On Fri, Sep 4, 2009 at 11:11 AM, Antti Palosaari<crope@xxxxxx> wrote: > On 09/02/2009 07:32 AM, Michael Krufky wrote: >> >> Over the course of the past year, a number of developers have >> expressed a need for giving the bridge drivers better control over >> dvb_frontend operations. Specifically, there is a need for the bridge >> driver to receive a DVB frontend IOCTL and have the opportunity to >> allow or deny the IOCTL to proceed, as resources permit. >> >> For instance, in the case of a hybrid device, only the bridge driver >> knows whether the analog functionality is presently being used. If >> the driver is currently in analog mode, serving video frames, the >> driver will have a chance to deny the DVB frontend ioctl request >> before dvb-core passes the request on to the frontend driver, >> potentially damaging the analog video stream already in progress. >> >> In some cases, the bridge driver might have to perform a setup >> operation to use a feature specific to the device. For instance, the >> bridge device may be in a low powered state - this new capability >> allows the driver to wake up before passing the command on to the >> frontend driver. This new feature will allow LinuxTV developers to >> finally get working on actual power management support within the >> v4l/dvb subsystem, without the fear of breaking devices with hybrid >> analog / digital functionality. >> >> In other cases, there may be situations in which multiple RF >> connectors are available to the tuner, but only the bridge driver will >> be aware of this, as this type of thing is specific to the device's >> hardware implementation. As there are many tuners capable of multiple >> RF spigots, not all devices actually employ this feature - only the >> bridge driver knows what implementations support such features, and >> how to enable / disable them. >> >> The possibilities are endless. I actually did all the heavy lifting >> involved in this a few months ago, but haven't had a moment to write >> up this RFC until now. >> >> The change to dvb-core that allows this new functionality is posted to >> my development repository on kernellabs.com. I have also included an >> example of how this can be used on a digital tuner board with multiple >> RF inputs. The multiple RF input switching is already supported in >> today's code, but I promised Mauro that I would present a better >> method of doing this before the upcoming merge window. For your >> review and comments, please take a look at the topmost changesets, >> starting with "create a standard method for dvb adapter drivers to >> override frontend ioctls": >> >> http://kernellabs.com/hg/~mkrufky/fe_ioctl_override > > Idea looks very good! I tested one DVB USB device need blocking IOCTLs when > demod and tuner are power save and didn't saw functionality problems. > > However, it was a little bit hard to add callback to DVB USB driver. Could > that callback be added to the struct dvb_usb_adapter_properties for simplify > things? I have feeling that this callback will be useful most DVb USB > devices - setting GPIOs and clock settings for power save. > > Name fe_ioctl_override sounds like whole IOCTL will be replaced with new one > which is not true. Still, I don't know which could be better name. Thank you for the feedback, Antti. Yes, I can add a generic mechanism into the DVB USB device framework to enable this functionality there. I'll take care of that after we merge the initial core changes, or maybe I will have time to do that over the next week. You're right, this feature will certainly be useful in the dvb-usb framework, as well as many others. As far as the name fe_ioctl_override, this does apply to the entire IOCTL, but currently only being used for SET_FRONTEND. As more needs arise, the possibilities of this feature will become more prominent. This is not a "replacement" for the IOCTL per se. As the comments in dvbdev.h explain, depending on the return value of the bridge callback, this will determine whether or not the previous default IOCTL handling should be allowed to continue, or whether the bridge's override returns back to userspace itself. So, if configured a certain way, this feature CAN be used to replace dvb-core's IOCTL handling, but in most cases it will simply allow the bridge to do pre-processing and post-processing of a given IOCTL. As fe_ioctl_override is actually most descriptive of the process going on, I'd like to keep that name as-is. Cheers, Mike -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
