Hi Russell, å¨ 2015/3/31 17:13, Russell King - ARM Linux åé: > On Tue, Mar 31, 2015 at 04:30:39AM -0400, Yang Kuankuan wrote: >>> + cs[0] = IEC958_AES0_CON_NOT_COPYRIGHT | IEC958_AES0_CON_EMPHASIS_NONE; >>> + cs[1] = IEC958_AES1_CON_GENERAL; >>> + cs[2] = IEC958_AES2_CON_SOURCE_UNSPEC | IEC958_AES2_CON_CHANNEL_UNSPEC; >>> + cs[3] = IEC958_AES3_CON_CLOCK_1000PPM | fs; >>> + >> Pretty good, also suitable to rockchip platform, but why not add the >> "IEC958_AES2_CON_CHANNEL_MASK" & "IEC958_AES2_CON_WORDLEN" ? >> >> Seems sample frequency & channle number & word length are the basic >> message :) > I was debating about the word length, and that's something I'll add > later to it - but only if length shows that we have the 5th byte > available in the buffer. Most users seem to only use the first four > bytes. > > As for the channel number, this is intentionally left to the driver - > most cases I've found either the driver isn't interested, or where > they are interested (the only case I know of is my dw_hdmi ahb audio > driver), it's more appropriate to generate a baseline channel status, > and let the driver iterate over the channels adding the appropriate > channel number in. Okay, agree with you to keep baseline channel status, but seems dw_hdmi i2s audio are interested in channle number (to fill in schnl resigeters).
Best regards. Yakir Yang
