Subversion Repositories shark

/*

 * Project: S.Ha.R.K.

 *

 * Coordinators:

 *   Giorgio Buttazzo    <giorgio@sssup.it>

 *   Paolo Gai           <pj@gandalf.sssup.it>

 *

 * Authors:

 *   Paolo Gai           <pj@gandalf.sssup.it>

 *   Massimiliano Giorgi <massy@gandalf.sssup.it>

 *   Luca Abeni          <luca@gandalf.sssup.it>

 *   Anton Cervin

 *

 * ReTiS Lab (Scuola Superiore S.Anna - Pisa - Italy)

 *

 * http://www.sssup.it

 * http://retis.sssup.it

 * http://shark.sssup.it

 */

/**

 ------------

 CVS :        $Id: edf.c,v 1.20 2004-09-15 11:44:11 giacomo Exp $

 File:        $File$

 Revision:    $Revision: 1.20 $

 Last update: $Date: 2004-09-15 11:44:11 $

 ------------

 This file contains the scheduling module EDF (Earliest Deadline First)

 Read edf.h for further details.

**/

/*

 * Copyright (C) 2000,2002 Paolo Gai

 *

 * This program is free software; you can redistribute it and/or modify

 * it under the terms of the GNU General Public License as published by

 * the Free Software Foundation; either version 2 of the License, or

 * (at your option) any later version.

 *

 * This program is distributed in the hope that it will be useful,

 * but WITHOUT ANY WARRANTY; without even the implied warranty of

 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

 * GNU General Public License for more details.

 *

 * You should have received a copy of the GNU General Public License

 * along with this program; if not, write to the Free Software

 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

 *

 */

#include <modules/edf.h>

#include <ll/stdio.h>

#include <ll/string.h>

#include <kernel/model.h>

#include <kernel/descr.h>

#include <kernel/var.h>

#include <kernel/func.h>

#include <tracer.h>

//#define EDF_DEBUG

#define edf_printf kern_printf

#ifdef EDF_DEBUG

char *pnow() {

  static char buf[40];

  struct timespec t;

  sys_gettime(&t);

  sprintf(buf, "%ld.%06ld", t.tv_sec, t.tv_nsec/1000);

  return buf;

}

char *ptime1(struct timespec *t) {

  static char buf[40];

  sprintf(buf, "%ld.%06ld", t->tv_sec, t->tv_nsec/1000);

  return buf;

}

char *ptime2(struct timespec *t) {

  static char buf[40];

  sprintf(buf, "%ld.%06ld", t->tv_sec, t->tv_nsec/1000);

  return buf;

}

#endif

/* Statuses used in the level */

#define EDF_READY      MODULE_STATUS_BASE    /* ready */

#define EDF_IDLE       MODULE_STATUS_BASE+1  /* idle, waiting for offset/eop */

#define EDF_WAIT       MODULE_STATUS_BASE+2  /* to sleep, waiting for eop */

#define EDF_ZOMBIE     MODULE_STATUS_BASE+3  /* to free, waiting for eop */

/* Task flags */

#define EDF_FLAG_SPORADIC    1   /* the task is sporadic */

#define EDF_FLAG_SPOR_LATE   2   /* sporadic task with period overrun */ 

/* Task descriptor */

typedef struct {

  int flags;                    /* task flags                         */

  TIME period;                  /* period (or inter-arrival interval) */

  TIME rdeadline;               /* relative deadline                  */

  TIME offset;                  /* release offset                     */

  struct timespec release;      /* release time of current instance   */

  struct timespec adeadline;    /* latest assigned deadline           */

  int dl_timer;                 /* deadline timer                     */

  int eop_timer;                /* end of period timer                */

  int off_timer;                /* timer offset                       */

  int dl_miss;                  /* deadline miss counter              */

  int wcet_miss;                /* WCET miss counter                  */

  int act_miss;                 /* activation miss counter            */

  int nact;                     /* number of pending periodic jobs    */

} EDF_task_des;

/* Level descriptor */

typedef struct {

  level_des l;                  /* standard level descriptor          */

  int flags;                    /* level flags                        */

  IQUEUE ready;                 /* the ready queue                    */

  bandwidth_t U;                /* used bandwidth                     */

  EDF_task_des tvec[MAX_PROC];  /* vector of task descriptors         */

} EDF_level_des;

/* Module function cross-references */

static void EDF_intern_release(PID p, EDF_level_des *lev);

/**** Timer event handler functions ****/

/* This timer event handler is called at the end of the period */

static void EDF_timer_endperiod(void *par)

{

  PID p = (PID) par;

  EDF_level_des *lev = (EDF_level_des *)level_table[proc_table[p].task_level];

  EDF_task_des *td = &lev->tvec[p];

  td->eop_timer = -1;

  if (proc_table[p].status == EDF_ZOMBIE) {

    /* put the task in the FREE state */

    proc_table[p].status = FREE;

    iq_insertfirst(p,&freedesc);

    /* free the allocated bandwidth */

    lev->U -= (MAX_BANDWIDTH/td->rdeadline) * proc_table[p].wcet;

    return;

  }

  if (proc_table[p].status == EDF_WAIT) {

    proc_table[p].status = SLEEP;

    return;

  }

  if (td->flags & EDF_FLAG_SPORADIC) {

    /* the task is sporadic and still busy, mark it as late */

    td->flags |= EDF_FLAG_SPOR_LATE;

  } else {

    /* the task is periodic, release/queue another instance */

    EDF_intern_release(p, lev);

  }

}

/* This timer event handler is called when a task misses its deadline */

static void EDF_timer_deadline(void *par)

{

  PID p = (PID) par;

  EDF_level_des *lev = (EDF_level_des *)level_table[proc_table[p].task_level];

  EDF_task_des *td = &lev->tvec[p];

  td->dl_timer = -1;

  TRACER_LOGEVENT(FTrace_EVT_task_deadline_miss,

                  (unsigned short int)proc_table[p].context,0);

  if (lev->flags & EDF_ENABLE_DL_EXCEPTION) {

    kern_raise(XDEADLINE_MISS,p);

  } else {

    td->dl_miss++;

  }

}

/* This timer event handler is called after waiting for an offset */

static void EDF_timer_offset(void *par)

{

  PID p = (PID) par;

  EDF_level_des *lev;

  lev = (EDF_level_des *)level_table[proc_table[p].task_level];

  EDF_task_des *td = &lev->tvec[p];

  td->off_timer = -1;

  /* release the task now */

  EDF_intern_release(p, lev);

}

/* This function is called when a guest task misses its deadline */

static void EDF_timer_guest_deadline(void *par)

{

  PID p = (PID) par;

  EDF_level_des *lev = (EDF_level_des *)level_table[proc_table[p].task_level];

  EDF_task_des *td = &lev->tvec[p];

  td->dl_timer = -1;

  TRACER_LOGEVENT(FTrace_EVT_task_deadline_miss,

                  (unsigned short int)proc_table[p].context,0);

  kern_raise(XDEADLINE_MISS,p);

}

/**** Internal utility functions ****/

/* Release (or queue) a task, post deadline and endperiod timers */

static void EDF_intern_release(PID p, EDF_level_des *lev)

{

  struct timespec temp;

  EDF_task_des *td = &lev->tvec[p];

  /* post deadline timer */

  if (lev->flags & EDF_ENABLE_DL_CHECK) {

    temp = td->release;

    ADDUSEC2TIMESPEC(td->rdeadline, &temp);

    if (td->dl_timer != -1) {

        kern_event_delete(td->dl_timer);

        td->dl_timer = -1;

    }

    td->dl_timer = kern_event_post(&temp,EDF_timer_deadline,(void *)p);

  }

  /* release or queue next job */

  if (proc_table[p].status == EDF_IDLE) {

    /* assign deadline, insert task in the ready queue */

    proc_table[p].status = EDF_READY;

    *iq_query_timespec(p,&lev->ready) = td->adeadline;

    iq_timespec_insert(p,&lev->ready);

#ifdef EDF_DEBUG

    edf_printf("At %s: releasing %s with deadline %s\n", pnow(),

       proc_table[p].name, ptime1(&td->adeadline));

#endif

    /* increase assigned deadline */

    ADDUSEC2TIMESPEC(td->period, &td->adeadline);

    /* reschedule */

    event_need_reschedule();

  } else {

    /* queue */

    td->nact++;

  }

  /* increase release time */

  ADDUSEC2TIMESPEC(td->period, &td->release);

  /* post end of period timer */

  if (td->eop_timer != -1) {

    kern_event_delete(td->eop_timer);

    td->eop_timer = -1;

  }

  td->eop_timer = kern_event_post(&td->release, EDF_timer_endperiod,(void *)p);

  TRACER_LOGEVENT(FTrace_EVT_task_timer,

                  (unsigned short int)proc_table[p].context,

                  (unsigned int)proc_table[p].task_level);

}

/**** Public generic kernel interface functions ****/

/* Returns the first task in the ready queue */

static PID EDF_public_scheduler(LEVEL l)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  return iq_query_first(&lev->ready);

}

/* Checks and decreases the available system bandwidth */

static int EDF_public_guarantee(LEVEL l, bandwidth_t *freebandwidth)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  if (*freebandwidth >= lev->U) {

    *freebandwidth -= lev->U;

    return 1;

  }

  else

    return 0;

}

/* Called by task_create: Checks task model and creates a task */

static int EDF_public_create(LEVEL l, PID p, TASK_MODEL *m)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  HARD_TASK_MODEL *h;

  if (m->pclass != HARD_PCLASS) return -1;

  if (m->level != 0 && m->level != l) return -1;

  h = (HARD_TASK_MODEL *)m;

  if (!h->wcet || !h->mit) return -1;

  if (h->drel > h->mit) return -1;  /* only D <= T supported */

  if (!h->drel) {

    td->rdeadline = h->mit;

  } else {

    td->rdeadline = h->drel;

  }

  /* check the free bandwidth... */

  if (lev->flags & EDF_ENABLE_GUARANTEE) {

    bandwidth_t b;

    b = (MAX_BANDWIDTH / td->rdeadline) * h->wcet;

    /* really update lev->U, checking an overflow... */

    if (MAX_BANDWIDTH - lev->U > b) {

      lev->U += b;

    } else {

      return -1;

    }

  }

  td->flags = 0;

  if (h->periodicity == APERIODIC) {

    td->flags |= EDF_FLAG_SPORADIC;

  }

  td->period = h->mit;

  if (td->rdeadline == td->period) {

    /* Ensure that D <= T-eps to make dl_timer trigger before eop_timer */

    td->rdeadline = td->period - 1;

  }

  td->offset = h->offset;

  td->dl_timer = -1;

  td->eop_timer = -1;

  td->off_timer = -1;

  td->dl_miss = 0;

  td->wcet_miss = 0;

  td->act_miss = 0;

  td->nact = 0;

  /* Enable wcet check */

  if (lev->flags & EDF_ENABLE_WCET_CHECK) {

    proc_table[p].avail_time = h->wcet;

    proc_table[p].wcet       = h->wcet;

    proc_table[p].control |= CONTROL_CAP; /* turn on measurement */

  }

  return 0; /* OK, also if the task cannot be guaranteed... */

}

/* Reclaim the bandwidth used by the task */

static void EDF_public_detach(LEVEL l, PID p)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  if (lev->flags & EDF_ENABLE_GUARANTEE) {

    lev->U -= (MAX_BANDWIDTH / td->rdeadline) * proc_table[p].wcet;

  }

}

/* Extracts the running task from the ready queue */

static void EDF_public_dispatch(LEVEL l, PID p, int nostop)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  iq_extract(p, &lev->ready);

}

/* Called when the task is preempted or when its budget is exhausted */

static void EDF_public_epilogue(LEVEL l, PID p)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  /* check if the wcet is finished... */

  if (lev->flags & EDF_ENABLE_WCET_CHECK) {

    if (proc_table[p].avail_time <= 0) {

      TRACER_LOGEVENT(FTrace_EVT_task_wcet_violation,

                      (unsigned short int)proc_table[p].context,0);

      if (lev->flags & EDF_ENABLE_WCET_EXCEPTION) {

        kern_raise(XWCET_VIOLATION,p);

      } else {

        proc_table[p].control &= ~CONTROL_CAP;

        td->wcet_miss++;

      }

    }

  }

  /* the task returns to the ready queue */

  iq_timespec_insert(p,&lev->ready);

  proc_table[p].status = EDF_READY;

}

/* Called by task_activate or group_activate: Activates the task at time t */

static void EDF_public_activate(LEVEL l, PID p, struct timespec *t)

{

  struct timespec clocktime;

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  kern_gettime(&clocktime);

  /* check if we are not in the SLEEP state */

  if (proc_table[p].status != SLEEP) {

    if (lev->flags & EDF_ENABLE_ACT_EXCEPTION) {

      /* too frequent or wrongful activation: raise exception */

      kern_raise(XACTIVATION,p);

    } else {

      /* skip the sporadic job, but increase a counter */

#ifdef EDF_DEBUG

      edf_printf("At %s: activation of %s skipped\n", pnow(),

                 proc_table[p].name);

#endif

      td->act_miss++;

    }

    return;

  }

  /* set the release time to the activation time + offset */

  td->release = *t;

  ADDUSEC2TIMESPEC(td->offset, &td->release);

  /* set the absolute deadline to the activation time + offset + rdeadline */

  td->adeadline = td->release;

  ADDUSEC2TIMESPEC(td->rdeadline, &td->adeadline);

  /* Check if release > clocktime. If yes, release it later,

     otherwise release it now. */

  proc_table[p].status = EDF_IDLE;

  if (TIMESPEC_A_GT_B(&td->release, &clocktime)) {

    /* release later, post an offset timer */

    if (td->off_timer != -1) {

        kern_event_delete(td->off_timer);

        td->off_timer = -1;

    }

    td->off_timer = kern_event_post(&td->release,EDF_timer_offset,(void *)p);

  } else {

    /* release now */

    EDF_intern_release(p, lev);

  }

}

/* Reinserts a task that has been blocked into the ready queue */

static void EDF_public_unblock(LEVEL l, PID p)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  /* Insert task in the correct position */

  proc_table[p].status = EDF_READY;

  iq_timespec_insert(p,&lev->ready);

}

/* Called when a task experiences a synchronization block */

static void EDF_public_block(LEVEL l, PID p)

{

  /* Extract the running task from the level

     . we have already extract it from the ready queue at the dispatch time.

     . the capacity event have to be removed by the generic kernel

     . the wcet don't need modification...

     . the state of the task is set by the calling function

     . the deadline must remain...

     So, we do nothing!!!

  */

}

/* Called by task_endcycle or task_sleep: Ends the current instance */

static int EDF_public_message(LEVEL l, PID p, void *m)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  switch((long)(m)) {

    /* task_endcycle() */

  case 0:

    /* if there are no pending jobs */

    if (td->nact == 0) {

      /* remove deadline timer, if any */

      if (td->dl_timer != -1) {

        kern_event_delete(td->dl_timer);

        td->dl_timer = -1;

      }

      if (td->flags & EDF_FLAG_SPORADIC) {

        /* sporadic task */

        if (!(td->flags & EDF_FLAG_SPOR_LATE)) {

          proc_table[p].status = EDF_WAIT;

        } else {

          /* it's late, move it directly to SLEEP */

          proc_table[p].status = SLEEP;

          td->flags &= ~EDF_FLAG_SPOR_LATE;

        }

      } else {

        /* periodic task */

        proc_table[p].status = EDF_IDLE;

      }

    } else {

      /* we are late / there are pending jobs */

      td->nact--;

      /* compute and assign absolute deadline */

      *iq_query_timespec(p,&lev->ready) = td->adeadline;

      iq_timespec_insert(p,&lev->ready);

      /* increase assigned deadline */

      ADDUSEC2TIMESPEC(td->period, &td->adeadline);

#ifdef EDF_DEBUG

      edf_printf("(Late) At %s: releasing %s with deadline %s\n",

         pnow(),proc_table[p].name,ptime1(&td->adeadline));

#endif

    }

    break;

    /* task_sleep() */

  case 1:

    /* remove deadline timer, if any */

    if (td->dl_timer != -1) {

      kern_event_delete(td->dl_timer);

      td->dl_timer = -1;

    }

    if (td->flags & EDF_FLAG_SPORADIC) {

      /* sporadic task */

      if (!(td->flags & EDF_FLAG_SPOR_LATE)) {

        proc_table[p].status = EDF_WAIT;

      } else {

        /* it's late, move it directly to SLEEP */

        proc_table[p].status = SLEEP;

        td->flags &= ~EDF_FLAG_SPOR_LATE;

      }

    } else {

      /* periodic task */

      if (!(td->nact > 0)) {

        /* we are on time. go to the EDF_WAIT state */

        proc_table[p].status = EDF_WAIT;

      } else {

        /* we are late. delete pending activations and go to SLEEP */

        td->nact = 0;

        proc_table[p].status = SLEEP;

        /* remove end of period timer */

        if (td->eop_timer != -1) {

          kern_event_delete(td->eop_timer);

          td->eop_timer = -1;

        }

      }

    }

    break;

  }

  if (lev->flags & EDF_ENABLE_WCET_CHECK) {

    proc_table[p].control |= CONTROL_CAP;

  }

  jet_update_endcycle(); /* Update the Jet data... */

  proc_table[p].avail_time = proc_table[p].wcet;

  TRACER_LOGEVENT(FTrace_EVT_task_end_cycle,

                  (unsigned short int)proc_table[p].context,(unsigned int)l);

  return 0;

}

/* End the task and free the resources at the end of the period */

static void EDF_public_end(LEVEL l, PID p)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  if (!(td->flags & EDF_FLAG_SPOR_LATE)) {

    /* remove the deadline timer (if any) */

    if (td->dl_timer != -1) {

      kern_event_delete(td->dl_timer);

      td->dl_timer = -1;

    }

    proc_table[p].status = EDF_ZOMBIE;

  } else {

    /* no endperiod timer will be fired, free the task now! */

    proc_table[p].status = FREE;

    iq_insertfirst(p,&freedesc);

    /* free the allocated bandwidth */

    lev->U -= (MAX_BANDWIDTH/td->rdeadline) * proc_table[p].wcet;

  }

}

/**** Private generic kernel interface functions (guest calls) ****/

/* Insert a guest task */

static void EDF_private_insert(LEVEL l, PID p, TASK_MODEL *m)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  JOB_TASK_MODEL *job;

  if (m->pclass != JOB_PCLASS || (m->level != 0 && m->level != l) ) {

    kern_raise(XINVALID_TASK, p);

    return;

  }

  job = (JOB_TASK_MODEL *)m;

  /* Insert task in the correct position */

  *iq_query_timespec(p, &lev->ready) = job->deadline;

  iq_timespec_insert(p,&lev->ready);

  proc_table[p].status = EDF_READY;

  if (td->dl_timer != -1) {

        kern_event_delete(td->dl_timer);

        td->dl_timer = -1;

  }

  td->period = job->period;

  if (!job->noraiseexc) {

    td->dl_timer = kern_event_post(iq_query_timespec(p, &lev->ready),

                                       EDF_timer_guest_deadline,(void *)p);

  }

}

/* Dispatch a guest task */

static void EDF_private_dispatch(LEVEL l, PID p, int nostop)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  /* the task state is set to EXE by the scheduler()

     we extract the task from the ready queue

     NB: we can't assume that p is the first task in the queue!!! */

  iq_extract(p, &lev->ready);

}

/* Called when a guest task is preempted/out of budget */

static void EDF_private_epilogue(LEVEL l, PID p)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  /* the task has been preempted. it returns into the ready queue... */

  iq_timespec_insert(p,&lev->ready);

  proc_table[p].status = EDF_READY;

}

/* Extract a guest task */

static void EDF_private_extract(LEVEL l, PID p)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  if (proc_table[p].status == EDF_READY)

    iq_extract(p, &lev->ready);

  /* we remove the deadline timer, because the slice is finished */

  if (td->dl_timer != -1) {

    kern_event_delete(td->dl_timer);

    td->dl_timer = -1;

  }

}

/**** Level registration function ****/

LEVEL EDF_register_level(int flags)

{

  LEVEL l;            /* the level that we register */

  EDF_level_des *lev;  /* for readableness only */

  int i;

  printk("EDF_register_level\n");

  /* request an entry in the level_table */

  l = level_alloc_descriptor(sizeof(EDF_level_des));

  lev = (EDF_level_des *)level_table[l];

  /* fill the standard descriptor */

  lev->l.private_insert   = EDF_private_insert;

  lev->l.private_extract  = EDF_private_extract;

  lev->l.private_dispatch = EDF_private_dispatch;

  lev->l.private_epilogue = EDF_private_epilogue;

  lev->l.public_scheduler = EDF_public_scheduler;

  if (flags & EDF_ENABLE_GUARANTEE)

    lev->l.public_guarantee = EDF_public_guarantee;

  else

    lev->l.public_guarantee = NULL;

  lev->l.public_create    = EDF_public_create;

  lev->l.public_detach    = EDF_public_detach;

  lev->l.public_end       = EDF_public_end;

  lev->l.public_dispatch  = EDF_public_dispatch;

  lev->l.public_epilogue  = EDF_public_epilogue;

  lev->l.public_activate  = EDF_public_activate;

  lev->l.public_unblock   = EDF_public_unblock;

  lev->l.public_block     = EDF_public_block;

  lev->l.public_message   = EDF_public_message;

  iq_init(&lev->ready, &freedesc, 0);

  lev->flags = flags;

  if (lev->flags & EDF_ENABLE_WCET_EXCEPTION) {

    lev->flags |= EDF_ENABLE_WCET_CHECK;

  }

  if (lev->flags & EDF_ENABLE_DL_EXCEPTION) {

    lev->flags |= EDF_ENABLE_DL_CHECK;

  }

  lev->U = 0;

  for (i=0;i<MAX_PROC;i++) {

    EDF_task_des *td = &lev->tvec[i];

    td->flags = 0;

    td->dl_timer = -1;

    td->eop_timer = -1;

    td->off_timer = -1;

    td->dl_miss = 0;

    td->wcet_miss = 0;

    td->act_miss = 0;

    td->nact = 0;

  }

  return l;

}

/**** Public utility functions ****/

/* Get the bandwidth used by the level */

bandwidth_t EDF_usedbandwidth(LEVEL l)

{

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  return lev->U;

}

/* Get the number of missed deadlines for a task */

int EDF_get_dl_miss(PID p)

{

  LEVEL l = proc_table[p].task_level;

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  return td->dl_miss;

}

/* Get the number of execution overruns for a task */

int EDF_get_wcet_miss(PID p)

{

  LEVEL l = proc_table[p].task_level;

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];

  return td->wcet_miss;

}

/* Get the number of skipped activations for a task */

int EDF_get_act_miss(PID p)

{

  LEVEL l = proc_table[p].task_level;

  EDF_level_des *lev = (EDF_level_des *)(level_table[l]);

  EDF_task_des *td = &lev->tvec[p];