Duplicate blocks in flash disk

[ali nasir]

Hi,

Details of the crash:
It is a bus error exception where a write is attempted to a region which is not in the memory map of the CPU. It occurs in function rtems_rfs_buffer_handle_request which is called from rtems_rfs_fs_open while mounting. 

We have a single flash device and use 5 sectors of the device for FFS. Each sector is 64KB.

We trying to find a way to send you the binary image of flash. Till then, if you can deuce anything would be helpful. 

Regards,
Ali

On Thu, Nov 17, 2011 11:52 AM IST Chris Johns wrote:

>Hi,
>
>Sorry about the delay. My internet connection has been up and down for the past couple of days and I have been snowed under (not literally; it is summer here).
>
>On 17/11/11 4:08 PM, ali nasir wrote:
>> Hi,
>> 
>> We are using the RFS on the flashdisk. Sometimes we observe that when the rtems_flashdisk_initialize (which is called from the rtems_io_register_driver) is called, there are some prints on the console which says:
>> 
>> fdisk:error:duplicate block: 0-0-104: duplicate: 0-0-108
>> fdisk:error:duplicate block: 0-0-104: duplicate: 0-0-112
>> 
>
>The code states:
>
>     /**
>      * @todo
>      * This may need more work later. Maybe a counter is stored with
>      * each block so we can tell which is the later block when
>      * duplicates appear. A power down with a failed wirte could cause
>      * a duplicate.
>      */
>
>> Since the return value of rtems_io_register_driver() is still 0
>
>Correct it because the code currently does not see an error. It sees the duplicate and selects one of them. The comment above indicates it maybe a counter is needed to help track the latest one.
>
>> the mounting continues further in our code and the code crashes in the function rtems_rfs_buffer_handle_request(). This crash happens only when the above error of duplicate blocks is seen.
>
>Do you have any details about the crash ?
>
>> When we see the RTEMS code, we found that even if the duplicate blocks are found, the return value of rtems_fdisk_recover_block_mappings() is still 0. This makes the code feel that everything is fine and the mounting can continue.
>
>The actual issue could be related to some other effect the is not handling correctly after the duplicate is found.
>
>> Is is OK, that we return some error if the duplicate blocks are found in the rtems_fdisk_recover_block_mappings() function? This can help in making some recovery like erasing the Flash device completely or some other thing.
>
>How big is the flash disk ?
>
>I would love a binary image of the data in the flash with the duplicate. I could then load it here and take a look.
>
>> Please guide.
>
>The error code is ok if we have an error. It is masking a bug we should find the bug.
>
>Chris