RE: MTD and jffs2 mount problem on Linux-2.6.23 rc2

RE: MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Dach Miroslaw]

Hi Grant,

Thanks for your hints. As regards the endianness. I have tried both little and big endian options.
In my case only Big endian (native for ppc405) is accepted. There is however the CRC calculation mismatch
between the calculated one (by the kernel) during mounting the jffs2 and that one which is calculated by mkfs.jffs2.

I am however not sure that the Flash memory access in my case is fine for all modes A8 A16 A32.
I have got than idea to test jffs2 image stored somehow in RAM instead of Flash to test if jffs2 image generated by mkfs.jffs2 is consistent from kernel 2.6.23 point of view.

this what I have tried was:

losetup /dev/loop0 img2.jffs2  
mount -t jffs2 /dev/loop0 /mnt

unfortunately mount did not work.

Maybe it is possible somehow to use RAMDISK to emulate the Flash to mount jffs2 image?

Best Regards

M.


-----Original Message-----
From: Grant Erickson [mailto:gerickson@nuovations.com]
Sent: Tue 3/4/2008 8:40 PM
To: Dach Miroslaw
Subject: Re: MTD and jffs2 mount problem on Linux-2.6.23 rc2
 
On 3/4/08 11:29 AM, Dach Miroslaw wrote:
> In fact I have NOR flash. The 'dd' command to copy data to flash and backwards
> to the file works fine so I thing that 'dd' is not the issue.

Seems reasonable.

> I suspect that crc32 works differently when generating the jffs2 image file
> and the kernel uses different calculation methode? Maybe I am wrong.

Are you specifying an endianness/byte ordering on the command line when
creating the JFFS2 file system?

> Maybe it is the problem to access the Flash memory? Is there any method to
> mount jffs2 image in RAM instead of Flash just to test the jffs2 image
> consistency?

If your kernel is so configured, you should be able to mount it via the
loopback device:

    mount -t loop root.img /mnt

Regards,

Grant

RE: MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Schlägl Manfred jun.]

Hi!
Am Mittwoch, den 05.03.2008, 09:24 +0100 schrieb Dach Miroslaw:
> Hi Grant,
> 
> Thanks for your hints. As regards the endianness. I have tried both little and big endian options.
> In my case only Big endian (native for ppc405) is accepted. There is however the CRC calculation mismatch
> between the calculated one (by the kernel) during mounting the jffs2 and that one which is calculated by mkfs.jffs2.
> 
> I am however not sure that the Flash memory access in my case is fine for all modes A8 A16 A32.
> I have got than idea to test jffs2 image stored somehow in RAM instead of Flash to test if jffs2 image generated by mkfs.jffs2 is consistent from kernel 2.6.23 point of view.
> 
> this what I have tried was:
> 
> losetup /dev/loop0 img2.jffs2  
> mount -t jffs2 /dev/loop0 /mnt
> 
> unfortunately mount did not work.
This can't work, because jffs2 is using the block-device mtdblock only
to get device-information about the mtd-device. jffs2 itself is using
the mtd-api directly. so jffs2 only on mtd.

> 
> Maybe it is possible somehow to use RAMDISK to emulate the Flash to mount jffs2 image?
> 
There is a nand-Simulator (nandsim). It simulates a nand in
kernel-virtual-memory. With nandsim you'll get a ram-mtd-device, on
witch jffs2 can be used.

Using nandsim: http://www.linux-mtd.infradead.org/faq/nand.html

Example: simulating 32MB 8bit nand-flash (page:512, eb: 16K):
 * nandsim & mtd
   # modprobe nandsim first_id_byte=0x20 second_id_byte=0x35
   # modprobe mtdblock
   # modprobe mtdchar
 * erase (not neccessary)
   # flash_eraseall /dev/mtd0
   Erasing 16 Kibyte @ 1ffc000 -- 99 % complete.
 * Create jffs2-Image
   # mkfs.jffs2 -p 512 -s 512 -e 16384 -d jffs2 -o jffs2.img
 * Write to nandsim
   # nandwrite /dev/mtd0 jffs2.img
   Writing data to block 0
   Writing data to block 4000
 * Mount
   # mount /dev/mtdblock0 /mnt/ -t jffs2


> Best Regards
> 
> M.
> 
> 
> -----Original Message-----
> From: Grant Erickson [mailto:gerickson@nuovations.com]
> Sent: Tue 3/4/2008 8:40 PM
> To: Dach Miroslaw
> Subject: Re: MTD and jffs2 mount problem on Linux-2.6.23 rc2
>  
> On 3/4/08 11:29 AM, Dach Miroslaw wrote:
> > In fact I have NOR flash. The 'dd' command to copy data to flash and backwards
> > to the file works fine so I thing that 'dd' is not the issue.
> 
> Seems reasonable.
> 
> > I suspect that crc32 works differently when generating the jffs2 image file
> > and the kernel uses different calculation methode? Maybe I am wrong.
> 
> Are you specifying an endianness/byte ordering on the command line when
> creating the JFFS2 file system?
> 
> > Maybe it is the problem to access the Flash memory? Is there any method to
> > mount jffs2 image in RAM instead of Flash just to test the jffs2 image
> > consistency?
> 
> If your kernel is so configured, you should be able to mount it via the
> loopback device:
> 
>     mount -t loop root.img /mnt
> 
> Regards,
> 
> Grant
> 
> 
> 
> 
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/

Greetings
	Manfred

Re: MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Jörn Engel]

On Wed, 5 March 2008 10:45:23 +0100, Schlägl Manfred jun. wrote:
> Am Mittwoch, den 05.03.2008, 09:24 +0100 schrieb Dach Miroslaw:
> > 
> > losetup /dev/loop0 img2.jffs2  
> > mount -t jffs2 /dev/loop0 /mnt
> > 
> > unfortunately mount did not work.
> This can't work, because jffs2 is using the block-device mtdblock only
> to get device-information about the mtd-device. jffs2 itself is using
> the mtd-api directly. so jffs2 only on mtd.

Some people use loop and block2mtd to mount jffs2 images.  Hardly
elegant, but it seems to work.

Jörn

-- 
Don't worry about people stealing your ideas. If your ideas are any good,
you'll have to ram them down people's throats.
-- Howard Aiken quoted by Ken Iverson quoted by Jim Horning quoted by
   Raph Levien, 1979

RE: MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Dach Miroslaw]

Hi Jörn,

Thank you for the hint. I have recompiled the kernel with built-in block2mtd driver.

Do you know where to find some examples how to use block2mtd with jffs2 image.

Best Regards

M.


-----Original Message-----
From: Jörn Engel [mailto:joern@logfs.org]
Sent: Wed 3/5/2008 1:03 PM
To: Schlägl Manfred jun.
Cc: Dach Miroslaw; linux-mtd@lists.infradead.org; Grant Erickson
Subject: Re: MTD and jffs2 mount problem on Linux-2.6.23 rc2
 
On Wed, 5 March 2008 10:45:23 +0100, Schlägl Manfred jun. wrote:
> Am Mittwoch, den 05.03.2008, 09:24 +0100 schrieb Dach Miroslaw:
> > 
> > losetup /dev/loop0 img2.jffs2  
> > mount -t jffs2 /dev/loop0 /mnt
> > 
> > unfortunately mount did not work.
> This can't work, because jffs2 is using the block-device mtdblock only
> to get device-information about the mtd-device. jffs2 itself is using
> the mtd-api directly. so jffs2 only on mtd.

Some people use loop and block2mtd to mount jffs2 images.  Hardly
elegant, but it seems to work.

Jörn

-- 
Don't worry about people stealing your ideas. If your ideas are any good,
you'll have to ram them down people's throats.
-- Howard Aiken quoted by Ken Iverson quoted by Jim Horning quoted by
   Raph Levien, 1979

Re: MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Jörn Engel]

On Wed, 5 March 2008 14:44:45 +0100, Dach Miroslaw wrote:
> 
> Thank you for the hint. I have recompiled the kernel with built-in block2mtd driver.
> 
> Do you know where to find some examples how to use block2mtd with jffs2 image.

I don't.  If someone wants to write some documentation, I'd gladly add
it.

But I believe this should do the trick:
echo block2mtd=/dev/loop0 > /sys/modules/block2mtd/parameters/block2mtd
or to explicitly set the erase size:
echo block2mtd=/dev/loop0,128KiB > /sys/modules/block2mtd/parameters/block2mtd

Jörn

-- 
The story so far:
In the beginning the Universe was created.  This has made a lot
of people very angry and been widely regarded as a bad move.
-- Douglas Adams

RE: MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Dach Miroslaw]

Hi Jorn,

I have followed your idea with block2mtd.

The steps I have done are as following:
# mkdir /mnt
# ./mkfs.jffs2 -n --pad=0xA0000 --eraseblock=0x20000 --root=MTEST-PC-M24 --output=image2.jffs2
# losetup /dev/loop0 image2.jffs2
# insmod block2mtd block2mtd=/dev/loop0,640ki

[ 2715.317777] block2mtd: mtd4: [d: /dev/loop0] erase_size = 640KiB [655360]

# mount -t jffs2 /dev/mtdblock4 /mnt
                                                                                
[ 2725.702309] jffs2: Too few erase blocks (1)
mount: mounting mtd4 on /mnt failed

I have no idea what is the meaning of "Too few erase blocks".

Best Regards

M.






-----Original Message-----
From: Jörn Engel [mailto:joern@logfs.org]
Sent: Wed 3/5/2008 2:50 PM
To: Dach Miroslaw
Cc: Jörn Engel; Schlägl Manfred jun.; linux-mtd@lists.infradead.org; Grant Erickson
Subject: Re: MTD and jffs2 mount problem on Linux-2.6.23 rc2
 
On Wed, 5 March 2008 14:44:45 +0100, Dach Miroslaw wrote:
> 
> Thank you for the hint. I have recompiled the kernel with built-in block2mtd driver.
> 
> Do you know where to find some examples how to use block2mtd with jffs2 image.

I don't.  If someone wants to write some documentation, I'd gladly add
it.

But I believe this should do the trick:
echo block2mtd=/dev/loop0 > /sys/modules/block2mtd/parameters/block2mtd
or to explicitly set the erase size:
echo block2mtd=/dev/loop0,128KiB > /sys/modules/block2mtd/parameters/block2mtd

Jörn

-- 
The story so far:
In the beginning the Universe was created.  This has made a lot
of people very angry and been widely regarded as a bad move.
-- Douglas Adams

Re: MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Jörn Engel]

On Thu, 6 March 2008 08:48:46 +0100, Dach Miroslaw wrote:
>
> # insmod block2mtd block2mtd=/dev/loop0,640ki
                                          ^^^^^
> [ 2715.317777] block2mtd: mtd4: [d: /dev/loop0] erase_size = 640KiB [655360]
                                                               ^^^^^^

That number is likely wrong.  Also you might have to pad your image with
0xff a bit.

Jörn

-- 
All art is but imitation of nature.
-- Lucius Annaeus Seneca

MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Dach Miroslaw]

Dear All,

I am just a beginner in MTD tools. I have downloaded from the web mtd-tools version 1.0.0.
I have crosscompiled the mtd-tools for ppc405. All went fine.

I have build the kernel 2.6.23 rc2 with mtd and jffs2 enabled:

CONFIG_MTD=y
CONFIG_MTD_PARTITIONS=y
CONFIG_MTD_CMDLINE_PARTS=y
 
CONFIG_MTD_CHAR=y
CONFIG_MTD_BLKDEVS=y
CONFIG_MTD_BLOCK=y

CONFIG_MTD_CFI=y
CONFIG_MTD_GEN_PROBE=y

CONFIG_MTD_MAP_BANK_WIDTH_1=y
CONFIG_MTD_MAP_BANK_WIDTH_2=y
CONFIG_MTD_MAP_BANK_WIDTH_4=y

CONFIG_MTD_CFI_I1=y
CONFIG_MTD_CFI_I2=y

CONFIG_MTD_CFI_INTELEXT=y

CONFIG_MTD_CFI_UTIL=y
CONFIG_MTD_RAM=y
CONFIG_MTD_ROM=y

CONFIG_MTD_PHYSMAP=y
CONFIG_MTD_PHYSMAP_START=0xFF000000
CONFIG_MTD_PHYSMAP_LEN=0x800000
CONFIG_MTD_PHYSMAP_BANKWIDTH=2
CONFIG_MTD_PLATRAM=y

CONFIG_BLK_DEV=y
CONFIG_BLK_DEV_LOOP=y

CONFIG_JFFS2_FS=y
CONFIG_JFFS2_FS_DEBUG=0
CONFIG_JFFS2_FS_WRITEBUFFER=y
# CONFIG_JFFS2_SUMMARY is not set
# CONFIG_JFFS2_FS_XATTR is not set
# CONFIG_JFFS2_COMPRESSION_OPTIONS is not set
CONFIG_JFFS2_ZLIB=y
CONFIG_JFFS2_RTIME=y

I have created on my target board the jffs2 image which is foreseen to be place on /dev/mtd3:

./mkfs.jffs2 --pad=0xA0000 --eraseblock=0x20000 --root=TEST-DIR --output=image2.jffs2
flash_erase /dev/mtd3
dd if=./image2.jffs2 of=/dev/mtd3 bs=2 count=327680

Up to this point all went fine.

The problem emerged when I have tried to mount the jffs2 partition:
mkdir /mnt
mount -t jffs2 /dev/mtdblock3  /mnt

---------------------------------------------------------------------------------------------------------
[ 7459.517787] jffs2_scan_eraseblock(): Node at 0x00000000 {0x1985, 0x2003, 0x0000000c) has invalid CRC 0xf060dc98 (calculated 0xf9d690b3)
[ 7459.665215] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000008: 0xf060 instead
[ 7459.772681] jffs2_scan_eraseblock(): Node at 0x0000000c {0x1985, 0xe001, 0x00000035) has invalid CRC 0xc44d1944 (calculated 0x92fedd67)
[ 7459.919977] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000014: 0xc44d instead
[ 7460.027434] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000028: 0x0d0a instead
[ 7460.134905] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0000002c: 0xe7db instead
[ 7460.242298] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000030: 0x326c instead
[ 7460.349685] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000034: 0x4750 instead
[ 7460.457079] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000038: 0x2e74 instead
[ 7460.564482] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0000003c: 0x706c instead
[ 7460.671504] Further such events for this erase block will not be printed
[ 7460.834424] jffs2_scan_eraseblock(): Node at 0x00020000 {0x1985, 0x2003, 0x0000000c) has invalid CRC 0xf060dc98 (calculated 0xf9d690b3)
[ 7460.981732] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020008: 0xf060 instead
[ 7461.089463] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0002000c: 0xffff instead
[ 7461.196950] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020010: 0xffff instead
[ 7461.304436] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020014: 0xffff instead
[ 7461.411890] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020018: 0xffff instead
[ 7461.519301] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0002001c: 0xffff instead
[ 7461.626689] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020020: 0xffff instead
[ 7461.734079] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020024: 0xffff instead
[ 7461.841467] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00020028: 0xffff instead
[ 7461.948428] Further such events for this erase block will not be printed
:
:
:
[ 7464.664307] jffs2_scan_eraseblock(): Node at 0x00080000 {0x1985, 0x2003, 0x0000000c) has invalid CRC 0xf060dc98 (calculated 0xf9d690b3)
[ 7464.811667] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080008: 0xf060 instead
[ 7464.919083] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0008000c: 0xffff instead
[ 7465.026794] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080010: 0xffff instead
[ 7465.134190] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080014: 0xffff instead
[ 7465.241665] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080018: 0xffff instead
[ 7465.349060] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0008001c: 0xffff instead
[ 7465.456461] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080020: 0xffff instead
[ 7465.563917] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080024: 0xffff instead
[ 7465.671325] jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080028: 0xffff instead
[ 7465.778305] Further such events for this erase block will not be printed
[ 7465.941101] Cowardly refusing to erase blocks on filesystem with no valid JFFS2 nodes
[ 7466.035471] empty_blocks 0, bad_blocks 0, c->nr_blocks 5
mount: mounting /dev/mtdblock3 on /mnt failed


I do not know if there is any mismatch between the mtd and jffs2 in the linux kernel 2.6.23 and MTD 1.0.0 ?

My image2.jffs2 looks like:

00000000  19 85 20 03 00 00 00 0c  f0 60 dc 98 19 85 e0 01  |.. ......`......|
00000010  00 00 00 35 c4 4d 19 44  00 00 00 01 00 00 00 00  |...5.M.D........|
00000020  00 00 00 02 00 00 00 04  0d 0a 00 00 e7 db b4 ae  |................|
00000030  32 6c 8c dc 47 50 4d 4d  2e 74 65 6d 70 6c 61 74  |2l..GPMM.templat|
00000040  65 ff ff ff 19 85 e0 02  00 00 00 6e 7f 54 eb e8  |e..........n.T..|
00000050  00 00 00 02 00 00 00 01  00 00 a1 ff 00 00 00 00  |................|
00000060  00 00 00 2a 00 00 00 04  00 00 00 04 00 00 00 04  |...*............|
00000070  00 00 00 00 00 00 00 2a  00 00 00 2a 00 00 00 00  |.......*...*....|
00000080  58 42 cb c7 7b 3f 69 70  2e 2e 2f 2e 2e 2f 2e 2e  |XB..{?ip../../..|
*
00020000  19 85 20 03 00 00 00 0c  f0 60 dc 98 ff ff ff ff  |.. ......`......|
00020010  ff ff ff ff ff ff ff ff  ff ff ff ff ff ff ff ff  |................|
*
00040000  19 85 20 03 00 00 00 0c  f0 60 dc 98 ff ff ff ff  |.. ......`......|
00040010  ff ff ff ff ff ff ff ff  ff ff ff ff ff ff ff ff  |................|
*
00060000  19 85 20 03 00 00 00 0c  f0 60 dc 98 ff ff ff ff  |.. ......`......|
00060010  ff ff ff ff ff ff ff ff  ff ff ff ff ff ff ff ff  |................|
*
00080000  19 85 20 03 00 00 00 0c  f0 60 dc 98 ff ff ff ff  |.. ......`......|
00080010  ff ff ff ff ff ff ff ff  ff ff ff ff ff ff ff ff  |................|


Best Regards

Miroslaw Dach

Re: MTD and jffs2 mount problem on Linux-2.6.23 rc2

[Grant Erickson]

On 3/4/08 2:34 AM, Dach Miroslaw wrote:
> I have created on my target board the jffs2 image which is foreseen to be
> place on /dev/mtd3:
> 
> ./mkfs.jffs2 --pad=0xA0000 --eraseblock=0x20000 --root=TEST-DIR
> --output=image2.jffs2
> flash_erase /dev/mtd3
> dd if=./image2.jffs2 of=/dev/mtd3 bs=2 count=327680
> 
> Up to this point all went fine.
> 
> The problem emerged when I have tried to mount the jffs2 partition:
> mkdir /mnt
> mount -t jffs2 /dev/mtdblock3  /mnt

Dach:

Based on recent experience, I found the following to work with my board and
a 2.6.25-rc2 kernel:

% mkfs.jffs2 -n -e 16 KiB -b -d /export/haleakala -n -o root.img -s 512 -p
  512

# flash_eraseall -q /dev/mtd9
# nandwrite -q /dev/mtd9 root.img
# mount -t jff2 /dev/mtdblock9 /import

So, based on your steps above, it looks like you might meet success by
replacing 'flash_erase' with 'flash_eraseall' and 'dd' with 'nandwrite'.

Regards,

Grant Erickson