On Thu, Sep 29, 2016 at 11:21:48PM +0300, Michael S. Tsirkin wrote:
> On Thu, Sep 29, 2016 at 10:05:22PM +0200, Maxime Coquelin wrote:
> >
> >
> > On 09/29/2016 07:57 PM, Michael S. Tsirkin wrote:
> Yes but two points.
>
> 1. why is this memset expensive?
I don't have the exact answer, but just
On Mon, Oct 10, 2016 at 06:04:32AM +0300, Michael S. Tsirkin wrote:
> > > So I guess at this point, we can teach vhost-user in qemu
> > > that version 1 implies any_layout but only for machine types
> > > qemu 2.8 and up. It sets a bad precedent but oh well.
> >
> > It should work.
> >
> > --
On Mon, Oct 10, 2016 at 02:20:22AM +0300, Michael S. Tsirkin wrote:
> On Wed, Sep 28, 2016 at 10:28:48AM +0800, Yuanhan Liu wrote:
> > On Tue, Sep 27, 2016 at 10:56:40PM +0300, Michael S. Tsirkin wrote:
> > > On Tue, Sep 27, 2016 at 11:11:58AM +0800, Yuanhan Liu wrote:
> >
On Sun, Oct 09, 2016 at 12:09:07PM +, Wang, Zhihong wrote:
> > > > Tested with testpmd, host: txonly, guest: rxonly
> > > > size (bytes) improvement (%)
> > > > 644.12
> > > > 128 6
> > > > 256 2.65
> > > > 512 -1.12
ged.
--yliu
> for TX-zero-copy enabled case. For TX-zero-copy disabled case, I just ran
> default testpmd(txd=512, rxd=128) without the patch.
> Could you help check if NIC2VM case?
>
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf
Add an option, dequeue-zero-copy, to enable this feature in vhost-pmd.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
drivers/net/vhost/rte_eth_vhost.c | 13 +
1 file changed, 13 insertions(+)
diff --git a/drivers/net/vhost/rte_eth_vhost.c
b/drivers/net/vhost
dequeue zero copy is enabled, guest Tx
used vring will be updated only when corresponding mbuf is freed. If mbuf
is not freed frequently, the guest Tx vring could be starved.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
v3: - handle missing "Tx --> dequeue" rena
Dequeue zero copy is disabled by default. Here add a new flag
``RTE_VHOST_USER_DEQUEUE_ZERO_COPY`` to explictily enable it.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
v2: - update release log
- doc dequeue zero copy in detail
---
doc/guides/prog_guide/vhost_lib.rst
small
packets, (e.g. 64B), it actually slows a bit down (well, it could up to
15%). That is expected because this patch introduces some extra works,
and it outweighs the benefit from saving few bytes copy.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
v3: - rebase on top of th
he mbuf is consumed. Thus, we need use another var to
note the last avail ring idx we have consumed.
Therefore, last_avail_idx is introduced.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
lib/librte_vhost/vhost.h | 2 +-
lib/librte_vhost/vhost_user.c | 6 --
lib/li
.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
v2: - use MAP_POPULATE option to make sure the page table will
be already setup while getting the phys address
- do a simple merge if the last 2 pages are continuous
- dump guest pages only in debug mode
---
lib/librte_vhost
: this patch removes orig_region_map and
redefines virtio_memory_region, to include all necessary info.
With that, we can simplify the guest/host address convert a bit.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
lib/librte_vhost/vhost.h | 49 ++--
lib/librte_vhos
uest phys memory region
- and few more trivial changes, please see them in the corresponding
patches
---
Yuanhan Liu (7):
vhost: simplify memory regions handling
vhost: get guest/host physical address mappings
vhost: introduce last avail idx for dequeue
vhost: add dequeue zero copy
On Fri, Oct 07, 2016 at 06:57:41PM +0530, Shreyansh Jain wrote:
> On Friday 07 October 2016 06:33 PM, David Marchand wrote:
> >The driver name has been lost with the eal rework.
> >Restore it.
> >
> >Fixes: c830cb295411 ("drivers: use PCI registration macro")
> >
> >Signed-off-by: David Marchand
>
On Thu, Sep 29, 2016 at 04:31:30PM -0400, Dey wrote:
> /*
> * dev_ops for virtio, bare necessities for basic operation
> */
> @@ -677,7 +685,6 @@ static const struct eth_dev_ops virtio_eth_dev_ops = {
> .allmulticast_enable = virtio_dev_allmulticast_enable,
> .allmulticast_disa
On Thu, Oct 06, 2016 at 02:37:27PM +, Xu, Qian Q wrote:
> this function copy_desc_to_mbuf has changed on the dpdk-next-virtio repo.
> Based on current dpdk-next-virtio repo, the commit ID is as below:
> commit b4f7b43cd9d3b6413f41221051d03a23bc5f5fbe
> Author: Zhiyong Yang
> Date: Thu Sep
On Thu, Sep 29, 2016 at 08:35:49PM +0800, Zhiyong Yang wrote:
> This feature adds vhost pmd extended statistics from per port perspective
> in order to meet the requirements of the applications such as OVS etc.
> RX/TX xstats count the bytes without CRC. This is different from physical
> NIC stats
On Thu, Sep 29, 2016 at 09:21:11AM +, Mcnamara, John wrote:
> > -Original Message-
> > From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Ciara Loftus
> > Sent: Tuesday, September 13, 2016 2:48 PM
> > To: dev at dpdk.org
> > Subject: [dpdk-dev] [PATCH] net/vhost: Add function to re
On Thu, Sep 29, 2016 at 08:48:40AM +, Loftus, Ciara wrote:
> Thanks for the patches Zhiyong. I've tested the size stats and they look good
> to me.
>
> Tested-by: Ciara Loftus
Ciara, thanks for testing.
Zhiyong, please fix the build issue so that I can apply. We have to make
sure every com
On Wed, Sep 28, 2016 at 09:26:47PM +0800, Zhiyong Yang wrote:
> The patch moves all stats counters to a new defined struct vhost_stats
> as follows, in order to manage all stats counters in a unified way and
> simplify the subsequent function implementation(vhost_dev_xstats_reset).
>
> Signed-off-
On Fri, Sep 23, 2016 at 11:56:17AM +0800, Yuanhan Liu wrote:
> On Thu, Sep 22, 2016 at 04:19:07PM +0800, Zhiyong Yang wrote:
> > Patch 1 moves all stats counters to a new defined struct vhost_stats,
> > in order to manage all stats counters in a consistent way.
> >
> >
On Tue, Sep 27, 2016 at 10:56:40PM +0300, Michael S. Tsirkin wrote:
> On Tue, Sep 27, 2016 at 11:11:58AM +0800, Yuanhan Liu wrote:
> > On Mon, Sep 26, 2016 at 10:24:55PM +0300, Michael S. Tsirkin wrote:
> > > On Mon, Sep 26, 2016 at 11:01:58AM -0700, Stephen Hemminger wrote:
>
On Tue, Jul 05, 2016 at 06:19:22PM +0530, Jerin Jacob wrote:
> This patch-set includes,
>
> 1) General cleanup of compile time dependency.
> 2) made vector handler section based on run-time cpuflags
> 2) Added NEON support for optimized Rx handling
>
> This patch-set is based on dpdk-next-virtio/
On Tue, Sep 27, 2016 at 05:24:35PM +, Chen, Jing D wrote:
> > On Mon, Sep 26, 2016 at 12:29:13PM +0800, Yuanhan Liu wrote:
> > > Interestingly, clang and gcc has different prototype for _mm_prefetch().
> > > For gcc, we have
> > >
> > >_mm_pr
On Tue, Sep 27, 2016 at 07:11:04PM +, Jianfeng Tan wrote:
> When virtio_user is used with OVS-DPDK (with mq disabled), it cannot
> receive any packets.
>
> It's because when vhost provides VHOST_USER_GET_PROTOCOL_FEATURES,
> all queue pairs are initialized in the disabled state. Quote
> QEMU/d
-yliu
>
> --
> Regards,
> Souvik
>
> -Original Message-
> From: Stephen Hemminger [mailto:stephen at networkplumber.org]
> Sent: Tuesday, September 27, 2016 2:57 PM
> To: Dey, Souvik
> Cc: Kavanagh, Mark B ; Yuanhan Liu at linux.intel.com>; dev at dpd
Can any PMD guys review it? It blocks a virtio patchset apply...
Thanks.
--yliu
On Mon, Sep 26, 2016 at 12:29:13PM +0800, Yuanhan Liu wrote:
> Interestingly, clang and gcc has different prototype for _mm_prefetch().
> For gcc, we have
>
>_mm_prefetch (const void *__P, e
On Tue, Sep 27, 2016 at 10:42:49AM +0200, Maxime Coquelin wrote:
> Indirect descriptors are usually supported by virtio-net devices,
> allowing to dispatch a larger number of requests.
>
> When the virtio device sends a packet using indirect descriptors,
> only one slot is used in the ring, even f
On Tue, Sep 27, 2016 at 05:05:41PM +0530, Pankaj Chauhan wrote:
> >Hi Pankaj,
> >
> >Again, apologize for late response: you see I was busy ;) Besides, I
> >need some time to think about it.
> >
>
> Hi YLiu,
>
> No issues with delayed response :)
Thanks!
>
> >Generally, I think your ideal prop
On Tue, Sep 27, 2016 at 05:14:44PM +0530, Pankaj Chauhan wrote:
> On 9/26/2016 9:42 AM, Yuanhan Liu wrote:
> >Besides the VMDq proposal, I got few more comments for you.
> >
> >On Mon, Sep 05, 2016 at 04:24:29PM +0530, Pankaj Chauhan wrote:
> >>Introduce suppo
On Thu, Sep 22, 2016 at 05:01:41PM +0800, Jianbo Liu wrote:
> On 22 September 2016 at 14:58, Wang, Zhihong
> wrote:
> >
> >
> >> -Original Message-
> >> From: Jianbo Liu [mailto:jianbo.liu at linaro.org]
> >> Sent: Thursday, September 22, 201
On Tue, Sep 27, 2016 at 12:16:32PM +0530, gowrishankar muthukrishnan wrote:
> On Friday 09 September 2016 09:29 PM, Thomas Monjalon wrote:
> >>>v7 changes:
> >>>- removed enforcing cache alignment for table hash structs and
> >>> instead check only for multiples of 64 bytes.
> >>>
> >>Acked-by: C
On Fri, Sep 23, 2016 at 10:28:23AM +0200, Maxime Coquelin wrote:
> + if (vq->desc[desc_indexes[i]].flags & VRING_DESC_F_INDIRECT) {
> + desc = (struct vring_desc *)gpa_to_vva(dev,
As mentioned before, this would break 32 bit OS build. It should be
(struct v
On Mon, Sep 26, 2016 at 10:24:55PM +0300, Michael S. Tsirkin wrote:
> On Mon, Sep 26, 2016 at 11:01:58AM -0700, Stephen Hemminger wrote:
> > I assume that if using Version 1 that the bit will be ignored
Yes, but I will just quote what you just said: what if the guest
virtio device is a legacy devi
On Mon, Sep 26, 2016 at 04:53:05PM +0200, Marcin Kerlin wrote:
> diff --git a/lib/librte_ether/rte_ether_version.map
> b/lib/librte_ether/rte_ether_version.map
> index 45ddf44..34e1109 100644
> --- a/lib/librte_ether/rte_ether_version.map
> +++ b/lib/librte_ether/rte_ether_version.map
> @@ -139,3
On Mon, Sep 26, 2016 at 03:25:35PM +0300, Michael S. Tsirkin wrote:
> On Mon, Sep 26, 2016 at 11:03:54AM +0800, Yuanhan Liu wrote:
> > On Fri, Sep 23, 2016 at 01:24:14PM -0700, Stephen Hemminger wrote:
> > > On Fri, 23 Sep 2016 21:22:23 +0300
> > > "Michael S. Ts
The feature VIRTIO_F_ANY_LAYOUT was actually supported by commit
dd856dfcb9e7 ("virtio: use any layout on Tx"). But it's not enabled.
Here this patch enables it.
Signed-off-by: Yuanhan Liu
---
drivers/net/virtio/virtio_ethdev.h | 1 +
1 file changed, 1 insertion(+)
diff --gi
pport this feature. But
it is not enabled.
Thus, this patch enables it.
Signed-off-by: Yuanhan Liu
---
lib/librte_vhost/vhost.c | 1 +
lib/librte_vhost/vhost.h | 3 +++
2 files changed, 4 insertions(+)
diff --git a/lib/librte_vhost/vhost.c b/lib/librte_vhost/vhost.c
index 46095c3..1d8d941 100644
The feature is actually supported both in virtio PMD and vhost lib.
We just haven't enabled it yet. This patchset simply enables it.
---
Yuanhan Liu (2):
vhost: enable any layout feature
net/virtio: enable any layout feature
drivers/net/virtio/virtio_ethdev.h | 1 +
lib/librte_vhost/vh
rtio: add vector Rx")
Fixes: c95584dc2b18 ("ixgbe: new vectorized functions for Rx/Tx")
Fixes: 9ed94e5bb04e ("i40e: add vector Rx")
Fixes: 7092be8437bd ("fm10k: add vector Rx")
Cc: Jerin Jacob
Cc: Chen Jing D(Mark)
Cc: Cunming Liang
Cc: Bruce Richardson
CC: Thomas
Besides the VMDq proposal, I got few more comments for you.
On Mon, Sep 05, 2016 at 04:24:29PM +0530, Pankaj Chauhan wrote:
> Introduce support for a generic framework for handling of switching between
> physical and vhost devices. The vswitch framework introduces the following
> concept:
>
> 1.
On Tue, Sep 20, 2016 at 02:28:17PM +0530, Pankaj Chauhan wrote:
> On 9/19/2016 8:13 PM, Yuanhan Liu wrote:
> >Firstly, sorry for being late on this discussion: I just got a chance
> >to follow what you guys were talking about.
> >
> >On Tue, Sep 13, 2016 at 02:51:31PM
On Fri, Sep 23, 2016 at 01:47:27PM +0100, Ferruh Yigit wrote:
> Alphabetically sorted items to check and
> added git log capitalization checks for LRO, NIC and PMD
>
> Signed-off-by: Ferruh Yigit
> ---
> scripts/check-git-log.sh | 21 -
> 1 file changed, 12 insertions(+), 9 d
= virtio_dev_info_get,
>
> .stats_get = virtio_dev_stats_get,
>
> .xstats_get = virtio_dev_xstats_get,
>
> --
>
> Please do let me know if this looks good to you all. Thanks
>
>
>
> --
&g
On Fri, Sep 23, 2016 at 09:23:11PM +, Wiles, Keith wrote:
>
> Regards,
> Keith
>
> > On Sep 23, 2016, at 12:26 AM, Yuanhan Liu
> > wrote:
> >
> > On Thu, Sep 22, 2016 at 06:43:55PM +0200, Thomas Monjalon wrote:
> >>>>>>>> T
On Fri, Sep 23, 2016 at 01:24:14PM -0700, Stephen Hemminger wrote:
> On Fri, 23 Sep 2016 21:22:23 +0300
> "Michael S. Tsirkin" wrote:
>
> > On Fri, Sep 23, 2016 at 08:16:09PM +0200, Maxime Coquelin wrote:
> > >
> > >
> > > On 09/23/2016 08:06 PM, Michael S. Tsirkin wrote:
> > > > On Fri, Sep
On Fri, Sep 02, 2016 at 03:23:34PM +, Mcnamara, John wrote:
>
>
> > -Original Message-
> > From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Gary Mussar
> > Sent: Friday, September 2, 2016 2:17 PM
> > To: dev at dpdk.org
> > Cc: Gary Mussar
> > Subject: [dpdk-dev] [PATCH] Tools
On Fri, Sep 23, 2016 at 10:31:12AM +0200, Thomas Monjalon wrote:
> 2016-09-23 14:15, Yuanhan Liu:
> > - _mm_prefetch((const void *)rused, _MM_HINT_T0);
> > + rte_prefetch0(rused);
>
> There are other occurences of _mm_prefetch in other drivers.
> It could deserve a
On Fri, Sep 23, 2016 at 10:43:20AM +0200, Thomas Monjalon wrote:
> 2016-09-23 12:26, Yuanhan Liu:
> > On Thu, Sep 22, 2016 at 06:43:55PM +0200, Thomas Monjalon wrote:
> > > > > > > > > There could be a similar need in other PMD.
> > > > > >
On Fri, Sep 23, 2016 at 10:50:06AM +0200, Thomas Monjalon wrote:
> 2016-09-23 15:10, Yuanhan Liu:
> > Commit c711ccb30987 ("ivshmem: remove library and its EAL integration")
> > removed ivshmem support, but seems David forgot to remove the another
&g
On Fri, Sep 23, 2016 at 09:02:49AM +, Jain, Deepak K wrote:
>
>
> > -Original Message-
> > From: Yuanhan Liu [mailto:yuanhan.liu at linux.intel.com]
> > Sent: Friday, September 23, 2016 7:33 AM
> > To: Kusztal, ArkadiuszX
> > Cc: dev at dpd
On Wed, Sep 21, 2016 at 06:45:05PM -0700, Stephen Hemminger wrote:
> On Thu, 22 Sep 2016 00:08:38 +
> "Dey, Souvik" wrote:
>
> > Answers inline.
> >
> > --
> > Regards,
> > Souvik
> >
> > -Original Message-
> > From: Stephen Hemminger [mailto:stephen at networkplumber.org]
> > Sent
On Thu, Sep 22, 2016 at 04:46:52PM +0200, Thomas Monjalon wrote:
> 2016-09-22 10:34, Mussar, Gary:
> > I am new at submitting patches to dpdk.org. This patch has been acked and
> > patchwork claims success on build it:
> > http://www.dpdk.org/dev/patchwork/patch/15595/
> >
> > Other users state
On Fri, Sep 23, 2016 at 09:16:49AM +0200, Maxime Coquelin wrote:
> + if (vq->desc[desc_indexes[i]].flags & VRING_DESC_F_INDIRECT) {
> + if (unlikely(!(dev->features &
> + (1ULL <<
> VIRTIO_RING_F_INDIRECT_DESC {
> +
e git grep shows no one else
references it; I think we can now rmeove it.
Cc: David Marchand
Signed-off-by: Yuanhan Liu
---
NOTE: only have done some build tests.
lib/librte_eal/common/eal_filesystem.h | 11 --
lib/librte_eal/common/eal_hugepages.h| 3 -
lib/librte_eal/linuxapp/eal/ea
On Fri, Sep 23, 2016 at 12:05:14PM +0530, Jerin Jacob wrote:
> On Fri, Sep 23, 2016 at 02:15:09PM +0800, Yuanhan Liu wrote:
>
> Hi Yuanhan,
>
> Thanks for this patch.
>
> > Interestingly, clang and gcc has different prototype for _mm_prefetch().
> > For gcc, we
On Fri, Sep 23, 2016 at 02:32:57PM +0800, Yuanhan Liu wrote:
> On Thu, Sep 22, 2016 at 04:54:57PM +0100, Arek Kusztal wrote:
> > This patch adds AES Galois Counter Mode test case for cryptodev
> > QAT tests. Test is performed with different buffer sizes, burst
> > size of
On Thu, Sep 22, 2016 at 04:54:57PM +0100, Arek Kusztal wrote:
> This patch adds AES Galois Counter Mode test case for cryptodev
> QAT tests. Test is performed with different buffer sizes, burst
> size of 32 and 128b key. Test vectors are placed in
> app/test/test_cryptodev_perf_vectors.h file.
>
>
r file, nothing else changed.
This patch fix the issue when Jerin's patchset is applied. Thus, I think
it's still needed.
Fixes: fc3d66212fed ("virtio: add vector Rx")
Cc: Jerin Jacob
Signed-off-by: Yuanhan Liu
---
v2: replace _mm_prefetch with rte_prefetch0 to make icc
On Tue, Aug 23, 2016 at 03:40:58PM +0800, Yuanhan Liu wrote:
> Interestingly, clang and gcc has different prototype for _mm_prefetch().
> For gcc, we have
>
>_mm_prefetch (const void *__P, enum _mm_hint __I)
>
> While for clang, it's
>
>#define _mm_prefetc
On Thu, Sep 22, 2016 at 06:43:55PM +0200, Thomas Monjalon wrote:
> > > > > > > There could be a similar need in other PMD.
> > > > > > > If we can get an opaque identifier of the device which is not the
> > > > > > > port id,
> > > > > > > we could call some specific functions of the driver not
>
Add an option, dequeue-zero-copy, to enable this feature in vhost-pmd.
Signed-off-by: Yuanhan Liu
---
drivers/net/vhost/rte_eth_vhost.c | 13 +
1 file changed, 13 insertions(+)
diff --git a/drivers/net/vhost/rte_eth_vhost.c
b/drivers/net/vhost/rte_eth_vhost.c
index 7539cd4
dequeue zero copy is enabled, guest Tx
used vring will be updated only when corresponding mbuf is freed. If mbuf
is not freed frequently, the guest Tx vring could be starved.
Signed-off-by: Yuanhan Liu
---
examples/vhost/main.c | 19 ++-
1 file changed, 18 insertions(+), 1 deletion
Dequeue zero copy is disabled by default. Here add a new flag
``RTE_VHOST_USER_DEQUEUE_ZERO_COPY`` to explictily enable it.
Signed-off-by: Yuanhan Liu
---
v2: - update release log
- doc dequeue zero copy in detail
---
doc/guides/prog_guide/vhost_lib.rst| 35
small
packets, (e.g. 64B), it actually slows a bit down (well, it could up to
15%). That is expected because this patch introduces some extra works,
and it outweighs the benefit from saving few bytes copy.
Signed-off-by: Yuanhan Liu
---
v2: - use unlikely/likely for dequeue_zero_copy check, as
he mbuf is consumed. Thus, we need use another var to
note the last avail ring idx we have consumed.
Therefore, last_avail_idx is introduced.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
lib/librte_vhost/vhost.h | 2 +-
lib/librte_vhost/vhost_user.c | 6 --
lib/li
.
Signed-off-by: Yuanhan Liu
---
v2: - use MAP_POPULATE option to make sure the page table will
be already setup while getting the phys address
- do a simple merge if the last 2 pages are continuous
- dump guest pages only in debug mode
---
lib/librte_vhost/vhost.h | 30
: this patch removes orig_region_map and
redefines virtio_memory_region, to include all necessary info.
With that, we can simplify the guest/host address convert a bit.
Signed-off-by: Yuanhan Liu
Reviewed-by: Maxime Coquelin
---
lib/librte_vhost/vhost.h | 49 ++--
lib/librte_vhos
even better: about 70% boost.
For small packets, the performance is worse (it's expected, as the extra
overhead introduced by zero copy outweighs the benefits from saving few
bytes copy).
---
Yuanhan Liu (7):
vhost: simplify memory regions handling
vhost: get guest/host physical address ma
On Mon, Aug 29, 2016 at 08:57:52AM +, Xu, Qian Q wrote:
> Btw, some good news: if I run a simple dequeue workload (running rxonly in
> vhost-pmd and runnin txonly in guest testpmd), it yields ~50% performance
> boost for packet size 1518B, but this case is without NIC.
> And similar case as
On Thu, Sep 22, 2016 at 04:19:07PM +0800, Zhiyong Yang wrote:
> Patch 1 moves all stats counters to a new defined struct vhost_stats,
> in order to manage all stats counters in a consistent way.
>
> Patch 2 adds the pmd xstats support.
Applied to dpdk-next-virtio.
Thanks.
--yliu
On Wed, Sep 21, 2016 at 03:07:51PM +0200, Thomas Monjalon wrote:
> 2016-09-18 16:27, Yuanhan Liu:
> > On Wed, Sep 14, 2016 at 10:35:53AM +0200, Thomas Monjalon wrote:
> > > 2016-09-14 15:21, Yuanhan Liu:
> > > > On Wed, Sep 14, 2016 at 09:10:48AM +0200, Thomas Monjal
On Wed, Sep 21, 2016 at 08:54:11PM +0800, Jianbo Liu wrote:
> >> > My setup consists of one host running a guest.
> >> > The guest generates as much 64bytes packets as possible using
> >>
> >> Have you tested with other different packet size?
> >> My testing shows that performance is dropping when
On Thu, Sep 22, 2016 at 01:42:14AM +, Yang, Zhiyong wrote:
> hi, yuanhan:
>
> > -Original Message-
> > From: Yuanhan Liu [mailto:yuanhan.liu at linux.intel.com]
> > Sent: Wednesday, September 21, 2016 6:57 PM
> > To: Yang, Zhiyong
> > Cc
On Wed, Sep 21, 2016 at 06:05:55PM +0800, Zhiyong Yang wrote:
> +static inline void
> +vhost_count_multicast_broadcast(struct vhost_queue *vq,
> + struct rte_mbuf **bufs,
> + uint16_t count)
Hmm.. why not just passing "struct rte_mbuf *mbuf"?
On Fri, Aug 05, 2016 at 02:18:42PM +0200, Maxime Coquelin wrote:
>
>
> On 08/03/2016 04:03 PM, Yuanhan Liu wrote:
> >On Tue, Jul 12, 2016 at 04:32:12PM +0200, Maxime Coquelin wrote:
> >>Indirect descriptors are usually supported by virtio-net devices,
> >>allow
Hi Maxime,
Do you have more comments about this set? If no, I think I could merge
it shortly.
Thanks.
--yliu
On Mon, Sep 19, 2016 at 10:00:11PM -0400, Zhihong Wang wrote:
> This patch set optimizes the vhost enqueue function.
>
> It implements the vhost logic from scratch into a single
r
indention.
You might want to send another version, to fix above two issues. Or,
if you don't mind, I could fix them for you and apply it.
--yliu
>
> -Original Message-----
> From: Yuanhan Liu [mailto:yuanhan.liu at linux.intel.com]
> Sent: Tuesday, September 20, 2016 3:
On Wed, Sep 21, 2016 at 01:32:15AM +, Wiles, Keith wrote:
> The concern I am having is on my standard Ubuntu 16.04 system these errors do
> not appear.
Yes, that's really weird. To me, it's such a solid error that could be
triggered on all linux platforms.
Anyway, I saw you removed the incl
(eth_pcap_rx_jumbo(pcap_q->mb_pool,
> mbuf,
> packet,
> - header.caplen) == -1))
> + header.caplen) == -1)) {
> + rte_pktmbuf_free(mbuf);
> break;
> + }
> }
>
> mbuf->pkt_len = (uint1
On Tue, Sep 20, 2016 at 05:36:45PM +0800, Zhiyong Yang wrote:
> +enum vhost_xstats_pkts {
> + VHOST_UNDERSIZE_PKT = 0,
> + VHOST_64_PKT,
> + VHOST_65_TO_127_PKT,
> + VHOST_128_TO_255_PKT,
> + VHOST_256_TO_511_PKT,
> + VHOST_512_TO_1023_PKT,
> + VHOST_1024_TO_1522_PKT,
>
On Tue, Sep 20, 2016 at 01:39:23PM +0300, Dror Birkman wrote:
> If rte_pktmbuf_alloc() fails on any segment that is not the initial
> segment, previously allocated mbufs are not freed.
You should put a "fixline" here.
Besides that, I think it's a good candidate for stable branch. Thus,
stable at
On Tue, Sep 20, 2016 at 05:36:45PM +0800, Zhiyong Yang wrote:
> diff --git a/drivers/net/vhost/rte_eth_vhost.c
> b/drivers/net/vhost/rte_eth_vhost.c
> index 9157bf1..c3f404d 100644
> --- a/drivers/net/vhost/rte_eth_vhost.c
> +++ b/drivers/net/vhost/rte_eth_vhost.c
> @@ -78,6 +78,7 @@ struct vhost_
On Tue, Sep 20, 2016 at 05:36:44PM +0800, Zhiyong Yang wrote:
> The patch moves all stats counters to a new defined struct vhost_stats
> as follows, in order to manage all stats counters in a unified way and
> simplify the subsequent function implementation(vhost_dev_xstats_reset).
>
> struct vhos
On Wed, Sep 14, 2016 at 12:15:37PM +, Kavanagh, Mark B wrote:
> >
> >>+{
> >>+?? struct rte_eth_dev_info dev_info;
> >>+?? uint32_t ether_hdr_len = ETHER_HDR_LEN + ETHER_CRC_LEN +
> >>+VLAN_TAG_LEN;
> >>+?? uint32_t frame_size = mtu + ether_hdr_len;
> >>+
> >>+?? virtio_dev_info
[ just got few more comments after some fiddling with the build issue ]
On Fri, Sep 16, 2016 at 11:22:13AM -0500, Keith Wiles wrote:
> diff --git a/drivers/net/tap/Makefile b/drivers/net/tap/Makefile
> new file mode 100644
> index 000..442a2fe
> --- /dev/null
> +++ b/drivers/net/tap/Makefile
>
On Mon, Sep 19, 2016 at 03:56:03PM +, Wiles, Keith wrote:
> All of the below errors are from Linux header files and not the Tap
> driver.
>
>
> Yes, but you are referencing them, so ...
>
>
> Yes I am referencing them, but still they are not interacting with the tap
>
On Mon, Sep 19, 2016 at 03:52:25PM +0200, Christian Ehrhardt wrote:
> On Mon, Sep 19, 2016 at 3:29 PM, Bruce Richardson <
> bruce.richardson at intel.com> wrote:
>
> > > This patch set fixes problem when mbufs are not released back to the
> > > original memory pool during Rx queue relese call lead
Firstly, sorry for being late on this discussion: I just got a chance
to follow what you guys were talking about.
On Tue, Sep 13, 2016 at 02:51:31PM +0800, Tan, Jianfeng wrote:
> >(2) we'd better not differentiate phys device and virtual
Agreed.
> >>device in generic framework (it's just an attr
On Sun, Sep 18, 2016 at 04:20:12PM +, Wiles, Keith wrote:
>
> Regards,
> Keith
>
> > On Sep 18, 2016, at 8:25 AM, Yuanhan Liu
> > wrote:
> >
> > On Fri, Sep 16, 2016 at 11:22:13AM -0500, Keith Wiles wrote:
> >> The rte_eth_tap.c PMD creates
On Thu, Sep 08, 2016 at 11:39:24PM -0400, Zhihong Wang wrote:
> This patch implements the vhost logic from scratch into a single function
> designed for high performance and better maintainability.
As always, your commit log just states what have been done, but doesn't
tell why such changes have b
On Fri, Sep 16, 2016 at 11:22:13AM -0500, Keith Wiles wrote:
> The rte_eth_tap.c PMD creates a device using TUN/TAP interfaces
> on the local host. The PMD allows for DPDK and the host to
> communicate using a raw device interface on the host and in
> the DPDK application. The device created is a T
On Wed, Sep 14, 2016 at 07:43:36AM +, Yang, Zhiyong wrote:
> Hi, yuanhan:
> Thanks so much for your detailed comments.
>
> > -Original Message-
> > From: Yuanhan Liu [mailto:yuanhan.liu at linux.intel.com]
> > Sent: Wednesday, September 14, 2016 2:20 PM
&g
On Wed, Sep 14, 2016 at 10:35:53AM +0200, Thomas Monjalon wrote:
> 2016-09-14 15:21, Yuanhan Liu:
> > On Wed, Sep 14, 2016 at 09:10:48AM +0200, Thomas Monjalon wrote:
> > > 2016-09-14 12:43, Yuanhan Liu:
> > > > On Tue, Sep 13, 2016 at 05:10:09PM +0200, Thomas Monjal
On Wed, Sep 14, 2016 at 04:32:46PM +0100, John Griffin wrote:
> Hi Liu,
> Comments embedded.
>
> Rgds,
> John.
>
> On 05/09/16 04:23, Yuanhan Liu wrote:
> >On Thu, Sep 01, 2016 at 11:21:38AM +0100, John Griffin wrote:
> >>Remove the dependency on dev->dri
On Thu, Sep 15, 2016 at 06:38:06PM +0200, Maxime Coquelin wrote:
> >>>+static inline void __attribute__((always_inline))
> >>>+flush_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq,
> >>>+ uint32_t used_idx_start)
> >>>+{
> >>>+ if (used_idx_start + vq->shadow_used_idx < vq->
On Wed, Sep 14, 2016 at 09:10:48AM +0200, Thomas Monjalon wrote:
> 2016-09-14 12:43, Yuanhan Liu:
> > On Tue, Sep 13, 2016 at 05:10:09PM +0200, Thomas Monjalon wrote:
> > > 2016-09-13 14:47, Ciara Loftus:
> > > > In some cases when using the vHost PMD, certain vHost
On Fri, Sep 09, 2016 at 04:15:27PM +0800, Zhiyong Yang wrote:
> +struct vhost_xstats {
> + uint64_t stat[16];
> +};
> +
> struct vhost_queue {
> int vid;
> rte_atomic32_t allow_queuing;
> @@ -85,7 +89,8 @@ struct vhost_queue {
> uint64_t missed_pkts;
> uint64_t rx_bytes
On Wed, Sep 14, 2016 at 04:46:21AM +, Liu, Changpeng wrote:
> > Few more generic comments:
> >
> > - you touched way more code than necessary.
> >
> > - you should split your patches into some small patches: one patch just
> > does one tiny logic. Doing one bunch of stuff in one patch is re
On Tue, Sep 13, 2016 at 05:10:09PM +0200, Thomas Monjalon wrote:
> 2016-09-13 14:47, Ciara Loftus:
> > In some cases when using the vHost PMD, certain vHost library functions
> > may still need to be accessed. One such example is the
> > rte_vhost_get_queue_num function which returns the number of
801 - 900 of 1959 matches
Mail list logo
