From: Maxime Coquelin <maxime.coquelin@redhat.com>
To: Tonghao Zhang <xiangxia.m.yue@gmail.com>,
"Tan, Jianfeng" <jianfeng.tan@intel.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH 2/2] vhost: add pipe event for optimizing negotiating
Date: Fri, 30 Mar 2018 09:57:15 +0200 [thread overview]
Message-ID: <148da049-ee99-ce61-0261-59c3917033a1@redhat.com> (raw)
In-Reply-To: <CAMDZJNU0NneWgi+GxfS5ZySAxjKCSRgcE2E-mHttpdC1YN66Ww@mail.gmail.com>
Hi Xiangxia,
On 03/29/2018 01:24 PM, Tonghao Zhang wrote:
> On Thu, Mar 29, 2018 at 3:32 PM, Tan, Jianfeng <jianfeng.tan@intel.com> wrote:
>> Hi Xiangxia,
>>
>>> -----Original Message-----
>>> From: xiangxia.m.yue@gmail.com [mailto:xiangxia.m.yue@gmail.com]
>>> Sent: Wednesday, March 28, 2018 1:49 PM
>>> To: Tan, Jianfeng
>>> Cc: dev@dpdk.org; Tonghao Zhang
>>> Subject: [PATCH 2/2] vhost: add pipe event for optimizing negotiating
>>>
>>> From: Tonghao Zhang <xiangxia.m.yue@gmail.com>
>>>
>>> When vhost—user connects qemu successfully, dpdk will call
>>
>> Typo: "-"
> v2 will update it. :)
>
>>> the vhost_user_add_connection to add unix socket fd to poll.
>>> And fdset_add only set the socket fd to a fdentry while poll
>>> may sleep now. In a general case, this is no problem. But if
>>> we use hot update for vhost-user, most downtime of VMs network
>>> is 750+ms. This patch adds pipe event, so after connections are
>>> ok, dpdk rebuild the poll immediately. With this patch, the
>>> most downtime is 20~30ms.
>>>
>>> Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
>>> ---
>>> lib/librte_vhost/fd_man.c | 49
>>> +++++++++++++++++++++++++++++++++++++++++++++++
>>> lib/librte_vhost/fd_man.h | 16 ++++++++++++++++
>>> lib/librte_vhost/socket.c | 14 ++++++++++++++
>>> 3 files changed, 79 insertions(+)
>>>
>>> diff --git a/lib/librte_vhost/fd_man.c b/lib/librte_vhost/fd_man.c
>>> index 181711c..7716757 100644
>>> --- a/lib/librte_vhost/fd_man.c
>>> +++ b/lib/librte_vhost/fd_man.c
>>> @@ -15,6 +15,7 @@
>>> #include <rte_log.h>
>>>
>>> #include "fd_man.h"
>>> +#include "vhost.h"
>>
>> This is a nice finding and solution, however, I don't think we shall include vhost header file in fd related files. Actually, I did not find out why you need to include this header file.
>
> Hi Jianfeng, thanks for your review. In the fdset_pipe_init function,
> I call the RTE_LOG with VHOST_CONFIG, when init pipe not successfully.
> So I included the vhost header file.
Maybe better to create VHOST_SOCKET in socket.c, than including vhost.h
that creates a layer violation:
#define RTE_LOGTYPE_VHOST_SOCKET RTE_LOGTYPE_USER1
Is it Ok for you? I so no need to resend, I can handle the change when
applying.
Thanks,
Maxime
>
>
>> Thanks,
>> Jianfeng
>>
>>>
>>> #define FDPOLLERR (POLLERR | POLLHUP | POLLNVAL)
>>>
>>> @@ -272,3 +273,51 @@
>>>
>>> return NULL;
>>> }
>>> +
>>> +static void
>>> +fdset_pipe_read_cb(int readfd, void *dat __rte_unused,
>>> + int *remove __rte_unused)
>>> +{
>>> + char charbuf[16];
>>> + read(readfd, charbuf, sizeof(charbuf));
>>> +}
>>> +
>>> +void
>>> +fdset_pipe_uninit(struct fdset *fdset)
>>> +{
>>> + fdset_del(fdset, fdset->u.readfd);
>>> + close(fdset->u.readfd);
>>> + close(fdset->u.writefd);
>>> +}
>>> +
>>> +int
>>> +fdset_pipe_init(struct fdset *fdset)
>>> +{
>>> + int ret;
>>> +
>>> + if (pipe(fdset->u.pipefd) < 0) {
>>> + RTE_LOG(ERR, VHOST_CONFIG,
>>> + "failed to create pipe for vhost fdset\n");
>>> + return -1;
>>> + }
>>> +
>>> + ret = fdset_add(fdset, fdset->u.readfd,
>>> + fdset_pipe_read_cb, NULL, NULL);
>>> +
>>> + if (ret < 0) {
>>> + RTE_LOG(ERR, VHOST_CONFIG,
>>> + "failed to add pipe readfd %d into vhost server
>>> fdset\n",
>>> + fdset->u.readfd);
>>> +
>>> + fdset_pipe_uninit(fdset);
>>> + return -1;
>>> + }
>>> +
>>> + return 0;
>>> +}
>>> +
>>> +void
>>> +fdset_pipe_notify(struct fdset *fdset)
>>> +{
>>> + write(fdset->u.writefd, "1", 1);
>>> +}
>>> diff --git a/lib/librte_vhost/fd_man.h b/lib/librte_vhost/fd_man.h
>>> index 3a9276c..76a42fb 100644
>>> --- a/lib/librte_vhost/fd_man.h
>>> +++ b/lib/librte_vhost/fd_man.h
>>> @@ -25,6 +25,16 @@ struct fdset {
>>> struct fdentry fd[MAX_FDS];
>>> pthread_mutex_t fd_mutex;
>>> int num; /* current fd number of this fdset */
>>> +
>>> + union pipefds {
>>> + struct {
>>> + int pipefd[2];
>>> + };
>>> + struct {
>>> + int readfd;
>>> + int writefd;
>>> + };
>>> + } u;
>>> };
>>>
>>>
>>> @@ -37,4 +47,10 @@ int fdset_add(struct fdset *pfdset, int fd,
>>>
>>> void *fdset_event_dispatch(void *arg);
>>>
>>> +int fdset_pipe_init(struct fdset *fdset);
>>> +
>>> +void fdset_pipe_uninit(struct fdset *fdset);
>>> +
>>> +void fdset_pipe_notify(struct fdset *fdset);
>>> +
>>> #endif
>>> diff --git a/lib/librte_vhost/socket.c b/lib/librte_vhost/socket.c
>>> index 95bed78..795239c 100644
>>> --- a/lib/librte_vhost/socket.c
>>> +++ b/lib/librte_vhost/socket.c
>>> @@ -231,6 +231,8 @@ struct vhost_user {
>>> pthread_mutex_lock(&vsocket->conn_mutex);
>>> TAILQ_INSERT_TAIL(&vsocket->conn_list, conn, next);
>>> pthread_mutex_unlock(&vsocket->conn_mutex);
>>> +
>>> + fdset_pipe_notify(&vhost_user.fdset);
>>> return;
>>>
>>> err:
>>> @@ -829,11 +831,23 @@ struct vhost_device_ops const *
>>> return -1;
>>>
>>> if (fdset_tid == 0) {
>>> + /**
>>> + * create a pipe which will be waited by poll and notified to
>>> + * rebuild the wait list of poll.
>>> + */
>>> + if (fdset_pipe_init(&vhost_user.fdset) < 0) {
>>> + RTE_LOG(ERR, VHOST_CONFIG,
>>> + "failed to create pipe for vhost fdset\n");
>>> + return -1;
>>> + }
>>> +
>>> int ret = pthread_create(&fdset_tid, NULL, fdset_event_dispatch,
>>> &vhost_user.fdset);
>>> if (ret != 0) {
>>> RTE_LOG(ERR, VHOST_CONFIG,
>>> "failed to create fdset handling thread");
>>> +
>>> + fdset_pipe_uninit(&vhost_user.fdset);
>>> return -1;
>>> } else {
>>> snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN,
>>> --
>>> 1.8.3.1
>>
next prev parent reply other threads:[~2018-03-30 7:57 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-03-28 5:49 [dpdk-dev] [PATCH 0/2] " xiangxia.m.yue
2018-03-28 5:49 ` [dpdk-dev] [PATCH 1/2] vhost: move stdbool header file to vhost.h xiangxia.m.yue
2018-03-30 8:09 ` Maxime Coquelin
2018-03-28 5:49 ` [dpdk-dev] [PATCH 2/2] vhost: add pipe event for optimizing negotiating xiangxia.m.yue
2018-03-29 7:32 ` Tan, Jianfeng
2018-03-29 11:24 ` Tonghao Zhang
2018-03-30 7:57 ` Maxime Coquelin [this message]
2018-03-30 8:07 ` Tonghao Zhang
2018-03-30 8:09 ` Maxime Coquelin
2018-03-30 8:17 ` [dpdk-dev] [PATCH 0/2] " Maxime Coquelin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=148da049-ee99-ce61-0261-59c3917033a1@redhat.com \
--to=maxime.coquelin@redhat.com \
--cc=dev@dpdk.org \
--cc=jianfeng.tan@intel.com \
--cc=xiangxia.m.yue@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).