[avrdude-dev] [bug #42267] jtag3isp fails to read lock and fuse bytes directly after changing lock byte

[Joerg Wunsch]

Update of bug #42267 (project avrdude):

                  Status:                    None => Confirmed              
             Assigned to:                    None => joerg_wunsch           

    _______________________________________________________

Follow-up Comment #2:

This appears to be an issue with the JTAGICE3 firmware itself.

You are right in that it's certainly timing-related: even 
adding -vvvv makes AVRDUDE slow enough to no longer trigger 
the issue.  If you sniff the USB traffic, you can see that 
the wrong answer comes straight from the device, so AVRDUDE 
itself is not at fault.

Btw., this happens for all the Atmel tools, ranging from 
the AVRISPmkII through STK600, JTAGICE3 to the new Atmel-ICE. 
Quite likely, they all use the same piece of firmware to 
implement the ISP protocol.  The only tool being different is 
the JTAGICEmkII which works correctly.

So far, this used to have been masked off by the stk500v2 
code using the old SPI_MULTI command for single-byte updates. 
Starting with the JTAGICE3, this command is no longer supported, 
so it had to be replaced using standard read/write commands in 
AVRDUDE.  Meanwhile, the code has been modified to always use 
this (more modern) implementation, so this firmware bug becomes 
apparent for all the tools.  (Besides, that change "fixes" the 
effect described in bug #21954.)

I'd tend to close the bug report as being "invalid" as it's 
clearly a firmware issue with those tools.  Adding arbitrary 
delays just to accomodate for broken tools seems like a crock 
to me, albeit it might be possible to do that.


    _______________________________________________________

Reply to this item at:

  <http://savannah.nongnu.org/bugs/?42267>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.nongnu.org/