Missing /dev/mpt

I have a target system that has a BOOT.BIN already loaded into its QSPI flash, but when I boot PetaLinux, there is no /dev/mtd. Doing "cat /proc/mtd" returns no results. Can I re-create the mtd partitioning from within PetaLinux? Or is the only solution to load a new BOOT.BIN that was built with the proper MTD settings via Xilinx tools? And how big does mtd0 need to be for a 7z045?

0 Former Member over 7 years ago

Hi,

This was an issue in the 2015.2 release that the factory test resident in the QSPI of some of our boards used (and did not have the fix applied). You can download the BSP for your board from the ZedBoard.org site and you should see the MTD partitions. You don't mention the board you are working with but assuming a Mini-ITX due to the 7z045 question you can find it here:

http://zedboard.org/support/design/2056/17

As for the size of the MTD partitions, it depends on what is going into them. For the boot partition, if the bitstream is included, it has to be big enough to accommodate the 13 MB + size for the 7045. Instructions for changing the partition sizes are in the Boot App note referenced in your other forum thread:

http://zedboard.org/content/split-boot-u-boot-qspi-linux-sd

-Gary
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel
0 zedhed over 7 years ago

Hi skibumgm,

Gary's approach listed above is the best course of action.

The fix mentioned involves adding an entry for the QSPI to the device tree in the system-top.dts file.

&qspi {
   flash0: flash@0 {
      compatible = "micron,n25q128a13";
   };
};

Regards,

-Kevin
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel
0 Former Member over 7 years ago in reply to zedhed

I've tried Kevin's suggestion with no change in results. Unfortunately, I'm working with custom hardware that could most closely be described as a PicoZed SOM with a 7045 in place of the 7030, so the BSP's available here don't fit (I assume). I'm working with a .hdf file exported from Vivado 2016.1, but the PetaLinux environment version I'm running is 2014.4, so versionitis might also be an issue. Any other ideas you might have will be much appreciated.

Thanks,
Gerald
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel
0 zedhed over 7 years ago

Hi skibumgm,

Mixing an matching Vivado and PetaLinux tool versions like that is definitely not recommended since there is no way to test all of the combinations and guarantee compatibility between the pairs.

Regards,

-Kevin
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel
0 Former Member over 7 years ago in reply to zedhed

I updated my PetaLinux environment to 2016.1, re-built everything, including re-exporting the .hdf file from Vivado, with no change.
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel
0 Former Member over 7 years ago

The problem turned out to be an inconsistency in the hardware description. If I used an HDF from the full design, an AXI DMA controller was included n the HDF, which caused PetaLinux to attempt to load the DMA driver, and poll one of its registers. But the bitstream hadn't been loaded, so there was no response. To fix that, I used an HDF from a dummy design that only included the Zynq PS, but it missed some clocking information that prevented the BOOT.BIN created with it from operating. So I mixed the BOOT.BIN from the full system with the image.ub from the PS-only system. It would boot to a linux command prompt, but was not correct. The missing /dev/mtd was just the most obvious symptom of the inconsistency.
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel