MMC devices do not work on MMC1 for linux boot

david atkins

Other Parts Discussed in Thread: DM3730

Hi,

Trying posting here as posting as advised in the linux forum has been a complete waste of time! this is a copy of a thread theer in the hope that someone at TI will haveto courtesy to take a look. I have been assured that 87% of posts get an answer so it would be good to see it happen......

Hi,

We have our hardware platform losely based on Beagle XM but massively cutdown, its has a Toshiba eMMC device (2GB) on MMC1, this is intened to act as the boot device.

We are doing a sector wise copy from an micro SD card to the eMMC, the SD card image is built using the DVSDK 4.02.00.06 and is tested on a Beagle XM and the same hardware with a SD card socket hardwired in place of the eMMC device, in both cases a successfully boot it achieved up to the linux prompt.

When we boot from the eMMC xload and uboot do the right thing and the kernel starts, when it gets to mount the file system from the second partition the device recognised as an MMC and the size is reported correctly. However thats it, it hangs at this point and never recovers.

We are currently adding printks to all the affected drivers but the methdo we have to use to clone the image onto the eMMC is very slow, so any help offered gratefully recieved.

Anyone else succesfully using eMMC to boot a DM3730?

David

my followup

We have now tested this further using a standalone MMC card and Begale card, we get exactly the same issue, the MMC card will not mount as a file system and the kernel hangs at that point.

As we have the issue with both our hardware and the Beagle hardware we assuem the problem is either the devic eor the driver, the beagle has a OM3530 so its a different chip suggetsign a basic drive issue.

Can anyone at TI confirm whether the MMC has been tested with the 4.x series of DVSDK releases?

Meant to post the snippet from the boot log, as follows

[   25.657836] List of all partitions:
[   25.661437] b300         1966080 mmcblk0 driver: mmcblk
[   25.666717]   b301           40131 mmcblk0p1
[   25.671081]   b302         1919767 mmcblk0p2
[   25.675415] No filesystem could mount root, tried:  ext3 ext2 ext4 squashfs vfat msdos
[   25.683502] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(179,2)

Original thread

http://e2e.ti.com/support/embedded/linux/f/354/t/161620.aspx#592619

over 13 years ago

0 Michael T over 13 years ago

TI__Expert 6735 points

David:

Sorry that it took too long to get to your question.

We have a script for formatting SD Cards as used with the EZSDK (the DVSDK may have a similar script).

http://processors.wiki.ti.com/index.php/How_to_Make_3_Partition_SD_Card

We format the boot partition (mmcblk0p1) as FAT32 and the rootfs partition (mmcblk0p2) as EXT3.

Could you post your u-boot env (output from u-boot printenv command)?

Beagleboard XM also requires a uEnv.txt file in boot partition (since it has no NANAD to save uboot env.

Boot partition should have x-loader and u-boot which would together be about 300KB. Why is your

mmcblk0p1 only 40131 bytes in size?

Regards,

Michael T

0 david atkins over 13 years ago in reply to Michael T

Intellectual 500 points

Hi,

we use the dvsdk script to create the file systems on a linux box, if we use the microSD card created this will boot on a hardwired micro sd slot attached to the MMC1 bus, this is hack we have used to make some progress. The production hardware has to vibration tolerant so we are using eMMC.

If we copy the sd card to eMMC then it will boot till it attempts to mount the rootfs file system. It appears to get snafued in the mount process. We have aquired a 2GB MMC card which due to its size can only be used in the older OM3530 beagle card. In this the MMC card has the same issue but the SD card works.

One of my software guys has sprayed the driver with printks and is trying to determine where it dies.

Are the numbers not sector or block counts?

David

0 Velan over 13 years ago in reply to david atkins

TI__Intellectual 2785 points

David,

User="David"

Can anyone at TI confirm whether the MMC has been tested with the 4.x series of DVSDK releases?

Before we dwell in to the subject, we would like to inform you that all the DVSDK 4.x releases are tested with MMC/SD boot. You could refer section "How to create an SD card" in the document TMS320DM3730_Software_Developers_Guide.pdf

attached (sdcard_boot_trace_DVSDK_4_02_00_06.txt)is the SD card boot trace with the DVSDK 4.02.00.06 in case if this would help.

Probably the issue you are facing could be due to the use of different variant of SD card or there could be issue with the driver handling these variants.

As mentioned to you in other post "http://e2e.ti.com/support/embedded/linux/f/354/t/155977.aspx#594802", this could be issue pertain to beagle board.

I will ask other folks to look in to this post.

Best Regards,

Velan

0 david atkins over 13 years ago in reply to Velan

Intellectual 500 points

Velan,

Perhaps you would be kind enough to confirm the part number of the devcies TI used to test this then, we can use the data sheet for that device to see what might be different.

We are using eMMC not SD, SD works find and has nto been issue, MMC doe snto work and does nto work on our hardware or on Beagle.

David

0 Velan over 13 years ago in reply to david atkins

TI__Intellectual 2785 points

David,

Part number is DM3730, you are right we tested against SD card only, not against MMC & eMMC. I am not familiar with the driver side of SD/MMC/eMMC, probably that is what missing here.

Best Regards,

Velan

0 david atkins over 13 years ago in reply to Velan

Intellectual 500 points

So, where do we go from here, is that it, your earlier email suggests that you had expected it to be tested, is this an oversight? Is there a timeline for this to be done?

Processors

Processors forum

MMC devices do not work on MMC1 for linux boot