09:11 <inode> i compiled OP-TEE and mainline u-boot for my device (rk3229 TX2-clone?) and i (foolishly?) attempted to replace the existing "uboot" partition (@ 0x2000 over rockusb). on boot, i got only the usual DDR init and OP-TEE init messages over UART2 followed by a freeze where u-boot would usually begin.

09:13 <inode> i put the device into maskrom mode, and poked around enough to screw up what i guess would be the initial loader or ID Blocks of the eMMC

09:14 <inode> now the device starts in maskrom mode by default

09:16 <inode> i can use either rkdeveloptool with a merged loader (rk322x_ddr*300mhz*.bin, rk322x_*usbplug*.bin, rk322x_*miniloader*.bin), or rkflashtool (l with ddr init bin, L with usbplug bin) to get enough access to see that the chip id and partitions on the eMMC still look ok

09:18 <inode> if i provide the miniloader to rkflashtool L, i can get the device to boot with uboot (0x2000 LBA) and trust (0x4000 LBA) as usual with the backup's restored - but the kernel bombs out with errors about mmc comms - which i guess is due to the fact that i started in maskrom mode as i can see the partition offsets listed are in raw and LBA sector counts

09:20 <inode> if i try to upgrade the loader with the merged loader that works for me, rkdeveloptool's "ul" says: "The device does not support this operation!"

09:21 <inode> if i try to write it to 0x40 via, rkdeveloptool wl 0x40, the device still enters maskrom mode by default

09:22 <inode> the same thing applies if i try to write a custom idbloader.img created as rksd or an rkimage

09:23 <inode> can anyone help me with some ideas about kind of metadata might need to be written into this area around 0x40 (idblocks according to rkflashtool?)

09:23 <inode> or where the initial miniloader might actually need to be placed for eMMC boot?

11:45 <mmind00> inode: the "merged loader" is not what gets written to the 0x40 position I think ... merely it extracts the relevant part from there ... what gets written to the 0x40 position should start with "RK32" in the binary file

11:47 <mmind00> inode: as for optee, you can probably see from the commit history that it didn't get much attention after the initial submission

11:47 <mmind00> inode: and all future stuff was sadly done binary-only

11:49 <mmind00> inode: and I don't know how far along the optee-loading patches for mainline uboot actually are

13:00 <inode> mmind00: thanks, i found that rk322x_ddr_300MHz_v1.08.bin begins with RK32, and writing that to 0x40 didn't prevent mask rom mode unfortunately

13:04 <inode> i found that when using: rkdeveloptool ul ..., the device doesn't like the write to sector 0x40 when it finally calls WriteLBA() with the prepared id blocks (200 sectors worth)

13:04 <inode> so i wrote those blocks out to a file (fopen/fwrite/fclose) and attempted to write them to 0x40 manually

13:05 <inode> no luck either, so i'm going to focus on getting a bootable sd card going instead and save this problem for later

13:07 <inode> mmind00: do you know if rk3229 shares jtag functionallity over any of the balls allocated for other purposes like sdmmc or something?

13:29 <inode> at least u-boot seems to work coherently enough from the sdcard