Hi Russell,
Thanks for pointing this. I based on pinctrl-armada-xp.c (it needs a
fix then, too) and it worked. I must have missed, because I got proper
registers' number and values in suspend/resume routines. As
pinctrl-armada-xp.c needs also a small fix and in order not to
duplicate code, how about a following solution:
- *mpp_saved_regs and *mpp_base become members of struct mvebu_pinctrl_soc_info
- common mvebu_pinctrl_suspend/resume functions in pinctrl-mvebu.c
(now there will be two users AXP and A38X)
Please let me know what you think.
Best regards,
Marcin
2015-10-18 0:29 GMT+02:00 Russell King - ARM Linux <linux at arm.linux.org.uk>:
> On Sat, Oct 17, 2015 at 11:28:48PM +0200, Marcin Wojtas wrote:
>> diff --git a/drivers/pinctrl/mvebu/pinctrl-armada-38x.c b/drivers/pinctrl/mvebu/pinctrl-armada-38x.c
>> index 6ec82c6..094cb48 100644
>> --- a/drivers/pinctrl/mvebu/pinctrl-armada-38x.c
>> +++ b/drivers/pinctrl/mvebu/pinctrl-armada-38x.c
>> @@ -23,6 +23,7 @@
>> #include "pinctrl-mvebu.h"
>>
>> static void __iomem *mpp_base;
>> +static u32 *mpp_saved_regs;
>
> I'm not a fan of unnecessary global variables. It adds to the bulk of
> the kernel when built-in (even though it's in .bss, it still has a cost)
> and these all add up when built-in.
>
> Please make it part of the driver data allocated at probe time.
>
>> static int armada_38x_mpp_ctrl_get(unsigned pid, unsigned long *config)
>> {
>> @@ -424,6 +425,7 @@ static int armada_38x_pinctrl_probe(struct platform_device *pdev)
>> const struct of_device_id *match =
>> of_match_device(armada_38x_pinctrl_of_match, &pdev->dev);
>> struct resource *res;
>> + int nregs;
>>
>> if (!match)
>> return -ENODEV;
>> @@ -441,11 +443,44 @@ static int armada_38x_pinctrl_probe(struct platform_device *pdev)
>> soc->modes = armada_38x_mpp_modes;
>> soc->nmodes = armada_38x_mpp_controls[0].npins;
>>
>> + nregs = DIV_ROUND_UP(soc->nmodes, MVEBU_MPPS_PER_REG);
>> +
>> + mpp_saved_regs = devm_kcalloc(&pdev->dev, nregs, sizeof(u32),
>> + GFP_KERNEL);
>> + if (!mpp_saved_regs)
>> + return -ENOMEM;
>> +
>> pdev->dev.platform_data = soc;
>
> The 'soc' is stored in platform data, not driver data, but...
>
>>
>> return mvebu_pinctrl_probe(pdev);
>> }
>>
>> +int armada_38x_pinctrl_suspend(struct platform_device *pdev, pm_message_t state)
>> +{
>> + struct mvebu_pinctrl_soc_info *soc = platform_get_drvdata(pdev);
>
> You access it through driver data. Isn't the driver data here a
> struct mvebu_pinctrl pointer? See platform_set_drvdata() in
> mvebu_pinctrl_probe().
>
>> + int i, nregs;
>> +
>> + nregs = DIV_ROUND_UP(soc->nmodes, MVEBU_MPPS_PER_REG);
>> +
>> + for (i = 0; i < nregs; i++)
>> + mpp_saved_regs[i] = readl(mpp_base + i * 4);
>> +
>> + return 0;
>> +}
>> +
>> +int armada_38x_pinctrl_resume(struct platform_device *pdev)
>> +{
>> + struct mvebu_pinctrl_soc_info *soc = platform_get_drvdata(pdev);
>
> Ditto.
>
> --
> FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
> according to speedtest.net.