source: trunk/kernel/kern/scheduler.c @ 629

/*
 * scheduler.c - Core scheduler implementation.
 * 
 * Author    Alain Greiner (2016,2017,2018)
 *
 * Copyright (c)  UPMC Sorbonne Universites
 *
 * This file is part of ALMOS-MKH.
 *
 * ALMOS-MKH. 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; version 2.0 of the License.
 *
 * ALMOS-MKH. 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 ALMOS-MKH.; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include <kernel_config.h>
#include <hal_kernel_types.h>
#include <hal_switch.h>
#include <hal_irqmask.h>
#include <hal_context.h>
#include <printk.h>
#include <list.h>
#include <rpc.h>
#include <core.h>
#include <thread.h>
#include <chdev.h>
#include <scheduler.h>


///////////////////////////////////////////////////////////////////////////////////////////
//         global variables
///////////////////////////////////////////////////////////////////////////////////////////

extern chdev_directory_t    chdev_dir;          // allocated in kernel_init.c
extern process_t            process_zero;       // allocated in kernel_init.c

///////////////////////////////////////////////////////////////////////////////////////////
//         private functions
///////////////////////////////////////////////////////////////////////////////////////////


////////////////////////////////////////////////////////////////////////////////////////////
// This static function does NOT modify the scheduler state.
// It just select a thread in the list of attached threads, implementing the following 
// three steps policy: 
// 1) It scan the list of kernel threads, from the next thread after the last executed one,
//    and returns the first runnable found : not IDLE, not blocked, client queue not empty.
//    It can be the current thread.
// 2) If no kernel thread found, it scan the list of user thread, from the next thread after
//    the last executed one, and returns the first runable found : not blocked.
//    It can be the current thread.
// 3) If no runable thread found, it returns the idle thread.
////////////////////////////////////////////////////////////////////////////////////////////
// @ sched   : local pointer on scheduler.
// @ returns pointer on selected thread descriptor
////////////////////////////////////////////////////////////////////////////////////////////
static thread_t * sched_select( scheduler_t * sched )
{
    thread_t     * thread;
    list_entry_t * current;
    list_entry_t * last;
    list_entry_t * root;
    bool_t         done;
    uint32_t       count;

    // first : scan the kernel threads list if not empty 
    if( list_is_empty( &sched->k_root ) == false )
    {
        root    = &sched->k_root;
        last    = sched->k_last;
        done    = false;
        count   = 0;
        current = last;

        while( done == false )
        {

// check kernel threads list
assert( (count < sched->k_threads_nr), "bad kernel threads list" );

            // get next entry in kernel list
            current = current->next;

            // check exit condition
            if( current == last ) done = true;

            // skip the root that does not contain a thread
            if( current == root ) continue;
            else                  count++;

            // get thread pointer for this entry
            thread = LIST_ELEMENT( current , thread_t , sched_list );

            // select kernel thread if non blocked and non THREAD_IDLE
            if( (thread->blocked == 0)  && (thread->type != THREAD_IDLE) ) return thread;

        } // end loop on kernel threads
    } // end kernel threads

    // second : scan the user threads list if not empty 
    if( list_is_empty( &sched->u_root ) == false )
    {
        root    = &sched->u_root;
        last    = sched->u_last;
        done    = false;
        count   = 0;
        current = last;

        while( done == false )
        {

// check user threads list
assert( (count < sched->u_threads_nr), "bad user threads list" );

            // get next entry in user list
            current = current->next;

            // check exit condition
            if( current == last ) done = true;

            // skip the root that does not contain a thread
            if( current == root ) continue;
            else                  count++;

            // get thread pointer for this entry
            thread = LIST_ELEMENT( current , thread_t , sched_list );

            // select thread if non blocked
            if( thread->blocked == 0 )  return thread;

        } // end loop on user threads
    } // end user threads

    // third : return idle thread if no other runnable thread
    return sched->idle;

}  // end sched_select()

////////////////////////////////////////////////////////////////////////////////////////////
// This static function is the only function that can actually delete a thread,
// (and the associated process descriptor if required).
// It is private, because it is only called by the sched_yield() public function.
// It scan all threads attached to a given scheduler, and executes the relevant
// actions for two types of pending requests: 
//
// - REQ_ACK : it checks that target thread is blocked, decrements the response counter
//   to acknowledge the client thread, and reset the pending request. 
// - REQ_DELETE : it removes the target thread from the process th_tbl[], remove it
//   from the scheduler list, and release the memory allocated to thread descriptor.
//   For an user thread, it destroys the process descriptor it the target thread is
//   the last thread in the local process descriptor.
//
// Implementation note:
// We use a while to scan the threads in scheduler lists, because some threads can
// be destroyed, and we want not use a LIST_FOREACH()
////////////////////////////////////////////////////////////////////////////////////////////
// @ core    : local pointer on the core descriptor.
////////////////////////////////////////////////////////////////////////////////////////////
static void sched_handle_signals( core_t * core )
{

    list_entry_t * iter;
    list_entry_t * root;
    thread_t     * thread;
    process_t    * process;
    scheduler_t  * sched;
    uint32_t       threads_nr;   // number of threads in scheduler list
    ltid_t         ltid;         // thread local index
    uint32_t       count;        // number of threads in local process

    // get pointer on scheduler
    sched = &core->scheduler;

    ////////////////// scan user threads to handle both ACK and DELETE requests
    root = &sched->u_root;
    iter = root->next;
    while( iter != root )
    {
        // get pointer on thread
        thread = LIST_ELEMENT( iter , thread_t , sched_list );

        // increment iterator
        iter = iter->next;

        // handle REQ_ACK 
        if( thread->flags & THREAD_FLAG_REQ_ACK )
        {

// check thread blocked
assert( (thread->blocked & THREAD_BLOCKED_GLOBAL) , "thread not blocked" );
 
            // decrement response counter
            hal_atomic_add( thread->ack_rsp_count , -1 );

            // reset REQ_ACK in thread descriptor
            thread_reset_req_ack( thread );
        }

        // handle REQ_DELETE only if target thread != calling thread
        if( (thread->flags & THREAD_FLAG_REQ_DELETE) && (thread != CURRENT_THREAD) )
        {
            // get thread process descriptor
            process = thread->process;

            // get thread ltid
            ltid = LTID_FROM_TRDID( thread->trdid);

            // take the lock protecting sheduler state
            busylock_acquire( &sched->lock );

            // update scheduler state
            threads_nr = sched->u_threads_nr;
            sched->u_threads_nr = threads_nr - 1;
            list_unlink( &thread->sched_list );
            if( sched->u_last == &thread->sched_list )
            {
                if( threads_nr == 1 ) 
                {
                    sched->u_last = NULL;
                }
                else if( sched->u_root.next == &thread->sched_list )
                {
                    sched->u_last = sched->u_root.pred;
                }
                else
                {
                    sched->u_last = sched->u_root.next;
                }
            }

            // release the lock protecting sheduler state
            busylock_release( &sched->lock );

            // release memory allocated for thread 
            count = thread_destroy( thread );

            hal_fence();

#if DEBUG_SCHED_HANDLE_SIGNALS
uint32_t cycle = (uint32_t)hal_get_cycles();
if( DEBUG_SCHED_HANDLE_SIGNALS < cycle )
printk("\n[%s] thread[%x,%x] on core[%x,%d] deleted (still %d threads) / cycle %d\n",
__FUNCTION__, process->pid, thread->trdid, local_cxy, thread->core->lid, count, cycle );
#endif

#if CONFIG_INSTRUMENTATION_PGFAULTS
uint32_t local_nr    = thread->info.local_pgfault_nr;
uint32_t local_cost  = (local_nr == 0)  ? 0 : (thread->info.local_pgfault_cost / local_nr);
uint32_t global_nr   = thread->info.global_pgfault_nr;
uint32_t global_cost = (global_nr == 0) ? 0 : (thread->info.global_pgfault_cost / global_nr);
uint32_t false_nr    = thread->info.false_pgfault_nr;
uint32_t false_cost  = (false_nr == 0)  ? 0 : (thread->info.false_pgfault_cost / false_nr);
printk("***** page faults for thread[%x,%x]\n" 
       "  - %d local  : %d cycles\n"
       "  - %d global : %d cycles\n"
       "  - %d false  : %d cycles\n",
       process->pid, thread->trdid,
       local_nr,  local_cost,
       global_nr, global_cost,
       false_nr,  false_cost );
#endif
            // destroy process descriptor if last thread
            if( count == 1 ) 
            {
                // delete process    
                process_destroy( process );

#if DEBUG_SCHED_HANDLE_SIGNALS
cycle = (uint32_t)hal_get_cycles();
if( DEBUG_SCHED_HANDLE_SIGNALS < cycle )
printk("\n[%s] process %x in cluster %x deleted / cycle %d\n",
__FUNCTION__ , process->pid , local_cxy , cycle );
#endif
            }
        }
    }  // end user threads

    ///////////// scan kernel threads for DELETE only
    root = &sched->k_root;
    iter = root->next;
    while( iter != root )
    {
        // get pointer on thread
        thread = LIST_ELEMENT( iter , thread_t , sched_list );

        // increment iterator
        iter = iter->next;

        // handle REQ_DELETE only if target thread != calling thread
        if( (thread->flags & THREAD_FLAG_REQ_DELETE) && (thread != CURRENT_THREAD) )
        {

// check process descriptor is local kernel process
assert( ( thread->process == &process_zero ) , "illegal process descriptor");

            // get thread ltid
            ltid = LTID_FROM_TRDID( thread->trdid);

            // take the lock protecting sheduler state
            busylock_acquire( &sched->lock );

            // update scheduler state
            threads_nr = sched->k_threads_nr;
            sched->k_threads_nr = threads_nr - 1;
            list_unlink( &thread->sched_list );
            if( sched->k_last == &thread->sched_list )
            {
                if( threads_nr == 1 ) 
                {
                    sched->k_last = NULL;
                }
                else if( sched->k_root.next == &thread->sched_list )
                {
                    sched->k_last = sched->k_root.pred;
                }
                else
                {
                    sched->k_last = sched->k_root.next;
                }
            }

            // release the lock protecting sheduler state
            busylock_release( &sched->lock );

            // get number of threads in local kernel process
            count = process_zero.th_nr;

// check th_nr value
assert( (process_zero.th_nr > 0) , "kernel process th_nr cannot be 0" );

            // remove thread from process th_tbl[]
            process_zero.th_tbl[ltid] = NULL;
            hal_atomic_add( &process_zero.th_nr , - 1 );
 
            // delete thread descriptor
            thread_destroy( thread );

#if DEBUG_SCHED_HANDLE_SIGNALS
uint32_t cycle = (uint32_t)hal_get_cycles();
if( DEBUG_SCHED_HANDLE_SIGNALS < cycle )
printk("\n[%s] thread[%x,%x] on core[%x,%d] deleted / cycle %d\n",
__FUNCTION__ , process_zero.pid , thread->trdid , local_cxy , thread->core->lid , cycle );
#endif
        }
    }
} // end sched_handle_signals()

////////////////////////////////////////////////////////////////////////////////////////////
// This static function is called by the sched_yield function when the RFC_FIFO
// associated to the core is not empty. 
// It search an idle RPC thread for this core, and unblock it if found. 
// It creates a new RPC thread if no idle RPC thread is found.
////////////////////////////////////////////////////////////////////////////////////////////
// @ sched   : local pointer on scheduler.
////////////////////////////////////////////////////////////////////////////////////////////
static void sched_rpc_activate( scheduler_t * sched )
{
    error_t         error;
    thread_t      * thread;  
    list_entry_t  * iter;
    lid_t           lid = CURRENT_THREAD->core->lid;
    bool_t          found = false;

    // search one IDLE RPC thread associated to the selected core   
    LIST_FOREACH( &sched->k_root , iter )
    {
        thread = LIST_ELEMENT( iter , thread_t , sched_list );

        if( (thread->type == THREAD_RPC) && 
            (thread->blocked == THREAD_BLOCKED_IDLE ) ) 
        {
            found = true;
            break;
        }
    }

    if( found == false )     // create new RPC thread     
    {
        error = thread_kernel_create( &thread,
                                      THREAD_RPC, 
                                              &rpc_server_func, 
                                      NULL,
                                          lid );
        // check memory
        if ( error )
        {
            printk("\n[ERROR] in %s : no memory to create a RPC thread in cluster %x\n",
            __FUNCTION__, local_cxy );
        }
        else
        {
            // unblock created RPC thread
            thread->blocked = 0;

            // update RPC threads counter  
            hal_atomic_add( &LOCAL_CLUSTER->rpc_threads[lid] , 1 );

#if DEBUG_SCHED_RPC_ACTIVATE
uint32_t cycle = (uint32_t)hal_get_cycles();
if( DEBUG_SCHED_RPC_ACTIVATE < cycle ) 
printk("\n[%s] new RPC thread %x created for core[%x,%d] / total %d / cycle %d\n",
__FUNCTION__, thread->trdid, local_cxy, lid, LOCAL_CLUSTER->rpc_threads[lid], cycle );
#endif
        }
    }
    else                 // RPC thread found => unblock it
    {
        // unblock found RPC thread
        thread_unblock( XPTR( local_cxy , thread ) , THREAD_BLOCKED_IDLE );

#if DEBUG_SCHED_RPC_ACTIVATE
uint32_t cycle = (uint32_t)hal_get_cycles();
if( DEBUG_SCHED_RPC_ACTIVATE < cycle ) 
printk("\n[%s] idle RPC thread %x unblocked for core[%x,%d] / cycle %d\n",
__FUNCTION__, thread->trdid, local_cxy, lid, cycle );
#endif

    }

} // end sched_rpc_activate()



///////////////////////////////////////////////////////////////////////////////////////////
//         public functions
///////////////////////////////////////////////////////////////////////////////////////////

////////////////////////////////
void sched_init( core_t * core )
{
    scheduler_t * sched = &core->scheduler;

    sched->u_threads_nr   = 0;
    sched->k_threads_nr   = 0;

    sched->current        = CURRENT_THREAD;
    sched->idle           = NULL;               // initialized in kernel_init()
    sched->u_last         = NULL;               // initialized in sched_register_thread()
    sched->k_last         = NULL;               // initialized in sched_register_thread()

    // initialise threads lists
    list_root_init( &sched->u_root );
    list_root_init( &sched->k_root );

    // init lock
    busylock_init( &sched->lock , LOCK_SCHED_STATE );

    sched->req_ack_pending = false;             // no pending request
    sched->trace           = false;             // context switches trace desactivated

}  // end sched_init()

////////////////////////////////////////////
void sched_register_thread( core_t   * core,
                            thread_t * thread )
{
    scheduler_t * sched = &core->scheduler;
    thread_type_t type  = thread->type;

    // take lock protecting sheduler state
    busylock_acquire( &sched->lock );

    if( type == THREAD_USER )
    {
        list_add_last( &sched->u_root , &thread->sched_list );
        sched->u_threads_nr++;
        if( sched->u_last == NULL ) sched->u_last = &thread->sched_list;
    }
    else // kernel thread
    {
        list_add_last( &sched->k_root , &thread->sched_list );
        sched->k_threads_nr++;
        if( sched->k_last == NULL ) sched->k_last = &thread->sched_list; 
    }

    // release lock 
    busylock_release( &sched->lock );

}  // end sched_register_thread()

//////////////////////////////////////////////////////////////////
void sched_yield( const char * cause __attribute__((__unused__)) )
{
    thread_t      * next;
    thread_t      * current = CURRENT_THREAD;
    core_t        * core    = current->core;
    lid_t           lid     = core->lid;
    scheduler_t   * sched   = &core->scheduler;
    remote_fifo_t * fifo    = &LOCAL_CLUSTER->rpc_fifo[lid]; 
 
#if (DEBUG_SCHED_YIELD & 0x1)
if( sched->trace )
sched_display( lid );
#endif

// This assert should never be false, as this check has been
// done before, by any function that can possibly deschedule...
assert( (current->busylocks == 0),
"unexpected descheduling of thread holding %d busylocks = %d\n", current->busylocks ); 

    // activate or create an RPC thread if RPC_FIFO non empty
    if( remote_fifo_is_empty( fifo ) == false )  sched_rpc_activate( sched );

    // disable IRQs / save SR in current thread descriptor
    hal_disable_irq( &current->save_sr );

    // take lock protecting sheduler state
    busylock_acquire( &sched->lock );
    
    // select next thread 
    next = sched_select( sched );

// check next thread kernel_stack overflow
assert( (next->signature == THREAD_SIGNATURE),
"kernel stack overflow for thread %x on core[%x,%d]", next, local_cxy, lid );

// check next thread attached to same core as the calling thread
assert( (next->core == current->core),
"next core %x != current core %x", next->core, current->core );

// check next thread not blocked when type != IDLE
assert( ((next->blocked == 0) || (next->type == THREAD_IDLE)) ,
"next thread %x (%s) is blocked on core[%x,%d]", 
next->trdid , thread_type_str(next->type) , local_cxy , lid );

    // switch contexts and update scheduler state if next != current
        if( next != current )
    {
        // update scheduler 
        sched->current = next;
        if( next->type == THREAD_USER ) sched->u_last = &next->sched_list;
        else                            sched->k_last = &next->sched_list;

        // handle FPU ownership
            if( next->type == THREAD_USER )
        {
                if( next == current->core->fpu_owner )  hal_fpu_enable();
                else                                    hal_fpu_disable();
        }

        // release lock protecting scheduler state
        busylock_release( &sched->lock );

#if DEBUG_SCHED_YIELD
if( sched->trace )
printk("\n[%s] core[%x,%d] / cause = %s\n"
"      thread %x (%s) (%x,%x) => thread %x (%s) (%x,%x) / cycle %d\n",
__FUNCTION__, local_cxy, lid, cause, 
current, thread_type_str(current->type), current->process->pid, current->trdid,next ,
thread_type_str(next->type) , next->process->pid , next->trdid , (uint32_t)hal_get_cycles() );
#endif

        // switch CPU from current thread context to new thread context
        hal_do_cpu_switch( current->cpu_context, next->cpu_context );
    }
    else
    {
        // release lock protecting scheduler state
        busylock_release( &sched->lock );

#if (DEBUG_SCHED_YIELD & 1)
if( sched->trace )
printk("\n[%s] core[%x,%d] / cause = %s\n"
"      thread %x (%s) (%x,%x) continue / cycle %d\n",
__FUNCTION__, local_cxy, lid, cause, current, thread_type_str(current->type),
current->process->pid, current->trdid, (uint32_t)hal_get_cycles() );
#endif

    }

    // handle pending requests for all threads executing on this core.
    sched_handle_signals( core );

    // exit critical section / restore SR from current thread descriptor
    hal_restore_irq( CURRENT_THREAD->save_sr );

}  // end sched_yield()


///////////////////////////////
void sched_display( lid_t lid )
{
    list_entry_t * iter;
    thread_t     * thread;

    core_t       * core    = &LOCAL_CLUSTER->core_tbl[lid];
    scheduler_t  * sched   = &core->scheduler;
    
    // get pointers on TXT0 chdev
    xptr_t    txt0_xp  = chdev_dir.txt_tx[0];
    cxy_t     txt0_cxy = GET_CXY( txt0_xp );
    chdev_t * txt0_ptr = GET_PTR( txt0_xp );

    // get extended pointer on remote TXT0 lock
    xptr_t  lock_xp = XPTR( txt0_cxy , &txt0_ptr->wait_lock );

    // get TXT0 lock 
    remote_busylock_acquire( lock_xp );

    nolock_printk("\n***** threads on core[%x,%d] / current %x / rpc_threads %d / cycle %d\n",
    local_cxy , lid, sched->current, LOCAL_CLUSTER->rpc_threads[lid],
    (uint32_t)hal_get_cycles() );

    // display kernel threads
    LIST_FOREACH( &sched->k_root , iter )
    {
        thread = LIST_ELEMENT( iter , thread_t , sched_list );
        if (thread->type == THREAD_DEV) 
        {
            nolock_printk(" - %s / pid %X / trdid %X / desc %X / block %X / flags %X / %s\n",
            thread_type_str( thread->type ), thread->process->pid, thread->trdid,
            thread, thread->blocked, thread->flags, thread->chdev->name );
        }
        else
        {
            nolock_printk(" - %s / pid %X / trdid %X / desc %X / block %X / flags %X\n",
            thread_type_str( thread->type ), thread->process->pid, thread->trdid,
            thread, thread->blocked, thread->flags );
        }
    }

    // display user threads
    LIST_FOREACH( &sched->u_root , iter )
    {
        thread = LIST_ELEMENT( iter , thread_t , sched_list );
        nolock_printk(" - %s / pid %X / trdid %X / desc %X / block %X / flags %X\n",
        thread_type_str( thread->type ), thread->process->pid, thread->trdid,
        thread, thread->blocked, thread->flags );
    }

    // release TXT0 lock
    remote_busylock_release( lock_xp );

}  // end sched_display()

/////////////////////////////////////
void sched_remote_display( cxy_t cxy,
                           lid_t lid )
{
    thread_t     * thread;

    // get local pointer on target scheduler
    core_t      * core  = &LOCAL_CLUSTER->core_tbl[lid];
    scheduler_t * sched = &core->scheduler;

    // get local pointer on current thread in target scheduler
    thread_t * current = hal_remote_lpt( XPTR( cxy, &sched->current ) );

    // get local pointer on the first kernel and user threads list_entry
    list_entry_t * k_entry = hal_remote_lpt( XPTR( cxy , &sched->k_root.next ) );
    list_entry_t * u_entry = hal_remote_lpt( XPTR( cxy , &sched->u_root.next ) );
    
    // get pointers on TXT0 chdev
    xptr_t    txt0_xp  = chdev_dir.txt_tx[0];
    cxy_t     txt0_cxy = GET_CXY( txt0_xp );
    chdev_t * txt0_ptr = GET_PTR( txt0_xp );

    // get extended pointer on remote TXT0 chdev lock
    xptr_t  lock_xp = XPTR( txt0_cxy , &txt0_ptr->wait_lock );

    // get TXT0 lock 
    remote_busylock_acquire( lock_xp );

    // get rpc_threads
    uint32_t rpcs = hal_remote_l32( XPTR( cxy , &LOCAL_CLUSTER->rpc_threads[lid] ) );
 
    // display header
    nolock_printk("\n***** threads on core[%x,%d] / current %x / rpc_threads %d / cycle %d\n",
    cxy , lid, current, rpcs, (uint32_t)hal_get_cycles() );

    // display kernel threads
    while( k_entry != &sched->k_root )
    {
        // get local pointer on kernel_thread
        thread = LIST_ELEMENT( k_entry , thread_t , sched_list );

        // get relevant thead info
        thread_type_t type    = hal_remote_l32 ( XPTR( cxy , &thread->type ) );
        trdid_t       trdid   = hal_remote_l32 ( XPTR( cxy , &thread->trdid ) );
        uint32_t      blocked = hal_remote_l32 ( XPTR( cxy , &thread->blocked ) );
        uint32_t      flags   = hal_remote_l32 ( XPTR( cxy , &thread->flags ) );
        process_t *   process = hal_remote_lpt ( XPTR( cxy , &thread->process ) );
        pid_t         pid     = hal_remote_l32 ( XPTR( cxy , &process->pid ) );

        // display thread info
        if (type == THREAD_DEV) 
        {
            char      name[16];
            chdev_t * chdev = hal_remote_lpt( XPTR( cxy , &thread->chdev ) );
            hal_remote_strcpy( XPTR( local_cxy , name ), XPTR( cxy , chdev->name ) );

            nolock_printk(" - %s / pid %X / trdid %X / desc %X / block %X / flags %X / %s\n",
            thread_type_str( type ), pid, trdid, thread, blocked, flags, name );
        }
        else
        {
            nolock_printk(" - %s / pid %X / trdid %X / desc %X / block %X / flags %X\n",
            thread_type_str( type ), pid, trdid, thread, blocked, flags );
        }

        // get next remote kernel thread list_entry
        k_entry = hal_remote_lpt( XPTR( cxy , &k_entry->next ) );
    }

    // display user threads
    while( u_entry != &sched->u_root )
    {
        // get local pointer on user_thread
        thread = LIST_ELEMENT( u_entry , thread_t , sched_list );

        // get relevant thead info
        thread_type_t type    = hal_remote_l32 ( XPTR( cxy , &thread->type ) );
        trdid_t       trdid   = hal_remote_l32 ( XPTR( cxy , &thread->trdid ) );
        uint32_t      blocked = hal_remote_l32 ( XPTR( cxy , &thread->blocked ) );
        uint32_t      flags   = hal_remote_l32 ( XPTR( cxy , &thread->flags ) );
        process_t *   process = hal_remote_lpt ( XPTR( cxy , &thread->process ) );
        pid_t         pid     = hal_remote_l32 ( XPTR( cxy , &process->pid ) );

        nolock_printk(" - %s / pid %X / trdid %X / desc %X / block %X / flags %X\n",
        thread_type_str( type ), pid, trdid, thread, blocked, flags );

        // get next user thread list_entry
        u_entry = hal_remote_lpt( XPTR( cxy , &u_entry->next ) );
    }

    // release TXT0 lock
    remote_busylock_release( lock_xp );

}  // end sched_remote_display()