I tried 5.7.4 & 5.8.7.0.1 - the latter updated uboot, but no changes in features that I could see. However, some strange behaviour:
1) If I flash update_PW2_5.4.3.2_initial.bin, I can get one to several successful main boots. Eventually I'll see a return to the battery ID check failing.
2) Newer firmwares crash the same way, but provide a bit more info:
Code:
<4>[ 4.896568] DEBUG:eink.c:727:eink_draw_sprite():eink_draw_sprite(x0:480,y0:983,s->x_size=24,s->y_size=41,s->x_off=0,s->y_off=0)
<4>[ 4.896915] INFO:eink.c:891:eink_printf():BATTERY INVALID: -22
<4>[ 5.511564] INFO:recovery-util.c:316:fl_control_fadedown():
<4>[ 5.511652] ===FL fadedown success===
<4>[ 5.511950] ERROR:recovery-util.c:346:battery_check():
<4>[ 5.511962] === BATTERY ID IS INVALID. STOPPING BOOT ===
I noticed that, if I run these newer firmwares & boot with the diags trick (main kernel booting the main rootfs in diags mode), the touchscreen doesn't work! Based on
this thread the behaviour seems to match up, although I've never dropped my unit. It would also explain why "ADC value out of thermistor range! setting display temp to 43 deg C" - there may be an intermittent flaw with the flex cable or display chip. I will reseat the connectors and try again.
It's also worth noting that, while I had the 5.4.3.2 firmware booting properly, I booted diags & mounted NFS over USBnet, taking a dd backup of the entire mmcblk0 in case it's something software causing the issue. I restored that backup and still ran into battery ID checks... at this point I'm beginning to suspect the hardware more and more. If it does turn out to be a hardware fault, I'm back to figuring out how to permanently change the diags bootargs to use mmcblk0p1 instead of 0p2 - this will give me a functional Kindle that can survive reboots, hardware be damned. Everything seems to work on 5.4.3.2 after all.
EDIT: I've now reseated every flex cable on the board. No dice
