lkml.org 
[lkml]   [2017]   [May]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH v3 2/4] soc: qcom: Introduce APCS IPC driver
On Thu, May 4, 2017 at 11:15 AM, Bjorn Andersson
<bjorn.andersson@linaro.org> wrote:
> On Wed 03 May 02:55 PDT 2017, Jassi Brar wrote:
>
>> Loic, thanks for adding me.
>>
>> On Wed, May 3, 2017 at 2:58 PM, Loic PALLARDY <loic.pallardy@st.com> wrote:
>> >
>> >
>> >> -----Original Message-----
>> >> From: linux-remoteproc-owner@vger.kernel.org [mailto:linux-remoteproc-
>> >> owner@vger.kernel.org] On Behalf Of Bjorn Andersson
>> >> Sent: Wednesday, May 03, 2017 7:29 AM
>> >> To: Andy Gross <andy.gross@linaro.org>; Rob Herring
>> >> <robh+dt@kernel.org>; Mark Rutland <mark.rutland@arm.com>; Ohad Ben-
>> >> Cohen <ohad@wizery.com>
>> >> Cc: linux-arm-msm@vger.kernel.org; linux-soc@vger.kernel.org;
>> >> devicetree@vger.kernel.org; linux-kernel@vger.kernel.org; linux-
>> >> remoteproc@vger.kernel.org
>> >> Subject: [PATCH v3 2/4] soc: qcom: Introduce APCS IPC driver
>> >>
>> >> This implements a driver that exposes the IPC bits found in the APCS Global
>> >> block in various Qualcomm platforms. The bits are used to signal inter-
>> >> processor communication signals from the application CPU to other masters.
>> >>
>> >> The driver implements the "doorbell" binding and could be used as basis for a
>> >> new Linux framework, if found useful outside Qualcomm.
>> >>
>> > Hi Bjorn,
>> >
>> > Even if Qualcom APCS IPC is limited, why don't you rely on existing mailbox framework.
>> > It is there to gather all IPC management under the same interface.
>> > No need to create a new one from my pov.
>> > If you don't provide message data, mailbox framework behaves as doorbell.
>> >
>> QCOM RPM reinvented the wheel for what mailbox framework already did,
>> despite my pointing it out =>
>> http://lkml.iu.edu/hypermail//linux/kernel/1406.2/03918.html
>>
>
> The RPM interface works by writing various information in shared DRAM
> and then invoking an interrupt on the remote processor.
>
This is what _most_ platforms do, and they use mailbox framework. If
mailbox framework doesn't suit RPM, then it doesn't suit most
platforms that use it.

> My argumentation against using the mailbox framework for the RPM case
> was that the message model is a software construct and doesn't fit the
> mailbox framework.
>
Mailbox framework works beneath protocol layer (software construct).
As I said years ago, the h/w bits of the controller should be under
mailbox api, while the message structures and signals are protocol
specific and be in QCOM specific location.


> Before "inventing" the function for acquiring a handle to a doorbell I
> did look at making this a mailbox controller, but as each mailbox
> channel related to a single bit and 1 is the only value we'll ever write
> it didn't feel like it matches the mailbox expectations. But as you seem
> to object I'll attempt to rewrite it using the mailbox framework
> instead.
>
Yes, please. 1-bit messages are just a special case of N-bits messages
that mailbox framework supports.


> But which one of these would be appropriate for a "mailbox channel" that
> doesn't have any actual messages?
>
> mbox_send_message(chan, NULL)
> or
> const int one = 1;
> mbox_send_message(chan, (void*)&one);
>
It depends upon your h/w.
If each physical channel is hardwired to work on a predefined single
bit, then the driver could do without that bit being explicitly
passed.
If a physical channel can be mapped onto any bit(s), then you do need
to pass that info via mbox_send_message().


>> The driver bypassed mailbox framework and was pushed via another tree.
>> Same is being attempted now, only now it is more expensive to switch
>> to generic mailbox framework having spent so much time on QCOM
>> specific implementation of controller and protocol drivers inside
>> drivers/soc/qcom/
>
> I'm not sure I follow this, there's no extensive rework going on here -
> all I'm trying to do is abstract the "writel(BIT(x), ipc_reg)" line of
> the RPM driver, because it's being used in other clients as well.
>
No, I meant ideally RPM should have used mailbox api for the
controller programming, but moving to that now will be expensive
because you already developed huge code base around QCOM specific
mailbox implementation.

Regards.

\
 
 \ /
  Last update: 2017-05-04 09:55    [W:0.831 / U:0.012 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site