On Mon, Jul 26, 2021 at 8:42 AM chihhao chen <chihhao.chen@xxxxxxxxxxxx> wrote: > > Hello, > > Attach USB descriptor of clock source and selectior for this earphone. > > AC Clock Source Descriptor: > ------------------------------ > > Value Valuename > 0x08 bLength > 0x24 bDescriptorType > 0x0A bDescriptorSubtype > 0x09 bClockID > 0x03 bmAttributes > 0x07 bmControls > 0x00 bAssocTerminal > 0x00 iClockSource > Hex dump: > 0x08 0x24 0x0A 0x09 0x03 0x07 0x00 0x00 > > AC Clock Selector Descriptor: > ------------------------------ > > Value Valuename > 0x09 bLength > 0x24 bDescriptorType > 0x0B bDescriptorSubtype > 0x0B bClockID > 0x02 bNrInPins > 0x09 baCSourceID(1) > 0x0A baCSourceID(2) > 0x03 bmControls > 0x00 iClockSelector > Hex dump: > 0x09 0x24 0x0B 0x0B 0x02 0x09 0x0A 0x03 0x00 > > AC Clock Source Descriptor: > ------------------------------ > > Value Valuename > 0x08 bLength > 0x24 bDescriptorType > 0x0A bDescriptorSubtype > 0x0A bClockID > 0x03 bmAttributes > 0x07 bmControls > 0x00 bAssocTerminal > 0x00 iClockSource > Hex dump: > 0x08 0x24 0x0A 0x0A 0x03 0x07 0x00 0x00 > > AC Clock Selector Descriptor: > ------------------------------ > > Value Valuename > 0x09 bLength > 0x24 bDescriptorType > 0x0B bDescriptorSubtype > 0x0C bClockID > 0x02 bNrInPins > 0x09 baCSourceID(1) > 0x0A baCSourceID(2) > 0x03 bmControls > 0x00 iClockSelector > Hex dump: > 0x09 0x24 0x0B 0x0C 0x02 0x09 0x0A 0x03 0x00 > > Thanks > Chihhao Thank you, Chihhao. So I was wrong about Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051) having a Clock Multiplier. There are two Clock Sources, both linked to the USB SOF with fixed sample rate. Plus two Clock Selectors which are host-programmable and can be set to either of the two Clock Sources. I'm still at a loss to explain what is going wrong here. Would a printk() reveal the first explicit uac_clock_selector_set_val() on the Clock Selector associated with USB_IN sets Clock Source ID to pin 1 with Clock Source ID 0x9? Or is it the other way around, i.e. it sets the Clock Source ID to pin 2 with Clock ID 0xA for the capture stream Clock Selector? Chihhao Chen, could you please try the following patch for debugging purposes and share what is printed in dmesg? Please try one time with your fix applied and one time without, i.e. with an otherwise unmodified vanilla kernel. Thank you, Geraldo Nascimento --- clock.c.orig 2021-07-17 12:15:06.416028360 -0000 +++ clock.c 2021-07-26 20:45:58.713881962 -0000 @@ -300,6 +300,7 @@ static int __uac_clock_find_source(struc /* the entity ID we are looking for is a selector. * find out what it currently selects */ ret = uac_clock_selector_get_val(chip, clock_id); + printk(KERN_ERR "FOR EP %x: Clock Selector %x has pin %d for Clock Source ID %x selected\n", (unsigned int)fmt->endpoint, clock_id, ret, sources[ret - 1]); if (ret < 0) { if (!chip->autoclock) return ret; @@ -324,6 +325,7 @@ static int __uac_clock_find_source(struc sources[ret - 1], visited, validate); if (ret > 0) { + printk(KERN_ERR "FOR EP %x: Found Source! Clock Selector %x has pin %d for Clock Source ID %x about to be reselected\n", (unsigned int)fmt->endpoint, entity_id, cur, sources[cur - 1]); err = uac_clock_selector_set_val(chip, entity_id, cur); if (err < 0) return err; @@ -344,6 +346,7 @@ static int __uac_clock_find_source(struc if (ret < 0) continue; + printk(KERN_ERR "FOR EP %x: Found source by trial and error! Clock Selector %x has pin %d for Clock Source ID %x about to be selected\n", (unsigned int)fmt->endpoint, entity_id, i, sources[i - 1]); err = uac_clock_selector_set_val(chip, entity_id, i); if (err < 0) continue;