Re: [PATCH v5 3/3] drivers:power:twl4030-charger: add deferred probing for phy and iio

[Sebastian Reichel]

Hi,

On Sun, May 21, 2017 at 12:38:18PM +0200, H. Nikolaus Schaller wrote:
> This fixes an issue if both this twl4030_charger driver and
> phy-twl4030-usb are compiled as modules and loaded in random order.
> It has been observed on GTA04 and OpenPandora devices that in worst
> case the boot process hangs and in best case the AC detection fails
> with a warning.
>
> Therefore we add deferred probing checks for the usb_phy and the
> iio channel for AC detection.
> 
> For implementing this we must reorder code because we can't safely
> return -EPROBE_DEFER after allocating any devm managed interrupt
> (it might already/still be enabled without working interrupt handler).
>
> So the check for required resources that may abort probing by
> returning -EPROBE_DEFER, must come first.

This sounds fishy. EPROBE_DEFER should not be different from
other error codes in this regard and devm_ requested resources
should be free'd on any error. Why is irq handler not working?

> Signed-off-by: H. Nikolaus Schaller <hns@xxxxxxxxxxxxx>
> ---
>  drivers/power/supply/twl4030_charger.c | 37 ++++++++++++++++++++--------------
>  1 file changed, 22 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/power/supply/twl4030_charger.c b/drivers/power/supply/twl4030_charger.c
> index 785a07bc4f39..945eabdbbc89 100644
> --- a/drivers/power/supply/twl4030_charger.c
> +++ b/drivers/power/supply/twl4030_charger.c
> @@ -984,6 +984,28 @@ static int twl4030_bci_probe(struct platform_device *pdev)
>  
>  	platform_set_drvdata(pdev, bci);
>  
> +	if (bci->dev->of_node) {
> +		struct device_node *phynode;
> +
> +		phynode = of_find_compatible_node(bci->dev->of_node->parent,
> +						  NULL, "ti,twl4030-usb");
> +		if (phynode) {
> +			bci->transceiver = devm_usb_get_phy_by_node(
> +				bci->dev, phynode, &bci->usb_nb);
> +			if (IS_ERR(bci->transceiver) &&
> +			    PTR_ERR(bci->transceiver) == -EPROBE_DEFER)
> +				return -EPROBE_DEFER;	/* PHY not ready */
> +		}
> +	}
> +
> +	bci->channel_vac = iio_channel_get(&pdev->dev, "vac");
> +	if (IS_ERR(bci->channel_vac)) {
> +		if (PTR_ERR(bci->channel_vac) == -EPROBE_DEFER)
> +			return -EPROBE_DEFER;	/* iio not ready */
> +		dev_warn(&pdev->dev, "could not request vac iio channel");
> +		bci->channel_vac = NULL;
> +	}
> +

You should not request non-devm managed iio_channel before
devm-managed power-supply. This could be fixed by switching to
devm_iio_channel_get(), which also cleans up some code.

I suspect, that this is also your IRQ problem, since iio_channel
is currently free'd before irqs are free'd, but its used in irq
code.

>  	bci->ac = devm_power_supply_register(&pdev->dev, &twl4030_bci_ac_desc,
>  					     NULL);
>  	if (IS_ERR(bci->ac)) {
> @@ -1017,25 +1039,10 @@ static int twl4030_bci_probe(struct platform_device *pdev)
>  		return ret;
>  	}
>  
> -	bci->channel_vac = iio_channel_get(&pdev->dev, "vac");
> -	if (IS_ERR(bci->channel_vac)) {
> -		bci->channel_vac = NULL;
> -		dev_warn(&pdev->dev, "could not request vac iio channel");
> -	}
> -
>  	INIT_WORK(&bci->work, twl4030_bci_usb_work);
>  	INIT_DELAYED_WORK(&bci->current_worker, twl4030_current_worker);
>  
>  	bci->usb_nb.notifier_call = twl4030_bci_usb_ncb;
> -	if (bci->dev->of_node) {
> -		struct device_node *phynode;
> -
> -		phynode = of_find_compatible_node(bci->dev->of_node->parent,
> -						  NULL, "ti,twl4030-usb");
> -		if (phynode)
> -			bci->transceiver = devm_usb_get_phy_by_node(
> -				bci->dev, phynode, &bci->usb_nb);
> -	}
>  
>  	/* Enable interrupts now. */
>  	reg = ~(u32)(TWL4030_ICHGLOW | TWL4030_ICHGEOC | TWL4030_TBATOR2 |

-- Sebastian

Attachment: signature.asc
Description: PGP signature