Re: [PATCH v5 1/3] dt-bindings: mailbox: add google,gs101-mbox

[Tudor Ambarus <tudor.ambarus@linaro.org>]

Hi, Jassi,

On 1/3/25 3:39 AM, Jassi Brar wrote:
>>> Looking at v6, I prefer this version... maybe modify it a bit.
>>
>> Just to summarize for the readers, in the end I chose for the
>> controllers to allow #mbox-cells = <0>; and for the clients to still use
>> the mboxes property, but just to reference the phandle to the controller:
>>         mboxes = <&ap2apm_mailbox>;
>>
> This was already supported, see drivers/mailbox/bcm2835-mailbox.c for example.

Thanks for the pointer. I was referring to the bindings patch:
https://lore.kernel.org/linux-arm-kernel/20241220-acpm-v4-upstream-mbox-v6-1-a6942806e52a@linaro.org/
> 
>> Then I updated the mailbox core to allow clients to request channels by
>> passing some args containing channel identifiers to the controllers,
>> that the controllers xlate() using their own method.
>>
> This is unnecessary.
> If you don't pass the doorbell number from DT, each channel populated
> by the driver is just a s/w construct or a 'virtual' channel. Make use
> of 'void *data'  in send_data() to specify the doorbell.
> 

I think this introduces concurrency problems if the channel identifiers
passed by 'void *data' don't match the virtual channel used for sending
the messages. Do we want to allow this?

Also, if we use 'void *data' to pass channel identifiers, the channel
checks will have to be made at send_data() time. Thus if passing wrong
channel type for example, the mailbox client will eventually get a
-ENOBUFS and a "Try increasing MBOX_TX_QUEUE_LEN" message, which I find
misleading.

Thanks,
ta