[no subject]

Anyway here I only share my opinion. It is not my role to give the direction.

> 
>> And What about using the "software flow control" instead? [1]
>>
>> [1] https://en.wikipedia.org/wiki/Software_flow_control
>>
>>> Â  Â Â Â Â Â  return 0;
>>> @@ -137,6 +152,7 @@ static int rpmsg_eptdev_open(struct inode *inode,
>>> struct file *filp)
>>> Â Â Â Â Â Â Â Â Â  return -EINVAL;
>>> Â Â Â Â Â  }
>>> Â  +Â Â Â  ept->sig_cb = rpmsg_sigs_cb;
>>> Â Â Â Â Â  eptdev->ept = ept;
>>> Â Â Â Â Â  filp->private_data = eptdev;
>>> Â  @@ -155,6 +171,7 @@ static int rpmsg_eptdev_release(struct inode
>>> *inode, struct file *filp)
>>> Â Â Â Â Â Â Â Â Â  eptdev->ept = NULL;
>>> Â Â Â Â Â  }
>>> Â Â Â Â Â  mutex_unlock(&eptdev->ept_lock);
>>> +Â Â Â  eptdev->sig_pending = false;
>>> Â  Â Â Â Â Â  /* Discard all SKBs */
>>> Â Â Â Â Â  skb_queue_purge(&eptdev->queue);
>>> @@ -265,6 +282,9 @@ static __poll_t rpmsg_eptdev_poll(struct file
>>> *filp, poll_table *wait)
>>> Â Â Â Â Â  if (!skb_queue_empty(&eptdev->queue))
>>> Â Â Â Â Â Â Â Â Â  mask |= EPOLLIN | EPOLLRDNORM;
>>> Â  +Â Â Â  if (eptdev->sig_pending)
>>> +Â Â Â Â Â Â Â  mask |= EPOLLPRI;
>>> +
>>> Â Â Â Â Â  mask |= rpmsg_poll(eptdev->ept, filp, wait);
>>> Â  Â Â Â Â Â  return mask;
>>> @@ -274,11 +294,30 @@ static long rpmsg_eptdev_ioctl(struct file *fp,
>>> unsigned int cmd,
>>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â  unsigned long arg)
>>> Â  {
>>> Â Â Â Â Â  struct rpmsg_eptdev *eptdev = fp->private_data;
>>> +Â Â Â  bool set;
>>> +Â Â Â  u32 val;
>>> +Â Â Â  int ret;
>>> Â  -Â Â Â  if (cmd != RPMSG_DESTROY_EPT_IOCTL)
>>> -Â Â Â Â Â Â Â  return -EINVAL;
>>> +Â Â Â  switch (cmd) {
>>> +Â Â Â  case TIOCMGET:
>>> +Â Â Â Â Â Â Â  eptdev->sig_pending = false;
>>> +Â Â Â Â Â Â Â  ret = put_user(eptdev->rsigs, (int __user *)arg);
>>> +Â Â Â Â Â Â Â  break;
>>> +Â Â Â  case TIOCMSET:
>>> +Â Â Â Â Â Â Â  ret = get_user(val, (int __user *)arg);
>>> +Â Â Â Â Â Â Â  if (ret)
>>> +Â Â Â Â Â Â Â Â Â Â Â  break;
>>> +Â Â Â Â Â Â Â  set = (val & TIOCM_DTR) ? true : false;
>>> +Â Â Â Â Â Â Â  ret = rpmsg_set_flow_control(eptdev->ept, set);
>>> +Â Â Â Â Â Â Â  break;
>> Could this directly be handled by the driver on open close?
>> If application wants to suspend the link it could just close de
>> /dev/rpmsgX.
> All clients may not require setting flow control.

Agree, but this could be conditioned by rpdrv->signals, right?
And this could avoid to expose controls 
- in open/close the rpmsg_set_flow_control would be called,
- in rpmsg_eptdev_write_iter an error would be returned ( or 0)
  if remote side has suspended the transmission.

But perhaps you need more that a ON/OFF flow control?

Regards,
Arnaud

>> Â  Regards,
>> Arnaud
>>
>>> +Â Â Â  case RPMSG_DESTROY_EPT_IOCTL:
>>> +Â Â Â Â Â Â Â  ret = rpmsg_eptdev_destroy(&eptdev->dev, NULL);
>>> +Â Â Â Â Â Â Â  break;
>>> +Â Â Â  default:
>>> +Â Â Â Â Â Â Â  ret = -EINVAL;
>>> +Â Â Â  }
>>> Â  -Â Â Â  return rpmsg_eptdev_destroy(&eptdev->dev, NULL);
>>> +Â Â Â  return ret;
>>> Â  }
>>> Â  Â  static const struct file_operations rpmsg_eptdev_fops = {