[dpdk-dev] examples/vhost: how to use vhost-switch between VMs?
Dear Mr. Liu, There are some descriptions in dpdk-16.11\doc\guides\sample_app_ug\vhost.rst, "The sample application performs simple packet switching between virtual machines based on Media Access Control (MAC) address or Virtual Local Area Network (VLAN) tag." So, we runned the vhost-switch sample in a VM with three virtio ports. First, we binded one virtio port to igb_uio. Then we runned the command "./vhost-switch -c f -n 4 --socket-mem 1024 -- --socket-file /tmp/sock0 �C-vm2vm 1 �C-dequeue-zero-copy" And then error occured. We found that the 'dev_info.max_vmdq_pools' of the virtio port was 0, and division by zero error occured at the statement 'queues_per_pool = dev_info.vmdq_queue_num / dev_info.max_vmdq_pools;' We fixed this by assigning 'queues_per_pool' to 0 if 'dev_info.max_vmdq_pools' was 0, but there were still other errors when running vhost-switch. So, we doubted whether we use the vhost-switch in a wrong way. Could you please give us some advice that how to use the vhost-switch between two or more VMs? I look forward to hearing from you. Yours faithfully, Yong Wang
[dpdk-dev] examples/vhost: how to use vhost-switch between VMs? Anybody who can help?
There are some descriptions in dpdk-16.11\doc\guides\sample_app_ug\vhost.rst, "The sample application performs simple packet switching between virtual machines based on Media Access Control (MAC) address or Virtual Local Area Network (VLAN) tag." So, we want to use vhost-switch between two VMs, one work as server and another as client. We runned the vhost-switch sample in a VM with three virtio ports. First, we binded one virtio port to igb_uio. Then we runned the command "./vhost-switch -c f -n 4 --socket-mem 1024 -- --socket-file /tmp/sock0 �C-vm2vm 1 �C-dequeue-zero-copy" And then error occured. We found that the 'dev_info.max_vmdq_pools' of the virtio port was 0, and division by zero error occured at the statement 'queues_per_pool = dev_info.vmdq_queue_num / dev_info.max_vmdq_pools;' We fixed this by assigning 'queues_per_pool' to 0 if 'dev_info.max_vmdq_pools' was 0, but there were still other errors when running vhost-switch. After we fixed all the errors while running vhost-switch as server, we didn't know how to run the vhost-switch as client in another VM, how could the client get the socket, the netdev and the virtio port that the server vhost-switch created? So, we doubted whether we use the vhost-switch in a wrong way. Could you please give us some advice that how to use the vhost-switch between two or more VMs? Anybody who can help us with this?
[dpdk-dev] 答复: Re: [PATCH] lib/librte_vhost: fix memory leak
> Yuanhan Liu
> 2017/01/04 12:02
>
> to
>
> Yong Wang ,
>
> cc
>
> dev@dpdk.org
>
> subject
>
> Re: [PATCH] lib/librte_vhost: fix memory leak
>
> On Tue, Jan 03, 2017 at 10:57:55PM -0500, Yong Wang wrote:
> > In function vhost_new_device(), current code dose not free 'dev'
> > in "i == MAX_VHOST_DEVICE" condition statements. It will lead to a
> > memory leak.
>
> Nice catch!
>
> Here are few minor stuff you might need pay attention to for future
> contribution:
>
> - a fix patch needs a fixline, like following
>
> Fixes: 45ca9c6f7bc6 ("vhost: get rid of linked list for devices")
>
> - the prefix for vhost lib is "vhost: ". And FYI, for PMD drivers, it's
> 'net/PMD_NAME', say 'net/virtio'.
>
>
> For you convenience, I have fixed the two while applying. And thanks
> for the fix.
>
> Applied to dpdk-next-virtio.
>
>--yliu
Thanks for your advice.
Re: [dpdk-dev] [PATCH] net/i40e: fix param check to avoid division by 0
> RE: [PATCH] net/i40e: fix param check to avoid division by 0
>
>
>
> > -Original Message-
> > From: Yong Wang [mailto:wang.yon...@zte.com.cn]
> > Sent: Monday, June 12, 2017 5:07 PM
> > To: Zhang, Helin ; Wu, Jingjing
>
> > Cc: dev@dpdk.org; Yong Wang
> > Subject: [PATCH] net/i40e: fix param check to avoid division by 0
> >
> > In function i40e_vsi_config_tc_queue_mapping(), if 'enabled_tcmap' is
> > 0, 'total_tc' might be 0. Then 'total_tc' might be used in a division
> > by 0 in "qpnum_per_tc = i40e_align_floor(vsi->nb_qps / total_tc)". Fix
> > it by adding a check to parameter 'enabled_tcmap'.
> >
> > Signed-off-by: Yong Wang
> > ---
> > drivers/net/i40e/i40e_ethdev.c | 5 +
> > 1 file changed, 5 insertions(+)
> >
> > diff --git a/drivers/net/i40e/i40e_ethdev.c
b/drivers/net/i40e/i40e_ethdev.c
> > index c18a93b..d41b213 100644
> > --- a/drivers/net/i40e/i40e_ethdev.c
> > +++ b/drivers/net/i40e/i40e_ethdev.c
> > @@ -4257,6 +4257,11 @@ enum i40e_status_code
> > int i, total_tc = 0;
> > uint16_t qpnum_per_tc, bsf, qp_idx;
> >
> > + if (enabled_tcmap == 0) {
> > + PMD_DRV_LOG(ERR, "Error! enabled tcmap shouldn't be 0");
> > + return I40E_ERR_PARAM;
> > + }
> > +
> > ret = validate_tcmap_parameter(vsi, enabled_tcmap);
> > if (ret != I40E_SUCCESS)
> >return ret;
> > --
> Thanks for the fix.
>
> Could you add the check in func validate_tcmap_parameter?
>
> Thanks
> Jingjing
>
>
Since func validate_tcmap_parameter() was called by 3 func, if I add
the check to func validate_tcmap_parameter(), I am not sure whether it
will impact the other two callers who didn't meet division by 0.
Or should I fix it by changing 'total_tc' from 0 to 1 just as func
i40e_vsi_update_queue_mapping() does?
Thanks
Yong
[dpdk-dev] 答复: Re: [PATCH] net/ark:add null point check
> Ferruh Yigit > 2017/09/14 17:33 > > to > > Yong Wang , shepard.sie...@atomicrules.com, > > cc > > dev@dpdk.org > > subject > > Re: [dpdk-dev] [PATCH] net/ark:add null point check > > On 9/14/2017 8:22 AM, Yong Wang wrote: > > In function ark_config_device(), there are several malloc without null > > point check. Fix it by adding null point check. > > > > Signed-off-by: Yong Wang > > Hi Yong, > > Thank you for the patch. > > I think you have sent new version of the patch because previous one was > giving checkpatch warning, but having two patches with same title is > being confusing. > > Next time would you mind sending next version as version 2 (v2) and as a > reply to previous version of the patch? So this makes clear which one is > the one to take into account. > Also adding a changelog to the commit log makes life easy. Thanks. > > DPDK Contributor's Guide has more details: > http://dpdk.org/doc/guides/contributing/patches.html > > Thanks, > ferruh Ok. Thanks for your advice.
