common bmc mailbox driver

common bmc mailbox driver

[Eugene.Cho]

Hey Cyril,
   I was recently following your upstream submission attempt for the aspeed mbox/mailbox, was wondering if there was been any progress since?

Reason I ask - is Nuvoton also has a "mailbox" for bmc<->host communication, which does the same thing (shared memory + interrupts both ways), and was wondering if we could just have 1 driver for both, instead 2 separate drivers.

Glancing at the aspeed driver, only key difference I see is the setting and handling the interrupt. Thoughts on this?

Re: common bmc mailbox driver

[Andrew Jeffery]

Hi Eugene,

On Thu, 10 May 2018, at 22:34, Eugene.Cho@dell.com wrote:
> Hey Cyril,
>    I was recently following your upstream submission attempt for the 
> aspeed mbox/mailbox, was wondering if there was been any progress since?

FYI Cyril's moved on from IBM and OpenBMC, and so may have lost some motivation here.

> 
> Reason I ask - is Nuvoton also has a "mailbox" for bmc<->host 
> communication, which does the same thing (shared memory + interrupts 
> both ways), and was wondering if we could just have 1 driver for both, 
> instead 2 separate drivers.
> 
> Glancing at the aspeed driver, only key difference I see is the setting 
> and handling the interrupt. Thoughts on this?

(Somewhat unfortunately) the ASPEED mailbox driver isn't upstream yet due to push-back from the upstream mailbox subsystem maintainer*. Similar to the misc driver it sounds like we should work together to get things moving.

* It turns out the mailbox subsystem as it stands doesn't fit our needs - it's written from the perspective of the host kernel communicating with some firmware black box on the other side of the mailbox, rather than the kernel of interest *being* the black box.

Andrew

Re: common bmc mailbox driver

[Cyril Bur]

On 11 May 2018 at 01:59, Andrew Jeffery <andrew@aj.id.au> wrote:
> Hi Eugene,
>
> On Thu, 10 May 2018, at 22:34, Eugene.Cho@dell.com wrote:
>> Hey Cyril,
>>    I was recently following your upstream submission attempt for the
>> aspeed mbox/mailbox, was wondering if there was been any progress since?
>
> FYI Cyril's moved on from IBM and OpenBMC, and so may have lost some motivation here.
>

Hi, sorry for the delayed response. Andrew is correct here.
Unfortunately I won't be continuing the upstreaming efforts for the
driver as I won't have the time to rewrite it thrice more but more
importantly I don't have the ability to do anything more than a
compile test.

You are more than welcome to take whatever is useful or was useful and
do with it whatever you want.

>>
>> Reason I ask - is Nuvoton also has a "mailbox" for bmc<->host
>> communication, which does the same thing (shared memory + interrupts
>> both ways), and was wondering if we could just have 1 driver for both,
>> instead 2 separate drivers.
>>
>> Glancing at the aspeed driver, only key difference I see is the setting
>> and handling the interrupt. Thoughts on this?
>
> (Somewhat unfortunately) the ASPEED mailbox driver isn't upstream yet due to push-back from the upstream mailbox subsystem maintainer*. Similar to the misc driver it sounds like we should work together to get things moving.
>
> * It turns out the mailbox subsystem as it stands doesn't fit our needs - it's written from the perspective of the host kernel communicating with some firmware black box on the other side of the mailbox, rather than the kernel of interest *being* the black box.
>

I wish you all the best of luck overcoming these     problems.

> Andrew



-- 
Cyril

p.s. Apologies if this comes through as html or anything strange -
I've resorted to using the gmail web interface.