Messages in this thread | | | Subject | Re: [PATCH v2] power: reset: at91-shdwc: add new shutdown controller driver | From | Nicolas Ferre <> | Date | Mon, 28 Sep 2015 11:44:28 +0200 |
| |
Le 24/09/2015 13:06, Alexandre Belloni a écrit : > On 22/09/2015 at 17:57:52 +0200, Nicolas Ferre wrote : >> +config POWER_RESET_AT91_SHDWC > > Maybe we should start numbering, at91-poweroff is already a driver for > an shdwc. > >> +#define SLOW_CLOCK_FREQ 32768 >> + > > The slow clock is an input to the shdwc, then you have to get it, > prepare and enable it. The you could use clk_get_rate.
Okay, sure: I modify the driver to take your enhancements into account.
> By duplicating the driver, you are repeating mistakes from the past ;)
Hey! look at the common parts that these drivers would share if it was merged and you would realize that it was not worth the effort: I tried and judged it was completely pointless.
>> +static void at91_poweroff(void) >> +{ >> + BUG_ON(!at91_shdwc); > > Is that condition even possible?
No it's not. I remove it.
> >> + >> + writel(AT91_SHDW_KEY | AT91_SHDW_SHDW, >> + at91_shdwc->at91_shdwc_base + AT91_SHDW_CR); > > This is not properly aligned.
Why? I prefer it like this and the second line doesn't cross the 80 column so I presume that it's valid.
> >> +} >> + >> +static u32 at91_shdwc_debouncer_value(struct platform_device *pdev, >> + u32 in_period_us) >> +{ >> + int i; >> + int max_idx = ARRAY_SIZE(sdwc_dbc_period) - 1; >> + unsigned long long period_us; >> + unsigned long long max_period_us = DBC_PERIOD_US(sdwc_dbc_period[max_idx]); >> + >> + if (in_period_us > max_period_us) { >> + dev_warn(&pdev->dev, >> + "debouncer period %u too big, reduced to %llu us\n", >> + in_period_us, max_period_us); >> + return max_idx; >> + } >> + >> + for (i = max_idx - 1; i > 0; i--) { >> + period_us = DBC_PERIOD_US(sdwc_dbc_period[i]); >> + dev_dbg(&pdev->dev, "%s: ref[%d] = %llu\n", >> + __func__, i, period_us); > ditto > > >> + if (in_period_us > period_us) >> + break; >> + } >> + >> + return i + 1; >> +} >> + >> +static u32 at91_shdwc_get_wakeup_input(struct platform_device *pdev, >> + struct device_node *np) >> +{ >> + struct device_node *cnp; >> + const char *pm; >> + u32 wk_input_mask; >> + u32 wuir = 0; >> + u32 wk_input; >> + >> + for_each_child_of_node(np, cnp) { >> + if (of_property_read_u32(cnp, "reg", &wk_input)) { >> + dev_warn(&pdev->dev, "reg property is missing for %s\n", >> + cnp->full_name); >> + continue; >> + } >> + >> + wk_input_mask = 1 << wk_input; >> + if (!(wk_input_mask & AT91_SHDW_WKUPEN_MASK)) { >> + dev_warn(&pdev->dev, >> + "wake-up input %d out of bounds ignore\n", >> + wk_input); >> + continue; >> + } >> + wuir |= wk_input_mask; >> + >> + if (!of_property_read_string(cnp, "atmel,wakeup-type", &pm)) { >> + if (!strcasecmp(pm, "high")) >> + /* >> + * only add a type if "high" is specified. Low >> + * is the default. >> + */ >> + wuir |= AT91_SHDW_WKUPT(wk_input); >> + } >> + dev_dbg(&pdev->dev, "%s: (child %d) wuir = %#x\n", >> + __func__, wk_input, wuir); > > ditto > >> +static int at91_shdwc_probe(struct platform_device *pdev) >> +{ >> + struct resource *res; >> + const struct of_device_id *match; >> + >> + if (!pdev->dev.of_node) >> + return -ENODEV; >> + >> + at91_shdwc = devm_kzalloc(&pdev->dev, sizeof(*at91_shdwc), GFP_KERNEL); >> + if (!at91_shdwc) >> + return -ENOMEM; >> + >> + platform_set_drvdata(pdev, at91_shdwc); >> + >> + res = platform_get_resource(pdev, IORESOURCE_MEM, 0); >> + at91_shdwc->at91_shdwc_base = devm_ioremap_resource(&pdev->dev, res); >> + if (IS_ERR(at91_shdwc->at91_shdwc_base)) { >> + dev_err(&pdev->dev, "Could not map reset controller address\n"); >> + return PTR_ERR(at91_shdwc->at91_shdwc_base); >> + } >> + >> + match = of_match_node(at91_shdwc_of_match, pdev->dev.of_node); >> + if (match != NULL) > > match will never be NULL, else you wouldn't be probed. You can remove that test.
I'll remove this and repost the driver as v3.
Thanks, bye, -- Nicolas Ferre
| |