PDU_LAST; /*Say it's the last one*/
+ header = kzalloc(TH_HEADER_LENGTH, gfp_type());
+
+ if (!header) {
+ printk(KERN_WARNING "ctcmpc: OUT OF MEMORY IN %s()"
+ ": Data Lost \n", __FUNCTION__);
+ spin_unlock(&ch->collect_lock);
+ fsm_event(priv->mpcg->fsm, MPCG_EVENT_INOP, dev);
+ goto done;
+ }
+
+ header->th_ch_flag = TH_HAS_PDU; /* Normal data */
+ ch->th_seq_num++;
+ header->th_seq_num = ch->th_seq_num;
+
+ if (do_debug_data)
+ ctcm_pr_debug("%s: ToVTAM_th_seq= %08x\n" ,
+ __FUNCTION__, ch->th_seq_num);
+
+ memcpy(skb_push(ch->trans_skb, TH_HEADER_LENGTH), header,
+ TH_HEADER_LENGTH); /* put the TH on the packet */
+
+ kfree(header);
+
+ if (do_debug_data) {
+ ctcm_pr_debug("ctcmpc: %s()trans_skb len:%04x \n",
+ __FUNCTION__, ch->trans_skb->len);
+
+ ctcm_pr_debug("ctcmpc: %s() up-to-50 bytes of trans_skb "
+ "data to vtam from collect_q\n", __FUNCTION__);
+ ctcmpc_dumpit((char *)ch->trans_skb->data,
+ min_t(int, ch->trans_skb->len, 50));
+ }
+
+ spin_unlock(&ch->collect_lock);
+ clear_normalized_cda(&ch->ccw[1]);
+ if (set_normalized_cda(&ch->ccw[1], ch->trans_skb->data)) {
+ dev_kfree_skb_any(ch->trans_skb);
+ ch->trans_skb = NULL;
+ printk(KERN_WARNING
+ "ctcmpc: %s()CCW failure - data lost\n",
+ __FUNCTION__);
+ fsm_event(priv->mpcg->fsm, MPCG_EVENT_INOP, dev);
+ return;
+ }
+ ch->ccw[1].count = ch->trans_skb->len;
+ fsm_addtimer(&ch->timer, CTCM_TIME_5_SEC, CTC_EVENT_TIMER, ch);
+ ch->prof.send_stamp = current_kernel_time(); /* xtime */
+ if (do_debug_ccw)
+ ctcmpc_dumpit((char *)&ch->ccw[0], sizeof(struct ccw1) * 3);
+ rc = ccw_device_start(ch->cdev, &ch->ccw[0],
+ (unsigned long)ch, 0xff, 0);
+ ch->prof.doios_multi++;
+ if (rc != 0) {
+ priv->stats.tx_dropped += i;
+ priv->stats.tx_errors += i;
+ fsm_deltimer(&ch->timer);
+ ctcm_ccw_check_rc(ch, rc, "chained TX");
+ }
+done:
+ ctcm_clear_busy(dev);
+ ctcm_pr_debug("ctcmpc exit: %s %s()\n", dev->name, __FUNCTION__);
+ return;
+}
+
+/**
+ * Got normal data, check for sanity, queue it up, allocate new buffer
+ * trigger bottom half, and initiate next read.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcmpc_chx_rx(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct mpc_group *grp = priv->mpcg;
+ struct sk_buff *skb = ch->trans_skb;
+ struct sk_buff *new_skb;
+ unsigned long saveflags = 0; /* avoids compiler warning */
+ int len = ch->max_bufsize - ch->irb->scsw.count;
+
+ if (do_debug_data) {
+ CTCM_DBF_TEXT_(TRACE, CTC_DBF_DEBUG, "mpc_ch_rx %s cp:%i %s\n",
+ dev->name, smp_processor_id(), ch->id);
+ CTCM_DBF_TEXT_(TRACE, CTC_DBF_DEBUG, "mpc_ch_rx: maxbuf: %04x "
+ "len: %04x\n", ch->max_bufsize, len);
+ }
+ fsm_deltimer(&ch->timer);
+
+ if (skb == NULL) {
+ ctcm_pr_debug("ctcmpc exit: %s() TRANS_SKB = NULL \n",
+ __FUNCTION__);
+ goto again;
+ }
+
+ if (len < TH_HEADER_LENGTH) {
+ ctcm_pr_info("%s: got packet with invalid length %d\n",
+ dev->name, len);
+ priv->stats.rx_dropped++;
+ priv->stats.rx_length_errors++;
+ } else {
+ /* must have valid th header or game over */
+ __u32 block_len = len;
+ len = TH_HEADER_LENGTH + XID2_LENGTH + 4;
+ new_skb = __dev_alloc_skb(ch->max_bufsize, GFP_ATOMIC);
+
+ if (new_skb == NULL) {
+ printk(KERN_INFO "ctcmpc:%s() NEW_SKB = NULL\n",
+ __FUNCTION__);
+ printk(KERN_WARNING "ctcmpc: %s() MEMORY ALLOC FAILED"
+ " - DATA LOST - MPC FAILED\n",
+ __FUNCTION__);
+ fsm_event(priv->mpcg->fsm, MPCG_EVENT_INOP, dev);
+ goto again;
+ }
+ switch (fsm_getstate(grp->fsm)) {
+ case MPCG_STATE_RESET:
+ case MPCG_STATE_INOP:
+ dev_kfree_skb_any(new_skb);
+ break;
+ case MPCG_STATE_FLOWC:
+ case MPCG_STATE_READY:
+ memcpy(skb_put(new_skb, block_len),
+ skb->data, block_len);
+ skb_queue_tail(&ch->io_queue, new_skb);
+ tasklet_schedule(&ch->ch_tasklet);
+ break;
+ default:
+ memcpy(skb_put(new_skb, len), skb->data, len);
+ skb_queue_tail(&ch->io_queue, new_skb);
+ tasklet_hi_schedule(&ch->ch_tasklet);
+ break;
+ }
+ }
+
+again:
+ switch (fsm_getstate(grp->fsm)) {
+ int rc, dolock;
+ case MPCG_STATE_FLOWC:
+ case MPCG_STATE_READY:
+ if (ctcm_checkalloc_buffer(ch))
+ break;
+ ch->trans_skb->data = ch->trans_skb_data;
+ skb_reset_tail_pointer(ch->trans_skb);
+ ch->trans_skb->len = 0;
+ ch->ccw[1].count = ch->max_bufsize;
+ if (do_debug_ccw)
+ ctcmpc_dumpit((char *)&ch->ccw[0],
+ sizeof(struct ccw1) * 3);
+ dolock = !in_irq();
+ if (dolock)
+ spin_lock_irqsave(
+ get_ccwdev_lock(ch->cdev), saveflags);
+ rc = ccw_device_start(ch->cdev, &ch->ccw[0],
+ (unsigned long)ch, 0xff, 0);
+ if (dolock) /* see remark about conditional locking */
+ spin_unlock_irqrestore(
+ get_ccwdev_lock(ch->cdev), saveflags);
+ if (rc != 0)
+ ctcm_ccw_check_rc(ch, rc, "normal RX");
+ default:
+ break;
+ }
+
+ if (do_debug)
+ ctcm_pr_debug("ctcmpc exit : %s %s(): ch=0x%p id=%s\n",
+ dev->name, __FUNCTION__, ch, ch->id);
+
+}
+
+/**
+ * Initialize connection by sending a __u16 of value 0.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+static void ctcmpc_chx_firstio(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+
+ if (do_debug) {
+ struct mpc_group *gptr = priv->mpcg;
+ ctcm_pr_debug("ctcmpc enter: %s(): ch=0x%p id=%s\n",
+ __FUNCTION__, ch, ch->id);
+ ctcm_pr_debug("%s() %s chstate:%i grpstate:%i chprotocol:%i\n",
+ __FUNCTION__, ch->id, fsm_getstate(fi),
+ fsm_getstate(gptr->fsm), ch->protocol);
+ }
+ if (fsm_getstate(fi) == CTC_STATE_TXIDLE)
+ MPC_DBF_DEV_NAME(TRACE, dev, "remote side issued READ? ");
+
+ fsm_deltimer(&ch->timer);
+ if (ctcm_checkalloc_buffer(ch))
+ goto done;
+
+ switch (fsm_getstate(fi)) {
+ case CTC_STATE_STARTRETRY:
+ case CTC_STATE_SETUPWAIT:
+ if (CHANNEL_DIRECTION(ch->flags) == READ) {
+ ctcmpc_chx_rxidle(fi, event, arg);
+ } else {
+ fsm_newstate(fi, CTC_STATE_TXIDLE);
+ fsm_event(priv->fsm, DEV_EVENT_TXUP, dev);
+ }
+ goto done;
+ default:
+ break;
+ };
+
+ fsm_newstate(fi, (CHANNEL_DIRECTION(ch->flags) == READ)
+ ? CTC_STATE_RXINIT : CTC_STATE_TXINIT);
+
+done:
+ if (do_debug)
+ ctcm_pr_debug("ctcmpc exit : %s(): ch=0x%p id=%s\n",
+ __FUNCTION__, ch, ch->id);
+ return;
+}
+
+/**
+ * Got initial data, check it. If OK,
+ * notify device statemachine that we are up and
+ * running.
+ *
+ * fi An instance of a channel statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from channel * upon call.
+ */
+void ctcmpc_chx_rxidle(fsm_instance *fi, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct mpc_group *grp = priv->mpcg;
+ int rc;
+ unsigned long saveflags = 0; /* avoids compiler warning */
+
+ fsm_deltimer(&ch->timer);
+ ctcm_pr_debug("%s cp:%i enter: %s()\n",
+ dev->name, smp_processor_id(), __FUNCTION__);
+ if (do_debug)
+ ctcm_pr_debug("%s() %s chstate:%i grpstate:%i\n",
+ __FUNCTION__, ch->id,
+ fsm_getstate(fi), fsm_getstate(grp->fsm));
+
+ fsm_newstate(fi, CTC_STATE_RXIDLE);
+ /* XID processing complete */
+
+ switch (fsm_getstate(grp->fsm)) {
+ case MPCG_STATE_FLOWC:
+ case MPCG_STATE_READY:
+ if (ctcm_checkalloc_buffer(ch))
+ goto done;
+ ch->trans_skb->data = ch->trans_skb_data;
+ skb_reset_tail_pointer(ch->trans_skb);
+ ch->trans_skb->len = 0;
+ ch->ccw[1].count = ch->max_bufsize;
+ if (do_debug_ccw)
+ ctcmpc_dumpit((char *)&ch->ccw[0],
+ sizeof(struct ccw1) * 3);
+ if (event == CTC_EVENT_START)
+ /* see remark about conditional locking */
+ spin_lock_irqsave(get_ccwdev_lock(ch->cdev), saveflags);
+ rc = ccw_device_start(ch->cdev, &ch->ccw[0],
+ (unsigned long)ch, 0xff, 0);
+ if (event == CTC_EVENT_START)
+ spin_unlock_irqrestore(
+ get_ccwdev_lock(ch->cdev), saveflags);
+ if (rc != 0) {
+ fsm_newstate(fi, CTC_STATE_RXINIT);
+ ctcm_ccw_check_rc(ch, rc, "initial RX");
+ goto done;
+ }
+ break;
+ default:
+ break;
+ }
+
+ fsm_event(priv->fsm, DEV_EVENT_RXUP, dev);
+done:
+ if (do_debug)
+ ctcm_pr_debug("ctcmpc exit: %s %s()\n",
+ dev->name, __FUNCTION__);
+ return;
+}
+
+/*
+ * ctcmpc channel FSM action
+ * called from several points in ctcmpc_ch_fsm
+ * ctcmpc only
+ */
+static void ctcmpc_chx_attn(fsm_instance *fsm, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct mpc_group *grp = priv->mpcg;
+
+ if (do_debug) {
+ ctcm_pr_debug("ctcmpc enter: %s(): cp=%i ch=0x%p id=%s"
+ "GrpState:%s ChState:%s\n",
+ __FUNCTION__, smp_processor_id(), ch, ch->id,
+ fsm_getstate_str(grp->fsm),
+ fsm_getstate_str(ch->fsm));
+ }
+
+ switch (fsm_getstate(grp->fsm)) {
+ case MPCG_STATE_XID2INITW:
+ /* ok..start yside xid exchanges */
+ if (!ch->in_mpcgroup)
+ break;
+ if (fsm_getstate(ch->fsm) == CH_XID0_PENDING) {
+ fsm_deltimer(&grp->timer);
+ fsm_addtimer(&grp->timer,
+ MPC_XID_TIMEOUT_VALUE,
+ MPCG_EVENT_TIMER, dev);
+ fsm_event(grp->fsm, MPCG_EVENT_XID0DO, ch);
+
+ } else if (fsm_getstate(ch->fsm) < CH_XID7_PENDING1)
+ /* attn rcvd before xid0 processed via bh */
+ fsm_newstate(ch->fsm, CH_XID7_PENDING1);
+ break;
+ case MPCG_STATE_XID2INITX:
+ case MPCG_STATE_XID0IOWAIT:
+ case MPCG_STATE_XID0IOWAIX:
+ /* attn rcvd before xid0 processed on ch
+ but mid-xid0 processing for group */
+ if (fsm_getstate(ch->fsm) < CH_XID7_PENDING1)
+ fsm_newstate(ch->fsm, CH_XID7_PENDING1);
+ break;
+ case MPCG_STATE_XID7INITW:
+ case MPCG_STATE_XID7INITX:
+ case MPCG_STATE_XID7INITI:
+ case MPCG_STATE_XID7INITZ:
+ switch (fsm_getstate(ch->fsm)) {
+ case CH_XID7_PENDING:
+ fsm_newstate(ch->fsm, CH_XID7_PENDING1);
+ break;
+ case CH_XID7_PENDING2:
+ fsm_newstate(ch->fsm, CH_XID7_PENDING3);
+ break;
+ }
+ fsm_event(grp->fsm, MPCG_EVENT_XID7DONE, dev);
+ break;
+ }
+
+ if (do_debug)
+ ctcm_pr_debug("ctcmpc exit : %s(): cp=%i ch=0x%p id=%s\n",
+ __FUNCTION__, smp_processor_id(), ch, ch->id);
+ return;
+
+}
+
+/*
+ * ctcmpc channel FSM action
+ * called from one point in ctcmpc_ch_fsm
+ * ctcmpc only
+ */
+static void ctcmpc_chx_attnbusy(fsm_instance *fsm, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct mpc_group *grp = priv->mpcg;
+
+ ctcm_pr_debug("ctcmpc enter: %s %s() %s \nGrpState:%s ChState:%s\n",
+ dev->name,
+ __FUNCTION__, ch->id,
+ fsm_getstate_str(grp->fsm),
+ fsm_getstate_str(ch->fsm));
+
+ fsm_deltimer(&ch->timer);
+
+ switch (fsm_getstate(grp->fsm)) {
+ case MPCG_STATE_XID0IOWAIT:
+ /* vtam wants to be primary.start yside xid exchanges*/
+ /* only receive one attn-busy at a time so must not */
+ /* change state each time */
+ grp->changed_side = 1;
+ fsm_newstate(grp->fsm, MPCG_STATE_XID2INITW);
+ break;
+ case MPCG_STATE_XID2INITW:
+ if (grp->changed_side == 1) {
+ grp->changed_side = 2;
+ break;
+ }
+ /* process began via call to establish_conn */
+ /* so must report failure instead of reverting */
+ /* back to ready-for-xid passive state */
+ if (grp->estconnfunc)
+ goto done;
+ /* this attnbusy is NOT the result of xside xid */
+ /* collisions so yside must have been triggered */
+ /* by an ATTN that was not intended to start XID */
+ /* processing. Revert back to ready-for-xid and */
+ /* wait for ATTN interrupt to signal xid start */
+ if (fsm_getstate(ch->fsm) == CH_XID0_INPROGRESS) {
+ fsm_newstate(ch->fsm, CH_XID0_PENDING) ;
+ fsm_deltimer(&grp->timer);
+ goto done;
+ }
+ fsm_event(grp->fsm, MPCG_EVENT_INOP, dev);
+ goto done;
+ case MPCG_STATE_XID2INITX:
+ /* XID2 was received before ATTN Busy for second
+ channel.Send yside xid for second channel.
+ */
+ if (grp->changed_side == 1) {
+ grp->changed_side = 2;
+ break;
+ }
+ case MPCG_STATE_XID0IOWAIX:
+ case MPCG_STATE_XID7INITW:
+ case MPCG_STATE_XID7INITX:
+ case MPCG_STATE_XID7INITI:
+ case MPCG_STATE_XID7INITZ:
+ default:
+ /* multiple attn-busy indicates too out-of-sync */
+ /* and they are certainly not being received as part */
+ /* of valid mpc group negotiations.. */
+ fsm_event(grp->fsm, MPCG_EVENT_INOP, dev);
+ goto done;
+ }
+
+ if (grp->changed_side == 1) {
+ fsm_deltimer(&grp->timer);
+ fsm_addtimer(&grp->timer, MPC_XID_TIMEOUT_VALUE,
+ MPCG_EVENT_TIMER, dev);
+ }
+ if (ch->in_mpcgroup)
+ fsm_event(grp->fsm, MPCG_EVENT_XID0DO, ch);
+ else
+ printk(KERN_WARNING "ctcmpc: %s() Not all channels have"
+ " been added to group\n", __FUNCTION__);
+
+done:
+ if (do_debug)
+ ctcm_pr_debug("ctcmpc exit : %s()%s ch=0x%p id=%s\n",
+ __FUNCTION__, dev->name, ch, ch->id);
+
+ return;
+
+}
+
+/*
+ * ctcmpc channel FSM action
+ * called from several points in ctcmpc_ch_fsm
+ * ctcmpc only
+ */
+static void ctcmpc_chx_resend(fsm_instance *fsm, int event, void *arg)
+{
+ struct channel *ch = arg;
+ struct net_device *dev = ch->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct mpc_group *grp = priv->mpcg;
+
+ ctcm_pr_debug("ctcmpc enter: %s %s() %s \nGrpState:%s ChState:%s\n",
+ dev->name, __FUNCTION__, ch->id,
+ fsm_getstate_str(grp->fsm),
+ fsm_getstate_str(ch->fsm));
+
+ fsm_event(grp->fsm, MPCG_EVENT_XID0DO, ch);
+
+ return;
+}
+
+/*
+ * ctcmpc channel FSM action
+ * called from several points in ctcmpc_ch_fsm
+ * ctcmpc only
+ */
+static void ctcmpc_chx_send_sweep(fsm_instance *fsm, int event, void *arg)
+{
+ struct channel *ach = arg;
+ struct net_device *dev = ach->netdev;
+ struct ctcm_priv *priv = dev->priv;
+ struct mpc_group *grp = priv->mpcg;
+ struct channel *wch = priv->channel[WRITE];
+ struct channel *rch = priv->channel[READ];
+ struct sk_buff *skb;
+ struct th_sweep *header;
+ int rc = 0;
+ unsigned long saveflags = 0;
+
+ if (do_debug)
+ ctcm_pr_debug("ctcmpc enter: %s(): cp=%i ch=0x%p id=%s\n",
+ __FUNCTION__, smp_processor_id(), ach, ach->id);
+
+ if (grp->in_sweep == 0)
+ goto done;
+
+ if (do_debug_data) {
+ ctcm_pr_debug("ctcmpc: %s() 1: ToVTAM_th_seq= %08x\n" ,
+ __FUNCTION__, wch->th_seq_num);
+ ctcm_pr_debug("ctcmpc: %s() 1: FromVTAM_th_seq= %08x\n" ,
+ __FUNCTION__, rch->th_seq_num);
+ }
+
+ if (fsm_getstate(wch->fsm) != CTC_STATE_TXIDLE) {
+ /* give the previous IO time to complete */
+ fsm_addtimer(&wch->sweep_timer,
+ 200, CTC_EVENT_RSWEEP_TIMER, wch);
+ goto done;
+ }
+
+ skb = skb_dequeue(&wch->sweep_queue);
+ if (!skb)
+ goto done;
+
+ if (set_normalized_cda(&wch->ccw[4], skb->data)) {
+ grp->in_sweep = 0;
+ ctcm_clear_busy_do(dev);
+ dev_kfree_skb_any(skb);
+ fsm_event(grp->fsm, MPCG_EVENT_INOP, dev);
+ goto done;
+ } else {
+ atomic_inc(&skb->users);
+ skb_queue_tail(&wch->io_queue, skb);
+ }
+
+ /* send out the sweep */
+ wch->ccw[4].count = skb->len;
+
+ header = (struct th_sweep *)skb->data;
+ switch (header->th.th_ch_flag) {
+ case TH_SWEEP_REQ:
+ grp->sweep_req_pend_num--;
+ break;
+ case TH_SWEEP_RESP:
+ grp->sweep_rsp_pend_num--;
+ break;
+ }
+
+ header->sw.th_last_seq = wch->th_seq_num;
+
+ if (do_debug_ccw)
+ ctcmpc_dumpit((char *)&wch->ccw[3], sizeof(struct ccw1) * 3);
+
+ ctcm_pr_debug("ctcmpc: %s() sweep packet\n", __FUNCTION__);
+ ctcmpc_dumpit((char *)header, TH_SWEEP_LENGTH);
+
+ fsm_addtimer(&wch->timer, CTCM_TIME_5_SEC, CTC_EVENT_TIMER, wch);
+ fsm_newstate(wch->fsm, CTC_STATE_TX);
+
+ spin_lock_irqsave(get_ccwdev_lock(wch->cdev), saveflags);
+ wch->prof.send_stamp = current_kernel_time(); /* xtime */
+ rc = ccw_device_start(wch->cdev, &wch->ccw[3],
+ (unsigned long) wch, 0xff, 0);
+ spin_unlock_irqrestore(get_ccwdev_lock(wch->cdev), saveflags);
+
+ if ((grp->sweep_req_pend_num == 0) &&
+ (grp->sweep_rsp_pend_num == 0)) {
+ grp->in_sweep = 0;
+ rch->th_seq_num = 0x00;
+ wch->th_seq_num = 0x00;
+ ctcm_clear_busy_do(dev);
+ }
+
+ if (do_debug_data) {
+ ctcm_pr_debug("ctcmpc: %s()2: ToVTAM_th_seq= %08x\n" ,
+ __FUNCTION__, wch->th_seq_num);
+ ctcm_pr_debug("ctcmpc: %s()2: FromVTAM_th_seq= %08x\n" ,
+ __FUNCTION__, rch->th_seq_num);
+ }
+
+ if (rc != 0)
+ ctcm_ccw_check_rc(wch, rc, "send sweep");
+
+done:
+ if (do_debug)
+ ctcm_pr_debug("ctcmpc exit: %s() %s\n", __FUNCTION__, ach->id);
+ return;
+}
+
+
+/*
+ * The ctcmpc statemachine for a channel.
+ */
+
+const fsm_node ctcmpc_ch_fsm[] = {
+ { CTC_STATE_STOPPED, CTC_EVENT_STOP, ctcm_action_nop },
+ { CTC_STATE_STOPPED, CTC_EVENT_START, ctcm_chx_start },
+ { CTC_STATE_STOPPED, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_STOPPED, CTC_EVENT_FINSTAT, ctcm_action_nop },
+ { CTC_STATE_STOPPED, CTC_EVENT_MC_FAIL, ctcm_action_nop },
+
+ { CTC_STATE_NOTOP, CTC_EVENT_STOP, ctcm_chx_stop },
+ { CTC_STATE_NOTOP, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_NOTOP, CTC_EVENT_FINSTAT, ctcm_action_nop },
+ { CTC_STATE_NOTOP, CTC_EVENT_MC_FAIL, ctcm_action_nop },
+ { CTC_STATE_NOTOP, CTC_EVENT_MC_GOOD, ctcm_chx_start },
+ { CTC_STATE_NOTOP, CTC_EVENT_UC_RCRESET, ctcm_chx_stop },
+ { CTC_STATE_NOTOP, CTC_EVENT_UC_RSRESET, ctcm_chx_stop },
+ { CTC_STATE_NOTOP, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+
+ { CTC_STATE_STARTWAIT, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_FINSTAT, ctcm_chx_setmode },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_TIMER, ctcm_chx_setuperr },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_STARTWAIT, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_STARTRETRY, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_STARTRETRY, CTC_EVENT_TIMER, ctcm_chx_setmode },
+ { CTC_STATE_STARTRETRY, CTC_EVENT_FINSTAT, ctcm_chx_setmode },
+ { CTC_STATE_STARTRETRY, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_STARTRETRY, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_FINSTAT, ctcmpc_chx_firstio },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_TIMER, ctcm_chx_setmode },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_SETUPWAIT, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CTC_STATE_RXINIT, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_RXINIT, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_RXINIT, CTC_EVENT_FINSTAT, ctcmpc_chx_rxidle },
+ { CTC_STATE_RXINIT, CTC_EVENT_UC_RCRESET, ctcm_chx_rxiniterr },
+ { CTC_STATE_RXINIT, CTC_EVENT_UC_RSRESET, ctcm_chx_rxiniterr },
+ { CTC_STATE_RXINIT, CTC_EVENT_TIMER, ctcm_chx_rxiniterr },
+ { CTC_STATE_RXINIT, CTC_EVENT_ATTNBUSY, ctcm_chx_rxinitfail },
+ { CTC_STATE_RXINIT, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_RXINIT, CTC_EVENT_UC_ZERO, ctcmpc_chx_firstio },
+ { CTC_STATE_RXINIT, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+
+ { CH_XID0_PENDING, CTC_EVENT_FINSTAT, ctcm_action_nop },
+ { CH_XID0_PENDING, CTC_EVENT_ATTN, ctcmpc_chx_attn },
+ { CH_XID0_PENDING, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CH_XID0_PENDING, CTC_EVENT_START, ctcm_action_nop },
+ { CH_XID0_PENDING, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CH_XID0_PENDING, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CH_XID0_PENDING, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CH_XID0_PENDING, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CH_XID0_PENDING, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CH_XID0_PENDING, CTC_EVENT_ATTNBUSY, ctcm_chx_iofatal },
+
+ { CH_XID0_INPROGRESS, CTC_EVENT_FINSTAT, ctcmpc_chx_rx },
+ { CH_XID0_INPROGRESS, CTC_EVENT_ATTN, ctcmpc_chx_attn },
+ { CH_XID0_INPROGRESS, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CH_XID0_INPROGRESS, CTC_EVENT_START, ctcm_action_nop },
+ { CH_XID0_INPROGRESS, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CH_XID0_INPROGRESS, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CH_XID0_INPROGRESS, CTC_EVENT_UC_ZERO, ctcmpc_chx_rx },
+ { CH_XID0_INPROGRESS, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CH_XID0_INPROGRESS, CTC_EVENT_ATTNBUSY, ctcmpc_chx_attnbusy },
+ { CH_XID0_INPROGRESS, CTC_EVENT_TIMER, ctcmpc_chx_resend },
+ { CH_XID0_INPROGRESS, CTC_EVENT_IO_EBUSY, ctcm_chx_fail },
+
+ { CH_XID7_PENDING, CTC_EVENT_FINSTAT, ctcmpc_chx_rx },
+ { CH_XID7_PENDING, CTC_EVENT_ATTN, ctcmpc_chx_attn },
+ { CH_XID7_PENDING, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CH_XID7_PENDING, CTC_EVENT_START, ctcm_action_nop },
+ { CH_XID7_PENDING, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CH_XID7_PENDING, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CH_XID7_PENDING, CTC_EVENT_UC_ZERO, ctcmpc_chx_rx },
+ { CH_XID7_PENDING, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING, CTC_EVENT_ATTNBUSY, ctcm_chx_iofatal },
+ { CH_XID7_PENDING, CTC_EVENT_TIMER, ctcmpc_chx_resend },
+ { CH_XID7_PENDING, CTC_EVENT_IO_EBUSY, ctcm_chx_fail },
+
+ { CH_XID7_PENDING1, CTC_EVENT_FINSTAT, ctcmpc_chx_rx },
+ { CH_XID7_PENDING1, CTC_EVENT_ATTN, ctcmpc_chx_attn },
+ { CH_XID7_PENDING1, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CH_XID7_PENDING1, CTC_EVENT_START, ctcm_action_nop },
+ { CH_XID7_PENDING1, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CH_XID7_PENDING1, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CH_XID7_PENDING1, CTC_EVENT_UC_ZERO, ctcmpc_chx_rx },
+ { CH_XID7_PENDING1, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING1, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING1, CTC_EVENT_ATTNBUSY, ctcm_chx_iofatal },
+ { CH_XID7_PENDING1, CTC_EVENT_TIMER, ctcmpc_chx_resend },
+ { CH_XID7_PENDING1, CTC_EVENT_IO_EBUSY, ctcm_chx_fail },
+
+ { CH_XID7_PENDING2, CTC_EVENT_FINSTAT, ctcmpc_chx_rx },
+ { CH_XID7_PENDING2, CTC_EVENT_ATTN, ctcmpc_chx_attn },
+ { CH_XID7_PENDING2, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CH_XID7_PENDING2, CTC_EVENT_START, ctcm_action_nop },
+ { CH_XID7_PENDING2, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CH_XID7_PENDING2, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CH_XID7_PENDING2, CTC_EVENT_UC_ZERO, ctcmpc_chx_rx },
+ { CH_XID7_PENDING2, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING2, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING2, CTC_EVENT_ATTNBUSY, ctcm_chx_iofatal },
+ { CH_XID7_PENDING2, CTC_EVENT_TIMER, ctcmpc_chx_resend },
+ { CH_XID7_PENDING2, CTC_EVENT_IO_EBUSY, ctcm_chx_fail },
+
+ { CH_XID7_PENDING3, CTC_EVENT_FINSTAT, ctcmpc_chx_rx },
+ { CH_XID7_PENDING3, CTC_EVENT_ATTN, ctcmpc_chx_attn },
+ { CH_XID7_PENDING3, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CH_XID7_PENDING3, CTC_EVENT_START, ctcm_action_nop },
+ { CH_XID7_PENDING3, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CH_XID7_PENDING3, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CH_XID7_PENDING3, CTC_EVENT_UC_ZERO, ctcmpc_chx_rx },
+ { CH_XID7_PENDING3, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING3, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING3, CTC_EVENT_ATTNBUSY, ctcm_chx_iofatal },
+ { CH_XID7_PENDING3, CTC_EVENT_TIMER, ctcmpc_chx_resend },
+ { CH_XID7_PENDING3, CTC_EVENT_IO_EBUSY, ctcm_chx_fail },
+
+ { CH_XID7_PENDING4, CTC_EVENT_FINSTAT, ctcmpc_chx_rx },
+ { CH_XID7_PENDING4, CTC_EVENT_ATTN, ctcmpc_chx_attn },
+ { CH_XID7_PENDING4, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CH_XID7_PENDING4, CTC_EVENT_START, ctcm_action_nop },
+ { CH_XID7_PENDING4, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CH_XID7_PENDING4, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CH_XID7_PENDING4, CTC_EVENT_UC_ZERO, ctcmpc_chx_rx },
+ { CH_XID7_PENDING4, CTC_EVENT_UC_RCRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING4, CTC_EVENT_UC_RSRESET, ctcm_chx_setuperr },
+ { CH_XID7_PENDING4, CTC_EVENT_ATTNBUSY, ctcm_chx_iofatal },
+ { CH_XID7_PENDING4, CTC_EVENT_TIMER, ctcmpc_chx_resend },
+ { CH_XID7_PENDING4, CTC_EVENT_IO_EBUSY, ctcm_chx_fail },
+
+ { CTC_STATE_RXIDLE, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_RXIDLE, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_RXIDLE, CTC_EVENT_FINSTAT, ctcmpc_chx_rx },
+ { CTC_STATE_RXIDLE, CTC_EVENT_UC_RCRESET, ctcm_chx_rxdisc },
+ { CTC_STATE_RXIDLE, CTC_EVENT_UC_RSRESET, ctcm_chx_fail },
+ { CTC_STATE_RXIDLE, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_RXIDLE, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_RXIDLE, CTC_EVENT_UC_ZERO, ctcmpc_chx_rx },
+
+ { CTC_STATE_TXINIT, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TXINIT, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_TXINIT, CTC_EVENT_FINSTAT, ctcm_chx_txidle },
+ { CTC_STATE_TXINIT, CTC_EVENT_UC_RCRESET, ctcm_chx_txiniterr },
+ { CTC_STATE_TXINIT, CTC_EVENT_UC_RSRESET, ctcm_chx_txiniterr },
+ { CTC_STATE_TXINIT, CTC_EVENT_TIMER, ctcm_chx_txiniterr },
+ { CTC_STATE_TXINIT, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_TXINIT, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_TXINIT, CTC_EVENT_RSWEEP_TIMER, ctcmpc_chx_send_sweep },
+
+ { CTC_STATE_TXIDLE, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TXIDLE, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_TXIDLE, CTC_EVENT_FINSTAT, ctcmpc_chx_firstio },
+ { CTC_STATE_TXIDLE, CTC_EVENT_UC_RCRESET, ctcm_chx_fail },
+ { CTC_STATE_TXIDLE, CTC_EVENT_UC_RSRESET, ctcm_chx_fail },
+ { CTC_STATE_TXIDLE, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_TXIDLE, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_TXIDLE, CTC_EVENT_RSWEEP_TIMER, ctcmpc_chx_send_sweep },
+
+ { CTC_STATE_TERM, CTC_EVENT_STOP, ctcm_action_nop },
+ { CTC_STATE_TERM, CTC_EVENT_START, ctcm_chx_restart },
+ { CTC_STATE_TERM, CTC_EVENT_FINSTAT, ctcm_chx_stopped },
+ { CTC_STATE_TERM, CTC_EVENT_UC_RCRESET, ctcm_action_nop },
+ { CTC_STATE_TERM, CTC_EVENT_UC_RSRESET, ctcm_action_nop },
+ { CTC_STATE_TERM, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_TERM, CTC_EVENT_IO_EBUSY, ctcm_chx_fail },
+ { CTC_STATE_TERM, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+
+ { CTC_STATE_DTERM, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_DTERM, CTC_EVENT_START, ctcm_chx_restart },
+ { CTC_STATE_DTERM, CTC_EVENT_FINSTAT, ctcm_chx_setmode },
+ { CTC_STATE_DTERM, CTC_EVENT_UC_RCRESET, ctcm_action_nop },
+ { CTC_STATE_DTERM, CTC_EVENT_UC_RSRESET, ctcm_action_nop },
+ { CTC_STATE_DTERM, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_DTERM, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+
+ { CTC_STATE_TX, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TX, CTC_EVENT_START, ctcm_action_nop },
+ { CTC_STATE_TX, CTC_EVENT_FINSTAT, ctcmpc_chx_txdone },
+ { CTC_STATE_TX, CTC_EVENT_UC_RCRESET, ctcm_chx_fail },
+ { CTC_STATE_TX, CTC_EVENT_UC_RSRESET, ctcm_chx_fail },
+ { CTC_STATE_TX, CTC_EVENT_TIMER, ctcm_chx_txretry },
+ { CTC_STATE_TX, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_TX, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_TX, CTC_EVENT_RSWEEP_TIMER, ctcmpc_chx_send_sweep },
+ { CTC_STATE_TX, CTC_EVENT_IO_EBUSY, ctcm_chx_fail },
+
+ { CTC_STATE_RXERR, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TXERR, CTC_EVENT_STOP, ctcm_chx_haltio },
+ { CTC_STATE_TXERR, CTC_EVENT_IO_ENODEV, ctcm_chx_iofatal },
+ { CTC_STATE_TXERR, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+ { CTC_STATE_RXERR, CTC_EVENT_MC_FAIL, ctcm_chx_fail },
+};
+
+int mpc_ch_fsm_len = ARRAY_SIZE(ctcmpc_ch_fsm);
+
+/*
+ * Actions for interface - statemachine.
+ */
+
+/**
+ * Startup channels by sending CTC_EVENT_START to each channel.
+ *
+ * fi An instance of an interface statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from struct net_device * upon call.
+ */
+static void dev_action_start(fsm_instance *fi, int event, void *arg)
+{
+ struct net_device *dev = arg;
+ struct ctcm_priv *priv = dev->priv;
+ int direction;
+
+ CTCMY_DBF_DEV_NAME(SETUP, dev, "");
+
+ fsm_deltimer(&priv->restart_timer);
+ fsm_newstate(fi, DEV_STATE_STARTWAIT_RXTX);
+ if (IS_MPC(priv))
+ priv->mpcg->channels_terminating = 0;
+ for (direction = READ; direction <= WRITE; direction++) {
+ struct channel *ch = priv->channel[direction];
+ fsm_event(ch->fsm, CTC_EVENT_START, ch);
+ }
+}
+
+/**
+ * Shutdown channels by sending CTC_EVENT_STOP to each channel.
+ *
+ * fi An instance of an interface statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from struct net_device * upon call.
+ */
+static void dev_action_stop(fsm_instance *fi, int event, void *arg)
+{
+ int direction;
+ struct net_device *dev = arg;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCMY_DBF_DEV_NAME(SETUP, dev, "");
+
+ fsm_newstate(fi, DEV_STATE_STOPWAIT_RXTX);
+ for (direction = READ; direction <= WRITE; direction++) {
+ struct channel *ch = priv->channel[direction];
+ fsm_event(ch->fsm, CTC_EVENT_STOP, ch);
+ ch->th_seq_num = 0x00;
+ if (do_debug)
+ ctcm_pr_debug("ctcm: %s() CH_th_seq= %08x\n",
+ __FUNCTION__, ch->th_seq_num);
+ }
+ if (IS_MPC(priv))
+ fsm_newstate(priv->mpcg->fsm, MPCG_STATE_RESET);
+}
+
+static void dev_action_restart(fsm_instance *fi, int event, void *arg)
+{
+ int restart_timer;
+ struct net_device *dev = arg;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCMY_DBF_DEV_NAME(TRACE, dev, "");
+
+ if (IS_MPC(priv)) {
+ ctcm_pr_info("ctcm: %s Restarting Device and "
+ "MPC Group in 5 seconds\n",
+ dev->name);
+ restart_timer = CTCM_TIME_1_SEC;
+ } else {
+ ctcm_pr_info("%s: Restarting\n", dev->name);
+ restart_timer = CTCM_TIME_5_SEC;
+ }
+
+ dev_action_stop(fi, event, arg);
+ fsm_event(priv->fsm, DEV_EVENT_STOP, dev);
+ if (IS_MPC(priv))
+ fsm_newstate(priv->mpcg->fsm, MPCG_STATE_RESET);
+
+ /* going back into start sequence too quickly can */
+ /* result in the other side becoming unreachable due */
+ /* to sense reported when IO is aborted */
+ fsm_addtimer(&priv->restart_timer, restart_timer,
+ DEV_EVENT_START, dev);
+}
+
+/**
+ * Called from channel statemachine
+ * when a channel is up and running.
+ *
+ * fi An instance of an interface statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from struct net_device * upon call.
+ */
+static void dev_action_chup(fsm_instance *fi, int event, void *arg)
+{
+ struct net_device *dev = arg;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCMY_DBF_DEV_NAME(SETUP, dev, "");
+
+ switch (fsm_getstate(fi)) {
+ case DEV_STATE_STARTWAIT_RXTX:
+ if (event == DEV_EVENT_RXUP)
+ fsm_newstate(fi, DEV_STATE_STARTWAIT_TX);
+ else
+ fsm_newstate(fi, DEV_STATE_STARTWAIT_RX);
+ break;
+ case DEV_STATE_STARTWAIT_RX:
+ if (event == DEV_EVENT_RXUP) {
+ fsm_newstate(fi, DEV_STATE_RUNNING);
+ ctcm_pr_info("%s: connected with remote side\n",
+ dev->name);
+ ctcm_clear_busy(dev);
+ }
+ break;
+ case DEV_STATE_STARTWAIT_TX:
+ if (event == DEV_EVENT_TXUP) {
+ fsm_newstate(fi, DEV_STATE_RUNNING);
+ ctcm_pr_info("%s: connected with remote side\n",
+ dev->name);
+ ctcm_clear_busy(dev);
+ }
+ break;
+ case DEV_STATE_STOPWAIT_TX:
+ if (event == DEV_EVENT_RXUP)
+ fsm_newstate(fi, DEV_STATE_STOPWAIT_RXTX);
+ break;
+ case DEV_STATE_STOPWAIT_RX:
+ if (event == DEV_EVENT_TXUP)
+ fsm_newstate(fi, DEV_STATE_STOPWAIT_RXTX);
+ break;
+ }
+
+ if (IS_MPC(priv)) {
+ if (event == DEV_EVENT_RXUP)
+ mpc_channel_action(priv->channel[READ],
+ READ, MPC_CHANNEL_ADD);
+ else
+ mpc_channel_action(priv->channel[WRITE],
+ WRITE, MPC_CHANNEL_ADD);
+ }
+}
+
+/**
+ * Called from device statemachine
+ * when a channel has been shutdown.
+ *
+ * fi An instance of an interface statemachine.
+ * event The event, just happened.
+ * arg Generic pointer, casted from struct net_device * upon call.
+ */
+static void dev_action_chdown(fsm_instance *fi, int event, void *arg)
+{
+
+ struct net_device *dev = arg;
+ struct ctcm_priv *priv = dev->priv;
+
+ CTCMY_DBF_DEV_NAME(SETUP, dev, "");
+
+ switch (fsm_getstate(fi)) {
+ case DEV_STATE_RUNNING:
+ if (event == DEV_EVENT_TXDOWN)
+ fsm_newstate(fi, DEV_STATE_STARTWAIT_TX);
+ else
+ fsm_newstate(fi, DEV_STATE_STARTWAIT_RX);
+ break;
+ case DEV_STATE_STARTWAIT_RX:
+ if (event == DEV_EVENT_TXDOWN)
+ fsm_newstate(fi, DEV_STATE_STARTWAIT_RXTX);
+ break;
+ case DEV_STATE_STARTWAIT_TX:
+ if (event == DEV_EVENT_RXDOWN)
+ fsm_newstate(fi, DEV_STATE_STARTWAIT_RXTX);
+ break;
+ case DEV_STATE_STOPWAIT_RXTX:
+ if (event == DEV_EVENT_TXDOWN)
+ fsm_newstate(fi, DEV_STATE_STOPWAIT_RX);
+ else
+ fsm_newstate(fi, DEV_STATE_STOPWAIT_TX);
+ break;
+ case DEV_STATE_STOPWAIT_RX:
+ if (event == DEV_EVENT_RXDOWN)
+ fsm_newstate(fi, DEV_STATE_STOPPED);
+ break;
+ case DEV_STATE_STOPWAIT_TX:
+ if (event == DEV_EVENT_TXDOWN)
+ fsm_newstate(fi, DEV_STATE_STOPPED);
+ break;
+ }
+ if (IS_MPC(priv)) {
+ if (event == DEV_EVENT_RXDOWN)
+ mpc_channel_action(priv->channel[READ],
+ READ, MPC_CHANNEL_REMOVE);
+ else
+ mpc_channel_action(priv->channel[WRITE],
+ WRITE, MPC_CHANNEL_REMOVE);
+ }
+}
+
+const fsm_node dev_fsm[] = {
+ { DEV_STATE_STOPPED, DEV_EVENT_START, dev_action_start },
+ { DEV_STATE_STOPWAIT_RXTX, DEV_EVENT_START, dev_action_start },
+ { DEV_STATE_STOPWAIT_RXTX, DEV_EVENT_RXDOWN, dev_action_chdown },
+ { DEV_STATE_STOPWAIT_RXTX, DEV_EVENT_TXDOWN, dev_action_chdown },
+ { DEV_STATE_STOPWAIT_RXTX, DEV_EVENT_RESTART, dev_action_restart },
+ { DEV_STATE_STOPWAIT_RX, DEV_EVENT_START, dev_action_start },
+ { DEV_STATE_STOPWAIT_RX, DEV_EVENT_RXUP, dev_action_chup },
+ { DEV_STATE_STOPWAIT_RX, DEV_EVENT_TXUP, dev_action_chup },
+ { DEV_STATE_STOPWAIT_RX, DEV_EVENT_RXDOWN, dev_action_chdown },
+ { DEV_STATE_STOPWAIT_RX, DEV_EVENT_RESTART, dev_action_restart },
+ { DEV_STATE_STOPWAIT_TX, DEV_EVENT_START, dev_action_start },
+ { DEV_STATE_STOPWAIT_TX, DEV_EVENT_RXUP, dev_action_chup },
+ { DEV_STATE_STOPWAIT_TX, DEV_EVENT_TXUP, dev_action_chup },
+ { DEV_STATE_STOPWAIT_TX, DEV_EVENT_TXDOWN, dev_action_chdown },
+ { DEV_STATE_STOPWAIT_TX, DEV_EVENT_RESTART, dev_action_restart },
+ { DEV_STATE_STARTWAIT_RXTX, DEV_EVENT_STOP, dev_action_stop },
+ { DEV_STATE_STARTWAIT_RXTX, DEV_EVENT_RXUP, dev_action_chup },
+ { DEV_STATE_STARTWAIT_RXTX, DEV_EVENT_TXUP, dev_action_chup },
+ { DEV_STATE_STARTWAIT_RXTX, DEV_EVENT_RXDOWN, dev_action_chdown },
+ { DEV_STATE_STARTWAIT_RXTX, DEV_EVENT_TXDOWN, dev_action_chdown },
+ { DEV_STATE_STARTWAIT_RXTX, DEV_EVENT_RESTART, dev_action_restart },
+ { DEV_STATE_STARTWAIT_TX, DEV_EVENT_STOP, dev_action_stop },
+ { DEV_STATE_STARTWAIT_TX, DEV_EVENT_RXUP, dev_action_chup },
+ { DEV_STATE_STARTWAIT_TX, DEV_EVENT_TXUP, dev_action_chup },
+ { DEV_STATE_STARTWAIT_TX, DEV_EVENT_RXDOWN, dev_action_chdown },
+ { DEV_STATE_STARTWAIT_TX, DEV_EVENT_RESTART, dev_action_restart },
+ { DEV_STATE_STARTWAIT_RX, DEV_EVENT_STOP, dev_action_stop },
+ { DEV_STATE_STARTWAIT_RX, DEV_EVENT_RXUP, dev_action_chup },
+ { DEV_STATE_STARTWAIT_RX, DEV_EVENT_TXUP, dev_action_chup },
+ { DEV_STATE_STARTWAIT_RX, DEV_EVENT_TXDOWN, dev_action_chdown },
+ { DEV_STATE_STARTWAIT_RX, DEV_EVENT_RESTART, dev_action_restart },
+ { DEV_STATE_RUNNING, DEV_EVENT_STOP, dev_action_stop },
+ { DEV_STATE_RUNNING, DEV_EVENT_RXDOWN, dev_action_chdown },
+ { DEV_STATE_RUNNING, DEV_EVENT_TXDOWN, dev_action_chdown },
+ { DEV_STATE_RUNNING, DEV_EVENT_TXUP, ctcm_action_nop },
+ { DEV_STATE_RUNNING, DEV_EVENT_RXUP, ctcm_action_nop },
+ { DEV_STATE_RUNNING, DEV_EVENT_RESTART, dev_action_restart },
+};
+
+int dev_fsm_len = ARRAY_SIZE(dev_fsm);
+
+/* --- This is the END my friend --- */
+