Hello,
FIRST QUESTION:
i have added char pkt_nr[48] field to struct skb{} to be able to recognize
packets on their way through linux.
Works fine (i mean reading skb->pkt_nr) from any of networking files
(skb->pkt_nr is first set in 8139too.c just after skb was allocated):
ip_input, ip_forward, ip_output.c etc.
BUT please tell me WHAT IS WRONG in sch_generic.c when i try to access
skb->pkt_nr from qdisc_restart()....
In function that dequeues skb's from output queues (qdisc_restart) (kernel
2.4.20):
82 /* Dequeue packet */
83 if ((skb = q->dequeue(q)) != NULL) {
84 if (spin_trylock(&dev->xmit_lock)) {
85 /* Remember that the driver is grabbed by us. */
86 dev->xmit_lock_owner = smp_processor_id();
87
>> skb->pkt_nr has PROPER value in here !!!
88 /* And release queue */
89 spin_unlock(&dev->queue_lock);
90
91 if (!netif_queue_stopped(dev)) {
92 if (netdev_nit)
93 dev_queue_xmit_nit(skb, dev);
94
95 if (dev->hard_start_xmit(skb, dev) == 0)
{
96 dev->xmit_lock_owner = -1;
97 spin_unlock(&dev->xmit_lock);
>> skb->pkt_nr has UNDEFINED value here !!!!!
Could you please tell WHY is that skb->pkt_nr is undefined (no value)
underneath???
SECOND QUESTION:
how can i transform
sprintf(info, "cycles[%lu], time[%d,%d]", cycles_count, time.tv_sec,
time.tv_usec);
which takes about 3000 cycles into something FASTER. I mean HOW can i
efficiently write (from kernel level) into "info" buffer whithout that much
cycles spent???
regards
adam
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
