| Message ID | 1454412668-6472-1-git-send-email-paul.durrant@citrix.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Feb 02, 2016 at 11:31:08AM +0000, Paul Durrant wrote: [...] > +static int xen_register_mcast_ctrl_watch(struct xenbus_device *dev, > + struct xenvif *vif) > +{ > + int err = 0; > + char *node; > + unsigned maxlen = strlen(dev->otherend) + > + sizeof("/request-multicast-control"); > + > + if (vif->mcast_ctrl_watch.node) > + return -EADDRINUSE; > + Sorry if I didn't make myself clear enough. I think EADDRINUSE should also be logged. Wei.
> -----Original Message----- > From: Wei Liu [mailto:wei.liu2@citrix.com] > Sent: 02 February 2016 11:47 > To: Paul Durrant > Cc: netdev@vger.kernel.org; xen-devel@lists.xenproject.org; Ian Campbell; > Wei Liu > Subject: Re: [PATCH net-next v2] xen-netback: implement dynamic multicast > control > > On Tue, Feb 02, 2016 at 11:31:08AM +0000, Paul Durrant wrote: > [...] > > +static int xen_register_mcast_ctrl_watch(struct xenbus_device *dev, > > + struct xenvif *vif) > > +{ > > + int err = 0; > > + char *node; > > + unsigned maxlen = strlen(dev->otherend) + > > + sizeof("/request-multicast-control"); > > + > > + if (vif->mcast_ctrl_watch.node) > > + return -EADDRINUSE; > > + > > Sorry if I didn't make myself clear enough. I think EADDRINUSE should > also be logged. > Ok. Not sure how much noise that will cause though. Paul > Wei.
On Tue, 2016-02-02 at 11:48 +0000, Paul Durrant wrote: > > -----Original Message----- > > From: Wei Liu [mailto:wei.liu2@citrix.com] > > Sent: 02 February 2016 11:47 > > To: Paul Durrant > > Cc: netdev@vger.kernel.org; xen-devel@lists.xenproject.org; Ian > > Campbell; > > Wei Liu > > Subject: Re: [PATCH net-next v2] xen-netback: implement dynamic > > multicast > > control > > > > On Tue, Feb 02, 2016 at 11:31:08AM +0000, Paul Durrant wrote: > > [...] > > > +static int xen_register_mcast_ctrl_watch(struct xenbus_device *dev, > > > + struct xenvif *vif) > > > +{ > > > + int err = 0; > > > + char *node; > > > + unsigned maxlen = strlen(dev->otherend) + > > > + sizeof("/request-multicast-control"); > > > + > > > + if (vif->mcast_ctrl_watch.node) > > > + return -EADDRINUSE; > > > + > > > > Sorry if I didn't make myself clear enough. I think EADDRINUSE should > > also be logged. > > > > Ok. Not sure how much noise that will cause though. If this is a guest trigerable path then don't forget to use the ratelimiting printk options.... Ian.
> -----Original Message----- > From: Ian Campbell [mailto:ian.campbell@citrix.com] > Sent: 02 February 2016 11:54 > To: Paul Durrant; Wei Liu > Cc: netdev@vger.kernel.org; xen-devel@lists.xenproject.org > Subject: Re: [PATCH net-next v2] xen-netback: implement dynamic multicast > control > > On Tue, 2016-02-02 at 11:48 +0000, Paul Durrant wrote: > > > -----Original Message----- > > > From: Wei Liu [mailto:wei.liu2@citrix.com] > > > Sent: 02 February 2016 11:47 > > > To: Paul Durrant > > > Cc: netdev@vger.kernel.org; xen-devel@lists.xenproject.org; Ian > > > Campbell; > > > Wei Liu > > > Subject: Re: [PATCH net-next v2] xen-netback: implement dynamic > > > multicast > > > control > > > > > > On Tue, Feb 02, 2016 at 11:31:08AM +0000, Paul Durrant wrote: > > > [...] > > > > +static int xen_register_mcast_ctrl_watch(struct xenbus_device *dev, > > > > + struct xenvif *vif) > > > > +{ > > > > + int err = 0; > > > > + char *node; > > > > + unsigned maxlen = strlen(dev->otherend) + > > > > + sizeof("/request-multicast-control"); > > > > + > > > > + if (vif->mcast_ctrl_watch.node) > > > > + return -EADDRINUSE; > > > > + > > > > > > Sorry if I didn't make myself clear enough. I think EADDRINUSE should > > > also be logged. > > > > > > > Ok. Not sure how much noise that will cause though. > > If this is a guest trigerable path then don't forget to use the > ratelimiting printk options.... > I *think* it is essentially an error path (i.e should not happen) and the function is only called when connecting backend and frontend, which is fairly well rate-limited by other things, but I'll stick a rate-limit in there just in case. Paul > Ian.
diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-netback/common.h index 0333ab0..1128252 100644 --- a/drivers/net/xen-netback/common.h +++ b/drivers/net/xen-netback/common.h @@ -251,6 +251,7 @@ struct xenvif { unsigned int stalled_queues; struct xenbus_watch credit_watch; + struct xenbus_watch mcast_ctrl_watch; spinlock_t lock; diff --git a/drivers/net/xen-netback/xenbus.c b/drivers/net/xen-netback/xenbus.c index 56ebd82..f8345ca 100644 --- a/drivers/net/xen-netback/xenbus.c +++ b/drivers/net/xen-netback/xenbus.c @@ -327,7 +327,7 @@ static int netback_probe(struct xenbus_device *dev, goto abort_transaction; } - /* We support multicast-control. */ + /* We support dynamic multicast-control. */ err = xenbus_printf(xbt, dev->nodename, "feature-multicast-control", "%d", 1); if (err) { @@ -335,6 +335,14 @@ static int netback_probe(struct xenbus_device *dev, goto abort_transaction; } + err = xenbus_printf(xbt, dev->nodename, + "feature-dynamic-multicast-control", + "%d", 1); + if (err) { + message = "writing feature-dynamic-multicast-control"; + goto abort_transaction; + } + err = xenbus_transaction_end(xbt, 0); } while (err == -EAGAIN); @@ -683,7 +691,8 @@ static void xen_net_rate_changed(struct xenbus_watch *watch, } } -static int xen_register_watchers(struct xenbus_device *dev, struct xenvif *vif) +static int xen_register_credit_watch(struct xenbus_device *dev, + struct xenvif *vif) { int err = 0; char *node; @@ -708,7 +717,7 @@ static int xen_register_watchers(struct xenbus_device *dev, struct xenvif *vif) return err; } -static void xen_unregister_watchers(struct xenvif *vif) +static void xen_unregister_credit_watch(struct xenvif *vif) { if (vif->credit_watch.node) { unregister_xenbus_watch(&vif->credit_watch); @@ -717,6 +726,73 @@ static void xen_unregister_watchers(struct xenvif *vif) } } +static void xen_mcast_ctrl_changed(struct xenbus_watch *watch, + const char **vec, unsigned int len) +{ + struct xenvif *vif = container_of(watch, struct xenvif, + mcast_ctrl_watch); + struct xenbus_device *dev = xenvif_to_xenbus_device(vif); + int val; + + if (xenbus_scanf(XBT_NIL, dev->otherend, + "request-multicast-control", "%d", &val) < 0) + val = 0; + vif->multicast_control = !!val; +} + +static int xen_register_mcast_ctrl_watch(struct xenbus_device *dev, + struct xenvif *vif) +{ + int err = 0; + char *node; + unsigned maxlen = strlen(dev->otherend) + + sizeof("/request-multicast-control"); + + if (vif->mcast_ctrl_watch.node) + return -EADDRINUSE; + + node = kmalloc(maxlen, GFP_KERNEL); + if (!node) { + pr_err("Failed to allocate memory for watch\n"); + return -ENOMEM; + } + snprintf(node, maxlen, "%s/request-multicast-control", + dev->otherend); + vif->mcast_ctrl_watch.node = node; + vif->mcast_ctrl_watch.callback = xen_mcast_ctrl_changed; + err = register_xenbus_watch(&vif->mcast_ctrl_watch); + if (err) { + pr_err("Failed to set watcher %s\n", + vif->mcast_ctrl_watch.node); + kfree(node); + vif->mcast_ctrl_watch.node = NULL; + vif->mcast_ctrl_watch.callback = NULL; + } + return err; +} + +static void xen_unregister_mcast_ctrl_watch(struct xenvif *vif) +{ + if (vif->mcast_ctrl_watch.node) { + unregister_xenbus_watch(&vif->mcast_ctrl_watch); + kfree(vif->mcast_ctrl_watch.node); + vif->mcast_ctrl_watch.node = NULL; + } +} + +static void xen_register_watchers(struct xenbus_device *dev, + struct xenvif *vif) +{ + xen_register_credit_watch(dev, vif); + xen_register_mcast_ctrl_watch(dev, vif); +} + +static void xen_unregister_watchers(struct xenvif *vif) +{ + xen_unregister_mcast_ctrl_watch(vif); + xen_unregister_credit_watch(vif); +} + static void unregister_hotplug_status_watch(struct backend_info *be) { if (be->have_hotplug_status_watch) { @@ -1030,11 +1106,6 @@ static int read_xenbus_vif_flags(struct backend_info *be) val = 0; vif->ipv6_csum = !!val; - if (xenbus_scanf(XBT_NIL, dev->otherend, "request-multicast-control", - "%d", &val) < 0) - val = 0; - vif->multicast_control = !!val; - return 0; }
My recent patch to the Xen Project documents a protocol for 'dynamic multicast control' in netif.h. This extends the previous multicast control protocol to not require a shared ring reconnection to turn the feature off. Instead the backend watches the "request-multicast-control" key in xenstore and turns the feature off if the key value is written to zero. This patch adds support for dynamic multicast control in xen-netback. Signed-off-by: Paul Durrant <paul.durrant@citrix.com> Cc: Ian Campbell <ian.campbell@citrix.com> Cc: Wei Liu <wei.liu2@citrix.com> --- v2: - Log error on failure to allocate memory for multicast watch --- drivers/net/xen-netback/common.h | 1 + drivers/net/xen-netback/xenbus.c | 87 ++++++++++++++++++++++++++++++++++++---- 2 files changed, 80 insertions(+), 8 deletions(-)