Hi Mathieu, On 1/22/21 9:59 PM, Mathieu Poirier wrote: > On Fri, Jan 22, 2021 at 02:05:27PM +0100, Arnaud POULIQUEN wrote: >> Hi Mathieu, >> >> On 1/22/21 12:52 AM, Mathieu Poirier wrote: >>> On Tue, Dec 22, 2020 at 11:57:14AM +0100, Arnaud Pouliquen wrote: >>>> Implement the ioctl function that parses the list of >>>> rpmsg drivers registered to create an associated device. >>>> To be ISO user API, in a first step, the driver_override >>>> is only allowed for the RPMsg raw service, supported by the >>>> rpmsg_char driver. >>>> >>>> Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@xxxxxxxxxxx> >>>> --- >>>> drivers/rpmsg/rpmsg_ctrl.c | 43 ++++++++++++++++++++++++++++++++++++-- >>>> 1 file changed, 41 insertions(+), 2 deletions(-) >>>> >>>> diff --git a/drivers/rpmsg/rpmsg_ctrl.c b/drivers/rpmsg/rpmsg_ctrl.c >>>> index 065e2e304019..8381b5b2b794 100644 >>>> --- a/drivers/rpmsg/rpmsg_ctrl.c >>>> +++ b/drivers/rpmsg/rpmsg_ctrl.c >>>> @@ -56,12 +56,51 @@ static int rpmsg_ctrl_dev_open(struct inode *inode, struct file *filp) >>>> return 0; >>>> } >>>> >>>> +static const char *rpmsg_ctrl_get_drv_name(u32 service) >>>> +{ >>>> + struct rpmsg_ctl_info *drv_info; >>>> + >>>> + list_for_each_entry(drv_info, &rpmsg_drv_list, node) { >>>> + if (drv_info->ctrl->service == service) >>>> + return drv_info->ctrl->drv_name; >>>> + } >>>> + >>> >>> I'm unsure about the above... To me this looks like what the .match() function >>> of a bus would do. And when I read Bjorn's comment he brought up the >>> auxiliary_bus. I don't know about the auxiliary_bus but it is worth looking >>> into. Registering with a bus would streamline a lot of the code in this >>> patchset. >> >> As answered Bjorn, we already have the RPMsg bus to manage the rpmsg devices. >> Look like duplication from my POV, except if the IOCTL does not manage channel >> but only endpoint. >> >> In my design I considered that the rpmsg_ctrl creates a channel associated to a >> rpmsg_device such as the RPMsg ns_announcement. >> >> Based on this assumption, if we implement the auxiliary_bus (or other) for the >> rpmsg_ctrl a RPMsg driver will have to manage the probe by rpmsg_bus and by the >> auxillary bus. The probe from the auxiliary bus would lead to the creation of an >> RPMsg device on the rpmsg_bus, so a duplication with cross dependencies and >> would probably make tricky the remove part. >> >> That said, I think the design depends on the functionality that should be >> implemented in the rpmsg_ctrl. Here is an alternative approach based on the >> auxiliary bus, which I'm starting to think about: >> >> The current approach of the rpmsg_char driver is to use the IOCTRL interface to >> instantiate a cdev with an endpoint (the RPMsg device is associated with the >> ioctl dev). This would correspond to the use of an auxiliary bus to manage local >> endpoint creations. >> >> We could therefore consider an RPMsg name service based on an RPmsg device. This >> RPMsg device would register a kind of "RPMsg service endpoint" driver on the >> auxiliary rpmsg_ioctl bus. >> The rpmsg_ctrl will be used to instantiate the endpoints for this RPMsg device. >> on user application request the rpmsg_ctrl will call the appropriate auxiliary >> device to create an endpoint. >> >> If we consider that one objective of this series is to allow application to >> initiate the communication with the remote processor, so to be able to initiate >> the service (ns announcement sent to the remote processor). >> This implies that: >> -either the RPMsg device has been probed first by a remote ns announcement or by >> a Linux kernel driver using the "driver_override", to register an auxiliary >> device. In this case an endpoint will be created associated to the RPMsg service >> - or create a RPMsg device on first ioctl endpoint creation request, if it does >> not exist (that could trig a NS announcement to remote processor). >> >> But I'm not sure that this approach would work with QCOM RPMsg backends... >> > > I don't think there is a way forward with this set without a clear understanding > of the Glink and SMD drivers. I have already spent a fair amount of time in the > Glink driver and will continue on Monday with SMD. > >>> >>> I'm out of time for today - I will continue tomorrow. >> >> It seems to me that the main point to step forward is to clarify the global >> design and features of the rpmsg-ctrl. >> Depending on the decision taken, this series could be trashed and rewritten from >> a blank page...To not lost to much time on the series don't hesitate to limit >> the review to the minimum. >> > > I doubt you will ever get clear guidelines on the whole solution. I will get > back to you once I am done with the SMD driver, which should be in the > latter part of next week. Thanks for your time past on this topic! I don't expect a clear guidance but that we clarify the objective of this RPMsg IOCTL. A first step would be sure that we are in line with the objective of the RPMsg IOCTL. For instance should we continue in a way to have the rpmsg_char more "rpmsg service" generic, relying on a rpmsg_ioctl for the control part? Or should we implement something independent (with is own API) to limit dependency with QCOM backends constraints? At the end, if implementing a IOCTL interface directly in the RPMsg TTY seems to you and Bjorn simpler, I can also go on this way... On my side I expect to find time this week to prototype a RPMSg ioctl using the auxiliary bus to better understand involved mechanism. Thanks, Arnaud > >> Thanks, >> Arnaud >> >>> >>> Thanks, >>> Mathieu >>> >>>> + return NULL; >>>> +} >>>> + >>>> static long rpmsg_ctrl_dev_ioctl(struct file *fp, unsigned int cmd, >>>> unsigned long arg) >>>> { >>>> struct rpmsg_ctrl_dev *ctrldev = fp->private_data; >>>> - >>>> - dev_info(&ctrldev->dev, "Control not yet implemented\n"); >>>> + void __user *argp = (void __user *)arg; >>>> + struct rpmsg_channel_info chinfo; >>>> + struct rpmsg_endpoint_info eptinfo; >>>> + struct rpmsg_device *newch; >>>> + >>>> + if (cmd != RPMSG_CREATE_EPT_IOCTL) >>>> + return -EINVAL; >>>> + >>>> + if (copy_from_user(&eptinfo, argp, sizeof(eptinfo))) >>>> + return -EFAULT; >>>> + >>>> + /* >>>> + * In a frst step only the rpmsg_raw service is supported. >>>> + * The override is foorced to RPMSG_RAW_SERVICE >>>> + */ >>>> + chinfo.driver_override = rpmsg_ctrl_get_drv_name(RPMSG_RAW_SERVICE); >>>> + if (!chinfo.driver_override) >>>> + return -ENODEV; >>>> + >>>> + memcpy(chinfo.name, eptinfo.name, RPMSG_NAME_SIZE); >>>> + chinfo.name[RPMSG_NAME_SIZE - 1] = '\0'; >>>> + chinfo.src = eptinfo.src; >>>> + chinfo.dst = eptinfo.dst; >>>> + >>>> + newch = rpmsg_create_channel(ctrldev->rpdev, &chinfo); >>>> + if (!newch) { >>>> + dev_err(&ctrldev->dev, "rpmsg_create_channel failed\n"); >>>> + return -ENXIO; >>>> + } >>>> >>>> return 0; >>>> }; >>>> -- >>>> 2.17.1 >>>>