RE: [PATCH V2 2/2] mailbox: introduce ARM SMC based mailbox

[Peng Fan]

Hi Andre,

> Subject: RE: [PATCH V2 2/2] mailbox: introduce ARM SMC based mailbox
> 
> Hi Andre,
> > Subject: Re: [PATCH V2 2/2] mailbox: introduce ARM SMC based mailbox
> >
> > On Mon, 3 Jun 2019 09:32:42 -0700
> > Florian Fainelli <f.fainelli@xxxxxxxxx> wrote:
> >
> > Hi,
> >
> > > On 6/3/19 1:30 AM, peng.fan@xxxxxxx wrote:
> > > > From: Peng Fan <peng.fan@xxxxxxx>
> > > >
> > > > This mailbox driver implements a mailbox which signals transmitted
> > > > data via an ARM smc (secure monitor call) instruction. The mailbox
> > > > receiver is implemented in firmware and can synchronously return
> > > > data when it returns execution to the non-secure world again.
> > > > An asynchronous receive path is not implemented.
> > > > This allows the usage of a mailbox to trigger firmware actions on
> > > > SoCs which either don't have a separate management processor or on
> > > > which such a core is not available. A user of this mailbox could
> > > > be the SCP interface.
> > > >
> > > > Modified from Andre Przywara's v2 patch
> > > > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2F
> > > > lo
> > > >
> >
> re.kernel.org%2Fpatchwork%2Fpatch%2F812999%2F&amp;data=02%7C01%
> > 7Cpen
> > > >
> >
> g.fan%40nxp.com%7C15c4180b8fe5405d3de808d6ea81d5f1%7C686ea1d3bc
> > 2b4c6
> > > >
> > fa92cd99c5c301635%7C0%7C0%7C636954240720601454&amp;sdata=1Cp
> > WSgTH7lF
> > > > cBKxJnLeIDw%2FDAQJJO%2FVypV1LUU1BRQA%3D&amp;reserved=0
> > > >
> > > > Cc: Andre Przywara <andre.przywara@xxxxxxx>
> > > > Signed-off-by: Peng Fan <peng.fan@xxxxxxx>
> > > > ---
> > >
> > > [snip]
> > >
> > > +#define ARM_SMC_MBOX_USB_IRQ	BIT(1)
> > >
> > > That flag appears unused.
> > >
> > > > +static int arm_smc_mbox_probe(struct platform_device *pdev) {
> > > > +	struct device *dev = &pdev->dev;
> > > > +	struct mbox_controller *mbox;
> > > > +	struct arm_smc_chan_data *chan_data;
> > > > +	const char *method;
> > > > +	bool use_hvc = false;
> > > > +	int ret, irq_count, i;
> > > > +	u32 val;
> > > > +
> > > > +	if (!of_property_read_u32(dev->of_node, "arm,num-chans", &val)) {
> > > > +		if (val < 1 || val > INT_MAX) {
> > > > +			dev_err(dev, "invalid arm,num-chans value %u
> > of %pOFn\n", val,
> > > > +pdev->dev.of_node);
> >
> > Isn't the of_node parameter redundant, because dev_err() already takes
> > care of that?
> 
> I'll remove that.
> 
> >
> > > > +			return -EINVAL;
> > > > +		}
> > > > +	}
> > >
> > > Should not the upper bound check be done against UINT_MAX since val
> > > is an unsigned int?
> >
> > But wouldn't that be somewhat pointless, given that val is a u32? So I
> > guess we could just condense this down to:
> > ...
> > 		if (!val) {
> > ...
> 
> make sense.
> 
> >
> > > > +
> > > > +	irq_count = platform_irq_count(pdev);
> > > > +	if (irq_count == -EPROBE_DEFER)
> > > > +		return irq_count;
> > > > +
> > > > +	if (irq_count && irq_count != val) {
> > > > +		dev_err(dev, "Interrupts not match num-chans\n");
> > >
> > > Interrupts property does not match \"arm,num-chans\" would be more
> > correct.
> >
> > Given that interrupts are optional, do we have to rely on this?
> 
> If there is interrupt property, the interrupts should match channel counts.
> 
> Do we actually
> > need one interrupt per channel?
> 
> I thought about this, provide one interrupt for all channels.
> But there is no good way to let interrupt handlers know which channel
> triggers the interrupt. So I use one interrupt per channel.
> 
> >
> > > > +		return -EINVAL;
> > > > +	}
> > > > +
> > > > +	if (!of_property_read_string(dev->of_node, "method", &method)) {
> > > > +		if (!strcmp("hvc", method)) {
> > > > +			use_hvc = true;
> > > > +		} else if (!strcmp("smc", method)) {
> > > > +			use_hvc = false;
> > > > +		} else {
> > > > +			dev_warn(dev, "invalid \"method\" property: %s\n",
> > > > +				 method);
> > > > +
> > > > +			return -EINVAL;
> > > > +		}
> > >
> > > Having at least one method specified does not seem to be checked
> > > later on in the code, so if I omitted to specify that property, we
> > > would still register the mailbox and default to use "smc" since the
> > > ARM_SMC_MBOX_USE_HVC flag would not be set, would not we want to
> > make
> > > sure that we do have in fact a valid method specified given the
> > > binding documents that property as mandatory?
> > >
> > > [snip]
> > >
> > > > +	mbox->txdone_poll = false;
> > > > +	mbox->txdone_irq = false;
> > > > +	mbox->ops = &arm_smc_mbox_chan_ops;
> > > > +	mbox->dev = dev;
> > > > +
> > > > +	ret = mbox_controller_register(mbox);
> > > > +	if (ret)
> > > > +		return ret;
> > > > +
> > > > +	platform_set_drvdata(pdev, mbox);
> > >
> > > I would move this above mbox_controller_register() that way there is
> > > no room for race conditions in case another part of the driver
> > > expects to have pdev->dev.drvdata set before the mbox controller is
> registered.
> > > Since you use devm_* functions for everything, you may even remove
> > > that call.
> > >
> > > [snip]
> > >
> > > > +#ifndef _LINUX_ARM_SMC_MAILBOX_H_ #define
> > > > +_LINUX_ARM_SMC_MAILBOX_H_
> > > > +
> > > > +struct arm_smccc_mbox_cmd {
> > > > +	unsigned long a0, a1, a2, a3, a4, a5, a6, a7; };
> > >
> > > Do you expect this to be used by other in-kernel users? If so, it
> > > might be good to document how a0 can have a special meaning and be
> > > used as a substitute for the function_id?
> >
> > I don't think we should really expose this outside of the driver. From
> > a mailbox point of view this is just the payload, transported according to the
> SMCCC.
> > Also using "long" here sounds somewhat troublesome.

Long on ARM64 is 64bit, and 32bit on ARM32, so I use long.
Do you forsee any issues?

> >
> > Also, looking at the SMCCC, I only see six parameters in addition to
> > the function identifier. Shall we reflect this here?

a0 is used as function id, not no arm,func-ids provided in dts. a1-a7 are
also passed to smc.
If arm,func-ids is provided, a0 will be omitted just for consistency as above.

You mean write comments in the code for it?

Thanks,
Peng.

> 
> I could move it to driver code. Jassi, do you have any comments?
> 
> Thanks,
> Peng.
> 
> >
> > Cheers,
> > Andre.