Re: Issue with mesh provisioning.

[Steve Brown <sbrown@ewol.com>]

Hi Luiz,

Would this be an appropriate fix.

It seems to solve the problem.

diff --git a/tools/mesh/gatt.c b/tools/mesh/gatt.c
index 2269a20cf..0a942d4eb 100644
--- a/tools/mesh/gatt.c
+++ b/tools/mesh/gatt.c
@@ -415,8 +415,11 @@ static bool sock_read(struct io *io, bool prov, void *user_data)
 	msg.msg_iovlen = 1;
 
 	while ((len = recvmsg(fd, &msg, MSG_DONTWAIT))) {
-		if (len <= 0)
+		if (len <= 0) {
+			if (errno == EAGAIN)
+				break;
 			return false;
+		}
 
 		res = buf;
 		len_sar = mesh_gatt_sar(&res, len);

If so, I'll submit a patch.

Steve

On Wed, 2019-05-15 at 13:48 +0300, Von Dentz, Luiz wrote:
> Hi Steve,
> 
> That specific patch doesn't change the meshctl, the following one
> does:
> 
> commit d6eec1b67d6d9e3336b7e3cd8f655baf408d864a
> Author: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> Date:   Mon Nov 19 17:35:54 2018 +0200
> 
>     meshctl: Switch from write to sendmsg for Acquire*
> 
>     Use sendmsg with MSG_NOSIGNAL to prevent crashes involving
> SIGPIPE.
> 
> Not sure what is the problem with that since we just switch from a
> pipe to a socketpair.
> 
> On Wed, May 15, 2019 at 1:01 PM Steve Brown <sbrown@ewol.com> wrote:
> > On Tue, 2019-05-14 at 15:25 -0400, Steve Brown wrote:
> > > On Tue, 2019-05-14 at 14:50 +0530, Shaik Mahaboob Nazeer wrote:
> > > > hi,
> > > > 
> > > > -> We tried Zephyr Mesh onoff example on nrf52 eval kit to test
> > > > the
> > > > Mesh Functionality.For provisioning of the node
> > > >    we are using meshctl, built from the latest current master
> > > > of
> > > > the
> > > > bluez source.So during the provisioning of the node when we
> > > >    start the "meshctl" and run the "discover-unprovisioned on"
> > > > we
> > > > are
> > > > able to get the Device UUID for the provisioning.
> > > > 
> > > >    Later when we try to provision the node using the "provision
> > > > <UUID>" command it initiate the provision and does not wait
> > > >    for the user to enter OOB .You can find the log as below.
> > > >    /*******************************************/
> > > > 
> > > 
> > > --- snip ---
> > > 
> > > >   /**************************************/
> > > > 
> > > > Could you elaborate a little on the status of the current
> > > > master of
> > > > bluez in regards to BLE mesh, is it usable or should we look to
> > > > older
> > > > versions?
> > > > 
> > > > Do you know if there is a common issue while provisioning of
> > > > the
> > > > node
> > > > with the current master and
> > > > is there anything that need to be configured before
> > > > provisioning?
> > > > 
> > > I get the same results.
> > > 
> > > I check out 48a2c3db1ccff28a615fac56066ca7d721f98d0a
> > > 
> > > That was the last time I used meshctl.
> > > 
> > > It requests an oob string and completes provisioning.
> > > 
> > > Steve
> > > 
> > 
> > I bisected to
> > 
> > commit 91df85690cc1943e46ef90561dd1ac0d673ee3d7
> > Author: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > Date:   Mon Nov 19 15:36:15 2018 +0200
> > 
> >     client: Switch from write to sendmsg for Acquire*
> > 
> >     Use sendmsg with MSG_NOSIGNAL to prevent crashes involving
> > SIGPIPE.
> > 
> > 
> >