[U-Boot] [PATCH] fs: fat: fix bug when write size is less than a sector size

[Josh Wu <josh.wu@atmel.com>]

Dear Wolfgang Denk,

Thank you for the quick reply.

On 7/18/2013 4:24 PM, Wolfgang Denk wrote:
> Dear Josh Wu,
>
> In message <1374134870-10154-1-git-send-email-josh.wu@atmel.com> you wrote:
>> Tested in at91sam9x5ek and sama5d3x-ek in mmc fat.
>>
>> Signed-off-by: Josh Wu <josh.wu@atmel.com>
>> ---
>> When I debugging this issue, I also find the thread talking about this:
>>    http://thread.gmane.org/gmane.comp.boot-loaders.u-boot/158698
>> But finally no patch is send for this bug. So I send this and cc all the
>> people in that thread. Hope it can be merged in 2013.07.
> Definitely not.  This needs thurough testing before it can go
> mainline.

got it.

>
>
>> -	__u32 startsect;
>> +	__u32 startsect, nr_sectors;
>>   
>>   	if (clustnum > 0)
>>   		startsect = mydata->data_begin +
>> @@ -571,7 +571,11 @@ set_cluster(fsdata *mydata, __u32 clustnum, __u8 *buffer,
>>   
>>   	debug("clustnum: %d, startsect: %d\n", clustnum, startsect);
>>   
>> -	if (disk_write(startsect, size / mydata->sect_size, buffer) < 0) {
>> +	nr_sectors = size / mydata->sect_size;
>> +	if (nr_sectors == 0)
>> +		nr_sectors = 1;
> Why would this be needed?  The case of a partial write is covered
> further down below in the code, in the "if (size % mydata->sect_size)"
> part...

The call of disk_write() will pass the sector size as 0. That will cause 
the mmc driver stalled.

And after a search, the fix patch is already sent by Ruud Commandeur and 
merge in mainline:
   "mmc write bug fix"
   commit a586c0aa211fb79ecaa06aee3299bfdd81329876
   Author: Ruud Commandeur <RCommandeur@clb.nl>
   Date:   Wed May 22 13:19:43 2013 +0200

So please forget this patch. Since it is fixed in mmc driver layer. I 
will test it in my board.
Sorry for the noise.

>
>
>
> Best regards,
>
> Wolfgang Denk
>

Best Regards,
Josh Wu