The Qualcomm Bluetooth driver is configuring the device address in
reverse order for none-ROME devices, which breaks user space tools like
btmgmt and the 'local-bd-address' devicetree property.
As these Qualcomm controllers lack persistent storage for the device
address, boot firmware can use the 'local-bd-address' devicetree
property to provide a valid address. The property should specify the
address in little endian order, but instead some boot firmware has been
reversing the address to match the buggy Qualcomm driver.
This specifically affects some Chromebook devices for which we now need
to maintain compatibility by deprecating the affected compatible string,
marking the corresponding devicetree properties as broken, and reversing
the addresses after parsing them. Fortunately, only 'qcom,wcn3991-bt' is
impacted and needs to be deprecated this way according to the Chromium
team [1].
Note that this series depends on the following revert:
https://lore.kernel.org/lkml/[email protected]/
Johan
[1] https://lore.kernel.org/lkml/[email protected]/
Changes in v2
- add quirk to handle deprecated devicetree compatibles that expect
broken address properties
- deprecate 'qcom,wcn3991-bt' and mark it as broken
Johan Hovold (4):
dt-bindings: bluetooth: add new wcn3991 compatible to fix bd_addr
Bluetooth: add quirk for broken address properties
Bluetooth: qca: fix device-address endianness
Bluetooth: qca: fix wcn3991 'local-bd-address' endianness
.../net/bluetooth/qualcomm-bluetooth.yaml | 29 +++++++++++--------
drivers/bluetooth/btqca.c | 8 +++--
drivers/bluetooth/hci_qca.c | 11 +++++++
include/net/bluetooth/hci.h | 10 +++++++
net/bluetooth/hci_sync.c | 5 +++-
5 files changed, 48 insertions(+), 15 deletions(-)
--
2.43.2
Several Qualcomm Bluetooth controllers lack persistent storage for the
device address and instead one can be provided by the boot firmware
using the 'local-bd-address' devicetree property.
The Bluetooth bindings clearly says that the address should be specified
in little-endian order, but due to a long-standing bug in the Qualcomm
driver which reversed the address some bootloaders have been providing
the address in big-endian order instead.
The only device out there that should be affected by this is the WCN3991
used in some Chromebooks. To maintain backwards compatibility, mark the
current compatible string as deprecated and add a new
'qcom,wcn3991-bt-bdaddr-le' for firmware which conforms with the
binding.
Signed-off-by: Johan Hovold <[email protected]>
---
.../net/bluetooth/qualcomm-bluetooth.yaml | 29 +++++++++++--------
1 file changed, 17 insertions(+), 12 deletions(-)
diff --git a/Documentation/devicetree/bindings/net/bluetooth/qualcomm-bluetooth.yaml b/Documentation/devicetree/bindings/net/bluetooth/qualcomm-bluetooth.yaml
index eba2f3026ab0..b6fce6d02138 100644
--- a/Documentation/devicetree/bindings/net/bluetooth/qualcomm-bluetooth.yaml
+++ b/Documentation/devicetree/bindings/net/bluetooth/qualcomm-bluetooth.yaml
@@ -15,18 +15,22 @@ description:
properties:
compatible:
- enum:
- - qcom,qca2066-bt
- - qcom,qca6174-bt
- - qcom,qca9377-bt
- - qcom,wcn3988-bt
- - qcom,wcn3990-bt
- - qcom,wcn3991-bt
- - qcom,wcn3998-bt
- - qcom,qca6390-bt
- - qcom,wcn6750-bt
- - qcom,wcn6855-bt
- - qcom,wcn7850-bt
+ oneOf:
+ - enum:
+ - qcom,qca2066-bt
+ - qcom,qca6174-bt
+ - qcom,qca9377-bt
+ - qcom,wcn3988-bt
+ - qcom,wcn3990-bt
+ - qcom,wcn3991-bt-bdaddr-le
+ - qcom,wcn3998-bt
+ - qcom,qca6390-bt
+ - qcom,wcn6750-bt
+ - qcom,wcn6855-bt
+ - qcom,wcn7850-bt
+ - enum:
+ - qcom,wcn3991-bt
+ deprecated: true
enable-gpios:
maxItems: 1
@@ -122,6 +126,7 @@ allOf:
- qcom,wcn3988-bt
- qcom,wcn3990-bt
- qcom,wcn3991-bt
+ - qcom,wcn3991-bt-bdaddr-le
- qcom,wcn3998-bt
then:
required:
--
2.43.2
Due to a long-standing bug in the Qualcomm Bluetooth driver, the device
address has so far been reversed when configuring the controller.
This has led to one vendor reversing the address provided by the
boot firmware using the 'local-bd-address' devicetree property.
The only device affected by this should be the WCN3991 used in some
Chromebooks. The corresponding compatible string has now been deprecated
so that the underlying driver bug can be fixed without breaking
backwards compatibility.
Set the HCI_QUIRK_BDADDR_PROPERTY_BROKEN quirk for the deprecated
compatible string and add the new 'qcom,wcn3991-bt-bdaddr-le' string to
the match table for boot firmware that conforms with the binding.
Fixes: 7d250a062f75 ("Bluetooth: hci_qca: Add support for Qualcomm Bluetooth SoC WCN3991")
Cc: [email protected] # 5.10
Cc: Balakrishna Godavarthi <[email protected]>
Signed-off-by: Johan Hovold <[email protected]>
---
drivers/bluetooth/hci_qca.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/drivers/bluetooth/hci_qca.c b/drivers/bluetooth/hci_qca.c
index f989c05f8177..346274fe66d8 100644
--- a/drivers/bluetooth/hci_qca.c
+++ b/drivers/bluetooth/hci_qca.c
@@ -1904,6 +1904,16 @@ static int qca_setup(struct hci_uart *hu)
case QCA_WCN6855:
case QCA_WCN7850:
set_bit(HCI_QUIRK_USE_BDADDR_PROPERTY, &hdev->quirks);
+
+ if (soc_type == QCA_WCN3991) {
+ struct device *dev = GET_HCIDEV_DEV(hdev);
+
+ if (device_is_compatible(dev, "qcom,wcn3991-bt")) {
+ set_bit(HCI_QUIRK_BDADDR_PROPERTY_BROKEN,
+ &hdev->quirks);
+ }
+ }
+
hci_set_aosp_capable(hdev);
ret = qca_read_soc_version(hdev, &ver, soc_type);
@@ -2597,6 +2607,7 @@ static const struct of_device_id qca_bluetooth_of_match[] = {
{ .compatible = "qcom,wcn3988-bt", .data = &qca_soc_data_wcn3988},
{ .compatible = "qcom,wcn3990-bt", .data = &qca_soc_data_wcn3990},
{ .compatible = "qcom,wcn3991-bt", .data = &qca_soc_data_wcn3991},
+ { .compatible = "qcom,wcn3991-bt-bdaddr-le", .data = &qca_soc_data_wcn3991},
{ .compatible = "qcom,wcn3998-bt", .data = &qca_soc_data_wcn3998},
{ .compatible = "qcom,wcn6750-bt", .data = &qca_soc_data_wcn6750},
{ .compatible = "qcom,wcn6855-bt", .data = &qca_soc_data_wcn6855},
--
2.43.2
Some Bluetooth controllers lack persistent storage for the device
address and instead one can be provided by the boot firmware using the
'local-bd-address' devicetree property.
The Bluetooth devicetree bindings clearly states that the address should
be specified in little-endian order, but due to a long-standing bug in
the Qualcomm driver which reversed the address some bootloaders have
been providing the address in big-endian order instead.
Add a new quirk that can be used to mark deprecated compatible strings
that expect such broken devicetree properties and use it to reverse the
address when parsing the property so that the underlying driver bug can
be fixed.
Fixes: 5c0a1001c8be ("Bluetooth: hci_qca: Add helper to set device address")
Cc: [email protected] # 5.1
Signed-off-by: Johan Hovold <[email protected]>
---
include/net/bluetooth/hci.h | 10 ++++++++++
net/bluetooth/hci_sync.c | 5 ++++-
2 files changed, 14 insertions(+), 1 deletion(-)
diff --git a/include/net/bluetooth/hci.h b/include/net/bluetooth/hci.h
index bdee5d649cc6..556cffed5698 100644
--- a/include/net/bluetooth/hci.h
+++ b/include/net/bluetooth/hci.h
@@ -176,6 +176,16 @@ enum {
*/
HCI_QUIRK_USE_BDADDR_PROPERTY,
+ /* When this quirk is set, the Bluetooth Device Address provided by
+ * the 'local-bd-address' fwnode property is incorrectly specified in
+ * big-endian order.
+ *
+ * This quirk can be set before hci_register_dev is called or
+ * during the hdev->setup vendor callback and must only be used for
+ * deprecated compatible strings.
+ */
+ HCI_QUIRK_BDADDR_PROPERTY_BROKEN,
+
/* When this quirk is set, the duplicate filtering during
* scanning is based on Bluetooth devices addresses. To allow
* RSSI based updates, restart scanning if needed.
diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
index 5716345a26df..283ae8edc1e5 100644
--- a/net/bluetooth/hci_sync.c
+++ b/net/bluetooth/hci_sync.c
@@ -3215,7 +3215,10 @@ static void hci_dev_get_bd_addr_from_property(struct hci_dev *hdev)
if (ret < 0 || !bacmp(&ba, BDADDR_ANY))
return;
- bacpy(&hdev->public_addr, &ba);
+ if (test_bit(HCI_QUIRK_BDADDR_PROPERTY_BROKEN, &hdev->quirks))
+ baswap(&hdev->public_addr, &ba);
+ else
+ bacpy(&hdev->public_addr, &ba);
}
struct hci_init_stage {
--
2.43.2
The WCN6855 firmware on the Lenovo ThinkPad X13s expects the Bluetooth
device address in big-endian order when setting it using the
EDL_WRITE_BD_ADDR_OPCODE command.
Presumably, this is the case for all non-ROME devices which all use the
EDL_WRITE_BD_ADDR_OPCODE command for this (unlike the ROME devices which
use a different command and expect the address in little-endian order).
Reverse the little-endian address before setting it to make sure that
the address can be configured using tools like btmgmt or using the
'local-bd-address' devicetree property.
Note that this can potentially break systems with boot firmware which
has started relying on the broken behaviour and is incorrectly passing
the address via devicetree in big-endian order.
Fixes: 5c0a1001c8be ("Bluetooth: hci_qca: Add helper to set device address")
Cc: [email protected] # 5.1
Cc: Balakrishna Godavarthi <[email protected]>
Cc: Matthias Kaehlcke <[email protected]>
Tested-by: Nikita Travkin <[email protected]> # sc7180
Signed-off-by: Johan Hovold <[email protected]>
---
drivers/bluetooth/btqca.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/drivers/bluetooth/btqca.c b/drivers/bluetooth/btqca.c
index b40b32fa7f1c..19cfc342fc7b 100644
--- a/drivers/bluetooth/btqca.c
+++ b/drivers/bluetooth/btqca.c
@@ -826,11 +826,15 @@ EXPORT_SYMBOL_GPL(qca_uart_setup);
int qca_set_bdaddr(struct hci_dev *hdev, const bdaddr_t *bdaddr)
{
+ bdaddr_t bdaddr_swapped;
struct sk_buff *skb;
int err;
- skb = __hci_cmd_sync_ev(hdev, EDL_WRITE_BD_ADDR_OPCODE, 6, bdaddr,
- HCI_EV_VENDOR, HCI_INIT_TIMEOUT);
+ baswap(&bdaddr_swapped, bdaddr);
+
+ skb = __hci_cmd_sync_ev(hdev, EDL_WRITE_BD_ADDR_OPCODE, 6,
+ &bdaddr_swapped, HCI_EV_VENDOR,
+ HCI_INIT_TIMEOUT);
if (IS_ERR(skb)) {
err = PTR_ERR(skb);
bt_dev_err(hdev, "QCA Change address cmd failed (%d)", err);
--
2.43.2
This is an automated email and please do not reply to this email.
Dear Submitter,
Thank you for submitting the patches to the linux bluetooth mailing list.
While preparing the CI tests, the patches you submitted couldn't be applied to the current HEAD of the repository.
----- Output -----
error: patch failed: drivers/bluetooth/hci_qca.c:1904
error: drivers/bluetooth/hci_qca.c: patch does not apply
hint: Use 'git am --show-current-patch' to see the failed patch
Please resolve the issue and submit the patches again.
---
Regards,
Linux Bluetooth
On Mon, 18 Mar 2024 at 13:09, Johan Hovold <[email protected]> wrote:
>
> Several Qualcomm Bluetooth controllers lack persistent storage for the
> device address and instead one can be provided by the boot firmware
> using the 'local-bd-address' devicetree property.
>
> The Bluetooth bindings clearly says that the address should be specified
> in little-endian order, but due to a long-standing bug in the Qualcomm
> driver which reversed the address some bootloaders have been providing
> the address in big-endian order instead.
>
> The only device out there that should be affected by this is the WCN3991
> used in some Chromebooks. To maintain backwards compatibility, mark the
> current compatible string as deprecated and add a new
> 'qcom,wcn3991-bt-bdaddr-le' for firmware which conforms with the
> binding.
>
> Signed-off-by: Johan Hovold <[email protected]>
> ---
> .../net/bluetooth/qualcomm-bluetooth.yaml | 29 +++++++++++--------
> 1 file changed, 17 insertions(+), 12 deletions(-)
>
> diff --git a/Documentation/devicetree/bindings/net/bluetooth/qualcomm-bluetooth.yaml b/Documentation/devicetree/bindings/net/bluetooth/qualcomm-bluetooth.yaml
> index eba2f3026ab0..b6fce6d02138 100644
> --- a/Documentation/devicetree/bindings/net/bluetooth/qualcomm-bluetooth.yaml
> +++ b/Documentation/devicetree/bindings/net/bluetooth/qualcomm-bluetooth.yaml
> @@ -15,18 +15,22 @@ description:
>
> properties:
> compatible:
> - enum:
> - - qcom,qca2066-bt
> - - qcom,qca6174-bt
> - - qcom,qca9377-bt
> - - qcom,wcn3988-bt
> - - qcom,wcn3990-bt
> - - qcom,wcn3991-bt
> - - qcom,wcn3998-bt
> - - qcom,qca6390-bt
> - - qcom,wcn6750-bt
> - - qcom,wcn6855-bt
> - - qcom,wcn7850-bt
> + oneOf:
> + - enum:
> + - qcom,qca2066-bt
> + - qcom,qca6174-bt
> + - qcom,qca9377-bt
> + - qcom,wcn3988-bt
> + - qcom,wcn3990-bt
> + - qcom,wcn3991-bt-bdaddr-le
This compatible doesn't describe new hardware kind. As such, I think,
the better way would be to continue using qcom,wcn3991-bt compatible
string + add some kind of qcom,bt-addr-le property.
> + - qcom,wcn3998-bt
> + - qcom,qca6390-bt
> + - qcom,wcn6750-bt
> + - qcom,wcn6855-bt
> + - qcom,wcn7850-bt
> + - enum:
> + - qcom,wcn3991-bt
> + deprecated: true
>
> enable-gpios:
> maxItems: 1
> @@ -122,6 +126,7 @@ allOf:
> - qcom,wcn3988-bt
> - qcom,wcn3990-bt
> - qcom,wcn3991-bt
> + - qcom,wcn3991-bt-bdaddr-le
> - qcom,wcn3998-bt
> then:
> required:
> --
> 2.43.2
>
>
--
With best wishes
Dmitry
On Mon, Mar 18, 2024 at 03:00:40PM +0200, Dmitry Baryshkov wrote:
> On Mon, 18 Mar 2024 at 13:09, Johan Hovold <[email protected]> wrote:
> >
> > Several Qualcomm Bluetooth controllers lack persistent storage for the
> > device address and instead one can be provided by the boot firmware
> > using the 'local-bd-address' devicetree property.
> >
> > The Bluetooth bindings clearly says that the address should be specified
> > in little-endian order, but due to a long-standing bug in the Qualcomm
> > driver which reversed the address some bootloaders have been providing
> > the address in big-endian order instead.
> >
> > The only device out there that should be affected by this is the WCN3991
> > used in some Chromebooks. To maintain backwards compatibility, mark the
> > current compatible string as deprecated and add a new
> > 'qcom,wcn3991-bt-bdaddr-le' for firmware which conforms with the
> > binding.
> This compatible doesn't describe new hardware kind. As such, I think,
> the better way would be to continue using qcom,wcn3991-bt compatible
> string + add some kind of qcom,bt-addr-le property.
No, you can't handle backwards compatibility by *adding* a property.
I wanted to avoid doing this, but if we have to support Google's broken
boot firmware for these devices, then this is how it needs to be done.
Johan
On Mon, Mar 18, 2024 at 02:17:47PM +0100, Johan Hovold wrote:
> On Mon, Mar 18, 2024 at 03:00:40PM +0200, Dmitry Baryshkov wrote:
> > On Mon, 18 Mar 2024 at 13:09, Johan Hovold <[email protected]> wrote:
> > >
> > > Several Qualcomm Bluetooth controllers lack persistent storage for the
> > > device address and instead one can be provided by the boot firmware
> > > using the 'local-bd-address' devicetree property.
> > >
> > > The Bluetooth bindings clearly says that the address should be specified
> > > in little-endian order, but due to a long-standing bug in the Qualcomm
> > > driver which reversed the address some bootloaders have been providing
> > > the address in big-endian order instead.
> > >
> > > The only device out there that should be affected by this is the WCN3991
> > > used in some Chromebooks. To maintain backwards compatibility, mark the
> > > current compatible string as deprecated and add a new
> > > 'qcom,wcn3991-bt-bdaddr-le' for firmware which conforms with the
> > > binding.
>
> > This compatible doesn't describe new hardware kind. As such, I think,
> > the better way would be to continue using qcom,wcn3991-bt compatible
> > string + add some kind of qcom,bt-addr-le property.
>
> No, you can't handle backwards compatibility by *adding* a property.
But you could add a property for the not broken case. That's a bit odd,
but so is your compatible.
> I wanted to avoid doing this, but if we have to support Google's broken
> boot firmware for these devices, then this is how it needs to be done.
Don't Chromebooks update everything together. So maybe we don't care in
this case?
Rob
On Mon, Mar 18, 2024 at 09:48:06AM -0500, Rob Herring wrote:
> On Mon, Mar 18, 2024 at 02:17:47PM +0100, Johan Hovold wrote:
> > On Mon, Mar 18, 2024 at 03:00:40PM +0200, Dmitry Baryshkov wrote:
> > > On Mon, 18 Mar 2024 at 13:09, Johan Hovold <[email protected]> wrote:
> > > > The only device out there that should be affected by this is the WCN3991
> > > > used in some Chromebooks. To maintain backwards compatibility, mark the
> > > > current compatible string as deprecated and add a new
> > > > 'qcom,wcn3991-bt-bdaddr-le' for firmware which conforms with the
> > > > binding.
> >
> > > This compatible doesn't describe new hardware kind. As such, I think,
> > > the better way would be to continue using qcom,wcn3991-bt compatible
> > > string + add some kind of qcom,bt-addr-le property.
> >
> > No, you can't handle backwards compatibility by *adding* a property.
>
> But you could add a property for the not broken case. That's a bit odd,
> but so is your compatible.
Sure, we could have a property that we only add for wcn3991-bt going
forward.
But we can't go back in time and add this property to all devicetrees
and say that 'local-bd-address' is big endian unless that property is
present (as that would leave all the non-wcn3991 devicetrees broken).
> > I wanted to avoid doing this, but if we have to support Google's broken
> > boot firmware for these devices, then this is how it needs to be done.
>
> Don't Chromebooks update everything together. So maybe we don't care in
> this case?
That was my hope, but Matthias seemed to suggest that we need to
continue supporting the current (broken) binding because doing such a
coordinated update may be easier said than done:
https://lore.kernel.org/lkml/[email protected]/
A new compatible string (or one-off property) would allow them do make a
change when they are ready (e.g. by only updating the devicetrees after
all boot firmware has been patched and pushed out).
Johan
Hi,
On Mon, Mar 18, 2024 at 8:10 AM Johan Hovold <[email protected]> wrote:
>
> > > I wanted to avoid doing this, but if we have to support Google's broken
> > > boot firmware for these devices, then this is how it needs to be done.
> >
> > Don't Chromebooks update everything together. So maybe we don't care in
> > this case?
>
> That was my hope, but Matthias seemed to suggest that we need to
> continue supporting the current (broken) binding because doing such a
> coordinated update may be easier said than done:
>
> https://lore.kernel.org/lkml/[email protected]/
Chromebooks update kernel and devicetree together, but not firmware.
Firmware is relatively hard to get updated trying to have kernel and
firmware updates coordinated at the exact same time has challenges.
This would further be complicated by the fact that firmware
qualification for each variant happens on its own timeline.
> A new compatible string (or one-off property) would allow them do make a
> change when they are ready (e.g. by only updating the devicetrees after
> all boot firmware has been patched and pushed out).
I have no real opinion about the exact way this is solved so happy to
let DT folks decide on how they want this. I will note, however, that
device trees are never shipped separately and thus we have no
intrinsic need for DT backward compatbility here. It would be OK from
a ChromeOS perspective to add a property or compatible string for the
broken case.
-Doug
Hi,
On Mon, Mar 18, 2024 at 8:26 AM Doug Anderson <[email protected]> wrote:
>
> Hi,
>
> On Mon, Mar 18, 2024 at 8:10 AM Johan Hovold <[email protected]> wrote:
> >
> > > > I wanted to avoid doing this, but if we have to support Google's broken
> > > > boot firmware for these devices, then this is how it needs to be done.
> > >
> > > Don't Chromebooks update everything together. So maybe we don't care in
> > > this case?
> >
> > That was my hope, but Matthias seemed to suggest that we need to
> > continue supporting the current (broken) binding because doing such a
> > coordinated update may be easier said than done:
> >
> > https://lore.kernel.org/lkml/[email protected]/
>
> Chromebooks update kernel and devicetree together, but not firmware.
> Firmware is relatively hard to get updated trying to have kernel and
> firmware updates coordinated at the exact same time has challenges.
> This would further be complicated by the fact that firmware
> qualification for each variant happens on its own timeline.
>
>
> > A new compatible string (or one-off property) would allow them do make a
> > change when they are ready (e.g. by only updating the devicetrees after
> > all boot firmware has been patched and pushed out).
>
> I have no real opinion about the exact way this is solved so happy to
> let DT folks decide on how they want this. I will note, however, that
> device trees are never shipped separately and thus we have no
> intrinsic need for DT backward compatbility here. It would be OK from
> a ChromeOS perspective to add a property or compatible string for the
> broken case.
Actually, I should probably say more about this to make it clear how it works.
Chromebooks ship the kernel as a FIT image which bundles the kernel
and device trees together. The firmware looks at all the bundled
device trees and picks the proper one based on the board name,
revision, and SKU ID. The firmware then looks for the bluetooth node
(I believe it finds it from the "aliases" section) and adds the MAC
address there.
...so we could update the DT to add a property (if that's desired)
even if we don't update the firmware.
-Doug
On Mon, Mar 18, 2024 at 08:31:09AM -0700, Doug Anderson wrote:
> On Mon, Mar 18, 2024 at 8:26 AM Doug Anderson <[email protected]> wrote:
> > > A new compatible string (or one-off property) would allow them do make a
> > > change when they are ready (e.g. by only updating the devicetrees after
> > > all boot firmware has been patched and pushed out).
> >
> > I have no real opinion about the exact way this is solved so happy to
> > let DT folks decide on how they want this. I will note, however, that
> > device trees are never shipped separately and thus we have no
> > intrinsic need for DT backward compatbility here. It would be OK from
> > a ChromeOS perspective to add a property or compatible string for the
> > broken case.
>
> Actually, I should probably say more about this to make it clear how it works.
>
> Chromebooks ship the kernel as a FIT image which bundles the kernel
> and device trees together. The firmware looks at all the bundled
> device trees and picks the proper one based on the board name,
> revision, and SKU ID. The firmware then looks for the bluetooth node
> (I believe it finds it from the "aliases" section) and adds the MAC
> address there.
>
> ...so we could update the DT to add a property (if that's desired)
> even if we don't update the firmware.
Thanks for the details. Sounds like we could get away with adding a new
property for the broken firmware in this case, which should resolve this
nicely without having to deprecate anything.
Could you carry such a devicetree patch out-of-tree until the firmware
has been fixed?
Johan
Hi,
On Mon, Mar 18, 2024 at 8:47 AM Johan Hovold <[email protected]> wrote:
>
> On Mon, Mar 18, 2024 at 08:31:09AM -0700, Doug Anderson wrote:
> > On Mon, Mar 18, 2024 at 8:26 AM Doug Anderson <[email protected]> wrote:
>
> > > > A new compatible string (or one-off property) would allow them do make a
> > > > change when they are ready (e.g. by only updating the devicetrees after
> > > > all boot firmware has been patched and pushed out).
> > >
> > > I have no real opinion about the exact way this is solved so happy to
> > > let DT folks decide on how they want this. I will note, however, that
> > > device trees are never shipped separately and thus we have no
> > > intrinsic need for DT backward compatbility here. It would be OK from
> > > a ChromeOS perspective to add a property or compatible string for the
> > > broken case.
> >
> > Actually, I should probably say more about this to make it clear how it works.
> >
> > Chromebooks ship the kernel as a FIT image which bundles the kernel
> > and device trees together. The firmware looks at all the bundled
> > device trees and picks the proper one based on the board name,
> > revision, and SKU ID. The firmware then looks for the bluetooth node
> > (I believe it finds it from the "aliases" section) and adds the MAC
> > address there.
> >
> > ...so we could update the DT to add a property (if that's desired)
> > even if we don't update the firmware.
>
> Thanks for the details. Sounds like we could get away with adding a new
> property for the broken firmware in this case, which should resolve this
> nicely without having to deprecate anything.
>
> Could you carry such a devicetree patch out-of-tree until the firmware
> has been fixed?
IMO we shouldn't try to fix the firmware at all. Given the fact that
it took me a year to get a firmware uprev completed for one trogdor
variant for fixes that actually had functional impact, it's possible
we'll never actually get an uprev completed that includes this fix or
it will happen years from now when nobody remembers about it. I'm also
certain this whole issue will also cause a bunch of debugging over the
years if we try to fix it in firmware like that. There are cases where
people end up running with old firmware since the developer workflow
doesn't automatically update it.
The handling should be added upstream and we should just accept that
the trogdor firmware gets it backward.
-Doug
On Mon, Mar 18, 2024 at 08:58:40AM -0700, Doug Anderson wrote:
> On Mon, Mar 18, 2024 at 8:47 AM Johan Hovold <[email protected]> wrote:
> > On Mon, Mar 18, 2024 at 08:31:09AM -0700, Doug Anderson wrote:
> > Thanks for the details. Sounds like we could get away with adding a new
> > property for the broken firmware in this case, which should resolve this
> > nicely without having to deprecate anything.
> >
> > Could you carry such a devicetree patch out-of-tree until the firmware
> > has been fixed?
>
> IMO we shouldn't try to fix the firmware at all. Given the fact that
> it took me a year to get a firmware uprev completed for one trogdor
> variant for fixes that actually had functional impact, it's possible
> we'll never actually get an uprev completed that includes this fix or
> it will happen years from now when nobody remembers about it. I'm also
> certain this whole issue will also cause a bunch of debugging over the
> years if we try to fix it in firmware like that. There are cases where
> people end up running with old firmware since the developer workflow
> doesn't automatically update it.
>
> The handling should be added upstream and we should just accept that
> the trogdor firmware gets it backward.
Fair enough.
Rob, are you OK with adding a 'qcom,local-bd-address-broken' or
similarly named property to indicate that the boot firmware passes the
address in the wrong order?
I'd then add that property to sc7180-trogdor.dtsi in mainline.
Johan