On 7/26/07, Oliver Endriss <o.endriss@xxxxxx> wrote: > Does it take a long time to download the firmware? > If not, I would go for the easiest way. (I guess the firmware has to be > loaded after a coldstart only.) > I also would prefer the easy method, but there are some pitfalls in that one. True that a download is necessary on a cold-start only. But being a USB device, the MASTER-SLAVE config one, there are more of cold-starts that we might think. That said, the firmware download is not too lengthy, but the Host when handling 2 Transfers simultaeneously in both directions, the performance might be degraded according to the vendor, though they do not have actual figures. > > In this case the firmware has to be copied from demod #0 to demod #1 > > for effeciency reasons, though a dual download can be used. But AFAICS > > the firmware getting downloaded to the demod would be necessary in the > > SLAVE - SLAVE configuration as found on PCI/PCIe devices, since there > > is no BIU in that configuration. > > > > I think it might be better then, that the firmware download in the > > demodulators is conditionally done through a BIU config data structure > > in such a case > > > > (Thoughts ?) > > You might add a firmware_request callback to the configuration struct > passed to the demod_attach routine (see tda1004x driver, for example). > > The card driver will set this entry before calling demod_attach. Yep, Overwriting the function pointer seems to be the only sane approach here, AFAICT. > For a dual download, it might point to the same routine for each demod. > For the download+copy approach, you might set the callback of the second > demod to a special firmware copy routine. Yep, the function pointer could be just overwritten here. Sounds good Thanks for the inputs, Manu _______________________________________________ linux-dvb mailing list linux-dvb@xxxxxxxxxxx http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
