RE: RE: [PATCH v3 1/2] can: m_can: Move mram init to mcan device setup

From: Vivek Yadav
Date: Wed Nov 30 2022 - 23:37:21 EST




> -----Original Message-----
> From: Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx>
> Sent: 24 November 2022 20:24
> To: Vivek Yadav <vivek.2311@xxxxxxxxxxx>
> Cc: rcsekar@xxxxxxxxxxx; krzysztof.kozlowski+dt@xxxxxxxxxx;
> wg@xxxxxxxxxxxxxx; davem@xxxxxxxxxxxxx; edumazet@xxxxxxxxxx;
> kuba@xxxxxxxxxx; pabeni@xxxxxxxxxx; pankaj.dubey@xxxxxxxxxxx;
> ravi.patel@xxxxxxxxxxx; alim.akhtar@xxxxxxxxxxx; linux-fsd@xxxxxxxxx;
> robh+dt@xxxxxxxxxx; linux-can@xxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx;
> linux-kernel@xxxxxxxxxxxxxxx; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; linux-
> samsung-soc@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx;
> aswani.reddy@xxxxxxxxxxx; sriranjani.p@xxxxxxxxxxx
> Subject: Re: RE: [PATCH v3 1/2] can: m_can: Move mram init to mcan device
> setup
>
> On 24.11.2022 14:36:48, Vivek Yadav wrote:
> > > Why not call the RAM init directly from m_can_chip_config()?
> > >
> > m_can_chip_config function is called from m_can open.
> >
> > Configuring RAM init every time we open the CAN instance is not
> > needed, I think only once during the probe is enough.
>
> That probably depends on you power management. If I add a regulator to
> power the external tcan4x5x chip and power it up during open(), I need to
> initialize the RAM.
>
Thanks for the clarification,
There is one doubt for which I need clarity if I add ram init in m_can_chip_config.

In the current implementation, m_can_ram_init is added in the probe and m_can_class_resume function.
If I add the ram init function in chip_config which is getting called from m_can_start, then m_can_init_ram will be called two times, once in resume and next from m_can_start also.

Can we add ram init inside the m_can_open function itself?
Because it is independent of m_can resume functionality.

> > If message RAM init failed then fifo Transmit and receive will fail
> > and there will be no communication. So there is no point to "open and
> > Configure CAN chip".
>
> For mmio devices the RAM init will probably not fail. There are return values
> and error checking for the SPI attached devices. Where the SPI
> communication will fail. However if this is problem, I assume the chip will not
> be detected in the first place.
>
> > From my understanding it's better to keep RAM init inside the probe
> > and if there is a failure happened goes to CAN probe failure.
>
> Marc
>
> --
> Pengutronix e.K. | Marc Kleine-Budde |
> Embedded Linux |
> https://protect2.fireeye.com/v1/url?k=2053d9ab-7fc8e0b4-205252e4-
> 000babdfecba-a8c309c53e3358f5&q=1&e=c0cfd0e2-a422-4821-a49d-
> 113cfa4da9cb&u=https%3A%2F%2Fwww.pengutronix.de%2F |
> Vertretung West/Dortmund | Phone: +49-231-2826-924 |
> Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |