* Re: [PATCH 1/1] net/hyperv: Add flow control based on hi/low watermark
2012-03-27 22:23 ` [PATCH 1/1] " Haiyang Zhang
@ 2012-03-27 22:20 ` Greg KH
2012-03-27 22:25 ` Haiyang Zhang
0 siblings, 1 reply; 6+ messages in thread
From: Greg KH @ 2012-03-27 22:20 UTC (permalink / raw)
To: Haiyang Zhang; +Cc: davem, netdev, devel, olaf, linux-kernel
On Tue, Mar 27, 2012 at 03:23:07PM -0700, Haiyang Zhang wrote:
> --- a/include/linux/hyperv.h
> +++ b/include/linux/hyperv.h
> @@ -274,6 +274,35 @@ struct hv_ring_buffer_debug_info {
> u32 bytes_avail_towrite;
> };
>
> +/* Amount of space to write to */
> +#define BYTES_AVAIL_TO_WRITE(r, w, z) \
> + (((w) >= (r)) ? ((z) - ((w) - (r))) : ((r) - (w)))
> +
That's a very bad #define to use in a .h file, please do not do that.
> +
> +/*
> + *
> + * hv_get_ringbuffer_availbytes()
> + *
> + * Get number of bytes available to read and to write to
> + * for the specified ring buffer
> + */
> +extern inline void
> +hv_get_ringbuffer_availbytes(struct hv_ring_buffer_info *rbi,
> + u32 *read, u32 *write)
What does "extern inline" mean?
Please fix.
greg k-h
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 0/1] net/hyperv: Add flow control based on hi/low watermark
@ 2012-03-27 22:23 Haiyang Zhang
2012-03-27 22:23 ` [PATCH 1/1] " Haiyang Zhang
0 siblings, 1 reply; 6+ messages in thread
From: Haiyang Zhang @ 2012-03-27 22:23 UTC (permalink / raw)
To: davem, netdev; +Cc: haiyangz, kys, olaf, linux-kernel, devel
This patch is targeting 'net-next' tree (when re-opened).
Haiyang Zhang (1):
net/hyperv: Add flow control based on hi/low watermark
drivers/hv/ring_buffer.c | 31 -----------------------------
drivers/net/hyperv/netvsc.c | 41 +++++++++++++++++++++++++++++++++++---
drivers/net/hyperv/netvsc_drv.c | 6 ++++-
include/linux/hyperv.h | 29 +++++++++++++++++++++++++++
4 files changed, 71 insertions(+), 36 deletions(-)
--
1.7.4.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 1/1] net/hyperv: Add flow control based on hi/low watermark
2012-03-27 22:23 [PATCH 0/1] net/hyperv: Add flow control based on hi/low watermark Haiyang Zhang
@ 2012-03-27 22:23 ` Haiyang Zhang
2012-03-27 22:20 ` Greg KH
0 siblings, 1 reply; 6+ messages in thread
From: Haiyang Zhang @ 2012-03-27 22:23 UTC (permalink / raw)
To: davem, netdev; +Cc: haiyangz, kys, olaf, linux-kernel, devel
In the existing code, we only stop queue when the ringbuffer is full,
so the current packet has to be dropped or retried from upper layer.
This patch stops the tx queue when available ringbuffer is below
the low watermark. So the ringbuffer still has small amount of space
available for the current packet. This will reduce the overhead of
retries on sending.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
---
drivers/hv/ring_buffer.c | 31 -----------------------------
drivers/net/hyperv/netvsc.c | 41 +++++++++++++++++++++++++++++++++++---
drivers/net/hyperv/netvsc_drv.c | 6 ++++-
include/linux/hyperv.h | 29 +++++++++++++++++++++++++++
4 files changed, 71 insertions(+), 36 deletions(-)
diff --git a/drivers/hv/ring_buffer.c b/drivers/hv/ring_buffer.c
index 8af25a0..7233c88 100644
--- a/drivers/hv/ring_buffer.c
+++ b/drivers/hv/ring_buffer.c
@@ -30,37 +30,6 @@
#include "hyperv_vmbus.h"
-/* #defines */
-
-
-/* Amount of space to write to */
-#define BYTES_AVAIL_TO_WRITE(r, w, z) \
- ((w) >= (r)) ? ((z) - ((w) - (r))) : ((r) - (w))
-
-
-/*
- *
- * hv_get_ringbuffer_availbytes()
- *
- * Get number of bytes available to read and to write to
- * for the specified ring buffer
- */
-static inline void
-hv_get_ringbuffer_availbytes(struct hv_ring_buffer_info *rbi,
- u32 *read, u32 *write)
-{
- u32 read_loc, write_loc;
-
- smp_read_barrier_depends();
-
- /* Capture the read/write indices before they changed */
- read_loc = rbi->ring_buffer->read_index;
- write_loc = rbi->ring_buffer->write_index;
-
- *write = BYTES_AVAIL_TO_WRITE(read_loc, write_loc, rbi->ring_datasize);
- *read = rbi->ring_datasize - *write;
-}
-
/*
* hv_get_next_write_location()
*
diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index d025c83..8b91947 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -428,6 +428,24 @@ int netvsc_device_remove(struct hv_device *device)
return 0;
}
+
+#define RING_AVAIL_PERCENT_HIWATER 20
+#define RING_AVAIL_PERCENT_LOWATER 10
+
+/*
+ * Get the percentage of available bytes to write in the ring.
+ * The return value is in range from 0 to 100.
+ */
+static inline u32 hv_ringbuf_avail_percent(
+ struct hv_ring_buffer_info *ring_info)
+{
+ u32 avail_read, avail_write;
+
+ hv_get_ringbuffer_availbytes(ring_info, &avail_read, &avail_write);
+
+ return avail_write * 100 / ring_info->ring_datasize;
+}
+
static void netvsc_send_completion(struct hv_device *device,
struct vmpacket_descriptor *packet)
{
@@ -455,6 +473,8 @@ static void netvsc_send_completion(struct hv_device *device,
complete(&net_device->channel_init_wait);
} else if (nvsp_packet->hdr.msg_type ==
NVSP_MSG1_TYPE_SEND_RNDIS_PKT_COMPLETE) {
+ int num_outstanding_sends;
+
/* Get the send context */
nvsc_packet = (struct hv_netvsc_packet *)(unsigned long)
packet->trans_id;
@@ -463,10 +483,14 @@ static void netvsc_send_completion(struct hv_device *device,
nvsc_packet->completion.send.send_completion(
nvsc_packet->completion.send.send_completion_ctx);
- atomic_dec(&net_device->num_outstanding_sends);
+ num_outstanding_sends =
+ atomic_dec_return(&net_device->num_outstanding_sends);
- if (netif_queue_stopped(ndev) && !net_device->start_remove)
- netif_wake_queue(ndev);
+ if (netif_queue_stopped(ndev) && !net_device->start_remove &&
+ (hv_ringbuf_avail_percent(&device->channel->outbound)
+ > RING_AVAIL_PERCENT_HIWATER ||
+ num_outstanding_sends < 1))
+ netif_wake_queue(ndev);
} else {
netdev_err(ndev, "Unknown send completion packet type- "
"%d received!!\n", nvsp_packet->hdr.msg_type);
@@ -519,10 +543,19 @@ int netvsc_send(struct hv_device *device,
if (ret == 0) {
atomic_inc(&net_device->num_outstanding_sends);
+ if (hv_ringbuf_avail_percent(&device->channel->outbound) <
+ RING_AVAIL_PERCENT_LOWATER) {
+ netif_stop_queue(ndev);
+ if (atomic_read(&net_device->
+ num_outstanding_sends) < 1)
+ netif_wake_queue(ndev);
+ }
} else if (ret == -EAGAIN) {
netif_stop_queue(ndev);
- if (atomic_read(&net_device->num_outstanding_sends) < 1)
+ if (atomic_read(&net_device->num_outstanding_sends) < 1) {
netif_wake_queue(ndev);
+ ret = -ENOSPC;
+ }
} else {
netdev_err(ndev, "Unable to send packet %p ret %d\n",
packet, ret);
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index dd29478..a0cc127 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -224,9 +224,13 @@ static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
net->stats.tx_packets++;
} else {
kfree(packet);
+ if (ret != -EAGAIN) {
+ dev_kfree_skb_any(skb);
+ net->stats.tx_dropped++;
+ }
}
- return ret ? NETDEV_TX_BUSY : NETDEV_TX_OK;
+ return (ret == -EAGAIN) ? NETDEV_TX_BUSY : NETDEV_TX_OK;
}
/*
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index 5852545..2c366f0 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -274,6 +274,35 @@ struct hv_ring_buffer_debug_info {
u32 bytes_avail_towrite;
};
+/* Amount of space to write to */
+#define BYTES_AVAIL_TO_WRITE(r, w, z) \
+ (((w) >= (r)) ? ((z) - ((w) - (r))) : ((r) - (w)))
+
+
+/*
+ *
+ * hv_get_ringbuffer_availbytes()
+ *
+ * Get number of bytes available to read and to write to
+ * for the specified ring buffer
+ */
+extern inline void
+hv_get_ringbuffer_availbytes(struct hv_ring_buffer_info *rbi,
+ u32 *read, u32 *write)
+{
+ u32 read_loc, write_loc;
+
+ smp_read_barrier_depends();
+
+ /* Capture the read/write indices before they changed */
+ read_loc = rbi->ring_buffer->read_index;
+ write_loc = rbi->ring_buffer->write_index;
+
+ *write = BYTES_AVAIL_TO_WRITE(read_loc, write_loc, rbi->ring_datasize);
+ *read = rbi->ring_datasize - *write;
+}
+
+
/*
* We use the same version numbering for all Hyper-V modules.
*
--
1.7.4.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* RE: [PATCH 1/1] net/hyperv: Add flow control based on hi/low watermark
2012-03-27 22:20 ` Greg KH
@ 2012-03-27 22:25 ` Haiyang Zhang
0 siblings, 0 replies; 6+ messages in thread
From: Haiyang Zhang @ 2012-03-27 22:25 UTC (permalink / raw)
To: Greg KH; +Cc: davem, netdev, devel, olaf, linux-kernel
> -----Original Message-----
> From: Greg KH [mailto:gregkh@linuxfoundation.org]
> Sent: Tuesday, March 27, 2012 6:21 PM
> To: Haiyang Zhang
> Cc: davem@davemloft.net; netdev@vger.kernel.org;
> devel@linuxdriverproject.org; olaf@aepfle.de; linux-kernel@vger.kernel.org
> Subject: Re: [PATCH 1/1] net/hyperv: Add flow control based on hi/low
> watermark
>
> On Tue, Mar 27, 2012 at 03:23:07PM -0700, Haiyang Zhang wrote:
> > --- a/include/linux/hyperv.h
> > +++ b/include/linux/hyperv.h
> > @@ -274,6 +274,35 @@ struct hv_ring_buffer_debug_info {
> > u32 bytes_avail_towrite;
> > };
> >
> > +/* Amount of space to write to */
> > +#define BYTES_AVAIL_TO_WRITE(r, w, z) \
> > + (((w) >= (r)) ? ((z) - ((w) - (r))) : ((r) - (w)))
> > +
>
> That's a very bad #define to use in a .h file, please do not do that.
>
> > +
> > +/*
> > + *
> > + * hv_get_ringbuffer_availbytes()
> > + *
> > + * Get number of bytes available to read and to write to
> > + * for the specified ring buffer
> > + */
> > +extern inline void
> > +hv_get_ringbuffer_availbytes(struct hv_ring_buffer_info *rbi,
> > + u32 *read, u32 *write)
>
> What does "extern inline" mean?
>
> Please fix.
Will do.
Thanks,
- Haiyang
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 0/1] net/hyperv: Add flow control based on hi/low watermark
@ 2012-03-26 22:18 Haiyang Zhang
2012-03-26 22:12 ` David Miller
0 siblings, 1 reply; 6+ messages in thread
From: Haiyang Zhang @ 2012-03-26 22:18 UTC (permalink / raw)
To: davem, netdev; +Cc: haiyangz, kys, olaf, linux-kernel, devel
This patch is targeting 'net-next' tree.
Thanks to Stephen Hemminger <shemminger@vyatta.com> for his suggestions.
Haiyang Zhang (1):
net/hyperv: Add flow control based on hi/low watermark
drivers/hv/ring_buffer.c | 15 +++++++++++++++
drivers/net/hyperv/hyperv_net.h | 3 +++
drivers/net/hyperv/netvsc.c | 23 +++++++++++++++++++----
drivers/net/hyperv/netvsc_drv.c | 16 +++++++++++++++-
include/linux/hyperv.h | 3 +++
5 files changed, 55 insertions(+), 5 deletions(-)
--
1.7.4.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 0/1] net/hyperv: Add flow control based on hi/low watermark
2012-03-26 22:18 [PATCH 0/1] " Haiyang Zhang
@ 2012-03-26 22:12 ` David Miller
0 siblings, 0 replies; 6+ messages in thread
From: David Miller @ 2012-03-26 22:12 UTC (permalink / raw)
To: haiyangz; +Cc: netdev, kys, olaf, linux-kernel, devel
From: Haiyang Zhang <haiyangz@microsoft.com>
Date: Mon, 26 Mar 2012 15:18:23 -0700
> This patch is targeting 'net-next' tree.
The merge window is still open, and therefore the net-next tree is
not open yet.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2012-03-27 22:25 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-03-27 22:23 [PATCH 0/1] net/hyperv: Add flow control based on hi/low watermark Haiyang Zhang
2012-03-27 22:23 ` [PATCH 1/1] " Haiyang Zhang
2012-03-27 22:20 ` Greg KH
2012-03-27 22:25 ` Haiyang Zhang
-- strict thread matches above, loose matches on Subject: below --
2012-03-26 22:18 [PATCH 0/1] " Haiyang Zhang
2012-03-26 22:12 ` David Miller
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®