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 For those that would prefer not to open up a web browser, here are the core changes (apologies if whitespace becomes broken): --- a/linux/drivers/media/dvb/dvb-core/dvb_frontend.c Tue Jun 16 16:08:17 2009 -0400 +++ b/linux/drivers/media/dvb/dvb-core/dvb_frontend.c Sat May 23 17:00:59 2009 -0400 @@ -1594,7 +1594,18 @@ struct dvb_device *dvbdev = file->private_data; struct dvb_frontend *fe = dvbdev->priv; struct dvb_frontend_private *fepriv = fe->frontend_priv; - int err = -EOPNOTSUPP; + int cb_err, err = -EOPNOTSUPP; + + if (fe->dvb->fe_ioctl_override) { + cb_err = fe->dvb->fe_ioctl_override(fe, cmd, parg, + DVB_FE_IOCTL_PRE); + if (cb_err < 0) + return cb_err; + if (cb_err > 0) + return 0; + /* fe_ioctl_override returning 0 allows + * dvb-core to continue handling the ioctl */ + } switch (cmd) { case FE_GET_INFO: { @@ -1858,6 +1869,13 @@ err = 0; break; }; + + if (fe->dvb->fe_ioctl_override) { + cb_err = fe->dvb->fe_ioctl_override(fe, cmd, parg, + DVB_FE_IOCTL_POST); + if (cb_err < 0) + return cb_err; + } return err; } --- a/linux/drivers/media/dvb/dvb-core/dvbdev.h Tue Jun 16 16:08:17 2009 -0400 +++ b/linux/drivers/media/dvb/dvb-core/dvbdev.h Sat May 23 17:00:59 2009 -0400 @@ -51,6 +51,8 @@ module_param_array(adapter_nr, short, NULL, 0444); \ MODULE_PARM_DESC(adapter_nr, "DVB adapter numbers") +struct dvb_frontend; + struct dvb_adapter { int num; struct list_head list_head; @@ -66,6 +68,32 @@ int mfe_shared; /* indicates mutually exclusive frontends */ struct dvb_device *mfe_dvbdev; /* frontend device in use */ struct mutex mfe_lock; /* access lock for thread creation */ + + /* Allow the adapter/bridge driver to perform an action before and/or + * after the core handles an ioctl: + * + * DVB_FE_IOCTL_PRE indicates that the ioctl has not yet been handled. + * DVB_FE_IOCTL_POST indicates that the ioctl has been handled. + * + * When DVB_FE_IOCTL_PRE is passed to the callback as the stage arg: + * + * return 0 to allow dvb-core to handle the ioctl. + * return a positive int to prevent dvb-core from handling the ioctl, + * and exit without error. + * return a negative int to prevent dvb-core from handling the ioctl, + * and return that value as an error. + * + * When DVB_FE_IOCTL_POST is passed to the callback as the stage arg: + * + * return 0 to allow the dvb_frontend ioctl handler to exit normally. + * return a negative int to cause the dvb_frontend ioctl handler to + * return that value as an error. + */ +#define DVB_FE_IOCTL_PRE 0 +#define DVB_FE_IOCTL_POST 1 + int (*fe_ioctl_override)(struct dvb_frontend *fe, + unsigned int cmd, void *parg, + unsigned int stage); }; As you can see from the changeset, this allows the bridge to assert control both before and after the frontend driver itself handles the operation. Depending on the value returned by the bridge driver, dvb-core can exit with error, proceed unchanged, or exit without error. In order to allow some drivers to use the new feature, I had to make a small change to videobuf-dvb: --- a/linux/drivers/media/video/videobuf-dvb.c Tue Jun 16 16:08:17 2009 -0400 +++ b/linux/drivers/media/video/videobuf-dvb.c Sat May 23 17:00:59 2009 -0400 @@ -144,7 +144,9 @@ struct device *device, char *adapter_name, short *adapter_nr, - int mfe_shared) + int mfe_shared, + int (*fe_ioctl_override)(struct dvb_frontend *, + unsigned int, void *, unsigned int)) { int result; @@ -159,6 +161,7 @@ } fe->adapter.priv = adapter_priv; fe->adapter.mfe_shared = mfe_shared; + fe->adapter.fe_ioctl_override = fe_ioctl_override; return result; } @@ -258,7 +261,9 @@ void *adapter_priv, struct device *device, short *adapter_nr, - int mfe_shared) + int mfe_shared, + int (*fe_ioctl_override)(struct dvb_frontend *, + unsigned int, void *, unsigned int)) { struct list_head *list, *q; struct videobuf_dvb_frontend *fe; @@ -272,7 +277,7 @@ /* Bring up the adapter */ res = videobuf_dvb_register_adapter(f, module, adapter_priv, device, - fe->dvb.name, adapter_nr, mfe_shared); + fe->dvb.name, adapter_nr, mfe_shared, fe_ioctl_override); if (res < 0) { printk(KERN_WARNING "videobuf_dvb_register_adapter failed (errno = %d)\n", res); return res; --- a/linux/include/media/videobuf-dvb.h Tue Jun 16 16:08:17 2009 -0400 +++ b/linux/include/media/videobuf-dvb.h Sat May 23 17:00:59 2009 -0400 @@ -42,7 +42,9 @@ void *adapter_priv, struct device *device, short *adapter_nr, - int mfe_shared); + int mfe_shared, + int (*fe_ioctl_override)(struct dvb_frontend *, + unsigned int, void *, unsigned int)); void videobuf_dvb_unregister_bus(struct videobuf_dvb_frontends *f); --- a/linux/drivers/media/video/cx23885/cx23885-dvb.c Tue Jun 16 16:08:17 2009 -0400 +++ b/linux/drivers/media/video/cx23885/cx23885-dvb.c Sat May 23 17:00:59 2009 -0400 @@ -856,7 +856,7 @@ /* register everything */ ret = videobuf_dvb_register_bus(&port->frontends, THIS_MODULE, port, - &dev->pci->dev, adapter_nr, 0); + &dev->pci->dev, adapter_nr, 0, NULL); /* init CI & MAC */ switch (dev->board) { --- a/linux/drivers/media/video/cx88/cx88-dvb.c Tue Jun 16 16:08:17 2009 -0400 +++ b/linux/drivers/media/video/cx88/cx88-dvb.c Sat May 23 17:00:59 2009 -0400 @@ -1180,7 +1180,7 @@ /* register everything */ return videobuf_dvb_register_bus(&dev->frontends, THIS_MODULE, dev, - &dev->pci->dev, adapter_nr, mfe_shared); + &dev->pci->dev, adapter_nr, mfe_shared, NULL); frontend_detach: core->gate_ctrl = NULL; --- a/linux/drivers/media/video/saa7134/saa7134-dvb.c Tue Jun 16 16:08:17 2009 -0400 +++ b/linux/drivers/media/video/saa7134/saa7134-dvb.c Sat May 23 17:00:59 2009 -0400 @@ -1509,7 +1509,7 @@ /* register everything else */ ret = videobuf_dvb_register_bus(&dev->frontends, THIS_MODULE, dev, - &dev->pci->dev, adapter_nr, 0); + &dev->pci->dev, adapter_nr, 0, NULL); /* this sequence is necessary to make the tda1004x load its firmware * and to enter analog mode of hybrid boards Finally, the actual implementation of this new feature in the cx23885 driver for the sake of RF input switching: --- a/linux/drivers/media/video/cx23885/cx23885-dvb.c Sat May 23 17:00:59 2009 -0400 +++ b/linux/drivers/media/video/cx23885/cx23885-dvb.c Fri May 08 21:39:24 2009 -0400 @@ -484,9 +484,33 @@ } break; } - return (port->set_frontend_save) ? - port->set_frontend_save(fe, param) : -ENODEV; + return 0; } + +static int cx23885_dvb_fe_ioctl_override(struct dvb_frontend *fe, + unsigned int cmd, void *parg, + unsigned int stage) +{ + int err = 0; + + switch (stage) { + case DVB_FE_IOCTL_PRE: + + switch (cmd) { + case FE_SET_FRONTEND: + err = cx23885_dvb_set_frontend(fe, + (struct dvb_frontend_parameters *) parg); + break; + } + break; + + case DVB_FE_IOCTL_POST: + /* no post-ioctl handling required */ + break; + } + return err; +}; + static int dvb_register(struct cx23885_tsport *port) { @@ -527,12 +551,6 @@ 0x60, &dev->i2c_bus[1].i2c_adap, &hauppauge_hvr127x_config); } - - /* FIXME: temporary hack */ - /* define bridge override to set_frontend */ - port->set_frontend_save = fe0->dvb.frontend->ops.set_frontend; - fe0->dvb.frontend->ops.set_frontend = cx23885_dvb_set_frontend; - break; case CX23885_BOARD_HAUPPAUGE_HVR1255: i2c_bus = &dev->i2c_bus[0]; @@ -856,7 +874,8 @@ /* register everything */ ret = videobuf_dvb_register_bus(&port->frontends, THIS_MODULE, port, - &dev->pci->dev, adapter_nr, 0, NULL); + &dev->pci->dev, adapter_nr, 0, + cx23885_dvb_fe_ioctl_override); /* init CI & MAC */ switch (dev->board) { --- a/linux/drivers/media/video/cx23885/cx23885.h Sat May 23 17:00:59 2009 -0400 +++ b/linux/drivers/media/video/cx23885/cx23885.h Fri May 08 21:39:24 2009 -0400 @@ -289,10 +289,6 @@ /* Allow a single tsport to have multiple frontends */ u32 num_frontends; void *port_priv; - - /* FIXME: temporary hack */ - int (*set_frontend_save) (struct dvb_frontend *, - struct dvb_frontend_parameters *); }; struct cx23885_dev { Again, the above changeset simply removes the old RF input path switching hack, and implements the feature using the new frontend ioctl override method instead. This is a much cleaner and more more robust way to handle this, which actually opening doors for other drivers to handle similar problems in a much more efficient manner. Please voice your comments about this particular change and any of its implications in this thread. If all goes well, I'd like to merge this into the master repository in time for the next merge window. Once this is in the master branch, I can start working on removing other similar hacks from other drivers in our tree, in favor of this better method. Cheers, Michael Krufky -- 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
