| Message ID | 20140828161447.6652.59388.stgit@dragon |
|---|---|
| State | Accepted, archived |
| Delegated to: | David Miller |
| Headers | show |
On Thu, 2014-08-28 at 18:14 +0200, Jesper Dangaard Brouer wrote: > Then testing the TX limits of the stack, then it is useful to > be-able to disable the do_gettimeofday() timetamping on every packet. > > This implements a pktgen flag NO_TIMESTAMP which will disable this > call to do_gettimeofday(). > > The performance change on (my system E5-2695) with skb_clone=0, goes > from TX 2,423,751 pps to 2,567,165 pps with flag NO_TIMESTAMP. Thus, > the cost of do_gettimeofday() or saving is approx 23 nanosec. I guess using local_clock() would provide an accurate and less expensive timestamp, if a timestamp is needed. -- 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
On Fri, 29 Aug 2014 09:17:06 -0700 Eric Dumazet <eric.dumazet@gmail.com> wrote: > On Thu, 2014-08-28 at 18:14 +0200, Jesper Dangaard Brouer wrote: > > Then testing the TX limits of the stack, then it is useful to > > be-able to disable the do_gettimeofday() timetamping on every packet. > > > > This implements a pktgen flag NO_TIMESTAMP which will disable this > > call to do_gettimeofday(). > > > > The performance change on (my system E5-2695) with skb_clone=0, goes > > from TX 2,423,751 pps to 2,567,165 pps with flag NO_TIMESTAMP. Thus, > > the cost of do_gettimeofday() or saving is approx 23 nanosec. > > I guess using local_clock() would provide an accurate and less expensive > timestamp, if a timestamp is needed. Sure, but I don't need this timestamp, thus I added an option to disable it.
From: Jesper Dangaard Brouer <brouer@redhat.com> Date: Thu, 28 Aug 2014 18:14:47 +0200 > Then testing the TX limits of the stack, then it is useful to > be-able to disable the do_gettimeofday() timetamping on every packet. > > This implements a pktgen flag NO_TIMESTAMP which will disable this > call to do_gettimeofday(). > > The performance change on (my system E5-2695) with skb_clone=0, goes > from TX 2,423,751 pps to 2,567,165 pps with flag NO_TIMESTAMP. Thus, > the cost of do_gettimeofday() or saving is approx 23 nanosec. > > Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> This is fine, applied, 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
On 08/28/2014 09:14 AM, Jesper Dangaard Brouer wrote: > Then testing the TX limits of the stack, then it is useful to > be-able to disable the do_gettimeofday() timetamping on every packet. > > This implements a pktgen flag NO_TIMESTAMP which will disable this > call to do_gettimeofday(). > > The performance change on (my system E5-2695) with skb_clone=0, goes > from TX 2,423,751 pps to 2,567,165 pps with flag NO_TIMESTAMP. Thus, > the cost of do_gettimeofday() or saving is approx 23 nanosec. > > Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> > + else if (strcmp(f, "NO_TIMESTAMP") == 0) > + pkt_dev->flags |= F_NO_TIMESTAMP; > + While porting pktgen changes into my own modified pktgen logic, I noticed that there is no handling for !NO_TIMESTAMP, so it seems you cannot disable this once it is enabled? Thanks, Ben
diff --git a/net/core/pktgen.c b/net/core/pktgen.c index 83e2b4b..d9acc16 100644 --- a/net/core/pktgen.c +++ b/net/core/pktgen.c @@ -202,6 +202,7 @@ #define F_QUEUE_MAP_CPU (1<<14) /* queue map mirrors smp_processor_id() */ #define F_NODE (1<<15) /* Node memory alloc*/ #define F_UDPCSUM (1<<16) /* Include UDP checksum */ +#define F_NO_TIMESTAMP (1<<17) /* Don't timestamp packets (default TS) */ /* Thread control flag bits */ #define T_STOP (1<<0) /* Stop run */ @@ -638,6 +639,9 @@ static int pktgen_if_show(struct seq_file *seq, void *v) if (pkt_dev->flags & F_UDPCSUM) seq_puts(seq, "UDPCSUM "); + if (pkt_dev->flags & F_NO_TIMESTAMP) + seq_puts(seq, "NO_TIMESTAMP "); + if (pkt_dev->flags & F_MPLS_RND) seq_puts(seq, "MPLS_RND "); @@ -1243,6 +1247,9 @@ static ssize_t pktgen_if_write(struct file *file, else if (strcmp(f, "!UDPCSUM") == 0) pkt_dev->flags &= ~F_UDPCSUM; + else if (strcmp(f, "NO_TIMESTAMP") == 0) + pkt_dev->flags |= F_NO_TIMESTAMP; + else { sprintf(pg_result, "Flag -:%s:- unknown\nAvailable flags, (prepend ! to un-set flag):\n%s", @@ -1251,6 +1258,7 @@ static ssize_t pktgen_if_write(struct file *file, "MACSRC_RND, MACDST_RND, TXSIZE_RND, IPV6, " "MPLS_RND, VID_RND, SVID_RND, FLOW_SEQ, " "QUEUE_MAP_RND, QUEUE_MAP_CPU, UDPCSUM, " + "NO_TIMESTAMP, " #ifdef CONFIG_XFRM "IPSEC, " #endif @@ -2685,9 +2693,14 @@ static void pktgen_finalize_skb(struct pktgen_dev *pkt_dev, struct sk_buff *skb, pgh->pgh_magic = htonl(PKTGEN_MAGIC); pgh->seq_num = htonl(pkt_dev->seq_num); - do_gettimeofday(×tamp); - pgh->tv_sec = htonl(timestamp.tv_sec); - pgh->tv_usec = htonl(timestamp.tv_usec); + if (pkt_dev->flags & F_NO_TIMESTAMP) { + pgh->tv_sec = 0; + pgh->tv_usec = 0; + } else { + do_gettimeofday(×tamp); + pgh->tv_sec = htonl(timestamp.tv_sec); + pgh->tv_usec = htonl(timestamp.tv_usec); + } } static struct sk_buff *pktgen_alloc_skb(struct net_device *dev,
Then testing the TX limits of the stack, then it is useful to be-able to disable the do_gettimeofday() timetamping on every packet. This implements a pktgen flag NO_TIMESTAMP which will disable this call to do_gettimeofday(). The performance change on (my system E5-2695) with skb_clone=0, goes from TX 2,423,751 pps to 2,567,165 pps with flag NO_TIMESTAMP. Thus, the cost of do_gettimeofday() or saving is approx 23 nanosec. Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com> --- net/core/pktgen.c | 19 ++++++++++++++++--- 1 files changed, 16 insertions(+), 3 deletions(-) -- 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