RE: [PATCH net-next, 1/1] hv_netvsc: call dump_rndis_message() only in netvsc debug mode
> -Original Message- > From: David Miller [mailto:da...@davemloft.net] > Sent: Tuesday, April 21, 2015 2:49 PM > To: Simon Xiao > Cc: KY Srinivasan; Haiyang Zhang; de...@linuxdriverproject.org; > net...@vger.kernel.org; linux-ker...@vger.kernel.org > Subject: Re: [PATCH net-next,1/1] hv_netvsc: call dump_rndis_message() only in > netvsc debug mode > > From: six...@microsoft.com > Date: Tue, 21 Apr 2015 15:58:05 -0700 > > > From: Simon Xiao > > > > Signed-off-by: Simon Xiao > > Reviewed-by: K. Y. Srinivasan > > Reviewed-by: Haiyang Zhang > > I just gave you feedback on this patch in response to your original > submission, > do not ignore it. Thanks for your feedback, David. In current netvsc driver, for each packet received, it will call dump_rndis_message() to try to dump the rndis packet information by netdev_dbg(). In non-debug mode, dump_rndis_message() will not dump anything but it still initialize some local variables and process the switch logic in the function of dump_rndis_message(), which is unnecessary, especially in high network throughput situation. My change is to have a run-time config flag to control the execution of dump_rndis_message() and avoid above unnecessary cost in non-debug mode. In the default case, it will be non-debug mode, and rndis_filter_receive() will not call dump_rndis_message() which saves the above extra cost for each packet received. ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
RE: [PATCH net-next,v2,1/1] hv_netvsc: introduce netif-msg into netvsc module
> -Original Message- > From: Joe Perches [mailto:j...@perches.com] > Sent: Friday, April 24, 2015 1:29 PM > To: Simon Xiao > Cc: KY Srinivasan; Haiyang Zhang; de...@linuxdriverproject.org; > net...@vger.kernel.org; linux-ker...@vger.kernel.org > Subject: Re: [PATCH net-next,v2,1/1] hv_netvsc: introduce netif-msg into > netvsc module > > On Fri, 2015-04-24 at 11:34 -0700, six...@microsoft.com wrote: > > From: Simon Xiao > > > > 1. Introduce netif-msg to netvsc to control debug logging output and > > keep msg_enable in netvsc_device_context so that it is kept > > persistently. > > 2. Only call dump_rndis_message() when NETIF_MSG_RX_ERR or above is > > specified in netvsc module debug param. > > In non-debug mode, in current code, dump_rndis_message() will not > dump > > anything but it still initialize some local variables and process the > > switch logic which is unnecessary, especially in high network > > throughput situation. > > [] > > > diff --git a/drivers/net/hyperv/netvsc_drv.c > > b/drivers/net/hyperv/netvsc_drv.c > [] > > @@ -888,6 +891,11 @@ static int netvsc_probe(struct hv_device *dev, > > > > net_device_ctx = netdev_priv(net); > > net_device_ctx->device_ctx = dev; > > + net_device_ctx->msg_enable = netif_msg_init(debug, default_msg); > > + if (netif_msg_probe(net_device_ctx)) > > + netdev_dbg(net, "netvsc msg_enable: %d", > > + net_device_ctx->msg_enable); > > Please use newlines to terminate formats. > > It helps prevent log content interleaving when multiple processes are > emitting output at the same time. > > This could be shortened to use netif_ like: > > netif_dbg(net_device_ctx, probe, net, "netvsc_msg_enable: %d\n", > net_device_ctx->msg_enable); > Thanks Joe. I would like to leave this to my next patch as there are some places else in netvsc (rndis_filter.c) have the same usage. I would like to fix them in one patch to make them consistent. Thanks, Simon ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
RE: [PATCH net-next,1/1] hv_netvsc: use per_cpu stats to calculate TX/RX data
> -Original Message-
> From: David Miller [mailto:da...@davemloft.net]
> Sent: Wednesday, May 13, 2015 9:52 PM
> To: Simon Xiao
> Cc: KY Srinivasan; Haiyang Zhang; de...@linuxdriverproject.org;
> net...@vger.kernel.org; linux-ker...@vger.kernel.org
> Subject: Re: [PATCH net-next,1/1] hv_netvsc: use per_cpu stats to calculate
> TX/RX data
>
> From: six...@microsoft.com
> Date: Tue, 12 May 2015 15:50:02 -0700
>
> > From: Simon Xiao
> >
> > Current code does not lock anything when calculating the TX and RX stats.
> > As a result, the RX and TX data reported by ifconfig are not accuracy
> > in a system with high network throughput and multiple CPUs (in my
> > test, RX/TX = 83% between 2 HyperV VM nodes which have 8 vCPUs and 40G
> Ethernet).
> >
> > This patch fixed the above issue by using per_cpu stats.
> > netvsc_get_stats64() summarizes TX and RX data by iterating over all
> > CPUs to get their respective stats.
> >
> > Signed-off-by: Simon Xiao
> > Reviewed-by: K. Y. Srinivasan
> > Reviewed-by: Haiyang Zhang
> > ---
> > drivers/net/hyperv/hyperv_net.h | 9 +
> > drivers/net/hyperv/netvsc_drv.c | 80
> > -
> > 2 files changed, 81 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/net/hyperv/hyperv_net.h
> > b/drivers/net/hyperv/hyperv_net.h index 41071d3..5a92b36 100644
> > --- a/drivers/net/hyperv/hyperv_net.h
> > +++ b/drivers/net/hyperv/hyperv_net.h
> > @@ -611,6 +611,12 @@ struct multi_send_data {
> > u32 count; /* counter of batched packets */ };
> >
> > +struct netvsc_stats {
> > + u64 packets;
> > + u64 bytes;
> > + struct u64_stats_sync s_sync;
> > +};
> > +
> > /* The context of the netvsc device */ struct net_device_context {
> > /* point back to our device context */ @@ -618,6 +624,9 @@ struct
> > net_device_context {
> > struct delayed_work dwork;
> > struct work_struct work;
> > u32 msg_enable; /* debug level */
> > +
> > + struct netvsc_stats __percpu *tx_stats;
> > + struct netvsc_stats __percpu *rx_stats;
> > };
> >
> > /* Per netvsc device */
> > diff --git a/drivers/net/hyperv/netvsc_drv.c
> > b/drivers/net/hyperv/netvsc_drv.c index 5993c7e..310b902 100644
> > --- a/drivers/net/hyperv/netvsc_drv.c
> > +++ b/drivers/net/hyperv/netvsc_drv.c
> > @@ -391,7 +391,7 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct
> net_device *net)
> > u32 skb_length;
> > u32 pkt_sz;
> > struct hv_page_buffer page_buf[MAX_PAGE_BUFFER_COUNT];
> > -
> > + struct netvsc_stats *tx_stats =
> > +this_cpu_ptr(net_device_ctx->tx_stats);
> >
> > /* We will atmost need two pages to describe the rndis
> > * header. We can only transmit MAX_PAGE_BUFFER_COUNT number
> @@
> > -580,8 +580,10 @@ do_send:
> >
> > drop:
> > if (ret == 0) {
> > - net->stats.tx_bytes += skb_length;
> > - net->stats.tx_packets++;
> > + u64_stats_update_begin(&tx_stats->s_sync);
> > + tx_stats->packets++;
> > + tx_stats->bytes += skb_length;
> > + u64_stats_update_end(&tx_stats->s_sync);
> > } else {
> > if (ret != -EAGAIN) {
> > dev_kfree_skb_any(skb);
> > @@ -644,13 +646,17 @@ int netvsc_recv_callback(struct hv_device
> *device_obj,
> > struct ndis_tcp_ip_checksum_info *csum_info)
> {
> > struct net_device *net;
> > + struct net_device_context *net_device_ctx;
> > struct sk_buff *skb;
> > + struct netvsc_stats *rx_stats;
> >
> > net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev;
> > if (!net || net->reg_state != NETREG_REGISTERED) {
> > packet->status = NVSP_STAT_FAIL;
> > return 0;
> > }
> > + net_device_ctx = netdev_priv(net);
> > + rx_stats = this_cpu_ptr(net_device_ctx->rx_stats);
> >
> > /* Allocate a skb - TODO direct I/O to pages? */
> > skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen); @@
> > -686,8 +692,10 @@ int netvsc_recv_callback(struct hv_device *device_obj,
> > skb_record_rx_queue(skb, packet->channel->
> > offermsg.offer.sub_channel_index);
> >
> > - net->stats.rx_packets++;
> > - net->stats.rx_bytes += packet->total_data_buflen;
> > + u64_stats_update_begin(&rx
[PATCH net-next, v2, 1/1] hv_netvsc: use per_cpu stats to calculate TX/RX data
Current code does not lock anything when calculating the TX and RX stats.
As a result, the RX and TX data reported by ifconfig are not accuracy in a
system with high network throughput and multiple CPUs (in my test,
RX/TX = 83% between 2 HyperV VM nodes which have 8 vCPUs and 40G Ethernet).
This patch fixed the above issue by using per_cpu stats.
netvsc_get_stats64() summarizes TX and RX data by iterating over all CPUs
to get their respective stats.
This v2 patch addressed David's comments on the cleanup path when
netdev_alloc_pcpu_stats() failed.
Signed-off-by: Simon Xiao
Reviewed-by: K. Y. Srinivasan
Reviewed-by: Haiyang Zhang
---
drivers/net/hyperv/hyperv_net.h | 9 +
drivers/net/hyperv/netvsc_drv.c | 85 +
2 files changed, 86 insertions(+), 8 deletions(-)
diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index 41071d3..5a92b36 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -611,6 +611,12 @@ struct multi_send_data {
u32 count; /* counter of batched packets */
};
+struct netvsc_stats {
+ u64 packets;
+ u64 bytes;
+ struct u64_stats_sync s_sync;
+};
+
/* The context of the netvsc device */
struct net_device_context {
/* point back to our device context */
@@ -618,6 +624,9 @@ struct net_device_context {
struct delayed_work dwork;
struct work_struct work;
u32 msg_enable; /* debug level */
+
+ struct netvsc_stats __percpu *tx_stats;
+ struct netvsc_stats __percpu *rx_stats;
};
/* Per netvsc device */
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 5993c7e..1570628 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -391,7 +391,7 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct
net_device *net)
u32 skb_length;
u32 pkt_sz;
struct hv_page_buffer page_buf[MAX_PAGE_BUFFER_COUNT];
-
+ struct netvsc_stats *tx_stats = this_cpu_ptr(net_device_ctx->tx_stats);
/* We will atmost need two pages to describe the rndis
* header. We can only transmit MAX_PAGE_BUFFER_COUNT number
@@ -580,8 +580,10 @@ do_send:
drop:
if (ret == 0) {
- net->stats.tx_bytes += skb_length;
- net->stats.tx_packets++;
+ u64_stats_update_begin(&tx_stats->s_sync);
+ tx_stats->packets++;
+ tx_stats->bytes += skb_length;
+ u64_stats_update_end(&tx_stats->s_sync);
} else {
if (ret != -EAGAIN) {
dev_kfree_skb_any(skb);
@@ -644,13 +646,17 @@ int netvsc_recv_callback(struct hv_device *device_obj,
struct ndis_tcp_ip_checksum_info *csum_info)
{
struct net_device *net;
+ struct net_device_context *net_device_ctx;
struct sk_buff *skb;
+ struct netvsc_stats *rx_stats;
net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev;
if (!net || net->reg_state != NETREG_REGISTERED) {
packet->status = NVSP_STAT_FAIL;
return 0;
}
+ net_device_ctx = netdev_priv(net);
+ rx_stats = this_cpu_ptr(net_device_ctx->rx_stats);
/* Allocate a skb - TODO direct I/O to pages? */
skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
@@ -686,8 +692,10 @@ int netvsc_recv_callback(struct hv_device *device_obj,
skb_record_rx_queue(skb, packet->channel->
offermsg.offer.sub_channel_index);
- net->stats.rx_packets++;
- net->stats.rx_bytes += packet->total_data_buflen;
+ u64_stats_update_begin(&rx_stats->s_sync);
+ rx_stats->packets++;
+ rx_stats->bytes += packet->total_data_buflen;
+ u64_stats_update_end(&rx_stats->s_sync);
/*
* Pass the skb back up. Network stack will deallocate the skb when it
@@ -753,6 +761,46 @@ static int netvsc_change_mtu(struct net_device *ndev, int
mtu)
return 0;
}
+static struct rtnl_link_stats64 *netvsc_get_stats64(struct net_device *net,
+ struct rtnl_link_stats64 *t)
+{
+ struct net_device_context *ndev_ctx = netdev_priv(net);
+ int cpu;
+
+ for_each_possible_cpu(cpu) {
+ struct netvsc_stats *tx_stats = per_cpu_ptr(ndev_ctx->tx_stats,
+ cpu);
+ struct netvsc_stats *rx_stats = per_cpu_ptr(ndev_ctx->rx_stats,
+ cpu);
+ u64 tx_packets, tx_bytes, rx_packets, rx_bytes;
+ unsigned int start;
+
+ do {
+ start = u64_stats_fetch_begin_irq(&tx_stats->s_sync);
+
[PATCH net-next, 1/1] hv_netvsc: change member name of struct netvsc_stats
Currently the struct netvsc_stats has a member s_sync
of type u64_stats_sync.
This definition will break kernel build as the macro
netdev_alloc_pcpu_stats requires this member name to be syncp.
(see netdev_alloc_pcpu_stats definition in ./include/linux/netdevice.h)
This patch changes netvsc_stats's member name from s_sync to syncp to fix
the build break.
Signed-off-by: Simon Xiao
---
drivers/net/hyperv/hyperv_net.h | 2 +-
drivers/net/hyperv/netvsc_drv.c | 16
2 files changed, 9 insertions(+), 9 deletions(-)
diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index 5a92b36..ddcc7f8 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -614,7 +614,7 @@ struct multi_send_data {
struct netvsc_stats {
u64 packets;
u64 bytes;
- struct u64_stats_sync s_sync;
+ struct u64_stats_sync syncp;
};
/* The context of the netvsc device */
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 0c858724..d9c88bc 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -580,10 +580,10 @@ do_send:
drop:
if (ret == 0) {
- u64_stats_update_begin(&tx_stats->s_sync);
+ u64_stats_update_begin(&tx_stats->syncp);
tx_stats->packets++;
tx_stats->bytes += skb_length;
- u64_stats_update_end(&tx_stats->s_sync);
+ u64_stats_update_end(&tx_stats->syncp);
} else {
if (ret != -EAGAIN) {
dev_kfree_skb_any(skb);
@@ -692,10 +692,10 @@ int netvsc_recv_callback(struct hv_device *device_obj,
skb_record_rx_queue(skb, packet->channel->
offermsg.offer.sub_channel_index);
- u64_stats_update_begin(&rx_stats->s_sync);
+ u64_stats_update_begin(&rx_stats->syncp);
rx_stats->packets++;
rx_stats->bytes += packet->total_data_buflen;
- u64_stats_update_end(&rx_stats->s_sync);
+ u64_stats_update_end(&rx_stats->syncp);
/*
* Pass the skb back up. Network stack will deallocate the skb when it
@@ -776,16 +776,16 @@ static struct rtnl_link_stats64
*netvsc_get_stats64(struct net_device *net,
unsigned int start;
do {
- start = u64_stats_fetch_begin_irq(&tx_stats->s_sync);
+ start = u64_stats_fetch_begin_irq(&tx_stats->syncp);
tx_packets = tx_stats->packets;
tx_bytes = tx_stats->bytes;
- } while (u64_stats_fetch_retry_irq(&tx_stats->s_sync, start));
+ } while (u64_stats_fetch_retry_irq(&tx_stats->syncp, start));
do {
- start = u64_stats_fetch_begin_irq(&rx_stats->s_sync);
+ start = u64_stats_fetch_begin_irq(&rx_stats->syncp);
rx_packets = rx_stats->packets;
rx_bytes = rx_stats->bytes;
- } while (u64_stats_fetch_retry_irq(&rx_stats->s_sync, start));
+ } while (u64_stats_fetch_retry_irq(&rx_stats->syncp, start));
t->tx_bytes += tx_bytes;
t->tx_packets += tx_packets;
--
1.8.5.2
___
devel mailing list
de...@linuxdriverproject.org
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
RE: linux-next network throughput performance regression
Thanks Eric to provide the data. I am looping Tom (as I am looking into his recent patches) and Olaf (from Suse). So, if I understand it correctly, you are running netperf with single TCP connection, and you got ~26Gbps initially and got ~30Gbps after turning the tx-usecs and tx-frames. Do you have a baseline on your environment for the best/max/or peak throughput? Again, in my environment (SLES bare metal), if use SLES12 default kernel as a baseline, we can see significant performance drop (10% ~ 50%) on latest linux-next kernel. Absolutely I will try the same test on net-next soon and update the results to here later. Thanks, Simon > -Original Message- > From: Eric Dumazet [mailto:eric.duma...@gmail.com] > Sent: Saturday, November 7, 2015 11:50 AM > To: David Ahern > Cc: Simon Xiao ; de...@linuxdriverproject.org; > net...@vger.kernel.org; linux-ker...@vger.kernel.org; David Miller > ; KY Srinivasan ; Haiyang > Zhang > Subject: Re: linux-next network throughput performance regression > > On Sat, 2015-11-07 at 11:35 -0800, Eric Dumazet wrote: > > On Fri, 2015-11-06 at 14:30 -0700, David Ahern wrote: > > > On 11/6/15 2:18 PM, Simon Xiao wrote: > > > > The .config file used to build linux-next kernel is attached to this > > > > mail. > > > > > > Thanks. > > > > > > Failed to notice this on the first response; my brain filled in. Why > > > linux-next tree? Can you try net-next which is more relevant for > > > this mailing list, post the top commit id and config file used? > > > > Throughput on a single TCP flow for a 40G NIC can be tricky to tune. > > > > Make sure IRQ are properly setup/balanced, as I know that IRQ names > > were changed recently and your scripts might have not noticed... > > > > Also "ethtool -c eth0" might show very different interrupt coalescing > > params ? > > > > I too have a Mellanox 40Gb in my lab and saw no difference in > > performance with recent kernels. > > > > Of course, a simple "perf record -a -g sleep 4 ; perf report" might > > point to some obvious issue. Like unexpected segmentation in case of > > forwarding... > > > > > > I did a test with current net tree on both sender and receiver > > lpaa23:~# ./netperf -H 10.246.7.152 > MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to > 10.246.7.152 () port 0 AF_INET > Recv SendSend > Socket Socket Message Elapsed > Size SizeSize Time Throughput > bytes bytes bytessecs.10^6bits/sec > > 87380 16384 1638410.0026864.98 > lpaa23:~# ethtool -c eth1 > Coalesce parameters for eth1: > Adaptive RX: on TX: off > stats-block-usecs: 0 > sample-interval: 0 > pkt-rate-low: 40 > pkt-rate-high: 45 > > rx-usecs: 16 > rx-frames: 44 > rx-usecs-irq: 0 > rx-frames-irq: 0 > > tx-usecs: 16 > tx-frames: 16 > tx-usecs-irq: 0 > tx-frames-irq: 256 > > rx-usecs-low: 0 > rx-frame-low: 0 > tx-usecs-low: 0 > tx-frame-low: 0 > > rx-usecs-high: 128 > rx-frame-high: 0 > tx-usecs-high: 0 > tx-frame-high: 0 > > lpaa23:~# ethtool -C eth1 tx-usecs 4 tx-frames 4 > lpaa23:~# ./netperf -H > 10.246.7.152 MIGRATED TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 > AF_INET to > 10.246.7.152 () port 0 AF_INET > Recv SendSend > Socket Socket Message Elapsed > Size SizeSize Time Throughput > bytes bytes bytessecs.10^6bits/sec > > 87380 16384 1638410.0030206.27 > ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
[PATCH net-next] hv_netvsc: Add feature flags NETIF_F_IPV6_CSUM and NETIF_F_TSO6 for netvsc
1. Adding NETIF_F_IPV6_CSUM and NETIF_F_TSO6 feature flags which are supported by Hyper-V platform. 2. Cleanup the coding style of flag assignment by using macro. Signed-off-by: Simon Xiao Reviewed-by: K. Y. Srinivasan Reviewed-by: Haiyang Zhang --- drivers/net/hyperv/netvsc_drv.c | 12 1 file changed, 8 insertions(+), 4 deletions(-) diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c index 1d3a665..0cde741 100644 --- a/drivers/net/hyperv/netvsc_drv.c +++ b/drivers/net/hyperv/netvsc_drv.c @@ -43,6 +43,12 @@ #define RING_SIZE_MIN 64 #define LINKCHANGE_INT (2 * HZ) +#define NETVSC_HW_FEATURES (NETIF_F_RXCSUM | \ +NETIF_F_SG | \ +NETIF_F_TSO | \ +NETIF_F_TSO6 | \ +NETIF_F_IP_CSUM | \ +NETIF_F_IPV6_CSUM) static int ring_size = 128; module_param(ring_size, int, S_IRUGO); MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)"); @@ -1081,10 +1087,8 @@ static int netvsc_probe(struct hv_device *dev, net->netdev_ops = &device_ops; - net->hw_features = NETIF_F_RXCSUM | NETIF_F_SG | NETIF_F_IP_CSUM | - NETIF_F_TSO; - net->features = NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_SG | NETIF_F_RXCSUM | - NETIF_F_IP_CSUM | NETIF_F_TSO; + net->hw_features = NETVSC_HW_FEATURES; + net->features = NETVSC_HW_FEATURES | NETIF_F_HW_VLAN_CTAG_TX; net->ethtool_ops = ðtool_ops; SET_NETDEV_DEV(net, &dev->device); -- 2.5.0 ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
[PATCH V2 net-next] hv_netvsc: cleanup netdev feature flags for netvsc
1. Adding NETIF_F_TSO6 feature flag; 2. Adding NETIF_F_HW_CSUM. NETIF_F_IPV6_CSUM and NETIF_F_IP_CSUM are being deprecated; 3. Cleanup the coding style of flag assignment by using macro. Signed-off-by: Simon Xiao Reviewed-by: K. Y. Srinivasan Reviewed-by: Haiyang Zhang --- drivers/net/hyperv/netvsc_drv.c | 11 +++ 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c index 1d3a665..c72e5b8 100644 --- a/drivers/net/hyperv/netvsc_drv.c +++ b/drivers/net/hyperv/netvsc_drv.c @@ -43,6 +43,11 @@ #define RING_SIZE_MIN 64 #define LINKCHANGE_INT (2 * HZ) +#define NETVSC_HW_FEATURES (NETIF_F_RXCSUM | \ +NETIF_F_SG | \ +NETIF_F_TSO | \ +NETIF_F_TSO6 | \ +NETIF_F_HW_CSUM) static int ring_size = 128; module_param(ring_size, int, S_IRUGO); MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)"); @@ -1081,10 +1086,8 @@ static int netvsc_probe(struct hv_device *dev, net->netdev_ops = &device_ops; - net->hw_features = NETIF_F_RXCSUM | NETIF_F_SG | NETIF_F_IP_CSUM | - NETIF_F_TSO; - net->features = NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_SG | NETIF_F_RXCSUM | - NETIF_F_IP_CSUM | NETIF_F_TSO; + net->hw_features = NETVSC_HW_FEATURES; + net->features = NETVSC_HW_FEATURES | NETIF_F_HW_VLAN_CTAG_TX; net->ethtool_ops = ðtool_ops; SET_NETDEV_DEV(net, &dev->device); -- 2.5.0 ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
[PATCH net-next] hv_netvsc: add software transmit timestamp support
Enable skb_tx_timestamp in hyperv netvsc.
Signed-off-by: Simon Xiao
Reviewed-by: K. Y. Srinivasan
Reviewed-by: Haiyang Zhang
---
drivers/net/hyperv/netvsc_drv.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index c72e5b8..202e2b1 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -550,6 +550,8 @@ do_send:
packet->page_buf_cnt = init_page_array(rndis_msg, rndis_msg_size,
skb, packet, &pb);
+ /* timestamp packet in software */
+ skb_tx_timestamp(skb);
ret = netvsc_send(net_device_ctx->device_ctx, packet,
rndis_msg, &pb, skb);
@@ -920,6 +922,7 @@ static const struct ethtool_ops ethtool_ops = {
.get_link = ethtool_op_get_link,
.get_channels = netvsc_get_channels,
.set_channels = netvsc_set_channels,
+ .get_ts_info= ethtool_op_get_ts_info,
};
static const struct net_device_ops device_ops = {
--
2.5.0
___
devel mailing list
de...@linuxdriverproject.org
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
[PATCH net-next] hv_netvsc: add ethtool support for set and get of settings
This patch allows the user to set and retrieve speed and duplex of the
hv_netvsc device via ethtool.
Example:
$ ethtool eth0
Settings for eth0:
...
Speed: Unknown!
Duplex: Unknown! (255)
...
$ ethtool -s eth0 speed 1000 duplex full
$ ethtool eth0
Settings for eth0:
...
Speed: 1000Mb/s
Duplex: Full
...
This is based on patches by Roopa Prabhu and Nikolay Aleksandrov.
Signed-off-by: Simon Xiao
---
drivers/net/hyperv/hyperv_net.h | 4 +++
drivers/net/hyperv/netvsc_drv.c | 56 +
2 files changed, 60 insertions(+)
diff --git a/drivers/net/hyperv/hyperv_net.h b/drivers/net/hyperv/hyperv_net.h
index fcb92c0..b4c6878 100644
--- a/drivers/net/hyperv/hyperv_net.h
+++ b/drivers/net/hyperv/hyperv_net.h
@@ -658,6 +658,10 @@ struct net_device_context {
struct netvsc_stats __percpu *tx_stats;
struct netvsc_stats __percpu *rx_stats;
+
+ /* Ethtool settings */
+ u8 duplex;
+ u32 speed;
};
/* Per netvsc device */
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 202e2b1..e703b9a 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -799,6 +799,58 @@ static int netvsc_set_channels(struct net_device *net,
goto do_set;
}
+static bool netvsc_validate_ethtool_ss_cmd(const struct ethtool_cmd *cmd)
+{
+ struct ethtool_cmd diff1 = *cmd;
+ struct ethtool_cmd diff2 = {};
+
+ ethtool_cmd_speed_set(&diff1, 0);
+ diff1.duplex = 0;
+ /* advertising and cmd are usually set */
+ diff1.advertising = 0;
+ diff1.cmd = 0;
+ /* We set port to PORT_OTHER */
+ diff2.port = PORT_OTHER;
+
+ return !memcmp(&diff1, &diff2, sizeof(diff1));
+}
+
+static void netvsc_init_settings(struct net_device *dev)
+{
+ struct net_device_context *ndc = netdev_priv(dev);
+
+ ndc->speed = SPEED_UNKNOWN;
+ ndc->duplex = DUPLEX_UNKNOWN;
+}
+
+static int netvsc_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
+{
+ struct net_device_context *ndc = netdev_priv(dev);
+
+ ethtool_cmd_speed_set(cmd, ndc->speed);
+ cmd->duplex = ndc->duplex;
+ cmd->port = PORT_OTHER;
+
+ return 0;
+}
+
+static int netvsc_set_settings(struct net_device *dev, struct ethtool_cmd *cmd)
+{
+ struct net_device_context *ndc = netdev_priv(dev);
+ u32 speed;
+
+ speed = ethtool_cmd_speed(cmd);
+ if (!ethtool_validate_speed(speed) ||
+ !ethtool_validate_duplex(cmd->duplex) ||
+ !netvsc_validate_ethtool_ss_cmd(cmd))
+ return -EINVAL;
+
+ ndc->speed = speed;
+ ndc->duplex = cmd->duplex;
+
+ return 0;
+}
+
static int netvsc_change_mtu(struct net_device *ndev, int mtu)
{
struct net_device_context *ndevctx = netdev_priv(ndev);
@@ -923,6 +975,8 @@ static const struct ethtool_ops ethtool_ops = {
.get_channels = netvsc_get_channels,
.set_channels = netvsc_set_channels,
.get_ts_info= ethtool_op_get_ts_info,
+ .get_settings = netvsc_get_settings,
+ .set_settings = netvsc_set_settings,
};
static const struct net_device_ops device_ops = {
@@ -1112,6 +1166,8 @@ static int netvsc_probe(struct hv_device *dev,
netif_set_real_num_tx_queues(net, nvdev->num_chn);
netif_set_real_num_rx_queues(net, nvdev->num_chn);
+ netvsc_init_settings(net);
+
ret = register_netdev(net);
if (ret != 0) {
pr_err("Unable to register netdev.\n");
--
2.5.0
___
devel mailing list
de...@linuxdriverproject.org
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
RE: [PATCH 1/1] hv_netvsc: fix a netvsc stats typo
Please ignore this patch. I will resubmit it to net-next. > -Original Message- > From: Simon Xiao [mailto:six...@microsoft.com] > Sent: Tuesday, February 7, 2017 10:03 AM > To: KY Srinivasan ; Haiyang Zhang > ; Stephen Hemminger > ; de...@linuxdriverproject.org; > net...@vger.kernel.org; linux-ker...@vger.kernel.org > Cc: Simon Xiao > Subject: [PATCH 1/1] hv_netvsc: fix a netvsc stats typo > > [This sender failed our fraud detection checks and may not be who they > appear to be. Learn about spoofing at http://aka.ms/LearnAboutSpoofing] > > Now, return the correct tx_errors stats in netvsc. > > Signed-off-by: Simon Xiao > Reviewed-by: Haiyang Zhang > --- > drivers/net/hyperv/netvsc_drv.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/drivers/net/hyperv/netvsc_drv.c > b/drivers/net/hyperv/netvsc_drv.c > index 72b0c1f..725ac19 100644 > --- a/drivers/net/hyperv/netvsc_drv.c > +++ b/drivers/net/hyperv/netvsc_drv.c > @@ -920,7 +920,7 @@ static void netvsc_get_stats64(struct net_device *net, > } > > t->tx_dropped = net->stats.tx_dropped; > - t->tx_errors= net->stats.tx_dropped; > + t->tx_errors= net->stats.tx_errors; > > t->rx_dropped = net->stats.rx_dropped; > t->rx_errors= net->stats.rx_errors; > -- > 2.7.4 ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
linux-next network throughput performance regression
I compared the network throughput performance on SLES12 bare metal servers, between SLES12 default kernel and latest linux-next (2015-11-05) kernel, based on the test results, I suspect there is a network regression exists on Linux-Next over the 40G Ethernet network: a) iperf3 reports 50% performance drop with single TCP stream on latest linux-next; b) iperf3 reports 10% ~ 30% performance drop with 2 to 128 TCP streams on latest linux-next; Another throughput benchmarking tool (ntttcp-for-linux) test result is also listed at the end of the email for reference. Server configuration: -- Two servers (one client and one server, cross linked by 40G Ethernet), which have: a) CPU: Intel(R) Xeon(R) CPU E5-2667 v3 @ 3.20GHz, 2 sockets, 16 CPUs, cache size : 20480 KB b) Memory: 64 GB c) Ethernet controller: Mellanox Technologies MT27520 Family [ConnectX-3 Pro], 40G Ethernet, default driver Test with iperf3: -- iperf3: https://github.com/esnet/iperf a) SLES12 default kernel, network throughput tested by iperf3: Test Connections1 2 4 8 16 32 64 128 Throughput (G bps) 36.737.337.637.737.737.737.7 25.7 b) SLES12 + Linux-Next 20151105, network throughput tested by iperf3: Test Connections1 2 4 8 16 32 64 128 Throughput (G bps) 18.232.234.632.827.632.027.0 21.3 Percentage dropped -50%-14%-8% -13%-27%-15%-28% -17% Test with ntttcp-for-linux: -- ntttcp-for-linux: https://github.com/Microsoft/ntttcp-for-linux a) SLES12 default kernel, network throughput tested by ntttcp-for-linux: Test Connections1 2 4 8 16 32 64 128 256 512 Throughput (Gbps) 36.19 37.29 37.67 37.68 37.737.72 37.74 37.76 37.81 37.9 b) SLES12 + Linux-Next 20151105, network throughput tested by ntttcp-for-linux: Test Connections1 2 4 8 16 32 64 128 256 512 Throughput (Gbps) 28.12 34.01 37.636.53 32.94 33.07 33.63 33.44 33.83 34.42 Percentage dropped -22%-9% 0% -3% -13%-12%-11% -11%-11%-9% ___ devel mailing list de...@linuxdriverproject.org http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
