Rtems_fsmount dosfs hanging
Andre Marques
andre.lousa.marques at gmail.com
Tue Aug 26 08:04:12 UTC 2014
Hello Ritesh,
My progress on the SD card support can be found on
https://bitbucket.org/asuol/rtems-graduation-project/src/
It contains the sources and a test case (it reads a file and write
another on the card), as well as patches so you can easily test that code.
It was based on
https://github.com/jncronin/rpi-boot/blob/master/emmc.c
and right now it may or may not work with your SD card. I got it working
with the SD card I have used during development (4GB class 4 SanDisk SD
card), but it does not work with the card I have now (Kingston 8GB
microSD card on a SD adapter). I suspect this is related with the
timeout the driver gives for each command to complete in, as it is an
hard coded value which should be calculated by parsing the card's CSD
register. Also the code needs some major cleanup, and some features were
overlooked such as memory barriers.
As for SD card documentation I have used the v3 simplified specifications
SD host controller -
https://www.sdcard.org/downloads/pls/simplified_specs/archive/partA2_300.pdf
SD physical layer -
https://www.sdcard.org/downloads/pls/simplified_specs/archive/part1_301.pdf
Besides the broadcom datasheet.
By the way, since you replied to this thread I am not sure if I ever
shared why it was hanging. Turns out that there was/is a conflict
between my emmc driver and the raspberry pi clock driver. All I had to
do was to disable the clock driver for it to work.
--André Marques.
On 08/26/14 06:38, Ritesh Harjani wrote:
> Hi Andre,
>
> I am too trying to understand SD card/MMC protocol on Raspberry pi for
> RTEMS. You said in your last mail that you have already done some
> progress on this.
> Could you please share your code at some place, so that others can get
> a help from your code in case anybody else too is trying to
> understand/develop SD card support for Raspberry pi.
>
>
> Thanks for the help!!
> Ritesh
>
>
> On Wed, May 14, 2014 at 12:30 AM, Andre Marques
> <andre.lousa.marques at gmail.com <mailto:andre.lousa.marques at gmail.com>>
> wrote:
>
> On 05/13/14 12:15, Andrey Mozzhuhin wrote:
>> Do you not forget to call rtems_blkdev_request_done in read
>> routine of block device driver?
>> Also try to increase CONFIGURE_MAXIMUM_TASKS - bdbuf need some
>> worker tasks.
>
> I call rtems_blkdev_request_done after each read request, and
> changing CONFIGURE_MAXIMUM_TASKS to 20 (instead of 2) did not had
> any impact.
>
>
>>
>> >> https://github.com/jncronin/rpi-boot/blob/master/emmc.c
>>
>> Hmm. Implementation specific to Raspberry Pi is not good idea. It
>> is better to separate MMC/SD protocol implementation and hardware
>> support like in FreeBSD or Linux.
>
> Yes, the point in the end is probably to have the MMC/SD protocol
> somewhere else, just like the spi-sd-card driver (SD card access
> through SPI interface protocol, at c/src/libchip/i2c/spi-sd-card*).
>
>
>>
>> >> Not sure how to approach this problem, as the logic seems
>> right. Will do more testing in the meanwhile.
>>
>
> Today I have re-formated the card (under Linux), and the block
> data read on RPi checks with the card hexdump on Linux.
>
> So far I tested:
>
> - The block data that is being read, and compared with the same
> block data read on Linux (small mismatch (4 bits) at the middle of
> a 512 byte block before, no mismatch now)
>
> - The request buffer contents
> (http://www.rtems.org/onlinedocs/doxygen/cpukit/html/structrtems__blkdev__sg__buffer.html).
> Looking at this I don't know what the User pointer is. I'm using
> the buffer pointer to store my readings.
>
> - Reading the partition table from the first block. I can read the
> first block and register the card partition as a logical disk, so
> reading and registering the partition table is no problem.
>
> - Tried also a busy wait between block readings.
>
> The result is always the same: when mounting the card partition,
> it reads the first two partition blocks and then hangs.
>
>
>> If you get stuck share your sources and let me know.
>>
>
> Will try to get the code to github soon.
>
>
>>
>>
>> 2014-05-13 14:03 GMT+04:00 Andre Marques
>> <andre.lousa.marques at gmail.com
>> <mailto:andre.lousa.marques at gmail.com>>:
>>
>> On 05/12/14 19:25, Andrey Mozzhuhin wrote:
>>> I think you need start with simple tests without file
>>> systems and block device driver.
>>> Try read and write blocks with known patterns and check it
>>> on PC and Raspberry Pi.
>>>
>>
>> I have checked the data read by my emmc driver from the first
>> two blocks (the ones it reads) on the RPi, and compared with
>> an hexdump of those same two blocks on Linux, and there is a
>> small mismatch at the middle of the first block.
>>
>> Not sure how to approach this problem, as the logic seems
>> right. Will do more testing in the meanwhile.
>>
>>
>>> What are you use for MMC/SD protocol? Can I see it and block
>>> device driver somewhere?
>>
>> I'm porting the following code
>>
>> https://github.com/jncronin/rpi-boot/blob/master/emmc.c
>>
>> and using the following SD card documentation
>>
>> Physical Layer Simplified Specification (3.01) -
>> https://www.sdcard.org/downloads/pls/simplified_specs/archive/part1_301.pdf
>>
>> Host Controller Simplified Specification (2.00) -
>> https://www.sdcard.org/downloads/pls/simplified_specs/archive/partA2_200.pdf
>>
>> and the raspberry soc datasheet: Broadcom BCM2835 Peripherals
>> Guide (Chapter 5 - EMMC)
>>
>>>
>>>
>>> 2014-05-12 20:23 GMT+04:00 Andre Marques
>>> <andre.lousa.marques at gmail.com
>>> <mailto:andre.lousa.marques at gmail.com>>:
>>>
>>> Hello,
>>>
>>> I have used the pc386 BSP to mount the card (before your
>>> suggestions) using the fileio sample and it could read
>>> the card contents perfectly.
>>>
>>> Then I filled the card with zeros using dd on Linux,
>>> modified the fileio sample to format the card with
>>> msdos_format(dev, NULL) and then mounted it on the pc
>>> 386 BSP fileio sample and I could write and read the card.
>>>
>>> On Linux I set the card with the necessary files for
>>> Raspberry, and when my driver tries to mount the card on
>>> the Raspberry Pi it reads the first 2 blocks (now 0 and
>>> 1, because the card has no partition table) and hangs,
>>> just as before.
>>>
>>> I guess my driver is having some reading problems.
>>>
>>>
>>> On 05/05/14 19:30, Andrey Mozzhuhin wrote:
>>>> You can try to use SD Card in opposite direction:
>>>> 0) fill SD Card with zeroes;
>>>> 1) format partition under RTEMS with msdos_format()
>>>> function;
>>>> 2) check that PC can read/write this SD Card;
>>>> 3) if PC fails to mount this partition - dump SD Card
>>>> and check that all data is on right sectors.
>>>>
>>>>
>>>>
>>>> 2014-05-05 22:28 GMT+04:00 Andrey Mozzhuhin
>>>> <nopscmn at gmail.com <mailto:nopscmn at gmail.com>>:
>>>>
>>>> Hi, Andre
>>>>
>>>> You can try to use SD Card in opposite direction:
>>>> 0) fill SD Card with zeroes;
>>>> 1) format partition under RTEMS with msdos_format()
>>>> function;
>>>> 2) check that PC can read/write this SD Card;
>>>> 3) if PC fails to mount this partition - dump SD
>>>> Card and check that all data is on right sectors.
>>>>
>>>>
>>>>
>>>> 2014-05-05 13:52 GMT+04:00 Andre Marques
>>>> <andre.lousa.marques at gmail.com
>>>> <mailto:andre.lousa.marques at gmail.com>>:
>>>>
>>>> Hello,
>>>>
>>>> Following the problem in
>>>>
>>>> http://www.rtems.org/pipermail/rtems-devel/2014-April/006585.html
>>>>
>>>> I am now trying to mount the SD card partitions
>>>> on RTEMS.
>>>>
>>>> Summarizing the process:
>>>>
>>>> 1. rtems_io_register_driver (by calling my
>>>> driver with CONFIGURE_APPLICATION_EXTRA_DRIVERS
>>>> on hello sample)
>>>>
>>>> 2. rtems_filesystem_make_dev_t (to get the
>>>> device file)
>>>>
>>>> 3. rtems_disk_io_initialize
>>>>
>>>> 4. rtems_disk_create_phys (to create the disk,
>>>> at /dev/sdc0)
>>>>
>>>> 5. rtems_bdpart_register_from_disk (to read and
>>>> register the partitions)
>>>>
>>>> 6. rtems_fsmount (to mount the partitions)
>>>>
>>>> For rtems_fsmount I am providing the following
>>>> fs_table:
>>>>
>>>> rtems_fstab_entry fs_table [] = {
>>>> {
>>>> .source = "/dev/sdc0",
>>>> .target = "/mnt/p1",
>>>> .type = "dosfs",
>>>> .options = RTEMS_FILESYSTEM_READ_WRITE,
>>>> .report_reasons = RTEMS_FSTAB_ANY,
>>>> .abort_reasons = RTEMS_FSTAB_OK
>>>> },{
>>>> .source = "/dev/sdc01",
>>>> .target = "/mnt/p2",
>>>> .type = "dosfs",
>>>> .options = RTEMS_FILESYSTEM_READ_WRITE,
>>>> .report_reasons = RTEMS_FSTAB_ANY,
>>>> .abort_reasons = RTEMS_FSTAB_NONE
>>>> }
>>>> };
>>>>
>>>> The SD card has only one partition starting at
>>>> block number 8192.
>>>>
>>>> After I call rtems_fsmount it calls my driver
>>>> to read block 8192 and 8193 and then hangs.
>>>>
>>>> I have tested the driver and It seems to have
>>>> no problem reading single or multiple blocks.
>>>>
>>>> Any tips?
>>>>
>>>> Also I am using the following confdefs
>>>> configuration (some values are exaggerated):
>>>>
>>>> #define CONFIGURE_APPLICATION_NEEDS_CLOCK_DRIVER
>>>> #define CONFIGURE_APPLICATION_NEEDS_CONSOLE_DRIVER
>>>> #define CONFIGURE_APPLICATION_NEEDS_LIBBLOCK
>>>>
>>>> #define CONFIGURE_FILESYSTEM_DOSFS
>>>> #define CONFIGURE_USE_IMFS_AS_BASE_FILESYSTEM
>>>>
>>>> #define CONFIGURE_APPLICATION_EXTRA_DRIVERS
>>>> SD_CARD_DRIVER_TABLE_ENTRY
>>>>
>>>> #define CONFIGURE_LIBIO_MAXIMUM_FILE_DESCRIPTORS 20
>>>>
>>>> #define CONFIGURE_MAXIMUM_TASKS 2
>>>> #define CONFIGURE_MAXIMUM_DRIVERS 10
>>>>
>>>> #define CONFIGURE_RTEMS_INIT_TASKS_TABLE
>>>>
>>>> #define CONFIGURE_INIT_TASK_STACK_SIZE (32 * 1024)
>>>>
>>>> #define CONFIGURE_INITIAL_EXTENSIONS
>>>> RTEMS_TEST_INITIAL_EXTENSION
>>>>
>>>> #define CONFIGURE_INIT
>>>>
>>>> --André Marques
>>>>
>>>> _______________________________________________
>>>> rtems-devel mailing list
>>>> rtems-devel at rtems.org
>>>> <mailto:rtems-devel at rtems.org>
>>>> http://www.rtems.org/mailman/listinfo/rtems-devel
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> rtems-devel mailing list
>>>> rtems-devel at rtems.org <mailto:rtems-devel at rtems.org>
>>>> http://www.rtems.org/mailman/listinfo/rtems-devel
>>>
>>>
>>
>>
>
>
> _______________________________________________
> rtems-devel mailing list
> rtems-devel at rtems.org <mailto:rtems-devel at rtems.org>
> http://www.rtems.org/mailman/listinfo/rtems-devel
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20140826/e37c5d7b/attachment-0002.html>
More information about the devel
mailing list