qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Jiachen Zhang <zhangjiachen.jaycee@bytedance.com>
To: Stefan Hajnoczi <stefanha@gmail.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>,
	Stefan Hajnoczi <stefanha@redhat.com>,
	 virtio-fs-list <virtio-fs@redhat.com>,
	QEMU <qemu-devel@nongnu.org>,  Connor Kuehl <ckuehl@redhat.com>
Subject: Re: Re: [PATCH] virtiofsd: support setting multiple request virtqueues
Date: Mon, 12 Dec 2022 23:51:50 +0800	[thread overview]
Message-ID: <CAFQAk7gj_6mZUJrqtf4Q0O1F+jtt9fesUPFevNN+12HBMqdU8w@mail.gmail.com> (raw)
In-Reply-To: <CAJSP0QW2fugP4PPZ=sN4bmPA8HPzQxc=AzPKHe+OUeu++TUKog@mail.gmail.com>

On Mon, Dec 12, 2022 at 10:48 PM Stefan Hajnoczi <stefanha@gmail.com> wrote:
>
> On Sun, 11 Dec 2022 at 05:49, Jiachen Zhang
> <zhangjiachen.jaycee@bytedance.com> wrote:
> >
> > Add an option '-o num_request_queues' to configure the queue number,
> > currently the total number of vqs should be (1 hiprio queue +
> > num_request_queues).
> >
> > The code is based on Connor's previous version in the virtio-fs
> > mailing-list [1], but change the semantic of the new option from total
> > queue number to request queue number.
> >
> > The corresponding virtio-fs kernel part modification can be found at the
> > mail [2].
> >
> > Link:
> > [1] https://www.mail-archive.com/virtio-fs@redhat.com/msg03333.html
> > [2] https://lore.kernel.org/linux-fsdevel/20221211103857.25805-1-zhangjiachen.jaycee@bytedance.com/
> > Suggested-by: Stefan Hajnoczi <stefanha@redhat.com>
> > Cc: Connor Kuehl <ckuehl@redhat.com>
> > Signed-off-by: Jiachen Zhang <zhangjiachen.jaycee@bytedance.com>
> > ---
> >  tools/virtiofsd/fuse_lowlevel.c  |  4 ++--
> >  tools/virtiofsd/fuse_lowlevel.h  |  2 +-
> >  tools/virtiofsd/fuse_virtio.c    | 20 +++-----------------
> >  tools/virtiofsd/fuse_virtio.h    |  2 +-
> >  tools/virtiofsd/helper.c         |  4 ++++
> >  tools/virtiofsd/passthrough_ll.c | 12 +++++++++++-
> >  6 files changed, 22 insertions(+), 22 deletions(-)
>
> QEMU's C virtiofsd is no longer actively developed. New development
> happens in the Rust https://gitlab.com/virtio-fs/virtiofsd codebase.
> Are you sure you want to continue using C virtiofsd?
>

Thanks for the reminder, I thought the C virtiofsd was the reference
implementation.

> >
> > diff --git a/tools/virtiofsd/fuse_lowlevel.c b/tools/virtiofsd/fuse_lowlevel.c
> > index 2f08471627..528492d2cf 100644
> > --- a/tools/virtiofsd/fuse_lowlevel.c
> > +++ b/tools/virtiofsd/fuse_lowlevel.c
> > @@ -2708,9 +2708,9 @@ out1:
> >      return NULL;
> >  }
> >
> > -int fuse_session_mount(struct fuse_session *se)
> > +int fuse_session_mount(struct fuse_session *se, unsigned int num_queues)
> >  {
> > -    return virtio_session_mount(se);
> > +    return virtio_session_mount(se, num_queues);
> >  }
> >
> >  int fuse_session_fd(struct fuse_session *se)
> > diff --git a/tools/virtiofsd/fuse_lowlevel.h b/tools/virtiofsd/fuse_lowlevel.h
> > index b889dae4de..aee02d3e91 100644
> > --- a/tools/virtiofsd/fuse_lowlevel.h
> > +++ b/tools/virtiofsd/fuse_lowlevel.h
> > @@ -1856,7 +1856,7 @@ struct fuse_session *fuse_session_new(struct fuse_args *args,
> >   *
> >   * @return 0 on success, -1 on failure.
> >   **/
> > -int fuse_session_mount(struct fuse_session *se);
> > +int fuse_session_mount(struct fuse_session *se, unsigned int num_queues);
> >
> >  /**
> >   * Enter a single threaded, blocking event loop.
> > diff --git a/tools/virtiofsd/fuse_virtio.c b/tools/virtiofsd/fuse_virtio.c
> > index 9368e292e4..a3ebcbea8e 100644
> > --- a/tools/virtiofsd/fuse_virtio.c
> > +++ b/tools/virtiofsd/fuse_virtio.c
> > @@ -762,20 +762,6 @@ static void fv_queue_set_started(VuDev *dev, int qidx, bool started)
> >               started);
> >      assert(qidx >= 0);
> >
> > -    /*
> > -     * Ignore additional request queues for now.  passthrough_ll.c must be
> > -     * audited for thread-safety issues first.  It was written with a
> > -     * well-behaved client in mind and may not protect against all types of
> > -     * races yet.
> > -     */
> > -    if (qidx > 1) {
> > -        fuse_log(FUSE_LOG_ERR,
> > -                 "%s: multiple request queues not yet implemented, please only "
> > -                 "configure 1 request queue\n",
> > -                 __func__);
> > -        exit(EXIT_FAILURE);
> > -    }
>
> Please include an explanation of why it's safe to drop this now. Have
> the thread-safety issues mentioned in the comment been resolved?

The struct fv_VuDev is protected by vu_dispatch_rwlock. Also the fuse
handlers in passthrough_ll.c are already concurrent with a (size > 1)
g_thread_pool.

>
> > -
> >      if (started) {
> >          /* Fire up a thread to watch this queue */
> >          if (qidx >= vud->nqueues) {
> > @@ -1011,7 +997,7 @@ static int fv_create_listen_socket(struct fuse_session *se)
> >      return 0;
> >  }
> >
> > -int virtio_session_mount(struct fuse_session *se)
> > +int virtio_session_mount(struct fuse_session *se, unsigned int num_queues)
> >  {
> >      int ret;
> >
> > @@ -1057,8 +1043,8 @@ int virtio_session_mount(struct fuse_session *se)
> >      se->vu_socketfd = data_sock;
> >      se->virtio_dev->se = se;
> >      pthread_rwlock_init(&se->virtio_dev->vu_dispatch_rwlock, NULL);
> > -    if (!vu_init(&se->virtio_dev->dev, 2, se->vu_socketfd, fv_panic, NULL,
> > -                 fv_set_watch, fv_remove_watch, &fv_iface)) {
> > +    if (!vu_init(&se->virtio_dev->dev, num_queues, se->vu_socketfd,
> > +                fv_panic, NULL, fv_set_watch, fv_remove_watch, &fv_iface)) {
> >          fuse_log(FUSE_LOG_ERR, "%s: vu_init failed\n", __func__);
> >          return -1;
> >      }
> > diff --git a/tools/virtiofsd/fuse_virtio.h b/tools/virtiofsd/fuse_virtio.h
> > index 111684032c..a0e78b9b84 100644
> > --- a/tools/virtiofsd/fuse_virtio.h
> > +++ b/tools/virtiofsd/fuse_virtio.h
> > @@ -18,7 +18,7 @@
> >
> >  struct fuse_session;
> >
> > -int virtio_session_mount(struct fuse_session *se);
> > +int virtio_session_mount(struct fuse_session *se, unsigned int num_queues);
> >  void virtio_session_close(struct fuse_session *se);
> >  int virtio_loop(struct fuse_session *se);
> >
> > diff --git a/tools/virtiofsd/helper.c b/tools/virtiofsd/helper.c
> > index f5f66f292c..b5138ce17d 100644
> > --- a/tools/virtiofsd/helper.c
> > +++ b/tools/virtiofsd/helper.c
> > @@ -191,6 +191,10 @@ void fuse_cmdline_help(void)
> >             "    -o killpriv_v2/no_killpriv_v2\n"
> >             "                               Enable/Disable FUSE_HANDLE_KILLPRIV_V2.\n"
> >             "                               (default: enabled as long as client supports it)\n"
> > +           "    -o num_request_queues=<num>\n"
> > +           "                               set maximum number of request virtqueues\n"
> > +           "                               supported by virtiofsd"
> > +           "                               default: 1\n"
> >             );
> >  }
> >
> > diff --git a/tools/virtiofsd/passthrough_ll.c b/tools/virtiofsd/passthrough_ll.c
> > index 20f0f41f99..f9d8075835 100644
> > --- a/tools/virtiofsd/passthrough_ll.c
> > +++ b/tools/virtiofsd/passthrough_ll.c
> > @@ -182,6 +182,7 @@ struct lo_data {
> >      /* Keeps track if /proc/<pid>/attr/fscreate should be used or not */
> >      bool use_fscreate;
> >      int user_security_label;
> > +    int num_request_queues;
> >  };
> >
> >  static const struct fuse_opt lo_opts[] = {
> > @@ -218,6 +219,8 @@ static const struct fuse_opt lo_opts[] = {
> >      { "no_posix_acl", offsetof(struct lo_data, user_posix_acl), 0 },
> >      { "security_label", offsetof(struct lo_data, user_security_label), 1 },
> >      { "no_security_label", offsetof(struct lo_data, user_security_label), 0 },
> > +    { "num_request_queues=%d",
> > +      offsetof(struct lo_data, num_request_queues), 1 },
> >      FUSE_OPT_END
> >  };
> >  static bool use_syslog = false;
> > @@ -4479,6 +4482,12 @@ int main(int argc, char *argv[])
> >
> >      lo.use_statx = true;
> >
> > +    if (lo.num_request_queues < 1) {
> > +        fuse_log(FUSE_LOG_ERR, "num_request_queues must be at least 1 (got %d)"
> > +                 "\n", lo.num_request_queues);
> > +        exit(1);
> > +    }
> > +
> >      se = fuse_session_new(&args, &lo_oper, sizeof(lo_oper), &lo);
> >      if (se == NULL) {
> >          goto err_out1;
> > @@ -4488,7 +4497,8 @@ int main(int argc, char *argv[])
> >          goto err_out2;
> >      }
> >
> > -    if (fuse_session_mount(se) != 0) {
> > +    /* There will be 1 hirpio queue plus lo.num_request_queues request queues */
>
> s/hirpio/hiprio/
>

Thanks, will fix it.

Best regards,
Jiachen

> > +    if (fuse_session_mount(se, lo.num_request_queues + 1) != 0) {
> >          goto err_out3;
> >      }
> >
> > --
> > 2.20.1
> >
> >


      reply	other threads:[~2022-12-12 15:53 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-12-11 10:47 [PATCH] virtiofsd: support setting multiple request virtqueues Jiachen Zhang
2022-12-12 14:48 ` Stefan Hajnoczi
2022-12-12 15:51   ` Jiachen Zhang [this message]

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=CAFQAk7gj_6mZUJrqtf4Q0O1F+jtt9fesUPFevNN+12HBMqdU8w@mail.gmail.com \
    --to=zhangjiachen.jaycee@bytedance.com \
    --cc=ckuehl@redhat.com \
    --cc=dgilbert@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=stefanha@gmail.com \
    --cc=stefanha@redhat.com \
    --cc=virtio-fs@redhat.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).