I'd love to have a run-time tuneable kernel parameter that lets me use
do_gettimeofday() instead of get_fast_time for received packet
timestamping. Does this seem reasonable?
This is especially useful for me when I perform round trip time
measurements (despite the affects of interrupt latency).
Here's a patch against the 2.4.17. Modified files are
include/linux/sysctl.h, net/core/dev.c, and net/core/sysctrl_net_core.c.
The sysctl tuneable parameter is /proc/sys/net/core/hires_rx_timestamp,
which is set to 0 by default.
Feedback, advice would be appreciated.
Patch follows:
diff -urN -X dontdiff linux-2.4.17/include/linux/sysctl.h linux-2.4.17-wilson/include/linux/sysctl.h
--- linux-2.4.17/include/linux/sysctl.h Mon Nov 26 05:29:17 2001
+++ linux-2.4.17-wilson/include/linux/sysctl.h Wed Jan 16 11:54:00 2002
@@ -204,7 +204,8 @@
NET_CORE_NO_CONG_THRESH=13,
NET_CORE_NO_CONG=14,
NET_CORE_LO_CONG=15,
- NET_CORE_MOD_CONG=16
+ NET_CORE_MOD_CONG=16,
+ NET_CORE_HIRES_RX_TIMESTAMP=17
};
/* /proc/sys/net/ethernet */
diff -urN -X dontdiff linux-2.4.17/net/core/dev.c linux-2.4.17-wilson/net/core/dev.c
--- linux-2.4.17/net/core/dev.c Wed Nov 7 14:39:36 2001
+++ linux-2.4.17-wilson/net/core/dev.c Wed Jan 16 11:53:30 2002
@@ -1069,7 +1069,6 @@
/*=======================================================================
Receiver routines
=======================================================================*/
-
int netdev_max_backlog = 300;
/* These numbers are selected based on intuition and some
* experimentatiom, if you have more scientific way of doing this
@@ -1082,6 +1081,7 @@
struct netif_rx_stats netdev_rx_stat[NR_CPUS];
+int hires_rx_timestamp;
#ifdef CONFIG_NET_HW_FLOWCONTROL
atomic_t netdev_dropping = ATOMIC_INIT(0);
@@ -1217,8 +1217,12 @@
struct softnet_data *queue;
unsigned long flags;
- if (skb->stamp.tv_sec == 0)
- get_fast_time(&skb->stamp);
+ if (skb->stamp.tv_sec == 0) {
+ if (hires_rx_timestamp)
+ do_gettimeofday(&skb->stamp);
+ else
+ get_fast_time(&skb->stamp);
+ }
/* The code is rearranged so that the path is the most
short when CPU is congested, but is still operating.
diff -urN -X dontdiff linux-2.4.17/net/core/sysctl_net_core.c linux-2.4.17-wilson/net/core/sysctl_net_core.c
--- linux-2.4.17/net/core/sysctl_net_core.c Tue Oct 10 10:33:52 2000
+++ linux-2.4.17-wilson/net/core/sysctl_net_core.c Wed Jan 16 11:52:39 2002
@@ -19,6 +19,7 @@
extern int netdev_fastroute;
extern int net_msg_cost;
extern int net_msg_burst;
+extern int hires_rx_timestamp;
extern __u32 sysctl_wmem_max;
extern __u32 sysctl_rmem_max;
@@ -62,6 +63,9 @@
{NET_CORE_MOD_CONG, "mod_cong",
&mod_cong, sizeof(int), 0644, NULL,
&proc_dointvec},
+ {NET_CORE_HIRES_RX_TIMESTAMP, "hires_rx_timestamp",
+ &hires_rx_timestamp, sizeof(int), 0644, NULL,
+ &proc_dointvec},
#ifdef CONFIG_NET_FASTROUTE
{NET_CORE_FASTROUTE, "netdev_fastroute",
&netdev_fastroute, sizeof(int), 0644, NULL,
-
: 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
