Fix for MediaTek reset affecting Focusrite audio interfaces
Subsystems (3)
cover Fix for MediaTek reset affecting Focusrite audio interfaces
This series (choose 1 of the 2 patches) addresses an issue where the MT7922 Bluetooth controller reset added in commit ccfc8948d7e4 ("Bluetooth: btusb: mediatek: reset the controller before downloading the fw") is causing Focusrite USB audio devices to fail to initialise when connected during boot on kernels 6.11 and newer.
Reported by three users here:
https://github.com/geoffreybennett/linux-fcp/issues/24
Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar change made in commit a7208610761a ("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if the reset should be reverted for the MT7925 as well as the MT7922.
Option 1: Revert the problematic reset behaviour entirely (MT7922 + MT7925)
Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller
before downloading the fw"
[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before
downloading the fw
--
2.45.0
Quoted 23 lines · Geoffrey D. Bennett
This series (choose 1 of the 2 patches) addresses an issue where the MT7922 Bluetooth controller reset added in commit
ccfc8948d7e4("Bluetooth: btusb: mediatek: reset the controller before downloading the fw") is causing Focusrite USB audio devices to fail to initialise when connected during boot on kernels 6.11 and newer.Reported by three users here:
https://github.com/geoffreybennett/linux-fcp/issues/24Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar change made in commit
a7208610761a("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if the reset should be reverted for the MT7925 as well as the MT7922.Option 1: Revert the problematic reset behaviour entirely (MT7922 + MT7925)
Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
Quoted 10 lines
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller
before downloading the fw"[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before
downloading the fw--
2.45.0
Hi Geoffrey,
I understand there's no apparent nexus of causality here.
However if three different users suddenly reported the same problem and the fix fixes it, we should take the report seriously and back down on the problematic commit until we figure for sure what the heck is going on.
My bet is this is bona fide bug in the USB subsystem, but either I'm not looking hard enough or I'm looking into the wrong places, because there's no way I can see in which way USB bluetooth driver from MediaTek could cause clock detection to fail.
No point in pressing this harder, but yeah, take the reports more than seriously, because if we don't understand in which way our own (Linux) code could be causing this, at least we should take cautionary measures.
In that sense, putting Takashi Iwai and Greg KH to Cc: in a modest but sincere belief that this issue is more than real, even though it looks like anticipated April 1st. Takashi can help with his expertise in clock detection and Greg could help pinpoint if this is indeed madness in the USB subsystem.
Hope you and them don't mind the escalation :)
Thanks,
Geraldo Nascimento
Luiz Augusto von Dentz
maintainerHi Chris, Sean, Hao,
Quoted 54 lines · Geraldo Nascimento
Quoted 23 lines · Geoffrey D. Bennett
This series (choose 1 of the 2 patches) addresses an issue where the MT7922 Bluetooth controller reset added in commit
ccfc8948d7e4("Bluetooth: btusb: mediatek: reset the controller before downloading the fw") is causing Focusrite USB audio devices to fail to initialise when connected during boot on kernels 6.11 and newer.Reported by three users here:
https://github.com/geoffreybennett/linux-fcp/issues/24Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar change made in commit
a7208610761a("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if the reset should be reverted for the MT7925 as well as the MT7922.Option 1: Revert the problematic reset behaviour entirely (MT7922 + MT7925)
Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
Quoted 10 lines
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller
before downloading the fw"[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before
downloading the fw--
2.45.0Hi Geoffrey,
I understand there's no apparent nexus of causality here.
However if three different users suddenly reported the same problem and the fix fixes it, we should take the report seriously and back down on the problematic commit until we figure for sure what the heck is going on.
My bet is this is bona fide bug in the USB subsystem, but either I'm not looking hard enough or I'm looking into the wrong places, because there's no way I can see in which way USB bluetooth driver from MediaTek could cause clock detection to fail.
No point in pressing this harder, but yeah, take the reports more than seriously, because if we don't understand in which way our own (Linux) code could be causing this, at least we should take cautionary measures.
Quoted 7 lines
In that sense, putting Takashi Iwai and Greg KH to Cc: in a modest but sincere belief that this issue is more than real, even though it looks like anticipated April 1st. Takashi can help with his expertise in clock detection and Greg could help pinpoint if this is indeed madness in the USB subsystem.
Hope you and them don't mind the escalation :)
Do you guys have any idea what could be possibly going on here? There really seems something is not right if one driver affects the other, especially if the devices are not actually related in any way.
Signature · 1 line
Luiz Augusto von Dentz
Quoted 81 lines · Luiz Augusto von Dentz
Hi Chris, Sean, Hao,
Quoted 77 lines · Geraldo Nascimento
Quoted 33 lines · Geoffrey D. Bennett
This series (choose 1 of the 2 patches) addresses an issue where the
MT7922 Bluetooth controller reset added in commitccfc8948d7e4("Bluetooth: btusb: mediatek: reset the controller before downloading
the fw") is causing Focusrite USB audio devices to fail to initialise
when connected during boot on kernels 6.11 and newer.Reported by three users here:
https://github.com/geoffreybennett/linux-fcp/issues/24Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar
change made in commita7208610761a("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if the
reset should be reverted for the MT7925 as well as the MT7922.Option 1: Revert the problematic reset behaviour entirely (MT7922 +
MT7925)Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller
before downloading the fw"Quoted 7 lines
[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before downloading the fw
--
2.45.0Hi Geoffrey,
I understand there's no apparent nexus of causality here.
However if three different users suddenly reported the same problem and
the fix fixes it, we should take the report seriously and back down on the problematic commit until we figure for sure what the heck is going on.My bet is this is bona fide bug in the USB subsystem, but either I'm not
looking hard enough or I'm looking into the wrong places, because there's no way I can see in which way USB bluetooth driver from MediaTek could cause clock detection to fail.No point in pressing this harder, but yeah, take the reports more than
seriously, because if we don't understand in which way our own (Linux)
code could be causing this, at least we should take cautionary measures.In that sense, putting Takashi Iwai and Greg KH to Cc: in a modest but
sincere belief that this issue is more than real, even though it looks
like anticipated April 1st. Takashi can help with his expertise in clock
detection and Greg could help pinpoint if this is indeed madness in the
USB subsystem.
Quoted 7 lines
Quoted 1 line
Hope you and them don't mind the escalation :)
Do you guys have any idea what could be possibly going on here? There really seems something is not right if one driver affects the other, especially if the devices are not actually related in any way.
Hello everyone,
I did the kernel bisection on this issue and tested Geoffrey's patches since I own the affected combination of hardware and can reliably reproduce the issue.
I was now able to capture both the sound device initialization failure and the successful initialization from initramfs using the usbmon module. One difference that is immediately noticeable is the amount of URB_CONTROL data that is sent to the mediatek btusb device and a considerable slow-down of the boot process (around 16s difference).
Following is some more information about the timing and other related messages from the kernel message log. The first excerpt is from an unpatched kernel 6.14.0-rc5:
[ 59.987242] lunar kernel: usb 1-2: new high-speed USB device number 2 using xhci_hcd [ 59.987975] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45 [ 59.988048] lunar kernel: usb 1-2: New USB device strings: Mfr=1, Product=3, SerialNumber=2 [ 59.988121] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 59.988197] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 60.571193] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002) [ 60.571314] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision: 79220010 [ 60.571501] lunar kernel: usbcore: registered new interface driver btusb [ 60.571513] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version: 0x8a108a10, Build Time: 20241106163228a [ 60.571627] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware Version: ____000000, Build Time: 20241106163310 [ 60.616673] lunar kernel: Bluetooth: hci0: HW/SW Version: 0x008a008a, Build Time: 20241106163512 [ 64.549017] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0 [ 70.700407] lunar kernel: usb 1-2: parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41) [ 81.264360] lunar kernel: usb 1-2: parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41) [ 81.264611] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to https://github.com/geoffreybennett/scarlett-gen2/issues [ 81.264782] lunar kernel: usb 1-2: Error initialising Scarlett Gen 3 Mixer Driver: -71 [ 81.264903] lunar kernel: snd-usb-audio 1-2:1.0: probe with driver snd-usb-audio failed with error -71 [ 81.265040] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0 [ 81.400030] lunar kernel: Bluetooth: hci0: Device setup in 20435397 usecs [ 81.400114] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 81.680052] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00 [ 81.680170] lunar kernel: Bluetooth: hci0: AOSP quality report is supported [ 87.373353] lunar kernel: usbcore: registered new interface driver snd-usb-audio
The following second excerpt is from the same kernel version with Geoffrey's second patch applied:
[ 139.603887] lunar kernel: usb 1-2: new high-speed USB device number 2 using xhci_hcd [ 139.604524] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45 [ 139.604598] lunar kernel: usb 1-2: New USB device strings: Mfr=1, Product=3, SerialNumber=2 [ 139.604672] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 139.604744] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 140.192488] lunar kernel: usbcore: registered new interface driver btusb [ 140.192501] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002) [ 140.192627] lunar kernel: Bluetooth: hci0: HW/SW Version: 0x008a008a, Build Time: 20241106163512 [ 140.192639] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision: 79220010 [ 140.192753] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version: 0x8a108a10, Build Time: 20241106163228a [ 140.192866] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware Version: ____000000, Build Time: 20241106163310 [ 140.240011] lunar kernel: Bluetooth: hci0: Device setup in 188119 usecs [ 140.240048] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 140.513341] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00 [ 140.513379] lunar kernel: Bluetooth: hci0: AOSP quality report is supported [ 144.115475] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0 [ 144.115747] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to https://github.com/geoffreybennett/scarlett-gen2/issues [ 144.115857] lunar kernel: usb 1-2: Firmware version 1605 [ 144.115954] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0 [ 147.843340] lunar kernel: usbcore: registered new interface driver snd-usb-audio
Comparing the USB captures from the hub, filtered to just the first few messages of the audio device yield the following exchanges. Again, first the broken case:
No. Time Source Destination Protocol Length Info
7 0.011098 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE
8 0.015218 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE
9 0.015272 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION
10 0.017966 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION
11 0.018020 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION
12 0.025965 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION
23 2.482241 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
24 2.484931 1.2.0 host USB 98 GET DESCRIPTOR Response STRING
25 2.484939 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
26 2.487916 1.2.0 host USB 84 GET DESCRIPTOR Response STRING
27 2.487941 host 1.2.0 USBAUDIO 65 SET CUR CX_CLOCK_SELECTOR_CONTROL request
2047 7.661154 1.2.0 host USBAUDIO 64 SET CUR CX_CLOCK_SELECTOR_CONTROL status
2048 7.661197 host 1.2.0 USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL request
4097 12.781553 1.2.0 host USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL[Malformed Packet]The first discrepancy can be found in the URB status field of packet no. 2047 (vs. 74 below) which is -2: No such file or directory (- ENOENT).
In the second (following) case where the sound device initializes okay, the URB status of the similar packet is Success (0) instead:
No. Time Source Destination Protocol Length Info
7 0.011565 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE
8 0.015471 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE
9 0.015523 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION
10 0.018477 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION
11 0.018537 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION
12 0.026331 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION
23 4.168190 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
24 4.171031 1.2.0 host USB 98 GET DESCRIPTOR Response STRING
25 4.171040 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
26 4.174021 1.2.0 host USB 84 GET DESCRIPTOR Response STRING
27 4.174041 host 1.2.0 USBAUDIO 65 SET CUR CX_CLOCK_SELECTOR_CONTROL request
74 4.476033 1.2.0 host USBAUDIO 64 SET CUR CX_CLOCK_SELECTOR_CONTROL status
75 4.476050 host 1.2.0 USBAUDIO 64 GET CUR CX_CLOCK_SELECTOR_CONTROL request
76 4.479034 1.2.0 host USBAUDIO 65 GET CUR CX_CLOCK_SELECTOR_CONTROL response
77 4.479050 host 1.2.0 USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL request
202 4.582043 1.2.0 host USBAUDIO 66 GET RANGE CS_SAM_FREQ_CONTROL responsePlease ask, if I should provide any more information or how and to whom I can provide the full USB captures.
Thank you,
Ben
Quoted 202 lines · Benedikt Ziemons
Quoted 81 lines · Luiz Augusto von Dentz
Hi Chris, Sean, Hao,
Quoted 77 lines · Geraldo Nascimento
Quoted 33 lines · Geoffrey D. Bennett
This series (choose 1 of the 2 patches) addresses an issue where the
MT7922 Bluetooth controller reset added in commitccfc8948d7e4("Bluetooth: btusb: mediatek: reset the controller before downloading
the fw") is causing Focusrite USB audio devices to fail to initialise
when connected during boot on kernels 6.11 and newer.Reported by three users here:
https://github.com/geoffreybennett/linux-fcp/issues/24Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar
change made in commita7208610761a("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if the
reset should be reverted for the MT7925 as well as the MT7922.Option 1: Revert the problematic reset behaviour entirely (MT7922 +
MT7925)Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller
before downloading the fw"Quoted 7 lines
[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before downloading the fw
--
2.45.0Hi Geoffrey,
I understand there's no apparent nexus of causality here.
However if three different users suddenly reported the same problem and
the fix fixes it, we should take the report seriously and back down on the problematic commit until we figure for sure what the heck is going on.My bet is this is bona fide bug in the USB subsystem, but either I'm not
looking hard enough or I'm looking into the wrong places, because there's no way I can see in which way USB bluetooth driver from MediaTek could cause clock detection to fail.No point in pressing this harder, but yeah, take the reports more than
seriously, because if we don't understand in which way our own (Linux)
code could be causing this, at least we should take cautionary measures.In that sense, putting Takashi Iwai and Greg KH to Cc: in a modest but
sincere belief that this issue is more than real, even though it looks
like anticipated April 1st. Takashi can help with his expertise in clock
detection and Greg could help pinpoint if this is indeed madness in the
USB subsystem.Quoted 7 lines
Quoted 1 line
Hope you and them don't mind the escalation :)
Do you guys have any idea what could be possibly going on here? There really seems something is not right if one driver affects the other, especially if the devices are not actually related in any way.
Hello everyone,
I did the kernel bisection on this issue and tested Geoffrey's patches since I own the affected combination of hardware and can reliably reproduce the issue.
I was now able to capture both the sound device initialization failure and the successful initialization from initramfs using the usbmon module. One difference that is immediately noticeable is the amount of
URB_CONTROLdata that is sent to the mediatek btusb device and a considerable slow-down of the boot process (around 16s difference).Following is some more information about the timing and other related messages from the kernel message log. The first excerpt is from an unpatched kernel 6.14.0-rc5:
[ 59.987242] lunar kernel: usb 1-2: new high-speed USB device number 2 using xhci_hcd [ 59.987975] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45 [ 59.988048] lunar kernel: usb 1-2: New USB device strings: Mfr=1, Product=3, SerialNumber=2 [ 59.988121] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 59.988197] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 60.571193] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002) [ 60.571314] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision: 79220010 [ 60.571501] lunar kernel: usbcore: registered new interface driver btusb [ 60.571513] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version: 0x8a108a10, Build Time: 20241106163228a [ 60.571627] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware Version: ____000000, Build Time: 20241106163310 [ 60.616673] lunar kernel: Bluetooth: hci0: HW/SW Version: 0x008a008a, Build Time: 20241106163512 [ 64.549017] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0 [ 70.700407] lunar kernel: usb 1-2: parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41) [ 81.264360] lunar kernel: usb 1-2: parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41) [ 81.264611] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to https://github.com/geoffreybennett/scarlett-gen2/issues [ 81.264782] lunar kernel: usb 1-2: Error initialising Scarlett Gen 3 Mixer Driver: -71 [ 81.264903] lunar kernel: snd-usb-audio 1-2:1.0: probe with driver snd-usb-audio failed with error -71 [ 81.265040] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0 [ 81.400030] lunar kernel: Bluetooth: hci0: Device setup in 20435397 usecs [ 81.400114] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 81.680052] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00 [ 81.680170] lunar kernel: Bluetooth: hci0: AOSP quality report is supported [ 87.373353] lunar kernel: usbcore: registered new interface driver snd-usb-audioThe following second excerpt is from the same kernel version with Geoffrey's second patch applied:
[ 139.603887] lunar kernel: usb 1-2: new high-speed USB device number 2 using xhci_hcd [ 139.604524] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45 [ 139.604598] lunar kernel: usb 1-2: New USB device strings: Mfr=1, Product=3, SerialNumber=2 [ 139.604672] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 139.604744] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 140.192488] lunar kernel: usbcore: registered new interface driver btusb [ 140.192501] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002) [ 140.192627] lunar kernel: Bluetooth: hci0: HW/SW Version: 0x008a008a, Build Time: 20241106163512 [ 140.192639] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision: 79220010 [ 140.192753] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version: 0x8a108a10, Build Time: 20241106163228a [ 140.192866] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware Version: ____000000, Build Time: 20241106163310 [ 140.240011] lunar kernel: Bluetooth: hci0: Device setup in 188119 usecs [ 140.240048] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 140.513341] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00 [ 140.513379] lunar kernel: Bluetooth: hci0: AOSP quality report is supported [ 144.115475] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0 [ 144.115747] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to https://github.com/geoffreybennett/scarlett-gen2/issues [ 144.115857] lunar kernel: usb 1-2: Firmware version 1605 [ 144.115954] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0 [ 147.843340] lunar kernel: usbcore: registered new interface driver snd-usb-audioComparing the USB captures from the hub, filtered to just the first few messages of the audio device yield the following exchanges. Again, first the broken case:
No. Time Source Destination Protocol Length Info 7 0.011098 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE 8 0.015218 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE 9 0.015272 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 10 0.017966 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION 11 0.018020 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 12 0.025965 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION 23 2.482241 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING 24 2.484931 1.2.0 host USB 98 GET DESCRIPTOR Response STRING 25 2.484939 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING 26 2.487916 1.2.0 host USB 84 GET DESCRIPTOR Response STRING 27 2.487941 host 1.2.0 USBAUDIO 65 SET CUR CX_CLOCK_SELECTOR_CONTROL request 2047 7.661154 1.2.0 host USBAUDIO 64 SET CUR CX_CLOCK_SELECTOR_CONTROL status 2048 7.661197 host 1.2.0 USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL request 4097 12.781553 1.2.0 host USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL[Malformed Packet]The first discrepancy can be found in the URB status field of packet no. 2047 (vs. 74 below) which is -2: No such file or directory (- ENOENT).
In the second (following) case where the sound device initializes okay, the URB status of the similar packet is Success (0) instead:No. Time Source Destination Protocol Length Info 7 0.011565 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE 8 0.015471 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE 9 0.015523 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 10 0.018477 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION 11 0.018537 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 12 0.026331 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION 23 4.168190 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING 24 4.171031 1.2.0 host USB 98 GET DESCRIPTOR Response STRING 25 4.171040 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING 26 4.174021 1.2.0 host USB 84 GET DESCRIPTOR Response STRING 27 4.174041 host 1.2.0 USBAUDIO 65 SET CUR CX_CLOCK_SELECTOR_CONTROL request 74 4.476033 1.2.0 host USBAUDIO 64 SET CUR CX_CLOCK_SELECTOR_CONTROL status 75 4.476050 host 1.2.0 USBAUDIO 64 GET CUR CX_CLOCK_SELECTOR_CONTROL request 76 4.479034 1.2.0 host USBAUDIO 65 GET CUR CX_CLOCK_SELECTOR_CONTROL response 77 4.479050 host 1.2.0 USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL request 202 4.582043 1.2.0 host USBAUDIO 66 GET RANGE CS_SAM_FREQ_CONTROL response
Quoted 5 lines
Please ask, if I should provide any more information or how and to whom I can provide the full USB captures.
Thank you,
Ben
Hi Ben,
This bit sticks out to me:
Quoted 1 line
[ 81.400030] lunar kernel: Bluetooth: hci0: Device setup in 20435397 usecs
20 seconds to initialise bluetooth, vs. 188ms when not doing the reset:
Quoted 1 line
[ 140.240011] lunar kernel: Bluetooth: hci0: Device setup in 188119 usecs
Does it still take 20 seconds when the Scarlett is not plugged in at boot time?
I presume lsusb -t shows both devices on the same USB bus?
Thanks,
Geoffrey.
Hi Geoffrey,
Quoted 316 lines · Geoffrey D. Bennett
Quoted 301 lines · Benedikt Ziemons
Quoted 97 lines · Luiz Augusto von Dentz
Hi Chris, Sean, Hao,
Quoted 93 lines · Geraldo Nascimento
Quoted 38 lines · Geoffrey D. Bennett
This series (choose 1 of the 2 patches) addresses an issue where
the
MT7922 Bluetooth controller reset added in commitccfc8948d7e4
("Bluetooth: btusb: mediatek: reset the controller before downloading
the fw") is causing Focusrite USB audio devices to fail to initialise
when connected during boot on kernels 6.11 and newer.Reported by three users here:
https://github.com/geoffreybennett/linux-fcp/issues/24Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar
change made in commita7208610761a("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if
the
reset should be reverted for the MT7925 as well as the MT7922.Option 1: Revert the problematic reset behaviour entirely (MT7922
+
MT7925)Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller
before downloading the fw"Quoted 8 lines
[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before
downloading the fw--
2.45.0Hi Geoffrey,
I understand there's no apparent nexus of causality here.
However if three different users suddenly reported the same problem
and
the fix fixes it, we should take the report seriously and back down
on the problematic commit until we figure for sure what the heck is
going on.My bet is this is bona fide bug in the USB subsystem, but either
I'm not
looking hard enough or I'm looking into the wrong places, because
there's no way I can see in which way USB bluetooth driver from MediaTek could cause clock detection to fail.No point in pressing this harder, but yeah, take the reports more
than
seriously, because if we don't understand in which way our own (Linux)
code could be causing this, at least we should take cautionary measures.In that sense, putting Takashi Iwai and Greg KH to Cc: in a modest
but
sincere belief that this issue is more than real, even though it
looks
like anticipated April 1st. Takashi can help with his expertise in
clock
detection and Greg could help pinpoint if this is indeed madness in
the
USB subsystem.Quoted 9 lines
Quoted 1 line
Hope you and them don't mind the escalation :)
Do you guys have any idea what could be possibly going on here? There
really seems something is not right if one driver affects the other,
especially if the devices are not actually related in any way.Hello everyone,
I did the kernel bisection on this issue and tested Geoffrey's patches
since I own the affected combination of hardware and can reliably reproduce the issue.I was now able to capture both the sound device initialization failure
and the successful initialization from initramfs using the usbmon module. One difference that is immediately noticeable is the amount ofURB_CONTROLdata that is sent to the mediatek btusb device and a considerable slow-down of the boot process (around 16s difference).Following is some more information about the timing and other related
messages from the kernel message log. The first excerpt is from an unpatched kernel 6.14.0-rc5:[ 59.987242] lunar kernel: usb 1-2: new high-speed USB device number 2 using
xhci_hcd
[ 59.987975] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45
[ 59.988048] lunar kernel: usb 1-2: New USB device strings:
Mfr=1, Product=3, SerialNumber=2
[ 59.988121] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 59.988197] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 60.571193] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002)
[ 60.571314] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision:
79220010
[ 60.571501] lunar kernel: usbcore: registered new interface driver btusb
[ 60.571513] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version:0x8a108a10, Build Time:20241106163228a
[ 60.571627] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware
Version: ____000000, Build Time: 20241106163310
[ 60.616673] lunar kernel: Bluetooth: hci0: HW/SW Version:0x008a008a, Build Time: 20241106163512
[ 64.549017] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0
[ 70.700407] lunar kernel: usb 1-2:parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41)
[ 81.264360] lunar kernel: usb 1-2:parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41)
[ 81.264611] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to
https://github.com/geoffreybennett/scarlett-gen2/issues
[ 81.264782] lunar kernel: usb 1-2: Error initialising Scarlett Gen 3 Mixer Driver: -71
[ 81.264903] lunar kernel: snd-usb-audio 1-2:1.0: probe with driver snd-usb-audio failed with error -71
[ 81.265040] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0
[ 81.400030] lunar kernel: Bluetooth: hci0: Device setup in 20435397 usecs
[ 81.400114] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 81.680052] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00
[ 81.680170] lunar kernel: Bluetooth: hci0: AOSP quality report is supported
[ 87.373353] lunar kernel: usbcore: registered new interface driver snd-usb-audioThe following second excerpt is from the same kernel version with Geoffrey's second patch applied:
[ 139.603887] lunar kernel: usb 1-2: new high-speed USB device number 2 using
xhci_hcd
[ 139.604524] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45
[ 139.604598] lunar kernel: usb 1-2: New USB device strings:
Mfr=1, Product=3, SerialNumber=2
[ 139.604672] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 139.604744] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 140.192488] lunar kernel: usbcore: registered new interface driver btusb
[ 140.192501] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002)
[ 140.192627] lunar kernel: Bluetooth: hci0: HW/SW Version:0x008a008a, Build Time: 20241106163512
[ 140.192639] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision:
79220010
[ 140.192753] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version:0x8a108a10, Build Time:20241106163228a
[ 140.192866] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware
Version: ____000000, Build Time: 20241106163310
[ 140.240011] lunar kernel: Bluetooth: hci0: Device setup in 188119 usecs
[ 140.240048] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 140.513341] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00
[ 140.513379] lunar kernel: Bluetooth: hci0: AOSP quality report is supported
[ 144.115475] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0
[ 144.115747] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to
https://github.com/geoffreybennett/scarlett-gen2/issues
[ 144.115857] lunar kernel: usb 1-2: Firmware version 1605 [ 144.115954] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0
[ 147.843340] lunar kernel: usbcore: registered new interface driver snd-usb-audioComparing the USB captures from the hub, filtered to just the first few
messages of the audio device yield the following exchanges. Again, first the broken case:No. Time Source Destination Protocol Length Info
7 0.011098 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE
8 0.015218 1.2.0 host
USB 82 GET DESCRIPTOR Response DEVICE
9 0.015272 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 10 0.017966 1.2.0 host
USB 73 GET DESCRIPTOR Response CONFIGURATION 11 0.018020 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 12 0.025965 1.2.0 host
USB 425 GET DESCRIPTOR Response CONFIGURATION 23 2.482241 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
24 2.484931 1.2.0 host
USB 98 GET DESCRIPTOR Response STRING
25 2.484939 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
26 2.487916 1.2.0 host
USB 84 GET DESCRIPTOR Response STRING
27 2.487941 host 1.2.0 USBAUDIO 65 SET CURCX_CLOCK_SELECTOR_CONTROLrequest 2047 7.661154 1.2.0 host USBAUDIO 64 SET CURCX_CLOCK_SELECTOR_CONTROLstatus 2048 7.661197 host 1.2.0 USBAUDIO 64 GET RANGECS_SAM_FREQ_CONTROLrequest 4097 12.781553 1.2.0 host USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL[Malformed Packet]The first discrepancy can be found in the URB status field of packet
no. 2047 (vs. 74 below) which is -2: No such file or directory (- ENOENT).
In the second (following) case where the sound device initializes okay,
the URB status of the similar packet is Success (0) instead:No. Time Source Destination Protocol Length Info
7 0.011565 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE
8 0.015471 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE
9 0.015523 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 10 0.018477 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION 11 0.018537 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 12 0.026331 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION 23 4.168190 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
24 4.171031 1.2.0 host USB 98 GET DESCRIPTOR Response STRING
25 4.171040 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
26 4.174021 1.2.0 host USB 84 GET DESCRIPTOR Response STRING
27 4.174041 host 1.2.0 USBAUDIO 65 SET CURCX_CLOCK_SELECTOR_CONTROLrequest 74 4.476033 1.2.0 host USBAUDIO 64 SET CURCX_CLOCK_SELECTOR_CONTROLstatus 75 4.476050 host 1.2.0 USBAUDIO 64 GET CURCX_CLOCK_SELECTOR_CONTROLrequest 76 4.479034 1.2.0 host USBAUDIO 65 GET CURCX_CLOCK_SELECTOR_CONTROLresponse 77 4.479050 host 1.2.0 USBAUDIO 64 GET RANGECS_SAM_FREQ_CONTROLrequest 202 4.582043 1.2.0 host USBAUDIO 66 GET RANGECS_SAM_FREQ_CONTROLresponseQuoted 6 lines
Please ask, if I should provide any more information or how and to whom
I can provide the full USB captures.Thank you,
BenHi Ben,
This bit sticks out to me:
Quoted 2 lines
[ 81.400030] lunar kernel: Bluetooth: hci0: Device setup in 20435397 usecs
Quoted 8 lines
20 seconds to initialise bluetooth, vs. 188ms when not doing the reset:
Quoted 2 lines
[ 140.240011] lunar kernel: Bluetooth: hci0: Device setup in 188119 usecs
Does it still take 20 seconds when the Scarlett is not plugged in at boot time?
Yes, it still takes 20 seconds without the Scarlett plugged in. Here are kernel messages about the mediatek device in this case:
[ 6.620016] lunar kernel: Bluetooth: Core ver 2.22 [ 6.666695] lunar kernel: NET: Registered PF_BLUETOOTH protocol family [ 6.666756] lunar kernel: Bluetooth: HCI device and connection manager initialized [ 6.666770] lunar kernel: Bluetooth: HCI socket layer initialized [ 6.666784] lunar kernel: Bluetooth: L2CAP socket layer initialized [ 6.666794] lunar kernel: Bluetooth: SCO socket layer initialized [ 6.912392] lunar kernel: usbcore: registered new interface driver btusb [ 6.912402] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002) [ 6.912519] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision: 79220010 [ 6.912627] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version: 0x8a108a10, Build Time: 20241106163228a [ 6.912739] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware Version: ____000000, Build Time: 20241106163310 [ 6.912847] lunar kernel: Bluetooth: hci0: HW/SW Version: 0x008a008a, Build Time: 20241106163512 [ 8.010016] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0 [ 12.723353] lunar kernel: usbcore: registered new interface driver snd-usb-audio [ 27.606693] lunar kernel: Bluetooth: hci0: Device setup in 20342576 usecs [ 27.606748] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 27.883398] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00 [ 27.883466] lunar kernel: Bluetooth: hci0: AOSP quality report is supported
Quoted 2 lines
I presume lsusb -t shows both devices on the same USB bus?
Yes, indeed, they are both on bus 1. For the captures I also made sure that no other devices are on the same bus.
I also just tried connecting the Scarlett to a different USB bus and cannot reproduce the original issue with this setup.
Quoted 3 lines
Thanks,
Geoffrey.
[Cc: +regressions, commit ccfc8948d7e4 was added in v6.11-rc1]
Am 13.03.25 um 02:43 schrieb Benedikt Ziemons:
Quoted 181 lines · Luiz Augusto von Dentz
Quoted 63 lines · Geraldo Nascimento
Quoted 61 lines · Geoffrey D. Bennett
Quoted 23 lines
This series (choose 1 of the 2 patches) addresses an issue where the MT7922 Bluetooth controller reset added in commit
ccfc8948d7e4("Bluetooth: btusb: mediatek: reset the controller before downloading the fw") is causing Focusrite USB audio devices to fail to initialise when connected during boot on kernels 6.11 and newer.Reported by three users here:
https://github.com/geoffreybennett/linux-fcp/issues/24Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar change made in commit
a7208610761a("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if the reset should be reverted for the MT7925 as well as the MT7922.Option 1: Revert the problematic reset behaviour entirely (MT7922 + MT7925)
Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
Quoted 10 lines
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller before downloading the fw"
[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before downloading the fw
--
2.45.0Hi Geoffrey,
I understand there's no apparent nexus of causality here.
However if three different users suddenly reported the same problem and the fix fixes it, we should take the report seriously and back down on the problematic commit until we figure for sure what the heck is going on.
My bet is this is bona fide bug in the USB subsystem, but either I'm not looking hard enough or I'm looking into the wrong places, because there's no way I can see in which way USB bluetooth driver from MediaTek could cause clock detection to fail.
No point in pressing this harder, but yeah, take the reports more than seriously, because if we don't understand in which way our own (Linux) code could be causing this, at least we should take cautionary measures.
In that sense, putting Takashi Iwai and Greg KH to Cc: in a modest but sincere belief that this issue is more than real, even though it looks like anticipated April 1st. Takashi can help with his expertise in clock detection and Greg could help pinpoint if this is indeed madness in the USB subsystem.
Quoted 5 lines
Quoted 1 line
Hope you and them don't mind the escalation :)
Do you guys have any idea what could be possibly going on here? There really seems something is not right if one driver affects the other, especially if the devices are not actually related in any way.
I did the kernel bisection on this issue and tested Geoffrey's patches since I own the affected combination of hardware and can reliably reproduce the issue.
I was now able to capture both the sound device initialization failure and the successful initialization from initramfs using the usbmon module. One difference that is immediately noticeable is the amount of
URB_CONTROLdata that is sent to the mediatek btusb device and a considerable slow-down of the boot process (around 16s difference).Following is some more information about the timing and other related messages from the kernel message log. The first excerpt is from an unpatched kernel 6.14.0-rc5:
[ 59.987242] lunar kernel: usb 1-2: new high-speed USB device number 2 using xhci_hcd [ 59.987975] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45 [ 59.988048] lunar kernel: usb 1-2: New USB device strings: Mfr=1, Product=3, SerialNumber=2 [ 59.988121] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 59.988197] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 60.571193] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002) [ 60.571314] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision: 79220010 [ 60.571501] lunar kernel: usbcore: registered new interface driver btusb [ 60.571513] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version: 0x8a108a10, Build Time: 20241106163228a [ 60.571627] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware Version: ____000000, Build Time: 20241106163310 [ 60.616673] lunar kernel: Bluetooth: hci0: HW/SW Version: 0x008a008a, Build Time: 20241106163512 [ 64.549017] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0 [ 70.700407] lunar kernel: usb 1-2: parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41) [ 81.264360] lunar kernel: usb 1-2: parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41) [ 81.264611] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to https://github.com/geoffreybennett/scarlett-gen2/issues [ 81.264782] lunar kernel: usb 1-2: Error initialising Scarlett Gen 3 Mixer Driver: -71 [ 81.264903] lunar kernel: snd-usb-audio 1-2:1.0: probe with driver snd-usb-audio failed with error -71 [ 81.265040] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0 [ 81.400030] lunar kernel: Bluetooth: hci0: Device setup in 20435397 usecs [ 81.400114] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 81.680052] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00 [ 81.680170] lunar kernel: Bluetooth: hci0: AOSP quality report is supported [ 87.373353] lunar kernel: usbcore: registered new interface driver snd-usb-audioThe following second excerpt is from the same kernel version with Geoffrey's second patch applied:
[ 139.603887] lunar kernel: usb 1-2: new high-speed USB device number 2 using xhci_hcd [ 139.604524] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45 [ 139.604598] lunar kernel: usb 1-2: New USB device strings: Mfr=1, Product=3, SerialNumber=2 [ 139.604672] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 139.604744] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 140.192488] lunar kernel: usbcore: registered new interface driver btusb [ 140.192501] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002) [ 140.192627] lunar kernel: Bluetooth: hci0: HW/SW Version: 0x008a008a, Build Time: 20241106163512 [ 140.192639] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision: 79220010 [ 140.192753] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version: 0x8a108a10, Build Time: 20241106163228a [ 140.192866] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware Version: ____000000, Build Time: 20241106163310 [ 140.240011] lunar kernel: Bluetooth: hci0: Device setup in 188119 usecs [ 140.240048] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 140.513341] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00 [ 140.513379] lunar kernel: Bluetooth: hci0: AOSP quality report is supported [ 144.115475] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0 [ 144.115747] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to https://github.com/geoffreybennett/scarlett-gen2/issues [ 144.115857] lunar kernel: usb 1-2: Firmware version 1605 [ 144.115954] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0 [ 147.843340] lunar kernel: usbcore: registered new interface driver snd-usb-audioComparing the USB captures from the hub, filtered to just the first few messages of the audio device yield the following exchanges. Again, first the broken case:
No. Time Source Destination Protocol Length Info
7 0.011098 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE 8 0.015218 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE 9 0.015272 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 10 0.017966 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION 11 0.018020 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 12 0.025965 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION 23 2.482241 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING 24 2.484931 1.2.0 host USB 98 GET DESCRIPTOR Response STRING 25 2.484939 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING 26 2.487916 1.2.0 host USB 84 GET DESCRIPTOR Response STRING 27 2.487941 host 1.2.0 USBAUDIO 65 SET CUR CX_CLOCK_SELECTOR_CONTROL request 2047 7.661154 1.2.0 host USBAUDIO 64 SET CUR CX_CLOCK_SELECTOR_CONTROL status 2048 7.661197 host 1.2.0 USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL request 4097 12.781553 1.2.0 host USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL[Malformed Packet]The first discrepancy can be found in the URB status field of packet no. 2047 (vs. 74 below) which is -2: No such file or directory (
-ENOENT). In the second (following) case where the sound device initializes okay, the URB status of the similar packet is Success (0) instead:No. Time Source Destination Protocol Length Info
7 0.011565 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE 8 0.015471 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE 9 0.015523 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 10 0.018477 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION 11 0.018537 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 12 0.026331 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION 23 4.168190 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING 24 4.171031 1.2.0 host USB 98 GET DESCRIPTOR Response STRING 25 4.171040 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING 26 4.174021 1.2.0 host USB 84 GET DESCRIPTOR Response STRING 27 4.174041 host 1.2.0 USBAUDIO 65 SET CUR CX_CLOCK_SELECTOR_CONTROL request 74 4.476033 1.2.0 host USBAUDIO 64 SET CUR CX_CLOCK_SELECTOR_CONTROL status 75 4.476050 host 1.2.0 USBAUDIO 64 GET CUR CX_CLOCK_SELECTOR_CONTROL request 76 4.479034 1.2.0 host USBAUDIO 65 GET CUR CX_CLOCK_SELECTOR_CONTROL response 77 4.479050 host 1.2.0 USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL request 202 4.582043 1.2.0 host USBAUDIO 66 GET RANGE CS_SAM_FREQ_CONTROL responsePlease ask, if I should provide any more information or how and to whom I can provide the full USB captures.
Hi Luiz,
Sorry for the late response.
Quoted 102 lines · Paul Menzel
External email : Please do not click links or open attachments until you have verified the sender or the content.
[Cc: +regressions, commit
ccfc8948d7e4was added in v6.11-rc1]Am 13.03.25 um 02:43 schrieb Benedikt Ziemons:
Quoted 93 lines · Luiz Augusto von Dentz
Quoted 92 lines · Geraldo Nascimento
Quoted 83 lines
Quoted 35 lines · Geoffrey D. Bennett
This series (choose 1 of the 2 patches) addresses an issue where the
MT7922 Bluetooth controller reset added in commitccfc8948d7e4
("Bluetooth: btusb: mediatek: reset the controller before downloading
the fw") is causing Focusrite USB audio devices to fail to initialise
when connected during boot on kernels 6.11 and newer.Reported by three users here:
https://urldefense.com/v3/__https://github.com/geoffreybennett/linux-fcp/issues/24__;!!CTRNKA9wMg0ARbw!hovfaKcQBzFKUu7X62SNU_berXiuXtSb2ndXGjAyf9WCptdz3D8jnprUqYc6b2fQE_3zrt88nGpKtJcYUCCcdA$Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar
change made in commita7208610761a("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if the
reset should be reverted for the MT7925 as well as the MT7922.Option 1: Revert the problematic reset behaviour entirely (MT7922 +
MT7925)Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller
before downloading the fw"Quoted 8 lines
[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before
downloading the fw--
2.45.0Hi Geoffrey,
I understand there's no apparent nexus of causality here.
However if three different users suddenly reported the same problem
and the fix fixes it, we should take the report seriously and back
down on the problematic commit until we figure for sure what the heck
is going on.My bet is this is bona fide bug in the USB subsystem, but either I'm
not looking hard enough or I'm looking into the wrong places, because
there's no way I can see in which way USB bluetooth driver from MediaTek could cause clock detection to fail.No point in pressing this harder, but yeah, take the reports more than
seriously, because if we don't understand in which way our own (Linux)
code could be causing this, at least we should take cautionary measures.In that sense, putting Takashi Iwai and Greg KH to Cc: in a modest but
sincere belief that this issue is more than real, even though it looks
like anticipated April 1st. Takashi can help with his expertise in
clock detection and Greg could help pinpoint if this is indeed madness
in the USB subsystem.Quoted 1 line
Hope you and them don't mind the escalation :)
Do you guys have any idea what could be possibly going on here? There
really seems something is not right if one driver affects the other,
especially if the devices are not actually related in any way.
I've discussed internally with Sean and Hao from MediaTek who submit the changes to Kernel. The original intention was to ensure controller's status is clean before Bluetooth driver doing firmware binary download when booting. However, it seems that the changes cause unexpected USB device enumeration problem at this timing with specific USB devices.
To avoid such problem, we also agree that reverting those changes would be a better choice. Thanks a lot.
Quoted 193 lines
Quoted 193 lines
I did the kernel bisection on this issue and tested Geoffrey's patches
since I own the affected combination of hardware and can reliably reproduce the issue.I was now able to capture both the sound device initialization failure
and the successful initialization from initramfs using the usbmon module. One difference that is immediately noticeable is the amount ofURB_CONTROLdata that is sent to the mediatek btusb device and a considerable slow-down of the boot process (around 16s difference).Following is some more information about the timing and other related
messages from the kernel message log. The first excerpt is from an unpatched kernel 6.14.0-rc5:[ 59.987242] lunar kernel: usb 1-2: new high-speed USB device number 2 using
xhci_hcd
[ 59.987975] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45
[ 59.988048] lunar kernel: usb 1-2: New USB device strings:
Mfr=1, Product=3, SerialNumber=2
[ 59.988121] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 59.988197] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 60.571193] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002)
[ 60.571314] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision:
79220010
[ 60.571501] lunar kernel: usbcore: registered new interface driver btusb
[ 60.571513] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version:0x8a108a10, Build Time:20241106163228a
[ 60.571627] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware
Version: ____000000, Build Time: 20241106163310
[ 60.616673] lunar kernel: Bluetooth: hci0: HW/SW Version:0x008a008a, Build Time: 20241106163512
[ 64.549017] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0
[ 70.700407] lunar kernel: usb 1-2:parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41)
[ 81.264360] lunar kernel: usb 1-2:parse_audio_format_rates_v2v3(): unable to retrieve number of sample rates (clock 41)
[ 81.264611] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to
https://urldefense.com/v3/__https://github.com/geoffreybennett/scarlett-gen2/issues__;!!CTRNKA9wMg0ARbw!hovfaKcQBzFKUu7X62SNU_berXiuXtSb2ndXGjAyf9WCptdz3D8jnprUqYc6b2fQE_3zrt88nGpKtJe1OM82_Q$
[ 81.264782] lunar kernel: usb 1-2: Error initialising Scarlett Gen 3 Mixer Driver: -71
[ 81.264903] lunar kernel: snd-usb-audio 1-2:1.0: probe with driver snd-usb-audio failed with error -71
[ 81.265040] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0
[ 81.400030] lunar kernel: Bluetooth: hci0: Device setup in 20435397 usecs
[ 81.400114] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 81.680052] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00
[ 81.680170] lunar kernel: Bluetooth: hci0: AOSP quality report is supported
[ 87.373353] lunar kernel: usbcore: registered new interface driver snd-usb-audioThe following second excerpt is from the same kernel version with Geoffrey's second patch applied:
[ 139.603887] lunar kernel: usb 1-2: new high-speed USB device number 2 using
xhci_hcd
[ 139.604524] lunar kernel: usb 1-2: New USB device found, idVendor=1235, idProduct=8212, bcdDevice= 6.45
[ 139.604598] lunar kernel: usb 1-2: New USB device strings:
Mfr=1, Product=3, SerialNumber=2
[ 139.604672] lunar kernel: usb 1-2: Product: Scarlett 4i4 USB [ 139.604744] lunar kernel: usb 1-2: Manufacturer: Focusrite [ 140.192488] lunar kernel: usbcore: registered new interface driver btusb
[ 140.192501] lunar kernel: mt7921e 0000:0a:00.0: enabling device (0000 -> 0002)
[ 140.192627] lunar kernel: Bluetooth: hci0: HW/SW Version:0x008a008a, Build Time: 20241106163512
[ 140.192639] lunar kernel: mt7921e 0000:0a:00.0: ASIC revision:
79220010
[ 140.192753] lunar kernel: mt7921e 0000:0a:00.0: HW/SW Version:0x8a108a10, Build Time:20241106163228a
[ 140.192866] lunar kernel: mt7921e 0000:0a:00.0: WM Firmware
Version: ____000000, Build Time: 20241106163310
[ 140.240011] lunar kernel: Bluetooth: hci0: Device setup in 188119 usecs
[ 140.240048] lunar kernel: Bluetooth: hci0: HCI Enhanced Setup Synchronous Connection command is advertised, but not supported. [ 140.513341] lunar kernel: Bluetooth: hci0: AOSP extensions version v1.00
[ 140.513379] lunar kernel: Bluetooth: hci0: AOSP quality report is supported
[ 144.115475] lunar kernel: mt7921e 0000:0a:00.0 wlp10s0: renamed from wlan0
[ 144.115747] lunar kernel: usb 1-2: Focusrite Scarlett Gen 3 Mixer Driver enabled (pid=0x8212); report any issues to
https://urldefense.com/v3/__https://github.com/geoffreybennett/scarlett-gen2/issues__;!!CTRNKA9wMg0ARbw!hovfaKcQBzFKUu7X62SNU_berXiuXtSb2ndXGjAyf9WCptdz3D8jnprUqYc6b2fQE_3zrt88nGpKtJe1OM82_Q$
[ 144.115857] lunar kernel: usb 1-2: Firmware version 1605 [ 144.115954] lunar kernel: usb 1-2: Quirk or no altset; falling back to MIDI 1.0
[ 147.843340] lunar kernel: usbcore: registered new interface driver snd-usb-audioComparing the USB captures from the hub, filtered to just the first few
messages of the audio device yield the following exchanges. Again, first the broken case:No. Time Source Destination Protocol Length Info
7 0.011098 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE
8 0.015218 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE
9 0.015272 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 10 0.017966 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION 11 0.018020 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 12 0.025965 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION 23 2.482241 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
24 2.484931 1.2.0 host USB 98 GET DESCRIPTOR Response STRING
25 2.484939 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
26 2.487916 1.2.0 host USB 84 GET DESCRIPTOR Response STRING
27 2.487941 host 1.2.0 USBAUDIO 65 SET CURCX_CLOCK_SELECTOR_CONTROLrequest 2047 7.661154 1.2.0 host USBAUDIO 64 SET CURCX_CLOCK_SELECTOR_CONTROLstatus 2048 7.661197 host 1.2.0 USBAUDIO 64 GET RANGECS_SAM_FREQ_CONTROLrequest 4097 12.781553 1.2.0 host USBAUDIO 64 GET RANGE CS_SAM_FREQ_CONTROL[Malformed Packet]The first discrepancy can be found in the URB status field of packet
no. 2047 (vs. 74 below) which is -2: No such file or directory (- ENOENT).
In the second (following) case where the sound device initializes okay, the URB status of the similar packet is Success (0) instead:No. Time Source Destination Protocol Length Info
7 0.011565 host 1.2.0 USB 64 GET DESCRIPTOR Request DEVICE
8 0.015471 1.2.0 host USB 82 GET DESCRIPTOR Response DEVICE
9 0.015523 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 10 0.018477 1.2.0 host USB 73 GET DESCRIPTOR Response CONFIGURATION 11 0.018537 host 1.2.0 USB 64 GET DESCRIPTOR Request CONFIGURATION 12 0.026331 1.2.0 host USB 425 GET DESCRIPTOR Response CONFIGURATION 23 4.168190 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
24 4.171031 1.2.0 host USB 98 GET DESCRIPTOR Response STRING
25 4.171040 host 1.2.0 USB 64 GET DESCRIPTOR Request STRING
26 4.174021 1.2.0 host USB 84 GET DESCRIPTOR Response STRING
27 4.174041 host 1.2.0 USBAUDIO 65 SET CURCX_CLOCK_SELECTOR_CONTROLrequest 74 4.476033 1.2.0 host USBAUDIO 64 SET CURCX_CLOCK_SELECTOR_CONTROLstatus 75 4.476050 host 1.2.0 USBAUDIO 64 GET CURCX_CLOCK_SELECTOR_CONTROLrequest 76 4.479034 1.2.0 host USBAUDIO 65 GET CURCX_CLOCK_SELECTOR_CONTROLresponse 77 4.479050 host 1.2.0 USBAUDIO 64 GET RANGECS_SAM_FREQ_CONTROLrequest 202 4.582043 1.2.0 host USBAUDIO 66 GET RANGECS_SAM_FREQ_CONTROLresponsePlease ask, if I should provide any more information or how and to whom
I can provide the full USB captures.
Quoted 115 lines · Chris Lu (陸稚泓)
Hi Luiz,
Sorry for the late response.
Quoted 102 lines · Paul Menzel
External email : Please do not click links or open attachments until you have verified the sender or the content.
[Cc: +regressions, commit
ccfc8948d7e4was added in v6.11-rc1]Am 13.03.25 um 02:43 schrieb Benedikt Ziemons:
Quoted 93 lines · Luiz Augusto von Dentz
Quoted 92 lines · Geraldo Nascimento
Quoted 83 lines
Quoted 35 lines · Geoffrey D. Bennett
This series (choose 1 of the 2 patches) addresses an issue where the
MT7922 Bluetooth controller reset added in commitccfc8948d7e4
("Bluetooth: btusb: mediatek: reset the controller before downloading
the fw") is causing Focusrite USB audio devices to fail to initialise
when connected during boot on kernels 6.11 and newer.Reported by three users here:
https://urldefense.com/v3/__https://github.com/geoffreybennett/linux-fcp/issues/24__;!!CTRNKA9wMg0ARbw!hovfaKcQBzFKUu7X62SNU_berXiuXtSb2ndXGjAyf9WCptdz3D8jnprUqYc6b2fQE_3zrt88nGpKtJcYUCCcdA$Two users confirmed they have an MT7922.
I'm providing two possible solutions as I note there was a similar
change made in commita7208610761a("Bluetooth: btmtk: Remove resetting mt7921 before downloading the fw"), so I'm not sure if the
reset should be reverted for the MT7925 as well as the MT7922.Option 1: Revert the problematic reset behaviour entirely (MT7922 +
MT7925)Option 2: Remove the reset only for the MT7922
Geoffrey D. Bennett (2):
[PATCH 1/2] Revert "Bluetooth: btusb: mediatek: reset the controller
before downloading the fw"Quoted 8 lines
[PATCH 2/2] Bluetooth: btmtk: Remove resetting mt7922 before
downloading the fw--
2.45.0Hi Geoffrey,
I understand there's no apparent nexus of causality here.
However if three different users suddenly reported the same problem
and the fix fixes it, we should take the report seriously and back
down on the problematic commit until we figure for sure what the heck
is going on.My bet is this is bona fide bug in the USB subsystem, but either I'm
not looking hard enough or I'm looking into the wrong places, because
there's no way I can see in which way USB bluetooth driver from MediaTek could cause clock detection to fail.No point in pressing this harder, but yeah, take the reports more than
seriously, because if we don't understand in which way our own (Linux)
code could be causing this, at least we should take cautionary measures.In that sense, putting Takashi Iwai and Greg KH to Cc: in a modest but
sincere belief that this issue is more than real, even though it looks
like anticipated April 1st. Takashi can help with his expertise in
clock detection and Greg could help pinpoint if this is indeed madness
in the USB subsystem.Quoted 1 line
Hope you and them don't mind the escalation :)
Do you guys have any idea what could be possibly going on here? There
really seems something is not right if one driver affects the other,
especially if the devices are not actually related in any way.I've discussed internally with Sean and Hao from MediaTek who submit the changes to Kernel. The original intention was to ensure controller's status is clean before Bluetooth driver doing firmware binary download when booting. However, it seems that the changes cause unexpected USB device enumeration problem at this timing with specific USB devices.
Quoted 2 lines
To avoid such problem, we also agree that reverting those changes would be a better choice. Thanks a lot.
Great, can someone please submit the reverts?
thanks,
greg k-h