[En-Nut-Discussion] Preview Fix 060626

[Harald Kipp]

Michael,

At 13:00 27.06.2006 +0200, you wrote:

>Therefore, as far as I am concerned the driver is not the problem.

Adding one line to static void NicInterrupt(void *arg) in
arch/avr/dev/lanc111.c

     if (isr & INT_RX_OVRN) {
         imr &= ~INT_RX_OVRN; <------------ This line
         nic_outlb(NIC_ACK, INT_RX_OVRN);
         //nic_outlb(NIC_MMUCR, MMU_TOP);
     }

Replacing

     while (td) {
         qhp = (NUTTHREADINFO **)(td->td_queue);
         if (qhp) {
             cnt = 0;
             for (;;) {
                 NutEnterCritical();
                 cnt += td->td_qpec;
                 td->td_qpec = 0;
                 tqp = *qhp;
                 NutExitCritical();
                 if (cnt == 0 || tqp == SIGNALED) {
                     break;
                 }
                 NutEventPostAsync((HANDLE *)qhp);
                 cnt--;
             }
         }
         td = td->td_next;
     }

by

     while (td) {
         NutEnterCritical();
         cnt = td->td_qpec;
         NutExitCritical();
         if (cnt) {
             qhp = (NUTTHREADINFO **)(td->td_queue);
             NutEnterCritical();
             td->td_qpec--;
             tqp = *qhp;
             NutExitCritical();
             if (tqp != SIGNALED) {
                 NutEventPostAsync((HANDLE *)qhp);
             }
         }
         td = td->td_next;
     }

reduces the context switch. The idea behind this is, that only the
top of the queue needs to be activated.

This is contrary to the demand of alternating threads of the same
priority. However, in this special case it is acceptable.

Harald