Re: [PATCH v4 1/1] virtio_bt: Fix alignment in configuration struct

[Igor Skalkin]

Hi Luiz Augusto,

On 10/24/22 22:54, Luiz Augusto von Dentz wrote:
> Hi Igor,
> 
> On Mon, Oct 24, 2022 at 6:41 AM Igor Skalkin
> <Igor.Skalkin@xxxxxxxxxxxxxxx> wrote:
>>
>> The current version of the configuration structure has unaligned
>> 16-bit fields, but according to the specification [1], access to
>> the configuration space must be aligned.
>>
>> Add a second, aligned  version of the configuration structure
>> and a new feature bit indicating that this version is being used.
>>
>> [1] https://ddec1-0-en-ctp.trendmicro.com:443/wis/clicktime/v1/query?url=https%3a%2f%2fdocs.oasis%2dopen.org%2fvirtio%2fvirtio%2fv1.1%2fvirtio%2dv1.1.pdf&umid=db3482bc-5b84-4bde-bbb0-41d837955a7a&auth=53c7c7de28b92dfd96e93d9dd61a23e634d2fbec-d27a9d4c2c971f9ecc5d00d40d5cd9b45c4b5f63
>>
>> Signed-off-by: Igor Skalkin <Igor.Skalkin@xxxxxxxxxxxxxxx>
>> ---
>>  drivers/bluetooth/virtio_bt.c  | 16 +++++++++++++---
>>  include/uapi/linux/virtio_bt.h |  8 ++++++++
>>  2 files changed, 21 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/bluetooth/virtio_bt.c b/drivers/bluetooth/virtio_bt.c
>> index 67c21263f9e0..35f8041722c8 100644
>> --- a/drivers/bluetooth/virtio_bt.c
>> +++ b/drivers/bluetooth/virtio_bt.c
>> @@ -306,7 +306,12 @@ static int virtbt_probe(struct virtio_device *vdev)
>>         if (virtio_has_feature(vdev, VIRTIO_BT_F_VND_HCI)) {
>>                 __u16 vendor;
>>
>> -               virtio_cread(vdev, struct virtio_bt_config, vendor, &vendor);
>> +               if (virtio_has_feature(vdev, VIRTIO_BT_F_CONFIG_V2))
>> +                       virtio_cread(vdev, struct virtio_bt_config_v2,
>> +                                    vendor, &vendor);
>> +               else
>> +                       virtio_cread(vdev, struct virtio_bt_config,
>> +                                    vendor, &vendor);
>>
>>                 switch (vendor) {
>>                 case VIRTIO_BT_CONFIG_VENDOR_ZEPHYR:
>> @@ -339,8 +344,12 @@ static int virtbt_probe(struct virtio_device *vdev)
>>         if (virtio_has_feature(vdev, VIRTIO_BT_F_MSFT_EXT)) {
>>                 __u16 msft_opcode;
>>
>> -               virtio_cread(vdev, struct virtio_bt_config,
>> -                            msft_opcode, &msft_opcode);
>> +               if (virtio_has_feature(vdev, VIRTIO_BT_F_CONFIG_V2))
>> +                       virtio_cread(vdev, struct virtio_bt_config_v2,
>> +                                    msft_opcode, &msft_opcode);
>> +               else
>> +                       virtio_cread(vdev, struct virtio_bt_config,
>> +                                    msft_opcode, &msft_opcode);
>>
>>                 hci_set_msft_opcode(hdev, msft_opcode);
>>         }
>> @@ -387,6 +396,7 @@ static const unsigned int virtbt_features[] = {
>>         VIRTIO_BT_F_VND_HCI,
>>         VIRTIO_BT_F_MSFT_EXT,
>>         VIRTIO_BT_F_AOSP_EXT,
>> +       VIRTIO_BT_F_CONFIG_V2,
>>  };
> 
> So this introduces a new flag which must be checked when attempting to
> config, right? But is this backward compatible? What happens if for
> some reason the userspace doesn't use the new struct are we able to
> detect that?

Yes, it's backwards compatible.

[q]Each virtio device offers all the features it understands. During
device initialization, the driver reads this and tells the device the
subset that it accepts. The only way to renegotiate is to reset the device.
This allows for forwards and backwards compatibility: if the device is
enhanced with a new feature bit, older drivers will not write that
feature bit back to the device. Similarly, if a driver is enhanced with
a feature that the device doesn’t support, it see the new feature is not
offered.[/q]

So, in our case:

old device - new driver:
The device does not offer VIRTIO_BT_F_CONFIG_V2 feature and uses the old
configuration structure.
The driver also uses the old configuration structure because
VIRTIO_BT_F_CONFIG_V2 bit was not negotiated.

new device - old driver:
The device offers this bit, the driver reads it but cannot support it,
so it does not write this bit back to the device during feature negotiation.
The device verifies that this bit is not negotiated and continues to use
the old configuration structure.


I tested this patch, it
a) works fine with a new device that supports VIRTIO_BT_F_CONFIG_V2.
b) uses the old configuration structure when working with an old device.
   Our device does not offer the VIRTIO_BT_F_VND_HCI feature bit, so the
driver does not tries to read unaligned "vendor" and "msft_opcode"
fields and everything is fine.
But, if the VIRTIO_BT_F_VND_HCI feature is set for the device for test
purposes, our middle layer asserts unaligned accesses to the
configuration space.

P.S. But, as Michael S. Tsirkin rightly stated, [q]Will a spec patch be
forthcoming?[/q], this patch requires a specification update.
I could not find any virtio_bt specification, do you have one?
That would be great. Otherwise, would you mind if I try to write some
initial draft?
>>  static struct virtio_driver virtbt_driver = {
>> diff --git a/include/uapi/linux/virtio_bt.h b/include/uapi/linux/virtio_bt.h
>> index a7bd48daa9a9..af798f4c9680 100644
>> --- a/include/uapi/linux/virtio_bt.h
>> +++ b/include/uapi/linux/virtio_bt.h
>> @@ -9,6 +9,7 @@
>>  #define VIRTIO_BT_F_VND_HCI    0       /* Indicates vendor command support */
>>  #define VIRTIO_BT_F_MSFT_EXT   1       /* Indicates MSFT vendor support */
>>  #define VIRTIO_BT_F_AOSP_EXT   2       /* Indicates AOSP vendor support */
>> +#define VIRTIO_BT_F_CONFIG_V2  3       /* Use second version configuration */
>>
>>  enum virtio_bt_config_type {
>>         VIRTIO_BT_CONFIG_TYPE_PRIMARY   = 0,
>> @@ -28,4 +29,11 @@ struct virtio_bt_config {
>>         __u16 msft_opcode;
>>  } __attribute__((packed));
>>
>> +struct virtio_bt_config_v2 {
>> +       __u8  type;
>> +       __u8  alignment;
>> +       __u16 vendor;
>> +       __u16 msft_opcode;
>> +};
>> +
>>  #endif /* _UAPI_LINUX_VIRTIO_BT_H */
>> --
>> 2.37.2