regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Bruno Pitrus]

I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a bluetooth adapter of ID 8087:0026 (Intel AX201).
I am using a Logitech M575 trackball mouse which only supports Bluetooth LE.

With kernel 6.4.2 or newer, scanning for devices — using either the KDE gui or bluetoothctl — most of the time returns only “classic bluetooth” devices like computers and phones, and not the mouse.
If the mouse was paired before, it works correctly however.

With previous kernel versions (6.3.9 and before), pairing the mouse worked correctly.

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Paul Menzel]

[Cc: +regressions@]

#regzbot introduced v6.3.9..v6.4.2

Dear Bruno,


Thank you for your report.

Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> Logitech M575 trackball mouse which only supports Bluetooth LE.
> 
> With kernel 6.4.2 or newer, scanning for devices — using either the
> KDE gui or bluetoothctl — most of the time returns only “classic
> bluetooth” devices like computers and phones, and not the mouse. If
> the mouse was paired before, it works correctly however.
> 
> With previous kernel versions (6.3.9 and before), pairing the mouse
> worked correctly.

Are you able to bisect the regression? Please also attach the Linux 
messages. (You can also create a ticket in the Linux kernel Bugzilla [1] 
and attach everything there.)


Kind regards,

Paul


[1]: https://bugzilla.kernel.org/

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Bruno Pitrus]

Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> [Cc: +regressions@]
> 
> #regzbot introduced v6.3.9..v6.4.2
> 
> Dear Bruno,
> 
> 
> Thank you for your report.
> 
> Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > 
> > With kernel 6.4.2 or newer, scanning for devices — using either the
> > KDE gui or bluetoothctl — most of the time returns only “classic
> > bluetooth” devices like computers and phones, and not the mouse. If
> > the mouse was paired before, it works correctly however.
> > 
> > With previous kernel versions (6.3.9 and before), pairing the mouse
> > worked correctly.
> 
> Are you able to bisect the regression? Please also attach the Linux 
> messages. (You can also create a ticket in the Linux kernel Bugzilla [1] 
> and attach everything there.)
> 
> 
> Kind regards,
> 
> Paul
> 
> 
> [1]: https://bugzilla.kernel.org/
> 

The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :

> Bluetooth: Enable all supported LE PHY by default
> This enables 2M and Coded PHY by default if they are marked as supported
> in the LE features bits.
> 
> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Luiz Augusto von Dentz]

Hi Bruno,

On Fri, Aug 11, 2023 at 4:25 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
>
> Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> > [Cc: +regressions@]
> >
> > #regzbot introduced v6.3.9..v6.4.2
> >
> > Dear Bruno,
> >
> >
> > Thank you for your report.
> >
> > Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > >
> > > With kernel 6.4.2 or newer, scanning for devices — using either the
> > > KDE gui or bluetoothctl — most of the time returns only “classic
> > > bluetooth” devices like computers and phones, and not the mouse. If
> > > the mouse was paired before, it works correctly however.
> > >
> > > With previous kernel versions (6.3.9 and before), pairing the mouse
> > > worked correctly.
> >
> > Are you able to bisect the regression? Please also attach the Linux
> > messages. (You can also create a ticket in the Linux kernel Bugzilla [1]
> > and attach everything there.)
> >
> >
> > Kind regards,
> >
> > Paul
> >
> >
> > [1]: https://bugzilla.kernel.org/
> >
>
> The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :
>
> > Bluetooth: Enable all supported LE PHY by default
> > This enables 2M and Coded PHY by default if they are marked as supported
> > in the LE features bits.
> >
> > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>

Do you have the btmon/HCI traces showing this, we also need the
firmware version as well since I suspect this is due to the controller
setting the PHYs as supported but being unable to to use them
properly.

-- 
Luiz Augusto von Dentz

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Bruno Pitrus]

Dnia piątek, 11 sierpnia 2023 19:11:50 CEST Luiz Augusto von Dentz pisze:
> Hi Bruno,
> 
> On Fri, Aug 11, 2023 at 4:25 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> >
> > Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> > > [Cc: +regressions@]
> > >
> > > #regzbot introduced v6.3.9..v6.4.2
> > >
> > > Dear Bruno,
> > >
> > >
> > > Thank you for your report.
> > >
> > > Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > > > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > > > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > > > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > > >
> > > > With kernel 6.4.2 or newer, scanning for devices — using either the
> > > > KDE gui or bluetoothctl — most of the time returns only “classic
> > > > bluetooth” devices like computers and phones, and not the mouse. If
> > > > the mouse was paired before, it works correctly however.
> > > >
> > > > With previous kernel versions (6.3.9 and before), pairing the mouse
> > > > worked correctly.
> > >
> > > Are you able to bisect the regression? Please also attach the Linux
> > > messages. (You can also create a ticket in the Linux kernel Bugzilla [1]
> > > and attach everything there.)
> > >
> > >
> > > Kind regards,
> > >
> > > Paul
> > >
> > >
> > > [1]: https://bugzilla.kernel.org/
> > >
> >
> > The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :
> >
> > > Bluetooth: Enable all supported LE PHY by default
> > > This enables 2M and Coded PHY by default if they are marked as supported
> > > in the LE features bits.
> > >
> > > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> 
> Do you have the btmon/HCI traces showing this, we also need the
> firmware version as well since I suspect this is due to the controller
> setting the PHYs as supported but being unable to to use them
> properly.
> 
> 
hciconfig says the below: (I assume “Revision” is the firmware version)

hci0:   Type: Primary  Bus: USB
        BD Address: F0:9E:4A:E6:62:15  ACL MTU: 1021:4  SCO MTU: 96:6
        UP RUNNING 
        RX bytes:150050 acl:8114 sco:0 events:270 errors:0
        TX bytes:5970 acl:69 sco:0 commands:173 errors:0
        Features: 0xbf 0xfe 0x0f 0xfe 0xdb 0xff 0x7b 0x87
        Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3 
        Link policy: RSWITCH SNIFF 
        Link mode: PERIPHERAL ACCEPT 
        Name: '[REDACTED]'
        Class: 0x6c010c
        Service Classes: Rendering, Capturing, Audio, Telephony
        Device Class: Computer, Laptop
        HCI Version: 5.2 (0xb)  Revision: 0x20ce
        LMP Version: 5.2 (0xb)  Subversion: 0x20ce
        Manufacturer: Intel Corp. (2)

I'm unsure how to attach the btmon output as it is a binary file that contains potentially private information (names of nearby devices)

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Luiz Augusto von Dentz]

Hi Bruno,

On Mon, Aug 14, 2023 at 3:00 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
>
> Dnia piątek, 11 sierpnia 2023 19:11:50 CEST Luiz Augusto von Dentz pisze:
> > Hi Bruno,
> >
> > On Fri, Aug 11, 2023 at 4:25 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > >
> > > Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> > > > [Cc: +regressions@]
> > > >
> > > > #regzbot introduced v6.3.9..v6.4.2
> > > >
> > > > Dear Bruno,
> > > >
> > > >
> > > > Thank you for your report.
> > > >
> > > > Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > > > > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > > > > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > > > > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > > > >
> > > > > With kernel 6.4.2 or newer, scanning for devices — using either the
> > > > > KDE gui or bluetoothctl — most of the time returns only “classic
> > > > > bluetooth” devices like computers and phones, and not the mouse. If
> > > > > the mouse was paired before, it works correctly however.
> > > > >
> > > > > With previous kernel versions (6.3.9 and before), pairing the mouse
> > > > > worked correctly.
> > > >
> > > > Are you able to bisect the regression? Please also attach the Linux
> > > > messages. (You can also create a ticket in the Linux kernel Bugzilla [1]
> > > > and attach everything there.)
> > > >
> > > >
> > > > Kind regards,
> > > >
> > > > Paul
> > > >
> > > >
> > > > [1]: https://bugzilla.kernel.org/
> > > >
> > >
> > > The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :
> > >
> > > > Bluetooth: Enable all supported LE PHY by default
> > > > This enables 2M and Coded PHY by default if they are marked as supported
> > > > in the LE features bits.
> > > >
> > > > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> >
> > Do you have the btmon/HCI traces showing this, we also need the
> > firmware version as well since I suspect this is due to the controller
> > setting the PHYs as supported but being unable to to use them
> > properly.
> >
> >
> hciconfig says the below: (I assume “Revision” is the firmware version)

You can find the firmware information on dmesg, e.g:

[    7.982585] Bluetooth: hci1: Found Intel DDC parameters: intel/ibt-20-1-3.ddc
[    7.984113] Bluetooth: hci1: Applying Intel DDC parameters completed
[    7.985111] Bluetooth: hci1: Firmware revision 0.3 build 144 week 11 2023

> hci0:   Type: Primary  Bus: USB
>         BD Address: F0:9E:4A:E6:62:15  ACL MTU: 1021:4  SCO MTU: 96:6
>         UP RUNNING
>         RX bytes:150050 acl:8114 sco:0 events:270 errors:0
>         TX bytes:5970 acl:69 sco:0 commands:173 errors:0
>         Features: 0xbf 0xfe 0x0f 0xfe 0xdb 0xff 0x7b 0x87
>         Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
>         Link policy: RSWITCH SNIFF
>         Link mode: PERIPHERAL ACCEPT
>         Name: '[REDACTED]'
>         Class: 0x6c010c
>         Service Classes: Rendering, Capturing, Audio, Telephony
>         Device Class: Computer, Laptop
>         HCI Version: 5.2 (0xb)  Revision: 0x20ce
>         LMP Version: 5.2 (0xb)  Subversion: 0x20ce
>         Manufacturer: Intel Corp. (2)
>
> I'm unsure how to attach the btmon output as it is a binary file that contains potentially private information (names of nearby devices)

Well you can collect it manually and redacted the addresses, etc, up
to you, I'm usually not that paranoid of sharing this kind of
information if you scan for a short time, in the other hand I usually
use gist.github.com to do that so I can delete it later if needed.

>


-- 
Luiz Augusto von Dentz

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Luiz Augusto von Dentz]

Hi Bruno,

On Mon, Aug 14, 2023 at 1:21 PM Luiz Augusto von Dentz
<luiz.dentz@gmail.com> wrote:
>
> Hi Bruno,
>
> On Mon, Aug 14, 2023 at 3:00 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> >
> > Dnia piątek, 11 sierpnia 2023 19:11:50 CEST Luiz Augusto von Dentz pisze:
> > > Hi Bruno,
> > >
> > > On Fri, Aug 11, 2023 at 4:25 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > > >
> > > > Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> > > > > [Cc: +regressions@]
> > > > >
> > > > > #regzbot introduced v6.3.9..v6.4.2
> > > > >
> > > > > Dear Bruno,
> > > > >
> > > > >
> > > > > Thank you for your report.
> > > > >
> > > > > Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > > > > > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > > > > > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > > > > > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > > > > >
> > > > > > With kernel 6.4.2 or newer, scanning for devices — using either the
> > > > > > KDE gui or bluetoothctl — most of the time returns only “classic
> > > > > > bluetooth” devices like computers and phones, and not the mouse. If
> > > > > > the mouse was paired before, it works correctly however.
> > > > > >
> > > > > > With previous kernel versions (6.3.9 and before), pairing the mouse
> > > > > > worked correctly.
> > > > >
> > > > > Are you able to bisect the regression? Please also attach the Linux
> > > > > messages. (You can also create a ticket in the Linux kernel Bugzilla [1]
> > > > > and attach everything there.)
> > > > >
> > > > >
> > > > > Kind regards,
> > > > >
> > > > > Paul
> > > > >
> > > > >
> > > > > [1]: https://bugzilla.kernel.org/
> > > > >
> > > >
> > > > The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :
> > > >
> > > > > Bluetooth: Enable all supported LE PHY by default
> > > > > This enables 2M and Coded PHY by default if they are marked as supported
> > > > > in the LE features bits.
> > > > >
> > > > > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > >
> > > Do you have the btmon/HCI traces showing this, we also need the
> > > firmware version as well since I suspect this is due to the controller
> > > setting the PHYs as supported but being unable to to use them
> > > properly.
> > >
> > >
> > hciconfig says the below: (I assume “Revision” is the firmware version)
>
> You can find the firmware information on dmesg, e.g:
>
> [    7.982585] Bluetooth: hci1: Found Intel DDC parameters: intel/ibt-20-1-3.ddc
> [    7.984113] Bluetooth: hci1: Applying Intel DDC parameters completed
> [    7.985111] Bluetooth: hci1: Firmware revision 0.3 build 144 week 11 2023
>
> > hci0:   Type: Primary  Bus: USB
> >         BD Address: F0:9E:4A:E6:62:15  ACL MTU: 1021:4  SCO MTU: 96:6
> >         UP RUNNING
> >         RX bytes:150050 acl:8114 sco:0 events:270 errors:0
> >         TX bytes:5970 acl:69 sco:0 commands:173 errors:0
> >         Features: 0xbf 0xfe 0x0f 0xfe 0xdb 0xff 0x7b 0x87
> >         Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
> >         Link policy: RSWITCH SNIFF
> >         Link mode: PERIPHERAL ACCEPT
> >         Name: '[REDACTED]'
> >         Class: 0x6c010c
> >         Service Classes: Rendering, Capturing, Audio, Telephony
> >         Device Class: Computer, Laptop
> >         HCI Version: 5.2 (0xb)  Revision: 0x20ce
> >         LMP Version: 5.2 (0xb)  Subversion: 0x20ce
> >         Manufacturer: Intel Corp. (2)
> >
> > I'm unsure how to attach the btmon output as it is a binary file that contains potentially private information (names of nearby devices)
>
> Well you can collect it manually and redacted the addresses, etc, up
> to you, I'm usually not that paranoid of sharing this kind of
> information if you scan for a short time, in the other hand I usually
> use gist.github.com to do that so I can delete it later if needed.

Seems to be working fine here (with AX201 and AX210):

https://gist.github.com/Vudentz/d80ab63f4f62f9f216796edb4838c855

> >
>
>
> --
> Luiz Augusto von Dentz



-- 
Luiz Augusto von Dentz

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Bruno Pitrus]

[Resending this for posterity so that it ends up in the mailing list archives]
Dnia wtorek, 15 sierpnia 2023 01:38:10 CEST piszesz:
> Hi Bruno,
> 
> On Mon, Aug 14, 2023 at 1:21 PM Luiz Augusto von Dentz
> <luiz.dentz@gmail.com> wrote:
> >
> > Hi Bruno,
> >
> > On Mon, Aug 14, 2023 at 3:00 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > >
> > > Dnia piątek, 11 sierpnia 2023 19:11:50 CEST Luiz Augusto von Dentz pisze:
> > > > Hi Bruno,
> > > >
> > > > On Fri, Aug 11, 2023 at 4:25 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > > > >
> > > > > Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> > > > > > [Cc: +regressions@]
> > > > > >
> > > > > > #regzbot introduced v6.3.9..v6.4.2
> > > > > >
> > > > > > Dear Bruno,
> > > > > >
> > > > > >
> > > > > > Thank you for your report.
> > > > > >
> > > > > > Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > > > > > > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > > > > > > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > > > > > > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > > > > > >
> > > > > > > With kernel 6.4.2 or newer, scanning for devices — using either the
> > > > > > > KDE gui or bluetoothctl — most of the time returns only “classic
> > > > > > > bluetooth” devices like computers and phones, and not the mouse. If
> > > > > > > the mouse was paired before, it works correctly however.
> > > > > > >
> > > > > > > With previous kernel versions (6.3.9 and before), pairing the mouse
> > > > > > > worked correctly.
> > > > > >
> > > > > > Are you able to bisect the regression? Please also attach the Linux
> > > > > > messages. (You can also create a ticket in the Linux kernel Bugzilla [1]
> > > > > > and attach everything there.)
> > > > > >
> > > > > >
> > > > > > Kind regards,
> > > > > >
> > > > > > Paul
> > > > > >
> > > > > >
> > > > > > [1]: https://bugzilla.kernel.org/
> > > > > >
> > > > >
> > > > > The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :
> > > > >
> > > > > > Bluetooth: Enable all supported LE PHY by default
> > > > > > This enables 2M and Coded PHY by default if they are marked as supported
> > > > > > in the LE features bits.
> > > > > >
> > > > > > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > > >
> > > > Do you have the btmon/HCI traces showing this, we also need the
> > > > firmware version as well since I suspect this is due to the controller
> > > > setting the PHYs as supported but being unable to to use them
> > > > properly.
> > > >
> > > >
> > > hciconfig says the below: (I assume “Revision” is the firmware version)
> >
> > You can find the firmware information on dmesg, e.g:
> >
> > [    7.982585] Bluetooth: hci1: Found Intel DDC parameters: intel/ibt-20-1-3.ddc
> > [    7.984113] Bluetooth: hci1: Applying Intel DDC parameters completed
> > [    7.985111] Bluetooth: hci1: Firmware revision 0.3 build 144 week 11 2023
> >
> > > hci0:   Type: Primary  Bus: USB
> > >         BD Address: F0:9E:4A:E6:62:15  ACL MTU: 1021:4  SCO MTU: 96:6
> > >         UP RUNNING
> > >         RX bytes:150050 acl:8114 sco:0 events:270 errors:0
> > >         TX bytes:5970 acl:69 sco:0 commands:173 errors:0
> > >         Features: 0xbf 0xfe 0x0f 0xfe 0xdb 0xff 0x7b 0x87
> > >         Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
> > >         Link policy: RSWITCH SNIFF
> > >         Link mode: PERIPHERAL ACCEPT
> > >         Name: '[REDACTED]'
> > >         Class: 0x6c010c
> > >         Service Classes: Rendering, Capturing, Audio, Telephony
> > >         Device Class: Computer, Laptop
> > >         HCI Version: 5.2 (0xb)  Revision: 0x20ce
> > >         LMP Version: 5.2 (0xb)  Subversion: 0x20ce
> > >         Manufacturer: Intel Corp. (2)

The relevant dmesg lines are:

[   10.355916] Bluetooth: hci0: Found device firmware: intel/ibt-19-0-4.sfi
[   10.356067] Bluetooth: hci0: Boot Address: 0x24800
[   10.356070] Bluetooth: hci0: Firmware Version: 206-22.23
[   10.356073] Bluetooth: hci0: Firmware already loaded


> > >
> > > I'm unsure how to attach the btmon output as it is a binary file that contains potentially private information (names of nearby devices)
> >
> > Well you can collect it manually and redacted the addresses, etc, up
> > to you, I'm usually not that paranoid of sharing this kind of
> > information if you scan for a short time, in the other hand I usually
> > use gist.github.com to do that so I can delete it later if needed.
> 
> Seems to be working fine here (with AX201 and AX210):
> 
> https://gist.github.com/Vudentz/d80ab63f4f62f9f216796edb4838c855
> 
> > >
> >
> >
> > --
> > Luiz Augusto von Dentz
> 
> 
> 
> 

Here is a btmon log of me trying to search for devices in a quiet location (the only device seen is the mouse — it shows up in the btmon logs but NOT in the KDE pairing window):
https://gist.github.com/brjsp/f6897ad1431a06820ecdd3773dd4a857

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Luiz Augusto von Dentz]

Hi Bruno,

On Mon, Aug 14, 2023 at 4:38 PM Luiz Augusto von Dentz
<luiz.dentz@gmail.com> wrote:
>
> Hi Bruno,
>
> On Mon, Aug 14, 2023 at 1:21 PM Luiz Augusto von Dentz
> <luiz.dentz@gmail.com> wrote:
> >
> > Hi Bruno,
> >
> > On Mon, Aug 14, 2023 at 3:00 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > >
> > > Dnia piątek, 11 sierpnia 2023 19:11:50 CEST Luiz Augusto von Dentz pisze:
> > > > Hi Bruno,
> > > >
> > > > On Fri, Aug 11, 2023 at 4:25 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > > > >
> > > > > Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> > > > > > [Cc: +regressions@]
> > > > > >
> > > > > > #regzbot introduced v6.3.9..v6.4.2
> > > > > >
> > > > > > Dear Bruno,
> > > > > >
> > > > > >
> > > > > > Thank you for your report.
> > > > > >
> > > > > > Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > > > > > > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > > > > > > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > > > > > > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > > > > > >
> > > > > > > With kernel 6.4.2 or newer, scanning for devices — using either the
> > > > > > > KDE gui or bluetoothctl — most of the time returns only “classic
> > > > > > > bluetooth” devices like computers and phones, and not the mouse. If
> > > > > > > the mouse was paired before, it works correctly however.
> > > > > > >
> > > > > > > With previous kernel versions (6.3.9 and before), pairing the mouse
> > > > > > > worked correctly.
> > > > > >
> > > > > > Are you able to bisect the regression? Please also attach the Linux
> > > > > > messages. (You can also create a ticket in the Linux kernel Bugzilla [1]
> > > > > > and attach everything there.)
> > > > > >
> > > > > >
> > > > > > Kind regards,
> > > > > >
> > > > > > Paul
> > > > > >
> > > > > >
> > > > > > [1]: https://bugzilla.kernel.org/
> > > > > >
> > > > >
> > > > > The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :
> > > > >
> > > > > > Bluetooth: Enable all supported LE PHY by default
> > > > > > This enables 2M and Coded PHY by default if they are marked as supported
> > > > > > in the LE features bits.
> > > > > >
> > > > > > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > > >
> > > > Do you have the btmon/HCI traces showing this, we also need the
> > > > firmware version as well since I suspect this is due to the controller
> > > > setting the PHYs as supported but being unable to to use them
> > > > properly.
> > > >
> > > >
> > > hciconfig says the below: (I assume “Revision” is the firmware version)
> >
> > You can find the firmware information on dmesg, e.g:
> >
> > [    7.982585] Bluetooth: hci1: Found Intel DDC parameters: intel/ibt-20-1-3.ddc
> > [    7.984113] Bluetooth: hci1: Applying Intel DDC parameters completed
> > [    7.985111] Bluetooth: hci1: Firmware revision 0.3 build 144 week 11 2023
> >
> > > hci0:   Type: Primary  Bus: USB
> > >         BD Address: F0:9E:4A:E6:62:15  ACL MTU: 1021:4  SCO MTU: 96:6
> > >         UP RUNNING
> > >         RX bytes:150050 acl:8114 sco:0 events:270 errors:0
> > >         TX bytes:5970 acl:69 sco:0 commands:173 errors:0
> > >         Features: 0xbf 0xfe 0x0f 0xfe 0xdb 0xff 0x7b 0x87
> > >         Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
> > >         Link policy: RSWITCH SNIFF
> > >         Link mode: PERIPHERAL ACCEPT
> > >         Name: '[REDACTED]'
> > >         Class: 0x6c010c
> > >         Service Classes: Rendering, Capturing, Audio, Telephony
> > >         Device Class: Computer, Laptop
> > >         HCI Version: 5.2 (0xb)  Revision: 0x20ce
> > >         LMP Version: 5.2 (0xb)  Subversion: 0x20ce
> > >         Manufacturer: Intel Corp. (2)
> > >
> > > I'm unsure how to attach the btmon output as it is a binary file that contains potentially private information (names of nearby devices)
> >
> > Well you can collect it manually and redacted the addresses, etc, up
> > to you, I'm usually not that paranoid of sharing this kind of
> > information if you scan for a short time, in the other hand I usually
> > use gist.github.com to do that so I can delete it later if needed.
>
> Seems to be working fine here (with AX201 and AX210):
>
> https://gist.github.com/Vudentz/d80ab63f4f62f9f216796edb4838c855

Can you try with the following patch:

https://patchwork.kernel.org/project/bluetooth/patch/20230821181848.3553863-1-luiz.dentz@gmail.com/

Looks like we were not storing the whole advertisement sometimes.

-- 
Luiz Augusto von Dentz

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Bruno Pitrus]

Dnia poniedziałek, 21 sierpnia 2023 22:08:45 CEST Luiz Augusto von Dentz pisze:
> Hi Bruno,
> 
> On Mon, Aug 14, 2023 at 4:38 PM Luiz Augusto von Dentz
> <luiz.dentz@gmail.com> wrote:
> >
> > Hi Bruno,
> >
> > On Mon, Aug 14, 2023 at 1:21 PM Luiz Augusto von Dentz
> > <luiz.dentz@gmail.com> wrote:
> > >
> > > Hi Bruno,
> > >
> > > On Mon, Aug 14, 2023 at 3:00 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > > >
> > > > Dnia piątek, 11 sierpnia 2023 19:11:50 CEST Luiz Augusto von Dentz pisze:
> > > > > Hi Bruno,
> > > > >
> > > > > On Fri, Aug 11, 2023 at 4:25 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > > > > >
> > > > > > Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> > > > > > > [Cc: +regressions@]
> > > > > > >
> > > > > > > #regzbot introduced v6.3.9..v6.4.2
> > > > > > >
> > > > > > > Dear Bruno,
> > > > > > >
> > > > > > >
> > > > > > > Thank you for your report.
> > > > > > >
> > > > > > > Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > > > > > > > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > > > > > > > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > > > > > > > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > > > > > > >
> > > > > > > > With kernel 6.4.2 or newer, scanning for devices — using either the
> > > > > > > > KDE gui or bluetoothctl — most of the time returns only “classic
> > > > > > > > bluetooth” devices like computers and phones, and not the mouse. If
> > > > > > > > the mouse was paired before, it works correctly however.
> > > > > > > >
> > > > > > > > With previous kernel versions (6.3.9 and before), pairing the mouse
> > > > > > > > worked correctly.
> > > > > > >
> > > > > > > Are you able to bisect the regression? Please also attach the Linux
> > > > > > > messages. (You can also create a ticket in the Linux kernel Bugzilla [1]
> > > > > > > and attach everything there.)
> > > > > > >
> > > > > > >
> > > > > > > Kind regards,
> > > > > > >
> > > > > > > Paul
> > > > > > >
> > > > > > >
> > > > > > > [1]: https://bugzilla.kernel.org/
> > > > > > >
> > > > > >
> > > > > > The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :
> > > > > >
> > > > > > > Bluetooth: Enable all supported LE PHY by default
> > > > > > > This enables 2M and Coded PHY by default if they are marked as supported
> > > > > > > in the LE features bits.
> > > > > > >
> > > > > > > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > > > >
> > > > > Do you have the btmon/HCI traces showing this, we also need the
> > > > > firmware version as well since I suspect this is due to the controller
> > > > > setting the PHYs as supported but being unable to to use them
> > > > > properly.
> > > > >
> > > > >
> > > > hciconfig says the below: (I assume “Revision” is the firmware version)
> > >
> > > You can find the firmware information on dmesg, e.g:
> > >
> > > [    7.982585] Bluetooth: hci1: Found Intel DDC parameters: intel/ibt-20-1-3.ddc
> > > [    7.984113] Bluetooth: hci1: Applying Intel DDC parameters completed
> > > [    7.985111] Bluetooth: hci1: Firmware revision 0.3 build 144 week 11 2023
> > >
> > > > hci0:   Type: Primary  Bus: USB
> > > >         BD Address: F0:9E:4A:E6:62:15  ACL MTU: 1021:4  SCO MTU: 96:6
> > > >         UP RUNNING
> > > >         RX bytes:150050 acl:8114 sco:0 events:270 errors:0
> > > >         TX bytes:5970 acl:69 sco:0 commands:173 errors:0
> > > >         Features: 0xbf 0xfe 0x0f 0xfe 0xdb 0xff 0x7b 0x87
> > > >         Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
> > > >         Link policy: RSWITCH SNIFF
> > > >         Link mode: PERIPHERAL ACCEPT
> > > >         Name: '[REDACTED]'
> > > >         Class: 0x6c010c
> > > >         Service Classes: Rendering, Capturing, Audio, Telephony
> > > >         Device Class: Computer, Laptop
> > > >         HCI Version: 5.2 (0xb)  Revision: 0x20ce
> > > >         LMP Version: 5.2 (0xb)  Subversion: 0x20ce
> > > >         Manufacturer: Intel Corp. (2)
> > > >
> > > > I'm unsure how to attach the btmon output as it is a binary file that contains potentially private information (names of nearby devices)
> > >
> > > Well you can collect it manually and redacted the addresses, etc, up
> > > to you, I'm usually not that paranoid of sharing this kind of
> > > information if you scan for a short time, in the other hand I usually
> > > use gist.github.com to do that so I can delete it later if needed.
> >
> > Seems to be working fine here (with AX201 and AX210):
> >
> > https://gist.github.com/Vudentz/d80ab63f4f62f9f216796edb4838c855
> 
> Can you try with the following patch:
> 
> https://patchwork.kernel.org/project/bluetooth/patch/20230821181848.3553863-1-luiz.dentz@gmail.com/
> 
> Looks like we were not storing the whole advertisement sometimes.
> 
> 

I applied your patch, along with https://www.spinics.net/lists/linux-bluetooth/msg105443.html since that defines the needed max_adv_len macro, on top of the regression-introducting commit. Unfortunately, it did not help — LE devices are still not detected (either at all or after one-two successful pairings).

Re: regression: kernel 6.4.x breaks scanning/pairing (but not use) of LE devices

[Luiz Augusto von Dentz]

Hi Bruno,

On Tue, Aug 22, 2023 at 7:29 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
>
> Dnia poniedziałek, 21 sierpnia 2023 22:08:45 CEST Luiz Augusto von Dentz pisze:
> > Hi Bruno,
> >
> > On Mon, Aug 14, 2023 at 4:38 PM Luiz Augusto von Dentz
> > <luiz.dentz@gmail.com> wrote:
> > >
> > > Hi Bruno,
> > >
> > > On Mon, Aug 14, 2023 at 1:21 PM Luiz Augusto von Dentz
> > > <luiz.dentz@gmail.com> wrote:
> > > >
> > > > Hi Bruno,
> > > >
> > > > On Mon, Aug 14, 2023 at 3:00 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > > > >
> > > > > Dnia piątek, 11 sierpnia 2023 19:11:50 CEST Luiz Augusto von Dentz pisze:
> > > > > > Hi Bruno,
> > > > > >
> > > > > > On Fri, Aug 11, 2023 at 4:25 AM Bruno Pitrus <brunopitrus@hotmail.com> wrote:
> > > > > > >
> > > > > > > Dnia sobota, 5 sierpnia 2023 09:31:08 CEST Paul Menzel pisze:
> > > > > > > > [Cc: +regressions@]
> > > > > > > >
> > > > > > > > #regzbot introduced v6.3.9..v6.4.2
> > > > > > > >
> > > > > > > > Dear Bruno,
> > > > > > > >
> > > > > > > >
> > > > > > > > Thank you for your report.
> > > > > > > >
> > > > > > > > Am 04.08.23 um 20:55 schrieb Bruno Pitrus:
> > > > > > > > > I am using openSUSE Tumbleweed on a Dell Precision 3551 which has a
> > > > > > > > > bluetooth adapter of ID 8087:0026 (Intel AX201). I am using a
> > > > > > > > > Logitech M575 trackball mouse which only supports Bluetooth LE.
> > > > > > > > >
> > > > > > > > > With kernel 6.4.2 or newer, scanning for devices — using either the
> > > > > > > > > KDE gui or bluetoothctl — most of the time returns only “classic
> > > > > > > > > bluetooth” devices like computers and phones, and not the mouse. If
> > > > > > > > > the mouse was paired before, it works correctly however.
> > > > > > > > >
> > > > > > > > > With previous kernel versions (6.3.9 and before), pairing the mouse
> > > > > > > > > worked correctly.
> > > > > > > >
> > > > > > > > Are you able to bisect the regression? Please also attach the Linux
> > > > > > > > messages. (You can also create a ticket in the Linux kernel Bugzilla [1]
> > > > > > > > and attach everything there.)
> > > > > > > >
> > > > > > > >
> > > > > > > > Kind regards,
> > > > > > > >
> > > > > > > > Paul
> > > > > > > >
> > > > > > > >
> > > > > > > > [1]: https://bugzilla.kernel.org/
> > > > > > > >
> > > > > > >
> > > > > > > The offending commit is https://github.com/SUSE/kernel/commit/288c90224eec55d13e786844b7954ef060752089 :
> > > > > > >
> > > > > > > > Bluetooth: Enable all supported LE PHY by default
> > > > > > > > This enables 2M and Coded PHY by default if they are marked as supported
> > > > > > > > in the LE features bits.
> > > > > > > >
> > > > > > > > Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
> > > > > >
> > > > > > Do you have the btmon/HCI traces showing this, we also need the
> > > > > > firmware version as well since I suspect this is due to the controller
> > > > > > setting the PHYs as supported but being unable to to use them
> > > > > > properly.
> > > > > >
> > > > > >
> > > > > hciconfig says the below: (I assume “Revision” is the firmware version)
> > > >
> > > > You can find the firmware information on dmesg, e.g:
> > > >
> > > > [    7.982585] Bluetooth: hci1: Found Intel DDC parameters: intel/ibt-20-1-3.ddc
> > > > [    7.984113] Bluetooth: hci1: Applying Intel DDC parameters completed
> > > > [    7.985111] Bluetooth: hci1: Firmware revision 0.3 build 144 week 11 2023
> > > >
> > > > > hci0:   Type: Primary  Bus: USB
> > > > >         BD Address: F0:9E:4A:E6:62:15  ACL MTU: 1021:4  SCO MTU: 96:6
> > > > >         UP RUNNING
> > > > >         RX bytes:150050 acl:8114 sco:0 events:270 errors:0
> > > > >         TX bytes:5970 acl:69 sco:0 commands:173 errors:0
> > > > >         Features: 0xbf 0xfe 0x0f 0xfe 0xdb 0xff 0x7b 0x87
> > > > >         Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
> > > > >         Link policy: RSWITCH SNIFF
> > > > >         Link mode: PERIPHERAL ACCEPT
> > > > >         Name: '[REDACTED]'
> > > > >         Class: 0x6c010c
> > > > >         Service Classes: Rendering, Capturing, Audio, Telephony
> > > > >         Device Class: Computer, Laptop
> > > > >         HCI Version: 5.2 (0xb)  Revision: 0x20ce
> > > > >         LMP Version: 5.2 (0xb)  Subversion: 0x20ce
> > > > >         Manufacturer: Intel Corp. (2)
> > > > >
> > > > > I'm unsure how to attach the btmon output as it is a binary file that contains potentially private information (names of nearby devices)
> > > >
> > > > Well you can collect it manually and redacted the addresses, etc, up
> > > > to you, I'm usually not that paranoid of sharing this kind of
> > > > information if you scan for a short time, in the other hand I usually
> > > > use gist.github.com to do that so I can delete it later if needed.
> > >
> > > Seems to be working fine here (with AX201 and AX210):
> > >
> > > https://gist.github.com/Vudentz/d80ab63f4f62f9f216796edb4838c855
> >
> > Can you try with the following patch:
> >
> > https://patchwork.kernel.org/project/bluetooth/patch/20230821181848.3553863-1-luiz.dentz@gmail.com/
> >
> > Looks like we were not storing the whole advertisement sometimes.
> >
> >
>
> I applied your patch, along with https://www.spinics.net/lists/linux-bluetooth/msg105443.html since that defines the needed max_adv_len macro, on top of the regression-introducting commit. Unfortunately, it did not help — LE devices are still not detected (either at all or after one-two successful pairings).

Seems I managed to reproduce:

https://gist.github.com/Vudentz/0988a78da9e5658a06edec159847db27

Looks like ax201 is not generating SCAN_RSP for some reason, will need
to follow up with Intel firmware folks to see what is going on.

-- 
Luiz Augusto von Dentz