• State
  • Locked Locked
  • Replies 9 replies
  • Subscribers 100 subscribers
  • Views 6096 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

jffs2, empty flash at ....

Arno Steffens
Intellectual 930 points


After adding a file to a jffs2 filesystem I get message like this:

"Empty flash at 0x001011e0 ends at 0x00101800"

What is the problem?

Are there a setting that is wrong or missed? I generated the jffs with

"mkfs.jffs2 -lqnp -e 128 -r /opt/rootfs/opt -o /share/fs_opt.jffs2"

I do use a Micron MT9F1G16ABC, so it is organized that way

– Page size • x16: 1056 words (1024 + 32 words)
– Block size: 64 pages (128K + 4K bytes)

The partitions are build like this, always mith multiples of 64*2048 (which is 128k)

static struct mtd_partition omap3evm_nand_partitions[] = {

    {

        .name           = "X-Loader",

        .offset         = 0,

        .size           = 2*(64*2048),

        .mask_flags     = MTD_WRITEABLE /* force read-only */

                                        /*

                                         * Rom code checks for x-loader

                                         * in the first four blocks.

                                         * In case NAND block 0 is bad

                                         * this feature can be useful.

                                         */

    },

    {

        .name           = "U-Boot",

        .offset         = MTDPART_OFS_APPEND,

        .size           =  4*(64*2048),         /*256 + 256 KB*/

//        .mask_flags     = MTD_WRITEABLE /* force read-only */

    },

  ....

 

 

  • 0
    Expert 1780 points

    Try to remove -p option from mkfs.jffs2. Then add -p option to nandwrite. Not sure it will work...

  • 0 in reply to Marco Braga
    TI__Expert 5170 points

    This message is generated if a block of data is partially written. It is generally not a sign of any problem.

    Regards, Sudhakar

  • 0 in reply to Sudhakar Rajashekhara
    Intellectual 930 points

    Thanks for answer:

    @: I will try this. Usually I use uboot to flash the system. Are there also option to compile in or to set?

    @Sudhakar: I know that is might not be a problem. But this messages while mounting are annoying.

    Anyhow I don't understand what is going on: An empty flash has 0xff in its cells. If I write or modify a file it is normal that the file is not as long as a  complete block. And than I would expect that the rest of the block keeps it 0xff. So isn't that the default behaviour?

  • 0 in reply to Marco Braga
    Intellectual 930 points

    I did some tests. 

    Did mkfs without -p. Than flashed with -p. Same "empty flash" message after adding  a file in system. If I flase without -p I get a message "Input file is not page aligned
    Data was only partially written due to error".

    Other way around I created with -p. The jffs2 fs becomes a bit bigger due to padding. I guess it is just necessary to do the padding once. Either in creating or flashing.

    I did in uboot a hexdump on the page which should contain the empty flash. The strange this - it is completly the same! Also the OOB are the same. The difference is the next block. It is empty before adding a file and filled after the fileoperation.

    So it seem it is not added gapless and that causes the message?

    Can I call a command which does some fileoperation to let disappear this message?

  • 0 in reply to Arno Steffens
    Expert 1780 points

    I think I've found your problem... The -e option you use in mkfs.jffs2.

    Please add "KiB" at the end or specify it in bytes. I think that -e 128 means 128 BYTES! In fact I use -e 0x20000. Please try it or "-e 128KiB" (case sensitive), but I am not sure about its syntax. Just go for 0x20000 to be sure.

    P.S: I'm not sure if you mkfs.jffs2 version matches mine. I've checked the source code of mine and this case is handled correctly. If you specify -e 128, it should mean 128 KB. But I am not sure that this is handled correctly in every version... So, just in case, use 0x20000.

     

  • 0 in reply to Marco Braga
    Intellectual 930 points

    Unfortunatly this is not the solution.

    The mkfs.jffs decides for a number smaller than xx to be used the parameter in kilobyte. To be sure I checked it with -e 128, -e 0x2000, -e 128KiB. All created filesystem files are identical (MD5Sum, length).

     

  • 0 in reply to Arno Steffens
    Prodigy 140 points

    DId you ever find a solution?  I am experiencing the same problem.

  • 0 in reply to Rishi 48200
    Intellectual 930 points

    No, I did not found a solution for that. I did have some hope, that TI's Linux team knows the ansers, but nothing happened.

  • 0 in reply to Arno Steffens
    Prodigy 140 points

    I ended up using flash_eraseall mtd utility on my target system.