[RFC PATCH] packet_mmap: expose hw packet timestamps to network packet capture utilities

[RFC PATCH] packet_mmap: expose hw packet timestamps to network packet capture utilities

[Mcmillan, Scott A]

This patch adds a setting, PACKET_TIMESTAMP, to specify the packet timestamp source that is exported to capture utilities like tcpdump by packet_mmap.  

PACKET_TIMESTAMP accepts the same integer bit field as SO_TIMESTAMPING.  However, only the SOF_TIMESTAMPING_SYS_HARDWARE
and SOF_TIMESTAMPING_RAW_HARDWARE values are currently recognized by PACKET_TIMESTAMP.  SOF_TIMESTAMPING_SYS_HARDWARE takes precedence over SOF_TIMESTAMPING_RAW_HARDWARE if both bits are set.

If PACKET_TIMESTAMP is not set, a software timestamp generated inside the networking stack is used (the behavior before this setting was added).

I am concurrently submitting a patch to the tcpdump / libpcap maintainers adding support for this capability.  

Thanks,
Scott

Signed-off-by: Scott McMillan <scott.a.mcmillan@intel.com>

--- a/include/linux/if_packet.h	2010-05-18 17:22:59.000000000 -0500
+++ b/include/linux/if_packet.h	2010-05-21 14:37:30.000000000 -0500
@@ -48,6 +48,7 @@
 #define PACKET_LOSS			14
 #define PACKET_VNET_HDR			15
 #define PACKET_TX_TIMESTAMP		16
+#define PACKET_TIMESTAMP		17
 
 struct tpacket_stats {
 	unsigned int	tp_packets;
--- a/net/packet/af_packet.c	2010-05-18 17:21:48.000000000 -0500
+++ b/net/packet/af_packet.c	2010-05-21 14:48:41.000000000 -0500
@@ -83,6 +83,7 @@
 #include <linux/if_vlan.h>
 #include <linux/virtio_net.h>
 #include <linux/errqueue.h>
+#include <linux/net_tstamp.h>
 
 #ifdef CONFIG_INET
 #include <net/inet_common.h>
@@ -202,6 +203,7 @@
 	unsigned int		tp_hdrlen;
 	unsigned int		tp_reserve;
 	unsigned int		tp_loss:1;
+	unsigned int		tp_tstamp;
 	struct packet_type	prot_hook ____cacheline_aligned_in_smp;
 };
 
@@ -656,6 +658,7 @@
 	struct sk_buff *copy_skb = NULL;
 	struct timeval tv;
 	struct timespec ts;
+	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
 
 	if (skb->pkt_type == PACKET_LOOPBACK)
 		goto drop;
@@ -737,7 +740,13 @@
 		h.h1->tp_snaplen = snaplen;
 		h.h1->tp_mac = macoff;
 		h.h1->tp_net = netoff;
-		if (skb->tstamp.tv64)
+		if ((po->tp_tstamp & SOF_TIMESTAMPING_SYS_HARDWARE)
+				&& shhwtstamps->syststamp.tv64)
+			tv = ktime_to_timeval(shhwtstamps->syststamp);
+		else if ((po->tp_tstamp & SOF_TIMESTAMPING_RAW_HARDWARE)
+				&& shhwtstamps->hwtstamp.tv64)
+			tv = ktime_to_timeval(shhwtstamps->hwtstamp);
+		else if (skb->tstamp.tv64)
 			tv = ktime_to_timeval(skb->tstamp);
 		else
 			do_gettimeofday(&tv);
@@ -750,7 +759,13 @@
 		h.h2->tp_snaplen = snaplen;
 		h.h2->tp_mac = macoff;
 		h.h2->tp_net = netoff;
-		if (skb->tstamp.tv64)
+		if ((po->tp_tstamp & SOF_TIMESTAMPING_SYS_HARDWARE)
+				&& shhwtstamps->syststamp.tv64)
+			ts = ktime_to_timespec(shhwtstamps->syststamp);
+		else if ((po->tp_tstamp & SOF_TIMESTAMPING_RAW_HARDWARE)
+				&& shhwtstamps->hwtstamp.tv64)
+			ts = ktime_to_timespec(shhwtstamps->hwtstamp);
+		else if (skb->tstamp.tv64)
 			ts = ktime_to_timespec(skb->tstamp);
 		else
 			getnstimeofday(&ts);
@@ -2027,6 +2042,18 @@
 		po->has_vnet_hdr = !!val;
 		return 0;
 	}
+	case PACKET_TIMESTAMP:
+	{
+		int val;
+
+		if (optlen != sizeof(val))
+			return -EINVAL;
+		if (copy_from_user(&val, optval, sizeof(val)))
+			return -EFAULT;
+
+		po->tp_tstamp = val;
+		return 0;
+	}
 	default:
 		return -ENOPROTOOPT;
 	}
@@ -2119,6 +2146,12 @@
 		val = po->tp_loss;
 		data = &val;
 		break;
+	case PACKET_TIMESTAMP:
+		if (len > sizeof(int))
+			len = sizeof(int);
+		val = po->tp_tstamp;
+		data = &val;
+		break;
 	default:
 		return -ENOPROTOOPT;
 	}
--- a/Documentation/networking/packet_mmap.txt	2010-05-18 17:24:18.000000000 -0500
+++ b/Documentation/networking/packet_mmap.txt	2010-05-21 14:39:48.000000000 -0500
@@ -493,6 +493,32 @@
     pfd.events = POLLOUT;
     retval = poll(&pfd, 1, timeout);
 
+-------------------------------------------------------------------------------
++ PACKET_TIMESTAMP
+-------------------------------------------------------------------------------
+
+The PACKET_TIMESTAMP setting determines the source of the timestamp in
+the packet meta information.  If your NIC is capable of timestamping
+packets in hardware, you can request those hardware timestamps to used.
+Note: you may need to enable the generation of hardware timestamps with
+SIOCSHWTSTAMP.
+
+PACKET_TIMESTAMP accepts the same integer bit field as
+SO_TIMESTAMPING.  However, only the SOF_TIMESTAMPING_SYS_HARDWARE
+and SOF_TIMESTAMPING_RAW_HARDWARE values are recognized by
+PACKET_TIMESTAMP.  SOF_TIMESTAMPING_SYS_HARDWARE takes precedence over
+SOF_TIMESTAMPING_RAW_HARDWARE if both bits are set.
+
+    int req = 0;
+    req |= SOF_TIMESTAMPING_SYS_HARDWARE;
+    setsockopt(fd, SOL_PACKET, PACKET_TIMESTAMP, (void *) &req, sizeof(req))
+
+If PACKET_TIMESTAMP is not set, a software timestamp generated inside
+the networking stack is used (the behavior before this setting was added).
+
+See include/linux/net_tstamp.h and Documentation/networking/timestamping
+for more information on hardware timestamps.
+
 --------------------------------------------------------------------------------
 + THANKS
 --------------------------------------------------------------------------------

[PATCH] packet_mmap: expose hw packet timestamps to network packet capture utilities

[Mcmillan, Scott A]

David,

There was no negative feedback on this RFC, and the corresponding tcpdump patch was well received.  Please apply.

Thanks,
Scott

-----Original Message-----
From: netdev-owner@vger.kernel.org [mailto:netdev-owner@vger.kernel.org] On Behalf Of Mcmillan, Scott A
Sent: Friday, May 21, 2010 3:25 PM
To: netdev@vger.kernel.org; davem@davemloft.net
Cc: tcpdump-workers@lists.tcpdump.org
Subject: [RFC PATCH] packet_mmap: expose hw packet timestamps to network packet capture utilities

This patch adds a setting, PACKET_TIMESTAMP, to specify the packet timestamp source that is exported to capture utilities like tcpdump by packet_mmap.  

PACKET_TIMESTAMP accepts the same integer bit field as SO_TIMESTAMPING.  However, only the SOF_TIMESTAMPING_SYS_HARDWARE
and SOF_TIMESTAMPING_RAW_HARDWARE values are currently recognized by PACKET_TIMESTAMP.  SOF_TIMESTAMPING_SYS_HARDWARE takes precedence over SOF_TIMESTAMPING_RAW_HARDWARE if both bits are set.

If PACKET_TIMESTAMP is not set, a software timestamp generated inside the networking stack is used (the behavior before this setting was added).

I am concurrently submitting a patch to the tcpdump / libpcap maintainers adding support for this capability.  

Thanks,
Scott

Signed-off-by: Scott McMillan <scott.a.mcmillan@intel.com>

--- a/include/linux/if_packet.h	2010-05-18 17:22:59.000000000 -0500
+++ b/include/linux/if_packet.h	2010-05-21 14:37:30.000000000 -0500
@@ -48,6 +48,7 @@
 #define PACKET_LOSS			14
 #define PACKET_VNET_HDR			15
 #define PACKET_TX_TIMESTAMP		16
+#define PACKET_TIMESTAMP		17
 
 struct tpacket_stats {
 	unsigned int	tp_packets;
--- a/net/packet/af_packet.c	2010-05-18 17:21:48.000000000 -0500
+++ b/net/packet/af_packet.c	2010-05-21 14:48:41.000000000 -0500
@@ -83,6 +83,7 @@
 #include <linux/if_vlan.h>
 #include <linux/virtio_net.h>
 #include <linux/errqueue.h>
+#include <linux/net_tstamp.h>
 
 #ifdef CONFIG_INET
 #include <net/inet_common.h>
@@ -202,6 +203,7 @@
 	unsigned int		tp_hdrlen;
 	unsigned int		tp_reserve;
 	unsigned int		tp_loss:1;
+	unsigned int		tp_tstamp;
 	struct packet_type	prot_hook ____cacheline_aligned_in_smp;
 };
 
@@ -656,6 +658,7 @@
 	struct sk_buff *copy_skb = NULL;
 	struct timeval tv;
 	struct timespec ts;
+	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
 
 	if (skb->pkt_type == PACKET_LOOPBACK)
 		goto drop;
@@ -737,7 +740,13 @@
 		h.h1->tp_snaplen = snaplen;
 		h.h1->tp_mac = macoff;
 		h.h1->tp_net = netoff;
-		if (skb->tstamp.tv64)
+		if ((po->tp_tstamp & SOF_TIMESTAMPING_SYS_HARDWARE)
+				&& shhwtstamps->syststamp.tv64)
+			tv = ktime_to_timeval(shhwtstamps->syststamp);
+		else if ((po->tp_tstamp & SOF_TIMESTAMPING_RAW_HARDWARE)
+				&& shhwtstamps->hwtstamp.tv64)
+			tv = ktime_to_timeval(shhwtstamps->hwtstamp);
+		else if (skb->tstamp.tv64)
 			tv = ktime_to_timeval(skb->tstamp);
 		else
 			do_gettimeofday(&tv);
@@ -750,7 +759,13 @@
 		h.h2->tp_snaplen = snaplen;
 		h.h2->tp_mac = macoff;
 		h.h2->tp_net = netoff;
-		if (skb->tstamp.tv64)
+		if ((po->tp_tstamp & SOF_TIMESTAMPING_SYS_HARDWARE)
+				&& shhwtstamps->syststamp.tv64)
+			ts = ktime_to_timespec(shhwtstamps->syststamp);
+		else if ((po->tp_tstamp & SOF_TIMESTAMPING_RAW_HARDWARE)
+				&& shhwtstamps->hwtstamp.tv64)
+			ts = ktime_to_timespec(shhwtstamps->hwtstamp);
+		else if (skb->tstamp.tv64)
 			ts = ktime_to_timespec(skb->tstamp);
 		else
 			getnstimeofday(&ts);
@@ -2027,6 +2042,18 @@
 		po->has_vnet_hdr = !!val;
 		return 0;
 	}
+	case PACKET_TIMESTAMP:
+	{
+		int val;
+
+		if (optlen != sizeof(val))
+			return -EINVAL;
+		if (copy_from_user(&val, optval, sizeof(val)))
+			return -EFAULT;
+
+		po->tp_tstamp = val;
+		return 0;
+	}
 	default:
 		return -ENOPROTOOPT;
 	}
@@ -2119,6 +2146,12 @@
 		val = po->tp_loss;
 		data = &val;
 		break;
+	case PACKET_TIMESTAMP:
+		if (len > sizeof(int))
+			len = sizeof(int);
+		val = po->tp_tstamp;
+		data = &val;
+		break;
 	default:
 		return -ENOPROTOOPT;
 	}
--- a/Documentation/networking/packet_mmap.txt	2010-05-18 17:24:18.000000000 -0500
+++ b/Documentation/networking/packet_mmap.txt	2010-05-21 14:39:48.000000000 -0500
@@ -493,6 +493,32 @@
     pfd.events = POLLOUT;
     retval = poll(&pfd, 1, timeout);
 
+-------------------------------------------------------------------------------
++ PACKET_TIMESTAMP
+-------------------------------------------------------------------------------
+
+The PACKET_TIMESTAMP setting determines the source of the timestamp in
+the packet meta information.  If your NIC is capable of timestamping
+packets in hardware, you can request those hardware timestamps to used.
+Note: you may need to enable the generation of hardware timestamps with
+SIOCSHWTSTAMP.
+
+PACKET_TIMESTAMP accepts the same integer bit field as
+SO_TIMESTAMPING.  However, only the SOF_TIMESTAMPING_SYS_HARDWARE
+and SOF_TIMESTAMPING_RAW_HARDWARE values are recognized by
+PACKET_TIMESTAMP.  SOF_TIMESTAMPING_SYS_HARDWARE takes precedence over
+SOF_TIMESTAMPING_RAW_HARDWARE if both bits are set.
+
+    int req = 0;
+    req |= SOF_TIMESTAMPING_SYS_HARDWARE;
+    setsockopt(fd, SOL_PACKET, PACKET_TIMESTAMP, (void *) &req, sizeof(req))
+
+If PACKET_TIMESTAMP is not set, a software timestamp generated inside
+the networking stack is used (the behavior before this setting was added).
+
+See include/linux/net_tstamp.h and Documentation/networking/timestamping
+for more information on hardware timestamps.
+
 --------------------------------------------------------------------------------
 + THANKS
 --------------------------------------------------------------------------------
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] packet_mmap: expose hw packet timestamps to network packet capture utilities

[David Miller]

From: "Mcmillan, Scott A" <scott.a.mcmillan@intel.com>
Date: Thu, 27 May 2010 09:58:19 -0700

> There was no negative feedback on this RFC, and the corresponding
> tcpdump patch was well received.  Please apply.

I think lack of feedback is not an indication of agreement.

You sent that patch right around the openning of the merge window when
people (myself included) are already overwhelmed with existing tasks.

New patches aren't likely to get serious consideration during that
time.

Re: [PATCH] packet_mmap: expose hw packet timestamps to network packet capture utilities

[David Miller]

From: "Mcmillan, Scott A" <scott.a.mcmillan@intel.com>
Date: Thu, 27 May 2010 09:58:19 -0700

> Subject: [RFC PATCH] packet_mmap: expose hw packet timestamps to network packet capture utilities
> 
> This patch adds a setting, PACKET_TIMESTAMP, to specify the packet
> timestamp source that is exported to capture utilities like tcpdump
> by packet_mmap.
> 
> PACKET_TIMESTAMP accepts the same integer bit field as
> SO_TIMESTAMPING.  However, only the SOF_TIMESTAMPING_SYS_HARDWARE
> and SOF_TIMESTAMPING_RAW_HARDWARE values are currently recognized by
> PACKET_TIMESTAMP.  SOF_TIMESTAMPING_SYS_HARDWARE takes precedence
> over SOF_TIMESTAMPING_RAW_HARDWARE if both bits are set.
> 
> If PACKET_TIMESTAMP is not set, a software timestamp generated
> inside the networking stack is used (the behavior before this
> setting was added).
> 
> Signed-off-by: Scott McMillan <scott.a.mcmillan@intel.com>

Ok, this looks fine, applied.

Thanks.