Re: [PATCH v11] mmc: support BKOPS feature for eMMC

[Chris Ball <cjb@laptop.org>]

Hi,

On Tue, Jul 31 2012, merez@codeaurora.org wrote:
>>> 512 byte stack? Isn't it really a problem?
>> How about this?
>>
>> +int mmc_read_bkops_status(struct mmc_card *card)
>> +{
>> +	int err;
>> +	u8 *ext_csd;
>> +
>> +	ext_csd = kmalloc(512, GFP_KERNEL);
>> +	if (!ext_csd) {
>> +		pr_err("%s: could not allocate a buffer to "
>> +			"receive the ext_csd.\n", mmc_hostname(card->host));
>> +		return -ENOMEM;
>> +	}
>> +
>> +	mmc_claim_host(card->host);
>> +	err = mmc_send_ext_csd(card, ext_csd);
>> +	mmc_release_host(card->host);
>> +	if (err)
>> +		goto out;
>> +
>> +	card->ext_csd.raw_bkops_status = ext_csd[EXT_CSD_BKOPS_STATUS];
>> +	card->ext_csd.raw_exception_status = ext_csd[EXT_CSD_EXP_EVENTS_STATUS];
>> +out:
>> +	kfree(ext_csd);
>> +	return err;
>> +}
>> +EXPORT_SYMBOL(mmc_read_bkops_status);
>
> I'm not sure it would be a good idea to allocate the buffer every time the
> ext_csd is read since with the periodic BKOPs we might do it more often to
> see if there is a need for BKOPs.
> How about keeping the buffer in the card structure?

It's a little large for that, but it would be worth it if we really do
use it regularly.  Perhaps go with the kmalloc option for now, add a
comment explaining that we should consider storing the entire ext_csd
in mmc_card later, and revisit it when the periodic bkops support is
submitted and we know how often we need to read the status?

Thanks,

- Chris.
-- 
Chris Ball   <cjb@laptop.org>   <http://printf.net/>
One Laptop Per Child