[Battlemesh] [FCC] What hardware still works?

Jonathan Morton chromatix99 at gmail.com
Tue Feb 23 19:04:28 CET 2016

> On 23 Feb, 2016, at 19:42, Adam Longwill <adam.longwill at metamesh.org> wrote:
> I do not have a good understanding of the difference between jtag/serial/ and tftp. Can someone briefly explain the difference for people like myself? Can JTAG flashing replace a locked firmware? I thought the chips themselves could be built to only cryptographically accept approved firmware? Or is that only with "higher level" flashing methods.
> Anyone have a Explain it Like I'm 5 version out there to help explain?

With TFTP, you’re relying on running firmware to load and write its own replacement.  This may be part of the bootloader or the main firmware - doesn’t matter.  Bootloaders are now large and sophisticated enough to support cryptographic signatures (simple, slow implementations of SHA256 and RSA aren’t very big), which means they can refuse to process images not originating from the original vendor, and they can also refuse to downgrade to an older version which ignores signatures.

Serial console access *might* let you bypass that, by installing and running a flash-writing program that doesn’t care about signatures.  Obviously this is a bit trickier than using the built-in utilities, but when the latter refuse to work, it’s a good option to try.  However, if you format the image wrong, this is an easy way to brick the device.

JTAG is a very low-level debugging interface.  It works completely independently of the firmware (even the bootloader!) and usually has sufficient access to rewrite the flash.  This makes it a great tool for un-bricking devices, where the only fault is in the firmware image.  However, setting up a JTAG is even more fiddly than setting up flash-writing software.

When JTAG doesn’t work - usually because it couldn’t be set up correctly, rather than because it doesn’t have access - the “buspirate” method is the last resort.  This takes direct electrical control of the flash EEPROM’s signals in order to rewrite its contents.  The only plausible defence against this that a vendor could implement would be an access-control password built into the flash chip itself.

 - Jonathan Morton

More information about the Battlemesh mailing list