On 1/5/21 2:03 AM, Bjorn Andersson wrote:
> On Tue 22 Dec 04:57 CST 2020, Arnaud Pouliquen wrote:
>
>> The name service announcement is not sent if no endpoint is created by
>> default. If the destination address is not precised by the
>> application when creating the device (thanks to the RPMsg CTRL interface),
>> it is not possible to have a valid RPMsg channel.
>>
>
> In the Qualcomm transports, the chinfo.name is used to identify the
> channel, so there it's valid to create a endpoint without dst.
So to be move in rpmsg virtio...either reporting an error or generating a NS
announcement.
Thanks,
Arnaud
>
> Regards,
> Bjorn
>
>> Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@xxxxxxxxxxx>
>> ---
>> drivers/rpmsg/rpmsg_char.c | 10 ++++++++++
>> 1 file changed, 10 insertions(+)
>>
>> diff --git a/drivers/rpmsg/rpmsg_char.c b/drivers/rpmsg/rpmsg_char.c
>> index 4b0674a2e3e9..8b1928594d10 100644
>> --- a/drivers/rpmsg/rpmsg_char.c
>> +++ b/drivers/rpmsg/rpmsg_char.c
>> @@ -305,6 +305,16 @@ static int rpmsg_chrdev_probe(struct rpmsg_device *rpdev)
>> struct device *dev;
>> int ret;
>>
>> + /* There is not ept created by default. As consequence there is not NS
>> + * announcement and the destination address has to be set.
>> + * this limitation could be solved in future by adding a helper in
>> + * rpmsg_ns.
>> + */
>> + if (rpdev->dst == RPMSG_ADDR_ANY) {
>> + dev_err(dev, "destination address invalid (%d)\n", rpdev->dst);
>> + return -EINVAL;
>> + }
>> +
>> eptdev = kzalloc(sizeof(*eptdev), GFP_KERNEL);
>> if (!eptdev)
>> return -ENOMEM;
>> --
>> 2.17.1
>>
