Re: [PATCH v3] MMC: Add JZ4740 mmc driver

[Lars-Peter Clausen <lars@metafoo.de>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi

Andrew Morton wrote:
> On Mon, 28 Jun 2010 03:20:41 +0200
> Lars-Peter Clausen <lars@metafoo.de> wrote:
> 
>> This patch adds support for the mmc controller on JZ4740 SoCs.
>>
>> Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
>> Cc: Andrew Morton <akpm@linux-foundation.org>
>> Cc: Matt Fleming <matt@console-pimps.org>
>> Cc: linux-mmc@vger.kernel.org
>>
>> ...
>>
>> +#define JZ4740_MMC_MAX_TIMEOUT 10000000
> 
> That was a really big timeout.  How long do 1e7 readw's take?  Oh well.
> 

Hm, right. It doesn't hurt though. I'll do some tests and to try to come up with a
more realistic value.

>> ...
>>
>> +static void jz4740_mmc_clock_disable(struct jz4740_mmc_host *host)
>> +{
>> +	uint32_t status;
>> +
>> +	writew(JZ_MMC_STRPCL_CLOCK_STOP, host->base + JZ_REG_MMC_STRPCL);
>> +	do {
>> +		status = readl(host->base + JZ_REG_MMC_STATUS);
>> +	} while (status & JZ_MMC_STATUS_CLK_EN);
>> +}
>> +
>> +static void jz4740_mmc_reset(struct jz4740_mmc_host *host)
>> +{
>> +	uint32_t status;
>> +
>> +	writew(JZ_MMC_STRPCL_RESET, host->base + JZ_REG_MMC_STRPCL);
>> +	udelay(10);
>> +	do {
>> +		status = readl(host->base + JZ_REG_MMC_STATUS);
>> +	} while (status & JZ_MMC_STATUS_IS_RESETTING);
>> +}
> 
> Maybe these should have a timeout too?

Its very unlikely that these will ever timeout. But right, to be on the safe, there
should be a timeout as well.

> 
>> ...
>>
>> +static inline unsigned int jz4740_mmc_wait_irq(struct jz4740_mmc_host *host,
>> +	unsigned int irq)
>> +{
>> +	unsigned int timeout = JZ4740_MMC_MAX_TIMEOUT;
>> +	uint16_t status;
>> +
>> +	do {
>> +		status = readw(host->base + JZ_REG_MMC_IREG);
>> +	} while (!(status & irq) && --timeout);
>> +
>> +	return timeout;
>> +}
> 
> This guy's too big to inline.  Recent gcc's know that and they tend to
> uninline such things behind your back anwyay.

Actually, even without the inline attribute and compiling with -Os gcc will inline
this function by itself.

> 
>> ...
>>
>> +struct jz4740_mmc_platform_data {
>> +	int gpio_power;
>> +	int gpio_card_detect;
>> +	int gpio_read_only;
>> +	unsigned card_detect_active_low:1;
>> +	unsigned read_only_active_low:1;
>> +	unsigned power_active_low:1;
>> +
>> +	unsigned data_1bit:1;
>> +};
> 
> The bitfields will all share the same word, so modification of one
> field can race against modification of another field.  Hence some form
> of locking which covers *all* the bitfields is needed.
> 
> Is that a problem in this driver?
> 
They are all read-only in the driver.


Thanks for reviewing the patch
- - Lars

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkwsuDQACgkQBX4mSR26RiOiMACeMMNj4koCYFAUnxyM0LBr+wOZ
x6oAnizk5CaAvZjQ2doXrD6ZYgDeNjSr
=92D2
-----END PGP SIGNATURE-----