On Tue, Sep 28, 2010 at 08:24:29PM -0700, Shirley Ma wrote:
Hello Michael,
On Wed, 2010-09-15 at 07:52 -0700, Shirley Ma wrote:
Don't you think once I address vhost_add_used_and_signal update
issue, it is a simple and complete patch for macvtap TX zero copy?
Thanks
On Wed, Sep 29, 2010 at 10:16:45AM +0200, Michael S. Tsirkin wrote:
On Tue, Sep 28, 2010 at 08:24:29PM -0700, Shirley Ma wrote:
Hello Michael,
On Wed, 2010-09-15 at 07:52 -0700, Shirley Ma wrote:
Don't you think once I address vhost_add_used_and_signal update
issue, it is a
On Wed, 2010-09-29 at 10:16 +0200, Michael S. Tsirkin wrote:
If you look at dev_hard_start_xmit you will see a call
to skb_orphan_try which often calls the skb destructor.
So I suspect this is almost equivalent to your original patch,
and has the same correctness issue.
I forgot to mention,
On Wed, 2010-09-29 at 10:28 +0200, Michael S. Tsirkin wrote:
I think you should try testing with guest to external communication,
this will uncover some of these correctness issues for you.
I think netperf also has some flag to check data, might
be a good idea to use it for testing.
I always
On Wed, 2010-09-29 at 10:16 +0200, Michael S. Tsirkin wrote:
I compared several approaches for addressing the issue being raised
here
on how/when to update vhost_add_used_and_signal. The simple approach
I
have found is:
1. Adding completion field in struct virtqueue;
2. when it is a
On Wed, 2010-09-29 at 10:28 +0200, Michael S. Tsirkin wrote:
1. Adding completion field in struct virtqueue;
2. when it is a zero copy packet, put vhost thread wait for
completion
to update vhost_add_used_and_signal;
3. passing vq from vhost to macvtap as skb destruct_arg;
4. when
On Tue, Sep 28, 2010 at 08:24:29PM -0700, Shirley Ma wrote:
Hello Michael,
On Wed, 2010-09-15 at 07:52 -0700, Shirley Ma wrote:
Don't you think once I address vhost_add_used_and_signal update
issue, it is a simple and complete patch for macvtap TX zero copy?
Thanks
On Wed, 2010-09-29 at 17:14 +0200, Michael S. Tsirkin wrote:
I guess I just don't understand what your patch does.
If you send it, I can take a look.
Ok, I will collect more performance data and send the patch.
Thanks
Shirley
--
To unsubscribe from this list: send the line unsubscribe kvm in
Hello Michael,
On Wed, 2010-09-15 at 07:52 -0700, Shirley Ma wrote:
Don't you think once I address vhost_add_used_and_signal update
issue, it is a simple and complete patch for macvtap TX zero copy?
Thanks
Shirley
I like the fact that the patch is simple. Unfortunately
I
@vger.kernel.org; linux-ker...@vger.kernel.org
Subject: Re: [RFC PATCH 2/2] macvtap: TX zero copy between guest and host
kernel
On Tue, Sep 14, 2010 at 09:00:25AM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 17:22 +0200, Michael S. Tsirkin wrote:
I would expect this to hurt performance
...@vger.kernel.org
Subject: Re: [RFC PATCH 2/2] macvtap: TX zero copy between guest and host
kernel
On Wed, Sep 15, 2010 at 10:46:02AM +0800, Xin, Xiaohui wrote:
From: Michael S. Tsirkin [mailto:m...@redhat.com]
Sent: Wednesday, September 15, 2010 12:30 AM
To: Shirley Ma
Cc: Arnd Bergmann
On Wed, 2010-09-15 at 07:27 +0200, Michael S. Tsirkin wrote:
For some of the issues, try following the discussion around
net: af_packet: don't call tpacket_destruct_skb() until the skb is
sent
out.
Summary: it's difficult to do correctly generally. Limiting ourselves
to transmit on
On Wed, 2010-09-15 at 07:12 +0200, Michael S. Tsirkin wrote:
Yes, I agree this patch is useful for demo purposes:
simple, and shows what kind of performance gains
we can expect for TX.
Any other issue you can see in this patch beside vhost descriptors
update? Don't you think once I address
...@vger.kernel.org
Subject: Re: [RFC PATCH 2/2] macvtap: TX zero copy between guest and host
kernel
On Tue, Sep 14, 2010 at 09:00:25AM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 17:22 +0200, Michael S. Tsirkin wrote:
I would expect this to hurt performance significantly.
We could do
On Tue, Sep 14, 2010 at 11:21:15PM -0700, Shirley Ma wrote:
On Wed, 2010-09-15 at 07:12 +0200, Michael S. Tsirkin wrote:
Yes, I agree this patch is useful for demo purposes:
simple, and shows what kind of performance gains
we can expect for TX.
Any other issue you can see in this patch
On Wed, 2010-09-15 at 12:10 +0200, Michael S. Tsirkin wrote:
Another issue is that macvtap can be bound to almost
anything, including e.g. a tap device or a bridge,
which might hang on to skb fragments for unlimited time.
Zero copy TX won't easily work there.
I can imagine either somehow
On Wed, 2010-09-15 at 17:39 +0200, Michael S. Tsirkin wrote:
In fact, I rechecked: both bridge and loopback have NETIF_F_HIGHDMA
set.
So maybe we should check NETIF_F_NETNS_LOCAL ...
macvtap in bridged mode is interesting as well.
I found that too, just wondered which flag to use is
On Wed, Sep 15, 2010 at 10:00:04AM -0700, Shirley Ma wrote:
On Wed, 2010-09-15 at 17:39 +0200, Michael S. Tsirkin wrote:
In fact, I rechecked: both bridge and loopback have NETIF_F_HIGHDMA
set.
So maybe we should check NETIF_F_NETNS_LOCAL ...
macvtap in bridged mode is interesting as
On Wed, 2010-09-15 at 19:30 +0200, Michael S. Tsirkin wrote:
At some level NETIF_F_NETNS_LOCAL makes sense: local packets
can get anywhere. OTOH one wonders whether there might be other
issues, e.g. in theory devices could hang on to frag pages
just by doing get_page. There might be other
On Tue, 2010-09-14 at 11:12 +0200, Avi Kivity wrote:
+base = (unsigned long)from-iov_base + offset1;
+size = ((base ~PAGE_MASK) + len + ~PAGE_MASK)
PAGE_SHIFT;
+num_pages = get_user_pages_fast(base, size,
0,page[i]);
+if ((num_pages !=
On Tuesday 14 September 2010, Shirley Ma wrote:
On Tue, 2010-09-14 at 11:12 +0200, Avi Kivity wrote:
That's what io_submit() is for. Then io_getevents() tells you what
a
while actually was.
This macvtap zero copy uses iov buffers from vhost ring, which is
allocated from guest kernel.
On Tue, 2010-09-14 at 17:22 +0200, Michael S. Tsirkin wrote:
I would expect this to hurt performance significantly.
We could do this for asynchronous requests only to avoid the
slowdown.
Is kiocb in sendmsg helpful here? It is not used now.
Shirley
--
To unsubscribe from this list: send the
On Tue, Sep 14, 2010 at 09:00:25AM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 17:22 +0200, Michael S. Tsirkin wrote:
I would expect this to hurt performance significantly.
We could do this for asynchronous requests only to avoid the
slowdown.
Is kiocb in sendmsg helpful here? It is
On Tue, 2010-09-14 at 18:29 +0200, Michael S. Tsirkin wrote:
Precisely. This is what the patch from Xin Xiaohui does. That code
already seems to do most of what you are trying to do, right?
I thought host pins guest kernel buffer pages was good enough for TX
thought I didn't look up xiaohui's
On Tue, Sep 14, 2010 at 10:02:25AM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 18:29 +0200, Michael S. Tsirkin wrote:
Precisely. This is what the patch from Xin Xiaohui does. That code
already seems to do most of what you are trying to do, right?
I thought host pins guest kernel buffer
On Tue, 2010-09-14 at 20:27 +0200, Michael S. Tsirkin wrote:
As others said, the harder issues for TX are in determining that it's
safe
to unpin the memory, and how much memory is it safe to pin to beging
with. For RX we have some more complexity.
I think unpin the memory is in kfree_skb()
On Tue, Sep 14, 2010 at 11:49:03AM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 20:27 +0200, Michael S. Tsirkin wrote:
As others said, the harder issues for TX are in determining that it's
safe
to unpin the memory, and how much memory is it safe to pin to beging
with. For RX we have
On Tue, 2010-09-14 at 21:01 +0200, Michael S. Tsirkin wrote:
On Tue, Sep 14, 2010 at 11:49:03AM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 20:27 +0200, Michael S. Tsirkin wrote:
As others said, the harder issues for TX are in determining that
it's
safe
to unpin the memory, and how
On Tue, 2010-09-14 at 21:01 +0200, Michael S. Tsirkin wrote:
I think that you should be able to simply combine
the two drivers together, add an ioctl to
enable/disable zero copy mode of operation.
That could work. But what's the purpose to have two drivers if one
driver can handle
From: Shirley Ma [mailto:mashi...@us.ibm.com]
Sent: Tuesday, September 14, 2010 11:05 PM
To: Avi Kivity
Cc: David Miller; a...@arndb.de; m...@redhat.com; Xin, Xiaohui;
net...@vger.kernel.org;
kvm@vger.kernel.org; linux-ker...@vger.kernel.org
Subject: Re: [RFC PATCH 2/2] macvtap: TX zero copy
From: Arnd Bergmann [mailto:a...@arndb.de]
Sent: Tuesday, September 14, 2010 11:21 PM
To: Shirley Ma
Cc: Avi Kivity; David Miller; m...@redhat.com; Xin, Xiaohui;
net...@vger.kernel.org;
kvm@vger.kernel.org; linux-ker...@vger.kernel.org
Subject: Re: [RFC PATCH 2/2] macvtap: TX zero copy between
On Wed, 2010-09-15 at 09:50 +0800, Xin, Xiaohui wrote:
I think what David said is what we have thought before in mp device.
Since we are not sure the exact time the tx buffer was wrote though
DMA operation.
But the deadline is when the tx buffer was freed. So we only notify
the vhost stuff
From: Michael S. Tsirkin [mailto:m...@redhat.com]
Sent: Wednesday, September 15, 2010 12:30 AM
To: Shirley Ma
Cc: Arnd Bergmann; Avi Kivity; Xin, Xiaohui; David Miller;
net...@vger.kernel.org;
kvm@vger.kernel.org; linux-ker...@vger.kernel.org
Subject: Re: [RFC PATCH 2/2] macvtap: TX zero copy
From: Shirley Ma [mailto:mashi...@us.ibm.com]
Sent: Wednesday, September 15, 2010 10:41 AM
To: Xin, Xiaohui
Cc: Avi Kivity; David Miller; a...@arndb.de; m...@redhat.com;
net...@vger.kernel.org;
kvm@vger.kernel.org; linux-ker...@vger.kernel.org
Subject: RE: [RFC PATCH 2/2] macvtap: TX zero copy
On Tue, Sep 14, 2010 at 12:36:23PM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 21:01 +0200, Michael S. Tsirkin wrote:
I think that you should be able to simply combine
the two drivers together, add an ioctl to
enable/disable zero copy mode of operation.
That could work. But
On Tue, Sep 14, 2010 at 07:40:52PM -0700, Shirley Ma wrote:
On Wed, 2010-09-15 at 09:50 +0800, Xin, Xiaohui wrote:
I think what David said is what we have thought before in mp device.
Since we are not sure the exact time the tx buffer was wrote though
DMA operation.
But the deadline is
On Tue, Sep 14, 2010 at 12:20:29PM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 21:01 +0200, Michael S. Tsirkin wrote:
On Tue, Sep 14, 2010 at 11:49:03AM -0700, Shirley Ma wrote:
On Tue, 2010-09-14 at 20:27 +0200, Michael S. Tsirkin wrote:
As others said, the harder issues for TX are
Add zero copy feature between userspace and kernel in macvtap when lower device
supports
high memory DMA.
Signed-off-by: Shirley Ma x...@us.ibm.com
---
drivers/net/macvtap.c | 136 +
1 files changed, 126 insertions(+), 10 deletions(-)
diff
From: Shirley Ma mashi...@us.ibm.com
Date: Mon, 13 Sep 2010 13:48:03 -0700
+ base = (unsigned long)from-iov_base + offset1;
+ size = ((base ~PAGE_MASK) + len + ~PAGE_MASK) PAGE_SHIFT;
+ num_pages = get_user_pages_fast(base, size, 0, page[i]);
+
39 matches
Mail list logo