Hello Jason,
Good work. Do you have local guest to guest results?
Thanks
Shirley
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, 2012-05-17 at 17:20 +0800, Liu Ping Fan wrote:
> Currently, the guest can not know the NUMA info of the vcpu, which
> will
> result in performance drawback.
>
> This is the discovered and experiment by
> Shirley Ma
> Krishna Kumar
>
On Mon, 2012-05-14 at 21:39 +0300, Michael S. Tsirkin wrote:
> > Hello Mike,
> >
> > Have you tested this patch? I think the difference between macvtap
> and
> > tap is tap forwarding the packet to bridge. The zerocopy is disabled
> in
> > this case.
> >
> > Shirley
>
> Testing in progress, but
On Mon, 2012-05-14 at 22:31 +0300, Michael S. Tsirkin wrote:
> On Mon, May 14, 2012 at 09:21:47PM +0200, Eric Dumazet wrote:
> > On Mon, 2012-05-14 at 22:14 +0300, Michael S. Tsirkin wrote:
> >
> > > They seem to be in net-next, or did I miss something?
> >
> > Yes, you re-introduce in this patch
On Sun, 2012-05-13 at 18:52 +0300, Michael S. Tsirkin wrote:
> Let vhost-net utilize zero copy tx when the experimental
> zero copy mode is enabled and when used with tun. This works on
> top of the patchset 'copy aside frags with destructors' that I posted
> previously. This is not using tcp so d
On Thu, 2012-04-05 at 08:22 -0700, Shirley Ma wrote:
> Haven't had time to focus on single stream result yet.
I forgot to mention that if I switch the vhost scheduling to per-device
based from per vq based, this minor single stream test regression will
be gone. However the improv
On Thu, 2012-04-05 at 15:28 +0300, Michael S. Tsirkin wrote:
> On Tue, Mar 27, 2012 at 10:43:03AM -0700, Shirley Ma wrote:
> > On Tue, 2012-03-27 at 18:09 +0800, Jason Wang wrote:
> > > Hi:
> > >
> > > Thanks for the work and it looks very reasonable, some
On Tue, 2012-03-27 at 18:09 +0800, Jason Wang wrote:
> Hi:
>
> Thanks for the work and it looks very reasonable, some questions
> below.
>
> On 03/23/2012 07:48 AM, Shirley Ma wrote:
> > Sorry for being late to submit this patch. I have spent lots of time
> > tr
Resubmit it with the right format.
Signed-off-by: Shirley Ma
Signed-off-by: Krishna Kumar
Tested-by: Tom Lendacky
---
drivers/vhost/net.c | 26 ++-
drivers/vhost/vhost.c| 300 --
drivers/vhost/vhost.h| 16
Signed-off-by: Shirley Ma
Signed-off-by: Krishna Kumar
Tested-by: Tom Lendacky
---
drivers/vhost/net.c | 26 ++-
drivers/vhost/vhost.c| 300 --
drivers/vhost/vhost.h| 16 ++-
3 files changed, 243 insertions
n.
If the host scheduler ensures that the VM's vcpus are not scheduled to
another socket (i.e. cpu mask the vcpus on same socket) then the
performance will be better.
Signed-off-by: Shirley Ma
Signed-off-by: Krishna Kumar
Tested-by: Tom Lendacky
---
drivers/vhost/net.c |
On Fri, 2011-07-29 at 16:58 -0700, Mike Waychison wrote:
> On Fri, Jul 29, 2011 at 3:55 PM, Shirley Ma
> wrote:
> > Resubmit it with a typo fix.
> >
> > Signed-off-by: Shirley Ma
> > ---
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virt
Resubmit it with a typo fix.
Signed-off-by: Shirley Ma
---
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 0c7321c..c8201d4 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -429,6 +429,22 @@ static int add_recvbuf_mergeable(struct virtnet_info *vi
To even the latency, refill buffer right after being used.
Sign-off-by: Shirley Ma
---
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 0c7321c..c8201d4 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -429,6 +429,22 @@ static int
On Tue, 2011-07-12 at 01:04 +0300, Michael S. Tsirkin wrote:
> So the following should do it, on top of Shirleys's patch, I think.
> I'm
> a bit not sure about using vq->upend_idx - vq->done_idx to check the
> number of outstanding DMA, Shirley, what do you think?
Yes, you can use this to track #
Hello Dave,
This patch clears tx zero-copy flag as needed.
Sign-off-by: Shirley Ma
---
net/core/skbuff.c |3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index a9577a2..d220119 100644
--- a/net/core/skbuff.c
+++ b/net/core
On Thu, 2011-07-07 at 04:01 -0700, David Miller wrote:
> From: Shirley Ma
> Date: Wed, 06 Jul 2011 15:22:12 -0700
>
> > + while (head) {
> > + put_page(head);
> > + head = (
On Wed, 2011-07-06 at 17:01 -0600, Zan Lynx wrote:
> On 7/6/2011 4:22 PM, Shirley Ma wrote:
> > This patch adds userspace buffers support in skb shared info. A new
> > struct skb_ubuf_info is needed to maintain the userspace buffers
> > argument and index, a callback is use
This patch maintains the outstanding userspace buffers in the
sequence it is delivered to vhost. The outstanding userspace buffers
will be marked as done once the lower device buffers DMA has finished.
This is monitored through last reference of kfree_skb callback. Two
buffer index are used for
Only 128 bytes is copied, the rest of data is DMA mapped directly from
userspace.
Signed-off-by: Shirley Ma
---
drivers/net/macvtap.c | 132
1 files changed, 121 insertions(+), 11 deletions(-)
diff --git a/drivers/net/macvtap.c b/drivers/net
added for zero-copy buffer check.
Signed-off-by: Shirley Ma
---
include/linux/skbuff.h | 16 ++
net/core/skbuff.c | 79 +++-
2 files changed, 94 insertions(+), 1 deletions(-)
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
Signed-off-by: Shirley Ma
---
include/net/sock.h |1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index ae56da6..396f735 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -563,6 +563,7 @@ enum sock_flags
This patchset add supports for TX zero-copy between guest and host
kernel through vhost. It significantly reduces CPU utilization on the
local host on which the guest is located (It reduced about 50% CPU usage
for single stream test on the host, while 4K message size BW has
increased about 50%). Th
On Wed, 2011-06-29 at 12:13 +0300, Michael S. Tsirkin wrote:
> On Sat, May 28, 2011 at 12:34:27PM -0700, Shirley Ma wrote:
> > Hello Michael,
> >
> > In order to use wait for completion in shutting down, seems to me
> > another work thread is needed to call vhost_ze
On Wed, 2011-06-29 at 12:13 +0300, Michael S. Tsirkin wrote:
> Assuming you mean vhost_zerocopy_signal_used, here's how I would do
> it:
> add a kref and a completion, signal completion in kref_put
> callback, when backend is set - kref_get, on cleanup,
> kref_put and then wait_for_completion_inter
I have submitted this patchset from Version 1 to Version 7 already in
the past few months.
Here is the link to the patchset:
http://lists.openwall.net/netdev/2011/05/28/
I am working on V8 now.
Thanks
Shirley
On Tue, 2011-06-28 at 16:35 -0700, David Miller wrote:
> From: Shirley Ma
>
On Mon, 2011-06-27 at 15:54 -0700, David Miller wrote:
> From: Shirley Ma
> Date: Mon, 27 Jun 2011 08:45:10 -0700
>
> > To support skb zero-copy, a pointer is needed to add to skb share
> info.
> > Do you agree with this approach? If not, do you have any other
> &
Hello Dave,
To support skb zero-copy, a pointer is needed to add to skb share info.
Do you agree with this approach? If not, do you have any other
suggestions?
Thanks
Shirley
On Sat, 2011-05-28 at 12:23 -0700, Shirley Ma wrote:
> From:
> Shir
On Sat, 2011-05-28 at 12:15 -0700, Shirley Ma wrote:
> include/net/sock.h |1 +
> 1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/include/net/sock.h b/include/net/sock.h
> index f2046e4..2229bd1 100644
> --- a/include/net/sock.h
> +++ b/include/net/sock
Hello Michael,
In order to use wait for completion in shutting down, seems to me
another work thread is needed to call vhost_zerocopy_add_used, it seems
too much work to address a minor issue here. Do we really need it?
Right now, the approach I am using is to ignore outstanding userspace
buffers
Only when buffer size is greater than or equal PAGE_SIZE, macvtap
enables zero-copy.
Signed-off-by: Shirley Ma
---
drivers/net/macvtap.c | 131
1 files changed, 120 insertions(+), 11 deletions(-)
diff --git a/drivers/net/macvtap.c b/drivers
? From the comments, this destructor_arg is
used for destructor, destructor calls doesn't wait for last reference to
that skb is gone.
Signed-off-by: Shirley Ma
---
include/linux/skbuff.h | 24
net/core/skbuff.c | 73
2
include/net/sock.h |1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index f2046e4..2229bd1 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -563,6 +563,7 @@ enum sock_flags {
SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_
This patchset add supports for TX zero-copy between guest and host
kernel through vhost. It significantly reduces CPU utilization on the
local host on which the guest is located (It reduced about 50% CPU usage
for single stream test on the host, while 4K message size BW has
increased about 50%). Th
On Thu, 2011-05-26 at 13:31 -0700, Shirley Ma wrote:
> On Thu, 2011-05-26 at 23:28 +0300, Michael S. Tsirkin wrote:
> > On Thu, May 26, 2011 at 01:00:20PM -0700, Shirley Ma wrote:
> > > 3. Add sleep in vhost shutting down instead of busy-wait for
> > outstanding
> &g
On Thu, 2011-05-26 at 22:55 +0200, Eric Dumazet wrote:
> Le jeudi 26 mai 2011 à 13:24 -0700, Shirley Ma a écrit :
>
> > I could reduce callback pointer by moving it to *arg, but not desc,
> this
> > indicates that which buffer DMA hasn't done yet in *arg.
>
>
&g
On Thu, 2011-05-26 at 23:28 +0300, Michael S. Tsirkin wrote:
> On Thu, May 26, 2011 at 01:00:20PM -0700, Shirley Ma wrote:
> > 3. Add sleep in vhost shutting down instead of busy-wait for
> outstanding
> >DMAs.
>
> I still think this is not much better. We need to use
On Thu, 2011-05-26 at 22:04 +0200, Eric Dumazet wrote:
> Le jeudi 26 mai 2011 à 12:36 -0700, Shirley Ma a écrit :
> > This patch adds userspace buffers support in skb shared info. A new
> > struct skb_ubuf_info is needed to maintain the userspace buffers
> > argument and inde
This patchset add supports for TX zero-copy between guest and host
kernel through vhost. It significantly reduces CPU utilization on the
local host on which the guest is located (It reduced about 50% CPU usage
for single stream test on the host, while 4K message size BW has
increased about 50%). Th
Hello Michael,
Let me anything I might miss from your preview's review.
Thanks
Shirley
---
This patch maintains the outstanding userspace buffers in the
sequence it is delivered to vhost. The outstanding userspace buffers
will be marked as done once the lower device buffers DMA has fi
Only when buffer size is greater than GOODCOPY_LEN (256), macvtap
enables zero-copy.
Signed-off-by: Shirley Ma
---
drivers/net/macvtap.c | 132
1 files changed, 121 insertions(+), 11 deletions(-)
diff --git a/drivers/net/macvtap.c b/drivers
any userspace apps to reference these userspace buffers,
then these userspaces buffers will be copied into kernel. This way we
can prevent userspace apps to hold these userspace buffers too long.
Signed-off-by: Shirley Ma
---
include/linux/skbuff.h | 26 +++
net/core/skbuff.c
Signed-off-by: Shirley Ma
---
include/net/sock.h |1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index 01810a3..ab09097 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -562,6 +562,7 @@ enum sock_flags
This patchset add supports for TX zero-copy between guest and host
kernel through vhost. It significantly reduces CPU utilization on the
local host on which the guest is located (It reduced about 50% CPU usage
for single stream test on the host, while 4K message size BW has
increased about 50%). Th
On Thu, 2011-05-26 at 11:49 +0300, Michael S. Tsirkin wrote:
> >
> > > - SG support
> > > - HIGHDMA support (on arches where this makes sense)
> >
> > This can be checked by device flags.
>
> OK, but pls note that SG can get turned off dynamically.
Tested the patch w/i SG dynmically on/off and
On Thu, 2011-05-26 at 11:49 +0300, Michael S. Tsirkin wrote:
> On Wed, May 25, 2011 at 03:49:40PM -0700, Shirley Ma wrote:
> > On Fri, 2011-05-20 at 02:41 +0300, Michael S. Tsirkin wrote:
> > > So the requirements are
> > > - data must be released in a timely fashio
On Fri, 2011-05-20 at 02:41 +0300, Michael S. Tsirkin wrote:
> So the requirements are
> - data must be released in a timely fashion (e.g. unlike virtio-net
> tun or bridge)
The current patch doesn't enable tun zero-copy. tun will copy data It's
not an issue now. We can disallow macvtap attach to
On Wed, 2011-05-18 at 10:00 -0700, Shirley Ma wrote:
> On Wed, 2011-05-18 at 19:51 +0300, Michael S. Tsirkin wrote:
> > > > Yes, I agree. I think for tcpdump, we really need to copy the
> > data
> > > > anyway, to avoid guest changing it in between. So we do th
On Wed, 2011-05-18 at 19:51 +0300, Michael S. Tsirkin wrote:
> > > Yes, I agree. I think for tcpdump, we really need to copy the
> data
> > > anyway, to avoid guest changing it in between. So we do that and
> then
> > > use the copy everywhere, release the old one. Hmm?
> >
> > Yes. Old one use
On Wed, 2011-05-18 at 19:36 +0300, Michael S. Tsirkin wrote:
> On Wed, May 18, 2011 at 09:07:37AM -0700, Shirley Ma wrote:
> > On Wed, 2011-05-18 at 18:47 +0300, Michael S. Tsirkin wrote:
> > > On Wed, May 18, 2011 at 07:38:27AM -0700, Shirley Ma wrote:
> > > > On
On Wed, 2011-05-18 at 18:47 +0300, Michael S. Tsirkin wrote:
> On Wed, May 18, 2011 at 07:38:27AM -0700, Shirley Ma wrote:
> > On Wed, 2011-05-18 at 13:40 +0200, Michał Mirosław wrote:
> > > >> >> Not more other restrictions, skb clone is OK.
> pskb_expand_head()
On Wed, 2011-05-18 at 07:38 -0700, Shirley Ma wrote:
> On Wed, 2011-05-18 at 13:40 +0200, Michał Mirosław wrote:
> > >> >> Not more other restrictions, skb clone is OK.
> pskb_expand_head()
> > looks
> > >> >> OK to me from code review.
> > &
On Wed, 2011-05-18 at 13:40 +0200, Michał Mirosław wrote:
> >> >> Not more other restrictions, skb clone is OK. pskb_expand_head()
> looks
> >> >> OK to me from code review.
> >> > Hmm. pskb_expand_head calls skb_release_data while keeping
> >> > references to pages. How is that ok? What do I miss?
macvtap.c so we can discard the desc even
in zero-copy case.
Signed-off-by: Shirley Ma
---
drivers/vhost/net.c | 42 +-
drivers/vhost/vhost.c | 49 +
drivers/vhost/vhost.h | 13 +
3 files
On Wed, 2011-05-18 at 00:28 +0300, Michael S. Tsirkin wrote:
> On Tue, May 17, 2011 at 01:50:19PM -0700, Shirley Ma wrote:
> > Resubmit the patch with most update. This patch passed some
> > live-migration test against RHEL6.2. I will run more stress test w/i
> > live migra
On Wed, 2011-05-18 at 00:58 +0200, Michał Mirosław wrote:
> W dniu 18 maja 2011 00:28 użytkownik Shirley Ma
> napisał:
> > On Tue, 2011-05-17 at 23:48 +0200, Michał Mirosław wrote:
> >> 2011/5/17 Shirley Ma :
> >> > Looks like to use a new flag require
On Tue, 2011-05-17 at 23:48 +0200, Michał Mirosław wrote:
> 2011/5/17 Shirley Ma :
> > Hello Michael,
> >
> > Looks like to use a new flag requires more time/work. I am thinking
> > whether we can just use HIGHDMA flag to enable zero-copy in macvtap
> to
> &
On Wed, 2011-05-18 at 00:28 +0300, Michael S. Tsirkin wrote:
> On Tue, May 17, 2011 at 01:50:19PM -0700, Shirley Ma wrote:
> > Resubmit the patch with most update. This patch passed some
> > live-migration test against RHEL6.2. I will run more stress test w/i
> > live migra
On Tue, 2011-05-17 at 23:58 +0300, Michael S. Tsirkin wrote:
> On Tue, May 17, 2011 at 01:50:19PM -0700, Shirley Ma wrote:
> > Resubmit the patch with most update. This patch passed some
> > live-migration test against RHEL6.2. I will run more stress test w/i
> > live migrati
On Tue, 2011-05-17 at 09:21 +0300, Michael S. Tsirkin wrote:
> Problem is, in your patch there are a set of restrictions on what the
> device can do with the skb that we need to enforce somehow.
> Also, how do we know it's a 'real NIC' and not a software device?
I checked macvtap newlink, it doesn
Resubmit the patch with most update. This patch passed some
live-migration test against RHEL6.2. I will run more stress test w/i
live migration.
Signed-off-by: Shirley Ma
---
drivers/vhost/net.c | 37 +++-
drivers/vhost/vhost.c | 55
Hello Michael,
Here is the patch I used to test out of order before: add used in a pend
array, and swap the last two ids.
I used to hit an issue, but now it seems working well.
This won't impact zero-copy patch since we need to maintain the pend
used ids anyway.
Signed-off-by: Shirl
On Tue, 2011-05-17 at 18:28 +0300, Michael S. Tsirkin wrote:
> Which is the order the descriptors are put on avail ring.
> By design, guest should not depend on used ring entries
> being in order with avail ring (and btw with virtio block,
> they aren't). If it does, it's a bug I think.
Ok, I thou
On Tue, 2011-05-17 at 08:55 +0300, Michael S. Tsirkin wrote:
> Something in your patch that overwrites the id in vhost
> and makes it put the wrong id in the used ring?
>
> By the way, need to keep in mind that a guest can
> give us the same head twice, need to make sure this
> at least does not c
On Tue, 2011-05-17 at 00:24 +0300, Michael S. Tsirkin wrote:
> On Mon, May 16, 2011 at 01:56:54PM -0700, Shirley Ma wrote:
> > On Mon, 2011-05-16 at 23:45 +0300, Michael S. Tsirkin wrote:
> > > > +/* Since we need to keep the order of used_idx as avail_idx,
>
Hello Michael,
Looks like to use a new flag requires more time/work. I am thinking
whether we can just use HIGHDMA flag to enable zero-copy in macvtap to
avoid the new flag for now since mavctap uses real NICs as lower device?
Thanks
Shirley
--
To unsubscribe from this list: send the line "unsub
On Tue, 2011-05-17 at 00:24 +0300, Michael S. Tsirkin wrote:
> Very strange, it sounds like a bug, but I can't tell where: in
> host or in guest. If it's in the guest, we must fix it.
> If in host, we should only fix it if it makes life simpler for us.
> Could you try to nail it down pls? Another
On Mon, 2011-05-16 at 23:45 +0300, Michael S. Tsirkin wrote:
> > +/* Since we need to keep the order of used_idx as avail_idx, it's
> possible that
> > + * DMA done not in order in lower device driver for some reason. To
> prevent
> > + * used_idx out of order, upend_idx is used to track avail_idx
Device can enable zero-copy flag when HIGHDMA is supported.
Signed-off-by: Shirley Ma
---
drivers/net/ixgbe/ixgbe_main.c |4
1 files changed, 4 insertions(+), 0 deletions(-)
diff --git a/drivers/net/ixgbe/ixgbe_main.c b/drivers/net/ixgbe/ixgbe_main.c
index 2dce3d0..7e9e881 100644
On Mon, 2011-05-16 at 20:35 +0100, Ben Hutchings wrote:
> Sorry, bit 31 is taken. You get the job of turning features into a
> wider bitmap.
:) will do it.
Thanks
Shirley
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More
Only when buffer size is greater than GOODCOPY_LEN (256), macvtap
enables zero-copy.
Signed-off-by: Shirley Ma
---
drivers/net/macvtap.c | 129
1 files changed, 118 insertions(+), 11 deletions(-)
diff --git a/drivers/net/macvtap.c b/drivers
This patch maintains the outstanding userspace buffers in the
sequence it is delivered to vhost. The outstanding userspace buffers
will be marked as done once the lower device buffers DMA has finished.
This is monitored through last reference of kfree_skb callback. Two
buffer index are used for
-by: Shirley Ma
---
include/linux/skbuff.h | 26 ++
net/core/skbuff.c | 13 +
2 files changed, 39 insertions(+), 0 deletions(-)
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 79aafbb..40faffe 100644
--- a/include/linux/skbuff.h
Signed-off-by: Shirley Ma
---
include/linux/netdevice.h | 10 ++
1 files changed, 10 insertions(+), 0 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index a134d80..2646251 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
Signed-off-by: Shirley Ma
---
include/net/sock.h |1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index f2046e4..2229bd1 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -563,6 +563,7 @@ enum sock_flags
This patchset add supports for TX zero-copy between guest and host
kernel through vhost. It significantly reduces CPU utilization on the
local host on which the guest is located (It reduced 30-50% CPU usage
for vhost thread for single stream test). The patchset is based on
previous submission and c
On Wed, 2011-05-04 at 16:13 -0700, Sridhar Samudrala wrote:
> On Wed, 2011-05-04 at 00:55 -0700, Shirley Ma wrote:
> > Signed-off-by: Shirley Ma
> > ---
> >
> > include/linux/netdevice.h | 10 ++
> > 1 files changed, 10 insertions(+), 0 deletions
On Wed, 2011-05-04 at 16:13 -0700, Sridhar Samudrala wrote:
> On Wed, 2011-05-04 at 01:14 -0700, Shirley Ma wrote:
> > Only when buffer size is greater than GOODCOPY_LEN (256), macvtap
> > enables zero-copy.
> >
> > Signed-off-by: Shirley Ma
> > ---
>
On Wed, 2011-05-04 at 19:12 +0300, Michael S. Tsirkin wrote:
> On Wed, May 04, 2011 at 08:37:29AM -0700, Shirley Ma wrote:
> > On Wed, 2011-05-04 at 17:58 +0300, Michael S. Tsirkin wrote:
> > > On Wed, May 04, 2011 at 01:14:53AM -0700, Shirley Ma wrote:
> > > > On
On Wed, 2011-05-04 at 17:58 +0300, Michael S. Tsirkin wrote:
> On Wed, May 04, 2011 at 01:14:53AM -0700, Shirley Ma wrote:
> > Only when buffer size is greater than GOODCOPY_LEN (256), macvtap
> > enables zero-copy.
> >
> > Signed-off-by: Shirley Ma
>
>
On Wed, 2011-05-04 at 12:56 +0300, Michael S. Tsirkin wrote:
> On Wed, May 04, 2011 at 01:11:24AM -0700, Shirley Ma wrote:
> > This patch maintains the outstanding userspace buffers in the
> > sequence it is delivered to vhost. The outstanding userspace
> buffers
> > wil
Only when buffer size is greater than GOODCOPY_LEN (256), macvtap
enables zero-copy.
Signed-off-by: Shirley Ma
---
drivers/net/macvtap.c | 126
1 files changed, 115 insertions(+), 11 deletions(-)
diff --git a/drivers/net/macvtap.c b/drivers
This patch maintains the outstanding userspace buffers in the
sequence it is delivered to vhost. The outstanding userspace buffers
will be marked as done once the lower device buffers DMA has finished.
This is monitored through last reference of kfree_skb callback. Two
buffer index are used for
skb has a 256 bytes copied data in head to
make sure we have enough room for head expanding; and mapped the rest of
userspace buffers in skb frags.
Signed-off-by: Shirley Ma
---
include/linux/skbuff.h | 26 ++
net/core/skbuff.c | 13 +
2 files
Signed-off-by: Shirley Ma
---
include/linux/netdevice.h | 10 ++
1 files changed, 10 insertions(+), 0 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 0249fe7..0808f1e 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
Signed-off-by: Shirley Ma
---
include/net/sock.h |1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/include/net/sock.h b/include/net/sock.h
index 01810a3..ab09097 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -562,6 +562,7 @@ enum sock_flags
This patchset add supports for TX zero-copy between guest and host
kernel through vhost. It significantly reduces CPU utilization on the
local host on which the guest is located (It reduced 30-50% CPU usage
for vhost thread for single stream test). The patchset is based on
previous submission and c
On Tue, 2011-05-03 at 10:42 -0700, Shirley Ma wrote:
> Better to prevent this kind of skbs to be used in skb_clone, expand
> head
> for now.
I looked at the code, skb_clone shouldn't have any problem since ubuf
callback is only called after the lower device DMA has done. I can
modif
On Mon, 2011-05-02 at 22:53 +0300, Michael S. Tsirkin wrote:
> On Mon, May 02, 2011 at 11:47:08AM -0700, Shirley Ma wrote:
> > On Mon, 2011-05-02 at 13:42 +0300, Michael S. Tsirkin wrote:
> > > This comment should specify what exactly is the promise the
> > > device
On Mon, 2011-05-02 at 13:53 +0300, Michael S. Tsirkin wrote:
> On Wed, Apr 20, 2011 at 12:47:57PM -0700, Shirley Ma wrote:
> > This patch adds userspace buffers support in skb. A new struct
> > skb_ubuf_info is needed to maintain the userspace buffers argument
> > and index,
On Mon, 2011-05-02 at 13:42 +0300, Michael S. Tsirkin wrote:
> This comment should specify what exactly is the promise the
> device makes by setting this flag. Specifically, the
> condition is that no skb fragments are used
> after the uinfo callback has been called.
>
> The way it's implemented,
Resubmit the patch with a bug being found by ,
when he tested 2.6.38.2 kernel.
When the virtio_net doesn't enable GSO, vnet hdr_len is 0, we can't
use vnet hdr_len to allocate linear skb in zerocopy.
Signed-off-by: Shirley Ma
---
drivers/net/macvta
On Wed, 2011-04-20 at 12:36 -0700, Shirley Ma wrote:
> I am collecting more test results against 2.6.39-rc3 kernel and will
> provide the test matrix later.
Single TCP_STREAM 120 secs test results over ixgbe 10Gb NIC results:
Message BW(Gb/s)qemu-kvm (NumCPU)vhost-net(NumCPU) PerfTop ir
On Thu, 2011-04-21 at 10:29 -0400, Jon Mason wrote:
> Why are only these 3 drivers getting support? As far as I can tell,
> the only requirement is HIGHDMA. If this is the case, is there really
> a need for an additional flag to support this? If you can key off of
> HIGHDMA, all devices that sup
Signed-off-by: Shirley Ma
---
drivers/net/ixgbe/ixgbe_main.c |1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/drivers/net/ixgbe/ixgbe_main.c b/drivers/net/ixgbe/ixgbe_main.c
index 6f8adc7..68f1e93 100644
--- a/drivers/net/ixgbe/ixgbe_main.c
+++ b/drivers/net/ixgbe
This patch adds userspace buffers support in skb. A new struct
skb_ubuf_info is needed to maintain the userspace buffers argument
and index, a callback is used to notify userspace to release the
buffers once lower device has done DMA (Last reference to that skb
has gone).
Signed-off-by: Shirley
On Wed, 2011-04-20 at 13:58 -0700, Shirley Ma wrote:
> This flag can be ON when HIGHDMA and scatter/gather support. I will
> modify the patch to make it conditionally.
Double checked, it only needs HIGHDMA condition, not scatter/gather.
thanks
Shirley
--
To unsubscribe from this list: se
On Wed, 2011-04-20 at 13:52 -0700, Dimitris Michailidis wrote:
> The features handling has been reworked in net-next and patches like
> this
> won't apply as the code you're patching has changed. Also core code
> now
> does a lot of the related work and you'll need to tell it what to do
> with
Resubmit it with 31 bit.
Signed-off-by: Shirley Ma
---
include/linux/netdevice.h |3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 0249fe7..0998d3d 100644
--- a/include/linux/netdevice.h
+++ b/include/linux
On Wed, 2011-04-20 at 13:24 -0700, Dimitris Michailidis wrote:
> Bit 30 is also taken in net-next.
How about 31?
Thanks
Shirley
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/m
1 - 100 of 319 matches
Mail list logo