|
Most of my testing has been on my older (for me) K1 SN B101*. For you, I switched to the newer one (B001) because it has the latest firmware on it AND it is registered to my amazon account, so most "normal" in its configuration.
I will set up an install script for systematic regression testing, so I can find what works and what doesn't in an orderly fashion. I will try copying other bins from rootfs, other destinations (such as /opt and /tmp) and also using a loop mounted EXT3 on userstore (both on /mnt/us and on /mnt/mmc).
I also posted the complete file tree listing above, showing that there is no /mnt/base-us/, nor is there a /var/local.
This device has a lot of partitions that are not mounted, AND (according to some scripts) some partitions that are NOT part of the BML file system (one of which gets flashed with a file called uImage where comments in the script call it a "recovery kernel", as I recall).
I tried to "dd" copy all the partitions (both BML raw and STL remapped versions) and most failed, and since then I see from install scripts that you need to ulock them with "magic cookies" first. Also, in some cases, you format partitions by sending magic cookies to the BML device driver and let *it* do the work, rather than using "normaL" file system utilities. As I reported earlier, these devices do NOT have eMMC, and need to do write wear levelling and bad block management in the BML filesystem (and the driver reports one bad block on my older K1, so a raw partition image of /mnt/us could have sectors out of order).
Regarding NO_EXEC, perhaps the SD card is mounted differently, and that may be where I was able to execute copied binaries. I will try shelling the binary. Thanks. I plan to attack this systematically (I am feeling unusually lucid this morning, even before my morning coffee, which is now brewing, so I hope to accomplish great things).
|