Look Ma' what I found on my harddisk ...

 o New faster syscalls for 2.1.x, too
 o Upgrade to 2.1.89.

Don't try to run this.  It's flaky as hell.  But feel free to debug ...

1 files changed, 311 insertions, 0 deletions

diff --git a/include/net/profile.h b/include/net/profile.h
new file mode 100644
index 000000000..82fad0a37
--- /dev/null
+++ b/include/net/profile.h
@@ -0,0 +1,311 @@
+#include <linux/config.h> /* for CONFIG_NET_PROFILE */
+#ifndef _NET_PROFILE_H_
+#define _NET_PROFILE_H_ 1
+
+#ifdef CONFIG_NET_PROFILE
+
+#include <linux/types.h>
+#include <linux/time.h>
+#include <linux/kernel.h>
+#include <asm/system.h>
+
+struct net_profile_slot
+{
+	char   id[16];
+	struct net_profile_slot *next;
+	struct timeval entered;
+	struct timeval accumulator;
+	struct timeval irq;
+	int    	hits;
+	int    	active;
+	int	underflow;
+};
+
+extern atomic_t net_profile_active;
+extern struct timeval net_profile_adjust;
+extern void net_profile_irq_adjust(struct timeval *entered, struct timeval* leaved);
+
+#if CPU == 586 || CPU == 686
+
+
+extern __inline__ void  net_profile_stamp(struct timeval *pstamp)
+{
+	__asm__ __volatile__ (".byte 0x0f,0x31"
+		:"=a" (pstamp->tv_usec),
+		"=d" (pstamp->tv_sec));
+}
+
+extern __inline__ void  net_profile_accumulate(struct timeval *entered,
+					       struct timeval *leaved,
+					       struct timeval *acc)
+{
+	__asm__ __volatile__ ("subl %2,%0\n\t" 
+			      "sbbl %3,%1\n\t" 
+			      "addl %4,%0\n\t" 
+			      "adcl %5,%1\n\t" 
+			      "subl " SYMBOL_NAME_STR(net_profile_adjust) "+4,%0\n\t" 
+			      "sbbl $0,%1\n\t" 
+			      : "=r" (acc->tv_usec), "=r" (acc->tv_sec)
+			      : "g" (entered->tv_usec), "g" (entered->tv_sec),
+			      "g" (leaved->tv_usec), "g" (leaved->tv_sec),
+			      "0" (acc->tv_usec), "1" (acc->tv_sec));
+}
+
+extern __inline__ void  net_profile_sub(struct timeval *sub,
+					struct timeval *acc)
+{
+	__asm__ __volatile__ ("subl %2,%0\n\t" 
+			      "sbbl %3,%1\n\t" 
+			      : "=r" (acc->tv_usec), "=r" (acc->tv_sec)
+			      : "g" (sub->tv_usec), "g" (sub->tv_sec),
+			      "0" (acc->tv_usec), "1" (acc->tv_sec));
+}
+
+extern __inline__ void  net_profile_add(struct timeval *add,
+					struct timeval *acc)
+{
+	__asm__ __volatile__ ("addl %2,%0\n\t" 
+			      "adcl %3,%1\n\t" 
+			      : "=r" (acc->tv_usec), "=r" (acc->tv_sec)
+			      : "g" (add->tv_usec), "g" (add->tv_sec),
+			      "0" (acc->tv_usec), "1" (acc->tv_sec));
+}
+
+
+#elif defined (__alpha__)
+
+extern __u32 alpha_lo;
+extern long alpha_hi;
+
+/* On alpha cycle counter has only 32 bits :-( :-( */
+
+extern __inline__ void  net_profile_stamp(struct timeval *pstamp)
+{
+	__u32 result;
+	__asm__ __volatile__ ("rpcc %0" : "r="(result));
+	if (result <= alpha_lo)
+		alpha_hi++;
+	alpha_lo = result;
+	pstamp->tv_sec = alpha_hi;
+	pstamp->tv_usec = alpha_lo;
+}
+
+extern __inline__ void  net_profile_accumulate(struct timeval *entered,
+					       struct timeval *leaved,
+					       struct timeval *acc)
+{
+	time_t usecs = acc->tv_usec + leaved->tv_usec - entered->tv_usec
+		- net_profile_adjust.tv_usec;
+	time_t secs = acc->tv_sec + leaved->tv_sec - entered->tv_sec;
+
+	if (usecs >= 0x100000000L) {
+		usecs -= 0x100000000L;
+		secs++;
+	} else if (usecs < -0x100000000L) {
+		usecs += 0x200000000L;
+		secs -= 2;
+	} else if (usecs < 0) {
+		usecs += 0x100000000L;
+		secs--;
+	}
+	acc->tv_sec = secs;
+	acc->tv_usec = usecs;
+}
+
+extern __inline__ void  net_profile_sub(struct timeval *entered,
+					struct timeval *leaved)
+{
+	time_t usecs = leaved->tv_usec - entered->tv_usec;
+	time_t secs = leaved->tv_sec - entered->tv_sec;
+
+	if (usecs < 0) {
+		usecs += 0x100000000L;
+		secs--;
+	}
+	leaved->tv_sec = secs;
+	leaved->tv_usec = usecs;
+}
+
+extern __inline__ void  net_profile_add(struct timeval *entered, struct timeval *leaved)
+{
+	time_t usecs = leaved->tv_usec + entered->tv_usec;
+	time_t secs = leaved->tv_sec + entered->tv_sec;
+
+	if (usecs >= 0x100000000L) {
+		usecs -= 0x100000000L;
+		secs++;
+	}
+	leaved->tv_sec = secs;
+	leaved->tv_usec = usecs;
+}
+
+
+#else
+
+extern __inline__ void  net_profile_stamp(struct timeval *pstamp)
+{
+	/* Not "fast" counterpart! On architectures without
+	   cpu clock "fast" routine is absolutely useless in this
+	   situation. do_gettimeofday still says something on slow-slow-slow
+	   boxes, though it eats more cpu time than the sobject of
+	   investigation :-) :-)
+	 */
+	do_gettimeofday(pstamp);
+}
+
+extern __inline__ void  net_profile_accumulate(struct timeval *entered,
+					       struct timeval *leaved,
+					       struct timeval *acc)
+{
+	time_t usecs = acc->tv_usec + leaved->tv_usec - entered->tv_usec
+		- net_profile_adjust.tv_usec;
+	time_t secs = acc->tv_sec + leaved->tv_sec - entered->tv_sec;
+
+	if (usecs >= 1000000) {
+		usecs -= 1000000;
+		secs++;
+	} else if (usecs < -1000000) {
+		usecs += 2000000;
+		secs -= 2;
+	} else if (usecs < 0) {
+		usecs += 1000000;
+		secs--;
+	}
+	acc->tv_sec = secs;
+	acc->tv_usec = usecs;
+}
+
+extern __inline__ void  net_profile_sub(struct timeval *entered,
+					struct timeval *leaved)
+{
+	time_t usecs = leaved->tv_usec - entered->tv_usec;
+	time_t secs = leaved->tv_sec - entered->tv_sec;
+
+	if (usecs < 0) {
+		usecs += 1000000;
+		secs--;
+	}
+	leaved->tv_sec = secs;
+	leaved->tv_usec = usecs;
+}
+
+extern __inline__ void  net_profile_add(struct timeval *entered, struct timeval *leaved)
+{
+	time_t usecs = leaved->tv_usec + entered->tv_usec;
+	time_t secs = leaved->tv_sec + entered->tv_sec;
+
+	if (usecs >= 1000000) {
+		usecs -= 1000000;
+		secs++;
+	}
+	leaved->tv_sec = secs;
+	leaved->tv_usec = usecs;
+}
+
+
+
+#endif
+
+extern __inline__ void net_profile_enter(struct net_profile_slot *s)
+{
+	unsigned long flags;
+
+	save_flags(flags);
+	cli();
+	if (s->active++ == 0) {
+		net_profile_stamp(&s->entered);
+		atomic_inc(&net_profile_active);
+	}
+	restore_flags(flags);
+}
+
+extern __inline__ void net_profile_leave_irq(struct net_profile_slot *s)
+{
+	unsigned long flags;
+
+	save_flags(flags);
+	cli();
+	if (--s->active <= 0) {
+		if (s->active == 0) {
+			struct timeval curr_pstamp;
+			net_profile_stamp(&curr_pstamp);
+			net_profile_accumulate(&s->entered, &curr_pstamp, &s->accumulator);
+			if (!atomic_dec_and_test(&net_profile_active))
+				net_profile_irq_adjust(&s->entered, &curr_pstamp);
+		} else {
+			s->underflow++;
+		}
+	}
+	s->hits++;
+	restore_flags(flags);
+}
+
+extern __inline__ void net_profile_leave(struct net_profile_slot *s)
+{
+	unsigned long flags;
+	save_flags(flags);
+	cli();
+	if (--s->active <= 0) {
+		if (s->active == 0) {
+			struct timeval curr_pstamp;
+			net_profile_stamp(&curr_pstamp);
+			net_profile_accumulate(&s->entered, &curr_pstamp, &s->accumulator);
+			atomic_dec(&net_profile_active);
+		} else {
+			s->underflow++;
+		}
+	}
+	s->hits++;
+	restore_flags(flags);
+}
+
+
+#define NET_PROFILE_ENTER(slot) net_profile_enter(&net_prof_##slot)
+#define NET_PROFILE_LEAVE(slot) net_profile_leave(&net_prof_##slot)
+#define NET_PROFILE_LEAVE_IRQ(slot) net_profile_leave_irq(&net_prof_##slot)
+
+#define NET_PROFILE_SKB_CLEAR(skb) ({ \
+ skb->pstamp.tv_usec = 0; \
+})
+
+#define NET_PROFILE_SKB_INIT(skb) ({ \
+ net_profile_stamp(&skb->pstamp); \
+})
+
+#define NET_PROFILE_SKB_PASSED(skb, slot) ({ \
+ if (skb->pstamp.tv_usec) { \
+   struct timeval cur_pstamp = skb->pstamp; \
+   net_profile_stamp(&skb->pstamp); \
+   net_profile_accumulate(&cur_pstamp, &skb->pstamp, &net_prof_##slot.accumulator); \
+   net_prof_##slot.hits++; \
+ }})
+
+#define NET_PROFILE_DECL(slot) \
+  extern struct net_profile_slot net_prof_##slot;
+
+#define NET_PROFILE_DEFINE(slot) \
+  struct net_profile_slot net_prof_##slot = { #slot, };
+
+#define NET_PROFILE_REGISTER(slot) net_profile_register(&net_prof_##slot)
+#define NET_PROFILE_UNREGISTER(slot) net_profile_unregister(&net_prof_##slot)
+
+extern int net_profile_init(void);
+extern int net_profile_register(struct net_profile_slot *);
+extern int net_profile_unregister(struct net_profile_slot *);
+
+#else
+
+#define NET_PROFILE_ENTER(slot) do { /* nothing */ } while(0)
+#define NET_PROFILE_LEAVE(slot) do { /* nothing */ } while(0)
+#define NET_PROFILE_LEAVE_IRQ(slot) do { /* nothing */ } while(0)
+#define NET_PROFILE_SKB_CLEAR(skb) do { /* nothing */ } while(0)
+#define NET_PROFILE_SKB_INIT(skb) do { /* nothing */ } while(0)
+#define NET_PROFILE_SKB_PASSED(skb, slot) do { /* nothing */ } while(0)
+#define NET_PROFILE_DECL(slot)
+#define NET_PROFILE_DEFINE(slot)
+#define NET_PROFILE_REGISTER(slot) do { /* nothing */ } while(0)
+#define NET_PROFILE_UNREGISTER(slot) do { /* nothing */ } while(0)
+
+#endif
+
+#endif