linux-kernel - Re: [PATCH] mailbox: add support for doorbell/signal mode controllers

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <59a05fcb-ff30-0683-144e-93521a7413f9@arm.com>
Date:   Wed, 1 Nov 2017 18:15:39 +0000
From:   Sudeep Holla <sudeep.holla@....com>
To:     Jassi Brar <jassisinghbrar@...il.com>
Cc:     Sudeep Holla <sudeep.holla@....com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        Arnd Bergmann <arnd@...db.de>,
        Bjorn Andersson <bjorn.andersson@...aro.org>
Subject: Re: [PATCH] mailbox: add support for doorbell/signal mode controllers



On 01/11/17 18:03, Jassi Brar wrote:
> On Wed, Nov 1, 2017 at 10:02 PM, Sudeep Holla <sudeep.holla@....com> wrote:
> 
>>
>> Such controllers don't need to transmit any data, they just transmit
>> the signal. In such controllers the data pointer passed to
>> mbox_send_message is passed to client via it's tx_prepare callback.
>> Controller doesn't need any data to be passed from the client.
>>
> Some controllers need a non-zero value written to a register in order
> to trigger the signal.

You are right, just right non-zero or whatever controller value to
trigger the interrupt to remote.

> That register is visible to the remote. While the data/packet is setup
> during tx_prepare() callback.

Agreed.

> You are overlooking this class of doorbell controllers.
>

Not sure what do you mean by that ?

>>
>> This is rough idea I have on extending mailbox interface to support
>> the doorbell requirements.
>>
> What doorbell requirements does the api not support?
> QComm's APCS IPC is what you call a "doorbell" controller and is
> already supported by the API. It could run SCMI even easier than MHU
> (your controller).
> 

Again agreed. But see below for reason to create this API.

>> The new API send_signal will eliminate the
>> issue Jassi has explained in earlier discussion with respect to generic
>> message format using Rockchip example.
>>
> Sorry I don't see how.
> Please explain how can send_signal() api be used by, say, rockchip to
> support SCMI?
> 

 80         writel_relaxed(msg->cmd, mb->mbox_base +
MAILBOX_A2B_CMD(chans->idx));
 81         writel_relaxed(msg->rx_size, mb->mbox_base +

 82                        MAILBOX_A2B_DAT(chans->idx));

 83

 will be replaced with

writel(whatever_value_to trigger_signal, MAILBOX_A2B_CMD(chans->idx));

in its send_signal function.

> I am not convinced we should clone an api just so that a client driver
> becomes simpler. Esp when it shifts, and not avoid, the additional
> code (to support the client) onto the provider side.
> 

It doesn't tie the data format with particular mailbox controller.
send_data has void *data and the interpretation is controller specific.
send_signal on the other handle can implemented by the controllers which
knows how and can trigger the specific signal to the remote.
-- 
Regards,
Sudeep