Hello,
I've run into a severe issue trying to flash your core-qte image onto emmc using the eflasher:
While the flashing process completes normally without any error messages, the ext4 root filesystem on emmc is corrupt afterwards. You will not notice this immediately, as the filesystem, even though corrupt "under the hoods", is still good enough to make the NanoPC-T3 boot normally and run the QtEmbedded test app just fine, but once you start doing a larger number of write operations to the root device on emmc, you will sooner or later horribly fail (on next reboot at the latest)...
I tracked this error further down as follows:
The attached script session output shows some commands that I issued from the eflasher environment, i.e. immediately after successful completion of the flashing task.
/dev/mmcblk1p1 (mountpoint: /boot) reports being clean on fsck.ext4, and this can indeed be confirmed by forcing a check with "-f".
But for /dev/mmcblk1p2 (mountpoint: /) aka rootfs, the "normal" fsck.ext4 also reports the partition to be clean, but this is a lie: When the file system check is forced with "-f", you see literally thousands of errors on the filesystem, i.e. it is severely corrupt...
I have managed to track this down a little further: To me, it looks like the file "images\core-qte\rootfs.img" itself is not corrupt: its md5 matches the md5 on your server. Also, the plain ext4 partition image that results from running the "simg2img" tool on the "sparse" image (rootfs.img) - resulting in the raw image (rootfs.img.raw) - passes a forced fsck.ext4 file system check just fine:
"fsck.ext4 -fn rootfs.img.raw" indeed reports the raw partition image as being free of errors.
So I assume that the bug seems to be hidden in "sd_update", your proprietary binary tool which seems to be used to write a sparse image onto a partition slot.
Can you please reproduce this, and then track the bug down to its root cause and fix it. In the meantime, please comment on whether the following workaround would be vaild:
1) expand sparse rootfs.img with simg2img to raw image rootfs.img.raw
2) directly write raw image onto /dev/mmcblk1p2 using dd (using which parameters?)
3) running resize2fs on /dev/mmcblk1p2
Many thanks for your kind help!
Best regards,
awl
I've run into a severe issue trying to flash your core-qte image onto emmc using the eflasher:
While the flashing process completes normally without any error messages, the ext4 root filesystem on emmc is corrupt afterwards. You will not notice this immediately, as the filesystem, even though corrupt "under the hoods", is still good enough to make the NanoPC-T3 boot normally and run the QtEmbedded test app just fine, but once you start doing a larger number of write operations to the root device on emmc, you will sooner or later horribly fail (on next reboot at the latest)...
I tracked this error further down as follows:
The attached script session output shows some commands that I issued from the eflasher environment, i.e. immediately after successful completion of the flashing task.
/dev/mmcblk1p1 (mountpoint: /boot) reports being clean on fsck.ext4, and this can indeed be confirmed by forcing a check with "-f".
But for /dev/mmcblk1p2 (mountpoint: /) aka rootfs, the "normal" fsck.ext4 also reports the partition to be clean, but this is a lie: When the file system check is forced with "-f", you see literally thousands of errors on the filesystem, i.e. it is severely corrupt...

I have managed to track this down a little further: To me, it looks like the file "images\core-qte\rootfs.img" itself is not corrupt: its md5 matches the md5 on your server. Also, the plain ext4 partition image that results from running the "simg2img" tool on the "sparse" image (rootfs.img) - resulting in the raw image (rootfs.img.raw) - passes a forced fsck.ext4 file system check just fine:
"fsck.ext4 -fn rootfs.img.raw" indeed reports the raw partition image as being free of errors.
So I assume that the bug seems to be hidden in "sd_update", your proprietary binary tool which seems to be used to write a sparse image onto a partition slot.
Can you please reproduce this, and then track the bug down to its root cause and fix it. In the meantime, please comment on whether the following workaround would be vaild:
1) expand sparse rootfs.img with simg2img to raw image rootfs.img.raw
2) directly write raw image onto /dev/mmcblk1p2 using dd (using which parameters?)
3) running resize2fs on /dev/mmcblk1p2
Many thanks for your kind help!
Best regards,
awl