[PATCH] ALSA: usb-audio: fix incorrect clock source setting
From: "chihhao.chen" chihhao.chen@mediatek.com
The following scenario describes an echo test for Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051).
We first start a capture stream(USB IN transfer) in 96Khz/24bit/1ch mode. In clock find source function, we get value 0x2 for clock selector and 0x1 for clock source.
Kernel-4.14 behavior Since clock source is valid so clock selector was not set again. We pass through this function and start a playback stream(USB OUT transfer) in 48Khz/32bit/2ch mode. This time we get value 0x1 for clock selector and 0x1 for clock source. Finally clock id with this setting is 0x9.
Kernel-5.10 behavior Clock selector was always set one more time even it is valid. When we start a playback stream, we will get 0x2 for clock selector and 0x1 for clock source. In this case clock id becomes 0xA. This is an incorrect clock source setting and results in severe noises. We see wrong data rate in USB IN transfer. (From 288 bytes/ms becomes 144 bytes/ms) It should keep in 288 bytes/ms.
This earphone works fine on older kernel version load because this is a newly-added behavior.
Signed-off-by: chihhao.chen chihhao.chen@mediatek.com --- sound/usb/clock.c | 6 ++++++ 1 file changed, 6 insertions(+)
diff --git a/sound/usb/clock.c b/sound/usb/clock.c index 52de522..14456f6 100644 --- a/sound/usb/clock.c +++ b/sound/usb/clock.c @@ -324,6 +324,12 @@ static int __uac_clock_find_source(struct snd_usb_audio *chip, sources[ret - 1], visited, validate); if (ret > 0) { + /* + * For Samsung USBC Headset (AKG), setting clock selector again + * will result in incorrect default clock setting problems + */ + if (chip->usb_id == USB_ID(0x04e8, 0xa051)) + return ret; err = uac_clock_selector_set_val(chip, entity_id, cur); if (err < 0) return err;
On Sat, 24 Jul 2021 06:23:41 +0200, chihhao.chen@mediatek.com wrote:
From: "chihhao.chen" chihhao.chen@mediatek.com
The following scenario describes an echo test for Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051).
We first start a capture stream(USB IN transfer) in 96Khz/24bit/1ch mode. In clock find source function, we get value 0x2 for clock selector and 0x1 for clock source.
Kernel-4.14 behavior Since clock source is valid so clock selector was not set again. We pass through this function and start a playback stream(USB OUT transfer) in 48Khz/32bit/2ch mode. This time we get value 0x1 for clock selector and 0x1 for clock source. Finally clock id with this setting is 0x9.
Kernel-5.10 behavior Clock selector was always set one more time even it is valid. When we start a playback stream, we will get 0x2 for clock selector and 0x1 for clock source. In this case clock id becomes 0xA. This is an incorrect clock source setting and results in severe noises. We see wrong data rate in USB IN transfer. (From 288 bytes/ms becomes 144 bytes/ms) It should keep in 288 bytes/ms.
This earphone works fine on older kernel version load because this is a newly-added behavior.
Signed-off-by: chihhao.chen chihhao.chen@mediatek.com
Thanks for the patch.
This looks like a regression introduced by the recent commit d2e8f641257d ("ALSA: usb-audio: Explicitly set up the clock selector"), which is a fix for certain devices. Too bad that the behavior really depends on the device...
Maybe we need to introduce some flag to handle this commonly, but for now, let's take the fix as is.
Takashi
On Sat, Jul 24, 2021 at 8:05 AM Takashi Iwai tiwai@suse.de wrote:
This looks like a regression introduced by the recent commit d2e8f641257d ("ALSA: usb-audio: Explicitly set up the clock selector"), which is a fix for certain devices. Too bad that the behavior really depends on the device...
Dr. Iwai, perhaps we could restrict the generalized fix for the Behringer UFX1604 / UFX1204 with some simple logic to devices that only have *one* clock source.
In that case the clock selector must be set to the only clock source.
This way we keep the generalization without breaking devices with more than one clock source.
Just an idea.
Thank you, Geraldo Nascimento
Dr. Iwai, perhaps we could restrict the generalized fix for the Behringer UFX1604 / UFX1204 with some simple logic to devices that only have *one* clock source.
Okay, rereading the original commit log from Cihhao Chen I gather Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051) actually has two clock selectors and only one clock source.
Correct me if I'm wrong.
This is complicated by the fact I haven't been able to find a lsusb -v of Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051)
Even so, my proposition still stands: devices with only one clock source and only one clock selector should be able to handle us selecting the clock selector to the only clock source.
I tried to convey in code what I had in mind.
It's a rough sketch and very much untested.
--- clock.5.14-rc2.c 2021-07-24 18:30:09.773718208 -0000 +++ clock-one-to-one.c 2021-07-24 18:35:52.276412366 -0000 @@ -54,6 +54,61 @@ static void *find_uac_clock_desc(struct return NULL; }
+/* Behringer UFX1604 / UFX1204 have a simple one-to-one + * topology where there is only one Clock Selector, only + * one Clock Source linked to USB SOF and no Clock Multipliers. + * + * This function checks for the presence of such a + * one-to-one clock selector / clock source topology + * so that it's possible to safely set the one and only + * Clock Selector to the one and only Clock Source + * upon sample rate change without breaking devices + * with more complicated topologies. + */ + +static bool one_to_one_clock_topology(struct usb_host_interface *iface, int proto) +{ + int clock_sources, clock_selectors, clock_multipliers = 0; + int source_version, selector_version, multiplier_version; + int found_count; + + void *cs = NULL; + + if (proto == UAC_VERSION_3) { + source_version = UAC3_CLOCK_SOURCE; + selector_version = UAC3_CLOCK_SELECTOR; + multiplier_version = UAC3_CLOCK_MULTIPLIER; + } + + else { + source_version = UAC2_CLOCK_SOURCE; + selector_version = UAC2_CLOCK_SELECTOR; + multiplier_version = UAC2_CLOCK_MULTIPLIER; + } + + if ((found_count = snd_usb_count_csint_desc(iface->extra, iface->extralen, + cs, source_version)) > 0) { + clock_sources = found_count; + } + + if ((found_count = snd_usb_count_csint_desc(iface->extra, iface->extralen, + cs, selector_version)) > 0) { + clock_selectors = found_count; + } + + if ((found_count = snd_usb_count_csint_desc(iface->extra, iface->extralen, + cs, multiplier_version)) > 0) { + clock_multipliers = found_count; + } + + if ((clock_sources == 1) && (clock_selectors == 1) && (clock_multipliers == 0)) { + return true; + } + + return false; +} + + static bool validate_clock_source(void *p, int id, int proto) { union uac23_clock_source_desc *cs = p; @@ -323,7 +378,7 @@ static int __uac_clock_find_source(struc ret = __uac_clock_find_source(chip, fmt, sources[ret - 1], visited, validate); - if (ret > 0) { + if (ret > 0 && one_to_one_clock_topology(chip->ctrl_intf, proto)) { err = uac_clock_selector_set_val(chip, entity_id, cur); if (err < 0) return err;
--- helper.5.14-rc2.c 2021-07-24 18:30:25.042526253 -0000 +++ helper-one-to-one.c 2021-07-24 18:35:45.019503597 -0000 @@ -64,6 +64,29 @@ void *snd_usb_find_csint_desc(void *buff }
/* + * find every class-specified interface descriptor with the given subtype + * and return how many did it find + */ +int snd_usb_count_csint_desc(void *buffer, int buflen, void *after, u8 dsubtype) +{ + int count = 0; + unsigned char *p = after; + + while ((p = snd_usb_find_desc(buffer, buflen, p, + USB_DT_CS_INTERFACE)) != NULL) { + if (p[0] >= 3 && p[2] == dsubtype) + count++; + } + + if (count > 0) { + return count; + } + + return 0; +} + + +/* * Wrapper for usb_control_msg(). * Allocates a temp buffer to prevent dmaing from/to the stack. */
--- helper.5.14-rc2.h 2021-07-24 18:30:35.219398312 -0000 +++ helper-one-to-one.h 2021-07-24 18:29:34.139166195 -0000 @@ -7,6 +7,8 @@ unsigned int snd_usb_combine_bytes(unsig void *snd_usb_find_desc(void *descstart, int desclen, void *after, u8 dtype); void *snd_usb_find_csint_desc(void *descstart, int desclen, void *after, u8 dsubtype);
+int snd_usb_count_csint_desc(void *descstart, int desclen, void *after, u8 dsubtype); + int snd_usb_ctl_msg(struct usb_device *dev, unsigned int pipe, __u8 request, __u8 requesttype, __u16 value, __u16 index, void *data, __u16 size);
On Sat, Jul 24, 2021 at 3:20 PM Geraldo Nascimento geraldogabriel@gmail.com wrote:
Dr. Iwai, perhaps we could restrict the generalized fix for the Behringer UFX1604 / UFX1204 with some simple logic to devices that only have *one* clock source.
Okay, rereading the original commit log from Cihhao Chen I gather Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051) actually has two clock selectors and only one clock source.
Correct me if I'm wrong.
This is complicated by the fact I haven't been able to find a lsusb -v of Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051)
Even so, my proposition still stands: devices with only one clock source and only one clock selector should be able to handle us selecting the clock selector to the only clock source.
On Sat, 24 Jul 2021 17:04:13 +0200, Geraldo Nascimento wrote:
On Sat, Jul 24, 2021 at 8:05 AM Takashi Iwai tiwai@suse.de wrote:
This looks like a regression introduced by the recent commit d2e8f641257d ("ALSA: usb-audio: Explicitly set up the clock selector"), which is a fix for certain devices. Too bad that the behavior really depends on the device...
Dr. Iwai, perhaps we could restrict the generalized fix for the Behringer UFX1604 / UFX1204 with some simple logic to devices that only have *one* clock source.
In that case the clock selector must be set to the only clock source.
This way we keep the generalization without breaking devices with more than one clock source.
Just an idea.
I don't think it's easy to generalize. All those bugs are more or less BIOS bugs, and a logic doesn't apply always, just because it's a bug :) For example, setting the clock selector itself should be a valid operation from the specification POV, while this leads to breakage on some devices. So, even if we add a more generic workaround, we need to see which side effect is more commonly seen at first.
Takashi
On Sun, Jul 25, 2021 at 7:44 AM Takashi Iwai tiwai@suse.de wrote:
On Sat, 24 Jul 2021 17:04:13 +0200, Geraldo Nascimento wrote:
On Sat, Jul 24, 2021 at 8:05 AM Takashi Iwai tiwai@suse.de wrote:
This looks like a regression introduced by the recent commit d2e8f641257d ("ALSA: usb-audio: Explicitly set up the clock selector"), which is a fix for certain devices. Too bad that the behavior really depends on the device...
Dr. Iwai, perhaps we could restrict the generalized fix for the Behringer UFX1604 / UFX1204 with some simple logic to devices that only have *one* clock source.
In that case the clock selector must be set to the only clock source.
This way we keep the generalization without breaking devices with more than one clock source.
Just an idea.
I don't think it's easy to generalize. All those bugs are more or less BIOS bugs, and a logic doesn't apply always, just because it's a bug :) For example, setting the clock selector itself should be a valid operation from the specification POV, while this leads to breakage on some devices. So, even if we add a more generic workaround, we need to see which side effect is more commonly seen at first.
Takashi
Hello,
Like I said in one of the other emails in this thread, it's hard to pinpoint a cause for the breakage of Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051) without the lsusb -v of the device in question.
But from the description Chihhao Chen gave in the original message, I'm *guessing* the Clock Source for the Samsung USB Headset (AKG) runs at 48000hz and that we'd see a 2x Clock Multiplier in the lsusb -v
This is all a wild guess, without the lsusb -v it's impossible to be sure, but if I'm right then the valid setting for the Microphone's Clock Selector is the Clock Multiplier, not the Clock Source, which, remember, runs at half the clock, hence why Chihhao Chen sees half the data rate for USB IN.
Unfortunately our kernel code presently *does* always set the Clock Selector to the Clock Source, which is a bad assumption to make in my humble opinion.
The only valid case for setting the Clock Selector to the Clock Source is when there's precisely one Clock Selector, precisely one Clock Source and no Clock Multipliers.
In that special case we may be able to touch the setting of the only Clock Selector to match the only Clock Source.
And, frankly, the only reason we're forced to do that explicitly is because some Behringer gear (Archwave AG DACs) gets confused and seems to somehow keep the old rate on the Clock Selector upon sample rate change.
Thank you, Geraldo Nascimento
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
On Mon, 2021-07-26 at 02:16 +0000, Geraldo Nascimento wrote:
On Sun, Jul 25, 2021 at 7:44 AM Takashi Iwai tiwai@suse.de wrote:
On Sat, 24 Jul 2021 17:04:13 +0200, Geraldo Nascimento wrote:
On Sat, Jul 24, 2021 at 8:05 AM Takashi Iwai tiwai@suse.de wrote:
This looks like a regression introduced by the recent commit d2e8f641257d ("ALSA: usb-audio: Explicitly set up the clock selector"), which is a fix for certain devices. Too bad that the behavior really depends on the device...
Dr. Iwai, perhaps we could restrict the generalized fix for the Behringer UFX1604 / UFX1204 with some simple logic to devices that only have *one* clock source.
In that case the clock selector must be set to the only clock source.
This way we keep the generalization without breaking devices with more than one clock source.
Just an idea.
I don't think it's easy to generalize. All those bugs are more or less BIOS bugs, and a logic doesn't apply always, just because it's a bug :) For example, setting the clock selector itself should be a valid operation from the specification POV, while this leads to breakage on some devices. So, even if we add a more generic workaround, we need to see which side effect is more commonly seen at first.
Takashi
Hello,
Like I said in one of the other emails in this thread, it's hard to pinpoint a cause for the breakage of Samsung USBC Headset (AKG) with VID/PID (0x04e8/0xa051) without the lsusb -v of the device in question.
But from the description Chihhao Chen gave in the original message, I'm *guessing* the Clock Source for the Samsung USB Headset (AKG) runs at 48000hz and that we'd see a 2x Clock Multiplier in the lsusb -v
This is all a wild guess, without the lsusb -v it's impossible to be sure, but if I'm right then the valid setting for the Microphone's Clock Selector is the Clock Multiplier, not the Clock Source, which, remember, runs at half the clock, hence why Chihhao Chen sees half the data rate for USB IN.
Unfortunately our kernel code presently *does* always set the Clock Selector to the Clock Source, which is a bad assumption to make in my humble opinion.
The only valid case for setting the Clock Selector to the Clock Source is when there's precisely one Clock Selector, precisely one Clock Source and no Clock Multipliers.
In that special case we may be able to touch the setting of the only Clock Selector to match the only Clock Source.
And, frankly, the only reason we're forced to do that explicitly is because some Behringer gear (Archwave AG DACs) gets confused and seems to somehow keep the old rate on the Clock Selector upon sample rate change.
Thank you, Geraldo Nascimento
On Mon, Jul 26, 2021 at 8:42 AM chihhao chen chihhao.chen@mediatek.com 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;
From: chihhao chen chihhao.chen@mediatek.com
Hello
<6>[ 150.347456][ T2768] __uac_clock_find_source: FOR EP 81: Clock Selector c has pin 2 for Clock Source ID a selected <6>[ 150.347517][ T2768] __uac_clock_find_source: FOR EP 81: Found Source! Clock Selector c has pin 2 for Clock Source ID a about to be reselected <6>[ 150.384289][ T2768] __uac_clock_find_source: FOR EP 81: Clock Selector c has pin 2 for Clock Source ID a selected <6>[ 150.390920][ T2768] __uac_clock_find_source: FOR EP 81: Found Source! Clock Selector c has pin 2 for Clock Source ID a about to be reselected <6>[ 150.438156][ T2768] __uac_clock_find_source: FOR EP 81: Clock Selector c has pin 2 for Clock Source ID a selected <6>[ 150.438226][ T2768] __uac_clock_find_source: FOR EP 81: Found Source! Clock Selector c has pin 2 for Clock Source ID a about to be reselected <6>[ 150.473547][ T2768] __uac_clock_find_source: FOR EP 81: Clock Selector c has pin 2 for Clock Source ID a selected <6>[ 150.480165][ T2768] __uac_clock_find_source: FOR EP 81: Found Source! Clock Selector c has pin 2 for Clock Source ID a about to be reselected <6>[ 150.513375][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.513439][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.546161][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.552678][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.584347][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.584418][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.617760][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.624253][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.657906][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.657982][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.689571][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.696109][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected
Chihhao
On Tue, Jul 27, 2021 at 10:28 AM chihhao.chen@mediatek.com wrote:
From: chihhao chen chihhao.chen@mediatek.com
Hello
<6>[ 150.347456][ T2768] __uac_clock_find_source: FOR EP 81: Clock Selector c has pin 2 for Clock Source ID a selected <6>[ 150.347517][ T2768] __uac_clock_find_source: FOR EP 81: Found Source! Clock Selector c has pin 2 for Clock Source ID a about to be reselected <6>[ 150.384289][ T2768] __uac_clock_find_source: FOR EP 81: Clock Selector c has pin 2 for Clock Source ID a selected <6>[ 150.390920][ T2768] __uac_clock_find_source: FOR EP 81: Found Source! Clock Selector c has pin 2 for Clock Source ID a about to be reselected <6>[ 150.438156][ T2768] __uac_clock_find_source: FOR EP 81: Clock Selector c has pin 2 for Clock Source ID a selected <6>[ 150.438226][ T2768] __uac_clock_find_source: FOR EP 81: Found Source! Clock Selector c has pin 2 for Clock Source ID a about to be reselected <6>[ 150.473547][ T2768] __uac_clock_find_source: FOR EP 81: Clock Selector c has pin 2 for Clock Source ID a selected <6>[ 150.480165][ T2768] __uac_clock_find_source: FOR EP 81: Found Source! Clock Selector c has pin 2 for Clock Source ID a about to be reselected <6>[ 150.513375][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.513439][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.546161][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.552678][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.584347][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.584418][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.617760][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.624253][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.657906][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.657982][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected <6>[ 150.689571][ T2768] __uac_clock_find_source: FOR EP 1: Clock Selector b has pin 2 for Clock Source ID a selected <6>[ 150.696109][ T2768] __uac_clock_find_source: FOR EP 1: Found Source! Clock Selector b has pin 2 for Clock Source ID a about to be reselected
Chihhao
Thank you. Chihhao Chen!
I see both EPs have their Clock Selectors (ID 0xC for EP 81 and ID 0xB for EP 1) selected to pin 2, i.e. Clock Source ID 0xA.
I'm assuming this log is for the vanilla kernel without Chihhao's fix. Please correct me if I'm wrong, Chihhao.
From the original commit message for the fix, we know the correct
setting for Clock Selector 0xB should be pin 1, with Clock Source ID 0x9.
Takashi Iwai already shared his perspective that this is a firmware bug on the device.
I have a hunch that the firmware bug is setting both Clock Selectors at the same time regardless of which one we want to select specifically.
Chihhao, please try the below patch and perform another echo test.
Let us know if the echo test works or if it still fails and please remember to share with us the relevant dmesg logs.
Thank you, Geraldo Nascimento
--- clock.c.orig 2021-07-17 12:15:06.416028360 -0000 +++ clock.c 2021-07-27 17:36:16.954774954 -0000 @@ -324,9 +324,45 @@ static int __uac_clock_find_source(struc sources[ret - 1], visited, validate); if (ret > 0) { - err = uac_clock_selector_set_val(chip, entity_id, cur); - if (err < 0) - return err; + if (chip->usb_id == USB_ID(0x04e8, 0xa051)) { + if (entity_id == 0xc) { + err = uac_clock_selector_set_val(chip, entity_id, 2); + if (err < 0) + return err; + + err = uac_clock_selector_get_val(chip, 0xc); + if (err > 0) { + printk(KERN_ERR "__uac_clock_find_source: Clock Selector 0xc has pin %d selected", err); + } + + err = uac_clock_selector_get_val(chip, 0xb); + if (err > 0) { + printk(KERN_ERR "__uac_clock_find_source: Clock Selector 0xb has pin %d selected", err); + } + } + + else if (entity_id == 0xb) { + err = uac_clock_selector_set_val(chip, entity_id, 1); + if (err < 0) + return err; + + err = uac_clock_selector_get_val(chip, 0xc); + if (err > 0) { + printk(KERN_ERR "__uac_clock_find_source: Clock Selector 0xc has pin %d selected", err); + } + + err = uac_clock_selector_get_val(chip, 0xb); + if (err > 0) { + printk(KERN_ERR "__uac_clock_find_source: Clock Selector 0xb has pin %d selected", err); + } + } + } + + else { + err = uac_clock_selector_set_val(chip, entity_id, cur); + if (err < 0) + return err; + } }
if (!validate || ret > 0 || !chip->autoclock)
Chihhao, please try the below patch and perform another echo test.
Let us know if the echo test works or if it still fails and please remember to share with us the relevant dmesg logs.
Chihhao Chen,
when I said echo test I meant just test the device and observe if there are noises related to incorrect clock setting.
I then realized a bit too late that probably in your programming culture "echo test" refers to the whole technique you used to hook the kernel and debug the issue to produce the fix.
If you call that technique "echo test" I'm sorry, I didn't know it was called that way.
I just meant to briefly test to hear if there is noise even when the clock parameters are correctly hardcoded.
My intent with that patch is to try to prove there's a firmware bug in action just like Takashi Iwai suggested.
My apologies, Geraldo Nascimento
From: chihhao chen chihhao.chen@mediatek.com
Hi Geraldo Nascimento,
For echo test, it means we use this earphone to receive and play sounds at the same time. We found in this case serious noise problem happens.
Log as follows with your patch <6>[ 175.960387][T401365] __uac_clock_find_source: Clock Selector 0xc has pin 2 selected <6>[ 175.966980][T401365] __uac_clock_find_source: Clock Selector 0xb has pin 2 selected <6>[ 176.026251][T400354] __uac_clock_find_source: Clock Selector 0xc has pin 1 selected <6>[ 176.032406][T400354] __uac_clock_find_source: Clock Selector 0xb has pin 1 selected
There is no noise and I think this should be a firmware bug.
Thanks Chihhao
On Thu, Aug 5, 2021 at 7:54 AM chihhao.chen@mediatek.com wrote:
From: chihhao chen chihhao.chen@mediatek.com
Hi Geraldo Nascimento,
Hi Chihhao Chen!
For echo test, it means we use this earphone to receive and play sounds at the same time. We found in this case serious noise problem happens.
That's what I understood initially, and only a little later I became afraid echo test was the name of the debugging technique you used :)
Thanks for the clarification.
Log as follows with your patch <6>[ 175.960387][T401365] __uac_clock_find_source: Clock Selector 0xc has pin 2 selected <6>[ 175.966980][T401365] __uac_clock_find_source: Clock Selector 0xb has pin 2 selected <6>[ 176.026251][T400354] __uac_clock_find_source: Clock Selector 0xc has pin 1 selected <6>[ 176.032406][T400354] __uac_clock_find_source: Clock Selector 0xb has pin 1 selected
There is no noise and I think this should be a firmware bug.
From the log I'm afraid my worst assumptions were right.
Regardless of which Clock Selector we want to set, the firmware will always set them both.
We should contact Samsung now that we have at least a sketch of a bug report...
Thanks, Geraldo Nascimento
Thanks Chihhao
Em Qui, 5 de ago de 2021 12:50, Geraldo Nascimento geraldogabriel@gmail.com escreveu:
There is no noise and I think this should be a firmware bug.
The fact that there's no noise during the echo test itself doesn't mean it's not a firmware bug.
It may just mean the MIC is able to support both 48KHz and 96KHz.
Because from the log we see both Clock Selectors end up selected to pin 1.
Thank you, Geraldo Nascimento
Chihhao Chen,
Since you're the one with the hardware I believe the bug report for Samsung should be done by your side.
Maybe Takashi Iwai has somebody in mind that develops audio code for Samsung and has already contributed to ALSA?
It's best if Takashi Iwai helps you out, and you take the shortcut to the corporate structure, otherwise the importance of our findings may be underestimated.
Chihhao Chen, please include the following link in your bug report to Samsung: https://lore.kernel.org/patchwork/patch/1466711/
I'd politely remind them that the headset requires a very specific quirk to work with recent Linux, that the patch has entered Linux-Next however it's a workaround because the headset is apparently breaking UAC compliance when it sets both Clock Selectors at the same time, regardless of what Clock Selector we tell it to set via USB Control Message.
But you're free to do as you wish. I'm sure you'll be fine.
Good luck, Geraldo Nascimento
On Thu, Aug 5, 2021 at 11:31 PM Geraldo Nascimento geraldogabriel@gmail.com wrote:
Em Qui, 5 de ago de 2021 12:50, Geraldo Nascimento geraldogabriel@gmail.com escreveu:
There is no noise and I think this should be a firmware bug.
The fact that there's no noise during the echo test itself doesn't mean it's not a firmware bug.
It may just mean the MIC is able to support both 48KHz and 96KHz.
Because from the log we see both Clock Selectors end up selected to pin 1.
Thank you, Geraldo Nascimento
participants (4)
-
chihhao chen -
chihhao.chen＠mediatek.com
-
Geraldo Nascimento -
Takashi Iwai