Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Arend Van Spriel]

+ Ilya
+ bt list

On 21-2-2016 15:17, Jérôme de Bretagne wrote:
> Hi,
>
> Since a commit from last April, the BCM4324(1) rev B5 chipset is
> supported in the Linux kernel and WiFi networking has been working
> like a charm on a Lenovo ThinkPad 8 tablet (running Linux
> 4.3.0-1-amd64 from Debian). A big "Thanks" for that.
>
> On the other hand, the Bluetooth part doesn't show up at all (empty
> result with hciconfig) and doesn't seem active with that same 4.3
> kernel. I don't even see any error message in dmesg about Bluetooth.
>
> bluetooth/btbcm.c mentions quite a few chipsets (UART/USB) but not the
> 4324 rev B5 so I'm not sure of the current Bluetooth status for that
> specific one. Is this chipset enabled for Bluetooth starting with
> Linux 4.4 or in the coming 4.5 version maybe? I couldn't check 4.4
> properly yet as this kernel version locks currently on my tablet (for
> other reasons).
>
> If it is not fully supported yet, feel free to ask any info that could
> help enable Bluetooth for this chipset on the ThinkPad 8 tablet!
>
> Thanks,
>
> Jérôme
>
>
> P.S. I guess the following page
> https://wireless.wiki.kernel.org/en/users/Drivers/brcm80211 could be
> updated to show that Wireless LAN is properly working.

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Jérôme de Bretagne]

[-- Attachment #1: Type: text/plain, Size: 3152 bytes --]

Hi,

I've been able to continue my investigation and I can confirm that I am
getting the same results on both Linux kernel:
- 4.4 - linux-image-4.4.0-1-amd64 from Debian (v4.4.2-3)
- 4.5 rc4 - linux-image-4.5.0-rc4-amd64 from Debian (4.5~rc4-1~exp1)
and in both cases no error message whatsoever related to Bluetooth.

It seems this BCM4324(1) rev B5 chipset is not initialized from a Bluetooth
point of view, while working great for Wi-Fi.

I've seen recent patches adding support for some devices simply by adding
their respective ACPI ID in hci_bcm.c (BCM2E7C for MacBook 2015, BCM2E65
for the Asus T100TAF, BCM2E54 for the HP Omni 10) so I've just taken a look
at the data listed by the ACPI subsystem on this tablet:

   $ ls /sys/bus/acpi/devices/BCM* -d1
   /sys/bus/acpi/devices/BCM2E3C:00
   /sys/bus/acpi/devices/BCM2E55:00

Neither of these 2 IDs (BCM2E3C, BCM2E55) is listed in hci_bcm.c. It's just
a guess at this stage but could one of these be the missing element
required in hci_bcm.c to initialize Bluetooth on the Lenovo ThinkPad 8
Tablet maybe? I don't know if the following commands might give a hint to
filter between the two IDs:

   $ cat /sys/bus/acpi/devices/BCM2E3C\:00/status
   0
   $ cat /sys/bus/acpi/devices/BCM2E55\:00/status
   15

If I can provide more information that would be useful for you, feel free
to ask as I would be glad to help.

Jérôme

P.S. I don't know if I should provide a binary firmware somehow to trigger
its initialization (as I've done for the Wi-Fi part) but I would expect to
see an error message complaining not to find the required binary file if
that was the case.


2016-02-22 10:17 GMT+01:00 Arend Van Spriel <arend.vanspriel@broadcom.com>:

> + Ilya
> + bt list
>
> On 21-2-2016 15:17, Jérôme de Bretagne wrote:
> > Hi,https://mail.google.com/mail/u/0/#inbox/1530430f49b01795
> >
> > Since a commit from last April, the BCM4324(1) rev B5 chipset is
> > supported in the Linux kernel and WiFi networking has been working
> > like a charm on a Lenovo ThinkPad 8 tablet (running Linux
> > 4.3.0-1-amd64 from Debian). A big "Thanks" for that.
> >
> > On the other hand, the Bluetooth part doesn't show up at all (empty
> > result with hciconfig) and doesn't seem active with that same 4.3
> > kernel. I don't even see any error message in dmesg about Bluetooth.
> >
> > bluetooth/btbcm.c mentions quite a few chipsets (UART/USB) but not the
> > 4324 rev B5 so I'm not sure of the current Bluetooth status for that
> > specific one. Is this chipset enabled for Bluetooth starting with
> > Linux 4.4 or in the coming 4.5 version maybe? I couldn't check 4.4
> > properly yet as this kernel version locks currently on my tablet (for
> > other reasons).
> >
> > If it is not fully supported yet, feel free to ask any info that could
> > help enable Bluetooth for this chipset on the ThinkPad 8 tablet!
> >
> > Thanks,
> >
> > Jérôme
> >
> >
> > P.S. I guess the following page
> > https://wireless.wiki.kernel.org/en/users/Drivers/brcm80211 could be
> > updated to show that Wireless LAN is properly working.
>
>

[-- Attachment #2: Type: text/html, Size: 4251 bytes --]

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Jérôme de Bretagne]

Hi,

I've been able to continue my investigation and I can confirm that I
am getting the same results on both Linux kernel:
- 4.4 - linux-image-4.4.0-1-amd64 from Debian (v4.4.2-3)
- 4.5 rc4 - linux-image-4.5.0-rc4-amd64 from Debian (4.5~rc4-1~exp1)
and in both cases no error message whatsoever related to Bluetooth.

It seems this BCM4324(1) rev B5 chipset is not initialized from a
Bluetooth point of view, while working great for Wi-Fi.

I've seen recent patches adding support for some devices simply by
adding their respective ACPI ID in hci_bcm.c (BCM2E7C for MacBook
2015, BCM2E65 for the Asus T100TAF, BCM2E54 for the HP Omni 10) so
I've just taken a look at the data listed by the ACPI subsystem on
this tablet:

   $ ls /sys/bus/acpi/devices/BCM* -d1
   /sys/bus/acpi/devices/BCM2E3C:00
   /sys/bus/acpi/devices/BCM2E55:00

Neither of these 2 IDs (BCM2E3C, BCM2E55) is listed in hci_bcm.c. It's
just a guess at this stage but could one of these be the missing
element required in hci_bcm.c to initialize Bluetooth on the Lenovo
ThinkPad 8 Tablet maybe? I don't know if the following commands might
give a hint to filter between the two IDs:

   $ cat /sys/bus/acpi/devices/BCM2E3C\:00/status
   0
   $ cat /sys/bus/acpi/devices/BCM2E55\:00/status
   15

If I can provide more information that would be useful for you, feel
free to ask as I would be glad to help.

Jérôme

P.S. I don't know if I should provide a binary firmware somehow to
trigger its initialization (as I've done for the Wi-Fi part) but I
would expect to see an error message complaining not to find the
required binary file if that was the case.

2016-02-22 10:17 GMT+01:00 Arend Van Spriel <arend.vanspriel@broadcom.com>:
> + Ilya
> + bt list
>
> On 21-2-2016 15:17, Jérôme de Bretagne wrote:
>> Hi,
>>
>> Since a commit from last April, the BCM4324(1) rev B5 chipset is
>> supported in the Linux kernel and WiFi networking has been working
>> like a charm on a Lenovo ThinkPad 8 tablet (running Linux
>> 4.3.0-1-amd64 from Debian). A big "Thanks" for that.
>>
>> On the other hand, the Bluetooth part doesn't show up at all (empty
>> result with hciconfig) and doesn't seem active with that same 4.3
>> kernel. I don't even see any error message in dmesg about Bluetooth.
>>
>> bluetooth/btbcm.c mentions quite a few chipsets (UART/USB) but not the
>> 4324 rev B5 so I'm not sure of the current Bluetooth status for that
>> specific one. Is this chipset enabled for Bluetooth starting with
>> Linux 4.4 or in the coming 4.5 version maybe? I couldn't check 4.4
>> properly yet as this kernel version locks currently on my tablet (for
>> other reasons).
>>
>> If it is not fully supported yet, feel free to ask any info that could
>> help enable Bluetooth for this chipset on the ThinkPad 8 tablet!
>>
>> Thanks,
>>
>> Jérôme
>>
>>
>> P.S. I guess the following page
>> https://wireless.wiki.kernel.org/en/users/Drivers/brcm80211 could be
>> updated to show that Wireless LAN is properly working.
>

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Loic Poulain]

Hi Jerome,

> It seems this BCM4324(1) rev B5 chipset is not initialized from a
> Bluetooth point of view, while working great for Wi-Fi.
I suppose this combo chip uses SDIO bus for WiFi and UART for Bluetooth.
If so, you need to attach the correct UART port to the hci_uart line 
discipline from user-space.

> I've seen recent patches adding support for some devices simply by
> adding their respective ACPI ID in hci_bcm.c (BCM2E7C for MacBook
> 2015, BCM2E65 for the Asus T100TAF, BCM2E54 for the HP Omni 10) so
> I've just taken a look at the data listed by the ACPI subsystem on
> this tablet:
>
>     $ ls /sys/bus/acpi/devices/BCM* -d1
>     /sys/bus/acpi/devices/BCM2E3C:00
>     /sys/bus/acpi/devices/BCM2E55:00

Usually, Broadcom BT controller is described in the ACPI table.
You can retrieve your ACPI table at /sys/firmware/acpi/tables/DSDT and 
decompile it with iasl.
You should have a Bluetooth node (BTHX):
Device (BTH1)
             {
                 Name (_HID, "BCM...")

With some gpio/irq resources. (0: wakeup, 1: power-, 2:host-wakeup).

> Neither of these 2 IDs (BCM2E3C, BCM2E55) is listed in hci_bcm.c. It's
> just a guess at this stage but could one of these be the missing
> element required in hci_bcm.c to initialize Bluetooth on the Lenovo
> ThinkPad 8 Tablet maybe? I don't know if the following commands might
> give a hint to filter between the two IDs:
>
>     $ cat /sys/bus/acpi/devices/BCM2E3C\:00/status
>     0
>     $ cat /sys/bus/acpi/devices/BCM2E55\:00/status
>     15

Check in your acpi table if one of them matches a bluetooth node.
Then, try to add it in the acpi device id list in hci_bcm.c.

Second step is to attach the device, but you need to find the correct 
tty to use.
If your ACPI table is correct, BCM acpi device should appear as a child 
of a tty device in sysfs.
The physical_node symlink in your device directory (sys/bus/acpi...) 
gives you the hierarchical path.

Now, you can run btattach (bluez tool):
$ btattach -B /dev/ttySX -P bcm

And check that hci0 has been created
$ hciconfig

Regards,
Loic

-- 
Intel Open Source Technology Center
http://oss.intel.com/

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Loic Poulain]

Jerome,

>>> I've just taken a look at the data listed by the ACPI subsystem on
>>> this tablet:
>>>
>>>      $ ls /sys/bus/acpi/devices/BCM* -d1
>>>      /sys/bus/acpi/devices/BCM2E3C:00
>>>      /sys/bus/acpi/devices/BCM2E55:00
>>
>> Usually, Broadcom BT controller is described in the ACPI table.
>> You can retrieve your ACPI table at /sys/firmware/acpi/tables/DSDT and
>> decompile it with iasl.
>> You should have a Bluetooth node (BTHX):
>> Device (BTH1)
>>              {
>>                  Name (_HID, "BCM...")
>>
>> With some gpio/irq resources. (0: wakeup, 1: power-, 2:host-wakeup).
>>
>> Check in your acpi table if one of them matches a bluetooth node.
> Both of them are matching a Bluetooth node it seems to me, with BTH0
> and BTH1 entries respectively. I've attached the original DSDT ACPI
> table (dsdt.dat) and the decompiled version (dsdt.dsl). Any useful
> information in there?

from your DSDT:
BTH0 _STA  (status) method returns 0x00 meaning that this device is not 
present.
BTH1 _STA method returns 0x0F meaning that this device is present.

So your BT controller is BTH1 (BCM2E55).

The ACPI table for this BTH1 looks good (1 irq + 2 gpios ) so it should 
be compatible with hci_bcm.

>> Second step is to attach the device, but you need to find the correct tty to
>> use.
>> If your ACPI table is correct, BCM acpi device should appear as a child of a
>> tty device in sysfs.
>> The physical_node symlink in your device directory (sys/bus/acpi...) gives
>> you the hierarchical path.
> Before applying any patches to hci_bcm.c , BCM2E55 has already the
> following hierarchical path:
>
>     /sys/bus/acpi/devices/BCM2E55:00$ ls -l physical_node
>     lrwxrwxrwx 1 root root 0 févr. 28 21:45 physical_node ->
> ../../../../platform/80860F0A:00/BCM2E55:00
80860F0A is the UART controller (8250_dw), looks good.
So you should also find BCM2E55 id into:
/sys/class/ttyS*/device/

> while BCM2E3C has none:
>
>     sys/bus/acpi/devices/BCM2E3C:00$ ls
>     hid  modalias  path  power  status  subsystem  uevent

Yes because BCM2E3C is not enabled.

Regards,
Loic

-- 
Intel Open Source Technology Center
http://oss.intel.com/

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Jérôme de Bretagne]

Hi,

> So your BT controller is BTH1 (BCM2E55).
> The ACPI table for this BTH1 looks good (1 irq + 2 gpios ) so it should b=
e
> compatible with hci_bcm.

I can confirm that it is compatible with hci_bcm; Bluetooth is indeed
fully working using the bluetooth-next tree on the ThinkPad Tablet 8.
Thanks again Marcel to have applied the patch and Loic to have helped
me find the right ACPI ID.

Maybe this is a naive question about ACPI ID, and not specific to
Broadcom by the way: could there be a more generic way to gather ACPI
IDs used by new devices and add them to the kernel on a regular basis?
They seem to be added almost one by one mainly upon user reports when
looking at the mailing list. So Bluetooth doesn't work out-of-the-box
on some systems which would be compatible otherwise, especially the
less popular ones. Is there no existing driver list to re-use perhaps?


> Now, you can run btattach (bluez tool):
> $ btattach -B /dev/ttySX -P bcm

Indeed, it is working fine with ttyS1 and the correct binary firmware
stored at /lib/firmware/brcm/BCM.hcd:

   # btattach -B /dev/ttyS1 -P bcm
   Attaching BR/EDR controller to /dev/ttyS1
   Switched line discipline from 0 to 15
   Device index 0 attached

Finally, is btattach the only command recommended and supported moving
forward, or should hciattach work as well with that specific
chipset/tablet (as a transition until most Linux distros switch to
btattach)?

Regards,
J=C3=A9r=C3=B4me


P.S. for information if it can be useful to someone else:

> And check that hci0 has been created
> $ hciconfig

$ hciconfig -a
hci0:    Type: BR/EDR  Bus: UART
    BD Address: B0:10:41:B8:16:6C  ACL MTU: 1021:8  SCO MTU: 64:1
    UP RUNNING
    RX bytes:110596 acl:4422 sco:0 events:328 errors:0
    TX bytes:35333 acl:36 sco:0 commands:273 errors:0
    Features: 0xbf 0xfe 0xcf 0xfe 0xdb 0xff 0x7b 0x87
    Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
    Link policy: RSWITCH SNIFF
    Link mode: SLAVE ACCEPT
    Name: 'thinkpad8'
    Class: 0x10011c
    Service Classes: Object Transfer
    Device Class: Computer, Unknown (reserved) minor device class
    HCI Version: 4.0 (0x6)  Revision: 0x3492
    LMP Version: 4.0 (0x6)  Subversion: 0x4606
    Manufacturer: Broadcom Corporation (15)

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Loic Poulain]

Hi Jerome,

> Maybe this is a naive question about ACPI ID, and not specific to
> Broadcom by the way: could there be a more generic way to gather ACPI
> IDs used by new devices and add them to the kernel on a regular basis?
> They seem to be added almost one by one mainly upon user reports when
> looking at the mailing list. So Bluetooth doesn't work out-of-the-box
> on some systems which would be compatible otherwise, especially the
> less popular ones. Is there no existing driver list to re-use perhaps?

New ACPI ID means that this is a different hardware. So it needs to be
tested before being added into the driver supported list. An already
supported hardware should keep the same ACPI ID, making it work out-of
the-box.

Since this driver is pretty new, this is not surprising that some IDs
are regularly added. This is a good thing. Meaning that people use this
driver and that Broadcom controllers keep good compatibility across
hardware versions.

> Indeed, it is working fine with ttyS1 and the correct binary firmware
> stored at /lib/firmware/brcm/BCM.hcd:
>
>     # btattach -B /dev/ttyS1 -P bcm
>     Attaching BR/EDR controller to /dev/ttyS1
>     Switched line discipline from 0 to 15
>     Device index 0 attached
>
> Finally, is btattach the only command recommended and supported moving
> forward, or should hciattach work as well with that specific
> chipset/tablet (as a transition until most Linux distros switch to
> btattach)?

Absolutely, think btattach is the only recommended way.
Even if you can make it work with hciattach by adding device entry with
correct protocol ID, it is not recommended since hciattach is dedicated
to user-space initialisation (old-way) and code is pretty deprecated.

Regards,
Loic

-- 
Intel Open Source Technology Center
http://oss.intel.com/

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Jérôme de Bretagne]

Hi Loic,

> New ACPI ID means that this is a different hardware. So it needs to be
> tested before being added into the driver supported list. An already
> supported hardware should keep the same ACPI ID, making it work out-of
> the-box.

That makes sense.

What about the other ACPI ID (BCM2E3C) that I've found in the ACPI
table of that ThinkPad 8 tablet? Is it included in its table to cover
another HW variant of that product, using a different Broadcom
chipset? Any value for you if I create a similar patch (without being
able to test it)? I guess the answer will be "No" from your message
above but just let me know.


>> Finally, is btattach the only command recommended and supported moving
>> forward
>
>
> Absolutely, think btattach is the only recommended way.
> Even if you can make it work with hciattach by adding device entry with
> correct protocol ID, it is not recommended since hciattach is dedicated
> to user-space initialisation (old-way) and code is pretty deprecated.

That's what I thought, thanks for the confirmation.

Regards,
J=C3=A9r=C3=B4me

Re: Broadcom BCM4324(1) rev B5 chipset - Bluetooth support in Linux Kernel 4.x ?

[Loic Poulain]

Hi Jerome


> What about the other ACPI ID (BCM2E3C) that I've found in the ACPI
> table of that ThinkPad 8 tablet? Is it included in its table to cover
> another HW variant of that product, using a different Broadcom
> chipset? Any value for you if I create a similar patch (without being
> able to test it)? I guess the answer will be "No" from your message
> above but just let me know.

Yes, ACPI table is often filled with different optional components so
that it can be compatible with several platform variants without impact
on the ACPI data. Then, components are enabled/disabled depending board ID.

This BCM2E3C is disabled on your platform, but seems compatible with 
hci_bcm UART driver (2GPIOs + 1IRQ). Should be only a revision of BCM2Ex.
Broadcom input is welcome here.

Regards,
Loic

-- 
Intel Open Source Technology Center
http://oss.intel.com/