Prev: [patch 008/164] NFSD: dont report compiled-out versions as present
Next: [patch 134/164] l2tp: Fix oops in pppol2tp_xmit
From: Andrew Morton on 1 Jul 2010 16:50
On Fri, 25 Jun 2010 22:06:44 +0400
Anton Vorontsov <avorontsov(a)mvista.com> wrote:

> There's nothing special, just SoC-specific ops and quirks.
>
> ...
>
> +static void sdhci_cns3xxx_set_clock(struct sdhci_host *host, unsigned int clock)
> +{
> + struct device *dev = mmc_dev(host->mmc);
> + int div = 1;
> + u16 clk;
> + unsigned long timeout;
> +
> + if (clock == host->clock)
> + return;

I assume that mmc core prevents this function from being exectued twice
at the same time?

> + sdhci_writew(host, 0, SDHCI_CLOCK_CONTROL);
> +
> + if (clock == 0)
> + goto out;
> +
> + while (host->max_clk / div > clock) {
> + /*
> + * On CNS3xxx divider grows linearly up to 4, and then
> + * exponentially up to 256.
> + */
> + if (div < 4)
> + div += 1;
> + else if (div < 256)
> + div *= 2;
> + else
> + break;
> + }
> +
> + dev_dbg(dev, "desired SD clock: %d, actual: %d\n",
> + clock, host->max_clk / div);
> +
> + /* Divide by 3 is special. */
> + if (div != 3)
> + div >>= 1;
> +
> + clk = div << SDHCI_DIVIDER_SHIFT;
> + clk |= SDHCI_CLOCK_INT_EN;
> + sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
> +
> + timeout = 20;
> + while (!((clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL))
> + & SDHCI_CLOCK_INT_STABLE)) {
> + if (timeout == 0) {
> + dev_warn(dev, "clock is unstable");
> + break;
> + }
> + timeout--;
> + mdelay(1);

Could we have used the more polite msleep() here?

> + }
> +
> + clk |= SDHCI_CLOCK_CARD_EN;
> + sdhci_writew(host, clk, SDHCI_CLOCK_CONTROL);
> +out:
> + host->clock = clock;
> +}
> +
>
> ...
>
> --- a/drivers/mmc/host/sdhci-pltfm.c
> +++ b/drivers/mmc/host/sdhci-pltfm.c
> @@ -158,6 +158,9 @@ static int __devexit sdhci_pltfm_remove(struct platform_device *pdev)
>
> static const struct platform_device_id sdhci_pltfm_ids[] = {
> { "sdhci", },
> +#ifdef CONFIG_MMC_SDHCI_CNS3XXX
> + { "sdhci-cns3xxx", (kernel_ulong_t)&sdhci_cns3xxx_pdata },
> +#endif

What the heck is this kernel_ulong_t thing and why did `struct
platform_device_id' see a need to invent it??


> { },
> };
> MODULE_DEVICE_TABLE(platform, sdhci_pltfm_ids);
>
> ...
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Anton Vorontsov on 8 Jul 2010 11:40
Sorry for the delayed response,

On Thu, Jul 01, 2010 at 01:48:39PM -0700, Andrew Morton wrote:
> On Fri, 25 Jun 2010 22:06:44 +0400
> Anton Vorontsov <avorontsov(a)mvista.com> wrote:
>
> > There's nothing special, just SoC-specific ops and quirks.
> >
> > ...
> >
> > +static void sdhci_cns3xxx_set_clock(struct sdhci_host *host, unsigned int clock)
> > +{
> > + struct device *dev = mmc_dev(host->mmc);
> > + int div = 1;
> > + u16 clk;
> > + unsigned long timeout;
> > +
> > + if (clock == host->clock)
> > + return;
>
> I assume that mmc core prevents this function from being exectued twice
> at the same time?

Yep, it's called under spin_lock_irqsave(&host->lock, flags).

[...]
> > + timeout = 20;
> > + while (!((clk = sdhci_readw(host, SDHCI_CLOCK_CONTROL))
> > + & SDHCI_CLOCK_INT_STABLE)) {
> > + if (timeout == 0) {
> > + dev_warn(dev, "clock is unstable");
> > + break;
> > + }
> > + timeout--;
> > + mdelay(1);
>
> Could we have used the more polite msleep() here?

Unfortunately not, we're in the atomic context.

[...]
> > --- a/drivers/mmc/host/sdhci-pltfm.c
> > +++ b/drivers/mmc/host/sdhci-pltfm.c
> > @@ -158,6 +158,9 @@ static int __devexit sdhci_pltfm_remove(struct platform_device *pdev)
> >
> > static const struct platform_device_id sdhci_pltfm_ids[] = {
> > { "sdhci", },
> > +#ifdef CONFIG_MMC_SDHCI_CNS3XXX
> > + { "sdhci-cns3xxx", (kernel_ulong_t)&sdhci_cns3xxx_pdata },
> > +#endif
>
> What the heck is this kernel_ulong_t thing and why did `struct
> platform_device_id' see a need to invent it??

It's not only platform_device_id's thing. Sometimes drivers just
pass a constant instead of a pointer (e.g. DEVICE_IS_FOO,
DEVICE_IS_BAR), for example see drivers/hwmon/lm75.c (enum
lm75_type).

Other than this I don't think that there's a good reason for it.

Thanks,

--
Anton Vorontsov
email: cbouatmailru(a)gmail.com
irc://irc.freenode.net/bd2
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 | 
Pages: 1
Prev: [patch 008/164] NFSD: dont report compiled-out versions as present
Next: [patch 134/164] l2tp: Fix oops in pppol2tp_xmit